Ultramafic rocks are exposed in many different tectonic settings in the ocean basins, including oceanic core complexes on slow-spreading ridges (e.g., Cannat, Karson, Miller, et al., 1995), the floors and walls of transform faults (e.g., Dick et al., 1991), the tips of propagating rifts such as Hess Deep and other extensional features such as Kings' Trough, the inner slopes of nonaccretionary trenches where mantle is exposed by tectonic erosion, and deep forearc grabens exposed by extensional tectonics. In every case, regardless of tectonic environment, the incursion of seawater leads to serpentinization, and, in some cases, the serpentinization is extensive. At sites of recent exposure, such as some places along the Tonga forearc, serpentinization has not proceeded very far. The form and composition of the serpentinites are quite similar in most of these settings because the fluid involved in serpentinization is seawater and the process of interaction is passive, leaving the outcrops intact for the most part. Some serpentinite in these settings is disaggregated. Although some controversy still exists as to whether this deformation is the result of formation of talus at the base of slopes or is the consequence of extrusive processes, most results of in situ inspection with submersibles and ODP drilling studies suggest the extrusive model for formation of these structures is less likely (Fryer, 2002).
Serpentinites in the IBM forearc setting have distinctive structural and compositional characteristics that reflect the dynamic nature of formation of the edifice and the unusual composition of the fluids that are the source of the serpentinization of the protoliths. As can be seen from geophysical, petrological, biological, and fluid-geochemical studies, the protoliths show evidence of more complicated origin. They may have been formed originally in an oceanic spreading center setting but show overprints of slab-derived constituents and suprasubduction-zone melting and deformation episodes. Trace element ratios of fluids associated with serpentinite muds indicate that these fluids cannot have formed exclusively as a consequence of simple seawater-rock interaction but must have involved a slab-derived component. The muds themselves are similar to some disaggregated deposits associated with exposures in other types of settings described above, but the form and spatial distribution of the mud flows on the serpentinite seamounts is distinctive. The muds most likely consist of highly comminuted fault gouge with angular to subangular fragments of all sizes, dominantly silt- to sand-sized grains, but also include clasts of rock from pebble to boulder size. The muds may derive from anywhere beneath the edifice, from the subducted slab, the region of the décollement, or the suprasubduction-zone mantle and crust. The mud is too weak (relative to surrounding unserpentinized mantle peridotite [see Phipps and Ballotti, 1992]) to displace surrounding mantle peridotite and rise under gravitational instability alone. Thus, we suggest that extensional faulting in the forearc is necessary for rise and emplacement of the muds.
The outer 50 km of the Izu-Bonin forearc is a ridge of variably serpentinized peridotite (Taylor, 1992; Taira et al., 1998; Takahashi et al., 1998; Sato et al., 2004; Kamimura et al., 2002). Torishima Forearc Seamount rests on top of this ridge, as do many others (Taylor, 1992). There is a suggestion that the low-velocity zone beneath the outer forearc, which on MCS profiles extends downdip along the base of the forearc wedge, may represent the source of the serpentinized peridotite that rises to form the Izu-Bonin serpentinite seamounts. This implies that the mechanism for formation of the serpentinite seamounts of the Izu-Bonin forearc may differ somewhat from that of the Mariana forearc. The idea that the Izu-Bonin seamounts are diapiric and represent the head of rising serpentine mush from deep along the décollement seems at odds with the observation of horizontal layering on MCS profiles beneath Torishima Forearc Seamount (see Horine et al., 1990), and this hypothesis will require more detailed study.
The deep low-velocity layer identified by Fryer et al. (1985) in the Mariana forearc also parallels the inferred top of the subducting slab. The interpretation that this low-velocity layer may be serpentinized peridotite of the mantle wedge seems reasonable, but we do not suggest a diapiric origin for the seamounts in the Mariana forearc. The mechanism of emplacement and formation of mud volcanoes on the Mariana forearc may differ from that of the Izu-Bonin system. The forearc of the Mariana system has undergone far more along-strike extension (Fryer et al., 1985; Fryer, 1992; Wessel et al., 1994) than the Izu-Bonin forearc (Taylor, 1992). Faulting in the Mariana forearc takes place on all scales, and conjugate faulting, likely associated with along-strike extension, is dominant in the southern half of the forearc, whereas trench-parallel faulting is dominant in the north (Stern and Smoot, 1998; Stern et al., 2004). Thus, the avenues for egress of serpentinite muds are more widely distributed in the Mariana forearc than in the Izu-Bonin part of the system.
Serpentinization of the mantle overlying the subducting plate may influence the nature of seismicity along the IBM margin because the phase of serpentine present along the décollement may influence its strength. Seismic velocity alone, however, is not sufficient to identify the phase of serpentine that comprises the zones of low velocity in the deep mantle overlying the subducted slab; thus, we must exercise caution in making predictions regarding the potential physical properties of serpentinized mantle near the décollement. O'Hanley (1996) notes that the most common phase of serpentine at the temperatures and pressures most likely along the décollement of the IBM system anywhere along strike is lizardite. Lizardite is also the dominant serpentinite phase present in samples cored from all of the seamounts examined to date. Lizardite has significantly different physical properties from chrysotile and would not be as likely to behave as a lubricant for the subduction zone, as suggested by Sato et al. (2004) and Kamimura et al. (2002). D'Antonio and Kristensen (2004) note that brucite may also be important in determining low-velocity zones in the forearc and in limiting down-slip earthquakes in subduction zones as suggested by Peacock and Hyndman (1999). Although geophysical data show a low-velocity region under the entire IBM forearc, we cannot yet determine the composition of the phases present nor their proportions from velocity information alone. The idea that serpentine and brucite in some combination provide an explanation for the aseismic character of the shallow mantle in the Izu-Bonin-Mariana subduction zone (e.g., Pacheco et al., 1993) has considerable merit. However, it remains to be determined how metamorphic minerals lubricate the décollement sufficiently to fit the seismological observations.
The pore fluids entrained in the mud flows erupting from these seamounts has a composition unique in the world (Mottl, 1992). There is no doubt that fluids rising with the serpentine muds in the IBM mud volcanoes have a slab-derived signature and that the composition of different seamounts varies with distance from the trench in a consistent manner (Mottl et al., 2003, 2004). There is also no doubt that seawater infiltrates the serpentinite muds once they are exposed at the seafloor following eruption (Mottl, 1992; Mottl et al., 2003, 2004; Wei et al., this volume). The extent to which these two fluids interact is still unknown, as is the hydrologic mechanism for seawater incursion into the interior of the edifices. Both drilling data from Conical Seamount (Sites 788–789; see Fryer and Mottl [1992] and Lagabrielle et al. [1992]) and side-scan sonar data from Conical and several other seamounts on the southern half of the forearc (Fryer et al., 1999) show that the seamounts erupt episodically, possibly in association with earthquake activity along the décollement (Fryer, 1990).
As we consider implications for the observed compositional variations in pore fluids, particularly with regard to estimates of fluid flux in the Mariana system, it is important to remember that we have sampled only a small number of the mud volcanoes present on this convergent margin. Further, we must realize that each sampling is merely a snapshot of the process at any given edifice. Stern and Smoot (1998) presented detailed maps of the Mariana forearc that show far more forearc seamounts than the nine already sampled (including the three that have been drilled). Because the seamounts are episodically active (Fryer, 1992; Fryer et al., 1999), the estimated output flux for Conical Seamount and South Chamorro Seamount (Mottl, 1992; Mottl et al., 2003) may be an underestimate. The overall output flux for slab-derived constituents through the Mariana forearc since subduction began in Eocene time may be far greater than the estimates proposed thus far.
Numerous deposits of serpentinites on land have been termed sedimentary serpentinites in the early literature and sometimes include marine fossils (Lockwood, 1971, 1972). Active sites like Conical and South Chamorro Seamounts support macrobiological communities and include chemosynthetic microbial communities below the seafloor, feeding on nutrients released during serpentinization and products of fluid-rock-microbial interactions (e.g., methane, other hydrocarbons, sulfate, and sulfides).
The nature of the microbiological communities present on the serpentine seamounts is reflected in the pore fluid compositions of the muds. We have yet to examine the rocks included in the matrix to determine whether they contain a similar microbial community. The differences between the Conical and South Chamorro Seamount communities most likely reflect differences in the rates of extrusion and fluid flow: Alvin dives on Conical Seamount reveal heavy manganese coatings, and although there are individual flows that lack sediment cover there are also areas that do have a thin veneer of sediment. This suggests the summit of Conical Seamount is not as active as the unsedimented summit knoll of South Chamorro Seamount. No biological communities have been identified at Torishima Forearc Seamount, suggesting that this seamount is extinct or at least dormant.
We note microbial communities have only been discovered to date near the surface, Bacteria have been observed to ~3 mbsf (Takai et al., 2005) and Archaea to ~ 30 mbsf (Mottl et al., 2003) at South Chamorro Seamount. Preliminary analyses of recently collected Jason 2 remotely operated vehicle (ROV) push cores from several other seamounts show the presence of similar communities on other active seamounts (Curtis and Moyer, 2005). The alkaliphilic bacteria Marinobacter alkaliphilus) (Takai et al., 2005) appear to be inoculated into the serpentinite muds from seawater as it infiltrates the newly erupted materials, but both the bacteria (and the more voluminous population of Archaea in the deeper muds [Moyer, pers. comm., 2004]) require the unique geochemical conditions provided by the pore fluids in these muds.
The geochemical characteristics of the Izu-Bonin and Mariana forearc serpentinites suggest a close link to subduction processes, both in terms of origin of the fluids responsible for the bulk of the serpentinization and in terms of the mechanism of formation of the mud volcanoes. There is evidence from drilling during Leg 125 that the serpentine seamounts may be long lived (Fryer, 1992). Eocene sediments recovered from Deep Sea Drilling Project (DSDP) Site 459 near one of the serpentine seamounts on the Mariana forearc contain serpentine (Despairies, 1982). Thus, it is possible that the seamount had formed and was shedding serpentinite muds onto the adjacent forearc in the earliest stages of subduction at this convergent margin (Fryer, 1992).
The wide range of rock types in the serpentinite mud flows from the volcanoes of the IBM forearc region indicates that they sample the entire forearc mantle and crust as well as the subducting plate. The clasts include metamorphosed peridotitic and mafic rocks (Maekawa et al., 1992, 1993, 1995; Fryer et al., 1999; Fryer and Todd, 1999; Todd and Fryer, 1999), altered mafic rocks from the subducted plate, including N-MORB, transitional MORB, ocean island basalt (OIB), and even hemipelagic cherts (presumably also derived from the subducted plate), as well as island-arc tholeiite (IAT) and boninite from the suprasubduction-zone igneous basement (Johnson and Fryer, 1990; Johnson, 1992; Johnson et al., 1991). Incipient blueschist metamorphic rocks recovered by drilling during Leg 125 from Conical Seamount provided the first proof that high-pressure/low-temperature metamorphism does indeed occur under active convergent margin systems (Maekawa et al., 1993; Fryer et al., 1999; Fryer and Todd, 1999; Todd and Fryer, 1999). Amphiboles with sodic rims and calcic cores display nearly ubiquitous prograde reactions (Fryer et al., 2000), suggesting a relatively short residence time at high-pressure/low-temperature conditions and a relatively rapid rise to the seafloor from the source region. Although retrograde metamorphic reactions are generally sluggish and high-pressure phases can persist metastably to low pressures, both the décollement source region for the rocks and the conduits of the mud volcanoes are likely to be strongly reactive environments because they are tectonically dynamic and fluid-charged and would thus favor retrograde reactions. If seismic activity along the faults underlying the seamounts triggers episodes of mud protrusion (Fryer, 1992), then the mud flows most likely contain rock clasts derived primarily from the slip surface but could also contain material from anywhere along the route to the seafloor. The subtle variations among flow units from the flank site drilled during Leg 125 show that discrete units, thus discrete source regions, can be distinguished in individual protrusion events. What we do not yet know is the rate of protrusion or the variation in volume of the individual units
There are several models for complexities of mantle flow in suprasubduction-zone regions, but D'Antonio and Kristensen (2004) and Savov et al. (this volume, 2005), based on REE abundances and systematics of cpx grains, propose that deep subarc mantle materials can reach the cold forearc ("corner") region. They suggest that differences in the degree of melting experienced by the subarc mantle could not explain the variations in the degree of depletion recorded in the Izu and Mariana arc lavas alone. They suggest that if this were so, the IBM arc lavas should be similar along strike, but this is not the case. Lavas from the Izu segment are significantly more depleted than Mariana segment lavas (Elliot et al., 1997; Ishikawa and Tera, 1999; Stern et al., 2004). Yamazaki and Yuasa (1998) proposed that there is some evidence from magnetic and gravity anomalies that there was a period of rifting in the middle Miocene that affected the Izu section of the Izu-Bonin arc. This could help to explain some of the along-strike differences in the two arcs. Whether this also explains some of the differences between forearc peridotitic compositions is still unresolved. Macpherson and Hall (2001) noted that the volume estimates for eruptive products in the IBM system given by Bloomer et al. (1995) may be low by a factor of 2 and stress the large extent and volume of boninitic lavas erupted essentially contemporaneously throughout the system. The formation of the ridges west of the Shikoku Basin (Oki-Daito Ridge and Amami Plateau) and the comparatively low volume of boninite in other arc systems may reflect a relatively short period during the early Eocene or earliest middle Eocene in which a hotspot (possibly the Manus plume) influenced volcanism in the nascent IBM system. Macpherson and Hall (2001) also note peculiarities of trace element compositions in the lavas from the early arc and forearc regions of the IBM system that suggest little input from a subducted slab (e.g., Pearce et al., 1999). One interesting corollary of the Macpherson and Hall (2001) hypothesis is that this hotspot model for early influence over the nature of magmagenesis in the IBM system also implies that thermally controlled upwellings (the hotspot plume) and downwellings (associated with the initiation of subduction) in the mantle may occasionally interact (Macpherson and Hall, 2003). If this hypothesis is tenable, how such interaction may have influenced the history of the forearc mantle remains to be determined.