Introduction to MagnetoHydroDynamics (MHD) Antoine Cerfon, Courant Institute, New York University

Transcription

1 Introduction to MagnetoHydroDynamics (MHD) Antoine Cerfon, Courant Institute, New York University SULI Introductory Course in Plasma Physics, June 6, 2016

2 PART I: DESCRIBING A FUSION PLASMA

3 METHOD I: SELF-CONSISTENT PARTICLE PUSHING Intuitive idea: solve for the motion of all particles iteratively, combining Newton s law with Maxwell equations At each time step i, solve m d2 x (i) k dt ( E ( ) = q k E(x k ) (i 1) + dx(i) k B (i 1) (x dt k ) ) i N = c 2 B (i 1) µ 0 c 2 dx i k q k dt δ(x xi k ) t ( ) B i = E i t k=1 k = 1,..., N Even with best available numerical solvers, problem still not tractable even with the most powerful computers when N as in magnetic fusion grade plasmas

4 METHOD II: COARSE-GRAIN AVERAGE IN PHASE SPACE (From G. Lapenta s: u /weather/pic.pdf ) For hot and diffuse systems with large # of particles, following every single particle is a waste of time and resources Replace the discrete particles with smooth distribution function f (x, v, t) defined so that f (x, v, t)dxdv is the # of particles in the infinitesimal six-dimensional phase-space volume dxdv.

5 DISTRIBUTION FUNCTION AND VLASOV EQUATION Macroscopic (fluid) quantities are velocity moments of f n(x, t) = f (x, v, t)dv Density nv(x, t) = vf (x, v, t)dv Mean flow P(x, t) = m (v V) (v V) fdv Pressure tensor Conservation of f along the phase-space trajectories of the particles determines the time evolution of f : df dt = f t + dx dv f + dt dt vf = 0 dx dt = v dv dt = q (E + v B) m f t + v f + q m (E + v B) vf = 0 This is the Vlasov equation

6 THE BOLTZMANN EQUATION In fusion plasmas, we separate, leading to the Boltzmann equation: f t + v f + q ( ) f m (E + v B) vf = t c This equation to be combined with Maxwell s equations: E = B t B = µ 0 J + 1 c 2 E t Nonlinear, integro-differential, 6-dimensional PDE Describes phenomena on widely varying length ( m) and time ( s) scales Still not a piece of cake, and never solved as such for fusion plasmas

7 MOMENT APPROACH f t + v f + q ( ) f m (E + v B) vf = t c Taking the integrals dv, mvdv and mv 2 /2dv of this equation, we obtain the exact fluid equations: n s t + (n sv s ) = 0 Continuity ( ) Vs mn + V s V s = q s n s (E + V s B) P s + R s Momen t ( ) d 3 dt 2 p s p s V s + π s : V s + q s = 0 (Energy) with P s = p s I + π s. Closure problem: for each moment, we introduce a new unknown End up with too many unknowns Need to make approximations to close the moment hierarchy

8 KINETIC MODELS VS FLUID MODELS For some fusion applications/plasma regimes (heating and current drive, transport), kinetic treatment cannot be avoided Simplify and reduce dimensionality of the Vlasov equation with approximations: Strong magnetization : Gyrokinetic equation Small gyroradius compared to relevant length scales : Drift kinetic equation Vanishing gyroradius : Kinetic MHD In contrast, fluid models are based on approximate expressions for higher order moments (off-diagonal entries in pressure tensor, heat flux) in terms of lower order quantities(density, velocity, diagonal entries in pressure tensor) We will now focus on the relevant regime and the approximations made to derive a widely used fluid model: the ideal MHD model

9 PART II: THE IDEAL MHD MODEL

10 The maximum p is limited by the stability properties Job of MHD The maximum τ E is determined by the confinement properties Job of kinetic models LAWSON CRITERION AND MHD Condition for ignition: pτ E 8 bar.s T min 15keV

11 PHILOSOPHY The purpose of ideal MHD is to study the macroscopic behavior of the plasma Use ideal MHD to design machines that avoid large scale instabilities Regime of interest Typical length scale: the minor radius of the device a 1m Wave number k of waves and instabilitities considered: k 1/a Typical velocities: Ion thermal velocity speed v T 500km/s Typical time scale: τ MHD a/v T 2µs Frequency ω MHD of associated waves/instabilities ω MHD 500kHz

12 EXAMPLE: VERTICAL INSTABILITY Figure from F. Hofmann et al., Nuclear Fusion (1997)

13 IDEAL MHD - MAXWELL S EQUATIONS a λ D, the distance over which charge separation can take place in a plasma On the MHD length scale, the plasma is neutral : n i = n e ω MHD /k c and v Ti v Te c so we can neglect the displacement current in Maxwell s equations: n i = n e B = 0 E = B t B = µ 0 J

14 IDEAL MHD - MOMENTUM EQUATION a λ D and a r Le (electron Larmor radius) ω MHD ω pe, ω MHD ω ce The ideal MHD model assumes that on the time and length scales of interest, the electrons have an infinitely fast response time to changes in the plasma Mathematically, this can be done by taking the limit m e 0 Adding the ion and electron momentum equation, we then get ρ dv dt J B + p = (π i + π e ) where ρ = m i n and V is the ion fluid velocity If the condition v Ti τ ii /a 1 is satisfied in the plasma ρ dv dt = J B p (Ideal MHD momentum equation)

15 IDEAL MHD - ELECTRONS In the limit m e 0, the electron momentum equation can be written as E + V B = 1 en (J B p e π e + R e ) This is called the generalized Ohm s law Different MHD models (resitive MHD, Hall MHD) keep different terms in this equation If r Li /a 1, v Ti τ ii /a 1, and (m e /m i ) 1/2 (r Li /a) 2 (a/v Ti τ ii ) 1, the momentum equation becomes the ideal Ohm s law E + V B = 0 The ideal MHD plasma behaves like a perfectly conducting fluid

16 ENERGY EQUATION Define the total plasma pressure p = p i + p e Add electron and ion energy equations Under the conditions r Li /a 1 and v Ti τ ii /a 1, this simplifies as ( ) d p dt ρ 5/3 = 0 Equation reminiscent of pv γ = Cst: the ideal MHD plasma behaves like a monoatomic ideal gas undergoing a reversible adiabatic process

17 IDEAL MHD - SUMMARY Valid under the conditions ( ) 1/2 (vi mi τ ) ii 1 a m e ρ t + (ρv) = 0 ρ dv dt = J B p ( ) d p dt ρ 5/3 = 0 E + V B = 0 E = B t B = µ 0 J B = 0 r Li a 1 ( rli a ) 2 ( me m i ) 1/2 a v Ti τ ii 1

18 VALIDITY OF THE IDEAL MHD MODEL (I) Are the conditions for the validity of ideal MHD ( ) 1/2 (vi mi τ ) ii r ( Li 1 a a 1 rli a m e mutually compatible? Define x = (m i /m e ) 1/2 (v Ti τ ii /a), y = r Li /a. ) 2 ( me m i ) 1/2 a v Ti τ ii 1 x 1 (High collisionality) y 1 (Small ion Larmor radius) y 2 /x 1 (Small resistivity) There exists a regime for which ideal MHD is justified (Figure from Ideal MHD by J.P. Freidberg, CUP, 2014) Is that the regime of magnetic confinement fusion?

19 VALIDITY OF THE IDEAL MHD MODEL (II) Express three conditions in terms of usual physical parameters: n, T, a For tokamak-like pressures and a = 1m, we find: The regime of validity of ideal MHD does NOT coincide with the fusion plasma regime (Figure from Ideal MHD by J.P. Freidberg, CUP, 2014) The collisionality of fusion plasmas is too low for the ideal MHD model to be valid. Is that a problem?

20 VALIDITY OF THE IDEAL MHD MODEL (III) It turns out that ideal MHD often does a very good job at predicting stability limits for macroscopic instabilities This is not due to luck but to subtle physical reasons One can show that collisionless kinetic models for macroscopic instabilities are more optimistic than ideal MHD This is because ideal MHD is accurate for dynamics perpendicular to the fields lines Designs based on ideal MHD calculations are conservative designs

21 FROZEN IN LAW E + V B = 0: in the frame moving with the plasma, the electric field is zero The plasma behaves like a perfect conductor The magnetic field lines are frozen into the plasma motion

22 MAGNETIC RECONNECTION Image from Principles of Magnetohydrodynamics With Applications to Laboratory and Astrophysical Plasmas by J.P. Goedbloed and S. Poedts, Cambridge University Press (2004) Magnetic reconnection: a key phenomenon in astrophysical, space, and fusion plasmas Cannot happen according to ideal MHD Need to add additional terms in Ohm s law to allow reconnection: resistivity, off-diagonal pressure tensor terms, electron inertia,... E + V B = ηj + 1 ne J B 1 ne P e + m e J ne 2 t Associated instabilities take place on longer time scales than τ MHD

23

24 PART III: MHD EQUILIBRIUM

25 EQUILIBRIUM STATE By equilibrium, we mean steady-state: / t = 0 Often, for simplicity and/or physical reasons, we focus on static equilibria: V = 0 A more condensed form is B = 0 B = µ 0 J J B = p B = 0 ( B) B = µ 0 p Note that the density profile does not appear

26 1D EQUILIBRIA (I) θ-pinch Z-pinch Combine the two to get... (Figure from Ideal MHD by J.P. Freidberg, CUP, 2014)

27 1D EQUILIBRIA (II) Screw pinch (Figure from Ideal MHD by J.P. Freidberg, CUP, 2014) Equilibrium quantities only depend on r Plug into B = 0, ( B) B = µ 0 p to find: ( d p + B2 θ + ) B2 z + B2 θ dr 2µ 0 µ 0 r = 0 Balance between plasma pressure, magnetic pressure, and magnetic tension Two free functions define equilibrium: e.g. B z and p, or B θ and B z

28 GENERAL EQUILIBRIA Equilibrium relation J B = p (J B = p) B B p = 0 Magnetic field is tangent to surfaces of constant pressure (J B = p) J = 0 J p = 0 Current density is tangent to surfaces of constant pressure

29 (Figure from Plasma Physics and Fusion Energy by J.P. Freidberg, CUP, 2008)

30 The regions of constant pressure are nested toroidal surfaces Magnetic fields and currents lie on these nested surfaces

31 GRAD-SHAFRANOV EQUATION Surfaces of constant pressure coincide with surfaces of constant magnetic flux Ψ These are given by the Grad-Shafranov equation R ( ) 1 Ψ + 2 Ψ R R R Z 2 = µ 0R 2 dp dψ F df dψ Second-order, nonlinear, elliptic PDE. Derived independently by H. Grad 1 and V.D. Shafranov 2. The free functions p and F determine the nature of the equilibrium In general, the GSE has to be solved numerically 1 Proceedings of the Second United Nations Conference on the Peaceful Uses of Atomic Energy, Vol. 31, p Sov. Phys. JETP 6, 545 (1958)

32 EXAMPLES (I) Z/R R/R 0 Grad-Shafranov equilibrium for JET tokamak

33 EXAMPLES (II) 5 Y X Grad-Shafranov equilibrium for Field Reversed Configuration

34 3D EQUILIBRIA / φ 0

35 3D EQUILIBRIA Equilibrium equations B = 0, ( B) B = µ 0 p still hold Existence of nested toroidal surfaces not guaranteed anymore Tokamak Stellarator (3D) Z/R R/R 0 Computing 3D equilibria fast and accurately still a challenge Several existing codes, based on different assumptions/approximations and used to design and study stellarators: VMEC, PIES, SPEC, HINT, NSTAB

36 PART IV: MHD STABILITY

37 WHAT DO WE MEAN BY MHD STABILITY? That the plasma is initially in equilibrium does not mean it is going to remain there The plasma is constantly subject to perturbations, small and large The purpose of stability studies is to find out how the plasma will react to these perturbations Will it try to return to the initial steady-state? Will it find a new acceptable steady-state? Will it collapse?

38 A MECHANICAL ANALOG Figure from J.P. Freidberg, Ideal MHD, Cambridge University Press (2014)

39 SOLVING FULL NONLINEAR MHD EQUATIONS Here is an idea to study stability of a magnetically confined plasma: Choose a satisfying plasma equilibrium Perturb it Solve the full MHD equations with a computer Analyze results Such an approach provides knowledge of the entire plasma dynamics There exist several numerical codes that can do that, for various MHD models (not only ideal): M3D, M3D-C1, NIMROD Computationally intensive Get more information than one needs?

40 LINEAR STABILITY (I) Ideal MHD dynamics can be so fast and detrimental that one may often require linear stability for the equilibrium This can simplify the mathematical analysis tremendously Start with an MHD equilibrium: B 0 = 0, ( B 0 ) B 0 = µ 0 p 0 Take full ideal MHD equations, and write Q = Q 0 (r) + Q 1 (r, t) for each physical quantity, where Q 1 is considered very small compared to Q 0 Drop all the terms that are quadratic or higher orders in the quantities Q 1 (linearization)

41 LINEAR STABILITY (II) ρ 1 t + (ρ 0v 1 ) = 0 v 1 ρ 0 = J 1 B 0 + J 0 B 1 p 1 t p 1 t + v 1 p + γp v 1 = 0 B 1 = (v 1 B 0 ) t B 1 = 0 µ 0 J 1 = B 1 By design, the system is now linear in the unknown quantities ρ 1, v 1, J 1, B 1, p 1 Much easier to solve in a computer or by hand!

42 ILLUSTRATION: WAVES IN IDEAL MHD (I) Consider the stability of an infinite, homogeneous plasma: Write v 1 = ξ(r)/ t B = B 0 ez J = 0 p = p 0 ρ = ρ 0 v = 0 Expand ξ(r) as ξ(r) = ˆξe ik r Plug into linear system, and solve by hand Dynamics is anisotropic because of the magnetic field: k + k e z

43 ILLUSTRATION: WAVES IN IDEAL MHD (II) Writing the expression for each component, we get the system ω 2 k 2 v2 A ω 2 k 2 v 2 A k2 v2 S k k v 2 S k k v 2 S ω 2 k 2 v2 S ˆξ x ˆξ y = ˆξ z Two key velocities appear: v A = B 2 0 µ 0 ρ 0 v S = γ p 0 ρ 0 v A is called the Alfvén velocity, in honor of Hannes Alfvén, the Swedish scientist who first described MHD waves v s is the adiabatic sound speed

44 ILLUSTRATION: WAVES IN IDEAL MHD (III) ω 2 k 2 v2 A ω 2 k 2 v 2 A k2 v2 S k k v 2 S k k v 2 S ω 2 k 2 v2 S ˆξ x ˆξ y ˆξ z = For nontrivial solutions, determinant of the matrix should be 0 This leads to the following three possibilities for ω 2 : ω 2 = k 2 v2 A, ω2 = k2 ( k v 2 2 A + v 2 S) 1 ± 2 v A v2 S k 2 ( v 2 S + v 2 2 A) One can see that ω 2 0 The infinite homogeneous magnetized plasma is always MHD stable Some of these modes become unstable in magnetic fusion configurations, because of gradients and field line curvature

45 SHEAR ALFVÉN WAVE Branch ω 2 = k 2 v2 A (Figure from Ideal MHD by J.P. Freidberg, CUP, 2014) Transverse wave Balance between plasma inertia and field line tension Incompressible often the most unstable MHD mode in fusion devices

46 FAST MAGNETOSONIC WAVE Fast magnetosonic wave given by ω 2 = k2 ( k v 2 2 A + v 2 S) v A v2 S k 2 ( v 2 S + v 2 2 A) Simplifies in the limit v 2 S v2 A : it becomes the compressional Alfvén wave, ω 2 = k 2 v 2 A (Figure from Ideal MHD by J.P. Freidberg, CUP, 2014) Oscillation between plasma kinetic energy and magnetic compressional energy

47 SLOW MAGNETOSONIC WAVE Slow magnetosonic wave given by ω 2 = k2 ( k v 2 2 A + v 2 S) 1 2 v A v2 S k 2 ( v 2 S + v 2 2 A) Physics simplifies in the limit v 2 S v2 A : it is then called the sound wave, with dispersion relation ω 2 = k 2 v 2 S (Figure from Ideal MHD by J.P. Freidberg, CUP, 2014) Plasma motion parallel to field lines, compressible Oscillation between plasma kinetic energy and plasma internal energy (plasma pressure)

48 COMMON MHD INSTABILITIES (I) Interchange instability (Figure from Plasma Physics and Fusion Energy by J.P. Freidberg, CUP, 2008)

49 (Figure from Plasma Physics and Fusion Energy by J.P. Freidberg, CUP, 2008) COMMON MHD INSTABILITIES (II) Ballooning instability

50 COMMON MHD INSTABILITIES (III) Kink instability (Figure from Plasma Physics and Fusion Energy by J.P. Freidberg, CUP, 2008)

51 BALLOONING MODES IN TOKAMAKS (Right image from by J.W. Connor et al.)

52 KINK BALLOONING MODES IN TOKAMAKS Reconstruction of experimentally observed kink ballooning mode (Figure from U.S. Burning Plasma Organization enews February 28, 2013 (Issue 69) by S.A. Sabbagh et al.)

53 KEY ROLE OF MHD IN REACTOR DESIGN Maximum achievable pressure set by ballooning modes Maximum achievable pressure set by kink modes Maximum elongation set by vertical instability

54 POST SCRIPTUM: THE COURANT INSTITUTE OF MATHEMATICAL SCIENCES (CIMS) AT NYU

55 CIMS IN MANHATTAN

56 Abel prize in 2005, 2007, 2009 and members of the National Academy of Sciences 5 members of the National Academy of Engineering

57 Specialization in applied math, scientific computing, mathematical analysis Particular emphasis on Partial Differential Equations PhD programs in Mathematics, Atmosphere and Ocean Science, Computational Biology Masters of Science in Mathematics, Masters of Science in Scientific Computing, Masters of Science in Data Science, Masters of Science in Math Finance 60 faculty 100 PhD students

58

59 MFD DIVISION AT CIMS Founded by Harold Grad in faculty, 3 post-docs, 2 graduate students Work on MHD, wave propagation, kinetic models Analytic pen and paper work Development of new numerical solvers Collaboration with colleagues specialized in scientific computing, computational fluid dynamics, stochastic calculus, etc. Funding currently available for PhD students Feel free to contact me if you have any questions