Continental underplating after slab break-off

HAL Id: hal-01785205

https://hal.uca.fr/hal-01785205

Submitted on 3 Sep 2020

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Distributed under a Creative Commons Attribution| 4.0 International License

Continental underplating after slab break-off

V. Magni, M.B. Allen, J. van Hunen, P. Bouilhol

To cite this version:

V. Magni, M.B. Allen, J. van Hunen, P. Bouilhol. Continental underplating after slab break-off.

Earth and Planetary Science Letters, Elsevier, 2017, 474, pp.59 - 67. �10.1016/j.epsl.2017.06.017�.

�hal-01785205�

Contents lists available atScienceDirect

Earth

and

Planetary

Science

Letters

www.elsevier.com/locate/epsl

Continental

underplating

after

slab

break-off

V. Magni

a

,

b

,

∗

, M.B. Allen

b

,

J. van Hunen

b

, P. Bouilhol

c

,

b

a_{TheCentreforEarthEvolutionandDynamics(CEED),UniversityofOslo,SemSælandsvei24,POBox1048,Blindern,NO-0316Oslo,Norway} b_{DepartmentofEarthSciences,DurhamUniversity,DH13LE,Durham,UnitedKingdom}

c_{UniversitéClermontAuvergne,CNRS,IRD,OPGC,LaboratoireMagmasetVolcans,F-63000Clermont-Ferrand,France}

a

r

t

i

c

l

e

i

n

f

o

a

b

s

t

r

a

c

t

Articlehistory:

Received19September2016 Receivedinrevisedform2May2017 Accepted9June2017

Availableonline30June2017 Editor:B.Buffett Keywords: continentalcollision underplating slabbreak-off numericalmodelling subductiondynamics

We present three-dimensionalnumerical models to investigatethe dynamicsofcontinental collision, and inparticular what happensto thesubductedcontinental lithosphere after oceanic slabbreak-off. Wefindthatinsomescenariosthesubductingcontinentallithosphereunderthruststheoverridingplate notimmediatelyafteritentersthetrench,butafteroceanicslabbreak-off.Inthiscase,thecontinental platefirstsubductswithasteepangleandthen,aftertheslabbreaksoffatdepth,itrisesbacktowards thesurfaceandflattensbelowtheoverridingplate,formingathickhorizontallayerofcontinentalcrust thatextendsforabout200kmbeyondthesuture.Thistypeofbehaviourdependsonthewidthofthe oceanic platemarginal to the collisionzone: wideoceanic margins promotecontinental underplating and marginalback-arcbasins;narrowmargins do notshow suchunderplatingunless afarfield force is applied. Ourmodels show that, as the subductedcontinental lithosphere rises, the mantle wedge progressivelymigratesawayfromthesutureandthecontinentalcrustheatsup,reachingtemperatures

>900◦C. This heating might lead to crustal melting, and resultant magmatism. We observe asharp peakinthe overridingplaterockupliftrightaftertheoccurrenceofslabbreak-off.Afterwards, during underplating,themaximumrockupliftissmaller,buttheaffectedareaismuchwider(upto350km). Theseresultscanbeusedtoexplainthedynamicsthatledtothepresent-daycrustal configurationof theIndia–Eurasiacollisionzoneanditsconsequencesfortheregionaltectonicandmagmaticevolution.

©2017TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).

1. Introduction

The dynamicsof continentalcollision are complex dueto the manyforcesactinginthesystematthesametimeandthemany factorsthatcanaffectthem.Differentscenariosarepossiblewhen a continent reaches the subduction zone trench. For instance,in

the Apennines and the Carpathians many studies suggested that

delaminationofthelithosphericmantlefromthecontinentalcrust occurred, leavinga thinlayer of crust asthe lithosphericmantle keepssubducting(e.g.Bird,1979; Cloos,1993; BrunandFaccenna, 2008; Gö˘gü ¸setal., 2011). In other cases(e.g. Zagros,Himalayas) slabbreak-offoccursasaresultofthehightensilestressescaused

by the oceanic slab pull at depth and the buoyant continental

crust that resists subduction at the surface (e.g. Davies and von Blanckenburg,1995; WongATonandWortel,1997; Replumaz et al., 2010). Seismicstudies that focused more on the architecture

of the Himalayan area, however, showed that the Indian

con-*

Correspondingauthorat:TheCentreforEarthEvolutionandDynamics(CEED), UniversityofOslo,SemSælandsvei24,POBox1048,Blindern,NO-0316Oslo, Nor-way.

E-mailaddress:valentina.magni@geo.uio.no(V. Magni).

tinental lithosphere lies sub-horizontally underneath Eurasia for about 200 km north of the suture zone (Nábˇelek et al., 2009; Chen etal., 2010). This configurationhas alsobeen suggestedto bepresentinancientorogenies(e.g.theSlaveProvinceinCanada,

Helmstaedt, 2009; the Variscan orogeny in France,Averbuch and Piromallo,2012).It isstill unclearhowthe underthrustingofthe subductingcontinentallithosphereandslabbreak-offcoexistinthe samesystem.Inparticular,itispoorlyknownhowunderthrusting evolvesduring continentalcollision, thedynamicsofthisprocess, andthefactorsthatcontrolitsoccurrence.

The long-term history of the plates prior to collision is one of the key factors that can affect the evolution of the continen-tal collision itself. Old and cold continents, such as cratons, are strongerthanyoungercontinents,whichare usuallycharacterised byaweak,“jellysandwich”-styleductilelowercrust(Burov,2011), andfavoursthe decouplingbetweenuppercrust andlithospheric mantle needed for delamination to occur (Bajolet et al., 2012; Magni et al., 2013). Moreover, the dynamics of collision might alsobeaffectedbytheinteractionwithmantleconvectioncaused

by features like mantle plumes, or the formation of slab

win-dows, orthe presence ofother subduction zonesnearby. This is, forinstance,thecasefortheIndia–Eurasiacollision,inwhichthe

http://dx.doi.org/10.1016/j.epsl.2017.06.017

60 V. Magni et al. / Earth and Planetary Science Letters 474 (2017) 59–67

presence ofan external force hasbeen arguedfor toexplain the sustainedconvergencebetweentheplates(Chemendaetal.,2000; BeckerandFaccenna,2011; CandeandStegman,2011).

In the last decade many 2D and 3D numerical andanalogue

experiments onslab break-off allowed us tohave a much better understanding of the break-off process and its consequences on

the evolution of topography, stress field and magmatism (Wong

A Tonand Wortel,1997; Gerya et al., 2004; Duretz etal., 2011; van Hunen and Allen, 2011; Pusok and Kaus, 2015). However,

less work has been done in studying what happens to the

sub-ducted lithosphereafter slab break-off.Several numericalstudies havefound the process ofeduction to be geodynamically plausi-ble,wheresubductedcontinentallithospherecoherentlyexhumes throughthesuturezoneafterslabdetachment(Duretzetal.,2012; Bottrilletal., 2014).Inthisstudyweuse3Dnumericalmodelsof continentalcollisiontoinvestigatewhatcontrolstheoccurrenceof underplating,anddiscussthepossibleapplicationsofourmodelto theIndia–Eurasiacollisionsystem.

2. Methodology

We investigate the dynamics of continental collision with 3D

numerical models of subduction using the finite element code

CITCOM that solves the conservation of mass, momentum,

ther-malenergy,andcomposition inaCartesiangeometry(Moresiand Gurnis, 1996) (seeMagni etal. (2012) andTable 1for used val-ues of defaultparameters). Our models simulate the collision of a 2000 km wide continentalblock witha continentaloverriding plateafteran initial stageof oceanicsubduction (Fig. 1). Oceanic lithosphereofvariablewidthflanksthecontinentalblock, totake intoaccountthecomplexity ofnaturalsubductionsystems,where often oceanicand continentalsubduction happen simultaneously alongthetrench,andtobetterunderstandhowtheyinteractwith

each other. We vary the width of the oceanic margin and the

density of the continental crust to investigate what controlsthe occurrence of underplating and its dynamics. Moreover, we also runanadditionalmodelwithanimposedcontinuousconvergence betweentheplates.

Toreducecomputationalcosts,weexploitthesystem’s symme-tryalongtheplanethroughthecentreofthecontinentalblock per-pendiculartothetrench(x–z plane).Therefore,inthe y-direction

we model only half of the domain. The reference model (which

hasa computational domain size of 3300

×

2180

×

660 km) has

a 500-km wide oceanic part within the subducting lithosphere

(Fig. 1). Other model calculations have different oceanic plate widths(200–2000km),andinthosemodelsthecomputational do-mainsize in y-direction isadjusted accordingly(1850–3960km). Theinitialpositionofthetrenchisimposed(atx

=

1850 km),but

Table 1

Symbols,unitsanddefaultmodelparameters.

Parameters Symbols Valueandunit Rheological pre-exponent A 6.52×106_[Pan_s] Activation energy E∗ 360[kJ/mol] Gravitational acceleration g 9.8[m/s2_]

Rheological power law exponent n 1(diff.c.),3.5(disl.c.)[–] Lithostatic pressure p0 [Pa]

Gas constant R 8.3[J/K/mol] Absolute temperature Tabs [K] Reference temperature Tm 1350[◦C] Compositional density contrast ρc 500(300)[kg/m3] Strain rate ε˙i j [s−1]

Second invariant of the strain rate ε˙I I [s−1] Effective viscosity η [Pa s] Reference viscosity ηm 1020[Pa s] Maximum lithosphere viscosity ηmax 1024[Pa s] Friction coefficient μ 0.1[–] Reference density ρ 3300[kg/m3_]

Yield stress τy [MPa]

Surface yield stress τ0 40[MPa] Maximum yield stress τmax 400[MPa]

Model geometry

Domain depth h 660[km]

Domain length l 3300[km]

Domain width w 1848(2180–3690)[km] Mesh resolution from8×8×8 to20×20

×20 [km3_] Continental block half-width – 1000[km] Oceanic side width – 200(500–2000)[km] Continental crust thickness Hc 40(30)[km]

duringthemodelevolutionthetrenchisfreetomoveinresponse to thesystemdynamics(Magnietal., 2012).Subduction is facili-tated by imposingan initial oceanicslabthat extendsto 200km depth. Theinitial temperaturefield forthe oceaniclithosphereis calculated following a half-space cooling solution for an 80-Myr old plate (Turcotte and Schubert, 2002). In the reference model, thecontinentallithosphereismodelledwitha40-kmthicklayerof positivelybuoyantcrust(



ρ

c

=

500 kg/m3or

ρ

c

=

2

.

8 g/cm3)and

itstemperatureextendslinearlyfrom0◦CatthesurfacetoT

=

Tm

at150kmdepth.Toallow possiblemantleflow aroundtheedge

oftheslabandavoidartefacts duetothelateralboundary condi-tionthecomputationaldomainiswiderthanthesubductingplate. Thisismodelledbyimposingatransformfaultwitha20kmwide low viscosityzone at y

=

660 km.Forsimplicity,we assume the twoplatestohavethesamewidth.

Thermalboundary conditionsare: T

=

0◦Catthetop, T

=

Tm

atthe bottomandleftboundary (atx

=

0), andinsulating condi-tionsalongtherestoftheboundaries.Mechanicalboundary condi-tions arefree-slipeverywhereexceptthebottomboundarywhere

Fig. 1. Initialsetupandboundaryconditionsofthereferencemodel.(a)Thewidthoftheoceanicsideis500kminthereferencemodel,butisvariedintheothermodels: 200or2000km,andaccordingly(b)thewidthofthedomainisvaried:1850or3960km.

a no slip-condition is applied to model the effect of a viscosity contrast between upper and lower mantle (Fig. 1). We run one modelwithanimposedvelocityof5cm/yratthesurfacebetween x

=

0 andx

=

30 kmtoinvestigatetheeffectofafarfieldpushon thesubductingplateoncontinentalcollisiondynamics.

2.1.Rheology

The viscosity ofthe systemis temperature and stress depen-dent.Thestrain rateisaccommodated byboth diffusionand dis-location creep (Hirth andKohlstedt, 2003; Korenaga andKarato, 2008).Theeffectiveviscosity

η

foreachmechanismisdefinedas:

η

=

A

ε

˙

1−n n I I exp



E∗ nR Tabs



(1)

withsymbolsasdefined inTable 1 andwith

ε

˙

I I, thesecond

in-variantofthestrainratedefinedas:

˙

ε

I I

=



1 2

ε

˙

2 i j (2) where

˙

ε

_{i j}

=

∂

vi

∂

xj

+

∂

vj

∂

xi (3)

Inaddition, a yield mechanismis implemented to reduce the strengthofthelithosphere, forwhichan effectiveviscosityis

de-finedas

η

=

τ

y

˙

ε

(4)

with

τ

y istheyieldstressdescribedas

τ

y

=

min

(

τ

0

+

μ

p0

,

τ

max

)

(5)

where

τ

max is the maximum yield stress and

τ

0

+

μ

p0 is Byer-lee’slaw,with

τ

0 istheyieldstressatthesurface,

μ

isthefriction coefficient,andp0 isthelithostaticpressure.Atanypointthe ef-fectiveviscosity isthe minimumofviscosityvalues derived from eachmechanismdescribedabove.Toaccountforotherweakening

mechanismsatlowtemperatures,we imposea maximum

viscos-ity of 1024 Pa s. The same rheology is assumed for mantle and crustal material without any internal rheological layering in the lithosphere.

Anarrowweak zone (witha viscosity

η

=

1020 Pa s)between theplatesallowsplatedecoupling(Magnietal.,2012).The

weak-ening of the mantle wedge due to slab dehydration and mantle

re-hydrationandmeltingissimulatedbyimposing anarea above the slab with a maximum viscosity of 1020 _{Pa s, which can get} weakerifthecomputedeffectiveviscosityislower.

3. Results

3.1.Referencemodel

Wefirstinvestigatetheabilityofslabpullfromtheoceanicpart ofthesubductinglithospheretodriveunderplating.Theevolution ofthe referencemodel (Oc500, Fig. 2and Animation A1in Sup-plementaryMaterial)canbedividedintothreemainphases: con-tinental subduction (the onset of which defines initial collision), slabbreak-off,andunderplating.Inthefirstphase,partofthe con-tinentalblockwithinthesubductingplateisdraggeddownbythe oceanicpartoftheslabatdepth.Thesubductedcontinentalcrust reachesdepths

>

200kmand,atthisstage,hasadipofabout35◦ (Fig. 2a).Aftertheonsetofcontinentalcollision,thesubduction ve-locitydecreases,becausethelow densityofthecontinentalcrust opposessubduction.Eventually,convergencecompletelystops,and theslabheats upandnecks(Fig. 2b).Elsewherealongthetrench,

where the subducting plate is oceanic, subduction is still active andthetrenchisretreating.Thisfastretreatproduceshightensile stresses inthe overridingplate andresultsin theformation ofa back-arcbasin intheoverridingplate. Atdepth theslabishighly curved asonly theoceanic partis retreating, whereas the conti-nentalpartis stationaryorslowly advancing.About 35Myrafter theonset ofcontinentalcollision, theslabbreaks off(Fig. 2c).At thispointthepositivelybuoyantcontinentalcrustisabletoascend

back towards the surface. However, it doesnot exhume through

the suture zone (i.e., eduction), but, instead, flattens underneath theoverridingplate, formingathick horizontallayerof continen-talcrustthatextendsforabout200kmbeyondtheoriginalsuture betweenthetwoplates(Fig. 2d).Wereferto thisprocess as ‘un-derplating’.

Fig. 3showsthemantleflow immediatelyafterslabbreak-off, whenthe oceanicslab atthe sideofthecontinentalblock isstill subducting.Weobservetwotoroidalflowcellstriggered and con-trolledby theretreatof theoceanicslab(Fig. 3a).Faraway from thecontinent,themantleflowsaroundtheslabedge,frombehind to the front of the slab. In the second toroidal flow cell, closer to the continent, themantle flows from behind the oceanicslab towardsthe continentalpartofthesubducting plate,pushing the continentalslabtowardstheoverridingplate.Simultaneously,since break-offhasjusthappened,thesubductedcontinentallithosphere is rising towards the surface (Fig. 3b). The combinationof these two forces makes possible for the continental crust to underlay sub-horizontallytheoverridingplateand,thus,forthe underplat-ingtooccur.

3.2. Controlsonunderplating

We investigate what controls the occurrence of underplating after slab break-off with additional models (Table 2). The first set of models aims at studying the effect of different oceanic

plate widths and external forces: model Oc200 has a 200 km

wideoceanicsidewithin thesubductingplate,narrowerthanthe

reference model, model Oc2000 has a much wider oceanic side

(2000 km), and a final model, Oc200v5, with a narrow oceanic

side (200 km) and a 5 cm/yr velocity imposed to the

subduct-ing plate.In addition,wetest thefeasibilityofunderplating with different buoyancies of the subducting continental lithosphere:

model Oc500thin has a thinner continental crust (30 km) and

lithosphere (100 km) compared to the reference model, similar

tothe‘unstretchedcontinent’usedbyCapitanioetal. (2010),and modelOc500layeredincludesa20km-thickuppercrust(with

ρ

=

2

.

8 g/cm3_{)anda20km-thickdenserlowercrust(}

_ρ

₌

₃

_.

_{0 g/cm}3_). Fromthefirstsetofmodels,wefirstdescriberesultsfrom mod-elsOc200andOc2000toseewhatitistheeffectofthewidthof theoceanicpartofthesubductingplate(andthereforetheamount of available slabpull) on the dynamicsof thesystem after colli-sionandslabbreak-off.Then,we usemodelOc200v5toexamine theroleofanexternallyimposedforceonunderplating.

In model Oc200 the narrow oceanic side retreats only a lit-tle, after collision occurred nearby, and both the trench andthe

slab remain almost undeformed. In thiscase we do not observe

anyunderplating ofthepreviouslysubductedcontinentalcrust.In fact,afterbreak-off,thecontinentallithosphererisesback, exhum-ing through the suture zone (Fig. 4a), displaying the behaviour termed eduction (Andersen etal., 1991). On the other hand, the evolutionofmodelOc2000,witha verywide oceanicpartofthe subducting plate, is very similar to the reference model: a first phaseofnormalcontinentalsubductionisfollowedbyslab break-offandsubsequentflatteningofthecontinentallithospherebelow theoverridingplate(Fig. 4b).Interestingly,inthismodelonlythe partoftheoceanicplatethatis closetothecontinentalindentor retreats,deforms,andtriggerstheopeningofaback-arcbasin.The

62 V. Magni et al. / Earth and Planetary Science Letters 474 (2017) 59–67

Fig. 2. Evolutionofthereferencemodelwithafirstphaseofcontinentalsubduction(a),followedbyslabnecking(b)andbreak-off(c),and,finally,thecontinentallithosphere risesbacktowardsthesurfaceandflattensbelowtheoverridingplate(underplating)(d).Thefirstcolumnshowsthetemperaturefieldandthepositionofthecontinental crust(whitecontour)inaverticalsectionalongthesymmetryaxis.Themiddleandleftcolumnsarethefrontandtopview,respectively:aT=1080◦Cisosurface represent-ingthelithosphereisshowninblueandthecontinentalcrustisshowningrey.Greenarrowsshowthevelocityfieldatthesurface.Theredthicklineintherightcolumn representsthetrench.Theorangelineindicateswheretheverticalsectionshowedintheleftcolumnistaken.Timeisaftertheonsetofcollision.(Forinterpretationofthe referencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

Table 2

Modelsparametersandresults. Model Oceanic sidewidth (km) Imposed velocity (cm/yr) Cont.crust thickness (km) Cont.plate thickness (km) Cont.crust density (g/cm3₎ Underplating afterslab break-off Oc500 500 – 40 150 2.8 Yes Oc200 200 – 40 150 2.8 No Oc2000 2000 – 40 150 2.8 Yes Oc200v5 200 5 40 150 2.8 Yes Oc500thin 500 – 30 100 2.8 Yes

Oc500layered 500 – 40 (20 uc, 20 lc)a _{150 2.8 uc, 3.0 lc}a _Yes a _{uc:uppercontinentalcrust,lc:lowercontinentalcrust.}

restof the oceanicplate isalmost stationary assubduction con-tinues.Thepartoftheoceanicplatethatdoesretreatis

∼

700km wide, which isenough to triggerthe same type of toroidal flow observedinthereferencemodelthatpushesthesubducted conti-nentalcrusttoflattenbelowtheoverridingplate.

WealsotookthemodelOc200,inwhichunderplatingdoesnot

happen, and we performed the same model but witha 5-cm/yr

velocityimposedattheleftboundarytosimulatethepushofafar field forceonthesubductingplate:modelOc200v5.Asformodel Oc200, the oceanic side is too narrow to have significant trench

Fig. 3. Mantleflowofthereferencemodel.(a)Mantle flowinahorizontalsectionatadepthof200km.Coloursshowthex-componentofthevelocity(positivevaluesare towardstheoverridingright);theblacklineshowswheretheslabis.(b)Temperaturefieldandmantleflowofaverticalsectionatthesymmetryaxis.Whitearrowsrepresent thegeneraltrendofthemantleflowthattriggeredbytheretreatoftheoceanicslabpushestherisingcontinentalcrustbelowtheoverridingplate.(Forinterpretationofthe referencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

Fig. 4. Resultsofthemodel(a)Oc200,withnarrowoceanicsides(200km),where eduction,thusnounderplating,occursafterslabbreak-off.Theprocessof under-platingoccursinbothmodels(b)Oc2000,withwideoceanicsides(2000km),and (c)Oc200v5,withnarrowoceanicsideandanimposedvelocityof5cm/yronthe subductingplate.Thesmallgreytriangleindicatestheoriginalsuturelocation.See

Fig. 2captionfordetailsonthecolourlegend.

retreat and slab deformation and, thus, to be able to affect the dynamics of the rest of the system laterally. However, after the slab breaks off,the continentallithosphere is unable to exhume through the suture zone because of the continued plate conver-gence; the subducted continentallithosphereflattens underneath theoverridingplate(Fig. 4c).

Finally,we testedthe effectofthecontinental platebuoyancy ontheunderplating process.ModelOc500thin hasa thinner con-tinental lithosphere(100 km instead of150 kmof the reference model) as well as a thinner continental crust (30 km instead of 40 km) (Fig. 5a). Results are similar to the reference model, as also herewe observe underplating occurringafter slab break-off. However, this process occurs faster than in Oc500. Because the lithosphereis thinner, it takeslesstime for theslab to break-off (23 Myr after collision) and the previously subducted continen-tal lithosphereunderlies sub-horizontallythe upperplatealready

28 Myr after collision. We observe underplating even when the

lower crustisslightlydenser, inOc500layered(Fig. 5b).However, the part of the subducting continental plate that underlies the overriding plateis lessinthiscase, sincethe continentalcrust is onaveragedenserthaninthereferencemodel,andthereforeitis moredifficult foritto drivetheunderplating. Some density con-trastbetweenthesubductedcontinentandthesurroundingmantle seemsnecessarytodriveriseandunderplatingafterslabbreak-off, andindeedtopromoteslabbreak-offinthefirstplace(vanHunen andAllen,2011).

3.3. Consequencesofunderplating

During the evolution of the reference model, from the ‘nor-mal’continentalsubductionstagetotheunderplating,thethermal structureoftheentiresubductionzoneundergoessignificant vari-ations(Fig. 6).Indeed,therisingofthesubductedcontinentalcrust andits flatteningunderneaththe overridingplateforce the man-tlewedgetoprogressivelymigrateawayfromthetrenchbyabout 200kmovera periodof

∼

35Myr.Weobservea slight tempera-turedecreaseatthetopoftherisingcontinentalcrustclosetothe suturezoneduetothemantlewedgemigration.Atthesametime, however,mostofthesubducted continentalcrustbecomes hotter duringthebreak-offandunderplatingstagesandthedeepestpart reachestemperatures

>

900◦C(Fig. 6b).

Weestimatetherockuplift(asdefinedbyEnglandandMolnar, 1990) inthe uppercrust aposteriori fromthe normalstresses at the topof themodel byassuming the relationship betweenrock

64 V. Magni et al. / Earth and Planetary Science Letters 474 (2017) 59–67

Fig. 5. Resultsofthemodel(a)Oc500thin,withathincontinentalcrust(30km)and lithosphere(100km),and(b)modelOc500layeredwitha20km-thickuppercrust (withρ=2.8 g/cm3_{)anda20km-thickdenserlowercrust(ρ=3.0 g/cm}3_).Both modelsshowtheoccurrenceofunderplatingafterslabbreak-off.SeeFig. 2caption fordetailsonthecolourlegend.

uplift (H ) and normalstress (

σ

zz) is,to first order, H

=

σ

zz/

ρ

g,

where

ρ

is thereferencedensityand g isthegravitational accel-eration. Thissimple approachignores effects fromelastic flexure, nearsurfacetectonics,anderosion/sedimentation,butgivesa use-ful first-order indication of rock uplift. Whilst we do not specif-ically model elevationchanges (i.e. surface uplift) orexhumation rates,rockupliftwouldimplyariseinsurface elevationunlessit

wascompletelymatchedbyexhumation.Weobserveasharppeak

intheoverridingplaterockupliftrightaftertheoccurrenceofslab break-off(Fig. 6a).Afterwards,duringunderplating,themaximum rockupliftissmaller,buttheaffectedareaislarger(upto350 km). 4. Discussion

Our results show how the subducting continental lithosphere can rise up after slab break-off and flatten below the overrid-ing plate. Previous 3D numerical studies with a buoyant inden-tor within an oceanic plate showed similar behaviours in terms of continental subduction, trench migration and overriding plate deformation (e.g.,Moresietal., 2014; PusokandKaus, 2015). In-deed,thesemodelsalsoshow thepartialsubductionofthe buoy-antindentorand,laterally,fasttrenchretreatoftheoceanicplate, causing the overriding plateto stretch enough to create a back-arcbasin.Duringslabrollback,themantleflowsaroundtheedges oftheoceanicslab,inatoroidalfashion,aspreviouslyshowedin many3D analogueandnumericalmodels (Funiciello etal., 2006:

Stegman et al., 2006). Moreover, the dynamics of slab break-off in our models is similar to previous numerical studies (Wong A Ton andWortel,1997; Duretz etal., 2011; vanHunen andAllen, 2011), withthe necking andconsequent rupture of theslab due to the two opposite forces ofthe oceanicslab pull atdepth and

Fig. 6. Evolutionoftherockupliftandcontinentalcrustposition(leftcolumn)andtemperaturechangeswithinthecontinentalcrust(rightcolumn)duringthethreemain phasesofthereferencemodel:continentalsubduction(lightblue),slabbreak-off(orange),withapeakinthesurfaceupliftclosetothesutureandtemperatureincreasein thesubductedcontinentalcrust,andunderplating(red)inwhichtheupliftislessbutextendsoverawiderareaandthecrustreachestemperatures>900◦Catdepth.The smallgreytriangleindicatestheoriginalsuturelocation.TimesareMyrafterinitialcontinentalcollision.(Forinterpretationofthereferencestocolourinthisfigurelegend, thereaderisreferredtothewebversionofthisarticle.)

thebuoyancyofthecontinentalmaterialatshallowerdepths.The

subsequent underplating, however, is not a commonly observed

behaviour and it has been documented, to our knowledge, only

inafew models(Lietal.,2013).This isduetothefact thatthis processdoesnot occurwhen thesystemisbuoyancy-drivenonly andnolateralvariationsarepresentalongthetrench.Instead,our resultsshowthat fortheunderplating tooccurafterbreak-off an additionalforce is needed; thisforce could be provided by a far fieldpushonthesubductingplateorbythemantleflowandslab deformationassociatedwiththeflankingoceanicsubduction.

Amore oftenmodelledbehaviour in continentalcollision set-tings is the process of eduction, in which after slab break-off, i.e. after the loss of slab pull, the subducted continental plate

coherently educts, leading to the exhumation of HP-UHP rocks

along the shearzone betweenthe plates (Andersen et al., 1991; Duretzet al., 2012). In thisstudy, we observethe mechanismof eductionwhenthereisnoimposedvelocitytothesubductingslab andtheoceanicsideadjacenttothecontinentistoonarrowto af-fect laterallythe dynamicsof thesystem (model Oc200,Fig. 4a).

In the model with an imposed velocity to the subducting plate

(modelOc200v5,Fig. 4c)orinthosewithwideoceanicsidesthere isaforcethat opposeseductionandmakesunderplating possible after the loss of slab pull due to slab break-off (model Oc2000,

Fig. 4b).Inthefirstcase, thisforce issimplytheimposed contin-uous pushto thesubducting plate. Inthe second case, thisforce is provided by the toroidal mantle flow and by the internal de-formationtriggered bytheretreatoftheoceanicslabclosetothe collisionzone.

The buoyancy of the subducting plate is an important factor in this process. Indeed, what drives the continental lithosphere to rise after the slab breaks off is its positive buoyancy. Conti-nents can have different structures (e.g. thickness, composition) andestimatingtheprecisedensityprofileofthecontinentalcrust isnota trivialtask,since geophysicalandpetrological modelsdo

not give a unique answer and do not always agree. Our results

showthatunderplatingafterslabbreak-offcanoccurevenwhena higherdensityforthelowercontinentalcrustisconsidered(model Oc500layered, Fig. 5b). For continents with even higher density crust, it would be increasingly more difficult for underplating to occur. However, the denser the continental crust, the easier it wouldbe for it tosubduct and, thus, the more unlikely it isfor break-off to happen. Although on-going subduction of continen-tallithospherehasbeensuggestedinsomecases(Capitanioetal., 2010),slabbreak-offisarguablyamuchmorecommonbehaviour. Oneofthemainconsequencesofthisprocessisthesignificant changeinthetemperaturefieldofthesubductionzone,whichwill

have important effects on the presence and type of magmatism

relatedtoit. Inparticular, the progressivemigration ofthe

man-tle wedge away from the suture zone would most likely result

ina shiftof volcanism as well. Furthermore, the warming-upof thecontinentalcrustduringitsunderplatingmightleadtocrustal meltingandthereforeaffectthecompositionofthemagmatismat thesurface.

Theprecise timing ofbreak-off(and subsequentunderplating) dependsonmanyfactors that we havenot explicitlyexplored in thisstudysuchastherheology,thicknessandcompositionofthe plates (Duretz et al., 2011; van Hunen and Allen, 2011).

More-over,inourmodelsweassume theboundarybetweenupperand

lowermantletobe impermeable,whereasinnatureslabs can ei-ther stagnate at or penetrate the upper–lower mantle transition (e.g., Fukaoand Obayashi, 2013). Ifthe slab could sink into the lowermantle,itspullwouldmostlikelybehigher,andweexpect slab break-off andthe subsequent underplating process to occur earlierwith respect to theonset of continental collision. Thisis, however,beyondthescopeofthisstudyand,althoughthetiming oftheprocessmightchange,thedynamicsprobablywouldnot.

4.1. ApplicationtoIndia–Eurasiacollision

Numerical models are inherently a more generic and simpli-fied version of reality. Unlike other studies, our model setup is not designed to specifically reproduce the complexmulti-phased India–Eurasia collision scenario (e.g. Bouilhol etal., 2013). How-ever,thesemodelscanstillbeusedtohelpusunderstandthe dy-namicsandsomeaspectsofthiscomplexsystem.ModelOc200v5,

withthe 200-km oceanic plateadjacentto a 2000 km

continen-tal block andwith an imposed velocity of5 cm/yr,is the model that best represents the active collision of the

∼

2000 km wide Indian continental block with the overriding Eurasian plate. The farfield forcesimulatesthecontinuous on-goingconvergence be-tween the plates of

∼

5 cm/yr (Copley et al., 2011). We do not investigate the nature of this force, which has previously been interpretedasridgepush,basaltraction,pullbyneighbouring sub-ducted slabs, or push of a plume associated “conveyor belt” in the mantle (Chemenda et al., 2000; Becker andFaccenna, 2011; CandeandStegman,2011).Thepresenceofnarrowoceanicplates adjacenttotheIndiancontinentalblockisconsistentwiththe tec-tonicreconstructionssuggestedforthisarea(Hall,2012).

In our model the underplating results in the formation of a thick horizontallayer of continental lithosphere that extends for about200kmbeyondthesuture. Sucha featureispresentin Ti-bet,wheregeophysicalstudiesimagedtheratherflatIndian litho-sphereasfaras250kmnorthofthemainsuture (e.g.Wittlinger etal.,2009; Chenetal., 2010) withan overallgeometrythat cor-respondstothoseobtainedfromthemodels.

As the Indian plate becomes sub-horizontal and underplates Eurasia,theasthenosphere abovetheremainingslabprogressively

migrates away from the suture zone and eventually the whole

slab flattens below the Eurasian continental lithosphere (Powell,

1986). This process should correspond to a change in

magma-tismwithin theLhasaterrane,sincemantlewedgemeltingwould

be shut down. It has been shown that the source of

magma-tism observed within theLhasa terrane evolve during thecourse of collision, shifting during the late Eocene from calc-alkaline – subduction dominated magmas to ultrapotassic lavas originating from a metasomatic mantle (e.g. Chung et al., 2005). From the Oligocene, the presence of the underthrusted Indian lithosphere

is documented by the presence in the magmatic rocks of the

southern Lhasa terrane of inherited zircon grains of Indian ori-gin (Bouilholet al., 2013). The observed temperatureincrease of theunderplatedcontinentalcrust beyondtypicalsolidus tempera-turesfitsthegeochemistryofvolcanismfromthenorthernTibetan Plateau, where acontribution fromcontinental crust is modelled tobepresentinthesourceregionoftheNeogene–Quaternary vol-canic rocks (Guo et al., 2006). Moreover, the end results ofslab break-off and underplating traps a non-negligible amount of the overridingplatemantlelithospherebetweenthetwoplates,which is demonstrated to be involved in the source of post-collisional magmas(e.g.Williamsetal.,2004).

Early–mid Miocene, synchronous thrusting and

extensional-senseshearintheHimalayasareadistincteventsintheevolution ofthecollision.Conceptualmodelsrequiringchannelflow (Grujic etal.,2002) needtoexplainthepreciseandratherrestricted time-frameforthistectonicphase.MiddleMiocenetorecenteast–west extensionacrossthebeltandthesouthernTibetanPlateau relates tothehighelevationandgravitationalpotentialenergyofthisarea (Elliottet al., 2010). Conventionalcrustal thickening alone might produce a state whereby the buoyancyforce associatedwith the elevated crust resisted further shortening in the elevated region, butit wouldnotproduce theobserved extensionunless a signif-icant change in boundary conditions took place (England et al., 1988).

66 V. Magni et al. / Earth and Planetary Science Letters 474 (2017) 59–67

Our modelledrock uplift, and consequent surface uplift,

aris-ing from slab break-off echoes very well with both High

Hi-malayanandTibetan Plateauevolution inthelate Cenozoic.First, the early stages of slab break-off results in a high peak in rock

uplift within

∼

80 km north of the suture. This would perturb

theHimalayanwedge,whichrespondsby extensionontheSouth

TibetanDetachmentSystemtorestorethecriticaltaperangle, be-tween

∼

23 and

∼

16 Ma (Kohn, 2014). After break-off, slab

re-bound and underplating further propagate the uplift away from

the suture, as far as

∼

350 km in our models. Since

∼

15 Ma,

rise of the southern Tibetan Plateau has resulted from slab flat-tening, and triggered east–west extension (Copley et al., 2011; Styronet al., 2015). Toexplain thegrowth of theentire plateau,

which is

∼

1000 km across, from north to south, other

mecha-nismshavetobe takenintoaccount.Forinstance,recent numer-ical studies showed how pre-existing heterogeneities within the overriding plate, especially in terms of rheology,play an impor-tantrole inbuildingtheorogenicplateau(PusokandKaus, 2015; ChenandGerya,2016).Ourmodelsdonotincludesuch complex-ities as inthis studywe focus on theeffect of the underplating process on the tectonics. Overall, oceanic slab break-off and the subsequentunderthrustingoftheIndiancontinentalplatebeneath Eurasiahelpexplain thelateCenozoicchangesintectonics inthe Himalayan–Tibetansystem.

5. Conclusions

Weperformed3D numericalmodelsto studythe dynamicsof

continentalcollisionand,morespecifically,theunderplatingofthe subductingcontinentalplatebeneaththeoverridingplate.Wefind thatthepreviouslysubductedcontinentallithospherecanriseback towards the surface and flatten below the overriding plate after slabbreak-off,formingathickhorizontallayerofcontinentalcrust thatextendsforabout200kmbeyondthesuture.Ourresultsshow that this process can occur only ifthere isa force that opposes eduction. Mechanisms that can provide this force could be, for instance,a far-field pushon the subducting plateor thetoroidal mantle flow triggered by the retreat of the oceanic slab nearby collision,iftheoceanicplateiswideenough.

Ourmodelsshowthatasthesubductedcontinentallithosphere flattensbelowtheoverridingplatethemantlewedgeprogressively migratesawayfromthesutureanditscrustheatsupandcanreach temperatures

>

900◦C,whichmightleadtocrustalmelting.A sig-nature ofthisheating would mostlikelyshow in theoccurrence andcompositionofsyn-collisionvolcanism.Immediatelyafterslab break-offthereisamarkedcontributiontorockupliftclosetothe suture.Afterwards, themaximumuplift effectisreduced, but af-fectsawiderareaintheoverridingplate.

Theseresultscan beused toexplain not onlythepresent-day crustal configurationof the India–Eurasiacollision zone, butalso thenorthwardshiftofvolcanismandthechangingdistributionand style of deformation in the Himalayas and Tibetan Plateau since

∼

20–15 Ma.

Acknowledgements

We thank M. Jadamec and A. Replumaz for their comments

andsuggestions that significantly improved the manuscript. This

studywas supported by theEuropean Research Council(ERC StG

279828).VMalsoacknowledges supportfromtheResearch

Coun-cil of Norway through its Centres of Excellence funding scheme,

Project Number 223272. MBA acknowledges Natural

Environ-mentResearchCouncilgrantNE/H021620/1.PBalsoacknowledges

his Auvergne Fellowships (French Government Laboratory of

Ex-cellence initiative no. ANR-10LABX-0006, ClercVoc contribution no. 252).ThisworkmadeuseofthefacilitiesofN8HPCprovided

andfundedbytheN8consortiumandEPSRC(grantEP/K000225/1) andtheUNINETTSigma2computationalresource allocation

(No-turNN9283KandNorStoreNS9029K).

Appendix A. Supplementarymaterial

Supplementarymaterialrelatedtothisarticlecanbefound on-lineathttp://dx.doi.org/10.1016/j.epsl.2017.06.017.

References

Andersen,T.B.,Jamtveit,B.,Dewey,J.F.,Swensson,E.,1991.Subductionandeduction ofcontinental crust:major mechanismsduring continent–continentcollision andorogenicextensionalcollapse,amodelbasedonthesouthNorwegian Cale-donides.TerraNova 3,303–310.

Averbuch,O.,Piromallo,C.,2012.IstherearemnantVariscansubductedslabinthe mantlebeneaththeParisbasin?ImplicationsforthelateVariscanlithospheric delaminationprocessand theParisbasinformation.Tectonophysics 558–559, 70–83.

Bajolet,F.,Galeano,J.,Funiciello,F.,Moroni,M.,Negredo,A.-M.,Faccenna,C.,2012. Continentaldelamination:insightsfromlaboratorymodels.Geochem.Geophys. Geosyst. 13,Q02009.

Becker,T.W.,Faccenna,C.,2011. MantleconveyorbeneaththeTethyancollisional belt.EarthPlanet.Sci.Lett. 310,453–461.

Bird,P.,1979.ContinentaldelaminationandtheColoradoPlateau.J.Geophys.Res., SolidEarth 84,7561–7571.

Bottrill,A.D.,vanHunen,J.,Cuthbert,S.J.,Brueckner,H.K.,Allen,M.B.,2014.Plate rotation during continental collisionand itsrelationship with the exhuma-tionofUHPmetamorphicterranes:applicationtotheNorwegianCaledonides. Geochem.Geophys.Geosyst. 15,1766–1782.

Bouilhol,P.,Jagoutz,O.,Hanchar,J.M.,Dudas,F.O.,2013.DatingtheIndia–Eurasia collisionthrougharcmagmaticrecords.EarthPlanet.Sci.Lett. 366,163–175. Brun, J.-P.,Faccenna,C., 2008.Exhumationofhigh-pressurerocks drivenbyslab

rollback.EarthPlanet.Sci.Lett. 272,1–7.

Burov, E.B., 2011. Rheologyand strength ofthe lithosphere. Mar. Pet. Geol. 28, 1402–1443.

Cande, S.C.,Stegman,D.R., 2011.Indianand Africanplatemotionsdrivenbythe pushforceoftheReunionplumehead.Nature 475,47–52.

Capitanio, F.A., Morra, G., Goes, S., Weinberg, R.F., Moresi, L., 2010. India–Asia convergence drivenby the subductionofthe Greater Indiancontinent.Nat. Geosci. 3(2),136–139.

Chemenda,A.I.,Burg,J.-P.,Mattauer,M.,2000.EvolutionarymodeloftheHimalaya– Tibet system:geopoem:based onnewmodelling,geologicaland geophysical data.EarthPlanet.Sci.Lett. 174,397–409.

Chen,L.,Gerya,T.V.,2016.Theroleoflaterallithosphericstrengthheterogeneities inorogenicplateaugrowth:insightsfrom3-Dthermo-mechanicalmodeling.J. Geophys.Res.,SolidEarth 121(4),3118–3138.

Chen,W.-P.,Martin,M.,Tseng,T.-L.,Nowack,R.L.,Hung,S.-H.,Huang,B.-S.,2010. Shear-wavebirefringenceand currentconfigurationofconverginglithosphere underTibet.EarthPlanet.Sci.Lett. 295,297–304.

Chung,S.-L.,Chu,M.-F.,Zhang,Y.,Xie,Y.,Lo,C.-H.,Lee,T.-Y.,Lan,C.-Y.,Li,X.,Zhang, Q.,Wang,Y.,2005.Tibetantectonicevolutioninferredfromspatialandtemporal variationsinpost-collisionalmagmatism.Earth-Sci.Rev. 68,173–196. Cloos, M., 1993. Lithospheric buoyancy and collisional orogenesis: subduction

of oceanic plateaus, continental margins, island arcs, spreading ridges, and seamounts.Geol.Soc.Am.Bull. 105,715–737.

Copley,A.,Avouac,J.-P.,Wernicke,B.P.,2011.Evidenceformechanicalcouplingand strongIndianlowercrustbeneathsouthernTibet.Nature 472,79–81. Davies,J.H.,vonBlanckenburg,F.,1995.Slabbreakoff:Amodeloflithosphere

de-tachmentanditstestinthemagmatismanddeformationofcollisionalorogens. EarthPlanet.Sci.Lett. 129,85–102.

Duretz,T.,Gerya,T.V.,Kaus,B.J.P.,Andersen,T.B.,2012.Thermomechanicalmodeling ofslabeduction.J.Geophys.Res.,SolidEarth 117,B08411.

Duretz,T.,Gerya,T.V.,May,D.A.,2011. Numericalmodellingofspontaneousslab breakoffandsubsequenttopographicresponse.Tectonophysics 502,244–256. Elliott,J.R.,Walters,R.J.,England,P.C.,Jackson,J.A.,Li,Z.,Parsons,B.,2010.Extension

onthe Tibetanplateau:recentnormalfaultingmeasuredbyInSARandbody waveseismology.Geophys.J.Int. 183,503–535.

England,P.,Molnar,P.,1990.Surfaceuplift,upliftofrocks,andexhumationofrocks. Geology 18(12),1173–1177.

England,P.C.,Houseman,G.A.,Osmaston,M.F.,Ghosh,S.,1988.Themechanicsof theTibetanPlateau[anddiscussion].Philos.Trans.R.Soc.Lond.Ser.A,Math. Phys.Sci. 326,301.

Fukao,Y.,Obayashi,M.,2013.Subductedslabsstagnantabove,penetratingthrough, andtrappedbelowthe660kmdiscontinuity.J.Geophys.Res.,SolidEarth 118, 5920–5938.

Funiciello,F.,Moroni,M.,Piromallo,C.,Faccenna,C.,Cenedese,A.,Bui,H.A.,2006. Mappingmantleflowduringretreatingsubduction:laboratorymodelsanalyzed byfeaturetracking.J.Geophys.Res.,SolidEarth 111(B3).

Gerya,T.V.,Yuen,D.A.,Maresch,W.V.,2004.Thermomechanicalmodellingofslab detachment.EarthPlanet.Sci.Lett. 226,101–116.

Gö˘gü ¸s, O.H., Pysklywec, R.N., Corbi, F., Faccenna, C., 2011. The surface tecton-icsofmantle lithosphere delaminationfollowing oceanlithosphere subduc-tion:insights fromphysical-scaledanalogueexperiments.Geochem.Geophys. Geosyst. 12,Q05004.

Grujic,D.,Hollister,L.S.,Parrish,R.R.,2002.Himalayanmetamorphicsequenceasan orogenicchannel:insightfromBhutan.EarthPlanet.Sci.Lett. 198,177–191. Guo,Z.,Wilson,M.,Liu,J.,Mao,Q.,2006.Post-collisional,potassicand

ultrapotas-sicmagmatismofthenorthernTibetanPlateau:constraintsoncharacteristics ofthemantlesource,geodynamicsettingandupliftmechanisms.J.Petrol. 47, 1177–1220.

Hall,R.,2012.LateJurassic–CenozoicreconstructionsoftheIndonesianregionand theIndianOcean.Tectonophysics 570–571,1–41.

Helmstaedt,H.,2009. Crust–mantlecouplingrevisited: theArcheanSlavecraton, NWT,Canada.Lithos 112(Supplement2),1055–1068.

Hirth,G.,Kohlstedt,D.,2003.Rheologyoftheuppermantleandthemantlewedge: aviewfromtheexperimentalists.In:InsidetheSubductionFactory,pp. 83–105. Kohn,M.J.,2014.Himalayanmetamorphismanditstectonicimplications.Annu.Rev.

EarthPlanet.Sci. 42,381–419.

Korenaga,J.,Karato,S.-I.,2008.Anewanalysisofexperimentaldataonolivine rhe-ology.J.Geophys.Res.,SolidEarth 113,B02403.

Li,Z.-H.,Xu,Z.,Gerya,T.,Burg,J.-P.,2013.Collisionofcontinentalcornerfrom3-D numericalmodeling.EarthPlanet.Sci.Lett. 380,98–111.

Magni,V.,Faccenna,C.,vanHunen,J.,Funiciello,F.,2013.Delaminationvs.break-off: thefateofcontinentalcollision.Geophys.Res.Lett. 40,285–289.

Magni,V.,vanHunen,J.,Funiciello,F.,Faccenna,C.,2012.Numericalmodelsofslab migrationincontinentalcollisionzones.SolidEarth 3,293–306.

Moresi,L.,Betts,P.G.,Miller,M.S.,Cayley,R.A.,2014.Dynamicsofcontinental accre-tion.Nature 508(7495),245–248.

Moresi,L.,Gurnis,M.,1996.Constraintsonthelateralstrengthofslabsfrom three-dimensionaldynamicflowmodels.EarthPlanet.Sci.Lett. 138,15–28. Nábˇelek,J.,Hetényi,G.,Vergne,J.,Sapkota,S.,Kafle,B.,Jiang,M.,Su,H.,Chen,J.,

Huang,B.-S.,Hi-CLIMBTeam,2009.UnderplatingintheHimalaya–Tibetcollision zonerevealedbytheHi-CLIMBexperiment.Science 325,1371.

Powell, C.M., 1986. Continental underplating model for the rise of the Tibetan Plateau.EarthPlanet.Sci.Lett. 81,79–94.

Pusok, A.E., Kaus, B.J.P., 2015. Development of topography in 3-D continental-collisionmodels.Geochem.Geophys.Geosyst. 16,1378–1400.

Replumaz,A.,Negredo,A.M.,Villaseñor,A.,Guillot,S.,2010.Indiancontinental sub-ductionandslabbreak-offduringTertiarycollision.TerraNova 22,290–296. Stegman,D.R., Freeman,J., Schellart,W.P.,Moresi, L.,May,D.,2006. Influenceof

trenchwidthonsubductionhingeretreatratesin3-Dmodelsofslabrollback. Geochem.Geophys.Geosyst. 7(3).

Styron,R.,Taylor,M.,Sundell,K.,2015.AcceleratedextensionofTibetlinkedtothe northwardunderthrustingofIndiancrust.Nat.Geosci. 8,131–134.

Turcotte,D.L.,Schubert,G.,2002.Platetectonics.In:Geodynamics,2ndedn. Cam-bridgeUniversityPress,Cambridge/NewYork,pp. 1–21.

vanHunen,J., Allen,M.B.,2011.Continental collisionand slabbreak-off:a com-parisonof3-Dnumericalmodelswithobservations.EarthPlanet.Sci.Lett. 302, 27–37.

Williams,H.M.,Turner,S.P.,Pearce,J.A.,Kelley,S.P.,Harris,N.B.W.,2004.Natureof the sourceregions for post-collisional,potassic magmatisminsouthern and northern Tibetfrom geochemicalvariations andinversetrace element mod-elling.J.Petrol. 45,555–607.

Wittlinger,G.,Farra,V.,Hetényi,G.,Vergne,J.,Nábˇelek,J.,2009.Seismicvelocitiesin SouthernTibetlowercrust:areceiverfunctionapproachforeclogitedetection. Geophys.J.Int. 177,1037–1049.

WongATon,S.Y.M.,Wortel,M.J.R.,1997.Slabdetachmentincontinentalcollision zones:ananalysisofcontrollingparameters.Geophys.Res.Lett. 24,2095–2098.