GEOPHYSICAL RESEARCH LETTERS, VOL. 25, NO. 4, PAGES 551-554, FEBRUARY 15, 1998 Large earthquake nucleation associated with stress exchange between middle and upper crust MireilleHuc, 1 Rind H•ss•ni,e •nd JeanChary Laboratoire de G•ophysiqueet Tectonique,UMR CNRS 5573, Universit• Montpellier II, France. Abstract. Earthquake nucleationand propagationcan be explainedusingtwo basicconcepts,crustal strength and fault theology. Despite numerousexperimentaland modelingstudies,the way in which a faulted crust generates a seismic cycle under loading is still controversial. We discussearthquake generation using a 2-D out that frictional stick-slip on a fault could be the main operating mechanism for earthquake generation and distribution. This view comes from the extrapo- lation of experimentson crustalrocksfrom laboratory to geologicalscale, which suggestsa velocity weaken- ing of the frictionlaw. The transitionfrom weakening to strengtheningwith velocityoccursat approximately of geodynamicalproblemssuch as theologicallayering, 3500 C [$tesky,1975]in quartz-richrocks,supportdynamic Coulomb faulting and relative plate velocity. ing the idea that the limit of earthquakesis causedby The occurrence of a deep nucleation is explained by the transition from unstable to stable slip with depth the strong viscosity drop at the base of the seismo- [$cholz,1988].Because we think that crustalstrength genic layer, without any transition of frictional prop- and fault rheologyare both deeplyinvolvedin the mechFinite Element Model which includes relevant features erties of the fault. Rather, viscosity contrast causes a anism of earthquakes, we combined them in a finite el- stress transfer ement model of continental around the fault that reloads the fault plane during the postseismicphase. We argue that this mechanismmay explain the seismiccyclefor large faults in the continental extension. 2. Mechanical modeling crust. A classicview of the strength of continental crust is providedby Brace and Kohlstedt[1980], who showed 1. Introduction pressure and temperature effects on differential stress for quartz-rich rocks. At shallow depth, low confining pressurein the brittle field restricts stresslevels to low A fairly good correlationexistsbetweenthe base of the seismogenic layer and expectedtemperaturesin the values.At intermediatedepths(4-12 km), higherconcontinental crust. Earthquakes that nucleate at 10-15 fining pressureand moderately high temperature perkm depth are typical for hot continental crust, like the mits higherstress.At greaterdepth, increasingtemperBasin and Range province in the Western US, while ature beyond 3500 C leads to stress decay due to visdeeperearthquakes(15-30 km) are mainly represented cousdissipation. In order to describethis peak strength in regionsof coldercrust(Rhinegraben,Baikal). Meiss- within the crust, we divide the crust into three visher and $trehlau[1982]and $ibson[1982],pointedout coelasticlayers(Figure 1). This layeredsystemis cut this fact, and proposed that the seismicity cut-off at by a planar,high-angle,deeply-rooted (16 km) fault in depth is controlled mainly by temperature. Beyond this the low-viscositylayer, with an infinite out-of-plane ex- correlation,$ibson[1982],stated that "the nucleation tension accordingto the plane strain geometry of our of larger earthquakesnear the base of the seismogenic zone, in the regionof inferredpeak resistance,suggestsa causal relationship". In this paper we show that such a relationship appears in our modeling of the earthquake cycle, and demonstrate how this modifies our current understandingof earthquakemechanics. While our approachimplicitly refers to crustal strength for explaining earthquake distribution, some authors model. We assumethat the dynamic friction coefficient /•d is lower(0.24) than the staticoneits (0.30). In any mechanical modeling, the way in which the crust-fault system behaves is strongly dependent on boundary conditions. Becausewe want to study the interaction between fault rheology and crustal strength, we do not prescribe any boundary condition near the fault zone. We rather impose a constant plate velocity [Braceand Byedee,1966;$cholz,1990],havepointed far from the fault [Lyzengaet al., 1991]as donefor a long-termgeological model(Figure1). The baseof the box (low viscosity zone) is chosendeepenough(30 km) 1Nowat Universityof Edinburgh,ScotlandUK. to not influence the numerical 2Now at ENSAM, University of Bordeaux,France. Copyright1998by theAmericanGeophysical Union. results. The crustal medium is discretizedusing3-node linear finite elements. The fault zone is simulated by contact betweennodesof differentelements [Jeanand Touzot, 1988]. Time discretizationis derivedusingthe dynamic Papernumber98GL00091. 0094-8534/98/98GL-00091505.00 551 552 HUC ET AL.' EARTHQUAKE Vn=6mm/y freesurface ::- NUCLEATION faultplane ;-'.........!!g....::• \" AND STRESS EXCHANGE Vn=6mm/y event• 3 ' 1.02Opa..;-:'"'•:• ...... :•" '............... '..... .-:._ ....... -::i;? ....... -'-': IN THE CRUST event•35 events62 [1',[I,',111 IIJIIIl'11illl I[1111UlI,II[111111 I !ø ß' "'"'•I'"TIII•I ' l'l'l"l'l' ':' 10600 5oho 6 •5boo Time (years) Vn=O Figure 1. Geometry,rheological layering,andbound- Figure 2. Earthquake distribution with time. Vertiary conditions of the model. Lateral and lower sides are cal linesrepresentfault-rupture length, and dots nuclefree slip, with a fixed normal velocity V, Mechanical ation points. Time before t=4500 yrs correspondsto a model correspondsto a finite-element mesh with 3000 quasi-static evolution where the critical shear stress on triangularelements,partly drawnin the lowerrightcor- the fault is not reached. Mean recurrence time stabiner. The stress-strainrelationsfor the crust are pro- lizes around 170 yrs after t=7000 yrs. Events3, 35 and 62 are shown on Figure 3. vided by the viscoelasticmodel: o o P_3Kd,•ot where • is the viscosityof the layer, G and K are shear and bulk modulusrespectively.Valuesof respectively at the first point wherethe condition]ertl= /• er• is 4.10lø and6.66.10lø Pa areusedfor all layers.da•,and satisfied. The condition lertl= tta er,,is then usedfor d•ot are the deviatoricand volumetricpart respectively slippingnodes. As extensionproceeds,earthquakesnuof the strain-rate tensord. aa• is the deviatoricpart cleateat deeperlevelsnear the baseof the high viscosity of the stresstensor a, and P is the pressure. layer (Figure2). Maximumcoseismic motionsreach3-5 m and are damped out within the first 1-2 km in the low viscositylayer (Figure3). All nucleationpointsare relaxationmethod [Cundalland Board,1988],that al- contained in the high-viscositylayer. Between seismic lows us to compute the velocity field for both inertial events,the fault is lockedbecauselertl</• er,•,and ex(coseismic) and quasi-static(inte. rseismic)phasesin a tension only results in a bulk viscoelasticstrain in the time continuum [Huc, 1997]. The time step At is ad- crust. Accumulated topography is similar to that ex- justedfromshort(10-2 s) to long(107s) timescales. pectedfor an half-graben, with an uplifted footwall and Earthquakeoccurrenceis controlledby the friction law a subsidinghangingwall. Coseismic motionof the 35th on the fault plane. As long as I•tl < • • every- event (Figure 2 and 3) is typicalof largeearthquakes where on the fault, relative fault motion is locked and the model behaves in a quasi-static mode. As stress in- creasesin the continuummedium,]ertlreaches/• er,at generated by our model: nucleation at depth, downward rupture which dampsin the middle crust, upward rupture that breaks the free surface. Slip envelopesof some point on the fault. The friction coefficient then Figure 3 result from a high-velocitymotion according switches to tta for this point, that causesearthquake to the elastodynamic equation of motion. Becausethe triggering and propagationwith respectto the friction total time of coseismicmotion is less than 10 seconds, law. The velocity (v) and displacement(u) computa- viscosity does not play a role, and coseismicdeforma- tions involvegravity forces,(F•), internal forces(Fi) t.ion is similar to an and contactforceson the fault (F•), as follows,for one viscosity medium. degreeof freedom: vt-[-1/2= vt--1/2q_[Feq_/7/q_Fcq_Fd] Ut+l At elastic dislocation even in the low The most surprising property of this mechanicalmodel is that earthquakesspontaneouslynucleate at depth, as for most large natural earthquakes. It is unexpectedbe- • ut q-vt+l/2At Event 35 Event 3 I .... 0 I Event 62 I where mi is the mass of the node i. The damping force Fa is added only during the quasi-static phase. The model is initially in isostatic equilibrium with no deviatoric stress. We apply a constant velocity of ex- tensionof 1.2 cm/yr betweenthe lateral sidesover a period of 15000 yrs. 3. Seismic cycle and stress change In the model, small earthquakesoccur at the begin- ning of extension(Figure 2), with a mean dislocation length of 4 km and a coseismicmotion of 1 m. These eventsnucleate under the top of the high viscositylayer 16 2 3 4 5 0 I 2 3 4 5 0 1 2 3 4 5 Coseismic slip (m) Figure 3. Coseismicslip occurring during inertial phase of modeling. left: Small event nucleation at the top of the high-viscositylayer. middle and right: Nucleation of large events mainly occursat depth near zone of viscosity drop. Rupture breaks the whole crust above and is rapidly damped below, within the low-viscosity layer. HUC ET AL.- EARTHQUAKE NUCLEATION MPa 150 120 90 60 30 -30 0 MPa 5 3 1 ---1 __i_3 .5 MPa 5 AND STRESS EXCHANGE lution mirrors IN THE CRUST coseismic evolution: 553 the middle crust is unloaded while the upper crust is reloaded in the faultplane vicinity. While coseismicstresschangeis sudden, postseismicstresschangein the middle crust is slowand is controlledby the relaxation time of the low viscosity layerthat scales with shear,,iscositv modulus ' Thisratiois about 2.5 109s (m 80 yrs) for our experiment.Althoughthe true time scale is probably dependent on geometric effects, one can predict that stress release of the middle crust is almost complete when the next event occurs with a mean recurrence time of 170 years. Because the relaxation time of the high-viscositylayer is 1000 tinms higher(80000years)than that of the middlecrust,the 3 upper crust respondselastically during the post-seismic 1 stage. ---1 5 We found by modelingthat stressdecreasein the middle crust is the cause of a rapid stress increase at the baseof the uppercrust(Figure4c). Fault-planereloadFigure 4. a: Deviatoric stressintensity beforethe ing during the postseismicphase is therefore largely 35tn event. High stressis mainlyconfinedin the high depth dependent because the main stresssource comes viscositylayer. b: Coseismic stresschange(10 s) asso- from below. This modeling result is well illustrated by ciatedto the35tnevent(between tl andt2 onfigure5). the transient evolution of • with time for three points Stressrelease(light areas)mainly occursin the upper on the fault at increasing depths(10.9, 12.0,and 13.2 crust. Stressincrease(dark areas)mainlyoccursin the kin, Figure 5). Becausethe stressincreaseof the shalmiddle crust. Labelled points p, q and r refer to stress lowerpoint (10.9 kin) is lowerthan the stressincrease evolutionon figure 5. c: Postseismic stresschange(53 of the deeperpoint of the seismogenic layer, the next yrs)following the35tnevent(between t2 andt3 onfigure earthquake occurs preferentially at a greater depth. Al5). Oppositeevolutionof the coseismic stresschangeis a variabilityin coseismic slip(comrecorded. Stress rate is controlled by relaxation time thoughweobserve of the middle crust. Fault reloading in the upper crust pare event 35 and 62, Figure 3) which correspondsto appears as very inhomogeneous. variable coseismicstressrelease,a rapid post-seismic stressincreaseat the baseof the seismogenic layer is a persistentfeattire of our model. Consequently, we cause1) the increasein normalstresswith depth pre- claim that, in our modeling,the stresstransfermechavents shear stressfrom increasingto its critical value nism,causedby a viscositydropat middlecrustallevels, Fs '•,,, and 2) constantpropertiesof the friction law favorsthe nucleationof deep earthquakes. with depth do not favor nucleationat any givendepth. In order to explain this phenomenon,we explorehow 4. Summary and Discussion stresschangesduring the 35tn event and the followAlthoughthe rheologicallayeringof our modelis siming interseismicphase. Figure 4 showssnapshotsof ilar to that usedin previousworks[e.g. Li and Rice, deviatoricstressintensitybeforethe earthquake(4a), 1987], we usea differentviewpointto setupboundary and changesrelatedto coseismic (4b) and postseismic conditions. As for a long term geodynamicalexper(4c) changes.As alreadystated, coseismicmotionon the normal fault is similar to a dislocation in an elastic iment, we assumethat only relative plate velocityis medium. Relative motion across the fault modifies the deviatoricstressnorm in the fault vicinity, mainly above andbelowthe nucleation point(Figure4b). Stressis releasedin the seismogenic layer, mainly in the footwall, and stressis gained in the low viscositylayer around the fault tip. Maxinmm stresschangeis about 50 Mpa. The far field stresschange is larger than what is expected for a finite dislocation, due to the 2D nature of our model. The zoneaffectedby k 5 MPa stresschange extends up to 10 km around fault. Coseismicstressincreasecorrespondsto a potential energy stored in the low-viscositymiddle crust. Becausethis layer is viscoelastic,this energystorageis only transient. Figure 4c showshow stressis redistributedduring the post- 0.3 0.0 tr 0 a) 100 gd 200 300 0 100 200 300 b) Time (years) 0 100 200 300 c) Figure 5. • evolutionon the fault after the 35tn event,for the three pointsshownon Figure 4. Depths of points p, q and r are 10.9, 12, and 13.2 km. Most rapid post-seismic stressloadingoccurson point q at the viscositydrop,corresponding to stressunloadingat seismicstage,53 yrs after the 35th event. This evo- point r in the middle crust below. 554 HUC ET AL.: EARTHQUAKE NUCLEATION AND STRESSEXCHANGE IN THE CRUST relevant.We thinkthat prescribing staticor kinematic References boundary conditions nearthefaultis misleading beW.F., and J.D. Byerlee,Stick-slipas a mechanism cause 1)these conditions arebasically unknown 2)such Brace, for earthquakes,Science,153, 990-991,1966. a choiceconstrains stressandslip historyaroundthe Brace,W.F. and D.L. Kohlstedt,Limitson lithospheric fault to be consistent with theseboundary conditions. stressimposedby laboratoryexperiments,J. Geophys. Weratherimpose a more"neutral"boundary condition Res., 85, 6248-6252, 1980. program that shouldcorrespond to the general hypothesis that Cundall,P.A., and M. Board,A microcomputer lowdeviatoric stress is always present at depthforgeologicalstrainrates.This hypothesis is consistent with for modellinglargestrainplasticityproblems,in Numeri- calmethods in Geomechanics, editedby G. Swoboda, pp. 2101-2108,Balkema, Rotterdam, 1988. theuseof linearviscosities lowerthan102•Pa.s, ac- Das, S., and C.H. Scholz,Why largeearthquakes do not cordingto viscousrheologyof quartz-rich-rocks. nucleateat shallowdepths,Nature,305,621-623,1983. par la mdthodedes Usinga constantvelocityof extension,linearvis- Huc, M., Moddlisationdu cyclesismique coelastic laws,andthesimplest frictionlaw dldmentsfinis, 192 pp., Th;esede Doctorat,Universitdde Montpellier II, 1997. bothconstants), wegenerate a seismic cyclethatisnei- Jean,M., and G. Touzot, Implementationof unilateralcontherperiodic, norchaotic.Asfor naturalearthquakes tactanddryfrictionin computercodesdealingwithlarge [Dasand$cholz,1983],almostall largeevents nucleate deformation problems,J. Mec. Theor.Appl.,7, suppl.1, 145-160, 1988. at depthnearthe baseof theviscosity transition.Anal- in greatCalysisof stresstransfermechanisms relatedto this model Li, V.C., and J.R. Rice,Crustaldeformation ifornia earthquakecycles,J. Geophys.Res., 92, 11533leadsus to the followingpropositions: 11551, 1987. 1. Deviatoricstressis storedat deptharoundthe fault Lyzenga,G.A., A. Raefsky,and S.G. Mulligan,Modelsof tipin anelastic manner during coseismic motion. Then, recurrent strike-slip earthquakecyclesand the state of thisstress isprogressively released in themiddle crust, crustalstress,J. Geophys.Res.,96, 21623-21640,1991. providing a newstressbuild-upat the baseof upper Meissner,R., and J. Strehlau, Limits of stressesin continentalcrustsand their relationsto the depth-frequency crust.Originally proposed by Nut andMavko[1974], distributionof shallowearthquakes,Tectonics,1, 73-89, this stresstransfereffectcouldbe a generalmecha- 1982. viscoelastic rebound, nismfor clustering and aftershocks, as alreadysug- Nur, A., and G. Mavko,Postseismic Science, 183, 204-206, 1974. gested[$ibson, 1982;Unruhet al., 1996;Rigoet al., Rigo, A., H. Lyon-Caen,R. Armijo, A. Deschamps, D. Hatzfeld, K. Makropoulos,P. Papadimitriou,I. and Kas2. The fault-loadingrate after an earthquake is theresaras,A microseismic study in the westernpart of the forea combination ofa long-term stress loading ratedue to relative plate velocity,and a short term stressloading rate due to stressrelaxationin the middlecrust. GulfofCorinth(Greece): implications forlarge-scale normalfaultingmechanisms, Geophys. J. Int., 126,663-688, 1996. C.H., Thebrittle-plastic transition andthedepth Thisshorttermrateis highnearthefaulttip, andis Scholz, seismic faulting, Geologische Rundschau, 77,319-328, thecause of inhomogeneous faultloading, asalready of 1988. shown by Lyzenga et al. [1991]forantiplane faults.We Scholz, C.H., Themechanics of earthquakes andfaulting, suggest that thisstress-rate partitioning isthephysical 439 pp., Cambridge,Universitypress,1990. R.H., Faultzonemodel,heatflow,andthedepth causefor thegeneration of largeearthquakes, whatever Sibson, the slip direction. distributionof earthquakesin the continentalcrustof the unitedstates,Bull.Seismol. Soc.Am., 72,151-163, 1982. 3. If we acceptthat viscosityrapidlydecreases with Stesky,R. M., The mechanicalbehaviorof faultedrock at hightemperature andpressure, Ph.D.Thesis,275pp., depthnearthe300- 3500C isotherm,/•d _</•sisonly requiredin the high-viscosity zone.The frictionlaw at Mass.Inst. of Technol.,Cambridge, Mass.,1975. greaterdepthis not of primaryimportance because the Unruh,J.R.,Twiss,R.J.,andHauksson, E., Seismogenic deformation in theMojaveblockandimplications fortecshearstresson fault planesneverreachesthe critical tonics of the eastern California shear zone, J. Geophys. value/• .a•. Thisinterpretation doesnot preclude slip-hardening law at hightemperature, but ratherindicatesthat thefrictionlawisnotnecessarily thecause for seismicitycutoffat depthas is oftenstated. Res., 101, 8335-8361,1996. M. HucandJ. Chdry,LaboGdophysique et Tectonique, 4 pl. Bataillon,34095,Universitd MontpellierII, France. (e-mail:[email protected]) R. Hassani, LEPT,EcoleNationale desArtset Metiers, Esplanade des Arts et Metiers, 33405 Talence Cedex. (email: We are gratefulto M. Jeanwhoprovidedus his contact hassani @lept-ensam. u-bordeaux.fr) Acknowledgments. algorithm to compute frictiononfaults.Thisworkwassupportedin 1995by INSU-CNRSthrougha PNRN granton (Received June26,1997;revised December 1, 1997; seismic risk. accepted December18, 1997.)
© Copyright 2026 Paperzz