Melt Segregation and Strain Partitioning: Implications for Seismic Anisotropy and Mantle Flow

See allHide authors and affiliations

Science  29 Aug 2003:
Vol. 301, Issue 5637, pp. 1227-1230
DOI: 10.1126/science.1087132


One of the principal means of understanding upper mantle dynamics involves inferring mantle flow directions from seismic anisotropy under the assumption that the seismic fast direction (olivine a axis) parallels the regional flow direction. We demonstrate that (i) the presence of melt weakens the alignment of a axes and (ii) when melt segregates and forms networks of weak shear zones, strain partitions between weak and strong zones, resulting in an alignment of a axes 90° from the shear direction in three-dimensional deformation. This orientation of a axes provides a new means of interpreting mantle flow from seismic anisotropy in partially molten deforming regions of Earth.

Most of the chemical exchange between the interior and the surface of Earth occurs at plate boundaries and above plumes. In these regions, the interactions resulting from coupled fluid migration and mantle flow must be understood in order to interpret mantle structure (from seismology) and processes of melt-rock reaction (from melt chemistry). Anisotropy in the velocity of elastic shear (S) waves and the polarization of compression (P) waves propagating through the upper mantle can be caused by preferred orientation of olivine and/or the alignment of melt pockets or melt-rich planes (1). Past and/or present flow directions in the mantle are generally assumed to be parallel to the seismically fast direction, corresponding to the a axis of olivine (24). This assumption has recently been questioned on the basis of shear deformation experiments on olivine under high stress and high water-content conditions (5). However, many partially molten mantle regions, such as spreading centers, are relatively dry, low-stress environments (6). We report the discovery of an alternative mechanism by which the parallelism between the seismic fast axis and the flow direction breaks down.

We deformed partially molten mantle rocks under anhydrous conditions. Samples were placed between pistons cut at 45° and deformed in a geometry approximating simple shear at high pressure and temperature (P = 300 MPa, T = 1523 K) in a gas-medium deformation apparatus (7, 8) (fig. S1; table S1). The samples consisted of olivine and mid-ocean ridge basalt (MORB) and olivine and MORB with either an additional melt (FeS) or solid (chromite) phase. The additional phases lower the permeability relative to that of olivine and MORB samples by partially plugging the melt channels, without substantially changing the rheological properties (7, 8). In sheared samples of olivine and MORB, melt pockets align at ∼20° to the shear plane and distribute uniformly across the sample (9). During deformation of samples with reduced permeability, melt segregates from an initially homogeneous distribution into melt-rich layers or “bands” oriented at ∼20° to the shear plane, separated by melt-depleted regions or “lenses.” (Fig. 1, A and B). When melt segregates, strain concentrates in the weak melt-rich bands. The interaction of strain partitioning and melt segregation leads to self-organization of the melt-rich bands. These bands form anastomosing networks (Fig. 1), with larger bands at higher angles connected by smaller bands at lower angles. Because the distribution of band angles remains independent of strain, the melt in the bands must be moving relative to the solid. The spacing between bands is governed by the compaction length of the sample [supporting online material (SOM) Text, section 2], consistent with theory for porous flow in a permeable viscous deforming medium (7, 10).


Microstructures. (A) Reflected-light image of a sample of olivine and chromite and 4% MORB sheared between tungsten pistons. The melt-rich bands are visible as the darker regions oriented 10° to 20° to the sample walls, forming an anastomosing network. (B) SEM backscattered electron image of one band. The chromite grains (white) are smaller than the olivine grains (gray). Chromite tends to sit in melt (black) tubules, reducing the permeability.

Crystallographic preferred orientations (CPOs) of olivine grains were measured for 1 undeformed and 10 sheared samples with electron backscatter diffraction (EBSD) analysis (8, 11). In samples of olivine and MORB deformed in shear, the CPOs develop an “axial” pattern in which the b axes concentrate normal to the sample shear plane (the walls of the pistons) and the a and c axes form girdles in the shear plane (Fig. 2A). In samples with an added third phase (olivine and MORB with either chromite or FeS melt), all of which formed networks of melt-rich bands, the CPO is characterized by (i) significant concentrations of a axes normal to the shear direction; (ii) tight clusters of b axes oriented 15° to 20° from the pole to the sample shear plane, back-rotated relative to the orientation caused by simple shear; and (iii) significant concentrations of c axes rotated 15° to 20° counterclockwise from the sample shear plane (90° from the b axes) (Fig. 2B). In six samples deformed to shear strains of γ = 1.1, 2.1, and 3.3 to 3.5, the intensity of CPOs increases with increasing strain. In addition, all samples progressively decrease in thickness by up to 20% with increasing strain, accommodated by minor but important lengthening of the sample normal to the shear direction.


Pole figures for a, b, c axes for sheared samples. The shear plane is horizontal and shear sense is at top right. (A) Sheared sample of olivine and 4% MORB. (B) Sheared sample of olivine and chromite and 6% MORB, shown in Fig. 1. The CPO measured in a sample of olivine and MORB with FeS melt is identical to this one. (C) Schematic diagrams of three sets of three pole figures for a melt-free olivine aggregate (top) [after (3)], olivine and oriented melt (middle), and olivine and segregated melt (bottom), all deformed at high T (1473 to 1573 K) and P (300 MPa) to shear strains larger than γ = 1. The reference position for the back-rotation of the CPO is indicated by the outlined areas. A similar obliquity between the shear plane and the axis aligned in the shear direction can develop due to recrystallization, discussed further in SOM Text, section 5, but is not the origin of the back-rotation observed here. The relative seismic velocities are Va > Vc > Vb. At right is a spatial representation of the third CPO.

Observations from detailed studies of CPOs and microstructures by scanning and transmission electron microscopy (SEM and TEM) constrain the microstructural processes active in the samples (Fig. 1B). A map of CPOs in the vicinity of a similar melt-rich band exhibits no significant spatial variations, suggesting that deformation in the bands does not modify the CPO (SOM Text, section 3). TEM observations reveal that dislocations with Burgers vectors parallel to the a axis outnumbered those with Burgers vectors parallel to the c axis (SOM Text, section 4). The density of dislocations is not elevated in olivine grains adjacent to stronger chromite grains, suggesting that the presence of chromite does not influence dislocation dynamics.

A commonly observed CPO for olivine deformed in simple shear at high temperature is characterized by a axes parallel to the shear direction, b axes normal to the shear direction, and c axes normal to the shear direction in the shear plane (Fig. 2C) (3) (SOM Text, section 5). This pattern reflects the fact that in dry olivine, a slip on b planes is significantly weaker than a slip on c planes, which is weaker than c slip on b planes (12). In contrast, in partially molten samples deformed at similar temperature and stress conditions as described in (3), the more diffuse CPO pattern is distinguished by a and c axis girdles in the shear plane with strong concentrations of b axes normal to the shear plane. Further, when melt segregates into bands, the b axes rotate 20° antithetic to the sense of shear, and concentrations of a and c axes appear 90° from their “normal” positions, referred to below as the “a-c switch” (Fig. 2C).

First, we must explain how the oriented melt pockets weaken a and c axis concentrations relative to CPOs from melt-free samples (Fig. 2B). Melt pockets provide a fast-diffusion path for olivine components, enhancing the importance of diffusion-accommodated creep and possibly grain boundary sliding in the direction of shear, whereas movement of dislocations accommodates the remainder of the interactions between grains, providing a mechanism for modifying the von Mises criterion (13). Strong anisotropy in melt-pocket orientation (MPO) may randomize the orientations of a and c axes, while leaving orientations of slip planes (b axes) intact (“the MPO-CPO effect”). As long as melt pockets are oriented, which should occur at shear stresses higher than 0.1 MPa (14), olivine CPOs in the mantle will be affected by the presence of melt.

Second, we argue that the transition in CPOs in going from samples of olivine and MORB (Fig. 2A) to samples with melt segregated into networks of bands (Fig. 2B) is due to changes in the flow pattern rather than a change in the deformation mechanism. The latter CPO is similar to type-B CPO defined in (5), which the authors of that paper attribute to a change in the behavior of dislocations at high water fugacity and high stress conditions. However, neither of these conditions applies to our experiments. Several lines of evidence disfavor a change in dislocation dynamics (e.g., a change from dominant a slip to c slip on b planes) as an explanation for the CPO observed in our experiments, discussed in (15). We propose a kinematic explanation for the a-c switch, on the basis of three points, as follows:

  1. ) The total strain in the sample partitions between the melt-rich bands and melt-depleted lenses. Although the bands comprise only ∼20% of the total sample volume, the strain rate, and thus the strain, is higher in the weaker bands than in the stronger lenses. In the bands, shear strain is oriented at ∼20° to the sample shear plane and, therefore, in the lenses the shear plane must be back-rotated relative to the sample shear plane (Fig. 3C). This back-rotation is observed in the orientation of b axes in the CPO.

  2. ) The observed CPO predominantly reflects deformation in the melt-depleted lenses, implying that the deformation in the bands is not contributing to or strongly modifying the CPO (16). Observations that the CPO is barely modified in the vicinity of a band (SOM Text, section 3) support this point.

  3. ) The deformation that produces the CPO in the melt-depleted lenses is not simple shear, but it involves substantial components of strain normal to the shear direction. Alignment of a axes normal to the shear direction suggests that a slip on the b plane occurs normal to the shear direction (17). Because much of the shear strain is accommodated in the bands, the components of strain normal to the shear direction in the lenses have much greater expression in the CPO than they would if the bands were absent. This expression may be enhanced further by the MPOCPO effect. Thus, the “mechanism” for a-axis orientation is the kinematic effect of strain partitioning, not a change in the dominant slip system.


Representation of observed melt distribution and internal strain partitioning in experimentally deformed samples. (A) Synthesis of the configuration of melt bands. Bands form anastomising networks with larger bands at higher angles relative to the shear plane (flat red arrow) connected by smaller bands at lower angles. Smaller arrows indicate that the samples flatten and widen with shear. In three dimensions, the melt-rich layers connect and surround melt-depleted lenses. (B) Strain partitioning between bands (anastomosing layers) and lenses. The flat arrows indicate the total shear and the component concentrated in the bands. The narrow arrows indicate alignment of olivine a axes normal to the shear direction in the lenses. The black lines mark the orientation of the shear plane in the lenses, “back-rotated” relative to the sample shear plane due to strain partitioning.

The influence of melt segregation and strain partitioning on CPO development will be even more effective in partially molten regions of Earth where deformation is more three-dimensional than in our experiments. The olivine a axes will rotate if the geometry of the overall flow in the region permits or requires an elongation of the melt-depleted lenses normal to the shear direction. An anisotropic network of melt-rich layers will affect the seismic properties of regions large in comparison to a seismic wavelength (λ) if the separation between layers (δS) is much less than a seismic wavelength (i.e., δS ≈ 100-3 m « λ ≈ 104 m). The configuration (comprising average thickness, spacing, angle, and topology) of the network depends on the physical properties of the solid and fluid, the kinetics of the processes governing their interactions and transport, and the geometric (kinematic) boundary conditions of regional flow. Some of these properties are encompassed in the first-order compaction length scaling argument discussed in (7) (SOM Text, section 2). Others remain to be studied experimentally and theoretically.

A literal extrapolation of the process discussed here may help to explain the complex seismic anisotropy observed in many partially molten regions of the upper mantle (18). Several potential examples include (i) vertical fast-direction measurements beneath the Reykjanes Ridge south of Iceland (19), (ii) trench parallel fast axes measured in the mantle wedge above subduction zones (20), or (iii) tangential patterns of anisotropy around plume heads, e.g., Iceland (21). Each of these observations has produced a range of hypotheses, a discussion of which is beyond the scope of this paper (22). However, the processes discussed here suggest detailed field-based (23) and seismological predictions and tests, which may influence our interpretations of the dynamics of partially molten regions of Earth.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S3

Table S1


References and Notes

View Abstract

Navigate This Article