Reflections On Magnetohydrodynamics: WWW - Moffatt.tc
Reflections On Magnetohydrodynamics: WWW - Moffatt.tc
Reflections On Magnetohydrodynamics: WWW - Moffatt.tc
tc
Reflections on Magnetohydrodynamics
H. K. MOFFATT
1 Introduction
Magnetohydrodynamics (MHD) is concerned with the dynamics of fluids
that are good conductors of electricity, and specifically with those effects
that arise through the interaction of the motion of the fluid and any ambient
magnetic field B ( x ,t ) that may be present. Such a field is produced by electric
current sources which may be either external to the fluid (in which case we
may talk of an ‘applied’ magnetic field), or induced within the fluid itself.
The induction of a current distribution j ( x , t ) by flow across the field B is the
result of Faraday’s ‘law of induction’. The resulting Lorentz force distribution
F ( x , t ) = j A B is generally rotational, i.e. V A F # 0, and therefore generates
vorticity in the fluid. There is thus a fundamental interaction between the
velocity field v and the magnetic field B, an interaction which not only leads
to modification of well-understood flows of ‘conventional’ fluid dynamics,
but also is responsible for completely new phenomena that simply do not
exist in non-conducting fluids.
There are three major fields of application of magnetohydrodynamics,
which will be discussed in this survey in the following order.
347
348 H . K . Mofatt
2 Fundamental principles
The electromagnetic field is governed by Maxwell’s equations, and it is
legitimate in all the above contexts to adopt the ‘magnetohydrodynamic
approximation’ in which displacement current and all associated relativistic
effects are neglected. The magnetic field is then related to current by Ampere’s
equation
V A B = poj with V*B =0, (2.1)
where po is constant (471 x lOP7 in SI units). Moreover B evolves according
to Faraday’s law of induction which may be expressed in the form
2Bl2t = -V A E, (2.2)
where E ( x , t ) is the electric field in the ‘laboratory’ frame of reference. The
current in this frame is related to E and B by Ohm’s law:
j = o(E +U AB), (2.3)
where u(x,t) is the velocity field and cr is the electrical conductivity of
the fluid. Like viscosity, o is temperature dependent, and will therefore in
general be a function of x and t in the fluid. We shall, however, neglect such
variations, and treat o as a given constant fluid property. Note that the field
+
E’ = E U A B appearing in (2.3) is the electric field in a frame of reference
*
moving with the local fluid velocity U.
From the above equations, we have immediately
(2.4)
350 H. K. Mofutt
where v] = (poo)-’, the ‘magnetic diffusivity’ (or ‘resistivity’) of the fluid.
Equation (2.4) is the famous ‘induction equation’ of magnetohydrodynamics,
describing the evolution of B if u ( x , t ) is known. The equation has a mar-
vellous generality : it holds quite independently of the particular dynamical
forces generating the flow (e.g. whether these are of thermal or compositional
origin, whether the Lorentz force is or is not important, whether Coriolis
forces are present or not); it holds also whether U is incompressible (V.u = 0)
or not. Equation (2.4) may be regarded as the vector analogue of the scalar
advection-diffusion equation
do
-= V A ( u A o ) + v V 2 o .
;t
The analogy is incomplete in that o is constrained by the relationship
o = V A U, whereas B and U in (2.4) suffer no such constraint: there is in
effect far more freedom in (2.4) than there is in (2.6)! Nevertheless a number
of results familiar in the context of (2.6) do carry over to the context of (2.4).
Rm Rm, , (2.8)
where R,, is a critical magnetic Reynolds number, presumably of order
unity; it must immediately be added that a condition such as (2.8) may
be necessary but by no means sufficient for dynamo instability, which may
require additional much more subtle conditions on the field U .
(c) R,>> 1: here we are into the domain of ‘nearly perfect conductivity’, in
which inductive effects dominate diffusion. The limit R,n + x (or CT -+ x
or q + 0) may be described as the perfect-conductivity limit. In this formal
limit, B satisfies what is known as the ‘frozen-field equation’
ZB
-= V A ( U A B) ,
2t
which implies that the flux @ of B across any material (Lagrangian) surface
S is conserved:
d@
-=0
dt
where @= 1 BqndS. (2.10)
s
XAv= AaBdV, (2.15)
dt
/A*BdV = 1 .BdV +/A. - (-)
D B pdV,
Dt P
(2.16)
AIA-BdV
dt
=
s V*[B(v-A-q)]dV
= i(n.B)(n.A-q)dS=O. (2.17)
Note that this result holds whether the fluid is incompressible or not; it
merely requires that the fluid be perfectly conducting.
This invariant admits interpretation in terms of linkage of the B-lines
(which are frozen in the fluid) (Moffatt 1969). To see this, consider the
simplest ‘prototype’ linkage (figure 1) for which B is zero except in two flux
7 Reflections on Magnetohydrodynamics 353
Figure 1. Linked flux tubes; here the linkage is right-handed and the linking number
n = 1.
(2.18)
The two line integrals are zero if C1 and C2 are unlinked (more strictly if
the flux of B across any surface spanning either C1 or C2 is zero); if C1 and
C2 are linked, with linking number n, then, from (2.18)
(2.19)
Oil
1.
NaK
Figure 2. Sketch of Northrup’s (1907) experiment: current flows through the NaK
under a layer of oil; the constriction leads to a Lorentz force distribution which
depresses the oil/NaK interface. ( a ) Plan; ( b ) elevation.
>
B
4 >
Figure 3. The pinch effect: axial current in the cylinder generates an azimuthal
magnetic field; the resulting magnetic pressure gradient causes radial contraction of
the cylinder.
j = -oVcp, (4.1)
where
8=0
Figure 5. The weldpool problem: the Lorentz force drives the fluid towards the axis
8 = 0 and a strong jet flow develops along this axis; the flow is subject to instability
involving swirl about this axis (BojareviEs et al. 1989).
This force, directed towards the axis 6 = 0, tends to drive the fluid towards
this axis; being incompressible, the fluid has no alternative but to flow out
along the axis in the form of a strong axisymmetric jet, which becomes
increasingly ‘focused’ for decreasing values of the fluid viscosity.
Experimental realization of this flow indicates a behaviour that is not
yet fully understood: the flow is subject to a strong symmetry-breaking
instability in which the fluid spontaneously rotates, in one direction or the
other, about the axial direction 6 = 0. It is believed that this rotation controls
the singularity of axial velocity that otherwise occurs if the dimensionless
parameter
K = poJ2/p~j2 (4.6)
exceeds a critical value K , of order 300; a recent discussion of this perplexing
phenomenon is given by Davidson et al. (1999).
I 1 Anodes
v1 / t f / f I
I I I
.pF
Cathode
Figure 6. Sketch of the aluminium smelting process (Moreau 1990). The interface
between the liquid cryolite and the molten aluminium is subject to instabilities of
magnetohydrodynamic origin, which limit the efficiency of the process.
limited by instabilities that may lead to contact between the aluminium and
the anodes, a 'short-circuiting' that would terminate the process. This is a
billion-dollar industry for which an understanding of the fundamentals of
magnetohydrodynamics would appear to be a first essential.
[^
B ( x , t ) = Re B(x)e-i"t
1, (4.71
C’
Figure 7. Stirring induced by AC fields; the sketch indicates the flow pattern that
would be driven by a travelling magnetic field from the source coils C and a
stationary AC field from the coils C ; a wide range of patterns of stirring may be
generated by appropriate engineering of the external coils and appropriate choice
of field frequency U .
and is small when CO is large. Outside the conductor, the field &x) is given
by
2 @ / 2 n = O on S . (4.10)
In effect, the field is perfectly excluded from the conductor in this limit. Of
course, the potential @ has ‘prescribed’ singularities at the external coils. Let
us regard this ‘external’ problem as solved; then Bs = (V@)S is known, as a
tangential field on the surface S . Note that the surface divergence of Bs is
non-zero in general; in fact
(4.11)
Within the skin inside the conductor, the field B and hence j^ can be
calculated by standard methods; the mean force f = ;Re(? A h) may then
be found; here, the * denotes the complex conjugate. Details may be found
7 Reflections on Magnetohydrodynamics 363
in Moffatt (198521);the result is
(4.12)
where i is a normal coordinate directed into the conductor. The first term
here is a strong inwardly directed normal component, which is responsible
for the shaping of any part of S that is free to move in the normal direction.
By contrast, stirring of the fluid is associated with the curl of F , given by
(4.13)
where
(4.14)
so that asymptotically
(4.16)
What this means is that, under the influence of the magnetic field, the no-slip
condition must be replaced by a condition of ‘prescribed tangential velocity’
as given by (4.16) for the driven flow in the interior of the conductor.
A corresponding analysis using a free-surface condition 2u/2C: = 0 on
5 = 0 leads to an effective ‘stress’ condition
7s = -PV ( 2 ~ l 2 iI;==) = $P~MQ~(.) (4.17)
as a boundary condition for the interior flow. Note that this effective stress
is independent of kinematic viscosity v.
Thus, in the high-frequency limit, the influence of the magnetic field is
confined to the magnetic boundary layer, or skin, in such a way as to simply
replace the no-slip (or zero-stress) conditions on S by an effective slip (or
effective stress) distribution on S, which is then responsible for generating an
internal flow, a flow whose topology will clearly be influenced, and indeed
364 H . K . Moflatt
controlled, by the function Qs(x), which is determined through (4.14) by the
surface field B s ( x ) .
As the frequency is decreased, the skin depth JLwincreases, and the above
simple description ceases to be valid; the force distribution penetrates more
and more into the interior of the fluid, and extends throughout the fluid
when 8-v increases to the scale L of the fluid domain.
Hence, from (4.14) and (4.15), the effective slip velocity just inside the skin is
21 = (11B;iPofwva) ee. (4.20)
The resulting motion is a rigid body rotation of the fluid inside this skin,
with angular velocity
SZ = YBi/popova2, (4.21)
a result obtained originally by Moffatt (1965).The viscous stress and resulting
couple G (per unit length of cylinder) acting on the cylinder are easily
calculated, with the result
Bi
G=2rr(?) -. (4.22)
POP
Remarkably, this couple, although of viscous origin, is independent of v ;
this is because the core angular velocity, given by (4.21), is proportional
to vl.In Braunbeck’s experiment, it is the torque (4.22) that is ultimately
in equilibrium with the torque transmitted by the torsion wire, which is
proportional to the net angle of rotation; thus measurement of this angle
provides a means of determination of 11, and hence of the conductivity g.
There are of course a number of limitations of the above type of analysis
that should be borne in mind. First, the high-frequency approximation is
valid only if 6.v << a, i.e. cc) >> q / a 2 . This is not however a serious restriction;
7 Re$ections on Magnetohydvodynamics 365
the problem can be solved exactly for arbitrary CO in terms of Bessel functions.
More seriously however, the analysis fails if the field strength Bo becomes
so strong that Q in (4.21) becomes comparable with C O ; for it is in fact not
the absolute angular velocity w of the field that is relevant to the induction
of currents, but rather the relative angular velocity w - Q between field and
fluid. This strong field situation requires a major modification of approach
(see, for example, Moreau 1990, where problems of this type are extensively
treated).
O=VACO-V
(;-+-U
?>
+F-vVACO (4.24)
(4.25)
i.e. inertia and pressure forces play no part in the equilibrium that is estab-
lished. In practice, the viscosity in liquid metals is small; put another way,
366 H . K . MofSatt
(a)
Axis of
symmetry
Levitated molten
current
the Reynolds number of the flow that is necessarily driven in the interior of
a levitated droplet of radius of the order of 10 mm or greater is very large.
It seems likely therefore that this interior flow will in these circumstances be
turbulent. The ‘great agitation’ referred to by Northrup (see $4.1 above) also
7 Reflections on Magnetohydrodynamics 367
presumably indicated a state of turbulent flow just inside the liquid metal -
5 Dynamo theory
Dynamo theory is concerned with explaining the origin of magnetic fields
in stars, planets and galaxies. Such fields are produced by currents in the
interior regions, which would in the normal course of events be subject to
ohmic decay, in the same way that current in an electric circuit decays if not
maintained by a battery. This decay can however be arrested, and indeed
reversed, through inductive effects associated with fluid motion; when this
happens, the fluid system acts as a self-exciting dynamo. The magnetic field
grows spontaneously from an arbitrary weak initial level, in much the same
way as any other perturbation of an intrinsically unstable situation.
The possibility of such ‘dynamo’ instabilities may be understood with
reference to the induction equation (2.4), which governs field evolution if
the velocity field v ( x , t ) in the fluid is regarded as ‘given’. Let us suppose
for simplicity that the fluid fills all space, and that the velocity field is
steady, i.e. U = u ( x ) . We must suppose also that B has no ‘sources at
infinity’ which would mitigate against the concept of an internally generated
dynamo.
We may seek ‘normal mode’ solutions of (2.4) of the form
where, by substitution,
pB = v A (U A B ) + yv2B. (5.2)
When coupled with the requirement that &x) should be either a local-
ized field of finite energy or, for example, a space-periodic field if U is
space-periodic, this constitutes an eigenvalue problem which (in principle)
determines a sequence of possibly complex eigenvalues p1, p2,. . . , which may
be ordered so that
Repl 3 Rep2 3 Rep3 3 ... , (5.3)
and corresponding ‘eigenfields’ B~(x),
&(x), . , . . If Repl > 0, then the corre-
sponding field
B ( x ,t ) = Re [B~(x)
ePlf] (5.4)
exhibits dynamo behaviour: it grows exponentially in intensity, the growth
368 H. K. Mofatt
being oscillatory or non-oscillatory according to whether Imp1 # 0 or = 0.
The mode of maximum growth-rate is clearly the one that will emerge from
an arbitrary initial condition in which all modes may be present.
If this exponential growth occurs, then it can persist only for as long as
the velocity field u ( x ) remains unaffected by the Lorentz force. This is the
‘kinematic phase’ of the dynamo process. Obviously, however, the Lorentz
force increases exponentially with growth rate 2Re p1 (considering only the
mode (5.4)) and so ultimately the back-reaction of the Lorentz force on
the fluid motion must be taken into account. This is the ‘dynamic phase’ of
dynamo action in which the nature of the supply of energy to the system (via
the dynamic equation of motion) must be considered. While great progress
has been made over the past 50 years towards a full understanding of the
kinematic phase, the highly nonlinear dynamic phase has proved far more
intractable from an analytical point of view, and is likely to remain a focus
of much research effort, both computational and analytical, over the next
few decades.
f(R,) - -
by contrast, is one for which f(R,) + 0 as R, -+ E; for example, if
R,1’2 as R, -+ x,then p ( c o / l ~ ) R ~ ’ ”
and
, diffusivity y continues
to influence the growth rate even in the limit y + 0. The distinction is an
important one because, on the galactic scale, R, is extremely large, and a
slow dynamo is likely to have little relevance in such contexts. This has led to
an intensive search for dynamos that can legitimately be described as fast (see
Childress & Gilbert 1995); however, in the strict sense indicated above, no
such dynamo has yet been found! All known dynamos are slow; diffusivity
remains important no matter how small y may be. The situation is again
somewhat analogous to that governing the vorticity equation in turbulent
flow: viscous effects remain important (in providing the mechanism for
7 Reflections on Magnetohydrodynamics 369
dissipation of kinetic energy) no matter how small the kinematic viscosity 1’
may be.
When a dynamo enters the dynamic phase (assuming that sufficient time
is available for it to do so) the distinction between fast and slow behaviour
disappears; in either case, the growth rate must decrease, ultimately to zero
when an equilibrium between generation of magnetic field by the (modified)
velocity field and ohmic dissipation of magnetic field is established.
there is some source of energy for this turbulence (e.g. through some random
stirring mechanism) so that v is statistically stationary in time. It is then
natural to adopt the notation ( . . . ) for a time average, over any interval
long compared with the time scale to = lo/t'o characteristic of the energy-
containing eddies of the turbulence; here lo is the scale of these eddies and
00 the r.m.s. value of v . For simplicity, we may suppose that ( U ) = 0.
The field B ( x ,t ) may be decomposed into mean and fluctuating parts:
+
where the terms indicated by . . . involve higher derivatives of Bo, and may
presumably be neglected when the scale of variation of Bo is large. It is clear
from the structure of (5.12) that P must be interpreted as an 'eddy diffusivity'
associated with the turbulence (although there is no guarantee from the above
treatment that P must invariably be positive!). The first term on the right of
(5.12) will however always dominate the evolution, provided x f 0 and the
scale of Bo is sufficiently large. Before going further it is therefore essential
to find a means of calculating 3 explicitly and determining the conditions
under which this key parameter is definitely non-zero.
372 H . K . MofSatt
5.4 First-order smoothing
To do this, it is legitimate to consider the situation in which Bo is constant;
the reason for this is that tl is independent of the field Bo(x,t), and we
are free to make any assumption about this field that simplifies calculation
of a. The assumption that Bo is constant is equivalent to considering the
conceptual limit L + sc, T + sc.
Under this condition, the fluctuation equation (5.8) becomes
c‘b
- = (Bo *
2t
V)U + V A (v A b - 8)+ yV2b. (5.13)
and Bo = (O,O, Bo) (see figure 9). Note that the vorticity associated with
(5.14) is
w=VAv=kv, (5.15)
and the associated helicity density,
(5.18)
For this case of a single ‘helicity’ wave (5.14), v A b turns out to be uniform,
so the awkward term in (5.13) is identically zero! It follows from (5.18) that
7 Reflections on Magnetohydrodynamics 373
B = rBo, where
(5.19)
It is obviously desirable to write the result in this form, both left- and
right-hand sides of the equation being pseudo-scalars.
It is important to note the origin of the ‘r-effect’ contained in (5.19): it
is the phase-shift between b and v caused by molecular diffusivity y which
leads to a non-zero value for r ; it is a strange fact that, although diffusion is
responsible for the decay of magnetic field in the absence of fluid motion, it
is also responsible for the appearance of an &-effectwhich, as shown below,
is a vital ingredient of the self-exciting dynamo process.
The velocity field (5.14) is of course rather special; it should be clear how-
ever that if we consider a random superposition of such waves with wave-
vectors k , frequencies CO, and amplitudes 6(k,CO) isotropically distributed,
then, provided the awkward term of (5.13) can be neglected, a result gen-
eralizing (5.19) can be obtained; all the contributions from different wave
374 H. K. Mofatt
where X(k,co) is the helicity spectrum function of the velocity field U , with
the property
3
The factor appears in (5.20) from averaging over all directions. The main
thing to note again is the direct relationship between x and the helicity of
the turbulent field.
This theory, usually described as ‘first-order smoothing’ theory, is limited
to circumstances in which, as indicated above, lbl << IBol. This condition is
satisfied if R, << 1; it is also satisfied under the alternative condition
Ivok/ol << 1 (5.22)
for values of k and co making the dominant contributions to the field of
turbulence. The condition (5.22) is relevant in a rapidly rotating system in
which the turbulence is more akin to a field of weakly interacting inertial
waves whose frequencies co are of the order of the angular velocity fi of the
system; in this context, the condition (5.22) is one of small Rossby number.
If neither of the above conditions is satisfied, then it is not legitimate to
neglect the awkward term in (5.13); no fully satisfactory theory is as yet
available for the determination of a in these circumstances. Nevertheless,
one may assert that the key property of turbulence required to provide a
non-zero value for a is that its statistical properties should lack reflectional
symmetry, i.e. should be non-invariant under change from a right-handed to
a left-handed frame of reference; it is this same property that is necessary
-
to provide non-zero mean helicity J? = (U U),which is indeed a simple
measure of ‘lack of reflectional symmetry’, and so a loose connexion between
x and .Atis to be expected.
The concept of isotropic turbulence that lacks reflectional symmetry is
quite novel! One often thinks of a turbulent vorticity field in pictorial terms
as like a random field of spaghetti; to picture turbulence lacking reflectional
symmetry, think rather of a pasta in which each pasta element is twisted
with the same sense of twist - say right-handed: if the vorticity field follows
the sense of these elements, it remains statistically isotropic but acquires
positive helicity. For a more realistic example, it suffices to consider Benard
convection in a rotating system (see, for example, Chandrasekhar 1961); it
7 Rejections on Magnetohydrodynamics 375
turns out that the average of v w over horizontal planes is non-zero and
antisymmetric about the centreplane (Moffatt 1978, chap. 10).
(5.23)
-
that K is chosen to have the same sign as x, i.e. xK > 0, and that IKI is
small, so that the scale L K-’ of B is large compared with the scale lo
of the underlying turbulence that gives rise to the x-effect. Of course (5.23)
implies that
v2 B = -V A(VAB) = -K~B, (5.26)
and so (5.23) becomes
(5.27)
so that
(5.31)
-
assuming 2 vi/lo, and note that in this lOw-Rm situation, p << q (actually
/3 = O(Ri)q), then the scale of maximum growth rate is given (from (5.29))
by
L - - K-‘ RG210. (5.32)
Hence a new magnetic Reynolds number Am defined in terms of L rather
than 10 is given by
- = Luo
Rm ~
Y
- R i 2 Rm - Ril. (5.33)
The condition Rm << 1 implies that R m >> 1: the field grows on a scale L at
which the corresponding R m is large.
We can now summarize the implications of the above discussion. If a
conducting fluid is in turbulent motion, the turbulence being homogeneous
and isotropic but having the crucial property of ‘lack of reflectional symme-
try’ (a property that, as observed above, can be induced in a rotating fluid
through interaction of buoyancy-induced convection and Coriolis forces),
then a magnetic field will in general grow from an arbitrarily weak initial
level, on a scale L large compared with the scale 10 of the turbulence. This is
one of these remarkable situations in which ‘order arises out of chaos’, the
order being evident in the large-scale magnetic field. It may be appropriate
to follow the example of Richardson (1922) by summarizing the situation in
rhyme :
Conuection and diffusion,
In turb’lence with helicity,
Yields order jkom confusion
In cosmic electricity!
This totally general principle applies no matter what the physical context
may be, whether on the planetary, stellar, galactic or even super-galactic
scale. It is this generality that makes the approach described above, which
derives from that pioneered by Steenbeck et al. (1966), so intensely appealing.
(5.34)
and becomes strong when the local mean field Bo grows strong so that
N >> 1. It has been shown (Moffatt 1967) that under this condition, and
provided NR; << 1, turbulence that is initially isotropic becomes locally
two-dimensional (invariant along the direction of Bo), with kinetic energy
ultimately equally partitioned between the components ( U , U ) perpendicular
to Bo and the component w parallel to Bo, i.e.
(w2)= ( U 2 ) + (02) * (5.35)
This obviously has implications for the r-effect, which is suppressed in mag-
nitude and which no longer remains isotropic. The effects can be exceedingly
complex, and it would be inappropriate to attempt to describe them here;
there can be no question however that, if the turbulence is maintained by a
prescribed force distribution f(x,t ) , then the large-scale magnetic field must
ultimately saturate at a level determined in principle by the statistics of
-
this forcing (in particular by the mean-square values ( f 2 ) and (f V A f ) ) in
conjunction with the relevant physical properties of the fluid, namely v and y ~ .
The second feature to note is that on scales of order 10 and smaller,
the spectrum T ( k ) of field fluctuations b is related to the spectrum E ( k )
of v in a way that can be derived from the fluctuation equation (5.8).
This calculation was first carried out by Golitsin (1960) (see also Moffatt
- -
1961); it shows that if there is an inertial range of wavenumbers in which
E ( k ) k - 5 / 3 (see Chapter 5), then in this range, r(k) k-l1l3. This is one
prediction where theory is now corroborated by experiment: Odier, Pinton
& Fauve (1998) found, in experiments on liquid gallium with values of R,
378 H . K . Moffatt
I t
Figure 10. Generation of toroidal fields by differential rotation: the poloidal field is
‘gripped’ by the flow, and ‘cranked’ around the axis of symmetry (see, for example,
Moffatt 1978).
in the range 1.3 to 15, that r(k)scales like k W , with ,u = 3.7 0.2, neatly
embracing the value 11/3 = 3.66. In a branch of magnetohydrodynamics in
which experimental results are sparse, this result stands out, and gives hope
that more central aspects of dynamo theory may also soon be subject to
experimental investigation and verification.
a-effect J
The ct-effect here operates through the action of helical convection (i.e.
convection influenced by Coriolis forces), which acts on the toroidal field in
much the same manner as described in $5.4 above. The non-axisymmetric
character of such a convective process is what allows the cto-dynamo to
escape the strait-jacket of Cowling’s theorem.
Many variants of the xo-dynamo, involving particular distributions of
~ ( r0), and o ( r , 0), have been investigated (see particularly Krause & Radler
1980), and much effort has been applied over the last 25 years to incorporate
the difficult back-reaction of Lorentz forces in these models. Many difficulties
still remain to be overcome, particularly in the two most prominent spheres
of application, the Earth and the Sun. But it seems likely that, in the ultimate
theories of the origins of both geomagnetism and heliomagnetism which may
be expected to evolve over the next 50 years, both a-effect and co-effect will
survive as essential ingredients.
F=jAB=Vp, (6.1)
i.e. the Lorentz force is balanced by a pressure gradient and the fluid is at rest:
the field is in magnetostatic equilibrium (see, for example, Biskamp 1993).
We shall look more closely at the details of this process in the following
subsections; for the moment it need merely be noted that we are faced
here with an intriguing class of problems of variational type: to minimize
a positive functional of the field B (here the magnetic energy) subject to
conservation of the field topology; this topological constraint is conveniently
captured by the requirement that only frozen-field distortions of B , i.e. those
governed by the frozen-field equation
2B
- = V A (U A B ),
2t
are to be considered. We know that this equation guarantees conservation
of magnetic helicity Z l v (62.2 above); but since (6.2) tells us that the field B
in effect deforms with the flow, all its topological properties (i.e. those prop-
erties that are invariant under continuous deformation) are automatically
conserved under this evolution.
We may look at this also from a Lagrangian point of view. Let the
Lagrangian particle path associated with the flow ~ ( xt ), be given by
j Vp = 0 and -
B Vp = 0, (6.5)
so that, in any region where Vp # 0, both the current lines (j-lines) and field
lines (B-lines) must lie on surfaces p = const.
In any region where Vp 0, it follows from (6.1) that j is parallel to B,
i.e.
j = y(x)B (6.6)
for some scalar field y(x). The field B is then described as ‘force-free’in this
region. There is a considerable literature devoted to the subject of force-free
fields (see particularly Marsh 1996); here we simply note that, on taking the
divergence of (6.6) and using V j = V B = 0, we obtain
B*Vy = 0 , (6.7)
so that now the B-lines lie on surfaces y = const. The only possible escape
from this (topological) constraint is when Vy 0 and y = const. The field B
is then a ‘Beltrami’ field. We have seen one example in (5.25) above; a more
general field satisfying V A B = yB may be constructed as a superposition of
circularly polarized Fourier modes in the form
where I$ = k/lkl and k . $ ( k ) = 0. With such a field, the B-lines are no longer
constrained to lie on surfaces; the particular field (5.25) exhibits the property
of chaotic wandering of B-lines (Henon 1966; Arnold 1966; Dombre et al.
1986), and it may be conjectured that this a generic property of force-free
fields of the general form (6.8).
(6.10)
(6.11)
the no-slip condition on $9, U -+ 0 also for all x E 9, i.e. the limit state is
indeed magnetostatic. Thus, as t + x,B + B E ( x ) ,j + j E ( x ) ,where
j E I P \ B E= V p E . (6.13)
This evolutionary process may be represented pictorially by a ‘trajectory’ $
in the function space F of solenoidal vector fields of finite energy (figure 11).
This trajectory is confined to a subspace 9 of those fields that can be
obtained from B o ( x ) by isotopic (frozen-field) deformation. Such a subspace,
described as ‘isomagnetic’ is indicated as a surface in figure 11, but this is a
misleading simplification, since it is in fact infinite-dimensional. Nevertheless,
9may be thought of as ‘foliated’ by such isomagnetic subspaces, the fields
on any such subspace being ‘topologically accessible’ one from another by
isotopic deformation. This concept of an ‘isomagnetic subspace’ is closely
7 Reflections o n Magnetohydrodynamics 383
Figure 12. Relaxation of linked flux tubes due to Maxwell tension along the axes of
the tubes; a tangential discontinuity of field develops when the tubes make contact
(Moffatt 1 9 8 5 ~ ) .
Figure 13. ( a ) Single flux tube with axial flux @ and uniform internal twist h ; the
helicity is &hQ2. ( b ) Multiply kinked tube, believed to be the preferred configuration
when h >> 1; the internal ‘twist’ is converted to ‘writhe’ which may be estimated by
the number of crossings.
V2 being conserved. Thus contact of the two tubes is inevitable in the limit
t --+ E, this contact being achieved by a ‘squeeze-film’ mechanism in which
the (viscous) fluid is squeezed from the space between the two tubes as these
contract towards contiguity.
Note that the fluxes @1 and @2, and also naturally the helicity
X . w = 2@1@2, are conserved during the relaxation process.
Figure 14. Conjectured minimum-energy configuration for the trefoil knot for the
same value of h (count the crossings!).
to (6.15) still applies (Moffatt 1990), but now the function m(h) may depend
on the knot K as well as on h. We therefore write it as m ~ ( h )a, real-valued
function which is a ‘property’ of the knot K . The argument leading to (6.17)
is now valid only for large h (since for h = O(1) there is now a topological
barrier to unlimited decrease of knot length L ) ; thus we have
mK(h) - h4I3 for h >> 1. (6.18)
Note also that there may be more than one minimum-energy state: two
possibilities for the trefoil knot with h = 6 are indicated in figure 14.
7 Concluding remarks
In this essay, I have attempted to convey the flavour, rather than the detail,
of three overlapping areas of magnetohydrodynamics in which I have been
successively involved over the last 40 years. In 1960, the subject was relatively
young and it attracted a huge cohort of researchers, active particularly in
the contexts of astrophysics and fusion (plasma) physics. Great progress
was made, notably in dynamo theory in the 1960s, the 1970s being more
a period of consolidation. In more recent years, as in other branches of
fluid mechanics, the computer has played an increasingly important role in
allowing the investigation of nonlinear effects frequently beyond the reach
of analytical study. Some fascinating new areas have emerged, most notably
the area treated in $ 6 above in which the global topology of the magnetic
field plays a central role. It may be noted that there is an exact analogy
between the magnetostatic equation (6.13) and the steady Euler equation
of ideal hydrodynamics, so that the magnetic relaxation technique provides
an indirect means of determining solutions of the steady Euler equations
7 Reflections on Magnetohydrodynamics 389
References
ALFV~N, H. 1950 Cosmical Electrodynamics. Oxford University Press.
ARNOLD,V. I. 1966 Sur la topologie des kcoulements stationnaires des fluides parfaits.
C. R. Acad. Sci. Paris 261, 17-20.
ARNOLD,V. I. 1974 The asymptotic Hopf invariant and its applications. In Proc.
Summer School in Diferential Equations, Erevan 1974. Armenian SSR Acad. Sci.
[English transl: Sel. Math. Sou. 5 (1986), 327-3451.
ARNOLD, V. I. & KHESIN,B. A. 1998 Topological Methods in Hydrodynamics. Springer.
BAJER,K. & MOFFATT, H. K. 1990 On a class of steady confined Stokes flows with
chaotic streamlines. J . Fluid Mech. 212, 337-363.
BATCHELOR, G. K. 1950 On the spontaneous magnetic field in a conducting liquid in
turbulent motion. Proc. R. Soc. Lond. A 201, 405-416.
BERGER,M. A. & FIELD, G. B. 1984 The topological properties of magnetic helicity.
J. Fluid Mech. 147, 133-148.
BISKAMP, D. 1993 Nonlinear Magnetohydrodynamics. Cambridge University Press.
BOJAREVIES, V., FREIBERGS, YA., SHILOVA, E. I. & SHCHERBININ, E. V. 1989 Electrically
Induced Vortical Flows. Kluwer.
BRAGINSKII, S. I. 1964 Theory of the hydromagnetic dynamo. Sou. Phys. J E T P 20,
1462-147 1.
BRAUNBEK, W. 1932 Eine neue Methode electrodenloser Leitfahigkeitsmessung.
Z . Phys. 73, 312-334.
CHANDRASEKHAR, S. 1961 Hydrodynamic and Hydromagnetic Stability. Clarendon.
CHILDRESS, S. & GILBERT, A. D. 1995 Stretch, Twist, Fold: the Fast Dynamo. Springer.
CHUI,A. Y. K. & MOFFATT, H. K. 1995 The energy and helicity of knotted magnetic
flux tubes. Proc. R. Soc. Lond. A 451, 609-629.
COOK,A. H. 1980 Interiors of the Planets. Cambridge University Press.
COWLING, T. G. 1934 The magnetic field of sunspots. Mon. Not. R . AstrON. Soc. 94,
39-48.
COWLING,T. G. 1957 Magnetohydrodynamics. Interscience; 2nd Edn 1975, Adam
Hilger Ltd., Bristol.
DOMBRE, T., FRISCH,U,, GREENE,J. M., H ~ N O NM., , MEHR,A. & SOWARD,A. M.
1986 Chaotic streamlines in the ABC flows. J. Fluid Mech. 167, 353-391.
DAVIDSON, P. A., KINNEAR, D., LINGWOOD, R. J., SHORT,D. J. & HE, X. 1999 The
role of Ekman pumping and the dominance of swirl in confined flow driven by
Lorentz forces. Eur. J. Mech. BIFluids 18, 693-711.
GOLITSIN,G. 1960 Fluctuations of magnetic field and current density in turbulent
flows of weakly conducting fluids. Sou. Phys. Dokl. 132, 315-318.
H ~ N O NM., 1966 Sur la topologie des lignes de courant dans un cas particulier. C. R .
Acad. Sci. Paris A 262, 312-314.
JACOBS, J. A. 1984 Recersals of the Earth’s Magnetic Field. Adam Hilger Ltd., Bristol.
390 H . K . Moffatt
KELVIN,LORD(then W. THOMSON) 1869 On vortex motion. Trans. R. Soc. Edin., 25,
217-260.
KRAUSE,F. & RADLER,K.-H. 1980 Magnetohydrodynamics and Dynamo Theory.
Pergamon.
LIXARDATOS, D. 1993 Determination of two-dimensional magnetostatic equilibria
and analogous Euler flows. J . Fluid Mech. 246, 569-591.
MARSH,G. E. 1996 Force-free Magnetic Fields: Solutions, Topology and Applications.
World Scientific.
MESTEL,A. J. 1982 Magnetic levitation of liquid metals. J. Fluid Mech. 117, 27-43.
MOFFATT, H. K. 1961 The amplification of a weak applied magnetic field by turbu-
lence in fluids of moderate conductivity. J. Fluid Mech. 11, 625-635.
MOFFATT, H. K. 1965 On fluid flow induced by a rotating magnetic field. J. Fluid
Mech. 22, 521-528
MOFFATT, H. K. 1967 On the suppression of turbulence by a uniform magnetic field.
J. Fluid Mech. 28, 571-592.
MOFFATT, H. K. 1969 The degree of knottedness of tangled vortex lines. J . Fluid
Mech. 35, 117-129.
MOFFATT, H. K. 1978 Magnetic Field Generation in Electrically Conducting Fluids.
Cambridge University Press.
MOFFATT, H. K. 1985a Magnetostatic equilibria and analogous Euler flows of arbi-
trarily complex topology. Part 1. Fundamentals. J. Fluid Mech. 159, 359-378.
MOFFATT, H. K. 19852, High frequency excitation of liquid metal systems. In Metal-
lurgical Applications of Magnetohydrodynamics (ed. H. K. Moffatt & M. R. E.
Proctor), pp. 180-189. Metals Society, London.
MOFFATT,H. K. & PROCTOR,M. R. E. (Eds.) 1984 Metallurgical Applications of
Magnetohydrodynamics. Metals Society, London.
MOFFATT, H. K. 1990 The energy spectrum of knots and links. Nature 347, 367-369.
MOFFATT, H. K. & RICCA,R. L. 1992 Helicity and the CglugHreanu invariant. Proc.
R. Soc. Lond. A 439, 411429.
MOREAU, R. 1990 Magnetohydrodynamics. Kluwer.
NORTHRUP, E. E 1907 Some newly observed manifestations of forces in the interior
of an electric conductor. Phys. Rev. 24, 474-497.
ODIER,P., PINTOX,J.-F. & FAUVE,S. 1998 Advection of a magnetic field by a
turbulent swirling flow. Phys. Rev. E 58, 7397-7401.
PARKER, E. N. 1955 Hydromagnetic dynamo models. Astrophys. J. 122, 293-3 14.
PARKER,E. N. 1979 Cosmical Magnetic Fields. Clarendon.
PARKER, E. N. 1994 Spontaneous Current Sheets in Magnetic Fields. Oxford University
Press.
PRIEST,E. R. 1982 Solar Magnetohydrodynamics. D. Reidel.
RICHARDSOK, L. F. 1922 Weather Prediction by Numerical Process. Cambridge Uni-
versity Press.
SCHUSTER, A. 1912 A critical examination of the possible causes of terrestrial
magnetism. Proc. Phys. Soc. Lond. 24, 121-137.
SKEYD,A. D. & MOFFATT,H. K. 1982 Fluid-dynamical aspects of the levitation
melting process. J. Fluid Mech. 117, 45-70.
STASIAK, A., KATRITCH, V., BEDKAR,J., MICHOUD,D. & DUBOCHET, J. 1996 Elec-
trophoretic mobility of DNA knots. Nature 384, 122.
STASIAK, A., KATRITCH, V. & KAUFFMAN, L. H. 1998 Ideal Knots. World Scientific.
STEEKBECK, M., KRAUSE,F. & RADLER,K.-H. 1966 Berechnung der mittleren
7 Rejections on Magnetohydrodynamics 391
Lorentz-Feldstarke fur ein elektrisch leitendes Medium in turbulenter, durch
Coriolis-Krafte beeinfluster Bewegung. 2. Naturforsch. 21a, 369-376.
VAINSHTEIN, S. I. & ZELDOVICH,
YA.B. 1972 Origin of magnetic fields in astrophysics.
SOU.Phys. USP.15, 159-172.
WOLTJER,L. 1958 A theorem on force-free magnetic fields. Proc. Nut1 Acad. Sci. 44,
4 8 9 4 9 1.
ZELDOVICH, YA. B., RUZMAIKIN,A. A. & SOKOLOFF, D. D. 1983 Magnetic Fields in
Astrophysics. Gordon and Breach.