Table of Contents

1. Executive Summary
2. What is the Subduction Factory?
3. What Do We Need to Know about the Subduction Factory?
   1. Subduction Parameters as Forcing Functions on Factory Output
   2. The Volatile Cycle through the Subduction Factory
   3. Towards Mass Balance
4. How Will We Study the Subduction Factory?
   1. The MARGINS Philosophy
   2. Strategy for Implementation of the Subduction Factory Initiative
   3. Field, Laboratory, and Experimental Efforts
      1. Bathymetry and Swath Mapping
      2. Seismic Methods
      3. Magnetotellurics
      4. Heatflow
      5. Submarine Sampling Strategies
      6. Geodesy
      7. Studies of Magmatic Arcs
      8. Experimental Studies
      9. Geodynamic Modeling
      10. Databases
5. Where Will We Study the Subduction Factory
   1. Focus Experiments
   2. Allied Studies
   3. Theoretical & Experimental Institute: "Inside the Subduction Factory"
6. How Long Will It Take to Study the SubFac?
7. How Much Will SubFac Cost?
8. How Will We Communicate Results and Opportunities for Cooperation
9. How Will International Cooperation be Established?
10. How Will SubFac be Evaluated
11. Selected References
12. List of Figures

I. Executive Summary

Subduction of oceanic plates causes earthquakes, tsunamis and explosive volcanism. Subduction also gives rise to beneficial products, such as ore deposits, geothermal energy and the very ground we live on. The Subduction Factory recycles raw materials from the subducting seafloor and overlying mantle, and creates products on the upper plate in the form of melts, aqueous fluids and gases. The Subduction Factory Initiative aims to study fluxes through the subduction zone to address three fundamental science themes: 1) How do forcing functions such as convergence rate and upper plate thickness regulate production of magma and fluid from the Subduction Factory? 2) How does the volatile cycle (H₂O and CO₂) impact biological, physical and chemical processes from trench to deep mantle? 3) What is the mass balance of chemical species and material across Subduction Factory, and how does this balance affect continental growth and evolution?

The Subduction Factory Initiative will proceed by focused investigations combining swath mapping of the incoming plate and fore-arc slope with both active and passive seismic experiments to image accretionary and slab structures, respectively. Heatflow measurements, magnetotelluric investigations and GPS plate and deformation rate estimates will combine with the other geophysical data to constrain the physical operation of the subduction system. Riserless drilling will provide samples of the input material seaward of the trench and output material in the forearc and arc. Riser drilling would permit deeper holes into the altered incoming crust, and riser or on-land drilling into the arc would sample a record of volcanic evolution and fluxes on the upper plate. Boreholes will be exploited to sample fluid outputs from the system. Field and analytical studies of the arc system will focus on the chemical composition and mass fluxes of lavas, melt inclusions and gases. Laboratory studies will provide element partitioning relationships, phase equilibria, and calibrations for rheological and seismological properties. A wide array of in situ observatories and multiple re-occupation campaigns, coupled with a strategy for rapidly responding to major events, round out the data collection strategy.

These diverse field and lab measurements will be integrated at every level with physico-chemical models for subduction, fluid flow, melting and melt flow. Phenomena predicted from geodynamic models will guide the early data acquisition efforts, and the data collected will provide constraints for further generations of models. In this way, modeling and observations will complement and propel each other.

Criteria for selection of subduction zones to be studied include the following: the margin should provide ample volcanic and seismic activity, accessibility to both input and output, along-strike variations in forcing functions, cross-arc and historical perspectives, minimal upper plate contamination of magmas, and ability to address the primary science objectives. These criteria are best met by studying two convergent margins, Central America and one of the intra-oceanic convergent margins of the Western Pacific.

Central America is a high priority location because it satisfies the criteria and provides excellent opportunities to address all of the science themes. Forcing functions and volcanic response vary systematically and dramatically along-strike from Nicaragua to Costa Rica. Extensive carbonate subduction and extremely water-rich eruptions enable unparalleled investigation of the carbon and water cycles through subduction zones. Lower crustal exposures and high-fidelity tracer studies will pave the way to element and mass balance. Many of the Subduction Factory objectives link very naturally with those of the SEIZE science plan in Central America.

Western Pacific margins provide ideal counterpoints to Central America. The slab subducting beneath Central America is relatively young, and parts of the fore-arc are sedimented. Central America offers excellent along-strike variations, but a weak cross-arc perspective. Western Pacific arcs are complementary, being characterized by old/cold slabs, carbonate-absent sediment subduction, and accessible outputs from fore-arc to back-arc. A choice between the front running candidates (Tonga, Izu-Bonin or Marianas) will be decided in future workshops.

The Subduction Factory Initiative will extend over ten years, with earlier geological and geophysical field programs and theoretical institutes paving the way for later drilling, arc refraction, slab seismic and geochemical efforts. A fully integrated study of the Subduction Factory will cost $15-20M, excluding shiptime and drilling. Extensive international cooperation will distribute some of these costs over a number of nations. A web site listing on-going programs will attract other synergistic projects in the same area; the web site will post data before formal publication. Results will also be communicated through international meetings and workshops.

II. What is the Subduction Factory?

The rumblings and emissions at convergent margins reveal the inner workings of the Subduction Factory (Fig. 1). The term Subduction Factory is used to encompass the fluxes of material into and out of subduction zones, together with the thermal, chemical and mechanical processes that shape convergent plate boundaries, the deep mantle beyond, and the air and water above. Raw materials -- seafloor sediments, oceanic crust, and mantle lithosphere -- are fed into the Subduction Factory at deep sea trenches. In the wedge above the slab, subducted materials are mixed with mantle, supplied by convection from the landward side of the arc. Output products - melts, aqueous fluids, metalliferous deposits, serpentine diapirs, volcanics, continental crust, gases, organic material, back-arc seafloor -- emerge from the Factory on the upper plate. The remainder of the material that is processed in the Subduction Factory sinks deep into the mantle, someday to be resurrected as mantle plumes. The Subduction Factory is thus powerful but well-hidden. We can examine its raw materials and its products, but the Factory itself is hidden from view.

Figure 1

Subduction of oceanic plates triggers a wide array of scientifically and societally important processes. It impacts society directly because it causes earthquakes and explosive volcanism, and whereas earthquakes (and the tsunamis they spawn) kill more people, explosive volcanism can change climate, potentially affecting the global population. Most of the world's important ore deposits and continental crust -the very ground we live on - have been formed in the past by the factory. An important potential new form of energy -- gas hydrates -- are generated by the factory.

Figure 2

From a scientific perspective, the Subduction Factory is central to the operation and evolution of the Earth System. Subduction of pore fluids and hydrous minerals in oceanic sediments and altered basalt, their distillation at depth, transport through the mantle, and re-emission from arcs represents earth’s deepest hydrologic cycle, one which has a profound impact on the global budgets of volatiles such as H₂O and CO₂. The return of subducted fluids and gases such as methane to the surface supports chemosynthetic biota, affects seawater chemistry, partially controls prism deformation, and serpentinizes shallow lithospheric mantle. Subducted ocean crust and sediments contribute to the chemistry of arc and some back-arc volcanoes, which contribute to crustal growth and provide a probe of physical and chemical processes operating deeper than can be drilled or imaged seismically. Subducted materials not returned to the surface by the Factory are carried into the deep mantle, where they alter its chemistry and rheology. And despite the central role for subduction in the evolution of the Earth and the fact that ours is the only planet where plate tectonics occurs, how subduction begins is understood in only the broadest terms.

The Subduction Factory is the dynamic site of mass and energy exchange between the asthenosphere, lithosphere, hydrosphere, atmosphere, and biosphere, with profound implications for the evolution of the Earth's surface and interior. It is huge, operates at depth, and involves complex physical and chemical interactions, the resolution of which requires close co-operation between scientists who do not usually work together. Thus, it has been difficult to investigate processes and measure fluxes through the factory owing to the sheer scale of the problem and the poor constraints on volumes of magmas, aqueous fluids and volatiles produced. The MARGINs approach is to focus an interdisciplinary study at convergent boundaries where geological and geophysical measurements promise to constrain processes within the Subduction Factory in real time.

III. What Do We Need to Know about the Subduction Factory?

Studies of subduction zones attract a wide range of scientists because the questions are globally significant. A number of key scientific themes to be addressed at subduction zones have been identified at various MARGIN workshops, the NSF FUMAGES workshop, the CONCORD conference on riser drilling and the Avalon JOI/USSAC workshop on Crustal Recycling: (1) How, why and where are new subduction zones started? (2) How are volatiles cycled through the subduction system? (3) What is the rate and mechanism of continental growth at convergent margins? (4) What is the impact of subduction on mantle evolution? (5) How does subduction lead to uni-directional changes in the composition of the continental crust? Answers to all these questions are the ultimate goal of Subduction Factory studies. Here we focus on a subset of these first-order questions that are increasingly tractable now, or that are necessary to pave the way for subsequent high priority science. At the Subduction Factory Workshop in June 1998, participants recognized the following three areas as critical for further progress.

1. Subduction Parameters as Forcing Functions on Factory Output

Sinking lithosphere powers the Subduction Factory, stirring the overriding mantle and bringing in mantle hot enough to melt, while also delivering ingredients essential for continental crust formation. The rate and angle of subduction and the physical and chemical properties of the subducting plate, such as its thermal structure, alteration profile, sediment load and volatile content are all likely to affect the type and quantity of Subduction Factory products. There are as many as 26 different physical parameters which vary among the world's 39 subduction zones. We still do not understand how the many independent and dependent variables control the factory output. Neither do we understand how these parameters affect intermediate and deep seismicity in subduction zones. Assessing the role of these various "forcing functions" is an important part of the Subduction Factory initiative. Along-strike variations in forcing functions within a single margin provide an efficient way to study cause and effect in the Subduction Factory. An alternative is to investigate paired margins with contrasting forcing functions.

Convergence Vectors: The behavior of the subducted lithosphere can be described as a vector, defining convergence rate and dip. The convergence rate should control the rate at which many processes operate within the Subduction Factory. The most obvious connection is with input parameters, such as the flux of material and volatiles in the subducting plate delivered to the factory. Other important processes that should be simply related to convergence rates include rates of induced convection in the mantle wedge, shear heating along the slab-mantle interface, conductive heating of the subducted slab, and seismic moment.” A looming question is whether faster convergence leads to faster growth of the arc crust. The existing growth rate estimates do not support such an connection, but they are also poorly known. In order to examine subduction rate as a forcing function, we need new approaches for measuring melt production and arc growth rates, and for using lava compositions to constrain thermal models. We also need an increasingly realistic geodynamic picture of convergent margins that includes dynamic rather than kinematic models, and a rheology structure that incorporates the effects of both temperature and volatiles.

The dip of the subducted slab defines the path-length of the slab from the trench to beneath the arc. Some theoretical models also predict different mantle and fluid flow fields associated with different subduction angles. Such models should be tested by comparing their predicted behavior with lava compositions and gravity and heatflow data from well-characterized arcs.

Slab Temperature: In addition to the convergence vector, the other major control on the thermal structure of the subduction zone is the age of the subducted lithosphere.” This is because conduction - the least efficient mode of heat transfer - largely controls heating of the subducted lithosphere. Old lithospheres are thick and cold through their upper part, leading to development of an entirely different thermal structure than young lithospheres, which are thin and hot (Fig. 3). Existing models predict that such different thermal structures will cause different loci in the slab for important metamorphic reactions, directly affecting fluid flow through the Subduction Factory. We know, for example, that hotter slabs lose most of their fluid-mobile elements (e.g., B, Cs, Sb) before they can be delivered to the melt generation zone, but it is unknown what metamorphic reactions control this distillation of elements out of the slab. We also don’t know how water behaves during this distillation, and if mantle melts are different above slabs of different ages. Very young slabs may melt and produce distinctive lavas known as adakites, but we do not know what thermal thresholds must be crossed before slab melting occurs. Since mantle and slab temperatures are central to the subduction factory output, it is desirable to study arcs with a range of parameters critical to these temperatures. A combination of geochemical tracer studies, slab metamorphic studies, thermal measurements and modeling, and seismic inversion techniques are needed to understand how slab thermal structure affects Subduction Factory operations.

Figure 3

Subduction Dynamics and Mass Transport to Depth: The crustal inputs to the Subduction Factory are another clear factor in controlling the mass, composition and distribution of outputs. The crustal inputs, in turn, depend not only on the supply to the trench, but also the dynamic processes that occur during subduction. Sediments may be bulldozed from the downgoing plate to form accretionary wedges, underplated beneath the fore-arc, subducted to the depths of magma generation or even joined by older material eroded from the fore-arc. The behavior of material through the upper 40 km of the subduction zone is intimately linked to the nature of the incoming sediment and rock sequence, its compaction dewatering, diagenesis and cementation, fore-arc deformation, and the nature of the seismogenic zone. Understanding all of these processes are objectives of both the Subduction Factory and SEIZE initiatives. The balance between accretion and subduction of sediments is also essential to resolving whether the continents are growing or shrinking, and to determining the flux of sediment-hosted chemical species into the mantle and back out the arc.

Many tools are required to estimate the material balance across the convergent margin. Seismic imaging can reveal the presence of a wedge-shaped sediment pile, or underplated sediment packets, but constrains neither the age of the sediments nor their source. Drilling and subsequent analyses can show if fore-arc sediment wedges are paleo-accretionary prisms, deformed piles of arc-derived sediments, or imbricate thrust packets of offscraped sediments. Neither technique can sample or image deeply enough, however, to constrain the full extent of accretionary prism dynamics, even with riser drilling. To investigate processes at greater depths, geochemical "imaging" is helpful. For example, only the youngest part of the subducting sediment column (<8-10Ma) contains ¹⁰Be, and so high ¹⁰Be concentrations in arc volcanoes indicate sediments subducted to depths appropriate for melt generation (Fig. 4). It is also possible to infer the partitioning of ¹⁰Be between frontally accreted and subducted sediment. If a discrepancy exists between what issues from arc volcanoes and what is thought to be subducted, then the geochemistry, drilling and seismic imaging may be used together to infer underplating or forearc erosion. In this way, volcanoes become flow monitors for material subducted past the seismogenic zone, into the Subduction Factory and beyond.

Figure 4

Upper Plate Thickness: The thickness of the overlying plate, including both lithosphere and crust, is another forcing function because it affects asthenospheric flow in the mantle wedge and its thermal regime. The overlying plate also controls the height of the mantle melting column beneath the arc, and so limits the amount of melting that can occur through decompression (Fig. 5). Arc lavas are consistently the most fractionated and petrographically complex on earth. This must be in part a reflection of the structure of overriding lithosphere. Aside from crustal thickness variations, we know little about the thickness of the upper plate, mostly because traditional seismic methods do not resolve well lithospheric thickness at convergent margins. This partly reflects the complex rheologies and thermal structures expected. For example, lithosphere is typically cold but strong, while asthenosphere is relatively hot but weak. The forearc mantle, however, may consist largely of serpentine, which is both cold and weak and so is not well described by this terminology. It may be more useful for understanding convergent margins to characterize the asthenosphere as the convecting portion of the mantle. Given these complications, how can we best map the boundary between the asthenosphere and lithosphere in the mantle wedge?

Figure 5

2. The Volatile Cycle through the Subduction Factory

A major goal of the Subduction Factory is to understand the deep water cycle of the blue planet and the role of subduction on Earth’s natural carbon cycle. Water and to a lesser extent CO₂ control the physical and chemical behavior of subduction. The effects of water on deformation, development of the décollement and behavior of the seismogenic zone in the 0-40 km depth interval are discussed extensively in the SEIZE Science Plan. Subduction Factory efforts will complement those of SEIZE at shallow depths, and extend to greater depths along the slab, to the arc and back-arc melting regimes, and to the deep mantle.

What is the distribution of water and CO₂-bearing alteration phases in the incoming ocean plate? The proportion of volatiles delivered to the subduction factory from the igneous slab is poorly known, but is expected to be larger than that in the sedimentary veneer when considering bound volatiles subducted to sub-arc depths. Paired CORKed sites on ocean ridge flanks reveal shallow sea floor hydrology, which can be combined with petrological and seismological studies to better investigate alteration and volatile budgets of the oceanic crust. ODP sites in exposed oceanic peridotites indicate that seawater penetrates to great depth. Heat flow and pore water chemistry from ODP sites outboard of some trenches indicate that seawater circulates to basement, presumably along fractures reactivated as the slab bends into the trench. Understanding aging of the oceanic crust, in general, is critical for reconstructing the volatile cycle at convergent margins. In particular, a focused experiment must include good heat flow surveys and drilling at least 300 meters into oceanic basement (the upper oxidative alteration zone) at more than one locality outboard of the trench.

Compaction in the uppermost part of the subduction zone wrings water that is trapped in pores and fractures from the slab. More water is released as minerals in the slab breakdown and reform in response to increasing pressure. These reactions also add selected elements, including hydrocarbons, to the water making its way back to the surface along faults and through diffuse fluid flow. Hydrological and geochemical studies of aqueous fluids venting in the fore-arc are critical for investigating gas hydrate composition and stability on convergent margins, the deep biosphere, the distribution of chemosynthetic vent communities and deformation within the seismogenic zone. Such studies are also essential to constrain the fluxes subducted to greater depth. Experimental studies of element partitioning between aqueous fluids and solid phases in the slab at low P and T are especially critical if we are to use aqueous fluids to interpret conditions occurring at depth. We also need dehydration experiments on natural mineral mixtures, including clay-rich, carbonate-rich, silica-rich, and carbonaceous sediments

What is the extent of fore-arc serpentinization? While the serpentinization of the shallow fore-arc mantle may be critical in controlling slip behavior across the seismogenic zone, it is also critical for material processing through the Subduction Factory. Serpentinite bodies are exposed across a wide section of the Izu-Bonin and Mariana forearcs, and represent a major sink for water distilled out of the slab (Fig. 6). Any effort to quantify the flux of water delivered to the depths of magma generation will need to account for the volatile flux out of the slab to the overlying serpentinized mantle. This leads to several key questions. Is sub-surface serpentinization typical of all arcs, but imaged and sampled easily only in sediment-starved and structurally distressed margins? Can laboratory calibration of P and S wave velocities for serpentinite, amphibolite and tonalite lead to seismic methods for determining the subsurface distribution of serpentinite? The serpentinites, and the aqueous fluids they host, record the volatile and chemical losses from the slab at about 20-50 km. They thus provide an important constraint on the effects of shallow subduction processes on the composition of the slab as it descends to greater depths. The effects of subduction on the shallow lithospheric mantle may be as profound as on the deeper mantle downstream of the volcanic arc.

Figure 6

What is the effect of subducted volatiles on mantle seismic velocity and viscosity, slab embrittlement, and intermediate depth earthquakes? Subducting slabs acts as heat sinks for the overlying mantle, and cool the adjacent mantle. The sub-arc asthenospheric mantle thus has an unusual thermal structure: it is hottest in the middle of the convecting mantle wedge and cools towards both the overlying lithosphere and downgoing slab. P-wave tomography across Japan shows an inclined layer parallel to and just above the subducting slab at about 75 to >150 km depth, which is lower velocity than the slab, but higher velocity than the shallower parts of the wedge (Fig. 7). Convection models that use a temperature-dependent viscosity structure show a higher viscosity layer in this cooled mantle, creating a halo of cold, stiff mantle that couples effectively to the down going slab. But what is the effect, if any, of volatiles from the slab on the seismic properties and viscosity structure of the mantle wedge? Identifying either the presence or absence of a volatile signature on the physical properties of the mantle wedge would be extremely useful if this information could be translated to limits on volatile form (hydrous minerals, free aqueous fluids), concentration or distribution. Realistic experiments, particularly those that examine the combined effects of temperature, melt, and volatile distribution are difficult, but essential.

Figure 7

The Seismogenic Zone initiative focuses on earthquakes occurring shallower than about 50 km. A significant fraction of seismic energy at convergent margins, however, is released in deeper events that occur within the subducting plate. Intermediate depth earthquakes, between about 50 and 350 km, often appear to be located near the top of the subducting plate. Are these earthquakes due to slab embrittlement during prograde metamorphism and dehydration of the altered oceanic crust? If so, then the frequency and depth distribution of intermediate depth earthquakes in subduction zones with different thermal parameters provides important clues about slab metamorphism, dewatering and rheology beneath, and also deeper than, the volcanic arc.

What is the stability of key hydrous and calcareous phases in the subducting slab and mantle wedge? Most existing models of arc magmagenesis emphasize the role of amphibole dehydration in the subducting basaltic crust and of amphibole and phlogopite stability in the overlying mantle. Recent experimental work, however, has revealed a menagerie of minerals that are stable in sediments, altered basalt and mantle peridotite compositions to relevant pressures and temperatures. Minerals such as phengite, lawsonite, aragonite, zoisite and chloritoid may be hosts for water and CO₂ in the subducting slab. It is critical that we understand the stability of phases in real systems during prograde metamorphism, as well as the partitioning of elements between these phases. In additional to laboratory experiments, seismic methods may also help to reveal the mineral reactions occurring in subduction slabs (Fig. 8).

Figure 8a

Figure 8b

What is the role of water in arc magma generation and volcanic explosivity? Of all the volatile species, water most affects the mantle solidus. It is clear that arc lavas are richer in water than lavas from other tectonic settings, and that water's depression of the mantle solidus abets melt generation in the mantle wedge. The recent discovery of water-poor (but non-degassed) arc magmas, however, means that melting is sometimes anhydrous, probably driven by decompression melting as in other tectonic settings. This raises questions as to the different roles of water and decompression in driving mantle melting in the subduction factory. Further analytical studies of the intrinsic water content of arc magmas, combined with further experimental studies of the effect of water on peridotite melting, are needed to better understand the role of fluids and mantle flow in arc magma generation and crustal growth.

In addition to melting in the mantle, water also affects magmatic evolution in the crust and the explosivity of volcanic eruptions. Because the solubility of water in melts decreases rapidly at pressures below 1-2 kilobars, much water may be lost as melts ascend. This leads to rapid crystallization of minerals and further degassing. At some point, the crystallizing melt is unable to release its water peacefully, leading to violent eruptions. Violent eruptions severely impact nearby populations, and hazard mitigation requires understanding the links between melt chemistry, dissolved water, and how melt ascent and cooling affects degassing. Direct correlations have been found between water content and explosivity. Further analytical studies, along with studies of the dynamics of magma degassing, are needed to develop models that show how magmatic water controls shallow fractionation and explosive eruptions.

How is CO₂ recycled in subduction zones? Arc magmas are clearly enriched in CO₂/³He relative to mid-ocean ridge basalts and ocean island basalts, and more than 80% of the CO₂ in arc magmas may be derived from the subducting slab. CO₂ released from arcs is a major return flux of subducted CO₂ to the atmosphere, comparable to the ocean ridge flux, and as such, is a potential driver of intermediate-term climate fluctuation. The mass of carbonate and organic material subducted, however, is extremely variable among convergent margins. Do arc magmas contain more CO₂ where more sedimentary carbon is subducting? How much volcanic CO₂ is derived from carbonate subducted as veins in the oceanic crust? Where do decarbonation reactions happen in the slab? What proportion of the subducted carbonate is recycled into the deep mantle? In order to answer these fundamental questions, integrated studies are needed of volcanic gases and melt inclusions, CO₂ solubility and degassing, carbonate metamorphism, and carbon budgets in the subducting plate.

What is the role of subducted volatiles and trace metals in ore-forming processes at convergent margins? Hydrothermal activity and ore-formation have been observed in the Kermadec, Izu-Bonin, Tonga, Mariana, and Bismarck arc systems. Isotopes and trace elements indicate that a significant fraction of the ore-forming fluids and the metals they carry have been exsolved from volatile-rich arc magmas rather than leached by seawater during hydrothermal circulation through the crust, as for ocean ridge deposits. These ore deposits thus represent a little known aspect of the mass and element fluxes out of the subduction factory. They also provide a unique window into economically significant ore-forming processes. Many world class ore deposits from the Tertiary through the Archean (e.g. Kuroko, Noranda, Sulfur Springs) are hosted by felsic volcanics that may have formed in a convergent margin setting.

3. Towards Mass Balance

An ultimate goal of Subduction Factory research is to understand how subduction builds the continents and affects mantle composition through time. A quantitative mass and element balance through the Subduction Factory would achieve this goal. To realize mass balance, however, we need to better understand how energy and matter move through the Subduction Factory. The greatest uncertainty is in the estimates of material output rates, such as fluid fluxes to the forearc and magma fluxes to the arc crust. In particular, there is a critical need to know the volumes and compositions of middle and lower arc crust. Crustal growth and mantle evolution models also rely, in part, on correct interpretation of the geochemical and petrological signatures of arc lavas. Many studies show that chemical components are fractionated from each other during distillation from the slab and transport through the mantle to the site of melt generation. Using element fluxes to obtain mass fluxes requires a better understanding of element partitioning than currently available. Thus the route to mass balance is paved through studies of lower and middle arc crust and experimental and theoretical studies of geochemical tracers.

What are the volume and production rates of middle and lower arc crust? Volumes and production rates of arc crust are critical for determining the fluxes out of the Subduction Factory, as well as for understanding how the continents grow. Estimating volcanic volumes is relatively straightforward, and requires integrating the volume of the volcanic edifice and surrounding volcaniclastic apron. On the other hand, the intrusive contribution to the arc is obtained from crude estimates of proportionality to the extrusive volume. Recent seismic refraction studies, with improved resolution, provide some new constraints. The seismic structure of the Izu-Bonin arc includes a mid-lower crustal layer with Vp=6 km/sec, which, based on seismic properties as well as exposures of correlative rocks in the Tanzawa Mountains of Japan, may consist largely of tonalite (Fig. 9). Seismic imaging of the Kyushu-Palau Ridge, a remnant arc isolated by Shikoku Basin spreading, reveals a similar Vp=6 km/sec layer, but only 2/3 the thickness of that imaged in the Izu-Bonin arc. These possibly tonalitic layers are in contrast to the largely basaltic lavas that erupt, and the mafic cumulates that are expected. Seismic studies of crustal structure, and experimental calibration of relevant seismic velocities, will be essential for constraining the volume of buried arc crust and crudely averaging its bulk composition. Direct sampling and analysis of this layer where tectonically exposed will be necessary. Mapping and dating of volcanic and plutonic rocks is also necessary to convert volumes to production rates.

Figure 9

What is the composition of the middle and lower arc crust? Numerous studies of the volcanic veneer of the arc crust show that the primary melt extracted from the mantle is mafic. The few exposed middle and lower arc crustal sections studied, however, are highly heterogeneous in composition. In some localities, lower and middle crust have compositions similar to the lavas, while in others they don’t. Evidence from the Aleutians indicates that while the lavas are dominantly basaltic, the exposed plutons are mainly intermediate to felsic; this is consistent with seismic structure of the Izu-Bonin arc discussed above. Thus different processes or different primary magmas may produce intrusive and extrusive rocks. This is important for understanding continental genesis, because the tonalitic plutonic rocks have compositions similar to average continental crust, whereas basaltic lavas do not. Field studies, geochronology, petrology and tracer geochemistry studies will be central in addressing this issue. Exposed plutons in the Aleutians, the Cordillera de Talamanca in Costa Rica, Kamchatka, the Tanzawa Mountains of Japan, the Kohistan terrane of Pakistan, and the accreted Talkeetna arc in Alaska are places where deeper crustal sections can be studied.

What is necessary to translate element fluxes to mass fluxes? Some mass balances are already available for a few chemical species at a few convergent margins. Various estimates indicate 20-50% of some elements (Th, Be, etc.) in subducted sediment are recycled to the arc. How do these elemental fluxes relate to the flux of mass from the slab to the mantle wedge? In theory, every geochemical process must obey mass balance, where the starting composition and fluid composition are related by the partition coefficient and fluid fraction. If we know the starting and final compositions and the partition coefficients, then we can calculate the mass fraction. In this way, geochemical tracers can constrain the mass fluxes of processes. In order to approach mass balance in this way, however, we need to determine the partition coefficients of crucial tracers in melts, solids, melts, and aqueous fluids through a wide range of P, T and composition space. High priority tracers that require better partitioning data are stable isotopes (B, Li, Cl, Be) and the U-series nuclides (U, Th, Ra, Pa). We also require experimental and partitioning studies of key high pressure phases in the subducting slab, such as lawsonite, phengite, and carbonates.

IV. How Will We Study the Subduction Factory?

1. The MARGINS Philosophy

The Subduction Factory Initiative is a component of the MARGINS program. The MARGINS approach concentrates resources on areas targeted for intense, multidisciplinary research. In these focus areas, interaction between researchers involved in field data collection, numerical simulation and laboratory analysis promises unparalleled synergism necessary to understand complex natural systems such as the Subduction Factory. The operation of the Subduction Factory involves lithospheric deformation, mass fluxes, sedimentation, melts, and aqueous fluids. The MARGINS philosophy promises to realize the goals of the Subduction Factory Initiative by sponsoring coordinated, interdisciplinary investigations in these areas.

2. Strategy for Implementation of the Subduction Factory Initiative

The fundamental goal of the Subduction Factory Initiative is to understand relationships between input and output mass and energy fluxes through a subduction zone, and to use this information to address the fundamental science questions outlined above. Realizing these goals requires a theoretical framework that quantitatively reproduces the observable consequences of plate subduction. Because subduction zones are among the most complex components of the solid earth system, achieving this goal will require a coordinated effort by a wide range of earth scientists interacting as members of an interdisciplinary team. The purpose of this section is to suggest guidelines for organizing the wide range of necessary studies.

Integration of the diverse science activities at a given convergent margin should lead to a robust physico-chemical model for how the Subduction Factory operates. We see the next generation of Subduction Factory models as progressing from preliminary model development to data acquisition to construction of refined models. For several convergent margins, sufficient data exist such that work can begin to develop specific physical and chemical models. Models will reveal areas of uncertainty to be addressed by subsequent field campaigns, laboratory analyses, and experiments. These results will constrain the construction, testing, and refinement of the next generation of models. The ultimate success of the Subduction Factory Experiment will be judged by the extent to which models become increasingly comprehensive, can be tested from the observables, and are able to predict behavior at other subduction zones. This will require an interdisciplinary scientific dialogue that promises to result in a quantum leap in our understanding of forcing functions, volatile cycling and mass balance in the Subduction Factory.

A series of early Subduction Factory Theoretical Institutes should focus on how best to develop the physical and chemical models. The Institutes should result in discussions of what geochemical and geophysical data are needed to drive the field and laboratory efforts. Periodic Subduction Factory Theoretical Institutes will maximize scientific team efficiency and recruit new team members.

3. Field, Laboratory, and Experimental Efforts

A wide range of data is needed to constrain and test physical and chemical models.

a. BATHYMETRY AND SWATH MAPPING

Improved swath mapping techniques are providing detailed and evocative bathymetric images of the subducting plate and the leading edge of the upper plate, along with submarine portions of the arc and back-arc (cross-chains and spreading centers). Such images are extremely useful for Subduction Factory research in a number of ways. They can identify fault scarps formed as the plate bends into the trench, where down-dropped grabens may become sediment-filled buckets, and horsts may affect the faulting structure of the upper plate. The faults themselves may be conduits for seawater flow to basement. Bathymetric images also help to identify places where smooth sedimented sea-floor vs. rougher topography (seamounts or ridges) is being subducted. Swath mapping reveals the response of the leading edge of the upper plate to topographic perturbations on the incoming plate. The images provide clues to processes such as prism evolution, frontal accretion, development of the deformation front, and subduction erosion, all important for understanding subduction dynamics and mass transport to depth. In non-accretionary margins, swath mapping can reveal the size, distribution, and to some extent structural setting of serpentinite mud volcanoes. In either case, swath mapping is a useful tool for identifying regions likely to be structurally complex enough that 3-D seismic surveys are important.

b. SEISMIC METHODS

Active-Source Seismology: Images and seismic velocities are obtainable at scales useful for probing the Subduction Factory to depths of about 25 km using newer experimental techniques and focused observational programs. Below 25 km, depth and resolution are limited by the difficulty in propagating broadband sound to great depths, and passive sources (seismicity) becomes the more powerful tool. Active-source imaging can elucidate the top of the down-going slab, delineate structures within the base of the overriding plate, and define structural and velocity details within the forearc and parts of the arc-mantle wedge. This is true for accretionary margins, where the subducted material may either be underplated or carried deep into the Subduction Factory. For many non-accretionary forearcs, imaging will be difficult, but low velocities associated with serpentinite bodies should be mappable. In addition, the initial volatile losses within the upper part of the oceanic crust will be detectable given sufficient velocity resolution. Recent seismic programs to study ocean ridges are applicable to the subduction factory, including active-source tomography.

Seismic sources must be large to penetrate to the needed depths and contain a broad-band spectrum of energy to preserve resolution and allow waveform inversions of the reflections. The best way to obtain high-quality images and velocities is by using 3-D seismic reflection acquisition methods and prestack depth migration. These techniques require high quality data as well as high-performance computing capability. Multichannel seismics, particularly 3-D acquisition and processing, have been shown to provide high-quality images of the décollement and structures above and below. For instance a 3-D data set from Barbados directly mapped the location of aqueous fluids along the décollement as well as in fault ramps splaying into the overlying accretionary prism (Fig. 10). Where the structures above the seismogenic zone are more complex, 3-D methods provide the essential first order corrections for the overlying structure. If the shallow structure is not properly accounted for, reflection amplitudes and waveforms of deeper events will be severely distorted.

Figure 10a

Figure 10b

Reflection and refraction techniques become more powerful when combined. Closely spaced ocean bottom seismographs/hydrophones (OBS/H) along modern normal incidence reflection lines have been used to extend structural imaging to depth as well as provide unique velocity data. These data also provide background velocities to combine with reflection waveform analysis. Active-source tomography has been used to map the structure of a seamounts, young oceanic crust, and rifted oceanic crust using 3-D arrays of closely-spaced OBSs with either conventional surface sources or explosive bottom sources. Similar experiments can help constrain shallow volatile losses from the subducting plate and the degree of serpentinization of non-accretionary forearcs.

Passive-Source Seismology: The wave trains of seismic signals sampling slabs can provide unique information about the structure and composition of subduction zones, at depths relevant to magmagenesis and tectonic driving forces. The development of high-fidelity broadband recording has made it possible to gain far more information from earthquake recordings than previously, particularly from the late parts of seismic signals.

At regional event-receiver distances, subduction zones produce several unusual seismic phenomena. Large P-to-S and S-to-P conversions are commonly observed, which reveal boundaries of the subducted crust and the associated material contrasts. These conversions map boundary locations at accuracies of a few km, and provide a good picture of the relationship between the earthquakes usually used to define subducting slabs and the actual material boundaries at depth. Body waves traveling along slabs are also severely distorted or dispersed, a phenomenon which is used to constrain otherwise inaccessible properties of the subducted plate such as seismic velocities within subducted crust and the thickness of that layer. These measures can provide in situ constraints on the extent to which the basaltic crust metamorphoses to eclogite, and the depths to which it persists at blueschist facies. Finally, large temperature variations in the subducting slab and mantle wedge generate strong changes in seismic attenuation, which is now observable over a wide range of frequencies. Attenuation studies can provide constraints on temperature variations beneath arc volcanoes independent of those provided by velocity tomography.

Teleseismic wave trains reveal strong converted-wave signals from the subducting slab at overlying stations, such as P-to-S and S-to-P conversions. These signals, usually analyzed as receiver functions, provide direct information on the location of the slab and other discontinuities, and on their impedance contrasts. They also provide sensitive measures of Poisson's ratio — useful because serpentinite has a anomalously high Poisson's ratio. Another kind of observation is provided by shear-wave splitting measurements, which constrain the flow-induced fabric of the mantle to the extent that olivine crystals align and produce bulk anisotropy. These observations provide tests of dynamical models in the mantle wedge and elsewhere. Both of these measurements are now being made routinely from portable PASSCAL-type deployments, and are rapidly expanding our understanding of the Earth's deep interior.

Requirements for Seismic Imaging of the Mantle Wedge: A good distribution of earthquakes and locations for seismic stations is essential for passive imaging of upper mantle structure associated with the Subduction Factory. Ideally, an arc would show a high level of seismic activity throughout the upper mantle to depths of ~600 km, and a broad land region for operating a land seismic network. In practice, the sub-arc magma production region can be well imaged as long as earthquake activity extends beneath the volcanic front to depths of 150-200 km. In addition, ocean bottom seismographs (OBSs) may be used in lieu of land seismic stations for arcs with limited land exposure, if seismicity rates are high. Some OBS deployment in the forearc and backarc are generally necessary in most arcs to image a wide region.

The following table lists the seismicity rates at various depths for the arcs under consideration.

Annual seismic events per degree length of subduction zone