Articulating a comprehensive plate-tectonic theory requires understanding how new subduction zones form (subduction initiation). Because subduction initiation is a tectonomagmatic singularity with few active examples, reconstructing subduction initiation is challenging. The lithosphere of many intra-oceanic forearcs preserves a high-fidelity magmatic and stratigraphic record of subduction initiation. We have heretofore been remarkably ignorant of this record, because the “naked forearcs” that expose subduction initiation crustal sections are distant from continents and lie in the deep trenches, and it is difficult and expensive to study and sample this record via dredging, diving, and drilling. Studies of the Izu-Bonin-Mariana convergent margin indicate that subduction initiation there was accompanied by seafloor spreading in what ultimately became the forearc of the new convergent margin. Izu-Bonin-Mariana subduction initiation encompassed ∼7 m.y. for the complete transition from initial seafloor spreading and eruption of voluminous mid-ocean-ridge basalts (forearc basalts) to normal arc volcanism, perhaps consistent with how long it might take for slowly subsiding lithosphere to sink ∼100 km deep and for mantle motions to evolve from upwelling beneath the infant arc to downwelling beneath the magmatic front. Many ophiolites have chemical features that indicate formation above a convergent plate margin, and most of those formed in forearcs, where they were well positioned to be tectonically emplaced on land when buoyant crust jammed the associated subduction zone. We propose a strategy to better understand forearcs and thus subduction initiation by studying ophiolites, which preserve the magmatic stratigraphy, as seen in the Izu-Bonin-Mariana forearc; we call these “subduction initiation rule” ophiolites. This understanding opens the door for on-land geologists to contribute fundamentally to understanding subduction initiation.


A better understanding of the mechanisms by which new subduction zones form is critical for advancing the solid Earth sciences. Until we can reconstruct how and why this happens, we cannot pretend to understand a wide range of important Earth processes and properties, including lithospheric strength, composition, and density, and the driving force behind plate motions. In spite of this, our understanding of the subduction initiation process has advanced slowly, for two important reasons: (1) Subduction initiation is an ephemeral process, so there are few active examples, and (2) nearly all of the evidence for tectonic, magmatic, and sedimentary responses to subduction initiation is preserved in forearcs, which are deeply submerged and buried beneath sediments. We would prefer to study subduction initiation in progress, but there are few places to do this. One such active region however, is the Puysegur subduction zone off the coast of southern New Zealand (LeBrun et al., 2003; Sutherland et al., 2006). However, as only a narrow segment of the Australia-Pacific transform plate margin is affected, studies of Puysegur cannot capture all of the processes that accompany major subduction initiation events, i.e., those that change the lithospheric force balance sufficiently to cause changes in plate motion and stimulate voluminous magmatism, as discussed herein. Such episodes shaped the western margin of North America in Mesozoic time (Dickinson, 2004), established a convergent margin along SW Eurasia in Late Cretaceous time (Moghadem and Stern, 2011), and engendered most of the active subduction zones of the western Pacific in Eocene time (Ishizuka et al., 2011). Major subduction initiation episodes are hemispheric in scale and necessarily reorganize upper-mantle flow, and in many cases are accompanied by widespread and voluminous igneous activity.

Here, we outline a strategy that promises to accelerate our understanding of processes associated with major subduction initiation episodes by considering both the subduction initiation record preserved in forearcs and insights from studying well-preserved ophiolites. The record of subduction initiation is preserved in igneous crust and upper-mantle residues and the associated sediments on the overriding plate next to the trench. These collectively comprise the forearc (Dickinson and Sealey, 1979) and provide the best record of subduction initiation. Significant parts of forearcs may be lost by tectonic erosion (Scholl and von Huene, 2009); nevertheless, whatever remains contains the best record of the processes that accompanied subduction initiation of that particular convergent margin. We explore why this record has been overlooked and summarize recent studies of forearc crust and upper mantle, and what the results reveal about subduction initiation. The expense and difficulty of directly studying forearc igneous rock exposures are huge obstacles to our progress, so we explore the potential of some ophiolites for illuminating forearc composition and magmatic stratigraphy. Ophiolites are exposed on land and so are vastly easier and cheaper to study than forearcs. We conclude that those ophiolites that formed in a forearc provide important opportunities for advancing our understanding of subduction initiation. The strategy of comparative study of igneous forearc crust and ophiolites, coupled with geodynamic modeling, promises to lead to major advancements in our understanding of subduction initiation processes.


Forearcs comprise the bulk of any arc-trench system, occupying the ∼150–200-km-wide region above the subducted plate between the trench and the magmatic arc. Forearcs are relatively stable and low standing—intra-oceanic forearcs lie entirely below sea level—and are morphologically unimpressive compared to spectacular volcanoes of the flanking magmatic arc and the tremendous gash of the trench. For these reasons, it is understandable that forearcs were either overlooked or misunderstood when the geologic implications of plate tectonics were first being explored in the late 1960s and 1970s. During this time, thinking about forearcs was dominated by examples on or near continents, such as California, Japan, Alaska, and Indonesia (for an account of early thinking about what would come to be called forearcs, see Dickinson, 2001). Even today, the textbook example of a convergent plate margin is provided by the Late Mesozoic of California, with the Franciscan mélange representing exhumed subduction-zone material, the Great Valley Group representing the forearc basin, and the Sierra Nevada Batholith representing the roots of the magmatic arc. This is indeed an excellent example of a sediment-rich convergent margin, but emphasis on California and other sediment-dominated forearc examples has inhibited appreciation of forearc crust itself.

Many—but not all—continental forearcs are excellent examples of convergent margins affected by high sediment flux. Some continental convergent margins—such as Peru-Chile and NE Japan—do not have high sediment flux, but these have not been textbook examples because the interesting outcrops are in very deep water, and thus are difficult and expensive to study. In contrast, forearcs away from continents are mostly sediment starved (Clift and Vannucchi, 2004). Such naked forearcs expose crust and upper mantle, which are readily accessed by drilling through thin sediment cover, as was done during Deep Sea Drilling Project (DSDP) Leg 60 and Ocean Drilling Program (ODP) Legs 125 and 126 in the Izu-Bonin-Mariana arc and ODP Leg 135 in the Tonga forearc (Bloomer et al., 1995).


Subordinate proportions of forearcs are accretionary, growing by deposition of large sediment loads from a flanking continent, which bury forearc crust beneath forearc basins and then overflow to the trench, where these sediments briefly ride on the subducting plate before some is scraped off to form an accretionary prism. Such situations of forearc thickening and widening are globally unusual, because most forearcs lose upper-plate crust to the subduction zone due to tectonic erosion, as a result of normal faulting, oversteepening, and basal fracturing and abrasion along the plate interface. Another misconception (due to bias toward studying sediment-rich convergent margins) is that all inner-trench slopes have very low slopes (<3°), when, in fact, erosive margins, especially those exposing igneous basement, are much steeper, typically with slopes of 3°–7° (Clift and Vannucchi, 2004). Estimates of the proportion of accretionary versus erosive convergent margins vary. According to Clift and Vannucchi (2004), 57% of the cumulative length of trenches is erosive and 43% is accretionary, whereas Scholl and von Huene (2007) estimated that 74% and 26% are erosive and accretionary, respectively. Thickness of sediment on the downgoing plate is the single most important control on whether a margin is erosive or accretionary. A sediment thickness of ∼500 m divides the two types of margins. Other factors favoring tectonic erosion include collision of large bathymetric features such as seamounts (Clift and Vannucchi, 2004) and presence of rasping grabens on the downgoing plate (Hilde, 1983).

Although most forearcs are erosive, they are more poorly known than accretionary forearcs because they are harder to study, and a smaller research community has been interested in them. Erosive forearcs are exclusively submarine, so studying them requires research vessels with technology to examine and sample the bottom (Fig. 1A). Compared to accretionary forearcs, erosive forearcs lie in deeper water, farther from the continents, making them more difficult and expensive to study. In contrast, accretionary forearcs lie in shallower water, and parts rise above sea level (e.g., Kodiak Island, Alaska; Shimanto Belt, SW Japan; and Nias and Mentawai, Sumatra), where they are relatively easy to study (Fig. 1B). Accretionary margins are found near continents and so are usually in some nation’s territorial waters or exclusive economic zone, which attracts study by the scientists of that nation. Most of the great accretionary margins lie in the Northern Hemisphere, facing the largest continents, biggest rivers, and greatest sediment flux. The Northern Hemisphere is also where the richest nations are—the ones most likely to support large-scale geoscientific efforts needed for marine tectonic studies. In contrast, erosive margins often lie far away from continents, rich nations, and scientists. Accretionary margins contain potential economic deposits of hydrocarbons, which attract the attention of oil companies; erosive margins do not. Finally, accretionary margins attract more geophysical interest than do erosive margins. Accretionary margins are characterized by progressive deformation of sedimentary layering, which spatially and temporally changes from flat lying and unconsolidated at the trench to steeply dipping and lithified arcward (Fig. 2A). Such lithostructural variations reward seismic-reflection profiling with spectacular images, on which structural interpretations, publications, and proposals can be based. In contrast, intra-oceanic forearcs are generally sediment starved because they are far from continents, and thus large sediment fluxes. These comprise a subclass of erosive margins known as “naked forearcs” (Stern, 2002), which lack thick sedimentary cover and obvious imbrication. Seismic-reflection profiling over naked forearcs yields much less interesting seismic reflection images, and they are more difficult to interpret (Fig. 2B). It is so much easier and more rewarding to study accretionary margins that it is a wonder that erosive margins are studied as much as they are.

In spite of the challenges of studying erosive forearcs, it is important that we do so. Naked forearcs expose igneous infrastructure in the inner trench wall and bury this crust under thin sediments of the forearc itself, providing the best opportunities for in situ investigations of this lithosphere by drilling, dredging, and diving.


An important characteristic of many naked forearcs is the presence of peridotite—exposed upper mantle—in the lower trench wall. Trench peridotite exposures demonstrate that the Moho is also exposed, along with a complete crustal section at shallower depths, and indicate the thickness of the crust.

Peridotite is exposed in the inner walls of intra-oceanic trenches at depths >8 km in the Tonga Trench (Bloomer and Fisher, 1987) but can be found as shallow as 5800 m in the southern Mariana Trench (Michibayashi et al., 2009). Such depths are mostly beyond the reach of manned submersibles, which currently cannot descend below 6500 m, so forearc peridotites are rarely sampled except by dredging. Still, we know about intra-oceanic peridotite exposures in four trenches: Izu-Bonin, Mariana, Tonga, and South Sandwich (Figs. 3 and 4). In addition, mantle peridotite is brought up by serpentine mud volcanoes, which are common in the Mariana forearc and are also known from the Izu forearc (Fryer, 2002).

Because of its importance for understanding the nature and origin of intra-oceanic forearcs, some basic concepts about mantle peridotite in general and forearc peridotite in particular are presented here. These are residues after partial melting and complement magmatic rocks such as lavas—especially basalts—which are more common subjects of marine petrologic study.

Because Earth has had mantle since shortly after it formed, but this has been modified by melt extraction and mixing with subducted materials, idealized compositions are useful for this discussion. For example, “primitive” mantle (PM) refers to an idealized chemical composition after the core segregated but before the continental crust was extracted. PM is also known as “bulk silicate earth” (BSE). Several studies have estimated PM compositions, including Mg# (100Mg/[Mg + Fe] = 89–90), CaO (2.8–3.7 wt%), and Al2O3 (3.5–4.5 wt%; see table 2 of Lyubetskaya and Korenaga, 2007). These estimates constrain minimum Mg# and maximum CaO and Al2O3 contents of the upper-mantle source region of most basalts. Because Earth has been recycling surface materials and melting to make basalt for several billion years, significant tracts of primitive upper mantle are unlikely to exist. Instead, the concept of “fertile mid-ocean-ridge basalt (MORB)–type mantle” (FMM; Pearce et al., 2000) is more useful. FMM is also an idealized upper-mantle composition, but one which acknowledges the extraction of the continental crust. “Pyrolite” is another idealized composition (Green and Falloon, 1998) that is very similar to FMM. FMM and pyrolite approximate the composition of upper mantle that partially melts to generate oceanic crust beneath divergent plate boundaries (spreading ridges) and to produce arc melts beneath convergent margins (note that other components such as pyroxenite exist in the mantle, but these melt almost completely, leaving no identifiable residue).

If PM, BSE, FMM, and pyrolite were rocks instead of ideas, they would be classified as lherzolite, an olivine-rich ultramafic rock that contains >5% clinopyroxene (cpx), the remainder being orthopyroxene and an aluminous phase (spinel or garnet; Fig. 5A). Changes in peridotite composition due to partial melting are simple and clear. Melt depletion diminishes abundances of cpx, CaO, and Al2O3 and increases Cr# (Cr/Cr + Al) in the residual spinel (Fig. 5). This is because basalts, which are rich in CaO and Al2O3 (∼12 and 16 wt%, respectively), are generated by partial melting of lherzolite, which is much poorer in CaO and Al2O3 (∼3 and ∼4 wt%, respectively for FMM; Fig. 5B). Clinopyroxene contains nearly all the CaO in peridotite, whereas spinel in FMM contains ∼35% Al2O3 (a few percent Al2O3 is also dissolved in pyroxene). Melt depletion decreases the proportion of cpx, so that the residue progressively changes from lherzolite to harzburgite (<5% cpx); extreme melt depletion yields dunite (>90% olivine; Fig. 5A).

Even though most forearc peridotite exposures are serpentinized, there are robust mineral and whole-rock compositional characteristics that are remarkably unaffected by such alteration. These include changes in the proportions of minerals (Fig. 5A), major-element bulk chemistry (Fig. 5B), and spinel compositions (Fig. 5C). All of these reflect the amount of melt depletion, as discussed already.

These approaches allow us to compare the “refractoriness” of peridotites from various tectonic settings, i.e., how much melt depletion they have experienced. This reveals that forearc peridotites are the most depleted ultramafic rocks from any modern tectonic environment, with the highest Cr# spinels and lowest proportion of cpx and whole-rock abundances of CaO and Al2O3 (Bonatti and Michael, 1989). Not all forearc peridotites are so depleted; for example, some from the South Sandwich forearc include lherzolites with up to 3.7% Al2O3 and 4.4% CaO, along with spinels with Cr# as low as ∼0.4 (Pearce et al., 2000). Morishita et al. (2011) documented two populations of spinel in Izu-Bonin forearc dunites, one group with moderate Cr# (0.4–0.6), and the other with high Cr# (>0.8). Variations in spinel compositions notwithstanding, forearc peridotites are dominated by ultradepleted compositions rarely found in other tectonic environments.

The unusually depleted nature of forearc peridotites requires unusual melting conditions: abnormally high temperature, volatile flux, or both. Whatever the cause, these depletions are all the more noteworthy because forearcs associated with mature arcs have unusually low heat flow (Stein, 2003) and rarely are volcanically active. Whatever conditions caused the unusually extensive melting beneath forearcs no longer exist. Such transitory conditions are linked to subduction initiation in the next section.

Igneous rocks of exposed forearc crust above peridotites (Fig. 4) are only now becoming the focus of geoscrutiny. At one time, forearc crust was thought to comprise oceanic crust that was trapped when subduction began (e.g., Dickinson and Sealey, 1979), so the origin of forearc igneous crust has not, until recently, been much studied. This is changing, partly because of what is now recognized as the unusually strong depletion of forearc peridotites and because of interest in boninites. Boninites are lavas with unusual combinations of high silica and magnesium coupled with low calcium and aluminum abundances. These compositional features reflect low-pressure melting of harzburgitic mantle (e.g., Falloon and Danyushevsky, 2000), which is not otherwise observed on modern Earth. In addition, boninites are enriched in large ion lithophile elements (LILEs; elements with low valence and large ionic radius) and light rare earth elements (LREEs) relative to high field strength elements (HFSEs; elements with high valence and small ionic radius). High LILE/HFSE ratios in boninites reflect metasomatism of the source mantle by hydrous fluid released from subducted crust and sediments (Gill, 1981; Stern et al., 1991; Pearce et al., 1992). This fluid lowers peridotite melting temperature at the same time that it re-enriches it in fluid-mobile elements, including LILEs and LREEs.

Not all naked forearcs have boninite, but at least one of them—the Izu-Bonin-Mariana arc system—does (Fig. 6A; Stern et al., 1991; Macpherson and Hall, 2001). Probably the best exposure of boninite in the world is found in the Bonin (Ogasawara) islands (Taylor et al., 1994). These boninites erupted on the seafloor ca. 46–48 Ma, shortly after subduction began ca. 52 Ma (Ishizuka et al., 2006, 2011). Recent studies of Izu-Bonin-Mariana forearc crust exposed in the inner-trench wall (Fig. 6B) reveal that boninite may be the uppermost component of forearc crust, underlain by thicker, slightly older tholeiitic basalts, which Reagan et al. (2010) called “forearc basalts.” Forearc basalt lavas and related dikes have chemical compositions similar to MORB. Forearc basalts were first recognized in the southern Mariana forearc SE of Guam, where they crop out trenchward of thin boninites and younger arc rocks (Reagan et al., 2010). This outcrop pattern, as well as the volcanic stratigraphy drilled in the Mariana forearc at DSDP Site 458 (Fig. 6B) indicates that forearc basalts here are older than the boninites and were likely the first lavas to erupt when the Izu-Bonin-Mariana subduction zone formed (Reagan et al., 2010; Ishizuka et al., 2011). Below the forearc basalts, there are gabbroic rocks, and then mantle peridotites, as summarized in Figure 6B (Ishizuka et al., 2011).

Forearc basalts have major-element compositions that are broadly MORB-like, although significant differences exist, at least for Izu-Bonin-Mariana forearc basalts. Forearc basalts are generally not as rich in TiO2 as is typical MORB, which often contains >1.4 wt% TiO2. Forearc basalts can also be more rich in SiO2 than typical MORB (most forearc basalts contain <51% but forearc basalt–boninite transitional lavas at DSDP Sites 458 and 459 are richer in silica). In spite of these differences, most forearc basalts also have trace-element (Figs. 7B and 7C) and isotopic compositions (Reagan et al., 2010) that are MORB-like, whereas younger forearc lavas have compositions suggesting that subducted fluids were involved in their genesis. For example, lavas generated by melting in the presence of a fluid from a subducting plate (e.g., Mariana arc lavas and boninites) typically have elevated Th/Yb compared to MORB. On a plot of Th/Yb versus Nb/Yb, most Izu-Bonin-Mariana forearc basalts plot with MORB along the unmodified “mantle array,” whereas younger, subduction-influenced lavas trend toward Th/Yb typical of Mariana arc lavas (Fig. 7E). Note that Figure 7 also plots the chemostratigraphies of ophiolitic lavas, a point which is discussed further below.

DeBari et al. (1999) studied the Izu-Bonin-Mariana inner-trench wall (6100–6500 m deep) near 32°N (Fig. 6A) and documented the presence of MORB-like basalts. They interpreted these to represent older oceanic crust that was trapped when subduction began. However, the composition of these lavas is identical to those of forearc basalts from elsewhere in the Izu-Bonin-Mariana forearc, and we prefer to interpret this as another exposure of basalts that formed when Izu-Bonin-Mariana subduction began.

To conclude this section, it is clear that our understanding of the composition of forearc crust is incomplete, largely due to the difficulty of accessing and studying this material. Much of what we think we know is based upon studies of the Izu-Bonin-Mariana forearc. We anticipate that Izu-Bonin-Mariana forearc crust is representative of igneous rocks that form when subduction begins, but we cannot be sure until we have studied the igneous rocks of other forearcs in similar detail.


In this section, we discuss igneous activity associated with the formation of a new subduction zone. As presented already, much of what we understand about the igneous crust of forearcs comes from studies of the Izu-Bonin-Mariana system. The fundamental question we address for this forearc, and one that is pertinent for many others is: Does subduction initiation generate this broad swath of crust, which ultimately forms the forearc (Fig. 8)? Also, what happens when the subduction zone itself forms? For the Izu-Bonin-Mariana arc system, we have documented the progression of igneous activity in the forearc about the time that the Pacific plate changed its motion, and we have concluded that this activity resulted from the dynamic response of the crust and upper mantle to subduction initiation. The basic idea is: at ca. 52 Ma, old, dense lithosphere of the Pacific plate began to sink, perhaps due to differential subsidence across a lithospheric weakness, such as an old fracture zone (Fig. 9A; Stern and Bloomer, 1992). We conclude that subduction initiation at this time was hemispheric in scale: much of the western Pacific, extending south from Izu-Bonin-Mariana to Fiji and the Tonga-Kermadec convergent margin, formed new subduction zones about this time. This was accompanied by voluminous generation of forearc basalts, boninite, and related igneous rocks, much of which is now preserved in these forearcs. We are not sure whether these new subduction zones were caused by, or were the cause of the change in the Pacific plate absolute motion, from NNW to WNW at ca. 50 Ma, as reflected by the bend in the Emperor-Hawaii seamount chain (Sharp and Clague, 2006), but both events happened about the same time.

There are many challenges to this summary of Izu-Bonin-Mariana subduction initiation. These include hypotheses that: (1) interaction with a mantle plume was responsible for boninite formation (Macpherson and Hall, 2001); (2) extrusion of Indian Ocean–Asian asthenosphere due to Tethys closure was a major cause of western Pacific arc rollback and basin opening (Flower et al., 2001); (3) boninite formation reflected the intersection of an active spreading ridge and a subduction-transform fault transition (Deschamps and Lallemand, 2003); (4) compression across a preexisting fracture zone was required (Hall et al., 2003); (5) lateral compositional buoyancy contrast within oceanic lithosphere controls subduction initiation (Niu et al., 2003); (6) the new Izu-Bonin-Mariana convergent margin cut across, rather than followed, preexisting lithospheric fabric, such as remnant arcs, fracture zones, and spreading ridges (Taylor and Goodliffe, 2004); and (7) subduction of the Pacific-Izanagi spreading ridge triggered a chain reaction of tectonic plate reorganizations that led to Izu-Bonin-Mariana and Tonga-Kermadec subduction initiation (Whittaker et al., 2007).

There is clearly a lot of uncertainty about what caused the Izu-Bonin-Mariana subduction initiation and attendant boninite volcanism, but there is no disagreement that it was accompanied by voluminous igneous activity. This formed Izu-Bonin-Mariana forearc crust (Figs. 6 and 8) as well as crustal tracts well to the west of the present volcanic front, including the West Mariana Ridge and Kyushu-Palau Ridge. Stern and Bloomer (1992) conservatively (i.e., assuming generation of 6-km-thick crust) estimated that 1200–1800 km3 of crust were produced per kilometer of arc during Izu-Bonin-Mariana subduction initiation, and that this episode lasted 10 m.y., for a crustal growth rate of 120–180 km3/km. This is equivalent to the volume of crust produced at a mid-ocean-ridge spreading at 2–3 cm/yr. In fact, Izu-Bonin-Mariana mean crustal thickness produced during this episode was probably greater than 6 km, as shown in Figure 8 for the Izu forearc. For the Mariana forearc, Calvert et al. (2008) concluded that 15-km-thick crust just east of the magmatic front formed during a brief magmatic episode early in the arc’s history. Furthermore, Izu-Bonin-Mariana subduction initiation probably lasted less than 10 m.y.; Ishizuka et al. (2011) estimated that it took ∼7 m.y. before a stable magmatic arc was established near the position that it still occupies. Finally, subduction erosion has removed a significant amount of forearc; a width of 1 km removed per million years translates into 240 km3/km of lost crust. All these considerations indicate that Izu-Bonin-Mariana crustal growth related to subduction initiation was significantly greater than 120–180 km3/km.

Such high crustal growth rates and the absence of large, central volcanoes of Eocene age in the Izu-Bonin-Mariana forearc imply that crust was formed by seafloor spreading, at least during the early, forearc basalts–dominated episode. This is consistent with the presence of dense diabase dike swarms at the base of the forearc basalts sequence in the Izu-Bonin-Mariana forearc (Reagan et al., 2010; Ishizuka et al., 2011). Furthermore, the composition of forearc basalts, which is similar to MORB, implies a similar origin by decompression melting beneath a spreading ridge (Plank and Langmuir, 1992). Small central volcanoes may have formed during later, boninitic volcanism, for example, at ODP Site 786B (Lagabrielle et al., 1992), but the thicker, older forearc basalt succession seems to have been emplaced by tectonomagmatic processes akin to seafloor spreading. There is no clear evidence about the arrangement of spreading ridges that might have existed; Stern (2004) inferred short segments aligned oblique to the evolving convergent plate boundary, although no good magnetic anomaly patterns have yet been identified in any forearc that could be interpreted as spreading fabric.

Given that seafloor spreading best explains the tectonic environment for Izu-Bonin-Mariana forearc crust formation, the logical conclusion is that a strongly extensional environment existed at that time. There is no evidence that early Izu-Bonin-Mariana subduction was accompanied by compression, as would be expected if the new subduction zone was caused by one plate being forced beneath the other (induced subduction initiation of Stern, 2004), although such evidence (uplift-related unconformity, thrust faulting) might have been obliterated by Eocene igneous activity. Such evidence of initial compression without forearc volcanism characterizes the Puysegur mini-subduction initiation episode, which serves as an excellent example of induced subduction initiation. The conclusion that a strongly extensional environment accompanied Izu-Bonin-Mariana subduction initiation leads logically to the idea that hinged subsidence of the older, thicker, and denser Pacific plate allowed asthenosphere to well up and fill the widening chasm, accompanied by extensive decompression melting (Fig. 8B). The sequence of events summarized in Figure 9 encapsulates the idea of spontaneous subduction initiation (Stern, 2004), including early extension and seafloor spreading. On the other hand, we cannot rule out the possibility that induced (forced) subduction initiation might also have been associated with early voluminous igneous activity, due to the likelihood of lithospheric collapse and asthenospheric upwelling, as modeled by Hall et al. (2003).

Does current knowledge about Izu-Bonin-Mariana subduction initiation serve as a useful model for reconstructing how other forearcs form? We cannot be sure because we know so little about their crust, but Izu-Bonin-Mariana serves as a useful analogue for the Tonga-Kermadec forearc. This crust is exposed only on the island of ‘Eua and was drilled near the trench at ODP Site 841. Arc tholeiite, gabbro, and harzburgitic peridotite have been dredged from the inner-trench wall, suggesting to Bloomer et al. (1995) that the Tonga forearc is floored by crust similar in age and composition to that of Izu-Bonin-Mariana, including Eocene boninite (Crawford et al., 2003). The oldest known rocks in the Tonga-Kermadec forearc are 46–40 Ma arc-type lavas occurring below Upper Middle Eocene limestones on ‘Eua (Ewart and Bryan, 1972; Duncan et al., 1985; Tappin and Balance, 1994). ODP drilling at Site 841 in the Tongan forearc recovered a thick sequence of low-K arc tholeiitic rhyolites (Bloomer et al., 1994), dated by McDougall (1994) at 44 ± 2 Ma. Farther north in the Tongan forearc, true low-Ca boninitic rocks and associated backarc basin–type basalts of probable Eocene age have been dredged at depths in excess of 4 km (Falloon et al., 1987) and have yielded ages between 45 and 35 Ma (Bloomer et al., 1998). It is not clear whether or not these boninitic lavas were generated in the same ca. 52 Ma subduction initiation event as that responsible for the subduction initiation boninite–refractory-forearc mantle package of the Izu-Bonin-Mariana arc system to the north.

We emphasize that not all subduction zones form by processes outlined here. As outlined by Stern (2004), some subduction zones form as a result of plate-boundary reconfigurations, for example, as a result of terrane accretion or continental collision. Collision of India with Eurasia is a good example of this, although a new subduction zone has not yet formed behind (south of) India, reflecting the great strength of Indian Ocean lithosphere (Stern, 2004). Collision of the Ontong-Java Plateau on the north side of the Solomon arc in Miocene time caused a new subduction zone to form to the south (Mann and Taira, 2004), in what may be the best actualistic example of subduction polarity reversal. Stern (2004) characterized this type of subduction initiation, which includes the Puysegur trench example, as reflecting “induced” nucleation of a subduction zone. Induced nucleation of a subduction zone conceptually contrasts with “spontaneous” nucleation of a subduction zone” such as Izu-Bonin-Mariana may be. Stern (2004) further suggested that induced nucleation of a subduction zone may not be associated with voluminous forearc igneous activity, such as that forming the Izu-Bonin-Mariana forearc. If so, forearcs formed by induced nucleation of a subduction zone may have fundamentally different crustal structures and origins than those formed during spontaneous nucleation of a subduction zone. The generalization that spontaneous nucleation of a subduction zone results in broad forearc magmatism, whereas induced nucleation of a subduction zone does not, may not be true. At least one example of induced nucleation of a subduction zone—Aleutian subduction initiation—may be an example of induced nucleation of a subduction zone with attendant forearc igneous activity. Scholl (2007) and Minyuk and Stone (2009) suggested that Aleutian subduction initiation exploited SSW-trending strike-slip faults related to the “tectonic escape” of Alaska, associated with motion along the North Pacific Rim Orogenic Stream (Redfield et al., 2007). A curved system of long strike-slip faults propagated southwestward across the North Pacific Rim throughout Cenozoic time, disrupting older subduction zones along the Bering Sea shelf edge. Recently obtained ca. 46 Ma 40Ar/39Ar ages from Aleutian forearc igneous rocks (Jicha et al., 2006; Minyuk and Stone, 2009) provide minimum age constraints for the timing of Aleutian subduction initiation, although Aleutian subduction initiation generated middle Eocene magmatic rocks as early as ca. 50 Ma (Scholl, 2007). The effects of tectonic erosion at convergent margins must be considered for any thoughtful subduction initiation analysis. Naked forearcs are likely to be trimmed back by subduction erosion, at rates that can vary from a few to several kilometers per million years (Clift and Vannucchi, 2004; Scholl and von Huene, 2007). Subduction erosion thus can remove all of the evidence for a magmatic forearc in several tens of millions of years, as may be the case for the Andean forearc.

Keeping such caveats and complications in mind, it seems reasonable to conclude that many forearcs form as a result of voluminous yet ephemeral igneous activity accompanying subduction initiation. This makes it worthwhile to reconsider the origins of forearc igneous rocks that formed about the same time as subduction initiation. For example, the ca. 55 Ma Siletzia terrane of the Oregon and Washington Coast Ranges formed about the same time as Cascadia subduction initiation and is variously interpreted as an accreted oceanic plateau (Duncan, 1982), produced by the Yellowstone plume head (Pyle et al., 2009), or due to a tear in the subducting slab (Humphreys, 2008). The possibility that Siletzia formed in situ as a forearc magmatic response to Cascadia subduction initiation should also be entertained.

Finally, we should consider the possibility of subduction “re-initiation,” i.e., the case where a subduction zone once existed, then was extinguished, and then resumed at about the same location, because of either induced or spontaneous nucleation of a subduction zone. The Cretaceous and younger evolution of SW Japan serves as an example of this. Subduction of Pacific seafloor beneath NE Japan has been continuous, but subduction of the Philippine Sea beneath SW Japan has been episodic. During Cretaceous time, a subduction zone dipped beneath what is now SW Japan, associated with a robust magmatic arc. Subduction there ceased and was replaced by a transform fault (shear margin) during Paleogene time. This lithospheric weakness was converted into a new subduction zone beginning in latest Oligocene time, with an attendant flare-up of forearc igneous activity (Kimura et al., 2005). Other likely examples of subduction re-initiation include the Paleogene Cascadia system and the Late Jurassic of California.

To conclude this section, the igneous crust and uppermost mantle of forearcs do not generally represent trapped oceanic lithosphere but in fact typically form during upper-plate spreading associated with subduction initiation (Shervais, 2001; Stern, 2004). There is no doubt that we need to better understand the composition and mode of formation of forearc crust, not only for its own sake but also to better reconstruct events accompanying subduction initiation. Such studies require studying naked forearcs, with all the challenges this entails. In the next section, we propose a complementary strategy that takes advantage of the fact that forearc lithosphere is commonly emplaced (obducted) when buoyant lithosphere—particularly continental crust—on the downgoing plate enters and clogs a subduction zone (Wakabayashi and Dilek, 2003). Introduction of buoyant lithosphere disrupts the normal operation of a subduction zone, so that subducted materials are partially regurgitated, and overlying forearc lithosphere is lifted above sea level (Glodny et al., 2005). Consequently, forearc crust (and accreted sediments) is a key component of orogens. These tracts of obducted forearc lithosphere are known as ophiolites.


Ophiolites are fragments of oceanic lithosphere that have been tectonically emplaced on land. A complete “Penrose” ophiolite includes tectonized peridotite, gabbro, sheeted dikes, and pillow basalt (Anonymous, 1972), but this idealization is rarely seen because ophiolites are faulted and fragmented during emplacement or because one, or more, unit was never generated. Nevertheless, ophiolites are key petrotectonic indicators, perhaps the single most important indicator of ancient plate-tectonic activity (Stern, 2005). Ophiolites mark tectonic sutures, indicating both the location of ancient oceans and convergent plate boundaries where buoyant lithosphere was partially subducted, also known as collision zones (Dilek, 2003). As a result, ophiolites are often highly altered and faulted, and much effort and imagination are needed to reconstruct the original crust and uppermost mantle section. Ophiolites are unequivocal evidence of seafloor spreading and have been an important part of modern geologic thought since the 1960s, but there is a lot of confusion about the tectonic environment in which they formed. Much of this misunderstanding results from a lack of appreciation of the disparate lines of evidence needed to reconstruct ophiolites, especially structural geology, igneous geochemistry, and marine geology. Ophiolites were originally thought to form at mid-ocean ridges, but we now understand that most sediments and all crust on the downgoing plate are subducted. If the downgoing plate includes very thick (>1 km; Clift and Vannucchi, 2004) sediments, some of this may be scraped off, and sometimes seamounts may be transferred from the subducting to the overriding plate, but normal oceanic crust itself has never been demonstrated to be transferred from downgoing to overriding plate (i.e., by seismic-reflection profiling) at any modern convergent margin (Fig. 10C). Changes in plate motion might trap mid-ocean-ridge crust in what may ultimately become a forearc (e.g., Macquarie Island; Varne et al., 2000), but this is likely a very unusual tectonic scenario.

In the 1970s, geoscientists (beginning with Miyashiro, 1973) began to recognize the similarity of some ophiolite lava compositions to those of convergent margins, leading to the concept of the “suprasubduction zone” ophiolite (Pearce et al., 1984). This created tension between conclusions based on structural studies, which indicated ophiolites form by seafloor spreading, and those based on geochemistry, which indicated ophiolites form at convergent margins. This tension was temporarily reconciled by the idea that suprasubduction zone ophiolites formed in backarc basins (Dewey, 1976; Pearce, 2003). This reconciled structural evidence of seafloor spreading with geochemical evidence for convergent margin setting, but it is difficult to see how a backarc basin, ∼200 km from the convergent plate boundary, would be emplaced (Fig. 10B). This logic train is further derailed for some ophiolites interpreted as fossil backarc basins because the associated arc and forearc, which are typically ∼200 km wide and ∼30 km thick, are usually not identified (e.g., Semail ophiolite interpretation of Godard et al., 2003). Furthermore, there are no known Cenozoic examples of backarc basin closure (Stern, 2004), which is required to emplace a backarc basin ophiolite.

A simple and elegant solution to the ophiolite emplacement problem is that whatever crust comprises the forearc is most likely to be emplaced during plate collision. This process is often referred to as “obduction,” although this term originally described how oceanic crust on the downgoing plate was thrust over the convergent margin (Coleman, 1971). Forearc lithosphere is optimally situated for obduction (Wakabayashi and Dilek, 2003). It is straightforward to envisage emplacement of forearc lithosphere above the same subduction system in which it was generated due to isostatic uplift following partial subduction of buoyant crust that jams the subduction zone (Fig. 10A). This process—akin to sliding a spatula under a pancake or fried egg to lift it—is also most likely to yield the least-disrupted ophiolites, and those most likely to approximate the Penrose ophiolite ideal, such as Troodos (Cyprus) and Semail (Oman). During the early days following the plate-tectonic revolution, when it was thought that forearc crust was relict, trapped oceanic crust, this provided an attractive way to emplace MORB-type ophiolites. As discussed previously herein, there is little evidence to support the idea that forearcs are composed of trapped oceanic crust that existed in the region prior to the formation of a new subduction zone. It is increasingly clear from the simple perspective of plausible emplacement mechanisms that forearcs are the most likely source of ophiolites (Casey and Dewey, 1984; Milson, 2003; Metcalf and Shervais, 2008).

Because most ophiolites are very incomplete and rarely satisfy the Penrose definition, we propose an alternative definition that better captures the important elements of what are called ophiolites in the literature. An ophiolite must consist of a significant proportion of harzburgitic peridotite (depleted mantle) and pillowed basalt. Gabbro and sheeted dikes may be missing, but there should be associated deep-sea sediments. These units should be exposed above sea level. This definition is more consistent with common usage of the term ophiolite than is the Penrose definition.

Based on emplacement mechanisms, Wakabayashi and Dilek (2003) distinguished four types of ophiolites: (1) “Tethyan” ophiolites, emplaced over passive continental margins; (2) “Cordilleran” ophiolites, emplaced over subduction complexes; (3) “ridge-trench intersection” ophiolites emplaced through complex processes resulting from the interaction between a spreading ridge and a subduction zone; and (4) the unique Macquarie Island ophiolite, which was subaerially exposed as a result of a change in plate-boundary configuration along a mid-ocean-ridge system. The first two types represent the overwhelming majority of ophiolites, with the second two types comprising miniscule proportions. Tethyan- and Cordilleran-type ophiolites reflect the fundamentally different nature of the subducted oceanic realms, with Tethyan ophiolites subducting relatively narrow oceans before colliding with continental fragments and Cordilleran ophiolites generally subducting Pacific seafloor, which contains no continental fragments.

Given that most ophiolites originate in forearcs, is it possible to independently determine whether or not a given ophiolite formed during subduction initiation or as a result of some other process, as outlined herein? Such an assessment would be very useful because of the difficulties involved in studying submerged forearc igneous rocks and reconstructing subduction initiation processes. If ophiolites could be so linked, it could pay huge dividends in advancing our understanding of the fundamental tectonic province of forearc and processes of subduction initiation. The next section summarizes a chemostratigraphic approach for evaluating whether or not a given ophiolite formed during subduction initiation.


The subduction initiation rule (Whattam and Stern, 2011) predicts that ophiolites that form as a result of subduction initiation processes consist of a sequence of igneous rocks formed by a magma source that changed progressively in composition by the combined effects of melt depletion and subduction-related metasomatism. Magmas erupted during subduction initiation progress from early decompression melts of unmodified fertile (lherzolitic) mantle to yield forearc basalts to younger hydrous flux melts of depleted (harzburgitic) mantle that has been strongly modified by subduction-related fluids to yield late high-Mg andesites and boninitic lavas. If the subduction initiation process continues long enough to generate steady-state subduction with downwelling mantle overlying the sinking plate, then normal arc volcanics will cap the subduction initiation sequence as the locus of magmatism retreats from the trench. This magmagenetic evolution is portrayed in Figures 7 and 11, which encapsulate the subduction initiation tectonic evolution shown in Figure 9. Whattam and Stern (2011) outlined the arguments and implications of the subduction initiation rule and proposed that ophiolites showing this progression be termed “subduction initiation rule ophiolites.”

It has long been recognized that some ophiolites contain igneous rocks with strong chemical affinities to both mid-ocean-ridge and arc basalts. Distinguishing compositions of MORB-like lavas include >1% TiO2, variable depletion in LREEs, absence of HFSEs (e.g., Nb, Ta), depletions on spider (primitive mantle– or N-MORB–normalized) diagrams, La/Nb <1, etc.; we regard these early MORB-like sequences as forearc basalts. In contrast, volcanic arc–like basalts have lower TiO2, are enriched in fluid-mobile elements (e.g., LILEs and LREEs), and have strong HFSE depletions relative to LREE (e.g., La/Nb >1, etc.; Pearce, 2003). Recognition of both arc-like and MORB-like compositions in some ophiolites has encouraged some workers to infer formation in a backarc basin (e.g., Beccaluva et al., 2004) or a complex, multistage tectonic history, for example, eruption of the two suites in discrete tectonic environments (e.g., Saccani and Photiades, 2004; Godard et al., 2003). The reasons against formation as a backarc basin are outlined in the previous section. The conclusion here that the basaltic sections in ophiolites are forearc basalts (Reagan et al., 2010) is supported by the observation that most Izu-Bonin-Mariana forearc basalts have lower Ti/V ratios than MORB, which probably results from the enhanced melting that commonly occurs in nascent subduction settings. The presence of transitional lavas with compositions that progress from forearc basalts to boninite with time at DSDP Sites 458 and 459 also supports this contention. Some ophiolites may reflect complex tectonic histories, but this cannot serve as the general explanation for subduction initiation rule ophiolites. Furthermore, any inference of complex tectonic histories is inconsistent with the absence of unconformities or significant sedimentary horizons between ophiolite lavas with MORB-like versus volcanic arc–like basalt compositions. Such interruptions might be difficult to identify because of ophiolite deformation, but such an important and distinctive feature should sometimes be recognized if this interpretation is generally correct. Conversely, the interpretation of a single, rapidly evolving magmagenetic environment is favored because such a hiatus is generally not observed. In addition, lavas with compositions that are transitional between forearc basalts and boninites have been observed in some ophiolites (e.g., Dilek and Furnes, 2009). As a result, ophiolite compositional variability is increasingly ascribed to progressive depletion and metasomatism of the mantle source as the ophiolite magmatic crust forms (e.g., Shervais, 2001; Beccaluva at al., 2005; Dilek and Furnes, 2009). Ophiolite lava compositions thus are increasingly interpreted to be products of magmatism in a single (suprasubduction zone) tectonic setting, albeit one that changed rapidly.

It is worth emphasizing that it is generally difficult to recognize a magmatic stratigraphy in ophiolites, because they are so often jumbled by faulting, and because lavas and associated intrusions with different chemistries appear similar in the field. We recognize four ophiolites that have been studied in sufficient detail to reconstruct their magmatic stratigraphies: Mirdita, Pindos, Troodos, and Semail (as summarized by Whattam and Stern, 2011). There are other examples of subduction initiation rule ophiolites that could also be considered, for example, the 485–489 Ma ophiolite belt that can be traced >1000 km from Newfoundland down into the Taconic suture of New York (Bédard et al., 1998; Schroetter et al., 2003; Pagé et al., 2009).

The four Tethyan ophiolites mentioned here show subduction initiation rule relationships, with lower lavas being more MORB-like and upper lavas being more arc-like, as Whattam and Stern (2011) showed. Troodos and Semail are at either end of the 3000-km-long ophiolite belt that marks the Late Cretaceous forearc of SW Asia, including the inner and outer ophiolite belts of Zagros, Iran (Moghadam and Stern, 2011). The magmatic chemostratigraphies of the many Zagros ophiolites have not been worked out, but these ophiolites in many cases have strong compositional affinities in some respect to arc magmas, for example, with respect to La/Nb (arc lavas usually have La/Nb >1.4 according to Condie, 1999). It is very likely that we will be able to work out magmatic stratigraphies in more ophiolites around the world, but this will require thoughtful integrated structural and petrologic studies.


Subduction initiation rule ophiolites provide wonderful opportunities to better understand the crust and upper mantle of magmatic forearcs and the ways in which new subduction zones form. Identifying and studying subduction initiation rule ophiolites will provide easy access to forearc lithosphere samples and structures, allowing many scientific perspectives to be engaged cheaply and easily. Where ophiolites can be firmly linked to forearcs through application of the subduction initiation rule and other approaches, geodynamic models for subduction initiation can be more easily developed and tested. Because subduction initiation rule ophiolites are fossil forearcs, they often can be traced across strike for tens of kilometers and along strike for hundreds of kilometers. Erosion of deformed subduction initiation rule ophiolites exposes various levels, allowing four-dimensional reconstructions of timing as well as vertical, across-strike, and along-strike magmatic variations. Sampling for geochronology is easier than for in situ forearcs because of this exposure, and relations of such dated samples to the surrounding rocks and fabrics provide context for interpreting the ages. The multiple perspectives and levels of detail allowed by these approaches mean that many aspects of forearc crust structure and subduction initiation can be understood much better by indirect study of subduction initiation rule ophiolites than by studying forearcs directly, as can be seen by mentally comparing the experiences captured in Figure 1.

Even as we recognize that studies of subduction initiation rule ophiolites are essential for understanding forearcs, studies of in situ forearc crust need to move forward. We are only beginning to understand the igneous crust of a single forearc in any detail—that of Izu-Bonin-Mariana—and this example seems to be unusual in terms of the abundance of early-arc boninites. We need to continue to test and refine what we know about forearc magmatic evolution. It is especially important that we drill and sample through the magmatic stratigraphy of an igneous forearc in order to better understand the sequence and timing of magmas at one location, allowing us to answer questions such as: What are the relative proportions of forearc basalts versus volcanic arc–like basalt/boninitic lavas? How does forearc basalt magmatism transition to volcanic arc–like basalt magmatism? Is it gradational or abrupt? How long does forearc volcanism last? Answering such questions for an in situ forearc and comparing these answers with those for subduction initiation rule ophiolites will be key tests of the ideas presented in this paper. Furthermore, there are aspects of forearc structure and evolution that cannot be understood without studying extant forearcs. For example, active serpentine mud volcanoes in the Mariana forearc serve as actualistic models for sedimentary serpentinite deposits associated with some forearcs (Fryer et al., 1995; Fryer, 2002). Knowing that these mud volcanoes exist has stimulated the search for ancient serpentine mudflows (e.g., Teklay, 2006). Another example is the problem of tectonic erosion, which can remove a few kilometers of forearc crust in a few million years. Studies of active systems are generally aware that some crust is missing, whereas studies of ophiolites rarely consider this complication.

Another point worth emphasizing is that there may be a wide range of subduction initiation magmatic products. Our models now are heavily biased toward the Izu-Bonin-Mariana convergent margin, which has abundant boninitic lavas and serpentinite mud volcanoes. The Izu-Bonin-Mariana forearc may serve as a useful example of intra-oceanic subduction initiation, but it could be less useful as a model for continental margin subduction initiation. Subcontinental mantle lithosphere is compositionally distinct from that beneath the ocean basins (Bonatti and Michael, 1989), and continental and oceanic mantle may melt to different extents during subduction initiation. Because there are likely to be significant differences in magmatic products from that shown by Izu-Bonin-Mariana and ophiolites, we should keep an open mind when considering the origin of any magmatic forearc built about the time that subduction began at a given convergent margin. The Siletzia terrane of coastal Oregon and Washington (Duncan, 1982) could be an example of continental margin subduction initiation where the magmatic sequence is somewhat different than that seen for Izu-Bonin-Mariana, for example, by having upper alkalic lavas instead of volcanic arc–like basalt overlying tholeiites.

Above all, combined studies of subduction initiation rule ophiolites and forearc crust need to be integrated with geodynamic modeling to learn about the ways in which new subduction zones form. Geodynamic modeling of subduction initiation needs to be firmly tethered in reality, and the more that such models attempt to explain the rocks making up a real forearc and ophiolite, the more rapidly our understanding of this fundamental Earth processes will advance.

We thank Dave Scholl for edifying comments on accretionary prisms and Steve Graham for the photo of a Franciscan chert exposure. Thoughtful reviews by Rod Metcalf, Jean Bédard, and editors John Shervais and John Goodge are much appreciated. This manuscript was written while Stern enjoyed a Blaustein Fellowship at Stanford University. Stern’s research on intra-oceanic arcs has been supported by the National Science Foundation, most recently by grant OCE-0961352. Reagan acknowledges research support from National Science Foundation grant EAR-0840862. This is University of Texas at Dallas Geosciences contribution 1228.