SOLAR MHD

Lecture 2
Plan
• Magnetostatic Equilibrium
ü Structure of Magnetic Flux Tubes
ü Force-free fields
• Waves in a homogenous magnetized
medium
ü Linearized wave equation
ü Alfvén wave
ü Magnetoacoustic waves
ü Phase diagrams
ü Poynting flux in MHD
• Waves in an inhomogenous magnetized
medium
 MHD Equations
∂ρ dρ
+ ∇ •( ρ v) = 0 or + ρ∇ • v = 0 (Continuity Equation)
∂t dt
dv 1
ρ = −∇p + (∇ × B) × B + ρg + ρν∇ 2 v (Momentum equation)
dt 4π
d% p (
' γ * = 0 (Energy equation for an adiabatic fluid),
d& ρ )
(where d/dt = ∂/∂t + v·∇).
∂B
= ∇ × (v × B) + η∇ 2B,
∂t
In the limit of infinite conductivity (η=0),
∂B/∂t =∇×(v×B) (induction equation)
=−B∇·v − (v·∇)B + (B·∇)v (since ∇·B = 0)
Magnetostatic Equilibrium
Structure of a cylindrically symmetric flux tube
Consider a cylindrically symmetric magnetic field of
the form, using cylindrical polars (R,φ,z) :

B = [0, Bφ(R), Bz(R)] (3.1)

Substituting the above in the magnetostatic

equation,
1
−∇p + (∇ × B) × B = 0. (3.2)
4π
(assuming that gravity can be neglected), the equilibrium equation becomes:

d !# Bφ + Bz $& Bφ
2 2 2
dp
+ + = 0, (3.3)
#
dR dR " 8π % 4π R &

where the second term represents the magnetic pressure and the third term the
magnetic tension.
The equation of a field line can be determined simply from the equation

Rdφ dz
= . (3.4)
Bφ Bz
Consider a section of a flux tube of length 2L. Thus the twist over 2L is:
2L
Bφ 2LBφ (R)
Φ(R) = ∫ dφ = ∫ RBz
dz =
RBz(R)
. (3.5)
0

The term 4πL/Φ is referred to as the pitch P of the field and gives the length over
which the field lines circle the axis once [since from Eq. (3.4), P = 2πRBz/Bϕ= 4πL/Φ].
Vertical Field (Bϕ=0)

When the azimuthal component of the field is zero, Eq. (3.3) becomes

d B2
(p + ) = 0,
dR 8π
which has the solution p+B2/8π = constant i.e. total pressure is constant in the
radial direction.
Azimuthal field (Bz=0)

When the vertical component of the field is zero, Eq. (3.3) becomes

d !# Bφ $& Bφ
2 2
dp
+ + = 0, (3.6)
# &
dR dR " 8π % 4π R

From Maxwell’s relation, Bϕ is related to the current density through:

1 d
jz = (RBφ ). (3.7)
4π R dR
Let the total current I flow within a cylinder of radius a. Integrating Eq. (3.7), gives:

Bϕ = 2IR/a2, R < a and Bϕ = 2I/R, R > a. (3.8)

The corresponding pressure is obtained by integrating Eq. (3.6):

2
" I % 2
p = p∞ + π $ 2 ' (a − R 2 ), R < a and p = p∞, R > a.
#πa &
Within the cylinder of radius a and Bϕ increase linearly with R, while the gas
pressure p decreases.
Force-free fields (p=0)
When p=0, Eq. (3.3) yields:
d !# Bφ + Bz $& Bφ
2 2 2

+ = 0, (3.9)
#
dR " 8π % 4π R &

Let us write:
B2 = f (R), (3.10)
Then Eq. (3.9), yields:
Bϕ2 = -½ R df/dR, (3.11)
and
Bz2 = B2 – Bϕ2. (3.12)
Uniform twist or constant pitch force free field(Gold and Hoyle 1957)
Substituting Bϕ/Bz= ΦR/2L=2πR/P [from Eq. (3.5)] in Eq. (3.9), we have

d ! 2π R 2 $
#[1+ ( ) ]Bz & = 0, (3.13)
dR " P %
which can be integrated to yield
B0 2π R B0
Bz = , Bφ = , (3.14)
1+ (2π R / P )2 P 1+ (2π R / P )2
where B0 is the field (purely vertical) on the axis and P = 4πL/Φ is the pitch of the field
that is assumed to be constant. Such a configuration was first proposed by Gold and
Hoyle (1958) in connection with a flare model.
Effect of twisting a tube (Parker 1977)
Consider the effect of twisting a force-free flux tube of radius a, that is confined by an external
plasma pressure pe=B2(a)/8π (follows from the condition of pressure balance R=a). During
twisting, assume that the field remains cylindrically symmetric, but the radius varies from its
initial value a(0) to a. Then the mean-square axial field <Bz2> is constant, since:
2 a
< Bz2 >= 2 ∫ RBz2dR
a 0
2 a" 1 df %
= 2 ∫ $ Rf + R 2 'dR from Eq. (3.12)
a 0# 2 dR &
2 a d
= ∫ (R 2f )dR
a2 0 dR
= f (a) = B 2 (a) using Eq. (3.10)
Consider the uniform twist case, given by Eq. (3.14), with B0=B(0)(1+Φ2a2/4L2)½, which varies with
the twist Φ in such a way as to keep the total field equal to a constant B(0) at the tube edge. B(0) is
the initial uniform axial field strength when the tube is untwisted. From flux conservation:
a
2π ∫ RBz dR = π [a (0) ]2 B (0) or (1+ Φ 2 a 2 / 4L2 )1/2 log(1+ Φ 2 a 2 / 4L2 ) =[a (0)Φ]2 / 4L2
0

It transpires from the above equation, that as the twist Φ increases, so do Bϕ and a increase.
At the same time Bz(0) increases (since an outwards magnetic pressure is required to
balance the inwards tension force produced by twisting), while Bz(a) decreases.
The mean longitudinal field is reduced, rather than enhanced by twisting the tube
because <Bz>2 = <Bz2> only if Bz is uniform - any inhomogeneity causes <Bz>2 < <Bz2>.
But <Bz2> does not increase – so the mean field is reduced.
Linear Force-Free Fields

For a force-free field, the force balance equation reduces to

j × B = 0. (3.15)

Since j ∝ curl B, Eq. (3.15) is satisfied for

curl B = αB, (3.16)

where α is in general a function of position. Taking the divergence of Eq. (3.16),
we find:
(B. ∇)α = 0, (3.17)

which implies that B lies on a surface of constant α (as well as j). When α is
constant, taking the curl of Eq. (3.16), we have:

(∇ 2 + α 2 )B = 0, (3.18)

which holds for linear force-free magnetic fields.

 Example of a constant α force-free field

Consider a field of the form B = [0, By(x),Bz(x)], for which Eq. (3.15) i.e. j × B = 0
becomes
d
(B y2 +Bz2 ) = 0
dx
which integrates to:

By2 +Bz2 = B02 (say) is constant.

The solution therefore is:

B = [0, By, (B02 - By2)½].

For the particular case of constant α, the z component of Eq. (3.16) (i.e. curl B = αB),
is:
dBy
= α (B02 − By2 )1/2 ,
dx
so that with the origin to be a zero of By, the solution is:

By=B0 sin αx, Bz=B0 cos αx (3.19)

 MHD WAVES IN A HOMOGENEOUS MEDIUM
Linearized Wave Equation
Consider a uniform vertical magnetic field in a homogeneous medium in which the
fluid is initially at rest everywhere (i.e. v = 0). The magnetic field vector defines a
preferred direction, but the two directions perpendicular to B0 are equivalent, so
the wave problem is effectively two-dimensional. In Cartesian coordinates (x,y,z),
take the initial magnetic field B0 along the z-axis, of the form:

B0 = [0, 0, B0] (4.1)

where B0 is constant.
Consider small departures from the initial equilibrium state of the form:
ρ = ρ0 + δρ, p = p0 + δp, v = δv1, B = B0 + δB,
where δρ is a perturbation in the density and so on. Substituting these in the
equations of motion and induction, we find:

∂δv 1
ρ0 = −∇δ p + (∇ × δB) × B0 . (4.2)
∂t 4π
∂δB
= ∇ × (δv × B0 ). (4.3)
∂t
The continuity equation becomes
∂δρ
+ ρ0 ∇ • δv = 0, (4.4)
∂t
where we have assumed that ρ0 is constant. We can relate δρ and δp, from the adiabatic
energy equation as follows:

∂δ p γ p0 ∂δρ ∂δ p ∂δρ
− = 0, or − cS2 = 0, (4.5)
∂t ρ0 ∂t ∂t ∂t
where γ is the ratio of specific heats and cS is the sound speed defined as
γ p0
cS2 = , (4.6)
ρ0
Taking the time derivative of Eq. (4.2), we have

∂2δv 1 % ∂δ p ( 1 % ∂δB (
2
= − ∇' *+ '∇ × * × B0
∂t ρ0 & ∂t ) 4πρ0 & ∂t )
B
= cS2 ∇(∇ • δ v) + ,-∇ × {∇ × (δ v × B0 )}./ × 0 , (4.7)
4πρ0

where we have used Eq. (4.5) to eliminate δp, Eq. (4.4) to eliminate δρ Eq.
(4.3) to eliminate δB.
We now look for plane wave solutions of the form:

δ v = v1 ei (k.r−ωt ) ,
and similarly for the other perturbed variables, where k is the wavenumber
vector and ω is the angular frequency of the wave. We can now replace ∂/∂t in
Eq. (4.7) by –iω and ∇ by ik. Eq. (4.7) now becomes:

2 2 B0
ω v1 = c k(k • v1) + {k × [k × (v1 × B0 )]} ×
S
. (4.8)
4πρ0
Sound Waves
When B0 = 0, so that the only force acting on the fluid is pressure, Eq. (4.8), becomes:

ω2v1 = cs2 k(k.v1)

We take the scalar product of Eq. (4.9) with k and assuming that k.v1 does not vanish, we
obtain,
ω2 =k2 cs2
This is the dispersion relation for acoustic waves, which propagate isotropically with a phase
speed cs.
Magnetic Waves
When the magnetic field is sufficiently large (i.e. β<<1), so that the gas pressure
can be neglected, Eq. (4.8) becomes:

ω 2 v1 = {k × [k × (v1 × B̂0 )]} × v A2B̂0 , (4.9)

where vA = B0 /√4πρ0 is known as the Alfvén speed. Let ez be a unit vector in

the z-direction i.e. directed along the field. Expanding the vector product in Eq.
(4.9), yields:

ω 2 v1 / v A2 = (k • ez )2 v1 − (k • v1)(k • ez )ez + "#(k • v1) − (k • ez )(ez • v1)$%k,

or in terms of the angle θ between the wave vector k and B0,

ω 2 v1 / v A2 = (k 2cos2θ )v1 − (k • v1)k cosθ ez + "#(k • v1) − k cosθ (ez • v1)$%k. (4.10)
From the condition div B = 0, we have k.B1=0, which means that the magnetic field
perturbation is normal to the direction of propagation. Taking the scalar production of Eq.
(4.10), with ez, we find:
ez.v1 = 0, (4.11)
from which it follows that the perturbed velocity is normal to the background field B0.
Now if we take the scalar product of Eq. (4.10) with k, we get:

(ω2 – k2vA2)(k.v1)=0, (4.12)

Properties of Alfvén Waves

Shear Alfvén Waves
If the perturbation is incompressible i.e. div v1 = 0, so that
k.v1=0. (4.13)
Substituting Eq. (4.13) in Eq. (4.10) yields,
ω = kvAcos θ, (4.14)

which is the dispersion relation for Alfvén waves (sometimes called shear Alfvén
waves) propagating in the positive z-direction (i.e. in the direction of the magnetic
field). These waves travel with a phase speed vAcos θ.
Eq. (4.14) can also be written as ω = kzvA. Differentiating with respect to kz, gives
the group velocity as vg = vA ez.
A polar diagram for Alfvén waves (solid curve) and The directions of v1 and B1 relative to B0 and the
compressional waves (dashed curve). The length of the k. For (a) the vectors v1 and B1 are normal to the
radius vector at an angle of inclination θ to B0 is equal to plane of k whereas for (b) v1 and B1 lie in the
the phase speed. same plane as k and B0.

The variation of the phase speed with θ can be seen in a polar diagram, which
takes the form of two circles of diameter vA. The waves propagate fastest along the
field, but not normal to it.
The direction of wave-energy is along the magnetic field. Energy propagates along
field lines, independent of direction, despite the fact that individual waves can travel
at any inclination to the field (other than 90°).
From Eq. (4.13), we find that Alfvén waves are transverse – the velocity
perturbation is perpendicular to the propagation direction. We also find from Eqs.
(4.4)-(4.5), that there are no density or pressure changes associated with these
waves.
From the induction equation, for plane waves, we find:

-ωB1= k × (v1× B0)

or
-ωB1= (k.B0) v1 - (k.v1) B0. (4.15)

Since k.v1=0 (from Eq. 4.11), Eqs. (4.14) & (4.15), give:
v1 = - B1/(4πρ0)½, (4.16)
which implies that B1 and v1 are in the same direction, both lying in a plane
parallel to the wave front.
Using Eq. (4.11), i.e. ez.v1 = 0 we find:
B0. B1=0, (4.17)
so that the magnetic field perturbations are normal to the field. From the Lorentz force,
j1 × B0/c = (k × B1) × B0/4π, we find:
= (k.B0) B1/4π - k (B0. B1)/4π. (4.18)
The first term comes from the magnetic tension and second from the magnetic pressure. Using
Eq. (4.17), we find that the driving force for Alfvén waves comes solely from the magnetic
tension. From Eq. (4.16), we find:
B12 1
= ρ0v12, (4.19)
8π 2
which shows that for Alfvén waves, there is equipartition between magnetic and kinetic energy.
Torsional Alfvén Waves

In cylindrical geometry with an axial field B0 ez, there exist waves which
possess only an azimuthal component in the field perturbation.
Let at time t = 0, a circular disk perpendicular to the (initially uniform)
magnetic field be put into rotation at a constant angular velocity Ω. A wave
front moves up at the Alfvén speed along the rotating field bundle.

Compressional Alfvén Waves

The second solution of Eq. (4.12) is

ω =k vA
which represents compressional Alfvén waves.
The phase speed is vA, regardless of the propagation angle. The group velocity is
vAk, so that energy is propagated isotropically.
Eqs. (4.10) and (4.11) imply that v1 lies in the (k,B0) plane and normal to B0. It
possesses components both along and transverse to k. From Eq. (4.15), B1 is in the
plane of v1 and B0, but is normal to k.
For θ=π/2, Eq. (4.10) shows that v1 is parallel to k and hence the wave is longitudinal
and from Eq. (4.18) only the magnetic pressure is the driver for the wave.
For θ=0, the wave is transverse and identical to an ordinary Alfvén wave; it is driven
solely by the magnetic tension and produces no compression (despite its name).
Magnetoacoustic Waves
The generalization of Eq. (4.10) when pressure and magnetic forces are important is:

ω 2 v1 / v A2 = k 2cos2θ v1 − (k • v1) k cosθ ez +

+"#(1+ cS2 / v A2 )(k • v1) − k cosθ (ez • v1)$%k.
Since v1 appears in the combinations k.v1 and ez.v1,let us take the scalar products with k and
ez to get:
(−ω 2 + k 2cS2 + k 2v A2 )(k • v1) = k 3v A2 cosθ (ez • v1)
and
k cosθ cS2 (k • v1) = ω 2 (ez • v1). (4.19)
If k.v1 vanishes, one recovers the Alfvén wave solution. Otherwise, from the two equations
above, the condition for a non-trivial solution yields the following dispersion relation:

ω 4 − ω 2k 2 (cS2 + v A2 ) + cS2v A2 k 4cos2θ = 0, (4.20)

For outward propagating waves (ω/k >0), there are two distinct solutions:
1/2
ω "1 2 2 1 4 4 %
= $ (cS + v A ) ± cS + v A − 2cS2v A2 cos 2θ ' . (4.21)
k #2 2 &

The higher frequency mode is known as the fast magnetoacoustic wave and the
other one the slow magnetoacoustic wave. The Alfvén wave phase speed lies
between the two.
Properties of Magnetoacoustic Waves
• The phase speed does not depend on frequency i.e. the waves are non-dispersive;
• They are anisotropic (i.e. have a dependence on θ) – also the phase speed is not the
same as the group speed.
• For θ = 0, the phase speed u(=ω/k) is either cS or vA, whereas for θ = π/2, the phase
speed is (cs2 +vA2)1/2 for the fast mode and 0 for the slow mode.
• Consider the limits vA >> cS or cS >> vA, when
uf →(cs2 +vA2)1/2, (vA >> cS or cS >> vA) (4.22)
i.e. uf is the larger of cS or vA and is independent of θ i.e. the fast mode propagates
isotropically. The slow mode speed becomes:
uS → cT cos θ, (vA >> cS or cS >> vA) (4.23)
cT = vAcs / √(cs2 +vA2)1/2 (4.24)
i.e. uS is the smaller of cS or vA and its angular dependence is the same as that of the
Alfvén wave.
• Consider the limiting case cos2θ << 1, that is for nearly perpendicular propagation to B0.
Expressions (4.22) and (4.23) hold in this limit as well, but they now apply for arbitrary cS
or vA. In this context, Eq. (4.22) represents the fast magnetoacoustic speed, while the
quantity given by Eq. (4.24) is referred to as the cusp speed.
• When cS >> vA, the fast mode is akin to a sound wave modified by the magnetic field. In
the opposite limit cS << vA, it propagates at the Alfvén speed, with fluid displacements
nearly perpendicular to B0 as in an Alfvén wave.
 Figure 1.12.: Propagation diagram for the MHD waves in a uniform medium, showing the ph
speed (length of the vector u) as a function of the angle ✓ of the wave vector w
respect to the direction of the magnetic field.

Phase velocities (left) and group velocities (right) for

fast & slow waves propagating at an angle θ to the 1.8.2. Properties of the magnetoacoustic waves
magnetic field direction: uS and uf are the slower of the
In the second set of waves, the density and pressure perturbations do not vanish,
Alfvén speed vA and sound speed cS. they share some properties with sound waves. The phase speed u (1.122) depends
the sound speed, the Alfvén speed and the angle ✓ between the wave vector and t
magnetic field. Introducing a dimensionless phase speed ũ,

u = ũ(cs vA )1/2 , (1.12

(1.122) becomes
cs vA 2
ũ4 ( + )ũ + cos2 ✓ = 0. (1.12
vA cs
This shows that the properties of the wave can be characterized by just two paramete
the angle of propagation ✓ and the ratio of sound speed to Alfvén speed, cs /vA
( /2)1/2 (cf. eqs. 1.65, 1.105). The phase speed does not depend on frequency: l

Figure 1.14.: Directions of fluid displacement (red) in the fast mode (top) and slow mode (bot-
Directionstom) at of fluid
low (left) displacement
and high (right) vA /cs . Grey: (red) in theof the
angular dependence fast
phase
Propagation diagrams
diagrams of of
the the fast(left)
mode (left)
modeand slows speed as in Fig. 1.12
mode (top) and slow mode 31 (bottom) at low (left) and
Figure 1.13.: Propagation fast mode and slow (right) for vA = cs .
mode Curves(right)(u) for vA=c
show the phase. speed
Curves
as in Fig.(u)
1.12. show the phase
Arrows: direction of the group
S
speed vg as a function of the direction (dotted) of the phase speed. high
The (right) vused
nomenclature A/csin. describing
Grey: angular
wave propagationdependence of the in
can be a bit confusing
speed, arrows the group speed vg as a function of the thephase speed
case of the Alfvén and slow waves. Because of their anisotropy, the wave vector
direction (dotted) of the phase speed. k is not the direction in which the energy of the wave flows. The case cos ⇥ 2
1 is
the Alfvén wave, the waves are nondispersive. Since they are anisotropic, however, the called ‘propagation perpendicular to B’, though in the Alfvén and the slow mode the
Poynting Flux in MHD
The Poynting flux S of electromagnetic energy:
S= c(E×B) /4π, (4.25)
is usually thought of in connection with electromagnetic waves in vacuum. It is in fact defined
quite generally, and has a MHD-specific interpretation. With the MHD expression for the electric
field, E = −v × B/c, we have
S= B×(v×B) /4π. (4.26)
S thus vanishes in flows parallel to B. Writing out the cross-products, and denoting by v⊥ the
components of v in the plane perpendicular to B, we have:
S = v⊥B2/ 4π. (4.27)
An MHD flow does not have to be a wave of some kind for the notion of a Poynting flux to apply.
It also applies in other time dependent flows, and even in steady flows
Also, the divergence of the Poynting flux S may be written as:

∇. (E × B) = -E. ∇ × B + B. ∇ × E
and transformed using Maxwell’s equations to:
-c∇. (E × B)/4π = E.j + ∂/∂t (B2/8π). (4.28)
The physical interpretation is that an inflow of electromagnetic energy ((E × B) produces
electrical energy (E.j) for the plasma an an increase in the magnetic energy (B2/8π). In turn the
electrical energy given to the plasma by the electromagnetic field may be rewritten after
substituting for E as:
E.j = j2 /σ + v.j × B/c. (4.29)
The electrical energy appears as heat by Ohmic dissipation and work done by the Lorentz force.
 MHD Equations
∂ρ dρ
+ ∇ •( ρ v) = 0 or + ρ∇ • v = 0 (Continuity Equation)
∂t dt
dv 1
ρ = −∇p + (∇ × B) × B + ρg + ρν∇ 2 v (Momentum equation)
dt 4π
d%p(
' γ * = 0 (Energy equation for an adiabatic fluid),
dt & ρ )
(where d/dt = ∂/∂t + v·∇).
∂B
= ∇ × (v × B) + η∇ 2B,
∂t
In the limit of infinite conductivity (η=0),
∂B/∂t =∇×(v×B) (induction equation)
=−B∇·v − (v·∇)B + (B·∇)v (since ∇·B = 0)
Linearized Wave Equation in an inhomogeneous medium
Consider a vertical magnetic field B0(x) ez in an unbounded medium in which the
fluid is initially at rest everywhere (i.e. v=0). The x-dependence of the magnetic
field leads to a similar dependence of the pressure p0 and density ρ0. The initial
equilibrium (static) state is such that:
d B02
p0 = p0 (x), ρ0 = ρ0 (x), (p + ) = 0, (5.1)
dx 0 8π
The last equation in (5.1) follows from the x-component of the momentum
equation for the initial equilibrium state.
Consider small departures from the initial equilibrium state of the form:
ρ = ρ0 + δρ, p = p0 + δp, v = δv, B = B0 + δB,
where δρ is a perturbation in the density and so on. Substituting these in the
equations of motion and induction, we find:

∂δv & 1 ) 1 1
ρ0 = −∇ (δ p + B0 ⋅ δB + + (B0 ⋅ ∇) δB + (δB ⋅ ∇) B0 . (5.2)
∂t ' 4π * 4π 4π
∂δB
= ∇ × (δv × B0 ). (5.3)
∂t
The continuity equation becomes
∂δρ
+ ∇ •( ρ0δv) = 0, (5.4)
∂t
We can relate δρ and δp, from the adiabatic energy equation as follows:

∂δ p !
2 ∂δρ
$
+ δ v ⋅ ∇p0 = cS # + δ v ⋅ ∇ρ0 &, (5.5)
∂t " ∂t %
where γ is the ratio of specific heats and cS= (γp0 / ρ0 )½ is the sound speed.
Let us define the following variables:
∂δv z B
Δ = ∇ ⋅ δ v, Γ = , pT = δ p + 0 δBz .
∂z 4π
From Eqs. (5.1)-(5.5), we find
∂δ p 2
$ dp ' ∂pT
= −ρ0cS Δ − & 0 )v x , = ρ0v A2 Δ − ρ0 (cS2 + v A2 )Δ, (5.6)
∂t % dx ( ∂t
and
∂δBx ∂δv x ∂δBy ∂δv y ∂δBz dB
= B0 , = B0 , = B0 (Γ − Δ) − ( 0 )δv x , (5.7)
∂t ∂z ∂t ∂z ∂t dx
where vA= B0 /√4πρ0 is the Alfvén speed.
The components of the momentum equation (5.2) give:
# ∂2 2 ∂
2 &
∂
ρ0 % 2 − v A 2 ( δv x = [ ρ0 (cS2 + v A2 )Δ − ρ0v A2 Γ], (5.8a)
$ ∂t ∂z ' ∂x
# ∂2 2 ∂
2 &
∂
ρ0 % 2 − v A 2 ( δv y = [ ρ0 (cS2 + v A2 )Δ − ρ0v A2 Γ], (5.8b)
$ ∂t ∂z ' ∂y
∂2δv z ∂Δ
2
= cS2 . (5.8c)
∂t ∂z
We now look for plane wave solutions of the form:
i (ωt−k y y −k z z)
δ v = v(x)e ,
and similarly for the other perturbed variables, where ky and kz are the y and z
components of the the wavenumber vector respectively and ω is the angular
frequency of the wave. We can now replace ∂/∂t in Eqs. (5.8) by iω, ∂/∂z by -ikz
and ∂/∂y by -iky. Eqs. (5.8) now can be reduced to a single equation for vx(x):
 d "$ ρ0 (x)[k z v A (x) − ω ] dv x &$
2 2 2
2 2 2
# 2 2
' − ρ 0
(x)[k v
z A
(x) − ω ]v x = 0, (5.9a)
dx $% [m0 (x) + k y ] dx $(
where
[k z2cS2 (x) − ω 2 ][k z2v A2 (x) − ω 2 ]
m02 (x) = 2 2 2 2 2
, (5.9b)
[c (x) + v (x)][k c (x) − ω ]
S A z T

2 cS2 (x) v A2 (x)

c (x) =
T 2 2
. (5.9c)
c (x) + v (x)
S A

For a uniform medium, when ρ0, cS, vA, cT and m02 are constants, Eq. (5.9)
becomes:
" d2v %
2 2 22 2
(k v − ω ) $$ 2 − (m0 + k y )v x '' = 0.
z A
x
(5.10)
# dx &
Eq. (5.10) has two possible solutions (for ky ≠ 0): either ω2 = kz2vA2, with vx(x)
arbitrary; or ω2 ≠ kz2vA2, and vx(x) satisfies:

d2v x
2
− (m02 + k y2 )v x = 0, (5.11)
dx
In a uniform unbounded medium, Eq. (5.10) admits plane wave solutions, with an x-
dependence e −ik x x satisfying:

(k z2v A2 − ω 2 )(m02 + k y2 + k x2 ) = 0,
the roots of which give the well-known dispersion relations for Alfvén waves
(ω2=kz2vA2) or magnetoacoustic waves discussed earlier.

For a non-uniform medium (assuming that the motions are independent of the y-
coordinate (i.e. ky=0 and the propagation is in the x-z plane), we see that vy is given
by (see Eq. 5.8b)
(k z2v A2 − ω 2 )v y (x) = 0,
and so unless vA is uniform (in which case Alfvén wave propagation is possible), we
have vy = 0. With ky=0 and vy = 0, we see that for 2-D motions v=(vx,0, vz), Eq. (5.9)
becomes:

d "$ ρ0 (x)[cS (x) + v A (x)][k z cT (x) − ω ] dv x &$

2 2 2 2 2
2 2 2
# 2 2 2
' − ρ 0
(x)[k v
z A
(x) − ω ]v x = 0, (5.12)
dx %$ [k z cS (x) − ω ] dx ($
Equation (5.12) describes the propagation of 2-D waves in a compressible non-
uniform medium (varying in the x-direction).

The vz component of the perturbation is related to vx as follows:

2 2 2 dv x2
(k c − ω )v z = ik c
z S z S
,
dx
and the amplitude of the total pressure perturbation using Eq. (5.6) is given by :
i
pT = − 3 (cS2 + v A2 )(k z2cT2 − ω 2 ) ρ0 Δ
ω
2 2 2
i ρ0 2 2 (k c − ω ) dv x
= (cS + v A ) 2 2
z T
2
.
ω (k z cS − ω ) dx
Eq. (5.12) has singular solutions at locations x=xs, where ω2=kz2cT2. Eq. (5.9) is
singular at ω2=kz2cT2(x), kz2vA2(x). Such singularities have important physical
repercussions such as continuous spectra.
Waves in a Sharply Structured Medium
Consider the behaviour of modes in a medium where the non-uniformity takes the
form of a single magnetic interface, the field being uniform (but of different
magnitude) on either side of the inteface. Suppose that the basic state of the
plasma is one in which the magnetic field changes discontinuously from B0 to Be, so
that
! B , x > 0,
#
B0 (x) = " e
#$ B0 , x < 0,

where B0 and Be are constants. From Eq. (5.1) which expresses pressure
balance, we have at the interface x = 0, the following relation:

Be2 B02
pe + = p0 + . (5.13)
8π 8π

Now the medium on either side of the discontinuity at x = 0 is uniform and so Eq.
(5.10) holds with constant coefficients.Thus, we can have an Alfvén wave in x < 0
with ω2=kz2vA2, vx(x) and vy(x) arbitrary, and vz, p, pT, and Δ are all zero. For
example, an Alfvén wave can propagate in x < 0 without disturbing the interface at
x = 0 or the fluid in x > 0.
In addition to the Alfvén wave there are magnetoacoustic modes which are
governed by Eq. (5.11). Thus, in x < 0, we have:
d2v x
2
− (m02 + k y2 )v x (x) = 0, x < 0, (5.14)
dx

where the constant m02 maybe positive or negative (for real ω2 and kz2). Similar
conditions apply in x > 0 with me2 being defined analogously.

Surface waves owe their existence to the presence of the discontinuity in B0(x)
and may arise if (m02 + ky2) and (me2 + ky2) are both positive. Solving Eq. (5.14)
and the corresponding equation in x > 0 for vx(x), gives:

" −(me2 +k y2 )1/2 x,

$ α ee x > 0,
v x (x) = # 2 2 1/2
(5.15)
$ α e(m0 +ky ) x, x < 0,
% 0
(m02 + ky2)½ > 0 and (me2 + ky2)½ > 0. In choosing the form of Eq. (5.15) we have
imposed the condition that vx(x) tends to zero at x = ± ∞ (we exclude laterally
propagating waves). Thus only surface modes arise.
Across the interface at x =0, vx(x) and pT(x) must be continuous. Thus, αe = α0,
while continuity of pT at x =0 yields:

ρ0 (k z2v A2 − ω 2 )(me2 + k y2 )1/2 + ρe (k z2v Ae

2
− ω 2 )(m02 + k y2 )1/2 = 0, (5.16)

valid for both (m02 + ky2)½ and (me2 + ky2)½ positive. We have written
(k z2cSe
2
− ω 2 )(k z2v Ae
2
− ω 2)
me2 (x) = 2 2 2 2 2 2
,
k c v
z Se Ae
− (c + v )ω
Se Ae

where cSe= (γpe / ρe )½ and vAe= Be /√4πρe are the sound and Alfvén speeds
respectively in x > 0. Eq. (5.14) is the dispersion relation for surface waves at a
single magnetic interface. Eq. (5.14) may also be rewritten as

ω2 2 R 2 2 2 1
2
= v A
− (v A
− v Ae
) = v Ae
+ (v A2 − v Ae
2
), (5.16) '
kz R +1 R +1
where
1/2
" m2 + k 2 %
R = ( ρe / ρ0 ) $ 02 y
' >0
$m + k2 '
# e y &

is a function of ω2. From the above we clearly see that the longitudinal phase speed
of a surface wave lies between vA and vAe.
There are a number of special cases of the general dispersion relation. Consider
the incompressible limit (cS → ∞), we obtain m2 → kz2 and from Eq. (5.16) we
have:
1/2
ω ! ρ0v A + ρev Ae $
2 2
= ## && .
k z " ρ0 + ρe %

Waves in Magnetic Flux Tubes

We consider the generalization of Eqs. (5.9) in cylindrical coordinates (r, θ, z), with
B0=B0(r) and ρ0=ρ0(r) and replacing vx with vr where
vr (r, θ, z,t) = vr (r) exp [i (ωt + nθ – kz z)] (5.17)
for mode number n = 0,1,2,….Then Eq. (5.9a) in cylindrical co-ordinates becomes
" &
d $$ ρ0 (r )[k z v A (r ) − ω ] 1 d(rv r ) $$
2 2 2
2 2 2
# 2
' − ρ 0
(r )[k z
v A
(r ) − ω ]v r = 0, (5.18)
dr $ n r dr $
[m 2 (r ) + 2 ]
$% r $(

where m2 is given by Eq. (5.9b).

Isolated Flux Tubes
Consider an isolated flux tube of radius a, for which

!# B , r < a, !# ρ , r < a,
0 0
B0 (r) = " , ρ0 (r) = " , (5.19)
$# 0, r > a, ρ
#$ e , r > a,

To determine the modes, it is convenient to solve for pT

instead of vr, which satisfies the following equation:
"$
1 d 1 dpT &$ n2
2 2 2 2
ρ0 (k v − ω )
z A
# 2 2 2
r ' = (m + 2 )pT , (5.20)
r drρ (k
%$ 0 z Av − ω ) dr $( r
For a medium in which ρ0, vA and m are constants, this is simply Bessel’s
equation, 2
d pT 1 dpT 2
n 2
2
+ − ( 2
+ m )pT = 0, (5.21)
dr r dr r
The relevant solution of Eq. (5.21) for region inside the flux tube is:
pT = A0 In (m0 r), r < a, (5.22)
where In denotes a modified Bessel function of order n. There is no restriction on
the sign of m02; modes with m02 > 0 are evanescent whereas m02 < 0 have an
oscillatory character. Modes with m02 > 0 are referred to as surface waves while
those with m02 < 0 are called body waves.
where m0 is given by

[k z2cS2 − ω 2 ][k z2v A2 − ω 2 ]

m02 = 2 2 2 2 2
, (5.9b)
[c + v ][k c − ω ]
S A z T

cS2 v A2
cT2 = 2 2
. (5.9c)
c +v
S A

Here cT is referred to as the cusp or tube speed.

The external atmosphere outside the flux tube is field
free. We consider Eq. (5.21) with vA = 0 and pT is the
perturbed gas pressure. Requiring pT → 0 as r → ∞, we
select the modified Bessel function Kn as the solution of n=1
Eq. (5.21), which yields
pT = A0 Kn (me r), r > a, (5.23)

where
2 ω2 2
m =k − 2 . (5.24)
e
cSe
z n=0

The requirement of a declining pressure function outside the tube means that we
are imposing the constraint me2 > 0 and so ω2 < kz2cSe2.
The two solutions (5.22) and (5.23) are matched by requiring that the radial velocity
component vr and the pressure perturbation pT be continuous at the interface r = a.
This results in the following dispersion relation:
2 In! (m0a) 2 2 2 K n! (mea)
ρeω m0 + ρ0 (k z v A − ω )me = 0, (5.25)
In (m0a) K n (mea)

where the dash denotes the derivative of the Bessel function with respect to x.
 vA < cSe vA > cSe

1990ApJ...3
Body modes have phase speeds ω/kz that that
are between cT and cS and correspond to slow
modes. The slow body modes may be viewed as
waves that are constrained within the tube,
bouncing from side to side of the tube as they
propagate along its interior. Body modes have
m02 < 0 and so are more conveniently obtained

1990ApJ...348..346E
from Eq. (5.25) if we rewrite as follows:

Jn! (n0a) K ! (m a)
ρeω 2n0 + ρ0 (k z2v A2 − ω 2 )me n e = 0, (5.26)
Jn (n0a) K n (mea)
where vA > cSe
2 2 2 2 2 2
(ω − k c )(ω − k v )
n02 = −m02 = 2
z S
2 2
z A
2 2
. (5.27)
(c + v )(ω − k c )
S A z T

In addition there are also slow surface waves. These can

be understood by taking the incompressible limit of Eq.
(5.25) (cS → ∞ and cSe → ∞). The result is

k z2v A2 " ρ % In( ( k z a)K n ( k z a)

2
= 1− $$ e '' , (5.28)
ω # 0 & In ( k z a)K n ( k z a)
ρ (
The phase speed as a function of
In the limit kza → 0 (thin tube limit), we find for the kink mode (n=1): kz a for waves in an isolated mag-
! 1 $ netic flux tube. Solid curves are
ρe 2 2
ω ~ k zcK #1+ ( )k z a K 0 ( k z a)& sausage modes (n = 0), dashed
" 2 ρ e
+ ρ 0 % curves are kink modes (n = 1) and
dotted curves are n = 2.
where cK2 = ρ0/(ρ0+ρe) vA2 is referred to as the kink speed (cK < vA).
Embedded Flux Tubes
Consider an equilibrium magnetic field and density of
the form:
! B , r < 0, ! ρ , r < 0,
# 0 #
B0 (r ) = " , ρ0 (r ) = " 0 , (5.29)
#$ Be , r > 0, #$ ρe , r > 0,

The dispersion relation for the case of an isolated flux

tube can be generalized to the present case by Magnetic flux tube idealized as
replacing me2 in Eq. (5.24) with: a cylinder of radius a with a
uniform vertical magnetic field
of strength B0 embedded in a
2 (k z2cSe
2
− ω 2 )(k z2v Ae
2
− ω 2) medium with a uniform vertical
m =
e 2 2 2 2 2
. (5.30)
(c + v )(k c − ω )
Se Ae z Te
field of strength B e . The
subscripts refer to equilibrium
It is straightforward now to obtain the dispersion relation quantities inside and outside
the flux tube respectively.
which is the generalization of Eq. (5.26) given by:

2 2 2 Jn" (n0a) 2 2 2 K n" (mea)

ρe (k v
z Ae
− ω )n0 + ρ0 (k z v A − ω )me = 0, (5.31)
Jn (n0a) K n (mea)

with n02 given by Eq. (5.27). When vAe = 0, Eq. (5.31) reduces to Eq. (5.26).
Coronal Tubes
In the solar corona, the Alfvén speed
typically exceeds the sound speed. Then
Eq. (5.31) possesses two sets of modes,
namely fast and slow body modes that
can be seen in the figure.
There are no surface waves (n02 < 0 ).
Both sets of modes are dispersive, with
the fast mode being strongly dispersive.
The fast body waves occur only if vAe > vA
i.e. with vAe,vA > cS, cSe, fast body waves
occur only for flux tubes that satisfy the The phase speed ω/kz as a function of kz
condition vAe > vA, and so body waves a for waves in a low β magnetic cylinder
embedded in an external field (typical in
occur in region of low Alfvén speeds.
the solar corona). Solid curves are
sausage modes (n = 0) and dashed
When B0 = Be, regions of low Alfvén speed curves are kink modes (n = 1).
correspond to regions of high gas density.
Coronal loops act as wave guides for fast
magnetoacoustic waves.
 Summary
MHD waves in magnetic tubes
• Dispersion relation ω(k) for tubes
2 2 2K n' (mea) 2 2 2 In' (moa)
ρ0 (k z v − ω )me
A
= ρe (k z v Ae − ω )m0
K n (mea) In (moa)
m02 > 0 à surface waves
2 2 2 K n' (mea) 2 2 2 Jn' (noa)
ρ0 (k v − ω )me
z A
= ρe (k z v Ae − ω )n0
K n (mea) Jn (noa)
m02 = -n02 < 0 à body waves. Note n=0
refers to sausage, n=1 to kink modes,
etc.

Main modes: n=0

• Fast sausage (|B|, ρ)
n=1
• Fast kink (almost incompressible)
Torsional waves
• Alfvén torsional (incompressible)
• Slow (acoustic) type (ρ, v)
 Summary ………………………………contd
MHD waves in magnetic tubes for typical conditions in the chromosphere

Note n=0 refers to sausage and n=1 to

kink modes.
Torsional waves

Main modes: n=0

• Alfvén torsional (incompressible) n=1

• Fast sausage (|B|, ρ)

• Fast kink (almost incompressible)
• Slow (acoustic) type (ρ, v)