Memory,CascadesandSpectrainModels of2DMHDandElasticTurbulence XiangFan1,PHDiamond1,LuisChacon2,andHuiLi2 1 UniversityofCalifornia,SanDiego 2 LosAlamosNationalLaboratory ThisresearchwassupportedbytheU.S.DepartmentofEnergy,OfficeofScience,Officeof FusionEnergySciences,underAwardNumberDE-FG02-04ER54738andCMTFO. 3/25/16 TTF2016 1 Why Care? • This is not a traditional study on transport inconfined plasma. It is about thefundamental physicsofmemory andcascades in elastic turbulence systems. Why should we care? • Virtually allmodelsofdrift-Alfven, EMITG,etc. turbulence arebaseduponavorticity equation, Ohm‘s Lawand(usuallymultiple) scalaradvection equations. Theappearance oftheAlfvenwaveintroduces acrucial element ofmemorytothedynamics.Such Alfvenization-induced-memory cansignificantly impact structure formation andtransport inturbulence. • Inthisposter,westudythefundamental physicsof memoryandcascadesinverysimplemodels suchas 2DMHDandelasticturbulence. 3/25/16 TTF2016 2 Why Care? • The2DCahn-HilliardNavier-Stokes (CHNS)model for spinodal decomposition is an example of generalized elastic system, and is analogousto2DMHD. • 2DMHDissimplebutsimilartosomeimportantmodels for transportstudies. 2DCHNS π" π + π£β ' π»π = π·π» + (−π + π . − π + π» + π) π+ π" π + π£β ' π»π = π΅4 ' π»π» + π + ππ» + π π 2DMHD π" π΄ + π£β ' π»π΄ = ππ» + π΄ 1 π" π + π£β ' π»π = π΅ ' π»π» + π΄ + ππ» + π π: π Atypical3-field modelforELM [Xi2012] 3/25/16 TTF2016 3 Why Care? • Spinodal decomposition is also related to zonal flow. Spinodal Decomposition • The pattern formation is because of negative diffusion/negative viscosity. Zonal Flow http://astronomy.nju.edu.cn/~li xd/GA/AT 4/AT41 1/HTM L/AT4 1102. htm 3/25/16 TTF2016 [Porter 1981] 4 Overview • TheCahn-HilliardNavier-Stokes(CHNS)modelforspinodaldecomposition in2Dsymmetricbinaryliquidmixtureand2DMagnetohydrodynamics (MHD)arebothexamplesofgeneralized elasticturbulentsystems. • Since2DMHDturbulence hasbeenwellstudied,itprovidesuswith potentialinsightandguidanceforexploringthephysicsof2DCHNS turbulence. • We compare and contrast the cascades and spectra of the two systems. • [New] Bydirectnumericalsimulation,wefindthatintheelasticrange,the 4 meansquareconcentrationspectrumπ»< ofthe2DCHNSsystemexhibits thesamepowerlaw(−7/3)asthemeansquaremagneticpotential spectrumπ»<= in the inverse cascade regime of2DMHD. • [New] Thekineticenergyspectrumofthe2DCHNSsystemisπΈ<? ~π B. if forcedatlargescale,suggestive ofthedirectenstrophycascadepowerlaw of2DNavier-Stokes (NS)turbulence.It is different from 2D MHD, and the difference could be explained by the difference of interface packing fraction. 3/25/16 TTF2016 5 Outline • Introduction: What is Spinodal Decomposition • The CHNSmodelforspinodaldecomposition • 2D CHNS turbulence compared to 2D MHD turbulence • • • • Basicequations Idealquadraticconserved quantities Cascades Linearelasticwave • Importantlengthscalesandrangesof2DCHNSturbulence • Simulation Results • • • • Simulation Setup Benchmark 4 π»< /π»<=spectrum power law Energyspectrum power law • Conclusion & discussion 3/25/16 TTF2016 6 Introduction:WhatIsSpinodalDecomposition • What is spinodal decomposition: binaryliquidmixtures canpassspontaneouslyfromonemisciblephasetotwo coexistingimmisciblephases followingatemperature drop, and thissecondorderphasetransitioniscalledthespinodal decomposition. • One miscible phase: like water + alcohol. • Two coexisting immiscible phases: like water + oil. • A real life example: the cyclohexanemethanol system, with the critical temperature of spinodal decomposition πE = 45.14β. [Huang et al. 1974] • TheCahn-HilliardNavier-Stokes (CHNS)modelisthe standardmodelforbinaryliquidmixtureundergoing spinodaldecomposition. 3/25/16 TTF2016 7 Introduction:WhatIsSpinodalDecomposition A-rich phase 3/25/16 TTF2016 B-rich phase 8 TheCHNSModelforSpinodalDecomposition • Second order phase transition -> Landautheory. • The order parameter: thelocal relative concentrationfield: π πβ, π‘ β [π= πβ, π‘ − πO πβ, π‘ ]/π • π = −1 means A-rich phase, π = 1 means B-richphase. • The free energy functional: 1 1 π+ + U F π = R ππβ π΄π + π΅π + π»π 2 4 2 + Phase Transition Curvature Penalty Where A and B are coefficients of Taylor expansion, π΄ = π΄: (π − πW ); π is a parameter that describes the strength of the interaction. • Forsimplicity,westudytheisothermalcasewhenthetemperatureisfixed belowTc,andsetB=−A=1: 1 + 1 U π+ F π = R ππβ(− π + π + |π»π|+ ) 2 4 2 3/25/16 TTF2016 9 TheCHNSModelforSpinodalDecomposition • Chemical potential: π = ]4 YZ 4 Y4 = −π + π . − π + π» + π. Fick’sLaw:π½β = −π·π»π. Continuityequation: + π» ' π½β = 0. Combiningtheabovetogether: ]" ππ = π·π» + π = π·π» + (−π + π . − π + π» + π) Cahn-HilliardEquation ππ‘ • Thefluidvelocitycomesin viatheconvectionterm π" = π" + π£β ' π». • ThesurfacetensionenterstheNavier-Stokes Eq asaforce: π»π π" π£β + π£β ' π»π£β = − − ππ»π + ππ» + π£β π • Forincompressiblefluid,itisconvenienttotakethecurlandworkwith vorticity.Finally,wegettheCahn-HilliardNavier-Stokes (CNHS)Equations: π" π + π£β ' π»π = π·π» + (−π + π . − π + π» + π) π+ π" π + π£β ' π»π = π΅4 ' π»π» + π + ππ» + π π With π£β=π§aβ×π»π, π = π» + π, π΅4 = π§aβ×π»π, π4 = π + π» + π. 3/25/16 TTF2016 10 Comparison & Contrast: Basic Equations • The2DCHNSEquations: −π:Negative diffusion term π" π + π£β ' π»π = π·π» + (−π + π . − π + π» + π) π+ π" π + π£β ' π»π = π΅4 ' π»π» + π + ππ» + π π With π£β=π§aβ×π»π, π = π» + π, π΅4 = π§aβ×π»π, π4 = π + π» + π π. :Self nonlinear term −π+ π»+ π:Hyper-diffusion term • The2DMHDEquations: π" π΄ + π£β ' π»π΄ = ππ» + π΄ 1 π" π + π£β ' π»π = π΅ ' π»π» + π΄ + ππ» + π π: π e With π£β=π§aβ×π»π, π = π» + π,π΅ = π§aβ×π»π΄,π = π» + π΄ π΄:Simplediffusion term fg • The force on fluid: hi j π΅4 ' π»π» + π e fg j π΅ ' π»π» + π΄ • Note that the magnetic potential π΄ is a scalar in 2D. 3/25/16 TTF2016 11 Comparison & Contrast: Idealquadraticconservedquantities • 2DMHD • 2DCHNS 1.Energy 1.Energy 2.MeanSquareMagnetic Potential 2.MeanSquare Concentration + + + π π΅4 + π£ ? O πΈ = πΈ + πΈ = R( + )π π₯ 2 2 + + π£ π΅ πΈ = πΈ ? + πΈ O = R( + )π + π₯ 2 2π: π» = = R π΄+ π + π₯ π»4 = R π + π+ π₯ 3.CrossHelicity 3.CrossHelicity π» E = R π£β ' π΅4 π + π₯ π» E = R π£β ' π΅π + π₯ • “Ideal”heremeansπ·, π = 0; π = 0. 3/25/16 TTF2016 12 Comparison&Contrast:Cascades • Turbulencecascadedirectionsaresuggestedbytheabsolute equilibrium distributions. • Thepeakoftheabsoluteequilibrium distributionforeach quadraticconservedquantityisagoodindicatorofthe correspondingcascadedirection. • The spectrum is peaked at high k -> excitation relaxes towards high k -> direct cascade. • The spectrum is peaked at small k -> excitation relaxes towards small k -> inverse cascade. • Thisapproachonlydependsontheidealquadraticconserved quantitiesofthesystem, so wecanfollowasimilarapproachas for2DMHDtoobtainanindicationofthecascadedirectionsin 2DCHNS. 3/25/16 TTF2016 13 Comparison&Contrast:Linear Elastic Wave • Alfven wave in 2D MHD: 1 1 π π =± π×π΅: − π π + π π+ π: π 2 • Linear elastic wave in 2D CHNS: CapillaryWave: Air Water π+ 1 π π =± π×π΅4: − π πΆπ· + π π+ π 2 Where • Thelinear elastic wave in 2D CHNS is like a capillary wave:it only propagates along the boundary of the two fluids, where the gradient of concentration π΅4: ≠ 0. Surfacetensiongenerates restoringforce. • The wave is similar to Alfven wave: they have similar dispersion relation; they both propagates along π΅: field lines; both magnetic field and surface tension act like an elastic restoring force. • Important difference: B fills the whole space; π΅4: is large only in the interface regions. 3/25/16 TTF2016 14 Importantlengthscalesandranges • Hinze scale: the balance between turbulentkineticenergyand jβ¨qi β© surfacetensionenergy: s/t ~1 where π is surface tension and π= u 8/9π [Kendon et al. 2001]. • Inthe2DNSdirectenstrophycascaderegime,thevelocity qi distributionis < ~πz+/.π{B.. So u π Be/. B+/} πΏ{ ~( ) πz π • We define thescalesbetweenπΏ{ anddissipationscale πΏ] to be theelasticrange,wheretheblobcoalescenceprocessdominates. • Define a dimensionless number: • tu t~ 3/25/16 tu t~ j h Be/e• = π»π~( )Be/. π Be/+ πz = π»π β« 1 isrequiredtoformalong enoughelasticrange. TTF2016 15 Importantlengthscalesandranges π» 4 Spectrum 4 π»< Hydrodynamic Range π•‚ ElasticRange π{ π] π • Intheelastic rangeofthe2DCHNSsystem,theblobcoalescenceprocessis analogoustothemagneticfluxblobcoalescence processin2DMHD. • Theformerleadstotheinversecascadeofπ» 4,andthelatterleadstothe inversecascadeofπ» = . • Intheelastic rangeofthe2DCHNSsystem,surfacetensioninduceselasticity andplaysamajorroleindefiningarestoringforce.Similarly,in2DMHD,the magneticfieldinduceselasticityandmakeMHDdifferentfromapurefluid. • The2DCHNSsystemismoreMHD-likeintheelasticrange. 3/25/16 TTF2016 16 Simulation Setup • The PIXIE2D code [Chacon 2002,2003] is used tosimulatethe system.PIXIE2D originally solves the 2D MHD equation, and now is modified to be able to solve the Cahn-Hilliard Navier-Stokes (CHNS) equations, too.Itis a Direct Numerical Simulation (DNS) code that solves thefollowingequationsinrealspace: π" π + π£β ' π»π = π·π» + −π + π . − π + π» + π + πΉ4 π+ π" π + π£β ' π»π = π΅4 ' π»π» + π + ππ» + π + πΉ„ π With π£β=π§βa ×π»π, π = π» + π, π΅4 = π§βa×π»π, π4 = π + π» + π • Initialcondition:π ineachcellis assignedto1 or−1 randomly; π = 0 everywhere. • Boundarycondition:doubly periodic. • Externalforceforeitherπ orπ:anisotropichomogeneous force thathasawavenumberπ•‚ : 3/25/16 TTF2016 17 Benchmark • We verified the length scale growth is πΏ π‘ ~π‘ +/. ; and the growth can be arrested. The saturation length scale is consistent with the Hinze scale (dashed line). 3/25/16 TTF2016 18 4 = π»< /π»< spectrum power law: -7/3 CHNS MHD π= Inverse cascade of π» 4 Inverse cascade of π» = 3/25/16 TTF2016 19 4 = π»< /π»< spectrum power law: -7/3 • Assumingaconstantmeansquaremagneticpotential dissipationrateπ{= ,according totheAlfvenic equipartition (πβ¨π£ + β© ∼ fe β¨π΅ +β©), the time scale for the g = = decay of π» (π{= ~π» /π) can be estimated by π~(π£π)Be ~(π΅π)Be. • Define the spectrum to be π» = = ∑< π»<= ~ππ»<= , so ‰ ‰ Š i • {‹ = i i = Š i = i π΅~ππ΄~π(π» ) ~(π»< ) π . Therefore π{= ~ ~(π»< ) π . Œ +/. BΕ½/. = π»< ~π{= π • Similarly, we can obtain the π»<4spectrum by assuming the elastic equipartition (πβ¨π£ + β© ∼ π + β¨π΅4+ β©): 4 +/. BΕ½/. π»< ~π{4 π 3/25/16 TTF2016 20 4 = π»< /π»< spectrum power law: -7/3 • The−7/3powerisrobust.Itdoesnotchangewiththe magnitudeofexternalforcing,aslongastheseparation between theHinzescaleandthedissipationscaleis maintained,sotheelasticrangeislongenough. π• 3/25/16 TTF2016 21 Energy Spectrum Power Law π• π• • πΈ<? ~π B. in 2D CHNS turbulence; on the other hand, it is well known that πΈ<? ~π B./+ in MHD turbulence. • The -3 power law is consistent with the direct enstrophy cascade in 2D NS turbulence. • The -3/2 power law comes from the Alfven effect (Iroshnikov-Kraichnan Theory). 3/25/16 TTF2016 22 Energy Spectrum Power Law • πΈ<? ~π B. for 2D CHNS; πΈ<? ~π B./+ for 2D MHD. Why? Why not -3/2? • Thisinitiallysurprisingresultisplausiblebecauseinthe2DCHNSsystem, π΅4 vanishesinmostregions. Back reaction is apparently limited. • On the other hand, themagneticfieldsinMHDarenotlocalizedatspecific regions,andAlfvenwavescanbeeverywhere. • Definetheinterfacepackingfractionπ tobetheratioofmeshgridnumber where |π΅4 | > π΅4’“” (or |π΅| > π΅’“”) tothetotalmeshgridnumber. MHD 3/25/16 CHNS TTF2016 23 Conclusion & Discussion • Comparison: • Contrast: 3/25/16 TTF2016 24 Conclusion&Discussion • Thestudy ofelasticturbulencegives us useful lessons. 4 • Theπ»< spectrum(inversecascadeofπ» 4)isrobust. It is an analogue to π»<=spectrum (inversecascadeof π» =) in 2D MHD. On the other hand, the energy spectrum is NOT so robust. This implies we should perhaps pay more attention on other spectra than the energy spectrum. • Thespatialrepresentations- i.e.trackingcontours ofbubblesurfaces- may be more revealing way of representingtheturbulence thetraditionalpower lawspectrum. 3/25/16 TTF2016 25