Comptes Rendus Mecanique...MWCNT PC

Embed Size (px)

Citation preview

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    1/26

    C. R. Mecanique 343 (2015) 371396

    Contents lists available at ScienceDirect

    Comptes Rendus Mecanique

    www.sciencedirect.com

    Thermo-mechanical characterization of multi-walled carbonnanotube

    reinforced

    polycarbonate

    composites:

    A molecular

    dynamics

    approach

    Sumit Sharma a,,1,

    Rakesh Chandra b,2,

    Pramod Kumar b,3,

    Navin Kumar c,2

    a SchoolofMechanicalEngineering,LovelyProfessionalUniversity,Phagwara,Indiab DepartmentofMechanicalEngineering,Dr.B.R.AmbedkarNationalInstituteofTechnology,Jalandhar,Indiac SchoolofMechanical,Materials&EnergyEngineering(SMMEE),IndianInstituteofTechnology,Ropar,India

    a r t i c l e i n f o a b s t r a c t

    Articlehistory:

    Received31January2015Accepted11March2015Availableonline23April2015

    Keywords:

    CarbonnanotubeDampingMechanicalpropertiesMoleculardynamicsPolycarbonate

    Thermal

    conductivity

    The

    present

    study

    aims

    at

    examining

    the

    mechanical

    properties

    of

    multi-walled

    carbon

    nanotubespolycarbonatecomposites (MWCNTPC), throughamoleculardynamics (MD)simulation.CompositesofMWCNTPCweremodeledusingMaterialsStudio 5.5software.Multiwallcarbonnanotubes (MWCNTs) compositions inpolycarbonate (PC)werevariedby

    weight

    from

    0.5%

    to

    10%

    and

    also

    by

    volume

    from

    2%

    to

    16%.

    Forcite

    module

    in

    MaterialsStudiowasused forfindingmechanicalproperties.Amarked increase in theelasticmodulus(upto89%)hasbeenobserved,evenwiththeadditionofasmallquantity(upto2weight %)ofMWCNTs.Also,uponadditionofabout2volume %ofMWCNTs,theelastic

    modulus

    increases

    by

    almost

    10%.

    The

    increase

    in

    mechanical

    properties

    is

    found

    to

    supplement

    earlier

    experimental

    investigations

    of

    these

    composites

    using

    nano-indentation

    techniques.

    Better

    load

    transfer

    property

    of

    MWCNTs,

    larger

    surface

    area

    and

    interactionbetween reinforcementwithbasematrixare thesuggested reasons for this increase in

    mechanical

    properties.2015Acadmiedessciences.PublishedbyElsevierMassonSAS.All rights reserved.

    1. Introduction

    Carbonnanotubesareexcellentreinforcements forpolymersbecauseof theiruniquemechanicalpropertiesand largesurfaceareaperunitvolume.Experimentsandcalculationsshowthatnanotubeshaveamodulusequaltoorgreaterthanthebestgraphitefibers,andstrengthsatleastanorderofmagnitudehigherthantypicalgraphitefibers.Forexample,themeasurementofthe tensilepropertiesof individualmulti-walledcarbonnanotubes(MWCNTs)gavevaluesof1163 GPafor

    the

    tensile

    strength

    and

    270950 GPa

    for

    Youngs

    modulus,

    as

    obtained

    by

    Yu

    et al. [1].

    For

    comparison,

    the

    modulus

    andstrengthofgraphitefibersare300800and5 GPa,respectively. Inaddition to theiroutstandingmechanicalproper-ties, the surfaceareaperunitvolumeofnanotubes ismuch larger than thatofembeddedgraphitefibers.Forexample,30-nm-diameternanotubeshave150timesmoresurfaceareathan5-m-diameterfibers forthesamefillervolume frac-tion,suchthatthenanotube/matrixinterfacialareaismuchlargerthanthatintraditionalfiber-reinforcedcomposites.The

    * Correspondingauthor.Tel.:+918146871758.E-mailaddress:[email protected](S. Sharma).

    1 Assistant Professor.2 Professor.3 AssociateProfessor.

    http://dx.doi.org/10.1016/j.crme.2015.03.002

    1631-0721/ 2015

    Acadmie

    des

    sciences.

    Published

    by

    Elsevier

    Masson

    SAS.

    All rights reserved.

    http://dx.doi.org/10.1016/j.crme.2015.03.002http://www.sciencedirect.com/http://www.sciencedirect.com/mailto:[email protected]://dx.doi.org/10.1016/j.crme.2015.03.002http://crossmark.crossref.org/dialog/?doi=10.1016/j.crme.2015.03.002&domain=pdfhttp://dx.doi.org/10.1016/j.crme.2015.03.002mailto:[email protected]://www.sciencedirect.com/http://www.sciencedirect.com/http://dx.doi.org/10.1016/j.crme.2015.03.002
  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    2/26

    372 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    unusualmechanicalstrengthofthecarbonnanotubeshasmotivatedscientiststofabricateandmodifyotherusefulmateri-alswhicharecheaplyavailableinbulkform,bycombiningthemascompositeswithcarbonnanotubes.Polycarbonate(PC)isa lightweightpolymerthat isavailable inbulk formand iswidelyused forseveralengineeringapplicationsdueto itsmoldability.Fortakingadvantageoftheusefulpropertiesofpolymersincombinationwithuniquestructuralpropertiesofcarbonnanotubes,multi-walledcarbonnanotubespolymercompositeshavebeenresearchedandfabricatedoverthepastfewyears.Inordertoexploittheusefulnessofthesecompositesforspecificmechanicalengineeringapplications,theirstaticanddynamicmechanicalpropertiesneedtobeevaluated.Amongthestaticproperties,theelasticmodulusofthespecimen

    is

    very

    important.AlotofworkhasbeenpublishedrelatedtotensiletestingofMWCNTPCcomposites.Thesetestshaveevidencedthat

    minorcompositions(upto2 wt. %)ofMWCNTinPCenhancethemodulusandtensilestrengthfrom10%toeven70%.Choiet al. [2] usedstyreneandacrylo-nitrile (SAN)graftedMWCNTswithPC insteadofpristineMWCNTsandobserved thatwhenSAN-graftedMWCNTs(1wt. %)wereusedwithPC,bothtensilestrengthandmodulus increasedbynearly5%and10%,respectively,incomparisontopristineMWCNTPCcomposites.Liuet al.[3] observedthatat3wt. %MWCNTsinaPC,thecompositesexhibitedanearly40%highertensilestrength incomparisontopurePC.However, fora5wt. %MWCNTcomposition, the strength reduced drastically. There are also contrary results obtained by Olek et al. [4], who reportednoimprovementinthepresenceofMWCNTsinthepolymerpoly-methyl-methacrylate(PMMA)foranystaticmechanicalproperty.Evenwith acompositionalchange from1% to 5%ofMWCNTs, theelasticmodulus remainedalmost the sameasthatofpurePMMA.However, iftheMWCNTswerecoatedwithsilica,thecompositeshowedremarkableresultsuponnano-indentation.Withonly4%MWCNTsilicainPMMA,themodulusmeasuredisaboutthreetimesthatofpurePMMA.Reinforcementonpoly-vinylalcohol(PVA)andPMMAwithfew-layergraphene(FG)wasalsotestedusinganano-indenter

    by

    Das

    et al. [5].LowcompositionsofFG(0.6%)inPVAmadethemodulusincreasebyabout20%.Vivekchandet al.[6] haveexplainedthe

    useofinorganicnanowires(NW)asreinforcementinPVAtobeasefficientasMWCNTs.Theelasticmodulusincreasedbyalmosttwotimesuponreinforcementby0.8%(involume)ofinorganicNW.However,MWCNTshaveaverysmoothsurface;makingthestrength impartedbyreinforcingwithMWCNTs lesserthanwithNW.Kimet al. [7]usedacompatibilizerastwopoly-g-polycapro-lactones (P3HT-g-PCLs)withbisphenol-A-PC-MWCNTcomposite.WhenaPC-MWCNTcompositewascombinedwithP3HT-g-PCL,thentherewasanincreaseofnearly22%intheYoungmodulusand30%inthetensilestrengthincomparisontopurePC.Forsmallconcentrations(0.10.5wt. %)ofMWCNTs,thisincreasewasfoundtobeconsistent.However,whentheMWCNTsconcentrationwasfurtherincreasedto1wt. %,thenbothYoungsmodulusandtensilestrengthwereconsiderablyreduced.

    Eitan et al. [8] used bisphenol-A-PC with MWCNTs as composites for mechanical characterization. Tensile tests wereperformedusingaUniversalTestingMachineanditwasfoundthatforcompositeswithsurface-modifiedMWCNTs(5 wt. %),themodulusimprovedby95%incomparisontopurePC.EvenforcompositesusingpristineMWCNTs(5wt. %),themodulus

    rise

    was

    nearly

    70%

    in

    comparison

    to

    pure

    PC.

    Ayatollahi

    et al. [9] have

    used

    epoxy-MWCNT

    composite

    under

    shear

    andbending loadusinga SantanUniversal TestingMachine. Theyhavealso found that there isagradual increase inelastic

    modulusandtensilestrengthasMWCNTcompositionincreasedinepoxy.Compositionsof0.1%,0.5%and1.0%MWCNTsinepoxywerefabricatedand,asthecompositionofMWCNTsincreased,bothelasticmodulusandtensilestrengthincreasedby10%. Montazeri et al. [10] usedaHounsfield machineand also evaluated the viscoelastic behavior of epoxy-MWCNTcomposite.TheyreportedthatwithfurtherincreaseinMWCNTcompositionof2%,theelasticmodulusincreasedbyabout20%ascomparedtopureepoxysamples.

    Computationalstudiescomplementexperimentbyprovidingeasymanipulation,analysisandinsightsintothemolecularlevel.Theextenttowhichmechanicalreinforcementcanbeachieveddependsonseveral factors, includinguniformityofdispersion,degreeofalignmentofCNTs,andthestrengthofpolymerCNTinterfacialbonding.Sinceitisdifficulttocontrolandmeasuremanyofthesepropertiesexperimentally,computationalmodelingcanprovidesomecrucialinsights.Forthisreason,theoreticalandcomputationalmethodshavebeenwidelyappliedtostudypolymer/CNTcomposites.

    Due to the difficulties in the experimental characterization of nanotubes, computer simulation has been regarded as

    a

    powerful

    tool

    for

    modeling

    the

    properties

    of

    nanotubes.

    Among

    the

    available

    modeling

    techniques,

    molecular

    dynamicssimulationhasbeenusedmostextensively.First-principlesmethodsareabletogeneratereasonablyaccuratedataofstruc-

    turesandenergies relevant to polymernanocomposite systems, but theycan only beused to study small systems overshorttimesduetotheircomputationalexpense. Incontrast,molecularsimulationmethodssuchasmolecularmechanics(MM)andmoleculardynamics (MD),whicharebasedonanalytic forcefields,arecomputationallycheapercompared tofirst-principlesmethods.Theycanthereforebeusedtostudylargermolecularsystemsforlongertimes.Asdescribedbelow,MDcanalsobeusedtoobtainmacroscopicpropertiessuchastheYoungmodulus.

    The present study was undertaken to investigate changes in the static mechanical properties under the influence ofvaryingcompositions(perweightandpervolume)ofpristineMWCNTsinPCbyemployingtheMDtechniquewithoutanyadditionalcomponentoranysurfacemodificationinthecomposite.Theelasticmodulushasbeenstudiedtosupplementthe experimental investigationson thesecomposites made by Kumaret al. [11].They preparedcomposites of MWCNTPCbyatwo-stepmethodofsolutionblending,followedbycompressionmolding.Multiwallcarbonnanotubes(MWCNTs)compositions inpolycarbonate (PC)werevariedbyweight from0.5% to10%.Nano-indentation techniqueswereused to

    evaluate

    mechanical

    properties

    like

    elastic

    modulus

    and

    hardness.

    A

    marked

    increase

    in

    the

    elastic

    modulus

    (up

    to

    95%)wasobservedwiththeadditionofsmallquantities(upto2wt. %)ofMWCNTs.

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    3/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 373

    Fig. 1. (Color online.) A carbonate monomer.

    Fig. 2. (Color online.) Polycarbonate chain containing ten repeat units of carbonate.

    Thesepropertieswillhelpus inthemechanicalcharacterizationofPC-basedcompositesand in theevaluationof theusefulnessofMWCNTsasreinforcementfromthepointofviewofapplications.

    2. Simulationstrategy

    OurMDsimulationswereperformedbyMaterialsstudio5.5.WeusedCondensed-phaseOptimizedMolecularPotentialsforAtomisticSimulationStudies (COMPASS) forcefield,which is implanted inMaterials studio5.5.COMPASS is thefirst

    force

    field

    that

    has

    been

    parameterized

    and

    validated

    using

    condensed

    phase

    properties

    in

    addition

    to

    empirical

    data

    for

    moleculesinisolation.Consequently,thisforcefieldenablesaccurateandsimultaneouspredictionofstructural,conforma-tional,vibrational,andthermo-physicalpropertiesforabroadrangeofmoleculesinisolationandincondensedphases.TheCOMPASSforcefieldconsistsoftermsforbonds(b),angles(),dihedrals(),out-of-planeangles( )aswellascross-terms,andtwonon-bonded functions,aCoulombic function forelectrostatic interactionsanda96LennardJonespotential forvanderWaalsinteractions.

    Etotal= Eb+ E+ E+ E +Eb,b + Eb,+ Eb, + E, + E, + E,, + Eq+ EvdW (1)

    where

    Eb=energy due to bond stretching,E=energy due to bond bending,E =energy due to bond torsion,

    E =energy due to out of plane inversion,Eq=electrostatic energy,EvdW=van der Waals energy andEb,b ,E, ,Eb,,Eb, ,E, ,E,, = crosstermsrepresentingtheenergyduetotheinteractionbetweenbondstretchbondstretch,bondbendbondbend,bondstretchbondbend,bondstretchbondtorsion,bondbendbondtorsionandbondbendbondbendbondtorsion,respectively.

    Allthesimulationsweredoneintheconstanttemperatureandconstantvolumecanonicalensemble(NVT).Theequationsof motions were integrated using the Verlet algorithm with an integration time step of 1 fs and the temperature wascontrolledbyAndersensthermostat.Thetotalsimulationtimeforthedynamicsrunwas50 ps,withatimestepof1 fs.Afterthecompletionofdynamicsrun,thestructureobtainedwasfirstsubjectedtogeometryoptimizationusingaconjugategradientalgorithmandthenastainof0.005wasappliedtoobtaintheelasticmoduli.Periodicboundaryconditionswereusedinallsimulations.

    Fig. 1and

    Fig. 2show,

    respectively,

    the

    chemical

    structure

    of

    the

    investigated

    carbonate

    and

    PC

    polymer.

    Firstly,

    a

    repeatunitwasbuiltusingthe BuildtoolinMaterialsstudio 5.5.Then,usingthesametool,thePCstructurewasobtainedby

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    4/26

    374 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Fig. 3. (Color online.) Simulation cell of pure PC containing 17,728 atoms.

    Fig. 4. (Color online.) Simulation cell of MWCNT (0.75% by weight) reinforced PC containing 20,273 atoms.

    Fig. 5. (Color online.) Simulation cell of MWCNT (2% by volume) reinforced PC containing 20,173 atoms.

    polymerizingthecarbonaterepeatunitandtakingthechainlengthasten.Fig. 3showsasimulationcellofpurePC.The

    densityofthecellwas1.5 g/cm3 andthecelldimensionsweretakenas:49.349.349.3 3

    .ThemolecularmodelofMWCNT/PCcompositeswithdifferentdimensionswasbuiltwiththeuseoftheAmorphouscellmodule.Firstly,aMWCNToflength41.81 wasconstructedusingtheBuildnanostructuredialogboxinMaterialsStudio.Theconfigurationoftheinnerandouternanotubeswastakenas(3,3)and(6,6)respectively.Thistubewastheninsertedinasimulationcelloftherequireddimensions.Lastly,PCwaspackedaroundtheMWCNTusingthePackingfeatureavailableinAmorphousCell.Inthisstudy,simulationswereperformedintwoways.Inthefirstapproach,MWCNTswerepackedinPCbyweightandinthesecondapproach,packingwasdonebyvolume.Fig. 4andFig. 5show,respectively,thestructuresobtainedafterpacking

    by

    weight

    and

    by

    volume.

    Similar

    to

    Figs. 45,

    a

    number

    of

    simulation

    cells

    were

    created

    in

    the

    software,

    and

    mechanicalpropertieswereobtained.

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    5/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 375

    Fig. 6. (Color online.) Simulation cell of MWCNT (Vf=0.04, l/d=30) reinforced PC composite containing 20,798 atoms.

    AnumberofsimulationswerealsoperformedforafixedMWCNTvolumefractionandbyvaryingtheaspectratio.Theaspect ratio of MWCNTs was varied from l/d=5 to l/d=100.Fig. 6 shows one of the simulation cells of an MWCNT-reinforcedPCcompositewith Vf= 0.04 and l/d=30.AsimilarapproachwasadoptedwhilemodelingzigzagandchiralMWCNT-reinforcedPCcomposites.

    3.Thermalconductivity

    Withthedimensionsofelectronicandmechanicaldevicesapproachingthenanometerscale,efficientheatremovalisofcrucialimportancetobothperformanceandfunction.Whileabasicunderstandingofheattransportindielectricshasal-readybeenachieved,manyimportantissuesremainunresolved.Theinterpretationofexperimentalresultsremainsdifficultbecausetypicallythecontributionsofindividualdefectscannotbedeconvoluted.MDsimulationsare ideal foraddressingsuch issuessince theycanbeused tostudy individualmicrostructuralelements, thereby identifying themost importantissues for thermal conductivity in polycrystalline materials. For example, by elucidating the correlation between grain-boundarystructureandthermal-transportproperties,onemayhope toeventuallydesignmaterialswith tailoredthermalproperties.However,priortoasystematicstudyofinterfacialeffects,itisnecessarytofirmlyestablishsuitablecomputa-tionalmethods.ThethermalconductivityrelatestheheatcurrenttothetemperaturegradientviaFourierslawas:

    J=

    T

    x(2)

    where J isacomponentofthethermalcurrent, isanelementofthethermalconductivitytensor,andT/x isthegradientofthetemperatureT.Experimentally, istypicallyobtainedbymeasuringthetemperaturegradientthatresultsfromtheapplicationofaheatcurrent.

    In MD simulations, thermal conductivity can be computed either using non-equilibrium MD (NEMD) or equilibriumMD (EMD). The two most commonly applied methods for computing thermal conductivity are the direct method andtheGreenKubomethod.Thedirectmethod is anNEMDmethod that relieson imposing a temperature gradient acrossthesimulationcellandisthereforeanalogoustotheexperimentalsituation.Bycontrast,theGreenKuboapproach isanEMDmethodthatusescurrentfluctuationstocomputethethermalconductivityviathefluctuationdissipationtheorem.Schellinget al. [12] studiedandcomparedthefeaturesofeachmethod.TheypointedoutthatNEMDmightcontainnon-lineareffectsduetotheapplicationoftherequiredtemperaturegradient.TheyalsonotedthatwhilebothEMDandNEMDapproachesexhibitfinitesizeeffects,theseeffectsaremuchmoresevereinNEMDduetothepresenceofinterfacesattheheatsourceandsink.Furthermore,EMD facilitatesthermalconductivityprediction inalldirectionsusingonesimulation,

    whereas

    NEMD

    requires

    the

    use

    of

    a

    thermal

    gradient

    and

    therefore

    only

    enables

    the

    calculation

    of

    thermal

    conductivityinonedirection.Therefore,EMDisparticularlyusefulforgeometrieswhereperiodicboundaryconditionscanbeapplied.

    However,thebasisofEMDisthefluctuationdissipationtheorem,whileNEMDsbasisinFouriersLawofconductionmakesNEMDanalogoustoexperimentalmeasurements.Furthermore,EMDhasbeenoftencomputationallymoreexpensiveandtheresultsaremoresensitivetosimulationparameters.

    InEMD,thesystemissettothedesiredtemperature,andthenaconstantenergyschemeisusedwiththewell-knownGreen[13]andKubo[14]relationstocalculatethethermalconductivitytensor.MDsimulationsmaybeappliedtodifferentstatisticalensembles,namelycanonical(NVT),grand-canonical(PT),andmicro-canonical(NVE).Itisworthnotingthattheirderivationshavebeendoneindifferentensembles:theformerinmicro-canonicalandthelatteringrand-canonical.Lepriet al. [15] resolvedthisdiscrepancybynotingthatifthemicro-canonicalensembleisused,thenthethermalconductivitymight diverge non trivially unless the velocity of the center of mass of the system is set to zero or alternatively sometermsare subtracted from thecalculatedheatflux vectors.However, thecanonical ensemblecanalsobeusedwith theGreenKubo formulatopredictthethermalconductivitybyapplyingadditionalthermalforcestoalltheatoms.Dynamic

    properties

    such

    as

    thermal

    conductivity

    are

    calculated

    in

    EMD

    based

    on

    the

    fluctuation

    dissipation

    and

    linear

    responsetheorem.Thismethodappliesthefactthattheheatflowinasystemofparticlesintheequilibriumstatefluctuatesaround

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    6/26

    376 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    zero. The heat flux vectors and their correlations are computed throughout the simulations. The time needed for HeatAuto-CorrelationFunctions(HACF)todecaytozero isthenusedthroughtheGreenKuborelationtopredictthethermalconductivity.

    NEMDsimulationsprovideameanstocalculatethermalconductivityinawayanalogoustotheexperimentalmeasure-mentseitherbyimposingathermalgradientintothesystemofparticlesorbyintroducingaheatflow.

    Inthereversenon-equilibriumMD(RNEMD)method,theenergyexchangeiscarriedoutbyexchangingthekineticen-ergy of two particles: the hottest particle in the cold layer and the coldest particle in the hot layer. The energyE is

    therefore

    variable

    and

    needs

    averaging

    over

    many

    exchanges.

    In

    the

    method

    of

    Jund[16],

    the

    energy

    E is

    fixed,

    and

    in-volvesallparticlesinthehotandcoldlayers.Bothmethodsconservetotallinearmomentumandenergyofthesystem.TheimposedfluxmethodofJund [16]alsoconservesthetotallinearmomentumofthehotandcoldlayer,i.e.nomomentumisexchanged.

    In this study, the imposed flux method was used to find thermal conductivity. The number of layers in which thedirectionofthefluxisdividedwasfixedat40.Increasingthenumberoflayerscanincreasetheaccuracyofthegradients,buttoomanylayerswillleadtolargefluctuationsinthelayertemperatures.Twotypesofexchangemethodwereusedinthestudy.ThefirstisVARIABLE,whichexchangesavariableenergybetweenoneobjectinthehotlayerandoneinthecoldlayer.FIXEDexchangesaconstantenergybetweenallhotobjectsinthehotlayerandallobjectsinthecoldlayer.TheamountofenergytoexchangeineachstepwhenusingtheFIXEDexchangetypewastakenas1kcal/mol.Thefluxwasdeterminedbytheratioofexchangeenergyandnumberofsteps.Thenumberofexchangesduringtheequilibrationstagewastakenas500.Duringtheequilibrationstage,athermostat(NVT)actsonthesystem.Thenumberofexchangesduringtheproductionstagewasequalto1000.Theproductionstagewascarriedoutatconstantenergy(NVE).Atimestepof1 fs

    was

    used

    in

    the

    simulation.

    The

    number

    of

    time

    steps

    in

    between

    two

    exchanges

    was

    fixed

    at

    100.

    Decreasing

    the

    numberofsteps leadstohigherfluxesandincreasesthetemperaturegradient.Toosmallvalueswereavoidedasthese introduce

    nonlineareffectsandmayimpactperformance.Tofindthermalconductivity,MDsimulationswereperformedwithMWCNTvolumefractionvaryingfromVf= 00.16

    and the aspect ratio was kept fixed at l/d=10. Also, simulations were performed with varying the aspect ratio (l/d)of MWCNT and the fixed volume fraction Vf= 0.04. Thermal conductivity results obtained using MD simulations werecompared with other models such as series model, parallel model, MaxwellGarnett model, LewisNielsen model, andHamiltonCrosser model. Comparison of MD results for thermal conductivity was made with Dengs model. A brief in-troductiontothesemodelswillbegiveninthenextsection.

    4. Modelstocalculatethermalconductivity

    Many theoreticalandsemi-theoreticalmodelsareavailable torepresent theeffective thermalconductivityofconven-

    tional

    polymer

    composites

    in

    which

    large-size

    fillers

    have

    been

    dispersed

    in

    a

    polymer

    matrix.

    Simple

    models

    such

    as

    theseriesmodelgivethelowerbound,whereastheparallelmodel(ruleofmixture)givestheupperboundofthethermalcon-

    ductivityofananocomposite.Asmightbeexpected,experimentalobservationssuggestthatrealvaluesfornanocompositesfallsomewhereinbetweenthesetwolimits.

    4.1. Parallelandseriesmodels

    Boththeseriesmodelandtheparallelmodelassumethateachphasecontributeindependentlytotheoverallthermalre-sistanceandconductance,respectively,andassumeaperfectinterfacebetweenanytwophasesincontact.Theseriesmodelappliesreadilytothethermalconductivityofalaminatedcompositealongthestackingdirection.However,ittypicallygivesanunderestimationofthermalconductivityduetothepresumablycompletelocalizationofthecontributionfromthefibersembeddedinthematrix,neglectingtheinteractionamongthefillers.Therefore,theseriesmodelgivesthelowerboundforthethermalconductivityofcomposites.Theseriesmodelofthermalconductivityisgivenby:

    k=(km kf)/

    kf (1 v f) + km v f

    (3)

    where,

    k=thermal conductivity of compositekm=thermal conductivity of polycarbonate matrix =0.15 W/m/Kkf=thermal conductivity of armchair MWCNT =30 W/m/Kvf=MWCNT volume fraction

    Incomparison,theparallelmodelpredictsthethermalconductivityofconventionalcontinuousfiber-reinforcedcompos-itesalongthefiberalignmentdirection.Theruleofmixture implicitlyassumesperfectcontactbetweenfibers.However,itgivesa largeoverestimationofthermalconductivityandgivesanupperboundforthermalconductivityofcomposites.Itisworthpointingoutthatthermalconductivitymeasurementresultsofcompositesshouldalwaysfallbetweenthepre-

    dictions

    of

    the

    series

    model

    (lower

    bound)

    and

    the

    parallel

    model

    (upper

    bound),

    except

    for

    the

    cases

    where

    interfacialphononscatteringinnano-laminatescanyieldevenlowerthermalconductivitythanthelowerboundbytheseriesmodel.

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    7/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 377

    Fortheparallelmodel,thermalconductivityisgivenas

    k=(1 v f) km+v f kf (4)

    wherethesymbolshavethesamemeaningaspreviouslydescribedintheseriesmodel.

    4.2. MaxwellGarnettmodel

    The

    problem

    of

    determining

    the

    effective

    transport

    properties

    of

    multiphase

    materials

    dates

    back

    to

    Maxwell.

    Based

    on

    thecontinuityofpotentialandelectriccurrentat the interface,and on theassumption that the interactionsamong thefibersarenegligible,whichmeansthat thedisseminatedfibersare located farenough fromeachother,Maxwellderivedananalytical formula for the effective specific resistance (K)of acompoundmediumconsistingofa substanceof spe-cificresistance K2 , inwhicharedisseminated small spheresof specificresistance K1 , the ratioof thevolumeofall thesmallspherestothatofthewhole,beingp.Whentransformedtothethermalconductivity(k)discussedhere,themodelgives:

    k= km+3 v f (kf km)/(2 km+ kf v f (kf km)

    (5)

    Eq.(5) givessatisfactoryresultsforcompositeswith:(i) verylowvf,(ii) gooddispersion,and(iii) nointerfacialthermalresistance.ItisalsoreferredtoasMaxwellGarnett(MG)equation.

    4.3.

    LewisNielsen

    model

    Inthismodel,conductivitybecomestheanalogofstiffnessortheelasticshearmodulus,andthedisturbanceofthefluxfieldbecomesanalogous to thedisturbanceof the stressfieldby thedispersed filler.Starting from theHalpinTsai [17]equations,whicharewidelyusedinmicro-mechanics,Nielsenappliedamodifiedequation,theNielsenLewisequation,tothemodelingofthermalconductivity:

    k= km(1 +A B v f)/(1 B v f)

    (6)

    where

    A= kE 1

    kE isthegeneralizedEinsteincoefficient,anddependsprimarilyupontheshapeofthefillersandhowtheyareoriented

    with

    respect

    to

    the

    direction

    of

    the

    heat

    flow.

    In

    our

    case,

    A

    =2 l/d

    =10,

    B=(kf/km) 1

    /(kf/km) +A

    =1 + 1.775 v f

    km=thermal conductivity of polycarbonate matrix =0.15 W/m/Kkf=thermal conductivity of armchair MWCNT =30 W/m/Kvf=MWCNT volume fraction.

    Although Nielsens model is a semi-empirical model, significant improvements in the model shouldbe appreciated.Theshape effect and to some extent theorientation effect are both taken into account. Reducedfiller loading () accountsfor themaximumpackingdensityof thefillerswitha specificshapeandsizedistribution,and isunique to thismodel.In comparison, most of the theoretical equations assume uniform changes of filler loading up to the point where the

    dispersed

    phase

    makes

    up

    the

    complete

    system,

    which

    is

    not

    realistic.

    The

    NielsenLewis

    equation

    gives

    a

    higher

    predic-tionthantheMaxwellGarnettequation,mainlybecauseofthereducedfiller loading.However,weshouldnotethatthemodelgivestoohighapredictionathighfillerloading.Inaddition,interfacialthermalresistanceisnotconsideredinthismodel.

    4.4. HamiltonCrossermodel

    TheHamiltonandCrossermodel [18] isanextensionofMaxwells theoryaccounting for thenon-sphericityoffillersthroughtheuseofashapefactor,n,definedasn=3/ ,with beingtheparticlesphericity.Thesphericity( = Ae/A)ofaparticleisdefinedastheratioofthesurfacearea(Ae)oftheequivalentspherehavingthesamevolumetotheactualsurfacearea(A)ofthenon-sphericalfiber.TheeffectivethermalconductivityoftheHamiltonandCrossermodelisgivenby:

    k= km

    kf+ 5km 5vf (km kf)kf+ 5km+v f (km kf)

    (7)

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    8/26

    378 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Fig. 7. (Color online.) Variation of Youngs modulus, E11 and E22, with the percentage of MWCNT (by weight) in PC.

    where

    km=thermal conductivity of polycarbonate matrix =0.15 W/m/K,kf=thermal conductivity of armchair MWCNT =30 W/m/K,

    vf=MWCNT volume fraction.

    Sinceneitherinterfacialresistancenorfiberfiberinteractionwastakenintoaccount,thefibersizewasfoundtohavenoinfluenceontheeffectivethermalconductivityofthecompositeinthismodel.

    4.5. Dengmodel

    Forestimatingthemechanicalandthermalpropertiesofinclusion-in-matrixcomposites,evenforthosewithhighcon-centratedinclusions,sothatthe interactionsamongtheinclusionsmustbeconsidered,somematuremethodshavebeenestablishedwithintheframeworkofmicro-mechanics.ForCNTcompositeswithlowloadingsofrandomlyorientedstraightCNTsofaverage length L and diameterd,an analytical estimate for theeffective thermal conductivities,ke , of the CNTcompositescanbegiveninthefollowingform:

    ke

    km =1 +

    vf3

    kmkf

    +H (8)

    where

    ke=equivalent thermal conductivity of the composite,km=thermal conductivity of polycarbonate matrix =0.15 W/m/K,kf=thermal conductivity of armchair MWCNT =30 W/m/K,vf=MWCNT volume fraction,

    H=1

    p2 1

    pp2 1

    lnp+

    p2 1

    1

    ,

    p= l/d.

    5.

    Results

    and

    discussion

    Thissectionhasbeendividedintotwoparts.InSection5.1,wehavediscussedtheresultsobtainedforelasticmoduli.InSection5.2,wehaveexplainedthedampingresultsandinSection5.3,wehaveexaminedtheresultsofthermalconductivity.

    5.1. Elasticmoduli

    Fig. 7 shows thevariationofYoungsmodulus E11 and E22 ,with thepercentageofMWCNT (byweight) inPC.BothmodulishowanincreasingtrendwithanincreaseintheweightpercentageofMWCNT.However,theincreaseintransversemodulus,E22 ,istoolowincomparisontothatofE11 .ThisconfirmsthefactthatMWCNTPCcompositeswhenloadedintransversedirectionbehavepoorly.Moreover,itcanbeinferredfromTable 1andFig. 7thatwhenincreasingtheMWCNTweightpercentageby05%,theincreaseinmoduliisgreaterthanwhentheMWCNTweightpercentageincreasesby510%.TheimprovementinmechanicalpropertieswasfoundforminorcompositionsofMWCNTsinPCandtheyfurtherincreased

    as

    the

    composition

    of

    MWCNTs

    was

    increased.

    However,

    beyond

    a

    certain

    limit

    of

    composition,

    this

    pattern

    was

    not

    fol-lowed.Ithasbeenarguedthatproperdispersionatlowercompositions,stronger interactionbetweenMWCNTsand

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    9/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 379

    Table 1

    PercentageincreaseinmoduliwhenincreasingtheweightpercentofMWCNT.

    MWCNT (weight %) Percentageincreaseinlongitudinalmodulus(E11)

    Percentageincreaseintransversemodulus(E22)

    05 89.65 54.13510 21.81 3.87

    Table 2

    PercentagedifferencebetweenKumaret al. [11] andpresentstudyforthelongitudinalmodulus(E11).

    MWCNT (weight %) Percentage difference

    0 2.030.50 5.400.75 5.752 5.205 4.18

    10 4.14

    Fig. 8. (Color online.) Variation of the shear modulus with the percentage of MWCNT (by weight) in PC.

    Table 3

    PercentageincreaseinshearmoduliwhenincreasingtheMWCNTweightpercentage.

    MWCNTrange(weight %)

    Increase in shear moduli (in %)

    G23 G31 G12

    02 110 83.33 63.33210 42.85 36.36 35.45

    basematrix,combinationoflargeaspectratioandhighsurface-to-volumeratioofMWCNTsandimprovedloadtransferca-pabilityofMWCNTsassistedintheimprovementoftheproperties,butforhighercompositions,aggregatesofMWCNTswereformedandthecurvyandslipperynatureofMWCNTsdidnotassistinfurtherimprovementofthemechanicalproperties.

    From

    Table 2and

    Fig. 7,

    it

    can

    be

    inferred

    that

    the

    results

    of

    our

    study

    supplement

    the

    experimental

    study

    conductedbyKumaretal.forthesamematerial.AscanbeobservedfromTable 2,MDsimulationresultsfromthisstudyareclose

    totheexperimentalresults.Fig. 8showsthevariationofshearmoduli,G23 ,G31 andG12 withMWCNTweightpercentage.ComparingFig. 7andFig. 8,itcanbeconcludedthatMWCNT-reinforcedPCcompositesareweakwhenloadedinshearincomparisontowhenloadedintension.ThoughtheshearmoduliincreasewiththeMWCNTweightpercentage,theriseismuch lesserwhencomparedwiththerisein longitudinalmodulus E11 .Table 3showsthatinitially,whenincreasingtheMWCNTweightpercentfrom02,theriseinalltheshearmoduliisgreaterincomparisontotheincreaseinshearmoduliwhenMWCNTweightpercentincreasesby2to10.ThisisduetothecurvyandslipperynatureofMWCNTs.

    Fig. 9showsthevariationofYoungsmoduli,E11 andE22 ,withthepercentageofMWCNT(byvolume)inPC.Itshowsthatthe longitudinalmodulus (E11) increaseswhen increasingtheMWCNTvolume fraction.Thepercent increase in E11when increasing theMWCNTvolume fraction isshown inTable 4.FromFig. 9 andTable 4, itcanbeobserved that E11increasessharplytilltheMWCNTvolumefractionis2%.Thereafter,theriseisminimalduetotheaggregationofMWCNTsandthecurvyandslipperynatureofMWCNTsdoesnotassistinfurtherimprovementofthemechanicalproperties.Fig. 9

    also

    shows

    that

    the

    increase

    in

    the

    transverse

    modulus

    (E22)

    is

    negligible

    in

    comparison

    to

    the

    rise

    in

    E11 .

    This

    is

    due

    tothefactthatthefibersarealignedinthelongitudinaldirection.AsthepercentageofMWCNT(byvolume)inPCincreases,

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    10/26

    380 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Fig. 9. (Color online.) Variation of Youngs modulus, E11 and E22, with the percentage of armchair MWCNT (by volume) in PC.

    Table 4

    Percentagevariationinlongitudinalmodulus(E11)whenincreasingtheMWCNTvolumefraction.

    MWCNT

    (volume %)

    Percentagedifferenceinlongitudinal

    modulus

    (E11)

    w.r.t.

    previous

    value

    Percentage increase in E11

    0 from Vf=0 to Vf=6% from Vf=6 to Vf=16%2 925.86

    1362.07 88.63

    4 20.506 18.278 15.09

    10 14.1612 13.5114 12.6916 12.22

    Fig. 10. (Color online.) Variation of the shear modulus with the percentage of armchair MWCNT (by volume) in PC.

    thevolumeofthefibersisincreasing.SinceMWCNTshaveahighmodulusinthelongitudinaldirection,thecompositehasimprovedpropertiesinthesaiddirection.Inthetransversedirection,thepropertiesofthecompositearematrixdominatedandhencelowervaluesofE22 areobserved.

    Fig. 10showsthevariationofshearmoduliwhenincreasingtheMWCNTvolumefraction.WhenincreasingtheMWCNTvolume fraction by 00.06%, the percent increase in transverse shear modulus G23 is approximately 183%, whereas thepercentincreasereducestomerely10%whentheMWCNTvolumefractionincreasesby0.060.16.Also,theriseinalltheshearmoduli isnegligible incomparison to the rise in E11 .This isagaindue to the fact that inevery shearmode, thepropertiesofthecompositearematrixdominatedandhencelowervaluesofallshearmoduliareobserved.Table 5showsthepercentageincreaseinshearmoduliwhenincreasingtheMWCNTvolumefraction.

    Fig. 11 showsthevariationofYoungsmodulus(E11)withthepercentageofMWCNT(byvolume)inPC, fordifferentconfigurationsofMWCNT.AllthesimulationswereperformedwithMWCNTaspectratio(l/d=10).Thecarbonatomsin

    a

    CNT

    are

    in

    sp2 configurations

    and

    connected

    to

    one

    another

    by

    three

    strong

    bonds.

    Due

    to

    the

    geometric

    orientationofthecarboncarbonbondsrelativetothenanotubeaxis,armchairMWCNT-reinforcedPCcompositesexhibithightensile

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    11/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 381

    Table 5

    PercentageincreaseinshearmoduliwhenincreasingtheMWCNTvolumefraction.

    MWCNTrange(volume %)

    Increase in shear moduli (in %)

    G23 G31 G12

    06 183.33 175 166.66616 9.73 9.69 9.15

    Fig. 11. (Color online.) Variation of Youngs modulus ( E11) with the percentage of MWCNT (by volume) in PC, for different configurations of MWCNT.

    Fig. 12. (Color online.) Variation of Youngs modulus ( E22) with the percentage of MWCNT (by volume) in PC, for different configurations of MWCNT.

    Table 6

    PercentageincreaseinelasticmoduliwhenincreasingMWCNTvolumefraction.

    MWCNTrange(volume %)

    Increase in moduli (in %)

    Armchair Zigzag Chiral

    E11 E22 E11 E22 E11 E22

    06 1362.07 13.10 1234.82 10.35 1137.58 8.27616 98.63 37.20 95.17 35.00 95.16 32.16

    strengthandYoungsmodulivaluescomparedtothezigzagMWCNT-reinforcedPCcomposites.ChiralMWCNT-reinforcedPCcompositeshavelowerelasticmodulibecauseofthecurvednatureofallC=Csp2 bonds.

    Fig. 12 showsthevariationofYoungsmodulus(E22)withthepercentageofMWCNT(byvolume) inPC,fordifferentconfigurationsoftheMWCNT.ComparingFig. 11andFig. 12,itcanbeobservedthattheincreaseinE22 issmallerthanthatofE11 .Table 6showsthepercentageincreaseinelasticmoduliuponincreasingtheMWCNTvolumefraction.FromFig. 11andTable 6,itcanbeinferredthatinitially,whenincreasingtheMWCNTvolumefractionby06%,theriseintheelasticmodulusE11 isfasterwhencomparedwiththeriseinmoduluslateron.FromFigs. 1112andTable 6,itcanbeconcludedthatthepercentageriseinmoduliisgreatestforarmchairMWCNT-reinforcedPCcomposites.

    Fig. 13shows

    the

    variation

    of

    the

    shear

    modulus

    (G23)

    with

    the

    percentage

    of

    MWCNT

    (by

    volume)

    in

    PC,

    for

    differentconfigurations of MWCNT. It can be observed that the increase in shear modulus is the greatest for armchair MWCNT-

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    12/26

    382 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Fig. 13.(Color online.) Variation of the shear modulus ( G23) with the percentage of MWCNT (by volume) in PC, for different configurations of MWCNT.

    Fig. 14.(Color online.) Variation of the shear modulus ( G31) with the percentage of MWCNT (by volume) in PC, for different configurations of MWCNT.

    Fig. 15.(Color online.) Variation of the shear modulus ( G12) with the percentage of MWCNT (by volume) in PC, for different configurations of MWCNT.

    reinforcedPCcomposites.Also,thevaluesareconsiderablysmallerincomparisontothevaluesofthelongitudinalelasticmodulus.

    Fig. 14showsthevariationoftheshearmodulus(G31)withthepercentageofMWCNT(byvolume)inPC,fordifferentconfigurationsofMWCNT.Thevaluesof G31 aresmaller incomparison tothevaluesofG23 .Fig. 15 showsthevariationoftheshearmodulus(G12)withthepercentageofMWCNT(byvolume)inPC,fordifferentconfigurationsofMWCNT.ThevariationissimilartothatobservedforG31.

    Table 7showsthepercentageincreaseinshearmoduliwithanincreaseintheMWCNTvolumefraction.Itshowsthat

    the

    percentage

    increase

    in

    modulus

    is

    the

    greatest

    for

    armchair

    MWCNT-reinforced

    PC

    composites.

    Initially,

    the

    increasein all the shear moduli is faster, when the MWCNT volume fraction increases by 06%, in comparison to the increase

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    13/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 383

    Table 7

    PercentageincreaseinshearmoduliwhenincreasingtheMWCNTvolumefraction.

    MWCNTrange(volume %)

    Increase in moduli (in %)

    Armchair Zigzag Chiral

    G23 G31 G12 G23 G31 G12 G23 G31 G12

    06 183.33 175.00 166.66 143.33 141.66 133.33 132.45 130.00 80.00616 28.23 19.70 13.87 20.00 19.31 13.57 15.83 13.40 13.14

    Fig. 16.(Coloronline.)VariationofYoungsmoduliE11 andE22,withtheaspectratio(l/d)ofarmchairMWCNT-reinforcedPCcompositewithfixedvolumefraction(Vf=0.04).

    Fig. 17. (Coloronline.)Variationoftheshearmoduliwiththeaspectratio(l/d)ofarmchairMWCNT-reinforcedPCcompositewithfixedvolumefraction(Vf=0.04).

    inmodulus lateronwhen theMWCNTvolume fraction increasesby616 %.ThecurvyandslipperynatureofMWCNTscausestheshearmodulustodecreasewhenincreasingtheMWCNTvolumefraction.Moreover,Table 7alsoshowsthatthepercentageincreaseinshearmoduliofthezigzagMWCNT-reinforcedPCcompositeisgreaterthanthepercentageincrease

    in

    the

    shear

    moduli

    of

    chiral

    MWCNT-reinforced

    PC

    composites.

    Comparing

    Table 6and

    Table 7,

    it

    can

    be

    concluded

    thatthepercentageincreaseinshearmoduliissmallerincomparisontothepercentageincreaseinlongitudinalelasticmodulus.

    Thus,MWCNT-reinforcedPCcompositesareweakerinshearthaninothermodesofloading.Fig. 16showsthevariationofYoungsmoduli E11 and E22 withtheaspectratio (l/d)ofarmchairMWCNT-reinforced

    PC composite with volume fraction (Vf=0.04). When increasing the MWCNT aspect ratio (l/d) till l/d=50, both thelongitudinalandtransverseelasticmoduliincreaserapidly.Thereafter,theincreaseissmaller.WhenthelengthoftheCNTsincreases,theefficiencyofloadtransferisincreaseddramatically.Butforlargerlengths,theincreaseinvanderWaalsforcescausesthemodulustoincreaseataslowerrate,ascanbeobservedfromFig. 16.Fig. 17showsthevariationoftheshearmoduluswiththeaspectratio(l/d)ofarmchairMWCNT-reinforcedPCcompositeswithafixedvolumefraction(Vf=0.04).Itcanbe inferred thatthe increase inshearmoduli issignificantlysmaller incomparisontothe increase in longitudinalandtransverseelasticmoduli.Alltheshearmoduliincreasetilll/d=50.Afterthat,theriseisminimal,whichmaybeduetotheincreaseinthevanderWaalsforceforhigheraspectratios.Fig. 17alsoshowsthattheincreaseinG31 andG12 issmallerthantheincreaseinG23 .

    Table 8 shows

    the

    percentage

    increase

    in

    elastic

    moduli

    when

    increasing

    the

    MWCNT

    aspect

    ratio

    (l/d)

    for

    a

    fixedvolumefraction(Vf= 0.04).FromFig. 16andTable 8,itcanbeconcludedthatE11 andE22 increasesharplyfroml/d=5

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    14/26

    384 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Table 8

    PercentageincreaseinelasticmoduluswhenincreasingtheMWCNTaspectratio(l/d).

    MWCNT aspect ratio (l/d) Percentage increase in E11 Percentage increase in E22

    550 418.07 381.5650100 6.43 46.41

    Table 9

    Percentage

    increase

    in

    shear

    moduli

    when

    increasing

    the

    MWCNT

    aspect

    ratio

    (l/d).

    MWCNTaspectratio(l/d)

    PercentageincreaseinG23

    PercentageincreaseinG31

    PercentageincreaseinG12

    550 220.34 169.04 132.8950100 19.42 25.00 26.83

    Fig. 18. (Coloronline.)Variationofthelongitudinalmodulus(E11)withtheaspectratio(l/d)fordifferentconfigurationsoftheMWCNT-reinforcedPCcompositewithfixedvolumefraction(Vf=0.04).

    to l/d=50. A possible reason for this is the presence of shear stress concentration at numerous fiber ends in a short

    fiber

    composite.

    However,

    for

    l/d

    >50,

    both

    moduli

    increase

    slowly.

    This

    is

    because,

    after

    l/d

    =50,

    higher

    fiber

    axial

    stressoccurs,whichcausesthemodulitoincreaseataslowerrate.

    Table 9 shows thepercentage increase inshearmoduliwhen increasing theMWCNTaspectratio (l/d). Itshows thatthepercentage increase inall the shearmoduli isgreaterwhen theaspectratiovaries from l/d=5 to l/d=50.This isprobablyduetothefactthatat lowfiberaspectratio,thenumberoffibersinthecompositewillbemoreforthesamevolumefraction.Thehigherthenumberofdiscontinuousfibersis,thehigherthenumberoffiber-endsavailableforstressconcentrationis.

    Thus,atlowvaluesofl/d,thestressconcentrationatnumerousfiberendsleadstoariseinthevaluesofalltheshearmoduli.Asthefiberaspectratioincreases,shearmodulibecomeconstant,asshowninFig. 17.

    Fig. 18 shows the variation of the longitudinal modulus (E11) with the aspect ratio (l/d) for different configurationsof MWCNT-reinforced PC composites with fixedvolume fraction (Vf= 0.04). It can be observed that armchair MWCNT-reinforced PC composites have a higher elastic modulus E11 , in comparison to zigzag and chiral MWCNT-reinforced PCcomposites. E11 increasesrapidlytilll/d=50,afterwhichtheincreasetakesplaceataslowerrate.Duetothegeometric

    orientation

    of

    the

    carboncarbon

    bonds

    relative

    to

    the

    nanotube

    axis,

    armchair

    MWCNT-reinforced

    PC

    composite

    exhibithigherYoungsmodulivaluescomparedtothezigzagMWCNT-reinforcedPCcomposites.ChiralMWCNT-reinforcedPCcom-

    positeshave lowerelasticmodulibecauseof thecurvednatureofallC=C sp2 bonds.At lowervaluesof l/d, thestressconcentrationatnumerousfiberendsleadstoariseinthevaluesofalltheshearmoduli.

    Fig. 19 shows the variation of transverse modulus (E22) with the aspect ratio (l/d) for different configurations ofMWCNT-reinforced PC composites with a fixed volume fraction (Vf=0.04). It can be observed here also that armchairMWCNT-reinforcedPCcompositeshavehigherelasticmodulus E22 incomparisontozigzagandchiralMWCNT-reinforcedPCcomposites.ItisinferredbycomparingFigs. 18 and 19thatthevaluesof E22 aresignificantlysmallerthanthecorre-spondingvaluesofE11 .ThisisbecauseMWCNTsarealignedinthedirectionoftheappliedstrain.

    Fig. 20showsthevariationoftheshearmodulus(G23)withtheaspectratio(l/d)fordifferentconfigurationsofMWCNT-reinforcedPCcompositeswithfixedvolumefraction(Vf= 0.04).ArmchairMWCNT-reinforcedPCcompositeshavehighershearmoduliincomparisontozigzagandchiralMWCNT-reinforcedPCcomposites.G23 increasesrapidlytilll/d=50,afterwhichtheincreaseoccursataslowerrate.

    Fig. 21shows

    the

    Variation

    of

    the

    shear

    modulus

    (G31),

    with

    the

    aspect

    ratio

    (l/d)

    for

    different

    configurations

    of

    MWCNT-reinforcedPCcompositeswithfixedvolumefraction(Vf= 0.04).G31 increasesrapidlytilll/d=50.Afterthat,theriseis

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    15/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 385

    Fig. 19. (Coloronline.)Variationofthetransversemodulus(E22)withtheaspectratio(l/d)fordifferentconfigurationsoftheMWCNT-reinforced PCcompositewithfixedvolumefraction(Vf=0.04).

    Fig. 20.(Color

    online.)

    Variation

    of

    the

    shear

    modulus

    (G23)

    with

    the

    aspect

    ratio

    (l/d)

    for

    different

    configurations

    of

    the

    MWCNT-reinforced

    PC

    composite

    withfixedvolumefraction(Vf=0.04).

    Fig. 21. (Color

    online.)

    Variation

    of

    the

    shear

    modulus

    (G31)withtheaspectratio(l/d)fordifferentconfigurationsoftheMWCNT-reinforcedPCcompositewith

    fixed

    volume

    fraction

    (Vf=0.04).

    minimal,whichmaybedue to the increase invanderWaals force forhigheraspect ratios.Fig. 22 shows thevariationshearmodulus (G12) with theaspect ratio (l/d) fordifferentconfigurations of MWCNT-reinforced PCcomposites withafixedvolumefraction(Vf=0.04).ThetrendissimilartothatofG23 .

    Table 10showsthepercentageincreaseinelasticmoduliwhenincreasingtheMWCNTaspectratio(l/d)forfixedvolume

    fraction

    (Vf= 0.04).

    From

    Figs. 1819and

    Table 10,

    it

    can

    be

    observed

    that

    the

    percentage

    increase

    in

    elastic

    moduli

    isgreater for MWCNT aspect ratio (l/d=5 to l/d=50) than the percentage increase in elastic moduli when l/d=50 to

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    16/26

    386 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Fig. 22.(Coloronline.)Variationoftheshearmodulus(G12)withtheaspectratio(l/d)fordifferentconfigurationsofMWCNT-reinforcedPCcompositewithfixedvolumefraction(Vf=0.04).

    Table 10

    PercentageincreaseinelasticmoduliwhenincreasingtheMWCNTaspectratio(l/d).

    MWCNT

    aspect

    ratio(l/d)Increase in moduli (in %)

    Armchair Zigzag Chiral

    E11 E22 E11 E22 E11 E22

    550 418.07 381.56 413.94 371.78 411.85 369.5450100 6.43 46.41 12.84 58.55 14.87 59.56

    Table 11

    PercentageincreaseinshearmoduliwhenincreasingtheMWCNTaspectratio(l/d).

    MWCNTaspectratio(l/d)

    Increase in moduli (in %)

    Armchair Zigzag Chiral

    G23 G31 G12 G23 G31 G12 G23 G31 G12

    550 220.34 169.04 132.89 203.57 151.25 124.48 180.24 132.25 121.4250100 19.42 25.00 45.48 27.05 36.81 47.87 36.96 48.88 53.54

    l/d=100.Table 11showsthepercentageincreaseinshearmoduliwhenincreasingtheMWCNTaspectratio(l/d).Itshowsthatthepercentageincreaseinalltheshearmoduliisgreaterwhentheaspectratiovariesfroml/d=5 tol/d=50.Thisisprobablyduetothefactthatatlowfiberaspectratios,thenumberoffibersinthecompositewillbemoreforthesamevolumefraction.Thehigherthenumberofdiscontinuousfibersis,thehigherthenumberoffiber-endsavailableforstressconcentrationis.Thus,atlowvaluesofl/d,thestressconcentrationatnumerousfiberendsleadstoariseinthevaluesofalltheshearmoduli.Asthefiberaspectratioincreases,theshearmodulibecomeconstant,asshowninFigs. 2022.

    5.2. Damping

    Whilemostoftheresearchesonnanocompositeswithcarbonnanotubesfocusonelasticproperties,relativelylittleef-

    fort

    to

    date

    has

    been

    put

    in

    the

    studies

    of

    their

    damping

    characteristics.

    In

    fact,

    there

    exists

    a

    great

    potential

    in

    developingnanocomposites with high damping capacity using carbon nanotubes, since the interfacial slips between nanotubes and

    polymerresinandbetweennanotubesthemselvesarebelievedtobesignificant.Thisisduetothenanoscaledimensionsand thehighaspectratioofnanotubes,whichresults ina large interfacialcontactareaandhigh frictionenergydissipa-tionduringtheslidingofnanotubesurfaceswithinthecomposite.Tomodelananocomposite,moleculardynamics(MD)methodsareoftenused.

    BuldumandLu[19] investigatedtheinterfacialslidingandrollingofcarbonnanotubesusingMDmethods.Itwasfoundthatananotubefirststickandthenslipssuddenlywhentheforceexertedonitissufficientlylarge.Suhret al. [20] inves-tigatedthedampingcharacteristicsofpolymersreinforcedwithmulti-wallednanotubesandconcludedthatthestick-slipmotionbetweenthenanotubesthemselvesisbelievedtobethemajorcontributiontotheoveralldampingofthenanocom-posite material. Because of the small size of nanotubes, the surface area to mass ratio (specific surface area) of carbonnanotubearrays isextremely large.Therefore, incompositeswithCNTfillers, it isanticipated thathighdampingcanbeachievedbytakingadvantageoftheweakbondingandinterfacialfrictionbetweenindividualCNTsandresin.

    In

    this

    study,

    damping

    properties

    of

    carbon

    MWCNT/PC

    composites

    have

    been

    expressed

    analytically

    from

    the

    theory

    ofshortfibercomposites.Theelasticviscoelastic correspondenceprinciplealinearelastostaticanalysiscanbeconvertedto

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    17/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 387

    Table 12

    DynamicpropertiesofMWCNTsandPC.

    Dynamic properties MWCNT Polycarbonate

    Storage modulus (E), GPa 1000 2.9(Esawi and Farag[26]) (Kumar et al. [11])

    Loss modulus (E), GPa 1.5 1.0Loss factor () 0.0015 0.345MWCNT volume fraction (Vf) 0.02

    avibratorylinearvisco-elasticanalysisbyreplacingtheelasticmoduliwiththecorrespondingcomplexmodulihasbeenusedtodefinethedynamicproperties.Thecomplexmodulushasarealandanimaginarypart.Thedampinglossfactorcanbeexpressedastheratiooftheimaginarytotherealpartofthecomplexmodulus.

    =C

    C (9)

    where

    C =C + iC =complex modulus of the compositeC =storage modulus of the composite

    C

    =loss modulus of the composite=damping loss factor of the composite.

    Bycombining the self-consistentapproachofHill [21] with the solutionsofHermans [22] andmakinga fewadditionalassumptions,HalpinandKardos[17]provideasimpleranalyticalformforpredictingmaterialpropertiesoffibercomposites.HalpinTsaiequationsneedonlyoneequationtofindallthecompositemoduliandthelongitudinalPoissonsratioissimplyfoundfromtheruleofmixtures.

    ThedynamicpropertiesofMWCNTsandpolycarbonatematrixasgivenbyKhatuaet al.[23]areusedasinputandshowninTable 12.

    TandonandWeng[24] derivedexplicitexpressionsfortheelasticconstantsofashortfibercompositeusingtheMoriTanakaapproach.Their formulae fortheplane-strainbulkmodulusk23 andthemajorPoissonratio12 arecoupled,andmust be solved iteratively. Using Tandon and Weng [24] equations, Eshelbys tensor E is evaluated and further used inEq. (10) to find the strain concentration tensor A. The strain concentration tensor is then used in Eq. (11) to find the

    stiffness

    and

    damping

    of

    the

    composite.

    AEshelby =

    I+ ESm

    CfCm1

    (10)

    C= Cm +Vf

    Cf CmAEshelby (11)

    where

    Vf=fiber volume fractionCm =stiffness matrix of PC matrixCf =stiffness matrix of MWCNTAEshelby =strain concentration tensor

    BywritingaprograminMatlab,wecansolvetheequationsandcanstudytheeffectofvariousparameterssuchasaspect

    ratio

    and

    fiber

    volume

    fraction

    on

    loss

    modulus,

    storage

    modulus

    and

    damping

    loss

    factor.

    Forcite

    mechanical

    propertiescalculationsusetheConstantstrainapproach.Theprocessstartsbyremovingsymmetryfromthesystem,followedbyan

    optionalre-optimizationofthestructure,wherethecellparameterscanbevaried.Optimizationatthisstageisalwaysad-vised,asincorrectresultscanbeobtainedifthestructureisfarfromitslowestenergyconfiguration.Foreachconfiguration,anumberofstrainsareapplied,resultinginastrainedstructure.Theresultingstructureisthenoptimized,keepingthecellparametersfixed.Forexample:

    numberofstepsforeachstrain= 4max.strainamplitude= 0.003strainpatterns100000,010000

    Thisdefinesarangeofvalues{0.003,0.001,0.001,0.003},whichareappliedtoeachstrainpattern:

    strain

    pattern

    100000

    gives

    e

    = {0.003,

    0,

    0,

    0,

    0,

    0},

    {0.001,

    0,

    0,

    0,

    0,

    0},

    {0.001,

    0,

    0,

    0,

    0,

    0},

    {0.003,

    0,

    0,

    0,

    0,

    0}strainpattern010000givese= {0,0.003,0,0,0,0},{0,0.001,0,0,0,0},{0,0.001,0,0,0,0},{0,0.003,0,0,0,0}

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    18/26

    388 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Fig. 23. (Color online.) Variation of loss factor (11) with the percentage of armchair MWCNT (by volume) in PC.

    Table 13

    Percentagevariationinlossfactor(11)whenincreasingtheMWCNTvolumefraction.

    MWCNT

    (volume %)

    Percentagedifferenceinlossfactor(11)

    w.r.t.

    previous

    value

    using

    MD

    Percentagedifferencebetween

    MoriTanaka

    model

    and

    MD0 2 5.57 3.914 2.58 5.146 2.39 6.818 2.03 9.05

    10 1.86 11.8112 1.76 12.9614 1.19 13.6216 0.35 13.64

    Each strain pattern represents the strainmatrix in Voigt notation. It is converted into the strain matrix E, such thatE(0,0)=e(0),E(1,1)=e(1),E(2,2)=e(2),E(2,1)=E(1,2)=0.5 e(3). . .

    ThesearethenusedtogeneratethemetrictensorG:

    G =H0[2E + I]H0

    where H0 is formed from the lattice vectors; I is the identity matrix and H0 is the transpose of H0. The new latticeparameterscanbederived fromG;theseare thenusedtotransformthecellparameters(fractionalcoordinatesareheldfixed).Followingthesesteps,thestructureisoptimizedandthestressiscalculated.Astiffnessmatrixisbuiltupbyfromalinearfitbetweentheappliedstrainandtheresultingstresspatterns.Inthecaseofatrajectory,thisisaveragedoverallframes.

    InMaterialsStudio5.5,theenergydissipatedcanbeobtainedaftercompletionofthemechanicalpropertycalculationtask. From the Forcite mechanical properties task, we get two values of energy. Energy in frame 1 gives that duringloading,whereasenergyinframe2givestheenergyduringunloadingcycle.Areabetween loadingandunloadingcurvesgivesus theenergydissipated.Dividing thisdissipatedenergyby theenergyobtainedbefore theunloadingcyclebeginsgives themeasureofdamping ().Resultsof MDhavebeencomparedwith the results obtained from theMoriTanakamodel(Eq. (11)).

    Fig. 23 shows

    the

    variation

    of

    the

    loss

    factor

    (11)

    with

    the

    percentage

    of

    armchair

    MWCNT

    (by

    volume)

    in

    PC.

    Itshowsthatthelossfactor(11)decreaseswhenincreasingtheMWCNTvolumefraction.Thepercentdecreasein11 when

    increasing the MWCNT volume fraction is shown in Table 13. From Fig. 23 and Table 13, it can be observed that 11decreasessharplytilltheMWCNTvolumefractionis2%.Thereafter,thefallsteadiesduetotheaggregationofMWCNTsandthecurvyandslipperynatureofMWCNTsdonotassistinafurtherdeclineofdampingproperties.ItcanalsobeobservedthatthedifferencebetweentheMoriTanakamodelandtheMDmodelgoesonincreasingasthevolumefractionincreases.

    Fig. 24 shows the variation of the loss factor (11) with the percentage of MWCNT (by volume) in PC, for differentconfigurationsofMWCNT.AllthesimulationswereperformedwiththeMWCNTaspectratio(l/d=10).ThecarbonatomsinaCNTareinsp2 configurationsandconnectedtooneanotherbythreestrong bonds.Duetothegeometricorienta-tionofthecarboncarbonbondsrelativetothenanotubeaxis,armchairMWCNT-reinforcedPCcompositesexhibithigherYoungsmodulivaluescomparedtozigzagMWCNT-reinforcedPCcomposites.SincethelossfactorisinverselyproportionaltoYoungsmodulus,thereforearmchairMWCNT-reinforcedPCcompositesexhibit lowestdampingvalues.ChiralMWCNT-reinforcedPCcompositeshavehighestlossfactorbecauseofthecurvednatureofallC=Csp2 bonds.

    Table 14shows

    the

    percentage

    decrease

    in

    loss

    factor

    when

    increasing

    the

    MWCNT

    volume

    fraction.

    From

    Fig. 24andTable 14, itcanbe inferred that initially,when increasing theMWCNTvolume fractionby06%, the fall in11 is faster

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    19/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 389

    Fig. 24. (Color online.) Variation of loss factor (11) with the percentage of MWCNT (by volume) in PC, for different configurations of MWCNT.

    Table 14

    PercentagedecreaseinlossfactorwhenincreasingtheMWCNTvolumefraction.

    MWCNTrange

    (volume %)

    Decrease in loss factor (in %)

    Armchair Zigzag Chiral

    06 11.37 10.15 8.93616 8.37 8.64 8.91

    Fig. 25. (Coloronline.)Variationoflossfactor(11)withtheaspectratio(l/d)ofarmchairMWCNT-reinforcedPCcompositeswithfixedvolumefraction(Vf=0.04).

    Table 15

    Percentagedecreaseinlossfactor(11)whenincreasingtheMWCNTaspectratio(l/d).

    MWCNTaspect

    ratio

    (l/d)

    Percentagedecrease

    in

    11

    Percentage difference between MoriTanaka model and MD model at

    l/d=5 l/d=60 l/d=100

    560 61.75 0.28 25.54 41.3060100 50.26

    whencomparedwiththefallinmoduluslateron.Fig. 25showsthevariationofthelossfactor(11)withtheaspectratio(l/d)ofthearmchairMWCNT-reinforcedPCcompositewithfixedvolumefraction(Vf=0.04).WhenincreasingtheMWCNTaspectratio(l/d)tilll/d=60,thelongitudinal lossfactordecreasesrapidly.Thereafter,thedecreaseissmaller.WhenthelengthoftheCNTsincreases,theefficiencyofloadtransferisincreaseddramatically.Butforlargerlengths,theincreaseinvanderWaalsforcecausesthelossfactortodecreaseataslowerrate.TheMoriTanakamodelgiveshighervaluesofthelossfactor(11)thanMD.

    Table 15showsthepercentagedecreasein11 whenincreasingtheMWCNTaspectratio(l/d)forafixedvolumefraction(Vf=0.04).FromFig. 25andTable 15,itcanbeconcludedthat11 decreasessharplyfroml/d=5 tol/d=60.Apossible

    reason

    for

    this

    is

    the

    presence

    of

    shear

    stress

    concentration

    at

    numerous

    fiber

    ends

    in

    a

    short

    fiber

    composite.

    However,

    forl/d>60,11 decreasesslowly.Thisisbecause,afterl/d=60,higherfiberaxialstressoccurs,whichcausesthelossfactor

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    20/26

    390 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Fig. 26. (Coloronline.)Variationoflossfactor(11)withtheaspectratio(l/d)fordifferentconfigurationsofMWCNT-reinforcedPCcompositeswithfixedvolumefraction(Vf=0.04).

    Table 16

    Percentagedecreaseinlossfactor(11)whenincreasingtheMWCNTaspectratio(l/d).

    MWCNTaspect

    ratio

    (l/d)

    Decrease in 11 (in %)

    Armchair Zigzag Chiral

    560 61.75 50.26 44.4660100 50.26 47.05 42.10

    todecreaseataslowerrate.Itcanalsobeobservedthatatlowaspectratios,theMoriTanakamodelandMDgivealmostsimilarresults.Atl/d=5,thepercentagedifferencebetweenthemisonly0.28%.

    Fig. 26 shows thevariationofthe loss factor(11)withtheaspectratio (l/d) fordifferentconfigurationsofMWCNT-reinforcedPCcompositeswithfixedvolume fraction(Vf= 0.04). ItcanbeobservedthatarmchairMWCNT-reinforcedPCcompositesexhibitlowervaluesoflossfactor(11),incomparisontozigzagandchiralMWCNT-reinforcedPCcomposites.11 decreasesrapidlytilll/d=60,afterwhichthedecreasetakesplaceataslowerrate.Duetothegeometricorientationof thecarboncarbonbondsrelative tothenanotubeaxis,thearmchairMWCNT-reinforcedPCcompositesexhibithigherYoungsmodulivaluescomparedtozigzagMWCNT-reinforcedPCcomposites.ChiralMWCNT-reinforcedPCcompositeshave

    lower

    elastic

    moduli

    because

    of

    the

    curved

    nature

    of

    all

    C=C

    sp2 bonds.

    Since

    the

    loss

    factor

    is

    inversely

    proportional

    toYoungsmodulus,thereforearmchairMWCNT-reinforcedPCcompositesexhibitlowestdampingvalues.

    Table 16showsthepercentagedecreasein11 whenincreasingtheMWCNTaspectratio(l/d)fora fixedvolumefraction(Vf= 0.04).FromFig. 26andTable 16,itcanbeobservedthatthepercentagedecreasein11 isgreaterfortheMWCNTaspect ratio (l/d=5 to l/d=60) than the percentage decrease in 11 when l/d=60 to l/d=100. Because armchairMWCNT-reinforcedPCcompositesexhibitthehighestelasticmoduli,thereforetheyhavethelowestdampingvalues.

    TherearetwopossiblemechanismsthatcouldberesponsiblefordampinginMWCNT-reinforcedpolymercompositesatthemolecularlevel:(a)energydissipationcausedbyinterfacialslidingatthenanotubepolymerinterface,and(ii)energydissipation caused by interfacial stick-slip sliding at the nanotubenanotube interface. When a normal tensile stress isappliedtoacomposite,itstartselongating.Asaresultoftheappliedstress,theresinstartsapplyingashearstressonthenanotube,thuscausingtheloadtobetransferredtonanotubes.Consequently,normalstrainstartsappearinginnanotubesandtheystartelongatingaccordingly.Whentheappliedstressissmall,thenanotuberemainsbondedtothematrix(stickingphase).Boththeresinandthenanotubemovetogetherduringthisphaseandthestrainsareequalinbothepoxyresinand

    nanotube.

    As

    the

    applied

    stress

    is

    increased,

    the

    shear

    stress

    on

    CNT

    increases.

    At

    a

    certain

    value

    of

    shear

    stress,

    called

    thecriticalshearstress,thenanotubedebondsfromtheresin.Whentheshearstressonthenanotube increasesbeyondthis

    value(asaresultofincreasedappliedstress),theepoxystartsflowingoverthesurfaceofthenanotube.The strain in the nanotube remains constant at its maximum level while the strain in the epoxy increases (slipping

    phase). In this phase, there is no transfer load between CNT and matrix, and because of thisenergydissipationdue toslippageoccurs,whichresultsinstructuraldamping.

    5.3. Thermalconductivity

    ThermalconductivityresultsobtainedusingMDsimulationswerecomparedwithothermodelssuchasseriesmodel,parallelmodel,MaxwellGarnettmodel,LewisNielsenmodel,andHamiltonCrossermodel.ComparisonofMDresultsforthevariationofthermalconductivitywithMWCNTaspectratiowasmadewithDengsmodel.Fig. 27showsthetemperatureprofilegeneratedwhilecalculatingthethermalconductivityforMWCNTPCcompositewithVf=0.10 andl/d=10.Fig. 28

    shows

    the

    variation

    of

    thermal

    conductivity

    with

    time

    for

    MWCNTPC

    composite

    with

    Vf=10%.

    It

    shows

    that

    the

    value

    ofthermalconductivitystabilizesafterapproximately60 ps.

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    21/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 391

    Fig. 27. (Color online.) Temperature profile generated while calculating thermal conductivity for MWCNTPC composites with Vf=10%.

    Fig. 28. (Color online.) Variation of the thermal conductivity with time for MWCNTPC composites with Vf=10%.

    Fig. 29. (Color online.) Variation of the energy flux with time for MWCNTPC composites with Vf=10%.

    Fig. 29showsthevariationofenergyfluxwithtimeforMWCNTPCcompositewithVf=10%.Itshowsthatthevalueofenergyfluxstabilizesafterapproximately60 ps.ThesizeoftheMWCNTsandtheirshapeplayanimportantroleintheheattransferbetweenpolymermatrixandtheincorporatedfiller.FillerswithahigherthermalconductivitythanPCimprovetheheattransferofcompositesonecanconsiderthatPC isathermalbarrier forheatpropagationwhilethefillermaterialtransmitstheheatmuchfaster.Thethermalconductivityofnanocompositesmighthaveacompletelydifferentmechanism

    in

    contrast

    to

    micro-composites.

    In

    the

    case

    of

    micro-composites,

    the

    heat

    is

    transported

    by

    micro-fibers

    much

    faster

    thaninPC.

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    22/26

    392 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Fig. 30. (Color online.) Variation of the thermal conductivity with the volume fraction for different types of MWCNTs with fixed aspect ratio ( l/d=10).

    Phonons,whichareresponsible forheatconduction indielectricmaterials,arescatteredattheinterfacebetweendis-similarmaterials.Theheatdissipatesonthesurfaceofnano-fiberstoahigherdegreethanonthesurfaceofmicro-fibers.Inthecaseofnanocompositesystemswitha surface-modifiedfiller,heattransportiscontrolledbytheinterfaceprovidedbyacouplingagentthatconnectsinorganicparticlesononesideandthepolymerhostontheotherside.Whennanosized

    fillers

    are

    used,

    the

    relative

    surface

    area

    of

    the

    interface,

    and

    thus

    the

    volume

    of

    the

    interfacial

    zone,

    is

    significant.

    Hence,theinterfacialzonewilldeterminethethermalconductivityofthesystem,sinceitcanconductheatmuchbetterthanthe

    constituentsthemselves.Thismeansthatultimatelythethermalconductivityisaffectedmorebytheinterfacialzonethanbythepolymerandnano-fibers.

    WhenassemblingCNTsintolarge-scalecomposites,itisdifficulttomakefulluseoftheexcellentthermalconductivityof individualCNTs (3000 W/mK).Weak tubetubecoupling,dangling tubeends,misalignmentand structuraldefectsallcontributetoquenchingofphononmodesandthusdecreasethethermalconductivityofthecomposites.LongerandthickerCNT would amount to a more efficient heat conduction path, which also allows the transport of phonons with longerwavelengths.Inaddition,larger-diameterMWCNTsaremorelikelytoformarigidandcompactstructureandthusreducethethermalcontactresistanceattheintertubeandtube-polymerinterfaces.

    Forshortnanotube-reinforcedcomposites,thephononmobilityisrestrictedattheinterfaceasaresultofthethermallyinsulating nature of the polymer. In such case, the overall thermal conductivity is mechanistically limited by the highCNT/polymerinterfacialthermalresistance.TheprerequisiteforinvestigatingCNTlengtheffectonthermalconductivitiesof

    composites

    is

    a

    near-ideal

    structure

    of

    the

    composites.

    Factors

    including

    CNT

    length,

    volume

    fraction,

    alignment

    and

    contactareaallcontributetotheefficiencyofphonontransportinthecomposites.

    Sincethemechanismofheatconductionbyphononsorelectronsdependsprofoundlyonthebandgapsofmaterials,theheattransfermechanismofCNTsisfoundtodependstronglyonthechirality,whichdeterminesthesizeoftheirbandgapsandelectronicproperties.The largestbandgap(on theorderof1.5 eV) is found innanotubeswith (n,m) chiral indicesdefiningthechiralvectorsatisfyingthecondition: |nm|=3p,wherep isaninteger.Forothertypesofnanotubes,thebandgapisconsiderablysmallerinthecaseofarmchairnanotubes(n=m).Thus,theelectroniccontributiontothethermalconductivitywillbesignificantinmetallicCNTswithasmallbandgap.Ontheotherhand,thermalconductivityofchiralCNTismainlygovernedbythephononcomponent.

    Fig. 30showsthevariationofthermalconductivitywiththevolumefractionofdifferenttypesofMWCNTswithfixedaspect ratio (l/d=10). It can be observed that armchair MWCNT-reinforced PC composites exhibit the highest thermalconductivityandthechiralMWCNT-reinforcedPCcompositeshavethelowestthermalconductivityamongstalltheconfig-urationsofMWCNT.BecauseofthelargebandgapinchiralMWCNTs,thecompositesreinforcedwiththesetubesdisplay

    poor

    thermal

    conductivity,

    as

    can

    be

    observed

    from

    Fig. 30.

    The

    thermal

    conductivity

    increases

    when

    increasing

    the

    volumeofMWCNTs.ItisbecauseoftheexcellentthermalconductivityofMWCNTsthattheconductivityofMWCNT-reinforcedPC

    compositesincreasessignificantly.Table 17showsthepercentageincreaseinthermalconductivitywithVf forarmchairMWCNTPCcompositeusingMD.

    ItshowsthatthethermalconductivityofarmchairMWCNT-reinforcedPCcompositeincreasesapproximatelyby380%whenincreasingtheMWCNTvolume fraction from0to16%.Table 18showsthepercentagedifference inthermalconductivityfordifferentconfigurationsofMWCNT-reinforcedPCcomposites. Itshowsthatthedifference inthermalconductivitybe-tweenthearmchairandothertypeofMWCNTsincreasesasVfincreases.ThisisduetothelargebandgapforzigzagandchiralMWCNTs in comparison witharmchair MWCNTs. Moreover, the percentagedifference betweenarmchairMWCNT-reinforced and zigzag MWCNT-reinforced PC composites is smaller in comparison to the percentage difference betweenarmchairMWCNT-reinforcedandchiralMWCNT-reinforcedPCcomposite.

    Fig. 31showsthecomparisonofMDresultsofthermalconductivityforarmchairMWCNT-reinforcedPCcompositewithothermodelswithfixedaspectratio(l/d=10).ThermalconductivityshowsanincreasingtrendwhenincreasingMWCNT

    volume

    fraction.

    The

    series

    model

    and

    the

    parallel

    model

    both

    assume

    that

    each

    phase

    contribute

    independently

    to

    theoverallthermalresistanceandconductance,respectively,andassumeaperfectinterfacebetweenanytwophasesincontact.

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    23/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 393

    Table 17

    PercentageincreaseinthermalconductivitywithVf foranarmchairMWCNTPCcompositeusingMD.

    MWCNTVf(in %)

    Thermalconductivity(W/m/K)

    Percentageincreasew.r.t.previousvalue

    Percentageincreasew.r.t.Vf=0

    0 0.15 2 0.19 26.66 26.664 0.24 26.31 60.006 0.30 25.00 100.008 0.37 23.33 146.66

    10 0.44 18.91 193.3312 0.52 18.18 246.6614 0.61 18.07 309.3316 0.72 17.26 380.00

    Table 18

    PercentagedifferenceinthermalconductivityfordifferentconfigurationsofMWCNT-reinforcedPCcomposites.

    MWCNTVf(in %)

    Percentage difference in thermal conductivity

    BetweenarmchairMWCNT-reinforcedandzigzagMWCNT-reinforcedPCcomposites

    BetweenarmchairMWCNT-reinforcedandchiralMWCNT-reinforced PCcomposites

    0 0 02 5.26 10.52

    4 8.33 16.666 13.33 23.338 16.21 27.0210 18.18 27.2712 19.23 28.8414 20.19 31.5916 22.22 33.33

    Fig. 31. (Coloronline.)ComparisonofMDresultsofthermalconductivityforarmchairMWCNT-reinforcedPCcompositeswithothermodelswithfixedaspectratio(l/d= 10).

    However,

    it

    typically

    gives

    an

    underestimation

    for

    a

    particulate

    composite

    due

    to

    the

    presumably

    complete

    localization

    ofthecontributionfromtheparticlesembeddedinthematrix,thatis,neglectingtheinteractionamongthefillers.Therefore,

    theseriesmodelgivesthelowerboundforthermalconductivityofcomposites.Incomparison,theparallelmodelpredictsthe thermal conductivity of conventional continuous fiber-reinforced composites along the fiber-alignment direction. Forcomposites with fibrous inclusions, the rule of mixture implicitly assumes perfect contact between particles in a fullypercolatingnetwork.

    However,itgivesalargeoverestimationofthermalconductivityforothertypesofcomposites,andgivesanupperboundfor the thermal conductivity of composites. It is important to point out that thermal conductivity measurement resultsofcompositesshouldalwaysfallbetweenthepredictionsoftheseriesmodel(lowerbound)andtheparallelmodel(upperbound)exceptforthecaseswhereinterfacialphononscatteringinnano-laminatescanyieldevenlowerthermalconductivitythanthelowerboundbytheseriesmodel.

    Basedonthecontinuityofpotentialandelectriccurrentatthe interface,andontheassumptionthatthe interactionsamongthesphericalfillersarenegligible,whichmeansthatthedisseminatedsmallspheresare located farenough from

    each

    other,

    Maxwell

    derived

    an

    analytical

    formula

    for

    the

    effective

    specific

    resistance

    (K)

    of

    a

    compound

    medium

    consistingof a substance of specific resistance K2 , in which are disseminated small spheres of specific resistance K1 , the ratio of

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    24/26

    394 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    Table 19

    ComparisonofMDresultsforthermalconductivityofarmchairMWCNT-reinforcedPCcompositewithothermodels.

    MWCNTVf(in %)

    Percentage difference in thermal conductivity between

    MaxwellGarnett model and MD MD and LewisNielsen model

    0 0 02 13.75 4.214 12.66 10.00

    6 10.52 16.338 10.00 21.8910 8.63 25.2212 6.15 28.4614 2.60 31.9216 1.25 34.86

    Fig.32. (Coloronline.)Variationofthethermalconductivitywiththeaspectratio(l/d)fordifferenttypesofMWCNTswithfixedvolumefraction(Vf=0.04).

    the volume of all the small spheres to that of the whole beingp. This model gives satisfactory results for compositeswith:(i) sphericalinclusions,(ii) verylowVf,(iii) gooddispersion,and(iv)nointerfacialthermalresistance.SinceneitherinterfacialresistancenorparticleparticleinteractionwastakenintoaccountintheHamiltonCrossermodel,thefibersize

    was

    found

    to

    have

    no

    influence

    on

    the

    effective

    thermal

    conductivity

    of

    the

    composite

    in

    this

    model.Although Nielsens model isa semi-empirical model, at least three important improvements to the model shouldbe

    appreciated.First,theshapeeffectandtosomeextenttheorientationeffectarebothtakenintoaccount.Second,reducedfiller loadingaccountsforthemaximumpackingdensityofthefillerswithaspecificshapeandsizedistribution,andisuniquetothismodel.Incomparison,mostofthetheoreticalequationsassumeuniformchangesoffillerloadinguptothepoint where the dispersed phase makes up the complete system, which is not realistic. Third, the earliest definition ofeffectiveunitisreflectedinthediscussiononaggregatesofspheres.

    Table 19 shows a comparison of MD results for thermal conductivity of armchair MWCNT-reinforced PC compositeswithothermodels.ItcanbeinferredthatatlowMWCNTvolumefractions,MDresultsagreewellwiththeLewisNielsenmodel.But at higher volume fractions, resultsofMDare inagreement with the MaxwellGarnett model. Fig. 32 showsthevariationofthermalconductivitywiththeaspectratio(l/d)fordifferenttypesofMWCNTswithfixedvolumefraction(Vf= 0.04).ArmchairMWCNT-reinforcedPCcompositesexhibitthehighestvaluesofthermalconductivityforvaryingl/dand fixed Vf. Heat transport in MWCNT/PC composites is carried out by phonons of various wavelengths. CNT lengths

    are

    much

    longer

    than

    the

    mean

    free

    path

    of

    phonons.

    The

    effect

    of

    short-wavelength

    phonons

    probably

    reaches

    a

    stablelevelwhilelong-wavelengthphononscontinuetocontributetotheheattransportprocess.Asaresult,alongerCNTwould

    amounttoamoreefficientheatconductionpath,whichalsoallowsthetransportofphononswithlongerwavelengths.Thisresultsinanincreaseofthermalconductivitywhenincreasingl/d.

    Fig. 33 shows the variation of thermal conductivity of armchair MWCNT-reinforced PC composite with other modelswith fixed Vf=0.04 and varying l/d. Longer CNTs with larger tubular diameter and more walls led to more efficientlong-distance phonon and electron conductions, resulting in higher thermal conductivities in the composites. It can beinferredfromFig. 33thatatlowMWCNTaspectratios,MDresultsareinagreementwiththeresultsgivenbyDengmodel.However,afterl/d=50,theresultsofMDstartdeviatingfromthoseobtainedbyDengmodel.

    AlargevolumeofliteraturereportsthecharacterizationofthePC-basedmaterialsbyusingauniversaltestingmachine.However,Kumaret al. [11] usedanano-indenter thathasalsobeenused for thecharacterizationofvariousother typesof materials. However, there is no study available at a molecular dynamic level for MWCNTPC composites, which cansupplementtheexperimentalfindings.Thecompositematerialsusedinthisstudycomprisenano-sizedfillersintheform

    of

    MWCNTs.

    Hence

    it

    becomes

    essential

    to

    study

    the

    properties

    of

    nano- or

    micro-scale

    owing

    to

    the

    size

    of

    the

    filler

    underobservationandthepropertiesatthislevelarecomplementarytothepropertiesatthemacrolevel.

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    25/26

    S. Sharma et al. / C. R. Mecanique 343 (2015) 371396 395

    Fig. 33. (Coloronline.)VariationofthethermalconductivityofarmchairMWCNT-reinforcedPCcompositeswithothermodelswithfixedvolumefractionVf=0.04.

    InanearlierworkbyJindalet al.[25],dynamicloadtestsonsimilarsamplesusingSplitHopkinsonPressureBar(SHPB)

    were

    reported,

    but

    the

    findings

    were

    somewhat

    different.

    It

    was

    found

    that

    for

    low

    concentrations

    of

    0.5%

    MWCNTs,

    theincreaseindynamicstrengthwasmostsignificantandremainedalmostthesameupto2%ofMWCNTs,whileforahigher

    concentrationof5%MWCNTs,thisincreasewasnegligible.ThemethodusedtodeterminethedatabasedonSHPBwasnotdependentonasingle-pointmeasurement,butratheronatotaleffectonthewholecrosssection,andtheextentoftheappliedloadwasmuchhigher,whichcausedcrushingofthespecimen.

    From the results of this paper, theelasticmodulus,which isa measureof the stiffness of anelasticmaterial and isaquantityusedtocharacterizematerials,increaseswhenincreasingtheweightandthevolumeofMWCNTs.Consideringacombinedeffectofbothweightandvolume studies, it issafe tostate that5%MWCNTcomposition inpurePC is themostsuitableonetoenhancethemechanicalpropertiestoasufficientextentforanymechanicalloadandpressure-relatedapplication.Plasticdeformation inamorphouspolymersoccursdue tonucleationandpropagationofshearbands. Intheunreinforcedpolymermatrix,shearbandspropagateuncheckedastherearenobarriersfortheirmovement.Ontheotherhand,thepresenceofMWCNTsinthecompositescouldofferresistanceforthepropagationofshearbands.ThereasonsfortheenhancementofelasticmodulusinpolymerMWCNTnanocompositesaregoodmechanicalinterlockingandthepresence

    of

    obstacles

    to

    the

    motion

    of

    shear

    bands.

    6. Conclusions

    ThestudypresentsresultsofMDsimulationforelasticmodulusobtainedbyusingMaterialsStudio5.5anditsupple-mentsanearlierexperimental studyconductedbyusinganano-indenteronaMWCNT-basedPCcomposite.The resultsindicateunambiguously thatstaticmechanicalpropertiesofpurePCaregreatlyenhancedbycomposing thesewith lessthan10%ofMWCNT.OnthebasisofthefindingsofthispaperontheelasticpropertiesofMWCNTPCcomposites,itissafetoconcludethataconcentrationaround2%ofMWCNTisagoodcompromisetoformcompositessuitable forenhancingbothstaticaswellasdynamicproperties.Themainfindingsofthestudyaresummarizedasbelow:

    (a) whentheMWCNTweightpercentincreasesfrom05%,thepercentageincreasein E11 and E22 isgreaterthanwhen

    the

    weight

    percent

    increases

    by

    510%;(b) our study supplements an earlier experimental study conducted by Kumar et al. [11] using a nano-indenter on aMWCNTbasedPCcomposite;

    (c) withanincreaseinweightpercentofMWCNTinPC,allthemodulishowanincreasingtrend,buttheincreaseinshearmoduliissmall;

    (d) withonly2%additionbyvolumeofMWCNTinPC, E11 increasesby925%.Thereafter,theincreaseoccursata lowerrate.TheincreaseinE22 ismuchlessthantheincreaseinE11;

    (e) armchairMWCNT-reinforcedPC compositesexhibit thehighestvalues of moduli in comparison to zigzagand chiralMWCNT-reinforcedPCcomposites;

    (f) tilll/d=50,bothlongitudinalandtransverseelasticmoduliincreaserapidly.Thereafter,theincreaseissmaller;(g) the loss factor (11)decreaseswhen increasing theMWCNTvolume fraction.11 decreases sharply till theMWCNT

    volumefractionis2%.Thereafter,thefallsteadiesduetotheaggregationofMWCNTsandthecurvyandtheslipperynatureofMWCNTsdonotassistinfurtherdeclineofdamping;

    (h) chiral

    MWCNT-reinforced

    PC

    composites

    exhibit

    the

    highest

    values

    of

    damping

    in

    comparison

    to

    zigzag

    and

    armchairMWCNT-reinforcedPCcomposites;

  • 7/26/2019 Comptes Rendus Mecanique...MWCNT PC

    26/26

    396 S. Sharma et al. / C. R. Mecanique 343 (2015) 371396

    (i) whenincreasingtheMWCNTaspectratio(l/d)tilll/d=60,thelongitudinallossfactor(11)decreasesrapidly.There-after,thedecreaseissmaller.Forlargerlengths,theincreaseinvanderWaalsforcecausesthelossfactortodecreaseataslowerrate;

    (j) because of the large band gap in chiral MWCNTs, the composites reinforced with these tubes display poor thermalconductivity.ThethermalconductivityofcompositesincreaseswhenincreasingthevolumeforalltypesofMWCNTs.Thethermalconductivityincreasesapproximatelyby27%uponadditionofonly2%byvolumeofMWCNTs;

    (k) theresultsofthermalconductivityobtainedfromMDareinagreementwiththeMaxwellGarnettandLewisNielsen

    models;(l) longerCNTswith largertubediameterandmorewallsleadtomoreefficient long-distancephononandelectroncon-

    ductions,resultinginhigherthermalconductivitiesinthecomposites.

    References

    [1] M.F.Yu,O.Lourie,M.Dyer,K.Moloni,T.F.Kelly,R.S.Ruoff,Strengthandbreakingmechanismofmultiwalledcarbonnanotubesundertensileload,Science287 (5453)(2000)637640.

    [2] W.S.Choi,R.S.Hun,ImprovementofinterfacialinteractionviaATRPinpolycarbonate/carbonnanotubenanocomposites,ColloidsSurf.A375 (13)(2011)5560.

    [3] S.P.Liu,S.S.Hwang,J.M.Yeh,K.W.Pan,EnhancementofsurfaceandbulkmechanicalpropertiesofpolycarbonatethroughtheincorporationofrawMWNTs-usingthetwin-screwextrudermixedtechnique,Int.Commun.HeatMassTransf.37 (7)(2010)809814.

    [4] M.Olek,K.Kempa,S.Jurga,M.Giersig,Nanomechanicalpropertiesofsilica-coatedmultiwallcarbonnanotubespoly(methylmethacrylate)composites,Langmuir21 (7)(2005)31463152.

    [5] B.Das,P.K.Eswar,U.Ramamurty,C.N.R.Rao,Nano-indentationstudiesonpolymermatrixcompositesreinforcedbyfew-layergraphene,Nanotechnol-

    ogy

    20 (12)

    (2009)

    125705125710.[6] S.R.C.Vivekchand,U.Ramamurty,C.N.R.Rao,Mechanicalpropertiesofinorganicnanowirereinforcedpolymermatrixcomposites,Nanotechnology

    17 (11)(2006)S344S350.[7] K.H.Kim,W.H.Jo,Astrategyforenhancementofmechanicalandelectricalpropertiesofpolycarbonate/multi-walledcarbonnanotubecomposites,

    Carbon47 (4)(2009)11261134.[8] A.Eitan,F.T.Fisher,R.Andrews,L.C.Brinson,L.S.Schadler,ReinforcementmechanismsinMWCNT-filledpolycarbonate,Compos.Sci.Technol.66 (9)

    (2006)11621173.[9] M.R.Ayatollahi,S.Shadlou,M.M.Shokrieh,Fracturetoughnessofepoxy/multi-walledcarbonnanotubenano-compositesunderbendingandshear

    loadingconditions,Mater.Des.32 (4)(2011)21152124.[10] A.Montazeri,N.Montazeri,Viscoelasticandmechanicalpropertiesofmultiwalledcarbonnanotube/epoxycompositeswithdifferentnanotubecontent,

    Mater.Des.32 (4)(2011)23012307.[11] N.Kumar,P.Jindal,M.Goyal,Mechanicalcharacterizationofmultiwalledcarbonnanotubes-polycarbonatecomposites,Mater.Des.54 (1) (2014)

    864868.[12] P.K.Schelling,S.R.Phillpot,P.Keblinski,Comparisonofatomic-levelsimulationmethodsforcomputingthermalconductivity,Phys.Rev.B65 (14)

    (2002)144306144317.[13] M.S.Green,Markovrandomprocessesandthestatisticalmechanicsoftimedependentphenomenon,J.Chem.Phys.20 (1)(1952)12811295.[14] R.

    Kubo,

    Statisticalmechanical

    theory

    of

    irreversible

    processes.

    I.

    General

    theory

    and

    simple

    applications

    to

    magnetic

    and

    conduction

    problems,

    J. Phys.

    Soc.Jpn.12(1957)570586.[15] S.Lepri,R.Livi,A.Politi,Thermalconductioninclassicallow-dimensionallattices,Phys.Rep.377 (1)(2003)180.[16] P.Jund,R.Jullien,Moleculardynamicscalculationofthethermalconductivityofvitreoussilica,Phys.Rev.B59(1999)1370713711.[17] J.C.Halpin,J.L.Kardos,TheHalpinTsaiequations:a review,Polym.Eng.Sci.16(1976)344352.[18] R.L.Hamilton,O.K.Crosser,Thermalconductivityofheterogeneoustwo-componentsystems,Ind.Eng.Chem.Fundam.1 (3)(1962)187191.[19] A.Buldum,J.P.Lu,Atomicscaleslidingandrollingofcarbonnanotubes,Phys.Rev.Lett.83(1999)50505053.[20] J.Suhr,N.Koratkar,P.Keblinski,P.Ajayan,Viscoelasticityincarbonnanotubecomposites,Nat.Mater.4(2005)134137.[21] R.Hill,Aself-consistentmechanicsofcompositematerials,J.Mech.Phys.Solids13(1965)213222.[22] J.J.Hermans,Theelasticpropertiesoffiberreinforcedmaterialswhenthefibersarealigned,Proc.K.Ned.Akad.Wet.,Ser.B,Phys.Sci.65(1967)19.[23] B.Khatua,N.K.Shrivastava,S.Maiti,Astrategyforachievinglowpercolationandhighelectricalconductivityinmelt-blendedpolycarbonate(PC)/mul-

    tiwallcarbonnanotube(MWCNT)nanocomposites:electricalandthermo-mechanicalproperties,eXPRESSPolym.Lett.7 (6)(2013)505518.[24] G.P.Tandon,G.J.Weng,Theeffectofaspectratioofinclusionsontheelasticpropertiesofuni-directionallyalignedcomposites,Polym.Compos.5

    (1984)327333.[25] P.Jindal,P.Shailaja,P.Sharma,V.Mangla,A.Chaudhury,D.Patel,B.P.Singh,R.B.Mathur,M.Goyal,Highstrainratebehaviorofmulti-walledcarbon

    nanotubespolycarbonatecomposites,Composites,PartB,Eng.45 (1)(2013)417422.[26] A.M.K.

    Esawi,

    M.M.

    Farag,

    Carbon

    nanotube

    reinforced

    composites:

    potential

    and

    current

    challenges,

    Mater.

    Des.

    28 (9)

    (2007)

    23942401.

    http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)00041-8/bib31s1http://refhub.elsevier.com/S1631-0721(15)0004