Mid-crustal shear zone development under retrograde conditions: pressure–temperature–fluid constraints from the Kuckaus Mylonite Zone, Namibia

The Kuckaus Mylonite Zone (KMZ) forms part of the larger Marshall Rocks–Pofadder shear zone system, a 550 km-long, crustal-scale strike-slip shear zone system that is localised in high-grade granitoid gneisses and migmatites of the Namaqua Metamorphic Complex. Shearing along the KMZ occurred c. 40 Ma after peak granulite facies metamorphism, during a discrete tectonic event, and affected the granulites that had remained at depth since peak metamorphism. Isolated lenses of 5 metamafic rocks within the shear zone allow the P–T–fluid conditions under which shearing occurred to be quantified. These lenses consist of an unsheared core that preserves relict granulite-facies textures, and is mantled by a schistose collar and mylonitic envelope that formed during shearing. All three metamafic textural varieties contain the same amphibolite-facies mineral assemblage, from which calculated pseudosections constrain the P–T conditions of deformation at 2.7–4.2 kbar and 450–480 ◦C, indicating that deformation occurred at mid-crustal depths through predominantly viscous flow. Calculated T– 10 MH2O diagrams show that the mineral assemblages were fluid-saturated, and that lithologies within the KMZ must have been rehydrated from an external source and retrogressed during shearing. Given that the KMZ is localised in strongly dehydrated granulites, the most likely source of external fluid is meteoric :::: fluid :::: must ::::: have :::: been :::::: derived ::::: from :: an ::::::: external :::::: source, with fluid flow allowed by local dilation and increased permeability within the shear zone. The absence of hydrothermal fractures or precipitates indicates that, even though the KMZ was fluid-bearing, the fluid–rock ratio and fluid pressure remained low. In 15 addition, the fluid could not have contributed to shear zone initiation, as an existing zone of enhanced permeability is required for fluid infiltration. We propose thatthe KMZ initiated by the reactivation of existing, favourably-oriented ductile structures, following which , :::::::: following :::::::: initiation, : fluid infiltration caused a positive feedback that allowed weakening and continued strain localisation. Therefore, the main contribution of the fluid was to produce retrograde mineral phases and facilitate grain size reduction. Features such as tectonic tremor, that are observed on active faults under similar conditions as described here, may 20 not require high fluid pressure, but could be explained by reaction weakening under hydrostatic fluid pressure conditions.

The problem of fluid flow and localised deformation in ductile shear zones is not restricted to exhumed examples, such as that presented here, but also a current question of interest in active plate boundaries. For example, the discovery of tectonic tremor and slow slip on the deep extensions of the Alpine and San Andreas Faults (Shelly, 2010;Shelly and Hardebeck, 2010;Wech et al., 2012) highlight the presence of localised structures below the brittle-viscous transition. The nature of these seismic signals also require the deep ductile root of these major faults to be significantly weaker than their surrounding rocks 5 (Shelly, 2010;Shelly and Hardebeck, 2010;Wech et al., 2012). As opposed to similar features in subduction zones, commonly associated with prograde metamorphism and high fluid pressures, the tremors on the deep San Andreas and Alpine Faults cannot be easily explained by in situ production of fluids in low porosity fault zones. It is possible that San Andreas Fault tremor is related to retrograde weakening mechanisms associated with the introduction of an external fluid (Fagereng and Diener, 2011). An additional question to address here, is therefore how fluid flow and shear zone weakening mechanisms on 10 the Kuckaus Mylonite Zone can serve as an analogue to those occurring on the deep extension of active faults exhibiting tremor and slow slip.
In the study area, the KMZ occurs in granitic gneisses that form part of the Aus granulite terrain (Jackson, 1976;Rennie et al., 2013). These rocks experienced peak metamorphic conditions of 5.5 kbar and 825 • C, with the timing of metamorphism 30 constrained at c. 1065-1045Ma (Diener et al., 2013. Metamorphism is inferred to have been dominated by heating and cooling, with only minor attendant crustal thickening and burial (Diener et al., 2013). The post-peak metamorphic retrograde path involved near-isobaric cooling, indicating that the terrain remained at depth as it cooled to a stable geotherm (Diener et al., 2013).
The shear zone core of the KMZ is about 1000 m in width, and consists of anastomozing high strain ultramylonite zones that wrap around lower-strain lozenges (Rennie et al., 2013). Rock types within the KMZ are dominated by granitic gneisses and mylonites, and only minor enclaves and lenses of retrogressed mafic granulite are present. These mafic lenses occur as 5 discrete units, and range from a few centimetres to 10-15 m long, and are up to 5 m in width (Rennie et al., 2013). Larger mafic lenses have a core of coarser-grained gneisses that are not pervasively mylonitised and in which remnant migmatitic granulitefacies textures can be recognised (Fig. 2a). The core is enveloped by more intensely sheared and retrogressed mylonitic schists, with the increase in strain occurring over a distance of 10-50 cm (Fig. 2b). The fabric in the coarser-grained gneisses and enveloping mylonites has a similar orientation to the penetrative subvertical foliation in the KMZ, and the weakly-developed 10 amphibole lineation is parallel to the subhorizontal quartz rodding lineation that is present in the granite gneisses and mylonites (Rennie et al., 2013). Whereas the volumetrically dominant felsic gneisses and mylonites do not contain mineral assemblages that record distinctive P -T conditions, the mafic enclaves provide a record of recrystallisation conditions from a preserved migmatitic core to a largely recrystallised mylonitic envelope. Therefore, to constrain the P -T -fluid conditions of shear zone deformation, three samples were chosen as a representative section from the core of a low strain lens, preserving peak fabrics 15 and mineral assemblages, into the well-developed retrograde mylonite zone, and these are described further below.
The low strain sample KMZ28 is coarse-grained and equigranular, with typical grain sizes on the order of 0.2-1 mm (Fig. 3a).

Mineral Chemistry
Mineral compositions were determined using a JEOL JXA-8100 electron microprobe housed at the University of Cape Town.

10
Analyses were carried out using a 15 kV acceleration voltage, 20 nA probe current and 2 to 3 µm spot size. Counting times were 5 seconds for both backgrounds and 10 seconds for peaks on all elements. Data were processed with ZAF matrix corrections and reduced using the PAP procedure. Compositions were quantified using natural mineral standards. Representative mineral compositions for the three samples are presented in Table 1.
All samples show the same trends in mineral compositions, with the amphibole in all samples being hornblende (sensu lato), 15 with appreciable Al and Na content (1.5-2 and 0.2-0.25 cations per formula unit, respectively), and X Fe of 0.47-0.65 (Table 1).

Mineral Equilibria Modelling
Mineral equilibria calculations were performed with the THERMOCALC program of Powell and Holland (1988) Table 2.

P -T pseudosections
Fluid-saturated P -T pseudosections for samples KMZ28, KMZ29 and KMZ30 are presented in Fig. 5. The phase relations in all three samples have a similar topology, and consist of the typical greenschist-facies assemblage actinolite-chlorite-epidote-30 albite-sphene-quartz at T below 450 • C, and contain the typical amphibolite-facies assemblage of hornblende-plagioclaseilmenite-quartz at T above 550-600 • C (Fig. 5). In the T range between 450 and 550-600 • C, these rocks undergo a series of phase changes, notably (1) the introduction of hornblende at the expense of actinolite, (2) the introduction of plagioclase and the demise of albite and epidote, (3) the replacement of sphene by ilmenite, and finally (4) the demise of chlorite (Fig. 5). Within this T zone, the inferred equilibrium assemblage of hornblende-plagioclase-chlorite-epidote-sphene occurs in a narrow, Tsensitive field at around 450 • C and between 2 and 4 kbar in KMZ28 and KMZ29 (Fig. 5a,b), but spans the entire P range of interest in KMZ30 (Fig. 5c). This field is bound by the removal of plagioclase to lower T and the loss of epidote to higher T .

5
Contours of X an calculated for KMZ29 indicate that the composition of plagioclase varies substantially, from X an = 0.32 to X an = 0.82, at the stability conditions of the inferred equilibrium assemblage (Fig. 5b).

T -M H2O pseudosections
Calculated T -M H2O pseudosections allow the degree of fluid-saturation in these rocks to be quantitatively evaluated, and are presented for the three samples in Fig. 6. The diagrams are calculated at constant P of 4 kbar in order to bracket the peak-

P -T conditions of shearing
The three samples described above were affected by KMZ-related retrograde deformation to different degrees, and can therefore be used to effectively constrain the P -T conditions under which quasi-plastic shearing in the KMZ occurred. The stability of 30 the observed assemblages are summarised in Fig. 7, and the overlap between the samples constrains the most likely P -T conditions experienced by these rocks at 2.7-4.2 kbar and 450-480 • C. Whereas T is tightly constrained, the P estimate is less precise, and straddles the kyanite-andalusite phase boundary. However, no aluminosilicates are present in the KMZ, nor have they been reported for other parts of the MRPSZ, and consequently the presence of kyanite or andalusite can not be used to refine the P estimate further.
The estimates indicate that the bulk of shearing in the KMZ occurred at 450-480 • C, at a depth of 10-16 km, assuming 5 overlying granitic crust. These conditions are significantly warmer than the brittle-viscous transition in quartz (Hirth et al., 2001), and the unstable-stable frictional transition in granitic rocks (Blanpied et al., 1995), and roughly coincides with the onset of crystal-plastic deformation of feldspar at geological strain rates (Pryer, 1993;Rybacki and Dresen, 2000). These areas (e.g. Fig. 2a), the mineralogical features have been pervasively overprinted at the P -T conditions constrained above.
One mineralogical aspect that could potentially be inherited from the earlier, high-T history is the anorthite-rich composition 20 preserved in plagioclase cores. However, calculated isopleths of anorthite content in plagioclase (only shown for KMZ29 in Fig. 5b) indicate that the plagioclase composition varies widely at, or near, the P -T constrained for the KMZ. This does not preclude the plagioclase cores being inherited from earlier, higher-T conditions, but shows that the core compositions are not far from equilibrium with the remaining assemblage at 450-480 • C. Consequently, it appears plausible that shearing along the KMZ occurred at near-constant T conditions of ∼ 450 • C at a depth within the range of 10 to 16 km, and that the KMZ did 25 not undergo progressive exhumation and cooling during its development. The KMZ is therefore related to a discrete tectonic event, and does not form a continuum of earlier granulite-facies metamorphism, which is supported by its large-scale structural discordance with the granulite-facies structures and fabrics (Jackson, 1976;Toogood, 1976;Moen and Toogood, 2007;Miller, 2008;Rennie et al., 2013). It also follows that the KMZ likely acted as a transcurrent shear zone, with little to no associated transpression or transtension.  -Ferrari et al., 2002). The thermal profile of these shear zones follow the continental geotherm that is stable at the 35 time, which in the case of the KMZ is estimated at 30-45 • C.km −1 . The uncertainty in this estimate is caused by uncertainty in the depth of the KMZ, but the entire range is significantly warmer than for stable continental lithosphere (∼ 20 • C.km −1 ; e.g. Cooper et al., 2002, and references therein). However, given that the Aus granulite terrain experienced granulite-facies metamorphism associated with lithospheric thinning only 40 Ma earlier, it is to be expected that the terrain was characterised by higher than average heat flow at the time of shear zone formation (e.g. Morrissey et al., 2014;Bial et al., 2015). In fact, the 5 estimated geotherm is comparable with those derived from crust underlain by young and thin lithosphere, such as the northern San Andreas Fault (Fagereng and Diener, 2011).

Fluid regime, fluid source and infiltration mechanisms
The rocks in which the KMZ localised were dehydrated and experienced partial melting and melt loss during preceding metamorphism (Diener et al., 2013). In the absence of rehydration, the rocks would have retained their (low) fluid content and blages with the minerals observed in the samples described here, they must have experienced an addition of H 2 O (Fig. 6). The 15 observed assemblages in all three samples straddle the H 2 O saturation line, indicating that they can occur at fluid-saturated or slightly fluid-undersaturated conditions. However, given that the absolute fluid content required by each assemblage differs widely between the samples, it appears unlikely that the samples were all rehydrated to be just fluid-undersaturated, and it is more plausible to conclude that they were fully rehydrated and fluid-saturated during shearing along the KMZ.
Kilometre-scale, open-system diffusion of water through crustal shear zones is not a unique phenomenon, and was proposed by Beach (1980) for major retrogression in ductile shear zones cross-cutting the Archaean Lewisian basement complex of northern Scotland, in similar setting and metamorphic conditions to the current study. Similarly, significant retrograde fluid influx has been inferred for shear zones at Broken Hill, Australia (Etheridge and Cooper, 1981), the French Pyrenees (McCaig et al., 1990) and the Yellowknife gold district, Canada (Kerrich et al., 1977); in the latter location, fluid flux is associated with hydrothermal vein mineralization as well as retrograde reactions. Based on these, and other examples, it has been estimated that typical fluid-rock ratios in retrogressed shear zones can exceed 10 2 (Etheridge et al., 1983), and can locally be much greater (Kerrich et al., 1977;McCaig et al., 1990). This is three orders of magnitude higher than the minimum ratio fluid-rock ratio 5 required for the KMZ, which again indicates that, although the KMZ was fluid-saturated, it was likely not inundated with fluid to the same extent as many other examples of retrograde shear zones.
The above examples all include an element of dip-slip displacement. In another example of fluids in a subvertical strike-slip 5 shear zone, the deep San Andreas Fault is suspected to be fluid-rich, based on high V p/V s in the lower crust (Ozacar and Zandt, 2009). A likely source for this fluid is a serpentinized mantle wedge, where serpentinization dates to past subduction, and the absence of a cool, insulating slab now leads to heating and dehydration (Kirby et al., 2002;Fulton and Saffer, 2009).
This San Andreas Fault case is, however, somewhat special in requiring a transition from subduction to strike-slip tectonics, and we do not see evidence for the presence of a similar serpentinized mantle wedge as a source of fluids for KMZ rehydration, 10 neither does the tectonic history of the Namaqua-Natal Belt call for an initial subduction origin for the MRPSZ. Thus, ::::: while :: we ::: do ::: not ::::::: preclude ::::: other :::::::::: alternatives, : our best hypothesis is that the external fluid source for KMZ rehydration was meteoric, and fluid flow was allowed by local dilation and/or increased fracture permeability within the shear zone, thus accounting for retrograde metamorphic reactions within the mylonites, and absence of such reactions outside the KMZ.
The San Andreas Fault is inferred to be weak and wet (e.g. Ozacar and Zandt, 2009), commonly explained by high fluid pressures (Rice, 1992). The Alpine Fault, New Zealand, has on the other hand been interpreted as frictionally strong (Boulton et al., 2014), and potentially weakened at depths by fluid presence (Wannamaker et al., 2002). Both of these faults are currently 20 active under retrograde metamorphic conditions, and exhibit tectonic tremor, a persistent low-frequency seismic signal characterised by lack of impulsive body wave arrivals, emanating from below the brittle-viscous transition (Shelly, 2010;Shelly and Hardebeck, 2010;Wech et al., 2012). These tremor signals have, in both places, been interpreted as fluid-enabled slip on the deep extension of the plate boundary faults, below the base of the seismogenic zone (Shelly and Hardebeck, 2010;Wech et al., 2012). The KMZ is exhumed from the same thermal and metamorphic regime as where San Andreas and Alpine Fault tremor is 25 occurring. The KMZ also deformed under fluid-present conditions, and was significantly weak compared to surrounding rock.
However, our observations and inferences along the KMZ require fluid presence, but do not require elevated fluid pressure to explain this weakening. As suggested by Fagereng and Diener (2011) for the San Andreas Fault, we therefore suggest that deformation localisation below the brittle-viscous transition at retrograde conditions may be a function of reaction weakening and metamorphic rehydration. At these conditions, localised slip may occur along weak planes characterised by aligned 30 phyllosilicates (Wintsch et al., 1995;Imber et al., 1997) The KMZ was localised in dry, high-grade mid-crustal gneisses, such that shear zone initiation most likely occurred through reactivation of existing favourably-oriented high-temperature, ductile structures (Worley and Wilson, 1996;Montesi, 2013).
Once shearing was active, it allowed fluids, most likely meteoric, ::::::::::::::: externally-derived ::::: fluids : to infiltrate the KMZ and activate a number of positive feedback processes that allowed weakening and continued strain localisation. The observed mineral 5 assemblages lead us to conclude that the KMZ was fluid-bearing during deformation, but the absence of hydrofractures and hydrothermal precipitates indicate that fluid pressures and the fluid-rock ratio remained low. In this regard the KMZ differs from most other exhumed and active continental retrograde shear zones for which high fluid-rock ratios have been suggested (Kerrich et al., 1977;Etheridge et al., 1983;McCaig et al., 1990), or for which high fluid pressures are inferred (Rice, 1992;Ozacar and Zandt, 2009). It consequently appears that retrograde shearing can be sustained under a variety of fluid regimes, 10 from dry and entirely fluid-absent (Tenczer et al., 2006), phases and facilitating grain size reduction. We also conclude that weak lower crustal, fluid-bearing shear zones do not need to imply high fluid pressures, but can also be significantly weaker than surrounding wall rocks from reaction weakening at hydrostatic fluid pressure conditions.