skip to content

# Timetable (MIMW02)

## From the grain to the continuum: two phase dynamics of a partially molten, polycrystalline aggregate

Monday 11th April 2016 to Friday 15th April 2016

 09:00 to 09:50 Registration 09:50 to 10:00 Welcome from John Toland (INI Director) INI 1 10:00 to 11:00 David Kohlstedt (University of Minnesota)From stress-driven to reaction-driven melt segregation – the frog’s eye view (1) Co-authors: Matej Pec (University of Minnesota), Ben Holtzman (Columbia University)Separation of melt from residual solid and its migration from beneath a mid-ocean ridge to Earth’s surface require a transition from porous to channelized flow in order to preserve chemical and radiogenic disequilibrium. Chemically isolated, high-permeability melt conduits in Earth’s mantle develop as a consequence of instabilities in the deformable and dissolvable porous media. Models for the formation of such flow instabilities include stress-driven and reaction-driven melt channelization.Melt rising from depth becomes under saturated in pyroxene with respect to the surrounding upper mantle; thus, pyroxene dissolves into the melt as it migrates toward the surface. Tabular rocks rich in olivine and depleted in pyroxene found in peridotite massifs serve as channels for rapid melt extraction from the mantle. Formation of such dunite channels involves dissolution-precipitation reactions between mantle rock and percolating reactive melt. Dunite channels also coincide with shear zones, indicating that deformation together with reaction plays an important role during melt channelization.Understanding stress-driven and reaction-driven melt segregation processes requires a close coupling of experiment with theory. My talk focuses on results from laboratory investigations of the formation and evolution of melt-enriched channels in mantle rocks. (i) The first part examines the formation of stress-driven, melt-enriched channels predicted by theory. Here, the experimentally observed alignment of channels motivated further development in theory. (ii) The second part considers experimental investigations of reaction infiltration instabilities in mantle rocks. In partially molten rock samples composed of olivine and pyroxene sandwiched between a source of reactive (pyroxene under-saturated) melt and a porous sink, finger-like melt-enriched channels composed of olivine + melt propagated into and often through the rock in response to a gradient in fluid pressure. INI 1 11:00 to 11:30 Morning Coffee 11:30 to 12:30 David Kohlstedt (University of Minnesota)From stress-driven to reaction-driven melt segregation – the frog’s eye view (2) Co-authors: Matej Pec (University of Minnesota), Ben Holtzman (Columbia University)Separation of melt from residual solid and its migration from beneath a mid-ocean ridge to Earth’s surface require a transition from porous to channelized flow in order to preserve chemical and radiogenic disequilibrium. Chemically isolated, high-permeability melt conduits in Earth’s mantle develop as a consequence of instabilities in the deformable and dissolvable porous media. Models for the formation of such flow instabilities include stress-driven and reaction-driven melt channelization.Melt rising from depth becomes under saturated in pyroxene with respect to the surrounding upper mantle; thus, pyroxene dissolves into the melt as it migrates toward the surface. Tabular rocks rich in olivine and depleted in pyroxene found in peridotite massifs serve as channels for rapid melt extraction from the mantle. Formation of such dunite channels involves dissolution-precipitation reactions between mantle rock and percolating reactive melt. Dunite channels also coincide with shear zones, indicating that deformation together with reaction plays an important role during melt channelization.Understanding stress-driven and reaction-driven melt segregation processes requires a close coupling of experiment with theory. My talk focuses on results from laboratory investigations of the formation and evolution of melt-enriched channels in mantle rocks. (i) The first part examines the formation of stress-driven, melt-enriched channels predicted by theory. Here, the experimentally observed alignment of channels motivated further development in theory. (ii) The second part considers experimental investigations of reaction infiltration instabilities in mantle rocks. In partially molten rock samples composed of olivine and pyroxene sandwiched between a source of reactive (pyroxene under-saturated) melt and a porous sink, finger-like melt-enriched channels composed of olivine + melt propagated into and often through the rock in response to a gradient in fluid pressure. INI 1 12:30 to 13:30 Lunch @ Wolfson Court 13:30 to 14:30 Wenlu Zhu (University of Maryland, College Park)Physical Properties of Partially Molten Rocks from Microtomography Experiments and Digital Rock Physics Co-authors: Kevin Miller (Stanford University), Laurent Montesi (University of Maryland), Glenn Gaetani (Woods Hole Oceanographic Institution)Better constraints on rates of melt migration within the partially molten regions of the upper mantle are required to advance our current understanding of various dynamic processes at ocean ridges. In this study, we synthesized texturally equilibrated mono- and polymineralic aggregates containing various amounts of partial melt and characterized the 3-dimensional (3-D) distribution of melt using synchrotron-based x-ray microtomography. With the availability of the high-resolution 3D melt distribution, we developed digital rock physics models to calculate the physical properties of partially molten rocks. We focus on the characteristic change in melt geometry as a function of melt fraction and lithological variation, and how they affect the transport and elastic properties. Our results indicate that 1) the permeability and melt fraction are related by a power-law relation with an exponent of ~2.7 and geometric factor of 57±28 (Miller et al., 2014); 2) the bulk electrica l conductivity also follows a power-law relationship with melt fraction, with the exponent is ~1.3 and the geometric factor 0.66±0.06 (Miller et al., 2015); 3) in a partially molten rock, in general, the fluid pathways differ from, and are more tortuous than the electric current pathways; 4) lithological melt partitioning is observed: the presence of pyroxene causes melt focusing in olivine-rich regions of partially molten harzburgite. We quantified the effect of lithological partitioning on transport properties. INI 1 14:30 to 15:30 Ken Golden (University of Utah)Multiscale analysis of sea ice - a partially melted, polycrystalline composite material Earth's sea ice packs are key players in the climate system and critical indicators of climate change, as evidenced by the recent precipitous losses of summer Arctic sea ice. As a material, frozen sea water is a polycrystalline composite of a pure ice matrix containing brine inclusions - the melt phase - whose volume fraction and connectivity depend strongly on temperature. The brine phase undergoes a percolation threshold at a critical temperature where the inclusions coalesce to form channels through which the melt phase can flow. Fluid transport through sea ice mediates key climatological and biological processes, and can enhance thermal transport via convection in the porous microstructure. During the Arctic melt season, the sea ice surface is transformed from vast expanses of snow covered ice to complex mosaics of ice and melt ponds. Sea ice albedo, a key parameter in climate modeling, is largely determined by melt pond evolution. As the ponds grow and coalesce, the melt phase undergoes a percolation threshold and the fractal dimension of the pond boundaries transitions from 1 to about 2 around a critical pond size. In the two lectures, I will discuss mathematical models of composite materials and statistical physics that we have been using to describe the effective fluid, thermal, and electromagnetic transport properties of sea ice, and to address other problems in sea ice physics such as melt pond evolution. I will cover a range of mathematical techniques, some of which may possibly shed light on similar questions for partially molten rock. They include percolation theory, multiscale homogenization, integral representations for effective transport coefficients of composite media, spectral measures and random matrix theory, homogenization for advection diffusion processes, and Ising models. These models have been developed in conjunction with our field experiments in the Arctic and Antarctic. A short video on a recent Antarctic expedition will be shown. INI 1 15:30 to 16:00 Afternoon Tea 16:00 to 17:00 Ken Golden (University of Utah)Multiscale analysis of sea ice - a partially melted, polycrystalline composite material Earth's sea ice packs are key players in the climate system and critical indicators of climate change, as evidenced by the recent precipitous losses of summer Arctic sea ice. As a material, frozen sea water is a polycrystalline composite of a pure ice matrix containing brine inclusions - the melt phase - whose volume fraction and connectivity depend strongly on temperature. The brine phase undergoes a percolation threshold at a critical temperature where the inclusions coalesce to form channels through which the melt phase can flow. Fluid transport through sea ice mediates key climatological and biological processes, and can enhance thermal transport via convection in the porous microstructure. During the Arctic melt season, the sea ice surface is transformed from vast expanses of snow covered ice to complex mosaics of ice and melt ponds. Sea ice albedo, a key parameter in climate modeling, is largely determined by melt pond evolution. As the ponds grow and coalesce, the melt phase undergoes a percolation threshold and the fractal dimension of the pond boundaries transitions from 1 to about 2 around a critical pond size. In the two lectures, I will discuss mathematical models of composite materials and statistical physics that we have been using to describe the effective fluid, thermal, and electromagnetic transport properties of sea ice, and to address other problems in sea ice physics such as melt pond evolution. I will cover a range of mathematical techniques, some of which may possibly shed light on similar questions for partially molten rock. They include percolation theory, multiscale homogenization, integral representations for effective transport coefficients of composite media, spectral measures and random matrix theory, homogenization for advection diffusion processes, and Ising models. These models have been developed in conjunction with our field experiments in the Arctic and Antarctic. A short video on a recent Antarctic expedition will be shown. INI 1 17:00 to 18:00 Welcome Wine Reception
 09:00 to 10:00 Grigorios Pavliotis (Imperial College London)Homogenization methods (1) INI 1 10:00 to 11:00 Grigorios Pavliotis (Imperial College London)Homogenization methods (2) INI 1 11:00 to 11:30 Morning Coffee 11:30 to 12:30 Yasuko Takei (University of Tokyo)Constitutive mechanical relations of a partially molten rock in terms of grain boundary contiguity: an approach with an internal state variable Mechanical and transport properties of a partially molten rock strongly depend on the grain scale melt geometry. To quantify the microstructural effects, constitutive mechanical relations for elasticity (Takei, 1998) and diffusion creep viscosity (Takei and Holtzman, 2009) are derived theoretically by considering a realistic microstructural model. The essential geometrical factor which determines these properties was found to be the grain boundary contiguity’’ which represents the area of grain-to-grain contacts relative to the total surface area of each grain. One of the most striking results is that while contiguity affects both elasticity and viscosity, the effect on viscosity is about 100 times larger than that on elasticity. When partially molten rock is texturally equilibrated, contiguity is determined as a function of melt fraction and dihedral angle. However, when it is deformed under a deviatoric stress, contiguity deviates from the equilibrium value an d evolves, resulting in a significant change in the matrix viscosity. Possible consequences of these microstructural evolution on the macroscopic dynamics can be studied within the framework of continuum mechanics by solving the governing equations of two phase flow together with the viscous constitutive relation which includes contiguity as an internal state variable. By applying this approach to the formation of stress-driven, melt-enriched channels, I will demonstrate the important role of microstructural processes in the macroscopic dynamics. INI 1 12:30 to 13:30 Lunch @ Wolfson Court 13:30 to 14:30 Ulrich Faul (Massachusetts Institute of Technology)Experimental constraints on melt distribution and its effect on the rheology and seismic properties of polycrystalline olivine Coauthors: Gordana Garapić (SUNY New Paltz), Ian Jackson (Australian National University) Bulk properties of partially molten rocks are significantly affected by the melt distribution and geometry. Surface energy minimisation determines the melt geometry, both locally at the junction of two grains and melt (dihedral angle), and for the aggregate as whole, in the form of grain growth. Grain growth is a continuous process and partially molten rocks therefore constitute a dynamic system. This contrasts with the assumptions of the model melt distribution in a static system with isotropic grains of a single size that can be fully characterised by measuring dihedral angles. In the dynamic system the local melt geometry is not always in its minimum energy configuration, as grain growth continuously creates new neighbours that need to adjust their grain boundary plane orientations. In this dynamic system, the melt distribution can not be characterised by only measuring dihedral angles.  As a somewhat more comprehensive assessment of the melt distribution, we measure the proportion of grain boundaries wetted by melt (grain boundary wetness/contiguity). While deformation experiments in the diffusion creep regime by necessity need to be carried out on fine-grained samples, the melt distribution can be determined on significantly more coarse-grained samples, hot-pressed at high temperatures up to two weeks in a piston cylinder apparatus. The wetness data from these samples, obtained at suitably high resolution, allows augmentation of the experimentally measured diffusion creep rheology. For the direct experimental determination of the effect of small amounts of melt on both large-strain rheology and seismic properties it is important to characterise genuinely melt-free materials for reference. For this reason we use synthetic Fo90 olivine aggregates that contain no melt or trace elements, unless deliberately added. Experiments with melt-bearing samples show that both the large strain rheology and seismic properties are substantially affected by small amounts of melt, consistent with the observations of the melt distribution described above. For seismic properties the presence of melt affects both the shear modulus and attenuation in the seismic frequency band. Important for the effect of melt on seismic properties are wetted grain boundaries with sufficiently low aspect ratio for local fluid flow to take place (i.e. on the scale of a single grain, ’melt squirt’). Similar materials are also used to determine the effect of water (hydroxyl) on the rheology and seismic properties of olivine. This allows comparison of the relative effects of water and melt in the upper mantle. INI 1 14:30 to 15:30 Sash Hier-Majumder (Royal Holloway, University of London)Cross Scale Modeling of Melt Migration The migration of melt over length scales of hundreds of kms takes place through grain edge tubules and films with typical dimensions of a few hundred nanometers to a few microns. The volume fraction, shape, and distribution of these tubules and films exert a strong influence on the effective physical properties, anisotropy, and the trajectory of melt migration in the mantle. In this talk, I will outline new techniques for modeling the microstructure in partially molten rocks,their influence on the effective physical properties, and the implications for large scale magma flow. INI 1 15:30 to 16:00 Afternoon Tea 16:00 to 17:00 Leila Hashim (Université d'Orléans)Reconciling macroscopic olivine grain growth with the microscopic physical properties of the intergranular medium Co-authors: Gardés Emmanuel (CNRS - Université Caen), Sifré David (CNRS - Université d'Orléans), Morales Luiz F.G. (GFZ Potsdam), Précigout Jacques (CNRS - Université d'Orléans), Gaillard Fabrice (CNRS - Université d'Orléans)Grain size is a critical parameter for the understanding of our planet’s mantle since it has considerable impact on seismological properties, on the permeability of mantle rocks and therefore on melt migration characteristics of Earth’s interior. Grain growth therefore becomes an important process that should be meticulously determined in order to better understand the dynamics of our planet’s interior.Olivine grain growth has therefore been experimentally determined by a wide number of grain growth studies where different effects have been considered (water, fO2, melt, secondary phases). However, no clear consensus on the values of the different material-dependent parameters in the empirical law has been reached. To increase the existing database on olivine grain growth, we experimentally investigated the effect of melt (from nominally melt-free to 12 wt.% melt) and water on San Carlos olivine under different pressure-temperature-duration conditions (0.3 GPa < P < 3.0 GPa, 1200°C < T < 1350°C, 1h < t < 360h).By combining the existing database on olivine grain growth and our experimental data, we have succeeded in modeling (i) genuinely dry olivine grain growth aggregates, through grain boundary diffusion-controlled processes, (ii) H2O-oversaturated olivine aggregates and (iii) melt-bearing olivine aggregates, from nominally melt-free to ∼ 50 wt.% melt. Different important parameters have been constrained by using our model, namely the dry effective grain boundary thickness (δ ∼ 6 nm), melt contents in nominally melt-free samples (Φ ≤ 0.5 wt.%) as well as the wetting properties (Ψ) of melt as a function of melt content, pressure and temperature. We expect that our results will not only have considerable implications on the grain size-dependent deformation mechanisms of mantle rocks but also reconcile macroscopic observations to microscopic-scale key processes governing the mantle behavior, particularly in intergranular zones impregnated by low melt content s. INI 1
 09:00 to 10:00 Grigorios Pavliotis (Imperial College London)Homogenization methods (3) INI 1 10:00 to 11:00 Grigorios Pavliotis (Imperial College London)Homogenization methods (4) INI 1 11:00 to 11:30 Morning Coffee 11:30 to 12:30 Gideon Simpson (Drexel University)Application to McKenzie model (1) INI 1 12:30 to 13:30 Lunch @ Wolfson Court 13:30 to 14:30 Gideon Simpson (Drexel University)Application to McKenzie model (2) INI 2 14:30 to 15:30 Yann Capdeville (CNRS (Centre national de la recherche scientifique)); (Université de Nantes)Non-periodic homogenization for seismic forward and inverse problems The modeling of seismic elastic wave full waveform in a limited frequency band is now well established with a set of efficient numerical methods like the spectral element, the discontinuous Galerking or the finite difference methods. The constant increase of computing power with time has now allow the use of seismic elastic wave full waveforms in a limited frequency band to image the elastic properties of the earth. Nevertheless, inhomogeneities of scale much smaller the minimum wavelength of the wavefield associated to the maximum frequency of the limited frequency band, are still a challenge for both forward and inverse problems. In this work, we tackle the problem of elastic properties and topography varying much faster than the minimum wavelength. Using a non periodic homogenization theory and a matching asymptotic technique, we show how to compute effective elastic properties and local correctors and how to remove the fast variation of the topography. The implications on the homogenization theory on the inverse problem will be presented. INI 2 15:30 to 16:00 Afternoon Tea
 10:00 to 11:00 Harro Schmeling (Goethe-Universität Frankfurt)Physics of mantle melting: two-phase flow, variable matrix viscosity and density effects In the introduction different partially molten regions within the earth's mantle will be identified. Then, the governing equations are introduced with emphasis on rheology, melt density, and solution strategies. The melt fraction and its geometrical distribution has an important influence on the shear and bulk viscosity of the matrix. A new semi-analytical model is introduced which may describe the geometrical distribution of the melt phase. Combined with a poro-elastic effective medium approach (Schmeling et al., 2012) effective shear and bulk viscosity can be estimated as a function of melt fraction. Models of 2D porosity waves are shown which use such effective viscosity laws. Another important quantity is the melt density which may be higher than the matrix density at transition zone depths. 1D models of a rising hot partially molten plume show that within a certain parameter regime standing porosity waves may develop. If there is time, I will briefly mention a simple mantle convection benchmark initiative with two-phase flow in its partially molten region. Schmeling, H., J.-P. Kruse, and G. Richard, 2012: Effective shear and bulk viscosity of partially molten rock based on elastic moduli theory of a fluid filled poroelastic medium. Geophys. J. Int., doi: 10.1111/j.1365-246X.2012.05596.x INI 1 11:00 to 11:30 Morning Coffee 11:30 to 12:30 Ralph Showalter (Oregon State University)Multiscale Systems for Flow and Transport An elliptic-parabolic system of partial differential equations describes the flow of a single-phase incompressible fluid and the transport of a dissolved chemical by advection and diffusion through a heterogeneous porous medium.  The objective is to develop an upscaled model of this system which represents the full range of scales observed.  After a review of homogenization results for the traditional low contrast and the $\epsilon^2$-scaled high contrast cases, the new discrete upscaled model based on local affine approximations is constructed. It reproduces the full range of scale contrasts observed in experiments. INI 1 12:30 to 13:30 Lunch @ Wolfson Court 13:30 to 14:30 Claude le Bris (ENPC - École des Ponts ParisTech); (INRIA Paris - Rocquencourt)Stochastic homogenization (1) INI 1 14:30 to 15:30 Claude le Bris (ENPC - École des Ponts ParisTech); (INRIA Paris - Rocquencourt)Stochastic homogenization (2) INI 1 15:30 to 16:00 Afternoon Tea 16:00 to 17:00 Stephen Morris (University of Toronto)The rippling instability of icicles Co-authors: Jake Wells (Dept. of Physics, University of Toronto, Toronto ON Canada M5S 1A7), Alina Barnett (Dept. of Physics, McMaster University of Toronto, Hamilton ON Canada L8S 4M1), Josh Calafato (Dept. of Physics, University of Toronto, Toronto ON Canada M5S 1A7), Ken Liao (Dept. of Physics, University of Toronto, Toronto ON Canada M5S 1A7), Antony Szu-Han Chen (Southern Alberta Institute of Technology, Calgary AB Canada T2M 0L4), John Ladan (Dept. of Physics, University of Toronto, Toronto ON Canada M5S 1A7) Icicles are a common ice formation, familiar to anyone who lives in a cold climate. The shape of an icicle emerges from a delicate dance between solidification, hydrodynamics and heat transport. Many, but not all, natural icicles are observed to be decorated around their circumference by ribs or ripples. These features are presumed to be the result of a morphological instability in the growth process of the ice. The sides of an icicle are covered by a thin supercooled water film which flows down their nearly vertical surface. The wavelength of the ripples, which is always found to be near 1~cm, is surprisingly constant, even under diverse growing conditions. A recent detailed study in which hundreds of icicles were grown in controlled laboratory experiments revealed that trace amounts of impurities are required for the formation of the ripples. Icicles grown from distilled water have no ripples. Ripples appear at a remarkably low concentration of impurity, becoming me asurable above a concentration of just 10−3 weight \% of salt. Thereafter, they grow at a rate which is roughly logarithmic in the concentration of the impurity. These effects are not explained by linear stability theory which does not account for impurities. In this talk, we will discuss our recent experiments in which the concentration and molecular species of the impurity were varied, as well as our progress toward a generalized linear stability analysis of the growing ice surface, which includes the effects of impurities. The theory crucially depends on the boundary conditions on the ice-water interface and the possible presence of a mushy layer near this interface. Related Links http://www.physics.utoronto.ca/Icicle_Atlas/ - Open source data archive http://dx.doi.org/10.1088/1367-2630/15/10/103012 - publication INI 1 19:30 to 22:00 Conference Dinner at Corpus Christi College
 10:00 to 11:00 Claude le Bris (ENPC - École des Ponts ParisTech); (INRIA Paris - Rocquencourt)Stochastic homogenization (3) INI 1 11:00 to 11:30 Morning Coffee 11:30 to 12:30 Claude le Bris (ENPC - École des Ponts ParisTech); (INRIA Paris - Rocquencourt)Stochastic homogenization (4) INI 1 12:30 to 13:30 Lunch @ Wolfson Court 13:30 to 14:30 Neil Ribe (CNRS (Centre national de la recherche scientifique)); (Université Paris-Sud 11)Evolution of Anisotropy in Olivine Polycrystals Progressive deformation of upper mantle rocks via dislocation creep causes their constituent crystals to take on a non-random orientation distribution (crystallographic preferred orientation or CPO) whose observable signatures include shear-wave splitting and azimuthal dependence of surface wave speeds. Comparison of these signatures with mantle flow models thus allows mantle dynamics to be unraveled on global and regional scales. However, existing self-consistent models of CPO evolution are computationally expensive when used in 3-D and/or time-dependent convection models. We propose a new method, called ANPAR, which is based on an analytical parameterization of the crystallographic spin predicted by the second-order (SO) self-consistent theory. Our parameterization runs ≈2–6\times 10^4 times faster than the SO model and fits its predictions for CPO and crystallographic spin with a variance reduction >99 per cent. We illustrate the ANPAR model predictions fo r the deformation of olivine with three dominant slip systems, (010)[100], (001)[100] and (010)[001], for three uniform deformations (uniaxial compression, pure shear and simple shear) and for a corner-flow model of a spreading mid-ocean ridge. INI 1 14:30 to 15:30 Lars Hansen (University of Oxford)The development of seismic anisotropy in partially molten rocks: Laboratory observations Co-authors: Chao Qi (University of Pennsylvania), David Wallis (University of Oxford), Benjamin Holtzman (Lamont-Doherty), David Kohlstedt (University of Minnesota)Seismic anisotropy is a key indicator of the kinematics of flow in the upper mantle. Much insight has been gained into seismic anisotropy that results from the crystallographic alignment of olivine during deformation. This anisotropy is primarily characterized by alignment of the seismically fast axis with the flow direction. This relationship between olivine anisotropy and the macroscopic kinematics allows detailed comparison between simulations of global mantle flow and seismic tomography. However, relatively little is known about the development of seismic anisotropy in partially molten rocks. Some experimental studies on partially molten rocks suggest that the seismically fast direction tends to lie at high angles to the flow direction, leading to a vastly different relationship between anisotropy and kinematics. Thus, the presence of a melt phase appears to fundamentally alter the grain-scale processes leading to crystallographic rotation of the solid phase.Here we present a new experimental data set detailing the evolution of anisotropy during deformation of partially molten peridotite. Torsion experiments were conducted on samples composed of San Carlos olivine and basaltic melt at a temperature of 1473 K and a confining pressure of 300 MPa. Seismically fast axes of olivine tend to lie at a high angle to the flow direction in a manner similar to previous experiments. The anisotropy in these samples is weak compared to that in dry, melt-free olivine deformed to similar strains. The anisotropy also exhibits relatively little change in strength and orientation with progressive deformation. Detailed microstructural analyses allow us to distinguish between competeing models for the grain-scale deformation processes, favoring one in which intergranular processes control grain rotations. Based on our observations, we extrapolate our results to flow in the oceanic upper mantle, demonstrating good correlation between predicted and obse rved seismic anisotropy. INI 1 15:30 to 16:00 Afternoon Tea 16:00 to 17:00 Andrea Tommasi (CNRS (Centre national de la recherche scientifique)); (Université de Montpellier)How do melts change texture and anisotropy of mantle rocks In a melt-free mantle, development of crystal preferred orientations (CPO or texture) in response to deformation is the major source of anisotropy of physical properties. Measurement of seismic (elastic) anisotropy is indeed the best available tool to unravel flow patterns at various depths in the mantle. Though it cannot be easily measured in situ, anisotropy is even more marked for thermal diffusion and viscosity. These anisotropies probably induce a memory-effect on the thermo-mechanical evolution of the upper mantle. In this presentation, we will address how the presence of melts may change the anisotropy of physical properties in the upper mantle. The presence of melts may: (1) induce an additional (probably stronger) component of anisotropy if the melt is concentrated in aligned pockets or lenses, (2) change the olivine texture evolution and (3) the mineralogical composition. Anisotropy due to melt alignment, though strong, is only effective while melts are present in t he system. The two latter processes induce weaker, but long-term changes in the anisotropy, which remain effective even after melt extraction or crystallization. Observations in naturally deformed peridotites suggest all three processes occur in the upper mantle. Analysis of the spatial arrangement of products of melt-rock reactions in mantle peridotites provides evidence for melt organization in planar lenses or layers parallel to the shear plane at both the grain boundary and larger (cm to tens of meters) scales. Such an arrangement may induce significant decrease in the shear viscosity parallel to the shear plane. Comparison of olivine crystal preferred orientations within and outside melt-focusing domains records changes in the deformation processes and hence on the resulting CPO-induced anisotropy, which depend on the nature of the melt-rock reactions. The latter also controls the crystallization of new minerals, which most often dilutes the anisotropy. INI 1