Abstract

Many large-slip faults, such as the San Andreas fault and low-angle normal faults (LANFs), appear to be weak relative to their surroundings or to laboratory friction measurements, and to be poorly oriented for slip in the regional stress field. Several models seek to explain the mechanics of slip and/or formation of such faults. Other models explain damage around faults as due to fault or earthquake rupture propagation or slip on nonplanar faults. Most of these models explicitly predict the near-fault stress field.

Exhumed footwalls of low-angle normal faults are advantageous natural laboratories for testing such models because they expose rocks that passed through the brittle-plastic transition and all or part of the seismogenic crust. We present reduced paleostress tensors derived from inversion of fracture and slip-line orientation data taken mainly from the fault cores and fractured damage zones in the upper footwalls of two LANFs, the Whipple and West Salton detachment faults of southern California. Frictionally weak materials probably were not significant along these faults except in the uppermost few kilometers of the crust, and pore-fluid pressure probably never approached lithostatic values.

Most results show that the faults were at a high angle to the near-fault maximum compressive stress (σ1) direction, in general accord with Andersonian extensional stress fields. Our results support a “strong-sandwich” mechanical model for slip in the upper crust, in which normal-friction LANFs are embedded in stronger surroundings and slip at high angles to σ1 and models of stress rotation across the thickness of the brittle crust, with moderately plunging σ1 near the brittle-plastic transition, provided that some mechanism allows the faults to propagate through the brittle crust at gentle dips as the footwalls are exhumed. Paleo-σ1 vectors oriented at moderate angles to the faults are sparse and may reflect early damage formed in the midcrust, while the angle between σ1 and the detachment was moderate or during along-strike LANF or earthquake rupture propagation. Coulomb plasticity due to granular flow, which predicts faults at ∼45° to σ1, is not well supported because many paleo-σ1 vectors with moderate angles to the LANFs are from fractures below the cataclastic fault cores. Our results are inconsistent with “weak-sandwich” models that predict reorientation of σ1 to low angles (∼30°) to the fault within the damage zone and/or fault core due to local pore-fluid pressure or elasticity changes. Fracturing due to slip on non-planar faults is generally consistent with our paleostress results. However, the roughness of the LANFs studied is not known, but they may have very low roughness. The stress state used in this wavy-fault model to constrain the expected damage region is nearly identical to that inferred in the strong-sandwich model from field measurements. Fractures in the damage zone probably do not record up-dip fault or earthquake rupture propagation, which is expected especially for earthquake propagation, but along-strike propagation may have controlled fracturing at some sites. Some paleostress fields are probably related to folding of the detachments about slip-parallel axes.

INTRODUCTION

Low-angle normal faults (LANFs) are considered to be weak because they apparently slip at low resolved shear traction and at a high angle to the regional maximum principal stress (σ1) direction (Reynolds and Lister, 1987, 1990; Axen and Selverstone, 1994; Axen, 2004). At typical fault friction values (μ = 0.6–0.85; Byerlee, 1978), slip should not occur in an Andersonian stress field unless other factors are important, yet much horizontal crustal extension in the world is accommodated on low-angle detachments (Wernicke, 1981, 1992, 1995; Wernicke et al., 1985; Howard and John, 1987; Davis, 1988; Davis and Lister, 1988; Spencer and Reynolds, 1989; Axen et al., 1990, 1993; Axen, 1993; Abers et al., 1997; Axen and Fletcher, 1998; Axen, 2004; Collettini et al., 2006; Reston, 2009; many others). The San Andreas fault also appears to be a weak fault that slips at a high angle to the inferred maximum horizontal stress (Lachenbruch and Sass, 1980; Mount and Suppe, 1987; Zoback et al., 1987; Lachenbruch and McGarr, 1990; Rice, 1992; Hardebeck and Hauksson, 2001) and may share similar mechanical properties to LANFs; therefore, this study has broad implications for fault mechanics.

Several testable hypotheses have been proposed to explain the weak-fault paradox; most hypotheses were developed for strike-slip faults and others for low-angle normal faults. Many hypotheses are applicable to both fault types. (1) Low-friction materials, such as aligned clay minerals, are common on fault surfaces or in fault cores (Numelin et al., 2007; Collettini et al., 2009; Carpenter et al., 2011; Haines and van der Pluijm, 2012; Lecomte et al., 2012). (2) High pore-fluid pressure, causing low effective normal stress, can reduce apparent friction (Hubbert and Rubey, 1959). (3) Dynamic weakening processes may reduce friction during seismogenic slip (Brodsky and Kanamori, 2001; Wibberley and Shimamoto, 2005; Rempel, 2006; Rice, 2006). (4) Physical-mechanical differences between fault-zone rocks and their surroundings may allow slip on misoriented faults (Figs. 1–3; Mandl and Fernandez Luque, 1970; Mandl et al., 1977; Axen, 1992; Byerlee and Savage, 1992; Rice, 1992; Lockner and Byerlee, 1993; Axen and Selverstone, 1994; Marone, 1995; Faulkner et al., 2006; Healy, 2008, 2009). Most of these models require rotation of σ1 to a lower angle to the detachment within the width of the fault zone (Mandl and Fernandez Luque, 1970; Mandl et al., 1977; Axen, 1992; Byerlee and Savage, 1992; Rice, 1992; Lockner and Byerlee, 1993; Marone, 1995; Faulkner et al., 2006). (5) In contrast, stress rotation may occur across the thickness of the brittle crust, with σ1 in a subvertical Andersonian orientation in the shallow crust and rotated to moderate plunges at and below the brittle-plastic transition, due to underlying plastic shear (Fig. 4; Yin, 1989; Buck, 1990; Westaway, 1999, 2001) or crustal thickness variations (Spencer and Chase, 1989). Such rotation would aid LANF slip in the deeper brittle crust, where slip is most difficult (Axen, 2004), but does not explain LANF slip in the shallower crust (e.g., Wernicke et al., 1985; Axen, 1993). In addition, two other models suggest that stresses may be reoriented near faults and may control fault-zone damage: (6) near-field stress rotation produced by fault propagation (Fig. 5; Vermilye and Scholz, 1998) and/or earthquake rupture propagation (e.g., Rice et al., 2005) and (7) near-field stress perturbations due to fault roughness (Fig. 6; Chester and Chester, 2000). Most models of all types above were developed from a combination of materials theory and experimental rock-mechanical results. Only a few incorporate fault-zone observations directly (Axen and Selverstone, 1994; Vermilye and Scholz, 1998; Chester and Chester, 2000; Healy, 2009), and relatively few have been tested using exposed fault-zone rocks.

Low-angle normal faults offer the opportunity to observe fault-zone rocks developed along weak faults: their generally large-magnitude slip commonly has delivered directly to the surface footwall shear- and fault-zone rocks that are the products of fault-zone processes beginning near or below the crustal brittle-plastic transition and overprinted during passage upward through the seismogenic brittle crust (e.g., Davis, 1983; Wernicke, 1985; Reynolds and Lister, 1987; Davis, 1988; Axen and Selverstone, 1994; Selverstone et al., 1995; Wawrzyniec et al., 1999, 2001; Cowan et al., 2003; Collettinni and Holdsworth, 2004; Hayman, 2006; Numelin et al., 2007; Smith et al., 2008; Selverstone et al., 2012).

We present inversions of macroscopic fracture and slip-vector orientations from the uppermost parts of the footwalls of two low-angle normal faults (the Whipple detachment fault [WDF] and the West Salton detachment fault [WSDF], southern California) to obtain local paleostress orientations. Fractures were measured at sites mostly from within the faults’ damage zones or breccia-ultracataclastic cores, within meters to tens of meters of the faults, with fewer sites below those levels providing “background” comparisons. These paleostress-field orientations can be compared to predictions made by several of the models above, specifically types 3, 5, 6, and 7. The models are discussed in the Weak Fault and Fault Damage section ; the geology of the Whipple and West Salton detachment faults is described in the Whipple and West Salton Detachment Faults section; and methods of inversion are discussed in the Paleostress Inversion section. Paleostress results are compared to the models in the Results section.

Similar stress-field inversions have been done around the San Andreas fault system using low-magnitude seismicity (e.g., Jones, 1988; Hardebeck and Hauksson, 1999; Townend and Zoback, 2001; Provost and Houston, 2001, 2003; Hardebeck and Michael, 2004). However, these studies suffer from map-view bins with sizes ≥2 km perpendicular to the faults; these bins limit the ability to separate “near-fault” from “background” events, and some studies have reached opposite conclusions with similar or identical data sets that were binned differently (Hardebeck and Michael, 2004). In order to capture sufficient numbers of earthquakes, narrow, near-fault bins are often longer than desirable, and bins along the faults narrower than 2 km yield data sets with focal mechanisms that are too similar for rigorous inversion.

WEAK-FAULT AND FAULT-DAMAGE MODELS

Low-Friction Materials

Low-friction materials are typically sheet silicates such as chlorite, illite, smectite, talc, or serpentine (e.g., Byerlee, 1978). Such materials have been found along several low-angle normal faults (Numelin et al., 2007; Collettini et al., 2009; Carpenter et al., 2011; Lecomte et al., 2012; Haines and van der Pluijm, 2012), and the weakness is enhanced by fault-core foliations subparallel to the slip surfaces (e.g., Collettini et al., 2009). For typical continental LANFs such as the WDF or WSDF, alteration to clays (as opposed to serpentine or talc) is expected and this is thought to occur in the upper ∼5 km of the crust (e.g., Numelin et al., 2007), or at temperature <180 °C (Haines and van der Pluijm, 2012; yielding, for example, a maximum depth of 6 km for a geotherm of 30 °C/km). Thus, it appears that weak, foliated gouge is unlikely to aid reduction of LANF strength in the strong, seismogenic crust (∼5–12 km depth).

Elevated Pore-Fluid Pressure

High pore-fluid pressure (Pf) has long been understood to aid fault slip by reducing the effective normal stress on faults (Hubbert and Rubey, 1959). This mechanism may explain the lack of a heat-flow anomaly across the San Andreas fault (Lachenbruch and Sass, 1980; Rice, 1992), but others argue that an along-strike anomaly exists (Scholz, 2000) or that heat-flow measurements need not be a robust indicator of fault strength (e.g., Saffer et al., 2003). Lithostatic Pf is suggested along misoriented reverse faults at seismogenic depths (Sibson et al., 1988), in accretionary prisms (Fagereng et al., 2010), and was likely along the Zuccale LANF that contains many shallowly dipping extensional veins (Collettini and Holdsworth, 2004; Smith et al., 2008). Many LANFs in the western U.S. Cordillera experienced large-volume fluid flux through the fault zones (e.g., review in Axen, 1992; Person et al., 2007; many others), but the levels of paleo-Pf are generally uncertain.

Elevated paleo-Pf is difficult to prove in the absence of appropriately oriented extensional veins, which are lacking in the detachments studied here. In an Andersonian extensional stress field (with σ1 vertical), subvertical tensile cracks should form if Pf exceeds rock tensile strength, causing Pf to drop. Also, steeply dipping upper-plate normal faults are common above LANFs and provide additional Pf escape routes. Thus, in the absence of special stress conditions within the fault damage zone that prevent hydrofracture (e.g., Rice, 1992; Axen, 1992; Healy, 2009; Weak-Sandwich Models section), it seems unlikely that supra-lithostatic Pf levels can be maintained in most LANFs. Axen and Selverstone (1994; Strong-Sandwich Model section) calculated elevated Pf (up to 0.7 of lithostatic pressure at 10 km depth) in mechanical models of brittle LANF slip on the Whipple fault, conditions that would not lead to pure tensile failure. Cataclastic textures along both the Whipple and West Salton detachments are consistent with constrained comminution (Sammis et al., 1987) and sublithostatic Pf (Luther et al., 2013). Elevated Pf may be considered a viable mechanism for LANF slip, if no other model explains data better, but should not be viewed as a universal solution or favored a priori.

Dynamic Weakening and Earthquake-Related Damage

Dynamic weakening during seismogenic slip (Brodsky and Kanamori, 2001; Di Toro et al., 2004; Wibberley and Shimamoto, 2005; Rempel, 2006; Rice, 2006) can explain observations of low heat flow around strike-slip faults (e.g., Brune et al., 1993) that suggests they are weak (Lachenbruch and Sass, 1980). Dynamic weakening also may explain the inferred weakness of LANFs: seismicity on some has been inferred from historical earthquake studies (Abers, 1991, 2001; Wernicke, 1995; Abers et al., 1997; Axen, 1999), fault-scarp studies (Johnson and Loy, 1992; Caskey et al., 1996; Axen et al., 1999) and presence of frictional melt-rock (pseudotachylyte) (John, 1987; Prante et al., 2014). Evidence for creep on some LANFs indicates that dynamic weakening mechanisms do not apply to all (Collettini et al., 2009).

Pseudotachylyte exists along both the Whipple (Wang, 1997) and West Salton detachment faults (Axen and Fletcher, 1998; Axen et al., 1998; Prante et al., 2014), indicating that both slipped seismogenically at least in part. For this study, the question is: what off-fault damage may have resulted from earthquakes?

The earthquake cycle can be broken into four parts: interseismic, preseismic, coseismic, and postseismic. Low-level seismicity occurs around many faults during the interseismic period between major events and presumably reflects local failure in the regional stress field controlled by long-term tectonic loading. Stress inversions of interseismic events around the San Andreas fault are interpreted various ways by different authors and may show rotation from higher angles from the fault (60°–85°) to moderate angles (40°–55°) closer to the fault, but low angles (∼30°) are rare (see review by Hardebeck and Michael, 2004). Preseismic foreshocks are rare on all types of faults, and none are reported from LANFs; but this may reflect the very sparse LANF earthquake catalog. Coseismic off-fault damage may be due to both rupture propagation and fault roughness. Earthquake rupture propagation (Di Toro et al., 2005; Rice et al., 2005) should create a near-fault stress field and resulting structures that are similar in orientation to those caused by initial fault propagation (Fig. 5 and the Fault and Earthquake Rupture Propagation section; Vermilye and Scholz, 1998). It is possible that seismic slip rates on nonplanar faults may create damage that is similar to that predicted by static elastic models of wavy fault slip (Chester and Chester, 2000; Damage Due to Slip on Nonplanar Faults section ), but we are not aware of such models. Most large earthquakes are followed by a period of aftershock activity, much of which is off-fault so should be reflected in off-fault damage. Aftershocks may reflect short-term stress perturbations due to the mainshock (King et al., 1994) or local stress perturbations at asperities. Aftershocks around the San Andreas fault may reflect a stress field rotated to higher angles to the fault, relative to interseismic seismicity, due to shear stress reduced by the main shock (Townend and Zoback, 2001; Hardebeck and Michael, 2004). We emphasize that “on-fault” seismicity in seismic stress-inversion literature typically includes the entire damage zone and may be 2 km to >5 km from the fault in question and therefore includes both on-fault and off-fault fracturing as used here, and the spatial resolution in seismic inversions is orders of magnitude larger than in our study.

Regardless, the static coefficient of friction on LANFs cannot be ignored, whether or not dynamic weakening occurs during earthquakes on LANFs: static friction must be overcome in order to initiate an earthquake rupture. Also, dynamically lowered friction must increase to arrest the propagating rupture and must return to static values during the interseismic period. Thus, if LANFs have low static friction and are misoriented with respect to the stress fields around them remain key questions.

Stress Changes in Fault Damage Zones or Cores

Weak-Sandwich Models

A variety of models propose that weak faults may slip because of a local stress rotation within the fault damage zone (∼100 m) or fault core (1–10 m), caused by strength changes perpendicular to the fault. One such “weak-sandwich model” (Fig. 1) was proposed by Rice (1992) for the San Andreas fault and applied by Axen (1992) to LANF slip. The fault zone has higher Pf than the surrounding host rocks, which, in turn, causes mean stress increase within the fault zone and σ1 to rotate to a shallow angle (∼30°) adjacent to the fault. Increased mean stress in the fault zone prevents hydrofracture. In this local stress field, the LANF has a primary Reidel shear orientation but secondary (antithetic) Reidel shears, and a mineralized fault core might be expected.

Another weak-sandwich model relies on Coulomb plasticity due to granular flow in the fault core (approximately one to tens of meters away from the LANF; Fig. 2), with σ1 oriented ∼45° to the fault, allowing slip at a somewhat lower apparent friction value (Mandl and Fernandez Luque, 1970; Mandl et al., 1977; Byerlee and Savage, 1992; Lockner and Byerlee, 1993; Marone, 1995).

Faulkner et al. (2006; see also Heap and Faulkner, 2008) presented a weak-sandwich model with near-fault stress rotation similar to that of Rice (1992) in which rotation is caused by damage rather than Pf, with elastic parameters changing inward through the damage zone due to progressive tensile cracking. They characterized microfracture density as a function of distance from faults, performed laboratory experiments on similar lithologies, and characterized the evolution of Young’s modulus and Poisson’s ratio as functions of microcrack density. Applying those elastic properties as if they were isotropic in a model of fault-parallel layers that are increasingly damaged toward the fault, they showed that σ1 around misoriented faults should rotate to lower angles to the fault plane as it is approached.

If damage must exist to cause stress rotation, but damage evolution is controlled by near-fault stress perturbations, then this model predicts that zones closer to a misoriented fault should show overprinting cracks with younger ones reflecting stress fields progressively rotated to lower angles to the fault. (In their field area, in fact, the microcracks show little change of orientation, being ∼30° from the fault at all distances and not providing evidence of any progressive stress rotation within the microfractured zone [Faulkner, 2007, written commun.].) Faulkner et al. (2006) treated the damaged rocks as elastically isotropic, which is almost certainly not true because tensile microfractures have preferred orientations perpendicular to the least principal stress and will cause anisotropic elastic properties.

Healy (2008) addressed the anisotropy of the damaged rock and concluded that the opposite effect, σ1 rotated to higher angles to a misoriented fault, is expected if elevated Pf is involved and for reasonable damage patterns. Healy (2009) extended this analysis to low-angle normal faults, assuming that weak, foliated (anisotropic) materials (shale, clays, talc, and serpentinite) form the fault core and that pulses of elevated fluid pressure pass upward from the footwall, through the LANF and into the hanging wall. This mechanism may apply to LANFs showing evidence of anisotropic fault cores and/or elevated Pf, such as the seismogenic Woodlark detachment fault (Abers, 1991; Abers et al., 1997; Floyd et al., 2001; Roller et al., 2001) or the Zuccale fault (Collettini and Holdsworth, 2004; Smith et al., 2008) but does not seem applicable to the WDF or WSDF, except possibly in the uppermost few km of the crust (see The Whipple and West Salton Detachment Faults section).

Strong-Sandwich Model

Axen and Selverstone (1994) proposed a model for slip on the Whipple detachment fault in which the mineralized material of the fault core (chlorite-epidote breccia zone) is stronger than both the fault and the surrounding heavily fractured damage zone (Fig. 3). This model is based upon observations of conjugate faults in the chlorite-epidote breccia zone, surrounding damage zone, and deeper “background” levels and is consistent with σ1 in a subvertical (Andersonian) orientation. The conjugate faults measured have abnormally low inter-fault (conjugate) angles, with many showing transtensile failure (opening plus shear offset). Assuming σ1 bisected the conjugate angle, then it was oriented at high angles (55°–80°) from the WDF (consistent with our results, General Paleostress Results and Steeply Plunging Extensional σ1 and LANF slip sections). The transtensile failure and low conjugate angles are consistent with elevated (but sublithostatic) Pf in the mineralized zone, where hydrous alteration minerals epidote and chlorite form a large percentage of the rock (see Selverstone et al., 2012). Using a Griffith failure envelope and assuming a tensile strength of 10 MPa for the mineralized zone, Axen and Selverstone (1994) showed that, for slip to occur on the WDF, Pf needed only to be moderately elevated (∼0.7 lithostatic) at ∼10 km depth but could be hydrostatic at shallower depth.

Crustal-Scale Stress Rotation

If σ1 rotates from subvertical in the upper crust to a moderate plunge in the midcrust, then LANFs at depth may slip without violating standard fault mechanics or rock friction values. Yin (1989) formulated an elastic model with σ1 plunging 45° at the brittle-plastic transition (Fig. 4), due to assumed subhorizontal shear traction applied to the base of the brittle crust by subjacent plastic flow. Spencer and Chase (1989) argued for similar crustal-scale stress rotation resulting from lateral crustal thickness changes. Buck (1990) objected to the analysis of Yin (1989), showing that the LANFs at depth would have lower resolved shear traction than steeper conjugates, which he argued would form preferentially. However, the steep conjugate faults should have reverse slip in the deeper, stronger brittle crust (R2 shears in Fig. 4) and, in extending lithosphere, the LANF orientations probably are favored by boundary conditions, nonelastic rheology, and/or energetic considerations. Westaway (1999, 2001) proposed a similar, but very complex model in which shear traction is greater on the LANF orientations. In crustal-scale stress rotation models, master faults are initially listric and curve gradually with depth, from steep dips (∼60°) in the near surface to gentle dips at the base of the elastic layer (Fig. 4). Differential stress should be high at the base of the brittle crust, consistent with common strength-depth profiles based upon experimental rock mechanics (e.g., Brace and Kohlstedt, 1980).

Selverstone et al. (2012) provided support for this class of models in their study of the paleo–brittle-plastic transition exposed in the Whipple Mountains, using sites and samples from the upper mylonite zone, tens of meters below the detachment. They argued that embrittlement in any given place was rapid, permanent, and caused by fluid infiltration and precipitation of epidote, with σ1 oriented ∼45° from mylonitic C planes both before and after embrittlement. C planes are the dominant shear plane during mylonitic flow and are thought to be planes of maximum shear stress; the WDF is subparallel to structurally deeper mylonitic C planes. C′ shears that formed during mylonitization were followed by development of brittle primary Reidel shears in the same orientation and by conjugate secondary Reidel shears (Fig. 4); if interpreted as conjugate Coulomb fractures, these orientations also imply σ1 oriented ∼45° from the C planes and the evolving WDF in the midcrustal brittle-plastic transition. Embrittlement occurred at temperature of 380–420 °C, while Pf dropped from lithostatic to hydrostatic levels, also consistent with high differential stress in the midcrust (Selverstone et al., 2012) and suggesting that lithostatic Pf might not exist at higher crustal levels.

Fault and Earthquake Rupture Propagation

The fault-propagation model of Vermilye and Scholz (1998; Fig. 5) is based upon detailed studies of damage around small faults and considers the stress field in the process zone ahead of propagating fault tips, where damage accumulates due to stress concentrations caused by slip gradients. Background studies near those faults suggest that the far-field stress was oriented ∼30°–40° away from them (measured in planes perpendicular to the faults and that contain the slip vectors), consistent with experimental results. For propagation of mode II fractures (where the fault tip-line is perpendicular to the slip line), stress concentrations and damage in the process zone are asymmetrical about the fault tip (Fig. 5). For mode III propagation (at the fault tip-line that is parallel to slip), σ1 should be oriented at 45° to the fault plane (Vermilye and Scholz, 1998).

Earthquake rupture propagation should cause similar patterns of damage that will overprint damage due to initial fault propagation and that may dominate fractures measured in the field (e.g., Rice et al., 2005). Both the WDF and WSDF were arguably seismogenic (Dynamic Weakening and Earthquake-Related Damage section); so rupture propagation may dominate over fault propagation in the stress inversions, if those processes were significant in forming the fractures studied here.

In large normal-fault earthquakes (mostly with dips >30°), the hypocenter is generally near the base of the crustal seismogenic zone (e.g., Jackson, 1987; Jackson and White, 1989; Abers, 1991; Abers et al., 1997; Axen, 1999); therefore, rupture propagation should have an up-dip (mode II) component. Earthquake ruptures on large, steep normal-faults are typically several times longer than their down-dip length; so damage formed during along-strike mode III rupture propagation may dominate. LANFs should have a greater down-dip length (Wernicke, 1995); thus their rupture areas may be more equant. If LANF hypocenters are also near the base of the seismogenic zone, then LANF earthquake ruptures also should propagate both up dip and laterally, and mode III rupture may be less dominant than for steep normal faults. For mode II propagation up the dip, we expect that our upper footwall sites should reflect stress orientations in the compressive quadrants, with σ1 at low angles to the LANFs (c in Fig. 5). If along-strike mode III propagation dominated, then σ1 oriented at 45° to the LANF should be recorded.

Damage Due to Slip on Nonplanar Faults

Chester and Chester (2000) show with an analytical elastic model that the orientation and magnitude of σ1 around a sinusoidally wavy fault can change dramatically within one-half wavelength perpendicular to the fault. They showed near-fault stress orientations for two scenarios (Fig. 6): a frictionally strong fault (μ = 0.6) at a low angle (45°) to σ1 and a frictionally weak fault (μ = 0.25) at a high angle (70°) to σ1. Using Coulomb failure criteria for intact granite, they concluded that failure is most likely near the fault in releasing bends where fault-normal compression is reduced, the failure region will expand along and away from the fault for higher friction, and the maximum principal stress in failure regions will be at a high angle to the fault in either scenario, with relatively little variation of orientation (20°–40°) within the region of failure (Fig. 6). In restraining bends, where failure is expected only for high friction, σ1 is subparallel to the regional applied stress and has similar magnitude.

THE WHIPPLE AND WEST SALTON DETACHMENT FAULTS

The Whipple detachment fault (WDF; Fig. 7A) in southern California is a broadly domed, shallowly dipping normal fault that accommodated ∼40–50 km of top-to-NE extension from early to middle Miocene (Davis et al., 1986; Howard and John, 1987; Davis, 1988; Davis and Lister, 1988). The footwall originated at ∼16 ± 4 km depth, below the brittle-plastic transition (Anderson, 1988). Extension-related mylonitization began after ca. 26 Ma and had ended in the central Whipple Mountains by ca. 19 Ma (Davis and Lister, 1988), and tilted rocks of the hanging wall are unconformably overlapped by essentially undeformed ca. 12–14 Ma old basalts.

The footwall of the WDF is broadly arched about a northwest-southeast axis, presumably due to isostatic rebound driven by lateral removal of the upper plate (e.g., Spencer, 1984). The eastern footwall, where our sites are located, remains with a generally gentle northeast dip and is composed of mylonitic metamorphic and intrusive rocks of mainly Precambrian or Cretaceous age (e.g., Anderson and Cullers, 1990).

The WDF where we worked has a 5–15-m-thick breccia zone (up to 300 m have been reported, but this includes chlorite-epidote altered rocks of the fractured damage zone; Davis et al., 1986) that is capped by an ultracataclasite layer (“microbreccia ledge”) up to 2 m thick. Together, these comprise the fault core. The fault-core breccia, underlying fractured damage zone, and preexisting mylonites are all retrogressively altered by epidote and chlorite. Clasts of mylonite are found in the breccia, and clasts of breccia are found in the ultracataclasite, indicating that the ultracataclasite is younger than the breccia, which is younger than mylonites. Mylonites and brittle fault rocks have structures indicative of top-to-NE transport (Davis et al., 1986). Pseudotachylite fault and injection veins are common (Wang, 1997; Hazelton, 2003), particularly along minidetachments: slip-surfaces below and subparallel to the main detachment fault that share many of its characteristics (Axen and Selverstone, 1994; Selverstone et al., 2012) and provide evidence for seismic slip on the WDF (McKenzie and Brune, 1972; Sibson, 1975; Spray, 1992).

The WDF and subjacent mylonitic foliation are folded broadly about gently NE-plunging axes (parallel to slip direction). Folding was synchronous with WDF slip (Davis, 1988; Davis and Lister, 1988; Yin and Dunn, 1992): in general, mylonitic foliation is more tightly folded and limbs dip more steeply than the detachment fault itself, suggesting that older mylonites record more fold-related, extension-perpendicular shortening than the younger detachment.

The top-east West Salton detachment fault (WSDF; Fig. 7B) in southern California bounds the western edge of the Salton trough, with the Peninsular Ranges in its footwall, and was part of the Pacific–North America plate boundary system, which is dominated by the dextral southern San Andreas fault farther east (Axen and Fletcher, 1998). The footwall of the WSDF is mostly composed of Cretaceous tonalite and granitoids that are overprinted by the Late Cretaceous Eastern Peninsular Ranges Mylonite Zone in northern exposures (Simpson, 1984; George and Dokka, 1994; Steely et al., 2009). The WSDF had ∼10–20 km of top-to-east normal slip from at least 5 to ca. 1 Ma (Lutz et al., 2006; Shirvell et al., 2009; Dorsey et al., 2012). The WSDF slipped seismogenically at least in part, as indicated by pseudotachylyte along it (Frost and Shafiqullah, 1989; Axen et al., 1998; Prante et al., 2014).

During the latest (final 200 k.y.) of slip on the WSDF, the San Jacinto, San Felipe, and Elsinore dextral fault zones (Fig. 7B) were also active, resulting in N-S shortening and gently E-plunging folds of the WSDF (Lutz et al., 2006; Janecke et al., 2011; Dorsey et al., 2012). However, some footwall arches, such as Whale Peak (Fig. 7B), are likely to be original corrugations (Steely et al., 2009). In general, the south limbs of these folds are controlled and/or modified by young faults of the Elsinore, San Felipe, and San Jacinto fault zones (Fig. 7B; Kairouz, 2005; Steely et al., 2009).

We chose these two faults for study because both: (1) are well characterized structurally and in terms of age, slip sense, and magnitude of slip; (2) have minimal subsequent tectonic overprinting; (3) have quartzo-feldspathic footwalls that represent typical continental midcrustal lithologies, reducing rheologic complexity; (4) probably did not experience lithostatic fluid pressure in the brittle crust; and (5) probably developed weak, foliated clay-rich gouges only in the upper crust, above the strong part of the seismogenic crust.

The stable Peninsular Ranges footwall of the WSDF suggests that it has had minimal structural modification by isostatic footwall uplift and the WSDF itself is nowhere west dipping (Axen and Fletcher, 1998; Kairouz, 2005; Steely et al., 2009) as would be expected if isostatic rebound had been significant. Additionally, the two footwalls were exhumed from different crustal levels. The WDF footwall resided at ∼16 ± 4 km depth (Anderson, 1988) prior to onset of extension and was processed through the brittle-plastic transition and the entire brittle crust by WDF slip. The WSDF footwall was at ∼5–10 km depth before extension (Shirvell et al., 2009) and exposes fault-related structures formed only in the brittle crust. Northern sites in the WSDF footwall contain Cretaceous reverse-sense mylonites of the Eastern Peninsular Ranges Mylonite Zone (Simpson, 1984; George and Dokka, 1994; Axen and Fletcher, 1998; Steely et al., 2009), but southern sites are in plutonic rocks that are unfoliated or display weak syn-to postplutonic foliation. Thus, southern WSDF footwall rocks can be treated as mechanically isotropic. We do not consider hanging-wall sites because of complex rotational and steep normal-fault histories that would unnecessarily complicate paleostress inversion.

Both detachments are well exposed locally (Fig. 8). The upper ∼6–50 cm (WSDF) or ∼0.5–2 m (WDF) of the footwalls are composed of ultracataclasites, which in turn are composed mainly of angular micron- to millimeter-scale fragments of footwall rocks, with some larger clasts, underlain by breccia. Below the WSDF, clasts are mostly unaltered from the Cretaceous or older protoliths, but the WDF cataclasites include chlorite- and epidote-rich fragments of altered detachment-related mylonites. The ultracataclasites of both faults contain clasts of cataclasite, ultracataclasite, and pseudotachylyte, indicating reprocessing of previously comminuted rocks. These “microbreccia ledges” are capped by sharp principal slip surfaces. The ultracataclasite layer below the WSDF commonly contains parallel, internal slip surfaces as well.

At most of our sites, the WSDF and WDF juxtapose sedimentary or volcanic rocks over granitoid or gneiss basement rocks (Figs. 8A and 8B), but both detachments also have areas where the upper plates contain quartzo-feldspathic crystalline rocks (Figs. 8C and 8D). Sedimentary-volcanic deposits in these areas are generally not more than a few km thick, with a maximum of ∼5.5 km in deeper parts of the WSDF upper-plate basins (Dibblee, 1954; Kerr and Beratan, 1991; Kidwell, 1991; Yin and Dunn, 1992; Nielson and Beratan, 1995; Dorsey and Roberts, 1996; Lutz et al., 2006; Dorsey et al., 2007, 2011, 2012; Janecke et al., 2011). Foliated and/or clay-rich gouge at our sites appears to lie above the principal slip surfaces (Figs. 8B and 8D) and to have been derived from upper-plate strata or basement from shallow depths below the basins. Thus, we conclude that low-friction phyllosilicates were unimportant in weakening these LANFs except at shallow crustal levels.

Chlorite and related sheet silicates are found along the West Salton detachment several kilometers north of our study area (Haines and van der Pluijm, 2012). There, the fault juxtaposes basement lithologies, and gouges are dominated by smectite derived mainly from chlorite. Haines and van der Pluijm (2012) also found clay minerals in gouges from high-angle normal faults in the base of the upper plate of the Whipple detachment, but they did not report weak materials from the detachment itself. In both places, it is likely that the gouges formed at shallow depths, and almost certainly at temperature <180 °C (Haines and van der Pluijm, 2012), or at <6 km depth (Low-Friction Models section). Thus, it seems unlikely that these gouges formed in the deep seismogenic brittle crust (∼10–12 km typical of rifts). For these reasons, we doubt that the materials along the Whipple and West Salton detachments were inherently weak as the footwall was carried up through the seismogenic crust.

PALEOSTRESS INVERSION METHODS

We collected data from 17 transects (from six areas) in the footwall of the WSDF and eight transects along the WDF. Transects have good three-dimensional exposure, are distant from postdetachment structures, are located at various positions on the transport-parallel folds, and have homogenous rock types. We mapped (on outcrop photographs; Fig. 9), described, and measured >5000 shear and tensile fractures in one to nine outcrops from each transect (∼65 total) and calculated the structural depths below and perpendicular to the detachments. We observed and measured fractures in outcrops with different aspects in order to reduce bias and measured fractures in a few lateral transects to check for local variability. We also measured ∼2000 striations and noted the slip or separation magnitude and sense when possible. We documented crosscutting relationships and fracture-fill information when possible. Fracture lengths range from 5 to 400 cm (most are ∼30–100 cm long). Slip magnitude on the fractures ranges from 1 to 100 cm (most ∼1–5 cm). Outcrops are generally 1–3 m high and 1–5 m wide.

We separated the database on location, depth, and relative age for inversion analysis in T-Tecto 3.0 (Žalohar and Vrabec, 2008; Žalohar, 2009). Data from different transects are not combined for paleostress analyses in order to test for spatial changes in paleostress orientation. Initially, each outcrop data set was inverted separately. We define three structural units: (1) the fault core, which is composed of ultracataclasite (or microbreccia) and underlying cataclasite and breccia; (2) the fractured damage zone, where fracture density is higher than background levels; and (3) the background where fracture density is not correlated with distance from the fault. Outcrop-level data sets from like structural units were often combined for several different reasons: (1) fewer than ten striated fractures were found; (2) the results of paleostress analyses from different outcrops were similar; and/or (3) the fracture orientations in different outcrops were similar. We also used crosscutting relationships within outcrops to determine if data needed to be separated. If one fracture set clearly formed before or after another, then they were separated; but such sets were rare.

We ran ∼400 paleostress inversions using T-Tecto 3.0 (Žalohar, 2009). Data input includes the orientation of the fault plane and striations and the slip sense (when known). T-Tecto output is a reduced stress tensor that includes the orientations of the three principal axes and a shape ratio: (σ2 – σ3)/(σ1 – σ3). We generally trust results from fracture sets with >25% known slip direction unless we had field data to support the slip directions generated by the program. Inversions find the best-fitting reduced stress tensor by minimizing the angular misfit (α, measured within the fault plane) between the predicted slip direction and the striation direction (assumed to reflect the actual slip direction) (e.g., Angelier, 1984, 1989; Žalohar and Vrabec, 2008) . In this study, we used α ≤ 30°. T-Tecto automatically separates heterogeneous databased on two criteria: (1) striae on fractures must be within 30° of that predicted by the model that best fits the largest number of data; and (2) the friction must be less than the ratio of shear stress to normal stress. Many input parameters in T-Tecto are adjustable, and we tested the stability of several of our models by varying them. The only parameter that significantly affected inversion results was the coefficient of static friction, μs. We modeled each data set twice at μs = 0.17 and μs = 0.7.

Paleostress inversion analysis is based on several assumptions. (1) The maximum resolved shear stress (τmax) on a plane is subparallel to the slip vector measured in the fault plane (Wallace, 1951; Bott, 1959). That this assumption is generally valid for our results is indicated by the generally small percentage of fractures not fit on the basis of α ≤ 30°. (2) All fractures being inverted formed at the same time in the same stress field. This assumption appears valid because fractures of different orientations commonly are mutually crosscutting. Also, fault-core breccia and ultracataclasite formed during brittle detachment slip (ca. 12 m.y. maximum for the WDF and ca. 6 m.y. for the WSDF); therefore, inverted fractures from the fault cores arguably formed over limited time spans and potential changes in the stress fields are probably predicted by the models considered in the Weak-Fault and Fault-Damage Models section. (3) Slip occurs on fractures of many orientations, not necessarily the optimum orientations as predicted by Coulomb failure criterion (Twiss and Unruh, 1998). This is supported by the fact that most sites yield well-dispersed fracture patterns. (4) Failure occurs in an isotropic rock body. Some sites have a preexisting Tertiary, detachment-related mylonitic fabric or older foliation. However, results from foliated versus non-foliated sites within a given transect and from transect to transect generally are similar; so this does not appear to be a problem. Brecciation and cataclasis within the fault cores have effectively destroyed any preexisting anisotropy in those sites.

T-Tecto does not include formal error analysis (Žalohar and Vrabec, 2008; Žalohar, 2009); thus, we evaluated accuracy and reproducibility of inversions the following ways (Luther, 2012). T-Tecto has six adjustable parameters (Žalohar, 2009), each of which was varied independently to assess impact on inversion stability. Inversions with fewer than ten fractures typically are poorly constrained and are not presented. Following Provost and Houston (2003), we randomly resampled five times 75% and 90% each of a large data set (68 fractures) and a moderate-size data set (33 fractures) and inverted them using both coefficients of friction to assess stability of inversions. The angular variance of σ1 orientation for the large data set ranged from 0° to 12.7°, and the variance of the medium-size data set ranged from 3.2° to 5.3°, with a combined average variance of 5.7° (Luther, 2012). Changing from high (0.7) to low (0.17) friction contributed ∼1° to 6° of variance, with more fractures being rejected as inconsistent with a single stress tensor for the high friction value. On this basis, we conservatively assign a maximum angular variance of 15° to our inversions, which is small compared to the ∼20°–30° ranges used to distinguish among different mechanical models (colored areas on stereonets, Figs. 10–12).

RESULTS

Footwall Damage

The West Salton detachment footwall typically contains a ∼0.5–1 m thick layer of ultracataclasite above 2–20 cm of breccia, above 2–20 m of fractured damage zone (altered and highly fractured tonalite, rare cataclasite), in turn overlying Cretaceous tonalite that locally has an older mylonite fabric. The detachment cuts the mylonitic foliation where it is present and typically dips ∼10°–30° shallower than the foliation (e.g., Steely et al., 2009); so its trajectory was not strongly controlled by foliation planes. Fracture density decreases exponentially away from the fault, with ∼50–100 fractures/m adjacent to the fault, dropping to ∼10–20 fractures/m by 5–10 m below the detachment. Fractures are dominantly shear fractures with striae that commonly rake >50°. Filled veins (typically gypsum or carbonate) are rare and cut the older fractures. Fracture orientations are variable depending on location along the fault, but damage zones generally are cut by two sets of mutually crosscutting shear fractures; one steep set and one shallow set that are subparallel to the detachment. Some transects have primary Reidel shears below the main fault.

The Whipple detachment fault (WDF) footwall typically consists of an ultracataclasite (∼1–2 m thick) and a thick cataclasite zone (∼5–15 m). Fracture densities near the fault core are ∼50–75 fractures/m and drop rapidly to ∼10 fractures/m by ∼5 m below the detachment. Most fractures cutting the WDF footwall are shear fractures. Similar to the WSDF, the fractures that cut the footwall appear to be mutually crosscutting, with many that are subparallel to the detachment and other steep fractures, commonly dominated by a steep set. A set of 1–5 m long, subvertical, unfilled tensile fractures with apertures of up to 2 cm are also common. These are probably late joints and are not considered further.

It does not appear that extremely elevated Pf played a role in slip on the WDF and WSDF. We found no evidence, such as common veins of any orientation, for supralithostatic Pf along the WDF and WSDF. This is consistent with breccia and ultracataclasite grain-size distributions that suggest constrained comminution, in which grains do not lift and roll over or slide past one another (Luther et al., 2013). Axen and Selverstone (1994) calculated hydrostatic to elevated (but sublithostatic) Pf on the basis of conjugate opening-shear fractures in the upper footwall of the WDF, and Selverstone et al. (2012) suggested, using fluid inclusions, that WDF Pf fell from lithostatic to approximately hydrostatic during rapid embrittlement at the brittle-plastic transition.

Paleostress Fields and Their Origins

General Paleostress Results

Paleostress inversion results are shown on lower-hemisphere, equal-area stereonets in Figures 10–12 as σ1 orientations, in which the σ1 vectors were rotated about the detachment strike such that the detachment is horizontal, allowing easy comparison among inversions from different sites and to the models considered. Most stress fields have a shape ratio ≤0.3 (68% from the WSDF and 86% from the WDF), indicating that σ1 was distinct from σ2 and σ3, which were subequal. For these low shape ratios, the positions of σ2 and σ3 may not be robust; thus they are not presented here.

In this section, we first describe the general inversion results in terms of orientations of footwall σ1 vectors (Fig. 10). We then focus on extensional stress fields, which are most common and yield appropriate shear sense on the detachments: σ1 vectors lie in or near a vertical girdle in the upper-plate transport direction and plunge in that direction. Several extensional vectors plunge steeply in the direction opposite of upper-plate transport, but these are mostly <10° from vertical and probably reflect inversion error. These results are compared to the mechanical models reviewed in the Weak-Fault and Fault-Damage Models section. Fewer fold-related stress fields were obtained and are discussed second: σ1 vectors plunging gently or moderately in girdles perpendicular to transport direction (e.g., Luther and Axen, 2013). The few extensional stress fields that imply the wrong sense of shear on the LANFs are discussed last.

Fifty-two WDF footwall σ1 vectors (Fig. 10A) define a fairly conical distribution with the densest cluster plunging nearly vertically to steeply northeast. Only a few plunge moderately to gently northeast. Several plunge moderately to gently northwest or southeast and suggest a weak girdle perpendicular to slip direction and fold hinges. Three plunge very gently either east or west. WSDF footwall σ1 vectors (Fig. 10B) cluster less tightly and form a weak, asymmetric girdle perpendicular to the transport direction and fold axes. The girdle and its asymmetry are largely defined by a group of <10 σ1 vectors that plunge gently south relative to the WSDF.

Steeply Plunging Extensional σ1 and LANF Slip

Most inversions from both faults yield σ1 orientations that are at high angles to the LANFs (∼70°–90°) and plunging toward the direction of slip. Steep σ1 orientations are consistent with Andersonian stress orientations in the brittle crust (Anderson, 1942), with the strong- sandwich model (Axen and Selverstone, 1994) and/or with weak materials that cannot support significant shear traction along the LANFs.

We favor the strong-sandwich model of Axen and Selverstone (1994; Fig. 3) to explain the steep σ1 orientations. An Andersonian extensional stress field combined with Coulomb fracture theory should produce initially steep normal faults, which are common from map to meter scales, consistent with steep σ1 (Axen and Selverstone, 1994; Strong-Sandwich Model section). As discussed in The Whipple and West Salton Detachment Faults section, weak materials are generally not present along these detachments, and foliated gouge that we have seen appears to have been derived mainly from supracrustal rocks (sedimentary and volcanic) and subjacent basement that were shallow at the onset of LANF slip. Furthermore, we have no evidence for lithostatic Pf, and most fractures in the upper footwalls are shear fractures. Under normal fault-mechanical theory that ignores cohesive and tensile strength of fault surroundings, slip at the low dips of these LANFs (or at high angles to σ1) would be difficult with normal friction and hydrostatic fluid pressure (e.g., Collettini and Sibson, 2001). The strong-sandwich model assumes that mineralization healed and strengthened the LANF surroundings; thus, it allows slip under hydrostatic Pf in the shallowest crust and requires only moderately elevated Pf at seismogenic depths (Axen and Selverstone, 1994), and the model is consistent with normal friction values and Andersonian stress orientation.

Crustal-scale stress-rotation models (Fig. 4; Crustal-Scale Stress Rotation section; e.g., Yin 1989; Westaway, 1999) also are consistent with Andersonian extensional stress fields in the upper brittle crust but do not predict formation of LANFs except near the brittle-plastic transition. Therefore, a mechanism, such as a rolling hinge (e.g., Spencer, 1984; Buck, 1988; Wernicke and Axen, 1988) or crustal-scale domino-style fault rotation (e.g., Davis, 1983), is required that allows LANFs to propagate into the upper crust if they have low initial dips only in the midcrust. If such a mechanism exists, then our results, particularly combined with those of Selverstone et al. (2012), who concluded that σ1 was 45° from the WDF in the midcrust, are consistent with crustal-scale stress rotation. The curvature and depth range of stress rotation is unknown but may not be as smooth as simple elastic model results suggest.

Healy (2009) argued that, if frictionally weak, elastically anisotropic, foliated fault gouge is present along faults, then σ1 may rotate to higher angles to the fault, contrary to many fault damage-zone models that assume mechanical isotropy. Healy (2009) envisioned Pf pulses moving from LANF footwalls, into the LANF core and then into the hanging wall, consistent with observations from LANFs that have weak, foliated materials in their core and evidence of high Pf (tensile veins). Thus his model may be applicable to the WDF and WSDF. However, we argue above that such materials were not present along these faults except at shallow crustal levels, and we do not have evidence for strongly elevated Pf (Axen and Selverstone, 1994; Selverstone et al., 2012; Luther et al., 2013). Thus, we do not favor the Healy (2008, 2009) model for these LANFs. In addition, the migrating Pf pulses envisioned by Healy (2009) predict that LANF slip events should predate slip on steep hanging-wall faults, whereas the sparse seismic record of LANF events suggests that the opposite occurs: slip on steep faults triggers LANF slip (Axen, 1999).

Gently to Moderately Plunging Extensional σ1 Results

Very few σ1 vectors plunge gently in the upper-plate transport direction (Fig. 10), as predicted by most weak-sandwich models (Weak-Sandwich Models section and Fig. 1; Rice, 1992; Axen, 1992; Faulkner et al., 2006). Thus, this class of models is not supported by our results, especially given that most extensional σ1 vectors make high angles to the LANFs.

Fault- and earthquake rupture–propagation models also predict low angles between σ1 and the LANFs, if mode II, up-dip propagation of either type was important (Vermilye and Scholz, 1998; Rice et al., 2005; Fig. 5). These models generally assume that regional σ1 was oriented at a low angle to the fault in question, raising concerns about their direct applicability to these LANFs, because our most common results suggest that assumption is invalid. In the absence of propagation modeling with regional σ1 at a high angle to the fault, we tentatively conclude that up-dip (mode II) fracture or rupture propagation was not important in formation of the damage zones.

A somewhat greater number of paleo-σ1 vectors plunge moderately in the transport direction (Fig. 10). Coulomb plasticity predicts σ1 at ∼45° to the fault plane (Fig. 2; e.g., Mandl et al., 1977; Byerlee and Savage, 1992). A few of our moderately inclined σ1 vectors are from fractures cutting the fault core (Fig. 11), where granular materials are present and Coulomb plasticity is permitted, but any granular flow must have ended by the time those fractures formed, and most σ1 vectors from the cores are steep. Most moderately plunging σ1 vectors are from fractures below the fault core (Fig. 11), in either the fractured damage zone or from background sites. These may reflect stress rotation in the damage zones, maintaining stress continuity with the fault cores, if granular flow occurred there before fractures crosscut the cores. Grain-size analysis of the fault cores supports constrained comminution, in which the cataclasites are sufficiently compact that grains cannot ride up and roll or slide over one another (Luther et al., 2013). Luther et al. (2013) interpreted this as indicating that Pf was too low to reduce the effective normal stress to levels allowing granular flow, which is consistent with a rapid drop of Pf to hydrostatic levels during rapid embrittlement in the brittle-plastic transition (Selverstone et al., 2012). However, a brief period(s) of granular flow and Coulomb plasticity cannot be ruled out entirely.

Two other explanations for fractures yielding moderately plunging σ1 vectors at sites outside the fault core also imply that those fractures may be relatively old. Along-strike fault or earthquake rupture propagation, in mode III (Fault and Earthquake Rupture Propagation section), could cause σ1 to be oriented ∼45° to the faults (e.g., Vermilye and Scholz, 1998). If due to along-strike LANF propagation, then the fractures inverted would be some of the most early-formed ones. If, however, the fractures formed due to earthquake rupture propagation, then they could be any age (within the period of fault activity). Both are difficult to reconcile with the very few σ1 vectors that plunge gently as expected from up-dip fault or rupture propagation.

Alternatively, Selverstone et al. (2012) argued that σ1 was oriented ∼45° from WDF mylonitic C planes during late mylonitic shearing in the crustal brittle-plastic transition, and that this orientation was maintained in the midcrust during embrittlement. It is not clear that this explanation is suitable for the WSDF because its footwall was at ∼5–10 km depth at the onset of extension, so probably at least a few kilometers above the brittle-plastic transition. Fracture sets that formed in this setting also would be early, and this stress orientation during embrittlement would have been consistent with Coulomb plasticity, if it occurred.

Slip on Nonplanar LANFs

The model of Chester and Chester (2000; Fig. 6) for damage arising from slip on a wavy fault predicts high angles between σ1 and the LANFs in the areas where failure is most likely; so the model is consistent with our results but difficult to evaluate completely with our data. In particular, the waviness (parallel to transport direction) of the LANFs in our study is not known, but we suspect it is small. Small-offset (<1 km) faults become less rough with increasing slip (Sagy et al., 2007), but how to extrapolate such results to faults with tens of kilometers of slip is unclear. At the outcrop scale, the WDF and WSDF appear to be very smooth and orientation differences over tens of meters are well below measurement accuracy of a hand-held compass. The generally small variation of ultracataclasite thickness over many kilometers is consistent with low roughness because significant asperities might be expected to disrupt or remove the ultracataclasite layer as slip juxtaposes asperities. The WDF is broadly arched about a northwest-southeast axis, probably due to isostatic rebound of the footwall, but our study sites are located on the northeast side of this arch in order to avoid fractures related to footwall uplift (compare to Axen et al., 1995; Selverstone et al., 1995; Wawrzyniec et al., 2001). Last, if the fractures used in our inversions formed due to transport-parallel waviness of the LANFs, then the fact that we obtain broadly similar stress tensors at sites separated by ∼3–10 km in the direction of transport (Fig. 7) suggests that the half wavelength is greater than that.

Chester and Chester (2000) did not present results for the combination of a strong fault (high friction) at a high angle to regional σ1; so the predicted variation in stress-field orientation is unknown for the case we favor (Steeply Plunging Extensional σ1 and LANF slip section). However, their model’s mechanical basis and results are consistent with our conclusion of strong LANFs at a high angle to the regional σ1. The variation of that angle decreases for increasing friction and, for a fault at 70° to regional σ1 with friction ≥0.4, is largest directly adjacent to the fault, with angular variation of 40° along the fault over one wavelength (their fig. 4). The variation would be smaller within the region of anticipated failure. Our inversions show comparable or closer clustering of σ1 orientations in the slip-parallel directions (Fig. 10). In addition, regions of failure predicted in their models involve tensile σ3, a stress state that is nearly identical to that inferred by Axen and Selverstone (1994; compare fig. 7b of Chester and Chester, 2000 to our Fig. 3B).

Folding-Related Stress Fields

Some paleo-σ1 vectors lie in girdles perpendicular to the transport direction of the WDF and WSDF. Both faults are folded about axes plunging parallel to transport, and these vectors are interpreted as reflecting fold-related stress fields (e.g., Luther and Axen, 2013). Dispersion perpendicular to fold axes is seen in inversions from the core and damage zone of both faults (Fig. 11) but is greater for the WSDF than for the WDF. “Background” sites from the WSDF footwall show little fold-related dispersion.

As discussed above, the WDF itself is less tightly folded than extension-related mylonites indicating a three-dimensional general strain field in which folding and normal slip on the WDF were synchronous. Of seven sites on the WDF (Fig. 12), four are near an antiformal crest (yielding ten σ1 vectors), and three sites (one each) are in a synformal keel (ten σ1 vectors), a north-facing limb (23 σ1 vectors) and a south-facing limb (eight σ1 vectors). Orientation of σ1 is most dispersed perpendicular to the fold axis in the antiformal crest sites, with both steep σ1 and steep to moderate northwest plunges relative to the WDF. Only about five moderately northwest-plunging σ1 vectors from antiformal crest sites are far enough from the orientations predicted by LANF-slip models to be clearly unrelated to northeast-directed slip on the WDF. Why only northwest plunges are obtained is unclear. Orientations of σ1 vectors in the synformal trough are mostly steep and fit generally with an extensional stress field, except two σ1 vectors that plunge gently northwest or southeast. These two may record fold-related horizontal shortening. Orientations of σ1 on northwest- and southeast-facing limbs range from perpendicular to the WDF to steeply northwest or southeast plunging, respectively, implying small components of top-northwest traction on the northwest-dipping limb and top-southeast traction on the southeast-dipping limb. This implies a component of upward motion of the core of the fold relative to the limbs but may simply reflect rotation of generally subvertical (Andersonian) σ1 during rotation of the detachment to horizontal for plotting on the stereonets. Moderately to gently northeast-plunging extensional σ1 vectors were obtained only from northwest-dipping fold limbs, but the reason for this is not clear.

Folds of the WSDF plunge gently east, parallel to upper-plate transport, and are open. Unlike the WDF, the south-dipping limbs are modified or controlled by younger, mostly dextral or dextral-oblique faults such as the San Felipe fault (Fig. 7B; Steely et al., 2009) and faults on the south side of Whale Peak (too small to show in Fig. 7B; e.g., Kairouz, 2005). We did not use sites from south-dipping limbs due to this complication. Dextral strike slip on these young strands of the San Andreas fault system overlapped in time with latest (final ∼200 k.y.) WSDF slip (Lutz et al., 2006; Janecke et al., 2011; Dorsey et al., 2012), but folding may have begun earlier because WSDF slip was concurrent with southern San Andreas fault slip (Axen and Fletcher, 1998), implying a regional stress field with σ1 north-south.

Orientations of σ1 vectors were determined from five sites along the WSDF (Fig 7B), with one site on an antiformal crest (ten σ1 orientations), one site in a synformal trough (18 σ1 orientations), and three sites on north-dipping limbs (50 σ1 orientations) (Fig.12). There is only moderate dispersion perpendicular to WSDF fold axes of σ1 from antiformal crest and synformal trough sites (Fig. 12) except one gently southeast-plunging outlier from the synclinal trough sites.

Several WSDF σ1 vectors from north-dipping fold limbs plunge gently south or moderately north (Fig. 12). Luther and Axen (2013) concluded that the gently south-plunging σ1 vectors reflect thrusting stress fields related to flexural slip (top-to-south) on the detachment during folding, consistent with many north- to north-northeast–plunging striations on the WSDF on the north limbs of folds (Axen and Fletcher, 1998; Steely et al., 2009). Luther and Axen (2013) argue that this thrusting stress field reflects north-south shortening within the regional strike-slip stress-strain field related to the dextral San Andreas fault system (e.g., Johnson et al., 1994; Savage et al., 1994), which overlapped in time with latest (final ∼200 k.y.) WSDF slip (Lutz et al., 2006; Dorsey et al., 2012). Thrusting apparently alternated with WSDF slip and extensional stress fields due to the earthquake cycle because many sites cannot be fit by a single stress tensor (Luther and Axen, 2013). Most σ1 vectors that plunge moderately north relative to the WSDF lie on north-facing fold limbs (Fig. 12) and were rotated ∼30° (the approximate dip of the WSDF there) to make Figure 12; so these vectors may simply reflect fracturing in extensional stress fields after folding and north tilting of the WSDF.

Extensional Stress Fields with Incorrect Shear Sense

A few extensional stress field σ1 orientations are located in the wrong quadrant for the known sense of slip on the LANFs (Fig. 10), implying top-to-west shear traction on the WSDF (six of 78) or top-to-southwest shear on the WDF (seven of 52). About half of these lie within 15° of the vertical plane that separates appropriate shear senses (top-east or top-northeast) from incorrect ones; thus, these orientations are admissible based upon our error analysis (Paleostress Inversion Methods section; Luther 2012). The remaining few presumably record localized complexities in the footwall stress or strain fields.

CONCLUSIONS

Inversions for reduced stress tensors were performed on fracture orientation data sets in the upper footwalls of the Whipple and West Salton detachment faults. Most inversions resulted in extensional stress fields with σ1 oriented at a high angle to the LANFs and with a steep plunge, compatible with Andersonian stress fields. Much smaller numbers of inversions yielded moderately plunging extensional σ1, and very few yielded gently plunging extensional σ1 vectors. In addition, some fracture sets appear to record stress related to extension-perpendicular folding.

Extensional stress inversions were compared to mechanical models for slip on weak faults and support best a strong-sandwich model of the type proposed by Axen and Selverstone (1994) for detachment slip in the brittle crust. In that model, LANFs can slip with normal friction values (∼0.6) under hydrostatic to only moderately elevated Pf and mineralization keeps the materials surrounding the LANF relatively strong.

Moderately plunging σ1 vectors from the WDF may reflect older fracture sets formed close to the brittle-plastic transition where elastic modeling (e.g., Yin, 1989) and field and lab studies (Selverstone et al., 2012) suggest σ1 was moderately plunging. Combined, these results support rotation of the stress across the thickness of the brittle crust, but the depth range and curvature of this rotation are not well constrained. Application of this concept to the WSDF is uncertain because its footwall originated at <10 km depth at onset of extension, above the brittle-plastic transition, but the exact depth is poorly constrained.

Coulomb plasticity (Mandl et al., 1977; Byerlee and Savage, 1992; Lockner and Byerlee, 1993; Marone, 1995), which predicts σ1 oriented at moderate angle (∼45°) within the cataclastic fault core, is only weakly supported because few such results were obtained from the fault core. More results were obtained from the fractured damage zone where granular flow did not occur but may reflect stress continuity between the cores and fractured damage zones during brief episode(s) of Coulomb plasticity in the core. Rapid Pf drop during embrittlement argues against this, however (Gently to Moderately Plunging Extensional σ1 Results section).

Weak-sandwich mechanical models that predict significant rotation of σ1 to low angles (∼30°) to the fault within the fault damage zone or core (Axen, 1992; Rice, 1992; Faulkner et al., 2006) are not supported.

Our results are compatible with the model by Chester and Chester (2000) of off-fault failure due to slip on wavy faults, but the amplitude and wavelength of roughness on these LANFs is not known; so this compatibility is not compelling. It is intriguing, and perhaps significant, that their model assumes a stress field near the fault that is nearly identical to that proposed by Axen and Selverstone (1994) based upon field measurements and mechanical reasoning.

The paucity of σ1 vectors inclined at low angles to the LANFs is not consistent with fractures forming due to up-dip (mode II) fault propagation (Vermilye and Scholz, 1998) or earthquake-rupture propagation (Rice et al., 2005). Along-strike (mode III) propagation may have controlled moderately plunging σ1 vectors, but this is difficult to reconcile with the sparse gently plunging results.

This research was supported by U.S. National Science Foundation grants EAR-0809638 (G.J.A.) and EAR-0809220 (J.S.) and salary from New Mexico Institute of Mining and Technology (G.J.A. and A.L.) and University of New Mexico (J.S.). Work on the WSDF was greatly aided by the staffs of Anza Borrego Desert State Park (ABDSP), especially George Jefferson, and Agua Caliente County Park (sampling and camping permits and access to the ABDSP library and research lab). Reviews by C. Scholz and C. Wibberley of an early version were very helpful, and we thank J. Spencer for a recent review. N. Khalsa and A. Mattox assisted in the field and with sample preparation.