Oceanic Detachment Faults Generate Compression in Extension

In extensional geologic systems such as mid-ocean ridges, deformation is typically 17 accommodated by slip on normal faults, where material is pulled apart under tension and 18 stress is released by rupture during earthquakes and magmatic accretion. However, at 19 slowly spreading mid-ocean ridges where the tectonic plates move apart at rates less than 20 80 km Ma -1 , these normal faults may roll over to form long-lived, low-angled 21 detachments that exhume mantle rocks and form corrugated domes on the seabed. Here 22 we present the results of a local microearthquake study over an active detachment at 23 13°20’N on the Mid-Atlantic Ridge to show that these features can give rise to reverse 24 faulting earthquakes in response to plate bending. During a six-month survey period we 25 observed a remarkably high rate of seismic activity, with more than 244,000 events 26 detected along 25 km of the ridge axis, to depths of ~10 km below seafloor. Surprisingly, 27 the majority of these were reverse faulting events. Restricted to depths of 3 - 6 km below 28 seafloor, these reverse events delineate a band of intense compressional seismicity 29 located adjacent to a zone of deeper extensional events. This deformation pattern is 30 consistent with flexural models of plate bending during lithospheric accretion. Our results

we present the results of a local microearthquake study over an active detachment at 23 13°20'N on the Mid-Atlantic Ridge to show that these features can give rise to reverse 24 faulting earthquakes in response to plate bending. During a six-month survey period we 25 observed a remarkably high rate of seismic activity, with more than 244,000 events 26 detected along 25 km of the ridge axis, to depths of ~10 km below seafloor. Surprisingly, 27 the majority of these were reverse faulting events. Restricted to depths of 3 -6 km below 28 seafloor, these reverse events delineate a band of intense compressional seismicity 29 located adjacent to a zone of deeper extensional events. This deformation pattern is 30 consistent with flexural models of plate bending during lithospheric accretion. Our results 31 indicate that the lower portion of the detachment footwall experiences compressive 32 stresses and deforms internally as the fault rolls over to low angles before emerging at the 33 seafloor. These compressive stresses trigger reverse faulting even though the detachment 34 itself is an extensional system. 35

INTRODUCTION 37 38
Oceanic lithosphere is formed at mid-ocean ridges by a combination of magmatism and 39 normal faulting, driven by far-field forces arising from processes including plate 40 subduction and mantle convection (Lachenbruch, 1976). In these extensional settings, a 41 portion of the strain is expected to be accommodated by slip on normal faults, which is 42 reflected in the focal mechanisms of earthquakes observed near the spreading axis (Sykes, 43 1967). At slow-spreading ridges, accounting for large parts of the lithosphere formed in 44 the Atlantic, Indian and Arctic Oceans, young lithosphere may be deformed by slip along 45 large-offset normal faults called detachments (Cann et al., 1997;Tucholke et al., 1998;46 Dick et al., 2003;Escartin et al., 2008b). Detachment faults can exhume lower crustal 47 gabbros and serpentinized mantle peridotites at the seabed and form kilometer-scale 48 dome-shaped features called oceanic core complexes (Cann et al., 1997;MacLeod et al., 49 2002;Escartin et al., 2003;Grimes et al., 2008). The mechanical behavior of detachment 50 faults is controversial because the domed fault surfaces emerge from the seafloor at low 51 angles that are that are incompatible with the physics of extensional faulting (Buck et al., 52 2005). There is evidence for fault initiation on a steeply dipping, deeply penetrating 53 rupture surface Morris et al., 2009;MacLeod et al., 2011), but the 54 mechanism by which the fault rolls over to low angles prior to seafloor exhumation is 55 poorly understood. Local earthquake surveys with ocean bottom seismographs (OBSs) 56 have the potential to address this issue; however, previous OBS deployments at oceanic 57 detachments had insufficient aperture and instrument density to resolve the mechanics of 58 fault rollover (deMartin et al., 2007;Collins et al., 2012;Grevemeyer et al., 2013). 59 Intriguingly, a few reverse faulting events were observed beneath the Logatchev core 60 complex on the Mid-Atlantic Ridge (MAR) at 14°40'N, but the relationship of these 61 events to the extensional fault system was unclear, and they were attributed to volume 62

MICROEARTHQUAKE EXPERIMENT 66
In 2014 we conducted the largest microearthquake experiment to-date at a slow-68 spreading ridge. A dense network of 25 short-period OBSs (instrument spacing of 2-3 69 km) was deployed for a six-month period along ~10 km of the ridge axis at 13°N on the 70 MAR ( Fig. 1). Detachment faults are common in this region, including two well-71 surveyed and sampled oceanic core complexes located at 13°20'N and 13°30'N (Smith et 72 al., 2006;MacLeod et al., 2009;Mallows and Searle, 2012;Escartin et al., 2017). Both 73 core complexes have well-developed, domed, corrugated surfaces and are accompanied 74 by a high level of hydroacoustically-recorded seismicity, suggesting that they are 75 currently active or have been in the recent geological past (Smith et al., 2008;MacLeod 76 et al., 2009;Mallows and Searle, 2012). 77 We recorded over 244,000 events on more than three stations during the 198 day 78 deployment, yielding a mean rate of ~1240 microearthquakes per day (see Methods), two 79 orders of magnitude greater than that observed at the Logatchev core complex 80 (Grevemeyer et al., 2013). This remarkably high rate of seismicity was fairly constant 81 throughout the deployment period (Fig. 2b). There was no evidence for foreshock-main 82 shock sequences, except for a small seismic swarm in the western band of events at Julian 83 day 280 within a region extending 3 km south from its northern tip. The locations and 84 focal mechanisms of these events are indistinguishable from the rest of the seismicity in 85 this area. Events have small local magnitude (ML), ranging between -1.0 and 2.7 and with 86 a modal average of 0.3 ( Supplementary Fig. 1). The high number of earthquakes, 87 combined with the dense seismic network, allowed us to estimate hypocenters and focal 88 mechanism solutions for a subset of 35,262 well-characterized events (see Appendix). 89 These reveal that reverse faulting was the most common mode of deformation near the 90 13°20'N detachment during our deployment (Fig. 1). The compressional events define a 91 distinct arc of intense seismicity that wraps around the detachment trace (on the eastern 92 edge of the corrugated surface), at depths of 3-6 km beneath the seabed ( Fig. 2 and 93 horizontal but there is no preferred orientation for the dip and strike of the fault planes 96 (Fig. 2c). Events within the reverse faulting band of seismicity have slightly smaller 97 magnitudes than those in the normal faulting band ( Supplementary Fig. 1). In contrast, 98 normal faulting is restricted to a narrow band of seismicity ~3 km east of the reverse 99 faulting zone, at depths of 5-12 km beneath the seabed ( Fig. 2 and Supplementary Fig. 2). Our observations indicate that lithospheric extension at the 13°20'N detachment generates 110 both compressional and extensional seismicity contemporaneously. The band of intense 111 reverse faulting at 3-6 km depth is located directly beneath the hanging wall cutoff, 112 where the gently-dipping corrugated surface emerges on the seafloor (Fig. 2e), hence 113 cannot lie on the detachment fault plane itself. Instead, this reverse faulting must be 114 occurring within the detachment footwall. An active high-temperature vent field is 115 located on the 1320 corrugated surface (Escartin et al., 2017), which could indicate 116 footwall emplacement of magma bodies (Fig. 1); however, the vent site is located 2.3 km 117 west from the band of reverse faulting (Fig. 2e), and cooling of a magma body should 118 generate tensile, rather than compressive, stresses. Thermal contraction associated with 119 heat extraction from a footwall magma body is therefore not a plausible source 120 mechanism for the shallow band of compressive seismicity. Our observations instead 121 support a model in which internal deformation of the lithosphere in response to flexural 122 bending stress results in a high level of seismicity at the point of maximum bending 123 (Lavier et al., 1999;Buck et al., 2005). The variability in the strike and dip of fault plane deeper, internal portion of the detachment footwall (Fig. 2c). In contrast, towards the 126 center of the axial valley and at greater depth (6-10 km), steep, ridge-parallel normal 127 faulting accommodates extensional deformation on the active detachment as new material 128 accretes into the footwall. The short distance between bands of reverse and normal 129 faulting (~2 km perpendicular to the fault plane) requires a rapid change in the footwall 130 stress field, from extensional stresses in the accretion zone to compressive stresses in the 131 region of fault rollover. This observation, combined with the spatially restricted zone of 132 reverse faulting, indicates that fault rollover may be a relatively abrupt, rather than 133 gradual, process, with tightening of curvature at progressively shallower sub-surface 134

depths. 135
We have developed a simple model based upon the deflection of a bending plate 136 with elastic-plastic rheology to reconcile our observations (see Supplementary Materials;137 McAdoo et al., 1978). The model is constrained by the location and dip of the corrugated 138 fault surface at the seafloor, the spatial distribution and focal mechanisms of observed 139 earthquakes, and a lithospheric slab thickness of 6 km inferred from the depth distribution 140 of seismicity. We use the distribution of earthquakes in the footwall to define a stress 141 profile, with 'plastic' failure at depths where seismic events are observed (in elastic-142 plastic models, deformation from earthquakes is treated as bulk 'plastic' yielding), and 143 assume that the initiating fault is likely to have a maximum dip of ~70°. We seek a 144 bending profile that satisfies these constraints, by varying the mechanical strength of the 145 plate in terms of its flexural rigidity, or effective elastic thickness (Te). We find that a 146 best fit is obtained if Te increases linearly from 0.7 km near the spreading axis, to 0.9 km 147 at the point where the footwall emerges at the seafloor (dashed line, Fig. 2e). This range 148 in Te, which is a modeling parameter rather than a physical property of the lithosphere, is 149 consistent with previous estimates from bathymetric profiles of detachment faulted 150 terrain (Schouten et al., 2010). Our simple model demonstrates that the location of the 151 reverse faulting is consistent with that predicted by bending of the detachment footwall 152 under a mechanically reasonable deflection profile (Fig. 3). 153 The two bands of seismicity show well-defined along-axis extents, the northern 154 ends of which lie within the OBS network and are therefore well resolved. The extent of 155 normal faulting extends ~3 km further north to 13°23'N, beyond which the seismicity 158 rate is remarkably low. These results demonstrate that the nature of seismically 159 accommodated deformation changes significantly at the northern limit of the 1320 core 160 complex, but the inability of our network to provide focal mechanism solutions in this 161 area makes it difficult to interpret this change in the context of fault structure and 162 deformation. A vigorous swarm of seismicity occurred over a 2-3 day period at 13°27'N, 163 just south of the 1330 core complex, which is suggestive of magmatic activity; however, 164 this interpretation is necessarily tentative because we cannot obtain focal mechanism 165 estimates from this area. 166 The apparent lack of seismicity on the upper surface of the detachment footwall at 167 shallow crustal depths is enigmatic. Extensional bending stresses are clearly high in this 168 region, and there must be slip between the footwall and hanging wall on the fault surface. 169 Rock samples recovered from the 13°20'N detachment fault scarp are dominated by 170 hydrothermal quartz-cemented basalt breccia, in addition to sheared serpentinites, talc 171 schists, incohesive cataclasites and hydrothermal deposits 172 Escartin et al., 2014172 Escartin et al., , 2017. This assemblage provides evidence for significant 173 hydrothermal alteration and mineralization in the fault zone, which may modify the 174 rheology of these rocks and preclude the generation of detectable seismicity (Reinen et al., 175 1992;Escartin et al., 2008a). Alternatively, we cannot rule out the possibility that 176 deformation in this zone occurs episodically over time intervals that are long compared to 177 the duration of our observations. Second, as the footwall rotates to lower angles, bending stresses lead to internal 188 compression in the lower half of the plate. As a result, reverse faults initiate within the 189 bending lithosphere at depths of 3 to 6 km below where the footwall emerges at the 190 seafloor to form a domed, corrugated fault surface (Fig. 3). This evolution of footwall 191 stress is consistent with kinematic models for detachment fault behavior (Buck, 1988) 192 and with direct observations for reverse faulting in detachment fault footwalls (Pressling 193 et al., 2012), suggesting that reverse faulting may be ubiquitous in mature, active oceanic 194 detachments. Our results provide a new framework for interpreting detachment seismicity, 195 and suggest that reverse faulting events reported at other core complexes may have been 196 triggered by bending stresses rather than volume expansion (e.g., serpentinization). The 197 mechanical regime we describe shows that plate bending associated with the exhumation 198 and formation of oceanic core complexes can generate compressional stresses leading to 199 reverse faulting, despite being situated in an extensional stress regime. 200

FIGURE CAPTIONS
Figure 1. Bathymetric map with seismicity and focal mechanisms at 13°20'N on the MAR. Inset shows location of study site (red box) and mid-ocean ridges (black lines). Main panel shows seismicity rate calculated in 100 x 100 m bins for 18,313 well constrained, relocated events detected by >9 instruments. Randomly selected first-motion focal mechanism solutions are plotted in lower-hemisphere projection; red line shows neovolcanic zone (NVZ); pink triangles show OBS positions; white triangle is Irinovskoe vent field. Location of along-axis adjacent corrugated oceanic core complexes shown by 1320 and 1330 labels, cross shows average 68% confidence level horizontal location uncertainty (0.9 km).

Figure 2.
Seismicity rate and cross-sections. a: Shaded-relief bathymetry (illuminated from NE) with cumulative seismic moment release in dyn cm -1 ; red/blue polygons delineate domains shown in b and c; black lines are transects shown in d and e; white triangle is vent field; red line is NVZ. b: Seismicity time series for domains 1 (blue) and 2 (red). c: Stereonets with P (black) and T (gray) axes for events in domains 1 and 2; gray shading is best-fitting fault plane solution for domain 2 (352° strike and 72° dip east). d and e: Cross-sections with hypocenters colored by domain as in (a) and representative focal mechanisms (cross-sections through lower hemisphere projection). Black solid line is seabed; thickened sections indicate corrugated fault scarp exposure; dashed line is calculated plate deflection from elastic-plastic model, applicable to spreading parallel profile in (e) only; arrows show location of hanging wall cutoff (HWC) and nearest along-strike projection of the neovolcanic zone (NVZ).    Click here to download Figure  ParnellTurner_Geology_Fig3_resub.pdf NonLinLoc software (Podvin and Lecomte, 1991;Lomax et al., 2000). Initial earthquake locations were determined using the grid-search algorithm (Tarantola and Valette, 1982;Lomax et al., 2000) for 183,762 events detected by more than four OBSs. After applying stations corrections, double-difference hypocenter relocation was carried out for 35,262 well-constrained events detected on more than nine OBSs with rms residual < 0.15 s, using differential travel times from the catalog and the program hypoDD (Waldhauser and Ellsworth, 2000), yielding 18,313 double-difference relocated hypocenters. Bestfitting first-motion focal mechanism solutions for the subset of 35,262 relocated events were obtained using HASH software (Hardebeck and Shearer, 2002). The seismic moment for each event was calculated using the long-period spectral level of vertical displacement spectra, and then converted into a local magnitude estimate.

ELASTIC-PLASTIC MODEL
The model for elastic-plastic bending allows us to calculate synthetic profiles for the detachment footwall surface. The deflection of a bending plate is defined in terms of the bending moment, M(x), which varies along the length of a bending profile, and the inplane force, T, which is the horizontal force applied to the end of the plate, and is constant along a profile (Fig. 3). Far-field forces give rise to the in-plane force, which is applied from outside the bending region (e.g. ridge push). The rheological parameters are expressed in terms of the depths and horizontal normal stresses, σ xx (z) at the top and base of the elastic core (z 1 and z 2 respectively). Mathematical details of the model are described elsewhere . We require the deflected surface to dip at 20° at the point of emergence at the seabed, and to have a maximum slope of no greater than ~70° adjacent to the spreading axis. The problem is simplified by assuming a constant stress profile and a constant yield stress of 52 MPa. We assume a Young's modulus and Poisson's ratio of 60 GPa and 0.25, respectively, and a density contrast between lithosphere and water of 3800 kg m -3 . We vary the flexural rigidity of the bending plate, expressed in terms of T e , in order to obtain a bending profile which best fits the observed seismicity and 20° dip of the corrugated surface on the seabed. T e represents the mechanical strength of a bending plate, which can be thought of as a response function that does not correlate to any geological or geophysical boundary within the lithosphere.