VOL.84, NO. B3 JOURNALOF GEOPHYSICALRESEARCH MARCH10, 1979 Back-ArcOpening and the Mode of Subduction SEIYA UYEDA EarthquakeResearch Institute, University of Tokyo,Tokyo, Japan HIROO K ANAMORI SeismologicalLaboratory, California Institute of Technology,Pasadena, California 91125 Trench-arcsystems (subduction zones) can be classified into two types depending onwhether or not activelyopening back-arc basins are associated with them. This suggests that subduction of an oceanic plateis not a sufficientcondition for back-arc opening, though it may be necessary one. Mechanisms that causethe distinction between the two types have been investigated. Earthquake studies suggest that there isa significantdifference inthe mode of plate motion at interplate boundaries between the two types of trench-arcsystems. Extreme cases are Chile, where plate motion is seismic, and the M arianas arc, where it is aseismic.This difference seems to indicatethat the stress state in theback-arc area differs between the twotypes: compression in the Chilean type and tension in theMarianas type. This difference in thestress stateis alsomanifested in other tectonic features, such as topography, gravity, volcanic activity, and crustalmovement. Two possible mechanisms forthe difference between the two types are suggested: (1) Thenature of thecontact zone between upper and lower plates chahges from tight coupling (Chile) to decoupling(the Marianas) through the evolutionary process ofsubduction. The decoupling results inan oceanwardretreat of the trench and back-arc opening. (2) The downgoing slab is anchored tothe mantle, sothat the position of a trenchis also fixed with respect to themantle. Since the motion in themantle is slowcompared tothat of surface plates, it is the motion of the landward plate which controls the opening and nonopeningof back-arcs. INTRODUCTION closelycorrelated with eachother because, from geometrical In platetectonics, the fundamental process occurring under considerations,it is clear that a subductionboundary can thetrench-arc systems is the subduction of an oceanicplate. easilychange toga 'leaky' transform boundary upon a collision Smalloceanic basins, such as the Japan Sea, Okhotsk Sea, and of a spreadingcenter with the trench.In fact, the openingof Philippine Sea are often found landward of arcs.The intent of the Gulf of Californiamay be an exampleof the combined the presentpaper is to discussthe genesisof thesebasins, effectsof 3 and 4. Until 20 m.y.B.P. the East PacificRise was calledback-arc basins or marginalseas. situatedto the westof Baja California[Atwater, 1970] while Wefirst note the fact that not all thetrench-arc systems have now it is in the gulf in the form of a seriesof ridgesand back-arc basins.The Peruvian and Chilean arcs do not have transforms.The ridgepossibly died duringits collisionwith back-arc basins.The processof back-arcbasin formation the trenchthat existedto the westof Bajaand reappearedin appearsto be inactiveat presentin somearcs as in the caseof theformer back-arc region by jumping rather than migrating theJapan Sea, whereas others are believed to havean actively underthe Baja. The AndamanSea may be anotherexample openingback-arc basin, for example,the M arianaTrough [Eguchiet al., 1979].Possibility 5 seemsto haveno petrolog- [Karig, 1971a].Any theoryfor the genesisof back-arcbasins ical support. must accountfor this fact. Clearly, the subductionof an oce- Possibility2 is our mainconcern in this paper.As we men- anic plate is not a sufficientcondition for the formation of tioned already, not all the subductionboundaries have back- back-arcbasins, although it maybe a necessaryone. arc basins, and not all subduction boundaries that have back- Severalpossibilities have been suggested to explainthe ori- arc basinsare presentlyactively opening. We will firstclassify gin of back-arcbasins: (1) entrapmentof a marginalpart of the subductionboundaries by back-arctype and then demon- preexistingocean by the formation of an island arc: Aleutian stratethat the differencesin back-arctype correlate with other Basin[Cooper et al., 1976]and west Philippine Sea [Uyeda and phenomenawhich are presumedto be relatedto the tectonic Ben-Avraham,1972], (2) back-arcspreading caused by or re- stressstate. We will next dicussthe possibilitythat the differ- latedto subduction:Lau Basinand M arianaTrough [Karig, encein the tectonicstress state in back-arcregions may be 1970,197 la], ScotiaSea [Barker, 1970], and Japan Sea [Karig, causedby a differencein themode of subductionand attempt 1971b;Matsuda and Uyeda, 1971], (3) openingrelated to to find out why such different modes of subductionoccur. 'leaky'•transform fault: Gulf of California[Wilson, 1965] and CLASSIFICATIONOF SUBDUCTIONBOUNDARIES AndamanSea [Curray et al., 1979],(4) openingrelated to the subductionof a ridge:Japan Sea [Uyeda and Miyashiro, 1974], Karig[1971b] classified marginal basins into active marginal and (5) subsidencecaused by oceanizationof continental basins(type 1), inactivemarginal basins with highheat flow crust:Japan and Okhotsk seas [Beloussov and Ruditch, 1961 ]. (type 2), and inactivemarginal basinswith normal heat flow Among the abovepossibilities, 1 is probablythe likeliest (type 3). He postulatedthat types2 and 3 wereactive in the mechanismprovided that theage of thebasin is olderthan that past,type 2 in the morerecent past than type3. We useideas of the arc. The Mesozoicsequence of magneticlineations in more or less similar to Karig's to classifythe subduction the AleutianBasin is a notableexample provided that the boundariesin Table1. First,subduction boundaries (trench- identifications.are correct. Possibilities3 and 4 seem to be arc-back-arcsystems) are groupedinto the continentalarcs and islandarcs. Continental arcs, by definition,have no back- Copyright¸ 1979by theAmerican Geophysical Union. arc basins.The back-arc area of the Sumatra and Java arc is Paper number 8B1006. 1049 0148-0227/79/008 B- 1006501.00 1050 UYEDAAND KANAMORI: BACK-ARc OPENING AND $UBDUCTION TABLE 1. Classification of Trench-Arc-Back-Arc Systems Classification SubductionBoundary Location Continental ,4 rc Continents Peru-Chile Andes Middle America Middle America (semi-active?) Proto-North America Basin and Range Province(semi-active?) Alaska Alaska Range Sumatra-Java Sunda Shelf Island Arc Back-arc active Back-arcspreading Mariana Mariana Trough Ryukyu Ryukyu Trough Tonga Lau Basin Scotia Scotia Sea New Hebrides 9 Bonin 9 Philippine 9 Leaky transform Proto-BajaCalifornia Gulf of California Andaman-Nicobar Andaman Sea Melanesian Borderlands Bismarck Sea, New Guinea New Hebrides Fiji plateau? Back-arc inactive Inactivated Antilles Grenada Trough? West Aleutian Kamchatka Basin Kurile Kurile Basin Japan JapanBasin, Tsushima Basin Proto-Izu-Bonin Shikoku Basin Proto-Mariana Parece Vela Basin Trapped Aleutian Bering Sea Proto-Mariana West PhilippineSea? Proto-Tonga South Fiji Basin? Middle America Caribbean Plate water covered,but this arc is groupedwith the continentalarcs posed to have existedbefore the collisionof the East Pacific because most of the back-arc area has only shallow water Ri•e with the North American continent [McKenzie and Mor- depth (about 100 m) and presumablya continentalcrust. gan, 1969;Scholz et al., 1971]. Island arcsare classifiedinto Conversely,the Middle Americanarc, thoughincluded in the those with active back-arc basins and those with inactive back- continentalarcs, might be better groupedwith the back-arc arc basins.Active and inactiverefer to whetherthe openingof spreadingtype becauseof the extensionalvolcanic grabens the basin is presentlyin progressor not. A back-arc basin found in Middle America [Plafker, 1976].A similar argument which has rift topographyand/or fault structuresuggestive of may hold for the Proto-North American arc, which is sup- tensional tectonicsis consideredto be actively opening if it showsone or more of the following characteristics:(1) thin +90' +120' +150' -+180' -150' -120' -90* -60' -30* sedimentcover of very young age over rugged volcanic base- ment relief, (2) shallow water depth comparable to that of +60 +60 active midoceanicridges, (3) high heat flow or highly variable heat flow [Watanabe et al., 1977], or (4) magneticlineations with zero age. • k •o C•-o1(14) The actual classificationin Table 1 was made using the referencesindicated in the legendof Figure 1. Figure 1 summa- 0 rizesthis information regardingthe time and rate of opening. (19 ' '' Typical examplesof the actively spreadingback-arc basins are the M ariana Trough and the Lau Basin.Among thosethat are supposedto be actively spreading,magnetic lineations of very young age have beendocumented for the Scotia Sea. On the whole, the magneticlineations in the back-arc basinsare I I • • { I { I I I [ I I I I { [ I I I • ] I { I I much less well developedthan those in oceans [Watts and +90* +120' +150' -+180' -150* -120* -90* -60' -30' Weissel,1975; Weisseland Watts, 1975;Kobayashi and lsezaki, Fig. 1. Summary of back-arc openingactivities. Reference' num- 1977; Lawyer and Hawkins, 1978], suggestingthat either the betsare as follows: 1, Barkerand Griffiths [1972]; 2, Malfaitand modeof spreadingin the two cases is different or theback-arc Dinkelman [1972]; 3, Plafker [1976]; 4, Larson [1972]; 5, Scholzet al. [1971]; 6, Christiansenand Lipman [1972], Smith [1977]; 7, Cormier spreading correspondsto the very early stage of oceanic [1975]; 8, Beloussoo[1968]; 9, HYde and Wageman[19731;-10, Uyeda spreading. and Miyashiro [1974]; l 1, Tomodaet al. [ 1975]; 12,