2016; Yamazaki & Karato, 2001b). Depending on the amount of Fp and the microstructure of the mantle and IWL type, microstructure may be relevant and Fp would control the strength of the aggregate. Given that estimates of the volume fraction of Fp are likely to be less than 30%, stability of an IWL microstructure should be largely dependent on the total strain experienced by the aggregate. High‐strain torsion experiments on a calcite + anhydrite mixture to shear strains of 57 at 1.5 GPa show extensive phase segregation that results in compositional layering similar to crustal mylonites (Cross & Skemer, 2017). If such phase segregation and compositional layering occurs in the lower mantle, it is likely that shear strain would be largely partitioned into the softer ferropericlase. Thus, in regions of high strain, such as in and around slabs, Fp is likely to control the strength of the mantle. If Fp in the mantle experiences a large increase in strength (Deng & Lee, 2017; Marquardt & Miyagi, 2015), this could lead to a mid‐mantle increase in viscosity consistent with geophysical observations (Mitrovica & Forte, 2004; Rudolph et al., 2015). This strengthening and consequent viscosity increase could explain observations of mid mantle slab stagnation (Fukao & Obayashi, 2013; C. Li et al., 2008) and plume broadening (French & Romanowicz, 2015).
Ca‐Pv, which is a relatively minor phase in the bulk of the lower mantle, can reach up 30% by volume in Mid‐Ocean Ridge Basalts (MORB) subducted into the lower mantle (Hirose et al., 2005; Ricolleau et al., 2010). As such, it is important to understand the rheological behavior of this mineral in the lower mantle and to resolve the discrepancies between the current measurements of Shieh et al. (2004) and Miyagi et al. (2009). One potential implication of a strong Ca‐Pv as suggested by the measurements of Miyagi et al. (2009) is that it could act to strengthen the slab, and it may behave similarly to carbide or ceramic particles in metal matrix composites, reducing texture development in the other phases. This could lead to reduced anisotropy in MORB rich regions of the slab. However if the strength Ca‐Pv is similar to the other phases then it will play a less profound role in controlling slab strength and anisotropy.
In deeper regions of the mantle, MgSiO3 pPv is likely the dominant phase in regions above the CMB. Although a direct comparison of experiments is difficult, as pressures do not overlap, pPv appears to be weaker than Brg in DAC experiments. Given that theoretical calculations of diffusion coefficients also indicate that pPv is significantly weaker than Brg (Ammann et al., 2010) and that analog materials show the pPv phase to be 5–10 time weaker than Pv (Dobson et al., 2012; Hunt et al., 2009), it is reasonable to expect that pPv is likely to be one of the weaker phase in the lower mantle. Interestingly the strength contrast between Pv and pPv analogs is as large or larger than that predicted for Brg and Fp (Kraych et al., 2016); thus, Fp and pPv may have similar strengths. If this is the case, aggregates of pPv + Fp would be considerably weaker than the rest of the mantle, and this would lead to shear localization in the D” layer. Shear localization above the CMB and resultant large strains and high strain rates could result in increased dislocation activity, consistent with presence of wide spread anisotropy in this region. Geodynamic models have also shown that a weak pPv layer will have important geodynamic implications for mantle convection in terms of heat transfer and temperature gradients, viscosity, and flow velocity (Nakagawa & Tackley, 2011; Samuel & Tosi, 2012).
Slip system activities and texture development in lower mantle phases have important implications for interpretation of seismic anisotropy. For a recent review of seismic anisotropy in the deep earth and methods to interpret anisotropy, see Romanowicz & Wenk (2017). The methodology that is typical used to relate slip systems to anisotropy has been to use a flow model combined with plasticity modeling to predict texture development in the deep earth. Single crystal elastic properties are then averaged over these textures to obtain polycrystal elastic properties. These combined with density can be used to calculate seismic properties. In the lower mantle, these properties are typically compared to shear wave splitting to assess the viability of that phase and deformation state to explain observed anisotropy. Flow models to calculate anisotropy may be obtained from convection simulations (Chandler et al., 2018; Cottaar et al., 2014; Immoor et al., 2018; Merkel et al., 2007; Miyagi et al., 2010; Wenk et al., 2011; Wu et al., 2017) or from instantaneous global flow models based on tomographic inversions (Ford & Long, 2015; Nowacki et al., 2013; Walker et al., 2011, 2018). Others have directly compared results from shear experiments to the mantle (e.g., Tsujino et al., 2016; Yamazaki et al., 2006) or from simple shear simulations as an approximation for mantle deformation (Merkel et al., 2006; Tommasi et al., 2018).
In the top of the lower mantle, single crystal elastic properties of ferropericlase are nearly isotropic (Marquardt et al., 2009). Furthermore, experiments seem to indicate that in the presence of Brg, Fp will not develop significant texture (Kaercher et al., 2016; Miyagi & Wenk, 2016; Wenk, Lonardelli, et al., 2006), and thus ferropericlase is unlikely to be a source of anisotropy in the top of the lower mantle. Ca‐Pv can have large single crystal elastic anisotropy at the top of the lower mantle (Karki & Crain, 1998), but its contribution to anisotropy has not been explored. Recent shear experiments on Brg show texture and anisotropy development that is qualitatively consistent with observations of seismic anisotropy in the top of the lower mantle(Tsujino et al., 2016). Thus, Brg with slip on (100)[001] appears to be the most likely candidate to explain anisotropy in this region.
In the lowermost mantle simulations of D” anisotropy do not provide a clear consensus on which phases and slip systems provide the best match. Studies by Chandler et al. (2018), Cottaar et al. (2014), and Wenk et al. (2011) find that pPv with slip on (001) planes provides the best match to anisotropy in the circum‐pacific region. In contrast, Nowacki et al. (2013) and Walker et al. (2011) found that pPv with slip on (010) provides the best match to global observations. Interestingly these result are most strongly correlated in regions of Large Low Shear Velocity Provinces (LLSVPs), where it is unclear if pPv is stable (e.g., Grocholski et al., 2012). Ford and Long (2015) tested the predictions of Walker et al. (2011) against their measurements of shear wave splitting at the eastern edge of the African LLSVP and found a large misfit between the model and observations. However, if constraints on the direction of the flow field were relaxed, pPv with slip on (010) provided the best match. Most recently, models including the effects of the the Brg to pPv phase transition on texture development have been performed using flow models from convection simulations (Chandler et al., 2018) and from instantaneous flow field models (Walker et al., 2018). Both of these studies find the best fit to global anisotropy with pPv deformed by dominant slip on (001) planes. A recent study on Fp at high pressure and temperatures found that slip on {100}〈011〉 could explain complex and laterally varying anisotropy near the edges of LLSVPs (Immoor et al., 2018).
In the mantle, polyphase interactions could play an important role in texture development however; to date most models of anisotropy are based on single phase experimental results. Additionally, most simulate deformation of single phases rather than deformation of a polyphase aggregate. The exceptions are the simulations of Cottaar et al. (2014), which contain Brg/pPv 75% and Fp 25%, Chandler et al. (2018), which uses an aggregate of Brg/pPv 60%, Fp 20%, and CaPv 20%, and Tommasi et al. (2018), which uses an aggregate of pPv 70% and Fp 25%. However, all of these studies use VPSC to simulate texture development. As a homogenization scheme (as opposed to full field methods), VPSC does not account for microstructural effects other than phase proportion, grain shape evolution, and grain orientation. Thus, VPSC cannot track spatial relationships between grains and phases and cannot distinguish between IWL or LBF microstructures and also cannot account for local variations in the stress–strain field. If Fp in the lower mantle does not develop texture when deformed with bridgmanite then any ansiotropy observed in the bulk of the lower mantle is likely due to bridgmanite. This would also imply that the observation of isotropy in the bulk of the lower mantle is likely due to a lack of texture devlopment in all the phases, e.g., due to diffusion creep (Karato et al., 1995) and/or pure climb creep (Boioli et al., 2017; Cordier & Goryaeva, 2018; Reali et al., 2019), rather than texturing where anisotropy patterns of the individual phases cancel each other (Wenk, Speziale, et al., 2006). It remains to be seen how texture develops in aggregates of pPv + Fp of Brg+pPv+Fp, but this will have important consequences for sources of anisotropy in the lowermost mantle.
2.7 CONCLUSIONS
Although considerable progress