Island arcs, deep-sea trenches, and back-arc basins.
* Swath bathymetry and sidescan acoustic imagery provided the first detailed maps of large seafloor areas;
* Marine seismic studies of sub-seafloor structure and stratigraphy increased in resolution and penetration;
* Manned submersibles and remotely operated vehicles carried our eyes and experiments to the seafloor for precise sampling and observations; and
* Deep ocean drilling provided cores and downhole measurements of previously unsampled formations.
These studies have brought fundamental changes to our understanding of volcanism, crustal deformation, fluid circulation, and sedimentation in trench/arc/back-arc systems.
Subduction is the major recycling process for Earth's crustal materials. Oceanic crust, covered with sediments and dotted with seamounts, plunges at deep-sea trenches back into the mantle (from which it was derived at a mid-ocean ridge). Where sediments are thick, most of them are scraped off onto the leading edge of the overriding plate, like snow piling up on a plow blade, forming an accretionary prism. In other areas, part of the upper plate may collapse onto the subducting plate and be carried down with it.
Not all of the subducted material is recycled into the mantle. Some underplates the crust, and most of the fluids in the pore spaces and hydrous minerals are released by squeezing and heating. At depths of about 100 kilometers, the expelled fluids lower the melting point of the overlying mantle rocks, forming magmas that rise to intrude beneath, and erupt at, island-arc volcanoes. Periodically, the island arc is stretched and splits. Seafloor spreading in the resulting back-arc basin separates an inactive (remnant) arc from the active arc and its deep mantle source.
Initial Arc Volcanism
The roots of the present western Pacific volcanic arcs are about 50 million years old. Had observers been present at their formation, two very different styles of volcanism would have been apparent. One type in Indonesia and Japan, for example, would have looked similar to today's volcanism. However, the volcanism that constructed the base of the intra-oceanic Izu-Bonin-Mariana and Tonga-Kermadec arcs was quite different, as confirmed by ocean-floor drilling there in 1989 and 1991. Volcanism during the first 10 million years of these arcs had characteristics intermediate between modern spreading centers and island arcs. It occurred over a vast terrain, covering up to 400 kilometers from the trench, rather than dominantly along a narrow (50 kilometers) volcanic line at a distance of about 200 kilometers from the trench. The old lavas have the chemical signature of volcanism above a subduction zone, but were fed by multiple dike intrusions in an extending region. Swarms of these dikes are exposed in the Bonin Islands, and one lava type from this environment, boninite, was named for the islands.
Boninites have a distinctive mineralogy and chemistry that requires their initial crystallization from melts at high temperatures (1,250 |degrees~ to 1,300 |degrees~ C) and shallow depths (3 to 10 kilometers)--conditions similar to spreading centers. However, boninites contain the chemical imprint of a source region invaded by fluids derived from a subduction zone (for example, they contain 10 times more water than mid-ocean ridge lavas).
Geologists have recognized many of these unusual characteristics in groups of seafloor rocks now exposed on land, called ophiolites. For many years, ophiolites were considered our best analogue of mid-ocean ridge crust. Their chemical distinctions from mid-ocean rocks were largely ignored, because they provided useful insights into the magmatic and structural processes thought to occur at spreading centers. Collecting the evidence and convincing the entrenched scientific hierarchy that such rock sequences typically form in a different tectonic environment is an ongoing battle. Many will now agree that most ophiolites formed above a subduction zone, but they still assume that it was at a (back-arc) spreading center. The concept that some might have formed in a broad immature arc terrain is hotly debated. The marine geoscience community also debates exactly how this might have occurred. All the models have problems, partly because some of the pieces of the puzzle (such as what was being subducted) have been lost, and there is no equivalent setting active today. One of the paradoxical inferences, however, is that there is a type of subduction initiation that does not cause compression of the upper plate, but, rather, extension and volcanism.
A Variety of Volcanoes
Ten years ago, the only detailed bathymetric maps of the western Pacific were classified. Although a few academics had clearance to see parts of the data, it was on a need-to-know basis and couldn't be discussed with colleagues. As small pieces of this data were released and as the scientific community conducted its own wide-swath mapping surveys, we discovered that not only are there many more submarine volcanoes than previously thought, but in some areas there are also whole new classes of volcanoes in addition to the main line that parallels the trench. One class forms lines of volcanoes in the back-arc, oblique to the main line. Others occur in the forearc, close to the trench. A series of geophysical mapping and dredge-sampling expeditions, Alvin dives, and ocean drilling investigations were made in the Mariana and Izu-Bonin arcs to discover more about what formed each of these two groups of volcanoes. The oblique lines of volcanoes are composed of lavas and volcanic sediments similar to those of the main line. Their distribution appears to be controlled by weaknesses in the upper plate, often inherited from the previous stretching of the arc, that allow arc magmas to erupt into the back-arc region. Nevertheless, we still know very little about the age, composition, and structural control of most of these volcanoes; there are hundreds awaiting study. The forearc volcanoes were an even greater surprise: most were not formed by lavas, but by green serpentine muds!
Serpentine Volcanoes and Diapirs
The forearc seamounts (mountains on the seafloor) are 5 to 30 kilometers across, 0.5 to 2 kilometers high, and occur within 100 kilometers of the trench. Most don't have the strong magnetic and gravity signal typical of volcanic seamounts; rocks dredged from them are composed primarily of the mineral serpentine. Serpentine is formed by the addition of water from the subducting plate to olivine in the peridotite mantle rock of the upper plate. This hydrated material, being less dense, is buoyant and slowly rises, often along faults in the overlying rocks. If these rising diapirs (upwardly mobile rock masses) reach the surface, they form seamounts with a range of characteristics. The finer-grained material with higher fluid content erupts onto the seafloor through a central conduit and forms a volcano of serpentine mud. The more massive, less mobile, material protrudes onto the seafloor like rising dough overflowing a small pan.
Fluids seeping through the surface of the serpentine volcanoes form chimneys up to 2.5 meters high, mainly composed of carbonate minerals. Near the surface the fluids are mixed with seawater, but their undiluted chemistry about 100 meters down into the volcanoes shows their subducted-plate origin. They contain light hydrocarbons (ethane and propane) and organic acids, as well as aromatic compounds (benzene and toluene) in fluid inclusions, that were produced by thermal maturation (baking) of organic matter subducted in sediments. The fluids are very alkaline, with a pH up to 12.6. Low strontium isotope ratios indicate that some of the fluids come from the crustal rocks beneath the subducted sediments.
Serpentinized peridotite seamounts on the inner-trench slope indicate the presence of mantle, and therefore a thin crust, in the outer forearc above the subducting plate. Some crustal rocks are entrained into the serpentine diapirs as they rise, resulting in a mixed sample of rocks from all crustal levels. Most of these are arc rocks from the initial (40- to 50-million-year-old) stage of volcanism, however, some 100-million year-old mid-ocean ridge rocks have also been found in the Mariana forearc, both in the diapir samples and exposed on large-offset faults. These rocks require that some fragments of oceanic crust from the subducting plate have been added to the Mariana forearc.
The trenchward boundary of arc volcanism is termed the volcanic front. The main chain of island arc volcanoes lies along this front. Although some volcanoes occur behind this line (in the back arc) none are supposed to occur under normal circumstances in front of it (in the forearc, closer to the trench). The few exceptions to this rule have been associated with conditions of unusually high temperature gradients, produced by either the subduction of a spreading center or by "back-arc" basin formation in the forearc. None of these exceptional circumstances have influenced the Izu-Bonin, Mariana, or Tonga arc-trench systems in at least the last 25 million years. Very old (more than 100-million-year-old), cold, Pacific crust is subducted along all three trenches. Nevertheless, multichannel seismic data indicate the presence of lava flows or sills in the Izu-Bonin forearc. In fact, recent drilling found young arc lavas in the forearcs of all three systems; the youngest, in the Marianas, is 1.7 million years old.
This leaves us in quite a quandary. First, the volcanic front, one of our fundamental reference points for other features (such as "back-arc"), is no longer a simple boundary. It can still be defined by the amount, but not the presence or absence, of arc volcanism. For example, arguing that material has been removed from a forearc just because some arc lavas are too close to the trench is no longer viable. Nor is postulating a changed geometry of subduction in the geologic past just because a few lavas are found on the "wrong" side of the arc line. Second, our models for the genesis of arc magmas may be in need of serious revision. The experimentally determined solidus temperature (above which melting begins) of mantle rock under water-saturated conditions is about 1,000 |degrees~ C. Nowhere in the forearc wedge above the subducted plate are temperatures close to these reached--measurements at three forearcs drilled by the Ocean Drilling Program in 1989 and 1991 ranged only between 0.5 heat-flow units near the trench and 1.3 heat-flow units closer to the arc. There seem to be only two areas hot enough to generate arc magmas: beneath the volcanic front, or deep in the mantle of the subducted plate. Magmas from the volcanic front would have to move laterally 80 to 120 kilometers, perhaps as near-horizontal intrusions (sills), to reach the forearc sites of eruption. The alternative is that fluids circulate down to the hot mantle of the subducting plate where they might generate magmas that could ascend vertically. Both scenarios seem unlikely, but given the inferred low temperatures beneath the forearc we haven't yet come up with a better explanation.
Arc Rifting and Back-Arc Spreading
When arcs are stretched they split, initially forming a rift (an elongate depression bounded by high-angle faults) and eventually a back-arc spreading center. The processes involved in arc rifting are similar to those involved in splitting continents, with the important difference of the presence of nearby arc volcanism. The causes of arc stretching are not fully understood, but we know some of the parameters. At a subduction zone the upper plate is horizontally coupled by suction to the downgoing plate. Most trenches migrate seawards as the subducting plate sinks, resulting in the upper plate being pulled seawards and stretched, depending on its other boundary conditions. Periodically, the stretching is sufficient to pull the upper plate apart. The weakest area of the upper plate is near the volcanic front, where the plate is both hottest and has the thickest crust (the stronger mantle is thinnest). The arc splits within about 50 kilometers of the volcanic front.
The resulting rift basins initially subside along a zigzag pattern (in plan view) of border faults. The rifts are often better developed between the arc volcanoes; magmatism, rather than stretching and subsidence, often fills the opening adjacent to arc volcanoes. The basins are rapidly filled by sediments shed from nearby volcanoes and are also intruded by magmas along faults. The magmas may be derived from the deep-mantle arc source or from pressure-release partial melting of the shallow mantle. The basins widen by continued stretching as well as by collapse of the weak arc margin. As they do, the zone of greatest subsidence and intrusion migrates with the retreating arc margin border faults. The result is an asymmetric basin in cross section, with a wider zone of stretched and intruded crust on the side away from the active arc. Activity along the volcanic front is often diminished, and sometimes ceases, as the frontal arc magma sources are bled off into the back-arc basin.
After continued stretching for 5 to 10 million years has produced a rift basin some 200 kilometers wide, the shallow and deep sources of mantle partial melts become horizontally separated. An organized spreading center, fed by the shallow mantle, forms in the back-arc basin, and the deep mantle arc magmas establish a new volcanic front along the rifted edge of the old one. This evolution is spatially transgressive; spreading is localized in some areas first and then propagates into adjacent rifting areas. Several variations on this model occur, depending on the interplay between the location, timing, and volume of back-arc and arc volcanism and the resulting strength of each region.
Volcanic-Hosted Massive Sulfide Deposits
A subduction signature distinguishes the sediments, lavas, and sulfide deposits of island arcs and back-arc basins from their mid-ocean equivalents. This same signature is associated with many ore deposits and their ophiolite host rocks, such as those mined in Cyprus, Oman, and the Kuroko district of Japan, suggesting that they formed in an arc/back-arc, rather than a mid-ocean, setting. Subduction-related volcanic rocks, including silica-rich lavas or intrusions, are common to all sites of major volcanic-hosted massive sulfide (VMS) ore bodies. Thus, studies of hydrothermal circulation and sulfide deposition at mid-ocean ridges cannot provide a comprehensive model of this important type of ore generation.
Therefore, Australia, Britain, Canada, France, Germany, Japan, Russia, and the United States have all sent expeditions to the western Pacific in the last five years to study arc/back-arc hydrothermal systems, and with great success. Dive programs, using submersibles from four nations, found active vents in all the regions investigated. The range of vent systems and sulfide types is much greater than at mid-ocean ridges, owing to the organic-rich volcanic sediments, and the silica- and volatile-rich rocks present in these environments. Modern Kuroko-type ore formation was discovered at depths of 1,300 to 1,600 meters in both the Okinawa and Izu-Bonin back-arc rifts. Silica, sulfate, and sulfide chimneys emitting clear, white, and black solutions at temperatures of 200 |degrees~ to 400 |degrees~ C were found on all the active back-arc spreading centers studied (Lau, North Fiji, Manus, and Mariana). The back-arc vent fluids have extremely low pH (2 to 5), and typically contain more zinc, barium, lead, cadmium, and arsenic than mid-ocean ridge vent fluids. The search for modern analogues of VMS ore bodies is over. The exploration and analysis of their diversity has just begun.
Fluids in Accretionary Prisms
Thick sediments subducted at a trench respond like a wet sponge fed through a wringer; their porosity decreases and most of the fluids that form 50 to 70 percent of their volume are expelled. This occurs in at least two stages: initially, at the toe of the slope where the upper sediments are scraped off and, subsequently, at depth beneath the forearc where many of the lower sediments are underplated. The fluids flow through and out of the accretionary prism in several ways. Pervasive dewatering by diffuse flow may dominate in the toe of the prism, as indicated by data from drilling at the Nankai Trench. Focused fluid flow occurs along permeable sedimentary layers and faults. Flow along fracture systems becomes dominant as the sediments are consolidated and cemented. Diapirs of mud and shale may also pierce the prism.
Only in the last five years have the extent and effects of this fluid flux been recognized. Biological communities consisting of macroscopic organisms living symbiotically with bacteria that derive their energy chemically are dramatic markers of fluid seeps. The near-surface precipitation of carbonates and the formation of gas hydrates indicate substantial movement of carbon as methane and carbon dioxide, derived from the breakdown of organic matter. Silica is dissolved from strongly cleaved rocks and later precipitated in quartz veins. Fluids warmer and fresher than seawater have been detected moving laterally up from deep in the accretionary prism where they probably formed by the dehydration of clay minerals. Fluid pressures control the style of deformation and the shape of the accretionary prism. Episodic fluid flow is likely linked to episodic faulting, and hence to the generation of the largest earthquakes and tsunamis on Earth.
We have only begun to characterize the flux of solids and fluids through a subduction system. We have order-of-magnitude estimates of the volumes, chemistry, and physical properties of the incoming materials. However, the changes in physical properties of this material and its partitioning between return flow to the oceans, addition to the crust, and absorption by the mantle is poorly known. The chemistries of the filtered fluids that escape to the ocean are extremely variable, ranging in pH from 2 to 12, for example. The mass flux of fluids of differing chemistries, and hence the chemistry of the residue crustal and mantle filters, are little constrained. So too are the fluid pressures and permeabilities in the rocks, which control the style of deformation and the earthquake cycle. Both the scale of temporal variability and the long-term averages need quantifying. One estimate of fluid flux through subduction zones suggests that they recycle the global ocean volume slowly, about once every 500 million years. Nevertheless, they may form a significant link in the global carbon cycle, transferring most of the carbon in subducted organic matter back to the ocean.
Current models hold that continental crust grows by accretion of island arcs to continental margins. Yet little net growth of the terrestrial mass appears to have occurred in the last several hundred million years. The estimated volume of terrestrial material subducted into the mantle as unaccreted sediment and collapsed upper plate rocks balances the estimated volume of mantle melts added to arc crust (both about 1.6 cubic kilometers each year).
Our present understanding of trench/arc/back-arc systems is only good for the near surface and the recent past. The plumbing systems of volcanoes (of magma or mud) and their mineral deposits, the fault zones that generate major earthquakes, and the deforming lower crust of forearcs are critical areas that we have yet to accurately image or effectively monitor. Future marine geoscience studies, including three-dimensional seismic imaging, seafloor observatories, and deep drilling will investigate these sub-seafloor areas that shape planet Earth.
Brian Taylor is a Professor in the Department of Geology and Geophysics in the School of Ocean and Earth Science and Technology at the University of Hawaii. He is an Australian, married to a New Yorker, with a daughter born in Hawaii. Brian coordinated the US-Japan program of Alvin dives in the Northwest Pacific and chaired the Western Pacific Panel of the Ocean Drilling Program. His research focuses on the tectonics, magmatism, and sedimentation of the circum-Pacific trench/arc/back-arc margins.