Shear deformation of bridgmanite and magnesiowüstite aggregates at lower mantle conditions

See allHide authors and affiliations

Science  08 Jan 2016:
Vol. 351, Issue 6269, pp. 144-147
DOI: 10.1126/science.aad3113

Mantle minerals won't share the strain

The deformation of a mixed block of material depends on the strength of the components of which it is made. Weak materials will deform more than the strong ones in a mixture that is squished or stretched. Girard et al. find a large difference in strength between the two primary minerals making up Earth's lower mantle (see the Perspective by Chen). Deformation in the convecting mantle may occur only near boundary layers as a result, leaving large regions potentially unaffected. This could explain long-lived chemical reservoirs in Earth's interior and the lack of seismic anisotropy in the lower mantle.

Science, this issue p. 144; see also p. 122


Rheological properties of the lower mantle have strong influence on the dynamics and evolution of Earth. By using the improved methods of quantitative deformation experiments at high pressures and temperatures, we deformed a mixture of bridgmanite and magnesiowüstite under the shallow lower mantle conditions. We conducted experiments up to about 100% strain at a strain rate of about 3 × 10−5 second−1. We found that bridgmanite is substantially stronger than magnesiowüstite and that magnesiowüstite largely accommodates the strain. Our results suggest that strain weakening and resultant shear localization likely occur in the lower mantle. This would explain the preservation of long-lived geochemical reservoirs and the lack of seismic anisotropy in the majority of the lower mantle except the boundary layers.

Earth’s large, rocky lower mantle is mostly composed of (Mg,Fe)SiO3 bridgmanite (~70%) and (Mg,Fe)O magnesiowüstite (~20%) (and a few percent of calcium perovskite CaSiO3) [e.g., (1)]. Many of Earth’s geochemical and geophysical questions depend strongly on the rheological properties of materials in this region. For instance, geochemical observations suggest that the lower mantle hosts a large amount of incompatible elements working as a reservoir of these elements (2, 3). The degree of preservation of these reservoirs is controlled by the nature of mixing or stirring of materials (4, 5), which strongly depends on the rheological properties of materials in this region. However, very little is currently known about the rheological properties of materials in the lower mantle because of the difficulties in quantitative experimental studies of deformation under the conditions of the lower mantle.

The main difficulties include the controlled generation of stress (or strain rate) and reliable measurements of stress and strain under the high-pressure and -temperature conditions [e.g., (6)]. Consequently, previous studies on plastic deformation of lower mantle minerals were either performed at high pressures and low temperatures (710), at high pressures and high temperatures without stress-strain rate control (11), or on analog materials at low pressures (1214). Applying low-temperature experiments to Earth’s interior is difficult because rheological properties are highly sensitive to temperature. Also the mechanisms by which deformation occurs are sensitive to temperature and strain rate, creating extrapolation issues for both low-temperature and poorly controlled strain-rate (stress) measurements [e.g., (15)]. Furthermore, microstructural evolution often leads to strain-dependent rheological behavior, which is particularly important for a sample containing two materials with a large strength contrast (16). The lower mantle approximates a two-phase mixture (bridgmanite and magnesiowüstite) with presumably a large strength contrast [e.g., (13, 17, 18)], and therefore large strain (>30%) experiments are essential to characterize the evolution of rheological behavior of the lower mantle.

We performed large strain deformation experiments on a mixture of bridgmanite and magnesiowüstite by using the rotational Drickamer apparatus (19) at the synchrotron x-ray radiation facility (fig. S1). We have made several technical developments on the cell assembly (Fig. 1) and anvil design to increase the maximum pressure and to improve the quality of x-ray signals and performed quantitative deformation experiments under the conditions of the shallow lower mantle of Earth.

Fig. 1 The rotational Drickamer apparatus (RDA) cell assembly used to reach the shallow lower mantle conditions.

(A) The anvils, gaskets, and cell assembly for the RDA deformation experiment; (B) a side view of the cell assembly; and (C) a top view of the cell assembly. Diffracted x-ray comes mostly from one side of a sample closer to the x-ray detectors.

We prepared starting materials (ringwoodite or bridgmanite and magnesiowüstite) from San Carlos olivine by using the Kawai-type multianvil press. We carried out in situ deformation experiments at the shear strain rate of ~3 × 10−5 s−1 on lower mantle mineral mixtures at pressures at 24 to 27.5 GPa and temperatures up to 2000 to 2150 K (Table 1). All the deformed samples are a mixture of bridgmanite and magnesiowüstite with 1:1 mole ratio (~70:30 volume ratio). We estimated the pressure and temperature conditions by in situ x-ray diffraction data using the equations of state of bridgmanite and magnesiowüstite. We also estimated stress from radial x-ray diffraction and strain from x-ray image analysis of a Pt strain marker sandwiched vertically between two half rings of the sample (Fig. 1).

Table 1 Summary of run conditions.

The total strain is the equivalent strain, εE, including both axial compression strain (εU) and shear strain (εS) as Embedded Image.

View this table:

Stress increases with strain, initially achieving nearly steady state at strains higher than ~40%, with some hint of strain weakening for bridgmanite (Fig. 2). Because of a small sample size, the intensity of diffraction peaks is relatively weak, and we were able to determine the stress only from a limited number of diffraction planes. We used (110) and (112) planes for bridgmanite and (220) and (200) planes for magnesiowüstite. We took an arithmetic average of stress values from different planes to estimate the strength of each material. This is a crude way to estimate the strength of a polycrystalline material, but the precise relation between the strength of various slip systems and the average strength of an aggregate is not known [e.g., (20)]. Although the errors in the estimated stress are large, the results show that bridgmanite is likely substantially stronger than magnesiowüstite.

Fig. 2 A plot of the equivalent stress in bridgmanite and magnesiowüstite as a function of strain.

Run conditions are given in Table 1. Stress in bridgmanite was estimated by using diffraction peaks (110) and (112). Stress in magnesiowüstite was estimated by using diffraction peaks (200) and (220). In both cases, the stresses shown here are the arithmetic average of stresses estimated from these planes. Some hint of strain weakening can be seen, particularly for bridgmanite (hatched regions are drawn to guide the eyes). Results from beta 74 are based on estimated strain (strain marker was not visible). Also, the pressure and temperature conditions for beta 74 are different from all others. Bars represent the errors. Errors are given for one standard deviation and are due to the uncertainties in the peak shift and the fitting errors to equation S1 (supplementary materials).

The microstructures of the deformed samples show evidence of substantial plastic deformation in both bridgmanite and magnesiowüstite (Fig. 3). Magnesiowüstite is isolated in most cases, whereas we infer bridgmanite to be interconnected on the basis of two-dimensional cross-sectional images. A large amount of strain accommodation by magnesiowüstite is evident from Fig. 3B, because magnesiowüstite is deformed more than the bulk strain. This is consistent with the conclusion drawn from the strength determination (Fig. 2). Our observations provide direct experimental verification that bridgmanite is substantially stronger than magnesiowüstite (13, 17, 18).

Fig. 3 SEM (scanning electron microscope) back-scattering images of the recovered sample from the run gamma 21.

(A) A back-scattered electron image of the RDA cell assembly cut along the diameter. The sample position, alumina ring, and TiC central electrode are labeled for clarity. The layers of material above the sample are also identified (ZrO2, Al2O3, TiC+Diamond, BN). (B) A back-scattered electron image of the recovered sample from the run gamma 21, deformed up to 100% strain. The light gray grains are mangesiowüstite, and the dark gray grains are bridgmanite. An oblate shape shows a strain ellipsoid corresponding to the bulk strain of 100%. Arrows indicate the sense of shear. (C) A back-scattered electron image of an undeformed sample from the run gamma 23. This sample was annealed at 27.4 GPa and 2140 K and quenched after 1.5 hours.

An obvious consequence of such a large rheological contrast is the evolution of strain partitioning with increasing strain, which may lead to shear localization [e.g., (16, 21)]. Shear localization would limit substantial deformation to only in or near the boundary layers. Hence, stirring and mixing would occur selectively in these regions, with the rest of the lower mantle deforming very little. Relatively undeformed regions in the lower mantle, where little mass transport takes place, may act as long-lived geochemical reservoirs suggested by geochemical studies [e.g., (2, 5)]. The possible localized deformation would also provide an alternative explanation for the absence of seismic anisotropy in the majority of the lower mantle other than the model proposed by Karato et al. (22), in which they attributed the absence of anisotropy in the majority of the lower mantle to the operation of diffusion creep.

Our conclusions are somewhat different from those by Wang et al. (14) on the CaGeO3 perovskite and MgO analog system. They deformed their samples to less than 20% strain and found that CaGeO3 perovskite is stronger than MgO by a factor of ~2. They concluded that the stronger phase CaGeO3 perovskite controls the strength of the two-phase aggregate by comparing the results with the strength of pure CaGeO3 perovskite. Although their observations on the strength contrast are similar to ours, we suggest that the larger strength contrast we observed for the lower mantle aggregate could promote shear localization more than an analog material where the strength contrast is less. Also, the difference in the strain magnitude may be responsible for the somewhat different results. Wang et al. (14) did not observe strain weakening but rather strain hardening at strain less than 20%. In that small strain regime, we also observed strain hardening. In contrast, we observed the hint of strain weakening at ~30% strain (Fig. 2). This suggests that the strength of the lower mantle aggregate could be controlled by a weaker phase at large strain. Similarly, a numerical study of deformation of a mixture of bridgmanite and magnesiowüstite by Madi et al. (23) showed no large strain partitioning to magnesiowüstite, but this study was made to very small strain (~10−4).

We must examine the validity of the necessary extrapolation in strain rate (or stress level) from laboratory data to deformation in Earth. We conclude that dislocation creep dominates under the present experimental conditions on the basis of the large variation in stress estimated for different diffraction planes (24). Seismological observations of distribution of anisotropy suggest that dislocation creep may dominate in the boundary layers in the lower mantle (25). The stress dependence of strain rate in dislocation creep regime is similar between orthorhombic perovskite and magnesiowüstite (26, 27). Consequently, the strength contrast between the two minerals deformed at geological strain rate would be similar to the contrast observed at laboratory strain rate in the boundary layer, where much of deformation occurs.

However, our current results contain major limitations, including (i) deformation mechanisms are not clearly identified, and the flow law was not determined in any detail; and (ii) the maximum strain is still low, and therefore the degree to which strain weakening could occur is unclear. In addition to well-characterized experimental studies to larger strains, characterization of strain weakening (and shear localization) may require numerical modeling, because small sample size might limit the full development of shear localization. The demonstration of significant stress-strain partitioning (and evolution) between the two major phases in the lower mantle, potentially resulting in shear localization, highlights the importance of experimental studies of deformation in interpreting geochemical and geodynamical observations.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

References (2841)

Database S1

References and Notes

Acknowledgments: This study was supported by a grant from NSF. The experimental facilities at the Brookhaven National Laboratory are supported by the Department of Energy and COMPRES (Consortium for Materials Properties Research in Earth Sciences). We thank D. Weidner and H. Chen for the support at the synchrotron facility at Brookhaven National Laboratory. Z. Jiang provided technical support for the microstructural analyses. All data presented in this paper are available in the supplementary materials.
View Abstract

Navigate This Article