Contrasting crustal production and rapid mantle ...

LETTER

doi:10.1038/nature09690

Contrasting crustal production and rapid mantle transitions beneath back-arc ridges Robert A. Dunn1 & Fernando Martinez2

The opening of back-arc basins behind subduction zones progresses from initial rifting near the volcanic arc to seafloor spreading1. During this process, the spreading ridge and the volcanic arc separate and lavas erupted at the ridge are predicted to evolve away from being heavily subduction influenced (with high volatile contents derived from the subducting plate)2–5. Current models4,6–8 predict gradational, rather than abrupt, changes in the crust formed along the ridge as the inferred broad melting region beneath it migrates away from heavily subduction-influenced mantle. In contrast, here we show that across-strike and along-strike changes in crustal properties at the Eastern Lau spreading centre are large and abrupt, implying correspondingly large discontinuities in the nature of the mantle supplying melt to the ridge axes. With incremental separation of the ridge axis from the volcanic front of as little as 5 km, seafloor morphology changes from shallower complex volcanic landforms to deeper flat sea floor dominated by linear abyssal hills, upper crustal seismic velocities abruptly increase by over 20%, and gravity anomalies and isostasy indicate crustal thinning of more than 1.9 km. We infer that the abrupt changes in crustal properties reflect rapid evolution of the mantle entrained by the ridge, such that stable, broad triangular upwelling regions, as inferred for mid-ocean ridges9,10, cannot form near the mantle wedge corner. Instead, the observations imply a dynamic process in which the ridge upwelling zone preferentially captures water-rich lowviscosity mantle when it is near the arc. As the ridge moves away from the arc, a tipping point is reached at which that material is rapidly released from the upwelling zone, resulting in rapid changes in the character of the crust formed at the ridge. The Lau back-arc basin is a triangular-shaped extensional basin bordered by the Lau Ridge remnant arc to the west and the active Tofua arc to the east1,2,11. Within the basin, the 400-km-long Eastern Lau spreading centre (ELSC) initially formed by propagating southward with its tip near the arc volcanic front12. Magnetic data13 show that the ELSC is spreading nearly symmetrically but with decreasing total opening rates from north to south (96 mm yr21 to 39 mm yr21); the northern end of the ELSC is presently more than twice as far from the Tofua arc than is the southern end (103 km versus 40 km). The present axis of the ELSC is a key region over which the mantle source composition changes from mid-ocean-ridge basalt (MORB)-like at the Central Lau spreading centre to the north to arc-like at the Valu Fa ridge to the south3. This and other systematic variations in axial crustal properties, such as ridge morphology14 and seismic structure15, correlate with the changing proximity of the ridge to the arc volcanic front3,4. The observations are consistent with a decreasing ‘subduction influence’ in the mantle as the ridge shifts away from the arc. Transitions in alongaxis crustal properties are known to occur near ridge-axis offsets, which discontinuously shift the ridge segments relative to the arc front. However, the across-axis evolution of the crust, which reveals the continuous history of changing mantle influence as each section of ridge has moved further from the arc, has not been investigated. We present results of seismic tomography (Fig. 1) and geophysical studies (Fig. 2) of the ELSC axis and flanks spanning two complete 1

ridge segments and portions of the adjoining segments. In plan view, the upper crust formed on the ELSC is characterized by distinct zones of relatively low, intermediate and high compressional wave velocities (Fig. 1b), which are most probably related to variations in both bulkrock porosity and composition15. The velocity variations spatially correlate with major-element changes in seafloor lavas (Fig. 1c), geochemical estimates of changes in source volatile content (and other arc-related components), and with changes in the vesicularity of seafloor samples2,3,8,16. Therefore we suggest that the seismic structure, via the effects of crustal porosity and composition, provides a record of changes in the initial volatile content of melts fed to the ridge. Changes in upper crustal seismic velocity (Fig. 1b) closely mirror changes in seafloor morphology and depth (Fig. 2a) and gravity (Fig. 2b), and imply abrupt changes in crustal production controlled by the volatile (H2O) content of the mantle17. Together with geochemical data2,3,8,16,18, these observations delineate two distinct crustal domains separated by a variable transition region. In the spreading direction, observed step-like transitions between these domains suggest that abrupt changes in mantle source composition take place over as little as 5 km of incremental spreading. In Fig. 2, the older crust into which the ELSC propagated is labelled as Domain I-type crust. Crust formed at the ELSC is labelled Domains II and III (Figs 1–2). Domain II is composed of crust that was produced soon after the ridge propagated southward near the arc volcanic front. Here, shallow and complex volcanic terrains imply high mantle heterogeneity and enhanced crustal production. The terrain is dominated by short arcuate segments with central highs and distal lows that are concave towards the present ridge axis. Domain II terrains are associated with low seismic velocities (,4.0 km s21 at 1 km depth), implying high porosities19 (.15%) and arc-related mineralogies15. These velocities are similar to velocities previously detected on the Valu Fa ridge to the south15,20 and are anomalously slow compared to mid-ocean ridges, which are typically 5–6 km s21 for this depth (Fig. 1 inset). The ,500-m shallower bathymetry and low gravity values suggest that the crust is on average about 1.9 km thicker than crust produced closer to the ELSC in more recent times. The few rock samples in these off-axis areas range from basalts to andesites2,8,18. The low velocities, geochemistry and thick crust indicate that during this period the ridge upwelling zone beneath the spreading centre entrained a significant degree of slab-derived water. The addition of water to the mantle is expected to increase the degree of melting by lowering the mantle solidus5,17,21, resulting in a thicker crust with more vesicular andesitic lavas2. A current morphologic analogue of Domain II terrains may be found along the Fonualei spreading centre in the northeastern Lau basin. The Valu Fa ridge may also be an active analogue, though it exhibits less variable morphology, suggesting relatively less heterogeneity in the mantle source. Domain III crust is characterized by sea floor ,500 m deeper (at ,2,600 m depth) and linear abyssal hill tectonic fabric typical of intermediate-spreading-rate mid-ocean ridges (Fig. 2a). Domain III spatially correlates with basaltic lavas8 and is further characterized by upper crustal velocities (on average 5 km s21 at 1 km depth) that are closer to typical oceanic upper crustal velocities (Fig. 1 inset); estimated

Department of Geology and Geophysics, University of Hawaii, Honolulu, Hawaii 96822, USA. 2Hawaii Institute of Geophysics and Planetology, University of Hawaii, Honolulu, Hawaii 96822, USA.

1 9 8 | N AT U R E | VO L 4 6 9 | 1 3 J A N U A RY 2 0 1 1

©2011 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH a

c

Doma in III

b

20° 15′ S

ΔVP (%)

SiO2 (%) 55.5

20

55.0

16

54.5

Domain II

20° 30′ S

12

54.0 53.5

8

53.0

6

ain I I

2

4

Dom

0

VP (km s–1) 4

Depth (km)

20° 45′ S

Domain III nELSC

176° 15′ W

50.0 49.5

–16

East Pacific Rise

176° 00′ W

50.5

–12

Juan de Fuca ridge 4

51.5

–4 –8

Valu Fa 21° 00′ S

52.0

51.0

Domain II 2

52.5

0

49.0 48.5

–20

176° 15′ W

176° 00′ W

176° 15′ W

Figure 1 | Layout of the seismic experiment and map view of compressional wave velocity variations in the upper crust. a, Square symbols indicate the positions of 83 ocean-bottom seismometers that recorded P-wave arrivals from 8,763 airgun shots that occurred along the black lines. b, A map-view tomographic image of the P-wave velocity (VP) structure at an average depth of 1 km beneath the seafloor. Dotted lines indicate the boundaries of the different crustal domains as described in Fig. 2 and the text. Velocities are given as a percentage variation from the average (4 km s21). Inset, a comparison of average Domain II and III velocity profiles with multi-channel-seismic-derived VP profiles from the axis of the Valu Fa ridge and the northern Eastern Lau

spreading centre, along with velocities of typical fast- and intermediatespreading-rate crust (after ref. 15). Domain II velocities are similar to those of the Valu Fa ridge and are much lower than for typical mid-ocean ridges; Domain II velocities are consistent with both a thicker seismic layer 2A and lower layer 2B velocities. c, Geological samples collected along the axis of the ridge exhibit major-element, trace-element, and isotopic-ratio variations that closely correlate with the geophysical observations8,16. Here we show SiO2 content (C.H. Langmuir, personal communication) as an indicator of the basalt-to-andesite transition that occurs along the ELSC.

porosities for this domain are ,10% (ref. 19). The deeper sea floor probably reflects an isostatic response to both higher rock density and thinner crust. The highest-velocity regions in the study area correspond to the deep crust formed along the northernmost spreading segment (Fig. 1b). On the basis of the seismic velocity and bathymetry maps, and the geochemical analyses of rocks collected in this area2,8,16, relatively low-volatile-content melts with basaltic compositions may have formed crust along this ridge segment for the past ,0.39 million years (16 km of spreading). A narrow region of Domain III crust is located along the entire axis of the ELSC segment just to the south (Fig. 1b), suggesting a recent switch-over (81 thousand years ago) from a ridge segment that formerly tapped volatile-richer melts to one that now taps relatively volatile-poorer melts. Sandwiched between the two domains is a transitional region, varying from ,5–10 km width, with intermediate-depth abyssal hill fabrics (Fig. 1a) and intermediate upper-crustal seismic velocities (Fig. 2b). Along the northern ELSC segment, the switch to transitional crust occurred over a few kilometres of spreading roughly 0.58 million years ago and a later switch to Domain III crust occurred abruptly at 0.39 million years ago. Along the central ELSC, the corresponding changes occurred at roughly 0.261 million years ago and 0.081 million years ago, respectively. Along the southern ELSC (south of the overlapping spreading centre at 20u 309 S), the ridge axis appears to be currently in the transitional phase and seafloor rock samples with

andesitic compositions are found here8 (Fig. 1c). The exception is that a small northern portion of this ridge section currently appears to tap Domain III melts. Here, the switch from transitional-type crust to Domain III crust (near 20u 399 S), as determined by the seismic image, occurs near a well-documented major-element, trace-element and isotopic geochemical boundary between more heavily subductioninfluenced lavas along the ridge axis to the south to less so along the ridge axis to the north8,16 (Fig. 1c). The combined data sets show significant correlation between crustal seismic velocity structure and changes in seafloor morphology, gravity and geochemistry. For a single spreading segment, abrupt transitions in crustal properties take place in a step-like fashion with only a few kilometres of incremental spreading. These abrupt transitions are unlike along-axis transitions that occur at mid-ocean ridges, which are thought to reflect dynamic threshold effects associated with changing spreading rate22,23 or mantle temperature24. We infer that the crustal transitions at the ELSC reflect abrupt changes in the composition of the mantle supplying melt to the back-arc ridge axes. Thus, the observations cast doubt on models of arc-proximal back-arc melting regions that are similar to the broad ,200-km-wide triangular regions imaged9 and modelled at mid-ocean ridges10. Such broad melting regimes would be expected gradually to entrain the various mantle wedge compositional domains as they migrate from arc-proximal to arc-distal settings and produce gradational changes in 1 3 J A N U A RY 2 0 1 1 | VO L 4 6 9 | N AT U R E | 1 9 9


RESEARCH LETTER b

a

30

-2

0

-40

Domain II

30

arc -50

30

0

-2 0

40

10

40

scar

scar

arc

-10

-30

Domain II

20

20° S

Dom ain III

Domain II

20

e ting ridg Propaga

Dom ain III

40

e ting ridg Propaga

Domain II

20

50

177° W –3,000

–2,000

-40

Tof u

a

176° W

30

30

21° S

Domain I

Tof u

Domain I

a

-20

vol

vol

can

can

0 -3

ic

40

ic

30

-20

40 -10

-4

175° W 177° W –1,000

Depth (m)

176° W –60

0 0

50

–40

100

–20

0

175° W 0

20

40

Gravity (mGal)

Distance (km)

Figure 2 | Seafloor bathymetry and Bouguer gravity anomalies in the study area. Red lines show the axis of the ELSC; the area of the seismic experiment is outlined by a box. a, The seafloor bathymetry is characterized by an older domain of pre-existing crust (Domain I) into which the ELSC propagated and two types of ELSC terrains labelled Domains II and III, separated by a variable transitional region demarcated by the dashed lines (see text). The two terrains closely follow the domains of low and high crustal seismic velocity shown in Fig. 1b. We note that near the active Tofua volcanic arc, volcaniclastic

sediments largely cover Domain II terrains. b, Bouguer gravity anomalies derived from ship and satellite data (contour interval is 2 mGal). The regional gradient associated with the slab was removed as described in ref. 4. Relative gravity lows and highs correspond to Domains II and III, respectively, separated by abrupt gradients. Local isostatic balance, assuming a crustal density of 2,700 kg m23 and a mantle density of 3,300 kg m23, indicates that Domain II crust is more than 1.9 km thicker than Domain III crust.

crustal properties. At the ELSC, the step-like changes in crustal properties suggest rapid transitions in the nature of the mantle entrained in the melting regime. We infer (Fig. 3) that these changes are enabled by a viscosity contrast of one to several orders of magnitude between a hydrous mantle near the arc and a relatively drier mantle further from the volcanic front25. We suggest that when near the volcanic arc, backarc spreading centres preferentially advect hydrous, low-viscosity mantle, possibly augmented by dynamic buoyant upwelling, as inferred for the arc itself26,27. This in turn produces the low-seismicvelocity, shallow, thick and volcanically complex crust observed. With increasing distance from the arc, however, a point is reached at which the connection to the low-viscosity hydrous mantle is broken and the spreading centre subsequently advects less-hydrous and higher-viscosity mantle. As the low-viscosity mantle is let go from the ridge upwelling zone, downwelling may occur on the arc side of the ridge, further suppressing melting of the higher-water-content mantle. At the ridge axis, mid-ocean-ridge-like crust is thereafter formed, although such crust still contains higher water content than at typical mid-ocean ridges and has trace-element enrichment and depletion characteristic of subduction, which can persist for hundreds of kilometres behind the arc volcanic front, giving back-arc basin basalts their distinctive character3,5,8. We hypothesize (Fig. 3) that geochemically and rheologically contrasting mantle domains with abrupt boundaries exist in arc-proximal regions of the mantle wedge, rather than gradational variations, such that individual spreading centres in this region can abruptly switch from advecting one to the other domain with only a few kilometres of incremental spreading. Even neighbouring ridge segments, separated by small non-transform discontinuities, can erupt lavas derived from contrasting mantle domains for long periods of time. For example, the juxtaposition of distinct domains across the northern overlapping

spreading centre is estimated to have lasted for roughly 0.31 million years. One can even see the duelling propagation ‘scars’ in the depth contrast (Fig. 2a) and velocity structure (Fig. 1b) as the W-shaped boundary that reflects the contrasting mantle domains and crust spread by the two neighbouring ridge segments. For an 8-km ridge offset to have acted as a persistent boundary between mantle domains implies either a sharp chemical distinction between the mantle regions that fed the two ridge axes or a rapidly changing mantle flow field, or both.

METHODS SUMMARY The L-SCAN experiment was carried out in January–March 2009, using the R/V Langseth and its 36-element, 6,600-cubic-inch airgun array towed at 9.0 m depth. The seismic data were recorded on 83 ocean-bottom seismometers (58 instruments, 25 of which were deployed twice) obtained via the National Ocean Bottom Seismograph Instrument Pool (https://obsip.ucsd.edu). Each unit contained a three-component seismometer and a hydrophone; the sample interval was 5 ms. Shots occurred every 450–500 m along 57 refraction lines. The ocean bottom seismometers were located using the travel times of the direct water wave for shots within 3 km of an instrument and a Bayesian grid search algorithm28; the average 1s location uncertainty was 6 m. At about 2 km to 12 km from an instrument, the first wave energy to arrive is a refraction from the upper crust (layer 2) with turning depths of 250 m to 2 km. Layer 3 refractions occur at greater distances (out to ,30 km). All layer 2 arrivals and layer 3 arrivals to ,20 km offset were used in this study. The arrivals are high quality and could be picked to within 1–3 samples; including all estimated experimental errors, the root-mean-square uncertainty of all travel times is 15 ms. The compressional (P)-wave velocity image was constructed using a nonlinear iterative tomographic technique29 that modelled 139,986 travel-time data points for threedimensional P-wave velocity structure and depth-varying anisotropy. Ray tracing was accomplished via a shortest path technique followed by a ray optimization procedure30. Although the number of data exceeds the number of unknowns by a



LETTER RESEARCH a

180°

175° W Samoa

b

e n t g

a

20° S 3

T o n

r i d g e

r

CLS C ELSC

To n g a

basin La u VFR

La u

r id g e

c h

FR

15° S

2 1

Pacific Plate

3

Figure 3 | Map of the Lau basin and cartoon interpretations of the formation of the crustal domains via changes in mantle wedge composition and flow patterns. a, Map of the Lau basin showing simplified interpretations of the velocity domains shown in Figs 1 and 2 extrapolated to the Valu Fa Ridge (VFR) (Domain II in orange, Domain III in blue, transition zone in green). The dotted box shows the area of Fig. 2a. The numbered lines show the locations of the vertical sections depicted in the panels of b. The dashed line is the boundary formed by the southward propagation of the ELSC. CLSC, Central Lau spreading centre. FR, Fonualei rift. b, (1) Formation of Domain II crust. Dynamic upwelling of the mantle occurs beneath the spreading axis when it is near the volcanic front and entrains low-viscosity mantle hydrated by water released from the subducting slab. Crust formed by the spreading centre in this position is arc-like, heterogeneous, thick, and has high porosity. (2) Formation of transitional crust. As the spreading centre separates from the volcanic front, the connection to the hydrous and low-viscosity mantle is abruptly cut off. Melt generation transitions to that of passive upwelling driven by plate separation at the ridge axis. Thinner and lower-porosity crust is formed at the spreading centre. (3) Formation of Domain III crust. At greater separation from the volcanic front, melt generation at the ridge results from passive upwelling and is separated from the hydrous and dynamic mantle upwelling beneath the arc. Mantle melted beneath the arc partly recirculates through the ridge system and the crust produced at the ridge has a lower porosity owing to lower slab water input, is thinner, and is depleted owing to prior melt extraction. The red dashes in panels 1–3 depict melt flow, which may be independent of solid mantle flow indicated by schematic streamlines. The blue area indicates hydrous mantle, the yellow area is nominally anhydrous mantle and green areas are lithospheric plates.

8. 9.

2

10. 11. 12.

13.

1

14.

15. 16.

factor of 2.5, the inversion was nevertheless regularized with weak smoothing and damping constraints. Seafloor bathymetry was included in the modelling procedure and its proper usage was extensively checked. (See also the Supplementary Information.)

17. 18. 19.

Received 16 August; accepted 18 November 2010. 1. 2. 3. 4. 5. 6. 7.

Karig, D. E. Ridges and basins of the Tonga-Kermadec island arc system. J. Geophys. Res. 75, 239–254 (1970). Hawkins, J. W. in Backarc Basins: Tectonics and Magmatism (ed. Taylor, B.) 63–138 (Plenum, 1995). Pearce, J. A. et al. in Volcanism Associated with Extension at Consuming Plate Margins (ed. Smillie, J. L.) Geol. Soc. Spec. Publ. 81, 53–75 (1995). Martinez, F. & Taylor, B. Mantle wedge control on back-arc crustal accretion. Nature 416, 417–420 (2002). Taylor, B. & Martinez, F. Back-arc basin basalt systematics. Earth Planet. Sci. Lett. 210, 481–497 (2003). Kincaid, C. & Hall, P. S. Role of back arc spreading in circulation and melting at subduction zones. J. Geophys. Res. 108, 2240, doi:10.1029/2001JB001174 (2003). Pearce, J. A., Stern, R. J., Bloomer, S. H. & Fryer, P. Geochemical mapping of the Mariana arc-basin system: implications for the nature and distribution of

20. 21. 22. 23. 24. 25. 26.

subduction components. Geochem. Geophys. Geosyst. 6, doi:10.1029/ 2004GC000895 (2005). Langmuir, C. H., Bezos, A., Escrig, S. & Parman, S. W. in Back-Arc Spreading Systems: Geological, Biological, Chemical, and Physical Interactions (eds Christie, D. M., Fisher, C. R., Lee, S.-M. & Givens, S.) Geophys. Monogr. AGU 166, 87–146 (2006). Forsyth, D. W. et al. Imaging the deep seismic structure beneath a mid-ocean ridge: the MELT experiment. Science 280, 1215–1218 (1998). Conder, J. A., Forsyth, D. W. & Parmentier, E. M. Asthenospheric flow and asymmetry of the East Pacific Rise, MELT area. J. Geophys. Res. 107, doi:10.1029/ 2001JB000807 (2002). Gill, J. B. Composition and age of Lau Basin and Ridge volcanic rocks: implications for evolution of an interarc basin and remnant arc. Geol. Soc. Am. Bull. 87, 1384–1395 (1976). Parson, L. M., Pearce, J. A., Murton, B. J., Hodkinson, R. A., &. the RRS Charles Darwin Scientific Party. Role of ridge jumps and ridge propagation in the tectonic evolution of the Lau back-arc basin, southwest Pacific. Geology 18, 470–473 (1990). Zellmer, K. E. & Taylor, B. A three-plate kinematic model for Lau Basin opening. Geochem. Geophys. Geosyst. 2, doi: 10.1029/2000GC000106 (2001). Martinez, F., Taylor, B., Baker, E. T., Resing, J. A. & Walker, S. L. Opposing trends in crustal thickness and spreading rate along the back-arc Eastern Lau Spreading Center: implications for controls on ridge morphology, faulting, and hydrothermal activity. Earth Planet. Sci. Lett. 245, 655–672 (2006). Jacobs, A. M., Harding, A. J. & Kent, G. M. Axial crustal structure of the Lau back-arc basin from velocity modeling of multichannel seismic data. Earth Planet. Sci. Lett. 259, 239–255 (2007). Escrig, S., Goldstein, S. L., Langmuir, C. H. & Michael, P. J. Mantle source variations beneath the Eastern Lau Spreading Center and the nature of subduction components in the Lau basin–Tonga arc system. Geochem. Geophys. Geosyst. 10, doi:10.1029/2008GC002281 (2009). Stolper, E. & Newman, S. The role of water in the petrogenesis of Mariana trough magmas. Earth Planet. Sci. Lett. 121, 293–325 (1994). Hawkins, J. W. & Melchior, J. T. Petrology of Mariana Trough and Lau Basin basalts. J. Geophys. Res. 90, 11431–11468 (1985). Carlson, R. L. & Herrick, C. N. Densities and porosities in the oceanic crust and their variations with depth and age. J. Geophys. Res. 95, 9153–9170 (1990). Turner, I. M., Pierce, C. & Sihna, M. C. Seismic imaging of the axial region of the Valu Fa Ridge, Lau Basin—the accretionary processes of an intermediate back-arc spreading ridge. Geophys. J. Int. 138, 495–519 (1999). Davies, J. H. & Bickle, M. J. A physical model for the volume and composition of melt produced by hydrous fluxing above subduction zones. Phil. Trans. R. Soc. Lond. A 335, 355–364 (1991). Small, C. A global analysis of mid-ocean ridge axial topography. Geophys. J. Int. 116, 64–84 (1994). Phipps Morgan, J. & Chen, Y. J. Dependence of ridge-axis morphology on magma supply and spreading rate. Nature 364, 706–708 (1993). Ma, Y. & Cochran, J. R. Transitions in axial morphology along the southeast Indian Ridge. J. Geophys. Res. 101, 15849–15866 (1996). Hirth, G. & Kohlstedt, D. L. in Inside the Subduction Factory (ed. Eiler, J.) 83–105 (American Geophysical Union, 2003). Marsh, B. D. Island-arc development: some observations, experiments, and speculations. J. Geol. 87, 687–713 (1979). 1 3 J A N U A RY 2 0 1 1 | VO L 4 6 9 | N AT U R E | 2 0 1


RESEARCH LETTER 27. Gerya, T. V. & Yuen, D. A. Rayleigh-Taylor instabilities from hydration and melting propel ‘cold plumes’ at subduction zones. Earth Planet. Sci. Lett. 212, 47–62 (2003). 28. Dunn, R. A. & Hernandez, O. Tracking blue whales in the eastern tropical Pacific with an ocean-bottom seismometer and hydrophone array. J. Acoust. Soc. Am. 126, 1084–1094 (2009). 29. Dunn, R. A., Lekic, V., Detrick, R. S. & Toomey, D. R. Three-dimensional seismic structure of the Mid-Atlantic Ridge at 35uN: focused melt supply and non-uniform plate spreading. J. Geophys. Res. 110, doi:10.1029/2004JB003473 (2005). 30. Nishi, K. A three-dimensional robust seismic ray tracer for volcanic regions. Earth Planets Space 53, 101–109 (2001). Supplementary Information is linked to the online version of the paper at www.nature.com/nature.

Acknowledgements We thank the captain, crew, and science parties of the R/V Langseth leg MGL0903 and R/V Kilo Moana leg KM0804. This work was funded as part of the NSF Ridge 2000 Program by grants OCE0426428 (R.A.D.) and OCE0732536 & OCE0727138 (F.M.). We thank J. Hammer, G. Ito, J. Sinton, B. Taylor and C. Wolfe for comments on an early version of the manuscript. Author Contributions R.A.D. carried out the seismic data collection, analysis and modelling; F.M. carried out bathymetry data collection and processing and gravity analysis. Both authors contributed to writing the paper. Author Information Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of this article at www.nature.com/nature. Correspondence and requests for materials should be addressed to R.A.D. ([email protected]).



RESEaRch NEWS & VIEWS while maintaining them as essential commercial pollinators. With due respect to Emily Dickinson1, ‘revery’ will not be enough if we want to see prairies, and other important terrestrial ecosystems, thriving in the future. ■ Mark J. F. Brown is in the School of Biological Sciences, Royal Holloway, University of London, Egham, Surrey TW20 0EX, UK. e-mail: [email protected]

1. Dickinson, E. The Complete Poems of Emily Dickinson (Little, brown, 1924). 2. Williams, P. H. & osborne, J. L. Apidologie 40, 367–387 (2009). 3. Thorp, r. W. & shepherd, M. D. in Red List of Pollinator Insects of North America CD-roM Version 1 (eds shepherd, M. D., Vaughan, D. M. & black, s. H.) (Xerces soc. Invertebrate Conserv., Portland, oregon, 2005); www.xerces.org/Pollinator_red_List/bees/ bombus_bombus.pdf 4. Cameron, s. A. et al. Proc. Natl Acad. Sci. USA

e ART h sCien Ce

A back-arc in time The Eastern Lau spreading centre in the Pacific Ocean is the subject of especial interest. The influence of the neighbouring subduction zone is considerable, but evidently has unexpected limits. See Letter p.198 PeTeR miChAel

I

n plate-tectonic theory, ocean crust and the associated lithosphere are recycled back into Earth’s mantle at the destructive plate boundaries called subduction zones. Several subduction zones also have submarine spreading centres that occur on the overriding plate lying behind the arc of surface volcanoes to be found above the site of subduction. These ‘back-arc’ spreading centres are the most rapidly changing plate-tectonic boundaries on the planet. New ocean crust is constructed by sea-floor spreading at back-arc spreading centres, just as occurs at mid-ocean ridges. But this spreading propels the back-arc spreading centre over the chemically diverse mantle of the subduction zone, and eventually away from the supply of subducted material that feeds the spreading. On page 198 of this issue 1, Dunn and Martinez describe a study of crustal thickness and structure at the Eastern Lau spreading centre (ELSC) in Tonga. Their work shows that back-arc spreading centres change even more rapidly than previously thought, suggesting that they are more active in capturing the subducted input from the mantle, and then rapidly releasing most of it when the spreading centre reaches a critical distance from the arc. Figure 3a of the paper (page 201) is a map of the region: the Tonga trench is the subduction zone’s intersection with the surface; triangles on the Tonga ridge show the associated volcanic arc; and the location of the ELSC is marked. The key ingredient in subduction zones is the mineralogically bound water that is carried into the mantle in the downgoing, subducted slab and then released into the overlying mantle wedge as the cold slab is heated. It promotes greater extents of mantle melting and the production of magmas that are progressively richer in silicon dioxide (SiO2) and water.

This results in crust that is thicker, seismically slower and more porous. It is these changes that are observed, in both crustal properties2 and rock composition3, southwards along the ‘zero-age’ axis of the ELSC, as the distance between the Tonga volcanic arc and the ELSC diminishes and the input of subducted materials to the back-arc increases. By examining the crustal structure across the axis as well as along it, Dunn and Martinez1 are peering back to a time when the back-arc basin was narrower. The volcanic morphology at the surface and the seismic velocities of the underlying few kilometres of crust show that, over a short period of time, as the back-arc spreading centre pushed itself away from the volcanic arc by sea-floor spreading, the volcanic crust abruptly became smoother, thinner, denser and probably less porous. In other words, it became less influenced by subducted water. These relatively shallow observations of the crust reflect what is happening in the deeper mantle wedge. The abruptness of the changes is the crucial factor here, as it suggests that the spreading centre is doing more than merely sampling whatever mantle it passes over. It remains to be seen how the concept of active capturing of subduction-influenced mantle and its rapid release at a critical distance will influence the increasingly sophisticated models that have been proposed for the formation of magmas behind the volcanic arc4,5. The conceptual cartoons that arise from these models of magma genesis are not yet sufficiently detailed. At the same time, geophysical imaging of the mantle wedge in other arcs6, and geodynamic models of the mantle wedge and slab that include dehydration and rheological changes7, are leading to more realistic models of subduction-zone processes and hint at a region in the mantle where conditions change rapidly over short distances. The time constraints from the

1 7 0 | N AT U R E | VO L 4 6 9 | 1 3 JA N UA Ry 2 0 1 1

© 2011 Macmillan Publishers Limited. All rights reserved

doi:10.1073/pnas.1014743108 (2011). 5. Keesing, F. et al. Nature 468, 647–654 (2010). 6. stout, J. C. & Morales, C. L. Apidologie 40, 388–409 (2009). 7. Vredenburg, V. T., Knapp, r. A., Tunstall, T. s. & briggs, C. J. Proc. Natl Acad. Sci. USA 107, 9689–9694 (2010). 8. otterstatter, M. C. & Whidden, T. L. Apidologie 35, 351–357 (2004). 9. Whitehorn, P. r., Tinsley, M. C., brown, M. J. F., Darvill, b. & Goulson, D. Proc. R. Soc. B doi:10.1098/rspb.2010.1550 (2011).

ELSC provided by Dunn and Martinez1 could improve the construction of these geodynamic models. The authors’ investigation1 is part of the Ridge 2000 programme sponsored by the US National Science Foundation8. This is an interdisciplinary initiative to study Earth’s oceanic spreading ridge system as an integrated whole, from its inception in the mantle to its manifestations in the biosphere and the water column. Intensive studies at three integrated study sites (including the ELSC) seek to establish links between different parts of these complex systems “from mantle to microbe”. The ELSC was chosen as a site because of the gradational nature of the effects of subduction (especially of water) along its axis. Hydrous magma degassing and crustal composition control the composition of hydrothermal fluids9, and therefore also strongly influence the microfauna and macrofauna at hydrothermal vents along the spreading centre. In their work, Dunn and Martinez exploited the expected link between crustal properties and mantle-source composition. Their research was made possible by the continually increasing investment in, and improvement of, ocean-bottom seismometers. Seismic-imaging studies use arrays of seismometers as receivers to provide a threedimensional view of travelling seismic waves, whether fast or slow. The deployment of seismometers on the sea floor is not new. But this study1 involved the largest, densest array of ocean-bottom seismometers deployed over an oceanic spreading centre anywhere on Earth, and permitted large-scale questions to be addressed at the ELSC. Dunn and Martinez1 used ship-borne airguns as the seismic-wave source to produce many relatively low-energy bursts that allowed the shallow crustal structure to be examined. Other seismic-imaging studies are under way with much longer deployments of the same seismometers, and using earthquake energy as a high-energy seismic-wave source to image the deeper mantle wedge and subducted slab. These investigations cover the same geographical area as the current shallow study, and may provide additional tests of the hypothesis of a critical distance in which volcanic-arc material is captured. Further tests of the Dunn and Martinez hypothesis will be forthcoming. If, as required by the hypothesis, there is an excellent

NEWS & VIEWS RESEaRch correlation between crustal properties and magma composition, there should be predictable, stepwise decreases in subduction-related elements such as barium across the axis as the rocks get younger. There will undoubtedly be further expeditions to sample rocks across the axis of the ELSC to test this hypothesis. Alongaxis changes in ‘zero-age’ rock chemistry are stepwise, but are not perfectly related to the distance between the ELSC and the volcanic arc3, suggesting a more complicated relationship. The rapid changes in the effects of subduction proposed by Dunn and Martinez should not be limited to the ELSC, but should be common in other subduction systems in which the back-arc spreading centre has migrated farther away from the volcanic arc. For example, detailed bathymetric and seismic studies of another back-arc spreading centre in the Pacific, the Mariana system, should show similar, sharply bounded domains. ■

Peter Michael is in the Department of Geosciences, The University of Tulsa, Tulsa, Oklahoma 74104, USA. e-mail: [email protected]

Diffusible molecules

a

1. Dunn, r. A. & Martinez, F. Nature 469, 198–202 (2011). 2. Jacobs, A. M., Harding, A. J. & Kent, G. M. Earth Planet. Sci. Lett. 259, 239–255 (2007). 3. Escrig, s., bézos, A., Goldstein, s. L., Langmuir, C. H. & Michael, P. J. Geochem. Geophys. Geosyst. 10, doi:10.1029/2008GC002281 (2009). 4. Langmuir, C. H., bézos, A., Escrig, s. & Parman, s. W. in Back-Arc Spreading Systems: Geological, Biological, Chemical, and Physical Interactions (eds Christie, D. M., Fisher, C. r., Lee, s.-M. & Givens, s.) Geophys. Monogr. AGU 166, 87–146 (2006). 5. Kelley, K. A. et al. J. Geophys. Res. 111, doi:10.1029/2005Jb003732 (2006). 6. rychert, C. A. et al. Geochem. Geophys. Geosyst. 9, doi:10.1029/2008GC002040 (2008). 7. van Keken, P. E. et al. Phys. Earth Planet. Inter. 171, 187–197 (2008). 8. www.ridge2000.org 9. Mottl, M. A. et al. Geochim. Cosmochim. Acta (in the press).

Gate1

Gate2

Gate3

Cell

b

Gate1

Gate2

Cell1

Cell2

synT heTiC biolo gy

Gate3

Division of logic labour

Cell3

Cellular compartmentalization is an effective way to build gene circuits capable of complex logic operations, in which binary inputs are converted into binary outputs according to user-defined rules. See Letters p.207 & p.212 boChong li & lingChong you

C

omputation forms the foundation of electronic devices that pervade our daily lives. Many of these devices run on digital circuits assembled from logic gates. All logic gates use an unambiguous rule to convert inputs of 0 or 1 into outputs — again, of 0 or 1. Depending on their basic operation, these gates are given names such as AND, OR and NOR. Computation is also fundamental to numerous biological functions, from information processing by neural networks to nutrient sensing by microbes. In this issue, Tamsir et al.1 and Regot et al.2 describe an innovative multicellular strategy for engineering complex logic circuits, which can potentially respond to combinations of biological signals by generating a useful read-out. In biological systems, cellular networks, which can often be thought of as assemblies of logic gates, underlie computation. To perform logic operations in such systems, researchers can engineer synthetic circuits in which biological substrates such as DNA, RNA and proteins are used as inputs, outputs and the information-processing hardware3 (Fig. 1a) in place of electronic components such as transistors and diodes. Compared with the assembly of their silicon-based counterparts, however, construction of complex biological logic circuits by the layering of elementary

gates is tremendously challenging4. This is due, in part, to crosstalk between cellular elements, the propagation of noise through networks5 and the metabolic burden that accompanies the expression of foreign genes in host cells6,7. The studies of Tamsir et al.1 and Regot et al.2, although different in their circuit implementation, share a common, crucial design concept: the compartmentalization of each elementary logic gate in a single cell (Fig. 1b). Here, each cell type is defined by the dedicated logic operation that it performs on inputs. And, for information flow, upstream gates produce signalling molecules that can diffuse across space into receiver cells, where these chemical ‘wires’ act on the downstream gate. Although a previous paper8 has used cellular compartmentalization in a synthetic predator–prey ecosystem, the current studies exploit it systematically. Specifically, Tamsir and colleagues’ work on the bacterium Escherichia coli (page 212) exploits quorum sensing — the process by which bacteria regulate their gene expression according to the local population density. The existence of multiple quorum-sensing systems with minimal crosstalk allows concurrent communication between more than one pair of sender–receiver E. coli populations. Regot et al. (page 207), meanwhile, use yeast pheromones (α-factors), taking advantage of the species-specific nature of their activity, which permits ‘secure’ communication in

Figure 1 | A multicellular approach to implementing complex logic circuits. a, Previously, systems of logic gates composed of DNA, RNA and proteins have been constructed within a single cell. b, New work1,2 compartmentalizes logic gates within different cells and uses diffusible output molecules to ‘wire’ such cell consortia together. (Adapted from a figure by Jeffrey Wong.)

cultures containing several yeast strains. Cellular compartmentalization is conceptually appealing. It conceals the implementation details of each encapsulated logic gate, which can be individually designed and optimized. As such, it can facilitate circuit implementation and reduce interference with the host cell’s physiology by minimizing the number of components introduced into each cell strain. Therefore, to assemble a complex multicellular circuit, the experimenter needs to be concerned with only two factors: the input–output function of each cellular gate and the output–input matching between layers. Encapsulation also physically insulates intracellular components of different gates, allowing them to be reused. To make an analogy to computer programming, each fully integrated circuit can be considered as a program. The cellular gates correspond to subroutines in the program, each dedicated to a specific function. Meanwhile, the individual components, such as promoters and transcription factors, are analogous to ‘local’ variables in the subroutines: the same variable can assume different roles in different subroutines without mutual dependency, thus lowering the demand for variable diversity. This design feature offers

1 3 JA N UA Ry 2 0 1 1 | VO L 4 6 9 | N AT U R E | 1 7 1

© 2011 Macmillan Publishers Limited. All rights reserved