Abstract Fault zones control the location, architecture and evolution of a broad range of geological features, act as conduits for the focused migration of economically important fluids and, as most seismicity is associated with active faults, they also constitute one of the most important global geological hazards. In general, the repeated localization of displacements along faults and shear zones, often over very long time scales, strongly suggests that they are weak relative to their surrounding wall rocks. Geophysical observations from plate boundary faults such as the San Andreas fault additionally suggest that this fault zone is weak in an absolute sense, although this remains a controversial issue. Our understanding of fault-zone structure and mechanical behaviour derive from three main sources of information: (1) studies of natural fault zones and their deformation products (fault rocks); (2) seismological and neotectonic studies of currently active natural fault systems; (3) laboratory-based deformation experiments using rocks or rock-analogue materials. These provide us with a basic understanding of brittle faulting in the upper crust of the Earth where the stress state is limited by the frictional strength of networks of faults under the prevailing fluid-pressure conditions. Under the long-term loading conditions typical of geological fault zones, poorly understood phenomena such as subcritical crack growth in fracture process zones are likely to be of major importance in controlling both fault growth and strength. Grain-size reduction in highly strained fault rocks produced in the plastic-viscous and deeper parts of frictional regime can lead to changes in deformation mechanisms and relative weakening that can account for the localization of deformation and repeated reactivation of crustal faults. Our understanding the interactions between deformation mechanisms, metamorphic processes and the flow of chemically active fluids is a key area for future study. An improved understanding of how fault- or shear-zone linkages, strength and microstructure evolve over large changes in finite strain will ultimately lead to the development of geologically more realistic numerical models of lithosphere deformation that incorporate displacements concentrated into narrow, weaker fault zones.

Abstract Analysis of stress orientation data from earthquake focal plane mechanisms adjacent to the San Andreas fault in the San Francisco Bay area and throughout southern California indicates that the San Andreas fault has low frictional strength. In both regions, available stress orientation data indicate low levels of shear stress on planes parallel to the San Andreas fault. In the San Francisco Bay area, focal plane mechanisms from within 5 km of the San Andreas and Calaveras fault zones indicate a direction of maximum horizontal compression nearly orthogonal to both subvertical, right-lateral strikeslip faults, a result consistent with those obtained previously from studies of aftershocks of the 1989 Loma Prieta earthquake. In southern California, the direction of maximum horizontal stress near the San Andreas fault is nearly everywhere at a high angle to it, similarly indicating that the fault has low frictional strength. Thus, along these two major sections of the San Andreas fault (which produced great earthquakes in southern California in 1857 and central and northern California in 1906), the frictional strength of the fault is much lower than expected for virtually any common rock type if near-hydrostatic pore pressure exists at depth, and so low as to produce no discernible shear-heating anomaly. Our findings in southern California are in marked contrast to recent suggestions by Hardebeck & Hauksson that stress orientations rotate systematically within c. 25 km of the fault, which prompted a high frictional strength model of the San Andreas fault. As we utilize the same stress data and inversion technique as Hardebeck & Hauksson, we interpret the difference in our findings as being related to the way in which we group focal plane mechanisms to find the best-fitting stress tensor. We suggest that the Hardebeck & Hauksson gridding scheme may not be consistent with the requisite a priori assumption of stress homogeneity for each set of earthquakes. Finally, we find no evidence of regional stress changes associated with the occurrence of the 1992 M7.4 Landers earthquake, again in apparent contradiction with the findings of Hardebeck & Hauksson.

Abstract In the western Woodlark Basin, off Papua New Guinea, the variation from continental rifting to sea-floor spreading has profound effects on the mechanical response of the lithosphere. The extension is well expressed in a seismically active, shallow-dipping detachment fault. Recent Ocean Drilling Program drilling (ODP Leg 180) in the area obtained cores from the hanging wall, detachment fault gouge, and footwall, of which samples underwent permeability testing in the laboratory. Permeability variation was found to be critically dependent on (1) flow direction, i.e. fabric anisotropy of the rocks, and (2) deformational structures in the hanging wall to the fault. Regarding the first, a slight but distinct increase in permeability has been recorded parallel to the fabric (compared with flow normal to this, as indicated by anisotropy indices of K horizontal / K vertical of >1.7). This phenomenon appears most profound directly above fault zones in the hanging-wall block, which are interpreted to represent splays to the main detachment fault plane at depth. Here, shear-enhanced compaction seals fluid flow to the sea floor, so that conductive flow parallel to the fault plane is favoured (in general one order of magnitude higher). The fault gouge, mainly consisting of highly serpentinized basalt and chlorite, exhibits an increase in permeability relative to the clay- and siltstones of the hangingwall block. In the metamorphic basalt from the tectonic footwall, permeability decreases again by three orders of magnitude (k is c. 6e–17 to 5e–18m 2 ). Consequently, the detachment and synthetic splays related to it are zones of enhanced fluid migration in the fault plane direction. Fluid overpressure, and hence fault activity, is suggested to be triggered by seal of the top of the fault zone, owing to both shear fabrics and cementation processes.

Abstract Deformation bands form in porous, clay-poor, sandstones in the top few kilometres of the Earth’s surface, involving the sequential growth of a set of discrete fault strands with minimal individual offset, ultimately culminating in the development of a slip surface with a large offset. We review some of our recent experimental results designed to reproduce the early stages of this sequence, obtained at room temperature and low confining pressure (P < 70MPa) in a large-capacity (10 cm core diameter) deformation rig. We examine the physical weakening and strengthening mechanisms at work in the experiments, and discuss the implications for fault sealing. We describe laboratory experiments where deformation occurs by the progressive formation of new bands with a finite small offset and a relatively constant fault gouge grain-size distribution, at a relatively constant stress measured at the sample boundaries. The friction coefficient is 0.6, i.e. within the standard range. No large-offset slip surfaces were observed. Cross-fault permeability is transiently increased during dynamic stress drop, associated with the ‘suction pump’ provided by rapid near-fault dilatancy under conditions of constant flow rate. As the deformation band develops quasi-statically, permeability is then reduced further by up to two orders of magnitude as a result of shear-enhanced compaction and porosity loss of the poorly sorted gouge fragments. A simple microstructural model successfully predicts the physical sealing rates in the post-failure stage. Finally, we estimate the chemical sealing rates from mass balance calculations based on direct measurement of the pore fluid chemistry during constant flow experiments at temperatures up to 120° C. When extrapolated to longer timescales, these account quantitatively for the differences between permeability reductions measured in the laboratory and in the field.

Abstract Transform faults that offset mid-ocean ridge (MOR) segments accommodate plate motion through deformations that involve complex thermal and mechanical feedbacks involving both brittle and temperature-dependent ductile rheologies. Through the implementation of a 3D coupled thermal-mechanical modelling approach, we have developed a more detailed picture of the geometry of plate boundary deformation and its dependence on plate velocity and the age offset of MOR transforms. The modelling results show that cooling of near-ridge lithosphère (lateral heat transfer) has significant effects in the ductile mantle lithosphere for both the location and style of deformation. The region where străin is accommodated in the subjacent mantle lithosphere is systematically offset from the position of the overlying linear transform fault in the brittle crust. This offset causes the boundary to be oblique to plate motions along much of the transform’s length, producing extension in regions of significant obliquity modifying the location of the surface fault segments. An implication of this complex plate-boundary geometry is that in the near-ridge region, the older (cooler) lithosphère will extend beneath the ridge tip, restricting the upwelling of mantle to the MOR. The melt to generate the oceanic crust adjacent to the transform must migrate laterally from its offset source, resulting in a reduced volume and thinner crust. This near-ridge plate boundary structure also matches the pattern of core-complex extension observed at inside corners of many slow-spreading ridges. The oblique extensional structure may also explain mag-matism that is observed along ‘leaky’ transforms, which could ultimately result in the generation of new ridge segments that effectively ‘split’ large transforms.

Abstract The formation of clay minerals within active fault zones, which results from the infiltration of aqueous fluids, often leads to important changes in mechanical behaviour. These hydrous phyllosilicates can (1) enhance anisotropy and reduce shear strength, (2) modify porosity and permeability, (3) store or release significant volumes of water, and (4) increase fluid pressures during shearing. The varying interplay between faulting, fluid migration, and hydrous clay mineral transformations along the central Alpine Fault of New Zealand is suggested to constitute an important weakening mechanism within the upper section of this crustal discontinuity. Well-developed zones of cataclasite and compacted clay gouge show successive stages of hydrothermal alteration, driven by the cyclic, coseismic influx of meteoric fluids into exhumed amphibolite-facies rocks that are relatively Mg rich. Three modes of deformation and alteration are recog-nized within the mylonite-derived clay gouge, which occurred during various stages of the fault’s exhumation history. Following initial strain-hardening and frictional melting during anhydrous cataclastic breakdown of the mylonite fabric, reaction weakening began with formation of Mg-chlorite at sub-greenschist conditions (<320 0 C) and continued at lo wer temperatures (<120°C) by growth of swelling clays in the matrix. The low permeability and low strength of clay-rich shears are suitable for generating high pore-fluid pressures during faulting. Despite the apparent weakening of the c . 6 km upper segment of the Alpine Fault, the upper crust beneath the Southern Alps is known to be actively releasing elastic strain, with small (<M 5) earthquakes occurring to 12 km depth. We predict that larger events will nucleate at c . 6–12 km along an anhydrous, strain-hardened portion of the fault.

Abstract A sequence of bentonite and shale samples in a gouge zone of the Lewis Thrust (Alberta, Canada) that display increasing degree of transformation of clay minerals toward the hanging wall of the thrust has been studied by X-ray diffraction (XRD), X-ray texture goniometry (XTG), scanning electron microscopy (SEM), and transmission and analytical electron microscopy (ΤΈΜ–ΑΕΜ), to examine the relations among mineral transformations, microfabrics and fault zone properties. TEM images of authigenic clays show abundant smectite in shale away from the hanging wall, characterized by anasto-mosing layers with an average orientation that parallels bedding, coexisting with uncom-mon RI illite–smectite (I–S). In the sample nearest the hanging wall, by contrast, the dominant clay is mixed-layered, illite-rich illite-smectite (R1 I–S), coexisting with discrete illite, occurring in individual packets of relatively straight layers with well-defined boundaries. Deformed clay packets are common. Pore space, where packets intersect at high angles to one another and to bedding, is abundant (c. 25%). The microfabric and proportion of illite of intermediate samples are transitional to these end-members. Interlayered bentonite samples show properties that are similar to those of shale. TEM observations are supported by quantification of the fabrics using XTG, which shows that the intensity of clay preferred orientation decreases significantly with increasing illitization. These relations imply that faulting was the cause of mineral transformations and formation of secondary pore space. The illitization reaction rate was enhanced both by stress-induced defects in clays, and by increased water/rock ratio resulting from deformation-related pore space, resulting in lowering of the effective stress. The deformation-enhanced reaction thus created a positive feedback for further faulting in clay gouge, leading to enhanced weakening of the fault zone.

Abstract Deformation in fault zones is commonly characterized by grain-scale microfracturing, with microcrack density typically increasing toward the middle of the zone. The cracks can form under a wide variety of conditions and need to be used with great caution in making tectonic interpretations, particularly in areas with a complex history of fault reactivation. Microcracks may be intragranular (contained within single grains) or intergranular (with a length of several grain diameters). Intragranular cracks formed under dominantly plastic deformation conditions are crystallographically controlled and may not be directly related to regional stresses. Intragranular cracks formed during initial fracturing under cataclastic conditions develop only in grains that are optimally oriented to the deforming stresses. Intergranular cracks form during progressive cataclasis as intragranular cracks grow to join one another: they may develop as transgranular cracks that cut across several grains or as grain-boundary cracks. Once formed, microcracks may be preserved in a variety of ways (e.g. sintering, healing, cementation) depending on postdeformation conditions, and may be distinguished from one another on the basis of microstructural characteristics. Distinguishing between successive generations of microcracks in areas of fault reactivation is particularly important in determining the deformation history and obtaining deformation conditions. For example, Proterozoic quartzites collected from the central Utah Sevier belt have undergone multiple episodes of contractional deformation followed by Basin-and-Range extension. The use of polarized and dark-field optical microscopy and scanning electron microscopy allows microcracks related to the separate episodes of deformation to be distinguished on the basis of morphology, mode of preservation and consistent cross-cutting relationships. Variations in microcrack density and volume of cataclasized rock for the different generations of microcracks are used to establish the patterns of overprinting during fault reactivation. Anastomosing zones of intense deformation formed during successive episodes of faulting may not coincide with one another, as grain-size reduction and cementing during each episode hardens the zones, causing deformation to shift to adjoining weaker rock. However, the fault zone as a whole is a sufficiently large inhomogeneity that it is reactivated during successive faulting events.

Abstract The Greiner shear zone (western Tauern Window) deformed a variety of metasedimentary, metavolcanic, and plutonic lithologies at conditions of c 525–575° C and 30–40 km depth. Microstructural relationships point to a succession of weakening and strengthening episodes. Stage I involved softening via a change in deformation mechanism. Grain-size réduction in plagioclase-rich horizons locally produced rocks with an average grain size of <30 μ m and microstructural features consistent with deformation via grain boundary diffusion creep (GBDC; a fluid-assisted deformation mechanism similar to pressure solution, which may result in superplastic behaviour). At constant stress, GBDC will result in a significant increase in strain rate relative to neighbouring layers. Stage II involved reaction-induced strengthening. Rapid bulk diffusion rates associated with GBDC allowed rapid growth of large (up to 20 cm) hornblende cystals. This growth of large cross-cutting crystals shut down grain-size-sensitive flow mechanisms in the plagioclase matrix and locally ‘locked’ the shear zone, shifting ductile deformation to weaker horizons. Stage III involved reaction-induced softening. Local variations in bulk composition and/or fluid availability caused large hornblende grains in some horizons to be partially replaced by biotite. These biotite-rich layers localized subsequent deformation, whereas adjacent layers with intact hornblende record minimal strain. Deformation and metamorphism together exert control on fluid availability, diffusion rates, and reaction kinetics, and these factors collectively control fabric development and rheology. The effects of interaction between small-scale deformational and metamorphic processes are difficult to predict, but can have an important influence on shear zone behaviour at depth. The resulting complexities need to be accounted for in models of crustal strength–depth relationships and shear zone rheologies.

Abstract The growth of normal fault arrays is examined in basins where sedimentation rates were higher than fault displacement rates and where fault growth histories are recorded by thickness and displacement variations within syn-faulting sequences. Progressive strain localization is the principal feature of the growth history of normal faults for study areas from the Inner Moray Firth, a sub-basin of the North Sea, and from the Timor Sea, offshore Australia. The kinematics of faulting are similar in both study areas. Fault displacement rates correlate with fault size, where size is measured in terms of either displacement or length. Small faults have higher mortality rates than larger faults throughout the growth of the fault system. Displacement and strain are progressively localized onto the larger faults at the expense of smaller faults at progressively larger scales. Strain localization and the preferential growth of larger faults are attributed to geometric factors, such as size and location, rather than to the mechanical properties of fault rock in individual faults. This conclusion is supported by numerical models that reproduce the main characteristics of fault system growth established from both study areas.

Abstract Mid- to upper-crustal fault zones often possess arrays of shear surfaces whose traces on sections perpendicular to the fault surface and parallel to the ac -plane conform with one or more of the ‘Riedel shear’ orientations. These shear surfaces often are oblique to the transport plane, however, so arrays exhibit monoclinic rather than orthorhombic symmetry. In a mudstone-dominated mélange in Humber Arm Supergroup strata in the Bay of Islands, Newfoundland, and in serpentinites from the base of the Bay of Islands complex, shear surfaces have orientations inclined to major fault-zone boundaries and an inferred ac -plane for macroscopic fault-related deformation. Deformation in these zones exhibits an overall monoclinic symmetry. The 3D, asymmetric character of shear surface fabrics suggests that a factor other than stress or the symmetric strain rate tensor controlled their formation. The velocity gradient tensor in a steady, non-coaxial shearing flow possesses a symmetry consistent with monoclinic fabrics. Shear surfaces in asymmetric arrays may initiate with predictable orientations relative to the velocity gradient tensor and then rotate toward flow apophyses, which identify stable positions in steady, 3D flows.

Abstract In common with many other regions of exposed continental basement, the Late Archaean to Palaeoproterozoic Lewisian Complex, NW Scotland, preserves numerous examples of faults that appear to reactivate pre-existing compositional and structural heterogeneities in the host gneisses. A regionally recognized set of late Laxfordian sinistral strike-slip faults and fractures are spatially associated with pre-existing NW–SE-trending ductile shear zones of Inverian and Laxfordian age. Field observations suggest that most of the sinistral displacements have been accommodated along laterally persistent faults (here termed principal displacement zones (PDZ)) that lie sub-parallel to the pre-existing foliation in the shear zones. Geometric and orientation data collected during structural logging of the PDZ faults have been used to quantitatively test the influence of lithology and pre-existing structural geometry on the spatial patterns of fault development. Stereographic analysis shows a strong geometrical correspondence between the intensity and form of the pre-existing anisotropy and the alignment of the PDZ brittle faults. Spatial clustering of PDZ faults varies depending on lithology (amphibolite v. acid gneiss v. Quartz–mica schist). A close correlation exists between the geometry and intensity of the pre-existing foliation and fault spatial clustering. The results demonstrate that reactivation of pre-existing anisotropies in typical continental basement gneisses exert a significant control on brittle fault development and growth in the upper crust.

Abstract The Salmon River suture zone, western Idaho, is a fundamental lithospheric boundary between the North American craton and the accreted terranes of the Cordilleran margin. The initial juxtaposition along this north–south-oriented structure occurred during Early Cretaceous time. This zone was potentially reactivated twice by subsequent tectonism, once during Cretaceous time and once during Miocene time. The Late Cretaceous western Idaho shear zone formed along the Salmon River suture zone, as denoted by a sharp gradient in the isotopic signature of the granitoids that intruded the lithospheric boundary zone. The reconstructed Late Cretaceous orientation of the western Idaho shear zone contains subvertical fabrics (lineation, foliation). The same boundary also acted as a locus for subsequent Miocene Basin and Range extensional deformation. Domino-style normal faulting and deep (2100 m) basin formation accommodated the motion between the extending accreted terranes to the west and the unextended Idaho batholith to the east. Whereas either the mantle boundary or a crustal-scale structuring controls the regional extent of the extensionally reactivated zone, locally crustal basement faults and lithological contacts control the orientation and precise location of faults that accommodate reactivation. The multiple reactivation of the Salmon River suture zone is critical for several reasons. The Early Cretaceous suture zone apparently created a fundamental lithospheric flaw, which was reactivated after terrane accretion. Whether this zone was a fracture or a shear zone, the fabric in the mantle lithosphere was apparently not ‘healed’ during orogenesis. Thus, juxtaposition of mantle lithosphere, which is inferred to occur by faulting in the uppermost mantle, acts as a weakness during later tectonism. Second, the paucity of strike-slip plate boundaries in the geological record makes sense in the context of reactivation. The vertical, lithospheric-scale nature of these structures makes them particularly susceptible to lithospheric-scale reactivation during both transcurrent and/or extensional deformation. These reactivations both overprint the earlier deformation and modify the original geometry. Steeply dipping fabrics, rather than vertical fabrics, may be the general signature of major, ancient strike-slip faults.

Abstract Metamorphic and plutonic basement rocks and cover sequences of the Eastern Sierras Pampeanas, Argentina, have undergone multiple episodes of fault reactivation. Faults take advantage of mid- to late Cambrian, NW-SE-striking, steeply east-dipping foliations in Vendian-aged accretionary prism metasedimentary rocks. Foliations in peraluminous schists, paragneisses and migmatites are deflected into late Cambrian amphibolite-grade high-strain zones. Greenschist-grade mylonite zones and thick retrogressed ultramylonite zones with mainly NNW strikes, easterly dips, and east-over-west movement, affect the metasedimentary rocks and Ordovician-aged intrusive rocks and are presumably related to early Devonian accretion of terranes to the west of Gondwana. pseudotachylyte veins occur in nearly all mylonite zones. Brittle deformation during Carboniferous to Triassic time produced major pull-apart basins located above terrane boundaries. Outcrop patterns of Triassic to Cretaceous sedimentary rocks are consistent with transtensional pull-apart basins followed by Andean transpressional deformation. The theoretical basis for fault reactivation and production of ‘short cuts’ is examined in the context of Tertiary to Recent basin inversion faults. The inversion faults follow the Palaeozoic trends and produce the present-day NNW-oriented, deep sedimentary basins and intervening ranges of basement rocks.

Abstract The Brevard Fault Zone is a linear, NE-trending, gently to moderately SE-dipping fault zone traceable some 750 km from Alabama to Virginia in the crystalline southern Appalachians. It ranges from 1 to 3 km wide and contains a mappable lithostratigraphy. The Brevard Fault Zone has been interpreted as a thrust, strike-slip fault (both dextral and sinistral), a suture and terrane boundary, and a fundamental crustal tectonic boundary. Deformation was partitioned in space and time, and motion was both strike-slip (dextral) and dip-slip (thrust). Early strike-slip and thrust movement was coupled to map-scale structures in the deep Inner Piedmont, late Palaeozoic dextral motion was confined to a zone of 1-3 km width, and the latest reactivation consisted of brittle thrusting confined to a zone of 100 m width. The fault zone is cut by undeformed NW-trending Mesozoic dolerite dykes. The Brevard Fault Zone is characterized by the presence of a prominent retrograde (chlorite-muscovite stable) S–C fabric that indicates dextral motion. This fabric is related to late Palaeozoic (Alleghanian) dextral reactivation of the fault zone, with an unknown displacement at a time when huge volumes of fluid were fluxed through the zone. The deformation overprints an earlier (Acadian) high-tempera-ture (garnet–staurolite–kyanite) fabric that also yields a dextral motion sense, and involved a component of thrusting. This mid-Palaeozoic deformation was coupled with west-directed, near-metamorphic peak thrusting and flow from the deep Inner Piedmont (to the east) that was buttressed against the primordial Brevard Fault Zone so that the motion became SW directed, and plastic flow became constricted in this narrow 1-3 km zone. Both of these plastic deformations were overprinted by late Alleghanian NW-directed dip-slip brittle deformation confined to the NW side of the Brevard Fault Zone. This last deformation involved at least 10 km of displacement and was related to reactivation of this block of crust as part of the late Alleghanian, NW-directed Blue Ridge-Pied-mont megathrust sheet, and formed out of sequence with respect to the megathrust sheet. The Brevard Fault Zone was clearly a zone of crustal weakness that had a suitable mech-anical stratigraphy that imparted sufficient anisotropy to localize the initial Acadian fault(s). Early Alleghanian fluid fluxing weakened the already strongly anisotropic fault zone and probably focused ductile reactivation at a shallower crustal depth during the early Alleghanian event. Late Alleghanian reactivation occurred even shallower as an almost discrete boundary in the brittle regime. This is one of the few faults in the Appalachians to have undergone deformational partitioning to permit multiple reactivation during Palaeozoic time.

Abstract Italy owes its complex geological structure to a double switch in tectonic regime, which involved the opening of the Tethys Ocean during Early Mesozoic time, its closure leading to development of the Apennine-Maghrebide fold-and-thrust belt during the Eocene-Recent interval, and the post-orogenic opening of the Tyrrhenian Sea since Miocene time. This history of tectonic inversion is partly preserved within two major fault zones, the Valnerina Line, in the central Apennines, and the Gratteri-Mount Mufara Line, in centrai-northern Sicily, which were repeatedly reactivated with different kinematic characters. The relatively long life of these structures indicates that strain was localized along anisotropies inherited from early deformation episodes. However, the progressive widening of both fault zones through time may result from strain-hardening fault-rock behaviour during subsequent deformations, thus suggesting that fault reactivation does not imply fault-zone weakening as is often assumed.

Abstract Examples of weak fault zones in and bordering Precambrian Sweden are reviewed and then analysed in terms of the factors that rendered them weak. The criterion taken here for weak zones is evidence of post-glacial uplift having reactivated old shear zones that are still active now. Strain analysis of the Singö shear zone demonstrates that it was already weak while it was deep and ductile between 1.86 and 1.6 Ga. Thus the orientations of străin ellipsoids indicate pure rather than simple shear across shear strands in which the dissolution of quartz and feldspar indicate high synshear fluid pressures. The characteristics of weak strands of the Baltic–Bothnian shear zones and the Senja–Bothnia shear zones are long histories of superimposed ductile, semi-ductile and brittle structures, indicating repeated reactivations with different kinematics. The shear zones that weakened are those that reactivated in several episodes during their history. Repeated reactivation renders major shear zones in the ‘brittle’ upper crust ‘ductile’ by rounding major angularities along subvertical and subhorizontal zones. This generates continuous seams of cataclasites in which pore fluids can pressurize so that internal strains can remain ductile and almost aseismic by frictional sliding or flow. This review ends by discussing how multiple reactivations weaken major faults and how reactivation remains focused on particular zones.

Abstract The rheology of crustal fault zones containing melts is governed primarily by two strain-dependent mechanical discontinuities: (1) a strength minimum parallel to mylonitic foliation just below the active brittle-viscous (b-v) transition; (2) the anatectic front, which marks the upper depth limit of anatectic flow. The mode of syntectonic melt segregation in fault zones is determined by the scale of strain localization and melt-space connectivity, to an extent dependent on strain, strain rate and melt fraction in the rock. Melt drains from the mylonitic wall rock into dilatant shear surfaces, which propagate sporadically as veins. Anatectic flow at natural strain rates therefore involves melt-assisted creep punctuated by melt-induced veining. On the crustal scale, dilatant shear surfaces and vein networks serve as conduits for the rapid, buoyancy-driven ascent of transiently overpressured melt from melt-source rocks at or just below the anatectic front to sinks higher in the crust. Strength estimates for natural rocks that experienced anatectic flow indicate that melts weaken the continental crust, particularly in depth intervals where they spread laterally beneath low-permeability layers or along active shear zones with a pronounced mylonitic foliation. However, acute weakening associated with strength drops of more than an order of magnitude occurs only during short periods (10 3 –10 5 a) of crustal-scale veining. Cooling and crystallization at the end of these veining episodes is fast and hardens the crust to strengths at least as great as, and in some cases greater than, its pre-melting strength. Repeated melt-induced weakening then hard-ening of fault zones may be linked to other orogenic processes that occur episodically (shifting centres of clastic sedimentation and volcanism) and has implications for stress transmission across orogenic wedges and magmatic arcs.