Journal of the Mechanics and Physics of Solids 112 (2018) 50–65
ContentslistsavailableatScienceDirect
Journal of the Mechanics and Physics of Solids
journalhomepage:www.elsevier.com/locate/jmps
A 2-D model for friction of complex anisotropic surfaces
Gianluca Costagliolaa, Federico Bosiaa, Nicola M.Pugnob,c,d,∗
a Department of Physics and Nanostructured Interfaces and Surfaces Centre, University of Torino, Via Pietro Giuria 1, Torino, 10125, Italy
b Laboratory of Bio-Inspired & Graphene Nanomechanics, Department of Civil, Environmental and Mechanical Engineering, University of Trento, Via Mesiano, 77, Trento 38123, Italy
c School of Engineering and Materials Science, Queen Mary University of London, Mile End Road, London E1 4NS, UK
d Ket Labs, Edoardo Amaldi Foundation, Italian Space Agency, Via del Politecnico snc, Rome 00133, Italy
a rt i c l e i n f o
Article history:
Received 24 July 2017 Revised 7 November 2017 Accepted 16 November 2017 Available online 20 November 2017 Keywords:
Friction Numerical models Microstructures Anisotropic materials
a b s t ra c t
The friction force observedat macroscale is the result of interactionsat various lower lengthscalesthataredifficulttomodelinacombinedmanner.Forthisreason,simplified approaches arerequired,dependingonthe specificaspectto beinvestigated.In partic- ular,thedimensionalityofthesystemisoftenreduced,especiallyinmodelsdesignedto provideaqualitativedescriptionoffrictionalpropertiesofelasticmaterials,e.g.thespring- blockmodel.Inthispaper,weimplementforthefirsttimeatwodimensionalextension ofthespring-blockmodel,applyingittostructuredsurfacesandinvestigatingbymeansof numericalsimulationsthefrictionalbehaviourofasurfaceinthepresenceoffeatureslike cavities,pillarsorcomplexanisotropicstructures.Weshowhowfrictioncanbeeffectively tunedbyappropriatedesignofsuchsurfacefeatures.
© 2017ElsevierLtd.Allrightsreserved.
1. Introduction
Thefrictionalbehaviorofmacroscopicbodiesarisesfromvarioustypesofinteractionsoccurringatdifferentlengthscales betweencontactsurfacesinrelative motion.Whileitisclearthat theirultimateoriginliesininter-atomicforces,itisdif- ficult to scale theseup tothe macroscopic levelandto includeother aspects such asdependenceon surfaceroughness, elasticityorplasticity,wearandspecificsurfacestructures (NosonovskyandBhushan,2007;Persson,2000).Moreover,the dependenceon“externalparameters”,e.g.relative slidingvelocity ofthesurfacesandnormalpressure,isneglectedinap- proximatemodelssuchasthefundamentalAmontons–Coulomblaw,andviolationsofthelatterhavebeenobserved(Deng etal.,2012;Katanoetal.,2014).
Forthesereasons,simplifiedmodelsarerequiredintheoreticalstudiesandnumericalsimulations,andfrictionproblems canbe addressedindifferentwaysdependingon thespecificaspectsunderconsideration.Inordertoimprovetheoretical knowledgeoffriction,ortodesignpracticalapplications,itisnotnecessarytosimulateallphenomenasimultaneously,anda reductionistapproachcanbeusefultoinvestigateindividualissues.Thus,despitetheimprovementincomputational tools, inmostcases itis still preferableto developsimplified models todescribe specific aspects,aiming to providequalitative understandingofthefundamentalphysicalmechanismsinvolved.
One ofthe mostusedapproaches todeal withfriction of elasticbodies consistsinthe discretization ofa materialin springsandmasses,asdonee.g.intheFrenkel–Kontorovamodel(BraunandKivshar,2004),ortheBurridge–Knopoff model
∗ Corresponding author.
E-mail addresses: [email protected] (G. Costagliola), [email protected] (F. Bosia), [email protected] (N.M. Pugno).
https://doi.org/10.1016/j.jmps.2017.11.015 0022-5096/© 2017 Elsevier Ltd. All rights reserved.
G. Costagliola et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 50–65 51
(BurridgeandKnopoff,1967), thelatteralsoknownasthespring-blockmodel.Forsimplicity,thesemodels areoftenfor- mulatedinonedimension alongthesliding direction,invarious versionsdependingonthe specificapplication. Inrecent years,interesting resultshave beenobtained withthese models, explainingexperimental observations (Amundsen etal., 2012;Bouchbinderetal.,2011;Maegawaetal.,2010;Rubinsteinetal.,2004;ScheibertandDysthe,2010;Trømborgetal., 2011).Theextensiontotwodimensionsisthestraightforwardimprovementtobetterdescribeexperimental resultsandto correctlyreproducephenomenaoccurringintwodimensions.Thishasalreadybeendoneforsomesystems,liketheFrenkel–
Kontorovamodel(Mandellietal.,2015;Norelletal.,2016)andthespring-blockmodelappliedtogeology(Andersen,1994;
Brownetal.,1991;Giaccoetal., 2014;Mori andKawamura,2008a;Mori andKawamura,2008b; Olamietal., 1992),but muchworkremainstobedonetodescribethefrictionofcomplexandstructuredsurfaces.
Theinterestofthisstudyliesnotonlyinthenumericalmodelingoffrictioninitself,butalsointhepracticalaspectsthat canbeaddressed:therearemanystudies relativetobio-inspiredmaterials(Baumetal.,2014;Lietal.,2016;Murarashet al.,2011;Yurdumakanetal.,2005)orbiologicalmaterials(Autumnetal.,2000;Labonteetal.,2014;StempfleandBrendle, 2006; Stempfle et al., 2009; Varenberg et al., 2010) that involve non-trivial geometries that cannot be reducedto one- dimensionalstructuresinordertobecorrectlymodeled.
Theonedimensionalspring-blockmodelwasoriginallyintroducedtostudyearthquakes(CarlsonandLanger,1989;Carl- son etal., 1994; Xiaet al., 2005) andhasalso been usedto investigatemany aspects ofdryfriction of elasticmaterials (Amundsen et al., 2015; Braun etal., 2009; Capozza and Pugno, 2015; Capozza et al., 2011; Capozza and Urbakh, 2012;
Pugnoetal.,2013;Trømborgetal.,2015).InCostagliolaetal.(2016),wehaveextensivelyinvestigatedthegeneralbehav- iorofthemodelandtheeffectsoflocalpatterning(regularandhierarchical)onthemacroscopicfriction coefficients,and inCostagliolaetal.(2017a) wehaveextendedthestudytocompositesurfaces,i.e.surfaceswithvaryingmaterialstiffness androughness;finallyinCostagliolaetal.(2017b)wehaveintroducedthemultiscaleextension ofthemodeltostudythe statisticaleffectsofsurfaceroughnessacrosslengthscales.
Inthis paper,we propose a 2-D extension of the spring-block modelto describe the frictional behavior ofan elastic materialslidingona rigidsubstrate.Ourmain aimistocomparetheresultswiththoseobtainedinthe one-dimensional caseandto extendthe studytomorecomplexsurfacestructures,e.g. isotropicoranisotropic arrangementsofcavities or pillars,simulatingthosefoundinbiological materials.Thetwo-dimensional spring-blockmodelallowstoconsidera more realisticcasesandcapturesavarietyofbehaviorsthatcanbeinterestingforpracticalapplications.
Thepaperisorganizedasfollows:inSection 2,wepresentthemodel;inSection 3.1,wediscussthemaindifferences withtheone-dimensionalcaseandweexploretheroleoftheparameterswithoutsurfacestructures,highlightingthephe- nomenologyofthe model;inSection3.2,wepresenttheresultsforstandard1-D and2-D surfacestructureslikegrooves andcavities;inSections3.3and3.4,weconsidermorecomplexcasesofanisotropicsurfacepatterning;finally,inSection4, conclusionsandfuturedevelopmentsarediscussed.
2. Model
Theequationofmotionforanisotropiclinearelasticbodydrivenby aslideronaninfinitelyrigidplane withdamping andfriction can be written as:ρu¨=μ∇2u+(λ+μ)∇(∇· u)−γ ρu˙, where uis the displacementvector, ρ isthe mass
density,γ isthedampingfrequency,λ,μaretheLamé constants.Thefollowingboundaryconditionsmustbeimposed:the topsurfaceofthebodyisdrivenatconstantvelocityv,thebottomsurfaceissubjectedtoaspatially variablelocalfriction force, whichwe discussbelow, representingthe surfaceinteractions betweentheelastic body andthe rigidplane, while freeboundaryconditionsaresetontheremainingsides.
Inordertosimulatethissystem,weextendthespring-blockmodeltothetwo-dimensionalcase:thecontactsurfaceis discretizedintoelementsofmassm,eachconnectedbyspringstotheeightfirstneighborsandarrangedinaregularsquare mesh(Fig. 1)withNx contactpointsalong thex-axisandNy contactpointsalongthe y-axis. Hence,the totalnumberof blocksisNb≡NxNy.Thedistancesontheaxisbetweentheblocksare,respectively,lxandly.Themeshadoptedinprevious studiesofthe2-Dspring-block model,e.g. Olamietal.(1992)andGiaccoetal.(2014),doesnotincludediagonalsprings, butweaddthemtoaccountforthePoissoneffect(ourmeshinsimilartothatusedinTrømborgetal.(2011)).
Inordertoobtaintheequivalenceofthisspring-masssystemwithahomogeneous elasticmaterialofYoung’smodulus E,thePoisson’sratiomustbefixedtoν=1/3(AbsiandPrager,1975),whichcorrespondstotheplanestresscase,lx=ly≡ l andKint=3/8Elz, where lz is the thickness ofthe 2-D layer andKint is the stiffness of the springs connectingthe four nearestneighborsofeachblock,i.e.thosealignedwiththeaxis.ThestiffnessofthediagonalspringsmustbeKint/2.Hence, the internal elastic force on the block i exerted by the neighbor j is F(inti j)=ki j(ri j− li j)(rj− ri)/ri j, whereri, rj are the positionvectorsofthetwoblocks,rijisthemodulusoftheirdistance,lijisthemodulusoftheirrestdistanceandkij isthe stiffnessofthespringconnectingthem.
All the blocks are connected, through springs of stiffness Ks, to the slider that is moving at constant velocity v in the xdirection, i.e. theslider vector velocity is v=(v,0). Giventhe initial restposition r0i of block i,the shear force is F(si)=Ks(vt+r0i− ri). Wedefine thetotaldriving forceoni asF(moti) =
jF(inti j)+F(si).ThestiffnessKs canbe relatedtothe macroscopicshearmodulusG=3/8E,sincealltheshearspringsareattachedinparallel,sothatbysimplecalculationswe obtainKs=Kintl2/lz2.Inthefollowing,forsimplicitywefixlz=l.Thisformulation,commonlyusedinspring-blockmodels, neglectsthelong-rangeinteractions thatmayarisefromwavepropagationthroughthe bulk(Elbanna 2011;Hulikaletal.,
52 G. Costagliola et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 50–65
Fig. 1. Discretization of a square surface into a 2-D spring-bock model, showing the mesh of the internal springs. The shear springs K s modeling the block interaction with the slider are not shown.
2015;Rice,1993;RiceandRuina,1983).Here,wesupposethatthelocalinteractionsaredominating,whichisareasonable assumptionforslowslidingvelocitiestypicaloftheexperimentswe useasbenchmarks(Baumetal.,2014;Lietal., 2016;
Murarashetal., 2011).Thisassumptionhasalreadyallowed toobtaincorrectdescriptions ofthephenomenaoccurringat thetransitionfromstatictodynamicfriction(Braunetal.,2009;Trømborgetal.,2015).
Theinteractions betweentheblocksandtherigidplanecanbe introducedinmanyways:intheoriginal paperonthe springblock model(BurridgeandKnopoff,1967) andinearthquakerelatedpapers,e.g. Olamietal.(1992)andMori and Kawamura(2008a), itisintroduced by meansofan effectivevelocity-dependent force(Dieterich,1979;Rice etal.,2001), infrictionstudies,e.g.Amundsenetal.(2012)andBraunetal.(2009),byspringsthatattachanddetachduringmotion,in Trømborgetal.(2011)andGiaccoetal.(2014)bymeansoftheclassicalAmontons–Coulomb(AC)frictionforce.Thesevari- ousapproachesgiverisetoslightlydifferentquantitativeresults,butiftheyareimplementedunderreasonableassumptions, theydonotsignificantlyaffecttheoverallpredictionsofthemodel,whichisthoughttoprovideaqualitativeunderstanding ofthebasicmechanismsoffriction.Thisistrueatleastforthesmallslidingvelocitiesweareconsideringcomparedtothe characteristicvelocityscalesofthesystem,i.e.l
Kint/m.Adifferentqualitativebehaviormayariseforhigherslidingveloc- ities,asshownforrate-and-state frictionlaws (Elbanna andHeaton,2012;Heaton,1990; ZhengandRice,1998).In these cases,acarefulevaluationoftheinterplaybetweenthefrictionlawandslidingvelocityofthesystemmustbeperformed.
Inthisstudy,weadoptaspring-blockmodelbasedontheACfrictionforce andastatisticaldistributiononthefriction coefficients (Costagliola et al., 2016; 2017a;2017b): while the block i is at rest,the friction force F(f ri) is opposedto the total drivingforce, i.e.F(f ri)=−F(moti),up to a thresholdvalue Ff r(i)=μsiFn(i),where μsi isthe staticfriction coefficient and Fn(i) isthenormalforceactingoni.Whenthislimitisexceeded,aconstantdynamicfrictionforceopposesthemotion,i.e.
F(f ri)=−μdiFn(i)r˙i,whereμdiisthestaticfrictioncoefficientandr˙iisthevelocitydirectionoftheblock.Inthefollowing,we willdropthesubscriptss,deverytimetheconsiderationsapplytoboththecoefficients.
Thefrictioncoefficientsareextractedfroma Gaussianstatisticaldistributiontoaccountfortherandomnessofthesur- faceasperities,i.e.p(μi)=(√2πσ)−1exp[−(μi−(μ)m)2/(2σ2)],where(μ)mdenotesthemeanofthemicroscopicfriction coefficientsandσ isitsstandarddeviation.Inordertosimulatethepresenceofpatterningorofstructuresonthesurface,
wesettozerothefrictioncoefficientsoftheblockslocatedonzonesdetachedfromtherigidplane.Themicroscopicstatic anddynamicfrictioncoefficientsarefixed conventionallyto (μs)m=1.0(1)and(μd)m=0.50(5),respectively,wherethe numbersinbracketsdenotethestandarddeviationsoftheirGaussiandistributions.
The macroscopic friction coefficientsare denoted with (μ)M. The staticfriction coefficient is calculated fromthe first maximum ofthetotal friction force, whilethe dynamicone asthetime average over thekinetic phase. Tocalculatethe frictioncoefficientsasratiobetweenlongitudinalforceandnormalforce,thenormofthelongitudinalforcevectormustbe calculated.When calculatingtime averages,care mustbe takenintheorderofthe operations:ifthereisan inversion of thefrictionforce(i.e.someblocksexceedtherestposition,asintheanalyticalcalculationsofCostagliolaetal.(2016))ora periodicmotiontakesplace,switchingthe operationsofnormandtimeaverageproducesdifferentresults.Inthesecases, the calculation closerto the realistic experimental procedure must be adopted. However, in the results below, we have checkedthatthe aboveconditionsdonotoccur andtheorderoftheoperationsisirrelevant.The modeldoesnot include roughnessvariationsduringslidingorotherlongtermeffects,sothattheresultsfordynamicfrictionaretobeconsidered withinthelimitsofthisapproximation.
Adampingforceisaddedto eliminateartificialblockoscillations: inBraunetal.(2009)andinthepapersbasedonit (e.g.CapozzaandPugno,2015)thisisdonebymeansofaviscousdampingforceproportionaltothevelocityoftheblock, i.e.F(di)=−γmr˙i.Anotheroption,asinthe2-DmodelinTrømborgetal.(2011),istoimposethedampingontheoscillations betweeneachpairofblocksiandj,i.e.F(di j)=−mγ (r˙i− ˙rj),thusemulatingthedescriptionusuallyadoptedforviscoelastic
G. Costagliola et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 50–65 53
Fig. 2. Time evolution for the total friction force and percentage of moving blocks for N = 20 , pressure P = 0 . 1 MPa, velocity v = 0 . 1 cm/s, γ/ω = 0 . 1 (a) or γ/ω = 0 . 5 (b), where ω is the internal frequency ω ≡
K int /m . The other parameters are set to the default values. Greater damping enhances the dynamic friction coefficient and reduces stick-slip oscillations.
materials(Banksetal.,2011;CarboneandPutignano,2013;Hunter,1961;Krushynskaetal.,2016;Persson, 2001;Tschoegl 1989).InSection 3.1,wediscussthe differentbehaviorobtainedwiththetwo approaches,butinthe followingweadopt theformerone,whichisthesimplesttoallowdampingofnon-physicaloscillations.
Thus,thecompleteequation ofmotionforblock i is:mr¨i=
jF(inti j)+F(si)+F(f ri)+F(di).The overallsystemofdifferential equationsis solved using a fourth-order Runge–Kuttaalgorithm. Inorder to calculatethe average ofany observable, the simulationmust beiterated, extractingeachtime newrandom friction coefficients. Inrepeatedtests, an integrationtime steph=10−8sprovestobesufficienttoreduceintegrationerrorsunderthestatisticaluncertaintyintherangeadoptedfor theparametersofthesystem.
Weconsideronly asquare mesh,i.e.Nx=Ny≡ N,andwe willspecifythenumberofblocks foreach considered case.
ThedefaultnormalpressureisP=0.05MPa,sothatthenormalforceoneachblockisFn(i)=Pl2 andthetotalnormalforce isFn=Pl2N2.Theslidervelocity isv=0.05cm/s.We willdiscussinSection3.1the motivationsforthesechoices, butin anycasetheresultsdisplaysmalldependenceontheseparameters.
Realisticmacroscopicelasticpropertiesarechosen,e.g.aYoung’smodulusE=10MPa,whichittypicalforasoftpolymer orrubber-like materialanda densityρ=1.2g/cm3. Thedistancebetweenblocks lin themodelis anarbitraryparame- ter representingthe smallestsurfacefeature that can be takeninto account andis chosen by defaultasl=10−3 cm,so thatthe orderofmagnitudematchesthose typicalofsurfacestructures usedin experiments(Baumetal., 2014;Lietal., 2016).
3. Results
3.1. Non-patternedsurface
Inthissection,wemodelfrictionproblemsrelativetohomogeneous,non-patternedsurfacesvaryingthefundamentalpa- rameterstounderstandtheoverallbehaviorandtocompareitwiththatofthe1-DmodelstudiedinCostagliolaetal.(2016). InFig.2,thefriction forcebehavior asa functionoftime isshownwiththedefaultsetofparameters:thereisalinearly growingstaticphase upto themacroscopicruptureevent, followedbya dynamicphaseinwhich thesystemslideswith smallstick-slip oscillationsatconstantvelocityv.Thepercentageofblocksinmotionasafunctionoftimeisalsoshown:
inthekinetic phase,singleblocksorsmallgroupsslipsimultaneouslybutnot ina synchronizedmannerwithrespectthe restofthesurface.
Thefirstdifferencewiththe1-D modelisthatthe2D arrayofspringsshowninFig.1allowstocorrectlysimulatethe Poissoneffect,i.e.adeformationinthetransversaldirectionduetothestretchinginthelongitudinalone.Secondly,dueto themodeldefinitionexplainedinSection2,thestiffnessesdonotdependonthetotalnumberofblocks,sothatincreasing Ndoesnot modifytheelasticproperties,butonlythesizeofthesystem.Sincethenumberofpointsgrowsasthesquare ofthe lateral size, N࣡100 can already be considered a large system, as shownin Fig. 3, where the size effects on the globalstaticfrictioncoefficientareshown.Similarresultsholdforthedynamicfriction.Intheleft(Fig.3a)andrightpanels (Fig.3b),theinfluenceoftheappliedpressurePandtheslidervelocityvisalsoshown,respectively.Inthetypicalrangesof