Slow‐to‐Fast Deformation in Mafic Fault Rocks on an Active Low‐Angle Normal Fault, Woodlark Rift, SE Papua New Guinea

Slip on the active Mai'iu low‐angle normal fault in Papua New Guinea that dips 15–24° at the surface has exhumed in its footwall a single, continuous fault surface across a >25‐km‐wide dome. Derived from a metabasaltic protolith, the fault zone consists of a <3‐m‐thick zone of gouges and cataclasites that overprint a structurally underlying carapace of extensional mylonites. Detailed microstructural and geochemical data, combined with chlorite‐based geothermometry, reveal changing deformation processes and conditions in the Mai'iu fault rocks as they were exhumed. The microstructure of nonplastically deformed actinolite grains inherited from the fine‐grained (6–35 µm) metabasaltic protolith indicates that shearing at depth was controlled by diffusion creep accompanied by grain‐boundary sliding of these grains together with chlorite neo‐crystallization at T > 275°C–370°C. In a foliated cataclasite unit at shallower crustal levels (T ≈ 150°C–275°C), metasomatic reactions accompanied fluid‐assisted mass transfer processes that accommodated aseismic, distributed shearing; pseudotachylites and ultracataclasites in the same unit indicate that such creep was punctuated by episodes of seismic slip—after which creep resumed. At the shallowest levels (T < 150°C), gouges contain abundant saponite, a frictionally weak mineral that promotes creep on the shallowest dipping (≤24°), most poorly oriented part of the Mai'iu fault. Our field, microstructural and geochemical data of freshly exhumed fault rocks support geodetic, seismological, and geomorphic evidence for mixed seismic‐to‐aseismic slip on this active low‐angle normal fault.

Deformational processes operating at different crustal levels along detachment faults are not well understood (e.g., Collettini, 2011;Smith et al., 2011). Most active continental LANFs are concealed beneath a cover of hangingwall rocks (Chiaraluce et al., 2014;Rigo et al., 1996), while ancient ones have commonly been overprinted and obscured by post-activity tectonic tilting, chemical alteration, and erosion (e.g., Axen, 2004;Axen & Bartley, 1997;Collettini, 2011;Whitney et al., 2013). The Mai'iu fault in SE Papua New Guinea is a rare exception-it is one of the best preserved of only a handful of known continental LANFs on Earth that are demonstrably active today (e.g., Spencer, 2010). This fault formed by extensional reactivation of a former subduction zone (e.g., Little et al., 2019;Webb et al., 2008). Dip slip on the Mai'iu fault at ∼10 mm/yr for the past 3-4 Myrs Wallace et al., 2014;Webber et al., 2018) has exhumed a >29 km width of little-eroded fault surface in its footwall. Along-and immediately beneath-this now abandoned fault surface, freshly exhumed microstructures in the metabasaltic footwall are well-preserved and can be sampled Mizera et al., 2019). Different types of fault rocks locally crosscut one another, and together they record spatiotemporal changes in the deformation mechanisms accommodating slip as the fault transited through the middle to uppermost crust .
The apparent conflict of active LANF slip with Andersonian fault mechanics can be resolved by understanding the fault rocks and fault orientation at different crustal levels. The active Mai'iu fault is an ideal natural laboratory in which to study microstructural and geochemical changes across a wide range of crustal levels. Here, we present data from fault rock samples collected from the exhumed Mai'iu fault. Mostly mafic in bulk composition, the samples comprise pre-extensional nonmylonitic schists (protolith), syn-extensional mylonites, foliated cataclasites, ultracataclasites, and gouges. We combine field observations with mineral phase and elemental composition data of the fault rocks, microstructural imaging and analyses (including grain size and crystallographic-preferred orientation), and chlorite-based estimates of paleo temperatures. By documenting changes in the microstructures and mineralogy of these fault rocks, we derive a syn-exhumational progression of deformation mechanisms and strain rates that accompanied slip on the Mai'iu fault at different structural levels. We also identify processes (such as metasomatic reactions and grain size reduction) that led to both weakening and transient strengthening of the fault rocks. Our aim is to evaluate as follows: (a) deformation mechanisms accommodating slip on a low-angle normal fault in space and time, and (b) metasomatic changes in the exhuming mafic rocks, in particular those that may have enabled slip on the most highly misoriented part of the fault near the Earth's surface.
Myr relative to the Australian plate about a nearby Euler pole (Wallace et al., 2014). In the active eastern Woodlark Basin, this rotation has resulted in north-south seafloor spreading (Eilon et al., 2015;Wallace et al., 2004). Farther west, the spreading center transitions into a zone of active continental rifting-the Woodlark Rift-which is thought to have initiated at 3.6-8.4 Ma (Figure 1a; Taylor & Huchon, 2002;Wallace et al., 2014;Cairns et al., 2015).
At the southwest margin of the Woodlark Rift (Figures 1a and 1b), almost all the regional extension is accommodated by slip on the Mai'iu fault. Cosmogenic nuclide studies on the exhumed scarp of this fault indicate a Holocene to present-day dip-slip rate of 11.7 ± 3.5 mm/yr (Webber et al., 2018)-a result that accords with geodetic data indicating a present-day slip-rate of 7.5-9.6 mm/yr (Wallace et al., 2014). The Mai'iu fault is thought to be an extensionally reactivated part of the Owen-Stanley thrust, a major Paleogene thrust fault (Figure 1a; Daczko et al., 2011;Davies, 1978;Little et al., 2019;Webb et al., 2008). The Owen-Stanley thrust accommodated the southwest-directed obduction of an oceanic and island arc in its hangingwall (the Late-Cretaceous Papuan Ultramafic Belt; PUB) over an accreted assemblage of oceanic marginal basin and Australian Plate-derived continental margin rocks (Daczko et al., 2009;Davies, 1978;Webb et al., 2008).

Geology of the Suckling-Dayman Metamorphic Core Complex
The Mai'iu fault bounds the Suckling-Dayman Metamorphic Core Complex (SDMCC). The exhumed lower plate of this MCC includes three antiformal culminations that coincide with peaks on the main divide of the Owen Stanley Ranges (Figure 1b): Mt Suckling (3,576 m), Mt Dayman (2,950 m), and Mt Masasoru (∼1,700 m). The footwall of the SDMCC consists of a >3-4 km thick section of mid-ocean ridge basaltderived metabasaltic rocks, known as the Goropu Metabasalt, together with minor interbeds of phyllitic metasediments, limestone, and chert Smith & Davies, 1976). The unusual thickness of the Goropu Metabasalt could be a result of previous tectonic compression and structural thickening (e.g., thrust duplexing). The Late Cretaceous metabasalt grades southward into unmetamorphosed equivalents called the Kutu Volcanics, parts of which are Eocene in age (Smith & Davies, 1976). Approximately mid-way up the northern flank of the SDMCC , a mapped pumpellyite-out isograd probably represents peak temperatures of 350°C-375°C (Figure 1b; Beiersdorfer & Day, 1995;Daczko et al., 2009). Farther south, pumpellyite-actinolite facies rocks are exposed as far as the Onuam fault (Figure 1b), beyond which they transition into prehnite-pumpellyite facies rocks (Davies, 1978) and ultimately the Kutu Volcanics.
Granitoid stocks (e.g., Mai'iu Monzonite, Suckling Granite; Davies & Smith, 1974) intrude the lower plate of the SDMCC near Mt Suckling and have syn-extensional U-Pb-based zircon crystallization ages of 1.9-3.7 Ma (Österle et al., 2020). The granitoids record melting of continental crust at depth (Österle et al., 2020). Geophysical surveys (based on seismic traveltimes) confirm that the Goropu Metabasalt is underlain by continental crust of felsic to intermediate composition extending to depths of ∼32 km near the SDMCC ( Figure  1c; e.g., Abers et al., 2016;Eilon et al., 2015;Ferris et al., 2006;Finlayson et al., 1977). A relict thrust flap of PUB overlies the Goropu Metabasalt on a southern part of the footwall of the Mai'iu fault. The PUB is the basement on which the unmetamorphosed alluvial sedimentary rocks of the Gwoira Conglomerate, of Plio-Pleistocene age, were deposited ( Figure 1b; Webber et al., 2020). This latter unit forms much of the hangingwall of the Mai'iu fault, although south of the active Gwoira splay fault, a synformal slice of the Gwoira Conglomerate (former hangingwall rocks) has been captured into the footwall of the SDMCC (Webber et al., 2020). This slice is floored by a now abandoned, inactive segment of the Mai'iu fault. Outcrops of this inactive segment of the Mai'iu fault provide a complete section through its exhumed fault rocks (Figure 1b; Little et al., 2019).
Outcrops and dip-slope geomorphic data indicate that the Mai'iu fault dips NNE from 15 to 24° (mostly 20-22°) along its trace at the rangefront (Figures 1d and 1f; Little et al., 2019;Mizera et al., 2019). Downdip of this, a linear alignment of microseismicity indicates that the fault steepens to a 30-40° dip in the subsurface below ∼12 km depth (Figure 1c; Abers et al., 2016). Wear striae on the exposed fault surface, including on fault-truncated cobble facets in the hangingwall, trend ∼012-015° . The striae are subparallel to the velocity of the Woodlark-Solomon Sea microplate relative to the Australian Plate farther south (Wallace et al., 2014). The striae trend is also subparallel to numerous megacorrugations in the exhumed footwall of the SDMCC (

Fault Rock Sequence
The immediate footwall of the Mai'iu fault contains a sequence of mostly mafic-composition fault rocks. Figure 2a shows a schematic section of this sequence, which is partly eroded along the active trace of the fault, but fully preserved in outcrops along the inactive segment of the Mai'iu fault (Figure 1g). The fault sequence includes five units: (a) nonmylonitic schist (protolith), (b) mylonite, (c) foliated cataclasite, (d) ultracataclasite, and (e) gouge unit (gouge) containing the principal displacement surface. Each unit was briefly described in Little et al. (2019), as summarized below: Most of the SDMCC footwall consists of schistose (nonmylonitic) mafic rocks (Goropu Metabasalt; Figure  2b) that decrease in metamorphic grade southward (Davies, 1978;Little et al., 2019). On the northern flank of the SDMCC, greenschist-facies schists contain a fine-grained assemblage of epidote, actinolite, chlorite, albite, titanite, relict clinopyroxene, ±quartz, ±calcite, ±stilpnomelane, ±pyrite, ±mica, ±apatite, and ±opaque minerals (Daczko et al, 2009;Little et al., 2019;Smith & Davies, 1976). The mafic schists have a NNW-trending stretching lineation and a top-to-the-SSE (thrust) sense of shear indicated by: (1) asymmetry of strain fringes filled with fibrous quartz ±actinolite or of strain shadows around epidote porphyroclasts filled with blocky chlorite (Figure 2b, inset), (2) extensional (C') shear bands, and (3) the shape-preferred orientation of actinolite porphyroclasts that have fine-grained, sigmoidal tails of chlorite. These shear fabrics developed during the Paleogene Papuan Orogeny when the Goropu Metabasalt was underthrusted northward beneath the PUB along the Owen-Stanley thrust .
Extensional mylonitic fabrics are exposed in the footwall within a ∼2 km wide band up-dip and to the south of the active fault trace (Figure 1b). The mylonite unit (mylonites) is at least 60 m-thick along the rangefront; whereas ∼2 km up-dip to the south, it narrows to as little as 1.5 m. The mylonites are LS-tectonites with a well-defined NNE-trending stretching lineation and normal-sense shear fabrics (Figure 2c). The northward transition from the schistose Goropu Metabasalt into Neogene-age extensional mylonites can be observed in outcrops. This transition is marked by a NNW to NNE rotation of the mafic minerals and stretching lineation in the footwall and a change in the shear sense from top-to-the-SSE (thrust sense) to top-to-the-NNE (normal sense) shear fabrics (Figure 1e; Little et al., 2019;Mizera et al., 2019). These microstructural relationships distinguish the pre-mylonitic fabric from the mylonitic fabric. Based on pseudosection modeling of the mineral assemblage (epidote, actinolite, chlorite, albite, titanite, ±quartz, and ±calcite), the mylonites show peak metamorphic conditions of T = 425°C ± 50°C and P = 5.9-7.2 kbar, reflecting exhumation from ∼25 ± 5 km depth (Daczko et al., 2009).
Above the mylonites, a zone of foliated cataclasite, ultracataclasite, and gouge reaching several meters in thickness overprint and rework the mylonites. The foliated cataclasite unit (foliated cataclasites) is 1.5-3 m-thick and hosts multiple generations of pseudotachylite veins (0.5-40 mm thick; Figures 2d and 2e). The pseudotachylite matrices are glassy and amorphous in some of the veins, as verified by quantitative X-ray powder diffraction (see Little et al., 2019). More commonly, however, they are devitrified. Five pseudotachylite veins with glassy matrices were dated by 40 Ar/ 39 Ar geochronology; the derived ages (interpreted as minimum ages for frictional melting) range from 2.24 ± 0.29 Ma to 3.00 ± 0.43 Ma (±2σ; Little et al., 2019).
The foliated cataclasites are structurally overlain by a 5-40 cm-thick dark gray to black (locally brick-red) ultracataclasite unit (ultracataclasite) (Figures 1f and 2f). Figure 1f shows the smooth top surface of the ultracataclasite. The fault surface is marked by fine wear striae trending NNE. The ultracataclasite is sharply overlain by a gouge containing one or more layers of incohesive clay-rich gouge that is preserved in inactive fault segment outcrops beneath the Gwoira Conglomerate    (Figure 1g).

Thin Section Preparation and Analytical Methods
The rock samples were collected in Little et al., 2019). Thin sections were cut parallel to the kinematic XZ-direction-perpendicular to foliation and parallel to mineral lineation or wear striae on the slip surfaces. Fragile fault rocks (e.g., ultracataclasites and gouges) were impregnated with epoxy resin. The mineralogy, microstructures, and grain sizes of 157 thin sections were initially described using optical microscopy. From these, we prepared polished thin sections of 47 representative samples for analysis using a field emission gun-scanning electron microscope (FEG-SEM) equipped with energy dispersive spectroscopy (EDS), electron backscatter diffraction (EBSD) detectors, and a cold-cathode cathodoluminescence (CL) microscope at the University of Otago. EDS analyses were used to quantify fault rock elemental compositions (see Data S1 and S2 for detailed EDS maps and descriptions).
Crystallographic orientation data for the phases actinolite, epidote, titanite, albite, and quartz were collected by EBSD. The fabric strength of individual phases in the analyzed fault rock samples is given by the misorientation index (M-index)-the stronger the fabric, the greater the M-index, which ranges from 0 (random fabric) to 1 (single crystal; Skemer et al., 2005;Mainprice et al., 2015). Estimated mean grain sizes (diameter of a circle with the equivalent area of a given grain, Ø, with errors given at one standard deviation, ±1σ) are based on EBSD-derived phase maps collected in the kinematic XZ plane (two dimensional). The data were collected with 0.4-2 µm step sizes, and processed using the MTEX toolbox for MATLAB (see Data S3 and S4 for method description, tables with estimated grain sizes and detailed EBSD phase maps). All fault rock images presented are arranged with the normal slip sense of the Mai'iu fault shown top-to-the-right.
In addition, one fault-truncated cobble from the Gwoira Conglomerate with a mirror-like slip surface was subjected to analysis by transmission electron microscope (TEM) at the University of New Brunswick and by SEM with a HITACHI SU-3500 at the Geological Survey of Japan. A further 12 mylonite and foliated cataclasite samples were subjected to electron probe microanalysis (EPMA) at the Victoria University of Wellington. A collection of 15 fault rock samples comprising ultracataclasites and gouges were analyzed with X-ray powder diffraction (XRD) at the Centre for Australian Forensic Soil Science (CAFSS), in Urrbrae, South Australia (see also Biemiller et al., 2020). Finally, whole rock major element concentrations of four mafic fault rocks including mylonite and gouge were analyzed using X-ray fluorescence (XRF) at the University of Waikato. All these methods are detailed in the Data S5-S9.

Chlorite-Based Geothermometry
Chlorite is a common mineral in the Mai'iu fault rocks, where it neocrystallized syntectonically in a range of dilational microstructural sites. Although various empirical geothermometers exploit the relationships between the composition and formation temperature of chlorite (see Bourdelle & Cathelineau, 2015), we employ the geothermometer by Cathelineau (1988) because it is best suited to low-grade metamorphic rocks of basaltic composition (e.g., Bevins et al., 1991). This geothermometer is based on the occupancy of Al (IV) in tetrahedral sites of the chlorite structure and was calibrated for mafic rocks from 150 to 300°C using combined microthermometric data, especially fluid inclusion studies on quartz coexisting with clays (Cathelineau, 1988;Cathelineau & Nieva, 1985).
We obtained chlorite compositions with EPMA from six mylonite and six foliated cataclasite samples (296 measurements; Mizera, 2019). The Windows© program WinCcac by Yavuz et al. (2015) was used to calculate the structural formulae of chlorite based on 14 oxygens and to estimate the formation temperatures by applying the geothermometer of Cathelineau (1988). Because chlorite shows considerable variation in structure at low temperatures, including the possibility of swelling varieties of trioctahedral structure (mixed-layer clays such as smectite-chlorite), we applied the method by W. Wise as presented in Bettison and Schiffman (1988) to calculate the relative fraction of chlorite to "swelling" component (X), where X = 1 MIZERA ET AL. 7 of 29 10.1029/2020GC009171 represents pure chlorite and X = 0 represents pure saponite. Bevins et al. (1991) showed that the Cathelineau (1988) thermometer is applicable for X > 0.55.
Deformed veins of monophase quartz are usually arranged sub-parallel to the mylonitic foliation ( Figure  3d). Elongated quartz grains within these veins define a grain shape fabric that is oblique to the mylonitic foliation (∼16°) and consistently indicate a top-to-the-south (thrust) sense of shear. In the veins, porphyroclastic quartz grains are surrounded by fine, ∼5-15 µm-sized quartz grains similar to a "core and mantle" MIZERA ET AL. Mafic mylonites are locally cross-cut by ∼0.5-1 mm thick seams of ultrafine-grained (optically irresolvable) rock containing fragments of epidote, albite, quartz, titanite, abundant pyrite, and mylonite wall-rock clasts. The seams are mostly discordant to the mylonitic foliation and have strongly indented, intrusive contacts indicating injection (Figure 3e). We interpret these to be pseudotachylite veins similar to the aforementioned ones hosted by the nearby foliated cataclasite unit.

Foliated Cataclasites
The contact between the mylonites and foliated cataclasites locally coincides with one or more, mm-to cm-thick, anastomosing bands of ultramylonite. The contact zone is also marked by upward increases in the intensity of microfaulting and brecciation and the degree of apparent bleaching of the metabasaltic protolith. The cataclasites contain a mm-to cm-spaced foliation defined by alternating dark-and light-colored folia (Figures 2d and 4). Darker, phyllosilicate-rich folia anastomose around light-colored domains and lenticular clasts that consist of albite-rich material. This spaced foliation is distinct from the much finer and more continuous foliation in the mylonites (Figures 2c and 2d). The spaced foliation is folded at the mm-to cm-scale and is locally cross-cut by (less folded) pseudotachylite veins and abundant, μm-to <1 mm-thick calcite gash veins ( Figure 4a). The calcite veins are typically oriented sub-perpendicular to the foliation. Backward-inclined gash veins (with respect to the sense of shear) are shortened and folded, while forward-inclined ones are stretched and show necking structures ( Figure 2d)-a change in strain type compatible with a normal-sense shearing of the wall rocks. Locally, bands of ultracataclasite cross-cut the folded mylonitic fabric, the calcite veins embedded within it (Figure 4b), and the pseudotachylite veins.
Most of the dark folia consist of phyllosilicates (<2 µm in size), epidote (Ø ≈ 17 ± 9 µm), actinolite (Ø ≈ 12 ± 7 µm), and titanite (Ø ≈ 9 ± 4 µm; Figures 4c and 4e). EDS and XRD analyses indicate that the phyllosilicates are dominated by chlorite and lesser 1Md and 2M1 illites (Data S1 and S2; also Little et al., 2019). Epidote and actinolite grains typically have indented or truncated phase boundaries in contact with chlorite. Bulging of chlorite inwardly into epidote and actinolite grains suggests growth of chlorite together with dissolution and removal of the neighboring phase (Figures 4c and 4e). Albite, calcite and quartz are rare in the dark-colored folia. Where preserved in these folia, the albite grains are either (1) anhedral with interlobate or sutured phase boundaries in contact with the other mafic minerals, or (2) preserved as inclusions inside larger epidote grains together with calcite, quartz and/or chlorite inclusions (Figure 4e). Based on these observations, we propose the following common mineral reaction responsible for the transformation of epidote and actinolite to chlorite in (sub)greenschist-facies metabasalt (Hashimoto, 1972;Skelton et al., 2000, and references therein): If this fluid-driven metasomatic reaction was operative, we would expect calcite and quartz to be formed at the expense of actinolite and epidote.
Indeed, in the light-colored, mm-thick albitic domains, abundant neorecrystallized calcite and quartz occurs in grain-scale dilation sites (Figures 4c and 4f). Optical cathodoluminescence microphotographs reveal   (100)  pervasive micro-veining and cementation of the light domains by calcite (Figure 4d). Albite grains (Ø ≈ 12 ± 9 µm) are usually anhedral, untwinned and cross-cut by either calcite veins or the dark-colored folia. EPMA analyses of albite porphyroclasts in the mylonite and of albite in the light-colored foliated cataclasite layers indicate an average composition of An 05 in both settings, but the albite in the foliated cataclasite shows a wider compositional spread (Figure 5a).
Epidote, actinolite and titanite in both the dark and light folia of the foliated cataclasites, show a decrease in grain size, shape preferred orientation and CPO strength relative to the mafic mylonites (Figure 5b).
Cathodoluminescence microscopy reveals intensive calcite veining in some of the ultracataclasites (Figures  6c and 6d). Calcite veinlets of <100 μm thickness are mostly unfractured. Instead, they are ductilely deformed in a directionally dependent manner that resembles calcite veins in the foliated cataclasites. Veinlets that are backwards-inclined with respect to the shear sense are folded, whereas those that are forward-inclined are unfolded and planar (Figure 6a, inset). This contrasting veinlet geometry suggests that veinlet stress was imposed by normal sense shearing in the weaker, surrounding matrix.

(b) (a)
to the foliated cataclasites and mylonites (Table S1.1). K-feldspar grains are >0.5 µm and up to ∼12 µm in diameter, subhedral, unfractured, and in part replace albite (Figure 6b). The observed increase of K and reduction in Na relative to the underlying foliated cataclasite, and growth of K-feldspar might be related to consumption of albite by the reaction: We infer that the K-feldspar grains grew authigenically in the ultrafine-grained corrensite-and saponiterich matrix of the ultracataclasite. The matrix encloses some ultracataclasite clasts that appear to have been derived from other parts of the same unit. The clasts are up to 100 µm in size, rounded and rarely fractured (Figure 6e), and occasionally coated with an outer cortex (or swirled into the clast) of smectite or magnetite (Figures 6f and 6g). These inherited fragments of ultracataclasite are generally more K-rich than the surrounding ultracataclasite host rock in which they are embedded ( Figure S2.5).
Some ultracataclasite samples contain distinctive magnetite-bearing clasts and veins (Figures 6g and 6h). SEM-EDS analysis show that the multiple generations of magnetite-bearing veins (∼20-60 µm-thick) crosscut each other. Some veins contain magnetite ±maghemite grains <1 µm in size (Figure 6h). Temperature-dependent magnetic susceptibility experiments on one of these ultracataclasite samples exposed just south of the Mai'iu fault trace (Data S5 for method and results) indicate a magnetic susceptibility of ∼13 × 10 −6 m 3 kg −1 and Curie temperatures between T = 580°C and 645°C for the magnetic fraction of this rock ( Figure 7). These results suggest the presence of cation-deficient magnetite (i.e., magnetite slightly oxidized to maghemite), an inference that also accords with our EDS data (Dunlop & Özdemir, 1997;Özdemir & Banerjee, 1984). Based on the similarity between the heating and cooling curves of the magnetic susceptibility experiment, the Curie temperatures and our EDS analyses, we infer that the principal magnetic mineral in the ultracataclasite is a stable (reversible), single domain (fine grained) magnetite.

Gouges
In contrast to the underlying fault rock units, all four analyzed gouge samples are devoid of calcite veins, although they do contain clasts of calcite or dolomite that were probably reworked from older veins. A foliation is not optically visible in the gouges. The gouge samples contain angular to sub-rounded monoand polyphase clasts of epidote (Ø ≈ 4 ± 3 µm), actinolite (Ø ≈ 4 ± 3 µm), titanite (Ø ≈ 3 ± 2 µm), and albite (Ø ≈ 3 ± 2 µm), as well as clinopyroxene, quartz, and calcite (<2 µm) embedded in a phyllosilicate matrix (Figures 8a and 8c). Based on XRD and EDS data, the 2 µm fraction of this matrix consists mainly of saponite (Table S7; see also Biemiller et al., 2020). Polyphase clasts are up to 1 cm in diameter and include fragments of mylonite, foliated cataclasite, and fine-grained K-rich fault rocks that resemble the ultracataclasite in texture and composition (Figure 2g, 8a, 8b, and 8d; Tables S1.2 and S1.3). Some of the K-rich clasts show an enhanced porosity not seen in the ultracataclasites (Figure 8d).
The gouges contain trace amounts of dolomite and chromite (based on EDS analysis)-two minerals strongly associated with hangingwall ultramafic rocks. Furthermore, XRF analyses show that the gouges contain elevated concentrations of Cr and Ni compared to the mafic schists or mylonites in the footwall of the SDMCC (Table  S8.1). Some gouges contain optically isotropic, tens of µm-thick silica-rich domains or patches (Figure 8e). Figure 8f shows the diffuse nature of the transition from saponite to a silica-rich domain. These relationships and the proximity of the silica phase to cracks and fractures in the samples suggest that the silica is an alteration product of saponite and/or it has precipitated together with saponite from hydrothermal fluids.

Hangingwall Conglomerate and Fault Faceted Cobbles
On the inactive segment of the Mai'iu fault, the principal displacement surface of the Mai'iu fault forms the uppermost contact of the fault rock sequence. This planar, mm-thick contact places unmetamorphosed Gwoira Conglomerate in the hangingwall against the gouges in the footwall. The principal displacement surface truncates cm-to dm-sized clasts in the Gwoira Conglomerate, forming    Figures 9a-9c show the striated surface of the analyzed dolerite including a groove lineation etched into a clinopyroxene-grain by a quartz fragment. The TEM analysis reveals that the mirror surface on the facets is a 2 µm-thick layer of amorphous material mainly consisting of Al and Si (Figures 9f, 9i, and 9j). The TEM images show multiple veins of this amorphous material emanating from the fault-parallel layer and injecting discordantly across host pyroxene grains in the truncated clast (Figures 9g and 9h).

Chlorite Geothermometry of the Fault Rocks
More than 100 chlorite grains (296 microprobe measurements) in 12 samples of mylonite and foliated cataclasite from eight localities in the Mai'iu fault zone were analyzed. Our analyses targeted chlorite grains infilling syntectonically created microstructural sites, including: (a) in the mylonites, asymmetrical (normal sense) strain shadows and shear bands as well as pulled-apart necks between boudins of epidote and albite; (b) in the foliated cataclasites, dark chlorite-rich folia or C' normal-sense shear bands or ultracataclasite bands; and (c) in both the mylonites and foliated cataclasites, little deformed chlorite veins that cross-cut the foliation and/or pseudotachylite veins. The results of estimated chlorite formation temperatures versus the fraction of chlorite (X = 1 is pure chlorite) to "swelling" component (X = 0 is saponite) are shown in Figure 10. is observed in all the mylonite samples (X > 0.85), whereas mixed-layer smectite-chlorite to saponite (X < 0.50) occurred in the dark folia of some of the foliated cataclasites. Overall there is an increase in "swelling" component (smaller X value) as the calculated temperature decreases. Following the recommendation of Bevins et al. (1991), we excluded 12 chlorite compositional measurements for which X < 0.55 in our calculation of the average chlorite formation temperature for that unit. The dark-shaded area in Figure 10 (226°C-329°C) indicates the overlap between the estimated temperatures in the mylonite samples and the later, cross-cutting chlorite veins.

Mai'iu Fault Structure and Fault Rock Assemblage
The Mai'iu fault exposes mafic mylonitic rocks that have been exhumed from ∼25 ± 5 km depth at peak metamorphic conditions of T = 425°C ± 50°C and P = 5.9-7.  EPMA mineral chemistry data, Daczko et al., 2009). The late Neogene and younger fabrics in the mylonitic unit were later overprinted in the narrower and structurally overlying <3 m thick fault rock sequence composed of (from bottom to top): foliated cataclasite, ultracataclasite and gouge (Figure 2). The upper mylonite and adjacent foliated cataclasite units are cross-cut by several generations of chlorite, calcite and pseudotachylite veins, many of them ductilely deformed (Figures 2-4). The brittle fault rocks (ultracataclasites and gouges) have a NNE-trending striation that is parallel to the ductile stretching lineation in the mylonitic rocks, and all fault rock units share the same normal shear sense. The fault rock units become thinner structurally upward, and each unit overprints the underlying one. Together, these observations record a progressive localization of slip that advanced structurally upward in time, and culminated with slip along the sharp and planar principal displacement surface at the base of the unmetamorphosed hangingwall.
The estimated chlorite formation temperatures (Figure 10) decrease from the mylonite (226°C-366°C and 310°C ± 31°C) to the foliated cataclasite unit (158°C-356°C and 273°C ± 46°C). We relate the wide temperature range of the foliated cataclasite, especially some very high-estimated temperatures in that unit, to reflect the largely inherited nature of the mylonitic rock that was reworked into that younger, partially brittle fault rock (Figure 4b). We interpret the overlap in estimated temperatures between the mylonites and late-stage chlorite veins that cut both it and the foliated cataclasite unit to indicate a temperature range of 226°C-329°C at the onset of brittle fracturing ( Figure 10). The mineral transformation of chlorite to mixed-layer smectite-chlorite ( Figure 10) and the neoformed corrensite in the ultracataclasite unit indicate sub-greenschist-facies temperatures of ∼150°C-225°C in this fault unit (e.g., Moore et al., 2016;Robinson et al., 2002;Surace et al., 2011). Furthermore, we suggest that saponite in the gouges (XRD data, Table S7.1) was stable at temperatures <150°C (e.g., Boulton et al., 2018;Lockner et al., 2011;Richard et al., 2014). Thus, the fault rock sequence and its associated temperatures track a temporal evolution of cooling and embrittlement of the footwall together with progressive localization of brittle slip during exhumation to the Earth's surface. The estimated temperatures for mylonite formation and the onset of brittle fracturing of these rocks are below what would be expected for basaltic rocks (Violay et al., 2012). Thus, the temperature estimates indicate that the metasomatic mineral assemblage (actinolite, chlorite, epidote, and albite) inherited from the schistose Goropu Metabasalt weakened the fault prior to reactivation.
The Mai'iu fault dips 15-24° along its trace but steepens northward to 30-40° at 12-25 km depth as indicated by a corridor of microseismicity downdip of it (Abers et al., 2016, Figure 1c). From these observations, we infer that the Mai'iu fault has a convex-upward geometry not only on the exhumed and abandoned southern part of the fault, but also on the active part of the fault in the subsurface to the north, a curvature that we attribute to the operation of rolling hinge-style deformation Spencer, 1984Spencer, , 2010. The convexity, structure, fault rock assemblage, and some deformation mechanisms for the Mai'iu fault resemble those of other continental detachment faults (e.g., Cooper et al., 2017;Platt et al., 2015), but an unusual aspect of the Mai'iu fault is that its footwall rock type is primarily metabasaltic. The next section outlines several deformation mechanisms, both seismic and aseismic, that we infer were active in these mafic rocks as they were exhumed by slip on the Mai'iu fault (Figure 11).

Late Neogene and Younger Exhumational Shearing and Strain Rates in the Mylonites, Precursor Shear Zone and Fault Reactivation
We do not observe any evidence for intracrystalline deformation of the primary mineral assemblage (epidote, actinolite, titanite, and albite) of the metabasaltic mylonites. Instead, we observe the following microstructures in these rocks (Figure 3): (a) an average grain size of 6-35 µm, (b) straight phase boundaries between epidote, actinolite and titanite grains and other phases along alignments that span a distance of several grain widths (Figure 3b), (c) phase-boundary parallel grain offsets (predominantly epidote, actinolite, and titanite) along these straight phase boundaries, (d) chemical zonation in epidote where the chemistry changes along the length of these grains, particularly toward crystal tips and fractures (Fe 3+ -rich rims; Figure 3b), (e) a strong CPO of nonplastically deformed actinolite grains (Figure 3c), and (f) a weak to random CPO of albite.
From these observations, we infer that shearing in the mylonites was controlled by diffusion creep accompanied by grain-boundary sliding (GBS) (cf. Aspiroz et al., 2007;Elyaszadeh et al., 2018;Getsinger et al., 2013). The strong CPO of actinolite in the mylonites resembles that described for clinoamphiboles in other shear zones of greenschist-to amphibolite-facies metabasites (Aspiroz et al., 2007;Getsinger et al., 2013;Getsinger & Hirth, 2014). These authors attribute strong preferred orientations in amphiboles to diffusion 10.1029/2020GC009171 Figure 11. a) Profile across the Mai'iu fault and schematic spatiotemporal changes of deformation mechanisms interpreted in the context of the seismic cycle (e.g., Sibson, 1992). Blue arrows: influx of a chemically active fluid phase (H 2 O + CO 2 rich). Mauve arrow with Chlorite, Corrensite, Saponite: Stability and change of these minerals as a function of depth. Microseismicity range (∼12-28 km depth) is based on Abers et al. (2016). Schematic curve of maximum (static) differential stress versus depth for the Mai'iu fault is based on Mizera (2019)

(b) (c)Interseismic (d)Late interseismic
creep accompanied by rigid-body rotation in an otherwise weak (plagioclase-rich), fine-grained matrix. Rigid-body rotation of actinolite in our rocks is consistent with widespread microfolding of the mylonitic foliation and fracturing of acicular actinolite grains (Figure 3a, inset). Diffusion creep accommodated by anisotropic dissolution and precipitation of amphibole may also contribute to a strong CPO of amphibole , and both processes are not mutually exclusive. The observed Fe 3+ -rich rims in epidote supports a diffusion creep model (e.g., Elyaszadeh et al., 2018) and are consistent with an elevated oxidation state during exhumational shearing (Daczko et al., 2009).
Based on the fine grain size of the mafic minerals, the strong preferred orientation (CPO and SPO) of actinolite, the syntectonic growth of chlorite (secondary phase admixture), we suggest that grain-size-sensitive creep was an important weakening mechanism in the mylonites. The fine grain size and preferred orientation of actinolite increases the cumulative surface area of the grains and decreases mean interparticle distances (e.g., Herwegh et al., 2011, and references therein), both of which would promote rates of diffusion creep. Newly grown interstitial chlorite may have helped to overcome local strain incompatibilities between grains (cf. Behrmann, 1985;Speckbacher et al., 2013;Stünitz & Tullis, 2001) and limit grain growth (Herwegh et al., 2011;Kilian et al., 2011;Menegon et al., 2015).
A maximum strain rate in the mafic mylonites that accommodated extensional slip on the Mai'iu fault can be calculated by the relation ɛ = ů/h, where ɛ is the strain rate, ů the slip rate and h the thickness of the deforming zone (e.g., Rowe et al., 2011). Using the fault dip-slip rate of ∼10 mm/yr (Webber et al., 2018) and the width of the mylonites (∼60 m and thinning to 1.5 m), calculated shear strain rates range from 5.3 × 10 −12 to 2.1 × 10 −10 s −1 on the assumption that the full width of the mylonites accommodate extension. Getsinger and Hirth (2014) showed that flow laws for wet plagioclase can be similarly applied to fine-grained amphibole. Applying experimentally derived diffusion creep flow laws for wet Ab 100 (H 2 O 0.2 wt.%) by Offerhaus et al. (2001), pure albite with an average grain size diameter of ∼15 µm in the mylonites at temperature of ∼400°C and differential stresses of ∼60-100 MPa Mizera, 2019) is predicted to deform at strain rates of 1.5 × 10 −13 to ∼2.5 × 10 −13 s −1 . This estimated strain rate is lower than that required to accommodate the full dip-slip rate. This shortfall may indicate that: (a) a natural polyphase mafic aggregate at greenschist facies conditions with its neoformed, interstitial chlorite (and with strong CPO and SPO) is weaker in diffusion creep than that predicted by extrapolations of laboratory experiments on wet albite , or (b) the sheared thickness of the mylonite unit was thicker than assumed above.
The transition from the precursor, nonmylonitic schist to the overlying Neogene and younger mylonites is marked by: (a) rotation of the stretching lineation from NNW to NNE trends , (b) a strengthening of the mafic fabric (especially the increase in CPO of actinolite, Figure 5b), and (c) a change from top-to-the-south (thrust sense) to top-to-the-north (normal sense) shear fabrics (Figures 2b and 2c). Both the nonmylonitic schist and the mylonites have similar fine grain sizes (Figure 5b; Tables S4.1-S4.4) indicating that the footwall of the Mai'iu fault was pre-conditioned for grain size sensitive (GSS) creep prior to the onset of extension when the slip-sense on the fault was reversed. In particular, the Mai'iu fault inherited the pre-existing contrast between these fine-grained mafic rocks in the footwall and the coarse-grained ultramafic rocks (predominantly harzburgite and minor serpentine; Smith & Davies, 1976) of the PUB in the adjacent hangingwall. Moreover, we infer that the greenschist-facies mafic minerals already present in the nonmylonitic schist accommodated normal faulting through diffusion creep accompanied by rotation and GBS. At the time of extensional fault inversion, this fault contact dipped >44°, not only at depth but also at the surface, where the oldest hangingwall sediments onlapped against the fault scarp Webber et al., 2020).
Geodynamic models have explored conditions under which a preexisting, failed continental subduction margin may evolve into a domal metamorphic core complex after changes in tectonic boundary conditions (e.g., Biemiller et al., 2019). According to these, extensional inversion is most likely if the precursor thrust is: (a) weak and/or strain softening (see also Choi & Buck, 2012;Choi et al., 2013;Lavier et al., 1999), and (2) originally moderately to steeply dipping . Our observations, together with other data for the Mai'iu fault Mizera et al., 2019;Webber et al., 2020), converge on a scenario of extensional inversion of the Owen-Stanley thrust shear zone, with slip on that inherited fault zone exhuming the footwall through a rolling-hinge process.

Formation of the Foliated Cataclasite and Implications for a Seismic Cycle on the Mai'iu Fault
The upward transition from mylonites to the foliated cataclasites is marked by: (a) bleaching of the greenschist-facies derived metabasalt (lighter in color and less mafic than in the mylonites, Figures 2c and 2d), (b) decreasing mean grain size ( Figure 5b) and increasing intensity of microfaulting and brecciation, (c) development of a mm-to cm-scale spaced foliation defined by alternating albite-rich light and chlorite-rich dark colored folia (Figures 2d and 4) (Figures 2e and 4a). These microstructures lead us to infer that the spaced foliation in the foliated cataclasite unit formed during deformation by the coupled fluid-assisted dissolution of epidote and actinolite, mostly in the dark-colored seams, where it was associated with growth and concentration of chlorite and subordinate clays; and precipitation of albite, calcite and quartz, mostly in the light colored domains.
One effect of dissolution of epidote and actinolite was bleaching of the foliated cataclasite relative to the mafic mylonite from which it was derived. Although this is a fluid-assisted reaction, we did not find evidence for sustained high pore-fluid pressures in the foliated cataclasites. Emplacement of calcite veins indicate that fluid pressures were at times high enough to induce hydrofracturing (i.e., P f >σ 3 ); however, fluid influx was not high enough to completely retrogress the metabasaltic mineral assemblage as observed in other MCCs such as the Moresby Seamount Detachment in the eastern Woodlark Rift (Speckbacher et al., 2012(Speckbacher et al., , 2013. We interpret temporal variations in fluid pressure, the formation and folding of the foliation in the foliated cataclasites, and multiple generation of pseudotachylite veins hosted in these rocks in the context of a seismic cycle-model as explained below (e.g., Scholz, 2002;Sibson, 1992, Figures 11b-11d).
Influx of a chemically active fluid phase (H 2 O + CO 2 ) during embrittlement of the mylonitic rocks (maybe during earthquakes) promoted the dissolution of mafic minerals (coseismic/postseismic period in Figure  11b). In metasomatic reaction R.1, actinolite and epidote react away, leaving residual chlorite ± illite to accumulate in the dark-colored folia. At the same time, the reaction products calcite and quartz were precipitated in dilation sites, particularly within the light-colored folia. Part of the reaction also involved dissolution of albite, diffusion of albite in an intergranular fluid, and re-precipitation of albite in dilation sites, especially in the light-colored folia together with calcite and quartz. Compositional changes of albite after dissolution and re-precipitation were insignificant (Figure 5a).
This mass transfer process (mineral transformation, diffusive mass transfer, and re-precipitation) ultimately led to the development of an interconnected chlorite-rich folia in the cataclasites; a case of paired reaction and textural softening (e.g., Collettini & Holdsworth, 2004;Imber et al., 2001;Moore & Lockner, 2004;Richard et al., 2014;Stewart et al., 2000). Aseismic creep, at least temporarily, was facilitated by stable frictional sliding in the chlorite-rich folia (interseismic; Figure 11c; e.g., Collettini et al., 2009;Jefferies et al., 2006). This is in some ways similar to the pressure solution-accommodated sliding model (frictional-viscous flow) of Bos and Spiers (2002), where minerals are dissolved at sites of stress concentration and re-precipitated in pressure shadows. Natural and experimental observations indicate that the kinetics of diffusive mass transfer is maximized along boundaries between phyllosilicates and more soluble phases (Gratier, 2011;Niemeijer & Spiers, 2005;Richard et al., 2014, and references therein).
To accommodate the Mai'iu fault's full slip rate in the ∼1.5-3.0 m-wide foliated cataclasites by aseismic creep of the interconnected chlorite-rich folia would require a bulk strain rate of 1.1 × 10 −10 to 2.1 × 10 −10 s −1 , but this is likely higher than the actual rate of long-term, frictional-viscous pressure solution creep (e.g., Bos & Spiers, 2002;Imber et al., 2008). Aseismic creep may locally be impeded where chlorite-rich folia become locked against stronger albitic domains. An explanation for the mm-scale differentiated foliation might be that the fluid activity of the precipitating minerals (albite, quartz, and calcite) was insufficient to transport mobile elements large distances in the intergranular fluid. Thus, deposition occurred in nearby voids (e.g., Gratier et al., 2013). Precipitation of albite, calcite and quartz in fractures and pores strengthened the rock, especially in the light-colored folia (e.g., Richard et al., 2014). Such reaction-strengthening would slow down rates of dissolution-precipitation creep (i.e., case of partial locking; see Biemiller et al., 2020) by increasing the mass transfer distance (e.g., Rutter, 1983; late interseismic period, Figure 11d). Increased sealing of the fractures with calcite may have led to a positive feedback loop, whereby the well-cemented, feldspar-rich light-colored folia formed frictionally strong asperities that accumulated stress elastically and ultimately yielded by brittle, cataclastic deformation (cf. Richard et al., 2014).
Ductilely deformed calcite and pseudotachylite veins in the foliated cataclasite unit (Figure 4a, inset) are strong evidence that some slip in this unit was seismic (causing fracturing, grain-size reduction, and frictional melting), and that fast-slip events were followed by postseismic and/or interseismic creep accommodated by slip along the phyllosilicate-rich folia. In other words, the rock experienced a mixed-mode style of slip that was variably seismic to aseismic (e.g., Collettini et al., 2011;Little et al., 2019). The spatial pattern of weakening (phyllosilicate enrichment and alignment, in the dark folia) and strengthening (calcite cementation and veining, in the light folia) caused the rock to become layered and mechanically anisotropic, a situation that contributed to development of fold instabilities along the spaced foliation fabric. Calcite veins reflect episodic periods of fluid inflow that may have influenced the activity and distribution of the different deformation mechanisms (Richard et al., 2014).
Dated pseudotachylite veins hosted by the foliated cataclasite unit yield 40 Ar/ 39 Ar ages as young as ∼2.2 Ma . At the known slip rate of ∼10 mm/yr on a 30-40° dipping fault, such ages imply pseudotachylite generation at 10-12 km depth. We infer that the foliated cataclasite and its pseudotachylite veins formed in a relatively strong, moderately dipping, mid-crustal domain of elevated differential stress ( Figure  11a) where this unit is pervasively folded on the mm-to cm-scale (Figure 4a; see also Little et al., 2019) and the phyllosilicate-rich folia are short and poorly connected. Moreover, the phyllosilicate-rich folia are cross-cut by abundant calcite veins, and partially cemented by calcite (Figures 2d and 4d). We interpret the corridor of microseismicity at ∼12-28 km depth (Abers et al., 2016) to reflect present-day aseismic creep taking place in the downdip, un-exhumed equivalents of the described mylonites and foliated cataclasites (e.g., Chiaraluce et al., 2014Chiaraluce et al., , 2007Vadacca et al., 2016). Microseismicity may occur within asperities encompassed by otherwise creeping material. In contrast, the shallower inferred zone of pseudotachylite formation (10-12 km depth) is interpreted to be currently locked and therefore microearthquake-free (Biemiller et al., 2020).

Grain-Size Reduction, Hydrous Alteration and Slow-To-Fast Slip Processes in the Ultracataclasites
The transition from the foliated cataclasite to the ultracataclasite is marked by: (a) extreme grain-size reduction (Figures 5b and 6b), (b) bulk chemical changes (e.g., loss of Na, Ca and gain of K, Mg; Table  S1.1), (c) mineral transformation from chlorite to trioctahedral mixed-layer smectite-chlorite ±corrensite to saponite and the K-feldspathization of albite (R.2), (d) randomization of all CPOs in actinolite, epidote and titanite (Figure 5b), and (e) formation of K-feldspar ( Figure 6b) and magnetite ±maghemite ( Figures  6g, and 7). We infer that the dark-colored, unfoliated, and clay-rich ultracataclasite unit with its sub-angular to rounded mafic clasts was derived from brittle fragmentation of the footwall. Bulk chemical changes, K-feldspathization of albite and neocrystallization of corrensite and K-feldspar indicate hydrous alteration of the metabasaltic footwall (XRD data of bulk mafic fault rocks Table S7.1). The well-rounded shape of the ultracataclasite lithic clasts in the ultracataclasite, and especially the coated structure of those clasts (clast-cortex aggregates, CCA; Figures 6e-6g), suggest clast rotation during granular flow and accretion of clay material (e.g., Boutareaud et al., 2008;Han & Hirose, 2012;Rempe et al., 2014). Granular flow of the ultracataclasite matrix is also suggested by attitude-dependant stretching versus folding of calcite veinlets embedded in it ( Figure 6a)-a relationship that indicates distributed shearing in that matrix .
We interpret the high magnetic susceptibility of ∼13 × 10 −6 m 3 kg −1 measured in one of the ultracataclasite samples from the Mai'iu fault (compared to the other Mai'iu fault rocks and Goropu Metabasalt; Watson, 2019) and the occurrence of abundant, nm-sized magnetite ±maghemite in ultracataclasite samples (in veins, as coatings, and swirled into reworked ultracataclasite clasts) to have formed by the breakdown of predominantly Fe-bearing mixed-layer smectite-chlorite during frictional shearing (thermomechanical decomposition). This interpretation is supported by changes in rock magnetic properties in other fault zones associated with recent earthquakes (e.g., 1999 Taiwan Chi-Chi earthquake, e.g., Mishima et al., 2009; Wenchuan earthquake in Sichuan Province, China, e.g., Cai et al., 2019) and high-velocity frictional experiments on crushed siltstones (containing quartz with a matrix of clay minerals, such as illite, kaolinite, smectite, and chlorite; e.g., Tanikawa et al., 2007). The high-velocity frictional experiment by Tanikawa et al. (2007) showed that the bulk magnetic susceptibility is proportional to the frictional work applied and increases with displacement due to thermal decomposition of paramagnetic clays in the powdered siltstone. Both thermally and mechanically driven mineral transformation reactions contribute to an anomalously high magnetic susceptibility, similar to that observed in the Mai'iu fault ultracataclasite unit (Figure 7; e.g., Tanikawa et al., 2008).
Frictional devolatilization of Fe-bearing smectite clay is expected at temperatures >400°C (Rowe & Griffith, 2015, and references therein). We did not observe breakdown of mixed-layer smectite-chlorite during the time-dependent magnetic susceptibility test; however, the test was performed under 1 atm and without any frictional work. Another possibility is that the magnetite was derived from external supersaturated hydrothermal fluids (i.e., external Fe-saturated fluids from ultramafic rocks in the hangingwall). This origin may be supported by the occurrence of magnetite along fractures in the ultracataclasites, but it does not explain the precipitation of single magnetite grains scattered within the fault rock or the magnetite nanograin coating around older ultracataclasite clasts (Data S5).
We infer that deformation of the ultracataclasite unit occurred at strain rates that varied with time. Hydrous alteration of the ultracataclasite matrix indicates that fluid inflow led to weakening of the matrix by retrogression of chlorite to corrensite (e.g., Moore, 2014). At the same time, cementation with unstable calcite would promote strengthening (e.g., Verberne et al., 2015). The observed suite of microstructures, including ductilely folded calcite veinlets, in the ultracataclasite unit indicate that periods of slow (probably aseismic) creep potentially alternated with fast (potentially seismic) slip events. CCAs similar to the ones observed in our fault rocks have been reproduced in rotary shear experiments on frictionally heated granular material at seismic velocities (Boutareaud et al., 2010(Boutareaud et al., , 2012. However, granular flow at a wide range of velocities can produce CCAs, even in the absence of thermal pressurization and fault rock fluidization at elevated temperature (Han & Hirose, 2012). While granular flow itself can facilitate CCA formation, the nanograin magnetite that coats clasts in the Mai'iu fault ultracataclasite and magnetite-bearing veins in this unit appear to have originated from the thermal decomposition of Fe-bearing minerals (predominantly mixed-layer smectite-chlorite) by frictional heating of a propagating rupture front during coseismic slip (Cai et al., 2019;Hirono et al., 2006;Mishima et al., 2006;Yang et al., 2012). We note that the inferred depth of the cataclasites (foliated cataclasites/ultracataclasites) of ∼6-15 km depth ( Figure  11a) overlaps with the modeled locking depth of ∼5-16 km depth by Biemiller et al. (2020). The presence of frictionally strong cataclasites prone to seismic slip is compatible with the inferred interseismic locked region that was sampled by densely spaced campaign GPS data gathered across the Mai'iu fault.

Frictionally Weak Fault Gouges and Fault-Faceted Surfaces of Cobbles Truncated by the Principal Displacement Surface
The gouges contain recycled clasts from the underlying mafic footwall (ultracataclasite, foliated cataclasite, and mylonite fragments), and from ultramafic rocks in the overlying hangingwall (dolomite and spinel-group minerals). This recycling is reflected by an increase in grain-size of mono-and polyphase clasts in the gouge relative to those in the underlying ultracataclasite ( Figure 8a). The recycled clasts in the gouges "float" in a fine-grained matrix dominated by saponite ( Figure 8c) and have random CPOs as indicated by a low M-index (Figure 5b). Incorporation of ultramafic components into the gouge potentially promoted saponite-forming reactions (e.g., Moore, 2014;Moore & Rymer, 2012). Saponite is typically derived from the breakdown of chlorite and/or corrensite under declining temperatures. In mafic rocks, its formation requires fluid, and it consumes Mg, Fe, Al, and (Ca + Na + K) derived from epidote, actinolite, albite (and, in our case potentially also the ultramafic hangingwall; e.g., Moore, 2014). Multiple phases of fluid flow might be recorded by our observation of saponitic domains overprinted by silica-rich ones (Figures 8e and 8f).
Our documentation of amorphous, mirror-like surfaces (nanograin coating) on the fault-faceted surfaces of cobbles truncated by the principal displacement surface may reflect seismic slip at shallow depths ( Figures  1g and 9)-although this interpretation is debatable (e.g., De Paola et al., 2015;Verberne et al., 2014, and references therein). In some laboratory experiments, a nanograin coating was produced by extreme grain comminution at seismic to subseismic creep velocities (e.g., Verberne et al., 2014). To assess whether the nanograins may have formed by comminution or shear heating, we calculated the minimum grain size that can be achieved by grinding (the grinding limit) for a basaltic rock, plagioclase and quartz based on fracture toughness (K IC ) and flow stresses at zero °K (τ 0 ; Equation 13 in Sammis & Ben-Zion, 2008). The estimated minimum grain sizes are > d minBasalt = 268 nm, >d minPlag = 53 nm and >d minQuartz = 86 nm, respectively (d minQuartz was calculated by Sammis and Ben-Zion (2008). This estimation is imprecize due to the uncertainties in material properties (Sammis & Ben-Zion, 2008), but it suggests that the interpreted amorphous mirror-like coating on the cobble-surfaces (Figures 9f, 9i, and 9j) cannot be achieved by comminution alone. For this reason, we interpret that this nano-coating of amorphous material (Si and Al rich) and the injection veins (Figures 9d, 9g, and 9h) have been generated by frictional heating and slip along a narrow principal displacement surface (e.g., Oohashi et al., 2011;Rowe & Griffith, 2015;Smith et al., 2013;Wu et al., 2020).

Conclusions
The Mai'iu fault zone, an extensional detachment formed from inversion of a pre-existing ophiolitic suture, developed in a pre-existing, fine-grained, footwall metabasaltic protolith. The upwardly narrowing arrangement of progressively lower-temperature fault rocks developed from this protolith display clear evidence that slip became progressively localized and more brittle as slip on the fault carried the footwall toward the surface. Microstructures in the fault rocks record transitions in dominant deformation mechanisms in metabasaltic rocks accommodating slip, and chlorite thermometry allows us to assign approximate temperatures to these transitions.
• Slip within the mylonites was accomplished by diffusion creep accompanied by rotation and GBS of pre-existing, fine-grained (∼6-35 µm in diameter) epidote, actinolite, titanite, and albite at temperatures >275°C-370°C and strain rates of 5.3 × 10 −12 to 2.1 × 10 −10 s −1 . • At shallower levels on the fault (T ≥ 150°C-275°C), fluid-assisted mass transfer of albite, quartz and calcite led to mineral transformation reactions with continuous chlorite growth, creating a 1.5-3 m thick zone of foliated cataclasites. This zone deformed in part by aseismic frictional-viscous creep at a maximum strain rate of 1.1 × 10 −10 to 2.1 × 10 −10 s −1 . • Build-up of elastic-strain in the foliated cataclasites eventually facilitated earthquakes propagation through them, as indicated by the injection of pseudotachylite veins. Later creep caused folding of those veins. • The foliated cataclasites formed at ∼6-15 km depth in a frictionally strong, mid-crustal part of the fault that dipped at least ∼30°. This zone maintained elevated differential stresses and deformed at slip rates that varied spatiotemporally from high (seismic) to low (aseismic). • Slip in the shallower-formed ultracataclasite unit (T∼150°C-225°C) was at least in part accomplished by distributed granular flow of the ultrafine-grained mafic minerals. • At the shallowest crustal levels (T < 150°C), clay-rich gouges contain abundant saponite, a velocity-strengthening, weak mineral (µ < 0.2). Given sufficient areal distribution on the fault plane, saponite gouges may have promoted aseismic slip on the shallowest dipping, most poorly oriented part of the Mai'iu fault (dipping ∼15-24°).