5.1. Mantle Sources and Relationship Between Genovesa Ridge and Island Lavas
 Lavas from Genovesa Island and Genovesa Ridge are among the most distinctive in the Galápagos Archipelago. Most notably, they are impoverished in the incompatible elements (including Ti, K, and P), closely resemble normal MORB, and many are ultraporphyritic with large (>1 cm) plagioclase phenocrysts [Harpp et al., 2002]. Therefore the depleted mantle source that feeds Genovesa Island [Harpp et al., 2002] is similar or even identical to that supplying magma to the submarine ridge, and consists primarily of depleted upper mantle.
 Most of the differences between the subaerial Genovesa Island and submarine Genovesa Ridge samples can be attributed to variations in the analyzed materials [i.e., glass versus whole rock), the accumulation of plagioclase, and shallow-level fractional crystallization (Figure 8). The compositional differences between the high-MgO porphyritic sample D22-D and the aphyric, lower MgO lavas are likely the result of fractional crystallization of olivine and plagioclase. Increasing Ca/Al ratios with evolution (Figure 8) rules out significant clinopyroxene fractionation, which is important because clinopyroxene is the liquidus phase in these compositions at pressures of greater than ∼1 kb (as calculated from the MELTS program [Ghiorso and Sack, 1995]). Most fractionation of the Genovesa Ridge magmas therefore must have taken place in the shallow crust, in contrast to magma from the western Galápagos volcanoes, which undergo extensive fractional crystallization in the upper mantle and lower crust [Geist et al., 1998].
 Although the major element compositions of lavas from Genovesa Island and Genovesa Ridge are similar and appear to be related primarily by shallow crystallization, subtle differences distinguish the two suites. In terms of major elements, the glass and whole rock major element compositions from sample D22-D are more MgO-rich than any of the subaerial lavas. For some incompatible trace elements and their ratios, the most depleted Genovesa Island lavas and the most enriched submarine-ridge compositions barely overlap (e.g., Lan/Smn and Smn/Ybn; Figure 10c). For other ratios (e.g., Hf/Ta, Zr/Nb, La/Ce), the compositions are more distinctive. The incompatible element ratios of most ridge lavas define a trend similar to that exhibited by the island but extending toward more depleted compositions (Figure 10c). Otherwise, the submarine ridge lavas most closely resemble basalts produced during the island's earliest shield-building phase [Harpp et al., 2002].
 With the limited set of samples and analytical data, it is not possible to determine whether the island magmas originate from slightly lower extents of melting, or whether they are derived from a slightly less depleted source. The lower Sm/Yb(n) and elevated ITE concentrations (e.g., Rb, Y) of the lavas (Figure 9) from the base of the ridge may be the result of shallower depths and smaller degrees of melting than lavas erupted at the ridge crest [e.g., Plank and Langmuir, 1992]. Nevertheless, the higher MgO and more depleted trace element signatures in Genovesa Ridge lavas lead us to believe that the magmatic plumbing systems feeding the ridge and island may not be connected.
 The relatively wide compositional range of the glass inclusions in the plagioclase in D22-D suggests that these phenocrysts originated from a number of host magmas, which hybridized and homogenized as the crystals formed. The glass rind composition is similar to the average composition of the inclusions, which supports this hypothesis. The SiO2 contents of the glass inclusions are problematic, however; all of the inclusions are richer in SiO2 than are the glass samples (Table 2). This discrepancy is not caused by post-entrapment crystallization. Plagioclase removal would decrease the liquid SiO2 content, and the crystallization trend is opposite that formed by plagioclase addition (Figure 8). Likewise, the SiO2 variations in the glass inclusions could not be the result of plagioclase dissolution, as the inclusions are slightly poorer in Al2O3 than is the glass (Figure 8). Thus, despite the absence of a discernable relationship between the inclusions and the glass, it is clear that the plagioclase phenocrysts did not simply grow from a liquid with the same composition as the glass [cf. Cullen et al., 1989]. Moreover, the fact that the inclusions exhibit both higher and lower Mg # than the glass indicates that hybridization occurred at shallow levels and between magmas both cooler and hotter than the final erupted magma.
 In summary, Genovesa Ridge lavas are derived from depleted mantle sources similar to those of Genovesa Island, have experienced shallow fractionation, and underwent mixing with magmas of variable temperatures prior to eruption. It is possible that the magmas supplying the ridge come from a central, stratified magmatic system beneath the island that intrudes laterally, as has been proposed for the Puna Ridge [e.g., Clague et al., 1995; Johnson et al., 2002]. The compositional distinctions instead suggest that the ridge magmas are fed from a broad melt zone directly underlying the ridge, and that the two systems are not directly connected.
5.2. Evidence of Volcanic Constructional Processes
 The morphology of the seafloor in the study area supports the conclusion that the Genovesa Ridge and its extensions toward Marchena and Pinta Islands result from constructional volcanic processes. The abundance of lava flows, volcanic terraces, and eruptive cones provide the most fundamental evidence for this hypothesis. Moreover, the submarine ridges and platforms exhibit markedly higher reflectivity and rougher surface textures than the surrounding sediment-covered seafloor. The hummocky terrain dotted with cones and prominent terraces is morphologically consistent with sequences of accumulated submarine flows (Figures 3–5).
 The boundary between the surrounding seafloor and the submarine ridges is abrupt at the base of the steeply sloping lower flanks of the ridge and is interfingered with young sediment, implying that the volcanic pile has accumulated relatively recently compared to the age of the ocean floor. The basement in this region is ∼1.5 Ma [Wilson and Hey, 1995] and should be thoroughly sediment-covered, given the high-productivity waters of the archipelago (which is indicated by our regional side scan survey). The distinct lack of sediment cover on the ridges suggests that the constructional volcanism occurred ≪1.5 Ma.
 Although volcanism is the dominant process for construction of the ridges in the study area, the eruptive activity has been tectonically controlled. The alignment of the cones along the ridge crests, the uniform orientation of flow fronts parallel to the ridge, the elongate structures between Pinta and Marchena that are also parallel to the NW-SE trend of the platform, and the marked orientation of the submarine ridges themselves suggest that the distribution of eruptive material is influenced by stress fields on a regional scale. Furthermore, the systematic shift in ridge orientation from 075° east of Genovesa Island to 088° between Genovesa and Pinta Islands suggests that the direction of the deviatoric stress fields must vary across the northern Galápagos.
5.3. Comparison With the Puna Ridge, Hawaii
 At first glance, the Genovesa Ridge appears comparable to other hot spot-related volcanic rift zones such as the Puna Ridge, the submarine extension of the East Rift Zone of Kilauea volcano. The Puna Ridge has been well studied [e.g., Malahoff and McCoy, 1967; Fornari et al., 1978; Holcomb et al., 1988; Lonsdale, 1989; Smith et al., 2001], most recently by high-resolution morphological analyses using near-bottom side-scan sonar and photoimagery [Smith et al., 2002]. The data for the Genovesa Ridge are far less extensive, having been collected only at the reconnaissance level. As such, it is not possible to compare the fine-scale morphological features of the ridges.
 Despite being comparable in size (the Genovesa Ridge is 2/3 the length of the Puna Ridge; Figure 6), and having similarly sharp, steep cross-sectional profiles, additional basic differences between the rift zones suggest they were not formed by the same fundamental mechanisms (Table 1). First, the volume of material erupted on the Puna submarine rift is dwarfed by the volume of the Kilauea structure [Holcomb, 1987; Holcomb et al., 1988; Fornari et al., 1978]. In contrast, Genovesa Island is merely the subaerial expression of ridge segment A's peak, and the ratio of the ridge's volume to that of the island is much greater at Genovesa than at the Puna Ridge.
 Second, even though Genovesa Ridge lava flow fields have similar volumes to those erupted at the Puna Ridge (Genovesa Ridge flow field F1, assuming ∼10 m thickness: >1.5 km3; Puna Ridge flows: ∼2 km3 each [Holcomb et al., 1988]), the thin, voluminous lava flows of the Puna Ridge are erupted primarily at depths greater than 4500 m [e.g., Smith et al., 2002; Holcomb et al., 1988; Fornari et al., 1978]; on the Genovesa Ridge, many of the large flows originate on or near the ridge crest.
 In addition, Genovesa Ridge cones are more abundant on the ridge crest, up to 100 m taller, and more unevenly distributed than they are on the Puna Ridge [Smith et al., 2002]. Many of the Puna axial cones are thought to be either rootless vents or secondary vents constructed over lava tubes supplying the large flank terraces [Smith et al., 2002]. This may be the case on Genovesa Ridge as well, on the basis of the apparent connection between some of the larger segment B cones and the younger, extensive lava flow fields (F2; Figures 5 and 7). The uneven, nonlinear distribution of cone vents on Genovesa Ridge, coupled with the en echelon structure and geochemical differences imply that a single dike complex does not underlie the entire structure, as is believed to be the case for the Puna Ridge [e.g., Malahoff and McCoy, 1967]. Instead, Genovesa Ridge may be supplied by a series of discontinuous dike swarms, each responsible for an individual ridge segment.
5.4. Magmatic Driving Forces and the Origins of Submarine Ridges and Platforms in the Northern Galápagos
 Submarine rifts on the Hawaiian volcanoes are believed to arise from lateral migration of magma from a central body, in which magmatic pressure provides the driving force for the propagation of the rift zone [e.g., Lonsdale, 1989]. Alternatively, ridge construction may be a passive phenomenon in which magma migrates vertically in response to extensional tectonics and fracture formation, analogous to a mid-ocean ridge system. In the former case, a single, focused magma source actively drives rift formation. In the latter, magma is drawn passively from melt pockets in the mantle immediately underlying the fractures in response to extension.
5.4.1. Active Rift Formation Model
 According to the current paradigm for Hawaiian volcanic rift zones, dike swarms transport magma along the ridge outward from a central body [e.g., Pollard et al., 1983; Knight and Walker, 1988]. Such ridges are usually characterized by highly linear structures, which extend from a subaerial volcano's shoreline and possess relatively uniform slopes of a few percent (7–8%) [Fialko and Rubin, 1999; Lonsdale, 1989; Fornari et al., 1978]. According to numerical and physical models, the along-axis slope is controlled primarily by laterally migrating dikes as they lose pressure with increasing distance from the magma source [Fialko and Rubin, 1999].
 One of the most striking features of the Kilauea East Rift Zone (ERZ) slopes (subaerial and submarine portions included) is the evenness of the crest over the first 100 km. There is only one major change in gradient, at the coastline, where the subaerial ERZ slope of 23 m/km steepens to 51 m/km along the Puna Ridge then continues nearly constantly to 2700 m depth [Lonsdale, 1989; Smith et al., 2002]. This pattern is consistent with models such as those of Lacey et al.  and Angevine et al.  that predict magma injected at the base of a volcano will follow paths of least resistance along a surface of constant hydraulic potential, the slope of which increases below sea level. Lonsdale  infers that the ERZ-Puna Ridge must therefore be supplied by a high, regular magma flux to maintain the equipotential surface for over 100 km along axis.
 Furthermore, the uniform slopes of ridges such as the Puna have been attributed to a mechanism in which the magmatic pressure from the central reservoir controls injection into the rift zone. Magma migrates along topographically controlled slopes with little or no variation in tectonic stress along strike [Fialko and Rubin, 1999]. Long rift zones (on the order of 100 km) must maintain relatively constant slopes for effective along-axis magma transport, because magmatic driving pressure decreases with distance from the source [Fialko and Rubin, 1998, 1999]. Consequently, the evenness of a ridge's slope provides a means of assessing the continuity of the intrusive dike system [e.g., Angevine et al., 1984]. In shorter dike systems or those with low magmatic driving pressures, eruptions are focused near the magma source, resulting in steeper, variable axial profiles near the vent that decrease progressively with distance [Fialko and Rubin, 1999].
 The distinct en echelon structure of Genovesa Ridge is not consistent with a lateral injection model. As described above, the ridge is not a straight rift, but a sequence of at least three segments that curve toward each other at their tips. The along-axis slope of Genovesa Ridge is highly variable (Figure 6), indicating that there is not an even surface of constant hydraulic potential along the ridge controlling magma distribution. The ragged axial profile of the Genovesa Ridge suggests instead that there must be variable, localized intrusions along the rift, consistent with an irregular magma supply along strike. Consequently, magma supply to the rift is sporadic in time (irregular, limited magma flux) and/or in space (isolated dike swarms).
 Further support for this conclusion comes from a comparison of the axes of the Puna and Genovesa Ridges. The crest of the Puna Ridge maintains a relatively constant width across its westernmost 38 kilometers (down to 2100 m depth), varying only between 2–4 km for the majority of the segment [Clague et al., 1994]. East of this point, the ridge crest becomes less well defined and considerably narrower. Fialko and Rubin  state that a topographically driven dike should be characterized by a constant thickness along any cross-section parallel to the direction of dike propagation; this translates to a constant crestal width throughout the portion of the rift supported by the main dike complex (except near the dike nose), as is observed along the Puna Ridge. Consistently, magnetic studies of Malahoff and McCoy  and subsequent work by Smith et al. [2001, 2002] confirm that virtually the entire 75-km long Puna Ridge is underlain by a dike complex ∼11 km in width.
 In contrast, crestal widths at Genovesa Ridge vary significantly over short distances (Figures 3 and 4). The segment adjacent to Genovesa Island (A) decreases from >6.5 km at its widest point (including the almost flat island) to near zero within ∼14 km; segment B pinches out completely from ∼6 km at its midpoint within a mere ∼9 km (Figure 3). This variability in crestal width implies discontinuous dike swarms along the axis, further ruling out a centralized, pressure-driven magma supply.
5.4.2. A Passive Extensional Rift Zone Model
 The structural, morphological, and geochemical characteristics of Genovesa Ridge are consistent with a passive rift zone model in which partial melts in the underlying mantle migrate in response to stresses induced by far-field tectonic forces. In an opening (mode I) crack, the local stress field at the propagating tip is complex and includes components of shear perpendicular to local maximum tensile stress. When two adjacent straight cracks overlap, the local shear component (mode II: sliding) causes the cracks to propagate out of plane, curving toward each other at increasingly sharp angles [Lange, 1968]. Theoretical calculations predict that interacting, en echelon fractures will have two defining characteristics: (1) as adjacent fractures approach each other, they will diverge initially before converging, taking on a hook-like path [e.g., Pollard and Aydin, 1984]; and (2) if propagation is driven by extensional forces, the region between the tips should experience focused extension, resulting in the formation of topographic depressions in the overlap region [Pollard and Aydin, 1984].
 Genovesa Ridge segments follow hooked paths, first diverging from each other before converging, reminiscent of overlapping spreading centers along fast spreading mid-ocean ridges (Figures 4 and 5) [e.g., MacDonald and Fox, 1983]. Topographic depressions exist between the segment tips, bounded by the crests of the ridge segments (Figures 4 and 5). These characteristics imply that the Genovesa Ridge segments are (1) interacting mechanically [Olson and Pollard, 1989]; (2) propagating from a central location; and (3) likely being formed by mode I opening caused by extensional stresses [Thomas and Pollard, 1993]. All of these observations are consistent with the focusing of magma by tectonic stresses as a passive response to extension.
 In the passive rift zone model, each segment is supplied by magma derived from the mantle immediately underlying the segment. Magma subsequently migrates toward the segment tips, resulting in the en echelon geometry. Consistently, lavas erupted from the base and crest of Genovesa Ridge appear to originate from a broad melt zone that taps the underlying depleted mantle, not necessarily from a single, interconnected magmatic system. The shallow fractionation signature (<1 kb) further precludes the existence of a long-lived magma chamber supplying the entire ridge. This process is reflected in the Genovesa Ridge morphology, including the uneven distribution of cone vents along its crest, variable crestal widths along axis, and a nonuniform axial slope, all indications that the ridge is underlain by a series of discontinuous dike swarms. The large-scale linearity of Genovesa Ridge (075°) and the western platforms (088°) further implies that regional stress fields control the volcanic constructional process, a phenomenon predicted by the passive rift zone model.
5.5. Passive Extension in the Northern Galápagos Region
 The mechanism responsible for passive rifting and for the major structural trends observed in the Northern Galapagos region may be deviatoric stresses generated by interaction between the Galapagos spreading center and the hot spot [e.g., Harpp and Geist, 2002]. Both geochemical [e.g., Verma and Schilling, 1982; Verma et al., 1983; Schilling et al., 1982; Detrick et al., 2002] and bathymetric observations [e.g., Canales et al., 1997] have shown that the Galápagos hot spot and the GSC are close enough to interact extensively. The flank of a spreading center adjacent to a hot spot is weakened by thermal effects of the warm, sublithospheric plume, encouraging rifting between the plume and the ridge [Small, 1995]. Lithospheric cracking may be initiated along crustal weaknesses caused by large-scale tectonic stresses centered on the 91°W transform fault [e.g., Clifton et al., 2000]. The tension-induced fractures could initiate passive upwelling of the underlying, depleted mantle, resulting in the observed rift zone formation.
 In a biaxial tensile loading model designed to reproduce the stress field around a transform fault, Gudmundsson  proposed that extensional structures form astride the ridge when ridge-parallel tensile stress accompanies the usual ridge-perpendicular stress. The stress field generated by the model successfully reproduces the observed distribution of oblique fractures near transform faults in Iceland (Figure 11). Where ridge-perpendicular stress is dominant, fractures are parallel to the ridge; where stresses parallel to the ridge are stronger, fractures become perpendicular to the ridge. In between, fractures oblique to both the transform fault and ridge are formed when the two stresses are close to equal (∼45° [Gudmundsson, 1995]) (Figure 11). Field studies in Iceland and on the Mid-Atlantic Ridge have noted oblique structures like those predicted by the model [e.g., Gudmundsson, 1995; Gudmundsson et al., 1993; Macdonald et al., 1986].
Figure 11. Conceptual model of the genesis and evolution of the Genovesa Ridge. (a) GSC-hot spot geometry over past ∼8 Ma (from Wilson and Hey ); Genovesa Ridge would have been constructed in the last <1.5 Ma. (b) Present-day regional stress field in the northern Galapagos, induced by adjacent transform fault at 91°W [Gudmundsson, 1995]. (c) Bathymetric map of the present-day northern Galapagos Archipelago; note broad similarity of distribution and orientation of volcanic structures to the stress field in B (see Figure 1 for reference).
Download figure to PowerPoint
 Many of the volcanic ridges in the northern Galapagos conform to the distribution and orientation of stresses predicted around the 91°W transform fault [Gudmundsson, 1995]. The structural trend defining Genovesa Ridge (075°) extends through Genovesa Island to West Genovesa Ridge. Genovesa Island is crosscut by major eruptive fissures whose alignment is also parallel to the overall trend of the ridge, lending support to the regional origin of the deviatoric stresses. These fissures are tension fractures with a minor normal component and have erupted lava virtually identical to the average composition of the shield-building phases of the island [Harpp et al., 2002]. The trend of the fractures shifts to nearly E-W (088°) between West Genovesa Ridge and Marchena Island, then the fractures turn northwestward toward Pinta Island.
 Although the trends of Genovesa Ridge and the Wolf-Darwin lineament almost exactly parallel those predicted by Gudmundsson's  calculations, the model does not conform to the trends of fissures and fractures directly south of the 91°W transform, near Pinta and Marchena Islands (see Figures 11b and 11c). This may be due to the boundary conditions set in the model; other calculations [e.g., Fujita and Sleep, 1978; Behn et al., 2002] predict alternative orientations in the area beyond the transform.
 Theoretically, similar deviatoric stress fields should occur at all transform faults; if so, then why are volcanic provinces like the Northern Galapagos not more common at mid-ocean ridge systems? We believe the difference lies in the particular setting of the northern Galápagos, where the hot spot is located within a mere 100–200 km of the spreading center. Consequently, the mantle between the central Galapagos platform and the GSC is hotter than the average, ambient mantle near MORs that are far removed from hot spots [e.g., Schilling, 1991]. Whereas all transform faults likely impose regional deviatoric stresses on the surrounding lithosphere and may induce extensive networks of fractures [e.g., Gudmundsson, 1995; Gudmundsson et al., 1993; Macdonald et al., 1986; Clifton et al., 2000], it is only in the presence of high heat flow and the attendant excess magma that significant eruptive activity can occur. The serendipitous combination of the transform fault and the adjacent hot spot results in the formation of major volcanic ridges along lines of deviatoric stress, essentially illuminating the regional stress field. According to this model, Genovesa Ridge is the conjugate to the Wolf-Darwin Lineament, which has been proposed to result from tension in the inside corner of an extensional transform zone [Harpp and Geist, 2002].
5.6. Implications of the Passive Rift Model in the Northern Galápagos Region
 The model in which the Genovesa Ridge formed as the passive response to regional plume-ridge interaction implies that the entire length of the ridge, including Genovesa Island, formed penecontemporaneously. The lithosphere underlying Genovesa Ridge is approximately 1.5 Ma [Wilson and Hey, 1995], placing an upper limit on its age. The lack of sediment cover on the ridges in the study area suggests that they are considerably younger than 1.5 Ma. Field studies by Harpp et al.  indicate that Genovesa Island emerged <350 ka. Whether the island's formation marks the final stage of the entire ridge's formation or only the last step in the construction of segment A, however, requires further detailed chronological analysis unattainable with the samples in hand.
 According to the passive rift model, each segment of the ridges taps only the underlying mantle, with little lateral transport. Consequently, variation in depth and volume of the segments indicates that the quantity of melt available in the underlying mantle differs along the fracture set, reaching a maximum at segment A (Genovesa Island) and a minimum at segment C for Genovesa Ridge. The substantially greater volumes of Pinta and Marchena Islands (including the submarine ridges that emanate from them) further suggest that magma supply is considerably more abundant west of Genovesa Ridge. This observation is supported by the fact that lavas also become progressively more enriched westward from Genovesa, reaching a peak at Pinta Island, indicating greater contribution from the Galapagos plume [White et al., 1993; Kurz and Geist, 1999; Harpp et al., 2002]. Even though these observations are based on sparse sampling, they provide important preliminary information regarding the geographic distribution of plume material in the mantle and the dynamics of plume-ridge magmatic communication.