Messengers from the deep: Fossil wadsleyite-chromite microstructures from the Mantle Transition Zone

Investigations of the Mantle Transition Zone (MTZ; 410–660 km deep) by deformation experiments and geophysical methods suggest that the MTZ has distinct rheological properties, but their exact cause is still unclear due to the lack of natural samples. Here we present the first direct evidence for crystal-plastic deformation by dislocation creep in the MTZ using a chromitite from the Luobusa peridotite (E. Tibet). Chromite grains show exsolution of diopside and SiO2, suggesting previous equilibration in the MTZ. Electron backscattered diffraction (EBSD) analysis reveals that olivine grains co-existing with exsolved phases inside chromite grains and occurring on chromite grain boundaries have a single pronounced crystallographic preferred orientation (CPO). This suggests that olivine preserves the CPO of a high-pressure polymorph (wadsleyite) before the high-pressure polymorph of chromite began to invert and exsolve. Chromite also shows a significant CPO. Thus, the fine-grained high-pressure phases were deformed by dislocation creep in the MTZ. Grain growth in inverted chromite produced an equilibrated microstructure during exhumation to the surface, masking at first sight its MTZ deformation history. These unique observations provide a window into the deep Earth, and constraints for interpreting geophysical signals and their geodynamic implications in a geologically robust context.

same ore body 8 further supports MTZ depths for these rocks. These observations indicate that Luobusa chromitites have experienced much higher pressures than the more common ophiolitic chromitites, whose history is restricted to the uppermost part of the mantle 9 .
The chromites commonly show exsolution of diopside, enstatite and coesite 4,10 , suggesting inversion from a high-pressure polymorph that incorporated Ca and Si, such as the Ca-Ferrite (CF; CaFe 2 O 4 ) structure (> 15 GPa at TZ temperatures of 1400-1500 °C) 8,10,11 . Several recent experimental studies show variations in the details of the phase diagram for chromite under MTZ conditions. Ishii et al. (2014Ishii et al. ( , 2015 12,13 suggested that chromite breaks down to an intermediate mLD-type (Mg, Fe) 2 Cr 2 O 5 + Cr 2 O 3 eskolaite assemblage (12)(13)(14)(15) before the CF structure is formed. However, these experiments may not be relevant to real chromite, which contains significant levels of aluminum. Multianvil experiments using natural Al-rich Luobusa chromitite + diopside or SiO 2 as a starting materials (14 GPa, 1600 °C) [14][15][16] found that chromite starts to transform into (Fe, Mg) 2 (Al, Cr) 2 O 5 with significant Ca or Si substitution on the Al site. Although further experiments are required, the intermediate phase is orthorhombic in all cases (as is the CF-type). Here we refer to the group of suggested high-pressure chromite phases as HPP (High Pressure Polymorph)-chromite.
The apparent deep recycling of some podiform chromitites (and their host peridotites) implies previously unrecognized types of mantle dynamics, and requires reinterpretation of the history of the lithospheric mantle in some ophiolites. This contribution presents an electron backscattered diffraction (EBSD) study of UHP massive chromitites from the Luobusa peridotite (E. Tibet, China), shedding light on the deformation behavior of high-pressure phases at MTZ conditions. We provide the first direct evidence of crystal-plastic deformation, evidenced by mineral CPOs and substructures interpreted to have formed by dislocation creep in the UHP phases wadsleyite (now olivine) and HPP-chromite, now normal spinel-structured chromite.
The sample consists of two zones: Zone 1 contains larger chromite grains (591vm vs 526 μ m) with slightly higher median aspect ratios (1.73 vs 1.70) and lower abundance of olivine (1.6 area% vs 2.2 area%) than zone 2 (Fig. 1e,f). The line separating these two zones is inferred to represent the trace of the foliation in this rock (Fig. 1d). Chromite subgrain and grain boundaries are commonly oriented perpendicular to foliation (S1) (white arrow in Fig. 1d), thus the longest grain axis of chromite grains is subhorizontal. The grain elongation perpendicular to foliation is interpreted to reflect late subgrain boundary evolution to form grain boundaries by recovery (black arrow in Fig. 1d).
A well-defined low-angle subgrain boundary (2-10° orientation change across the boundary) is shown in Figs 2(c) and S2. The trace of this boundary does not lie in a plane perpendicular to the misorientation axis as determined by the dispersion paths depicted in the pole figure. Thus, a twist-boundary model can be eliminated for this subgrain boundary. A steep plane that contains the subgrain-boundary trace and the determined misorientation axis [100] is the probable subgrain boundary plane, in this case (001) comprising an array of edge dislocations lying in (001). Hence, subgrain-boundary characteristics are consistent with the activation of the (010)[001] slip system. Olivine (both inclusions and interstitial grains) also shows the same, pronounced CPO with the (010) plane parallel to foliation, consistent with the activation of (010)[001] slip.
The small olivine inclusions in chromite commonly occur in areas where the chromite host contains exsolved needles of diopside (Figs 1a, 2d, S1). All diopside needles show a uniform crystallographic orientation relative to the host chromite, suggesting that they exsolved from a HPP-chromite in a topotaxial relationship with the host: [100] dio //< 111> chr , [010] dio //< 110> chr , and [001] dio //< 112> chr (Figs 2d, S3). No significant crystal bending of diopside needles was detected, however it should be noted that the needles are too thin to allow unequivocal quantification of substructures.

Significance and Discussion
The abundance of exsolved diopside needles in the cores of chromite grains, vs their absence in the rims (Fig. 1a), indicates that grain growth occurred after the transition from the orthorhombic HPP-chromite to cubic chromite. Grain growth occurs by grain boundary migration between pre-existing grains, allowing readjustment of mineral compositions in the "swept" area to one in equilibrium in the prevailing PT conditions, while maintaining crystallographic near-continuity with the core [29][30][31][32][33] . The grain-size difference between the enclosed and interstitial olivine grains is consistent with the evidence for grain boundary migration, because grain boundaries can readily migrate across small grains but can be pinned by larger grains [34][35][36][37][38][39] . A decrease in the rate of grain growth due to the presence of grains of a second phase also is consistent with the finer grain size and higher abundance of olivine in zone 2 relative to zone 1 (Fig. 1e).
While deformation features such as continuous lattice bending and subgrains perpendicular to foliation are present in the chromite, the chromite CPO patterns do not resemble typical deformation-or growth-induced textures known from cubic phases 40 . In contrast, the patterns resemble nearly point maxima concentrations commonly formed by dislocation creep i.e. crystal-plastic deformation, in orthorhombic phases such as olivine or pyroxene 1 . In addition, subgrain boundaries in orthorhombic phases are commonly formed perpendicular to the mineral's long axis and foliation 41 . We thus infer that the observed CPO and substructures were inherited from an orthorhombic precursor phase, and were preserved during subsequent inversion to the lower-pressure cubic chromite structure 42,43 . The exsolution of diopside + SiO 2 suggests transformation from the HPP, which has orthorhombic symmetry [11][12][13] . Chromite domains exhibiting diopside exsolution are smaller than the average grain size. The observed grain-size and orientation characteristics imply that a finer-grained orthorhombic HPP was deformed by dislocation creep, and thus could develop a noticeable CPO. Chromite shows subtle deformation features unrelated to late decompression fracturing (Figs 1d, S1), confirming crystal plastic deformation as the origin of the chromite CPO. The latter deformation characteristics, along with the presence of rims of chromite without exsolution, suggest that grain growth under static conditions (lack of significant differential stress) occurred in the upper mantle, i.e. in the low pressure chromite stability field, resulting in some recovery of internal deformation structures, growth-related substructures in the rim 31,44 , increase in chromite grain size, and the present equilibrated triple junctions and straight grain boundaries (Fig. 1a). The preserved significant chromite CPO is the result of a combination of dynamic recrystallization of the HPP chromite precursor and significant static recrystallization at lower pressure 45 .
The diopside needles exsolved in a topotaxial relationship with host chromite, as shown in previous studies 10 (Figs 2d, S3). This relationship is also shown by clinopyroxene lamellae in garnet from Norway, formed during the breakdown of a majoritic-garnet precursor 46 . Diopside does not show any significant crystal bending, implying that the diopside needles are likely to have exsolved after the crystal-plastic deformation recorded by the chromite CPO (Fig. 2d).
The strong CPO of the olivine indicates deformation by dislocation creep, as crystal alignment during compaction and/or flow of partially molten systems produces a weaker CPO 47 . Grain-boundary sliding and diffusion creep would produce even weaker CPOs 48,49 . In the last decade, there has been significant progress in understanding the crystal-plastic deformation of olivine at high pressure, as well as that of wadsleyite in its stability field. Olivine CPO characterized by [001] axes aligned parallel to the shear direction (i.e. foliation) has been observed in experiments 50,51 and predicted by numerical modelling of crystal plasticity 52 , and has been attributed to the presence of water and/or pressure. In this study, olivine shows the (010)[001] fabric, which we interpret to record the peak PT conditions. This was preserved during late exhumation due to a general lack of significant differential stress at lower P, as shown by the lack of significant deformation overprint on the inverted chromite, and entrapment in the rheologically stronger chromite host, shielding it from subsequent deformation. If later deformation changed CPO patterns, we would expected this to be more pronounced in interstitial olivine grains, as inclusions would be protected. However, the CPO patterns of inclusions and interstitial olivines are identical (Fig. 2b). The fact that inclusions with the characteristics shown here are only found in the exhumed UHP chromite, (identified by the exsolution of diopside and SiO 2 ), further supports our interpretation that the olivine CPO was produced by deformation under UHP conditions.
Olivines hosted in chromite (both inclusions and interstitial grains) have a (010)[001] fabric, which we interpret as reflecting deformation by dislocation creep within the MTZ. This implies that these olivine grains were deformed and aligned while in one of the higher-P polymorph states. Wadsleyite (or β -olivine) is volumetrically the main component of the upper transition zone, between 410 and 520 km depths, and is not stable at shallower depths. Deformation experiments on wadsleyite show several active slip systems 53 55,56 . The phase transformation from olivine (ol) to wadsleyite (wds) with topotaxial relations was studied at 13 Pa and 1400 °C using a multi-anvil apparatus 58 , and occurs as two types; (1) (101) wds //(100) ol and [010] wds //[001] ol , (2) (021) wds //(100) ol , (011) wds //(010) ol and [100] wds //[001] ol , respectively. Although these topotaxial relationships are the result of the phase transformation from olivine to wadsleyite, crystallographic considerations predict that they are expected also to be applicable to that from wadsleyite to olivine, because the close-packed oxygen layers of olivine and wadsleyite cause these topotaxial relations. Integration of our data indicates that the observed (010)[001] fabric preserves the (011)[100] fabric of wadsleyite despite the transformation. Such a fabric is most consistent with simple shear deformation of "dry" wadsleyite under MTZ conditions 54 .

Conclusions
The significant CPO of included and interstitial olivine grains in chromitite implies the activation of the (011)[100] slip system in wadsleyite during crystal-plastic deformation by dislocation creep under MTZ P-T conditions. The transformation from an orthorhombic HPP-chromite to cubic chromite during exhumation drove the exsolution of diopside and coesite, and the wadsleyite grains transformed to olivine during this process, while preserving the wadsleyite CPO. Significant grain-boundary migration under static conditions during passive exhumation produced equilibrated microstructures but preserved the core-rim structure with exsolved diopside only in the cores (Fig. 3), as well as the high-PT fabric. The relict olivine-chromite fabric of this Luobusa chromitite provides the first direct evidence of significant crystal-plastic deformation in the Mantle Transition Zone.

Methods
Chemical compositions of host chromite and olivine inclusions in the thin section were analyzed with Cameca SX-100 electron microprobe at Geochemical Analysis Unit (GAU) at the Australian Research Council Center (ARC) of Excellence for Core to Crust Fluid Systems (CCFS) in Macquarie University, Australia). Quantitative analyses were performed with an accelerating voltage of 15 kV, 12 nA beam current. Crystallographic orientation measurements were obtained using the SEM-EBSD facilities at GAU at the ARC Center of Excellence for CCFS (Macquarie University, Australia) and Australian Centre for Microscopy and Microanalysis (University of Sydney, Australia). The EBSD patterns were generated by the interaction of a vertical electron beam with a polished thin section, tilted at 70° to the horizontal in a scanning electron microscope (Zeiss EVO MA15 and Zeiss Ultra Plus). The operating conditions were a voltage of 20 kV, a current of 8.2 nA and working distance of 12-13 mm. The diffraction pattern was projected onto a phosphor screen and recorded using a digital CCD camera. The resulting image was then processed and indexed in terms of crystal orientation using the AZtec software distributed by Oxford Instruments. Maps were acquired with sampling step size of 20 μ m, 1-5 μ m or 200 nm. Raw indexation rates were > 85%. Data treatment allowed the rare non-indexed pixels to be filled, if up to six identical neighbors existed with this orientation. We present the resulting data in the form of color-coded maps and pole figures. For the maps that show crystal orientation changes relative to the specific direction of the sample reference frame, full red, green, and blue colors are assigned to the grains whose < 100> , < 110> or < 111> axes are parallel to the projection of the inverse pole figure. Intermediate orientations are colored as a mixture of the primary axes. For our analysis, we define (i) a grain as an area that is completely surrounded by boundaries with a misorientation of 10, and (ii) a grain size of equivalent circle diameter calculated by grain area. One point per grain data are plotted on pole figures using programs developed by D. Mainprice of Université Montpelier II, France. Transformation from HPP structure to chromite; diopside/coesite exsolution; transformation of wadsleyite to olivine. (C) Grain growth through static grain-boundary migration to produce coarse equilibrated microstructure, with exsolution-rich cores and exsolution-free rims. Larger interstitial olivine grains pin the migrating boundaries.