The Effect of the Garnet Content on Deformation Mechanisms and Weakening of Eclogite: Insights From Deformation Experiments and Numerical Simulations

We performed deformation experiments on omphacite‐garnet aggregates at a temperature of 1000°C, a confining pressure of 2.5 GPa, and a strain rate of 3 × 10−6 s−1 and complemented them by numerical simulations to gain insight into the role of garnet fraction for the deformation behavior of dry eclogite, with a focus on strain weakening mechanisms. We determined the spatial and temporal evolution of strain and strain rate by basing numerical simulations on experimentally derived microstructures, and thereby identified characteristic deformation mechanisms. Pure omphacite and garnet aggregates deform by two different mechanisms. Internally strained clasts and low‐angle grain boundaries indicate crystal plasticity for omphacitite; the fracture dominated fabric of garnetite documents brittle deformation. Electron channeling contrast imaging, however, revealed low‐angle grain boundaries and free dislocations in garnet crystals, suggesting that minor crystal plasticity accompanies the brittle failure. Eclogitic aggregates show varying deformation behavior between the two end‐members shifting from crystal plastic toward brittle deformation with increasing garnet content. All samples exhibit strain weakening. The intensity of weakening shows a positive correlation with the garnet content. Our combined experimental, numerical, and microstructural investigations suggest that the majority of strain weakening is associated with crystal plastic processes in omphacite. Numerical simulations and experiments show that a garnet content above 25% enhances the activity of crystal plastic processes in omphacite and results in strain localization, which subsequently weakens the eclogite.

Recently, fluid-supported formation and deformation of eclogites have been suggested to result in strain weakening when the main strain-accommodating mechanism is dissolution-precipitation creep (Cao et al., 2020;Rogowitz & Huet, 2021;Stünitz et al., 2020). Such fluid-assisted eclogitization and deformation is highly relevant in subduction zones where hydrous conditions are likely (Wassmann & Stoeckhert, 2013, and references therein); yet, several convergent settings are assumed to be rather dry (Altenberger & Wilhelm, 2000;Engvik et al., 2007) and there, the rheology is expected to be dislocation-creep dominated (Engvik et al., 2007;Kurz et al., 2004). Experimental investigations showed that eclogite (Jin et al., 2001;Zhang & Green, 2007;Zhang et al., 2006) and its two main constituents, omphacite (Moghadam et al., 2010;Zhang et al., 2006) and garnet (Li et al., 2006;Mei et al., 2010), are not expected to creep at differential stresses below 1 GPa for tectonically relevant strain rates. Nevertheless, highly localized shear zones and mylonitic fabrics are frequently observed in eclogite (Engvik et al., 2007;Keppler et al., 2016;Kurz et al., 1998Kurz et al., , 2004, raising the question of how and why localization occurred. These mylonites are interpreted to have formed after eclogitization, and therefore, transformation weakening can be excluded as a localization mechanism. Instead, the onset of localization might be facilitated by the presence of brittle precursors, as often observed in upper to lower crustal shear zones (Mancktelow & Pennacchioni, 2005;Menegon et al., 2017;Rogowitz et al., 2014;Segall & Simpson, 1986). Brittle precursors at dry eclogitic conditions have often been linked to seismic strain rates (Austrheim et al., 2017;Jamtveit et al., 2019;Petley-Ragan et al., 2019). In the absence of seismicity, a garnet content above 40% in eclogite can initiate brittle failure and potentially introduce precursors for further localization (Yamato et al., 2019). However, it remains unclear as to which degree the change from crystal plastic to brittle deformation with increasing garnet content influences the rheological evolution of eclogite, in particular its weakening. We combined deformation experiments with numerical simulations to investigate deformation behavior and potential weakening mechanisms of dry garnetite, omphacitite and eclogite with varying garnet fractions to gain insights into small-scale processes operating at HP conditions in convergent settings. Sample compositions are representative of eclogites found in LT-HP to UHP settings (Angiboust et al., 2011;Engvik et al., 2007;Hertgen et al., 2017;Ji et al., 2003;Terry & Heidelbach, 2006). Our approach allows us to analyze the extent to which omphacite and garnet mutually affect their deformation behavior and consequently the overall deformation regime. Detailed microstructural investigations are combined with numerical simulations to unravel location and timing of deformation localization and the operative mechanisms in eclogite depending on the content of its two major phases.

Starting Material
Natural eclogite from the Koralpe (UTM33, 511146°E/5175207°N; Eastern Alps, Austria) served as the starting material (Miller & Thöni, 1997). The rock was crushed and ground to below the natural grain size. Garnet and omphacite were separated according to their magnetic and gravimetric properties; the separated powders were dry-sieved to a grain size fraction between 40 and 70 μm. Powders of omphacite, garnet, and garnet-omphacite mixtures containing 25, 50, and 75 volume percent garnet were uniaxially cold pressed with 3 MPa into gold capsules with a wall thickness of 0.2 mm, which were mechanically sealed. Prepared samples with a height of 10-12 mm and a diameter of 3.6 mm were hot pressed in a piston cylinder apparatus at 1100°C and 3 GPa for 72 hr.

Deformation Experiments
Triaxial deformation experiments were performed in a servohydraulically controlled modified Griggs-type deformation apparatus (Griggs, 1967;Rybacki et al., 1998) at a confining pressure of 2.5 GPa, a temperature of 1000°C, and a strain rate of 3 × 10 −6 s −1 (Table 1). Synthesized samples were prepared in a cylindrical shape with 10.1029/2022GC010743 3 of 23 a diameter of 3.1-3.4 mm and a height of 6.5-7.5 mm and placed in a second gold jacket, again mechanically sealed. Samples were stored in a drying oven at 80°C; however, minor amounts of water might have entered the sample in the course of the experiment (Joachim et al., 2012). Details on the experimental procedure and data treatment can be found in Text S1 of Supporting Information S1.

Numerical Simulations
We conducted a series of numerical simulations simulating sample deformation using a digitized representative area (minimum input size 743 × 860 μm) of thin sections of undeformed samples (S10_25grt, S10_grt50, S28_grt25 and S28_grt75) to complement the microstructural analyses of deformed samples. The focus of the simulations lies on the spatial distribution of strain and deformation mechanisms until peak stress is reached. In the simulations, rheological weakening is limited to brittle deformation so that a comparison of post-peak behavior of numerical simulations and experiments will allow us to deduce the importance of brittle and ductile weakening mechanisms.
We employed the code LaMEM , which applies a marker-in-cell finite difference staggered grid approach, to solve Stokes equations for incompressible media where τ ij are the components of the deviatoric stress tensor, p the pressure, and v i the components of the velocity vector. The rheology of the constituents, omphacite and garnet, is modeled as a combination of elastic, viscous and brittle deformation. We assume that all three deformation mechanisms operate independently and thus add up to the total strain rate :̇=̇+̇+̇ The elastic strain rate is given bẏ= where μ is the shear modulus and ∂ /∂t is the Jaumann objective stress rate defined as with the spin tensor Viscous deformation is modeled using a power law of the form: where A, E a , and n are material parameters (Table 2). Finally, the strain rate due to brittle failure is given bẏ where ̇ is the plastic multiplier and Q the plastic potential. For simplicity, we used a von Mises failure criterion with a yield function F and a plastic potential given by where τ y is the maximum strength of the material and = √ 1 2 is the effective stress. For F > 0, brittle failure occurs and ̇ and therefore also assume finite values (de Borst & Vermeer, 1984). In addition to prescribing the maximum strength of both omphacite and garnet, we also employ a strain weakening parameterization that reduces τ y as a linear function of the accumulated irreversible strain ε irr , which is defined as follows: The maximum strength τ y is then given by the equation: where τ y,0 is the initial maximum strength, ε 1 the irreversible strain at which weakening initiates and ε 2 the irreversible strain at which weakening stops. Once the irreversible strain reaches ε 2 , strength is reduced by the factor f r .
Dislocation creep stress exponents and activation energies were chosen based on the most suitable published creep laws (see Table 2), whereas the dislocation creep prefactors, the elastic shear modulus and the failure stress were modified by trial and error to approximately fit the stress-strain curves for the three experiments. As the numerical models are conducted as plane-strain two-dimensional deformation experiments, these modified prefactors comprise the conversion factors otherwise necessary to account for the difference in deformation geometry (e.g., Cyprych et al., 2016;Gerya, 2009). To compare numerical simulations with laboratory data, the average stress of the sample was computed by integrating the normal stresses at the top boundary of the model domain and dividing the result by the domain width.

Microstructural Analysis
Synthesized and deformed eclogite samples were cut in half along the cylindrical axis to prepare mechanically polished 30 μm thin sections for optical microscopy, scanning electron microscopy (SEM), and electron microprobe analysis. Thin sections selected for electron backscatter diffraction (EBSD) mapping, orientation contrast (OC) imaging, and electron channeling contrast imaging (ECCI) were chemo-mechanically polished for a minimum of 4 hr with an alkaline colloidal silica suspension (Syton©) and subsequently carbon coated. Mineral chemistry was measured with a CAMECA SX Five SE electron microprobe operated at an acceleration voltage of 15 kV and a beam current of 20 nA. Structural formulae were calculated using the MINSORT software (Petrakakis, 1985).
Secondary electron microscopy, including EBSD mapping, OC and ECC imaging, was performed on a FEI Quanta 3D FEG scanning electron microscope equipped with secondary electron-(SE), backscatter electron-(BSE), energy dispersive x-ray-(EDX), forward scatter-(FSD) and EBSD-detectors. The SEM was operated at an accelerating voltage of 15 kV, a beam current of 4 nA in analytical mode, a working distance of 14-16 mm and a sample tilt of 70°. For all maps, a 9 × 9 convolution mask was applied to determine 3 to 14 peaks at a minimum peak distance of 8 or 10 pixels in Hough space. Based on pattern quality and map size, a binning of 4 × 4 or 8 × 8 was used. A hexagonal grid and a step size ranging from 0.18 to 0.50 μm was used for EBSD mapping. The OIM Data Collection and Analysis software was used for indexing and processing the EBSD data. Data were cleaned based on the confidence index (CI > 0.1) and by applying the neighbor orientation correlation (clean-up level 2). Orientation distribution functions were calculated and plotted as equal angle upper hemisphere projections. Data were plotted as color coded grain average misorientation maps, grain boundary maps, and misorientation deviation angle maps representing the misorientation of each pixel with respect to the grain average misorientation. All color coding is overlying image quality maps, where darker gray shades denote lower image quality.
For ECCI, the SEM was operated at 15-25 kV accelerating voltage, a beam current between 2 and 4 nA in analytical mode at a working distance between 7 and 8.5 mm. To obtain visibility of lattice defects, the crystal in consideration was tilted such that strong channeling occurs (i.e., the primary beam falls in under the Bragg angle of a low-indexed lattice plane; Rogowitz et al., 2018;Zaefferer & Elhami, 2014). Samples were tilted between −4 and 4° in steps of 0.5°.

Starting Material
The synthesized eclogite samples are characterized by a weak foliation defined by a minor shape-preferred orientation of omphacite grains (Figures 1a and 1b). Omphacite grains are characterized by straight grain boundaries and range in size from 8 to 100 μm, exhibit deformation twins and minor undulatory extinction. Minor intracrystalline strain is confirmed by EBSD maps revealing minor grain average misorientations ( Figure S1 in Supporting Information S1). Omphacite exhibits a weak crystallographic preferred orientation (CPO) with [001] aligned on a girdle in the foliation, while poles to (010) and (100) appear randomly orientated ( Figure 1c). Garnet often occurs in clusters composed of grains ranging in size between 15 and 80 μm devoid of a shape preferred orientation (Figures 1a and 1b). Locally, fragmented garnet grains show a minimum fragment size of 3 μm. The crystallographic orientation of garnet crystals resembles a random pattern with no detectable preferred alignment ( Figure 1d).
The jadeite content of the omphacite in the natural rock has an average of Xjd = 0.37; the garnet is iron rich and has an average composition with Xalm = 0.46, Xpyr = 0.28, Xgrs = 0.25, and Xsps = 0.01. After hot-pressing, a chemical zoning along grain boundaries and partly along garnet-omphacite interfaces occurred; omphacite is enriched in iron and depleted in magnesium while the opposite applies to garnet (Table S1 in Supporting Information S1). A similar variation in garnet composition occurs along healed micro-fractures.

Mechanical Data-Deformation Experiments
The mechanical behavior of the synthetic eclogite samples varies systematically with their garnet content. Differential strength increases with increasing garnet content following a nearly log-linear behavior ( Figure 2). The nearly linear initial increase in differential stress with axial strain is similar for the samples with 0% and 25% garnet fraction but increases with garnet content above 25%. The yield strength of samples (i.e., point where the differential stress-strain curve deviates from a linear relation) is reached between 3% and 5% strain, while the peak stress, the compressive strength, is reached between 6% and 8.5% strain (Table 1), without systematic correlation to specimen composition. All samples exhibit nearly constant stress for about 5% strain before strain weakening sets in, as indicated by a continuous decrease of differential stress with increasing axial strain. A comparison of the differential stress drop after a strain of 18% shows that the amount of weakening increases in samples with a garnet content above 50% (Table 1, Figure 2b).

Microfabrics
All deformed samples show a strain gradient from the less deformed top to the more deformed bottom, as evidenced by variations in sample diameter. Reported strain values represent the average sample strain. Microstructural investigations focus on the lower half of the samples where strain is likely to be higher than the average value.

Garnetite and Omphacitite
Garnetite shows a cataclastic fabric. Angular to sub-angular fractured garnet clasts are partly offset along conjugated micro-faults ( Figure 3a). The micro-faults are characterized by traces of fine-grained garnet fragments with varying crystallographic orientation (see OC in Figure 3b). Low-angle grain boundaries detected by EBSD are often identified as micro-fractures by ECCI. However, locally minor variation in crystal orientation does occur within garnet clasts. Frequently, OC is cross-cut by micro-fractures (Figures 3b and 3c). Such zones are characterized by dislocation arrays arranged to low-angle boundaries ( Figure 4a). In accord with low misorientation between clast and subgrain (below 2°), the density of dislocations within the low-angle grain boundary is rather low. An increased dislocation density is only present next to micro-fractures ( Figure 4b) or in the vicinity of micro-fracture networks ( Figure 4c).
The pure omphacite sample exhibits a mylonitic fabric. A minor foliation is defined by elongated omphacite clasts with a maximum grain average misorientation of 4.3°. Clasts often reveal a higher misorientation angle along grain boundaries than in grain interiors, locally reaching maxima of 33° ( Figure 3d). Omphacite grains with low average misorientation display continuous linear misorientation patter (Figures 3g and 3h), whereas highly strained omphacite clasts exhibit irregular shaped subgrains and bulges (<3 μm) at grain boundaries (Figures 3e and 3f). Such highly strained clasts are often surrounded by fine-grained, close to euhedral-shaped omphacite, devoid of internal distortion ( Figure 3g).

Eclogite
The deformed eclogite samples are characterized by a distinct foliation whose characteristics vary with garnet content. In samples with 25% and 50% of garnet content, garnet clusters and omphacite clasts that are elongated perpendicular to the maximum principal stress define the foliation (Figures 5a and 5c). Garnet clusters are composed of heterogeneously sized crystals and commonly exhibit fine grained (<15 μm) tails perpendicular to the shortening direction ( Figure 5c). In samples with a volume fraction of 75% garnet, flattened omphacite (a) Differential stress-strain curves for garnet-omphacite aggregates experimentally deformed at a confining pressure of 2.5 GPa, a temperature of ∼1000°C, and a strain rate of ∼3 × 10 −6 s −1 . (b) Maximum differential stress and differential stress at an axial strain of 18% plotted against the garnet content in the specimen. In all deformed eclogite samples, omphacite grains show deformation twins and undulatory extinction ( Figure 5). Crystal lattice distortion is confirmed by EBSD analysis that reveals a grain average misorientation of up to 4.1° for omphacite ( Figure 5b) that exceeds the grain average misorientation of garnet with a maximum of 3.0° (Figures 5d and 5f). Despite an overall low average misorientation of garnet crystals, elongated garnet grains display an OC (Figures 6b and 6c). The low-angle grain boundary density increases in both garnet and omphacite grains with increasing garnet content and tends to be higher in grains with a higher grain average misorientation (Figures 5b, 5d, and 5f). In some cases, low-angle grain boundaries detected by EBSD correspond to micro-fractures. In all experimentally deformed eclogites, we observe that crystal distortion of omphacite is most pronounced at phase boundaries, whereas garnet crystals show the highest internal misorientation at grain contacts in clusters or along micro-fractures ( Figure 6a). With the exception of the 25% garnet sample, deformed eclogites show a network of conjugated MSZ with respect to the maximum compressive stress direction. The microstructure of those MSZs is variable and described in detail in Section 3.3.3.
In all eclogite samples, omphacite shows a CPO with (010) poles orientated sub-perpendicular to the foliation, while [001] axes are either aligned on a girdle in the foliation (25% and 50% garnet fraction) or in single maxima within the foliation (75% garnet fraction; Figure 7). The orientation of (100) poles is not as systematic and varies with the garnet fraction. The crystallographic orientation of garnet grains is variable for all samples and texture strength is weak with a maximum of 2.2. Nevertheless, in eclogite containing 50% garnet, a crystallographic dependence is observed between euhedral small grains and adjacent clasts (Figure 8f).
Similar to the starting material, garnet in deformed eclogite has an average composition with Xalm = 0.45, Xpyr = 0.28, Xgrs = 0.26, and Xsps = 0.01, and omphacite an average composition of Xjd = 0.38, Xq = 0.62, and Xae = 0.01, where q represents the sum of enstatite, ferrosilite and wollastonite (Table S2 in Supporting Information S1). Minor chemical variation occurs along grain and phase boundaries and healed fractures. However, the zoning is not systematic and appears to be a relic from synthesis experiments.

Conjugated Micro-Shear Zones in Eclogite
Conjugated MSZ approximately 45-55° inclined to the direction of the maximum compressive stress occur in samples with a garnet content ≥50% (Figures 8 and 9). Micro-shear zones are between 1 and 10 μm wide and show a maximal displacement of 20 μm. The density of these conjugated MSZ increases with the garnet content in the samples.
Micro-shear zones in samples with 50% garnet are characterized by omphacite grains smaller than 3 μm with irregular grain boundaries. Grains are elongated and orientated with their long axis oblique to the MSZ boundary  indicating the direction of shearing (Figure 8c). In the vicinity of the MSZ, omphacite clasts show evidence of significant intracrystalline plastic deformation. Orientation contrast images reveal irregular changes in crystal orientation that are confirmed by misorientation deviation angles of up to 31° (Figures 8a and 8b). Grains bordering MSZs are composed of irregularly shaped subgrains with a size up to 3 μm (Figure 8d). When cut by MSZs, garnet is characterized by irregular shaped fragments and micro-porosity (Figure 8e). Furthermore, garnet grains in MSZs are characterized by local distortion of the crystal lattice, development of 2-5 μm subgrains, and fracturing (Figures 8d and 8e). The size of subgrains in garnet clasts is comparable to that of the smallest observed grains in clusters (Figure 8d). Shape-wise the subgrains are similar to the euhedral to roundish grains in clusters, but appear slightly more irregular. Strained clasts and adjacent grains exhibit similar crystallographic orientation (Figure 8f).
In the sample containing 75% garnet, localization of deformation in MSZs is dominated by fracturing (Figure 9). The OC observed in garnet clasts can be directly linked to fracture networks associated with nano-sized fragments (Figures 9b and 9c). Elongated omphacite clasts show irregular misorientation patterns and are offset along distinct zones traced by micro-to nanometer sized grain fragments and increased nano-porosity ( Figure 9d). Locally, subgrains and deformation twins occur in omphacite grains; however, these are cross-cut by MSZ-associated fractures (Figure 9e).

Numerical Simulations
Differential stress-strain curves obtained from numerical simulations show an increase in tangent elastic modulus and compressive strength with garnet content (Figure 10). Yield strength is reached after a strain of about 2.5% independent of eclogite composition. While the simulations for 25% and 50% garnet indicate steady state behavior with constant stress, eclogite with 75% garnet shows minor strain weakening.
Numerical simulations for all eclogite compositions show initial strain partitioning with strain rates in omphacite grains exceeding that in garnet grains. In eclogite with a 25% garnet fraction, strain and strain rate continue to be relatively homogeneously distributed in the omphacite matrix (Figures 11a and 11b top). With ongoing deformation, minor strain localization occurs in omphacite next to phase boundaries, while garnet crystals behave close to rigid and rarely exhibit creep. After a strain of about 5%, fracturing of garnet crystals occurs. Fractures commonly trace zones in garnet that initially evidence higher creep strain.
In eclogite with a 50% garnet fraction, deformation starts to localize at a strain of about 2.5%, with omphacite grain boundaries being the loci of strain localization (Figure 11a, center). The onset of localization coincides with yielding; further localized zones evolve into a conjugated network (Figures 11a and 11b center) and form until peak stress is reached. Subsequently, no further localized zones are formed; instead, existing zones widen to accommodate the strain. Omphacite creeps throughout the simulation while garnet begins to fracture after a strain of 5%. Overall, the garnet in the 50% sample exhibits a higher creep strain than that when the garnet fraction is only 25%.
In eclogite with 75% garnet, strain localization starts at a strain of about 4%. The initial post-yield deformation is characterized by the creep of garnet and omphacite (Figure 11b, bottom). Eventually, the increase in differential stress results in brittle failure of garnet crystals (Figures 11a and 11b bottom). Localized strain zones are initiated up to the peak stress leading to a final strain and strain rate pattern comparable to that for eclogite with a garnet content of 50%.

Deformation Mechanisms Operating in Eclogite and Its Two Main Constituents
At the applied deformation conditions (1000°C, 2.5 GPa, 3 × 10 −6 s −1 ), the two main constituents of eclogite, omphacite and garnet, deform dominantly by two different mechanisms. The pure omphacite sample is characterized by flattened, internally strained omphacite clasts, grain-boundary bulges and low-angle grain boundaries,   all indicative of crystal plastic deformation mechanisms (Bascou et al., 2001(Bascou et al., , 2002Buatier et al., 1991;Kurz et al., 2004;Moghadam et al., 2010;Zhang et al., 2006). The high density of euhedral, nearly strain-free omphacite crystals adjacent to highly misorientated clasts indicates the activation of dynamic recrystallization (Keppler et al., 2016). The occurrence of bulges and subgrains suggests a concomitant activation of recrystallization by bulging and subgrain rotation. In accord with dislocation-accommodated deformation, omphacite develops a CPO. The observed texture with a preferred alignment of (010) poles normal to the foliation and [001] axes aligned in the foliation plane resembles an S-to SL-type CPO (Godard & van Roermund, 1995;Helmstaedt et al., 1972) common for coaxial straining (Godard & van Roermund, 1995;Zhang et al., 2006). The slight deviation from the typical S-type CPO toward SL-type might be caused by local stress heterogeneity.
In contrast to the mylonitic omphacitite, the fracture-dominated fabric of synthetic garnetite evidences brittle deformation. Locally, micro-faults offsetting grains are traced by nano-to micro-sized garnet fragments indicating cataclastic flow. While the high density of fractures and micro-faults, and garnet fragments, heterogeneous in size and orientation, are indicators of cataclasis (Trepmann & Stöckhert, 2002), OC images and EBSD mapping reveal local intracrystalline misorientation. Moreover, ECC images show the presence of free dislocations in garnet crystals. The majority of dislocations can be directly linked to brittle fracturing, either being emitted from the fracture tip or accommodating crystal lattice rotation induced by intracrystalline fracture (Hawemann et al., 2019;Rogowitz et al., 2018). However, the presence of aligned dislocations in low-angle grain boundaries also indicates the activation of thermally activated dislocation climb (Voegelé, Ando, et al., 1998;. A predominantly brittle behavior of garnet accompanied by minor crystal plasticity is a common observation for garnet naturally deformed at temperatures below 750°C and has previously been directly linked to seismically  induced deformation (Angiboust et al., 2012;Austrheim et al., 2017;Hawemann et al., 2019;Petley-Ragan et al., 2019;Trepmann & Stöckhert, 2002). In the absence of seismic strain rates, naturally deformed garnets deform dominantly by dislocation creep (Bestmann et al., 2008;Vollbrecht et al., 2006) or dissolution-precipitation creep (Rogowitz & Huet, 2021;Smit et al., 2011). The latter is bound to the presence of fluids and can be excluded as active deformation mechanism during our experiments. Experiments on garnet single crystals showed a switch from brittle mechanisms to crystal plasticity at a temperature of 1000°C (Voegelé, Ando, et al., 1998), the temperature of our experiments. In contrast to the results of Voegelé, Ando, et al. (1998), the garnet crystals retrieved from our experiments show predominately brittle deformation, which might have several causes. Firstly, the garnet composition in our samples deviates from that of the single crystals in previous experimental investigations possibly affecting mineral strength (Katayama & Karato, 2008). Secondly, the fast strain rates (3 × 10 −6 s −1 ) in our experiments may promote the activation of brittle deformation (Yamato et al., 2019). Finally, locally increased stresses at grain boundaries can exceed the garnet strength and cause brittle failure (Yamato et al., 2019).
The omphacite-garnet mixtures show deformation behaviors between that of their two constituents. In eclogites with a garnet fraction of 75%, a load bearing framework of garnet dominates the overall deformation behavior. Cataclasis of the garnet grains is accompanied by minor dislocation activity. Strained omphacite clasts are dissected by micro-faults, which likely originate from fractures in garnet crystals. With increasing omphacite content, crystal plasticity becomes more relevant for the deformation behavior of eclogite. In samples containing 50% omphacite, either constituent shows evidence of dislocation creep. Strong OC, CPO and high sub-grain density next to small elongated grains indicate dynamic recrystallization of omphacite (Keppler et al., 2016;Piepenbreier & Stöckhert, 2001). Low-angle grain boundaries in flattened and elongated garnet aggregates together with a crystallographic relation of euhedral small garnet grains and adjacent garnet clasts suggest minor activation of subgrain rotation (Bestmann et al., 2008;Prior et al., 2002;Vollbrecht et al., 2006).
The deformation behavior of samples with 25% garnet is dominated by the interconnected omphacite matrix. Heterogeneous misorientation, CPO and deformation twins indicate crystal plastic deformation (Bascou et al., 2001(Bascou et al., , 2002Buatier et al., 1991;Keppler et al., 2016;Moghadam et al., 2010;Zhang et al., 2006). Low average misorientation, limited OC, and a lack of fractures in garnet suggest that neither dislocation glide nor cataclasis accommodates the observed strain in garnet. Instead, the flattening of garnet clusters occurs by rigid body rotation and grain boundary sliding (Bascou et al., 2001;Kim et al., 2018;Piepenbreier & Stöckhert, 2001). The latter is in accordance with the heterogeneous grain size in garnet tails.

Coupled Deformation Behavior of Garnet-Omphacite Aggregates
The deformation behavior of two-phase aggregates has been at the focus of a large number of experimental, numerical and phenomenological studies (Barnhoorn et al., 2005;Bruhn & Casey, 1997;Bruhn et al., 1999;Dimanov & Dresen, 2005;Handy, 1990Handy, , 1994Li et al., 2007;Madi et al., 2005), all indicating a strong interaction of the two phases in the course of deformation. In accord with these previous studies, the microstructural record of the experimentally deformed omphacite-garnet aggregates shows that omphacite and garnet crystals mutually affect their deformation behavior. Omphacite crystals generally evidence the activity of intracrystalline plasticity but localized, recrystallized zones and/or micro-faults develop with increasing garnet content. Our numerical simulations indicate the development of conjugated localized high strain zones with increasing garnet content in accord with the microstructural observations.
The initial strain partitioning with higher deformation rates in omphacite than in garnet crystals is reflected by a glide and twinning dominated omphacite fabric in experimentally deformed eclogite (Zhang & Green, 2007). Preferred strain accommodation by omphacite is a common observation in experimentally (Jin et al., 2001;Zhang & Green, 2007) and naturally deformed eclogite with varying modal compositions (Ábalos, 1997;Keppler et al., 2016;Kurz et al., 2004;Philippot & van Roermund, 1992), and can be assigned to the strong contrast in Figure 10. Numerically determined differential stress-strain curves of eclogite aggregates with indicated garnet contents. The vertical gray lines denote strains that correspond to the snapshots shown in Figure 11. Figure 11. Snapshots of numerical simulations at 0%, 2.5%, 5.5%, and 6.8% axial strain. (a) The effective strain rate and (b) the strain for three different samples with differing garnet content, as indicated on the left of each row (S28_grt75, S10_grt50 and S28_grt25). The three rows of strain maps at the bottom present the amount of strain accommodated by crystal plastic creep ("creep strain," blue) and by brittle failure ("brittle strain," red). The first row shows the digitized phase map, omphacite in green and garnet in red. See Section 4 for details. The color range was adapted to highlight the relevant features and to facilitate comparisons between different samples. In the first three rows, the color range for the strain rate covers 10 −8 to 10 −3 1/s. In the bottom three rows, the color range in the first column covers 0 to 10 −3 , as creep strains are low and brittle failure does not occur. In the remaining three columns, the color ranges for both creep and brittle strain cover 0 to 0.1. mineral strength between omphacite and garnet. However, our investigations show that the evolution of the strain and strain-rate distribution depends on the garnet fraction in eclogite. In eclogite with a garnet fraction of 25%, garnet crystals dominantly behave as rigid objects. Only minor dislocation creep was detected in the garnet crystals of deformed samples (Figure 11b, top) and the fabric and sample strength is dominated by an interconnected weak layer of omphacite (Handy, 1994). Nevertheless, simulations indicate that minor fracturing in garnet occurs at an early deformation stage (5% strain). Fractures commonly trace zones of higher creep strain in garnets, indicating a strain hardening effect ( Figure S2_1 in Supporting Information S1; Hansen et al., 2019). With simulation progress, fractures act as brittle precursors for minor localization of further dislocation creep, a feature previously reported to occur from macro-to micro-scale in naturally deformed rocks (Rawling et al., 2002;Rogowitz et al., 2018;Segall & Simpson, 1986). Structures associated with strain hardening also occur in our experiments in the form of fractures cross-cutting OC in garnet (Figure 6c and Figure S2_1e in Supporting Information S1).
Eclogite with a garnet fraction of 50% and 75% show similar final strain and strain rate patterns with a conjugated network of localized zones (Figure 11) in the experimentally deformed samples and the numerical simulations. The localized zones correspond to micro-faults in garnet-rich eclogite (75% garnet fraction) and a combination of micro-faults in garnet and/or recrystallized shear zones in omphacite. Recrystallized zones can only be observed in eclogite with 50% garnet fraction. Simulations reveal that in eclogites with a garnet fraction of 50%, phase boundaries are the loci of strain localization. The strong competence contrast between garnet and omphacite introduces a rheological heterogeneity, which is known to enhance strain localization ( Figure S2 in Supporting Information S1; Kenkmann & Dresen, 1998;Zhang & Green, 2007). According to the numerical simulations, the onset of localization starts at a strain of about 2.5%, which corresponds to yielding in the simulations ( Figure 10). The increase in differential stress with strain is accompanied by an increasing creep strain in garnet, well in line with the experimentally observed recrystallization processes. Strain hardening is followed by fracturing of garnet, leading to brittle-plastic deformation of eclogite. Contrary to eclogite with 50% garnet, the higher garnet content in the 75% garnet sample causes a switch from distributed glide to localized frictional deformation once the maximal compressive strength is reached. Fractures and micro-faults initiate in garnet and spread into omphacite where they cross-cut older structures, such as OC, deformation twins, and minor low-angle grain boundaries (Figures 9d and 9e; Figure S2 in Supporting Information S1). Our numerical simulations in combination with microstructural investigations suggest that whether fractures act as precursors for strain localization or if fracturing is initiated by crystal plastic-induced strain hardening depends on the garnet fraction in eclogite.

Implications for Small-Scale Processes Operating in Convergent Settings
The deformation behavior of eclogitic rocks has been the subject of a large number of field-based studies from LT-HP to UHP terrains (Angiboust et al., 2011;Austrheim & Griffin, 1985;Boundy et al., 1992;Bras et al., 2021;Cao et al., 2020;Engvik et al., 2007;Hertgen et al., 2017;Keppler et al., 2016;Kurz et al., 1998;Macente et al., 2017;Philippot & van Roermund, 1992;Rogowitz & Huet, 2021;Stünitz et al., 2020;Zertani et al., 2019). Diffusion-based mechanisms, such as dissolution precipitation and Coble creep, dislocation creep, and cataclasis are invoked, and their dominance is likely controlled by fluid content, temperature, pressure, and strain rate. Dislocation creep and cataclasis, supposedly both operate at eclogite facies conditions (Hertgen et al., 2017), a superposition of mechanisms typically assigned to strain-rate fluctuations due to seismic events or episodes of increased pore-fluid pressure (Angiboust et al., 2012;Broadwell et al., 2019;Hertgen et al., 2017). Our investigations add garnet content as a potential cause for the large variety of deformation mechanisms observed in natural eclogites deformed at similar pressure and temperature conditions. Indeed, eclogite with relatively high garnet fractions or incorporated garnetite bodies exhibit brittle deformation while omphacite dominated layers from the same localities are characterized by a mylonitic fabric (Angiboust et al., 2011;Behr et al., 2018;Broadwell et al., 2019;Hertgen et al., 2017). Thus, we suggest that the identified switch from distributed creep to friction-dominated deformation in eclogite with increasing garnet fractions can potentially explain frequently observed fractured and brecciated mylonitic eclogites in the absence of seismic strain rates or fluid. Correspondingly, the switch from frictional to viscous deformation, for example, by a fracture that propagated from a garnet-rich domain into an omphacite dominated layer, can act as a brittle precursor for the nucleation of shear zones (Engvik et al., 2007;Mancktelow & Pennacchioni, 2005;Menegon et al., 2017).

Strain-Weakening in Garnet-Omphacite Aggregates
The experimental and numerical data on eclogite strength presented here are in line with previous deformation experiments on eclogite samples that also revealed an increase in strength with garnet content (Jin et al., 2001;Zhang & Green, 2007). The correlation between strength and garnet content is further in accord with theoretical mixing models for two-phase aggregates Tullis et al., 1991), garnet constituting the strongest of the two phases. However, these models neglect a deformation-induced microstructure evolution and a potential switch in deformation behavior. Our experimental and numerical investigations demonstrate a switch from crystal plastic to brittle dominated deformation with increasing garnet content and thus support the results of Yamato et al. (2019) who demonstrated a variation in deformation behavior of eclogite with varying volume fraction of the constituents based on numerical simulations.
Our deformation experiments document a decrease in differential stress with increasing strain, indicating an onset of strain weakening. The intensity of this strain weakening increases for a garnet content >50%. In samples with a garnet content ≥50%, the weakening is probably related to the MSZ that are traced by fine-grained material formed either by cataclasis (75% garnet fraction) or combined recrystallization of omphacite and fracturing of garnet (50% garnet fraction). The increase in weakening in eclogite with garnet content might be related to two processes. (a) Cataclasis, which is more common in eclogite with a higher garnet fraction, might favor strain localization. (b) The overall MSZ density increases with increasing garnet fraction. Numerical models performed by Poliakov and Herrmann (1994) showed that the density of conjugated fractures and MSZs in a rock depends dominantly on two factors: the deformation rate and the bulk shear modulus of the rock. Experiments and simulations impose a constant strain rate, leaving the shear modulus as the only variable. The experimental and numerical stress-strain curves suggest an increase in the effective shear modulus with increasing garnet content. The shear modulus of omphacite-dominated eclogite (25% garnet fraction) is almost identical to that of pure omphacite, inhibiting localization at the given conditions. In the absence of MSZ (25% garnet sample and omphacitite), weakening can be assigned to dislocation glide in the interconnected omphacite matrix and the development of a CPO, as previously observed for rock forming minerals (Barnhoorn et al., 2004;Handy, 1994).
In our numerical simulations, strain weakening only occurs by micro-faulting and an associated local loss in cohesion. Only simulations for 75% garnet reveal minor fracture-associated weakening, whereas all experimentally deformed eclogites show strain weakening. The difference in the results of the simulations and the experiments after the peak stress (compare Figure 2 and Figure 10) suggests that brittle deformation alone does not suffice to explain observed strain weakening. Instead, it appears that dislocation glide (25% garnet sample and omphacitite) and grain size reduction via omphacite recrystallization (50% garnet sample) are key processes for strain weakening in eclogite. Nevertheless, numerical and experimental investigations highlight that a garnet content of >25% is necessary to initiate strain localization in eclogite at high pressure and temperature conditions. Our investigations suggest that the garnet content in eclogite increases the activity of crystal plastic deformation of omphacite, which subsequently weakens the eclogite.

Rheological Implications for Convergent Settings
Besides temperature, the relative amount, distribution and strength of the major minerals control the rheological behavior of rocks. At convergent plate boundaries, where rocks are subducted to high pressure conditions, mafic rocks transform to eclogite with a composition that is commonly simplified as a mixture of garnet and omphacite. The early experimental study of Green and Ringwood (1967) demonstrated that the final eclogitic phase/mineral assemblage does not only depend on protolith chemistry but also on peak metamorphic conditions; in particular, garnet content tends to increase with pressure. Furthermore, natural eclogites frequently show variations in garnet content on scales from millimeters to several decimeters, for example, reported from LT-HP terrains like Syros, the Mt. Emilius klippe, and the Lago Superiore Unit (Angiboust et al., 2011;Hertgen et al., 2017;Kotowski & Behr, 2019), HP terrains such as the Western Gneiss and Koralpe Regions (Engvik et al., 2007;Rogowitz & Huet, 2021;Terry & Heidelbach, 2006) and from UHP eclogites from the Sulu Region (Ji et al., 2003). The spatial variability may reflect protolith heterogeneity, variability in mineral reactions, or deformation-related processes, such as dynamic recrystallization (Ji et al., 2003;Mauler et al., 2000;Philippot & Van Roermund, 1992;Terry & Heidelbach, 2006). Previous deformation experiments on synthetic eclogite samples showed that their strength increases with increasing garnet content (Jin et al., 2001). Our combined numerical and experimental investigations showed that the garnet content of eclogites not only influences their strength (Jin et al., 2001;Zhang & Green, 2007) but also the dominant deformation mechanism and associated weakening processes.
Despite experimental and natural evidence of the heterogeneous nature of eclogitic rocks, geodynamic models of convergent zones generally assume a uniform power-law rheology for the deformation of the subducting crust (e.g., Behr et al., 2022). However, the thermo-mechanical modeling by Ioannidi et al. (2021), inspired by exhumed high-pressure mélanges, demonstrated that the effective rheology and deformation type in subducted mélange zones strongly depend on the distribution of strong blocks from outcrop to regional scale (i.e., m-to km-scale). With increasing block density, the brittle-ductile transition zone is shifted to a greater depth (Ioannidi et al., 2021). Based on our results, we suggest that the rheological behavior of the subducted crust might be additionally controlled by the compositional and microstructural variability of eclogites. Our results reveal the complex interplay of crystal plastic and frictional deformation mechanisms in eclogite and potentially help to understand the complex spatial and temporal distribution of slow slip events and episodic tremor and slip in subduction zones (Behr et al., 2018;Fagereng et al., 2014;Peng & Gomberg, 2010;Rogers & Dragert, 2003).

Conclusions
Our integrated experimental and numerical study reveals how the garnet content affects the rheology and deformation behavior of eclogite at conditions representative of intracontinental subduction zones. We determined the timing, location and mechanism of strain localization and identified potential weakening mechanisms in dry eclogite.
The findings are summarized below: • In accord with phase-mixing models, eclogite strength increases with increasing garnet content following a log-linear behavior. • The deformation regime of eclogites switches from overall crystal plastic to brittle-dominated with increasing garnet content. • At 1000°C and 2.5 GPa, the garnet content in eclogite is the controlling factor for the occurrence as well as spatial and temporal evolution of strain localization. • In eclogite with a garnet fraction of 50%, strain localization initiates in omphacite, whereas in eclogite with a 75% garnet fraction brittle deformation in the interconnected garnet network is the loci of strain localization. • Comparison of simulations and experiments after maximal compressive strength is reached emphasizes the importance of omphacite creep for strain weakening of eclogite. However, the presence of a garnet fraction >25% is needed to initiate deformation localization at here applied conditions and enhance crystal plasticity in omphacite.
Our findings on small-scale processes operating in eclogite with varying garnet fractions have the following implications for the deformation behavior in convergent settings at high pressure and temperature conditions: • In garnet-rich domains initiated fractures can serve as brittle precursors for eclogite facies shear zones. • Compositional changes in eclogite bodies can result in mixed viscous-frictional deformation behavior in the absence of seismic strain rates (i.e., earthquakes) or high fluid pressure. • Rheological heterogeneities based on the mineralogical composition in eclogites might contribute to the understanding of slow slip events and associated low-frequency earthquakes.

Data Availability Statement
Data for this study can be found in the Open Access Collection-Phaidra of the University of Vienna (https:// phaidra.univie.ac.at/detail/o:1603569). The code LaMEM is available at https://bitbucket.org/bkaus/lamem/src/ master/ (commit:b537d7f724ef9ceeb8a28e3e5e12e9698ea40957) . This study was financed by the Austrian Science Fund (FWF) Grant P 29539-N29 (to AR), P 36034-N (to AR), and DFG Grant TH 2076/8-1 (to MT). We thank Marieke Rempe for support during synthesis and deformation experiments, and Sarah Incel for help during mechanical data processing. Discussions with Martin Schöpfer and Benjamin Huet were highly appreciated. We are grateful for constructive reviews by Luiz Grafulha Morales, an anonymous reviewer and an anonymous associate editor that highly improved the quality of the manuscript and appreciate editorial handling by Paul D. Asimow.