Fracturing and crystal plastic behaviour of garnet under seismic stress in the dry lower continental crust (Musgrave Ranges, Central Australia)

Garnet is a high-strength mineral compared to other common minerals such as quartz and feldspar in the felsic crust. In felsic mylonites, garnet typically occurs as porphyroclasts that mostly evade crystal plastic deformation, except under relatively high-temperature conditions. The microstructure of granulite facies garnet in felsic lowercrustal rocks of the Musgrave Ranges (Central Australia) records both fracturing and crystal plastic deformation. Granulite facies metamorphism at ∼ 1200 Ma generally dehydrated the rocks and produced millimetre-sized garnets in peraluminous gneisses. A later ∼ 550 Ma overprint under sub-eclogitic conditions (600–700 C, 1.1–1.3 GPa) developed mylonitic shear zones and abundant pseudotachylyte, coeval with the neocrystallization of fine-grained, high-calcium garnet. In the mylonites, granulite facies garnet porphyroclasts are enriched in calcium along rims and fractures. However, these rims are locally narrower than otherwise comparable rims along original grain boundaries, indicating the contemporaneous diffusion and fracturing of garnet. The fractured garnets exhibit internal crystal plastic deformation, which coincides with areas of enhanced diffusion, usually along zones of crystal lattice distortion and dislocation walls associated with subgrain rotation recrystallization. The fracturing of garnet under dry lowercrustal conditions, in an otherwise viscously flowing matrix, requires transient high differential stress, most likely related to seismic rupture, consistent with the coeval development of abundant pseudotachylyte. Highlights. – Garnet is deformed by fracturing and crystal plasticity under dry lower-crustal conditions. – Ca diffusion profiles indicate multiple generations of fracturing. – Diffusion is promoted along zones of higher dislocation density. – Fracturing indicates transient high-stress (seismic) events in the lower continental crust.

geospeedometry (Lasaga, 1983;Caddick et al., 2010;Baxter and Scherer, 2013). Being a high-strength mineral (Karato et al., 1995;Wang and Ji, 1999), both brittle and crystal plastic deformation are rarely observed in garnet when compared to the common matrix minerals of the crust, such as quartz and feldspar. However, Dalziel and Bailey (1968) already interpreted elongate garnets in high-grade mylonites to be the result of crystal plastic behaviour, and advancements since then in electron microscopy, and especially EBSD (electron backscatter diffraction), have allowed for a detailed investigation of garnet textures (Kunze et al., 1993;Prior et al., 2000Prior et al., , 2002. Experimental deformation of garnet indicates that differential stresses on the order of a few gigapascals are required to produce shear fractures and that the onset of crystal plastic behaviour for strain rates typical of actively deforming regions (10 −12 -10 −15 s −1 ; e.g. Behr and Platt, 2011) should only occur at corresponding temperatures above ca. 640-750 • C (Karato et al., 1995;Wang and Ji, 1999). The observation of fractured garnets in natural samples may therefore be linked to seismic stresses, as suggested by Austrheim et al. (1996), who described the fracturing of garnets during pseudotachylyte formation and fluid-assisted eclogitization of granulites. Trepmann and Stöckhert (2002) also interpreted the microstructure of fractured and offset garnets as evidence for synseismic loading and post-seismic creep. More recently, both brittle (Austrheim et al., 2017;Engi et al., 2017;Angiboust et al., 2017;Hawemann et al., 2018; and associated crystal plastic behaviour (Austrheim et al., 2017; of garnet have been related to seismic events in lower continental crust or deeply subducted continental fragments. Papa et al. (2018) interpreted similar deep-seated dilatant fracturing of garnet immediately adjacent to pseudotachylyte to be related to thermal shock due to frictional heating rather than to damage associated with the propagation of the seismic rupture. Konrad-Schmolke et al. (2007) described enhanced diffusion of Mg along subgrain boundaries in garnet (but not of slowdiffusing elements, such as Ca, Ti and Y) from high-pressure meta-granitoids of the deeply subducted Sesia Zone (Western Alps). However, in contrast to more recent studies in the Sesia Zone, which propose that precursor fracturing was crucial for dissolution-precipitation and diffusion processes in garnet , they considered that there were no signs of crystal plastic deformation in their garnet samples and concluded that a diffusion-induced dislocation migration and/or diffusion-induced recrystallization process was responsible for the development of the observed subgrain texture.
Garnets can retain their microstructure and chemical composition during retrograde deformation and metamorphism and can therefore preserve indicators for seismic events, which are otherwise possibly erased from the rock record.
Here we present a study of garnet microstructures from lower crustal rocks of the Musgrave Block in Australia, which 1. illustrates the close association between brittle and crystal plastic deformation of garnet under well-established pressure-temperature conditions; 2. infers deformation mechanisms from the observed microstructure; 3. explores the close link between deformation and diffusion in garnet; 4. complements other independent observations indicating transient high stresses in the lower crust.
2 Geological setting 2.1 Regional geology The Musgrave Block is located in an intraplate position close to the centre of the Australian continent (inset Fig. 1). Amalgamation of the different cratonic blocks took place during the Musgravian Orogeny (1120-1200 Ma), which pervasively overprinted ca. 1550 Ma gneisses (Gray, 1978;Camacho and Fanning, 1995). The Petermann Orogeny (∼ 550 Ma) produced a series of crustal-scale fault zones, most prominently the Woodroffe Thrust and the Mann Fault (Collerson et al., 1972;Major, 1973;Bell, 1978;Camacho and Fanning, 1995;Raimondo et al., 2010;Hawemann et al., 2018Hawemann et al., , 2019Wex et al., 2017Wex et al., , 2018Wex et al., , 2019. The south-dipping Woodroffe Thrust has a top-to-the-north sense of shear, and it juxtaposes the Fregon Subdomain in the south (hanging wall) against the Mulga Park Subdomain in the north (footwall). During the Musgravian Orogeny, the Mulga Park Subdomain attained amphibolite facies conditions, while the Fregon Subdomain reached granulite facies (Camacho and Fanning, 1995; and depleted the rocks of OH-bearing minerals Hawemann et al., 2018). The Woodroffe Thrust hosts one of the largest occurrences of pseudotachylyte worldwide , but all larger-scale shear zones in the hanging wall also show abundant pseudotachylyte that developed under lower-crustal conditions (Camacho, 1997;Hawemann et al., 2018). Deformation in the Fregon Subdomain associated with the Petermann Orogeny is concentrated along the sub-eclogitic (∼ 650 • C, 1.2 GPa) Davenport Shear Zone and the North Davenport Shear Zone (Fig. 1), with little discernible overprint of the earlier granulites in between . The Davenport Shear Zone is a WNW-ESE-striking strike-slip zone, with a near horizontal stretching lineation. Deformation inside the Davenport Shear Zone itself is heterogeneous and strongly localized .

Sample description
Fractured garnet is ubiquitous in the Fregon Subdomain and is not exclusively found in association with pseudotachylyte veins. However, this study focuses on a representative outcrop for which field relationships, metamorphic and deformation conditions have been well established (F68, Hawemann et al., 2018;26.3849 • S, 131.7067 • E). This outcrop consists of a quartzo-feldspathic mylonite with millimetresized granulite facies garnets and includes multiple pseudotachylyte veins and breccias. Pseudotachylytes in the studied outcrop are sheared, as indicated by elongated clasts (Fig. 2a,  c), and show the same stretching lineation as the host mylonite. The original discordant relationship to the host foliation is still preserved, with the crosscutting relationship most obvious in sections perpendicular to the stretching lineation (Fig. 2b).
The synmylonitic assemblage associated with the Petermann overprint of the felsic granulites is Qz + Kfs + Pl + Gt + Bt + Ky + Ilm + Rt (mineral abbreviations following Whitney and Evans, 2010) and is similar to that of the associated sheared pseudotachylyte (Qz + Kfs + Pl + Gt + Bt + Ky + Rt) . The fine-grained garnet growing within the pseudotachylyte gives the rock its macroscopic caramel colour (Fig. 2). Larger fractured garnets within the granulites are clearly recognizable in polished hand specimens (Fig. 2c) and are very apparent in a thin section (Fig. 3). The metamorphic conditions during shearing of this pseudotachylyte are estimated at ∼ 600 • C and ∼ 1.1 GPa (Fig. 7 of Hawemann et al., 2018).

Optical microstructure
Granulite facies garnet porphyroclasts in Musgravian peraluminous gneisses that were mylonitized during the Petermann Orogeny are almost invariably fractured, irrespective of their proximity to pseudotachylyte (Fig. 3). Large garnet porphyroclasts (> 1 mm) are typically slightly elongated with their long axis parallel to the foliation, which is attributed at least partially to resorption. Fractures in garnets often show offsets on the order of a few hundred micrometres. It is not possible to determine whether these offsets are primarily due to the initial shear fracture or result from subsequent sliding during ongoing ductile shearing. Moreover, no consistent sense of shear can be derived from the offsets (Fig. 3a, b). These discrete fractures are sub-planar, commonly have a consistent orientation at a moderate angle to the foliation and locally occur in conjugate sets (Fig. 3b). Wide fractures are filled with biotite, kyanite and quartz (Fig. 4b). A later generation of unfilled fractures, without any discernible offset, is oriented perpendicular to both the foliation and stretching lineation (Fig. 3b). Garnet porphyroclasts commonly contain rutile exsolution lamellae and inclusions of monazite and kyanite ( Fig. A1 in Appendix). The latter are present as aggregates with an overall prismatic shape, possibly represent-ing pseudomorphs after sillimanite (Camacho and Fitzgerald, 2010).

Analytical techniques
Quantitative mineral compositions were measured with a JEOL JXA-8200 electron probe micro-analyser (EPMA), equipped with a tungsten filament, at the Institute of Geo- chemistry and Petrology at ETH Zurich (Switzerland). Natural standards were used for quantification, and, when available, natural garnet standards were preferred. To reach a spatial resolution of about 1 µm, an acceleration voltage of 10 kV was set (Fig. 8 in Hofer and Brey, 2007). Elemental maps were acquired using energy wavelength-dispersive spectrometers in parallel for calcium, to increase the signal-to-noise ratio. Backscatter electron (BSE) images, energy-dispersive spectrometry (EDS) and electron backscatter diffraction (EBSD) mapping was carried out on a Quanta 200F field emission gun scanning electron microscope (FEG-SEM) at the ScopeM (Scientific Center for Optical and Electron Microscopy, ETH Zurich). EBSD maps were collected with an acceleration voltage of 20 kV, a sample tilt of 70 • and a working distance of 15 mm. Data were post-processed using chemical indexing with the OIM Analysis 7 software by EDAX. When necessary, three different clean-up techniques were used: the neighbour confidence index correlation, neighbour orientation correlation and grain dilation. Point and map analyses, as well as BSE images, were combined for correlation with optical microscope images in a project using the QGIS software programme from the Open Source Geospatial Foundation. Two lamellae were cut with a focused ion beam (FIB) for transmission electron microscopy (TEM). The microscope used for TEM is a Tecnai F30 with an FEG source operated at 300 kV and equipped with a Gatan 794 MultiScan charge-coupled device (CCD) (ScopeM, ETH Zurich).

Compositional gradients
Granulite facies garnet has a homogeneous composition of X Alm 0.54, X Pyp 0.40, X Grs 0.03 and X Sps 0.03, whereas garnet neocrystallized during the Petermann Orogeny is more Ca-rich (X Alm 0.48, X Pyp 0.28, X Grs 0.22 and X Sps 0.02). Grain boundaries of granulite facies garnet and fractures are decorated with a Ca-enriched rim, 20 to 40 µm wide (Fig. 4c). The length scale for variation in Fe (X Alm ) and Mg (X Pyp ) is identical to that for Ca (X Grs ), whereas the Mn content (X Sps ) does not show any variation (Fig. 4d). Neocrystallized garnet is present where the grain boundary is in contact with or close to plagioclase. The outermost rim of remnant garnet has the same composition as the neocrystallized garnet (Fig. 4d, profile 1). The granulite facies plagioclase is partially transformed to a more Na-rich plagioclase with needle-shaped inclusions of kyanite (bottom of Fig. 4e). This reaction provides Ca for the observed diffusion into garnet (Camacho et al., 2009).
Along fractures across the porphyroclasts, the Ca enrichment is narrower than along the grain boundaries, and the grossular component only reaches up to about X Grs 0.1 (Fig. 4d, profile 2). Compositional gradients are also present around inclusions in garnet connected to the outer garnet boundary, providing evidence of Ca diffusion along grain boundaries (right part of Fig. 4c, profile 3 in Fig. 4d). Profile 4 (Fig. 4d) was measured next to a kyanite inclusion: the diffusion length is still comparable to those of profiles 1-3, but Ca concentrations are much lower. Ca probably diffused along fractures (invisible in the plane of the thin section) towards the inclusion. In summary, the diffusion length at the original grain boundaries is maximized where it is in contact with plagioclase, and otherwise its width is constant at about 20 µm. However, variations in diffusion lengths do occur around garnet fragments, without any correlation with the proximity to plagioclase, although the exact relationship in the third dimension is unknown. Surfaces with limited diffusion can often be identified as fracture surfaces, which were exposed to diffusion for a shorter time than original grain boundaries (Fig. 4e). Fractures oriented perpendicular to the foliation and stretching lineation lack any signs of diffusion and are therefore interpreted as later-stage extensional fractures. Some garnets display more complicated compositional patterns, with zones > 100 µm of Ca enrichment extending into the porphyroclast's interior, which are not associated with fractures (e.g. the garnet fragment on the far right in Fig. 4e). EBSD analysis highlights that the three fragments in the right part of Fig. 4e most likely originated from the same grain, as they share a common rotation axis (Fig. 4f). The colours in the inverse pole figure map are not solid, reflecting slight variations of orientation within the crystal. Furthermore, the image quality map shows areas of suppressed Kikuchi patterns (grey value), which is suggestive of higher dislocation density and therefore possible subgrain boundaries (Fig. 4f). The misorientation angle map (Fig. 4g) reveals a complex pattern of varying crystal orientation (all within the order of 5 • ) in the fragments, with very distributed zones connected to the edges of the crystal, triangular-shaped zones of misorientation (upper left of Fig. 4g) and discrete zones (lower right of Fig. 4g). The discrete zones of misorientation, about 5 µm wide, correlate well with the Caenriched zones (compare Fig. 4e, f, garnet fragment on the right).

Texture of deformed garnets
Two to three orientations of fractures are generally present in a single garnet crystal and coincide with the trace of the (101) plane derived from EBSD data (Fig. 5a, b). Fracture set (I) in the example of Fig. 5a is often associated with a relative rotation of both sides, as visible from the difference in colour. In the lower part of the grain, where the fracture density is very high, more subgrains are present. The subgrain spatial density increases towards the original grain boundary, and some subgrains are "eroded" by ductile shearing and strung out along the foliation. This demonstrates that ductile shearing outlasted subgrain formation and fracturing. The fractures described above are all crosscut by extensional fractures (set II in Fig. 5a), which are oriented perpendicular to the stretching lineation and foliation. These do not show any associated distortion of the crystal lattice.
The garnet porphyroclast of Fig. 5c shows a central fracture as well as a set of two other parallel fractures. The central fracture is the only one with a significant offset and is filled with kyanite and quartz. This fracture displays misorientations of more than 5 • towards the right-hand side of the scan, but it has none towards the left-hand side. In the lower left corner of the fragment, subgrains are observed with misorientations, relative to the average orientation, typically in the range of 10 • . Misorientation axes are often parallel to (111) and (101). The lowermost fragment shows a wide zone of progressive rotation. The chemical profile in Fig. 5e shows the highest Ca counts towards the boundaries of the porphyroclasts and, internally, towards two fractures. The larger fracture with an apparent offset of the two garnet fragments exhibits a less well-developed zone of Ca enrichment when compared to the tight fracture with introduced lattice distortion.

TEM investigations
The garnet fragment of Fig. 4g was further investigated using TEM, as it includes a narrow zone of misorientation without fractures, and it is therefore suitable for the preparation of FIB lamellae. As visible in Fig. 6a (around profile 1), the image quality map shows a well-defined narrow, darker-grey band, possibly indicating high dislocation density. The zone is even more evident in the misorientation plot (Fig. 6b) and changes from about 5 µm wide, with discrete boundaries to the right, to a wider (> 10 µm) band towards the left of the image. In the upper left part of the image, a subgrain boundary with > 5 • misorientation transitions into a zone of gradual misorientation. The misorientation axis is consistently parallel to (101) with minor rotation around (111) (Fig. 6c,  Fig. A2). Misorientation profiles reveal a slight asymmetry within the narrow band, where the lower boundary appears to be sharper. Misorientation changes more gradually within the wider portion of the misorientation band. Locally, subgrains developed with discrete boundaries, documenting a misorientation of usually around 5-10 • (profile 3 in Fig. 6d). The FIB lamella was cut across the narrow band of misorientations (Fig. 6e). The lower boundary corresponds to a narrow discrete zone, without visible dislocations (Fig. 6f). The upper boundary is marked by a series of dislocation walls, and only a few free dislocations are visible, which are often organized in arrays (Fig. 6g, h). The existence of dislocation walls and subgrain boundaries indicates recovery by dislocation climb (e.g. Hobbs, 1968;Passchier and Trouw, 2005).

Discussion
Garnets in this study show evidence for both brittle and ductile deformation under relatively low temperatures of about 600 • C, as inferred from synchronous diffusion and ductile shearing of pseudotachylyte . This is below the experimentally determined values for the onset of crystal plastic deformation of garnet (Wang and Ji, 1999) at the higher strain rates considered typical of mylonitic shear zones (> 10 −14 s −1 ). In contrast to experiments, many natural examples (Vollbrecht et al., 2006;Bestmann et al., 2008;Austrheim et al., 2017) indicate that crystal plasticity of garnet can occur at lower temperatures between 650 and 700 • C.
The presence of microstructures and textures consistent with dislocation climb and recovery, as well as subgrain rotation, in garnet at around 600 • C is in agreement with previous studies (Bestmann et al., 2008;Massey et al., 2011). No evidence for grain boundary sliding is observed, since subgrains show rotation around a specific crystallographic axis. Rotation around (111) and (101) is in accordance with the slip systems described by Voegelé et al. (1998). Multiple generations of overprinting fractures with different orientations demonstrate repeated fracturing events. Extensional fractures do not show any induced lattice distortion or diffusion and therefore occurred after the temperature had decreased to values too low for diffusion (Camacho et al., 2009), possibly during exhumation (compare Prior, 1993 andJi et al., 1997).
In  Fig. 4f) with darker zones that can be interpreted as areas of high dislocation density and the location of the FIB lamella. (b) A misorientation plot with respect to the reference point (marked with the white x) shows a discrete zone of misorientation, which has discrete boundaries in the right part of the image but is more distributed towards the left. (c) A misorientation axis plot with respect to the average orientation of the grain shows a consistent rotation around the (101)  ing", "shattering" or "fragmentation" of garnet is observed in relict porphyroclasts immediately adjacent to pseudotachylyte. The fractures described here are generally planar and often consistently oriented, in some cases showing single and conjugate shear offsets. Fractured garnet is not restricted to the boundary with pseudotachylyte and is still present even in samples without pseudotachylyte, where the nearest pseudotachylyte is possibly many metres or more away. Fracturing in this case cannot be related to thermal shock (Papa et al., 2018) or localized high stress due to (seismic) frac-ture propagation (Austrheim et al., 2017;, but it must reflect a larger-scale distribution of differential stresses in the lower crust that were, at least transiently, high enough to cause brittle-garnet failure . This could be due to stress pulses from earthquakes in the shallower brittle regime (Trepmann and Stöckhert, 2002;Ellis and Stöckhert, 2004;Jamtveit et al., 2018aJamtveit et al., , b, 2019 or a more local, lower-crustal source due to jostling of lessdeformed strong blocks within an irregular shear zone network . The narrower Ca diffusion profiles on some fractures relative to garnet rims and crosscutting relationships suggest that fracturing was recurrent under sub-eclogite facies metamorphic conditions, as also indicated by the occasional presence of kyanite in some fractures. The presence of kyanite needles and the absence of zoisite/clinozoisite or epidote, as a breakdown product of plagioclase during sub-eclogitic metamorphism (Fig. 3b), indicate relatively dry lower-crustal conditions . According to Wayte et al. (1989), this indicates a water activity of < 0.004, calculated for rocks of comparable composition and pressuretemperature conditions. However, new biotite did form in fractures across relict garnet, so conditions were probably not strictly anhydrous. The sheared and recrystallized pseudotachylyte developed a similar synkinematic assemblage as the host mylonite, demonstrating that there is also no marked partitioning of water into the frictional melt, which implies little free or bound water available in the original source rock (e.g. Wex et al., 2018). The effect of pore fluid pressure on the effective confining pressure must therefore have been negligible.
As reported in Hawemann et al. (2019), the dynamically recrystallized quartz grain size and microstructure in the host rock mylonites indicates that long-term flow stresses were not particularly high, on the order of less than 10 MPa. The ambient pressure of ca. 1.1-1.2 GPa determined for the host rocks should therefore be close to the lithostatic value (Mancktelow, 2008c). Figure 7 shows a simple linear plot of the Mohr-Coulomb failure criterion for an angle of internal friction of 30 • (coefficient µ = 0.6), a lithostatic load of 1.2 GPa and no pore fluid pressure. This plot is only qualitative, since the angle of internal friction could decrease towards higher pressure (Shimada et al., 1983 ). However, the summary of experimental results in Byerlee (1978) indicates that there may be little change at least up to pressures similar to those considered here. It follows that the differential stress for fracture initiation must have been on the same order as the confining pressure (Fig. 7). As discussed in detail in Hawemann et al. (2019), such high differential stresses, leading to garnet fracture and the development of abundant pseudotachylyte, can only have been transient and presumably related to repeated short-term seismic events in the lower continental crust Jamtveit et al, 2018a, b;Menegon et al., 2017). The lack of shattered garnet adjacent to pseudotachylyte in these samples may reflect drier condi- tions relative to those in the Bergen Arc (Austrheim et al., 2017) and Mont Mary (Papa et al., 2018). The samples studied could therefore represent one end-member of the lower continental crust, where deformation occurs without the initial presence or influx of free water during fracturing and subsequent crystal plastic deformation.

Conclusions
In dry lower continental crust deformed under conditions of ca. 600 • C and 1.1 GPa, garnet shows both single and conjugate sets of shear fractures, fractures with associated subgrains and induced lattice damage around fractures, subgrain formation without fracturing, and late-stage extensional fractures. Most of these fractures show a strong crystallographic control, with fracturing preferentially occurring along the (101) planes of garnet. Dynamic recrystallization is evident from inferred subgrain rotation recrystallization, and recovery is manifested by the presence of dislocation walls. The observed microstructures of garnets are interpreted to record transient high stresses during deep seismic events in the lower-crustal Fregon Subdomain. This is also indicated by the abundance of pseudotachylyte which developed under similar lower-crustal conditions and, possibly, by the variability of recrystallized quartz grain sizes including values down to a few micrometres (Hawemann et al., 2009). The studied example represents one end-member of lowercontinental-crustal behaviour where, due to earlier metamorphic dehydration and the intracratonic position well removed from the plate margin, rocks were initially dry and water was not introduced during fracturing and crystal plastic deformation.
Data availability. All data used in this paper can be accessed through the depository of the Open Science Framework at https://doi.org/10.17605/OSF.IO/YRZGH (Hawemann, 2019).
Author contributions. All authors listed took part in at least two of the three field seasons. NM assisted FH in data collection and interpretation. AC's and GP's knowledge in the field of garnet deformation and diffusion processes were crucial in preparing the paper. SW contributed to the microprobe work and research involving scanning electron microscopy. FH prepared the paper with contributions from all co-authors.