You are on page 1of 36

ASDEX Upgrade

MaxPlanckInstitut fr Plasmaphysik

2D Fluid Turbulence

Florian Merz Seminar on Turbulence, 08.09.05

2D turbulence?
strictly speaking, there are no two-dimensional ows in nature approximately 2D: soap lms, stratied uids, geophysical ows, magnetized plasmas

2D turbulence?
a simplied situation (compared to 3D), more accessible to theoretical, experimental and computational approaches, interesting for developing and testing general ideas about turbulence much easier to visualize than 3D-turbulence interesting new phenomena (e.g. dual cascade)

Outline

Basic equations Cascades in 2D Coherent structures

Basic equations: velocity

2D-Navier-Stokes equations for an incompressible uid:

Dv Dt

= t + v v =0

v = 1 p + fext + 2v

where v is in the (x,y) plane, (v z = 0). For viscosity = 0 and fext = 0 these equations are called Euler equations.

Basic equations: vorticity

Taking the curl of the NS equations and discarding the zero x and y components of the equation gives D Dt = t + v v z.

= g + 2

for the vorticity =

D If g = ( fext) z = 0 and = 0 (Euler equation), we have Dt = 0 vorticity is conserved

Basic equations: energy and enstrophy


Important quantities:

mean energy E = 1 u2 2 enstrophy Z = 1 2 2 (mean square vorticity)

dE = 2Z dt dZ = 2 ( dt

)2

in 2D with curl free forcing, energy and enstrophy can only decrease with time, they are conserved in the inviscid case ( = 0) for 0 we get dE 0, energy is a robust invariant in 2D dt dZ does not necessarily go to zero for 0, enstrophy is a fragile dt invariant (dissipation anomaly!)

Dissipation anomaly
as time evolves, vorticity patches get distorted by background velocity and generate smaller and smaller laments the vorticity gradient increases and dZ = 2 ( )2 becomes sizeable dt dissipation even for 0: fragile invariant in the enstrophy cascade, the dissipation rate is independent of but depends only on the enstrophy transfer rate

only determines the enstrophy dissipation scale, not the enstrophy dissipation rate

Exact results
For forced, isotropic and homogenous turbulence:

for small scales: Corrsin-Yaglom-relation for passive tracers dS2 4 DL(r) 2 = r dr 3

(DL(r) = (vL(x) vL(x + r))((x) (x + r))2 , S2(r) = ((x) (x + r))2 ) , = ( )2 )

for large scales: 2D-analogue of the Kolmogorov relation (4/5-law) 3 F3 (r) = r 2

(F3 (r) = (vL(x) vL(x + r))3 , sign of the prefactor reversed!

= ()2 = 2Z)

Cascades

Direction of the energy/enstrophy transfer

In spectral space, the expressions for energy and enstrophy read E = Z = E(k, t)dk k2 E(k, t)dk

E1 ... k1 Z1 k2

E3 Z3 k3 ...

E2 = E1 + E3 ,

E2 Z2

Z2 = Z1 + Z3 .

Energy and enstrophy conservation for three Fourier modes k1 , k2 = 2k1 , k3 = 3k1

Direction of the energy/enstrophy transfer

E1 + E2 + E3 = 0
2 2 2 k1 E1 + k2 E2 + k3 E3 = 0

with Ei = E(ki , t2 ) E(ki, t1 ). Combining the equations gives

5 E1 = E2 8 5 2 2 k1 E1 = k2 E2 32

3 E3 = E2 8 27 2 2 k3 E3 = k2 E2 32

enstrophy goes to higher k (direct enstrophy cascade),

energy goes to lower k (inverse energy cascade)

Direction of the energy/enstrophy transfer

Alternatively:

E = Z =

E(k, t)dk k2 E(k, t)dk

an evolution of E(k, t) to larger k conicts with the boundedness of enstrophy

E(k, t) must evolve towards small wave numbers / large

scales

KBL theory: dual cascade


Kraichnan, Batchelor, Leith proposed the existence of a dual cascade in steady state turbulence (1968):

energy/enstrophy is constantly injected at some intermediate k i direct enstrophy cascade to higher k dissipation scale kd = (/ 3)1/6 (equivalent to Kolmogorov microscale in 3D)

inverse energy cascade to lower k condenses in the lowest mode (for bounded domain) / is stopped by Ekman friction (v-term in NS-equation) at (kE = 3/ )1/2 E(k) is stationary, the transfer rates of energy and enstrophy far from the dissipation scales are independent of k (self-similarity)

KBL theory: inertial ranges


inertial range of the energy cascade: for kE spectrum can only depend on . Dimensional analysis: k = [L]1 ; E(k) = [L]3 [T ]2 ; k ki, the energy

= [L]2 [T ]3 E(k) = C 2/3k5/3 k kd ): the energy

inertial range of the enstrophy cascade (ki spectrum can only depend on . Dimensional analysis:

k = [L]1 ; E(k) = [L]3 [T ]2 ; = [T ]3 E(k) = C 2/3 k3 C, C constant and dimensionless.

zero enstrophy transfer in the energy inertial range, zero energy transfer in the enstrophy inertial range

KBL: Inertial ranges in steady state turbulence

Experiments: soap lms


vertically owing soap lms are approximately 2D (thickness variations of about 10% - condition of incompressibility is slightly violated)

turbulence is generated by grids/combs inserted in the ow

Experiments: soap lms


Energy spectrum in soap lm

Correction to the energy spectrum


the k3 -spectrum of the enstrophy cascade gives rise to inconsistency: infrared divergence of for the (k-dependent) enstrophy transfer rate (k) for ki 0 reason: contributions of larger structures to the shear on smaller structures (nonlocality in k-space).

Kraichnan (TFM-closure approximation): logarithmic correction to restore constant transfer rate


2/3 k3 ln k E(k) = C ki 1/3

Correction to the energy spectrum

attempts to measure the corrections are being made: k 3 E(k) and (enstrophy ux)/ for various Reynolds numbers (direct numerical simulation [DNS] results)

Cascades: 2D vs. 3D

Vorticity equation in 3 dimensions (analogous to MHD kinematic equation for B): D = ( )v + g + 2 Dt the additional vortex-stretching term changes the behaviour signicantly:

gradients in the velocity eld stretch embedded vortex tubes as the cross section decreases, the vorticity increases

Cascades: 2D vs. 3D

enstrophy in 3D is not conserved even for = 0 but increases with time (in 2D: fragile invariant) no enstrophy cascade in 3D! energy in 3D follows dE = 2Z and is a fragile invariant (in 2D: dt robust invariant)

energy cascades to smaller scales in 3D (direct cascade), in 2D to large scales (inverse cascade) E(k) has the same k 5/3 -dependence in the inertial range of the energy cascade

Cascades: 2D vs. 3D

experimental results for grid turbulence in 3D/2D

Coherent structures

Coherent structures

physical and numerical experiments: long lived vortical structures (lifetime turnover time) spontaneously emerging from the turbulent background

these coherent structures alter the cascading behavior especially important in freely decaying turbulence (they can be inhibited / destroyed in forced systems)

clear denition/identication of coherent structures dicult, several competing methods: e.g. simple threshold criteria, Weiss criterion, wavelet decomposition..

Coherent structures: axisymmetrisation

experiments: elliptical structures are not stable but become circular exact theoretical result: circular patches of uniform vorticity are nonlinearly stable

Coherent structures: vortex merging


experiments: if like sign vortices of comparable strength get too close, they merge (axisymmetrization)

theory: analytical solution for the merging of two identical vortices

Coherent structures: vortex break-up

experiments: the strain caused by the stronger vorices distorts weaker vortices up to destruction vorticity adds to background vorticity

Coherent structures: time evolution


by vortex merging and break-up, the coherent structures in a freely decaying system become fewer and larger

observables: evolution of vortex density , typical radius a, intervortical distance r, extremal vorticity ext

Universal decay theory for dilute vortex gas


empirical approach by Carnevale et al. assumption: two invariants
2 E exta4, contributions outside vortices negligible

vorticity extremum ext of the system (observation) length scale l = E/ext, time scale = 1/ext

Dimensional reasoning gives = l2 g(t/ ). Assumption g(t/ ) = (t/ ) gives ( is to be measured) r l(t/ )/2 , l2(t/ ) , a l(t/ )/4 , v E, Z 2(t/ )/2

Universal decay theory: comparison with DNS


left: decay law for the number of vortices right: inverse vortex density, intervortical distance, size, extremal vorticity (lines for = 0.75)

Decay of vortex populations

vortex merging and vortex break-up lead to ever larger and fewer coherent structures

the system evolves towards a nal dipole

Intermittency in 2D
experimental results: no intermittency in 2D turbulence no theoretical explanation yet PDFs for longitudinal, transverse velocity increment (energy cascade) and the vorticity increment (enstrophy cascade) for several scales

v = vL (x)vL (x+r)

v = vT (x)vT (x+r)

= (x)(x +r)

Intermittency in 2D
F hyperatness H2n (l) = F2n(l) , Fn(l) = v (l)n (energy cascade) n 2(l)

and structure functions of vorticity Sn(l) = (l)n (enstrophy cascade)

no intermittency, slight deviations from gaussianity are assumed to stem from coherent structures

Summary

the existence of a dual enstrophy-energy cascade (KBL-theory) is experimentally conrmed

coherent structures play an important role (especially in decaying turbulence) and modify the energy spectrum predicted by KBL-theory

there is no intermittency found in experiments several systems of interest (e.g. geophysical ows, magnetized plasmas) are approximately 2-dimensional - results of 2D uid turbulence are applicable

Further reading
General 2D turbulence: P.A. Davidson, Turbulence, Oxford University Press (2004) M. Lesieur, Turbulence in Fluids, Kluwer (1997) U. Frisch, Turbulence, Cambridge University Press (1995) P. Tabelling, Two-dimensional turbulence: a physicist approach, Phys. Rep. 362, 1-62 (2002) Cascade classics Kraichnan, Inertial Ranges in Two-Dimensional Turbulence, Phys. Fluids 10, 1417 (1967) Leith, Diusion Approximation for Two-Dimensional Turbulence, Phys. Fluids 11, 1612 (1968) Batchelor, Computation of the Energy Spectrum in Homogenous Two-Dimensional Turbulence, Phys. Fluids 12, II-233 (1969)

You might also like