66
13. FOUNDATIONS OF MHD
In this chapter, we extend our fluid treatment to address
current-carrying fluids in magnetic fields. The extra
force term, from the j × B Lorentz force, adds – shall
we say – an interesting richness to the problem. I am
using several references for this part of the course, as I
have not found one single book which covers the mate-
rial usefully. Thus, some of my references come from
labratory plasma work (confinement, stability, toroids),
and others come from space physics and astrophysics
(basic equations, flows, reconnection).
A. Remember your E&M?
We start with the full Maxwell:
4π 1 ∂E
∇×B= j+
c c ∂t
∇·B=0
(13.1)
1 ∂B
∇×E=−
c ∂t
∇ · E = 4πρq
where ρq is the charge density. A couple of simplifica-
tions are common in MHD. First, note that we usually
work with highly conductive fluids, that cannot sup-
port a finite space-charge density (why??). Thus, the E
fields present are those induced, which are O(vB/c) ≪
B in subrelativistic flows (our limit in this course).
Thus, ∂E/∂t ∼ O[(v/l)(vB/c)] = O[(cB/l)(v 2 /c2 )],
i.e. small, and we can also drop the displacement cur-
rent in most of what follows. With these approxima-
tions, we can write the “reduced Maxwell equations:
4π
∇×B= j
c component of E across B; and b̂ is the unit vector along
∇·B=0
(13.2)
1 ∂B
∇×E=−
c ∂t
∇·E=0
which will govern almost everything we do in this
course.
1. CONDUCTIVITY AND OHM ’ S LAW: I
To close the (reduced) system we need to connect fields
to currents. In most situations we can assume a vector
Ohm’s law, which in the absence of inductive effects
would be written j = σE, with scalar electrical con-
ductivity σ. How do we include inductive effects here?
In our reduced MHD, we can write 4πj = c∇ × B.
Now go to a frame moving at v (relative to the lab);
and keep v ≪ c to avoid worrying about relativity. In
that frame there will be a simple Ohm’s law, j′ = σE′ .
But we know the fields transform as
 v   v 
E′ = E + × B ; B′ = B − × E
c c
(13.3)
(recall γ ≃ 1 here). Now, we know that any charge den-
sity ρq in the lab frame will contribute an extra current
ρq v in the comoving frame ... however with E small,
and v also small, any extra j′ is (v 2 /c2 ). We therefore
have j = j′ , and thus in the lab we can write
 v 
j = σE′ = σ E + × B (13.4)
c
We can also interpret (13.4) by noting that E′ is the
field felt by an observer (or an electron) moving with
the fluid element.
2. CONDUCTIVITY AND OHM ’ S LAW: II
In a high density fluid collisions usually dominate par-
ticle motions, so that a scalar conductivity σ and the
vector form of Ohm’s law, (13.4), is enough. Occa-
sionally, however, we need to extend the description to
consider microphysical effects in a low density plasma.
In particular, applying a E field to a magnetized plasma
will result in currents parallel and perpendicular to the
background field Bo . The details are in the Appendix;
here, just note that the Generalized Ohm’s law can be
written as
j = σo Ek + σ⊥ E⊥ + σH n̂ × E (13.5)
Here, Ek is the component of E along B; j, E⊥ is the
B. The first term uses the normal (collisional) conduc-
tivity; the second uses the Pederson conductivity; and
the last (usually the smallest) uses the Hall conductiv-
ity.
B. Field Evolution: Induction Equation
How does the field evolve? We use (13.4), but consider
only the inductive part;1 and we work with ”reduced”
Maxwell (13.2):
v c
E=− ×B+ ∇×B (13.6)
c 4πσ
1
Why? because we don’t expect any unshielded free charge to
stick around ... right?
 67

Alternatively, using vector algebra and defining the The first term describes intrinsic changes in B, while
quantity η = c2 /4πσ, this becomes the second describes advection of the field through the
boundary. Now, Stokes’ theorem rewrites this as
∂B
= ∇ × (v × B) + η∇2 B
Z  
(13.7) dΦB ∂B
∂t = − ∇ × (v × B) · dS (13.12)
dt S ∂t
which is often called the induction equation. Two com- Now, (13.7) allows us to write this, generally, as
ments on this equation.
dΦB
Z
• Compare this to equation (4.8), which describes = η∇2 B · dS (13.13)
the evolution of vorticity, in the barotropic limit. They dt S
are just the same: and so we should expect the magnetic Thus, as η → 0 (the perfect conductivity limit, an
field and the vorticity to behave similarly in hydrody- “ideal” plasma), ΦB becomes a constant of the mo-
namic situations. One example is flux freezing, which tion. That is, the magnetic flux through a loop which
we derive immediately below: it’s the MHD analog of is “attached to the plasma”, stays constant as that loop
Kelvin’s theorem. Another example is the existence of is stretched (or compressed) by the flow field. The
magnetic flux ropes: a field subject to fluid forces tends mean magnetic field in that loop therefore decreases (or
to bunch into linear, high-field regions (think of a tor- grows), in proportion to the change of loop area. This
nado or a vortex line). is Flux Freezing.2
• Equation (13.7) has two important limits, as fol- Magnetic Field Lines? (A discussion from
lows. When we worked with fluids, we noted that the Schmidt...) The concept of a magnetic line of
behavior of a system is sensitive to the ratio of the ad- force is an abstraction. IN general no identity can
vective to dissipative terms. We expressed this ratio in be attached to these lines (they cannot be labelled
terms of the Reynolds number. The induction equation in a varying field), nor can we speak of “motion”
admits a similar separation of effects. We take the mag- of field lines. In a perfect conductor, however, the
netic Reynolds number concept of field lines becomes meaningful, due to
flux freezing.
LV Consider a material line in the fliud (say a chain
Rm = (13.8) of labelled droplets, or particles painted pink), de-
η
fined by intersecting material surfaaces. Choose
where L, V are characteristic length and velocity scales these surfaces everywhere tangential to B at t =
of the problem. This quantity again measures the radio 0. The flux through both surfaces is therefore zero
of inertial/advection, to dissipative, terms. to start, and their intersection defines a field line at
that point. Flux freezing guarantees that these sur-
1. IDEAL LIMIT: FLUX FREEZING faces continue to ΦB = 0 at any later time. Thus,
their intersection continues to define a field line,
An imporant limit in MHD is the limit when Rm ≫ 1, in fact the same field line – it has become identi-
called ideal MHD . This is the case of a highly con- fiable; labelling the material (painting it pink) has
ducting plasma. Here, the induction equationk (13.7), labelled the field line, and the local fluic velocity
reduces to v(x, t) is also the velocity of that section of the
field line. The field line is attached to – “frozen
∂B into” – the fluid.
≃ ∇ × (v × B) (13.9)
∂t 2. RESISTIVE LIMIT: FLUX ANNIHILATION
while Ohm’s law becomes In a fluid with finite conductivity, flux freezing no
v longer holds. We can explore this by going to the other
E+ ×B≃0 (13.10) limiting case, when Rm ≪ 1. This is diffusive limit.
c
If we simply ignore the advection term, equation (13.7)
Consider, now, a closed curve (C) bounding a surface becomes
(S) which is moving withR the plasma. The magnetic ∂B
flux through S is ΦB = B · dS; and its rate of change = η∇2 B (13.14)
∂t
can be written
dΦB ∂B
Z I
2
Compare Kelvin’s Theorem, from chapter 4; you will see that the
= · dS − B · (v × dl) (13.11)
dt S ∂t C math is identical, as is the nature and interperation of the result.
 68

This describes the effect of Ohmic dissipation on the 1. ENERGETICS OF THE E AND B FIELDS
magnetic field; note that it is a standard diffusion equa-
tion. To remind yourself of this...go to Jackson.3 Keep all of
Diffusion of field lines? We know how solutions the terms – the full Maxwell’s equations. Consider the
to (18.4) behave: an initial field will decay on a quantity E · j: it expands as
timescale ∼ L2 /η. Some authors discuss this in ∂E
terms of field line “diffusion” or “slippage” out 4πj · E = cE · (∇ × B) − E ·
of the fluid. Remember that the density of field ∂t
lines is related to the strength of the field; so a But now, use the vector identity ∇·(E×B) = B·(∇×
lower density of field lines, with time, should cor- E) − E · (∇ × B), and Maxwell again, to get
respond to field lines “diffusing” out of the field.
In particular, when η is finite, field lines are no c ∂E ∂B
−j · E = ∇ · (E × B) + E · +B·
longer tied to parcels of the plasma; some authors 4π ∂t ∂t
talk of field lines “moving through” the plasma in Now: we identify the the energy density in the fields, as
dissipative regions. uE = E 2 /8π; and uB = B 2 /8π. And we integrate the
C. Fluid Equations: Lorentz force equation above over some volume V , with a surface S,
giving
The effect of the B field on the force equation is Z Z
d
straightforward. We simply add the Lorentz force to − j · EdV = (uE + uB ) dV
the momentum equation: V V dt
(13.17)
c
Z
Dv j + (E × B) · dS
ρ = −∇p + × B (13.15) S 4π
Dt c
Note, I have ignored viscosity here, as well as any ex- Thus: the first term is the rate of change of field energy
ternal forces (such as gravity). Now: expand out the in V . The second term – involving the Poynting flux,
Lorentz force as S = (c/4π)(E×B) – is the flow of EM energy through
j 1 the surface. The last term, then, is the work done by
×B= (∇ × B) × B the fields on the sources in the volume (and it can have
c 4π (13.16)
1 1 either sign ... thinking about driving vs. dissipation .. I
2
= − ∇B + (B · ∇)B think). This can also be written in differential form:
8π 4π
This is an important breakdown of the Lorentz force; d
(uE + uB ) + ∇ · S = −j · E (13.18)
it demonstrates that the field exerts a magnetic ten- dt
sion and a magnetic pressure on the fluid. The first which looks much like any other conservation equation.
term in (13.16) represents the gradient of a scalar pres-
sure, pB = B 2 /8π. It appears in the momentum 2. ENERGETICS OF THE FLUID
equation parallel to the fluid pressure....you can think
Schmidt has a nice approach: I’ve redone it in cgs,
of trying to compress a magnetic field, parallel to it-
keeping all the terms. To start, dot v into the force
self, with the field resisting the compression (“fight-
equation (13.15):
ing back”). The second term in (13.16) is non-zero
only if the field varies parallel to itself. A simple il- Dv 1
ρv · + v · ∇p = v · j × B
lustration is a curved field line. The curvature means Dt  c (13.19)
there is a current flowing along the field line; the j × B

1 1 ∂E
force points inwards (relative to the curvature). Thus, = v · (∇ × B) − ×B
4π c ∂t
curved field lines “want to straighten out”...Some au-
thors combine both effects by describing magnetic field where I’ve used the full Maxwell on the right. With
lines as “elastic bands within the fluid”, which resist some algebra, the terms on the LHS can be written:
being stretched: either pushed together, or pulled trans- 
∂v

verse to their length. ρv· + (v · ∇)v
∂t
    (13.20)
D. Fluid Equations: Energetics ∂ 1 2 1 2
= ρv + ∇ · ρv v
∂t 2 2
We need two items here: first recall the energetics of
the fields, by themselves; then their effects on the ener-
3
getics of the fluid. Preferably the 1st edition, where he works in in cgs.
 69

and just the same as the first part of the course. Now, con-
sider the last term, using vector algebra to reorganize
1 ∂p γ each term:
v · ∇p = + ∇ · (pv) (13.21)
γ − 1 ∂t γ−1
where we’ve used p ∝ ργ in this last; otherwise these
are both general, for compressible fluids. So far, this is

 
1 1 ∂E 1 1 ∂E
v · (∇ × B) − × B = − (v × B) · (∇ × B) + (v × B) · (13.22)
4π c ∂t 4π 4πc ∂t

Now ... use v × B/c = j/σ − E; and E · (∇ × B) = B · ∇ × E − ∇ · (E × B); expand out and collect terms, to get
 
v 1 2 1 ∂E ∂B c
· (j × B) = j − E· +B· − ∇ · (E × B) (13.23)
c σ 4π ∂t ∂t 4π

Collecting everything, we get

   
∂ 1 2 1 1 2 2 1 2 γ c 1
ρv + p+ (E + B ) + ∇ · ρv v + pv + E × B = − j2 (13.24)
∂t 2 γ−1 8π 2 γ−1 4π σ

This is still fully general; it combines the EM and fluid
terms in a conservative form... We can now follow the
usual MHD approach and drop E 2 in the ∂/∂t term (as
Schmidt does). Also note that Parker writes the Poynt-
ing flux, in the MHD case, as S = B × (v × B)/4π =
v⊥ B 2 /4π, explicitly assuming ideal and an induction-
only E field.
References
Once again, much of this basic material is “just from
me”. Possibly useful references might be Priest (who
has a very good MHD introduction), also Woods and
Davidson. I followed Schmidt for the energetics analy-
sis, and cannot for the life of me remember where I first
found the Generalized Ohm’s law arguments (probably
Chen’s plasma book? ... sorry).
tween collisions, tcoll , or its inverse the collision rate,
νcoll = 1/tcoll , as follows. Consider a free electron, in
a plasma, subjected to an external electric field E. The
net force on the particle can be estimated,
∆p
Fnet ≃ eE − (13.25)
∆t
where ∆p/∆t is the mean rate of momentum change
per collision. But if the charges have a net drift velocity
vD , we can estimate ∆p/∆t ∼ me vD /tcoll ; then, in a
steady state we have Fnet ≃ 0, so that
eE ≃ nvνcoll (13.26)
and the drift velocity must be vD = eEtcoll /me . Next,
we can use this in the (static) Ohm’s law, to relate the
conductivity to the drift velocity:
j = ne evD = σE (13.27)

where the second equality defines σ. Collecting every-

thing, we end up with

E. Appendix I: Conductivity notes ne e2 ωp2

σ= = τcoll (13.28)
me νcoll 4π
1. COLLISIONAL CONDUCTIVITY
In the last expression the plasma frequency is used:
First, simple collisional (parallel) conductivity, σ. We ωp2 = 4πne2 /m. The simplest physics is when the col-
can find σ simply, in terms of the mean time be- lision time is determined by Coulomb collisions in the
 70

plasma, as in chapter 1 (check §1.F). Alternatively, the • microscopic. Alternatively, consider single parti-
collisions may be with microturbulence in the plasma cle motion. The particles undergo gyromotion around
... in which case the collision rate usually has to be put B; they undergo an E × B drift; and they suffer col-
in by hand (rather than calculated from first principles). lisions which disrupt these ordered motions. To illus-
trate (I take this from Park), let B = (0, 0, B) and
2. CROSS - FIELD CONDUCTIVITY E = (Ex , 0, Ez ). The equations of motion, in the ab-
Consider a general situation in which E and B exist; sence of collisions, are
there will be Ek and E⊥ components, relative to B. We
can anticipate the three terms in the Generalized Ohm’s dvx
law, as follows: (1) collisional conductivity, giving a m = qEx + qBvy
dt
field-aligned current proportional to Ek , as usual; (2) dvy
quasi-collisional conductivity, giving a cross-field cur- m = qBvx
dt
rent proportionao to E⊥ ; and (3) a transverse current, dvx
perpendicular to both E and B, based on single-particle m = qEz
dt
E × B drift.4 Note that this drift is independent of par-
ticle charge and mass .. but we get a net current be-
cause the collision frequencies are not. There are two with solutions
approaches, macro and micro.
• macroscopic. Now: just consider one species, vyo + qEx
vx = vxo cos Ωt + sin Ωt
still, and extend (??eqn13.26) as qB
vyo + qEx qEx
 v 
vy = cos Ωt − vxo sin Ωt − (13.33)
e E + × B + mvνcoll = 0 (13.29) qB qB
c
qEz
Retain the definition j = nev; so that vz = vzo + t
m
mνcoll 1
j + j × B = −eE (13.30)
ne nc (This shows the combination of gyromotion and E ×
Bdrift). Now, include collisions. Let νcoll be the colli-
Because we have cross products, things get complicated
sion frequency, and assume collisions occur randomly
.. choose the form,
in time. The probability of a collision is e−νcoll t n and
j = σo Ek + σ⊥ E⊥ + σH n̂ × E (13.31) the probability that a particle will escape a collision in
time t, t + dt is given by νcoll e−νcoll t dt. Inbetween col-
Put this into (13.30), do the algebra and collect the lisions the trajectories of (13.33) are followed. Thus,
terms5 We find, the drift speeds are
ne2
σo = Z
mνcoll hvx i = A νcoll e−νcoll t (Ex /B) sin Ωtdt
ν2
σ⊥ = σo 2 coll 2 (13.32) Ex νcoll Ω
νcoll + Ω = 2 + Ω2
B νcoll
νcoll Ω
σH = σo 2 Z
νcoll + Ω2 hvy i = A νcoll e−νcoll t (Ex /B)(cos Ωt − 1)dt
where Ω is the gyrofrequency, as usual. These are the Ex Ω2
collisional, Pederson and Hall conductivities. =− 2
B νcoll + Ω2
Z
4
Quick physics here: a single particle undergoes gyromotion hvz i = A νcoll e−νcoll t (Ex /B) sin Ωtdt
around the local magnetic field, at a frequency Ω = eB/mc. If
there is also a perpendicular E field, the center of the gyro-orbit
Ez
= (13.34)
shifts during one orbit. Following the particle’s trajectory, it’s Bνcoll
easy to show that the orbit-center moves at a steady drift speed,
∝ E × B.
if the normalizing factor A−1 = νcoll e−νcoll t dt. Fi-
R
5
Remember the definitions: ωp2 = 4πne2 /m, and Ω = eB/mc;
that’s how the frequencies ωp and Ω come into the expression. nally, use this in the definition of current density, j =
 71

nqhvi. We get where, in SI, η = 1/µo σ is the magnetic diffusivity;

note that it differs from the electrical resistivity, which
ne Ωνcoll is 1/σ.
jx = Ex
B ν 2 + ω2 Note of caution: just to confuse us all, the usual
ne Ω2 cgs definition is η = 1/σ, and lightspeed terms
jy = − Ex (13.35) are carried through the equations explicitly. In
B ν 2 + ω2
ne Ω addition, some authors working in SI also take
jz = Ez η = 1/σ. Sigh . . . be sure to check the defi-
B νcoll
nitions in each different book you use.]
Compare this to (13.32); you can see we get the same
answers. 3. FORCE EQUATION , MAGNETIC TENSION AND
PRESSURE
F. Appendix II: Do It in SI
The force equation is
Both SI and cgs are used in the literature, which makes
it confusing to go between sources. For reference, I
store the main equations in SI.
Dv
ρ = −∇p + j × B + F (13.38)
1. MAXWELL IN SI: Dt
with the separation of terms, as
1 ∂E
∇ × B = µo j +
c2 ∂t 1 1
j×B=− ∇B 2 + (B · ∇)B (13.39)
∇·B=0 2µo µo
∂B (13.36)
∇×E=−
∂t
1 As disscussed in §13.C, the first term represents the gra-
∇·E= ρq dient of a scalar pressure, which in SI ispB = B 2 /2µo .
ǫo
The second term is non-zero of B varies parallel to its
2. INDUCTION EQUATION length, and represents magnetic tension.

If we eliminate E from Maxwell # 1, Maxwell # 3 and 4. ENERGY EQUATION

(13.3), we have
∂B After dropping the charge-based E, but keeping the
= ∇ × (v × B) + η∇2 B (13.37) Poynting flux explicitly:
∂t

   
∂ 1 2 1 1 2 1 γ 1 1
ρv + p+ B + ∇ · ρv 2 v + pv + E × B = − j 2 (13.40)
∂t 2 γ−1 2µo 2 γ−1 µo σ

and that’s the end of this chapter, folks.