It is generally admitted that oceanic basalts are formed by continuous melting of the mantle . However, most parameterizations are calibrated on batch melting experiments. Using pMELTS , I modelled the adiabatic decompression of a heterogeneous mantle (peridotite + pyroxenite) and compared the melt composition and productivity produced by both batch and continuous melting.
Calculations reveal two major flaws in using parameterizations based on batch melting rather than continuous melting:
(1) the proportion of solid pyroxenite in the mantle source might be strongly underestimated. Because most proposed mantle pyroxenite compositions are denser than peridotite , an increase in the pyroxenite proportion may affect the buoyancy of the mixture in the upper mantle. Applying the corrected proportion to Iceland, I show that maintaining a positive buoyancy of the mantle source requires either a potential temperature TP ≈ 1600°C notably higher than the recent estimates (i.e., 1450-1510°C [4,5]) or a significant proportion of less dense material, such as harzburgite (e.g., for TP = 1500°C, buoyancy is maintained if the mantle contained > 20% harzburgite).
(2) The Ni proxy used to constrain pyroxenite abundance in mantle source  might not be suitable for continuous melting regime. In fact, the Ni content of the integrated melt produced by continuous melting of a given lithology can be significantly lower than the one produced by batch melting, and Ni-rich, olivine-free pyroxenites might produce melts that are not enriched in nickel when compared with peridotite melts if continuous melting is considered.
References:  McKenzie (1984) J. Petrol 25;  Ghiorso et al. (2002), GGG 3;  Shorttle et al. (2014), EPSL 395;  Herzberg & Asimow (2015), GGG 16;  Matthews et al. (2016), GGG 17;  Sobolev et al. (2007), Science, 412-417.
Oral Presentation: 13 August
Session No. 05a: Morb Petrogenesis: From mantle partial melting to fractional crystallization
13-18 August 2017
Studying mineralogy is fundamental for understanding the composition and physical behavior of natural materials in terrestrial and extraterrestrial environments. However, some students struggle and ultimately get discouraged with course material because they lack well-developed spatial visualization skills that are needed to deal with three-dimensional (3D) objects, such as crystal forms or atomic-scale structures, typically represented in two-dimensional (2D) space. Fortunately, spatial visualization can improve with practice. Our presentation demonstrates a set of experiential learning activities designed to support the development and improvement of spatial visualization skills in mineralogy using commercially available magnetic building tiles, rods, and spheres. These instructional support activities guide students in the creation of 3D models that replicate macroscopic crystal forms and atomic-scale structures in a low-pressure learning environment and at low cost. Students physically manipulate square and triangularly shaped magnetic tiles to build 3D open and closed crystal forms (platonic solids, prisms, pyramids and pinacoids). Prismatic shapes with different closing forms are used to demonstrate the relationship between crystal faces and Miller Indices. Silica tetrahedra and octahedra are constructed out of magnetic rods (bonds) and spheres (oxygen atoms) to illustrate polymerization, connectivity, and the consequences for mineral formulae. In another activity, students practice the identification of symmetry elements and plane lattice types by laying magnetic rods and spheres over wallpaper patterns. The spatial visualization skills developed and improved through our experiential learning activities are critical to the study of mineralogy and many other geology sub-disciplines.
* Denote the speakers
Poster: 13 December
Moscone South - Poster Hall
ED21A: Activities and assignments that promote student learning in undergraduate earth and environmental science classes
AGU Fall meeting San Francisco
12-16 December 2016
Cracking caused by reaction-driven volume increase is an important process in many geological settings. In particular, the interaction of brittle rocks with reactive fluids can create fractures that modify the permeability and reactive surface area, leading to a large variety of feedbacks. The conditions controlling reaction-driven cracking are poorly understood, especially at geologically relevant confining pressures. We conducted two sets of experiments to study the effects of confining pressure on cracking during the formation of gypsum from anhydrite CaSO4 + 2H2O = CaSO4∙2H2O, and portlandite from calcium oxide CaO + H2O = Ca(OH)2. In the first set of experiments, we cold-pressed CaSO4, or CaO powder to form cylinders. Samples were confined in steel, and compressed with an axial load of 0.1 to 4 MPa. Water was allowed to infiltrate the initially unsaturated samples through the bottom face via capillary and Darcian flow across a micro-porous frit. The height of the sample was recorded during the experiment, and serves as a measure of volume change due to the hydration reaction. We also recorded acoustic emissions (AEs) using piezoelectric transducers (PZTs), to serve as a measure of cracking during an experiment. Experiments were stopped when the recorded volume change reached 80% - 100% of the stoichiometrically calculated volume change of the reaction. In a second set of experiments, we pressed CaSO4 powder to form cylinders 8.9 cm in length and 3.5 cm in diameter for testing in a tri-axial press with ports for fluid input and output, across the top and bottom faces of the sample. The tri-axial experiments were set up to investigate the reaction-driven cracking process for a range of confining pressures. Cracking during experiments was monitored using strain gauges and PZTs attached to the sample. We measured permeability during experiments by imposing a fluid pressure gradient across the sample. These experiments elucidate the role of cracking caused by crystallization pressure in many important hydration reactions.
* Denotes the speaker
Poster: 15 December
Moscone South - Poster Hall
Session: MR41A: Physical Properties of Earth Materials (PPEM): Rock Deformation over Various Time and Spatial Scales II
AGU Fall meeting San Francisco
12-16 December 2016
Thanks to numerous experimental studies, parameterizations are available to model the melting behavior of peridotite  and pyroxenite  compositions that are thought to be present in the mantle. Based on these parameterizations, numerous studies have attempted to estimate the proportion of pyroxenites in magmatic sources. However, while these parameterizations are mostly based on batch melting experiments, oceanic basalts are likely to be formed by near fractional melting rather than batch melting . Using pMELTS , I investigated the effect of near-fractional melting of pyroxenite and applied the results to Iceland's basalt for which the proportion of pyroxenite-derived melt has been estimated to 20-30 % [5,6]. Calculations suggest that 30% of pyroxenite-derived melt in the aggregated magma can be generated by 5% pyroxenite in the source in a case of batch melting but highlight that at least 12% (i.e, more than twice the amount required with batch melting) pyroxenite in the source may be required for near-fractional melting. The buoyancy of a mantle with 5% of pyroxenite can simply be maintained with increasing the mantle potential temperature, TP, from 1300 to to 1440°C, but 12% of pyroxenite in the mantle required either TP ≈ 1600°C, i.e., a temperature significantly higher than the recent estimates (i.e., 1450-1510°C [7,8]) or a significant proportion of less dense material, such as harzburgite (e.g., for TP = 1500°C, buoyancy is maintained if the mantle contained > 20% harzburgite). Note these estimates are minimal as the plume buoyancy beneath Iceland is supposed to be strongly positive to meet the volume flux estimate .
 Katz et al. (2003), GGG 4;  Lambart et al. (under review), JGR - Solid Earth;  Hirose & Kawamura (1994), Geophy. Res. Let 21, 2139-2142;  Ghiorso et al. (2002), GGG 3, 1030;  Sobolev et al. (2007), Science, 412-417;  Shorttle et al. (2014) EPSL 395, 24-40;  Courtier et al. (2007), EPSL 264, 308-316;  Herzberg & Asimow (2015), GGG 16, 563-578;  Jones et al. (2014), EPSL 386, 86-97.
Oral Presentation: 30 June
Session 04g: Do basalts accurately sample mantle source rocks?
The role of lithologic heterogeneity in the sub-oceanic upper mantle during melt generation at ridges and hot spots
26 June - 01 July 2016
A variety of data requires that the mantle source for basaltic magmatism is heterogeneous. Thanks to numerous experimental studies, parameterizations are available to model the melting behavior of peridotite and pyroxenite compositions that are thought to be present in the mantle (e.g., 1, 2). Based on these parameterizations, numerous studies have attempted to estimate the proportion of pyroxenites in magmatic sources. However, while almost all melting experiments correspond to a batch melting process, it is likely that oceanic basalts are formed by near fractional melting rather than batch melting (e.g., 3). Due to the limited extent of melting of peridotites under upper mantle conditions, their magmatic productivity and melt compositions are similar for batch and fractional melting (e.g., 4). In contrast, pyroxenites undergo much higher meting degrees during decompression of a heterogeneous, peridotite-rich mantle source. Using pMELTS, I investigated the effect of near-fractional melting of pyroxenite. Results suggest that the nature of the melting process for pyroxenites can significantly affect (1) the melt productivity of pyroxenites and thus their potential contribution in basalt genesis, (2) the major element composition of melts and thus their interaction with the surrounding peridodite, and (3) the concentration of minor elements such as Ni and consequently the estimation of pyroxenite proportion in magma-source (e.g., 5). In particular, calculations imply that the proportion of solid pyroxenite in the magma source is likely to be underestimated using "batch melting" rather than "fractional melting" parameterization. An increase in the pyroxenite proportion may affect the buoyancy of the mixture in the upper mantle and have important geodynamical implications.
1-Katz et al., 2003, GGG 4; 2-Lambart et al., 2013, Lithos 160-161; 3- Hirose & Kawamura, 1994, Geophy. Res. Let 21; 4-Johnson et al., 1990, J. Geophy. Res. 95; 5-Sobolev et al., 2007, Science 316
Poster: Wed. December 16th
Paper No. DI31B-2582
Session: DI31B Melt and Liquids in Earth and Planetary Interiors
AGU Fall Meeting, San Francisco
14-18 December 2015
It is generally believed that melt-peridotite interaction in the upper mantle plays an important role in melt transport. A major aspect of these interactions is the coupled dissolution of pyroxene and precipitation of olivine, or vice versa. Pyroxene dissolution and olivine precipitation increase the proportion of melt in the system and facilitate its transport by increasing the porosity and permeability of mantle rocks [1,2]. This process has also been suggested for the extraction of pyroxenite-derived melt with little interaction with the peridotite mantle [3,4]. Natural occurrences of this reaction are found, for example, in the Krivaja-Konjuh massif in Bosnia , in mantle xenoliths from French Polynesia , and in the Balmuccia massif in Italy . Inversely, pyroxene crystallization at the expense of olivine would lead to a porosity reduction and increase the lithological variability of the mantle [8,9]. However, melt-peridotite interaction processes depend on many chemical and physical parameters such as the composition of the melts, and the nature of the transport mechanism (pervasive porous flow, focused flow or magma transport in dikes), and are very difficult to model in the laboratory.
First I will review the experimental and theoretical efforts that have been published to constrain the effect of melt-rock interaction on the lithological variability of the mantle as functions of the melt composition, the melt-rock ratio and the pressure-temperature conditions. Then I will discuss the conditions to reproduce the rock suite observed in the Krivaja-Konjuh massif and how they differ from the conditions in the Balmuccia massif and the French Polynesia xenoliths.
 Kelemen et al., 1990, J. Petrol. 31(1), 51-98;  Daines & Kohlstedt, 1994, GGG 21(2), 145-148;  Lundstrom et al., 2000, Chem. Geol. 162, 105-126;  Hirschmann et al., 2003, Geology 31(6), 481-484;  Faul et al., 2014, Lithos 202-203, 283-299;  Tommasi et al., 2004, EPSL 227, 539-556;  Mazzucchelli et al., 2009, J. Petrol 50(7), 1205-1233;  Lambart et al., 2012, J. Petrol. 53(3), 451-476;  Sobolev et al., 2005, Nature 434, 590-597.
Geological Society of America Abstracts with Programs. Vol. 47, No. 7
Oral Presentation: 3 November
Paper No. 195-13
Session No. 195:
T162. A Lower Crustal Perspective on Magmatic Arc Processes
GSA Annual Meeting
1-4 November 2015
Carbonation of peridotite may be an important sink in the global carbon cycle. Some natural systems attain 100% carbonation. In these systems, all Mg, Ca and Fe is extracted from silicate minerals to form carbonate minerals, perhaps as a result of reaction-driven cracking processes that maintain or enhance permeability and reactive surface area. Without reaction-driven cracking, in situ mineral carbonation could fill pore space and armor reactive surfaces, limiting reaction progress and leaving much of the silicate unaltered (Kelemen and Hirth, 2012).
The reactive-cracking process will occur if stress resulting from anisotropic volume changes during fluid-rock interactions, i.e., the "crystallization pressure", is sufficient to fracture rocks. The crystallization pressure is usually described as:
P' = -ΔGr/ΔVs (1)
where P' is the crystallization pressure, ΔGr is the Gibbs Free Energy of a reaction, and ΔVs is the change in solid volume resulting from this reaction (Scherer, 2004; Steiger, 2005). Kelemen and Hirth (2012) proposed that use of Helmholtz Free Energy of reaction ΔFr may be preferable, yielding:
P' = -ΔFr/ΔVs = -ΔGr/ΔVs + PΔVr/ΔVs (2)
where P is the confining pressure and ΔVr is the volume change of a stoichiometric reaction including fluid components.However, the conditions most favorable for reaction-driven cracking are poorly understood, especially at confining pressures relevant to CO2 storage. Fundamental uncertainties remain in estimating the crystallization pressure that drives reaction-driven cracking and, so far, experiments on peridotite hydration or carbonation have not produced reactive cracking, possibly due to limited reactive surface area in low porosity samples. To address this, we explore crystallization pressure in simple systems under controlled laboratory conditions.
We have performed experiments on reaction (1): solid CaO + H2O = solid Ca(OH)2, with a 100% increase in the solid volume and high initial porosities (Φ). Using equation (1), the calculated pressure of crystallization of Ca(OH)2 is 4.1 GPa. However, equations (1) and (2) do not incorporate energy sinks such as exothermic heating and/or thermal diffusion, implicitly assuming that all chemical potential energy is converted into stress. Thus, both equations yield an upper bound for P'.
We cold-pressed CaO powder to form ∼20 mm long cylinders 4 to 12 mm in diameter, with initial Φ ranging from 0.36 to 0.53. These cylinders were confined in steel, and compressed with an axial load of 0.1 to 27 MPa while water was introduced through a micro-porous frit. Without expansion of the total volume, the reaction would stop when Φ = 0, producing Φ = 0, 2∙Φ∙VCa(OH)2+ (1-2∙Φ)∙VCaO.
Instead, in all experiments the volume of cylinders increased with time, maintaining Φ > 20%. Hence, the stress exerted by the crystallization of Ca(OH)2 is higher than the maximum axial load (27 MPa). 27 MPa is also higher than the tensile strengths of most of the rocks, suggesting that, without confined pressure, reactive-cracking can happen with this system. The evolution of ΔV is best fit by a power law. The reaction slows over time, either because of a reduction of permeability or consumption of the reactant. The fact that all experiments show the same slope (ΔV = btm with m ∼ constant) whatever if the porosity was decreasing or increasing during the reaction, supports the latter explanation. Hence, experiments were stopped at reaction extents from 42 to 100% but all would certainly reach 100% at longer durations.
To investigate the conditions leading to reaction-driven cracking in this system, we’ve begun a series of high pressure experiments in a triaxial deformation apparatus. We will map stress and strain resulting from CaO hydration at geologically relevant combinations of confining pressure, temperature, and volume change. The configuration that we chose is a CaO cylinder embedded in a relatively inert, porous, sandstone (i.e., mainly SiO2). CaO solid + H2Ofluid = solid Ca(OH)2 and CaO solid + CO2 fluid = solid CaCO3. Both have a big volume change. The configuration with has already been tested without confined pressure and the first macroscopic fractures appears after only 3min. Additional fractures continued to form during the next 15min, after what the sample fell apart.
Kelemen PB, Hirth G, 2012, Reaction-driven cracking during retrograde metamorphism: Olivine hydration and carbonation: Earth Planet. Sci. Lett., v. 345-348, p. 81–89.
Scherer GW, 2004, Stress from crystallization of salt: Cement and Concrete Research, v. 34, p. 1613-1624.
Steiger M, 2005, Crystal growth in porous materials – I: The crystallization pressure of large crystals: J. Crystal Growth, v. 282, p. 455-469.
* Denotes the speaker
Keynote Presentation 3:
23 June - 8:30am-9:00am
Fifth International Conference on Accelerated Carbonation for Environmental and Material Engineering
New York City
21-24 June 2015
The presence of the Hawaiian plume is manifested by the Hawaiian swell  and voluminous eruption of Ni-rich lavas  with enriched isotopic compositions . Here we estimate the conditions of melt generation needed to reproduce both features.
We used thermodynamic treatment for fractional melting  and melting parameterizations for pyroxenites  and peridotite  to determine pyroxenite contribution in magmas Xpx as functions of potential temperature TP, pyroxenite abundance in the source P, radius of the melting zone R and distance to the plume axis. The final pressure of melting is set to correspond with the base of the lithosphere (3 GPa) at the plume axis and increases with the distance from the axis . The Hawaiian plume axis is thought to be currently between Loihi (L), Kilauea (K) and Mauna Loa (ML), which are 25 km, 32 km and 44 km radially away from the plume axis, respectively . To determine Xpx, we assumed that magmas are accumulated melts produced on a circular sampling zone of 50 km diameter centered beneath each volcano .
Preliminary calculations show that for TP = 1525°C, P = 0.07 and R = 55 km, XpxML = 0.59, XpxK = 0.49 and XpxL = 0.45. XpxML and XpxK are similar to values suggested by . Computed liquidus temperatures at 3 GPa are consistent with those of Hawaiian parental melts (1500-1520°C; ). XpxL is higher than suggested by  (XpxL = 0.09) but their estimate is based on only one glass analysis. Our model is also consistent with isotopic compositions: K and L have similar εNd, while ML is more enriched . Finally, we can compute the density deficit using parameterization of  and relate it to the volume flux volume flux : we obtain 3.2 km3/Yr, a value similar to the estimations based on the Hawaiian swell model .
 Ribe & Christensen 1999, EPSL;  Sobolev et al. 2005, Nature;  DePaulo et al. 2001, GGG;  Phipps Morgan 2001, GGG;  Lambart et al. in prep, EPSL;  Katz et al. 2003 GGG;  Ito & Mahoney 2005, EPSL;  DePaulo & Stolper 1996, JGR;  Putirka 2008, Miner. Soc. Amer. Geochem. Soc.;  Turcotte & Schubert 2002, Cambridge Press.
Poster: 17 December
Session: The geochemical diversity of the mantle inferred from hotspots: Five decades of debate
AGU Fall Meeting
15-19 December 2014
Carbonation of peridotite may be an important sink in the global carbon cycle. In some natural systems, 100% carbonation of rocks is attained via reaction-driven cracking processes. Engineered systems that emulate such processes may provide relatively inexpensive CO2 capture and storage [1,2].
Volume changes during fluid-rock reaction lead to stresses in elastic host rocks, known as the "pressure of crystallization", that can cause fracture. Cracking can maintain or enhance permeability and reactive surface area in a positive feedback mechanism. So far, experiments on peridotite hydration or carbonation have not produced reactive cracking, possibly due to limited reactive surface area in low porosity samples. To address this, we've begun experiments on reaction (1): solid CaO + H2O = solid Ca(OH)2, with a 100% increase in the solid volume and high initial porosities (φ).
We cold-pressed CaO powder to form ~20 mm long cylinders 6 mm in diameter, with initial φ ranging from 0.36 to 0.53. These cylinders were confined in steel, and compressed with an axial load of 0.1 to 4.2 MPa while water was introduced through a micro-porous frit. Without expansion of the total volume, the reaction would stop when φ = 0, producing 2⋅φ⋅Ca(OH)2 + (1-2⋅φ)⋅CaO.
Instead, in all experiments the volume of cylinders increased with time, maintaining φ > 20%. Experiments were stopped at reaction extents from 82 to 100% and would probably reach 100% at longer durations. The pressure of crystallization for reaction (1) is then > 4.2 MPa. Experiments performed on boreholes containing demolition mortar, largely CaO, demonstrate that this pressure is sufficient to break rocks .
Reaction (1) is rapid, which allowed us to perform numerous experiments but at higher axial loads Ca(OH)2 may flow viscously. In the future we plan similar experiments on ground, cold pressed olivine + H2O + CO2. At elevated temperatures, reaction progress in unconfined powders is > 80% in a few hours .
 Kelemen and Hirth 2012 EPSL;  Kelemen et al. 2011 AREPS;  Kelemen et al. 2013 AGUFM;  Gadikota et al. 2014 Phys Chem Chem Phys.
* Denotes the speaker
Poster: 16 December 2014
Session: Achieving negative carbon emissions: Distributed carbon capture from air and surface water using geologic materials and/or storage reservoirs
AGU Fall Meeting
15-19 December 2014
Melting of mantle sources with multiple rock types, each with their own melting behavior and chemical and isotopic properties, is believed to be an important factor in producing the range of magma types characteristic of individual igneous provinces. An important example of such a compound source is a peridotitic mantle with minor pyroxenitic veins. Melting models of such mixtures require knowledge of the relationships between melt fraction (F), temperature (T), pressure (P), and bulk composition (X) for both peridotites and pyroxenites. While various parameterizations are available to model the melting behavior of peridotites, none yet exist to model pyroxenite melting.
Although empirical melting parameterizations are a useful and simple way to incorporate melting into tectonic models and have the advantage that they often better reproduce existing experimental data than exclusively thermodynamically models, current experimental data for pyroxenite melting are too sparse and show a large amount of scatter. Hence, we took an approach that incorporates thermodynamical concepts, using pMELTS  to guide our choice of functional forms, and calibrates them against the current experimental database. In this way, our parameterization becomes a reliable tool for compositions not used to calibrate the model.
Our model, PX-MELT is calibrated on 193 experiments (20 bulk compositions) in which F has been determined up to the cpx-out for pyroxenites over a substantial range of P and T (0.9-5 GPa; 1150-1675°C) and the parameterization is of the form of:
F = A·T′2 + B·T′ + 5 with T′=(T - T5%)/(Tcpx - T5%)
where F is the weight fraction of melt present, T5% is the temperature at F = 5%, Tcpx is the temperature of clinopyroxene exhaustion (cpx-out) in the pyroxenite assemblage and A and B are functions of the pressure (P, in GPa) and of the bulk composition of the pyroxenite.
PX-MELT, reproduces the melting degree undergone by a pyroxenite with a mean standard deviation of 11% absolute, T5% and Tcpx are reproduced with a standard deviation of 30°C and 32°C over a temperature range of ∼ 500°C, respectively. To measure the success of our melting parameterization, we used the variance reduction , i.e. the percentage of the total variance of experimentally determined F explained by the parameterization. The mean variance reduction on the 18 compositions used in the model is 83%. This value is similar to those obtained with parameterizations on peridotite melting :(72-85%) and is much higher than the value obtained with pMELTS for this experimental dataset (55%).
While remaining mathematically simple, PX-MELT succeeds in capturing the important features of the behavior of pyroxenites melting. It can be used to model the partial melting of multilithologic mantle sources, including the effects of varying the composition and the modal proportion of pyroxenite in such source regions. Several examples will be presented and the potential implications for the magma production and transport will be discussed.
Oral presentation: 8 May
Session: Subcontinental mantle lithosphere / melt-rock reactions
Sixth International Orogenic Lherzolite Conference
4-15 May 2014
As the primary flux of material from the mantle to the surface, the basalts erupted at mid-ocean ridges (MORB) are a key resource for investigating the mantle's chemical composition. However, despite the large volumes of oceanic lithosphere returned to the mantle by subduction, it has proven difficult using basalt chemistry alone to quantify this material's involvement in melt production. Even more enigmatic is the signal of refractory material in the source, which may barely melt if other more fusible lithologies are present and so be difficult to identify from many common chemical tracers. Here we demonstrate how combining thermodynamic models of melting, the density of phase assemblages at high pressure and geochemical observations, can allow the proportion of refractory and enriched material in the mantle source to be estimated and place limits on mantle potential temperature.
We focus on determining the abundance of recycled material in the mantle beneath Iceland, where we have excellent geophysical and geochemical constraints on the melting process and the chemical variability in the mantle. Firstly, the lithologies contributing to melting are identified by quantitative comparison of the major element composition of erupted basalts to a database of experimental partial melts (Shorttle and Maclennan, 2011). Secondly, a mass balance is calculated between the endmember basalt compositions and the fully mixed melt to obtain the relative proportion, by mass, of enriched and depleted melts. A three-lithology melting model is then developed (peridotite-harzburgite-basalt), which uses the appropriate melting parametrisations to account for the differences in productivity between each lithology. The melting model enables the calculated abundance of the different endmember melt compositions to be projected back into mass fractions of solid mantle domains.
Applying this method to Iceland demonstrates that ~10% of the source is recycled basaltic material and at least 20% must be highly refractory and essentially un-melting. Combining geophysical constraints with the modelled high pressure densities of the three lithologies' assemblages constrains excess mantle temperature beneath Iceland to be at least 150°C. We extend the crustal thickness and geochemical constraints south along the Reykjanes ridge to show that Iceland represents a long wavelength lithological and thermal anomaly in the mantle - and that both lithology and temperature must be varying along ridge to match observations. Density modelling shows that the proportion of recycled basaltic material carried in the Iceland plume is near the limit of what maintains plume buoyancy in the shallow mantle.
References: Shorttle O. and Maclennan J. Compositional trends of Icelandic basalts: Implications for short-lengthscale lithological heterogeneity in mantle plumes. Geochemistry, Geophysics and Geosystems, 12(11):Q11008, 2011.
Poster: 10 December 2013
Session: Linking the Earth's Surface With the Deep Interior: Comparing Predictions and Observations of Mantle Plumes II
AGU Fall Meeting
9-13 December 2013
As the primary flux of material from the mantle to the surface, the basalts erupted at mid-ocean ridges (MORB) are a key resource for investigating the mantle's chemical composition. However, despite the large volumes of oceanic crust returned to the mantle by subduction, it has proven difficult to estimate the abundance of this recycled material in the mantle source using the chemistry of MORB. This is a significant problem, as fundamental questions about the dynamics of our planet cannot be answered without quantifying the abundance and spatial distribution of the mantle's chemical heterogeneity: is Earth's convection layered, or does it involve the whole mantle? What is the eventual fate of recycled oceanic material? What are the fluxes of elements from the deep Earth to the surface? Here, we present a method to estimate the proportion of enriched material in mantle source regions by combining geochemical observations with simple bilithologic models of mantle melting.
Working backwards from the chemistry of an erupted basalt to the proportions of peridotite and basalt lithologies in the source requires a number of critical pieces of information. Firstly, the geochemical variability at a ridge segment needs to be used to identify the types of lithology melting. Secondly, the relative mass of the enriched and depleted melts needs to be determined. Finally, a bi-lithological melting model needs to be run in an attempt to account for the over-representation of fusible, productive lithologies in the final mixed melt (Hirschmann and Stolper, 1996; Shorttle and Maclennan, 2011). This last step involves a large number of secondary assumptions/inputs to make the melting problem tractable, such as the mantle flow field and mantle potential temperature.
We determine the abundance of recycled material in the mantle beneath Iceland and at other ridges and ocean island settings in three stages. (1) The lithologies contributing to melting are identified by quantitative comparison of the major element composition of erupted basalts to a database of experimental partial melts (Shorttle and Maclennan, 2011). (2) A mass balance is calculated between the endmember basalt compositions and the fully mixed melt to obtain the relative proportion of enriched and depleted melts. (3) The bilithologic melting model from Shorttle and Maclennan (2011) is then used with the appropriate lithological melting parametrisations to account for the differences in productivity.
Applying this method to Iceland demonstrates that ∼ 10 % of the source is recycled basaltic material. However there are large uncertainties on this number, and our results demonstrate that the ability to constrain the mass fraction of lithologies contributing to melting depends heavily on the dynamics of mantle flow, melting and melt transport/reaction.
Hirschmann M.M., Stolper E.M. A possible role for garnet pyroxenite in the origin of the garnet signature in MORB. Contributions to Mineralogy and Petrology, 124:185-208, 1996.
Shorttle O. and Maclennan J. Compositional trends of Icelandic basalts: Implications for short-lengthscale lithological heterogeneity in mantle plumes. Geochemistry, Geophysics and Geosystems, 12(11):Q11008, 2011.
Poster: 11 April 2013
Session: Materials, evolution and dynamics of the mantle
European Geosciences Union, General Assembly
7-12 April 2013
Geochemical and isotopic data suggest that the source regions of MORBs and OIBs may be mixtures of peridotite and pyroxenite . Models of the melting of such mixtures require knowledge of the relationships between melt fraction (F; expressed as percent), temperature (T), pressure (T), and bulk composition (X) for both peridotites and pyroxenites. While various parameterizations are available to model the melting behavior of peridotites [2,3], none yet exist to model pyroxenite melting. We have used 243 high-P experiments (22 bulk compositions) in which F has been determined up to the disappearance of clinopyroxene (i.e., cpx-out) for pyroxenites over a substantial range of P and T (0.9-5 GPa; 1150-1675°C) to constrain a model that can be used to estimate T from near the solidus (i.e., T5% the T at which F = 5%) up to cpx-out (Tcpx) for mantle pyroxenites spanning a wide compositional range extending to fertile peridotites.
Following , we assume that F is a monotonic, quadratic function of T from the solidus up to cpx-out. We first fit F as a function of T for each set of experiments at constant P on a given bulk composition; using these fits of F vs T, we then estimated T5% and Tcpx for each bulk composition. Because the uncertainty on the calculated T (T5% or Tcpx) increases with the difference between this T and the nearest experiment (in T-F space), fits where this distance is >50°C have not been used. Four studies on pyroxenites where Tcpx was closely bracketed but F not determined were also included in the database, yielding a total of 42 "T5% values" and 56 "Tcpx values".
Guided by pMELTS calculations , we have previously determined the functional form of T5% f(P,X) . Based on previous studies , Tcpx was parameterized as a linear function of P and Xusing the results of high-pressure experiments from the database of . We then used pMELTS to guide our modeling of the P and X dependences of the coefficients of the quadratic F vs T function between T5% and Tcpx. Tcpx =f(P,X). Although our modeling is not thermodynamic, our use of pMELTS, which is thermodynamically based, reduces the hazards of an arbitrary functional form.
When applied to the 243 experimental data, the model reproduces experimental F values with an average uncertainty of 15%, absolute; T5% of pyroxenites are reproduced with an average uncertainty of 18°C over a temperature range of ≈500°C. For fertile peridotite KLB-1 , our parameterization predicts T5% values of 1243, 1502, and 1689°C at 1, 3, and 5 GPa, respectively. For comparison, at the same pressures, the parameterization of  for fertile peridotite predicts solidus temperatures of 1248, 1473, and 1657°C.
In conjunction with estimates of heats of fusion for pyroxenite and peridotite lithologies, these parameterizations will permit calculations of how multilithologic mantle sources melt during adiabatic decompression, including the effects of varying the composition and the modal proportion of pyroxenite in such source regions.
 Hofmann 1997, Nature;  McKenzie and Bickle 1988, JPet;  Katz et al. 2003 GGG;  Pertermann & Hirschmann 2003, JGR;  Ghiorso et al. 2002, GGG;  Lambart et al. 2011, AGUFM #V32B-04;  Putirka 2008, Rev.Miner. Geochem.;  Hirschmann & al. 2008, GGG;  Takahashi 1986, JGR;  Hirschmann 2000, GGG.
Presentation: 7 December 2012
Session: Mantle Plumes: What Do We Really Know?
AGU Fall Meeting
3-7 December 2012
Models of the contribution of mantle eclogites and pyroxenites to basalt petrogenesis depend on knowledge of the relationship between melt fraction (F), temperature (T), pressure (P), and bulk composition (X) for these lithologies and how this relationship compares to that of mantle peridotite. Here we present a parameterization of experimentally determined T at 5% melting for eclogites and pyroxenites as functions of P and X.
Using existing experimental data (14 bulk compositions, 162 experiments at 1-7 GPa and 1165-1750°C), we fit F as a function of T (at constant P) for each set of experiments on a given bulk composition and used these fits to estimate T at F = 5% melting (T5%). This exercise yielded a total of 32 "data points", each of which represents T5% for one of the 14 bulk compositions in the experimental data set at a given pressure. We limited our treatment to 5% melting since there are few experiments for any of the bulk compositions at F < 5%; moreover, at such low melt fractions, unknown factors such as dissolved water content in the melt can substantially influence the T vs F function.
We used pMELTS  to guide our choice of a functional form of T5% = f(P, X). The two most significant compositional parameters proved to be the bulk Na2O+K2O content [alk, in wt%] and the Mg# (= Mg/(Mg+Fe*), atomic, Fe* equals total Fe). The pMELTS calculations suggested the following functional form: T5% = P² (a·[alk] + b) + P (c·[alk] + d·Mg# + e) + F where a-f are fit parameters. When applied to the 32 experimental "data points", this expression reproduces the input T at F = 5% with an average uncertainty of 19°C over a T range of ~500°C; the maximum misfit is 41°C (see figure).
For the range of experimental bulk compositions used in our fit ([alk] = 0.57-3.71 wt%, Mg# = 0.590-0.898), the parameterization predicts T at F = 5% of 1124-1229°C at 1 GPa; 1251-1544°C at 3 GPa; and 1386-1967°C at 7 GPa. For comparison, the parameterization of  for a fertile peridotite predicts solidus temperatures (TS) of 1248, 1473, and 1799°C at 1, 3, and 7 GPa. Note that with increasing P, calculated TS values decrease relative to the range of T5% values.
Applying our parameterization to a global data set of mantle eclogites and pyroxenites suggests that for P of 4-7 GPa, approximately half of the compositions have T5% > peridotite TS at the equivalent P. The pyroxenites and eclogites with T5% < TS plot below the line: Mg# = 0.091·[alk] + 0.66, in [alk] vs Mg# space. As P decreases from 4 GPa, the percentage of compositions in the global data set with T5% < TS increases, reaching 100% at 1 GPa. Hence, our analysis suggests that not all olivine-poor heterogeneities in mantle sources can contribute substantial amounts of melt (i.e., have T5% < TS); the size of this contribution depends not only on the potential temperature of the upwelling mantle and the thickness of the overlying lithosphere, but also on the bulk composition of the pyroxenite/eclogite heterogeneities.
Geological Society of America Abstracts with Programs. Vol. 47, No. 7
Oral Presentation: 7 December 2011
Session: Mantle Melts: Innovative Approaches and Constraints to Modeling the Melting Regime II
AGU Fall Meeting
5-9 December 2011
Trace element and isotopic characteristic of alkali (i.e nepheline normative) basalts from ocean islands demonstrate that their source-region is heterogeneous . However, the matter of lithologic or major element heterogeneity is still debated. A fundamental characteristic in terms of major elements of these basalts is their very wide variation in silica content for a given MgO content: the silica content ranges from less than 35 wt% to more than 50 wt%. To explain this variation, several assumptions have been suggested such as the variation of the melting rate of a peridotite ±CO2/H2O or various proportions of pyroxenites/eclogites in the source. However, these assumptions have a flaw: they do not take into account the compositional variability of mantle rocks. In fact, each "component" proposed as a possible source of OIB has been considered with a constant composition in terms of major element. Unlike peridotites, pyroxenites show a wide compositional range. Here we investigate if the compositional variability of mantle pyroxenites can explain the variability of alkalic oceanic basalts in terms of major elements.
We present an experimental study at 2 and 2.5 GPa on two natural pyroxenites from Beni Bousera: M5-40 is close to the mean of the natural pyroxenite population; M7-16 represents an iron-rich and silica-poor pole of this population. Solidi of M7-16 and M5-40 are located close to 1275°C and 1225°C at 2 GPa and close to 1325°C and 1375°C at 2.5 GPa, respectively. Melt compositions were analysed with the microdike technique  adapted for ½-inch assemblages. Pyroxenites present contrasted melting behaviours as a function of the phase proportion at the solidus. The high mode of garnet and the presence of olivine in the subsolidus assemblage of M7-16 yield to low-degree melts with very low silica content (36.7 - 41.0 wt%) and low aluminium content (9.3 - 13.2 wt%), very high FeO content (21.7 - 25.5 wt%) and high CaO/Al2O3 ratio (0.85 - 1.36). These major element features well correpond to HIMU-type basalts. Otherwise, the low-degree melts of M5-40 show typical EM1-type compositions according to Jackson and Dasgupta , characterized by high K2O contents (1.4 - 4.6 wt%), high K2O/TiO2 (0.7 - 3.8) and low CaO/Al2O3 (0.22 - 0.58) at moderate high SiO2 and FeO contents .
Thus we show that a large part of the major element variability of alkali oceanic basalt could be explained by the low degree partial melts of mantle pyroxenite if the compositional variability of the latter is taken into account. This has important geodynamical implications for the heterogeneous mantle. Solidi of pyroxenites are located at much lower temperature than the peridotite solidus at a given pressure and low-degree melts of pyroxenites for the origin of alkalic OIB implies strong constraints on mantle temperatures.
Oral Presentation: 13 December 2010
Session: Generation and Evolution of Alkaline to Subalkaline Magmas II
AGU Fall Meeting
13-17 December 2010
The involvement of pyroxenites has often been suggested to solve major problems in basalt petrogenesis (e.g., Hirschmann and Stolper, 1996, CMP, 124: 185-208). Numerous high pressure partial melting experiments of these rock-types allowed to better constrain their melting relations at pressures > 2 GPa. As the melting of pyroxenites can also contribute to basalt petrogenesis at pressures < 2 GPa, we performed partial melting experiments on three natural pyroxenites from Beni Bousera ultramafic massif (Morocco) at 1 and 1.5 GPa in a piston-cylinder apparatus: a garnet clinopyroxenite (M7-16), a garnet pyroxenite (M5-40), and a websterite (M5-103); M7-16 is silica-undersaturated, while M5-40 and M5-103 are silica-saturated. These three rocks cover the compositional cloud of pyroxenites worldwide, and can provide an insight into the links between bulk compositions and melting relations. Run products were characterized by scanning electron microscopy and electron probe microanalysis, and phase proportions were computed by mass balance. Our results suggest that bulk pyroxenite compositions control mineralogical assemblages, that in turn control solidus temperatures, melt productivities, and ultimately the relative contribution of pyroxenites and peridotites to basalt genesis. Melting intervals range from 110° C to 240° C and solidus temperatures range from 1160° C to 1220° C at 1 GPa and from 1200° C to 1265° C at 1.5 GPa; in the three compositions investigated, the solidus temperatures are lower and the productivities higher than those of dry peridotite. Depending on bulk composition and degree of melting, partial melting of pyroxenites at 1-1.5 GPa produces two kinds of liquids: (1) liquids with compositions close to those produced by peridotite melting, which may ascend to the surface without significant chemical exchanges with the surrounding peridotitic mantle; and (2) very silica-undersaturated and Fe-rich liquids, which are in strong disequilibrium with the peridotitic mantle and which may induce major changes of the composition of their host. Thus partial melting of pyroxenites may have significant effects on mantle source compositions and on the physicochemical processes of basalt genesis, while the contribution of pyroxenite-derived melts may not lead to major changes of the compositions of primitive basaltic magmas. The implications of these results for the generation of mid-ocean ridge basalts will be discussed.
Oral Presentation: 19 December
Session: Geochemical Heterogeneities in OIB and MORB Sources: Implications for Melting Processes and Mantle Dynamics III
AGU Fall Meeting
15-19 December 2008
The involvement of pyroxenites has often been suggested to solve major problems in basalt petrogenesis (e.g., Hirschmann and Stolper, 1996). Numerous high pressure experiments were devoted to the study of partial melting of these rock-types, and allowed to better constrain their melting relations at pressures > 2 GPa (where garnet and pyroxenes are the principal residual phases). As the melting of pyroxenites can also contribute to basalt petrogenesis at pressures < 2 GPa, we performed partial melting experiments on three natural pyroxenites from Beni Bousera ultramafic massif (Morocco) at 1 and 1.5 GPa in a piston-cylinder apparatus: a garnet clinopyroxenite A, a garnet pyroxenite B, and a websterite C; A is silica-undersaturated, while B and C are silica-saturated. These three rocks correspond to frequency maxima of the compositional cloud of pyroxenites worldwide, and can provide an insight into the links between bulk compositions and melting relations. Run products were characterized by scanning electron microscopy and electron probe microanalysis, and phase proportions were computed by mass balance. Our main results may be summarized as follows:
(1) Despite important differences in their bulk compositions, the three investigated pyroxenites show similar melting relations and melt evolutions at moderate pressures (1-1.5 GPa). In particular, the clinopyroxene (cpx) contribution to the melting reaction increases with increasing temperature (at fixed pressure), and the cpx fraction in residual phase assemblages increases with increasing pressure.
(2) Pyroxenite-derived melts have higher FeO and lower MgO and SiO2 than peridotite-derived melts. Moreover, they are nepheline-normative on the whole melting interval for silica-undersaturated composition A, and up to melting degrees of ≈ 40% for silica-saturated compositions (B and C), suggesting that the production of nepheline-normative primary melts is facilitated at low pressure in the presence of pyroxenitic rock-types in the basalt source regions. Nevertheless, pyroxenite- and peridotite-derived melt compositions show a wide overlap, and the variation ranges of pyroxenite-devived melts for many oxides (Al2O3, CaO, TiO2, Na2O, K2O) are close to that of peridotite-derived melts.
(3) Despite these similarities in melt compositions, we observed important differences in melt productivities and solidus temperatures as a function of bulk composition and pressure. For instance, the large fraction of plagioclase (30 wt%) in B at 1 GPa increases the solidus temperature and the melt productivity. On the contrary, the increase of the cpx fraction in residual assemblages with increasing pressure tends to decrease the melt productivity.
Hence, our results suggest that, independently of their bulk composition, partial melting of pyroxenites at 1-1.5 GPa produces liquids with major element patterns close to those produced by peridotite melting. However, bulk pyroxenite compositions control mineralogical assemblages, that in turn control solidus temperatures and melt productivities, and so the relative contributions of various pyroxenites and peridotites to basalt genesis. The implications of these results for MORB petrogenesis will be discussed.
Hirschmann MM, Stolper EM (1996), Contrib. Mineral. Petrol., 124: 185-208.
* Denotes the speaker
Oral Presentation: 9 September 2008
XIIth International Symposium on Experimental Mineralogy, Petrology and Geochemistry
7-10 December 2008
The compositions of primitive mid-ocean ridge basalts (MORBs) are not only controlled by partial melting processes at depth, but also by magma-rock interactions en route to the surface (as well as shallow-level fractional crystallization). During their ascent, basalts dissolve pyroxenes in the surrounding peridotites, ultimately leading to the formation of high-porosity dunitic channels. Focused flow in dunitic channels may play a major role in the dynamics of magma transport beneath mid-ocean ridges (Kelemen PB & al., 1995, Nature,375: 747-753).
We performed a series of experiments in a piston-cylinder apparatus to determine the effects of focused magma transport on the composition of the ascending basalt and on the formation of dunitic channels. We assumed that the system follows an adiabatic decompression path and that magma focusing occurs instantaneously at 1.25 GPa: at this pressure, the total mass of liquid in the system is multiplied by a factor , referred to as the focusing factor. We first determined the equilibrium melt composition of fertile mantle MBK at 1.25 GPa-1310°C; this composition was then synthesized as a gel and added in different proportions to peridotite MBK to simulate focusing factors equal to 3 and 6. Peridotite MBK and the two basalt-enriched compositions were equilibrated at 1 GPa-1290°C, 0.75 GPa-1270°C, and 0.5 GPa-1250°C.
Our main result is that, at 0.5 GPa-1250°C and Ω=6, the liquid composition is very close to primitive MORB compositions and is in equilibrium with olivine only. Therefore, this strengthens the hypothesis that magma transport by focused flow can explain both the formation of dunite channels beneath mid-ocean ridges and the fact that MORBs are not in equilibrium with a harzburgitic residue at low pressure.
Poster: 16 April 2007
Geophysical Research Abstracts, Vol. 9, 03387, 2007
European Geosciences Union, General Assembly
15-20 April 2007
The gravitational load of a volcanic edifice is recognised to cause modifications to the underlying substrate and ductile intrusive complexes, producing lateral motion away from the centre and in turn resulting in the development of structures and deformation fields on the volcano. Previous work has concentrated on laterally homogenous systems, resulting in relatively simple structures, which although clear are not necessarily comparable to real-life settings.We use a series of analogue models to show the structures and deformation fields deriving from variations in both the edifice shape and the weak substrate dimensions.
There are significant differences in the structure and deformation fields of central volcanoes (e.g. Mount Etna; Concepción) compared to “ridge-like” volcanic complexes (e.g. Poas, Kilauea). Single volcanoes with realistic topography have a system of summit grabens and conjugate strike-slip faults at the base, which relay the deformation away from the volcano. “Ridge-like” edifices present a straight deformation front, which forces the formation of basal thrusts, which in turn act as the limit for deformation. The form of structures and deformation on the edifice is also influenced by the slope of the underlying substrate, and may cause a situation where spreading is confined to a sector of the cone, increasing the potential of catastrophic collapse of the volcano flank. We use morpho-structural comparisons with the global SRTM dataset to show that the features developed in the models are present on numerous volcanoes in differing tectonic settings, and as such volcano spreading should be considered one of the dominant modifying processes at volcanoes.
* Denotes the speaker
Poster: 28 April 2005
Geophysical Research Abstract, vol. 7, 03045
European Geosciences Union, General Assembly
24-29 April 2005