V24A-01 INVITED 16:00h
Asthenospheric Melt Segregation and Channelization: The Influence of Differential Yielding and Disaggregation on Fluid Flow in Ductile Rocks
Asthenospheric melting processes produce small quantities of melt, yet geochemical evidence suggests this melt must be collected from its source and transported to the surface on extraordinarily short time scales. We propose that a mechanical flow channeling instability, which arises because of viscoplastic behavior of the rock matrix, may resolve this dilemma. To characterize differential yielding we employ a model in which the ratio of the matrix viscosity during decompaction to the viscosity for compaction is treated as a free parameter, {\it R}. Numerical solutions of the compaction equations for porous flow in a matrix with differential yielding reveal that solitary dike-like porosity waves initiate from vanishingly small perturbations to a uniform background porosity. The waves grow and accelerate as they propagate by drawing fluid from the background porosity. For {\it R} $<<$ 1, the wave evolution is self-similar in time such that the growth rate of the waves is $\sim {\it R}^{-3/8}/\tau$, where $\tau$ is the viscous compaction time scale. This result implies an initial melt fraction of $10^{-3}$ could be amplified to the conditions for matrix disaggregation in $\sim 10^{4}$ y for {\it R} = $10^{-8}$, a value expected for the case that decompaction is limited by the melt viscosity. From such an initial melt fraction, wave velocities (10 m/y) may be too low for the waves to be the primary mechanism for rapid melt transport, however the waves offer a mechanism by which by melt may be localized to form a magmatic suspension from which dikes initiate. For flow originating from a high volume source, as in metamorphic devolatilization or some lithospheric melting processes, differential yielding causes flow to be channelled into the tails of dike-like porosity waves with a characteristic spacing corresponding to the viscous compaction length $\delta$ and widths $\sim \delta {\it R}^{1/2}$. The numerical results can be understood in the context of an analytical solution of the compaction equations that is completely general with respect to the constitutive relations used to define the matrix rheology and permeability. This solution combines the porosity dependence of the rheology and permeability in a single hydromechanical potential, which can be used to construct phase diagrams depicting the conditions for smooth pervasive flow, wave propagated melt extraction and matrix disaggregation (dike formation).
V24A-02 16:20h
Getting on the Band Wagon: analysis of melt localization instabilities due to mechanical shear
Recent experiments by Holtzman et. al, (G-cubed, 2003) demonstrate that partially molten aggregates of mantle materials undergoing simple shear (from $\sim 100-500$% strain) can spontaneously develop localized melt-rich bands. These bands have been suggested as a source of seismic anisotropy in the shallow upper mantle, but the physics of their formation is not well understood. These experiments, can be modeled using the equations governing flow in deformable porous media, however, and thus provide an important opportunity to both validate this theory and to gain a better understanding of the rates and processes of melt-band formation in the Earth. Here, we present linear and numerical analysis of these equations and compare them to results of the experiments. The important feature of the experiments is that the melt-rich bands develop at small strains and persist at low angles to the plane of shear ($\sim15$--25$\deg$) even at large shear strains. They also appear to form localized weak regions that act as strain guides for the solid matrix flow. To model these, we consider the evolution of a deformable, permeable solid undergoing simple shear with a variable shear viscosity that weakens with increasing porosity. The linear analysis calculates the growth in porosity of a plane wave perturbation that starts at an initial angle $\theta_0$ to the plane of shear and grows with increasing strain. The perturbations grow exponentially with a rate that depends on the strain and the derivative of shear viscosity with respect to porosity $\alpha=\partial \eta/\partial\phi$. For a strain of 300%, the maximum growing melt band initiates at 16.8$\deg$ (but rotates to $\sim70\deg$ in the linear analysis). We also calculate the additional solid shear induced by the localized weak regions and show that it develops a sense of shear consistent with observations only for melt bands less that 45$\deg$. These results suggest that low angle bands are favored under shear, consistent with the observations. The linear analysis, however, does not allow the growth of the melt bands to interact with the enhanced shear and thus the initially low angle bands are rotated to higher angles by the background simple shear. To follow the further development of the bands requires solutions of the full non-linear equations and we present numerical solutions that also show the spontaneous development of melt rich bands. These bands tend to form near $45\deg$ and rapidly saturate until the inter-band regions are compacted dry (which currently halts the solution). Future work will explore the quantitative differences between the numerical solutions and experiments and explore additional physics such as surface energy and grain-boundary interactions that may also be important for a full description of melt-band formation.
V24A-03 16:35h
The Effect of Melt Pressure on the Rheology of Compacting, Partially Molten Peridotite
The rheology of partially molten rock controls rock strength beneath spreading centers, deformation of the mantle wedge under subduction zones, and migration of melt to hot spots and volcanic arcs. Our understanding of these regions has been predominately shaped by chemical analyses of rocks and by remotely collected geophysical data. To interpret these data, however, requires knowledge of the relationships among deformation, melt topology, and melt migration. Most previous experimental studies of these relationships in partially molten rocks were conducted using undrained experiments, i.e., where melt cannot leave the matrix during deformation. For this configuration, melt pressure is inferred to roughly equal the minimum principle stress, but is actually unknown. By contrast, we have performed drained tests in which both melt pressure and compaction rates were measured independently. First, samples were synthesized by hot-isostatic pressing (HIP) fine-grained olivine power (10-38 $\mu$m) with a prescribed amount of mid-ocean ridge basalt (MORB) powder ($<$ 15 $\mu$m) in a gas-medium apparatus at $1200\deg$C and 300 MPa for 10 hours. Melt fractions (MORB contents) ranged from 0-30 $%$. Subsequently, samples were reinserted into the apparatus and deformed in the standard triaxial configuration. Melt flow out of the sample was accommodated by a glassy carbon bead reservoir (grain size 80-200 $\mu$m) located above the sample. A small alumina cylinder centered within the reservoir transferred the load from the pistons to the sample. Melt pressure was controlled by regulating the pressure of argon gas in contact with the melt in the reservoir. Sample compaction was measured by recording the position of a piston with the pore pressure generator. Variations of melt fraction on the strength of drained samples at P$_{m}$ = 30 or 50 MPa, where P$_{m}$ is the melt pressure, affect strength in the same way as previously observed under undrained conditions. Within the uncertainty of our measurements, changing P$_{m}$ from 30 to 50 MPa did not affect the strength of the rock. However, undrained rocks (P$_{m}$ = 300 MPa) are approximately 5 times weaker than those with the same initial melt fraction, but deformed at P$_{m}$ =30 MPa. Finally, compaction rates of drained triaxially loaded samples are significantly greater than those under isostatic compressive loads (i.e., hydrostatic melt extraction). This last result suggests that shear-enhanced compaction may play in important role in deformation and melt extraction at oceanic spreading centers and the mantle wedge of subduction zones.
V24A-04 16:50h
An experimental study of the kinetics of lherzolite reactive dissolution: Implications for contrasting styles of melt transport in the mantle.
It has been suggested that dunite dikes or veins found in harzburgite and lherzolite hosts in the mantle sections of ophiolites are high porosity channels through which basaltic magmas were extracted from their source regions. The formation of such channels may involve pervasive melt flow and reactive dissolution. In order to better understand the kinetics of reactive dissolution we conducted two series of lherzolite dissolution experiments: one in an alkali basalt and the other in a basaltic andesite. Dissolution experiments were run at 1300$\deg$C and 1 GPa using lherzolite-melt reaction couple method. The lherzolite dissolution experiments produce a reactive boundary layer (RBL) that consists of distinct lithological units separated by sharp mineralogical interfaces. The details of the RBL depend on the relative stabilities of the lherzolite minerals with respect to the reacting melt. Dissolution of lherzolite in the basaltic andesite resulted in 2 distinct regions: harzburgite (45% ol, 45% opx, 10% melt) and lherzolite (45% ol, 35% opx, 12% cpx, 8% melt). In contrast, dissolution of lherzolite in the alkali basalt resulted in 3 distinct rock units: dunite (75% ol, 25% melt), harzburgite (60% ol, 30% opx, and 10% melt), and lherzolite (50% ol, 30% opx, 10% cpx, 10% melt). The average grain size of the dunite is greater than the average grain size of unreacted lherzolite, whereas the average grain size of the harzburgite in the two sets of dissolution experiments are nearly the same as the average grain size of the lherzolite. This implies that the permeability of the dunite is larger than either the newly created harzburgite or the unreacted lherzolite, and that the permeabilities of the harzburgite and lherzolite are about the same within the DHL sequence. Hence dunite dikes in the mantle are capable of serving as melt channels, whereas harzburgites may not. Systematic compositional variations in the interstitial melt, olivine, and to a lesser extent, pyroxenes in the RBL as functions of distance and time were also observed in the two sets of experiments. The Mg\# of olivine, for example, increases from 88 at the harzburgite-melt interface to 91 at the harzburgite-lherzolite interface in the basaltic andesite-lherzolite dissolution experiment. The systematic variations in mineralogy and mineral chemistry resulted from preferential dissolution of cpx and opx and precipitation of olivine in the alkali basalt dissolution experiments, and dissolution of cpx and precipitation of opx in the basaltic andesite dissolution runs. Results of our lherzolite dissolution experiments underscore the importance of reacting melt composition in determining the lithology and composition of the mineralogical regions developed during melt-rock reaction and in controlling the style of melt transport in the mantle. If a magma is only olivine saturated then a high permeability dunite channel will develop. If, on the other hand, the magma is more silica-rich or multi-saturated with olivine and opx, only harzburgite will be produced upon reaction with the host lherzolite. We suggest that reaction between slab-derived or mantle wedge-derived magmas with lherzolite may not form high permeability channels and therefore different styles of melt migration mechanism are required to transport such magmas through the mantle wedge.
V24A-05 17:05h
Compositional variations across a dunite - harzburgite - lherzolite - plagioclase lherzolite sequence at the Trinity ophiolite: Evidence for multiple episodes of melt flow and melt-rock reaction in the mantle.
In the preceding report we showed experimentally that the dunite-harzburgite-lherzolite (DHL) sequence found in the mantle sections of ophiolite could be formed by reactive dissolution of lherzolite in a basaltic liquid. The most striking results of our lherzolite dissolution experiments are the sharp mineralogical boundaries between adjacent lithologies and simple monotonic composition variations in minerals across the DHL sequence. Here we present a detailed compositional traverse across a dunite (3.64 m wide) - harzburgite-lherzolite (5.64 m) - plagioclase lherzolite ($>$ 10 m) sequence (referred to as DHL-PL) at the Trinity ophiolite that shows complicated composition trends and melt flow history. With the exception of a small (1 m wide) anomalous region within the dunite, less than 1 m away from the dunite-harzburgite contact, the Mg\#s of olivine (90), cpx (92.8), opx (90.4 in harzburgite) and spinel (40), as well as Al$_{2}$O$_{3}$ and TiO$_{2}$ abundance in cpx, opx and spinel are essentially constant from dunite to lherzolite. The CaO content in olivine (0.02%), opx (1%) and cpx (23.5%) are also uniform throughout the harzburgite-plagioclase lherzolite sequence. However, the Mg\# of olivine and opx, Al$_{2}$O$_{3}$, TiO$_{2}$ and Cr$_{2}$O$_{3}$ in cpx and opx, as well as Na$_{2}$O in cpx increase 2 meters into the plagioclase lherzolite. In addition, asymmetric concentration gradients are observed for CaO in olivine and Cr$_{2}$O$_{3}$, Al$_{2}$O$_{3}$, MgO, and FeO in spinel. These asymmetric concentration gradients are mostly in the dunite-side of the dunite-harzburgite contact. And finally, the 1 m wide anomalous region within the dunite is characterized by elevated Mg\# and NiO in olivine, Al$_{2}$O$_{3}$, TiO$_{2}$, Cr$_{2}$O$_{3}$, and REE in cpx, and very distinct elemental abundance in spinel. The composition variations reported here are substantially different from those of Quick (1981) who measured a smaller (1 m) DHL-PL sequence at the Trinity ophiolite. Together these two Trinity data sets show a large variation in DHL-PL chemistry within the same ophiolite. Concentration gradients across DHL-PL sequences have been observed in the mantle sections of ophiolites around the world. Based on our lherzolite dissolution experiments, preliminary numerical calculations, and previous studies we suggest that the DHL-PL sequence at Trinity was formed by pervasive melt flow and reactive dissolution of a plagioclase lherzolite in basaltic liquids. Although the details are still unknown, the complicated concentration profiles reported in this study can be explained by a model that involves multiple episodes of melt flow and melt-rock reaction in an evolving dunite channel system. Multiple episodes of melt flow in the mantle, each with distinct elemental and isotopic characteristics, have already been documented in olivine-hosted melt inclusions. The spatial distributions of the compositional variations reported here can be used to further constrain the time interval between different episodes of melt flow in the dunite channel. For example, the two (or more) episodes of melt flow that created the anomalous region in the dunite and the DHL-PL sequence could at most be separated by less than a few hundred years.
V24A-06 17:20h
A Theoretical Study on the Dissolution Mechanisms of Forsterite
{\it Ab initio} quantum chemistry and molecular dynamics simulation methods (including both classic and {\it ab initio} MD) have been used to study the dissolution progress of forsterite and other olivine series minerals at the molecule-scale. Several very interesting points were found from this study: (1) Protonation of bridging oxygen (BO) of a forsterite surface will lengthen the Mg-O bond but not necessarily break it. Mg-O bond breaking will take place as a result of more dramatic thermal fluctuations. However, at the edges or kinks of the surface, protonation is capable of directly making the Mg-O bond break. Hence the dissolution rate can be much faster at such kink or edge sites. (2) The Mg-O breaking does not take place on the first Mg-O layer. It happens on the second or third layers in the mineral bulk structure. Similar results were reported recently (e.g. Rustad, et al. 2004). This means the cations far from ($>$ 4 \AA) the surface may have strong effects on the dissolution process. (3) The slowest step of the dissolution is the dissociation of Si(OH)$_{4}$ -Mg(OH$_{2}$)$_{5}$$^{2+}$. Hence, it is the controlling step. (4) During dissolution, Mg dissociates from forsterite by forming Mg(OH$_{2}$)$_{6}$$^{2+}$ and Si dissociates from forsterite by forming neutral Si(OH)$_{4}$(aq). Every time the Si-O(H)-Mg linkage breaks, Si gets the OH while Mg combines with a water molecule from the solution. This way is more energetically favorable to further dissolution. It reduces the positive charge on the surface and hence helps to the further protonation process. (5) Different dissolution rates of the olivine series minerals can be explained by differences in the Si(OH)$_{4}$-M$^{2+}$(H$_{2}$O)$_{5}$ bond strengths.
V24A-07 17:35h
Diffusion of Ca in San Carlos Olivine at 800 to 1200 $\deg$C
Calcium is a minor constituent in magnesian olivine that has potential as both a geothermobarometer [1] and a geospeedometer [2]. Fully exploiting the potential of Ca in olivine requires knowledge of it's diffusivity as a function of temperature. Here we present results from new experiments in which the diffusivity of Ca in San Carlos olivine has been experimentally determined over a temperature range of 400 $\deg$C. These experiments demonstrate that a significant anisotropy develops at temperatures below 950 $\deg$C, with diffusion parallel to the {\it c} crystallographic axis becoming significantly faster than parallel to either the {\it a} or {\it b} axes. We conclude that this anisotropy is due to the influence of low angle subgrain boundaries. Experiments were carried out on oriented pieces of gem-quality San Carlos olivine at 1 bar and temperatures of 800 $\deg$ to 1200 $\deg$C using the powder-source technique. The fugacity of oxygen was controlled at the NiNiO buffer. Diffusion profiles generated at 800 $\deg$ to 1000 $\deg$C were analyzed using Rutherford Backscattering Spectroscopy. Analyses of run products from higher temperature experiments were carried out using either Secondary Ion Mass Spectrometry or electron microprobe. At temperatures of 950 $\deg$C or greater, our experiments show no evidence for significant anisotropy with respect to diffusion. Activation energies for diffusion parallel to the {\it a} (440$\pm$60 kJ/mol), {\it b} (457$\pm$14 kJ/mol), and {\it c} (520$\pm$90 kJ/mol) crystallographic axes are indistinguishable. However, at temperatures below 950 $\deg$C the activation energy for diffusion parallel to the {\it c} axis decreases to only 270$\pm$40 kJ/mol. This change occurs only for diffusion parallel to {\it c}, resulting in considerable anisotropy. At 800 $\deg$C diffusion parallel to the {\it c} axis is faster than diffusion parallel to either {\it a} or {\it b} by more than an order of magnitude. The change in diffusivity parallel to {\it c} is attributed to the presence of (100) tilt boundaries, producing a regime in which type B kinetics dominate. Diffusion is anisotropic owing to pipe diffusion along the cores of (010)[100] edge dislocations within the low angle boundaries. The cores of edge dislocations in the (010)[100] system are parallel to the c (i.e., [001]) direction. References: [1] K\"{o}hler and Brey (1990) Geochim Cosmochim Acta 54:2375-2388. [2] Pan and Batiza (2002) J Geophys Res DOI 10.1029/2000JB000435.