Abstract. In a study of geological and geophysical data from the Mid-Atlantic Ridge, we have identified 17 large, domed edifices (megamullions) that have surfaces corrugated by distinctive mullion structure and that are developed within inside-corner tectonic settings at ends of spreading segments. The edifices have elevated residual gravity anomalies, and limited sampling has recovered gabbros and serpentinites, suggesting that they expose extensive cross sections of the oceanic crust and upper mantle. Oceanic megamullions are comparable to continental metamorphic core complexes in scale and structure, and they may originate by similar processes. The megamullions are interpreted to be rotated footwall blocks of low-angle detachment faults, and they provide the best evidence to date for the common development and longevity (-1-2 m.y.) of such faults in ocean crust. Prolonged slip on a detachment fault probably occurs when a spreading segment experiences a lengthy phase of relatively amagmatic extension. During these periods it is easier to maintain slip on an existing fault at the segment end than it is to break a new fault in the strong rift-valley lithosphere; slip on the detachment fault probably is facilitated by fault weakening related to deep lithospheric changes in deformation mechanism and mantle serpentinization. At the segment center, minor, episodic magmatism may continue to weaken the axial lithosphere and thus sustain inward jumping of faults. A detachment fault will be terminated when magmatism becomes robust enough to reach the segment end, weaken the axial lithosphere, and promote inward fault jumps there. This mechanism may be generally important in controlling the longevity of normal faults at segment ends and thus in accounting for variable and intermittent development of inside-corner highs.

that the position and longevity of segments are controlled primarily by the subaxial position of buoyant mantle diapirs or focused zones of rising melt. Within segments, there are distinct differences in seafloor depth, morphology, residual mantle Bouguer gravity anomaly, and apparent crustal thickness between inside-corner and outside-corner crust. This demands fundamentally asymmetric crustal accretion and extension across the ridge axis, which we attribute to low-angle, detachment faulting near segment ends. Cyclic variations in residual gravity over the crossisochron run of segments also suggest crustal-thickness changes of at least 1-2 km every 2-3 m.y. These are interpreted to be caused by episodes of magmatic versus relatively amagmatic extension, controlled by retention and quasiperiodic release of melt from the upwelling mantle. Detachment faulting appears to be especially effective in exhuming lower crust to upper mantle at inside corners during relatively amagmatic episodes, creating crustal domes analogous to "turtleback" metamorphic core complexes that are formed by low-angle, detachment faulting in subaerial extensional environments.

[1] An extensive deep-tow magnetic survey of the TAG ridge segment on the MidAtlantic Ridge reveals new information about the relationship between the magnetic anomaly field and the TAG hydrothermal deposits. Results show the strongest magnetization is located over the neovolcanic axis and asymmetrically toward the western side of the central Brunhes anomaly. A well-defined linear magnetization low is located over the eastern rift valley wall of the TAG segment. The near-bottom data show no direct correlation between this crustal magnetization low and the hydrothermal deposits. The magnetization low is explained by crustal thinning caused by 4 km of horizontal extension along a normal fault. Previous observations and sampling indicate exposures of gabbros and dikes in the eastern rift valley wall, suggesting slip along a normal fault has revealed this crust. Modeling suggests the fault has been active since 0.35 ± 0.1 Ma at a horizontal slip rate of roughly half the spreading rate of 22 km/Myr. The TAG hydrothermal system is located on the hanging wall of this fault within 3 km of its termination. Over the past several hundred thousand years, movement on the detachment fault may have episodically increased the permeability of the hanging wall reactivating the overlying hydrothermal systems. Significant vents like TAG may be typically associated with hanging walls of long-term detachment faults near seafloor spreading centers. This would imply that it is the reactivation of permeability in the hanging wall related to repeated fault movement that controls the longevity of these hydrothermal systems.

We have tested and corroborated the propagating rift hypothesis with high-resolution Sea Beam and Deep-Tow data collected over the Galapagos 95.5øW propagating rift system. The propagating rift is continuously breaking through the Cocos plate at a velocity of about 50 km/m.y. with an azimuth of about 273 ø, away from the Galapagos hotspot. This process transfers Cocos lithosphere to the Nazca plate, including the preexisting spreading center which fails as it is gradually preempted by the propagating riff. The spreading center azimuth is being changed by about 8 ø clockwise by this rift propagation. The active propagating and failing rift axes overlap by about 15 km and are joined by a broad and anomalously deep zone of nonrigid plate deformation rather than by a transform fault. Although rift propagation is continuous on a scale of a few kilometers (50,000 years), and lithospheric transferral from one plate to the other occurs mostly continuously, as the preexisting rift fails, it breaks into discrete segments. The geometry of seafloor fabric and tectonic elements such as active and failed rifts, the V-shaped pattern of pseudofaults, and the oblique tectonic fabric in the sheared zone of transferred lithosphere are observed as predicted. This quantitative verification provides strong support for the propagating rift hypothesis. amplitude anomalies to the west [Vogt and Johnson, 1973; Anderson eta!., 1975]. The magnetic telechemistry hypothesis was developed to explain this pattern of anomaly amplitudes as a function of rock chemistry [Vogt and Johnson, 1973]. Iron and titanium-rich (FeTi) basalts with unusually strong remanent magnetization dredged from the high-amplitude magnetic (HAM) zone supported this hypothesis [Anderson eta!., 1975; Byerly eta!., 1976; Vogt and de Boer, 1976]. Schilling et al. [1976] proposed that these basalts, distinctive relative to normal mid-ocean ridge basalts (MORB), probably resulted from unusually extensive shallow level fractional crystallization, and Vogt and de Boer proposed that this anomalous zone was growing westward at a velocity of 100 mm/yr. When the tectonic evolution of this geochemically interesting area was analyzed, it appeared that sequential jumps of the spreading axis were reorienting the spreading center azimuth from 265 ø to 273 ø and systematically transferring lithosphere from the Cocos to the Nazca plate. On successive profiles from east to west, each jump appeared slightly younger and slightly longer than on the preceding profile, with the most recent jump observed near 95.5øW. The propagating rift hypothesis, illustrated in Figure 2, was developed to explain this pattern of jumps in the young magnetic anomalies [Hey and Vogt, 1975, 1977]. A similar hypothesis was developed independently by Shih and Molnar [1975] to explain missing and extra magnetic anomalies in the older seafloor of the North Pacific, and other propagating rifts have since been recognized on the Juan de Fuca ridge [e.g., Hey, 1977b], the southeast rift zone of Iceland [e.g., Schilling eta...

A model for propagation of an active spreading axis into preexisting lithosphere has been developed based on intensive detailed studies using SeaMARC II, Sea Beam, Deep‐Tow, and Alvin near the tip of the propagator growing westward at ∼52 km/m.y. along the east‐west trending Galapagos spreading axis near 95.5°W. Initial lithospheric rifting appears to be accommodated along reactivated roughly east‐west abyssal hill faults at the “tectonic tip.” Extension leads to downdropping of a keystone block which is pervasively cut by normal faults and fissures. The initial volcanism (“initial volcanic tip”) occurs about 6.5 km behind the tectonic tip as basalt erupts through these fissures, covering most of the keystone block. The “neovolcanic axial tip” occurs another 4.5 km back at the western termination of a well‐defined axial pillow ridge along which volcanism is localized. Spreading along the propagator axis accelerates to the full rate (29 mm/yr half rate) at the “full rate tip,” interpreted to be ∼10 km behind the neovolcanic axial tip, or ∼21 km behind the tectonic tip. A complex pattern of strain in the tip area results from the interaction of plate spreading, transform shear stress, crack propagation, dynamic depression of the western end of the axial valley and the transform zone, and variations in the strength, thickness, heterogeneity, and anisotropy of the lithosphere. Southwesterly curvature of the neovolcanic axis near its tip is inferred to result primarily from the response of young, thin, weak, relatively homogeneous lithosphere to counterclockwise rotation of the maximum tensional stress direction, relative to the spreading direction, due to the influence of shear stress from the transform zone. Similar stress interactions likely occur at the tectonic tip but are inferred to be accommodated in this older, thick, strong, heterogeneous, and anisotropic lithosphere by oblique slip along a set of left‐stepping, reactivated, preexisting east‐west abyssal hill structures. Although on a time scale of ∼106 years the overall process of propagation is mostly continuous, rifting at the tectonic tip appears discontinuous at a scale of ∼1500 m or 30,000 years, based on the east‐west spacing of grabens formed as rifting begins. The scale of discontinuity of propagation of the neovolcanic axis is ∼3–5 km or ∼50,000–100,000 years, based on the distribution of northeast trending features interpreted as beheaded western terminations of paleoneovolcanic axes. The extension involved in rifting is ∼10–20%, based on summing the horizontal normal separation of all normal faults and fissures integrated and averaged over the zone of rifting. These values are approximately doubled if simple shear listric faulting is dominant rather than pure shear normal faulting.