Role of orthopyroxene in rheological weakening of the lithosphere via dynamic recrystallization

Significance

The theory of plate tectonics is well established for Earth; however, why it operates is less well known. Here we propose that the deformation of a realistic two-phase aggregate (olivine and orthopyroxene) contributes to the rheological weakening and localization in the lower lithosphere and upper mantle. In particular, in a certain temperature range, the experimentally deformed aggregates show substantial weakening and the partitioning of strain in fine-grained regions composed of both olivine and orthopyroxene, possibly governed by grain-boundary pinning effects. However, at higher temperatures, the polyphase aggregates deform more homogeneously and are subject to less weakening. Using a model, we predict the conditions under which localized deformation may occur in the Earth.

Abstract

For plate tectonics to operate on a terrestrial planet, the surface layer (the lithosphere) must have a modest strength (Earth, ≤200 MPa), but a standard strength profile based on olivine far exceeds this threshold value. Consequently, it is essential to identify mechanisms that reduce the strength of the lithosphere on Earth. Here we report results of high-strain laboratory deformation experiments on a representative olivine–orthopyroxene composition that show the addition of orthopyroxene substantially reduces the strength in the ductile regime within a certain temperature window. The reduction in strength is associated with the formation of small orthopyroxene and olivine grains. Our samples show heterogeneous microstructures similar to those observed in natural peridotites in shear zones: fine-grained regions containing both orthopyroxene and olivine that form interconnected bands where a large fraction of strain is accommodated. A model is developed to apply these results to geological conditions. Such a model, combined with our experimental observations, suggests that orthopyroxene may play a key role in the plastic deformation of the lithosphere in a critical temperature range, leading to long-term weakening associated with strain localization in the lithosphere.

Plate tectonic style of convection is characterized by regions of localized deformation, such as subduction zones, and the origin of these regions needs to be investigated to understand why plate tectonics operates on Earth. Localized deformation at plate boundaries involves not only brittle fracture at shallow portions but also localized deformation in the ductile shear zones (1–4). Although the basic processes of localized deformation in the brittle regime are well understood (5, 6), mechanisms of localized deformation in the ductile regime remain elusive. The strength profile of the lithosphere in a standard model (7) predicts stresses that are much higher than the critical strength below which plate tectonic style of convection would occur (8, 9).

Among the possible mechanisms of shear localization (10), grain size reduction is most often recognized in ductile shear zones and has been given particular attention in previous studies (11–15). Grain size reduction can occur during high-temperature deformation by dynamic recrystallization during dislocation creep (16, 17). In many cases, small grains are formed along preexisting grain boundaries (18). If the degree of grain-size reduction is large enough and these fine-grained regions are connected, then regions of small grain size will be deformed by grain-size–sensitive creep processes—diffusion creep or dislocation creep accommodated by grain-boundary sliding (Dis-GBS) introduced by Hirth and Kohlstedt (1995) (19). Deformation in both regimes results in the weakening of a rock, although less so in the Dis-GBS regime. In addition, for this weakening to cause substantial shear localization, the influence of grain growth must be minimal. Pinning of grain boundaries by a second phase is an obvious possibility, and therefore it is essential to understand the microstructural development during deformation of multiphase rocks. One of the key observations in naturally deformed peridotitic rocks in shear zones is the association of degree of mixing with localization: well-developed shear zones (mylonites, ultramylonites) are in most cases comprised of a well-mixed mineralogy with smaller grain size and larger volume fraction of secondary phases (3).

Previous theoretical studies have addressed various shear localization processes (20, 21). However, experimental studies are essential to make further progress in this area because the essence of shear localization is the development of heterogeneous microstructure that is difficult to be captured by theoretical studies. In this article, we describe experimental observations on the microstructural development and mechanical behavior of olivine + orthopyroxene aggregates, and we present a theoretical model to interpret the results and extrapolate them to geological conditions.

Using a Griggs apparatus, we conducted deformation experiments on hot-pressed olivine (75%) and orthopyroxene (25%) aggregates in nearly simple-shear geometry at pressures between 1.3 and 2.0 GPa and temperatures between 1,173 and 1,540 K (Fig. 1). Deformation is nearly simple shear because the sample is oriented on the plane in the direction of maximum shear stress (Fig. S1). However, because of the presence of a compressive stress normal to the sample plane, there is always a small (<20%) contribution from axial compression. We deformed samples to shear strains of 0.7–3.0 at constant displacement rates equivalent to strain rates between 3.5 × 10⁻⁵ and 2.3 × 10⁻⁴⋅s⁻¹ (Methods Summary and Materials and Methods).

Fig. 1.

Shear stress–shear strain curves for deformation at constant strain rate on a select number of key experiments. The colored curves correspond to current experiments and are labeled by run number and temperature. The gray curves correspond to deformation of olivine polycrystals at 1,473 K and similar strain rates carried out in nearly simple shear geometry (Zh00, Fo₉₀, dashed curve) (22) and in torsional geometry [By00, Fo₉₀, dotted curve (23) and Ha12, Fo₅₀, dash-dotted curve (24)]. In this study, peak flow stresses were achieved at 1,373 K and higher temperatures.

In experiments at intermediate temperatures (1,373–1,500 K), we observed a progressive increase in shear stress τ followed by strain weakening (Fig. 1). Upon termination of the experiments, we observe at 1,373 K ∼38% weakening and observe at 1,473 K ∼60% weakening following the peak stress. In contrast, at low temperatures (≤1,273 K), the specimens kept strain hardening. At the highest temperatures (>1,500 K), no appreciable strain weakening was observed up to shear strains of γ ∼3.0 (GA270/1).

We compare our data for olivine–orthopyroxene aggregates at 1,473 K (Fig. 1) with three previous studies that have explored large strain deformation of polycrystalline olivine, one in nearly simple-shear geometry (22) and two in torsional geometry (23, 24). Overall, we see that the peak strength of our olivine–orthopyroxene mixture is comparable to pure, but dry, Fo₉₀ olivine aggregates deformed at 1,473 K and similar strain rates (∼10⁻⁴⋅s⁻¹). Previous studies on Fo₉₀ olivine show between 0% and 15% weakening after the peak flow stress was reached (22, 23), whereas Fo₅₀ shows up to ∼50 weakening (24). To be clear, strain weakening does not necessarily suggest localization occurred, and can happen via a range of mechanisms but is generally associated with an overall reduction in grain size.

To investigate how strain weakening occurred in our samples, we conducted microstructural observations using the scanning electron microscope (SEM) and the electron-probe microanalyzer (EPMA). Additional electron backscatter diffraction (EBSD) data is presented in Figs. S2 and S3. Fig. 2 shows the SEM images and element (Si) maps of representative regions of selected specimens deformed at four temperatures. The hot-pressed starting materials show near-uniform distribution of orthopyroxene grains among olivine grains with an average grain size ∼ 14 µm, with near-equilibrium grain-boundary morphology and equant grain shapes (Fig. 2A). However, the deformed microstructures for GA265 and GA267, in particular, show a bimodal grain size distribution characteristic of heterogeneous deformation. In contrast, at ∼1,540 K, the sample GA270 mostly recrystallized homogeneously (Fig. 2 E and J). Dynamic recrystallization is the likely main cause for grain size reduction and weakening. We do not observe much grain-boundary bulging, which suggests most recrystallization occurred via subgrain rotation mechanism (25). Grain shape analysis of some orthopyroxene grains for GA267 suggests plastic strains larger than the average strain of the sample also occurred (Materials and Methods).

Fig. 2.

Backscattered electron images and microprobe Si element maps showing a hot-pressed microstructure (A) and deformation microstructures at a series of temperatures: 1,273 K (B), 1,373 K (C), 1,473 K (D), and 1,540 K (E). Dark shaded grains are generally orthopyroxene and correspond to the pink shaded regions in the Si element maps (G–J), whereas the yellow regions correspond to olivine grains. The cartoon in F highlights the principle behind recrystallization and mixing of new grains under an applied shear strain. Recrystallized grain size, d₀, typically increases with stress as described by a paleopiezometer (26).

By comparing the backscattered electrons images directly with the Si element maps (Fig. 2 B–D versus Fig. 2 F–H), we observe that for all samples the minerals olivine and orthopyroxene recrystallized, but at different grain sizes. The mixing of recrystallized grains is highlighted in Fig. 2 by yellow regions representing olivine and the pink regions representing orthopyroxene (Fig. S4). The extent of recrystallization depends on temperature and strain, where GA265, GA267, and GA270 have recrystallized to ∼13%, ∼28%, and ∼71%, respectively (Fig. S5). In the mixed-phase regions of specimens GA265 and GA267, the recrystallized grain size of olivine is somewhat larger than that of orthopyroxene by 10–30%, and the average recrystallized grain size is systematically smaller than the recrystallized grain size of olivine in the monophase regions by a factor of ∼2–3 (Table 1). In contrast, the olivine and orthopyroxene grain size in the polyphase regions of a recrystallized sample deformed at the highest temperatures (Fig. 2J, Fig. S4, and Fig. S5C) is more similar to the olivine grain size in the monophase regions (Table 1). Finally, at intermediate temperatures, the fine-grained and mixed-phase regions may be partially interconnected (Fig. 2 B–D and Fig. S4D). Additional connectivity in the third dimension is possible, but only advanced techniques, such as X-ray tomography or Focused Ion Beam milling combined with electron backscatter diffraction (EBSD), may reveal more.

Table 1.

Deformation run parameters and specimen analysis

Run no.	P, T, GPa, K	Shear strain, γ	Shear strain rate, s−1	Shear stress, τ, MPa*	Water content before/after, wt. ppm H2O†	Initial grain size, d, µm‡	Recrystallized grain size, d0, polyphase olivine/opx, µm‡	Recrystallized grain size, d0, monophase olivine, µm‡
GA256	1.5, 1,173(20)	—	∼10−4	604§	54(22)/56 (18)	5(3)	—	—
GA248	1.3, 1,273(20)	1.0	2.3 × 10−4	545§	317(61)/110 (32)	5(3)	ca. 0.3	1.3(0.8)
GA254	1.5, 1,273(20)	0.7	4.5 × 10−5	380§	182(55)/54 (22)	5(3)	—	—
GA258	1.5, 1,373(20)	0.4	3.5 × 10−5	174§	413(74)/119 (26)	15(10)	—	—
GA265	1.5, 1,373(20)	1.3	1.2 × 10−4	590	398(85)/118(64)	14(9)	0.44(0.20)/ 0.37(0.17)	1.5(0.7)
GA267	2.0, 1,473(20)	1.4	1.2 × 10−4	255	398(85)/96 (18)	14(9)	1.52(0.88)/ 0.96(0.55)	2.8(0.9)
GA268	2.0, 1,498(20)	0.7	1.1 × 10−4	190	398(85)/77 (19)	14(9)	3.36(2.30)/ 2.90(1.00)	5.2(4.2)‡
GA270	2.0, 1,540(50)	3.0	2.1 × 10−4	150§	398(85)/227(67)	14(9)	3.18(1.57)/ 2.78(1.70)	3.0(1.9)‡
GA271	2.0, 1,540(50)	2.1	2.2 × 10−4	140	213(78)/106 (30)	11(9)	—	4.0(1.0)‡

Uncertainty, where given in parentheses, is 1 SD. Opx, orthopyroxene.

^*Shear stress corrected for linear increase in friction and strength of nickel capsule.

^†Paterson 1982 calibration.

^‡No clear regions of localized deformation developed.

^§Shear stress at maximum strain.

To understand why rheological weakening occurred in a certain temperature range but not for other temperatures, we plot our experimental observations in a deformation mechanism map of olivine (see Table S1 for flow laws used) combined with a recrystallized grain size–stress relationship (26) (Fig. 3). Fig. 3 consists of six panels where the first row of panels (Fig. 3 A–C) only include dislocation creep and diffusion creep for wet conditions and second row of panels (Fig. 3 D–F) include an additional deformation mechanism, Dis-GBS (27). We note that Dis-GBS creep may play an important role, but the exact flow laws are poorly constrained, particularly under “wet” conditions. The effect of including this flow law in the deformation mechanism map is that the stress predicted (for pure olivine) at the onset of our experiments should already be nearly an order of magnitude lower at a constant strain rate of ∼1 × 10⁻⁴⋅s⁻¹. Furthermore, deformation still proceeds into the diffusion creep regime as highlighted by the position of the data points (except at the highest temperature). We emphasize that only in diffusion creep can a rock be truly weakened, an important aspect for the strength evolution of the lithosphere.

Fig. 3.

Deformation mechanism maps based on empirical wet olivine flow laws (Table S1) corresponding to three different temperatures at which specimens GA265, GA267, and GA270 were deformed. Stresses are recalculated for compressional geometry, σ (stress) = 2τ(shear stress). The solid blue line indicates the boundary between diffusion and dislocation creep regimes, the gray lines are constant strain rate contours (labeled), and the purple dashed line is the paleopiezometer for olivine (26). In A–C we used the flow laws for diffusion creep and dislocation creep in wet olivine, whereas in D and E we added an additional deformation mechanism: Dislocation accommodated grain-boundary creep (Dis-GBS) (Table S1). In each deformation mechanism map (A–F) two to three data points are plotted as follows: peak flow stress for original hot-pressed grain size (orange square), flow stress at maximum shear strain in mixed-phase regions (orange circle), and olivine-only regions (blue asterisk). Note that for GA270, no true mixed-phase regions could be identified. The arrows indicate the path of microstructural evolution.

Low-temperature experiments (≤1,273 K) were terminated before substantial deformation and recrystallization occurred (Table 1, Fig. 2 B and F), because deformation at lower temperatures (T ≤ 1,273 K) required plastic flow stresses to exceed the confining pressure of the experiments potentially leading to the onset of brittle deformation processes. At temperatures above ∼1,273 K, the recrystallized grain size is increasingly smaller than the critical grain size (Fig. 3 A and B). In this latter case, rheological weakening is possible. The recrystallized grain sizes (for mixed regions) plotted by the circular symbols in Fig. 3 are systematically smaller than the recrystallized grain sizes in pure olivine (asterisk symbols). Finally, at the highest temperature explored in this study (T ∼ 1,540 K), the dynamically recrystallized grain size is close to the critical grain size between diffusion and dislocation creep and not much weakening is observed (Fig. 1).

The essential cause for rheological weakening in our experiments is grain-size reduction. At low temperatures, no appreciable recrystallization occurs, whereas at high temperatures, recrystallized grain size is too large for weakening. In the intermediate temperature range, substantial recrystallization occurs to produce small enough grains (Fig. 2 and Fig. S4). The detailed microstructural analyses show that samples deformed at intermediate temperatures develop heterogeneous microstructures, and large strains in some of the favorably oriented grains (expressed by the Schmid factor) that experienced a large degree of recrystallization (Fig. S6). The local strain was determined from the shape of grains (circularity), assuming the initial shape of each grain was circular (Materials and Methods). We find large variability in grain shapes of parent orthopyroxene grains. The highly elongated (orthopyroxene) grains typically develop long tails or bands composed of small grains that presumably recrystallized from the elongated parent grain(s). Local strain in these recrystallized regions can therefore be much larger than the macroscopically imposed strain. Large strains provide an effective mechanism of grain mixing through grain-switching events (28). Evidence for Zener pinning is also found, suggesting the importance of mixing to maintain small grain size (Materials and Methods and Fig. S7).

Our laboratory results are obtained at much higher stresses (strain rates) than those operating in Earth. Consequently, we need to understand the conditions for weakening based on a physical model and use that model to extrapolate the results to the Earth’s interior. Let us consider a case where stress suddenly increases (this could be due to bending of lithosphere at a trench). Upon the increase in stress, new small grains will be formed at preexisting grain boundaries. If the size of these newly formed grains is smaller than the previously discussed critical grain size, then rheological weakening associated with localization will occur. However, in those regions where diffusion creep takes over, dynamic recrystallization ceases and grain growth follows. If grain growth is fast, then no substantial deformation occurs in the fine-grained regions. Therefore, grain growth must be slow enough to allow for the accumulation of large strains in fine-grained regions. Such a concept was formulated by Karato (2008, chap. 16) (29), where he characterized the conditions for localized deformation using two parameters. One corresponds to the relation between recrystallized grain size and the critical grain size for transition between diffusion and dislocation creep. Another is a parameter expressing the competition between deformation and grain growth (Materials and Methods). Karato (2008) also discussed that in pure olivine, grain growth is so fast that shear localization will not occur.

The presence of a secondary phase slows down the kinetics of grain growth due to Zener pinning. As a result, a secondary phase reduces the dynamically recrystallized grain size. These effects enhance the degree of rheological weakening and promote localization. We use a preliminary dataset on the grain-growth kinetics in an olivine + protoenstatite aggregate (30), but the data given on the influence of secondary phase on the size of recrystallized grains of a primary phase are limited to a narrow range in grain size (0.6–2.9 µm); thus, we only offer qualitative discussions on this effect.

Fig. 4 shows the conditions for shear localization with grain-growth kinetics included for both olivine + orthopyroxene mixture and olivine alone. Unlike in Fig. 3, we used the flow laws for dry olivine (Table S1) to calculate the boundaries in Fig. 4 (Materials and Methods) because two-phase grain-growth measurements are only available for dry conditions, the flow laws for the relevant deformation mechanisms are better constrained for dry conditions, and typically the lithosphere is assumed to be (nearly) dry. We constructed this figure by translating the variables in the Karato model (29) from two nondimensional parameters to temperature and stress. In this figure, the maximum temperature at which substantial localization associated with weakening can occur corresponds to the condition where recrystallized grain size falls inside the diffusion creep regime. The low temperature limit shown for pure olivine corresponds to the temperature above which grain-growth kinetics is too fast compared with diffusion creep (using parameters in Table S2). This lower temperature limit does not exist for an olivine–orthopyroxene mixture with very slow grain-growth kinetics.

Fig. 4.

Model describing the regimes for homogeneous deformation and strain localization in temperature–stress space. In both regimes, various amounts of rheological weakening are possible by different mechanisms, but long-term weakening is possible once the strain has fully localized. The model is based on dry olivine flow laws (Table S1), a paleopiezometer (26), and olivine (+orthopyroxene) grain-growth kinetics (Table S2) (30). The gray lines (olivine) and purple dashed lines (orthopyroxene) (36) are constant strain rate contours for deformation in the dislocation creep and transition to Peierls (high stress) creep regimes (where shown by an inflection point in the curve). Panel A shows a large regime where localization is possible, including the temperature–stress space for geological conditions for mylonite and ultramylonite formation. This regime exists only when the grain-growth kinetics is controlled by a given volume fraction of 25% orthopyroxene with Zener pinning grain size exponent of 4 and activation energies of 300 kJ⋅mol⁻¹ or higher. Panel B demonstrates that development of a heterogeneous microstructure in pure olivine with fast grain-growth kinetics (H_gg ∼ 200 kJ⋅mol⁻¹) (29) is restricted to certain laboratory conditions only.

The model based on the experimental data on olivine + orthopyroxene grain-growth kinetics (30) agrees with observations from this experimental study and with observations of natural peridotites deformed under lower temperature and strain rate conditions (Fig. 4A). Furthermore, the observed efficient mixing of recrystallized olivine and orthopyroxene grains in our study is consistent with the microstructures observed in sheared peridotites (2, 3), suggesting that our observations and model offer a reasonable means of assessing the potential for shear weakening across a range of temperature–stress–strain rate conditions.

However, natural observations are inconsistent with a model prediction based on pure olivine grain-growth kinetics (29) (Fig. 4B). Therefore, we conclude that the pure olivine model of the lithosphere alone cannot explain shear localization in the lithosphere. In addition, our experimental observations are considerably different from those on pure olivine under similar conditions (22, 23). Bystricky et al. (2000) (23) observed only modest strain weakening (∼10%) associated with nearly homogeneous recrystallized microstructure. We show that the presence of orthopyroxene results in a highly heterogeneous microstructure and that the recrystallized grain size in mixed regions is smaller than in unmixed regions (Table 1). Therefore, we interpret that the difference is mainly caused by the role of orthopyroxene in the dynamically recrystallizing mixture.

More recent experiments on polycrystalline Fo₅₀ olivine at 1,473 K (i.e., higher homologous temperature of 0.78 T/T_m) show that this olivine type deformed homogeneously under constant strain rate conditions with 25–50% weakening after peak stress (24, 31). Hansen et al. (2012) (31) suggested that such amount of weakening was possible via both recrystallization and development of a strong lattice preferred orientation (LPO) in their microstructure, the latter lacking in our study (Materials and Methods). Even though a stress reduction up to a factor of 2 is possible in pure olivine deformed to large strains (γ ∼10), we infer that the polyphase aggregates weaken more efficiently at smaller strains due to a switch in deformation mechanism and sluggish dynamic grain growth. In the long term, recrystallized grains in pure olivine will quickly recover to larger sizes, whereas recrystallized grains in a polyphase mixture persist, likely over geological time, enabling deformation to remain localized.

Considerable weakening associated with mixed-phase dynamic recrystallization observed in our experiments helps to explain the weakness of the oceanic lithosphere on Earth implied by the operation of plate tectonics. The similarity in microstructure between our samples in the laboratory and naturally deformed peridotites in shear zones suggests that similar physical processes may operate in these two cases. However, application of our results to explain the presence or absence of plate tectonics on various planets is not straightforward. First, although the physical mechanisms and conditions for localization are well constrained by our study, a quantitative estimate of the degree of weakening is difficult. Studies, such as the one by Montési (2013) (32) on modeling the distribution and interconnectivity of a network of shear zones, are needed to estimate the magnitude of rheological weakening in the lithosphere. Our results provide a critical dataset for such modeling. Second, our samples contain a small amount of water (Fig. S8 and Materials and Methods). Although we were unable to measure water content in olivine and orthopyroxene grains individually, due to too small recrystallized grain size, orthopyroxene is known to take in more water than olivine under lithospheric pressures (<9 GPa) (33). It is likely that a small amount of water enhances the rate of deformation and recrystallization of orthopyroxene as it does for olivine and quartz (24, 35). Furthermore, higher water content may lead to larger or smaller recrystallized grains, depending on the competition between nucleation and growth kinetics affected by water. The outcome may therefore prevent a switch in deformation mechanism if the recrystallized grain sizes of different phases are too large or if nucleation-growth kinetics of different phases is too different. The influence of water on deformation of olivine–orthopyroxene mixtures is currently unknown and open for further investigation.

Methods Summary

For a detailed methodology, see Materials and Methods. San Carlos olivine and orthopyroxene grains were hand-picked for gem quality, crushed to micrometer grain size, and ultrasonically mixed for 75% olivine–25% orthopyroxene weight composition. The powders were left in laboratory air (∼30% humidity) to ensure grain growth during hot pressing. A Griggs-type solid-medium apparatus was used both for hot isostatic pressing of the olivine–orthopyroxene mixture and for deformation with a soft caesium chloride (CsCl) pressure medium (Fig. S1). A typical hot-press experiment was carried out at 1,573 K for 7 h at 1.3 GPa using nickel as capsule material to control oxygen fugacity. After quenching and depressurisation, the recovered hot-pressed material was cut into slices at 45° with a diamond blade, leaving some material for microstructural analysis and water content measurement. Procedures for deformation closely follow those described by Ohuchi et al. (2011) (36) with the following modifications. Each specimen slice was cut in half for a nickel foil strain marker and sandwiched at 45° between two grooved alumina pistons inside a nickel capsule. The loaded capsule was placed inside the Griggs solid-salt (CsCl) deformation cell and pressurized and heated to target conditions (P = 1.3–2.0 GPa; T = 1,173–1,540 K). At target conditions, nearly simple shear deformation experiments were carried out at constant strain rate (typically ∼10⁻⁴⋅s⁻¹). At the highest temperatures (∼1,540 K) and pressures (2 GPa), temperature was estimated from the power–temperature relationship with an uncertainty of 50 K as the thermocouples repeatedly failed under these conditions. The deformed specimens were recovered, polished, and analyzed for water content (FTIR spectroscopy) and microstructure using the SEM with EBSD and EPMA.