Strong seismic anisotropy observed in the Earth's lowermost mantle, the D″ layer, is likely attributed to the lattice-preferred orientation (LPO) of its predominant mineral, magnesium-silicate post-perovskite (PPv). Here we report simultaneous high-pressure and -temperature plastic deformation experiments on MnGeO3 PPv from 77 to 111 GPa and from 83 to 105 GPa at 2000 K using the membrane-type diamond-anvil cell (DAC). Radial X-ray diffraction measurements demonstrate that the (001) plane aligned perpendicular to the compression direction, indicating deformation dominated by slip on (001). In strong contrast to other slip planes proposed previously, the (001) slip plane can cause significant shear-wave splitting with polarization in line with seismic observations. D″ anisotropy can therefore be primarily reconciled with LPO of the PPv phase.
 MgSiO3 PPv has SiO6 octahedral sheet-stacking structure with interlayer Mg ions [Murakami et al., 2004; Oganov and Ono, 2004], and thus the (010) layering plane was first suggested to be a dominant slip plane [Iitaka et al., 2004]. This idea was supported by subsequent low-pressure deformation experiments using CaIrO3-PPv as analogue under various conditions [Miyajima et al., 2006; Yamazaki et al., 2006; Niwa et al., 2007; Miyagi et al., 2008]. In contrast, first-principles modeling of plastic deformation involving the formation of stacking fault implied (110) as a slip plane [Oganov et al., 2005]. Other theoretical study based on Peierls-Nabarro model showed that (010) appears as the easiest slip system for both CaIrO3-PPv and MgSiO3-PPv, while (001) would be the easiest slip system for MgGeO3-PPv [Carrez et al., 2007; Metsue et al., 2009]. Radial X-ray diffraction (XRD) measurements on both MgGeO3- and (Mg,Fe)SiO3-PPv at room temperature indicated that lattice planes near (100) or (110) planes became aligned perpendicular to the compression axis, suggesting the slip on (100) or (110) [Merkel et al., 2006, 2007], consistently with the predictions of Oganov et al. . Nevertheless, the LPO observed by Merkel et al. [2006, 2007] was originally formed upon phase transition to PPv and did not change during further compression. It could thus represent a phase transformation-induced texture with topotactic relationship with the pre-phase [Okada et al., 2010], which is different from the deformation fabric [Walte et al., 2009]. More recently, the high-pressure deformation experiments by Okada et al.  newly proposed the (001) slip plane for MgGeO3-PPv at room temperature, based on the Rietveld analysis of axial XRD data.
 Regardless of the effects of phase transformations and pre-phase textures, most of the previous works were performed under ambient temperature. However, the dominant slip systems and deformation mechanisms at high temperature can be different from those at room temperature (e.g., olivine by Carter and Ave Lallemant ). Therefore, the calculations and experiments conducted at T = 0 or 300 K are possibly not applicable to the Earth's lower mantle [Yamazaki and Karato, 2007], even though the main slip plane of CaIrO3-PPv appears largely independent of temperature [Miyajima et al., 2006; Yamazaki et al., 2006; Niwa et al., 2007; Miyagi et al., 2008; Walte et al., 2009]. Here we conducted plastic deformation experiments on MnGeO3-PPv at simultaneously high pressure and high temperature corresponding to the condition at the Earth's lower mantle. Arguably, MnGeO3 may be the best low-pressure analogue to MgSiO3 [Ringwood and Seabrook, 1963], because Ge is one of the 4B group elements same as Si and the ratio of ionic radius of Mn/Ge is very similar to that of Mg/Si.
2. Experimental Methods
 Present deformation experiments were carried out by a combination of laser-heated membrane-type DAC techniques and radial XRD measurements. Polycrystalline MnGeO3 orthopyroxene was mixed with platinum black that served as a pressure standard and a laser absorber. The cubic boron nitride (inner) + beryllium (outer) composite gasket was used to maximize the sample thickness so that the sample can plastically deform at 100 GPa pressure range [Funamori and Sato, 2008]. The gasket was preindented to about 80-μm in thickness by drilling a circular truncated conical pit. The sample mixture was loaded into a 50-μm hole in the cBN inner gasket and compressed by beveled diamond anvils with 200-μm culet size. After pressurized to a target pressure at room temperature, the sample was heated with a focused TEM01*-mode continuous-wave Nd:YLF laser from both sides of the sample [Ohishi et al., 2008]. Temperature was measured by the spectroradiometric method. By increasing the gas pressure in membrane system, we were able to compress the sample at high temperature during laser heating.
 Angle-dispersive XRD spectra were collected on a charge-coupled device (CCD) detector (Bruker APEX) at the BL10XU, SPring-8. Typical exposure time was 10 sec. A monochromatized X-ray beam with a wavelength of 0.410 and 0.422 Å was collimated to 15-μm in diameter. Pressures were determined based on the equation of state of platinum [Jamieson et al., 1982]. In addition to the axial XRD pattern (Figure S1 of the auxiliary material), the radial XRD images were collected at room temperature, by irradiating the X-ray to the sample perpendicular to the compression axis through gasket, before (at the synthesis) and after deformation (Figure 1). For data analyses, see auxiliary material.
 In the first run, the orthopyroxene starting material was compressed at room temperature and subsequently heated to 1900 K for 9 min at 72 GPa. MnGeO3-PPv was obtained as a single phase at a pressure of 63 GPa upon cooling to 300 K [Tateno et al., 2006] (Figure S1 of the auxiliary material). The sample was then heated again and subjected to plastic deformation at high temperature by further compression; MnGeO3-PPv was pressurized in membrane DAC at 2000 K from 77 to 111 GPa in 21 min. After cooling to 300 K, pressure was 105 GPa. In the second set of experiment, MnGeO3-perovskite (Pv) was initially synthesized and then mostly converted into the PPv phase by heating to 2000 K for 15 min at 77 GPa. After quenching temperature to 300 K, pressure was 68 GPa. The sample was then heated again and further squeezed, at 2000 K, from 83 to 105 GPa in 20 min. Upon cooling to room temperature, pressure decreased to 100 GPa.
 In both runs, the radial XRD patterns demonstrated substantial variations in the diffraction peak positions and intensities as a function of azimuth angle (Figure 1), which indicates the stress and LPO of the sample. The agreement between the experimental and recalculated textures is particularly good (Figure 2). In run #1, the XRD data were fairly smooth, with a large number of relatively small grains within the X-ray beam. Differential stress increased from 2.4 GPa after synthesis at 63 GPa to 7.3 GPa at 105 GPa (see Table S1 of the auxiliary material). In run #2, PPv grains were much larger, and minor peaks from Pv could be also observed. Differential stress in PPv increased from 2.2 GPa after synthesis at 68 GPa to 10.0 GPa at 100 GPa.
 In run #1, just after synthesis, PPv displayed a texture characterized by (001) lattice planes nearly perpendicular to the compression direction, with an inverse pole figure (IPF) maximum of 2.65 multiples of a random distribution (m.r.d.) (Figure 3a and Table S1 of the auxiliary material). After compression to 111 GPa at 2000 K, the strength of the 001 maximum increased remarkably to 5.24 m.r.d. (Figure 3b). In run #2, just after synthesis, PPv displayed a broad texture maximum near 100, with an IPF maximum of 2.10 m.r.d. (Figure 3c). After compression to 106 GPa at 2000 K, we can notice a rotation of texture towards an 001 maximum, although this evolution does not seem to be completed yet (Figure 3d).
4.1. Deformation Texture
 Differences between those two runs highlight the effect of transformation texture in radial XRD measurements, as inferred by Walte et al.  and Okada et al. . PPv showed a weak 001 texture already at the time of synthesis, and it was significantly strengthened by subsequent deformation in run #1 (Figures 3a and 3b). Textures observed after transformation from Pv to PPv in run #2 were similar to those obtained by Merkel et al. [2006, 2007] on MgGeO3- and (Mg,Fe)SiO3-PPv, but they were evolved towards 001, consistent with the results of run #1.
 In contrast with the previous experiments [Merkel et al., 2006, 2007] but in agreement with Okada et al. , we observed a clear texture evolution with compression that led to a strengthening of the 001 maximum (Figure 3). Moreover, our samples have been compressed at 2000 K. We can therefore argue that our texture evolution can be directly attributed to the plastic deformation of MnGeO3-PPv at high pressure and high temperature, corresponding to the condition in the Earth's lower mantle. Development of texture in PPv deformed in compression has been simulated using the viscoplastic self-consistent polycrystal plasticity method (VPSC) multiple times [Merkel et al., 2006, 2007]. Such simulations only depend on the starting texture, the deformation geometry, and the plastic properties of the sample. By comparing our experimental results and those simulations, we can infer that such an alignment can be interpreted to indicate the slip on the (001) plane.
4.2. Origin of D″ Anisotropy
 Our results suggest that the (001) plane dominates the plastic deformation of MnGeO3-PPv under the lower mantle pressure and temperature. This is consistent with the results of Okada et al.  in which LPO was estimated by simulation from the axial XRD data. The knowledge about predominant slip plane of PPv has important implications for the origin of strong seismic anisotropy observed in the D″ layer, although the deformation of analogue material by fast strain rate in the present experiments may not be relevant to the Earth.
 The shear-wave polarization anisotropy, frequently observed in fast VS regions of the lowermost mantle, in the lowermost mantle is characterized by VSH is faster than VSV ((VSH–VSV)/<VS> = 1.01–1.03, where VSH and VSV are horizontally and vertically polarized shear velocities, respectively, and <VS> is the velocity for an isotropic aggregate) [Panning and Romanowicz, 2004; Wookey and Kendall, 2008]. Such shear-wave splitting is likely attributed to the LPO of elastically anisotropic minerals. Under the horizontal shear flow expected in the lowermost mantle, PPv develops the LPO with (001) parallel to the horizontal plane. The perfect alignment produces 8 to 15% shear splitting with the same polarization as observed, (VSH–VSV)/<VS> = 1.08–1.15, at 125 GPa and 2500 K, depending on the predicted high-temperature elastic constants [Wentzcovitch et al., 2006; Stackhouse and Brodholt, 2007]. Note that such splitting is indeed consistent with the observations, much more than those assuming other slip planes proposed previously [e.g., Oganov et al., 2005; Merkel et al., 2006, 2007; Walte et al., 2009]. If the dominant slip plane is (010), the resulting shear-wave splitting is calculated to be only 2% [Stackhouse and Brodholt, 2007] or of opposite polarization [Wentzcovitch et al., 2006]. The LPO involving (100) and (110) slip planes also produces the opposite shear-wave polarization.
 These results suggest that the LPO of PPv can cause much larger shear-wave splitting than previously thought, with the right sign of polarization. The maximum 3% polarization anisotropy observed at the D″ layer is thus attributed to the LPO of PPv with less than 25% alignment based on the elastic constants predicted by Stackhouse and Brodholt , considering that the PPv phase constitutes about 80% of the lowermost mantle [Murakami et al., 2005]. Deformation experiments on the CaIrO3 analogue have demonstrated that the PPv-type phase is at least five times weaker [Hunt et al., 2009]. More recent calculations predicted that MgSiO3-PPv is softer than Pv by four orders of magnitude [Ammann et al., 2010]. Plastic deformation is thus much easier in PPv than the corresponding Pv phase [Niwa et al., 2007], suggesting that such 25% alignment is feasible. In addition, the role of (Mg,Fe)O ferropericlase in the formation of D″ anisotropy has been repeatedly stressed [Wentzcovitch et al., 2006; Yamazaki and Karato, 2007]. It was claimed that ferropericlase is highly elastically anisotropic [Yamazaki and Karato, 2002], but even a complete alignment with horizontal shear could result in only about 1% VSH > VSV anisotropy when its abundance (∼20%) is considered. While LPO of ferropericlase certainly helps in some degree, strong shear splitting in the D″ region is more likely attributed primarily to the LPO of PPv. This is consistent with the seismological observations that the shear velocity anisotropy exists only at and below the D″ discontinuity [Lay et al., 1998].
 We thank N. Sata for assistance in the XRD measurements and T. Komabayashi and D. Yamazaki for valuable suggestions. Experiments were carried out at SPring-8 (proposals 2009A0087 and 2009B0087).