or collective redistirbution other or collective or means this reposting, of machine, is by photocopy article only anypermitted of with portion the approval of T SPECIAL ISSUE FEATURE in published been is has article Oceanography

MANTLE MELTING BENEATH , Volume 1, a quarterly 20, Number journal of MID-OCEAN RIDGES

BY CHARLES H. LANGMUIR AND DONALD W. FORSYTH T

The plate-tectonic revolution was initially “kinematic”— pressures equivalent to thousands of atmospheres in the inte- Society. Copyright e Oceanography 2007 by a description of plate motions across Earth’s surface. Plate rior. As mantle ascends beneath the mid-ocean ridge, less and tectonics is now recognized as the surface manifestation of a less rock lies above it, so large pressure changes occur, which greater process—circulation of the solid earth. Magma ascends leads to melting. The melt is less dense than the solid, and rises to the surface at mid-ocean-ridge spreading centers to cool and to the surface to form the oceanic crust. form oceanic crust, which millions of years later returns to the Figure 1 shows how rising mantle crosses the “solidus” (the mantle at subduction zones. Formation of oceanic crust is the transition from complete solid to partial melt) and melts pro- greatest contribution of fl ow from our planet’s interior, as two- gressively towards the surface. Note that because the mantle is T thirds of the earth is resurfaced about every 100 million years. a solid consisting of many different molecules, it does not melt Permission is Society. Allreserved. rights granted e Oceanography to in teaching copy this and research. for use article Republication, systemmatic reproduction, T Partial melting of the mantle at spreading centers is the mecha- entirely at a single temperature, but progressively over a range all correspondence to: Society. Send [email protected] e Oceanography or nism by which this fl ow takes place, and thus is fundamental to of temperatures—from 0 percent melting at the solidus to understanding solid-earth circulation. 100 percent melting several hundred degrees higher at the liq- Melting is a primary means by which the earth cools: sea- uidus. Thus, partial melting is possible. fl oor spreading brings hot mantle from depth to the colder sur- Several lines of evidence provide information about this face. Because we normally think that melting occurs through melting process. New maps generated over the past 25 years heating (e.g., putting a slab of butter in a frying pan), it may show the variations in shape and depth of thousands of volca- seem paradoxical to say the earth melts while cooling down. noes distributed along the ridge. These maps have enabled sci- The explanation for this paradox is that melting temperatures entists to sample the volcanic rocks—ocean-ridge basalt—that are dependent on pressure as well as temperature. Just as in- make up the surface pavement of the oceanic crust. About creased temperature excites atoms so they free themselves from 80 percent of the 60,000-kilometer-long mid-ocean ridge their ordered, solid, crystalline state, so increased pressure has been mapped and sampled at least to 100-km spacing T squeezes atoms, making it more diffi cult for them to transi- (see www.petdb.org). At the same time, experimental studies MD 20849-1931, USA. 1931, Rockville, Box Society, PO e Oceanography tion from solid to liquid. Thus, temperature and pressure exert have led to quantitative models of how melt composition and opposite effects on melting, and melting can occur by decreas- amount vary with temperature and pressure (e.g., Jacques and ing pressure at a given temperature as well as by increasing Green, 1980; Kinzler and Grove, 1992; Baker and Stolper, 1994; temperature at a given pressure. The reason melting by pres- Pickering-Witter and Johnston, 2000). And seismic studies, sure release seems foreign to common experience is because which are able to probe Earth’s interior directly, provide infor- human life on Earth is lived in an environment of almost con- mation about the “melting regime” beneath the ridge axis. This stant pressure caused by the weight only of the atmosphere. article synthesizes some of these developments, and outlines a The solid earth, however, is subject to huge changes in pressure, set of major questions for future research. because the weight of hundreds of kilometers of rock exerts

78 Oceanography Vol. 20, No. 1 AXIS 0 15% 4 Ocean Crust

5% 30% 3 50 Solidus 20%

2 pth (km) De 10% 1 100 Solidus

Cold Ascending Mantle Hot Ascending Mantle

Temperature Temperature

4

3 3

Pr 2 Pr 2 essu essu

Ma Ma re re nt nt le So le Solidu 1 1 lidu

As Ma s 30% s 30% cending nt 20% 20%

le As 10% Ma 10%

cend

0% nt

le

ing 0%

Figure 1. Diagrams illustrating the melting mechanisms beneath ocean ridges. At any one pressure, the mantle melts over a temperature range of several hundred degrees. T e boundary between melt absent and melt present is called the mantle solidus. As mantle ascends beneath the ocean ridge, it begins melting as the solidus is crossed, and melts progressively dur- ing further ascent. T us, the mantle melts by pressure decrease rather than by temperature increase. Hot mantle crosses the solidus at greater depths, leading to a larger melting regime, greater extents of melting, and thicker crust than that produced by cold mantle. T e numbers on the bottom diagrams correspond to the pressures where melting stops for the numbered fl ow lines on the upper diagrams.

Oceanography March 2007 79 A FIRSTORDER MODEL FOR larger range of pressures, leading to the water. The same principle applies on OCEANRIDGE MELTING greater extents of melting. Here, then, Earth to crust “floating” on the denser The deep mantle is solid, but not brittle. the common-sense intuition holds true: mantle. Oceanic crust is denser and At the high temperatures and pres- hot mantle melts more than cold man- thinner than continental crust, and for sures characteristic of Earth’s interior, tle. The greater quantity of melt from this reason the ocean floor is generally the mantle beneath the plates flows like hot mantle thus produces thicker oce- at a lower elevation than continents. a very viscous liquid at rates of up to anic crust than is produced from colder Variations in the thickness of the oceanic several tens of centimeters per year. As mantle (see Figure 1). All of this can be crust along ocean ridges lead to varia- the rigid plates separate at mid-ocean understood as a consequence of how the tions in the elevation of the ocean floor, ridges, the deeper mantle rises to fill the pressure-temperature diagram relates with thick crust, in areas such as Iceland, “gap” created by spreading. The ascend- to the melting regime created by mantle actually rising above sea level, and very ing mantle crosses its melting point flow driven by plate separation. thin crust lying as much as 5000 meters and begins to melt. The mantle-melting The actual extent of mantle melt- below sea level. region beneath the ridge, the “melting ing can be estimated from the chemical Putting these considerations together, regime,” is roughly triangular in shape compositions of the basalts that rise to the first-order prediction is that hotter (see Figure 1). The total amount of melt the surface and are sampled at ocean mantle leads to lower sodium content, that can be produced by any particular ridges. Elements that are preferentially thicker crust, and shallow water depths, part of the mantle within the melting concentrated into the liquid (that is to and colder mantle to higher sodium con- regime is proportional to how far this say, elements that are incompatible with tents, thinner crust, and greater depths. mantle rises after crossing the solidus. the crystals remaining in the solid man- Observations are in agreement with this The melting regime ranges in extent of tle, called “magmaphile” elements) have prediction, as Figure 2 shows. Quanti- melting, therefore, from zero at the bot- concentrations that are inversely propor- tative models of this simple picture of tom where the mantle begins to melt, tional to the extent of melting. The most mantle melting can to first order suc- to a maximum at the shallowest point abundant element with this behavior is cessfully account for the composition of melting. The remarkable fact is that sodium. High extents of melting lead to and thickness of the oceanic crust, and as the mantle melts more and more, it liquids with low sodium concentrations, the global variations in depths of the is at lower and lower temperatures, so it and low extents of melting to high con- ocean ridges (Klein and Langmuir, 1987; actually melts while cooling down rather centrations, because most of the “incom- Langmuir et al., 1992). Thus, a simple than while heating up. patible” sodium is partitioned into the bathymetric map of the ocean-ridge Because the melts produced in the first small melt fraction. Further melting system could be considered to reflect melting regime are buoyant and fluid, then dilutes the sodium concentration. largely the temperature structure of the they separate from the solid and rise to A physical measure of the extent of underlying mantle. the surface to form the oceanic crust. melting is the amount of crust produced If the mantle is hotter, it starts to melt per increment of spreading, which is COMPLEXITIES OF THE deeper, and therefore can melt over a the crustal thickness. Crustal-thickness MELTING PROCESS measurements are difficult and expensive This first-order understanding is only an CHARLES H. LANGMUIR (langmuir@ because they require seismic experiments initial approach to what we now know is eps.harvard.edu) is Professor, Department using instruments deployed on the sea- a more diverse and complex set of pro- of Earth and Planetary Sciences, Harvard floor. A useful proxy for crustal thick- cesses, such as the complexities of mantle University, Cambridge, MA, USA. DONALD ness comes from Archimedes’s buoyancy flow, mantle composition, and the de- W. FORSYTH is Professor, Department principle: A thick piece of wood sticks tailed processes of melt segregation. Let of Geological Sciences, Brown University, up higher out of the water than a thin us consider some factors that are not tak- Providence, RI, USA. piece, and also extends deeper below en into account by the first-order model.

80 Oceanography Vol. 20, No. 1 4 some scale (Allegre and Turcotte, 1986). COLD, LESS Major chemical heterogeneities can also MELTING be caused by other mantle processes 3.5 such as recycling back into the mantle of the cold lithosphere beneath continents, and within the mantle the movement of 3 melts that do not reach the surface. 8.0

Na Another important aspect of mantle

2.5 composition that affects how it melts beneath a spreading center is its con- HOT, MORE tent of volatiles, principally water and MELTING Normal Ridges carbon dioxide. Both have very low 2 Hot Spot Margins melting temperatures, and addition of Hot Spot Centers these compounds to the mantle can sub- 1.5 0 1000 2000 3000 4000 5000 6000 stantially increase the pressure where Depth (m) melting begins, as we will examine in Figure 2. Plots of average compositions of ocean-ridge basalts (each point represents more detail below.

about 100 km of ridge length) vs. the average depth of the ridge. Na8.0 is the composi- tion of basalt normalized to a constant MgO content of 8 wt.% to correct for shallow- level differentiation. High Na contents reflect small extents of melting, while lower Na Variations in Spreading Rate contents reflect higher extents of melting. High extents of melting lead to low Na con- Ridges vary in the rate at which they tents, greater crustal thickness, and shallower depths below sea level, consistent with a model of varying mantle temperature. After Langmuir et al., 1992 produce new seafloor, from less than 10 mm yr-1 to nearly 200 mm yr-1. At spreading rates of 100 mm yr-1, the up- welling mantle rises from the depth of melting onset to the surface in only Complexities of Melt Segregation Variations in Mantle Composition about one million years. At 10 mm yr-1, The model assumes that melt is delivered Sodium contents of erupted magmas it takes about ten million years, long to the surface without significant inter- are influenced not only by the extent of enough for the mantle to lose heat by action with the surrounding mantle and melting, but also by source composition. conduction to the surface while it is still crustal rocks that it traverses. Although Although the mean mantle composition rising (Figure 3). At the slowest spread- this assumption may seem simplistic, is quite well constrained (McDonough ing rates, therefore, the extent of melting it became more conceivable with the and Sun, 1995), the operation of plate decreases, and the average depth of melt- discovery that melt can be transported tectonics inevitably leads to variations on ing increases, compared to fast-spreading through the mantle in channels of pure a variety of scales, called mantle hetero- rates (Shen and Forsyth, 1995). olivine (Kelemen et al., 1995), a mineral geneity. Melting beneath an ocean ridge that has little effect on chemical compo- creates some 6 km of crust enriched in Tectonic Complexities sition. The potential chemical complexi- elements such as sodium and titanium, The Figure 1 cartoon shows a two- ties and ramifications of melt transport and 60–100 km of mantle that is de- dimensional slice across a spreading are still only beginning to be understood, pleted in these elements. When this plate ridge, which is assumed to extend long however, and a full model of mantle is recycled into the mantle at convergent distances along the ridge axis. The real melting must ultimately consider both margins, these heterogeneities gradually world is far more interesting. Ridges are melt generation and melt transport and become mixed, but differences in density offset by transform faults every 100 km the chemical consequences of each. and stiffness will preserve variations on or so, creating a segmented fabric, as

Oceanography March 2007 81 shown in Figure 4. Upwelling is expected cally by “hotspots,” such as those found beneath the ridge, and a different rela- to be slowed in the vicinity of transform near Iceland, the Galápagos, and the tionship between extent of melting and faults, and the rising mantle should be Azores islands (see Dyment et al., this crustal thickness. And there is evidence cooled by proximity to the colder, older issue). Most scientists consider hotspots from trace elements and radiogenic iso- lithosphere across the transform (Fox to be generated by hot plumes rising topes that the mantle at hotspots may and Gallo, 1984; Bender et al., 1984). from the deep mantle in “active” mantle have a composition different from that Thus, on a local scale, we would also flow, rather than the passive mantle found under normal ridges, creating an expect a truncation of the top of the flow at ridges that we considered in the additional complexity (e.g., Schilling, melting regime. first-order model. Active flow generates 1975; Schilling et al., 1982). Ridges are also punctuated periodi- a different pattern of mantle upwelling Finally, there are the spreading axes that are closely associated with con- vergent margins, called back-arc basin spreading centers (see Martinez et al., this issue). Down-going slabs influence the pattern of mantle flow and prevent the kind of simple upwelling seen in Figure 1. Back-arc spreading centers are also influenced by the flux of water and other elements that come from the slab as it subducts (e.g., Gill, 1976; Sinton and Fryer, 1987; Stolper and Newman, 1994; Taylor and Martinez, 2003). Simple consideration of the geometry also indi- cates that there may not be enough room in the mantle wedge above the slab to accommodate a melting regime such as is observed at open-ocean ridges (Kelley et al., 2006; Langmuir et al., 2006b).

TESTING MULTI AXIS AXIS DIMENSIONAL CONTROLS Oceanean Cr Crusust t Ocean Crust ON MANTLE MELTING

15% This overview of the diverse ocean- ridge environments shows that there 10% 10% are many “forcing functions” that in- 5% 5% fluence the ridge. Understanding the diverse influences of all these forc- Solidus Solidus ing functions continues to be a focus of ocean-ridge research. Figure 3. Map of the Arctic Ocean’s Gakkel Ridge, which is the slowest major spreading ridge on Earth. The classic approach to such ques- Spreading rate decreases progressively towards Siberia, as evident from the narrowing of the basin tions in many other areas of scientific created by the spreading (delimited by the red lines). As spreading rate declines, slower upwelling prevents melting all the way to the surface, and the melting regime becomes progressively truncated, research is to carry out experiments leading to a melting trapezoid rather than a melting triangle such as seen in Figure 1. in the laboratory where the boundary

82 Oceanography Vol. 20, No. 1 conditions can be controlled, and the influences on mantle melting and crust place, traversing the platform created by experiment completed in weeks to years. formation. Many other examples from the Galápagos hotspot in the equatorial Much research in earth science cannot be other portions of the ocean-ridge system Pacific (Cushman et al., 2004). addressed in this way because the rele- can be found in the literature. Around the Azores and Galápagos, the vant scales of time and space are millions ridge varies in depth from greater than of years and hundreds of kilometers. It Transects Across the Azores 3000 m some thousand kilometers from is not possible to go to a ridge and turn and Galápagos Platforms the hotspot to less than 1500 m where the up the spreading rate or turn down the The first project of the InterRidge pro- ridge most closely approaches the hot- mantle temperature to see what happens gram was a targeted series of investiga- spot. In earlier studies of these regions, or to build a functioning ridge in the lab. tions across the Azores platform in the Schilling et al. (1980, 1982) showed that Instead, earth scientists have to make use Atlantic Ocean to see how a near-ridge the contents of volatiles, such as water of “natural experiments” provided by hotspot influenced the ridge (Detrick et in the spreading-axis magmas, increased the earth. The following are examples of al., 1995; Langmuir et al., 1997; Asimow toward the hotspots’ centers as ridge such experiments and a brief discussion et al., 2004). More recently, a similar in- depth shallowed, and that “hotspots” of what they reveal about the diverse vestigation of the Galápagos Rise took were also “wet spots.” Bonatti (1990)

-3˚00'

-3˚30'

-4˚00'

-4˚30'

-5˚00'

253˚30' 254˚00' 254˚30' 255˚00' 255˚30' 256˚00' 256˚30' 257˚00' 257˚30'258˚00'

2400 2500 2600 2700 2800 2900 3000 3100 3200 3300 3400 3500 3600 3700 3800 3900 4000 4100 4200 4300 4400 4500 4600 4700 4800 Bathymetry (m) Figure 4. Map of the East Pacific Rise where there are large transform offsets, illustrating the three-dimensional complexity of the ridge system. Melting regimes cannot be continuous across such long transforms, and the cooling effects of the transforms also truncate the top of the melting regime, leading to melting trapezoids near the transform edges, such as seen in Figure 3. From Forsyth et al. (2006)

Oceanography March 2007 83 inferred that mantle temperatures Water is significant to mantle melting partitioned into the first melts formed, beneath these shallow ridges might even because it acts as a melting flux in the its influence decreases markedly with be colder than usual, rather than hotter. mantle and is a “carrier phase” for mag- increasing extent of melting. This leads Schilling also showed that concentra- maphilic elements. The melting point of to very low melt production in the deep tions of various magmaphile elements the mantle is substantially lowered by the part of the melting regime, and a very also substantially increased, and that the addition of water, even in small amounts. different distribution of melt with depth. gradient in water depth was associated Each 0.1 percent of water added to the Therefore, water and temperature both with a gradient in chemical composition. mantle lowers the melting point of the lead to increases in total melt production The natural experiment here, then, was first liquid produced by 150°C–250°C and crustal thickness, and therefore shal- to explore the influence of a hotspot on (e.g., Gaetani and Grove, 1998; Katz et lower ridge depths, but the effect of wa- the spreading center. From the “hot,” one al., 2003 and references therein). If this ter is in the deep, low-degree melts. would infer higher mantle temperatures, effect is added to the melting diagram Experimental data quantify the effects and yet the mantle composition, includ- shown in Figure 1 (Figure 5), it becomes of water on mantle melting (e.g., Gaetani ing the particularly significant volatile evident that water causes melting to and Grove, 1998). The challenge is to abundances, changes towards the hot- begin at much higher pressures and leads produce models that yield both the cor- spot. Both water and temperature aug- to greater extents of melting at the top of rect crustal thickness and contents of ment the extent of mantle melting. Can the melting regime (Hirth and Kohlstedt, water and other elements in the magmas the relative importance of these variables 1996; Asimow and Langmuir, 2003). at the hotspot center. Such models show be separated? Because water, like sodium, is strongly that the shallow depths and increased

Melt Fraction 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 a 0 b Mean 12 5 Dry to Wet, Melt 35°C T Fractions difference Dry Only, 10 75°C T 11 difference 15 Azores Platform Crust 10 20 re

25 ickness (km) essu

T 9

Pr Wet melting with 700 ppm Dry melting 30 H O

ustal 2 Cr 35 8

40 7 45 1000 ppm water Normal Atlantic Crust 100 ppm water 50 6 1325 1350 1375 1400 1425 1450 1475 Potential Temperature °C

Figure 5. Illustration of the effects of water on melting beneath ridges. Te left panel shows that addition of water creates a deep “tail” of low extents of melting, which contributes additional melt and causes greater crustal thickness. Although adding water causes the maximum extent and total amount of melt to increase, the average extent of melting across the whole depth of melting decreases because of the large, deep region of low-degree melts. Te right panel shows how this effect can help to explain data from the Azores hotspot. Having a water-rich mantle source reduces the needed temperature excess to explain crustal thickness variations from about 75°C to 35°C. Right-hand panel modified from Asimow et al (2004)

84 Oceanography Vol. 20, No. 1 crustal thickness are the result of both perature structure. Direct ground-truth the surface (Figure 6). This lack of a hotspot and wet-spot effects (Asimow tests of how the mantle actually melts deep root confirms that upwelling at et al., 2004). Temperature differences require imaging the melting regime this spreading center is passively driven alone would suggest that the ridge near beneath the ridge axis. The Mantle Elec- by plate separation rather than by active the Azores was some 75°C hotter than tromagnetic and Tomography (MELT) convective upwelling. Seismic shear- normal. Inclusion of water shows the seismic experiment was designed to im- wave velocities dropped dramatically at addition of some 750 parts per mil- age the mantle beneath a spreading axis depths above 100–150 km, indicating the lion of water to the mantle leads to the using seismic and electromagnetic meth- onset of melting (shear or “s” waves can- appropriate chemical compositions ods (MELT Seismic Team, 1998). The not propagate through liquid). Because and crustal thickness, and this reduces ideal location for this investigation was experiments on peridotite melting show the temperature differences required to the equatorial East Pacific Rise, which is that volatile-free mantle should begin to some 35°C (Asimow et al., 2004). characterized by a long ridge with few melt only at 60–80 km, the presence of transform offsets, a high spreading rate melt at greater depths inferred from the The MELT Seismic Experiment on and therefore maximum rate of melt MELT results indicate that, even in the the Southern East Pacific Rise production, and good weather. absence of a hotspot or wet-spot influ- The various models of mantle melting The experiment showed that, unlike ence, the effects of water on melting outlined above are based on information beneath hotspot islands, there was no regime are important. derived from experiments on peridotite deep “root” of anomalously hot mantle Some theoretical models had sug- melting and calculations of mantle tem- extending hundreds of kilometers below gested that once melting began, mantle

West DISTANCE FROM AXIS (km) East 400 200 0 200 400

101 mm/yr 45 mm/yr Crust

0 Figure 6. Schematic of results from the MELT experiment, showing that melting begins at about 150-km depth and that the melting regime extends preferentially Lithospheric to the west of the East Pacific Rise. Tis Mantle experiment was the first attempt to image an ocean-ridge melting regime, testing 100 the conceptual cartoons seen in Figure 1. Embedded Modified from the MELT Seismic Team, 1998 Heterogeneity Primary Melting

DEPTH (km) Incipient 200 Melting

300

400 410 km Discontinuity

Oceanography March 2007 85 upwelling might concentrate into a very (see Snow and Edmonds, this issue). The primary spreading centers (Figure 4). narrow zone directly beneath the spread- Gakkel is the slowest-spreading ridge This area provides a natural laboratory ing axis. Instead, the low-velocity zone in the world, with spreading rates from to test models of the three-dimensional was more than 100 km across, consistent 6–15 mm yr-1. It has no transform offsets pattern of mantle flow and melt migra- with the roughly triangularly shaped re- or hotspots that would lead to perturba- tion. If melting occurs in broad, trian- gion of melting shown in Figure 1. The tions in mantle temperature. This simple gular zones beneath the ridges as shown electromagnetic experiment showed geometry is ideal for an investigation in Figure 1, how does the melt from the that the upper 60 to 70 km of the mantle of the role of spreading rate on mantle distal portions of the melting regime away from the spreading center itself is melting (Michael et al., 2003). Simple migrate laterally back to the ridge axis? highly resistive, indicating that the man- thermal models suggest that at these Several ideas have been suggested, such tle above this depth has been depleted of low spreading rates there would be a as melt migrating vertically to the top water by the removal of melt (Evans et substantially increased lid of cold litho- of the melting region, then flowing up al., 2005). Perhaps the most surprising sphere and lower extents of melting. along the base of the sloping lithosphere result was the degree of asymmetry of If variations in lithospheric thickness back to the spreading center (Magde and structure beneath the East Pacific Rise, rather than mantle temperature cause Sparks, 1997). Another suggestion is that with lower seismic velocities, shallower the variations in crustal thickness and the melts are driven by pressure gradi- seafloor, and more pronounced seismic magma chemistry observed along ridges, ents within the deforming solid mantle anisotropy associated with alignment then variations along the Gakkel Ridge (Phipps Morgan, 1987; Spiegelman and of olivine crystals. This result suggested should mirror global trends. Or, can the McKenzie, 1987). These models make that deeper flow in the mantle (to supply important variables of spreading rate different predictions for the composi- the material flowing in to fill the gap left and mantle temperature be separated? tion and volume of melt that would by the separating plates) comes primarily Analyses of the wealth of data from this be delivered to each intra-transform from the west, perhaps supplied by hot- expedition are well underway. The criti- spreading center. spots beneath Tahiti and other islands cal chemical parameters turned out to A research cruise in April 2006 (For- far from the spreading center (Mahoney be the basalts’ iron and silica concentra- syth et al., 2006) mapped the bathymetry et al., 1994; Phipps Morgan et al., 1995; tions, and the distribution of the rare in detail, made gravity measurements Toomey et al., 2002; Conder et al., 2002). earth elements, whose pattern of distri- to estimate differences in crustal thick- bution is sensitive to melting pressure. ness, and sampled basalts. The composi- Investigation of the Gakkel Ridge, The results clearly show the effects of tion of the basalts will reveal the relative Arctic Ocean increased lithospheric thickness on the contributions of deep and shallow melt- One of the most interesting spreading melting regime, as distinct from those ing to the crust that is formed at each centers on Earth, the Gakkel Ridge, lies caused by mantle temperature differ- intra-transform center, thus helping to beneath the ice of the Arctic Ocean and ences (Langmuir et al., 2006b). decipher the plumbing system that pipes extends from Greenland to Siberia (see magma out of the mantle and the effect Figure 3). Studying this ridge is fraught The Quebrada-Discovery-Gofar of transform offsets on melt production. with operational difficulties. InterRidge Fracture Zone System scientists invested considerable time and The Quebrada, Discovery, and Gofar The Lau Basin Back-Arc effort in devising a plan to begin study- fracture zones are a set of transform Spreading Center ing this ridge, a plan that came to frui- faults on the fast-spreading East Pacific Back-arc basins provide another natural tion in 2001 with the two-ship AMORE Rise. Within each transform, there are experiment for ocean-ridge processes expedition involving the new US ice- short, intra-transform spreading cen- because of their unique thermal and breaker, the Coast Guard cutter Healy, ters from 5–15 km in length, which are tectonic environments. The Lau back- and the German icebreaker Polarstern offset by different distances from the arc basin formed behind the northeast-

86 Oceanography Vol. 20, No. 1 southwest trending Tonga volcanic arc, the arc front also increases. These condi- understanding of this region. which itself was formed by subduction tions make the Lau Basin an ideal back- Back-arc basins are important from of the Pacific Plate at the Tonga trench arc environment for addressing issues the mantle melting perspective because (Figure 7). Situated west of the arc, the such as transport of fluid components water is added from dehydration of the southern Lau Basin is relatively young from slab to mantle wedge, the timing of subducting slab, and the influence of (< 5.5 million years old) oceanic crust these transport processes, and the influ- water contrasts markedly with its influ- created by two main rifts, the Central ence of slab fluids on mantle melting. To ence at open-ocean ridges. Plotting an Lau Spreading Center and the Eastern take advantage of this unique tectonic index of the extent of melting (such as

Lau Spreading Center. From south to setting, the US Ridge 2000 program the TiO2 or Na2O contents) vs. H2O con- north, the spreading rate increases from designated the Lau Basin an “integrated tents, indicates opposite trends in the 65 mm yr-1 at 21°S to 90 mm yr-1 at study site.” Japanese, British, German, back-arc and open ocean. This result can 18°S (Taylor et al., 1996; Martinez et al., and Australian investigators have also be understood visually by placing the 2006), as distance between the ridge and made substantial contributions to the melting regime seen in Figure 1 within


 
 
  
 Figure 7. Te left-hand panel shows the Lau Back-Arc Basin cre- 1000 ated by spreading along the Eastern Lau Spreading Center (ELSC) 500 0 behind the Tonga arc, where the Pacific Plate is being subducted. -500 -1000 Te right-hand panel shows a cross section at about 22°S. Te grey -1500 -2000 diamonds are earthquake locations that indicate the position of CLSC -2500 Late -3000 the cold, subducting plate. Tere is no room for a triangular melt- -3500 -4000 ing regime in this environment—the slab truncates it. Tus, the 
 -4500 effects of hydrous melting seen in Figure 5, which are the result of -5000 ILSC -5500 the deep “wings” of the melting regime, are prevented from tak- -20 00 0 -6000 0 0 -2 -6500 ing place in the back-arc environment. Map from Martinez et al. -7000 - 2 0 0 00 -7500 (2006); right-hand panel modified from Langmuir et al. (2006a) -200 Kao -8000 -8500 Tofua 0 -9000 -9500 -10000  -10500 !)*%(&$* &% *(% "$ -11000   


Nomuka 00 - 20 2000 - (,

0 $#*!% 0 0 -2 Arc
* ga ELSC n $#*!%

0 -2000 00 
 -2 To

0

-2

0 00 0 Tongatapu 0 0

-6  ench Tr '* "$

0 -2 0 0 0 -20 0 0 -2 00 nga 4 00 -

0 

-2  To Valu Fa  Ata

 +)!% Map after Martinez and TaTaylorylor (2002) #+(!     

Oceanography March 2007 87 the context of a back-arc such as the science remains a frontier. ing that incorporate mantle flow, melt southern Lau Basin (Figure 7b). There With sustained funding, it seems like- migration, and crystallization in the is no room for it! In the open ocean, the ly that over the next 10–15 years, better context of three-dimensional mantle effects of water lowering the mean ex- constraints from all parts of the system flow and tectonic environments that tent of melting come from the “wings,” will lead to an integrated understanding reflect the fascinating complexities of or distal edges, of the melting regime. of mantle melting beneath ridges. These the real earth. And in the back-arc, there is no room for constraints will likely arise from the fol- the extremities of the melting regime on lowing complement of directions: ACKNOWLEDGEMENTS the arc side of the system, exactly where 1. Although there are many samples This manuscript benefited from reviews the subducting plate might be introduc- from ocean ridges, few of them have by Colin Devey and John Sinton, and the ing water into the mantle. Therefore, in relatively complete geochemical anal- careful editing of Kristen Kusek, whose the back-arc it seems likely that melts yses, which are crucial to test models efforts are much appreciated. Research from the two halves of the melting re- of melting and mantle heterogeneity. was supported by the National Science gime have very different characteristics. Comprehensive data sets are likely to Foundation. On the back-arc, or dry side, melting provide much clearer constraints. is similar to open-ocean ridges. On the 2. Over the next 10–15 years, the ridges REFERENCES arc side, or wet side, the flux of water may become completely sampled on Allegre, C., and D.L. Turcotte. 1986. Implications of a two component marble cake mantle. Nature occurs at shallow pressures, leading to a global scale. Several “natural experi- 323:123–127. substantially increased melting. Mixing ments” remain in logistically remote Asimow, P.D., and C.H. Langmuir. 2003. The im- of melts from these two diverse environ- regions that will provide a broader portance of water to oceanic mantle melting regimes. Nature 421:815–820. ments may create the distinctive back- range of “forcing functions” with Asimow, P.D., J.E. Dixon, and C.H. Langmuir. 2004. arc data arrays. which to constrain melting models. A hydrous melting and fractionation model 3. Imaging of the ridge on global and for mid-ocean ridge basalts: Application to the Mid-Atlantic ridge near the Azores. Geochem- FUTURE PERSPECTIVES local scales will be improved by steady istry, Geophysics and Geosystems 5:Q01E16, Ocean ridges are the largest tectonic progress in seismology. Global seismic doi:10.1029/2003GC000568. landforms on earth. They are remark- arrays will lead to much better con- Baker, M.B., and E.M. Stolper. 1994. Determining the composition of high-pressure mantle melts ably inaccessible because they are far straints on temperature, melt produc- using diamond aggregates. Geochimica Cosmo- from land, hidden beneath thousands tion, and mantle flow. Local seismic chimica Acta 58:2,811–2,827. of meters of water, and often lie in the experiments focused on different Bender, J.F., C.H. Langmuir, and G.N. Hanson. 1984. Petrogenesis of basalt glasses from the most remote portions of the globe where ridges with the full range of spreading Tamayo region, East Pacific Rise. Journal of Pe- weather conditions are some of the most characteristics will enable imaging of trology 25:213–254. difficult. While studies over the last three the melting regime and how it varies. Bonatti, E. 1990. Not so hot “hot spots” in the oce- anic mantle. Science 250:107–111. decades have led to substantial progress, 4. More experimental data, particularly Conder J.A., D.W. Forsyth, and E.M. Parmentier. we are still at a very early stage of under- on the influence of volatiles on melt 2002. Asthenospheric flow and asymmetry of standing. No ocean-ridge volcano has compositions and the consequences the East Pacific Rise, MELT area. Journal of Geo- physical Research 107(B12):2344, doi:10.1029/ anywhere near the monitoring and his- of melting of a lithologically heteroge- 2001JB000807. torical record that are common on land. neous mantle, will provide constraints Cushman, B., J. Sinton, G. Ito, and J.E. Dixon. For example, the Smithsonian Catalogue on how melt composition varies with 2004. Glass compositions, plume-ridge interac- tion, and hydrous melting along the Galápagos of Volcanoes of the World (Simkin the composition and lithology of the Spreading Center, 90.5°W to 98°W. Geochem- and Siebert, 1994) reports thousands mantle source. istry, Geophysics and Geosystems 5:Q08E17, of eruptions from subaerial volcanoes. 5. All of these data can be best assimi- doi:10.1029/2004GC000709. Detrick, R.S., H.D. Needham, and V. Renard. 1995. In contrast, only a handful of under- lated and new hypotheses generated Gravity anomalies and crustal thickness varia- sea eruptions are known. Ocean-ridge by integrated models of mantle melt- tions along the Mid-Atlantic Ridge between

88 Oceanography Vol. 20, No. 1 33°N and 40°N. Journal of Geophysical Research Morgan, D.K. Blackman, and J.M. Sinton, eds, 140:190–211. 100:3,767–3,788. American Geophysical Union, Washington, DC. Schilling, J.G. 1975. Azores mantle blob: Rare-earth Evans, R.L., G. Hirth, K. Baba, D. Forsyth, A. Chave, Langmuir, C.H., A. Bézos, S. Escrig, and S.W. Par- evidence. Earth and Planetary Science Letters and R. Mackie. 2005. Geophysical evidence man. 2006a. Chemical systematics and hydrous 25:103–115. from the MELT area for compositional controls melting of the mantle in back-arc basins, Pp. Schilling, J.G., M.B. Bergeron, and R. Evans. 1980. on oceanic plates. Nature 437:249–252. 87–146 in Back-Arc Spreading Systems: Geologi- Halogens in the mantle beneath the North-At- Forsyth, D.W., N. Harmon, R.C. Pickle, and A. Saal. cal, Biological, Chemical, and Physical Interac- lantic. Philosophical Transactions of the Royal 2006. Stability and instability in an evolving tions. D.M. Christie, C.R. Fisher, S.-M. Lee, and Society, Series A: Mathematical, Physical, and oceanic transform fault system. Eos, Transac- S. Givens, eds, Geophysical Monograph Series, Engineering Sciences 297:147–178. tions, American Geophysical Union 87(52), Fall Volume 166, American Geophysical Union, Schilling, J.G., H. Kingsley, and J.D. Devine. 1982. Meeting Supplement, Abstract T42C-01. Washington, DC. Galapagos hot spot - spreading center system. 1. Fox, P.J., and D.G. Gallo. 1984. A tectonic model for Langmuir, C.H., J. Standish, P. Michael, and S. Spatial petrological and geochemical variations ridge-transform-ridge plate boundaries: Impli- Goldstein. 2006b. Constraints on ocean ridge (83°W–101°W). Journal of Geophysical Research cations for the structure of oceanic lithosphere. basalt generation from Gakkel Ridge basalts. 87:5,593–5,610. Tectonophysics 104(3-4): 205–242. Eos, Transactions, American Geophysical Union Shen, Y., and D.W. Forsyth. 1995. Geochemical Gaetani, G.A., and T.L. Grove. 1998. The influ- 87(52), Fall Meeting Supplement, Abstract constraints on initial and final depths of melt- ence of water on melting of mantle perido- V12c-01. ing beneath ocean ridges. Journal of Geophysical tite. Contributions to Mineralogical Petrology Magde, L.S., and D.W. Sparks. 1997. Three-di- Research 100 (B2):2,211–2,237. 131:323–346. mensional mantle upwelling, melt generation Simkin, T., and L. Siebert. 1994. Volcanoes of the Gill, J.B. 1976. Composition and age of Lau Basin and melt migration beneath segmented slow- World, second edition. Geoscience Press, Inc., and ridge volcanic rocks: Implications for evo- spreading ridges. Journal of Geophysical Re- Tucson, Arizona, 349 pp. lution of an interarc basin and remnant arc. search 102(B9):20,571–20,585. Sinton, J.M., and P. Fryer. 1987. Mariana Trough Geological Society of America Bulletin 87:1,384– Mahoney, J.J., J.M. Sinton, M. Kurz, J.D. Macdou- lavas from 18°N: Implications for the origin of 1,395. gall, K.J. Spencer, and G.W. Lugmair. 1994. back-arc basin basalts. Journal of Geophysical Hirth, G., and D.L. Kohlstedt. 1996. Water in the Isotopic and trace element characteristics of Research 92:12,782–12,802. oceanic upper mantle: Implications for rheol- a super-fast spreading ridge: East Pacific Rise, Spiegelman, M., and D. McKenzie. 1987. Simple ogy, melt extraction and the evolution of the 13-23°S, Earth and Planetary Science Letters 121: 2-D models for melt extraction at mid-ocean lithosphere. Earth and Planetary Science Letters 173-193. ridges and island arcs. Earth and Planetary Sci- 144:93–108. Martinez, F., B. Taylor, E.T. Baker, J.A. Resing, and ence Letters 83:137–152. Jaques, A.L., and D.H. Green. 1980. Anhydrous S.L. Walker. 2006. Opposing trends in crustal Stolper, E., and S. Newman 1994. The role of water melting of peridotite at 0–15kb pressure and thickness and spreading rate along the back-arc in the petrogenesis of Mariana Trough magmas. genesis of tholeiitic basalts. Contributions to Eastern Lau Spreading Center: Implications for Earth and Planetary Science Letters 121:293–325. Mineralogical Petrology 73:287–310. controls on ridge morphology, faulting, and hy- Taylor, B., and F. Martinez. 2003. Back-arc basin Katz, R.F., M. Spiegelman, and C.H. Langmuir. drothermal activity. Earth and Planetary Science basalt systematics. Earth and Planetary Science 2003. A new parameterization of hydrous man- Letters 245:665–672. Letters 210:481–497. tle melting. Geochemistry, Geophysics, Geosys- McDonough, W.F., and S.-S. Sun. 1995. Com- Taylor, B., K. Zellmer, F. Martinez, and A. Good- tems 4:9, doi:10.1029/2002GC000433. position of the Earth. Chemical Geology liff. 1996. Sea-floor spreading in the Lau back- Kelemen, P.B., N. Shimizu, and V.J.M. Salters. 1995. 120:223–253. arc basin. Earth and Planetary Science Letters Extraction of MORB from the mantle by fo- MELT Seismic Team. 1998. Overview of the seis- 144:1, 35–40. cused flow of melt in dunite channels. Nature mological component of the MELT Experiment. Toomey, D.R., W.S.D. Wilcock, J.A. Conder, D.W. 375:747–753. Science 280:1,215–1,218. Forsyth, J. Blundy, E.M. Parmentier, and W.C. Kelley, K.A, T. Plank, T.L. Grove, E.M. Stolper, S. Michael, P.J., C.H. Langmuir, H.J.B. Dick, J.E. Snow, Hammond. 2002. Asymmetric mantle dynam- Newman, and E. Hauri. 2006. Mantle melting S.L. Goldstein, D.W. Graham, K. Lehnert, G. ics in the MELT region of the East Pacific Rise. as a function of water content beneath back- Kurras, W. Jokat, R. Mühe, and H.N. Edmonds. Earth and Planetary Science Letters 200:287–295. arc basins. Journal of Geophysical Research 111: 2003. Magmatic and amagmatic seafloor gener- B09208, doi:10.1029/2005JB003732. ation at the ultraslow-spreading Gakkel Ridge, Kinzler, R.J., and Grove, T.L. 1992. Primary mag- Arctic Ocean. Nature 423:956. mas of mid-ocean ridge basalts. 1. Experiments Phipps Morgan, J. 1987. Melt migration beneath and methods. Journal of Geophysical Research 9: mid-ocean spreading centers. Geophysical Re- 6885-6906. search Letters 14:1238–1241. Klein, E.M., and C.H. Langmuir. 1987. Global cor- Phipps Morgan, J., W.J. Morgan, Y.-S. Zhang, and relations of ocean ridge basalt chemistry with W.H.F. Smith. 1995. Observational hints for a axial depth and crustal thickness. Journal of plume-fed, sub-oceanic asthenosphere and its Geophysical Research 92:8,089–8,115. role in mantle convection. Journal of Geophysi- Langmuir, C.H., E.M. Klein, and T. Plank. 1992. cal Research 100:12,753–12,767. Petrological systematics of mid-ocean ridge Pickering-Witter, J., and A.D. Johnston. 2000. The basalts: Constraints on melt generation beneath effect of variable composition on the melt- ocean ridges. Pp. 183–280 in Mantle Flow and ing systematics of fertile peridotitic assem- Melt Generation at Mid-Ocean Ridges. J. Phipps blages. Contributions to Mineralogical Petrology

Oceanography March 2007 89