Simulations and Theory in FZD, February 25, 2010 Theoretical and Numerical Problems of Magnetohydrodynamics Frank Stefani Institute of Safety Research, Department Magnetohydrodynamics with thanks to: Gunter Gerbeth, André Giesecke, Uwe Günther, Thomas Gundrum, Thomas Wondrak, Mingtian Xu, Agris Gailitis (Riga), Karl-Heinz Rädler (Potsdam), Günther Rüdiger (Potsdam), Rainer Hollerbach (Leeds), Oleg Kirillov (Darmstadt), Sasa Kenjeres (Delft)... Outline of the talk • Motivation: Magnetic fields in the cosmos...and in the lab • Basic theory: Induction equation and Navier-Stokes equation • Simulations: From 3D to 0D...and slowly back • Prospects: New experiments and numerical challenges Motivation: Cosmic magnetic fields... ...are self-excited by the hydromagnetic dynamo effect ...help in cosmic structure formation (stellar systems, black holes) by virtue of the magnetorotational instability (MRI) in accretion disks Motivation: Cosmic magnetism in the liquid metal lab... Dynamo experiments: Riga, Karlsruhe, Cadarache, Madison, Maryland, Perm... 1 H1 H2 F 2 H3 3m 3 H4 4 H5 H6 5 0.8 m MRI related experiments: Maryland, Princeton, Socorro, Grenoble, Rossendorf Basic theory: What is left out...? ...A LOT !!!...sorry • ...plasma foundation of MHD (kinetic models, Boltzmann (Vlassov) equation, Landau’s collision integral, Particle-in-cell techniques...) • ...derivation of the electrical conductivity of liquid metals (Ziman’s formula, Born approximation, electron-ion pseudo-potentials...) We restrict ourselves to the one-fluid description of MHD, and take the conductivities as given! Basic theory: Induction equation j: E: σ: electric current density; electric field; electrical conductivity; B: v: µ0 : magnetic field velocity of the fluid magnetic permeability Pre-Maxwell equations: Ampère’s law Faraday’s law: Magnetic field is source free: Ohm’s law in moving conductors: =⇒ Induction equation: ∇ × B = µ0(j + Ḋ) ∇ × E = −Ḃ ∇·B=0 j = σ(E + v × B) Ḃ = ∇ × (v × B) + µ10σ ∆B Basic theory: Induction equation • Induction equation describes two competing effects – Diffusion (decay) of magnetic field: Ḃ ∼ 1 µ0 σ ∆B – ”Stretch, Twist and Fold” (production): Ḃ ∼ ∇ × (v × B) • Ratio of production to decay: ∼ magnetic Reynolds number Rm := µσvL • Rm must be large enough (Rm > 10) for self-excitation to occur Example: Rm = 10 means for liquid sodium: vL = 1 m2/s Basic theory: The interaction between B and v • Induction equation for evolution of B under the influence of v: 1 Ḃ = ∇ × (v × B) + ∆B µ0 σ • Navier-Stokes-equation for evolution of v, including the (back-)reaction of B on v v̇ + (v · ∇)v = − 1 ∇p + (∇ × B) × B + ν∆v + fextern ρ µ0 ρ Basic theory: Dimensionless parameters Reynolds: Magnetic Reynolds: Re Rm Hartmann: Ha Magnetic Prandtl: Rossby: Ekman: Pm Ro Ek = RV /ν = RV µ0σ = Re Pm q σ = BR ρν = νµ0σ =: ν/λ = 21 Ωr dΩ dr = ν/(2ΩR2) Basic theory: Self-excitation B0 ω vxB jxB v j 0 B1 Inverse amplification B /B Basic theory: The disk dynamo as a simple model 1 0.8 0.6 0.4 0.2 0 Self−excitation ↓ 0 0.2 0.4 0.6 0.8 1 Normalized rotation rate ν L/R Basic theory: Self-excitation in homogenous media Right hand rule (a). α-effect (b). α2-dynamo (c) (b) (a) (c) B B10 v I I2 I1 B 1+ B I B2+ B B20 Basic theory: From kinematic to saturated regime Basic theory: Some distinctions for numerical treatment • Degree of coupling: Kinematic, inductionless, or fully coupled system • Spatial dimensions: 0D (dispersion relation, WKB), 1D (e.g. in r), 2D (e.g. in r, z), 3D (e.q. in r, ϕ, z) • Treatment of small scales: Direct numerical simulations, Mean-field models, hyperdiffusivities, turbulence models (Large eddy simulations (LES), Reynolds averaged Navier-Stokes (RANS)) • Numerical methods: Finite differences, Finite elements, Finite volumes, Spectral elements, Integral equations, Boundary element methods Simulations: The state of the art Roberts and Glatzmaier (Nature 1995): First fully 3D simulation of an Earth’s magnetic field reversals Simulations: The state of the art Kageyama et al. (Nature 2008): Simulation on a 2 × 511 × 514 × 1538 grid on 4096 processors (Earth Simulator): Vorticity ωz for Ek=2.3 10−7 (left) and Ek=2.6 10−7 (right) Simulations: Stepping back from 3D to 0D Simulation (MRI): Accretion disks in the cosmos Jets Black hole Credit: NASA/ESA (Hubble) Jet Accretion disk Black Hole Companion Accretion disc Credit: NASA Supermassive BH Jet X-ray binaries Young star (HH 30) Accretion disk Young stars Simulations (MRI): The problem with Rayleigh’s criterion • Central problem: How do accretion disks work? Central object (Star, Black Hole) • Necessary outward angular momentum transport is not explainable by molecular viscosity. 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 0000000000000000000000000000000000 1111111111111111111111111111111111 Accretion disk • Turbulence could explain it. However: Kepler-rotation is hydrodynamically stable (angular momentum increases outward)! Ω (r)~r −3/2 2 1/2 L(r)= Ω(r)r ~r Simulations (MRI): A (too) simple spring model • In Keplerian disks, inner masses orbit more rapidly than outer mass Magnetic field acts similar as a massless spring • Inner masses slow down =⇒ reduce angular momentum =⇒ move to lower orbit • Outer masses speed up =⇒ increase angular momentum =⇒ move to higher orbit • Spring tension increases =⇒ runaway process Ω = i r: Ω Ω o 11111111111111111111111111111111111111 00000000000000000000000000000000000000 • Balbus/Hawley 1991: Revealed the astrophysical relevance of MRI. ha ek dr an Ch v− ho ik el V • Flow between inner and outer cylinders with radii ri and ra and angular frequencies Ωi and Ωo 3/2 ro o Ω 11111111111111111111111111111111111111 3/2 = 00000000000000000000000000000000000000 00000000000000000000000000000000000000 ir i 11111111111111111111111111111111111111 00000000000000000000000000000000000000 11111111111111111111111111111111111111 Ω : always 00000000000000000000000000000000000000 11111111111111111111111111111111111111 r le 00000000000000000000000000000000000000 11111111111111111111111111111111111111 2 p stable 00000000000000000000000000000000000000 11111111111111111111111111111111111111 e r K o 00000000000000000000000000000000000000 11111111111111111111111111111111111111 2=Ωo 00000000000000000000000000000000000000 11111111111111111111111111111111111111 00000000000000000000000000000000000000 11111111111111111111111111111111111111 r i i MRI Ω 00000000000000000000000000000000000000 11111111111111111111111111111111111111 : 00000000000000000000000000000000000000 11111111111111111111111111111111111111 igh e l 00000000000000000000000000000000000000 11111111111111111111111111111111111111 ay 00000000000000000000000000000000000000 11111111111111111111111111111111111111 R 00000000000000000000000000000000000000 11111111111111111111111111111111111111 00000000000000000000000000000000000000 11111111111111111111111111111111111111 always 00000000000000000000000000000000000000 11111111111111111111111111111111111111 00000000000000000000000000000000000000 11111111111111111111111111111111111111 unstable 00000000000000000000000000000000000000 11111111111111111111111111111111111111 as • E.P. Velikhov, July 1, 1958, at Riga MHD conference: prove of MRI for Taylor-Couette flow. o Simulations (MRI): Some history Ωi Simulations (MRI): Standard MRI and helical MRI • Experiments on MRI with large Rm (pure axial fields) in Maryland, Princeton, New Mexico, Moscow • Hollerbach and Rüdiger (2005): Helical MRI: Bz replaced by Bz + Bϕ: Recrit ∼ 103 instead of 106 • Potsdam ROssendorf Magnetic Experiment (PROMISE) InStability • The ”only” Problem: Axial current ∼ 5000 Amp. 1111 0000 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 Ωi 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 B 1111 0000 1111 ϕ 0000 1111 0000 1111 00000000 11111111 00000000 11111111 0000 1111 00000000 11111111 00000000 11111111 0000 1111 00000000 11111111 00000000 11111111 Ωo Bz Simulations (MRI): PROMISE 2 Outer ring Inner ring Inner rings rotate with the inner cylinder, outer rings with the outer cylinder Simulations (MRI): Numerics with increasing complexity • Local analysis (WKB): Liu et al. 2006, HMRI characterized as a weakly destabilized inertial oscillation • 1D global analysis: Hollerbach and Rüdiger 2005, Rüdiger et al. 2005, Priede et al. 2007, own computations • 1D global analysis with complex wavenumbers: Priede and Gerbeth 2009, convective versus absolute instability (vanishing group velocity) • 2D simulations: Szklarski and Rüdiger 2006, Szklarski 2007, 2008, Liu et al. 2007, 2008 Simulations (MRI): PROMISE 2 - Increasing Icoil (Ha) fin = 0.06 Hz. µ = 0.27, Irod = 8000 A Fingering of angular momentum. Simulation by Jacek Szklarski. Simulations (MRI): Comparison measurements/numerics fin = 0.06 Hz. µ = 0.26, Icoil = 60 A. Irod = 6000 A Measurement Simulation (J. Szklarski) For many more results see PRL 2006, NJP 2007, PRE 2009 Simulations (MRI): An strange paradox R. Hollerbach 2008 Rayleigh line – Continuous and monotonic transition between standard MRI (SMRI) and helical MRI (HMRI) – HMRI is a destabilized inertial oscillation while SMRI is a destabilized magneto-Coriolis wave log(Re c ) • β = Bφ(ri)/Bz Simulations (MRI): Dispersion relation from WKB γ 4 + a1γ 3 + a2γ 2 + (a3 + ib3)γ + a4 + ib4 = 0 a1 = a2 = a3 = a4 = b3 = 2 „ √ Pm + √ 1 Pm « , a21 2 2 2 2 + 2(1 + Ha∗ ) + 4β ∗ Ha∗ + 4Re∗ Pm(1 + Ro), 4 √ ∗2 ∗2 ∗2 ∗2 a1 (1 + Ha ) + 2a1β Ha + 8Re (1 + Ro) Pm, “ ” ∗2 2 ∗2 ∗2 ∗2 ∗2 ∗2 1 + Ha + 4β Ha + 4Re + 4Re Ro(PmHa + 1), ∗ ∗2 −8β Ha ∗√ Re Pm, ∗ ∗2 b4 = −4β Ha ∗ Re (2 + (1 − Pm)Ro). Simulations (MRI): Resolving the paradox HMRI appears at an EXCEPTIONAL POINT of the SMRI branch with the inertial wave branch, at small but finite Pm (Kirillov and F.S., APJ 2010) Simulations (MRI): Resolving the paradox Another view on the exceptional point Simulations (MRI): Resolving the paradox Still, SMRI and HMRI are continuosly connected! Simulations: Going from 0D to 1D Simulations (reversals): Polarity reversals during the last 150 Myr • Reversal rate between nearly zero (during superchrons) and ∼4/Myr presently. Last reversal occurred 780 kyr ago. Mean number of field reversals per Million years Sequence of normal (black) and reversed (white) polarity Simulations (reversals): Asymmetry, Clustering, Stochastic resonance Last five reversals • Slow decay, but very fast recreation (∼ 5 kyr) of dipole with opposite polarity Brunhes-Matuyama Upper Jaramillo Lower Jaramillo Upper Olduvai Lower Olduvai Average 10 VADM [1022 A m2] • Saw-tooth shape of reversals 8 6 4 2 from: Valet et al. 2005 0 -80 -60 -40 -20 0 Time from transition [kyr] 20 Simulations (reversals): Questions to address • Is there a generic and universal mechanism of reversals? • What is a minimum model to explain the three reversal features? • How to tailor dynamo experiments so that they show reversals? • Can reversal features be used to constrain essential parameters of the geodynamo? PRL 2005, EPSL 2006, ..., Inverse Probl. 2009 Simulations (reversals): A simple reversal model • Mean-field dynamo source is the helical turbulence parameter part α • Induction equation: Ḃ = ∇ × (αB) + (µ0σ)−1 ∆B. Decomposition in toroidal and poloidal field components: B = −∇ × (r × ∇S) − r × ∇T • α is assumed spherically symmetric (this is not true for the Earth!) =⇒ Two partial differential equations for sl(r, t) and tl(r, t) ∂sl ∂t ∂tl ∂t = = 1 ∂2 l(l + 1) (rs ) − sl + α(r, t)tl , l 2 2 r ∂r r 1∂ ∂ d l(l + 1) (rtl) − α(r, t) (rsl) − [tl − α(r, t)sl] . r ∂r ∂r dr r2 Simulations (reversals): Implementing back-reaction • Algebraic α-quenching with averaged magnetic energy α(r, t) = C 1+Q 2s21 (r,t) r2 + r12 αkin ∂(rs1 (r,t)) ∂r 2 + t21(r, t) +ξ1(t) + ξ2(t) r 2 + ξ3(t) r 3 + ξ4(t) r 4 , • Noise: hξi(t)ξj (t + t1)i = D 2(1 − |t1|/τ )Θ(1 − |t1|/τ )δij . • C - normalized dynamo number, D - noise amplitude, Q - quenching parameter, τ - noise correlation time Simulations (reversals): Relaxation oscillations van der Pol oscillator ∂ 2y ∂t2 = µ(1 − y 2 ) ∂y −y ∂t equivalent to: ∂y ∂t ∂z ∂t = z = µ(1 − y )z − y Spherically symmetric α2 dynamo with αquenching and noise ξ(r, τ ) (intensity D) ∂sl ∂τ ∂tl ∂τ = = 2 α(r, τ ) = 1 ∂2 l(l + 1) (rs ) − sl + α(r, τ )tl , l r ∂r 2 r2 „ « 1 ∂ ∂ ∂ (rtl ) − α(r, τ ) (rsl ) r ∂r ∂r ∂r − l(l + 1) [tl − α(r, τ )sl ] r2 C αkin (r) + ξ(r, τ ) 0 1 + Emag (r, τ )/Emag 2 y’ 1 0 -1 -2 µ=0 µ=1 µ=2 -3 -4 -2 -1.5 -1 -0.5 0 y 0.5 1 1.5 2 0.8 0.6 0.4 α 2 dynamo 3 1 s’(r=1) 4 van der Pol oscillator Simulations (reversals): Relaxation oscillations 0.2 0 -0.2 -0.4 -0.6 C=6.8 C=7.0 C=7.237 -0.8 -1 -0.4 -0.3 -0.2 -0.1 0 s(r=1) 0.1 0.2 0.3 0.4 10 15 25 20 25 0 5 10 15 2 0 -2 0 5 10 15 20 t 20 t 35 40 µ=2 30 35 40 µ=5 t 2 0 -2 30 25 30 35 40 µ=20 25 30 35 40 van der Pol oscillator s(r=1) 0 0.4 0.2 0 -0.2 -0.4 0.4 0.2 0 -0.2 -0.4 1 0.5 0 -0.5 -1 0 0 2 2 2 4 6 t C=7.2 t C=7.237 4 6 4 6 8 10 8 10 8 10 8 10 α 2 dynamo 5 20 t s(r=1) 15 -0.1 s(r=1) 10 2 0 -2 0 y 5 C=6.8 0 s(r=1) y 0 y 0.1 µ=0 2 0 -2 van der Pol oscillator y Simulations (reversals): Relaxation oscillations t C=7.24 0 2 4 6 t αkin(r) = C · (1 − 6r 2 − 5r 4) Simulations (reversals): Spectral features 1.4 40 1.2 20 Growth rate 1 α(r) 0.8 0.6 0.4 0 n=1 -20 n=2 -40 n=3 -60 n=4 0.2 -80 0 -100 0 0.2 0.4 0.6 0.8 1 r Constant α along r yields .... n=5 0 2 n=6 4 6 8 C ... a purely real spectrum 10 Simulations (reversals): Spectral features 15 40 10 20 Growth rate 5 α(r) 0 -5 0 n=1 -20 n=2 -40 n=3 -15 -60 n=4 -20 -80 -25 -100 -10 0 0.2 0.4 0.6 0.8 r Varying α along r yields .... 1 n=5 0 0.5 n=6 1 C n=7 1.5 ... also complex eigenvalues (remember Uwe Günthers talk on PT-symmetric QM !) 2 Growth rate Simulations (reversals): The basic mechanism Noise triggered Unstable jump fix point Stable fix point 0 y r o t la il c Os y or t a l Sign change and field recovery il c s o − n No Self−accelerating field decay −o Non sc ry illato Rapid α −quenching Exceptional point Simulations (reversals): Parameter fit for the geodynamo Three functionals to be minimized simultaneously: 1. Mean deviation from averaged asymmetric reversal curve (Valet 2005) 2. Mean deviation from clustering curve (Cande/Kent 95, Carbone 2005) 3. Mean deviation from residence time distribution curve (Consolini/De Michelis 2003) Simulations (reversals): Parameter fit for the geodynamo • Downhill simplex method is used to determine the following parameters: 1. C - Dynamo strength; C/Ccrit - Supercriticality 2. δ - Shift parameter 3. ǫ - Relative strength of periodic forcing (Milankovic cycle 95 kyr) 4. D - Noise intensity 5. τdif f - Effective diffusive time scale: τdif f = µ0σturbR2 (β-effect) Simulations (reversals): Parameter fit for the geodynamo • Various results depending on relative weight of individual functionals • Parameters of three versions resulting from the downhill simplex inversion. Version 1 Version 2 Version 3 C 57.0 96.8 147.8 δ -0.03 -0.16 -0.18 Ccrit 13.0 8.99 8.73 C/Ccrit 4.4 10.8 16.9 D 6.3 8.0 7.8 Td 86.8 kyr 63.7 kyr 55.1 kyr ǫ 0.133 0.123 0.094 #/Myr 2.05 3.07 2.91 VADM [1022 A m2] Simulations (reversals): 1. Asymmetry 8 7 6 5 4 3 Real Version 1 2 Version 2 1 Version 3 0 -60 -50 -40 -30 -20 -10 0 Time from transition [kyr] 10 20 Simulations (reversals): 2. Clustering 1 P(h>H) 0.8 0.6 Real 0.4 Version 1 Version 2 0.2 Version 3 Poisson 0 0 0.2 0.4 0.6 H 0.8 1 Probability density [1/Myr] Simulations (reversals): 3 - Residence time distribution 4 Real Version 1 Version 2 Version 3 3 2 1 0 0 0.2 0.4 0.6 0.8 1 τ [Myr] Stochastic resonance with Milankovich cycle of Earth’s orbit eccentricity Simulations (reversals): Best results for ... • Supercriticality of the geodynamo: Factor 10 • Relative strength of periodic forcing: around 10 per cent • Diffusion time: 64 kyr, i.e reduction by a factor 3.5 compared to 225 kyr resulting from molecular conductivity. This is in rough agreement with recent numerical simulations (Schrinner et al. 2007). Simulations (reversals): Conclusions for reversals • Reversals are (very likely...) noise triggered relaxation oscillations in the vicinity of exceptional points of the spectrum of the dynamo operator. • Highly supercritical dynamos self-tune into a reversal-prone state (”edge of chaos”?, ”self-organized criticality”?, ”punctuated equilibrium”?). • Asymmetry, clustering, and stochastic resonance can be used as proxies for constraining essential parameters of the geodynamo. • The estimated conductivity reduction, if anisotropic, would be important for dynamo simulations (selection of axial or equatorial dipole, etc.) Simulations: Going from 1D to 2D and 3D Simulations (dynamo experiments): Riga • 2 m3 sodium • Propeller (1). Helical flow (2). Back flow (3). Sodium at rest (4). 1 H1 H2 F H3 3m • Two motors (200 kW) produce flow up to 20 m/s. 2 3 H4 4 H5 H6 5 0.8 m Simulations (dynamo experiments): Riga • Optimized flow (maximum helicity!) • Self-excitation at propeller rotation rate of ∼ 1900 rpm. • Field structure rotates slowly (1-2 Hz) around vertical axis. • Axial field component: up to 120 mT Simulations (Riga): A simple saturation model • Weak-field dynamo: Linearization of Navier-Stokes-equation, Perturbation δv (and δp) with respect to measured unperturbed velocity v̄ ∂δv ∇δp 1 + (v̄ · ∇)δv + (δv · ∇)v̄ = − + (∇ × B) × B + ν∆δv . ∂t ρ µ0 ρ • Inviscid approximation • Restrictions to m = 0 mode in the Lorentz force and to m = 1 mode in the magnetic field • Radial and axial Lorentz force components are assumed to be absorbed into pressure (to a large extend) Simulations (Riga): A simple saturation model • Focus on azimuthal component (cannot be absorbed into pressure) (v̄z 1 ∂ )δvφ = [(∇ × B) × B]φ ∂z µ0 ρ • Fix the field strength to the measured excess power (via energy balance) d dt Z B2 dV = − 2µ0 Z j2 dV − σ Z v · (j × B) dV . • Insert the perturbed velocity field into induction equation solver and compare the output with observational facts Simulations (Riga): Computed Lorentz force 3 20 40 60 80 100 3 fz[kN/m ]: -14-12-10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 3 fp[kN/m ]: -6 -4 1 1 0 0 0 -1 z 1 z z fr[kN/m ]: -100 -80 -60 -40 -20 0 -1 -2 0 2 4 6 8 10 12 -1 0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2 r r r frLorentz fzLorentz fϕLorentz Simulations (Riga): Braking of vφ • Hence, the differential rotation is decreased (Lenz’s rule !) Vp[m/s]: -12-11-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 dVp[m/s]: -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Vp[m/s]: -12-11-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 1 0 0 0 -1 z 1 z z • The azimuthal component is braked in the inner cylinder and accelerated in the back-flow cylinder. -1 -1 0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2 r r r v̄φ dvφ vφ Simulations (Riga): saturation model • Simultaneous solution of 2D-induction equation ∂B 1 2 = ∇ × (v(t) × B) + ∇ B ∂t µ0 σ and the simplified equation for δvφ: 1 ∂ [(∇ × B(t)) × B(t)]φ (v̄z )δvφ(t) = ∂z µ0 ρ • Gailitis et al.: Comptes Rend. Phys. 9, 721-728 (2008), Stefani et al.: ZAMM 88, 930-954 (2008), Stefani et al.: Theor. Comp. Fluid Dyn. 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1.2 1.8 1.6 Kinematic: numerical Kinematic: measured Saturated: numerical (old) Saturated: numerical (new) Saturated: measured -1.4 -1.6 1000 1200 1400 1600 1800 2000 2200 2400 2600 2800 Rotation rate [1/min] Frequency [1/s] Growth rate [1/s] Simulations (Riga): Comparison Experiment/Simulation 1.4 1.2 1 0.8 0.6 0.4 0.2 Kinematic: numerical Kinematic: measured Saturated: numerical (old) Saturated: numerical (new) Saturated: measured 1000 1200 1400 1600 1800 2000 2200 2400 2600 2800 Rotation rate [1/min] Good correspondence with measured growth rate and frequency in the kinematic and saturation regime Simulations (dynamo experiments): Cadarache (VKS) VKS experiment (Von Karman sodium): Helical flow with two poloidal and two toroidal eddies driven by counter-rotating impellers Lid layers 90 200 80 360 Side layer Simulations (dynamo experiments): Cadarache (VKS) Simulation for an analytical test flow Isosurface of magnetic energy Magnetic field lines Simulations (dynamo experiments): Cadarache (VKS) • Detrimental role of rotating layers behind impellers on the m = 1 mode (EJM/B 2006) • Using soft iron impellers (µrel ∼ 70) self-excitation was achieved (Monchaux et al. 2007) Simulations (dynamo experiments): Cadarache (VKS) Very nice field reversals (Berhanu et al. 2007), and extremely rich dynamics Simulations (dynamo experiments): Cadarache (VKS) • Recent attempts to explain low Rmcrit ∼ 31 and axial dipole (How to bypass Cowling’s theorem???) – Non-linear back-reaction producing non-axisymmetric flow components (Gissinger et al., PRL 2008) – α − Ω dynamo with α concentrated at impellers (Laguerre et al., PRL 2008). However: Problem with normalization of α... – Role of azimuthally drifting vortices – Role of soft-iron impellers ??? Simulations (dynamo experiments): Cadarache Take into account the correct geometry and the soft-iron impellers Simulations (dynamo experiments): Cadarache (VKS) We have to solve the generalized induction equation, including µrel and the helical turbulence parameter α: ∂B 1 B = ∇ × v × B + αB − ∇× ∂t µ0 σ µrel New Finite-Volume/Boundary element code: A. Giesecke et. al. Magnetohydrodynamics 2009, PRL 2010 Simulations (dynamo experiments): VKS Rm=30, α=−1.5 Rm=70, α=0 Rm=30, α=0 growing m=0 growing m=1 decaying m=0 Simulations (dynamo experiments): VKS • The soft-iron impellers work strongly in favour of the m=0 mode • Still, some small amount of helical turbulence is needed to excite it (A. Giesecke et al., Phys. Rev. Lett. 2010) Prospects: New experiments and numerical challenges Prospects: Azimuthal MRI (AMRI) 00 11 µ=0.25 00 11 00 11 00 11 00 11 µ=0.26 00 11 00 11 00 11 µ=0.27 00 11 00 11 00 11 µ=0.28 00 11 00 11 00 11 Hollerbach et al. PRL 2010 Prospects: Tayler instability Filling tube 0−8000 A Copper 0 1 0 electrode 1 0 1 Copper electrode GaInSn 75 cm 10 cm Measuring electrodes 0 1 Water cooling 6 µB=500 5 2 5.5 Ifluid [kA] UDV sensor 5 4.5 4 3.5 3 0 0.05 0.1 0.15 0.2 0.25 0.3 00 00 11 rin 11 /rη out 00 11 00 11 Prospects: Spherical Couette V 111111 000000 000000 111111 000000 111111 000000 111111 000000 111111 1111111111 0000000000 0000000000 1111111111 0000000000 1111111111 0000000000 1111111111 0000000000 1111111111 UDV 6 cm Bz 111111111 000000000 000000000 111111111 000000000 111111111 000000000 111111111 UDV 000000000 111111111 GaInSn 00 11 00 11 00 11 00 11 00 11 00000 11111 00000 11111 00000 11111 00000 11111 stable 00 11 00 11 Ha 00 11 00 11 V UD 1111111111111111 0000000000000000 0000000000000000 1111111111111111 0000000000000000 1111111111111111 0000000000000000 1111111111111111 0000000000000000 1111111111111111 0000000000000000 1111111111111111 0000000000000000 1111111111111111 0000000000000000 1111111111111111 0000000000000000 1111111111111111 0000000000000000 1111111111111111 0000000000000000 1111111111111111 111111 000000 000000 111111 000000 111111 000000 111111 000000 111111 000000 111111 000000 111111 000000 111111 000000 111111 unstable 00000 11111 00000 11111 00000 11111 00000 11111 unstable Recrit 18 cm 000000 111111 000000 111111 0−0.2 Hz 000000 111111 UD 00 11 11 00 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 00 11 Hollerbach 2009 Prospects: Alfvén wave experiment at HLD Ampoule with liquid Rubidium (hmmmm!!!) Interaction of shear and compressional Alfvén waves: Much more numerics needed... Prospects: DRESDYN, MRI+Tayler in sodium Prospects: DRESDYN, Precession experiment • Goal: Large experiment • J. Léorat (Meudon): Water experiment in cylinder (ATER) • A. Tilgner: Phys. Fluids crit 2005, Rm around 200 (based on unstable modes) • Much more numerical studies needed before experiment can be designed Some review papers • Gailitis A., Lielausis, O., Platacis, E., Gerbeth, G., Stefani, F. (2002): Colloquium: Laboratory experiments on hydromagnetic dynamos, Rev. Mod. Phys. 74, 973-990 • Stefani, F., Gailitis, A., Gerbeth, G. (2008): Magnetohydrodynamic experiments on cosmic magnetic fields, ZAMM 88, 930-954 • Stefani, F., Giesecke, A., Gerbeth, G. (2009): Numerical simulations of liquid metal experiments on cosmic magnetic fields, Theor. Comp. Fluid Dyn. 23, 405-429