Rapid conversion of an oceanic spreading centerto a subduction zone inferred fromhigh-precision geochronologyTimothy E. Keenana, John Encarnacióna,1, Robert Buchwaldtb, Dan Fernandezc, James Mattinsond,Christine Rasoazanamparanye, and P. Benjamin Luetkemeyera

aDepartment of Earth and Atmospheric Sciences, Saint Louis University, St. Louis, MO 63108; bDepartment of Earth and Environment, Boston University,Boston, MA 02215; cSchlumberger-WesternGeco, Geosolutions-Interpretation, Houston, TX 77042; dDepartment of Earth Science, University of California,Santa Barbara, CA 93106; and eDepartment of Geology and Environmental Earth Science, Miami University, Oxford, OH 45056

Edited by W. G. Ernst, Stanford University, Stanford, CA, and approved October 11, 2016 (received for review June 20, 2016)

Where and how subduction zones initiate is a fundamental tectonicproblem, yet there are few well-constrained geologic tests thataddress the tectonic settings and dynamics of the process. Numericalmodeling has shown that oceanic spreading centers are some of theweakest parts of the plate tectonic system [Gurnis M, Hall C, Lavier L(2004) Geochem Geophys Geosys 5:Q07001], but previous studieshave not favored them for subduction initiation because of the pos-itive buoyancy of young lithosphere. Instead, other weak zones,such as fracture zones, have been invoked. Because these modelsdiffer in terms of the ages of crust that are juxtaposed at the site ofsubduction initiation, they can be tested by dating the protoliths ofmetamorphosed oceanic crust that is formed by underthrusting atthe beginning of subduction and comparing that age with the ageof the overlying lithosphere and the timing of subduction initiationitself. In the western Philippines, we find that oceanic crust was lessthan∼1My old when it was underthrust andmetamorphosed at theonset of subduction in Palawan, Philippines, implying forced sub-duction initiation at a spreading center. This result shows that youngand positively buoyant, but weak, lithosphere was the preferred sitefor subduction nucleation despite the proximity of other potentialweak zones with older, denser lithosphere and that plate motionrapidly changed from divergence to convergence.

subduction initiation | tectonics | geochronology | Philippines | ophiolite

Subduction is the major driver of plate motion (1), and pro-cesses at subduction zones are largely responsible for the

growth and evolution of continents through accretion, collision,and magmatism. Despite its importance, the process of subductioninitiation is still debated partly because evidence from the earlystages of subduction is often obscured by later deformation andmagmatism. The lack of geologic constraints on how subductionzones initiate remains a significant void in our understanding ofEarth’s tectonics.Subduction initiation has been addressed primarily through nu-

merical modeling (2–9). These studies have demonstrated the needfor a weak zone in the lithosphere to facilitate subduction. Based onthis, subduction has been proposed to initiate in a variety of settingssuch as transform faults or fracture zones (10), passive continentalmargins (6, 11), oceanic detachment faults (12), and oceanicspreading centers (13). Spreading centers have been the least fa-vored, however, because the lithosphere there is positively buoyant.Two contrasting ideas regarding the dynamics (i.e., the forces) of

subduction initiation have been explored. In “spontaneous” sub-duction initiation, a plate’s increasing density with age may even-tually cause it to sink into the underlying asthenosphere (10, 14),whereas in forced subduction initiation, external plate forces arerequired to initiate subduction (3, 5, 15). Oceanic plates that are atleast ∼10 My old are negatively buoyant (16) and may undergoeither forced or spontaneous subduction initiation. Subductioninitiation within very young lithosphere near a spreading center,however, can only be forced because the plate is still positively

buoyant. Interestingly, some numerical models predict that despitethe buoyancy of the plate and the ridge push force, the forces re-quired to initiate underthrusting within the young, thin lithosphereof a spreading center are lower than within older, thicker litho-sphere, which requires increasingly larger forces to cause down-bending of the stronger plate (1). Self-sustained subduction, drivenby a plate’s negative buoyancy, might eventually be achieved afterinitiation at a spreading center, if forced convergence is sustaineduntil older, denser lithosphere finally enters the trench.Well-constrained geologic tests of the aforementioned models

are necessary to carry the debate forward. Because transform faults,fracture zones, and continental margins juxtapose lithosphere ofdifferent ages, whereas plates of equal and approximately zero ageare adjacent at spreading centers, determining where subductionhas initiated may be possible by comparing the ages of the un-derthrust and overriding lithosphere at the time of subductioninitiation, in relation to the time of subduction initiation itself.The timing of subduction initiation in some paleo-subduction

zones may be determined by constraining the timing of high tem-perature metamorphism—associated with the initiation of sub-duction—of the uppermost portions (i.e., the crust) of the initialsubducted plate. This metamorphic material may be transferred to(or “welded”) and preserved underneath the mantle peridotitehanging wall of the nascent subduction zone forearc of the upperplate as heat from the overlying mantle, and the resulting ductileshearing, progressively propagates down into the cold underthrust

Significance

Subduction, the process by which tectonic plates sink into themantle, is a fundamental tectonic process on Earth, yet thequestion of where and how new subduction zones form remainsa matter of debate. In this study, we find that a divergent plateboundary, where two plates move apart, was forcefully andrapidly turned into a convergent boundary where one plateeventually began subducting. This finding is surprising because,although the plate material at a divergent boundary is weak, it isalso buoyant and resists subduction. This study suggests thatbuoyant, but weak, plate material at a divergent boundary canbe forced to converge until eventually older and denser platematerial enters the nascent subduction zone, which thenbecomes self-sustaining.

Author contributions: T.E.K. and J.E. designed research; T.E.K., J.E., R.B., D.F., J.M., C.R.,and P.B.L. performed research; R.B. contributed new reagents/analytic tools; T.E.K., J.E.,R.B., D.F., J.M., and C.R. analyzed data; J.E., D.F., C.R., and P.B.L. performed fieldwork;and T.E.K., J.E., and R.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: encarnjp@slu.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1609999113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1609999113 PNAS | Published online November 7, 2016 | E7359–E7366

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 4

, 202

0

crust (17). These high-temperature “metamorphic soles” that formduring the initiation of subduction (18, 19) are found at the base ofophiolites—sections of oceanic lithosphere now on land—that areoften thought to represent the trapped forearc (i.e., the upper plate)of subduction zones (20, 21). The high-temperature (>700 °C)amphibolite-granulite facies rocks preserved in metamorphic soles(18, 22, 23) indicate that they formed during the very early stagesof subduction, before the development of depressed isothermsthat produce typical blueschist facies rocks during more maturesubduction (24). Also important, the high pressures (∼10 kbar)(18, 22, 23) associated with the formation of some metamorphicsoles indicate that they formed at oceanic mantle depths duringsubduction initiation rather than by some other process like oce-anic core complex formation or intraoceanic thrusting unrelatedto subduction initiation.Metamorphic soles commonly have metamorphic cooling ages

that are similar to the igneous crystallization ages of their overlyingophiolites (19, 25, 26), implying that the overlying ophiolite was stillvery young or formed during or shortly after subduction initiated.However, because these cooling ages reflect the time of meta-morphic cooling and not necessarily the time when the original ig-neous crust formed, the similarity in ages between the metamorphicsole and overlying ophiolite may be interpreted in several ways:(i) subduction initiation between older lithospheric plates followedby rapid slab rollback and seafloor spreading that generates theophiolite eventually preserved with the sole (10); (ii) subductioninitiation of variably older lithosphere beneath an already activespreading center (27, 28); (iii) subduction initiation along weakdetachment faults at some distance from a spreading center (12), or

(iv) underthrusting of young lithosphere at a spreading center (13)(Fig. 1). Without the age of the initially subducted plate, it may notbe possible to differentiate between these models.The geochemical affinities of the ophiolite-sole pairs could vary

depending on where, and how, the subducting and overridingplates were generated. Thus, the geochemistry of the sole andoverlying ophiolite could be similar or different and may providesome constraint on the different models outlined in Fig. 1. Al-though geochemical investigations of sole-ophiolite pairs mayprove useful, the age relationships discussed above can providefurther tests on the subduction initiation scenarios. Although allfour scenarios listed above predict similar ages for the initiation ofsubduction (i.e., the metamorphic age of the sole) and the over-lying ophiolite, each case predicts a different age relationship be-tween the initially subducted crust (i.e., the metamorphic sole’sprotolith age) and subduction initiation. A determination of theigneous crystallization age of the protolith of the high temperaturemetamorphic sole has been the missing piece of information in allprevious studies of subophiolitic metamorphic soles and is key totesting the various models of subduction initiation.

Cenozoic Subduction and Collision in PalawanWe applied the foregoing test to the metamorphic sole of anophiolite associated with a young, short-lived subduction zone inPalawan, western Philippines (Fig. 2). The central Palawanophiolite was trapped in the forearc of the subduction zone thatgenerated the Cagayan arc (18). Subduction began at ∼34 Ma andlasted for ∼20 My before it was terminated by microcontinent-arccollision (18, 29, 33). The subduction zone’s relatively young age

Subducting Plate

Crust

Mantle

Mantle upwelling

TF/FZ

or

A

D E

Metamorphic soleprotolith significantlyolder than both thesole and theophiolite

Metamorphic soleprotolith older thanthe sole and slightlyolder than theophiolite

Metamorphic sole protolith age similar to(or indistinguishable) from the sole and the ophiolite ages Ages are compatible

with Model (D)

CB

Metamorphic soleprotolith significantlyolder than both thesole and theophiolite

detachment fault

? ?

Pal

awan

op

hiol

iteS

ole

Igneous crystallization of ophiolite34.1 ± 0.1 MaCooling of themetamorphic solethrough 550-400°C34.2 ± 0.6 MaIgneous crystallization of sole protoliths35.25 ± 0.15 Ma35.242 ± 0.062 Ma35.862 ± 0.048 Ma

Fig. 1. Models of subduction initiation that explain similar ages between the formation of metamorphic soles and associated ophiolites (in cross-section andmap view). The high temperature metamorphic sole (shown as a thick, black line) is generated from the crust of the subducting plate during subductioninitiation. It may then be preserved at the base of the upper plate (future ophiolite, shown in cross-hatched pattern). Each model predicts a different agerelation between the initially subducted crust, the overlying ophiolite, and the time of subduction initiation. Plate ages are schematically shown with darkershades representing older lithosphere. White arrows on subducting plate indicate relative plate motion. (A) Sinking of the subducting plate along a transformfault (TF) or fracture zone (FZ) drives extension in the upper plate, generating the future ophiolite (10). (B) Subduction initiation of distinctly older lithospherenear an active spreading center (27, 28). (C) Subduction initiation along an oceanic detachment fault near a spreading center (12). (D) Subduction initiationvery close to or at a spreading center axis (13). Hacker et al. (25) proposed a variant of this in which subduction initiates across a transform or fracture zonewith underthrusting directed parallel to an active spreading center axis. Both options are shown in map view. (E) Schematic of the Palawan ophiolite, itsmetamorphic sole, and the dated lensoid pods preserved within the sole. U-Pb zircon ages of the pods and ophiolite obtained from this study are displayedalong with the metamorphic cooling age of the sole (18).

E7360 | www.pnas.org/cgi/doi/10.1073/pnas.1609999113 Keenan et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 4

, 202

0

and short duration make it an ideal site to study subduction initi-ation because it was terminated before a potentially complicatedhistory might have ensued.The island of Palawan flanks the southern margin of the South

China Sea (Fig. 2). Its geology consists of ophiolite-related rocks(31, 32) and continental crust [the North Palawan continentalterrane (NPCT)] that rifted from southeast China during theopening of the South China Sea Basin (SCSB) (29, 34). TheCagayan Ridge is located to the southeast of, and trends parallelto, Palawan. It is composed of calc-alkaline volcanic rocks (33) andis the volcanic arc associated with the subduction zone whose in-ception is preserved on Palawan (18, 29, 33, 34). To the northwestof central and southwest Palawan is the Palawan trough (Fig. 2), alinear depression that, based on seismic reflection observations, is

interpreted as a downwarped segment of the southern edge of theNPCT and contiguous proto-SCSB that is underthrust beneath thePalawan ophiolite (35, 36). The proto-SCSB is Cretaceous oceaniclithosphere that existed south of the NPCT–southeast China pas-sive margin before opening of the SCSB (29). Allochthonousremnants of this older Cretaceous ophiolite are found in tectonicwindows beneath the ∼34 My old (see below) Palawan ophiolite(18, 31, 36). Gravity data are consistent with the younger CentralPalawan ophiolite being rooted in the south (37) and thrustnorthward on to the continental crust of the NPCT (Fig. 2D) (32)in a manner similar to Tethyan ophiolites (20).In summary, all of the available evidence (onshore geologic and

offshore drill hole, seismic, and gravity data) from Palawan andthe surrounding areas is consistent with a south-southeast dipping

N

PuertoPrincesa

Aborlan

Quezon

Penacosa Point(plagiogranite)

South China Sea

SuluSea

Basalt / Diabase

Gabbro

Peridotite

Central PalawanOphiolite

Reverse Fault

Foliated flyschoid metasediments; somepillow basalts (L. Cretaceous to E. Eocene)

Clastics and Carbonates (Miocene)

Sediments(Upper Eocene? - Miocene)

Alluvium/ River Deposits (Pliocene - Recent)

15 km

Inferred Fault

China

ChinaSouth

Sea

Basin

PalawanNPCT

SuluSea

Basin

Palawan

Trou

ghCag

ayan

Ridg

e

Borneo CelebesSea Basin

Philippine SeaLuzon

PhilippineTrench

Man

ila

Trenc

h

LandPalawan Ophiolitesubmerged Eurasiancontinental crustoceanic crust ofmarginal basinsspreading axisNW limit of Palawanthrust wedge

margin ofNPCT Ulugan Bay

?

Red River-shear zone

Indo-china

Normal fault

TaiwanA B

D

Sagasa Point

metamorphic sole(panel C)

9°45’

10°15’118°00’ 118°45’110° 120°

20°

10°

N

Dalrymple Point1

43

2

55

4

100 m

?

B

C

DE,F

?N

ULUGANBAY

Epidote amphibolite1 -2 -

3 -4 -

5 -Inferred contactInferred fault

Amphibolite gneiss with hornblendite lensesGarnet amphiboliteHornblendite, quartzite, kyanite schistMantle peridotite

?

C

Location of photos in Figure 3

NPCT PALAWANTROUGH PALAWAN

OPHIOLITE(eroded)

ophioliteperidotite

klippe

clastic rocks

PlatformLimestone

NPCT BASEMENT

10 km

metamorphicsole

Schematic NW-SE section across Ulugan Bay area (see panel “b”)

DEFORMED METASEDIMENTSDERIVED FROM NPCT

NPCT

Fig. 2. (A) Present tectonic setting of Palawan island, Philippines (18, 29, 30). Rectangle outlines the area shown in B. (B) General geology of central Palawanshowing locations of sample sites. The general structure consists of an ∼34-Ma ophiolite (the Central Palawan ophiolite) thrust over deformed Cretaceous-Eocene turbiditic sedimentary rocks of the NPCT. Remnants of the older Early Cretaceous proto-SCSB ophiolite are found as occasional pillow lavas in tectonicwindows in the younger Palawan ophiolite (geology from our field observations and refs. 18 and 31). (C) Geologic map of the metamorphic sole at DalrymplePoint. See B for location. Background image from Google Earth (Digital Globe, CNES/Astrium). Apparent metamorphic grade decreases away from the mantleperidotite. (D) Schematic NW-SE cross section of Palawan in the Ulugan Bay area after ref. 32.

Keenan et al. PNAS | Published online November 7, 2016 | E7361

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 4

, 202

0

subduction zone that formed the Cagayan arc, which places thePalawan ophiolite in the forearc. This subduction zone was ter-minated when the southern edge of the rifted margin of China(the NPCT) jammed the trench in the Middle Miocene (33),causing obduction of the Palawan ophiolite. The question of howsubduction initiated in Palawan is unresolved. Mitchell et al. (32)speculated that subduction may have initiated at a spreadingcenter, whereas the model of Encarnación et al. (18) shows initi-ation at a transform fault or fracture zone. Both of these proposalslack the necessary age data to properly test these models.

Age and Geochemistry of the Palawan Ophiolite and Its SoleThe central Palawan ophiolite is a relatively coherent section ofoceanic lithosphere consisting of pillowed and massive lavas, diabasedikes, plagiogranite, gabbro, troctolite, and mantle harzburgite (31,32) (Fig. 2). Samples of pillow lavas, dikes, and gabbroic intrusions,as well as felsic intrusions (plagiogranite) (Fig. 3) collected fromthe ophiolite, show slight light rare earth element (LREE) de-pletions and relatively flat middle REE and heavy REE patterns,about 10 times chondritic values, consistent with midocean ridge(MORB)-like magma (Fig. 4A). A similar MORB to transitionalMORB-island arc basalt (IAB) or suprasubduction zone signatureis evident in other multielement plots. Tectonic discriminationdiagrams that use robust statistical tests (38) assign the data withinthe MORB field (Fig. 4B) or to a transitional MORB-IAB field(Fig. 4C), consistent with a backarc basin basalt (BABB)-typegeochemistry (39) and consistent with many other ophiolites (40)formed at oceanic divergent-type boundaries.Based on zircon U-Pb geochronology from a plagiogranite

sampled near Penacosa Point (Fig. 2), the best estimate for theage of the ophiolite is 34.1 ± 0.1 Ma (Fig. 1 and Table S1). Theregional distribution of rock types (Fig. 2) shows that the outcropsof gabbroic and tonalitic intrusives in the Penacosa Point area arelocated in the upper levels of the main gabbroic crustal sectiontransitional to the extrusive section of the ophiolite. The PenacosaPoint plagiogranite exhibits field relations that are consistent withthe felsic magma being comagmatic with the dominant maficmagmas comprising the bulk of the oceanic crustal section here(Fig. 3). In addition, the geochemistry of the plagiogranite isconsistent with simple fractional crystallization from the dominantbasaltic magma (Fig. 4A).

The metamorphic sole of the ophiolite is exposed in the Dal-rymple Point area and is composed of ductiley strained garnetamphibolites, hornblendites, amphibolites, epidote amphibolites,quartzites, and kyanite schists with isoclinal folds and a distinct,penetrative mineral lineation in most rocks (Fig. 2). These rocksrepresent metamorphosed igneous, basaltic oceanic crust and as-sociated sediments (cherts and mudstone). As in other meta-morphic soles, the higher temperature garnet amphibolites andhornblendites are located closer to the basal mantle harzburgites,whereas lower temperature epidote amphibolites tend to bestructurally lower. At several locations, more highly strained solerocks enclose less deformed (or isotropic) and more competent,irregularly shaped lensoid pods of epidote amphibolite (decimeter-meter scale) and smaller lensoid pods of amphibolite (a few mil-limeters to centimeters in thickness and several centimeters todecimeters in diameter) (Fig. 3). Thermobarometric determina-tions show that the garnet amphibolites of the sole reached peakmetamorphic temperatures of 700–760 °C and pressures exceeding9 kbar (∼27 km depth in the mantle) (18), conditions consistentwith those that the crust of a subducting plate would be subject toduring the earliest stages of subduction (23). Critical to this study,previous work (18) has constrained the minimum age for sub-duction initiation by determining 40Ar-39Ar cooling ages on twohornblende samples and one white mica sample (from garnetamphibolite, amphibolite, and kyanite schist, respectively) in thesole. These ages are indistinguishable at 34.2 ± 0.5, 34.2 ± 0.6, and34.25 ± 0.3 Ma, respectively (Fig. 1) (corrected for new revisedages of the neutron flux monitors) (41), and indicate rapid coolingof the sole to 550–400 °C after reaching peak metamorphic tem-peratures (18). Lower grade, altered greenstones, and pillow lavasstructurally beneath the metamorphic sole are exposed furthersouth in the Sagasa Point area (Figs. 2 and 3). These rocks are lessmetamorphosed oceanic components underthrust beneath thePalawan ophiolite and its sole sometime after the initial un-derthrusting associated with subduction initiation.Overall, metabasite samples from the metamorphic sole are

geochemically similar to the ophiolite in that they plot in theMORB-like to transitional MORB-IAB fields (Fig. 4). Severalsamples of the epidote amphibolite pods are depleted (∼2–3 timeschondritic values) relative to the MORB-like samples and havepositive Eu anomalies, indicating they were probably cumulate

50 um 100 um 100 um 100 um 100 um 100 um

A B C D E F

G H I J K L

Fig. 3. Photographs of outcrops from the central Palawan ophiolite (A) and its metamorphic sole (B–F) and cathodoluminescence images of extracted zirconsfrom selected samples (G–L). (A) Magma-mingling structures exhibited by light-colored tonalite (plagiogranite) and diorite-gabbro (darker) at Penacosa Point.The tonalite yielded zircons with a crystallization age of 34 Ma. Pencil for scale. (B) Layered chert/quartzite and amphibolite showing sheath folds.(C) Amphibolite gneiss with hornblendite domains exhibiting isoclinal folding. (D) Foliated and lineated epidote amphibolite, looking ∼west; mountainsacross Ulugan Bay are mantle harzburgite of the Palawan ophiolite structurally overlying metamorphic sole rocks; strike and dip symbol indicates foliation.(E) Smaller, foliation-parallel, light-colored lensoid pods of amphibolite (sample PL-14-07). (F) Competent, light-colored pod of epidote amphibolite (withcumulate gabbro-like REE signatures; sample PL-14-05) enclosed within the strongly foliated amphibolite. These pods yielded zircons with crystallization agesof 35.242 Ma. Hammer for scale. (G and H, I and J, and K and L) Cathodoluminescence images of zircons, showing magmatic oscillatory zoning (samples PL-14-05, PL-14-06, and PL-14-07, respectively).

E7362 | www.pnas.org/cgi/doi/10.1073/pnas.1609999113 Keenan et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 4

, 202

0

rocks (gabbros, based on major element chemistry) of the mid-lower crust. One of the smaller pods of amphibolite displays anegative Eu anomaly and is slightly more enriched than the largerepidote amphibolite pods. This sample may have crystallized froma magma following the extraction of plagioclase. The rocks of thesole are therefore not geochemically unlike the overlying ophioliteand probably formed in a similar petro-tectonic setting.Based on their geochemistry and geologic context within the

high temperature sole, we interpret the more competent pods tobe middle- to lower-level crustal rocks of the leading edge of thesubducted plate. During the period of underthrusting to sub-duction, the underthrust crust was sheared, thinned, and ductileyfolded, resulting in the transposition of middle to lower crustalgabbroic rocks with upper crustal basaltic rocks, both of whichwere subject to the high temperature and high pressure meta-morphism that formed the metamorphic sole (Fig. 1).Zircons from two of the larger pods (PL-14-05 and PL-14-06)

and one smaller pod (PL-14-07) from the high temperaturemetamorphic sole were extracted for U-Pb CA-ID-TIMS geo-chronology. The four analyzed zircons from sample PL-14-05yielded internally and externally concordant ages with an error-weighted mean age of 35.242 ± 0.062. Three of five analyzed

zircons from sample PL-14-06 yielded internally and externallyconcordant ages with a mean of 35.862 ± 0.048 Ma (Fig. 1, TableS2, and Fig. S1). Three externally discordant but internally con-cordant ages of 37.00 ± 0.16, 35.97 ± 0.11, and 35.25 ± 0.15 Ma(Fig. 1, Table S2, and Fig. S1) were obtained from the smaller pod.The age of the youngest zircon, 35.25 ± 0.15 Ma, is taken as thebest estimate of the final crystallization age of the protolith of thissample. The small spread in ages seen in this sample is similar tothose revealed by high precision U-Pb geochronology in theSamail ophiolite (44) and may be due to prolonged zircon crys-tallization in a replenished magma chamber or assimilation ofslightly older wall rock.Establishing that these zircons are igneous, and not meta-

morphic, is critical because the age of metamorphic zircons wouldmerely represent the age of metamorphic sole formation (sub-duction initiation) instead of the crystallization age of the meta-morphic sole protoliths. Although metamorphic zircon growth hasbeen shown to occur under amphibolite facies conditions (45),cathodoluminescence imaging shows no evidence of metamorphicovergrowths in these zircons. Instead, they are euhedral, prismatic,and have distinct, fine, oscillatory zoning, a feature that is char-acteristic and unique to magmatic zircons (46) (Fig. 3). Eventhough Th/U ratios are not completely reliable indicators ofmagmatic vs. metamorphic zircon, we note that the Th/U ratios inthese zircons (>0.1) are consistent with many magmatic zircons(47). We are therefore confident that these zircons are igneous andthat their ages reflect the original igneous crystallization age of theoceanic crust that was underthrust, and then metamorphosed, atthe onset of subduction.

Forced Subduction of Young, Buoyant LithosphereAs discussed earlier, a critical test to constrain the tectonic settingof subduction initiation is a comparison of the igneous ages of theunderthrust and overriding lithosphere in relation to the time ofsubduction initiation. A positive test for subduction initiation at anactive spreading center is to find all three events very close in age.We find that the age differences between the upper plate (thePalawan ophiolite), the subducting plate (protoliths of the sole),and metamorphism of the sole are less than ∼1 My (Fig. 1). Thevery small age difference between formation of the sole protolithand its metamorphism during subduction initiation leads us to re-ject outright the models shown in Fig. 1 A and B. Furthermore, themodel shown in Fig. 1C is rejected for subduction initiation atdetachment faults that are far from the spreading center. Wetherefore conclude that subduction must have initiated in veryclose proximity to, or at, a spreading center (Fig. 1D). Our age datado not differentiate between the ridge parallel and ridge normalsubduction initiation scenarios in Fig. 1D in cases where the ridge-transform fault segments in the right-hand scenario are very short.The similarity in the geochemistry of the sole and overlyingophiolite also supports subduction initiation close to a spreadingcenter, where the eventual upper plate and lower plate (meta-morphic sole) are not expected to be geochemically different.Our data show that oceanic crust was formed at 35.24 Ma and

was then underthrust/subducted almost immediately, reaching∼27 km depth, metamorphosed to amphibolite, and subsequentlycooled to ∼400 °C by 34.25 Ma, a remarkably short interval be-tween crust formation and subduction. Assuming a slab dip con-trolled by an ∼30° dipping lithosphere–asthenosphere boundary,this underthrusting would require a convergence rate on the orderof 5 cm/y.In cases where the detachment fault (Fig. 1C) is located very

close to, or at the spreading center, the models shown in Fig. 1 Cand D become indistinguishable using age data alone. The weak-ness at which subduction initiated could have been a detachmentfault very near the spreading center axis or at the spreading centeraxis itself. Although the exact nature of the weak zone could bedebated, the high-precision age data from our sample site tightly

MORB

OIB

IAB

- Metamorphic Sole- Palawan ophiolite

MORB

OIB

IAB

Ti / 50

V

Ti / 50

V50 * Sm 5 * Sc

1

10

100

La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu

Sam

ple

/ Cho

ndrit

e- Metamorphic Sole (dated samples)- Metamorphic Sole- Palawan ophiolite (dated sample)- Palawan ophiolite PW-00-18

(34.1 Ma)

PL-14-07(35.25 Ma)

PL-14-06(35.862 Ma)

PL-14-05(35.242 Ma)

A

CB

Fig. 4. Geochemical data on samples from the central Palawan ophioliteand its metamorphic sole. Palawan ophiolite samples plotted in A–C includepillow lavas, mafic dikes, gabbroic intrusions, and (A) felsic intrusions (pla-giogranite). Metamorphic sole samples plotted in A–C include amphibolites,epidote amphibolites, and garnet amphibolites. (A) Chondrite normalizedREE concentrations in the samples. Samples that were selected for U-Pbzircon geochronology are symbolized by diamonds and their ages are in-dicated next to the data. The majority of samples have REE patterns re-sembling MORB and possible differentiates of MORB. Two samples (PL-14-06and PL-14-05) display positive Eu anomalies indicating cumulate plagioclasein the samples. The geochemistry of the plagiogranite (PW-00-18) is consis-tent with simple fractional crystallization from the MORB-like basalticmagmas. (B) Ti-V-Sm and (C) Ti-V-Sc tectonic discrimination diagrams (38).Basaltic samples from the ophiolite and sole are similar and plot as MORB ortransitional MORB-IAB.

Keenan et al. PNAS | Published online November 7, 2016 | E7363

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 4

, 202

0

constrains subduction initiation at, or very close to the spreadingaxis. If a detachment fault was the weakness, it would have to havebeen located very close to the ridge where nascent oceanic crustwas generated and then underthrust and metamorphosed at adepth of ∼27 km less than 1 My later.The much younger age of the Palawan ophiolite compared with

the Cretaceous proto-SCSB suggests that subduction was initiatedwithin a young marginal oceanic basin hosted in older Cretaceousproto-SCSB (Fig. 5). Because of the positive buoyancy of youngoceanic lithosphere, subduction initiation must have been forcedin this case. Following forced underthrusting of the young, buoy-ant lithosphere, the old, cold, and dense Cretaceous lithosphere ofthe proto-SCSB would have eventually entered the trench andenabled the transition to self-sustained subduction until the NPCTcollided with the trench, caused obduction of the Palawanophiolite and its sole, and terminated subduction.Regionally, the origin of the force that converted a divergent

boundary to a convergent boundary was probably the collision ofIndia with Asia. Tapponnier et al. (49) proposed that this collisionresulted in the extrusion of fairly rigid continental lithosphericblocks from the southeast margin of Asia along large strike-slipfaults, such as the Red River shear zone (Fig. 2). The timing of theonset of strike-slip movement along the Red River shear zonehas been estimated by U-Pb ages on monazite included in shear-related rotated garnets at ∼34 Ma (48). This age coincides wellwith the timing of the initiation of subduction in Palawan. Seafloorspreading in the South China Sea, which accompanied the south-ward motion of the NPCT, began around 32–30 Ma (30), a time

also compatible with initiation of subduction in Palawan south ofthe NPCT (18) and southward convergence of the NPCT with thePalawan subduction zone–Cagayan arc.If the reconstruction of Fig. 5 is correct, the initiation of sub-

duction at the Palawan spreading center, forced by the extrusion ofIndochina, supports the results of numerical modeling (1), whichpredict that the thin lithosphere at oceanic spreading centers re-quires less force to converge compared with areas of thicker lith-osphere. Several potential weak zones existed in the area: (i) thesoutheast China passive margin, (ii) the contact between the olderCretaceous proto-SCSB lithosphere and the much younger Pala-wan marginal basin, and (iii) the spreading center of the Palawanophiolite. Despite the presence of older and denser oceanic litho-sphere, our results indicate that subduction nucleated within theyoung, buoyant, but weaker Palawan marginal basin. Presumably,after the eventual underthrusting of the older Cretaceous litho-sphere and an additional critical convergence of 100–130 km,subduction became self-sustaining (1).There is currently no known modern analog for subduction

initiation exploiting an active oceanic spreading center. Intra-oceanic subduction appears to be initiating along several sectionsof the Australian–Pacific plate boundary south of New Zealand,exploiting weaknesses associated with extinct spreading centersand/or fracture zones undergoing transpressional deformation (50–52). However, unlike the case in Palawan where deep un-derthrusting occurred within 1 My of oceanic crust formation,spreading at the boundary south of New Zealand had ceased andwas followed by strike-slip deformation several million years before

Rifting

NPCTCretaceous oceanic lithosphere

(Proto-SCSB)SE Chinalithospehre

PalawanSpreading Center

Detachment of Palawan ophiolite atridge and initiation of subduction

Indochina

Onset of Red Rivershear zone (~34 Ma)

Continent-ocean

boundary

Cretace

ous

ocean

ic litho

sphere

(proto

-SCSB)

Eurasiancontinentallithosphere

Break-up rifting.Seafloor spreading starts in

South China Sea at 32-30 Ma

Palawan

spreading center

Detachment of Palawan ophiolite at the spreading center and forced initiation of subduction (~35-34 Ma)

due to extrusion of Indochina

A

A’

A A’

A

BPassive margin

Extinct fault or ‘intrusive’ boundary

Palawan ophiolite

NPCT

Fig. 5. Subduction initiation (∼34–35 Ma) at the spreading center that generated the Palawan ophiolite. (A) Schematic map view of the area showing thetiming of initial strike-slip movement on the Red River shear zone (48), seafloor spreading in the South China Sea (30), and initiation of subduction at thePalawan ophiolite spreading center. Palawan ophiolite (yet to be obducted onto the rifted continental crust) shown in cross-hatched pattern. (B) Cross-sectional schematic view of the area along A-A′ during the onset of subduction. Weak zones where subduction had the potential to initiate, but did not, arealso shown.

E7364 | www.pnas.org/cgi/doi/10.1073/pnas.1609999113 Keenan et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 4

, 202

0

incipient subduction. It should be noted that there are only a fewlocations where subduction initiation might be happening (ref. 1and references therein), and therefore it is unlikely that theserepresent the full range of potential mechanisms for intraoceanicsubduction initiation. The process of subduction initiation may laston the order of only 2–3 My (assuming a convergence rate of∼5 cm/y) (1), a small fraction of the lifetime of a subduction zone,and is therefore unlikely to be captured by present day observa-tions. Ophiolites that have similar crystallization ages to themetamorphic age of their sole, like the Palawan ophiolite, are notuncommon (19, 25, 26). These other examples, however, lack datesfor the protolith of the sole. Dating these protoliths may show thatsubduction initiation at spreading centers has been more pervasivethan currently recognized.The subduction zone that existed in Palawan and the nascent

Puysegur trench (50) are possibly the only well-constrained geo-logic examples that address the dynamics of subduction initiation.In both of these examples, a forced initiation has been inferred.Although spontaneous subduction has been simulated by numer-ical models (9), corresponding geologic tests have been questioned(53). Our study suggests that subduction initiation in the modernday plate tectonic regime requires extant subduction, the maindriver of plate motion, to force new subduction. Furthermore, itappears that forces generated in the interior of major continentalcollisions zones can be transmitted out to the oceanic realm,rapidly causing diverging plates at a spreading center to convergeand lead to incipient subduction in less than ∼1 My.

MethodsGeochemical Analysis.Whole rock powders were analyzed by a combination ofa lithium metaborate/tetraborate fusion followed by inductively coupled

plasma (ICP) methods for major elements abundances and inductively coupledplasmaMS (ICP-MS)methods for trace elements abundances. The analysesweredone at Activation Laboratories (Ontario, Canada) (ACTLAB). A detailed de-scription of sample preparation methods is given on the ACTLAB Website(www.actlabs.com).

U-Pb Zircon Geochronology. All reported ages are 206Pb/238U error-weightedmean ages, because the uncertainty of 235U/207Pb ages are pronouncedly largedue to the very young and low 235U content of these zircons. They are cor-rected for initial Th/U disequilibrium, and errors are reported as 2σ.

Twelve zircon grains from the epidote amphibolite pods (PL-14-05, PL-14-06,and PL-14-07) were analyzed by chemical abrasion thermal ionization MS(CA-TIMS) at the radiogenic isotope laboratory atMassachusetts Institute ofTechnology. Samples PL-14-05 and PL-14-06 yielded weighted mean ages of35.242 ± 0.062 and 35.862 ± 0.048 Ma, respectively, whereas sample PL-14-07displays three distinct ages of 37.00 ± 0.16, 35.25 ± 0.15, and 35.97 ± 0.11 Ma.For sample PL-14-06, the youngest cluster of zircon ages was used for theweighted mean age because this best approximates the timing of magmacrystallization (54).

Three zircon fractions from the plagiogranite (PW-00-18) were analyzed byconventional TIMS at the University of California, Santa Barbara. One fractionwas analyzed as is, whereas the other two were air abraded to remove anyexterior zones with possible Pb-loss. The unabraded fraction is only slightlyyounger than the two abraded fractions that give amean age of 34.1± 0.1Ma.Additional zircon fractions were analyzed using the stepwise chemical abra-sion technique (55), and all analyzed fractions after the first few steps yieldedidentical 34.1 ± 0.1 Ma ages.

ACKNOWLEDGMENTS. We thank John Spray and two anonymous reviewersfor helpful and constructive comments. Freddie Dela Cruz provided excellentboatmanship in Ulugan Bay. This work was supported, in part, by Saint LouisUniversity and the Geological Society of America. The geochronologylaboratories at University of California Santa Barbara and MassachusettsInstitute of Technology are supported by the National Science Foundation.

1. Forsyth D, Uyeda S (1975) On the relative importance of the driving forces of plate

motion. Geophys J R Astron Soc 43(1):163–200.2. Gurnis M, Hall C, Lavier L (2004) Evolving force balance during incipient subduction.

Geochem Geophys Geosyst 5(7):Q07001.3. McKenzie DP (1977) The initiation of trenches: A finite amplitude instability. Island

Arcs, Deep Sea Trenches and Back-Arc Basins, Maurice Ewing Series, eds Talwani M,

Pitman WC IIII (American Geophysical Union, Washington, DC), Vol 1, pp 57–61.4. Mueller S, Phillips RJ (1991) On the initiation of subduction. J Geophys Res 96(B1):

651–665.5. Toth J, Gurnis M (1998) Dynamics of subduction initiation at preexisting fault zones.

J Geophys Res 103(B8):18053–18067.6. Regenauer-Lieb K, Yuen DA, Branlund J (2001) The initiation of subduction: Criticality

by addition of water? Science 294(5542):578–580.7. Ueda K, Gerya T, Sobolev SV (2008) Subduction initiation by thermal-chemical plumes:

Numerical studies. Phys Earth Planet Inter 171(1–4):296–312.8. Leng W, Gurnis M (2011) Dynamics of subduction initiation with different evolu-

tionary pathways. Geochem Geophys Geosys 12(12):Q12018.9. Leng W, Gurnis M (2015) Subduction initiation at relic arcs. Geophys Res Lett 42(17):

7014–7021.10. Stern RJ, Bloomer SH (1992) Subduction zone infancy: Examples from the Eocene Izu-

Bonin-Mariana and Jurassic California arcs. Geol Soc Am Bull 104(12):1621–1636.11. Erickson SG (1993) Sedimentary loading, lithospheric flexure, and subduction initia-

tion at passive margins. Geology 21(2):125–128.12. Maffione M, et al. (2015) Dynamics of intraoceanic subduction initiation: 1. Oceanic

detachment fault inversion and the formation of supra-subduction zone ophiolites.

Geochem Geophys Geosyst 16(6):1753–1770.13. Spray JG (1983) Lithosphere-asthenosphere decoupling at spreading centres and

initiation of obduction. Nature 304(5923):253–255.14. Turcotte DL, Haxby WF, Ockendon JR (1977) Lithospheric instabilities. Island Arcs,

Deep Sea Trenches and Back-Arc Basins, Maurice Ewing Series, eds Talwani M, PitmanWC

IIII (American Geophysical Union, Washington, DC), Vol 1, pp 63–69.15. Hall CE, Gurnis M, Sdrolias M, Lavier LL, Dietmar Müller R (2003) Catastrophic initi-

ation of subduction following forced convergence across fracture zones. Earth Planet

Sci Lett 212:15–30.16. Cloos M (1993) Lithospheric buoyancy and collisional orogenesis: Subduction of

oceanic plateaus, continental margins, island arcs, spreading ridges, and seamounts.

Geol Soc Am Bull 105(6):715–737.17. Hacker BR (1991) The role of deformation in the formation of metamorphic gradi-

ents: Ridge subduction beneath the Oman ophiolite. Tectonics 10(2):455–473.18. Encarnación JP, Essene EJ, Mukasa SB, Hall CH (1995) High-Pressure and –Temperature

Subophiolitic Kyanite-Garnet Amphibolites Generated during Initiation of Mid-Ter-

tiary Subduction, Palawan, Philippines. J Petrol 36(6):1481–1503.

19. Wakabayashi J, Dilek Y (2003) What constitutes ‘emplacement’ of an ophiolite?:Mechanisms and relationship to subduction initiation and formation of metamorphicsoles. Geol Soc Lond Spec Publ 218:427–447.

20. Dilek Y, Furnes H (2014) Ophiolites and their origins. Elements 10(2):93–100.21. Stern RJ (2004) Subduction initiation: Spontaneous and induced. Earth Planet Sci Lett

226(3–4):275–292.22. Jamieson RA (1986) P-T paths from high temperature shear zones beneath ophiolites.

J Metamorph Geol 4(1):3–22.23. Wakabayashi J, Dilek Y (2000) Spatial and temporal relations between ophiolites and

their metamorphic soles: A test of models of forearc ophiolite genesis. Ophiolites andOceanic Crust: New Insights From Field Studies and the Ocean Drilling Program, edsDilek Y, Moores EM, Elthon D, Nicolas A (Geological Society of America, Boulder, CO),GSA Special Papers 349, pp 33–64.

24. Peacock SM (1988) Inverted metamorphic gradients in the westernmost Cordillera.Metamorphism and Crustal Evolution of the Western United States, ed Ernst WG(Prentice Hall, Englewood Cliffs, NJ), Vol 7, pp 954–975.

25. Hacker BR, Mosenfelder JL, Gnos E (1996) Rapid emplacement of the Oman ophiolite:Thermal and geochronologic constraints. Tectonics 15(6):1230–1247.

26. Spray JG (1984) Possible causes and consequences of upper mantle decoupling andophiolite displacement. Geol Soc Lond Spec Publ 13:255–268.

27. Hacker BR (1994) Rapid emplacement of young oceanic lithosphere: Argon geo-chronology of the oman ophiolite. Science 265(5178):1563–1565.

28. Searle M, Cox J (1999) Tectonic setting, origin, and obduction of the Oman ophiolite.Geol Soc Am Bull 111(1):104–122.

29. Holloway NH (1982) North Palawan Block, Philippines—Its relation to Asian mainlandand role in the evolution of South China Sea. AmAssoc Pet Geol Bull 66(1):1355–1383.

30. Briais A, Patriat P, Tapponnier P (1993) Updated Interpretation of Magnetic Anom-alies and Seafloor Spreading Stages in the South China Sea: Implications for theTertiary Tectonics of Southeast Asia. J Geophys Res 98(B4):6299–6328.

31. Raschka H, Nacario E, Rammlmair D, Samonte G, Steiner L (1985) Geology of theophiolite of Central Palawan Island, Philippines. Ofioliti 10(1):375–390.

32. Mitchell AHG, Hernandez F, Dela Cruz AP (1986) Cenozoic evolution of the PhilippineArchipelago. J Southeast Asian Earth Sci 1(1):3–22.

33. Rangin EA, Silver EA (1991) Neogene tectonic evolution of the Celebes and SuluBasins: New insights from Leg 124 drilling. Proceedings of ODP Scientific Results, edsSilver EA, Fisk M, Rangin C, von Breymann MT (Ocean Drilling Program, CollegeStation, TX) Vol 124, pp 51–63.

34. Hamilton W (1979) Tectonics of the Indonesian region. US Geol Surv Prof Pap 1078(US Government Printing Office, Washington, DC), pp 197–201.

35. Hinz K, Schluter HU (1985) Geology of the dangerous grounds, South China Sea, andthe continental margin off southeast Palawan: Results of Sonne Cruises SO-23 and SO-27. Energy 10(3–4):297–315.

36. Letouzey J, Sage L (1988) Geological and Structural Map of Eastern Asia (AmericanAssociation of Petroleum Geologists, Tulsa, OK).

Keenan et al. PNAS | Published online November 7, 2016 | E7365

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 4

, 202

0

37. Milsom J, et al. (2009) The gravity fields of Palawan and New Caledonia: Insights

into the subsurface geometries of ophiolites. J Geol Soc (London, U.K.) 166(6):985–988.

38. Vermeesch P (2006) Tectonic discrimination diagrams revisited. Geochem Geophys

Geosys 7(6):Q06017.39. Hawkins JW (1995) Evolution of the Lau Basin—Insights fromODP Leg 135.ActiveMargins

and Marginal Basins of theWestern Pacific, eds Taylor B, Natlund J (American GeophysicalUnion, Washington, DC), pp 125–173.

40. Dilek Y, Furnes H (2011) Ophiolite genesis and global tectonics: Geochemical and

tectonic fingerprinting of ancient oceanic lithosphere. GSA Bull 123(3–4):387–411.41. Renne PR, et al. (1994) Intercalibration of astronomical and radioisotopic time.

Geology 22(9):783–786.42. Jaffey AH, Flynn KF, Glendenin LE, Bentley WC, Essling AM (1971) Precision mea-

surements of half-lives and specific activities of 235U and 238U. Phys Rev C Nucl Phys

4(5):1889–1906.43. Hiess J, Condon DJ, McLean N, Noble SR (2012) 238U/235U Systematics in terrestrial

uranium-bearing minerals. Science 335(6076):1610–1614.44. Rioux M, et al. (2013) Tectonic development of the Samail ophiolite: High-precision

U-Pb zircon geochronology and Sm-Nd isotopic constraints on crustal growth and

emplacement. J Geophys Res Solid Earth 118(5):2085–2101.45. Zeh A, Gerdes A, Will TM, Frimmel HE (2010) Hafnium isotope homogenization

during metamorphic zircon growth in amphibolite-facies rocks: Examples from theShackleton Range (Antarctica). Geochim Cosmochim Acta 74(16):4740–4758.

46. Cavosie AJ, Wilde SA, Liu D, Weiblen PW, Valley JW (2004) Internal zoning and U-Th-Pb chemistry of Jack Hills detrital zircons: A mineral record of early Archean toMesoproterozoic (4348-1576 Ma) magmatism. Precambrian Res 135(4):251–279.

47. Harley SL, Kelly NM, Möller A (2007) Zircon behavior and the thermal histories ofmountain chains. Elements 3(1):25–30.

48. Gilley LD, et al. (2003) Direct dating of left-lateral deformation along the Red Rivershear zone, China and Vietnam. J Geophys Res 108(B2):14-1–14-21.

49. Tapponnier P, Peltzer G, Le Dain AY, Armijo R, Cobbold P (1982) Propagating ex-trusion tectonics in Asia: New insights from simple experiments with plasticine.Geology 10(12):611–616.

50. House MA, Gurnis M, Kamp PJJ, Sutherland R (2002) Uplift in the Fiordland region,New Zealand: implications for incipient subduction. Science 297(5589):2038–2041.

51. LeBrun JF, Lamarche G, Collot JY (2003) Subduction initiation at a strike-slip plateboundary: The Cenozoic Pacific-Australian plate boundary, south of New Zealand.J Geophys Res 108(B9):15-1–15-18.

52. Meckel TA, et al. (2003) Underthrusting at the Hjort Trench, Australian-Pacific plateboundary: Incipient subduction? Geochem Geophys Geosyst 4(12):1–30.

53. Keenan T, Encarnación J (2016) Unclear causes for subduction. Nat Geosci 9(5):338.54. Schaltegger U, et al. (2009) Zircon and titanite recording 1.5 million years of magma

accretion, crystallization and initial cooling in a composite pluton (southern Adamellobatholith, northern Italy). Earth Planet Sci Lett 286(1–2):208–218.

55. Mattinson J (2005) Zircon U-Pb chemical abrasion (“CA-TIMS”) method: Combinedannealing and multi-step partial dissolution analysis for improved precision and ac-curacy of zircon ages. Chem Geol 220(1–2):47–66.

E7366 | www.pnas.org/cgi/doi/10.1073/pnas.1609999113 Keenan et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 4

, 202

0