Chapter 4

Magnetostatics
Introduction

As previously stated, in the static regime (or steady state), where the sources {, J } do not
vary in time, the coupling between the electric and magnetic field vanishes. The magnetic field,
which concerns us here, is only created by constant currents. This is the area of Magnetostatics.
Chapter 1 contains, in principle, all laws and properties necessary to solve any problem in
electromagnetism, so that, very much like what we have done in chapter 2, we will first briefly go
over the relations valid in magnetostatics.

4.1
4.1.1

Maxwells equations applied to Magnetostatics and consequences

Magnetic force

The force exerted by a magnetostatic field B (M, t) on a point charge q placed at point M at

time t and having a velocity

v (M, t) is given by:

F = q
v (M, t) B (M, t)

(4.1)

Property : The magnetic force does not produce any work since W = F . dl = q
v B . dl =


0. From the work-energy theorem, one can infer that the magnetic force does not change the kinetic
energy of a particle. In particular, a magnetic field cannot set a particle in motion.

A nice visualization of this force can be obtained in so-called bubble chambers 1 A bubble
chamber consists in a liquid that is kept, in a metastable phase, at a temperature above its boiling
point through an increased pressure. When a particle goes through the chamber, it may collide with
the electrons of the atoms forming the liquid. The energy given by the particule to the electrons
locally heats up the liquid and hence bubbles form on the particle trajectory. The bubbles are
therefore the signature of a particle going through the chamber. When the particle is charged, an
applied magnetic field allows to distinguish between positively and negatively charges species (see
figure 4.1) through the direction of curling trajectories.

1. Bubble chambers were invented by Donald Glaser in 1952. For more information on this technique and its
use, see http://cerncourier.com/cws/article/cern/29120.

91

92

CHAPTER 4. Magnetostatics

Figure 4.1: Trajectories of an electron and a positron created by an incoming particle visualized
in a bubble chamber. The particle hits the electron of a hydrogen atom. The atomic electron is
ejected while a pair electron/positron is created. The two latter particles curl in opposite directions
thanks to the applied magnetic field. Taken from http://www.alternativephysics.org/book/
MatterEnergy2.htm.

Comment on magnetic monopoles : An electric field is created by an ensemble of charges. Each

one of these can be defined as a positive or a negative charge. In other words, one can define
electric monopoles. However, a magnetic field is created by a current - whether macroscopic
or local as we will see in the next chapter -. In other words, in standard materials, if you cut a
magnet into two parts, you will always obtain both a North pole and a South pole 2 . Indeed, there
has, so far, been no consensus on the existence or not of magnetic monopoles, although
their existence has been theoretically predicted by Paul Dirac in 1931 who claimed that they could
explain charge quantization 3 . Experimentally, numerous claims have been made for the discovery
of these monopoles. To date, they have all been disproved.... except the latest one by David Hall
and Michael Ray in 2014 4 (at least not yet !).

4.1.2

Corollary : Laplace force

portion of linear conductor dl through which flows a current I is given by the Laplace force:

Property #1: The elemental magnetic force exerted by a magnetostatic field B on a

dF = I dl B

(4.2)

2. As can be seen from the equation div B = 0, there is no magnetic charges.

3. See for example the discussion in J.D. Jackson, Classical Electrodynamics, section 6.11.
4. They have observed syntheticmagnetic monopoles in Bose-Einstein condensates. See Ray et al., Nature, 505,
657, 2014, or https://www.amherst.edu/aboutamherst/news/faculty/node/532493.

4.1. MAXWELLS EQUATIONS APPLIED TO MAGNETOSTATICS AND CONSEQUENCES93

Proof :

dF =

ni q i
vi B dV

J B dV

J dSdl B

JdS dl B

I dl B

=
=
=

dF =

Property #2: From the above demonstration, one can also extract the force exerted by a

magnetostatic field B on a conducting surface dS through which flows a surface current density

Js :

dF = J s B dS

Application to the definition of the Amp`ere: 1 Amp is the intensity

of the currents I1 and I2 flowing through two infinite wires, separated
by a distance of 1 m and exerting a force on one another of 2.107 N
per unit length.
Proof : We will see in section 4.2.2 that the magnetic field created

, so that the force

0 I1
u
by the first infinite wire at a distance d is B1 = 2d
y
exerted by the first wire on the second one is :

df 12

=
=

df 12

0 I1

I2 dl
u
y
2d
0 I1 I2

dlu
u
z
y
2d
0 I1 I2 dl
u

x
2d

This definition is equivalent as defining 0 = 4.107 u.S.I.

4.1.3

Typical values for magnetic fields

The table below gives the orders of magnitude for a few magnetic fields 5 :

Residual magnetic field within an excellently shielded box

Earths magnetic field at its surface
Magnetic field of a usual magnet
Magnetic field generated by a large electromagnet
Magnetic field generated by a large superconducting magnet
Magnetic field generated by a neutron star

1014 T
2-4105 T
102 T
2T
10-50 T
108 T

5. The international unit for the magnetic field is the tesla (T). However the gauss is also commonly used : 104
G = 1 T.

94

CHAPTER 4. Magnetostatics

4.1.4

Maxwells equations in magnetostatics

The decoupling between the electric and magnetic fields reduces the number of equations to be
taken into account. The Maxwells equations of magnetostatics are :

4.1.4.a

Conservation of magnetic flux

Maxwell-Amp`
ere

div B = . B = 0

curl B = B = 0 J

(4.3)
(4.4)

Conservation of the magnetic flux

Definition : The magnetic flux B of a magnetic field B through a surface (S) is :

B =

B .dS

(4.5)

B is in unit of webers (Wb) 6 .

As we have seen on numerous occasions, we have :

div B = 0

(4.6)

which implies that B is a vector which conserves its flux 7 .

4.1.4.b

Amp`
eres law

The integral form of Maxwell-Amp`eres equation in the steady-state regime gives Amp`
eres
law :

B . dl = 0 I(S)
(4.7)
(C)

where (C) is a closed contour delimiting a surface (S), I(S) is the current encircled by (C).
4.1.4.c

Discontinuity equations at interfaces

The discontinuity equation for the magnetic field is :

B 2 B 1 = 0 Js
n 12

(4.8)

where B i is the magnetic field in the medium i, Js is the surface current density at the interface

between the two media, and n 12 is the unit vector, normal to the interface, directed from medium
(1) to medium (2).

6. 1 Wb = 1 T.m2 .
7. See section 2.1.3.c for the consequences of this equation.

4.1. MAXWELLS EQUATIONS APPLIED TO MAGNETOSTATICS AND CONSEQUENCES95

4.1.5

Vector potential and Biot-Savart law

4.1.5.a

Vector potential and potential propagation

In chapter 1, we have seen that the magnetic field derives from a vector potential :

B = curl A

(4.9)

In section 1.3.4, we have derived the general equation for the propagation of the vector potential :



1 2 A
1
A 2

grad
div
A
+
= 0 J
2
2
c t
c t
In magnetostatics, this equation, within the Coulomb or Lorenz gauge 8 , this equation amounts to
a Poisson equation:

A = 0 J
(4.10)
which solution is :

0
A (M ) =
4

J (P )
dV
PM

(4.11)

Proof : In electrostatics, one had a Poisson equation for the electrostatic potential: = 0

which solution was expressed as (M ) =

40 P M dV . By analogy, one derives the solution of
the Poisson equation for the vector potential.
4.1.5.b

Biot-Savart Law

Hence, the magnetic field created by the current density J can be expressed under the form of
9
the Biot-Savart law :

0
J (P ) P M
B (M ) =
dV
(4.12)
4
PM3
Proof :

Jx (P )
0
 A
 x = 4
P M dV


Jy (P )


0
A =  Ay = 4
P M dV



Jz (P )
 Az = 0
4
P M dV



 Ax
 x






curl A =  y
 Ay





 Az
z

For the component along z :

Bz

=
=
=

Ax
Ay

x
y





0
Jy (P )
Jx (P )
dV
dV
x 4
PM
y 4
PM





0

1
1
Jy (P )
Jx (P )
dV
4
x P M
y P M

= div A = 0.
8. In statics, both gauges are equivalent: div A + c12
t
9. This law was derived by Jean-Baptiste Biot and F
elix Savart in 1820.

96

CHAPTER 4. Magnetostatics

and if

PM

PM

1
PM

=
=


 x



 y



 z
p
x2 + y 2 + z 2
1 P M
1
x

=
P M 2 x
PM2 PM

so that :
Bz
Bz

4.1.5.c

0
4

0
4

x 
y

J
(P
)
dV
y
PM3
PM3
!

PM
dV
J
.
u
z
PM3

Jx (P )

Biot-Savart law for a wire

In the case of a wire through which flows a constant current I, one can express the magnetic
field created by I :

0 I
dl P M
B (M ) =
(4.13)
4
PM3
This is the Biot-Savart law.
Proof :

B (M )

0
4

B (M )

0
4

B (M )

0
4

J (P ) P M
dV
PM3

J (P ) P M
Sdl
PM3

J(P )S dl P M
PM3

where S is the section of the wire.

4.1.6

Field lines and symmetries

4.1.6.a

Field lines

Definition : Magnetic field lines are the lines that are, at all points in space, tangent to

the magnetic field B . They can be determined through the relation :

B (M ) dl = 0

where dl is an infinitesimal vector along the field line, centered around point M .

(4.14)

4.2. EXAMPLES OF USUAL CHARGE DISTRIBUTIONS

4.1.6.b

97

Properties

Property #1 : The vector potential and the magnetic field reproduce all symmetries and

invariances of the current distribution J (P ) that creates this potential and this field.
Property #2 : The field lines for the magnetic field are closed (possibly at infinity) and

enlace the current sources. B is an axial vector, whereas A is a polar vector. They then obey
the symmetry properties described in section 1.2.1.

Proof : The field lines must be closed, otherwise
B .dS0 which is not possible, due to the

conservation of flux. The field lines must embrace the current sources :
B . dl = 0 I due to
Maxwell-Amp`eres equation.
4.1.7

Energetics in magnetostatic

4.1.7.a

Magnetostatic energy

Definition : The volume density of magnetostatic energy is given by :

u=

B2
20

(4.15)

This is a particular case of equation 1.24.

As a consequence, the total electrostatic energy U in a volume V can be written as :

U=
u dV
(V )

U=

1
2

(V )

B2
dV 0
0

(4.16)

This energy is always positive. It implies that the presence of a field adds energy to the system.

4.2
4.2.1

Examples of usual charge distributions

Example #1 : Circular loop

Let us consider a circular loop of radius R and axis Oz through which flows a current I. We
will determine the field on the axis of this circular loop.

98

CHAPTER 4. Magnetostatics
Figure 4.2: Circular loop.

The problem is of cylindrical symmetry. We will therefore use the cylindrical coordinates. Since

the current distribution J (r,
, z), and the magnetic field conserves the symmetries of the current

distribution, B (r,
, z). For a point M on the Oz axis, M belongs to all (

u
r uz ) antisymmetry
planes. On the Oz axis, the magnetic field therefore is along the Oz direction. Using Biot-Savart
law :

=
=

0 I
dl P M
4
PM3
2
(R
+ z
)
0 I
Rd
u
u
u

r
z
3
4 =0
2
2
2
(r + z )

will therefore cancel out

But we know that the field will be along Oz. The contribution of
u
u

z
in the integration. Hence :

=
=
=

4.2.2

0 I
4

0 I
4

(R
)
Rd
u
u

r
3

(r2 + z 2 ) 2

R2 d
u

=0

=0 (R2

20 IR2
4 (R2

z2) 2

+
0 I

sin3
u
z
2R

z2) 2

u
z

Example #2 : Infinite wire

Let us consider an infinite wire through which flows a current I. We will determine the magnetic
field created by this wire.

Figure 4.3: Infinite wire.

4.2. EXAMPLES OF USUAL CHARGE DISTRIBUTIONS

99

The problem is of cylindrical symmetry. We will therefore use the cylindrical coordinates. Since

the current distribution J (r,
, z),

and the magnetic field conserves the symmetries of the current

distribution, B (r,
, z).

All
points
M in space belong to a (
u
r uz ) symmetry plane. The magnetic

field therefore is along u . Then, by taking the Amp`ere contour drawn on the above figure :

B . dl = 0 I
B(r)2r

=
=

0 I
0 I

2r

Given the symmetries and invariances, the vector potential can be written as : A = A(r,
, z)

uz .
Then :

A . dl =
B .dS

0 I
dz dr

u
u
Az (r1 )h Az (r2 )h =

2r
r2
0 Ih
1
Az (r1 )h Az (r2 )h =
dr
2 r=r1 r
 

r
0 I

u
ln
A (r) = A (r0 )
z
2
r0

4.2.3

Example #3 : Infinite solenoid

Let us consider an infinite solenoid of radius R and axis Oz, having n turns per unit length, and
through which flows a current I. We will determine the field created by this infinite solenoid 10 .

Figure 4.4: Infinite solenoid.

The problem is of cylindrical symmetry. We will therefore use the cylindrical coordinates. Since

the current distribution J (r,
, z),

and the magnetic field conserves the symmetries of the current
10. The case of a finite solenoid will be seen in the exercise sheet.

100

CHAPTER 4. Magnetostatics

distribution, B (r,
, z).

All points M in space belong to a (ur , u ) symmetry plane (because the
solenoid is infinite). The magnetic field therefore is along the Oz direction. An Amp`ere contour
which does not include any current shows that the field is homogeneous outside the solenoid. Since

B (r +) = 0 , the field is null everywhere outside the solenoid. An Amp`ere contour including
a current gives :

B . dl = 0 I(S)
B(r1 )h B(r2 )h

0 nhI

B(r1 )

0 nI

0 nI
u
z

Noticing that all points M in space belong to a (

u
r uz ) antisymmetry plane and given the

previously established symmetries and invariances, the vector potential can be written as : A =

A(r,
, z)

u . Then, along a circle of axis Oz encompassing the solenoid :

4.3

A . dl

A 2r

A (r)

A 2r

A (r)

B .dS

0 nI R2 for r > R
0 nIR2
for r > R
u

2r
0 nI r2 for r < R
0 nIr
for r < R
u

Magnetic dipole

The notion of magnetic dipole is important. Indeed, we will see in the next chapter that it is
at the basis of the microscopic description of magnetic materials.

4.3.1

Definition
Definition : A magnetic dipole is a localized distribution of
current loops, of finite spatial extension , centered around point A.
This distribution can then be modeled by a closed loop of electric current

for which one can define a magnetic moment M (in A.m2 ) :

M=I

dS

(4.17)

(S)

A few comments :

In the case of a plane current loop of normal

n:

M = IS
n

(4.18)

101

4.3. MAGNETIC DIPOLE

The dipole approximation can here also be used (see section 2.4.1). It consists in considering the effect of a magnetic dipole at point M such that r = AM . In other words, one
neglects the spatial extension of the current distribution around point A and replaces it by

a dipole M placed in A.
This model can account for Earths magnetic field (see additional reading section 4.6).
This model can account for orbital magnetic moments. Indeed, a simple semi-classical
model describes the electronic motion in an atom as a circular movement due to the attractive
Coulomb force exerted by the nucleus. The period T of such a motion is very short, so that
it can be associated to an effective current I e
T . Then :
eS

n
M = IS
n =
T
Since the motion derives from a central force :
M=

S
T

L
2m .

Hence :

e
eL

L =L
n =
2m
2m

e
where = 2m
is the gyromagnetic ratio.

4.3.2

Magnetic field and potential created by an magnetic dipole

4.3.2.a

Vector potential created by an magnetic dipole

Figure 4.5: Schematic representation of a dipole.

In the dipole approximation, the vector potential created at point M by an magnetic dipole M

placed at point A such that AM = r

u , with
u an unit vector, is 11 :

u
0 M
(4.19)
A (M ) =
4 r2
4.3.2.b

Magnetic scalar potential created by an magnetic dipole

In the regions of space where there are no sources ( J = 0 ), one has rot B = 0 . One can
therefore define a magnetic scalar potential m such that :

B = grad m
(4.20)

11. The proof of this expression is troublesome so that we will not derive it here. However, it can be found in
J.D. Jackson, Classical Electrodynamics, section 5.6.

102

CHAPTER 4. Magnetostatics

The magnetic field is then said to derive from the magnetic scalar potential.
In the case of a magnetic dipole, in the dipole approximation:

0 M.
r
m (M ) =
4 r3
4.3.2.c

(4.21)

Magnetic field created by an magnetic dipole

In the dipole approximation, the magnetic field created at point M by an electrostatic dipole

M placed at point A such that AM = r

u , with
u an unit vector, is 12 :



M.
u
.
u M
3

0
(4.22)
B (M ) =
4
r3
4.3.2.d

Field lines for an magnetic dipole

One can note that, within the dipole approximation, the magnetic field created by a magnetic
dipole (equation 4.22) has the same mathematical structure than the one giving the electric field
created by an electrostatic dipole (equation 2.30). Thus, the field lines for a magnetic dipole can
be directly derive. They are schematized figure 4.6.

Figure 4.6: Schematic representation of the field lines created by a magnetic dipole. Taken from
http://en.wikipedia.org/wiki/Magnetic_dipole. The dipole is here modeled by a closed current loop.

4.3.3

Mechanical action of an external field on a magnetic dipole

4.3.3.a

Magnetic dipole in an external magnetic field

In the dipole approximation, an magnetic dipole M, of center A, placed in an external magnetic

field B is submitted to forces such that 13 :

12. The proof of this expression is also troublesome so that we will not derive it here. However, it can be found
in J.D. Jackson, Classical Electrodynamics, section 5.6.
13. The proof of this expression is also troublesome so that we will not derive it here. However, it can be found
in J.D. Jackson, Classical Electrodynamics, section 5.7.

103

4.4. METHODS IN MAGNETOSTATICS

the resultant force is :
the angular momentum is :

4.3.3.b

F (A) = M. grad A B (A)

(4.23)

(A) = M B (A)

(4.24)

Application in the case of a rigid magnetic dipole

If the considered dipole is rigid, the current loop is not deformable. Then the norm of the

magnetic moment M is time-independent. Then the resulting force is :

F (A)

Um

grad A (Um )

M. B (A)

(4.25)
(4.26)

The force exerted on the dipole therefore derives from the potential energy Um . The dipole

tends to align with the external field B (A).

There again, you can notice the resemblance between these expressions and the ones seen for
the resultant force, angular momentum and potential energy in the case of an electrostatic dipole
(section 2.4.3.a).

4.4

Methods in magnetostatics

There are numerous solving methods in magnetostatics which we will not review here 14 . For
the purpose of this course, you will mainly be using one of the following methods :
Directly compute the magnetic field using symmetries, invariances and :

0
B (M ) =
4

J (P ) P M
dV
PM3

Directly compute the vector potential using symmetries, invariances and :

0
A (M ) =
4

J (P )
dV
PM

Determine the magnetic field using symmetries, invariances and Amp`eres theorem :

B . dl = 0 Iint
Moreover, as stated in section 2.5.2, there are numerous numerical methods available to compute
more complex problems.

4.5

General field analogies

In the course of these four chapters, we have seen similar mathematical expressions for different

quantities : div B = 0 ; = 0 ; 1 2 = E . dl ; ... You may also have noted that you
had encountered the same expressions in different occasions outside electromagnetism : whether
in electricity, in thermal conduction, gravitation, or particle diffusion, you have the same kind of
14. An introduction to those can be found in chapters 5 of J.D. Jackson, Classical Electrodynamicsand in chapters
5 to 8 of E. Weber, Electromagnetic Theory.

equations. The way to solve them is therefore similar and one can establish analogies between
the different quantities at play. Weber has written a very nice chapter on this 15 and below is
an excerpt giving the correspondance between different problems dealing with scalar potential
fields.ext dddpdage

4.6

Additional reading

The following article 16 , describes how a dynamo field can simulate the Earths magnetic field,
including the switching between two directions.
15. E. Weber, Electromagnetic Theory, chapter 3.
16. Berhanu et al., ArXiv:physics/0701076.

91.25.Cw Origins and models of the magnetic field; dynamo theories

47.65.+a Magnetohydrodynamics and electrohydrodynamics

Magnetic induction properties of turbulent swirling

flows (VKS and others) have been widely studied experimentally [8, 10]. Dynamo action by these flows has
also received strong support from numerical simulations [11, 12]. The VKS experimental set-up, shown
in Figure 1, has been described in [9]. The fluid is
liquid sodium, chosen for its high electrical conductivity
( = 107 ohm1 m1 ), low density ( = 930 kg m3 ) and
low melting point (98 C). The net volume of sodium is
roughly 160 L. A turbulent von K
arman flow is generated
by two counter-rotating iron impellers (rotation frequencies F1 and F2 ). Its mean structure has the following
characteristics: the fluid is ejected radially from the
disks by centrifugal force and loops back towards the
axis in the mid-plane between the impellers. A strong
differential rotation is superimposed on this poloidal

field in Riga). No further dynamical evolution was observed. The search for more complex dynamics, such as
exhibited by natural objects, has motivated most teams
working on the dynamo problem to design experiments
with less constrained flows and a higher level of turbulence [8]. The von K
arman sodium experiment (VKS) is
one of them. It has recently shown regimes where a statistically stationary dynamo self-generates [9]. We report
here the existence of other dynamical regimes and describe
below the occurence of irregular reversals and excursions.

p-1

Dynamo action is the instability mechanism by which

mechanical energy is partially converted into magnetic energy by the motion of an electrically conducting fluid [1].
It is believed to be at the origin of the magnetic fields
of planets and most astrophysical objects. One of the
most striking features of the Earths dynamo, revealed by
paleomagnetic studies [2], is the observation of irregular
reversals of the polarity of its dipole field. This behaviour
is allowed from the constitutive equations of magnetohydrodynamics [1] and has been observed in numerical models [3]. On the other hand, industrial dynamos routinely
generate currents and magnetic fields from mechanical motions. In these devices, pioneered by Siemens [4], the path
of the electrical currents and the geometry of the (solid)
rotors are completely prescribed. As it cannot be the
case for planets and stars, experiments aimed at studying
dynamos in the laboratory have evolved towards relaxing these constraints. Solid rotor experiments [5] showed
that a dynamo state could be reached with prescribed motions but currents free to self-organize. A landmark was
reached in 2000, when the experiments in Riga [6] and
Karlsruhe [7] showed that fluid dynamos could be generated by organizing favourable sodium flows, the electrical
currents being again free to self-organize. For these experiments, the self-sustained dynamo fields had simple time
dynamics (a steady field in Karlsruhe and an oscillatory

Abstract. - We report the first experimental observation of reversals of a dynamo field generated
in a laboratory experiment based on a turbulent flow of liquid sodium. The magnetic field randomly switches between two symmetric solutions B and B. We observe a hierarchy of time scales
similar to the Earths magnetic field: the duration of the steady phases is widely distributed, but
is always much longer than the time needed to switch polarity. In addition to reversals we report
excursions. Both coincide with minima of the mechanical power driving the flow. Small changes
in the flow driving parameters also reveal a large variety of dynamo regimes.

PACS

PACS

Laboratoire de Physique Statistique, Ecole

Normale Superieure CNRS UMR8550, 24 rue Lhomond, F-75005 Paris
(France)
2

Service de Physique de lEtat

Condense, Direction des Sciences de la Mati`ere
CNRS URA 2464, CEA-Saclay, F-91191 Gif-sur-Yvette (France)
3

Laboratoire de Physique, Ecole

Normale Superieure de Lyon, CNRS UMR5672, 46 allee dItalie, F-69364 Lyon
(France)

M. Berhanu1 , R. Monchaux2 , S. Fauve1 , N. Mordant1 , F. P

etr
elis1 , A. Chiffaudel2 , F. Daviaud2
B. Dubrulle2 , L. Mari
e2,+ , F. Ravelet2, , M. Bourgoin3,# , Ph. Odier3 , J.-F. Pinton3 , R. Volk3

arXiv:physics/0701076v1 [physics.flu-dyn] 6 Jan 2007

Magnetic field reversals in an experimental turbulent dynamo

epl draft

15

20

40

25

30

60
t [sec]

Rm1

80

35

40

100

45

50

120

(c)

(b)

(a)

Fig. 2: (a) Preliminary inspection of dynamo regimes observed as the impeller rotation frequencies are independently
set. Symbols: (o): statistically stationary dynamos, (+) no
dynamo, i.e., magnetic field less than 10 gauss at the measurement location and for measurement times longer than
180 s. (): dynamo with reversals. (b) Examples of the time
variations of the main magnetic field component for rotation
frequencies of the disks, 22 22 Hz (Rm1 = Rm2 = 42.5) and
14 22 Hz (Rm1 = 27.5, Rm2 = 43.5). Note that on measurement time scales of the order of 180 s, the regimes can depend
on the path followed to reach them.

20

27.5-43.5

42.5-42.5

-50
0

50

100

-150
150

-100

-50

50

10
10

15

20

25

30

35

40

45

50

p-2

flow, which generates a high shear in the mid-plane. The

flow maximum driving power is 300 kW, and cooling
is performed using an oil flow inside the copper walls
of the vessel. It allows experimental runs at constant
temperatures between 110 C and 160C. The integral
Reynolds numbers are defined as Rei = 2KR2 Fi / and
take values up to 5 106 where is the fluid viscosity and
K = 0.6 is a coefficient that measures the efficiency of the
driving impellers [12]. Corresponding magnetic Reynolds
numbers, Rmi = 2K0 R2 Fi , up to 49 at 120C are
reached 0 is the magnetic permeability of vacuum. parameter space accessible when the flow is driven with
The magnetic field is measured with local Hall probes disks rotating at different speeds. As said above, only stainserted inside the fluid.
tistically stationary dynamos are observed in the counterrotating case (Figure 2b). Another statistically stationary
When the impellers are operated at equal and opposite dynamo mode is observed when the frequency of one imrotation rates F , a fully turbulent dynamo is observed peller is increased from zero (say F1 ), the other being kept
when F is larger than about 17 Hz (Rm = 31) [9]. The fixed at 22 Hz, thus Rm2 in the range 4243 depending on
self-sustained magnetic field is statistically stationary with the sodium temperature (Figure 2c). Note however that
either polarity in this case. In the experiment, the rota- its relative fluctuations are much smaller (compare Figtion rates (F1 , F2 ) of the driving impellers can be inde- ures 2b, c), an effect possibly ascribed to global rotation.
pendently adjusted and this gives an additional degree of This regime undergoes secondary bifurcations when the
freedom. Starting from a symmetric flow forcing, F1 = F2 , slower impeller frequency is increased further. In a small
one can progressively change the rotation frequency of one parameter range, Fi /Fi 20%, a variety of dynamical
disk and explore regimes in which the faster disk imposes regimes, oscillations, intermittent bursts (not shown), as
some kind of global rotation to the flow, a feature common well as dynamos with random reversals (Figure 3) are obto most natural dynamos.
served. We also find pockets of parameters for which we
We show in Figure 2 a preliminary inspection of the could not record the growth of a dynamo during 3-minute

Fig. 1: The VKS2 set-up is designed to generate a dynamo

flow in an electrically conducting fluid. The overall vessel is
a copper cylinder of radius 289 mm and length 604 mm. The
flow itself is confined within an inner copper cylinder (radius
R = 206 mm, length 524 mm, thickness 5 mm), with sodium at
rest between the inner and outer cylinders. An annulus of inner
radius 175 mm (thickness 5 mm) is fixed along the inner cylinder in the mid-plane between the disks. The counter-rotating
iron impellers have radius 154.5 mm and are set 371 mm apart
in the inner vessel; they are fitted with 8 curved blades of height
h = 41.2 mm. Their rotation frequencies are independently adjustable, up to 26 Hz. Magnetic measurements are made using
a temperature controlled, 3D Hall probe mounted flush on the
flow boundary, at the inner cylinder.

Berhanu et al.

Rm2
B [gauss]
B [gauss]

One important discovery in these measurements is that

however that the amplitude of the magnetic field, that is

much larger than the Earth field, is the same for both polarities. Standard deviations are of the same order of magnitude as the mean values, although better statistics may
be needed to fully converge these estimates. The mean
duration of each reversal, 5 s, is longer than magnetohydrodynamics time scales: the flow integral time scale
is of the order of the inverse of the rotation frequencies,
i.e. 0.05 s, and the ohmic diffusive time scale is roughly
0.4 s. Concerning the dynamics of field reversals, a
natural question is related to the connection between B
and B in time. The equations of magnetohydrodynamics are symmetric under the transformation B to B so
that the selection of a polarity is a broken symmetry at the
dynamo bifurcation threshold. The sequences of opposite
polarities act as magnetic domains along the time axis,
with Ising-type walls in-between them: the magnetic field
vanishes during the polarity change rather than rotating
as in a Bloch-type wall.

p-3

long runs corresponding to over 3000 forcing time scales.

We now describe reversals of the magnetic field. In Figure 3, we show a time series that corresponds to F1 =
16 Hz and F2 = 22 Hz. In this regime, the magnetic field
reverses at irregular time intervals. All three components
of the dynamo field switch polarity in perfect synchrony,
so that B changes to B. For each polarity, the amplitude
of the magnetic field has strong fluctuations, with an rms
fluctuation level of the order of 20% of the mean. This
level of fluctuations is due to the very intense turbulence
of the flow, as the kinetic Reynolds number exceeds 106 .
Reversals occur randomly and have been followed for up
to 45 minutes, i.e. 54000 characteristic time scales of the
flow forcing.
In the regime reported in Figure 3, the polarities do not
have the same probability of observation. Phases with
a positive polarity for the largest magnetic field component have on average longer duration (hT+ i = 120 s) than
phases with the opposite polarity (hT i = 50 s). This
asymmetry can be due to the ambient magnetic field. Note

Fig. 3: Magnetic field measured inside the flow vessel, by a 3-dimensional Hall probe. No external magnetic field is applied,
other than the ambient field, whose amplitude is about 0.2 gauss across the measurement volume.The temperature of the outer
copper cylinder is T = 123 o C. (Main): Time evolution of all three magnetic field components. The main component (red) is
the azimuthal one. Note that all components decay to zero at a reversal. The bottom graph shows synchronous recordings of
the power driving the flow. (Right): detail of the time series of the main magnetic field and simultaneous power consumption
(arrows mark the synchronous events). (Top): Chronos of the magnetic field orientation, white for a positive direction, black for
the negative direction, for 2 successive recordings 900 and 1800 seconds long (separated by the shaded area, the first sequence
corresponds to the main graph). In this regime, the von K
arm
an flow is driven with counter-rotating disks at frequencies
F1 = 16 Hz and F2 = 22 Hz.

Magnetic field reversals in an experimental turbulent dynamo

conditions both of them can be experimentally adjusted.

Others are related to the dynamics of the magnetic reversals. From inspection of paleomagnetic data, it has
been proposed that reversing dynamos and non-reversing
ones are metastable states in close proximity [13]. In
geodynamo simulations (convective dynamos in rapidly
rotating spheres), the flow is often laminar and reversals
have been associated to interaction between dipole and
higher order modes, with the possibility of reversal
precursor events [14]. Fields reversals have also been
observed in turbulence driven numerical 2 and dynamos based on mean-field magnetohydrodynamics [15].
In these, the role of noise was found to be essential,
together with the proximity of steady and oscillating
states. In many cases, the existence of several dynamo
regimes in a narrow region of parameter space has been
considered as essential. Our experiment displays this
feature: two different stationary dynamo modes bifurcate
for F1 = F2 and respectively F1 6= F2 . Their interaction
gives rise to a variety of different dynamical regimes in
parameter space. This is a general feature for bifurcations
of multiple codimension. The most striking aspect of our
observation is that the low dimensional dynamics that
result from the interaction of a few modes of the magnetic
field is preserved despite strong fluctuations of the flow
that generates the field. On average, the largest scales of
the flow change more than a thousand times during each
phase of given polarity. The large scale magnetic field
cannot follow turbulent fluctuations and display features
characteristic of low dimensional dynamical systems.
Flows generating the magnetic fields of planets or stars
involve dimensionless parameters (Reynolds and Eckman
numbers) orders of magnitudes different from the ones
of the present experiment and even further in the case
of numerical simulations. However, a weak coupling
between the large scale dynamics of their magnetic
field and hydrodynamic fluctuations may explain why
similar features are observed in some natural dynamos,
in numerical modeling and in this experiment.

p-4

We are indebted to Marc Moulin for the technical design of the experiment, and for making so many parts. We
thank Cecile Gasquet for data acquisition development
and for her participation to several campaigns. We thank
Jean-Baptiste Luciani and Andre Skiara for their skills in
operating the sodium-related equipments. We gratefully
Fig. 4: Superimposition of 5 successive reversals from down acknowledge Didier Courtiade and Jean-Francois Point
to up polarity. For each of them the origin of time has been for their assistance with the cooling system, Pascal Metz
shifted such that it corresponds to B = 0.
for instrumentation development and Vincent Padilla
for making parts of the experiment. We thank the
Further investigation of this regime will help address Dynamo GDR 2060, and numerous colleagues with
from an experimental perspective persistent questions whom we have had fruitful discussions over the years,
about magnetic field reversals. Some of these concern in particular Emmanuel Dormy for useful discussions
the role of hydrodynamics and electromagnetic boundary about the characteristics of the dynamics of the Earth

reversals of magnetic field are correlated with the global

energy budget of the flow. The total power P (t) delivered by the motors driving the flow fluctuates in time in
a strongly asymmetric manner: the record shows short
periods when P is much smaller than its average. They
always coincide with large variations in the magnetic field,
as shown in Figure 3. Either a reversal occurs, or the magnetic field first decays and then grows again with its direction unchanged. Similar sequences, called excursions [2],
are observed in recordings of the Earths magnetic field.
The variation of power consumption during the weakening of the magnetic field is in agreement with the power
required to sustain a steady dynamo in the VKS2 experiment [9] (drops by over 20%, that is 20 kW out of 90 kW).
However, we note that in other regions of the parameter
space, different regimes also involve changes in polarity
without noticeable modification of power.
We have also observed that the trajectories connecting
the symmetric states B and B are quite robust despite
the strong turbulent fluctuations of the flow. This is
displayed in Figure 4: the time evolution of reversals
from down to up states can be neatly superimposed by
shifting the origin of time such that B(t = 0) = 0 for
each reversal. Despite the asymmetry due to the Earth
magnetic field, up-down reversals can be superimposed in
a similar way on down-up ones if B is plotted instead
of B. For each reversal the amplitude of the field first
decays exponentially. A decay rate of roughly 0.8 s1 is
obtained with a log-lin plot (not shown). After changing
polarity, the field amplitude increases linearly and then
displays an overshoot before reaching its statistically
stationary state.

Berhanu et al.

p-5

[1] See for instance, Moffatt H. K., Magnetic field generation in electrically conducting fluids (Cambridge University
Press, Cambridge) 1978
[2] Dormy E., Valet J.-P. and Courtillot V., Geochemistry, Geophysics and Geosystems (G-cubed), 1 (2000) 62;
Valet J.-P., Rev. Geophys., 41 (2003) 1004.
[3] Glatzmaier G. and Roberts P. H., Nature, 377 (1995)
203; Kutzner C. and Christensen U. R., Physics of the
Earth and Planetary Interiors, 131 (2002) 29.
[4] Siemens W., Nature, (1871) December 28th.
[5] Lowes F. J. and Wilkinson I., Nature, 198 (1963) 1158;
Nature, 219 (1968) 717
[6] Gailitis A. et al., Phys. Rev. Lett., 86 (2001) 3024
[7] Stieglitz R. and M
uller U., Phys. Fluids, 13 (2001) 561
[8] Bourgoin M. et al., Phys. Fluids, 13 (2001) 561; Peffley
N. L., Cawthorne A. B. and Lathrop D. P., Phys. Rev.,
E 61 (2000) 5287; Nornberg M. D. et al., Phys. Rev.
Lett., 97 (2006) 044503; Frick P. et al., Magnetohydrodynamics, 38 (2002) 136
[9] Monchaux R. et al., Phys. Rev. Lett., in press, (2006)
[10] P
etr
elis F. et al., Phys. Rev. Lett., 90 (2003) 174501;
Volk R., Odier P. and Pinton J.-F., Phys. Fluids, 18
(2006) 085105; Volk R. et al., Phys. Rev. Lett., 97 (2006)
074501
[11] Duddley M.L. and James R. W., Phys. Rev. Lett., 97
(2006) 044503; Nore C. et al., Phys. Plasmas, 4 (1997)
1; Ponty Y. et al., Phys. Rev. Lett., 94 (2005) 164502;
Minnini P.D. et al., Astrophys. J., 626 (2005) 853; Laval
J. P. et al., Phys. Rev. Lett., 96 (2006) 204503
[12] Ravelet F. et al., Phys. Fluids, 17 (2005) 117104
[13] MacFadden P.L. and Merrill R.T., Physics of the
Earth and Planetary Interiors, 91 (1995) 253
[14] Sarson G. R. and Jones C. A., Physics of the Earth and
Planetary Interiors, 111 (1999) 3; Wicht J. and Olson
P., Geochemistry, Geophysics and Geosystems (G-cubed),
5 (2004)
[15] Giesecke A., R
udiger G. R. and Elstner D., Astron.
Nachr., 326 (2005) 693; Stefani F. and Gerbeth G.,
Phys. Rev. Lett., 94 (2003) 184505; Widrow L., Rev. Mod.
Phys., 74 (2002) 775

REFERENCES

present addresses:
(L.M., +) IFREMER / Laboratoire de Physique des
Oceans, CNRS UMR 6523, BP70, F-29280, Plouzane
(France)
(F.R. *) Laboratory for Aero and Hydrodynamics,
TU-Delft (The Netherlands)

(M.B. #) Laboratoire des Ecoulements

Geophysiques et
Industriels, CNRS UMR 5519, BP53, F-38041 Grenoble
(France)

magnetic field. This work is supported by the French

institutions: Direction des Sciences de la Mati`ere and
Direction de lEnergie Nucleaire of CEA, Minist`ere de la
Recherche and Centre National de Recherche Scientifique
(ANR 05-0268-03).
The experiment is operated at
CEA/Cadarache DEN/DTN.

Magnetic field reversals in an experimental turbulent dynamo