Understanding and Controlling Plasma Rotation in Tokamaks: Proefschrift

Understanding and controlling
plasma rotation in tokamaks
Proefschrift
ter verkrijging van de graad van doctor aan

de Technische Universiteit Eindhoven, op gezag
van de Rector Magnificus, prof.dr.ir. C.J. van Duijn,
voor een commissie aangewezen door het College
voor Promoties in het openbaar te verdedigen op
maandag 22 januari 2007 om 16.00 uur
door
Maarten De Bock
geboren te Sint-Niklaas, België
Dit proefschrift is goedgekeurd door de promotoren:
prof.dr. N.J. Lopes Cardozo

en
prof.dr. F.C. Schüller
Copromotor:
dr. R. Jaspers
CIP-data library Technische Universiteit Eindhoven

De Bock, Maarten
Understanding and controlling plasma rotation in tokamaks/ by Maarten De Bock.–
Eindhoven : Technische Universiteit Eindhoven, 2007.– Proefschrift.
ISBN-10: 90-386-2172-8
ISBN-13: 978-90-386-2172-2
NUR 926
Trefwoorden: plasmafysica / kernfusie / fusieplasma’s / plasmarotatie / magnetische

velden / magnetohydrodynamica / plasmadiagnostiek
Subject headings: plasma / nuclear fusion / plasma rotation / magnetic fields / magne-
tohydrodynamics / plasma diagnostics
The work described in this thesis was performed as part of a research programme
of Stichting voor Fundamenteel Onderzoek der Materie (FOM) with financial support
from the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO), the
Forschungszentrum Jülich GmbH and EURATOM. It was carried out at the Institut
für Plasmaphysik at the Forschungszentrum Jülich GmbH in collaboration with the
FOM-Institute for Plasma Physics Rijnhuizen.
Typeset in LATEX 2ε .
Printed by B-SG, Forschungszentrum Jülich GmbH, D-52425 Jülich, Germany.
Cover design by Wim De Pagie, artist impression of an excited carbon ion that passes by.
Just remember that you’re standing
on a planet that’s evolving, and revolving
at 900 miles an hour.
That’s orbiting at 19 miles a second,

so it’s reckoned,
a sun that is the source of all our power.
The sun and you and me,

and all the stars that we can see,
are moving at a million miles a day.
In an outer spiral arm,

at 40,000 miles an hour,
of the galaxy we call the Milky Way.
Galaxy song, by Monty Python
Understanding and controlling plasma rotation in tokamaks
Contents
Contents v
1 Introduction 1
1.1 Thermonuclear Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Why do we need fusion? . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Fusion reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Tokamaks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Importance of plasma rotation . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 This thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 List of publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2 Theory of plasma rotation 11

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 From single particle motion to plasma rotation . . . . . . . . . . . . . . . . 12
2.2.1 Magnetohydrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Tokamak equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.3 Definition of plasma rotation . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Momentum transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.1 Momentum confinement time . . . . . . . . . . . . . . . . . . . . . . 19
2.3.2 Momentum sources and sinks . . . . . . . . . . . . . . . . . . . . . . 20
2.3.3 Momentum balance . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Neoclassical transport and rotation . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.1 Neoclassical theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4.2 Poloidal flow damping . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.3 Spontaneous neoclassical rotation . . . . . . . . . . . . . . . . . . . . 27
2.5 Impurity and MHD Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5.1 Impurity rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5.2 MHD rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3 TEXTOR, DED and diagnostics 35

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 The TEXTOR tokamak and its heating systems . . . . . . . . . . . . . . . 35
3.3 The Dynamic Ergodic Divertor . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.4 Co- and counter directions in TEXTOR . . . . . . . . . . . . . . . . . . . . 42
3.5 Plasma diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5.1 Charge exchange recombination spectroscopy . . . . . . . . . . . . . 44
iii
3.5.2 Electron Cyclotron Emission . . . . . . . . . . . . . . . . . . . . . . 45
3.5.3 Thomson Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5.4 Interferometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4 Charge Exchange Recombination Spectroscopy 49

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2 Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.3 Additional effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.3.1 Energy dependence of the emission rate . . . . . . . . . . . . . . . . 54
4.3.2 Non-thermal line broadening . . . . . . . . . . . . . . . . . . . . . . 56
4.3.3 Non-CX emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.4 CXRS at TEXTOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5 Plasma rotation at TEXTOR 67

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.2 Momentum and energy confinement time in TEXTOR . . . . . . . . . . . . 67
5.3 Ohmic rotation in TEXTOR . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.4 Impurity rotation in TEXTOR . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.5 MHD rotation in TEXTOR . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6 Measurements of plasma rotation during DED operation 75

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.2 A general overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.3 Change in rotation as a function of DED current . . . . . . . . . . . . . . . 79
6.4 Plasma rotation locked to the DED frequency . . . . . . . . . . . . . . . . . 85
6.5 Transition threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.6 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
7 Theory of perturbation fields 95

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.2 Penetration of modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2.1 Tearing modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2.2 Rutherford equation, EM torque and Viscous torque . . . . . . . . . 97
7.2.3 Effect of an external helical perturbation . . . . . . . . . . . . . . . 101
7.2.4 A locked mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.2.5 Mode penetration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7.3 The effect of an external perturbation in the plasma edge . . . . . . . . . . 111
7.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
7.3.2 The stochastic torque in the plasma edge . . . . . . . . . . . . . . . 111
7.3.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
7.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
8 Momentum input by a stochastic edge field 115

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
8.2 12/4 DED operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
8.3 Analysis of discharge #95592 . . . . . . . . . . . . . . . . . . . . . . . . . . 117
iv
8.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
9 Mode penetration 125

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2 Force balance model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2.1 Electromagnetic Torque . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2.2 Force balance at q = 2 . . . . . . . . . . . . . . . . . . . . . . . . . . 127
9.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
10 Changing the plasma rotation by EM waves 137

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
10.2 Changing the momentum by ion cyclotron radiation . . . . . . . . . . . . . 137
10.3 Influence of ECRH on rotation . . . . . . . . . . . . . . . . . . . . . . . . . 140
10.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
11 Discussion, conclusions and outlook towards future devices 143

11.1 Exciting tearing modes with the DED in TEXTOR . . . . . . . . . . . . . . 143
11.2 Influence of ICRH and ECRH on rotation in TEXTOR . . . . . . . . . . . 146
11.3 Rotation in ITER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Summary 151
Samenvatting 155
Curriculum Vitae 159
Acknowledgements 161
A List of the discharges used in this thesis 163
Bibliography 165
v
Chapter 1
Introduction
1.1 Thermonuclear Fusion

1.1.1 Why do we need fusion?
The world’s economy runs on fossil fuel. It heats our houses, it drives our cars and delivers
our electricity. However, the supply of oil, coal and gas is exhaustible. If no action is taken
an energy crisis is imminent in the next couple of centuries, maybe decades. Apart from the
inevitable fact that we will run out of fuel, burning coal, oil and gas also have an impact on
the environment. The signs of global warming due to CO2 and the accompanying climate
change become increasingly more alarming.
Nevertheless, mankind will not give up the present standard of living, so the energy
consumption is not expected to reduce. It will rather increase. Sources of energy, other
than fossil ones, are therefore needed. Those that are presently available, however, have
a too low energy density (solar, wind and bio-energy) or produce long term radioactive
waste (nuclear fission).
Thermonuclear fusion holds the promise of an abundant supply of energy, without
affecting the climate and with a minimum of short term radioactive waste.
1.1.2 Fusion reactions

Fusion is what keeps the sun and all the stars burning. It is a process where two light
nuclei fuse to form a heavier nucleus. It is found that the sum of the masses of the two light
nuclei is larger than the mass of the fusion products. The difference in mass is, according
to Einstein’s well-known formula E = mc2 , converted into energy. There are several fusion
reactions possible. The reaction that is the best candidate for energy production on earth
is the deuterium-tritium reaction, because its cross-section is reasonable large. In this
reaction two hydrogen isotopes, deuterium and tritium, fuse resulting in a helium nucleus
and a neutron:
D + T → He + n + 17.6 M eV (1.1)
The produced energy is divided over the fusion products: the neutron receives a kinetic
energy of 14.1 M eV , the helium nucleus, also called α-particle, has an energy of 3.5 M eV .
Deuterium is largely available in the oceans on earth. Tritium is not, but can be produced
by a nuclear reaction of lithium – also widely available – and an energetic neutron. The
‘waste’ of this fusion reaction is helium: a non-toxic, non-radioactive gas.
1
Chapter 1 - Introduction
T He
n
D
Figure 1.1 : The two hydrogen isotopes, deuterium and tritium, fuse to form
helium and a neutron. In this reaction 17.6 M eV of energy is released.
The nuclei of deuterium and tritium do not fuse spontaneously. Because they both
have a positive charge, the repelling Coulomb force prevents the reaction. A sufficiently
high kinetic energy of the nuclei is needed to overcome the Coulomb force. This high
kinetic energy is achieved in a gas that has a temperature T of about 100 million degrees
centigrade. At this kind of temperature gasses are ionised. We do no longer call them a
gas, but call them a plasma. It is common to express temperatures in a plasma with eV ,
where 1 eV ≈ 12000 C ◦ .
Not only a high temperature is needed. In order to have enough collisions between the
highly energetic nuclei the density n of the particles has to be high enough as well.
A third important parameter is the energy loss. If a fusion plasma loses its energy
faster to the outside world than it can gain energy, from fusion reactions and/or from
external heating, then the process will die out. The rate at which a plasma loses its
energy is given by 1/τE , where τE is called the energy confinement time.
A fusion reaction will be self-sustained if the product of the above three parameters –
temperature T , density n and confinement time τE – is sufficiently high. For the deuterium-
tritium reaction it is found to be:
n T τE > 5 · 1021 keV s/m3 (1.2)
The above criterion is the so-called Lawson criterion. The triple product n T τE is a
figure-of-merit for a fusion reactor: the higher it is, the better.
In order to have a reasonable cross-section for the deuterium-tritium reaction, a tem-
perature of 1 to 10 keV is needed. In order to keep a plasma at such high temperatures
we should avoid contact between the plasma and any material wall. A possible way of
doing this, is to use a magnetic trap that confines the plasma in the centre of a vacuum
chamber.
2 Section 1.1 - Thermonuclear Fusion

1.2 Tokamaks
The fact that a plasma mainly consists of charged particles can be used to confine it. Due
to Lorentz forces the movement of charged particles perpendicular to magnetic field lines
is restricted, causing the charge particle α to gyrate around the field line with cyclotron
frequency ωc = eα B/mα and a gyroradius, also called Larmor radius, ρL = u⊥ /ωc (see
fig. 1.2), where B is the magnetic field, eα the charge of the particle, mα the mass of the
particle and u⊥ the particle velocity perpendicular to the magnetic field.
Figure 1.2 : Charged particles gyrate closely around

the magnetic field lines. Hence, the geometry of
the magnetic field determines the geometry of the
plasma that consists of charged particles.
The movement of a charged particle parallel to the magnetic field is not restricted. In
order to confine a plasma effectively, the field lines should therefore close in themselves,
hence form a toroidal geometry. However, just a torus shaped vacuum vessel with a
toroidal magnetic field is insufficient to confine a plasma. The curvature of the magnetic
field causes electrons and ions to drift to the bottom, respectively the top of the torus,
resulting in an electric field. This electric field in its turn leads to an outward drift
of all particles, and therefore to a loss of confinement. To neutralise this electric field,
particles that drifted to the top of the machine should be brought to the bottom and
vice versa. This can be achieved by adding a poloidal component to the magnetic field.
In a tokamak configuration the poloidal magnetic field is generated by a toroidal plasma
current (see fig. 1.3). This plasma current is induced by a transformer, using the plasma as
the secondary winding1 . On top of these poloidal and toroidal field components radial and
vertical components are added to the magnetic field by external positioning and shaping
coils.
Poloidal and toroidal components add up to a set of helical field lines. These field lines
lie on magnetic surfaces. The magnetic structure in a tokamak can be seen as an infinite
set of nested magnetic flux surfaces. An important parameter in tokamak physics is the
safety factor q:
m r Bφ
q= ≈ (1.3)
n R Bθ
It describes the number of toroidal windings of a field line needed for one poloidal
winding (e.g. m=3, n=2 means that, when you follow a field line on this q = 3/2 surface,
you need 3 toroidal windings and 2 poloidal windings to reach your starting point again,
or 1.5 toroidal windings for 1 poloidal winding). At places with a (low) rational q-number
instabilities can easily arise. The effect of a minor fluctuation at a certain position is
amplified because the particles pass every q toroidal revolutions at that same position.
1
Another method to generate a poloidal field is using additional or specially shaped coils. This is done
e.g. in a stellarator. Those configurations will not be discussed in this thesis
Section 1.3 - Tokamaks 3

Figure 1.3 : The principle of magnetic confinement in a tokamak. For stability rea-
sons a helical magnetic field is needed. The toroidal component of this field comes
from field coils, the poloidal component is induced by the plasma current. By
making the plasma column the secondary winding of a transformer, high plasma
currents can be reached easily by feeding a current ramp to the primary winding of
the transformer.
1.3 Importance of plasma rotation

A fusion relevant plasma must fulfill three main criteria:
• The first criterion is the Lawson criterion, that says that the energy and particle
confinement must be high. This is achieved if the energy that is put into the plasma
stays sufficiently long inside the plasma. In a fusion reactor we thus have to minimise
the energy transport to the wall.
• The confinement must not only be high, it also has to be stable. Because the plasma
in a tokamak is confined by a magnetic field, a good understanding of the magnetic
stability is necessary. In other words: we need to know which magnetic instabilities
can occur and how we can avoid them.
• The power that goes into a fusion reactor eventually ends up at the wall of the
reactor. A good knowledge of the plasma-wall interaction is needed in order to (a)
optimise the first wall materials, so that they can cope with the power load, and (b)
keep impurities and dust out of the core plasma.
Several experiments around the world have shown that a high plasma rotation can
attribute to the first two criteria: the improvement of both stability and confinement.
The confinement in a tokamak is governed by the radial transport of energy from the
plasma centre to the plasma edge. A large part of this transport is driven by turbu-
4 Section 1.3 - Importance of plasma rotation

lence. Empirically it is observed that the confinement decreases when the power input
is increased. It is generally agreed upon that this confinement degradation is a result of
increased turbulent transport. An important breakthrough in fusion physics was the dis-
covery of the H-mode – H stands for high confinement – in the early 80’s [87]. When the
power input reached a certain threshold the confinement suddenly almost doubled. One
of the observations made during this transition was a sudden change in plasma rotation in
the edge. More precisely the gradient in poloidal rotation vθ and the radial electric field
Er – which is linked with rotation – increased. This observation led to the question: Does
a sheared rotation reduce turbulent transport?
Figure 1.4 : A simulation of turbulence with (A) and without (B) sheared plasma flow. The
contours of the fluctuation potential in a poloidal cross section are plotted. Large turbulent
transport exists along the iso-potential contours. It is clearly seen that the turbulent cells in
the case of sheared flow are much smaller than in the case without flow, which means that
the overall radial transport is lower. [58]
Now, twenty years later, the mechanism of turbulence suppression by rotational shear
is widely accepted. The physical picture of turbulence suppression by flow shear can
be looked upon as follows: turbulent transport is a result of fluctuations with a radial
correlation length ∆rc and a decorrelation time τc . These fluctuations can be seen as
turbulent cells with a radial extension of ∆rc and a lifetime τc . The transport induced by
the turbulent cells is given by the diffusion coefficient Dturb ∝ ∆rc2 /τc . A sheared flow,
i.e. a different velocity at each radial point, will shred these turbulent cells apart, leading
to smaller cells ∆rsheared < ∆rc . Consequently the diffusion coefficient, hence the radial
transport, is smaller [7]. In figure 1.4 the fluctuation potential is shown for a simulation
with and without sheared flow [58]. It is clearly seen that the turbulent cells are much
smaller in the case with flow (A) than in the case without flow (B).
A sheared plasma flow will thus improve the confinement. A straightforward method
for increasing the velocity gradient is increasing the velocity. Because the vacuum vessel
of a tokamak does not move, a velocity gradient exist between the plasma and the wall.
This gradient will be large if the plasma rotates fast.
Apart from improving the confinement, a fast rotating plasma also increases the sta-
bility of the magnetic configuration. In an ideal world the plasma in a tokamak is confined
in a set of perfectly nested flux surfaces. In the real world sources of free energy in the
plasma can break up and reconnect a flux surface, hence changing the magnetic topology
(see figure 1.5). These reconnected flux surfaces are called tearing modes or magnetic
islands. Tearing modes have an unfavourable effect on plasma confinement and can even
cause minor and major disruptions. We therefore usually try to avoid them.
Section 1.3 - Importance of plasma rotation 5

Figure 1.5 : In (a) the ideal magnetic topology is shown: a set of nested flux surfaces. In (b) two
of the flux surfaces – the q=1 and q=2 surfaces – are broken up and reconnected, resulting
in m/n = 2/1 (dark grey) and m/n = 1/1 (black) islands.
Not only an excess in free energy can drive tearing modes. They can also be formed
by an externally applied perturbation field. This perturbation field can be applied on
purpose, but usually it is the result of misalignment of coils, a non-axisymmetric wall of
the vacuum vessel, et cetera. In this case we call it an error field. Because no machine is
perfectly aligned or perfectly symmetric, every fusion device has an error field. And due
to this error field tearing modes can develop.
Error fields are static, and as a result the tearing modes they excite do not move
as well. Tearing modes, however, have to rotate with the plasma velocity. In rotating
plasmas, therefore, an error field does not excite tearing modes: the tearing modes are
suppressed. The interaction between the suppressed tearing modes and the error field will,
however, slow down the plasma. Once the suppressed tearing modes are at rest in the
frame of the error field large tearing modes will develop. This is called mode excitation.
Because the error field needs to slow down the plasma to great extent, we expect a high
threshold for mode excitation in fast rotating plasmas. The threshold for mode excitation
by a static perturbation field in TEXTOR, as a function of the plasma rotation is given
in figure 1.6. One sees that for fast rotating plasmas the threshold is indeed higher than
for slow rotating plasmas.
Another type of magnetic instabilities are the resistive wall modes (RWM). These
RWM’s occur in plasmas with a high plasma energy (β). When they lock, i.e. when they
do not move with respect to the vessel wall, RWM’s cause a disruption. In high β devices,
like ITER, the prevention of RWM’s is therefore crucial.
Although RWM’s have a different topology than tearing modes, they can be treated
with the same philosophy as tearing modes: (a) in a fast rotating plasma RWM’s will be
suppressed, (b) the interaction between the suppressed RWM and the conducting (i.e.
resistive) tokamak wall will slow down the plasma and (c) once the RWM is at rest with
respect to the wall the RWM grows and causes a disruption.
A fast rotating plasma rotation has both good plasma confinement, through flow shear
turbulence suppression, and stability, due to the high threshold for error field driven tearing
modes and resistive wall modes. In order to optimise the performance of fusion devices, it
is vital to know (a) how plasma rotation improves confinement and stability and (b) how
a large plasma rotation can be driven.
6 Section 1.4 - Importance of plasma rotation

Mode threshold (IDED / kA)

1.5
0.5
0
−1 −0.5 0 0.5 1 1.5
Ωφ,0(q=2) (rad/s) 4
x 10
Figure 1.6 : The threshold at which an externally applied, static perturbation field excites a 2/1
tearing mode at the q=2 surface in TEXTOR, is plotted as a function of the plasma rotation
frequency at the q=2 surface. At high rotation velocities, in both directions, the threshold is
high. At low rotation velocities, the threshold is lower.
1.4 This thesis

Above, two benign aspects of plasma rotation were mentioned: turbulence suppression by
flow shear and increased stability against error field driven tearing modes and resistive
wall modes. In this thesis will concentrate on the error field driven tearing modes. The
reason for selecting this specific topic is the fact that the TEXTOR tokamak, where the
work presented in this thesis was executed, is equipped with two tangential neutral beams,
that allow us to control the plasma rotation, and a dynamic ergodic divertor (DED), that
allows us to create and control the external perturbation field.
For the next generation of fusion reactors, like ITER, the stability against resistive
wall modes (RWM) will be more important than the stability against tearing modes. The
investigation of RWM’s is, however, difficult on TEXTOR due to the low β. Nevertheless
we can gain insight in the process of RWM locking by investigating the excitation of
tearing modes, because both processes are very similar.
When we have a closer look at figure 1.6, we see that the simple statement ‘the higher
the rotation, the higher the threshold’ is not completely valid. The minimal threshold
is not located at zero rotation. Furthermore the increase in threshold depends on the
direction of the rotation; at the right side of the minimum the threshold increases faster
with increasing rotation than at the left side of the minimum. The relation between error
field modes and plasma rotation seems to be more complex than we have thought. It is
advisable to have a closer look into this subject.
A first question we will try to answer in this thesis is:
How do we avoid the excitation of unfavourable modes by an
external perturbation field?
Section 1.4 - This thesis 7

To answer this question we can reverse it:
How can we excite modes with an external perturbation field?
Because it can be noticed on several tokamaks that the mode excitation is linked with the
plasma rotation, we will narrow down the above question to:
How does the threshold for mode excitation by an external

perturbation field depend on the plasma rotation?
with as specific sub-questions:
- Why does the minimal threshold occur at a finite rotation

velocity?
- Why is the threshold plot asymmetric around the minimum?
In the previous section it was said that the dependence of the threshold on the plasma
rotation was a result of the fact that the error field had to slow down the plasma first. We
should thus also ask ourselves how this slowing down mechanism works:
How does the external perturbation field change the plasma

rotation?
Because the plasma rotation is driven by torques, the aim of this thesis is to identify
the different torques that a perturbation field can exert onto a plasma. Once these
torques are known the interplay between these perturbation field torques and the torques
that are present in plasma without external perturbation field have to be investigated.
As a result we hope to be able to reconstruct the change in plasma rotation as a function
of the external perturbation, and consequently predict the threshold for mode excitation
as a function of the plasma rotation.
The thesis is structured as follows. In the second chapter the basic theoretical frame-
work is introduced. Therefore a selection is made from what is known about plasma
rotation in the literature. In chapter 3 the TEXTOR tokamak and the DED are intro-
duced. The TEXTOR tokamak, with its two neutral beams, and the DED, that provides
us with a fully adjustable perturbation field, are the tools that allow us to perform dedi-
cated experiments on the interplay between a perturbation field and the plasma rotation.
In order to investigate plasma rotation we have to be able to measure it. The technique
used for rotation measurements is charge exchange recombination spectrometry and is
introduced in chapter 4. Now that the theory, the machine and the measurement tech-
nique are introduced, chapter 5 briefly looks into what plasma rotation typically means at
TEXTOR. Some theoretical predictions from chapter 2 are applied to TEXTOR and are
compared with measurements.
In chapter 6 we have a look into the measurements of plasma rotation during DED
operation. These measurements do not fully match our expectations. In chapter 7 we
therefore have a look into the literature. We find that our expectation is based on the
assumption that there is only one type of force exerted on the plasma by the perturbation
field: an electromagnetic force positioned at the rational q=surfaces – comparable with
the slip force in an electric induction motor. We also find in the literature that another
8 Section 1.4 - This thesis

force is possible. The perturbation field can create a stochastic region in the plasma edge.
Parallel electron loss in the stochastic region causes subsequently a stochastic force in the
plasma edge. To get a good description of the plasma rotation and the mode excitation
both the electromagnetic force and the stochastic force have to be taken into account.
In order to get a better insight into the stochastic force, an experiment was carried
out under conditions in which the electromagnetic force could be neglected. This is pre-
sented in chapter 8. It allowed us to compare the theory of the stochastic force given in
chapter 7 to the experimental observation. It also led to a straightforward expression for
the stochastic force.
Finally, in chapter 9 the interplay between perturbation field and plasma rotation is
investigated in detail. Including both the electromagnetic and the stochastic force we
were able to get a prediction for the change in plasma rotation and the threshold for mode
excitation that is in agreement with the measurements.
The work presented in chapters 6 to 9 confirms that a fast rotating plasma has a better
resistance against the excitation of tearing modes than a slow rotating one. In present day
devices high rotation velocities can be achieved using neutral beam injection. This is not
the case for the next generation of fusion reactors . Extra sources of momentum input,
apart from neutral beam injection, are therefore necessary. In the literature ion cyclotron
heating (ICRH) is often mentioned as a good candidate for momentum input. And if
ion cyclotron waves can change the plasma rotation, it is a small step to assume that
also electron cyclotron waves (ECRH) will have an influence on plasma rotation. At the
TEXTOR tokamak both ICRH as ECRH systems are available. In chapter 10 experiments
are presented that investigate the influence of ICRH and ECRH on the plasma rotation.
In a final chapter a prospect is given on what the results of this thesis mean for future
devices like ITER.
1.5 List of publications

In this section a list of publications and conference contributions is given. Those related
to this thesis are marked with an asterisk.
Journal publications
[1] M. De Bock, K. Jakubowska, M. von Hellermann et al. “Measuring one-dimensional
and two-dimensional impurity density profiles on TEXTOR using combined charge
exchange-beam emission spectroscopy and ultrasoft x-ray tomography.” Review of Sci-
entific Instruments, vol. 75 pp. 4155—4157 (2004).
∗ [2]
M.F.M. de Bock, I.G.J. Classen et al. “The influence of plasma rotation on tearing
mode excitation in TEXTOR.” submitted to Nuclear Fusion.
∗ [3]
I.G.J. Classen, M.F.M. de Bock et al. “Dynamics of tearing modes in the presence of
a perturbation field.” submitted to Nuclear Fusion.
∗ [4]K.H. Finken, S.S. Abdullaev, M.F.M. De Bock et al. “Toroidal Plasma Rotation
Induced by the Dynamic Ergodic Divertor in the TEXTOR Tokamak.” Physical Review
Letters, vol. 94 pp. 15003-1–15003-5 (2005).
Section 1.5 - List of publications 9

∗ [5]
Y. Kikuchi, M.F.M. de Bock, K.H. Finken et al. “Forced Magnetic Reconnection
and Field Penetration of an Externally Applied Rotating Helical Magnetic Field in the
TEXTOR Tokamak.” Physical Review Letters, vol. 97 pp. 085003-1–085003-4 (2006).
∗ [6]
H. Koslowski, E. Westerhof, M.F.M. de Bock et al. “Tearing mode physics studies
applying the Dynamic Ergodic Divertor on TEXTOR.” Plasma Physics and Controlled
Fusion, vol. 48 pp. B53–B61 (2006).
∗ [7]
M. Lehnen, S. Abdullaev, W. Biel, M.F.M. de Bock et al. “Transport and divertor
properties of the dynamic ergodic divertor.” Plasma Physics and Controlled Fusion, vol.
47 pp. B237–B248 (2005).
∗ [8]R. Wolf, W. Biel, M.F.M. de Bock et al. “Effect of the Dynamic Ergodic Divertor
in the TEXTOR Tokamak on MHD Stability, Plasma Rotation and Transport.” Nuclear
Fusion, vol. 45 pp. 1700–1707 (2005).
Conference contributions
∗ [1]M. de Bock, R. Jaspers, M. von Hellermann et al. “Plasma rotation during operation
of the dynamic ergodic divertor in TEXTOR.” 31st EPS Conference on Controlled Fusion
and Plasma Physics, vol. 28G, pp. P-1.117. (2004).
∗ [2]
M. de Bock, C. Busch, K.H. Finken et al. “Plasma rotation during operation of the
dynamic ergodic divertor in TEXTOR.” 32nd EPS Conference on Controlled Fusion and
Plasma Physics, vol. 29C, pp. P-2.042. (2005).
∗ [3]
C. Busch, M. de Bock, K.H. Finken et al. “Impact of the DED on ion transport and
poloidal rotation at TEXTOR.” 32nd EPS Conference on Controlled Fusion and Plasma
Physics, vol. 29C, pp. P-1.016. (2005).
10 Section 1.5 - List of publications

Chapter 2
Theory of plasma rotation
2.1 Introduction
In order to tackle the questions raised in the previous chapter, a basic theoretical
framework is needed. The basic theory of the dynamics of a plasma is very well described
in the literature [88, 32, 2]. In this chapter we have selected all the information needed
when discussing the rotation of a plasma.
We start with the general description of a plasma as a collection of individual particles,

with each their own velocity, in an electro-magnetic environment. This description can
be transformed to a multiple fluids description, in which each species – ions, electrons
and neutrals – have a collective fluid velocity and a thermal velocity distribution. A
further simplification is made when only two plasma species are considered, electrons and
ions. These two are combined to describe the plasma as a single, conducting fluid: the
magnetohydrodynamic (MHD) equations of a plasma.
At this point the magnetic configuration comes into discussion. It will be shown that
the magnetic configuration in a tokamak consists of nested flux surfaces. The fluid velocity
can now be split in two components: a fluid velocity perpendicular to the flux surfaces is
referred to as convection, whereas fluid velocity on a flux surface is called rotation.
Having defined rotation as being fluid velocity on a flux surface, we have a closer look
into the direction of the rotation. We can split the rotation in a toroidal and poloidal
component, compatible with the symmetry of a tokamak, or in a component parallel and
a component perpendicular to the magnetic field, which has more physical relevance. We
also discuss the sources, the sinks and the transport of rotation. More specifically the
force or momentum balance equation is derived. This will turn out to be a powerful tool
to interpret the experiments on plasma rotation.
Up to this point, we have assumed the transport coefficients for particle, energy and
momentum to be given. We have not looked into the underlying physical mechanisms.
This means whether classical transport (collisional diffusion), neoclassical transport
(taking into account the toroidal geometry of a tokamak) or turbulent transport is
considered. When following neoclassical transport theory, we find out that poloidal
plasma rotation is strongly damped, so that we can concentrate in this thesis on the
toroidal rotation. It also follows from neoclassical theory that plasmas have a tendency
11
Chapter 2 - Theory of plasma rotation
to rotate spontaneously. This intrinsic or spontaneous rotation could be very helpful in

establishing a fast rotating, well-confined plasma. The influence that turbulent transport
has on rotation – e.g. zonal flows – falls out of the scope of this thesis.
Although the single fluid MHD picture provides good insight into the physics, plasma
diagnostics measure the properties of the several plasma species rather than those of the
single plasma fluid. In the last section we therefore return to the multiple fluid picture
and try to link the electron fluid rotation and the impurity fluid rotation, to the single
fluid plasma rotation.
2.2 From single particle motion to plasma rotation

2.2.1 Magnetohydrodynamics
A plasma is a large collection of several types of particles: ions, neutrals and electrons. The
statistical behaviour of a large number of particles is governed by the Boltzmann equation
(2.1). It describes the change of the distribution function fα (x, u, t) of the particles of the
type α (α being ions or electrons or neutrals), where x describes the position and u the
velocity of each particle.

∂fα F ∂fα
+ u · ∇fα + · ∇u fα = (2.1)
∂t mα ∂t c
The distribution function fα (x, u, t) changes as a result of the forces F and collisions
∂fα
∂t . In a plasma the main forces are long-range Lorentz forces F = qα (E + u × B).
c
Collisions between different plasma species result in friction forces (momentum sources
and sinks), viscosity (momentum transport) and resistivity.
In order to get a full description of the plasma the above Boltzmann equation has to
be solved for each plasma species, together with the Maxwell equations for E and B and
an appropriate description for the collisions. This set of Boltzmann, Maxwell and colli-
sion equations is called the kinetic model. Solving these equations is quite cumbersome
and usually it is not necessary to describe plasma behaviour in terms of the distribution
function and microscopic quantities. A simplified model is therefore derived: magnetohy-
drodynamics.
A first step in simplifying the above kinetic model is evaluating the appropriate mo-
ments of equation (2.1). A moment hAi is defined by:
Z
1
hAi ≡ Afα du (2.2)
nα
R
where nα = fα du is the density distribution. Multiplying equation (2.1) by A and
integrating over the velocity space, we get the conservation equations for mass (A = 1)
and momentum (A = mu):
dnα
+ nα ∇ · v α = 0 (2.3)
dt
dv α
nα mα = nα qα (E + v α × B) − ∇ · P α + Rα (2.4)
dt
12 Section 2.2 - From single particle motion to plasma rotation

d ∂
The total time derivative in equations (2.3)-(2.4) is defined as dt = ∂t + v α · ∇. It
is seen that by evaluating the moments of equation (2.1) we now have a set of equations
that describes the plasma by macroscopic quantities:
vα = hui the fluid velocity

Pα = nα mα hwwi the pressure tensor1
Rα the momentum transfer due to collisions between
different plasma species (i.e. friction)
1 w = u − v is the thermal velocity
α
A next step is to consider only 2 particle species in the plasma: electrons and ions.
When we combine equation (2.3) for electrons and ions, the 1 fluid conservation of mass
relation is found:
dρ
+ ρ∇ · v = 0, (2.5)
dt
where ρ = mi ni + me ne is the mass density and v = minniim
v i +me ne v e
i +ne me
the centre of mass
velocity. It is clear that me mi and that, due to quasi-neutrality, electron and ion
density are equal n = ne = ni . Hence it can be said that ρ ≈ nm, m ≈ mi and v ≈ v i .
The combination of equation (2.4) for electrons and ions results in a new momentum
conservation and a current conservation equation:
dv
= j × B − ∇p − ∇ · Π
ρ (2.6)
dt
1 me ∂j
E+v×B = − + j × B − ∇pe − ∇ · Πe + Re (2.7)
ne e ∂t
In short it can be said that the description of the momenta of the two fluids is replaced
by the momentum equation of the centre of mass velocity v – which is almost equal to the
ion fluid velocity due to the large difference in electron and ion mass – and a description of
the electrical current density j – which relates to the difference in ion and electron velocity.
In equations (2.6)-(2.7) the pressure tensor P = P e + P i is split up in the scalar

pressure p = pe + pi and the anisotropic part Π = Πe + Πi , so that P = p · I + Π.
The scalar pressure relates to temperature via pα = nα Tα and p = nT , which implies
T = Te + Ti . Πα depends – like Rα – on collisions, but – in contrast to Rα – also on
velocity gradients (see subsection 2.3.3 and [80, 88, 9]). Πα describes the viscosity and is
therefore also called the viscous stress tensor. Both Πe and Re in equation (2.7) give rise
to resistivity.
Equation (2.6) is also known as the force balance equation: it is indeed nothing more
than Newton’s law m dv dt = F . Equation (2.7) is the generalised Ohm’s law, which links
currents, flows, resistivity and electromagnetic fields.
As seen in (2.3) and (2.4) in each conservation equation of a specific moment there
appears a higher order moment. A term to close the fluid equations is therefore needed.
Section 2.2 - From single particle motion to plasma rotation 13

Usually the assumption of an adiabatic fluid

is made: no heat is transferred to or from
the plasma fluid. This is expressed by dt ρpγ = 0.
d
The adiabatic assumption, equations (2.5)-(2.7) and the Maxwell equations together
form the MHD equations: a single fluid description of the plasma. The complete set of
MHD equations is given below:
Mass conservation:
dρ
+ ρ∇ · v = 0 (2.8)
dt
Momentum conservation or force balance:
dv
ρ = j × B − ∇p − ∇ · Π (2.9)
dt
Generalised Ohm’s law:
1 me ∂j
E+v×B = − + j × B − ∇pe − ∇ · Πe + Re (2.10)
ne e ∂t
Assumption of an adiabatic fluid:

d p
=0 (2.11)
dt ργ
Faraday’s law:
∂B
∇×E =− (2.12)
∂t
Ampère’s law:
1 ∂E
∇ × B = µ0 j + 2 (2.13)
c ∂t
Gauss’ law:
∇·B =0 (2.14)
Charge conservation:
∇·j =0 (2.15)
2.2.2 Tokamak equilibrium

The description of the equilibrium of a magnetically confined plasma in a tokamak allows
a further reduction of the MHD equations. Obviously an equilibrium is a stationary
∂
situation, so all partial derivatives to time can be set to zero ( ∂t = 0). Furthermore
hot tokamak plasmas have low resistivity and viscosity. We can therefore – in a first
approximation – neglect them. Finally we assume an incompressible plasma fluid. By
definition incompressibility means dρdt = 0. This is, according to mass conservation (2.8),
equivalent with ∇ · v = 0. The application of these reductions to the MHD equations leads
to the ideal MHD equations. The most used equation is the ideal force balance equation,
that expresses the balance between the kinetic pressure gradient and the Lorentz force:
∇p = j × B (2.16)
From equation (2.16) it can easily be seen that in an equilibrium magnetic field lines
and current lines lie on surfaces of constant pressure:

B · ∇p = 0
j · ∇p = 0 (2.17)
These surfaces of constant pressure are also surfaces of constant poloidal magnetic flux
(see below) and are therefore called flux surfaces. Most important plasma parameters, like
pressure, density, temperature, field line helicity, are flux functions, i.e. they are constant
on a flux surface. The exact geometry of these flux surfaces can be calculated with the
Grad-Shafranov equation; this is the expression for the ideal force balance equation (2.16)
in a toroidal geometry [32, 77, 88].
Figure 2.1 : Magnetic flux surfaces in a tokamak
Grad-Shafranov equation To calculate the magnetic R configuration in a tokamak it is

useful to introduce the poloidal flux function Ψ = S B θ · dS. It describes the poloidal
flux per radian toroidal angle φ through a surface going from the magnetic axis to a point
(Z, R) (see fig. 2.1). The definition of the flux function and the right-handed cylindrical
coordinate system (Z, R, φ) given in figure 2.1 allows us to rewrite the magnetic field as:
1 1 ∂Ψ 1 ∂Ψ
B= ∇Ψ × eφ + Bφ eφ ⇔ BR = − , BZ = (2.18)
R R ∂Z R ∂R
Substitution of (2.18) in (2.17) leads to ∇Ψ × ∇p = 0, which implies Ψ = Ψ(p) or
p = p(Ψ). The latter form is more customary; it states that pressure is a flux function.
Defining F ≡ RBφ and using Ampère’s law (2.13), j can be rewritten as:
1 1 ∂F 1 ∂F
j= ∇F × eφ + jφ eφ ⇔ jR = − , jZ = (2.19)
µ0 R µ0 R ∂Z µ0 R ∂R
Substituting (2.19) in (2.17) gives ∇F × ∇p = 0. This proves F = F (p) and using
p = p(Ψ), it follows that F = F (Ψ) is a flux function as well. Including (2.18) and (2.19)
in the force balance equation, we end up with an partial differential equation describing
the equilibrium of in a tokamak: the Grad-Shafranov equation [32, 77, 88].
∂2Ψ

∂ 1 ∂Ψ dp dF
R + 2
= −µ0 R2 −F (2.20)
∂R R ∂R ∂Z dΨ dΨ

This equation has three variables – Ψ, F and p, corresponding with B, j and p – so that
two out of the three variables have to be known in order to resolve the third one. In reality
initial estimates are taken for p, j and B, based on measurements of density, temperature,
the external B-fields and q, which is a measure for the field line helicity: q ≈ rBφ /RBθ .
In an iterative process p, j and B subsequently converge to a self-consistent solution. As a
result one gets the total magnetic equilibrium field within the plasma, the current density
profile, and the pressure p as a flux function. The topology of the equilibrium field shown
in figure 2.1. It shows that the magnetic flux surfaces are nested. Figure 2.2 gives typical
values for the toroidal field Bφ , the toroidal current density jφ and, corresponding to that,
the q-profile.
4
Shafranov shift
3.5
3
q
Bφ (T), jφ (MA/m2), q
Bφ
2.5
1.5
1
jφ
0.5
0
−0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4
R − R0 (m)
Figure 2.2 : Magnetic field, current density and q-profile for a typical ohmic
discharge in TEXTOR (Bφ = 2.25 T, Ip = 350 kA, Te = 1 keV, ne =
3 · 1019 m−3 ).
One of the results of the Grad-Shafranov equation – or more precisely: one of the
consequences of the toroidal shape of a tokamak – is that the magnetic axis lies not in the
centre of the plasma. It is shifted to the outside of the tokamak. This shift is called the
Shafranov shift. It is also indicated in figure 2.2.
Influence of plasma velocity on the equilibrium The Grad-Shafranov equation

expresses the balance between the force due to the pressure gradient and the Lorentz
force. When a plasma rotates, there is another force we should take into account: the
centrifugal force. Including the centrifugal force into the force balance will result in a
different equilibrium. In other words: fluid velocity has an influence on the tokamak
equilibrium.
The Grad-Shafranov equation given above does not take plasma velocity into account.
In [8] the Grad-Shafranov equation, including plasma velocity is derived. It is shown that
the pressure p is no longer a flux function, it is a function of flux Ψ and R: p(Ψ, R). There-
fore a single modified Grad-Shafranov equation does not suffice to get a full description

of the equilibrium; a set of equations is needed. In [8] a pure toroidal plasma velocity is
assumed; Ωφ = vφ /R. The set of equations describing the equilibrium then are:
∂2Ψ

∂ 1 ∂Ψ ∂p dF
R + = −µ0 R2 −F
∂R R ∂R ∂Z 2 ∂Ψ dΨ
1 ∂p 1 2
|Ψ = Ω (Ψ) (2.21)
ρ ∂R 2 2 φ
Equation (2.21) shows that plasma rotation has an influence on the magnetic equi-
librium in a tokamak. The equilibrium determines to great extent the plasma stability
and behaviour. For large rotation velocities the plasma flow should therefore be taken
into account when calculating the equilibrium configuration. It is however shown in [8]
that for subsonic velocities solving the ‘static’ Grad-Shafranov equation (2.20) suffices. In
most tokamaks the plasma velocity is subsonic, so it is justified to neglect the influence of
plasma velocity on the equilibrium.
2.2.3 Definition of plasma rotation

Starting from the velocity of individual particles u, over the fluid velocity of each plasma
species v α , we ended up with the fluid velocity v of the single plasma fluid described by
the MHD equations. The last step is to define which part of this fluid velocity v is rotation
and which part represents particle flux.
The total plasma fluid velocity v consists of a toroidal component vφ and a component
in the poloidal plane. The component in the poloidal plane can in its turn be split up in a
part parallel to the flux surfaces vθ and a part perpendicular to the flux surfaces vr . For
circular, unshifted flux surfaces θ and r are polar coordinates in the poloidal plane, for
non-circular, shifted flux surfaces the situation is a bit more complicated. For simplicity
we will use polar coordinates in the poloidal plane: vθ and vr will be called the poloidal
and radial components of the plasma velocity respectively.
Thanks to the symmetries of a tokamak configuration, the toroidal and poloidal veloc-
ities describe plasma rotation. Because both the poloidal and the toroidal component of
the plasma velocity lie on a flux surface the definition of plasma rotation is:
Plasma rotation is the part of the fluid velocity that lies on a

flux surface.
Plasma rotation is often expressed by the angular frequency rather than by the rotation
velocity: Ωφ ≡ vφ /R, Ωθ ≡ vθ /r.
The radial component of the plasma velocity – perpendicular to the flux surfaces –
describes particle convection. It is usually expressed as a particle flux: Γ = nmvr .
Perpendicular and parallel rotation The θ and φ coordinates correspond with the
geometry of a tokamak. From a physics point of view it is also interesting to describe
the plasma rotation by the velocity component vk parallel to the magnetic field and the
component v⊥ perpendicular to the field.

B
Bθ
vθ v
v⊥ vk
vφ Bφ
Figure 2.3 : Plasma velocity tangential to a flux surface – i.e. rotation – is shown here.
For this we cut open a flux surface. The rotation can be expressed in toroidal and
poloidal components, or in components parallel or perpendicular to the magnetic
field. Due to the fact that Bφ /Bθ 1, the parallel rotation is mainly in the toroidal
direction, the perpendicular rotation is mainly in the poloidal direction.
The velocity evolution is given by the force balance equation (2.9). For the parallel
velocity vk the only term at the right hand side of (2.9) is (∇ · Π)k . This means that the
parallel velocity depends on the viscosity and has no direct driving terms.
The perpendicular rotation velocity – i.e perpendicular to B, but not perpendicular
to the flux surface – also has a dependence on viscosity (∇ · Π)⊥ , but has a direct driving
term j r × B as well.
Viscosity influences the rotation profiles, but it will never by itself make the plasma
rotate. This means that we can only directly drive plasma rotation in the direction per-
pendicular to the magnetic field. This does not mean there is no parallel rotation – in fact
the parallel rotation is usually quite high – but the parallel rotation is always a result of
the redistribution of perpendicular velocity by the viscosity.
E × B-drift An example of the importance of perpendicular rotation is the E × B-drift.

In an ideal case – i.e. steady state and no resistivity – Ohm’s law (2.10) reduces to:
E+v×B =0 (2.22)
Taking the cross product of equation (2.22) with the magnetic field one gets,
E×B
v⊥ = . (2.23)
B2
This velocity perpendicular to the magnetic field is called E ×B-drift. If we look at E ×
B-rotation – i.e the velocity component of v ⊥ that is tangential to the flux surface and thus
perpendicular to the radial direction – it means that a radial electric field will contribute to
the perpendicular rotation. Vice versa it can be said that the perpendicular part of plasma
rotation induces a radial electric field. Because Bφ /Bθ 1, the perpendicular rotation is
mainly in the poloidal direction (see figure 2.3). In the plasma centre this poloidal rotation
is strongly damped (see later), which means that, in presence of an electric field, a strong
parallel velocity exists so that the poloidal component of the parallel velocity compensates
the poloidal component of the E × B-rotation. One finds:
Er vk Bφ
vθ = 0, vφ = and = (2.24)
Bθ v⊥ Bθ

2.3 Momentum transport

In this section we have a closer look into the force balance equation (2.9), that describes
the evolution and distribution of rotation. One could do this by evaluating the parallel and
perpendicular rotation velocities, or by looking into the toroidal and poloidal components.
The latter option will be taken here. In fact we will just discuss the toroidal rotation.
The reason for neglecting poloidal rotation will be given in section 2.4.2, where it will be
shown that the poloidal rotation in a tokamak is strongly damped.
2.3.1 Momentum confinement time

As said previously, equation (2.9) is Newton’s law. Considering only the toroidal compo-
nent and a constant force it can be rewritten as:
dρvφ
= Fφ (x). (2.25)
dt
Integrated over the whole plasma this becomes:
Z dρv Z
φ
dV = Fφ (x) dV
plasma dt plasma
Z a Z a
3 3
d (mnΩφ R) Rdr = Fφ (r)Rdr dt
0 0
Z Lφ Z τφ
dL0φ = Tφ dt
0 0
Lφ
τφ = , (2.26)
Tφ
where Lφ is the total angular momentum of the plasma, Tφ is the total input torque and
τφ is the momentum confinement time [46].
Equation (2.26) basically states that the momentum confinement time τφ is the time
necessary to reach a certain angular momentum Lφ , given a certain input torque Tφ .
Consequently it is also the time in which the total momentum is lost when the input
torque is turned off. It is therefore a characteristic time for (radial) momentum transport.
Momentum transport is described by the momentum diffusion coefficient Dφ 1 . When
the plasma radius is a, the average momentum diffusion coefficient is given by Dφ ∝ a2 /τφ .
The definition of τφ is very similar to that of the energy confinement time τE =

3 3
R
2 nT dr /P ,
P being the total input power [88]. Empirically it has been observed that the
relationship between τφ and τE is actually stronger than just a similarity. Measurements
of τφ and τE done in several tokamaks show that increasing τE corresponds with increasing
τφ (see e.g. figure 2.4 and [46]). When comparing the ion energy confinement time τEion
– i.e. the energy transport in the ion channel – with the momentum confinement time τφ
the agreement is even better [19, 86].
Apart from the experimental evidence that τφ and τEion are related, several theoretical
models, as e.g. described in [60], predict the ion thermal diffusivity and the momentum
diffusion to be equal; χi = Dφ .
1
In the literature both Dφ and χφ are used for denoting the momentum diffusion coefficient
Section 2.3 - Momentum transport 19

Figure 2.4 : Momentum confinement time τφ

plotted against energy confinement time τE
for different plasma conditions. Measure-
ments were done in L-mode, neutral beam
heated shots in ASDEX, during steady state
phases.[46]
Many experiments on machines throughout the world have shown that, both in
L-mode and in H-mode regimes, the energy confinement time τE reduces with increasing
power input. Scaling laws of τE describe the relation between τE and the input power
P as τE ∝ P −α , α = 0.5 − 0.73 [67]. Given the fact that the momentum confinement τφ
is observed to be scaling with the ion energy confinement time τEion , we can state that
the momentum confinement τφ also reduces with increasing power input [19, 18]. When
investigating the plasma rotation in discharges with varied power input this dependence
of τφ on the power input has to be taken into account.
Exceptions on the tight relation between τφ and τE are found at low densities, in
transient phases (when e.g. extra momentum sources like neutral beams are suddenly
switched on or off) and towards the plasma edge [46, 86, 91]. Also for plasmas rotating
in the direction counter to the plasma current the energy confinement time is observed to
be lower than the momentum confinement time [4, 90].
2.3.2 Momentum sources and sinks

The right hand side of equation (2.9) shows three forces that drive the plasma velocity v:
j × B, −∇p and −∇ · Π. The ∇p term is a force perpendicular to the flux surfaces and
hence has no contribution to the plasma rotation. The two other force terms do have an
influence on plasma rotation.
The divergence of the anisotropic part Π of the pressure tensor – also known as the
viscous stress tensor – has components in the toroidal and poloidal direction. As said
in subsection 2.2.1 the different elements of Π depend – among other things – on the
gradients of the velocity. This means that the plasma rotation at one position causes a
driving force at an other position in the plasma. In other words: the term −∇ · Π is
responsible for momentum transport. In a steady state situation it determines the shape
of the rotation profile. It will be discussed in more detail in subsection 2.3.3.
j × B is a force perpendicular to the magnetic field. There are two mechanisms to
apply a j × B-force to the plasma. A first method is creating a radial current jr . This
will result in both a toroidal force jr Bθ and a poloidal force −jr Bφ . A second possibility
is to perturb the magnetic field such that a radial component Br is created. If a radial
field Br exists, the flux surface picture breaks down and we can no longer define plasma
20 Section 2.3 - Momentum transport

rotation as the velocity tangential to flux surface. We therefore define the plasma rotation
as being the toroidal and poloidal components of the plasma velocity. The cross-product
of this Br with the main plasma current (jφ , jθ ) results again in poloidal (jφ Br ) and
toroidal (−jθ Br ) forces.
In order to come to the force balance equation the assumption was made that the
plasma only consists of electrons and ions. This assumption is not valid in the edge of
the plasma where the temperature is not high enough to ionise all atoms, hence neutral
particles exist. Also neutral beam injectors add neutral particles to the plasma. These
neutrals cause a need for two extra force terms in equation (2.9).
The neutrals at the plasma edge are essentially a momentum sink: charge-exchange
reactions replace the fast plasma ions by slow neutral particles, which of course slows
down the plasma. The fast neutralised plasma ions lose their momentum due to collisions
with other neutrals and the wall. Without the neutral particles in the edge, the plasma
would never lose its momentum. The neutral particles introduce a force term into the force
balance equation that slows down the plasma. This force depends on the rate coefficient
for charge-exchange reactions hσvicx , the neutral density n0 , the plasma density n and the
plasma velocity v: F cx−edge = −nmhσvicx n0 v.
A second source of neutrals is provided by the neutral beams. In the previous section
it was said that the only way to drive plasma rotation was through the j × B term: a
force term perpendicular to the magnetic field. This is no longer true when tangential
neutral beams are present. The neutral particles injected by a tangential beam have a
velocity component in the toroidal direction. Ultimately these neutral particles are all
ionised, transferring their toroidal momentum to the plasma. In [91] the process of the
momentum transfer is discussed in detail. It is shown that a part of the neutral beam
force is a perpendicular j × B-force, but partly the NBI also directly drives momentum
parallel to the magnetic field. In present day devices tangential neutral beams are the
main sources of rotation. We will denote this beam force as F N BI (r).
2.3.3 Momentum balance

In a steady state situation and with the force terms due to neutral beam injection and
friction with edge neutrals included, the force balance equation (2.9) becomes:
nmv · ∇v = j × B − ∇p − ∇ · Π − nmhσvicx n0 v + F N BI (2.27)
The steady state form of the conservation of mass equation (2.8): ∇ · nmv = 0, allows
us to rewrite the left hand side of equation (2.27) as nmv · ∇v + v∇ · nmv = ∇ · (nmvv).
This leads to the general momentum balance equation:
∇ · (nmvv) + ∇ · Π = j × B − ∇p − nmhσvicx n0 v + F N BI (2.28)
The right hand side of the momentum balance equation (2.28) contains momentum
sources and sinks. The left hand side contains the divergence of two tensors that describe
inertia (nmvv) and viscosity (Π); they account for momentum transport.
The viscous stress tensor Π depends on viscosity coefficients η and on the rate-of-strain
tensor W [80, 88, 9]. The rate-of-strain tensor W comes from conventional fluid theory
and has following definition:
Section 2.3 - Momentum transport 21

∂vα ∂vβ 2
W αβ = + − δαβ ∇ · v (2.29)
∂xβ ∂xα 3
The earlier mentioned dependence of Π on the velocity gradients comes from the con-
tribution of W to Π. The viscosity coefficients η provide the link between the stresses in
Π and velocity gradients in W . The viscosity coefficients η rely on the underlying phys-
ical transport mechanisms. Both the neoclassical transport model (see section 2.4) and
anomalous, turbulent transport models can be used to derive these viscosity coefficients
[15, 60, 79].
For an orthogonal coordinate system (x, y, z) with the z-coordinate along the magnetic
field the relation between Π, η and W is as follows [88]:
1 1
Πxx = − η0 (Wxx + Wyy ) − η1 (Wxx − Wyy ) − η3 Wxy
2 2
1 1
Πyy = − η0 (Wxx + Wyy ) − η1 (Wyy − Wxx ) + η3 Wxy
2 2
Πzz = −η0 Wzz
1
Πxy = Πyx = −η1 Wxy + η3 (Wxx − Wyy )
2
Πxz = Πzx = −η2 Wxz − η4 Wyz
Πyz = Πzy = −η2 Wyz + η4 Wxz , (2.30)
where η0 is a parallel viscosity term, η1 and η2 describe perpendicular viscosity and η3

and η4 represent gyro-viscosity.
We now consider the toroidal component of equation (2.28) and take the average over
the flux surfaces [74]:
D E D E D E D E
(∇ · nmvv)φ + (∇ · Π)φ = (j × B)φ − hnmhσvicx n0 vφ i + (F N BI )φ , (2.31)
1
R 2π
where the definition of the flux-surface average on circular surfaces is hAi = 2π 0 Adθ.
The expression for the toroidal component of the divergence of a tensor in large aspect
ratio, circular flux-coordinates is given by [80]:
1 ∂ 1 ∂Tθφ 1 ∂Tφφ
(∇ · T )φ = (rTrφ ) + + (2.32)
r ∂r r ∂θ R0 ∂φ
∂T ∂T
Due to axi-symmetry ∂φφφ = 0. If the assumption ∂θθφ = 0 is made, and flux surface
averaging is applied, the terms on the left hand side of equation (2.31) become:
D E 1 ∂
(∇ · Π)φ = (rΠrφ ) (2.33)
r ∂r
D E 1 ∂
(∇ · nmvv)φ = (rnmvr vφ ) (2.34)
r ∂r
We first have a look at equation (2.33). The rφ-component of Π is [80]:
∂ η4 R ∂
vφ R−1 − vφ R−1

Πrφ = −η2 R
∂r r ∂θ
∂Ωφ
= −η2 R0 , (2.35)
∂r
22 Section 2.3 - Momentum transport

∂v R−1 ∂Ω
where large aspect ratio and φ∂θ = ∂θφ = 0 were assumed. Neglecting convective
momentum transport, the perpendicular viscosity η2 can be defined as η2 ≡ nmDφ , Dφ
being the momentum diffusion coefficient.
The first term on the left hand side of equation (2.31) is given by (2.34). The nmvr
in (2.34) is a particle flux Γ = nmvr . Assuming only anomalous, diffusive transport the
following expression is valid: Γ = −mDp ∂n ∂r , Dp being an anomalous particle diffusion
coefficient.
We now assume, as was done in [74], that Dφ = Dp = D. When we use the above
assumptions and the equations (2.33) - (2.34) , the left hand side of equation (2.31)
becomes:

D E D E 1 ∂ ∂Ωφ
(∇ · nmvv)φ + (∇ · Π)φ = rΓvφ − rη2 R0
r ∂r ∂r

1 ∂ ∂n ∂Ωφ
= r(−mD )R0 Ωφ − r(nmD)R0
r ∂r ∂r ∂r

R0 ∂ ∂n ∂Ωφ
= − rmD Ωφ + n
r ∂r ∂r ∂r

R0 ∂ ∂
= − rD (nmΩφ ) (2.36)
r ∂r ∂r
This leads to the following angular momentum balance equation:

R0 ∂ ∂
− rD (nmΩφ ) = jr Bθ − jθ Br − nmhσvicx n0 vφ + FN BI , (2.37)
r ∂r ∂r
where we assumed all terms on the right hand side to be independent of θ. When all forces
are known, this angular momentum balance equation allows us to calculate the toroidal
rotation profile.
2.4 Neoclassical transport and rotation

The viscosity coefficients η and the momentum diffusion coefficient Dφ depend on the phys-
ical mechanism of momentum transport. There are several ways to describe the transport
of momentum, energy and particles: classical, neoclassical and turbulent. The classical
model describes the transport caused by collisions in a cylindrical plasma. The plasma
in a tokamak has however a toroidal shape. When we take the torus shape into account,
we find that the collision driven transport is higher than in the classical, cylindrical case.
The collision driven transport in a torus shaped, magnetically confined plasma is called
the neoclassical transport. Apart from collisions, also turbulence can drive transport.
The neoclassical model thus gives us the lower limit of particle, energy and momentum
transport: the transport in absence of turbulence.
In this chapter we will discuss the neoclassical transport and especially focus on the
neoclassical momentum transport. From neoclassical theory it follows that the poloidal
velocity is strongly damped. The mechanism of poloidal flow damping is often referred to
as parallel viscosity. In section 2.3 only the toroidal momentum transport was discussed.
The reason for neglecting poloidal momentum transport is this poloidal flow damping: due
to poloidal flow damping the effect of a poloidal force on the poloidal rotation is small and
Section 2.4 - Neoclassical transport and rotation 23

strongly localised to the position where the force is applied. There is virtually no poloidal
momentum transport.
Another result from neoclassical transport is the existence of a spontaneous rotation in
a tokamak plasma, that will be discussed in section 2.4.3. As said in the chapter 1 a fast
rotating plasma has a better confinement and stability. In devices where the momentum
input by e.g. neutral beams is limited, spontaneous rotation is therefore very important.
The mechanisms of poloidal flow damping and the neoclassical expressions for sponta-
neous plasma rotation will be given in the following sections.
2.4.1 Neoclassical theory

Classical transport is governed solely by collisions. It is a diffusion process with Dα ∝
ρ2L,α /τc,α the diffusion coefficient for particles α. Here τc,α is the collision time and the
step size ρL,α is the Larmor radius. Classical transport is only valid when the magnetic
field inside the plasma is straight, homogeneous and stationary.
Due to the toroidal shape of a tokamak, the magnetic field is not homogeneous: it is
stronger at the inner side of the torus than at the outer side. The curvature and gra-
dient of the field leads to extra forces and drifts that are not present in e.g. cylindrical
configurations. The classical model can therefore not describe the transport of particles,
energy and momentum in a tokamak. Neoclassical theory takes into account this inhomo-
geneous, curved field, which results in the correct description of the collisional transport
in a tokamak [2].
Figure 2.5 : Particles with a total velocity uα bounce back at a position Rmin = 2µB 0 R0
mα u2α . The
orbit they follow is banana shaped due to drifts; here the poloidal projection is shown. Parti-
cles with a small pitch angle (u⊥ /uk 1) follow their trajectory through the high-field side
and are called “passing”. Due to drifts the orbits do not lie exactly on the magnetic flux
surface, but are slightly shifted.
One of the most visible aspects of neoclassical theory are so-called trapped particles.
Charged particles that travel, parallel to the magnetic field, from the outer side of the
torus to the inner side, go from a region with lower magnetic field to a region with higher
magnetic field. In other words: these particles see a ∇B and experience a force F k =
−µ∇k B, where µ = mα u2⊥ /2B is the magnetic moment which is a constant of motion.
This force will slow down the parallel velocity of the particles when they are moving
24 Section 2.4 - Neoclassical transport and rotation

towards the high-field side of the tokamak. Particles with a low enough uk will be stopped
and reflected before reaching the high-field side of the tokamak: they oscillate in a so-
called banana orbit (see fig. 2.5) and are called trapped particles. Particles that are fast
enough do not bounce back at the high-field side, but continue their trajectory parallel
to the field line; they are called passing particles. Due to drifts the orbits of passing and
trapped particles are shifted with respect to the magnetic field lines they are connected
to. For trapped particles this means they have a certain banana width wB . Below three
important parameters for trapped particles are given:
• the width of a banana orbit wB = ε−1/2 ρL q

√
• the density of trapped particles ntrapped = ntotal ε
√
• the bounce frequency ωb = vth ε/qR◦ ,
where = a/R0 is the inverse aspect-ratio of the plasma. For large aspect-ratio tokamaks
like TEXTOR < 1.
The existence of trapped particles and the banana orbits have a significant influence on
the transport. Let us, as an example, look at collisional transport of particles. If a trapped
particle undergoes collisions in the time it needs to complete its banana orbit – defined by
ωb – it does not know it is trapped and the transport will be close to classical, with a step
size ρL . When the collision frequency is lower than the bounce frequency, the trapped
particle can at least complete one banana orbit without colliding. The step that trapped
particles then make in a collision is no longer the Larmor radius ρL , but the much larger
banana width wB . This results in a completely different diffusion coefficient. In a tokamak
one defines three different transport regimes: the low collisionality banana regime, the high
collisionality Pfirsch-Schlüter regime and the intermediate plateau regime. All three have
different transport coefficients. In figure 2.6 the diffusion coefficient is drawn as a function
of the collisionality. The three transport regimes are easily recognised.
D
te r
- S chlü
sch
Plateau Pfir
na
na
Ba
a l
Classic
ν*
ε3/2 1
Figure 2.6 : The variation of the diffusion coefficient with the collisionality throughout the three
neoclassical transport regimes. As a comparison also the classical diffusion coefficient is
drawn. One sees that neoclassical theory predicts a larger transport than classical theory.
Also for investigating rotation trapped particles are important. Trapped particles
bounce back and forward, so they never make a poloidal turn. Their poloidal momentum

is therefore zero. Toroidally trapped particles do have a rotation, because the banana
orbits can have a precession around the torus. The poloidal flow damping that follows
from neoclassical theory is discussed in the next section.
2.4.2 Poloidal flow damping

The reason for poloidal rotation damping is the toroidal shape of a tokamak. This
causes an inhomogeneous magnetic field, stronger at the inside of the torus, weaker at
the outside. A toroidally rotating plasma in its rest frame does not see a change in the
magnetic field. However a poloidally rotating plasma does see a time-varying field when
moving from the outside to the inside of the torus. Through the mechanism of magnetic
pumping the energy of a time-varying field will be transferred to the plasma [6, 81]. In
case of poloidal movement through a spatial periodic field, the energy of the time-varying
field in the rest frame of the plasma is the kinetic energy of the rotating plasma. In other
words: the poloidal rotation decays, because the kinetic energy of the poloidal rotation is
used to heat the plasma.
The magnetic pumping can be split in a collisional and a collision-free part. The
collisional scheme is applied when the collision time is smaller than the period of the
perturbation. The magnetic moment of the particle is µ = mα u2⊥ /2B. As long as particle
does not collide this is a constant of motion. Hence, when the particle moves to the
high-field side, B increases and the particle will gain perpendicular energy mα u2⊥ /2. Due
to the conservation of energy, the parallel velocity uk will reduce. When returning to
the low-field side the particle will again lose its extra perpendicular energy and the net
change in perpendicular energy over one period is zero. Also uk will increase again when
moving to the low field side, so also the net change in uk over one period is zero. However,
when a collision occurs when the particle is at the high-field side, it will redistribute its
extra energy, effectively heating the plasma. This means that, after the collision, µ is
lower than before the collision. The force F k = −µ∇k B, that has to accelerate uk when
the particle moves to the low field side, will therefore also be lower and there will be a
net reduction of uk over one period. [6]
Also in a collision-free situation, where the collision time is larger than the perturbation
period, the poloidal rotation is damped. This is due to the trapped particles: those
particles for which uk gets zero before the high-field side is reached. Because trapped
particles do not make a full poloidal turn they do not contribute to the poloidal momentum.
This form of poloidal flow damping is usually called transit-time magnetic pumping. Seen
from within the frame of a single particle, this process is identical to the Landau damping
of an electromagnetic wave.
One could think that the toroidal momentum of trapped particles is also zero, but
this is not the case because the banana orbits of the trapped particles have a toroidal
precession around the torus. Collisions between trapped and passing particles will further
reduce the poloidal flow (the collisional part of poloidal flow damping).
A more rigourous derivation of the decay of poloidal rotation is given in [81]. Using a
drift-kinetic equation for ions it describes the cross-B currents due to the magnetic pertur-
bation and viscous drag. These currents are then responsible for j × B-forces that brake
the poloidal rotation. The above damping occurs over the whole plasma volume. Depend-
ing on local density and temperature more transit-time or more collisional damping occurs.

In the edge of the plasma – at low temperature – high electron-ion collisionality leads to
strong electric fields and a non-zero poloidal (and toroidal) velocity. This spontaneous
rotation in an ohmic plasma2 will be discussed in subsection 2.4.3.
In general it can be said that the poloidal rotation is strongly damped. This means
that any externally induced poloidal rotation decays towards the neoclassical value3 within
a decay time which is in the order of the ion-ion collision time.
The effect of poloidal flow damping is included in the viscous term h(∇ · Π)θ i of the
poloidal force balance. This viscous term, taking into account poloidal damping, has
following form [11, 71]:
h(∇ · Π)θ i = α(vθ − vθneo ), (2.38)

where vθneo
is the neoclassical rotation (see section 2.4.3) and α is a factor that depends on
the parallel viscosity – in contrast to the momentum diffusion coefficient D in the toroidal
balance equation (2.37) that depended on perpendicular viscosity. Poloidal flow damping
is therefore often referred to as being the result of strong parallel viscosity. The exact
values of vθneo and α depend on the transport regime: banana, plateau or Pfirsch-Schlüter
regime.
The poloidal force balance equation, where inertia has been neglected, becomes [11, 71]:
α(vθ − vθneo ) = −jr,⊥ Bφ − (1 + q 2 )nmνcx vθ . (2.39)
In contrast to the the toroidal balance equation (2.37) this equation has no radial derivative
of the poloidal velocity. This means that a poloidal velocity at one radial position does
not affect the poloidal rotation at another position. The effect of a poloidal force on the
poloidal rotation is strongly localised to the position where the force is applied. Because
no poloidal force terms are present in the plasma core and because the neoclassical poloidal
velocity is low, the poloidal velocity can be neglected in the plasma core.
2.4.3 Spontaneous neoclassical rotation

If no external momentum is provided (e.g. by neutral beams), plasma rotation is still
possible due to density gradients, temperature gradients and local, radial electric fields.
One could assume that the parallel electric field, that exerts a parallel force Fk = eEk on
the ions, is also a possible source of momentum. However, due to the higher mass of the
ions compared with the electrons, Ek will predominantly accelerate the electrons, giving
rise to current rather than rotation.
For the derivation of this ohmic rotation we must depart from MHD one fluid model
and return to the description of the ion fluid. The ion fluid velocity that follows will be
approximately equal to the fluid velocity in the MHD model. First, in order to get a
physical picture of the mechanisms at work, we will consider the mass and momentum
conservation equations (2.3) and (2.4) for the ion fluid. In a steady state situation, with
subsonic velocities, no momentum exchange with other species (Rie ≈ 0) and a isotropic
pressure (P i ≈ pi I), these can be reduced to [13]:
ni ∇ · v i = 0 (2.40)
∇pi = eni (E + v i × B) (2.41)
2
in contrast to beam or RF heated plasmas, as beams and RF waves also supply extra momentum input
3
see section 2.4.3

Taking the cross product of equation (2.41) with B a perpendicular velocity is found:
E×B 1 ∇pi × B
v⊥
i = − (2.42)
B2 eni B2
k
Through equation (2.40) the parallel velocity v i is also partly determined. v ⊥
i is the
⊥ 1 F ×B
drift velocity vi = eni B 2 , due to the forces eni E and −∇pi . The first term on the right
hand side of (2.42) is the E × B drift that was discussed earlier. The second term in (2.42)
is, for the one fluid MHD picture, hidden in the diamagnetic current:
∇pi × B ∇pe × B
ni ev ∗i − ne ev ∗e = − −
B2 B2
∇p × B
j∗ = − (2.43)
B2
The above derivations indicate that electrical fields and pressure gradients can cause
both poloidal and toroidal rotation. In [14, 37] neoclassical theory is used to derive a
natural neoclassical poloidal velocity vθneo and the derivative of the neoclassical toroidal
velocity vφneo :
K1 ∂Ti
vθneo = (2.44)
eBφ ∂r
∂vφneo
2
K1 Ti q 2

∂ ln Ti
= 0.107 , (2.45)
∂r eBθ ∂r
where K1 depends on the collisionality: K1 = 1.17 in the banana regime, K1 = −0.5 in
the plateau regime and K1 = −1.83 in the Pfirsch-Schlüter regime according to [14, 37].
Also in [49] neoclassical expressions for the poloidal and toroidal velocities are derived.
The expression for vθneo is the same as the one derived in [14, 37], but the K1 parameter
given in [49] goes from 0.5 in the banana regime down to approximately −2 in the Pfirsch-
Schlüter regime, where the exact values depend on the amount of impurities in the plasma.
In [14] only the derivative of the toroidal plasma velocity is given (see (2.45)), whereas in
[49] a expression for the local toroidal velocity is given. The expression for the toroidal
velocity in [49] needs, apart from density and temperature measurements, an extra input:
the radial electric field Er . This Er can e.g. be the result of momentum transport from
spontaneous rotation at other positions in the plasma. The expression for vφneo in the
banana regime is:
neo Er Ti 1 dni 1 − K1 dTi
vφ = − + . (2.46)
Bθ eBθ ni dr Ti dr
Also in [49] the effect of the parallel electric field Ek is considered. As expected, the
parallel electric field has little influence on the rotation of the bulk ions, but, surprisingly,
impurity ions will experience a rotation in the counter-current direction due to the
parallel electric field.
Because plasma rotation improves the confinement and stability of a plasma, having
spontaneous rotation is very beneficial. The above shows that a spontaneous rotation is
expected in tokamaks. However, even in a well-established model like the neoclassical
transport model, the actual values of spontaneous rotation can differ depending on the

assumptions made when evaluating the model. Moreover the models used to derive the
above expressions are far from complete; the friction forces due to neutrals were for ex-
ample not included. Also turbulent transport was not taken into account. Due to the
complexity of the subject there exists a wide range of models that yield an equally wide
range of predictions for the spontaneous rotation.
2.5 Impurity and MHD Rotation

The MHD equations allow us to describe the dynamics of a plasma by regarding it as a
single conducting fluid. Although this picture gives us physical insight, the physical truth
is that a plasma consists of many different species, that each have a fluid like behaviour.
When a plasma is diagnosed, the properties of one of the plasma species are measured,
instead of the properties of the single plasma fluid. It is therefore necessary to know the
relationship between the plasma velocity v used in the MHD equations, and the fluid
velocities of the main ions v i , the electrons v e and the impurities v I .
As said in section 2.2.1 the plasma fluid velocity is the centre of mass velocity of all
fluids: P
mi ni v i + me ne v e + I mI nI v I
v= P . (2.47)
ni mi + ne me + I mI nI
If we assume that the impurity density is low nI ni ≈ ne and take into account the
mass difference between ions and electrons me mi , we can approximate the plasma fluid
velocity by the main ion velocity: v ≈ v i .
Measurements of plasma rotation are mostly done through spectroscopic analysis of
the emission of impurity ions. This means the v I is measured. Other diagnostics – like
e.g. electron cyclotron emission (ECE) or reflectometry – measure fluctuations in electron
temperature or density. These fluctuations are usually related to the velocity v e of the
electron fluid. In the following sections we will discuss how measured impurity velocities v I
can be translated into the plasma velocity v and how the frequency of electron temperature
or density fluctuations can be linked to v.
2.5.1 Impurity rotation

The collision frequency between impurities and the main ions is usually high enough to
assume that the temperature of the impurities and the main ions is equal, within an energy
confinement time: Ti = TI . However the fluid velocity of the ions is not necessarily the
same as that of the main ions. When there is a difference in pressure gradient ∇p/ZI ,
ZI being the nuclear charge of the impurity ions, then a differential rotation between
the impurities and main ions exists. This occurs typically in a situation with continuous
momentum input, e.g. by neutral beams.
The neoclassical moment approach of Hirshman and Sigmar [40, 49] provides an ex-
pression for the differential rotation between the main ions and one impurity. In [49, 82]
this difference in rotation ∆v = v i − v I is given by:

1 dT LT Zi Lp,i
∆vθ ≈ −3K2 + 2 1−
2eBφ dr Lp,i ZI Lp,I
3 dT
∆vφ ≈ − K2 , (2.48)
2eBθ dr
Section 2.5 - Impurity and MHD Rotation 29

where we have neglected the θ-dependence. LT is the temperature gradient length and Lp,i
and Lp,I are the pressure gradient lengths of the main ions and impurity ions respectively.
K2 is a function depending on the collisionality.
One sees that the difference in poloidal rotation depends on the impurity pressure
and main ion pressure, whereas the difference in toroidal rotation only depends on the
temperature gradient. One also notices that the difference in toroidal rotation will be
significant in discharges with a strongly peaked temperature profile (dT /dr 0) and a
low plasma current (1/Bθ 0). The validity of the above equations (2.48) has been tested
on several machines [5, 82].
2.5.2 MHD rotation

The difference between electron and ion velocities, where we neglect impurities, is ex-
pressed by the current in a plasma: j = ne(v i − v e ). If we measure the current density
and the electron fluid velocity, then also the main ion velocity is known.
The main part of the current is parallel to the magnetic field, but equation (2.43) shows
that a pressure gradient results in a diamagnetic current, and thus the difference between
the electron and ion fluid velocity also has a component perpendicular to the magnetic
field.
It is quite difficult to directly measure the electron fluid velocity, although it can be
done with tangential Thomson scattering [47]. Diagnostics like electron cyclotron emission
(ECE) or reflectometry are, however, capable of measuring fluctuations in electron den-
sity are temperature. The frequency of these fluctuations can be translated into plasma
velocity.
These fluctuations in temperature and density occur when so-called MHD modes are
present within the plasma. As said in subsection 2.2.2 and shown in figure 2.1, the magnetic
topology in a confined plasma consist of a set of nested flux surfaces. Under certain
conditions flux surfaces can break up and reconnect, thus changing the magnetic topology.
This can be seen in figure 2.7. These reconnected flux surfaces are a specific class of MHD
modes, called tearing modes or magnetic islands. Magnetic islands occur at flux surfaces
with a rational q = m/n, and have therefore a periodicity given by the poloidal and
toroidal mode numbers m and n.
Figure 2.7 : At rational q-surfaces the nested flux surfaces can break up. Here m/n = 2/1 (dark
grey) and m/n = 1/1 (black) islands are shown at the q=2 and q=1 surfaces.
30 Section 2.5 - Impurity and MHD Rotation

Inside an island the electron density and temperature is usually flat or slightly peaked.
The velocity of the electrons perpendicular to the magnetic field is restricted: they gyrate
around the field. This means that, when a magnetic island rotates the electrons rotate
along. As a result the flat electron temperature and density periodically appears in the
measuring volume of the ECE or reflectometry, which is seen as a fluctuation. The fre-
quency of the fluctuation depends on the island rotation velocity and the poloidal and
toroidal mode numbers m and n.
In order to get an expression for the rotation of these magnetic perturbations, an
inverse approach is used: looking a the change of magnetic flux through a surface S, with
contour ∂S, moving through the plasma with a velocity v MHD .
Z I
dΦ ∂B
= · dS − (v MHD × B) · dl (2.49)
dt S ∂t ∂S
In an ideal flux conserving plasma dΦ ∂B

dt is zero. Using Faraday’s law ∂t = −∇ × E and
the generalised Ohm’s law (2.10) E + v × B = ne 1
(j × B − ∇pe ) – where the terms − mee ∂j
∂t ,
∇ · Πe and Re were neglected – equation (2.49) can be rewritten as [85]:
I
dΦ
= (v + v ∗e − v MHD ) × B · dl = 0, (2.50)
dt ∂S
∗
je 1 ∇pe ×B
where v ∗e = − ne = en B2
is the electron diamagnetic drift. Equation (2.50) implies
that dΦ
dt = 0 if the MHD mode moves with a perpendicular velocity v MHD = v ⊥ + v ∗e .
Measuring the MHD velocity is rather complicated, one usually measures the MHD
frequency, i.e. the frequency of the fluctuations on the ECE or reflectometry measure-
ments. When the poloidal and toroidal mode numbers m and n are known, however, one
finds the following relation between plasma velocity, electron diamagnetic drift and MHD
frequency:
nvφ,MHD mvθ,MHD
fMHD = +
2πR0 2πr
∗
mvθ,e ∗
nvφ,e ∗
mvθ,e
nvφ mvθ
fMHD ≈ + + , because:
2πR0 2πr 2πr 2πR0 2πr
nvφ mvθ
fMHD ≈ + + fe∗ (2.51)
2πR0 2πr
The electron diamagnetic frequency is given by:
m Bφ dpe m 1 dpe
fe∗ = 2
≈ . (2.52)
2πr ne eB dr 2πr ne eBφ dr
Because the pressure gradient and local density are often not very well diagnosed, the
so-called natural profile shape of density and pressure can be used [34, 76]. The natural
density and pressure profiles are given by ne (r) = ne (0).f (r) and pe = ne (0).Te (0).f 3 (r),
with:
2 −2/3
qa r
f (r) = 1 + −1 . 2 , (2.53)
q0 a
where a is the plasma radius, q0 the value of q at the magnetic axis, qa the value of q at
the plasma edge and the q-profile is assumed to be quadratic. The resulting diamagnetic
Section 2.5 - Impurity and MHD Rotation 31

frequency at a specific q-surface is then given by:

−7/3
2m Te (0)[keV] qa q
fe∗ = −1 [kHz] (2.54)
π Bφ [T] a2 [m] q0 q0
When q0 is also unknown, it can be approximated by q0 = qa /(qa + 1) for a sawtoothing

plasma [34, 76].
Discussions on the influence of vθ on equation (2.51) indicated that the poloidal term
can be neglected due to strong poloidal flow damping and the difficulty of relating non-
uniform poloidal velocity to a poloidal frequency [92]. So equation (2.51) can be simplified
to:
vφ
fMHD = n + fe∗ . (2.55)
2πR0
When a magnetic island is present in the plasma, and its toroidal mode number n is
known, then the toroidal plasma velocity vφ can be derived from the diamagnetic frequency
fe∗ and the measured MHD frequency fMHD .
2.6 Conclusion
In this chapter we followed the derivation of the single fluid MHD equations as it is
described in the literature. The toroidal and poloidal components of the fluid velocity of
the plasma were defined as plasma rotation.
The toroidal momentum balance in steady state – equation (2.37) – allows us to calcu-
late the toroidal rotation profile when the toroidal forces are known. The radial derivatives
in (2.37) indicate that a toroidal force at one position will have an influence over the whole
plasma. How strong that influence is and how far it goes depends on the level of toroidal
momentum transport. The momentum transport is governed by the diffusion coefficient
Dφ . The averaged momentum diffusion coefficient is given by Dφ ∝ a2 /τφ , where a is
the plasma radius and τφ the momentum confinement time. In this thesis it is assumed
that the momentum diffusion coefficient Dφ equals the particle diffusion coefficient D. It
is also often observed that the ion energy confinement time τEion equals the momentum
confinement time τφ .
Poloidally the rotation is strongly damped. This is a neoclassical effect; trapped par-
ticles can not make a poloidal turn and therefore do not contribute to the poloidal mo-
mentum. Collisions of passing with trapped particles will further slow down the poloidal
rotation. The result of this poloidal flow damping is the poloidal momentum balance given
in equation (2.39). This equation shows that the poloidal rotation will only change at the
position where a poloidal force is applied. There is virtually no radial transport of poloidal
momentum. Because there are no strong poloidal forces present in the core of the plasma,
the poloidal rotation velocity can assumed to be zero in the plasma core.
Also without external momentum input a plasma will rotate. The neoclassical ex-
pressions for spontaneous poloidal and toroidal rotation are given in section 2.4.3. It is
generally believed that for a good description of the spontaneous rotation, the neoclassical
theory does not suffice. More advanced transport models including turbulence, however,
give a wide range of predictions on spontaneous rotation.
Diagnostics do not see the plasma as the single fluid that MHD equations describe;
they measure the properties of the different species in the plasma. If we want to compare
the measurements with the MHD theory, it is necessary to convert the fluid velocity of the
32 Section 2.6 - Conclusion

measured species to the MHD fluid velocity. In section 2.5.1 the link between impurity
rotation and main ion rotation is given. In section 2.5.2 the relationship between MHD
frequency and toroidal plasma rotation is derived.
In this chapter we have summarised the theoretical tools to deal with plasma rotation.
Section 2.6 - Conclusion 33

Chapter 3
TEXTOR, DED and diagnostics
3.1 Introduction
In order to investigate the influence that plasma rotation has on the excitation of modes by
an external perturbation field in a tokamak, we need a tokamak, a way to control the rota-
tion and a way to control the perturbation field. The TEXTOR1 tokamak provides us with
all three. TEXTOR is a circular, medium-sized tokamak (R0 = 1.75 m , a = 0.47 m). It
has two tangential neutral beam injectors (NBI), injecting neutral particles in two opposite
directions. By balancing the two NBI’s, toroidal rotation profiles can be established going
from full rotation in the direction of the plasma current, via a plasma with no rotation,
to full ‘counter-rotation’. Furthermore, TEXTOR is equipped with a set of helical per-
turbation coils at the high-field side, called the Dynamic Ergodic Divertor (DED). These
coils can be used to set up a static or dynamic perturbation field. In a first section we will
discuss the general setup of the TEXTOR tokamak and its heating systems. Section 3.3
will introduce the properties of the DED.
We do not only wish to control the plasma rotation and the perturbation field, we
also need to measure the plasma response to the perturbation field. The most important
measurement is that of the plasma rotation. For this, charge exchange recombination
spectroscopy (CXRS) is used. Due to its importance for this thesis it is discussed in
a separate chapter (chapter 4). The diagnostics that measure plasma parameters like
temperature and density are introduced in section 3.5.
This chapter gives a short introduction to the TEXTOR tokamak, at which the work
presented in this thesis was carried out. For a more detailed overview we refer to [26] and
[73].
3.2 The TEXTOR tokamak and its heating systems

TEXTOR is a limiter tokamak. Three types of limiters can be used:
• A toroidal bumper limiter at the high-field side. This limiter that also protects about
one third of the inner wall in case of a disruption.
• A pumped, toroidal belt limiter, located near the bottom of the vacuum vessel at
the low-field side.
1
Tokamak EXperiment for Technology Oriented Research
35
Chapter 3 - TEXTOR, DED and diagnostics
• The poloidal limiters that are located at one toroidal position. These poloidal lim-
iters can be remotely moved, hence changing the plasma radius a.
During a plasma discharge of course only one of these three limiters is really limiting the
plasma. [64]
A set of 16 coils around the vacuum vessel induce the toroidal magnetic field Bφ of
TEXTOR. This field can go up to 2.9 T . The iron core transformer provides the flux
swing needed for plasma breakdown and plasma current Ip . In TEXTOR the maximum
plasma current is 800 kA, but typically the machine is operated with a plasma current
of 400 kA. The coil system of TEXTOR is completed by a set of vertical and horizontal
plasma positioning and shaping coils. A TEXTOR discharge can last up to 10 s, however
most TEXTOR discharges last about 6 s. [64]
The heating of the plasma in TEXTOR is provided by several sources. First of all, the
plasma current supplies Ohmic heating, which is typically 0.3 M W . Additional heating is
provided by ion cyclotron heating (ICRH), electron cyclotron heating (ECRH) and neutral
beam injection (NBI).
Ion Cyclotron Resonance Heating When a wave is launched into the plasma, of
which the electric field rotates with an angular frequency ωIRCH , the power of the wave
will be absorbed by ions gyrating around the magnetic field lines with a cyclotron frequency
ωi = ωIRCH . This process is called Ion Cyclotron Resonance Heating (ICRH).
In TEXTOR, two independent antenna systems are capable of each coupling 2 M W of
ICRH power into the plasma. The ICRH antennae generate waves with frequencies in the
range of 25−38 M Hz. The power can be injected continuously for up to 3 s. In TEXTOR
usually minority heating is used. In that case the ICRH power is transferred to a minority
of H ions (about 10 %) in a D plasma. Through collisions the H ions subsequently release
their energy to the plasma bulk. [50]
Electron Cyclotron Resonance Heating The principle of electron cyclotron reso-

nance heating (ECRH) is comparable to that of ICRH, where the power is in the ECRH
case absorbed by electrons. The frequencies are therefore higher – at TEXTOR there is
e.g a 140 GHz ECRH system – and the wavelength is small – 2 mm for the ECRH system
at TEXTOR. Thanks to the low wavelength the heat deposition of the ECRH power is
very well localised.
When the electron cyclotron waves are injected under a specific toroidal angle, Doppler
shift causes ECRH to preferentially heat the electrons that move in the direction of the
ECRH antenna. The heated electrons have less collisions and thus lose their momentum
slower than other electrons. As a result a net current is driven: electron cyclotron current
drive (ECCD). The possibility of current drive, and its localisation, make ECRH/ECCD
a very powerful tool for tailoring the current profile in a tokamak.
In TEXTOR the ECRH power is produced by a 800 kW gyrotron. This gyrotron
generates 140 GHz waves during a pulse of up to 10 s. The waves are in the extra-ordinary
polarisation mode (X-mode) and are absorbed in the plasma at the second harmonic of
the electron cyclotron frequency ωe . A specially designed steerable launcher allows to
deposit the ECRH power at various radial positions, by changing the injection angle in
the vertical direction.
36 Section 3.2 - The TEXTOR tokamak and its heating systems

The launcher also allows to change the toroidal injection angle, thus enabling ECCD. At
full gyrotron power and a toroidal injection angle of ±10◦ , depending on the temperature
and density up to 50 kA of current can be driven, either co or counter to the inductive
plasma current Ip . [89]
Neutral Beam Injection The two Neutral Beam Injectors (NBI) form a third heating
source at the TEXTOR tokamak. A NBI generates a beam of highly energetic neutral
particles, that are injected into the plasma. In the plasma the neutral beam particles are
ionised through charge exchange reactions with the plasma ions. The fast ionised beam
particles transfer their energy and their momentum to the bulk ions and electrons through
collisions. Whether mostly the ions or the electrons are heated depends on the ratio of
the beam energy and the electron temperature.
When a NBI is directed tangentially to the magnetic axis of a tokamak, it will not
only heat the plasma, but it will also supply an net toroidal momentum input. Tangential
NBI’s are commonly used to create fast rotating plasmas. Because the momentum transfer
to ions and electrons is usually not equal, tangential neutral beams drive current as well.
NBI 2 (counter-direction)
Bφ
) n
tio
Ip
rec
-di
(co
I1
NB
Figure 3.1 : Top view of TEXTOR tokamak. The two neutral beams and the usual
direction of the plasma current Ip and the toroidal magnetic field Bφ are indicated.
At TEXTOR two tangential heating beams are installed. One (NBI1) injects its neutral
particles in the direction of the plasma current Ip , the other (NBI2) injects its particles
Section 3.2 - The TEXTOR tokamak and its heating systems 37

counter to the plasma current. Figure 3.1 gives a top view of TEXTOR with the two
NBI’s and the typical direction of plasma current and toroidal magnetic field indicated.
The type of particles that can be injected are H, D and He – the usual injection species
is H. Each of the NBI’s launch a maximum of 1.5 M W into the plasma at a maximum
energy of 55 keV /amu. The duration of the beam pulses can be as long as 10 s.
To regulate the power of the neutral beams, V-targets are used. This V-target consists
of two plates in a V-shaped configuration at the end of the injector. By bringing the plates
closer together the aperture is reduced and the beam profile is partially scraped off. This
means the power of both neutral beams can be tuned by changing the V-target opening.
[83]
In table 3.1 the TEXTOR machine parameters are summarised, together with some
typical plasma parameters.
Table 3.1 : TEXTOR machine and typical plasma parameters
Major radius (R0 ) 1.75 m

Plasma radius (a) 0.47 m
Plasma volume 7.0 m3
Magnetic field (B) 1.1 − 2.9 T (typical 2.25 T )
Plasma current (Ip ) 200 − 800 kA (typical 400 kA)
Pulse length < 10 s (typical 5 s)
Ohmic power 0.3 − 0.5 M W
Neutral beams 2 tangential beams
NBI1 1.5 M W , co-current direction
NBI2 1.5 M W , counter-current direction
ICRH 4.0 M W , 25 − 38 M Hz
ECRH 0.8 M W , 140 GHz
ECCD < 50 kA
Electron temperature (Te ) 1 keV
Ion temperature (Ti ) 1 keV
Electron density (ne ) 3 × 1019 m−3
Toroidal rotation frequency (Ωφ ) −6 − +6 × 104 rad/s
38 Section 3.2 - The TEXTOR tokamak and its heating systems

3.3 The Dynamic Ergodic Divertor

A poloidal divertor – called the Dynamic Ergodic Divertor or DED – is installed at TEX-
TOR [26]. This DED consist of a set of helical perturbation coils located at the high-field
side of the torus, behind the bumper limiter that takes up the task of divertor target. The
main objectives of the DED are:
• Study and control of transport in the plasma edge. One important aim is the dis-
tribution of the heat load over the large area of the divertor target. Another is the
creation of a region with a stochastic magnetic field and large transport near the
plasma edge, which could be beneficial for e.g. impurity screening.
• The study of the effect that external perturbations have on the plasma; how external
fields influence the plasma stability, the plasma rotation, the confinement et cetera.
18 magnetic coils are installed at the high-field side of TEXTOR: 16 perturbation
coils, grouped in four sets of four coils, and two compensation coils, above and below the
perturbation coils, that reduce the stray field. The perturbation coils enter the torus at
the bottom, make one toroidal turn around the torus – in clockwise direction when looking
from the top – , and leave at the top. Each set of four coils enters, and leaves, at a different
toroidal position, separated by 90◦ . The numbering of the coils goes from top to bottom.
The pitch angle of the perturbation coils matches that of the field lines at the q=3 surface.
In figure 3.2 the coils of the DED are schematically drawn.
Figure 3.2 : The DED coils are located inside the TEXTOR vacuum vessel at the high-field
side. There are 16 perturbation coils (yellow, black, red and gray), in 4 groups of 4 coils,
and 2 compensation coils (green). Coils with the same colour carry identically phased cur-
rents. Figure (a) shows the coil configuration in 3/1 mode, figure (b) represents the coil
configuration in 12/4 mode.
When we apply current to these divertor coils we form a magnetic perturbation field.
Four differently phased currents (0◦ , 90◦ , 180◦ and 270◦ ) can be fed to the coils. By feeding
identically phased currents to several coils simultaneously, the principal mode numbers of
the perturbation field can be changed. There are three modes of operation:
3/1 mode All coils in one group have the same phase. The next group has a 90◦ phase
shift. This is sketched in figure 3.2 (a).
Section 3.3 - The Dynamic Ergodic Divertor 39

12/4 mode Each coil in the group has a different phasing; the first 0◦ , the second 90◦ ,
the third 180◦ and the fourth 270◦ . The first coil of next group then has 0◦ phasing
again. This is shown in figure 3.2 (b).
6/2 mode The operational mode in between 3/1 and 12/4. The first two coils of the
first group have 0◦ phasing, the last two have 90◦ phasing. The first two coils of the
second group have a 180◦ phase, the last two 270◦ . For the third and fourth group
the situation is the same.
The work presented in this thesis will focus on the 3/1 mode of DED operation, because
the effects on plasma rotation and mode stability are found to be influenced strongest by
this mode of operation. The 12/4 DED mode only affects the very edge of the plasma and
does not create any detectable island activity. This operational mode is used in chapter 8
to look at the influence of the DED on the plasma rotation, in the situation where tearing
modes are not excited. The 6/2 DED mode will not be discussed in this work.
−3
x 10
−3 4
x 10
2
Br (T)
2
Br (T)
0
0
15 1
2 14 2 6
13 5
3 12 3
4 11 n−number 4 4
n−number 5 10 3
9 5 2
6 8 1
7 m−number m−number
Figure 3.3 : The amplitudes of the dif- Figure 3.4 : The amplitudes of the dif-
ferent (m, n) components of the DED ferent (m, n) components of the DED
vacuum field in 12/4 mode at the vacuum field in 3/1 mode at the q = 3
q = 3 surface. The calculation surface. The DED coil current is now
was done for a DED coil current of 1.5 kA, plasma current and toroidal
7.5 kA, a plasma current of 400 kA field are 400 kA and 1.9 T , respec-
and a toroidal magnetic field of 1.9 T . tively. Important Fourier components
Important Fourier components are: are: Br1,1 = 2.9 · 10−3 T , Br2,1 =
Br8,4 = 0.7 · 10−3 T and Br12,4 = 3.7 · 10−3 T and Br3,1 = 3.2 · 10−3 T .
1.9 · 10−3 T . [27] [27]
For each of these operational modes the vacuum field of the DED can be calculated
[1]. The vacuum field is defined as the superposition of the field induced by the DED
coils and the TEXTOR equilibrium field. The amplitudes of the toroidal (n) and poloidal
(m) Fourier components of the radial field Br at the q = 3 surface are given in figure 3.3
for 12/4 operation and in figure 3.4 for 3/1 operation. The Fourier components are –
as expected – centred around the desired mode numbers (m = 12 and n = 4 for 12/4
operation, m = 3 and n = 1 for 12/1 operation). For the toroidal mode number n,
the band is very narrow. For the poloidal mode number m the band is wider and the
components with mode numbers different from m = 12 or m = 3 cannot be ignored.
In figure 3.5 (a) a Poincaré plot of the vacuum field in 3/1 mode is superimposed on
a complete poloidal cross section. Each dot represents the crossing of a field line through
40 Section 3.3 - The Dynamic Ergodic Divertor

the plotted poloidal cross section. One can observe a strong m/n = 2/1 structure, as
could be expected from the wide m-band in the Fourier spectrum shown in figure 3.4.
Figure 3.5 (b) zooms in on the edge of the plasma and unfolds the poloidal angle. One
can distinguish three areas in this plot:
• Very close to the edge there is a zone where the field lines connect to the wall after
a limited number of toroidal turns. This is the laminar zone.
• A bit further into the plasma the field lines do no longer have a short connection
length to the wall, but they do have a radial component. This causes them to fill a
volume rather than laying on a (flux) surface. This is the stochastic (ergodic) zone.
• Closer to the plasma core the field lines become regular; they form circular, flux
surfaces.
Figure 3.5 : A Poincaré plot of the vacuum field in 3/1 DED operation. In (a) a poloidal cross
section of the vacuum field in the TEXTOR vessel is drawn; also the DED coils are indicated.
In (b) the same vacuum field is drawn, but now with the poloidal direction unfolded. On this
plot the three regions – laminar, stochastic and regular – can easily be recognised. The
parameters for which this vacuum field was calculated are: Ip = 300 kA, Bφ = 2.25 T and
IDED = 3.75 kA. [27]
In 3/1 operation the current supplied to the DED coils can go up to 3.75 kA, in 12/4
operation a coil current up to 15 kA is allowed. Both the current amplitude and the mode
of operation determine the strength of the perturbation field. For high mode numbers –
e.g. in 12/4 operation – the field rapidly decays with increasing distance from the DED
coils. Therefore the 12/4 mode will only affect the very edge of the plasma, whereas in
3/1 operation the influence of the DED penetrates deeply into the plasma.
The current fed to the DED coils can be either DC or AC. For DC DED the resulting
perturbation field is static. When we apply AC current the phase in each coil constantly
changes, resulting in a dynamic perturbation field. The phase can increase (AC+ ) or
decrease (AC− ). Consider the 3/1 coil configuration sketched in figure 3.2 (a), with as a
starting point the first group of coils (yellow) at 0◦ , the second (black) at 90◦ , the third
Section 3.3 - The Dynamic Ergodic Divertor 41

(red) at 180◦ and the fourth (gray) at 270◦ . When we apply AC+ , the phase in each group
will increase, such that after a quarter period the phases are as follows: yellow = 90◦ , black
= 180◦ , red = 270◦ and gray = 0◦ . This means that the field created by the DED has
rotated. This rotation is mainly in the poloidal direction and directed from top to bottom
at the high-field side. In a poloidal cross section, with the high-field side at the left (like is
the case in figure 3.5 (a)), the poloidal rotation of the AC+ field is counterclockwise. The
rotation of the field also has a small component in the toroidal direction. The toroidal
rotation of the AC+ field is clockwise when looking on top of the torus. With AC− DED
the field also rotates, but the direction is reversed. For 12/4 DED operation the result is
of course the same.
To summarise: an AC+ field rotates counterclockwise in the poloidal direction and
clockwise in toroidal direction. An AC− field rotates poloidally clockwise and toroidally
counterclockwise. The rotation frequency of the AC fields can go up to 10 kHz, but
usually 1 kHz and 3.75 kHz are used.
To conclude this section, the main DED parameters are summarised in table 3.2.
Table 3.2 : DED parameters
Set of 16 perturbation coils (and 2 compensation coils)

Modes of operation:
4 consecutive coils with same phasing - 3/1 operation
2 consecutive coils with same phasing - 6/2 operation
Phase shift for each coil - 12/4 operation
DED coil currents:
DC - Static perturbation field
fDED = 0 kHz
AC+ - Rotating perturbation field
toroidal component clockwise
poloidal component from top to bottom at HFS
fDED = 1 kHz or 3.75 kHz
AC− - Co-rotating perturbation field
toroidal component counterclockwise
poloidal component from bottom to top at HFS
fDED = 1 kHz or 3.75 kHz
3.4 Co- and counter directions in TEXTOR

In this thesis directions play an important role. We therefore give some attention to the
definition of positive, co- and negative, counter-directions in the TEXTOR tokamak.
Throughout this thesis, a righthanded coordinate system (r, θ, φ) is used, where r
represents the radial direction and acts as the flux surface coordinate; θ and φ represent
42 Section 3.4 - Co- and counter directions in TEXTOR

the poloidal and toroidal direction, respectively.

The radial coordinate r starts at 0 in the plasma centre and increases towards the
plasma edge. This means a vector with a positive radial component points outwards,
while a vector with a negative r-component points in the direction of the plasma centre.
Toroidally the counterclockwise direction, when looking on top of the TEXTOR toka-
mak, is defined as the positive or co-direction; consequently the toroidal, negative or
counter-direction is clockwise.
With the positive direction defined for r and φ, and taking into account the righthanded
(r, θ, φ) coordinate system, the positive direction in the poloidal plane is now fixed. When
we look at a poloidal cross section, with the high-field side at the left, the poloidal,
positive or co-direction is clockwise; the negative or counter-direction is counterclockwise
in a poloidal plane.
Co, Ip, DED AC−
Counter, Bφ, DED AC+
eφ
Counter, f*e, DED AC+
er
Co, Bθ, f*i , DED AC−
eθ
Figure 3.6 : Definition of co- (positive) and counter-directions (negative) in TEX-

TOR, both toroidally and poloidally.
Although it is possible to reverse the direction of the plasma current Ip and the toroidal
magnetic field Bφ , all data presented in this thesis were obtained from discharges using
the common TEXTOR settings. Toroidally these settings imply that the plasma current
Ip is in the co-direction, hence positive. The toroidal magnetic field Bφ is directed in the
counter-direction and negative. In figure 3.1 it is seen that NBI1 injects momentum in
the co-direction, inducing positive, toroidal plasma rotation, while NBI2 causes negative
or counter-rotation. In the poloidal plane the poloidal magnetic field Bθ is directed in the
co-direction.
Diamagnetic drifts of ions and electrons are perpendicular to the magnetic field:
1 ∇pα × B
v ∗α = − , (3.1)
eα n B2
with α = i or e. Due to the large toroidal component of the B-field, compared to the
poloidal component, the diamagnetic drift of ions and electrons is almost completely
poloidal. When we take into account the charge of the particle in question, the fact
that ∇pα is always directed towards the plasma centre and that Bφ is in the counter-
Section 3.4 - Co- and counter directions in TEXTOR 43

direction, equation (3.1) results in an ion diamagnetic drift in the poloidal, co-direction
and an electron diamagnetic drift in de counter-direction.
In section 3.3 it was seen that the AC+ DED field rotates poloidally in the counter-
clockwise direction, and toroidally it rotates clockwise. With the definition of co- and
counter-direction given above, this means that the rotation of the AC+ field is, both
poloidally and toroidally, in the counter-direction and thus negative. Consequently, the
AC− DED field is rotating in the positive, co-direction.
In figure 3.6 the co- and counter-directions for TEXTOR are shown. In table 3.3 the
typical directions of several plasma parameters are summarised.
Table 3.3 : Directions of plasma parameters in TEXTOR
Co-direction, Positive Counter-direction, Negative

Toroidal Plasma current Ip Toroidal magnetic field Bφ
Momentum input NBI1 Momentum input NBI2
Toroidal rotation AC− DED field Toroidal rotation AC+ DED field
Poloidal Poloidal magnetic field Bθ
Ion diamagnetic drift vi∗ ∝ fi∗ Electron diamagnetic drift ve∗ ∝ fe∗
Poloidal rotation AC− DED field Poloidal rotation AC+ DED field
3.5 Plasma diagnostics

The plasma parameters that are of importance for this thesis are the plasma rotation veloc-
ity, the ion and electron temperature and the density. In the following paragraphs several
diagnostics capable of measuring these parameters will be introduced. A description of all
the diagnostics available at TEXTOR can be found in [10] and [20].
3.5.1 Charge exchange recombination spectroscopy

Charge exchange recombination spectroscopy (CXRS) is the principle diagnostic used in
this thesis. It is a technique that analyses the light emitted by impurity ions in the plasma
core when they undergo a charge exchange reaction. In such a charge exchange reaction,
a fully stripped impurity ion receives an electron from a neutral particle, e.g. provided by
a neutral beam. The impurity ion receives this electron in an excited state and will return
to the ground state, losing its excess energy through line-emission. This line spectrum can
be analysed with a spectrometer, yielding following information:
Ion temperature The width of the emission line is caused by Doppler broadening. It
is therefore a measure of the thermal velocity distribution, and hence of the ion
temperature.
Plasma rotation The Doppler shift of the emission line gives the velocity in the direction
of the line-of-sight. Taking into account the geometry of the CXRS system this
velocity component along the line-of-sight can be transformed into a toroidal or
poloidal rotation.
44 Section 3.5 - Plasma diagnostics

Impurity content The intensity of the emission line is proportional to the density of the
impurity ions.
At TEXTOR there are two main CXRS systems. The NBI-CXRS system uses the
NBI1 heating beam as a source of neutrals. The system has lines-of-sight with a toroidal
geometry, which means that the measured Doppler shifts can be translated in toroidal
rotation frequencies Ωφ . The time resolution of the NBI-CXRS system is 50 ms. The ne-
cessity of NBI1 prohibits CXRS measurements with this system during Ohmic discharges.
The RUDI-CXRS system uses a low power, radially injecting, diagnostic neutral beam.
The influence of this neutral beam on the global plasma behaviour is negligible, such that
this diagnostic can be used to investigate Ohmic discharges. However, due to the lower
beam power, the CXRS signal is low and a longer exposure time is needed to collect enough
signal. This results in a time resolution of over 1 s. The lines-of-sight of the RUDI-CXRS
system are located in a poloidal plane, allowing for poloidal rotation measurements.
In chapter 4 CXRS is discussed in detail.
3.5.2 Electron Cyclotron Emission
The electrons that gyrate around the magnetic field lines emit electron cyclotron radia-
tion. At the electron cyclotron frequency the plasma is optically thick. This means that
the electron cyclotron emission (ECE) is a black body radiation. According to Raleigh-
Jeans law, the long wavelength approximation of Planck’s law, the intensity of black body
radiation is related to the temperature. The ECE intensity is therefore proportional to
the electron temperature Te . The frequency of the ECE radiation – ωe = eB/me – is
proportional to the magnetic field B that in turn depends on the major radius R. This
means that the ECE intensity as a function of frequency can be transformed to the elec-
tron temperature as a function of major radius: Te (R). Microwave detectors with a high
sampling rate allow for measurements of Te (R, t) profiles with a fast time resolution. An
absolute measurement of the intensity is needed, which means that an intensity calibration
of the detectors is needed. This can be done by using a ‘hot’ source or by cross calibration
with another Te -diagnostic like e.g. Thomson scattering.
Several ECE systems are installed at TEXTOR. The two commonly used systems are:
EC11 and ECE-imaging. The EC11 system is a 11-channel, fixed frequency system. The
frequencies of the 11 channels lie in the range of 105 − 145 GHz. For TEXTOR operation
with a central, toroidal magnetic field of Bφ = 2.25 T this means the whole plasma region
R = 1.55−2.14 m is covered by the EC11 system. The time resolution of the EC11 system
goes up to 10 kHz.
The ECE-imaging system provides 2D Te profiles. The radial position R comes from
8 frequency channels, separated by 0.5 GHz. The vertical Z-coordinate is resolved by
imaging a plasma column of 16 cm in height onto an array with 16 detectors. The result
is a matrix of 8 × 16 channels that images the Te in an area of 8 × 16 cm2 . The measuring
area can be radially positioned by changing the frequency of the central channel over
95−125 GHz in steps of 0.5 GHz. For a standard TEXTOR Bφ field of 2.25 T this means
from R = 1.40 m to R = 1.84 m with a step of approximately 1 cm. For measurements
with a high time resolution, sampling at 500 kHz is possible for a duration of 2 s. At
200 kHz sampling the whole plasma discharge – typically 6 s – is covered.
Section 3.5 - Plasma diagnostics 45

3.5.3 Thomson Scattering

The Thomson scattering diagnostic is based on the scattering of laser light at free electrons.
The velocity of the electrons causes the scattered light to have a Doppler shift with respect
to the wavelength of the incoming laser light. For a Maxwellian velocity distribution of
the electrons, the spectrum of the scattered light has a Gaussian shape. The width of the
spectrum gives the electron temperature Te .
The intensity of the scattered light is proportional to the electron density ne . So apart
from being a Te measurement, Thomson scattering provides a ne measurement as well.
For the density measurement an absolute calibration is needed.
The scattered light originates where the laser passes through the plasma. Imaging the
scattered light therefore gives the local Te and ne along the path of the laser light. For
Thomson scattering measurements a powerful laser is needed, so that the intensity of the
scattered light is high enough. It also has to be a pulsed laser, so that the scattered light
rises significantly above the plasma background light.
At TEXTOR a multi-pulse Thomson scattering system is installed. It uses a high-
power, intra-cavity ruby laser. The advantage of an intra-cavity laser is the fact that all
laser light is used, whereas in a normal laser only a few percent of the laser light is coupled
out of the cavity. The technical challenge of an intra-cavity laser is the enormous length
of the cavity, that has to include the plasma. In TEXTOR the laser cavity is 18 m long.
This unique system allows the measurement of Te and ne with a very high spatial and
time resolution.
The intra-cavity laser currently allows for up to 40 laser pulses over a period of 4 ms
with each an energy of 15 J. The laser passes vertically through the plasma and is shifted
9 cm from the centre of the tokamak, towards the low-field side. The core Thomson
scattering system images the full plasma diameter – 90 cm – on 120 channels. This means
that during a 4 ms laser burst, Te (z, t) and ne (z, t) profiles are obtained with a spatial
resolution of 7.5 mm and a 10 kHz time resolution. When the plasma position and
equilibrium are known, the z-coordinate can be transformed in a radial r-coordinate.
Apart from the core system, also a high-resolution edge system is present at TEXTOR.
It images 16 cm in the plasma edge onto 98 channels, yielding a spatial resolution of
1.7 mm. Because the core and edge system use the same detection equipment, they
cannot be used simultaneously.
3.5.4 Interferometer
Another technique for measuring the electron density is interferometry. In interferometry
the light of a laser beam that has gone through the plasma is compared with that of a
reference beam that travelled over the same length, but in vacuum. As the phase velocity
of laser light depends on the refractive index of the medium it is going through – plasma
for one beam, vacuum for the reference beam – there is a phase difference between the
light of the plasma beam and the reference beam. From the interference pattern of the two
beams the phase difference and thus the line-averaged refractive index of the plasma can
be determined. The refractive index of a plasma is proportional to the electron density, so
that interferometry yields the electron density ne averaged over the line along which the
laser passes through the plasma.
In TEXTOR the light of a far-infrared HCN laser goes vertically through the plasma
at 9 different radial positions, yielding the line-integrated density at 9 different positions.
46 Section 3.5 - Plasma diagnostics

The time resolution is quite high – 20 kHz – and the line-integrated density signals are
used for real time density and positioning control. Using Able-inversion the line-integrated
densities can be transformed into density profiles ne (r, t). The estimated error in the local
density due to the Able inversion is about 5%.
Section 3.5 - Plasma diagnostics 47

Chapter 4
Charge Exchange Recombination

Spectroscopy
4.1 Introduction
For the study of plasma rotation it is necessary to have a reliable method to measure that
rotation. Charge exchange recombination spectroscopy (CXRS) is a powerful technique
that allows local measurements of the ion velocity distribution, from which the random
thermal velocity – or ion temperature – and the fluid velocity are derived. In CXRS the
line-emission by ions in the plasma core is analysed. The Doppler shift and broadening of
this line-emission are a measure for the fluid velocity and the ion temperature.
Line-emission from ions in the plasma core does not occur spontaneously. Due to the
high temperature the ions in the centre of the plasma are – except for heavy impurities
– fully stripped. In order to emit light, they must at least receive one electron. This
happens when they undergo a charge exchange reaction. Neutral atoms, usually provided
by neutral beam injection, will in such a charge exchange reaction deliver one of their
electrons to the plasma ion. The ion receives this electron in a high quantum level. It
will return from its excited state to the ground state, losing its excess energy through
line-emission.
Charge exchange recombination spectroscopy is well described in the literature [42, 38].
In this chapter we bring together the elements that are essential in the interpretation of
the measurements at TEXTOR. The CXRS setup used at TEXTOR will be discussed in
the last section of this chapter.
4.2 Principle
In the charge exchange process a neutral atom loses an electron to a plasma ion [42].
Depending on the type of the plasma ion and the neutral particle, the electron is most
likely to be transferred to a specific, preferential quantum level of the plasma ion. The
probability for electron transfer to both higher and lower levels decreases monotonically.
In case of a charge exchange reaction H + C 6+ → H + + C 5+ the C 5+ level most likely to
be populated is n = 4, but also for the n = 7 and n = 8 levels the chances of population
are significant [38]. These quantum levels differ from the ground state, which means that
a charge exchange reaction will yield a plasma ion in an excited state. It will return to
49
Chapter 4 - Charge Exchange Recombination Spectroscopy
H + A+ → H + + A∗
→ H + + A + hν1 + hν2 + ... (4.1)
the ground state through a cascade of transitions.

The radiation emitted by the transitions varies over a wide wavelength range, depend-
ing on the populated states. For diagnostic use of charge exchange, especially transitions
that emit visible light are of interest. The advantage of visible light is that it can be
easily collected by common optical elements and also plenty spectroscopical methods are
available for the analysis of visible light.
In the TEXTOR tokamak the n : 8 → 7 transition of C 5+ is used. It has a wavelength
of 529 nm, which is in the visible range. The reason for analysing the charge exchange light
of an impurity – carbon – instead of the charge exchange light of the bulk ions – deuterium
or hydrogen – is the high complexity of the hydrogen-deuterium spectrum. Usually the
bulk plasma is a mixture of deuterium and hydrogen, where the charge exchange (CX)
emission lines of deuterium and hydrogen overlap. Apart from CX emission, neutral
deuterium and hydrogen in the plasma edge have a strong line-emission due to excitation
by electrons. This emission lies in the same wavelength range as the charge exchange
emission. Finally the neutral beam species is commonly hydrogen as well, resulting in a
beam spectrum on top of the charge exchange spectrum. As a result one gets a spectrum
that is the sum of two charge exchange lines, electron excited emission and beam emission.
Decomposing this spectrum into its components in order the regain the charge exchange
line results in large error bars on the derived quantities.
Therefore, as said above, CXRS is usually done at impurity ions. Commonly used
are helium and carbon. Helium is investigated because of its relevance: it is a reaction
product of the fusion reaction. Carbon is looked into because its spectrum is – compared
to the spectrum of hydrogen or helium – relatively easy to analyse. Also, carbon is always
present in tokamaks because it is commonly used as a wall material.
Of course, when analysing the CX spectrum of carbon, the temperature, rotation
velocity and density that result from this analysis are those of carbon, not those of the
bulk ions. The energy transfer time between carbon and hydrogen is, however, quite low.
More specifically: it is shorter than the energy confinement time τE . This means that
the assumption that the carbon temperature equals the bulk ion temperature is valid,
within the energy confinement time. The carbon rotation velocity does differ from the
bulk rotation velocity. In section 2.5.1 the expression for this difference is given, so that
from the carbon fluid rotation the plasma bulk rotation can be calculated. The fact that
50 Section 4.2 - Principle

carbon CXRS gives us a measure for the carbon density, allows us to use CXRS data for
impurity transport studies.
In spectroscopy the light emitted by C 5+ ions is called C VI emission. The C VI
(n : 8 → 7) emission has a natural wavelength of λn = 529 nm and a natural line width of
∆λn = 5 . 10−5 nm [63]. If an ion moves with a velocity uz in the direction of the observer,
the light emitted by this ion is Doppler shifted by ∆λd = uz /cλn . For an ion temperature
of 10 eV the Doppler shift due to the thermal velocity is ∆λd = 5 . 10−2 nm. Compared
to this Doppler shift the natural line width ∆λn can be neglected, so that we can use a
delta-function f (λ) = δ (λ − λn (1 + uz /c)) to describe the emission line.
Let us consider a Cartesian observation system, with the z-axis in the direction of
observation. The intensity of the emission coming from a volume (dx, dy, dz) around the
point (x, y, z), depends on the density of neutral hydrogen nH , the density of the carbon
ions nC and the effective emission rate hσvi. This effective emission rate depends on the
cross section σ for charge exchange reactions and expresses the likelihood of interactions
between carbon and hydrogen in which the n = 8 state is populated.
The velocity of the carbon ions is given by a Maxwellian distribution:
s
mC 3 (ux − vx )2 + (uy − vy )2 + (uz − vz )2

g(ux , uy , uz ) = exp − , (4.2)
2πTC 2TC /mC
where vx , vy and vz are the components of the carbon fluid velocity.

The spectrum of the light coming from the volume (dx, dy, dz) emitted in the direction
of the observer then is:
Z Z Z
1 uz
flocal (λ) = nH nC hσviδ λ − λn 1 + g(ux , uy , uz )dux duy duz ,
4π c

(λ − λ0 )
= I exp − , (4.3)
∆λ2
where
vz
λ0 = λn 1 +
r c
2TC λn
∆λ =
mC c
hσviλn
I = nH nC 3/2 . (4.4)
4π c ∆λ
The above learns us that we can derive vz – the carbon velocity in the direction of the
observer – from the peak position of the Gaussian line-emission, that the carbon tem-
perature TC can be derived from the width of the Gaussian and that the intensity of the
Gaussian emission line gives us the carbon density nC .
The total charge exchange spectrum that is observed is the sum of all the light locally
emitted along the line-of-sight:

(λ − λ0 )
Z
f (λ) = I exp − dz, . (4.5)
∆λ2
In the largest part of the plasma there is no neutral hydrogen present, thus nH = 0.
There are however two regions in the plasma where there is a considerable amount of
Section 4.2 - Principle 51

neutral particles: neutral particles that come from the wall will penetrate over a certain
distance into the plasma before being ionised, and the neutral beam will inject highly en-
ergetic neutral particles all the way to the plasma centre. This means that at two positions
on the line-of-sight the neutral density is non-zero: where the line-of-sight goes through the
plasma edge and at the position where the line-of-sight crosses the neutral beam. The CX
spectrum can therefore be split in two components: a passive charge exchange component
coming from the plasma edge and an active charge exchange component caused by the
neutral beam. This is shown in figure 4.1. The passive part of the spectrum contains of
course only data on the edge properties. The active part of the charge exchange light orig-
inates from the crossing of neutral beam and line of sight, ensuring a local measurement
of nC , TC and vz . Because the neutral beam supplies neutral particles up to the plasma
centre, we can measure nC , TC and vz over the whole plasma. With multiple lines-of-sight,
profiles of nC , TC and vC can be derived from the active part of the CX spectrum.

(λ − λ0 ) (λ − λ0 )
Z Z
f (λ) = I exp − dz + I exp − dz . (4.6)
edge ∆λ2 beam ∆λ2
| {z } | {z }
passive active
passive CX emission
active CX emission
passive CX emission
m
y
bea
x
z
tral
Neu
Figure 4.1 : A top view of a tokamak with a CX line-of-sight. The passive component
of the CX spectrum is caused by neutral particles in the edge, the active CX signal
is emitted where the line-of-sight crosses the neutral beam.
In an optimised viewing geometry the line-of-sight is tangential to a flux surface where

it crosses the neutral beam, like it is drawn in figure 4.1. Because nC , T and vz are flux
functions, they can be taken out of the integral and the active part of the CX spectrum
becomes:
(λ − λ0 )
factive (λ) = I exp − , (4.7)
∆λ2
52 Section 4.2 - Principle

with
vz
λ0 = λ n 1 +
r c
2TC λn
∆λ =
mC c
hσviλn
Z
I = nC 3/2 nH dz. (4.8)
4π c ∆λ beam
When we use CXRS as a diagnostic tool, we are mainly interested in the active CX
emission. The active part of the CX spectrum gives us the local values of TC , nC and
vz . If we assume that the radial part of the carbon fluid velocity v C is a lot smaller
than the toroidal and poloidal component, vz can be written as: vz = vφ cos α + vθ cos β,
where α is the angle between the line-of-sight and the toroidal direction and β is the angle
between the line-of-sight and the poloidal direction. When we choose a line-of-sight that is
perpendicular to the poloidal direction, vz is proportional to the toroidal rotation velocity
and vice versa.
1600
1400
1200
Intensity (a.u.)
1000
800
600
active CX emission
400
passive CX emission
200
Background radiation CIII impurity line
0
526 527 528 529 530 531 532
Wavelength (nm)
Figure 4.2 : A typical CX spectrum. The black dots represent the measured spectrum,
the solid gray line is the result of a fitting routine. The different components of the
CX spectrum are indicated: the active CX in blue, the passive CX in red, a CIII
impurity line in green and the background radiation is yellow.
The undesirable passive component can be dealt with in two ways. The first is using
a modulated neutral beam. During the period that the neutral beam is switched off, the
only emission left is the passive emission. When the passive spectrum is subtracted from
the total spectrum when the neutral beam is switched on, only the active CX spectrum
remains. This method is widely used. There are, however, some disadvantages to this
method. First of all there is a loss in time resolution, because the neutral beam is switched
off half of the time. Another weak point is the fact that it assumes that the passive emission
during neutral beam is the same as without neutral beam. Because a neutral beam can
change the conditions in the edge of the plasma, this is not necessarily the case. Modulation
Section 4.2 - Principle 53

also increases the errorbars: with A the total spectrum and B p the passive spectrum the
error on the subtracted spectrum A − B becomes: ∆(A − B) = (∆A)2 + (∆B)2 , which
is larger than the error on both A and B.
A second method for subtracting the passive emission is modelling of the passive CX
spectrum. For this modelling measurements of temperature and density in the plasma
edge can be used, as well as e.g. the spectrum measured by a line-of-sight that does not
cross the neutral beam. When the passive spectrum is modelled a continuous neutral
beam can be used. One needs, however, a very accurate model of the passive emission.
In the wavelength range of passive and active CX emission, also light from other sources
is present; e.g. a continuum background of Bremsstrahlung and line radiation from partly
ionised impurity ions. Figure 4.2 a measured charge exchange spectrum is given. When
this spectrum is fitted, all different components – active CX, passive CX, impurity lines
and background radiation – have to be taken into account.
4.3 Additional effects

The simple picture given above, where the width of the active CX line only depends on
the temperature, the peak position is a simple function of the plasma rotation and the
impurity density solely determines the intensity of the line-emission, is not fully correct.
Several other mechanisms cause a broadening and shift of the CX emission line. Also,
as was already mentioned, passive CX and active CX are not the only emissions in the
observed wavelength range: line-emission from impurities at the edge, secondary emission
by the CX ions, background emission, et cetera, they all complicate the charge exchange
spectrum. These complicating effects will be discussed in this section.
4.3.1 Energy dependence of the emission rate

In the previous section the effective emission coefficient hσvi was, for simplicity, considered
to be constant. In reality hσvi depends on the collision velocity between the neutral particle
and the plasma ion [38]:
q
vcol = vb2 + u2x + u2y + u2z − 2vb (cos δuy − sin δuz ), (4.9)
where the same Cartesian system is used as above, vb is the velocity of the neutral beam
particle and δ is the angle between the neutral beam and the y-axis.
In most situations vb is constant and vb (ux , uy , uz ); in TEXTOR e.g. the beam
energy is 50 keV whereas the thermal energy is in the order of 1 keV . In a first approxi-
mation it is acceptable to say hσvi(vcol ) ≈ hσvi(vb ), thus equation (4.3), where hσvi was
taken out of the velocity integral, is correct.
Equation (4.9) shows however that vcol has a distribution around vb . If the value
of hσvi changes significantly around vb or it the distribution of vcol around vb is wide –
i.e. at high temperatures – the effective emission coefficient hσvi can not be taken out
of the integral of equation (4.3). In figure 4.3 the effective emission rate hσvi for C VI
(n : 8 → 7) emission is plotted as a function of collision velocity vcol . In the same figure
also the distribution of vcol is given for a temperature of 1 keV in a TEXTOR CXRS
geometry with the TEXTOR beam energy of 50 keV and for a temperature of 15 keV in
ITER with the ITER beam energy of 100 eV . It is seen that hσvi changes significantly
over the FWHM of the vvol distributions. Especially in the ITER case.
54 Section 4.3 - Additional effects

−14 −6
x 10 x 10
1.4 6
1.2
5
1
4
σ v (m3/s)
0.8
fcollision
3
0.6
2
0.4
1
0.2
TEXTOR ITER
1 keV 15 keV
0 0
0 1 2 3 4 5 6 7 8 9 10
v (m/s) 6
x 10
collision
Figure 4.3 : The effective emission coefficient hσvi is as a function of the collision
velocity vcol (dashed line). The distribution of vcol is also plotted (solid lines), for
a temperature of 1 keV in a TEXTOR geometry and for a temperature of 15 keV
in ITER. The neutral beam in TEXTOR has a beam energy of 50 keV , whereas
the beam energy in ITER is 100 keV . This explains the different positions of both
vcol distributions. The gray regions indicate the FWHM of both distributions. [38]
The result of the vcol distribution and dependence of hσvi on vcol is a change in Doppler
broadening and Doppler shift of the emission line. This has an effect on the measurement
of temperature, rotation velocity and intensity. Usually the width of the emission line is
smaller than the width that one would expect in the case of a constant hσvi, hence the
observed temperature is lower than the true temperature. In figures 4.4 and 4.5 the effect
of the energy dependence of hσvi on temperature and rotation is given for the TEXTOR
core CXRS system and for the ITER core CXRS system. Due to the low temperatures
in TEXTOR – below 5 keV – we can neglect this effect in TEXTOR. Despite the higher
temperatures that can be achieved in ITER the effect of the energy dependence of hσvi
is less significant than in the TEXTOR case. The reason for this is the geometry of
the core CXRS system in ITER. The lines-of-sight in the ITER core CXRS system are
almost perpendicular to the neutral beam. The shape of the CX spectrum comes from the
Doppler shift due to movement in the direction of the observer uz . The collision velocity
vcol however depends strongly on the velocity towards and away from the neutral beam,
thus on uy in the ITER geometry. Therefore the shape of the CX spectrum – depending
on uz – and the intensity of the spectrum – depending on the collision velocity, thus uy –
are independent in the ITER geometry. As a result the effect on rotation and temperature
measurement is negligible.
Gyration of plasma ions The energy dependence of hσvi is also responsible for a
gyration-dependent shift of the poloidal rotation. The gyromotion of ions is predominantly
in the poloidal direction, due to the large toroidal magnetic field, so the effect is only seen
in the poloidal rotation, not in the toroidal rotation. During one part of its gyro-obit a
C 6+ ion will move towards the neutral beam, having a larger collision velocity, in the other
Section 4.3 - Additional effects 55

4
x 10
15
TEXTOR 10 TEXTOR, Te = 1 keV
ITER TEXTOR, Te = 2 keV
9
TEXTOR, Te = 5 keV
8
ITER, Te = 2 keV
Ωobserved (rad/s)
10 7 ITER, Te = 15 keV
Tobserved (keV)
4
5 3
0
0
0 5 10 15 0 2 4 6 8 10
Ttrue (keV) Ωtrue (rad/s) x 10
4
Figure 4.4 : The observed ion tempera- Figure 4.5 : The observed toroidal rota-
ture, due to the energy dependence of tion, due to the energy dependence
hσvi plotted against the true tempera- of hσvi plotted against the true ro-
ture for the TEXTOR (blue) and the tation for the TEXTOR (blue) and
ITER (red) case. The dotted line in- the ITER (red) case. For ITER we
dicates Ttrue = Tobserved . assumed that no poloidal rotation is
present. The dotted line indicates
Ωtrue = Ωobserved .
part of the gyro-obit the collision velocity is lower. This means that in one part of the
gyro-orbit more excited C 5+ ions are produced due to the higher effective emission rate
hσvi than in the other part. If the lifetime τ of the excited state is larger than the gyro-
period this does not matter. However, with τ being of the order of 10−9 s and the cyclotron
frequency ωc being of the order of 108 Hz for carbon impurities and typical TEXTOR
magnetic fields of 2.25 T , the excited impurity ions only perform about a tenth of their
gyration before emitting the charge exchange photon. This means that predominantly
ions with a gyromotion that has a high hσvi will emit CX light, leading to a shift in the
measured poloidal rotation.
4.3.2 Non-thermal line broadening

In this section we discuss two broadening effects of the CX emission line that are inde-
pendent of the ion temperature. These effects are the Zeeman effect, due to the magnetic
field, and so-called l-state mixing. They are both discussed in more detail in [42] and [75].
Non-degeneracy of the l-states In the previous section we assumed that the transi-
tion between the n = 8 and the n = 7 level was only one transition between two levels with
each only one corresponding energy. As a result the emission consisted of one single line
with a natural wavelength of λn = 529 nm and a natural line width of ∆λn = 5.10−5 nm
[63]. In reality the n = 8 and n = 7 levels of the C 5+ ion have a substructure. As a result
there is not just one transition between the n = 8 and n = 7 level, but there is a number
of possible transitions between the sublevels of n = 8 and the sublevels of n = 7.

In figure 4.6 the energy level diagram of the C 5+ ion is given. It shows that each
principle quantum shell – defined by the quantum number n – has n subshells – defined
by the orbital quantum number l = 0, . . . , n − 1 and also referred to as s, p, d, f, g, . . . .
When electron spin is taken into account each l-level – except l = 0 – splits in two levels,
defined by J = l + 1/2 and J = l − 1/2. The allowed radiative transitions between the
n = 8 and n = 7 level are those for which ∆l = ±1 and ∆J = 0, ±1.
s p d f g h i k
(l)
0 1 2 3 4 5 6 7
J=7/2 (0.36) J=9/2 (0.40) J=11/2 (0.40) J=13/2 (0.43) J=15/2 (0.43)
J=5/2 (0.32) J=13/2 (0.43)
J=3/2 (0.25) J=7/2 (0.36) J=9/2 (0.40) J=11/2 (0.40)
J=5/2 (0.32)
J=3/2 (0.25)
J=1/2 (0.00)
n=8
481.895 eV J=1/2 (0.00)
J=13/2 (0.65)
J=7/2 (0.54) J=9/2 (0.58) J=11/2 (0.61)
λn = 529 nm J=5/2 (0.47)
J=7/2 (0.54) J=9/2 (0.58)
J=11/2 (0.61)
J=3/2 (0.36) J=5/2 (0.47)
J=3/2 (0.36)
J=1/2 (0.00)
n=7
479.553 eV J=1/2 (−0.04)
Figure 4.6 : Energy level diagram for the principle (n = 8)- and (n = 7)-shells of C 5+ . Each
principle n-shell is divided in l-subshells and J-levels. The energy within the principle n-shell
differs slightly. That difference is exaggerated in this plot. The energy difference with the
(l = 0)-level, within the same n-shell, is given in parentheses (10−4 eV ). The energy level
of the (l = 0)-level, for both n = 8 and n = 7 is given at the left (eV ). [63]
If all these different sublevels within the same n shell would have the same energy,
i.e. if they would be degenerated, the sum of the emission by all the allowed transitions
between n = 8 and n = 7 would still be one single line with a wavelength of λn = 529 nm.
However, the energy of the different sublevels within one principle n-shell is not exactly
the same; there is a fine-structure. Due to the fine-structure each allowed transition has
a slightly different wavelength, and the total emission spectrum consists of a set of lines
instead of one line.
Which of the l-levels of the n = 8 shell of C 5+ are populated in a charge exchange reac-
tion depends on the beam energy. In TEXTOR the l-levels are approximately statistically
populated due to the 50 keV neutral beam. Furthermore, if the lifetime τ of the excited
states is considerably larger than the ion-ion collision time, collisions will cause a transfer
between the different l-states of n = 8 shell before the charge exchange electron drops
to a lower level and emits a photon. This phenomenon is called collisional l-mixing. It
means that, even if the population of the l-levels would not be statistic, collisional l-mixing
would make sure that the l-levels are statistically populated. The statistic population of
all l-levels results in 37 allowed transitions, instead of one. The spectrum of these 37 lines
is given by the black lines in figure 4.7. These lines cover a wavelength range of about

0.015 nm. This corresponds with the Doppler broadening caused by a ion temperature of
4 eV .
Zeeman effect In a tokamak a strong magnetic field – in the order of B = 2 T – is

present. Without a magnetic field the energy levels within the same n shell differ slightly
due to the fine-structure. The presence of a magnetic field will cause Zeeman splitting
of one J-level into 2J + 1 energy levels separated by ∆Ez = µb gBmj , mj = −J, . . . , +J,
where µb is the Bohr magneton (5.789 .10−5 eV /T ) and g is called the Landé factor (of
the order 1, depends on l, J and the spin). The allowed transitions are now given by
∆l = ±1, ∆J = 0, ±1 and ∆mj = 0, ±1. When, for simplicity, we set g = 1, then every
line in the fine-structure spectrum is split into three lines: λ → λ − ∆λZ , λ, λ + ∆λZ , with
∆λZ = µb B. For B = 2 T the Zeeman split is ∆λZ = 0.03 nm. This is twice as large as
the wavelength range of the fine-structure spectrum, which implies that in the presence
of a strong magnetic field the total line spectrum of the C VI (n : 8 → 7) transition
consists of three groups of fine-structured lines, where the distance between the groups
is determined by Zeeman splitting. In figure 4.7 the total line spectrum in a magnetic
field of 2 T is drawn. It covers a wavelength range of 0.07 nm, which corresponds with a
Doppler broadening caused by an ion temperature of 90 eV .
529 529.02 529.04 529.06 529.08 529.1

λ (nm)
Figure 4.7 : Line spectrum of the C VI (n : 8 → 7) transition. In case there is no magnetic

field present the line spectrum is given by the black lines. It is determined solely by the
fine-structure. When a strong magnetic field is present, Zeeman splitting causes every line
to split in three lines. The resulting spectrum has three groups of fine-structure lines: the
original – in black – and the groups shifted by ±∆λZ – in grey.
The total CX spectrum, where non-thermal broadening due to l-mixing and Zeeman
splitting is included, is the sum of the Doppler spectra for every emission line, where the
relative intensity of every transition and the population of every sublevel is taken into
account. When we treat every emission line separately the analysis of a CX spectrum
gets quite complicated. Therefore the total profile of all transition lines is presented as a
single, but broadened Gaussian, of which the width and the peak position depends on the
l-mixing and Zeeman splitting. This Gaussian replaces the δ-function in equation (4.3).
Apart from the introduction of this non-thermal Gaussian into the CX spectrum, non-
thermal broadening also puts a lower limit to ion temperature measurements with CXRS.
Temperatures below 100 eV are usually hard to resolve.
4.3.3 Non-CX emission

In the same wavelength range as the charge exchange emission there are other sorts of
emission that have nothing to do with the CX reaction or are a secondary effect of charge
exchange. Especially when the active CX emission is low – e.g. in the plasma centre where

the neutral beam density is low due to beam attenuation – these non-CX emissions can
swamp active CX signal and thus make the measurement of TC , nC and Ωφ,C virtually
impossible. Below, we give a short overview of some of these non-CX emissions.
Halo and plume In a charge exchange interaction a plasma ion receives an extra elec-
tron. After some the time it will of course lose this electron again. However, if the
ionisation time is long enough, the plasma ion can be excited again by electron impact.
This causes a delayed, secondary emission. This secondary emission is especially impor-
tant for low Z ions, like hydrogen and helium, because they don’t have to be excited to a
high level in order to emit visible light. For C 5+ the electron impact should excite the ion
from the ground state n = 1 to the n = 8 state in order to have a secondary emission at
529 nm. The chance for such an excitation by electron impact is negligible.
If the plasma ion remains charged after the CX interaction, it still moves along the
magnetic field and the delayed emission is seen as a ‘plume’. If the plasma ion is neutralised
in the CX interaction, e.g. when H + or He+ undergoes a CX reaction, it is no longer
confined to the magnetic field and it can escape in all possible directions. The delayed
emission is therefore called ‘halo’.
Impurity lines In the edge of the plasma the temperature is too low to fully ionise all
particles. The non-fully stripped ions – usually impurity ions – can be excited by electron
impact and thus emit line radiation. If the wavelength of these impurity lines overlaps with
the wavelength of the CX emission, the analysis of the CX spectrum is difficult. Getting
rid of these unfavourable impurity lines can be done in the same way as the passive CX
emission is treated: by beam modulation or modelling of the impurity emission.
Impurity emission is however not always an annoyance. If the wavelength of the
impurity line does not overlap with the CX emission but lies within the vicinity, the
impurity line can be used for real-time wavelength calibration. Line-emission by impurity
ions comes from the very edge of the plasma. There the temperature is very low, resulting
is a narrow emission line who’s peak position can be easily determined. Also the plasma
rotation in the very edge of the plasma is negligible, such that the wavelength of the
impurity emission is the natural wavelength λn . The measured pixel position of this
impurity line on the camera, together with the known λn , is then used to derive the
relation between the pixels on the camera and the wavelengths.
Continuum emission The whole spectrum of active and passive CX emission and
impurity line-emission is superposed onto a background of continuum emission. This con-
tinuum emission is Bremsstrahlung due to the relative movement of the charged particles
in the plasma. The formula for the intensity of Bremsstrahlung in the visible region is
[88]:
dI ḡn2 Zef f
= 3 × 10−21 e√ , (4.10)
dλ λ Te
where Te is in keV√ , ne in m−3 and λ in nm. ḡ is called the Gaunt factor and can be
approximated by 2 3/π. It shows that the intensity of the Bremsstrahlung can be used
to determine Zef f , i.e. the purity of the plasma. If carbon is the major impurity in the
plasma, then Zef f ≈ 1 + ZC (ZC − 1)nC /ne . Combining the Zef f from Bremsstrahlung and
Zef f from the intensity of the carbon CX emission, an intensity calibration of the system
can be performed.

The continuum radiation also depends on the square of the density. This can pose
a diagnostic problem in fusion reactors with a high density like ITER. For a TEXTOR
plasma typical values for Te , ne are 1 kev and 2 . 1019 m−3 . The resulting continuum
intensity is quite small, as can be seen in figure 4.2. For ITER both the density and the
temperature are much higher: Te ≈ 10 keV and ne ≈ 1020 m−3 . This means that the
local intensity of Bremsstrahlung is eight times higher in ITER. Also the length of the
line-of-sight through the plasma is about four times larger in ITER than in TEXTOR. As
a result the line-integrated continuum intensity is 30 times higher in ITER than it is in
TEXTOR. Because Bremsstrahlung is a statistical process the noise level goes with the
square root of the signal, resulting in a five times higher noise level in ITER compared to
TEXTOR. This continuum noise contaminates the active CX signal. When we take into
account that the beam attenuation in ITER is stronger than in TEXTOR and that the
effective emission rate hσvi is lower for the 100 kV beams at ITER than it is for the 50 kV
beam at TEXTOR (see figure 4.3), we expect the active CX signal to be much lower in
ITER than in TEXTOR. Therefore in ITER the high noise due to the high Bremsstrahlung
intensity could possibly drown the low active CX signal.
4.4 CXRS at TEXTOR

There are two main CXRS systems at TEXTOR. One system uses the NBI1 heating beam
as a source of neutrals. We will therefore refer to it as the NBI-CXRS system. The other
system uses a low power, radially injecting, diagnostic neutral beam. This system is called
RUDI1 . The NBI-CXRS system has lines-of-sight tangential to the toroidal direction and
is mainly used for ion temperature, carbon density and toroidal rotation measurements.
The RUDI system has poloidal lines-of-sight. Apart from ion temperature and carbon
density, it measures poloidal rotation. In this thesis we are mainly interested in the
toroidal plasma rotation. Therefore most plasma rotation data presented in this work
comes from the NBI-CXRS system.
NBI-CXRS system The NBI-CXRS system has three sets of lines-of-sight. They are
plotted in figure 4.8. These lines-of-sight are all located in the equatorial plane and are
directed tangentially to the flux surfaces, which means that the measured Doppler shifts
can be translated in toroidal rotation frequencies Ωφ . Two sets of lines-of-sight – the red
core system and the blue the edge system – observe the plasma in the counter-direction
and one set – the green lines in figure 4.8 – looks in the co-direction.
The NBI-CXRS system uses the heating beam NBI1 as a source of neutral particles.
Because this is a high power beam, there is a large neutral density up to the plasma centre
and the resulting CX signal is high. This allows us to reduce the exposure time of the
camera and hence have a reasonably fast time resolution. In normal operating conditions
a time resolution of 50 ms is used. Faster sampling is possible – up to 10 ms – but the
quality of the CX spectra reduces significantly with shorter exposure times.
A disadvantage of the heating beam is its large width. It was said previously that
the CXRS measurement is a local measurement as long as the line-of-sight is tangential
to a flux surface where it crosses the neutral beam. When a wide neutral beam is used,
this condition is difficult to reach. The neutral particle density of NBI1 in TEXTOR has
1
this stands for RUssian DIagnostic, which immediately reveals its origin.
60 Section 4.4 - CXRS at TEXTOR

(a) (b)
0
−0.5
−1
Y (m)
−1.5
−2
−2.5
0 0.5 1 1.5 2 2.5 1.7 1.8 1.9 2 2.1 2.2

X (m) R (m)
Figure 4.8 : The NBI-CXRS setup at TEXTOR. The left figure shows a top view of
TEXTOR with the three sets of lines-of-sight (core in red, edge in blue and ). The
gray area represents the full width of NBI1 at half maximum. In the right figure an
arbitrary CXRS profile is plotted, with the radial resolution. The horizontal lines
indicate the FWHM of the neutral beam over the lines-of-sight.
a Gaussian shape perpendicular to the beam axis. In the equatorial plane, where the
lines-of-sight are located, the full width half maximum (FWHM) is in the order of 20 cm.
The grey area in figure 4.8 indicates the FWHM of NBI. In an optimised geometry the
major radius of the flux surface where the line-of-sight enters the beam – i.e. where the
neutral beam density is at half maximum – does not differ from the major radius of the
flux surface where the line-of-sight crosses the centre of the beam and the major radius
of the flux surface where the line-of-sight leaves the neutral beam again. In that case the
measurement is truly local. In reality the line-of-sight does cross several flux surfaces when
it goes through the neutral beam. As a result the CX signal is the line integrated emission
over the crossed flux surfaces. The range of flux surfaces that are crossed determines the
radial resolution of the CXRS system. In figure 4.8 (b) an arbitrary Gaussian profile is
shown to indicate the radial resolution of each of the sets of lines-of-sight. It is clear
from this figure that the co-observing system and the edge system yield the best radial
resolution. These two systems were however both in development when the experiments
for this thesis were done. The CXRS data presented in this work therefore comes only
from the core, counter-observing system.
RUDI system Apart from the NBI-CXRS system, there is also a CXRS system installed
at TEXTOR that uses a low power, radially injecting, diagnostic neutral beam: the RUDI-
CXRS system. The lines-of-sight of this system are radially as well, which means that
the RUDI system measures poloidal rotation. Of course also the ion temperature and the
carbon density are diagnosed with this system.
Because a low power beam is used, the influence on the plasma is minimal. The RUDI
beam does not drive rotation and it hardly heats or fuels the plasma. This makes the
RUDI system suitable for measurements of TC , nC and vθ,C in e.g. Ohmic discharges. The
Section 4.4 - CXRS at TEXTOR 61

RUDI beam is also modulated. The passive CX component and the impurity lines are thus
removed from the spectrum by subtracting the spectrum measured during the off-phase
of the beam from the spectra obtained during the on-phase of the beam.
The disadvantage of a low power, diagnostic beam is the low CXRS signal. As a result
a long exposure time is needed to collect enough signal. For RUDI the time resolution is
rather poor: in the order of 1 s.
In this thesis the main part of the data comes from the NBI-CXRS system, that
measures toroidal rotation. The poloidal rotation measurements from the RUDI system
are only used on a few occasions.
Spectrometers, cameras and calibration As mentioned above the NBI-CXRS sys-

tem has three sets of lines-of-sight. The light of each of these sets goes through a bundle of
optical fibres to a spectrometer. The spectrometers are located well away from TEXTOR.
The spectrometers that are used are of the Littrow type. In figure 4.9 the spectrometer
setup is drawn schematically. The light comes from TEXTOR through an optical fibre
bundle and is focussed with two lenses on the entrance slit of the spectrometer. The
entrance slit is either 50 µm, 100 µm or 300 µm. The width of the entrance slit is important
because it greatly determines the width of instrument function. A narrow entrance slit
gives a narrow instrument function, but also reduces the intensity of the light.
Littrow spectrometer
camera
lens grating
movable mirror wavelength calibration lamp
Fibre bundle to TEXTOR
Figure 4.9 : The spectrometer setup for the NBI-CXRS system at TEXTOR. The light
comes from TEXTOR through the a fibre bundle. It is focussed on the entrance
slit of the spectrometer. In the spectrometer a large lens assures a full illumination
of the grating by a parallel bundle of light. The dispersed light is finally imaged,
through the same lens, on the camera.
Behind the entrance slit there is a small mirror that reflects the light approximately
90◦ .Subsequently the light goes through a large lens that causes the grating at the back
of the spectrometer to be fully illuminated by a parallel bundle of light. The grating of the
62 Section 4.4 - CXRS at TEXTOR

spectrometers that we used has 1200 lines per mm. The light reflected from the grating
is imaged on the camera. The wavelength of the reflected light depends on the number
of lines on the grating g, the angle of the grating γ and the order in which you want to
measure m. To calculate the dispersion – ‘wavelength per meter’ at the position of the
camera – also the focal length f of the large lens is needed. For the used spectrometer
this is f = 750 mm. The imaged wavelength and dispersion are then:
2 m sin γ
λ = (4.11)
g
s
mλg 2

dλ 1
= 1− (4.12)
dx mf g 2
For CXRS emission of carbon (λ = 529 nm) and a measurement in first order (m = 1),
the dispersion is: dλ/dx = 1 nm/mm.
Two types of cameras are used at TEXTOR:
• The Wright cameras have a CCD with 1152 × 298 pixels. Only half of the chip
(576 × 298) is illuminated. The pixels have an area of 22.5 µm × 22.5 µm. The
camera is positioned such that the narrow side (298 pixels) is in the direction of the
wavelengths. Along the other direction (576 pixels) the different fibres are imaged.
In this direction we can bin rows of pixels that are illuminated by the light of the
same fibre. The output of the Wright camera than consists of n spectra – n being the
number of fibres. Each spectrum has 298 data points, representing the wavelengths
525.5 nm to 532.5 nm with a dispersion of 0.0237 nm/pixel (for carbon).
The read-out time of the camera depends on the binning of the pixels. For 10 fibres
the fastest read-out time is 30 ms. This limits the time resolution of the CXRS
system.
• The other camera is a PixelVision camera. The active region of the CCD chip is
655 × 496 pixels. The pixels are 11.8 µm × 11.8 µm. Here the camera is positioned
such that the wide side (655 pixels) is in the wavelength direction. Again binning is
possible along the other direction – the direction of the different fibres. The output
of the PixelVision camera also consists of n spectra – n being the number of fibres.
Each spectrum now has 655 data points, representing the wavelengths 524.9 nm to
533.1 nm with a dispersion of 0.0124 nm/pixel.
The PixelVision camera is faster than the Wright camera. It can read out the
spectra of 20 fibres in 8 ms. The PixelVision camera can therefore reach a higher
time resolution, if the intensity of the CX light allows it.
The spectra that are the output of the cameras, give the intensity per pixel, not per
wavelength. The properties of the spectrometer (γ, g, f and m) give us an idea about
the pixel to wavelength conversion, but this is far from accurate. A good wavelength
calibration is therefore needed. To do so, we use a wavelength calibration lamp that emits
a spectrum of which the wavelengths are well documented. Ideally this calibration lamp
should be put inside TEXTOR, such that the light of the calibration lamp follows the
same path as the light from a plasma discharge would. This is however only possible
during major service-shutdowns of TEXTOR, when the TEXTOR vessel is accessible.
Section 4.4 - CXRS at TEXTOR 63

Instead of placing the calibration lamp inside TEXTOR, we put the calibration lamp
close to the spectrometer. In between TEXTOR discharges a mirror can be placed in
the optical beam path of the CXRS setup, as is shown in figure 4.9. The light of the
calibration lamp is than focussed on the entrance slit of spectrometer and a calibration
spectrum is recorded. Because the wavelengths of the lines of the calibration lamp are
known, we can determine the pixel to wavelength conversion from this calibration spec-
trum. The conversion can then be applied to the spectra of the plasma discharges. This
only works however if the optical axis of the ‘calibration lamp – spectrometer’ system
exactly coincides with the optical axis of the ‘TEXTOR – spectrometer’ system. Because
there is always some deviation between these optical axes, we empirically found that it is
not possible to increase the accuracy above ±10 µm at the camera position. For carbon
this corresponds with systematic wavelength error of ±0.01 nm or a systematic error in
the velocity measured by Doppler shift of ±5.5 km/s. This error is the same for every
position and every time frame, such that rotation gradients and the time evolution of the
rotation are not influenced by this inaccuracy.
For the intensity calibration a integrating sphere is used. The intensity of the light
emitted in every direction by such an integrating sphere is known for every wavelength.
The sphere is put inside TEXTOR. The intensity measured by the camera can then be
compared to the intensity emitted by the integrating sphere. Because the sphere has to
be placed inside TEXTOR, an intensity calibration can only be done during a service-
shutdown.
4.5 Summary
We introduced the technique of charge exchange recombination spectroscopy. This tech-
nique analyses the light emitted when a fully stripped impurity ion receives an electron
form a neutral particle. The Doppler shift of this line-emission is proportional to the
plasma rotation velocity. Apart from plasma rotation, also the ion temperature and the
impurity density can be determined via CXRS.
A charge exchange spectrum consists of an active CX part, caused by the neutral
beam, passive CX emission and impurity emission from the edge. The most interesting
data comes from the active CX light, which means that we have to filter the passive and
impurity light out of the spectrum. This can be done by beam modulation or by modelling
the passive and impurity emission.
The effective emission coefficient of the CX light depends on the collision velocity
between the neutral particles and the plasma ions. This means that ions moving with a
different velocity will emit light with a different intensity. Because the plasma ions have a
thermal velocity distribution, the measured temperature and rotation may therefore differ
from the true temperature and rotation. For low temperatures (order 1 keV ) the difference
can be neglected. For high temperatures this effect can be important. One can however
outsmart this effect by choosing the lines-of-sight of the CXRS system perpendicular to
the neutral beam. This is e.g. the case for the core CXRS system at ITER.
Zeeman splitting and l-state mixing are causes of non-thermal broadening of the CX
spectrum. To take this non-thermal broadening into account the ideal CX spectrum has to
be convoluted with a Gaussian of which the width is determined by the Zeeman splitting
and l-state mixing. Non-thermal broadening makes it difficult to measure temperatures
below 100 eV with CXRS.
64 Section 4.5 - Summary

The light emitted by impurities can be annoying, when it is located at the same
wavelength position as the CX emission. But if the impurity emission is located in the same
wavelength range but not at the same wavelength position it can be used for wavelength
calibration.
The continuum radiation is Bremsstrahlung. Combining the Bremsstrahlung with the
intensity of the CX emission the impurity density can be determined without the need
for an intensity calibration. The Bremsstrahlung can however also be malignant. For
large machines with a high plasma density, like ITER, the Bremsstrahlung is very high.
The noise on the continuum radiation goes with the square root of the intensity and will
therefore be high as well. A high enough active CX signal is necessary in ITER, because
it will otherwise drown in the continuum noise.
In TEXTOR there are two CXRS systems. One uses the heating beam NBI1 as a source
of neutrals, the other system – RUDI – has its own diagnostic neutral beam. The NBI-
CXRS system is capable of measuring the toroidal rotation, the RUDI system measures
poloidal rotation. Because of our interest in toroidal rotation, mainly the NBI-CXRS
system was used for this work. There are three sets of lines-of-sights for the CXRS-
NBI system, but only one of these sets – the 9 channel core system – is used for the
measurements in this thesis. The time resolution of this system is 50 ms. The limited
accuracy of the wavelength calibration for this system, results in a systematic error in the
rotation velocity of ±5.5 km/s. Because this error is systematic the rotation gradients
and the time evolution of the rotation are not influenced by this inaccuracy.
Section 4.5 - Summary 65

Chapter 5
Plasma rotation at TEXTOR
5.1 Introduction
In chapter 2 several statements on the rotation of a plasma were made:
• The momentum confinement time τφ scales with the energy confinement time τE .
• Plasmas rotate even when no external momentum input is applied.
• Impurities rotate at a different speed than the main plasma ions.
• The frequency of MHD modes is the toroidal rotation frequency plus the electron
diamagnetic drift frequency.
The experimental verification of this is given here for the TEXTOR case.
5.2 Momentum and energy confinement time in TEXTOR

In several tokamaks it is experimentally found that very often the momentum confinement
time τφ is strongly related to the energy confinement time τE (see section 2.3.1). Especially
when the ion energy confinement time is compared to the momentum confinement time, it
is observed that for most steady-state discharges, with densities above 1019 m−3 , τφ = τEion .
In transient phases and at low densities (n < 1019 m−3 ), the relation τφ = τEion does not
hold [86].
Figure 5.1 shows that also in TEXTOR τφ and τE are coupled. The momentum
confinement time τφ was calculated for shots with a known momentum input by the
neutral beams:
RNBI1 PNBI1 RNBI2 PNBI2
TNBI = − , (5.1)
vNBI1 vNBI2
with RNBI the impact radius of the neutral beam, PNBI the neutral beam power and vNBI
the velocity of the neutral beam particles.
The total angular momentum of the plasma, Lφ , is calculated on the basis of rotation
and density measurements during these shots. The resulting momentum confinement time
τφ then is:
Lφ
τφ = . (5.2)
TNBI
67
Chapter 5 - Plasma rotation at TEXTOR
The energy confinement time τE = Edia /Ptotal is calculated with the measured total
power input and the stored plasma energy measured with the standard diamagnetic loop
diagnostic. Due to poor diagnostic coverage it was not possible to determine the ion energy
confinement time τEion in TEXTOR.
0.06
Figure 5.1 : The measured momentum con-
measured τφ (s)
finement time τφ is plotted against the
measured energy confinement time τE . 0.04
Just like in other tokamaks, we find that
τφ scales with τE . Within the error of
the measurement we can even say τφ =
0.02
τE . This allows us to determine an un-
known torque Tφ if τE and the rotation
L
are measured: Tφ = τEφ .
0
0 0.02 0.04 0.06
measured τE (s)
The fact that also in TEXTOR τφ ≈ τE allows us to determine the momentum input Tφ
from the measured Lφ and the measured τE . This is interesting when the total momentum
input is not known. The dynamic ergodic divertor at TEXTOR, for example, is expected
to exert a torque onto the plasma. The total momentum input of the plasma during the
application of the DED is the sum of the known TNBI and the unknown TDED . This total
momentum input is also Lφ /τE . Hence the total torque exerted by the DED that we find
is: TDED = Lφ /τE − TNBI .
5.3 Ohmic rotation in TEXTOR

In chapter 2 we indicated that plasmas rotate even when no external momentum input is
applied. In present day tokamaks the plasma rotation is mainly driven by neutral beams
and the small fraction of spontaneous rotation is usually neglected. In ITER neutral beams
will not be able to drive a large amount of toroidal momentum, due to large plasma mass,
and the spontaneous rotation will become very important. It is therefore worthwhile to
look into spontaneous rotation, even in a medium-size tokamak like TEXTOR.
We will apply the neoclassical expressions given in section 2.4.3 to the TEXTOR
case in order to get an estimate for the spontaneous rotation in TEXTOR. According
to neoclassical theory, the poloidal rotation is proportional to the gradient of the ion
temperature. For the toroidal rotation several different expressions are available in the
literature. The expressions for the toroidal velocity are not ‘closed’: they are given as a
derivative of the toroidal velocity, or an assumption on the radial electric field has to be
made.
In figure 5.2 the density, temperature, magnetic field and the collisionality in an Ohmic
TEXTOR discharge are given. It is seen that over a large part of the plasma the collision-
ality is low and the transport regime of interest is the banana regime. Only in the edge
of the plasma, the expressions for transport in the plateau regime must be used. We use
the collisionality factor K1 (ν ∗ ) as it is given in [49].
68 Section 5.3 - Ohmic rotation in TEXTOR

B B
Bφ (T) / Bθ (0.1 T) / q 4 q φ θ
−2
4
T (keV) / n (1019 m−3)
T = Ti = Te n = ni = ne
3
0
8 Pfirsch−Schlüter regime Figure 5.2 : The magnetic fields, the den-
Collisionality ν*
sity and temperature of a typical Ohmic

6 discharge in TEXTOR are given. From
4 Plateau regime
this the collisionality follows. This colli-
sionality defines which transport regime
2 has to be used when the spontaneous neo-
Banana regime classical rotation is calculated.
0
1.4 1.6 1.8 2 2.2
R (m)
The sign of K1 determines the direction of the poloidal rotation. In the banana regime
the poloidal rotation is in the ion diamagnetic drift direction and in the Pfirsch-Schlüter
regime the rotation is in the electron diamagnetic drift direction. In the plateau regime the
poloidal rotation goes from the ion to the electron diamagnetic drift direction. Because
the largest part of the TEXTOR plasma is in the banana regime, the neoclassical poloidal
rotation in TEXTOR is in the ion diamagnetic drift direction. The expected poloidal
rotation in TEXTOR is also very low. At its maximum it reaches only 0.8 km/s. The
poloidal rotation is given by the blue line in figure 5.3.
For the toroidal neoclassical velocity there are several expressions. The expression
given by equation (2.45) in chapter 2 was found in [14]. Here, not the toroidal velocity
vφ is given, but the derivative dvφ /dr. We made the assumption that the vφ is zero at
the plasma edge and that equation (2.45) is valid over the whole plasma. The resulting
toroidal velocity is in the counter-current direction and has a maximum velocity of 6 km/s.
The toroidal velocity profile according to [14] is drawn by the solid red line in figure 5.3.
Although the assumption vφ = 0 at the edge is reasonable, one should be careful when
stating that equation (2.45) is valid over the whole plasma. The equation considers a pure
ion-electron plasma, while in the edge neutral and not fully ionised particles are present
as well. It also assumes the plasma to be confined in nested flux surfaces, but this picture
breaks down at the plasma edge.
If we use the expression (2.46) found in [49] we have a local expression of vφ . The as-
sumption we have to make here lies in the radial electric field Er . A possible assumption
is Er = 1/eni dpi /dr, which basically states that the electric field, by balancing the force
due to the ion pressure gradient, prevents the loss of ions. If we substitute Er into equa-
tion (2.41), we see that this assumption also means that we assume a parallel plasma flow.
Section 5.3 - Ohmic rotation in TEXTOR 69

Equation (2.46) then returns: vφ = (K1 /eBθ ) dTi /dr, which together with the poloidal ro-
tation vθ = (K1 /eBφ ) dTi /dr indeed yields a parallel rotation vk = (K1 B/eBθ Bφ ) dTi /dr.
In figure 5.3 vφ , according to [49] and with assumption of a parallel flow, is given by the
dashed red line. Because vθ is in the co-direction and the total plasma flow is parallel, vφ
is counter-rotating. The toroidal rotation velocity in the plasma centre is 8 km/s.
8000 Poloidal velocity (vθ)

Toroidal velocity (vφ) according to [14]
Toroidal velocity (vφ) according to [49]
4000
Toroidal velocity from MHD frequency
v (m/s) / v (m/s)
Ohmic sawtooth precursor
0
φ
−4000 Banana regime

θ
Plateau regime
Plateau regime
−8000
−12000
1.4 1.6 1.8 2 2.2
R (m)
Figure 5.3 : Poloidal and toroidal velocity profiles calculated for an Ohmic discharge at TEX-
TOR (Ip = 400 kA, Bφ = 1.9 T , T and n as given in figure 5.2).
The poloidal rotation is in the ion diamagnetic drift direction (i.e. co-rotation) due to the
fact that almost the whole plasma is in the banana regime.
For the toroidal rotation we used two different theoretical expressions. If we follow the
approach of [14] and assume vφ = 0 in the edge, the toroidal rotation in the centre will be in
the counter direction. Again due to the banana regime.
If we use the expression given in [49] and use the assumption that Er = 1/eni dpi /dr, we
find a parallel flow. The resulting vφ is in the counter-direction.
An indirect measurement (red star), using the MHD frequency of the sawtooth precursor
in an Ohmic discharge, shows that toroidal rotation in the plasma centre is indeed in the
counter-direction.
The NBI driven rotation is a factor 5 to 10 higher than the expected spontaneous rotation in
TEXTOR.
The NBI-CXRS system at TEXTOR can not be used to compare the theoretical ex-
pectation with a measurement of the rotation velocity, because this system uses a neutral
beam. Neutral beam driven plasmas reach rotation velocities up to 100 km/s. This is a
lot more than the expected spontaneous rotation (6 to 8 km/s) and therefore, whenever
the neutral beam is used, the plasma rotation velocity is dominated by the beam driven
rotation.
We can however use MHD frequency measurements in TEXTOR to derive the
toroidal rotation velocity at a specific q surface (see section 5.5). When we measure
the MHD frequency of the sawtooth precursor, located at the q = 1 surface, we can
measure the central toroidal rotation velocity in an the Ohmic discharge. This measured
rotation velocity is 9 km/s in the counter direction, which is consistent with the expected
spontaneous toroidal rotation.
70 Section 5.3 - Ohmic rotation in TEXTOR

The subject of spontaneous rotation is a very important issue in plasma physics, due to
its relevance for the next generation of fusion devices like ITER. To get a neoclassical value
for this rotation, quite large assumptions have to be made. On top of that the neoclassical
theory does not take turbulence into account. It is however expected that turbulence is
a key parameter in the description of spontaneous rotation. This section served as an
illustration on the amount of spontaneous rotation we could expect in TEXTOR. For
a detailed discussion on spontaneous rotation we would like to refer to the literature
[17, 41, 62, 69].
5.4 Impurity rotation in TEXTOR

In TEXTOR the plasma rotation is measured using charge exchange recombination spec-
troscopy. The line-emission that is analysed is emitted by carbon ions. The measured
velocity therefore is the bulk velocity v C of the carbon fluid. What we are actually inter-
ested in is the plasma velocity v; i.e. the centre of mass velocity of all plasma species. For
low impurity densities, the plasma velocity v can be approximated by the fluid velocity of
the main plasma ions. In the case of TEXTOR the main ions are deuterium ions. Zef f ,
which is a measure for the impurity content, is usually about 2. The plasma velocity is
then given by v = 0.7v D + 0.3v C .
In section 2.5.1 the neoclassical expression for the difference between the velocity of the
impurity ions and the velocity of the main ions is given. For TEXTOR we can calculate
the difference ∆v D,C = v D − v C from equation (2.48). The relation between the measured
v C and the plasma velocity v then is: v = v C + 0.7∆v D,C .
According to the neoclassical theory the difference between the toroidal rotation of
the carbon ions and the plasma fluid is large for plasma discharges with a low plasma
current and a peaked temperature profile. For TEXTOR the plasma current is usually
around 400 kA, which is not particulary low, and the temperature is not extremely high
or peaked (central temperatures in the order of 1 keV ). The difference between vφ,C and
vφ is therefore expected to be low.
In figure 5.4 the measured carbon toroidal rotation frequency and the resulting plasma
rotation frequency when using (2.48) is plotted for a typical TEXTOR, beam driven
discharge. The magnetic field and plasma current for the presented discharge are the
same as those given in figure 5.2; the electron temperature (Tcentral = 1.5 keV ) and
density (ncentral = 5 · 1019 m−3 ) are higher due to the fact that this is a beam driven
plasma, not an Ohmic discharge. The beam used was NBI1, which injected 600 kW of
power in the co-direction.
The calculated difference between the carbon rotation velocity and the plasma rotation
velocity is approximately 2 km/s = 103 rad/s. It is directed in the co-direction. We could
choose to neglect this correction, because it is very close to the statistical error that is made
when the CXRS spectra are fitted. The correction is, however, systematic and therefore
must be taken into account.
In chapter 4 it was seen that the wavelength calibration of the CXRS system only
allows for a measurement accuracy of about 5 km/s. This means that the plasma rotation
velocity lies within the systematic error of the carbon rotation velocity measurement.
Instead of trying to get a better calibration for the carbon rotation velocity and then
apply the neoclassical correction to get the plasma rotation velocity, usually a different
Section 5.4 - Impurity rotation in TEXTOR 71

4
x 10
2.5
1.5
Ωφ (rad/s)
1
0.5
1.75 1.8 1.85 1.9 1.95 2 2.05

R (m)
Figure 5.4 : The measured carbon rotation frequency plotted together with the statistical fitting
error (dotted line and error bars). The solid line represents the profile of the plasma rotation,
that was calculated with eq. (2.48). It shows that the plasma rotation is higher, or more
precisely is more in the co-current direction, than the carbon rotation. The correction is
low: ∆Ωφ ≈ 103 rad/s, which is of the order of the statistical error and lower than the
calibration error for the CXRS system (indicated by the shaded area). Nevertheless, the
correction should be taken into account because it is systematic.
approach is used at TEXTOR. As will be shown in the next section the frequency of
MHD modes is directly linked with the toroidal plasma velocity. From the measurements
of the MHD frequency the plasma velocity at one location – the q surface of the mode
– can be determined. By calculating the difference, the carbon rotation velocity at this
location is derived from the plasma rotation velocity. This gives us a accurate calibration
of the carbon rotation velocity profile. Subsequently we use the now calibrated profile of
the carbon rotation velocity and the neoclassical difference between carbon and plasma
rotation velocity, to determine the profile of the plasma rotation velocity.
5.5 MHD rotation in TEXTOR

In both previous sections we mentioned that the frequency of MHD modes, measured with
e.g. ECE, can be used to determine the toroidal plasma rotation at the q surface where
the mode is located. This allows us to figure out the direction of the rotation in Ohmic
discharges and to calibrate the measured carbon rotation velocity.
The relationship between the toroidal rotation velocity and the MHD frequency is
given in section 2.5.2: fMHD = (nvφ )/(2πR) + fe∗ . In TEXTOR the electron diamagnetic
drift is in the counter-direction, which means that fe∗ < 0. Because, in the literature, the
diamagnetic frequency is always given as a positive number, we will do so as well in this
thesis. With fe∗ positive the above formula for TEXTOR becomes:
nvφ
fMHD = − fe∗ . (5.3)
2πR
When both fe∗ , calculated with equation (2.54), and fMHD , measured with ECE or a
72 Section 5.5 - MHD rotation in TEXTOR

magnetic diagnostic, are known, then the toroidal velocity is known as well.
The validity of the above equation has already been checked in [85]. We have taken
over the results from [85] here as an illustration.
In figure 5.5, for a number of discharges, the MHD frequency of the sawtooth precursor
vφ
(m/n = 1/1) is plotted against the toroidal rotation frequency (fφ = 2πR ) at the q = 1
surface. Positive frequencies correspond to co-rotation, negative frequencies to counter-
rotation. In these discharges the electron temperature (Te = 1.2 keV ), the magnetic field
(Bφ = 2.25 T ), q at the edge (qa = 4) and q at the magnetic axis (q0 ≈ qa /(qa + 1) = 0.8)
where kept constant, such that the diamagnetic frequency calculated with (2.54) is also
constant: fe∗ = 3.6 kHz.
Figure 5.5 : The MHD frequency of the 1/1 sawtooth precursor plotted against the toroidal ro-
vφ
tation frequency fφ = 2πR . It shows a linear relationship: fMHD = fφ − 2.0 kHz.
This plot proves that the expression fMHD = nfφ − fe∗ is valid. Thus that fMHD measurements
can be used to determine vφ .
The offset of the line gives us the local fe∗ . The value of the local fe∗ is observed to be lower
than the fe∗ that is calculated with equation (2.54). A lower local pressure gradient, due to
the presence of modes itself, is probably responsible for this.
With the local fe∗ known, the measurement of fMHD during an Ohmic discharge gives us the
toroidal rotation in an Ohmic discharge. This point is indicated by Ω. The corresponding
toroidal rotation velocity is 104 m/s in the counter-direction. [85].
According to equation (5.3) we expect to see a straight line when we plot the precur-
sor frequency against the toroidal rotation frequency. Figure 5.5 indeed shows a linear
relationship between the measured fφ and the measured fMHD . We also expect the slope
of this line to be 1, because n = 1 for the sawtooth precursor. Linear regression analysis
of the data points learns that the slope is 0.998, which is consistent with this.
For fφ = 0 we expect to see fMHD = fe∗ . Figure 5.5 shows us then that fe∗ = 2.0 kHz.
The calculation of fe∗ with equation (2.54) resulted in fe∗ = 3.6 kHz. This is higher
than the measured fe∗ . For the calculation of fe∗ the local pressure gradient is needed.
In equation (2.54) this local pressure gradient was derived by assuming natural density
and temperature profiles. The presence of modes can however locally change pressure and
density profile. Over the O-point of a magnetic island the pressure gradient is zero, only
through the X-point there is a pressure gradient. The averaged pressure gradient over the
Section 5.5 - MHD rotation in TEXTOR 73

island region could therefore be lower than the one that follows from the natural pressure
profile. A lower fe∗ , typically a factor two, would result from that.
From figure 5.5 also the toroidal rotation at the q = 1 surface during an Ohmic
discharge can be determined. The precursor frequency during an Ohmic discharge
was measured to be 3.0 kHz. Because this frequency decreased when the co-beam
(NBI1) was switched on, we find that the precursor rotated in the counter-direction:
fMHD = −3.0 kHz. With the local fe∗ = 2.0 kHz given, equation (5.3) yields a toroidal
frequency of fφ = −1.0 kHz, or a rotation velocity of vφ = −104 m/s at the q = 1 surface.
This is a rotation in the counter-direction. It is indicated by the Ω-mark in figure 5.5 and
by the red star in figure 5.3.
The work presented in [85], and repeated here, shows that measurements of the MHD
frequency of modes is a very valuable addition to the rotation measurements with CXRS.
5.6 Conclusion
In this chapter we experimentally verified some of the topics presented in chapter 2 for
the specific case of the TEXTOR tokamak.
First of all we found that for steady state TEXTOR discharges τφ ≈ τE . This allows us
to use τE instead of τφ for discharges where τφ can not be determined from measurements;
e.g. when we want to determine the momentum input by the DED.
Secondly we tried to find out how much spontaneous rotation is expected in TEXTOR
based upon the neoclassical approach of [14] and [49]. The expected spontaneous, toroidal
rotation velocity is approximately 6 to 8 km/s in the counter-direction. In TEXTOR,
CXRS can not be used to measure the toroidal rotation in an Ohmic discharge, but a
measurement of the sawtooth precursor frequency allowed us to derive a central toroidal
rotation velocity of 9 km/s in the counter-direction, which is consistent with the expec-
tation. The velocity of the spontaneous rotation is an order of magnitude smaller than
the beam driven velocities. Therefore, in TEXTOR spontaneous rotation is usually neg-
ligible. However, it is interesting to study the phenomenon of spontaneous rotation in a
quantitative manner, as it will be a very important effect in ITER.
The toroidal rotation velocity of the carbon ions, that is measured with CXRS, differs
about 2 km/s with the toroidal plasma rotation. This difference is of the order of the
statistical error when fitting a CX spectrum, but has to be taken into account because it
is a systematic deviation: the plasma rotation is always more in the co-direction than the
carbon rotation. Together with the measurement of the MHD frequency, the correction
on the carbon velocity can be used to get an absolute calibration of the plasma velocity
profile.
A final topic was the relationship between the toroidal velocity and the frequency of
MHD modes. The validity of fMHD = nfφ − fe∗ was tested. With the above expression
the measurement of the MHD frequency at one position – the rational q surface – gives
us the toroidal plasma rotation at that location. This value can then be used to calibrate
the velocity profile measured with CXRS. The measurement of fMHD also showed that the
plasma rotates in the counter-direction during an Ohmic discharge.

Chapter 6
Measurements of plasma rotation

during DED operation
6.1 Introduction
Whereas most tokamaks in the world have a fixed error field, the ergodic dynamic divertor
(DED) on TEXTOR allows us to manipulate the error field. We will therefore not call it
an error field anymore but a controlled perturbation field.
In the introduction we mentioned that error or perturbation fields can excite MHD
modes. Because these modes degrade the plasma confinement and stability – they can
even lead to a disruption – we want to avoid their excitation. The best way to know how
to avoid them, is to know how to excite them. The controlled perturbation field of the
DED will be used to carefully study the threshold at which mode excitation occurs.
In this chapter we try to find an answer to the questions raised in the introduction:
• How does the application of a perturbation field change the plasma rotation?
• How does the excitation of tearing modes depend on the plasma rotation velocity?
In the literature the prediction is made that the plasma rotation velocity changes with
increasing perturbation strength [28, 55]. This change in rotation velocity is expected
to be monotonic towards the plasma rotation velocity for which the MHD modes are at
rest in the frame of the perturbation field. Once this plasma velocity is reached, the
MHD modes grow and form large magnetic islands: the so-called mode excitation. To test
these hypotheses, we carried out a research programme in which we applied a systematic
variation of DED perturbation level and the momentum input by the neutral beams. The
rotation velocity during the application of the DED and the threshold for mode excitation
were measured. The results of these measurements are then compared to the predictions
that are found in the literature.
6.2 A general overview

A linear increase of the current in the DED coils causes a linear increase of the perturbation
field. And, as said in the introduction, we expect to see a change in the plasma rotation
when the perturbation is increased.
75
Chapter 6 - Measurements of plasma rotation during DED operation
In figure 6.1 time traces of the toroidal rotation frequency are given during a linear
ramp up of the DED current in three different modes: AC− (co), DC (static) and AC+
(counter). In figures 6.2 to 6.4 the evolution of the toroidal rotation profile is given during
the DED current ramp up, for all three modes of operation (AC− , DC and AC+ ). For
the DC and AC+ case, one can observe two phases: a phase where the rotation frequency
increases slightly with increasing DED current and where the shape of the rotation profile
is conserved; and a phase where the rotation profile is flattened and the frequency stays
constant even when the DED current is further increased. When the DED is operated in
AC− , only the first phase of increasing rotation frequency with increasing DED strength
is observed.
4
x 10
4 Co−rotating DED field (AC−)

Static DED field (DC)
Counter−rotating DED field (AC+)
3
Ωφ (rad/s)
2
IDED (kA)
0
1.5 2 2.5 3 3.5
time (s)
Figure 6.1 : Time traces of the toroidal rotation frequency at R = 1.83 m, for 3 plasma
discharges with a static (DC), co-rotating (AC− ) and counter-rotating (AC+ ) per-
turbation field induced by the DED. For the co-rotating field the rotation increases
with the perturbation strength (∼ IDED ). For the static and counter-rotating fields
the increase in plasma rotation is cut short at the transition threshold.
The time traces of the toroidal rotation frequency in figure 6.1 are taken at the position
R = 1.83 m (r/a = 0.1) for three TEXTOR discharges with DC, AC+ and AC− DED
operation. All other plasma parameters where identical in these discharges. The plasma
rotation before the DED is applied is in the co-direction (positive). One observes an initial
increase in rotation frequency when the DED current is increased. For the co-rotating DED
field (AC− ) the increase in rotation frequency continues up to the maximum value of the
DED current; the plasma rotation stays in the first phase.
For a static (DC) or counter-rotating DED field (AC+ ), after a slight initial increase, a
sharp drop of the rotation frequency is observed. After this sharp drop the toroidal plasma
rotation frequency stays constant, even though the DED current is further increased. The
plasma rotation frequency during this constant phase depends on the DED frequency: it is
76 Section 6.2 - A general overview

4
x 10
t = 1.65 s, I = 0.06 kA
DED
4 t = 1.90 s, I = 0.28 kA
DED
t = 2.15 s, IDED = 0.50 kA
3.5
t = 2.40 s, IDED = 0.73 kA
3 t = 2.65 s, I = 0.96 kA
DED
Ωφ (rad/s) t = 2.90 s, IDED = 1.18 kA
2.5
1.5
1 q=1 q=2
0.5
0
1.75 1.8 1.85 1.9 1.95 2 2.05
R (m)
Figure 6.2 : The evolution of the toroidal rotation profiles during a ramp up of the current in
the DED coils in AC− operation.
The rotation frequency increases when the DED current is increased. The shape of the
rotation profile does not change: the rotation profile is just lifted.
4
x 10
t = 1.65 s, IDED = 0.06 kA
4 t = 1.90 s, I = 0.28 kA
DED
t = 2.15 s, IDED = 0.50 kA
3.5
t = 2.40 s, IDED = 0.73 kA
3 t = 2.65 s, IDED = 0.96 kA
t = 2.90 s, IDED = 1.18 kA
Ωφ (rad/s)
2.5
1.5
1 q=1 q=2
0.5
0
1.75 1.8 1.85 1.9 1.95 2 2.05
R (m)
the DED coils in DC operation.
black profiles: Initial phase, increasing rotation with increasing DED current.
grey profiles: Second phase, plasma rotation locked with DED frequency.
More towards the edge – around the q = 2 – the plasma rotation monotonically increases
towards the final locked rotation. In the plasma centre the – around the q = 1 – the plasma
rotation velocity first increases and then suddenly drops to the locked value.
Section 6.2 - A general overview 77

4
x 10
t = 1.65 s, I = 0.06 kA
DED
4 t = 1.90 s, I = 0.28 kA
DED
t = 2.15 s, IDED = 0.50 kA
3.5
t = 2.40 s, IDED = 0.73 kA
3 t = 2.65 s, I = 0.96 kA
DED
t = 2.90 s, IDED = 1.18 kA
Ωφ (rad/s)
2.5
1.5
1 q=1 q=2
0.5
0
1.75 1.8 1.85 1.9 1.95 2 2.05
R (m)
the DED coils in AC+ operation.
black profiles: Initial phase, increasing rotation with increasing DED current.
grey profiles: Second phase, plasma rotation locked with DED frequency.
Here both around the q = 2 surface and around the q = 1 surface the plasma rotation velocity
first increases and then decreases again. There is no location in the plasma where the change
in plasma rotation is monotonic.
lower for AC+ DED operation than for DC operation. From figure 6.1 one can determine
the difference between the plasma rotation frequency in AC+ and DC operation. This
difference is about 6.3 · 103 rad/s or 1 kHz. Taking into account that an AC+ DED field
rotates in the counter direction (i.e. negative) with a frequency of 1 kHz and the DC field
has of course zero frequency, this indicates that in the constant phase the plasma rotation
frequency is directly related to the DED frequency.
The transition between the increasing phase and the constant phase is reasonably
fast; in the order of 100 ms. The threshold of the perturbation amplitude at which this
transition occurs is different for different plasma conditions.
In figures 6.2 to 6.4 give the evolution of the toroidal rotation frequency profile during
the linear increase of the DED current. The data in these figures corresponds with the
AC− , DC and AC+ time traces in figure 6.1. The black profiles are in the initial phase; the
rotation frequency increases with increasing DED current. For the DC and AC+ current
ramp, figures 6.3 and 6.4, the increase is marginal. In case of a current ramp of the DED
in AC− operation, figure 6.2, the increase in rotation frequency is substantial. We also
observe that the shape of the rotation profile does not change.
The grey profiles in figures 6.3 and 6.4 are in the second phase where the rotation
frequency does not change anymore with increasing DED current. The grey profiles are
significantly flattened in comparison with the peaked profile before DED operation and
during the initial phase. The rotation frequency in this flat region is related to the DED
frequency.
The transition from the peaked, black profiles to the flattened, grey profiles is different
in DC and in AC+ operation. For DC DED operation the plasma rotation in the core
78 Section 6.2 - A general overview

drops when the transition threshold is reached, while the rotation more towards the edge
increases. In the case of AC+ DED operation the rotation decreases over the whole
measured region. Due to the flattened profile after the transition, the drop in rotation is
of course larger in the centre than it is near the edge of the plasma.
6.3 Change in rotation as a function of DED current

In the previous section we saw that, when increasing the DED current, the response of
the plasma rotation can be separated in two clearly different phases. We now focus on the
initial phase where the change in plasma rotation depends on the DED current.
When increasing the current in the DED coils, in a first phase the rotation changes as a
function of the DED current. The shape of the rotation profile does not change during this
phase; the profile is just lifted as can be clearly seen in figure 6.2. This can be attributed
to momentum transport. The toroidal momentum balance equation (2.37) shows that if
the toroidal rotation changes at one position, this change will have an effect on the whole
plasma. How strong this effect is depends on how large the momentum transport, defined
by the momentum diffusion coefficient D, is. A toroidal torque localised near the plasma
edge is able to lift the whole rotation profile.
If we look at the change in rotation for a large number of discharges, we notice that co-
rotating plasmas show an increase in rotation velocity, when the DED current is increased.
Counter-rotating plasmas will slow down. This means the rotation does not spin up when
the DED current is increased, but that there is a change of the plasma velocity in the co-
direction. In most cases this change in co-rotation is a monotonically increasing function
of the DED current; it lasts until the threshold for the transition to the constant phase or
the maximum DED current is reached.
For plasmas that have a large rotation velocity in the co-direction, it is observed that
they have an initial increase in co-rotation. This increase however slows down and reverses
at higher DED currents, so finally the change of rotation is in the counter-direction.
In figure 6.5 the change in toroidal rotation is plotted versus effective DED current for
different plasma discharges, where the effective DED current is defined as:
(
Icoil (DC)
IDED = 2√ (6.1)
2+ 2
× I coil (AC),
The data sets in figure 6.5 cover a wide range of plasma parameters: low and high
power input, discharges rotating in co-direction or counter-direction before the DED was
applied and discharges where DC, AC+ and AC− DED was used. The data can be divided
in two groups:
• three sets of three discharges where in each set all plasma parameters where kept
constant, only the frequency of the DED was changed form AC+ over DC to AC− .
Between the sets of discharges the plasma parameters are different.
– The data points plotted in black come from discharges with low power and low
momentum input by the neutral beams.
– The data points plotted in red come from discharges with strongly counter-
rotating plasmas and high power input.
Section 6.3 - Change in rotation as a function of DED current 79

4
x 10
2
3/1 DC, P = 0.25 MW, T = 0.15 Nm
NBI NBI
+
3/1 AC , P = 0.25 MW, T = 0.15 Nm
NBI NBI
1.5 3/1 AC−, P = 0.25 MW, T = 0.15 Nm
NBI NBI
3/1 DC, P = 1.65 MW, T = −0.65 Nm
NBI NBI
3/1 AC+, P = 1.65 MW, T = −0.65 Nm
1 NBI NBI
3/1 AC−, P = 1.65 MW, T = −0.65 Nm
∆ Ωφ (rad/s)
NBI NBI
3/1 DC, PNBI = 1.30 MW, TNBI = 0.20 Nm
0.5 +
3/1 AC , PNBI = 1.30 MW, TNBI = 0.20 Nm
3/1 AC−, P = 1.30 MW, T = 0.15 Nm

NBI NBI
3/1 DC, PNBI = 1.35 MW, TNBI = 0.85 Nm
0
3/1 DC, P = 1.40 MW, T = 0.75 Nm
NBI NBI
3/1 DC, PNBI = 1.30 MW, TNBI = 0.35 Nm
−0.5 3/1 DC, PNBI = 1.30 MW, TNBI = 0.20 Nm
3/1 DC, PNBI = 1.35 MW, TNBI = −0.15 Nm
3/1 DC, P = 1.35 MW, T = −0.35 Nm
NBI NBI
−1
0 0.5 1 1.5 2 2.5 3
Effective IDED (kA)
Figure 6.5 : The change in toroidal rotation versus effective DED current for several discharges.
The change in rotation is mostly in the co-direction. Only for discharges with a high momen-
tum input in the co-direction, the toroidal rotation changes in the counter-direction during
the application of the DED. The change in rotation frequency also differs for discharges with
different power input.
1
3/1 DC, PNBI = 0.25 MW, TNBI = 0.15 Nm
3/1 AC+, P = 0.25 MW, T = 0.15 Nm

NBI NBI
0.8
3/1 AC−, PNBI = 0.25 MW, TNBI = 0.15 Nm
3/1 DC, P = 1.65 MW, T = −0.65 Nm
NBI NBI
0.6 +
3/1 AC , PNBI = 1.65 MW, TNBI = −0.65 Nm
DED torque (N m)
3/1 AC−, PNBI = 1.65 MW, TNBI = −0.65 Nm

0.4 3/1 DC, P = 1.30 MW, T = 0.20 Nm
NBI NBI
+
3/1 AC , PNBI = 1.30 MW, TNBI = 0.20 Nm
0.2 3/1 AC−, PNBI = 1.30 MW, TNBI = 0.15 Nm
3/1 DC, PNBI = 1.35 MW, TNBI = 0.85 Nm
0 3/1 DC, PNBI = 1.40 MW, TNBI = 0.75 Nm
3/1 DC, PNBI = 1.30 MW, TNBI = 0.35 Nm
3/1 DC, PNBI = 1.30 MW, TNBI = 0.20 Nm
−0.2
3/1 DC, PNBI = 1.35 MW, TNBI = −0.15 Nm
3/1 DC, PNBI = 1.35 MW, TNBI = −0.35 Nm
−0.4
0 0.5 1 1.5 2 2.5 3
Effective IDED (kA)
Figure 6.6 : The torque exerted by the DED versus effective DED current for several discharges.
This torque was calculated by dividing the measured change in rotation by the momentum
confinement time. One sees that the influence of the power input on the change in rotation
was mainly due to the difference in momentum confinement time.
80 Section 6.3 - Change in rotation as a function of DED current

– The data points plotted in green come from discharges with similar momentum
input as the ‘black’ data, but with higher power input.
• a rotation scan where the power input was kept constant and the momentum input
was changed from discharge to discharge. The DED was operated in DC mode.
These data are plotted in blue.
Despite the variety in plasma parameters the change in co-rotation shows the same
kind of behaviour for most discharges. The change is mostly in the co-direction and has
a non-linear dependence on the DED current. Only for strongly co-rotating plasmas –
indicated by the blue + and × signs – the plasma slows down; i.e. the change in rotation
is in the counter-direction.
Figure 6.5 also shows that the in rotation depends on the power input as well. This
is not unexpected; as mentioned in section 2.3.1 and section 5.2 the toroidal momentum
confinement time τφ is related to the energy confinement time τE , and hence decreases
with increasing power input. This means that for plasmas with high power input the
momentum input by the DED is lost faster than in low power plasmas. In other words,
the change in plasma rotation will be different for discharges with a different power input,
even if the torque exerted by the DED does not depend on the power input. So, in order
to find out whether the torque exerted by the DED is influenced by the power input, we
should plot the change in torque, instead of the change in rotation, versus the effective
DED current. This change in torque is calculated from the rotation measurements using
the definition of the momentum confinement time ( see also equation (2.26)):
R02 ρ∆Ωφ dr
R
∆TDED = , (6.2)
τφ
where the momentum confinement time τφ is determined before the DED is applied, as-
suming that it does not change during the initial DED phase.
Figure 6.6 shows the torque exerted by the DED versus the effective DED current. It
reveals that the influence of the power input on the torque caused by the DED is marginal.
The non-linear dependence on the DED current and the sign reversal for highly co-rotating
plasmas of course remains. The amount of change in torque due to the DED can go up
to 1 N m, which is of the same order of magnitude as the torque caused by the neutral
beam. Usually the change in torque is lower because the threshold for transition to the
locked phase is reached before the effective DED current is at maximum.
Not only the toroidal rotation of the core plasma is affected during DED operation.
The poloidal velocity at the edge of the plasma - measured with both CXRS and passive
emission of CIII just inside the last closed flux surface - shows a change in the co-direction
during DED as well, as is shown in figure 6.7 [11]. Also turbulence rotation measured
with reflectometry shows an increase in the ion diamagnetic drift direction [53]. A fourth
observation made during DED operation is a positive change ∆Er of radial electric field
at the very edge of plasma. This is seen in figure 6.8 where the floating potential profile in
the outer few centimetres of the plasma gets less steep during DED operation [43]. Also
sign reversal of the edge electric field has been observed in some cases. A positive ∆Er
is, through the E × B drift, linked with an increase in co-rotation, both toroidally and
poloidally.
An important remark has to be made concerning the measurements of poloidal rotation,
turbulence rotation and floating potential in the edge of the plasma. For the toroidal

Figure 6.7 : The change of poloidal rotation as function of the DED current.
For DC, AC+ and AC− these changes are in the co-direction, i.e. the
ion diamagnetic direction or the direction of the poloidal magnetic field.
[11]
rotation in the core an increase in co-rotation was observed in most cases, but for some
plasmas that were rotating fast in the co-direction, a decrease of the co-rotation was
Figure 6.8 : The profile of the floating potential Vf l near the plasma bound-
ary in the case of no DED, AC+ DED and AC− DED. For both AC+
and AC− the profile of Vf l gets less steep. This means that the radial
electric field Er gets less negative and hence the change in electric field
∆Er is positive.[43]

seen. This is not seen in the measurements poloidal rotation and floating potential near
the plasma edge; they always show an increase in co-direction. Also in contrast to the
measurements of the central toroidal rotation, these measurements do not show as clearly
the two phases below and above the DED threshold. Where the core plasma rotation is
locked to the rotation frequency of the perturbation field when the DED current exceeds
the threshold, the edge measurements still see an increase of the rotation in co-direction.
Moreover this increase in co-rotation with increasing DED current is even slightly enhanced
when the threshold is exceeded [11].
Summary and discussion

In summary, these experiments show that, when the level of the perturbation is increased,
the plasma rotation changes. In other words, the perturbation field of the DED exerts a
torque onto the plasma. We have seen that the total power input has little influence on
this torque. We have also seen that the direction of this torque is mostly directed in the
co-direction, resulting in a slowing down of counter-rotating plasmas and a spin-up of
co-rotating plasmas. An exception to this behaviour was found for plasmas that rotated
fast in the co-direction: after an initial increase in rotation velocity, these plasmas started
to slow down with increasing perturbation strength. This exception was only found for the
toroidal rotation in the plasma core: the poloidal rotation in the edge and the floating po-
tential profile in the edge were for all discharges compatible with an increase in co-rotation.
Also in the literature it is found that the plasma rotation is expected to change when
the level of perturbation is increased [28, 55]. This change in rotation should monotonically
go towards the rotation velocity for which the tearing modes are at rest in the frame of
the perturbation field. As seen in section 5.5 the rotation frequency of tearing modes in
TEXTOR is given by:
ωt = nΩφ − 2πfe∗ , (6.3)
where fe∗ is the local electron diamagnetic frequency.
Hence the modes will be at rest when the slip frequency ω = 0, with:
ω = ωt − 2πfDED = nΩφ − 2π(fe∗ + fDED ). (6.4)
The change in ω is given in [55] as:

" 2 #1/2
ω 1 1 IDED
= + 1− for IDED < Ithreshold , (6.5)
ω0 2 2 Ithreshold
and hence |ω| > |ω0 /2|,
In the above formula ω0 is the initial slip frequency before the perturbation field is
applied, and Ithreshold is the DED current at which the slip frequency is half the initial slip
frequency ω = ω0 /2. At this point the theory predicts that ω jumps to zero. In [55] it is
found that Ithreshold is proportional to ω0 .
At ω = 0 the toroidal rotation frequency is given by Ωφ = (2π/n) (fe∗ + fDED ). In
TEXTOR fe∗ is typically about 2 kHz and fDED = 0, 1 or −1 kHz for DC, AC− or AC+
operation, respectively. This means that in TEXTOR ω = 0 corresponds with a positive,
i.e. co-rotating, toroidal rotation frequency Ωφ . The lowest Ωφ corresponding with ω = 0,
occurs for AC+ operation, where fDED = −1 kHz, hence Ωφ ≈ 6.3 · 103 rad/s for a toroidal
mode number n = 1.

In other words, when the perturbation level is increased the rotation is expected
to change according to (6.5), which means towards a positive toroidal rotation fre-
quency Ωφ . Counter-rotating plasmas and plasmas with a co-rotation frequency below
Ωφ = (2π/n) (fe∗ + fDED ) will experience a change in the co-direction. Plasmas that rotate
in the co-direction with Ωφ larger than (2π/n) (fe∗ + fDED ) are expected to slow down,
hence experience a change in the counter-direction. Qualitatively this seems to agree
with our experimental results.
To make a quantitative comparison, we need to know at which radial position we need

to measure the rotation frequency. Equation (6.5) predicts the rotation at the position of
the MHD modes. In the next section we will see that 1/1 and 2/1 modes are excited at
the q = 1 and q = 2 surfaces. We will therefore compare the measurements of the plasma
rotation frequency at the location of the q = 1 and q = 2 surfaces to the ω-evolution given
by (6.5).
In figures 6.3 and 6.4 we immediately see that the rotation at the position of the q = 1
surface does not correspond with the rotation predicted in (6.5). Instead of monotonic
change in rotation, the rotation first increases in the co-direction and than suddenly drops.
We therefore conclude that influence of the interaction between the perturbation field and
modes at the q = 1 surface will have little influence on the change in plasma rotation.
The interaction between the perturbation field and modes at the q = 2 surface is a far
more promising candidate to explain the change in plasma rotation. We see that at the
q = 2 surface there is a monotonic change in rotation frequency in the DC case (fig. 6.3).
For AC+ operation (fig. 6.4) there still is an initial increase followed by a decrease, but
this non-monotonic change is in the order of the measurement error.
4
x 10
ω = Ωφ(q=2) − 2 π (fe + fDED) (rad/s)
0.5
*
−0.5
0 0.5 1 1.5 2
IDED (kA)
Figure 6.9 : A comparison between the measured slip frequency ω as a function of IDED and the
ω-evolution as predicted in [55]. One immediately sees that the prediction does not agree
with the measurement.

In figure 6.9 we compare the slip frequency ω calculated from the rotation frequency
Ωφ at the q = 2 surface, with the expected ω that follows from equation (6.5). This is
done for two situations: with an initial ω0 < 0 and with a ω0 > 0. It clearly shows that
there is a discrepancy between the expected change in rotation and the measured change
in rotation.
So at first sight there seems to be an agreement between the measurements and the
prediction: a change in co-rotation for counter-rotating and slowly co-rotating plasmas and
a change in the counter-direction for plasmas that rotate fast in the co-direction. A quan-
titative comparison, however, shows that the prediction and measurement do not agree –
the change in not monotonic if ω0 > 0 and is faster than expected if ω0 < 0 (see figure 6.9).
The discrepancy could possibly be explained if, apart from a torque trying to bring
ω to zero, the perturbation field is responsible for a second torque acting on the plasma.
This torque should be directed in the co-direction, such that for ω0 < 0 it would bring ω
faster to zero, while for ω0 > 0 it would oppose the torque that tries to reduce |ω|. Because
for ω0 , ω initially increases, for low DED currents this co-torque should be stronger than
the torque that tries to bring ω to zero. On the other hand, because ω eventually becomes
zero, for higher DED currents the torque reducing |ω| prevails.
The measurements of the poloidal rotation and the floating potential in the plasma
edge, suggest that this co-torque could be located in the edge region: the edge measure-
ments always show a change of the rotation in the co-direction and this change continues
even when the core toroidal rotation has already reached the constant phase. In the next
section it will be shown that in this constant phase ω = 0 at the q = 2 surface. The fact
that the poloidal rotation and the floating potential still change, although the rotation at
the q = 2 surface is kept constant, suggests that a co-torque is present outside the q = 2
surface that increases with increasing DED current.
6.4 Plasma rotation locked to the DED frequency

Once the DED perturbation reaches a high enough amplitude the rotational behaviour of
the plasma changes very rapidly. In 50 to 100 ms the toroidal rotation profile flattens over
practically the whole measured area. This can be seen in figures 6.1, 6.3 and 6.4. A flat
area in a rotation profile means that the plasma rotates as a rigid body within this area.
In figures 6.3 and 6.4 the positions of the q = 1 and q = 2 surfaces are indicated, where
a quadratic q-profile was assumed. The rigid body rotation stretches out between these
two q surfaces.
After the transition the rotation frequency of stays constant, even with increasing DED
current. Figure 6.1 shows that for discharges with a counter-rotation AC+ DED field the
rotation frequency is 6.3 rad/s or 1 kHz lower than for discharges with a static DC field.
This strongly suggests that, in this constant phase, the toroidal rotation frequency Ωφ is
such that tearing modes within the plasma are locked to the DED field: ωt = ωDED =
2πfDED , where ωt is the tearing mode frequency. The relation between Ωφ and ωt is given
by equation (6.3): ωt = nΩφ − 2πfe∗ .
In 3/1 DED operation the main Fourier components of the perturbation field have
n = 1 as toroidal mode number. Figure 6.10 shows ωDED versus the MHD frequency ωt
for n = 1 at the q = 2 surface. It is clearly seen that ωDED indeed equals ωt , as would be
expected when a 2/1 mode is locked to the DED field.
Section 6.4 - Plasma rotation locked to the DED frequency 85

4
x 10
3
AC− DED 3.75 kHz
2
ωDED = 2 π fDED (rad/s)
1
AC− DED 1.00 kHz
DC DED 0.00 kHz

0
AC+ DED 1.00 kHz
Figure 6.10 : The data points,
−1 with errorbars, show the fre-
quency of the DED field ver-
sus the tearing frequency in the
−2 AC+ DED 3.75 kHz locked phase. The data points
coincide with the dashed line
−3 indicating ωDED = ωt .
−3 −2 −1 0 1 2 3
ωt = Ωφ − 2 π f*e (rad/s)
4
x 10
Of course a plasma rotation frequency that corresponds with the presence of locked
modes is not a solid prove for the existence of these modes. The direct observation of
tearing modes locked to the DED in the phase with a constant plasma rotation, is therefore
needed.
Because the electron temperature Te and density ne are uniform or slightly peaked in
a magnetic island, diagnostics that measure electron temperature and/or density are well
suited for mode detection.
In figure 6.11 (a) the Te profile with Thomson scattering during DC DED operation
is shown. A flat region, indicating the presence of an island, is seen at the position where
q = 2 is expected. The fact that the island is only seen in the lower part of the profile
(z < 0), and not, as the 2/1 island symmetry requires, in both lower and upper part, is
due to the off-axis measurement of the Thomson scattering system.
When the time evolution of the Te profile is measured during DC DED operation, no
changes in the profile are observed as a function of time. This is expected, because the
island is locked to the static DED field and hence does not move. During AC DED oper-
ation the island is locked to a rotating field, therefore the time evolution of the Thomson
measurements should show the appearance and disappearance of the flat Te region with
the frequency of the perturbation field. These fluctuations in Te , but also in ne , are shown
in figure 6.11 (c). From this plot an island width in the order of W ∼ 10 cm can be
determined.
Islands can also be detected with ECE measurements. During DED in AC operation,
fluctuations in Te are seen at specific locations in the plasma. Taking into account the
rigid body rotation in the toroidal direction measured with charge exchange and assuming
no poloidal rotation, the time information of the ECE signals corresponds with different
toroidal positions. Through the q-profile these toroidal angles can be projected onto
poloidal angles, thus reconstructing the 2 dimensional spatial electron temperature profile
86 Section 6.4 - Plasma rotation locked to the DED frequency

Figure 6.11 : Thomson Scattering measurements of the 2/1 islands during the locked DED
phase:
(a) Te profile during DC DED operation. The flattened region indicates a magnetic island.
(b) The flattening in (a) is only seen at the lower part side of the profile because the mea-
surement of the Thomson scattering system does not go through the plasma centre.
(c) The time evolution of electron density and temperature during AC+ DED operation.
Both Te and ne show fluctuations with the DED frequency of 1 kHz; ne is peaked in the
island O-points, Te is flattened. The horizontal dotted lines indicate the position of the q = 2
surface, the vertical dashed lines indicate the time points of the measurements. [84]
in a poloidal cross section. The result of such a rotational reconstruction is given in

figure 6.12, where a quadratic q-profile is assumed. One can clearly see a 2/1 mode, but
apart from that more towards the centre of the plasma a 1/1 internal kink mode is visible.
Thomson scattering and ECE are not the only diagnostics that observe the excitation
of islands when the rotation jumps into the constant phase. Measurements with soft X-ray
cameras indicate the presence of a 2/1 island with an island width of about W = 8 cm
and a coupled 1/1 internal kink mode [57]. And also magnetic probes and coils located
just outside the plasma, pick up a change in magnetic flux when the transition is made
to the locked phase.
Apart form the Ωφ profile, that is constant and flat, and the observation of 2/1 and 1/1
modes with ECE, Thomson scattering and other diagnostics, a third observation is made
in the phase with constant rotation frequency. For TEXTOR discharges with sawteeth,
the sawteeth disappear quite abruptly, when the constant rotation phase is reached; this

2000
(b)
Te (eV)
1500
Te (eV)
(a)
0.2 X 2/1
1000
0.1 O 2/1
O 1/1
500
−0.2 −0.1 0 0.1 0.2
r (m)
0 r (m)
2.5
X 1/1 (c)
−0.1 O 2/1
2
−0.2
X 2/1
1.5
q
−0.2 −0.1 0 0.1 0.2
r (m) 1
0.5
−0.2 −0.1 0 0.1 0.2
r (m)
Figure 6.12 : Rotational reconstruction of ECE fluctuations.

(a) Contour plot of the electron temperature in a poloidal cross section during the locked
phase in AC+ DED operation. Both a 2/1 island as a 1/1 mode can be observed.
(b) The blue profile goes through the X points of the 2/1 island and through the X and O
point of the 1/1 island. The red profile goes through the O points of the 2/1 island and goes
twice partially through the O point of the 1/1 island.
(c) The q profile used for the reconstruction.
can be seen in figure 6.13 (a). Before the transition the sawteeth have a precursor with
a frequency of about 2 kHz. This can be seen in figure 6.13 (b). After the transition to
the constant phase, the sawteeth have disappeared, a small fluctuation with a frequency
of 1 kHz is still visible in the ECE signal (see figure 6.13 (c) ). This 1 kHz fluctuation
comes from the internal kink mode that is shown in figure 6.12.

When the plasma rotation makes the transition to the constant phase three major obser-
vations are made:
• The Ωφ profile stays constant, even when the DED current is further increased. The
constant rotation frequency at the q = 2 surface is compatible with a 2/1 tearing
mode frequency equal to the DED frequency. In between the q = 1 and the q = 2
surface, the Ωφ profile is flattened.
• Several diagnostics (ECE, Thomson scattering, soft X-ray, et cetera) observe large
2/1 islands, that are locked to the DED field. Some diagnostics (ECE and soft X-
ray) observe a 1/1 internal kink mode, also locked to the DED field and consequently
to the 2/1 island.
• The sawteeth suddenly disappear.
88 Section 6.4 - Plasma rotation locked to the DED frequency

10
Te (a.u.)
8
6
1.5 1.52 1.54 1.56 1.58 1.6 1.62
time (s)
10 8.5
Te (a.u.)
Te (a.u.)
9 8
8 7.5
7 7
1.537 1.538 1.539 1.61 1.612 1.614
time (s) time (s)
Figure 6.13 : Sawtooth behaviour during DED.

(a) Before the locked DED phase sawteeth are present. When the transition to the locked
phase occurs the sawteeth very suddenly disappear.
(b) The sawtooth precursor has a frequency of about 2 kHz.
(c) In the locked phase a small fluctuation is present. The sine function with frequency of
−1 kHz – AC+ DED – is added to guide eye.
All these observations are compatible with the process of mode excitation as it is
described in the literature [28, 55]. When a perturbation field interacts with a plasma, it
will first change the plasma rotation towards the rotation velocity for which the tearing
modes are at rest in the frame of the perturbation field (see section 6.3). Once the tearing
modes are at rest in the frame of the perturbation field, the perturbation field will keep
the plasma rotation constant, so that the tearing modes stays at rest. This corresponds
with the first observation. As long as the tearing modes move within the frame of the
perturbation field, they will be undetectable. Once the tearing modes are at rest in the
frame of the perturbation field, they are excited: they become large enough to be detected
– the second observation. The transition from suppressed to excited modes is called mode
excitation or mode penetration.
The fact that the profile of the toroidal rotation frequency is flattened from the q = 1
to the q = 2 surface, suggests that modes are excited and locked to the DED at both the
q = 1 and the q = 2 surface. The measurements of ECE, Thomson scattering and soft
X-ray, provide the evidence for both a 2/1 island and an 1/1 internal kink mode, both
locked to the DED.
The coupling between the 2/1 island and the 1/1 internal kink mode also explains
the sudden loss of the sawteeth. The sawtooth crash is the result of the growth of the
sawtooth precursor. This sawtooth precursor is an unstable 1/1 internal kink mode

[85, 88]. The mode grows until it fills the whole region inside the q = 1 surface, at which
moment the sawtooth crash occurs [61, 44, 88]. The coupling between the 1/1 internal
kink and the 2/1 island in the locked phase, stabilises the 1/1 internal kink, preventing it
from causing a sawtooth crash.
The following summary describes the mode excitation process in TEXTOR:
• Because we know that a 2/1 island and a 1/1 internal kink mode are excited, we
expect that the perturbation field will try to bring the rotation frequency Ωφ towards
2π (fe∗ + fDED ) both at the q = 1 surface and the q = 2 surface.
• We assume that the influence of the perturbation field at the q = 1 surface is

significantly weaker than at the q = 2 surface. This assumption is justified by the
fact that the q = 1 surface lies further away form the DED coils.
• When the perturbation level is increased, the rotation frequency Ωφ at the q = 2

surface goes towards 2π (fe∗ + fDED ). Because we assume that the influence of the
perturbation field at the q = 1 surface is weak, the change in Ωφ at q = 1 is mainly
governed by momentum transport. At the q = 1 surface Ωφ does therefore not
necessarily go towards 2π (fe∗ + fDED ).
• Once Ωφ = 2π (fe∗ + fDED ) at the q = 2 surface, a large 2/1 island is excited.

Measurements show that this island has a width of about 10 cm.
• The large, excited 2/1 island has a perturbation field of its own. The total pertur-
bation field – the DED field plus the perturbation caused by the 2/1 island – is now
large enough to have a significant influence at the q = 1 surface. As a result the
rotation at the q = 1 surface changes abruptly so that also at the q = 1 surface
Ωφ = 2π (fe∗ + fDED ) and a 1/1 internal kink mode is locked to the DED and the 2/1
island.
6.5 Transition threshold

We now know that the transition from the phase with changing rotation as a function of
DED current to the locked phase is a mode penetration or mode excitation process. The
threshold for mode excitation can thus be determined from the sudden change in rotation
measured with CXRS. In TEXTOR it can also be determined by a transient drop in the
density signal, because the presence of islands reduces the confinement.
The threshold for mode excitation depends on several plasma parameters. Empirically
it has been observed at TEXTOR that the threshold is proportional with the plasma
energy β and the density ne [51].
We are interested in the dependence of the threshold on the plasma rotation. Therefore
experiments were carried out in which the density (ne = 1.5 1019 m−3 ) and power input
(P = 1.5 M W ) were kept constant, so that the influence of ne or β did not play any
role, while the combination of co-beam NBI1 and counter-beam NBI2 was used to change
the rotation from discharge to discharge. The result of these experiments is shown in
figure 6.14. The threshold, determined both from the rotation and from the density
measurement, is plotted as a function of the rotation frequency, before the DED was
applied, at the q = 2 surface. The blue data points come from a plasma rotation scan
90 Section 6.5 - Transition threshold

during DC DED operation, the red data points come from AC+ DED (thus a negative
DED field rotation of −1 kHz).
1.5
Effective IDED (kA)
0.5
0.6 104 rad/s

≈
1 kHz
0
−1 −0.5 0 0.5 1 1.5 2
Ωφ,0(q=2) (rad/s) x 10
4
Figure 6.14 : Threshold for mode excitation plotted against the rotation at q = 2 before the
DED is applied. The density (1.5 1019 m−3 ) and the power input (∼ β) were kept constant.
The data for DC DED operation are given in blue, the data for AC+ DED (−1 kHz) are
given in red. The dotted lines are cubic fits to the data, to guide the eye.
It is observed that the threshold for both the DC and the AC+ scan has a minimum at a
certain rotation velocity; Ωφ,0 = 104 rad/s and Ωφ,0 = 0.4 104 rad/s for DC and AC+ ,
respectively. With the slip frequency ω = ωt − ωDED defined as the difference between
the DED frequency ωDED = 2πfDED and the tearing mode frequency ωt = Ωφ − 2πfe∗ ,
then the minimal threshold corresponds in both cases with ω0 = 0, where ω0 is the slip
frequency before the DED is applied. The diamagnetic frequency for these discharges was
fe∗ = 1.65 kHz.
For |ω0 | > 0 the threshold for mode excitation increases. The sign of the slip frequency
ω0 however seems to be important: it can be observed in figure 6.14 that the slope of the
threshold curve left and right of the minimum differs. If the plasma at the q = 2 is rotating
faster than 2π(fDED + fe∗ ) before the DED is applied, a higher DED current is needed to
excite the mode than in the case were the plasma would be rotating an equal amount
below 2π(fDED + fe∗ ).

Figure 6.14 was already introduced in the first chapter. It stresses the importance of
plasma rotation, when it comes to avoiding the excitation of tearing modes. One finds
from figure 6.14 that the excitation of tearing modes can be avoided when the plasma
rotation is high. The direction of the plasma rotation, however, is also of importance,
because
(a) the minimum is not located at Ωφ,0 = 0 but at ω0 = 0. This means that one has to
take into account the frequency and the direction of the electron diamagnetic drift
Section 6.5 - Transition threshold 91

(fe∗ ) as well as the frequency and the direction of the perturbation field rotation
(fDED ).
(b) the threshold increases faster with increasing |ω0 | for ω0 > 0 than for ω0 < 0.
In the literature it is found that the threshold for mode excitation is proportional with
|ω0 | [28, 55]. This means that from theory we expect that the sign of ω0 has no influence
on the mode excitation threshold. Our measurements however show that the sign of ω0
does play a role. The discrepancy between theory and measurement becomes clear in
figure 6.15, where the predicted threshold and the measured threshold in the DC DED
case our plotted in the same figure.
1.5
IDED, threshold (kA)
0.5
0
−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5
ω0 = Ωφ,0(q=2) − 2 π f*e (rad/s)
4
x 10
Figure 6.15 : The measured and the theoretically expected threshold for mode excitation plotted
against the ω0 , for DC DED operation. One clearly sees that theory and measurement do
not fully agree.
6.6 Summary and conclusion

In summary, we have carried out several experiments in an effort to elucidate the relation
between perturbation fields and plasma rotation.
First, we found that, when the DED field is increased, the rotation of the plasma
changes. The change in rotation is for most discharges in the co-direction. Only for
discharges with a fast plasma rotation in the co-direction, the rotation shows a non-
monotonic change. An initial change in the co-direction is followed by a change in the
counter-direction. During the DED current ramp, the shape of the rotation profile does
not change.
This change in rotation frequency as a function of DED current comes to an abrupt
halt at a certain threshold. At this threshold the rotation frequency changes very rapidly
92 Section 6.6 - Summary and conclusion

towards: Ωφ = 2π (fe∗ + fDED ) at the q = 2 surface. 2/1 magnetic islands are excited by
the perturbation field at this moment. These islands are large and locked to the DED
frequency. The large 2/1 mode immediately couples to an internal kink mode at the q = 1
surface. This causes a flattening of the rotation profile in between the q = 1 and q = 2
surface.
Finally, we examined the dependence of the threshold, at which the mode excitation
occurs, on the plasma rotation. We found that fast rotating plasmas are more resistant
towards mode excitation. The minimal threshold, however, does not occur at Ωφ,0 = 0,
where Ωφ,0 is the toroidal plasma rotation frequency in absence of a perturbation field.
The minimal threshold occurs at ω0 = 0, with ω = ωt − ωDED the slip frequency between
the rotation frequency of tearing modes in the plasma ωt = nΩφ − 2πfe∗ and the rotation
frequency of the perturbation field ωDED = 2πfDED . This means that the frequency and the
direction of the electron diamagnetic drift (fe∗ ) as well as the frequency and the direction
of the perturbation field rotation (fDED ) have to be taken into account. We also found
that the sign of ω0 is important: the threshold increases faster with increasing |ω0 | for
ω0 > 0 than for ω0 < 0.
4
x 10
2
ω = Ωφ(q=2) − 2 π (f*e + fDED) (rad/s)

1
1.5
0.5
1
0
0.5
−0.5
0
−2 −1 0 1 0 0.5 1 1.5 2
ω0 = Ωφ,0(q=2) − 2 π f*e (rad/s) IDED (kA)
4
x 10
Figure 6.16 : The left plot shows that the measurements of the mode excitation threshold (data
points) deviate from the expected threshold found in the literature (solid line). In the right
plot it is seen that also the measured change in rotation differs from what is predicted by
mode penetration theory.
We confronted our observations with the theoretical predictions that are found in the
literature [28, 55]. In [55] it is found that the external perturbation field will cause a
monotonic change of the plasma rotation such that the slip frequency |ω| between the
perturbation field and the resonant mode in the plasma reduces. Once ω = 0 the modes
are excited. If the initial slip frequency |ω0 | is large, e.g. when the plasma rotation is
large or when the perturbation field frequency is large, then a large change in rotation is
needed, with a large threshold for mode excitation as a consequence. Again in [55] it is
found that the threshold for mode excitation is therefore proportional to |ω0 |.
Our observations indeed show a change in ω towards zero and the subsequent
Section 6.6 - Summary and conclusion 93

excitation of modes. Also a higher threshold for a larger |ω0 | is observed. However, the
predicted monotonic change of the plasma rotation does not agree with the results of
our experiments. Also the proportionality of the mode excitation threshold to |ω0 | is not
observed. These discrepancies between the predictions found in the literature and the
results of our experiments are summarised in figure 6.16.
In conclusion, because of this discrepancy between the theory and the measurements,
we need to have a closer look into the theory of mode excitation. We have already suggested
in section 6.3 that there might be more than one force working on the plasma. We should
therefore also search for a force that is related to the presence of an external field, but
has little to do with the mode penetration itself. In the next chapter we will present an
overview of the theory on mode penetration and the theory on stochastic edge fields due to
an external perturbation field. This will lead us to an expanded model on mode excitation
that will be used in chapter 9 in order to understand these measurements.
94 Section 6.6 - Summary and conclusion

Chapter 7
Theory of perturbation fields
7.1 Introduction
In TEXTOR, the DED changes the plasma rotation and excites tearing modes. The
measurements on the change in rotation and the threshold for mode excitation, however,
do not agree with the change in rotation and the threshold that is expected from existing
theories. To find the reason for the discrepancy between the TEXTOR data and the
theory, it is necessary to have a closer look into the theory of mode penetration.
The measured rotation data itself already hinted a possible explanation. The non-
monotonic change of the plasma rotation suggested that maybe not one, but two forces
depending on the DED current work on the plasma. Again this calls for a closer look into
the literature, in order see which forces could be related to the presence of a perturbation
field.
In the literature two effects are found that can be attributed to the presence of an
external perturbation field:
• In the plasma core shielding currents will be induced that change the stability of
the plasma with respect to tearing modes. The presence of these shielding currents
is also responsible for a force at the rational q surface that is similar to the slip
force in an electric induction motor: it tries to bring the slip frequency, between the
modes in the plasma and the perturbation field itself, to zero. As long as this slip
frequency is larger than zero, the modes in the plasma will be suppressed. Once
the slip frequency is zero the modes are excited. The theory that describes this slip
force and the excitation of modes can be found in [28]. It will be summarised in
section 7.2.
• Near the edge of the plasma the equilibrium field will be significantly altered. A
stochastic zone is formed where the field lines no longer lie on flux surfaces, but
rather fill a volume. Parallel electron transport in this stochastic region changes
the radial electric field, leading to a change in rotation. The description of this
loss current and the resulting electric field is given in [45]. It will be discussed in
section 7.3.
95
Chapter 7 - Theory of perturbation fields
7.2 Penetration of modes

7.2.1 Tearing modes
In this first section we will follow the theory given in [28, 29] and [31] that describes
the process of tearing mode excitation. Before going into the theory of describing their
excitation, let us first describe what a tearing mode is.
Tearing modes are caused by deviations δB from the perfectly axisymmetric equilib-
rium magnetic field B. It is conventional to describe the field perturbations δB by its
Fourier harmonics in terms of poloidal (θ) and toroidal (φ) coordinates. Each Fourier
harmonic is defined by its poloidal and toroidal mode number (m and n), and has an
associated wave vector k. If in the plasma a flux surface exists where B · k = 0 this
Fourier harmonic is called resonant. The flux surface itself is also called resonant. By
definition a resonant flux surface associated with a Fourier harmonic with mode numbers
m and n, has a rational q = m/n. Fourier harmonics for which no corresponding rational
flux surface exists are non-resonant.
The resonant and non-resonant harmonics will give rise to a slight displacement of
the flux surfaces, but for small δB this can be ignored. Resonant harmonics change, in
addition to the displacement, the magnetic topology by tearing apart the equilibrium field
lines, hence the name tearing modes, and reconnect them again, thus creating an extra
magnetic axis. In figure 7.1 a schematic representation of tearing modes is given. Due to
their shape tearing modes are often called magnetic islands. Parameters that describe a
magnetic island are its periodicity, given by the poloidal and toroidal mode numbers m
and n, its width W and the separatrix, which is the ‘border’ of the magnetic island.
Figure 7.1 : At rational q surfaces the nested flux surfaces can break up – i.e. mag-
netic reconnection occurs – and form magnetic islands, a.k.a. tearing modes. The
reconnection can take place due to intrinsic instability of the plasma, or can be
‘forced’ by an external error field.
Due to its width W the field lines of a magnetic island connect radially separated
positions. Fast parallel transport of particles and energy along the separatrix makes islands
responsible for large radial transport and thus degradation of confinement. The difference
between the fast parallel transport around the island, along the separatrix, and the slow
perpendicular transport across the island, is also the reason for the flat temperature and
density profiles over the island.
Deviations δB from the perfectly axisymmetric equilibrium magnetic field B are always
present in a tokamak, so there will always be islands in a tokamak. Usually δB, and thus
96 Section 7.2 - Penetration of modes

the island, is very small. If the island width W is so small that the perpendicular transport
across the island is of the same order as the parallel transport around the island, the plasma
does not ‘see’ the island. In that case the island is suppressed.
7.2.2 Rutherford equation, EM torque and Viscous torque

Now that we know what a tearing mode is, we will give the theory that describes the
evolution of a tearing mode. That evolution can be split in two parts:
• An island can grow or shrink. This is described by the Rutherford equation.
• An island can rotate at a certain rotation frequency. This is caused by a balance

between an electromagnetic torque and a viscous torque.
The rotation frequency of a tearing mode is given in section 2.5.2. We immediately

apply this for the TEXTOR situation (see section 5.5):
ωt = nΩφ − 2πfe∗ . (7.1)
The above expression links the plasma rotation to the rotation of magnetic islands in the
lab frame. The final goal of this chapter is to describe mode penetration by a external
perturbation field. This perturbation field can also have a rotation frequency ωp = 2πfp in
the lab frame. For the theory not the rotation of the modes in the lab frame is important,
but the rotation with respect to the perturbation field. We will therefore describe the
rotation of tearing modes by the slip frequency ω:
ω = ωt − ωp . (7.2)
Deviations δB from the perfectly axisymmetric equilibrium field B are described by

the perturbed flux Ψ. Usually Ψ consists of a number of Fourier harmonics; however for
reasons of simplicity we assume that Ψ has only one general (m,n) component. Seen from
within the tearing frame, i.e. moving along with the mode, Ψ takes the following form:
Ψ = ψ(r) exp (i[mθ + nφ]). In a set of right handed flux coordinates (r, θ, φ) the perturbed
magnetic field in the plasma can therefore be written as:

m ∂Ψ(r)
δB = ∇ × Ψeφ = i Ψ, − ,0 (7.3)
r ∂r
ψ(r) itself can be written as ψ = |ψ| exp(iϑ), where |ψ| is the amplitude of the flux
and ϑ is the phase.
Because the plasma is a highly conductive medium the equations of ideal MHD can
be used to describe it. Using the ideal force balance equation ∇p = j × B, a differential
equation for ψ can be derived [88, 28]:
m2 µ0 jφ0

1 ∂ ∂ψ
r − 2 ψ− ψ = 0 (7.4)
r ∂r ∂r r Bθ 1 − nq m
The above equation has a singularity at positions r = rs where q = m/n; i.e. at the
rational q surface. In a thin layer [rs− , rs+ ] around the rational or resonant surface ideal
MHD breaks down and a separate analysis using resistivity is required. Due to the thin,
‘sheet-like’ nature of these resistive layers, one can show that the perturbed magnetic field
Section 7.2 - Penetration of modes 97

is in the radial direction. The associated parallel, perturbed current can be approximated
by a toroidal, perturbed current: δB = δB⊥ e⊥ ≈ δBr er , δj = δjk ek ≈ δjφ eφ [28].
Outside the resistive layer equation (7.4) can be solved using ψ(0) = ψ(∞) = 0 as ‘well
behaved’ boundary conditions. Over the resistive layer one gets a discontinuity of ∂ψ/∂r,
r
conventionally denoted by the ‘tearing stability index’ ∆0 = [∂(ln ψ)/∂r]rs+ s− . It is called
the ‘tearing stability index’ because it is a measure for the amount of free-energy available
for driving a tearing mode: conventional tearing theory predicts spontaneous reconnection
for ∆0 > 0; for ∆0 < 0 the plasma is intrinsically stable [33]. Through Ampère’s law ∆0 is
associated with the currents δjφ flowing within the resistive layer (see equation (7.7) ).
Within the resistive layer it is common to use the ‘constant ψ’ approximation. This
means ψ(rs− ) = ψ(rs+ ) = ψs is assumed to be constant across the resistive layer.
A last concept that needs to be introduced is the island width. If ψ(r) is non-zero at
a rational q surface, a chain of magnetic islands is formed. The maximum width of these
islands is proportional to the square root of ψ: W ∼ ψ 1/2 .
To summarise:
ψs = ψ(rs− ) = ψ(rs+ ) (7.5)

rs+
0 1 ∂ψ
∆ = (7.6)
ψ ∂r rs−
Z rs+ Z 2π Z 2π
δjφ exp [−i(mθ + nφ)] rR0 drdθdφ
rs+ 0 0
4π 2 R0 ∂ψ rs+ 4π 2 R0

=− =− rs ∆ 0 ψ s (7.7)
2µ0 ∂r rs− 2µ0
1/2 1/2
R0 m ψs
W =4 (7.8)
Bφ n [rq 0 /q]rs
The concepts given in equations (7.5) to (7.8) are used to describe the two main pro-
cesses attributed to tearing modes: their rotation, which is linked with an electromagnetic
torque and the growth of magnetic islands.
The part of the current δjφ flowing in the resistive layer that is in phase with the per-
turbation flux ψ is responsible for the time evolution of the width of the island. Regarding
δjφ as a purely inductive current one finds the island evolution is given by the Rutherford
equation [72, 28, 29]:
dW
0.8227 τR = rs2 ∆0 (7.9)
dt
It is clearly seen that for ∆0 > 0 the island grows, while for ∆0 < 0 the island is
stabilised. In general ∆0 has a dependence on the island width that can be simulated by
∆0 (W ) = ∆0 (0) (1 − W/W0 ). Equation (7.9) then results in a growth to a saturated island
width of W0 .
The part of the δjφ that is in phase quadrature with ψ gives rise to a j × B torque
onto the plasma. The poloidal and toroidal components of this electromagnetic torque

are:

1 ∗ ∗

hTθ,EM i = r δjφ δBr + δjφ δBr (7.10)
2

1 ∗ ∗
hTφ,EM i = R0 (δjθ δBr + δjθ δBr )
2
n
= hTθ,EM i, (7.11)
m
where the 12 (A+A∗ ) defines the real part of a complex A, and the average over the resistive
layer hAi is defined by:
RRR R rs+ R 2π R 2π
layer AdV rs− 0 0 A rR0 drdθdφ
hAi = = 2
(7.12)
Vlayer 4π R0 rs (rs+ − rs− )
Using equations equations (7.3), (7.7), (7.10), (7.11) and (7.12) the expression for the
torques becomes:
n
hTφ,EM i = hTθ,EM i (for rs− < r < rs+ )
m !
∂ψ rs+ ∗ ∂ψ ∗ rs+

in
= r ψ − r ψs
4µ0 rs (rs+ − rs− ) ∂r rs− s ∂r rs−
!
∂ψ rs+ ∗ ∂ψ ∗ rs+

in
≈ r ψ − r ψs δ(r − rs ) (7.13)
4µ0 rs ∂r rs− s ∂r rs−
The above equation gives the flux averaged torque. To find the total torque in [N m], it
has to be integrated over the volume of the resistive layer. Due to the δ-function in (7.13)
this means one should multiply (7.13) – without the δ-function – by 4π 2 R0 rs . This results
in: !
rs+ ∗ rs+

in ∂ψ ∂ψ
Tφ,EM = 4π 2 R0 r ψ∗ − r ψs [N m] (7.14)
4µ0 ∂r rs− s ∂r rs−
As seen in equation (2.37) of chapter 2, in a steady state situation the toroidal force
balance over the whole plasma is given by:
R2 d

dρΩφ
− 0 rD = hTφ,EM i + hTφ (r 6= rs )i, (7.15)
r dr dr
where hTφ,EM i is only non-zero at the rational surface (see equation (7.13) ) and hTφ (r 6=
rs )i is the sum of all torques applied onto the plasma – friction due to neutrals coming from
the wall, torque induced by neutral beams, EM torques at other resonant surfaces, biasing
with radial electric fields, et cetera . . . – outside of the position of the rational surface.
The effective diffusion coefficient D in (7.15) is related to the anomalous, perpendicular
viscosity η⊥ = ρD, ρ being the mass density.
Integrating equation (7.15) over the resistive layer, taking into account that hTφ (r 6=
rs )i is assumed zero over this layer, a torque balance between the EM torque and a viscous

torque over the resistive layer is derived:

Z Z Z 2
1 R0 d dρΩφ
hTφ,EM i + rD dV = 0
Vlayer layer r dr dr
hTφ,EM i + hTφ,V S i = 0
2 dρΩφ rs+

R0
with : hTφ,V S i = rD δ(r − rs ) (7.16)
rs dr rs−
The viscous torque hTφ,V S i is associated with a discontinuity in the first derivative of
the rotation over the hresistiveilayer. In case there is no EM torque present, no viscous
dρΩ rs+
torque is needed and rD dr φ = 0. Therefore the viscous torque can be rewritten as:
rs−
R02 dρ ∆Ωφ rs+

hTφ,V S i = rD δ(r − rs ), (7.17)
rs dr rs−
with ∆Ωφ = Ωφ − Ωφ,0 , Ωφ,0 being the ‘natural’ rotation, i.e. in the rotation in the case
no EM torque is present.
Let us now assume that the rotation at the edge is zero both in presence and in absence
of an EM torque. This means that when an EM torque is present the natural rotation
profile inside rs is lifted, or reduced, by ∆Ωφ , while outside rs the rotation goes from Ωφ
to zero. This is sketched in figure 7.2.
Ωϕ ∆Ωϕ
∆Ωϕ
⇒
rs a rs a
Figure 7.2 : At the left a rotation profile without EM torque and the rotation profile in presence
of an EM torque is plotted. The corresponding ∆Ωφ profile is plotted on the right. It is
seen that inside rs , d∆Ωφ /dr = 0, and outside rs , d∆Ωφ /dr is proportional to ∆Ωφ . This
explains why the viscous torque is proportional to ∆Ωφ .
r
It is reasonable to assume that [dρ/dr]rs+
s− = 0, such that (7.17) can be rewritten as:
R02 Dρ
hTφ,V S i = − ∆Ωφ δ(r − rs ). (7.18)
a − rs
The above expression shows that the viscous torque is proportional to ∆Ωφ and
opposes any change in rotation. This proportionality does not depend on the specific
profile sketched in figure 7.2: as long as the torques acting on the plasma outside q = m/n
do not depend on the perturbation, the viscous torque is proportional to ∆Ωφ . Because
the plasma rotation is through equation (7.1) linked with the mode rotation ωt , the
viscous torque is also proportional to ∆ωt = ωt − ωt,0 . And, assuming the frequency of

the perturbation field is a constant, it is even proportional to the change in slip frequency
∆ω = ω − ω0 .
Finally we summarise the above section by giving the three equations are necessary to
describe the growth and the rotation of an island:
dW
0.8227 τR = rs2 ∆0 (7.19)
dt !
∂ψ rs+ ∗ ∂ψ ∗ rs+

2 in
Tφ,EM = 4π R0 r ψ − r ψs (7.20)
4µ0 ∂r rs− s ∂r rs−
Tφ,V S ∝ ∆ω (7.21)
7.2.3 Effect of an external helical perturbation

The general equation for the island growth and the electromagnetic torque were given in
the previous section. In this section we will try to find the practical form of the Rutherford
equation and the EM torque in the situation of an external perturbation field.
Because the plasma is a conducting medium, it does not allow a change in magnetic
flux. When applying a perturbation field, Lenz’ law predicts that shielding currents will
be induced that cancel out the external field.
For the (m,n) Fourier component of the external perturbation a shielding current with
(m,n) helicity is needed. This current will flow there where the resistivity is lowest, i.e.
where it can flow parallel to the magnetic field. For a (m,n) helical current this is at the
q=m/n rational surface. This means that the (m,n) component of the external field is
shielded inside the q=m/n surface, as can be seen in figure 7.3.
1
ψvac
q = m/n
ψcoil
0.8
0.6
|ψ|
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
r/rc
Figure 7.3 : The (m,n) component of the external perturbation field in vacuum (ψvac )
and in presence of plasma (ψcoil ).
The (m,n) perturbation can be seen as produced by a helical current in a coil at a

position r = rc outside the plasma. In the experiments with the DED this is exactly the

case, for intrinsic error fields this is an approximation. In the absence of plasma this coil
current causes a perturbed vacuum flux Ψvac = ψvac exp(i[mθ + nφ]):
m
r
ψvac (r, t) = ψc (t) (0 < r < rc− )
rc
−m
r
ψvac (r, t) = ψc (t) (rc+ < r), (7.22)
rc
where ψc (t) is the value of ψvac inside the coil and is proportional to the current applied
to the coil.
In presence of plasma the total perturbed flux ψ, that is resonant at rational surface
rs , can be split up into a part representing the flux in absence of an external perturbation
(ψfree ) and a part attributed to the perturbation current in the coil (ψcoil ). As mentioned
above the part attributed to the perturbation current in the coil is shielded from the
resonant surface inwards. So a first constraint applied to ψcoil is that, due the shielding
currents at the resonant surface, it is zero for r < rs+ .
As said before a helical current with helicity (m,n) in a thin layer is linked with a jump
in the derivative of ψ over this layer (see equation (7.7) ). At the position of the coil this
means:
∂ψ rc+

r = −2mψc . (7.23)
∂r rc−
As ψfree is the flux in absence of an external perturbation, its derivative shows no jump
at rc , and thus (7.23) puts a second constraint on ψcoil . With the above two constraints,
the expression for ψcoil becomes:
ψcoil (r, t) = 0 (0 < r < rs+ )
" m −m #
r r
= ψvac (rs+ , t) − (rs+ < r < rc− )
rs+ rs+
" m −m # −m
rc rc r
= ψvac (rs+ , t) − (rc+ < r)
rs+ rs+ rc
(7.24)
This is plotted in figure 7.3.
Moving along with the mode, the total perturbed flux ψ has a stationary phase ϑ:
ψ = |ψ| exp(iϑ). The part of the flux attributed to the external perturbation, has the phase
of the vacuum field: ψvac (r, t) = |ψvac (r)| exp (i(ϑc + ωt)), where ϑc is the stationary phase
and ωt is de time-dependent phase due to the difference between perturbation field rotation
and mode rotation. Here it is assumed that the slip frequency ω is quasi-stationary.
The current δjφ in the resistive layer can now be written as:
Z Z Z
2µ0
− 2 δjφ exp [−i(mθ + nφ)] dV
4π R0 Vlayer
∂ψ rs+

0
= rs ∆ ψ s = r
∂r rs−
∂ψfree rs+ ∂ψcoil rs+

0
= rs ∆ ψ s = r + r
∂r rs− ∂r rs−
= rs ∆0 ψs = rs ∆0free ψs + 2m ψvac (rs+ , t), (7.25)

where ψs = ψ(rs ) = ψfree (rs ), because ψcoil = 0 for r < rs+ .

Using (7.25) and the Rutherford equation (7.9), one can easily rewrite the Rutherford
equation as:
Wvac 2

dW 2 0
0.8227 τR = rs ∆free + 2m rs cos ∆ϑ, (7.26)
dt W
where Wvac ∼ |ψvac (rs )|1/2 is the vacuum island width and ∆ϑ = ϑc + ωt − ϑ is the phase
shift between the actual plasma island and the vacuum island.
When including equation (7.25) into the definition of the flux averaged EM torque
(7.13), the expression becomes:
mn
hTφ,EM i = − |ψs ||ψvac (rs )| sin ∆ϑ δ(r − rs ) (7.27)
µ0 rs
The total torque over the resonant layer is:
mn
Tφ,EM = −4π 2 R0 |ψs ||ψvac (rs )| sin ∆ϑ (7.28)
µ0
The above equations (7.25), (7.26) and (7.27) show that due to the presence of an
external perturbation the current in the resistive layer is changed. The part of the current
attributed to the external perturbation – the shielding current – is responsible for an extra
term in the Rutherford equation and a non-zero torque onto the plasma.
The effect of the external perturbation depends of course on the strength |ψvac | of the
perturbation field, but also, through the phase difference ∆ϑ = ϑc + ωt − ϑ, on the slip
frequency ω. In the next two sections we will consider two situations: ω = 0 and |ω| 0.
It will be found that the plasma response in both situations differs significantly, and that
the transition between the two situations is non-linear.
7.2.4 A locked mode

First we have a look at the case where ω = 0. The phase difference ∆ϑ is then stationary
and the mode is locked to the external perturbation field. The Rutherford equation (7.26)
has now two constant driving terms on the right hand side. This means that the island
width will grow until it reaches its equilibrium value. This equilibrium island width can
be found if we set the left hand side of equation (7.26) to zero.
In figure 7.4 the equilibrium width of the locked island is plotted as a function of the
vacuum width, (a), and as a function of the phase difference, (b). In general the width
can be approximated as:
r
m √
W ≈ W0 + f (Wvac ) cos ∆ϑ Wvac , (7.29)
2 rs ∆0free (0)
where f (Wvac ) is approximately one and only weakly depends on Wvac .

If the plasma has strongly saturated tearing modes without an external perturbation
applied (rs ∆0free (0) 0), then it can be seen that the width of the locked mode is hardly
affected by the perturbation field: W ≈ W0 . If the plasma is just marginally stable
with respect to tearing modes (rs ∆0free (0) ≈ 0), the locked island width is in most cases
enhanced with respect to the vacuum island Wvac and the natural island width W0 .
Due to the stationary phase difference ∆ϑ, the EM torque given by equation (7.27) is
stationary as well. In section 7.2.2 is was shown that the viscous torque is proportional

15 rs ∆’free(0) = 0.5 10
(a) (b)
rs ∆’free(0) = 1 8
10
6
W (cm)
W (cm)
rs ∆’free(0) = 4
4
5 rs ∆’free(0) = 32
rs ∆’free(0) = 512
2
W = W0
W = Wvac
0 0
0 5 10 π/4 π/2
Wvac (cm) ∆ ϑ (rad)
Figure 7.4 : The width of a locked island:

(a) The width as a function of the vacuum island width for different values of rs ∆0free (0)
and the phase difference ∆ϑ = 0.
(b) The width of the locked island as a function of the phase difference ∆ϑ between vacuum
and locked island. The width of the vacuum island
√ and rs ∆0free (0) were kept constant (5 cm
and 0.5 respectively). The island width has a cos ∆ϑ dependence.
to ∆ω and in the locked case ∆ω = −ω0 is constant. Thus a constant EM torque is

balanced by a constant viscous torque, when ω = 0. The balance of EM and viscous
torque reveals that |ψs ||ψvac | ∝ 1/ sin ∆ϑ. This means that the phase difference ∆ϑ will
increase when the perturbation field and/or the island width is decreased. This holds up
until ∆ϑ = 90◦ . Once the phase difference reaches 90◦ the constant EM torque can no
longer balance the viscous torque and the mode unlocks.
To conclude: if ω = 0 a locked island will be present at the rational q surface. The is-
land width is always larger than the natural island width in absence of a perturbation field.
If the island already had large islands before the perturbation is applied (rs ∆0free (0) 0),
the perturbation causes only a slight increase in island width. If only small or no islands
were present in the plasma before the perturbation (rs ∆0free (0) ≈ 0) , the perturbation
field will create large islands. When the perturbation field is decreased again the phase
difference ∆ϑ between the island and the perturbation field will increase towards 90◦ . At
90◦ the island unlocks (|ω| > 0).
7.2.5 Mode penetration
Non-linear theory A second situation is that where |ω| 0. The phase difference
∆ϑ is then time-dependent. For simplicity we assume ∆ϑ = ωt. A result of the time-
dependent phase difference is that the second driving term in the Rutherford equation
(7.26) oscillates. Because in the measurements done at TEXTOR there were no islands
present in the plasma before the DED field was switched on, we will here look into the
situation where ∆0free (0) ≤ 0. For a very small perturbed flux ψ, and thus W ≈ 0, only
the second term on the right hand side of equation (7.26) is of importance:
2
dW Wvac
0.8227 τR = 2m rs cos ωt, for W ≈ 0, (7.30)
dt W

Equation (7.30) shows that the island will periodically be stabilised and destabilised,
leading to an island that constantly grows and shrinks (figure 7.5) [29]:
2
1/3
mrs Wvac
W (t) = 1.939 | sin ωt|1/3 (7.31)
|ω|τR
−3
x 10
7
6 ω = 2 π x 2 kHz, Wvac = 20 mm, rs = 0.3 m, τR = 1.5 s, m = 2

4
5
Tφ, EM (Nm)
W (mm)
4 3
3 2
2
1
1
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time (ms)
Figure 7.5 : The solid line shows the time evolution of the width of a magnetic island, with slip
frequency |ω| 0, that is driven by an external perturbation field. In half a period the island
grows, reaches a maximum and shrinks to zero again. The higher the slip frequency |ω|,
the lower the maximum width that the island can reach. The maximum width is also a lot
smaller than the width of the vacuum island. We therefore call the island ‘suppressed’.
The dashed line shows the EM torque. This torque also oscillates between zero and a max-
imum value. The torque is always directed such that it tries to reduce |ω|. The maximum
torque depends on the maximum island width, and is therefore smaller for higher |ω|.
Not only the island width oscillates, also the EM torque (7.28) oscillates: due to the
time-dependent phase and due to the oscillating island width. This oscillating torque, also
drawn in figure 7.5, tries to reduce |ω|.
Due to its viscosity the plasma is not capable of following the fast oscillation. The
viscous torque thus balances the EM torque averaged over one period. The oscillating and
averaged torqued are given by [29]:
2/3
4π 2 R0 m n

mrs
Tφ,EM (t) = −3.7597 |ψvac (rs )|2 | sin ωt|5/3 (7.32)
µ0 ωτR Wvac
2/3
4π 2 R0 m n

mrs
Tφ,EM = −2.0137 |ψvac (rs )|2 (7.33)
µ0 ωτR Wvac
It is important to notice that the maximum width of the island is inversely proportional
to |ω|1/3 and therefore very small for |ω| 0. In fact the maximum width of the islands
is too small and their growth and shrinking too fast, for them to be ‘seen’ by the plasma:
the parallel transport from one side of the island to the other along the separatrix is of
the same order as the cross-field transport over the island. Because the magnetic islands
cannot be observed in this phase it is called the ‘suppressed’ island phase.

Is the non-linear theory valid? The fact that in the suppressed phase the islands
are very small has some serious consequences. In the derivation of equation (7.26) it is
assumed that the island moves along with the electron fluid: ψ has a stationary phase
ϑ. This is referred to as the ‘no slip’ constraint. If the island is large enough the ‘no
slip’ constraint is fulfilled: there is no net electron flow over the island separatrix. More
correctly we should say that there is a small convection layer around the separatrix; over
this convection layer there is no flow, but inside this convection layer there is. If the island
becomes so small that the width of the convection layer is of the same order as the island
width, then there can be a net electron flow over the separatrix: the island slips through
the electron fluid and the ‘no slip’ constraint is relaxed. This is shown in the cartoon of
figure 7.6 where an unperturbed situation, the situation with a small perturbed flux and
a relaxed ‘no slip’ constraint, and the situation where a large island is subject to the ‘no
slip’ constraint are depicted.
(a) (b) (c)
Figure 7.6 : Cartoon of the relation between the electron flow and magnetic perturbations.
(a) No perturbation, free electron flow.
(b) Small perturbation, electron fluid can slip over the island.
(c) Large perturbation, ‘no slip’ constraint: the island moves with the electron fluid.
If the ‘no slip’ constraint is relaxed, the derivation of Rutherford equation (7.26) is
no longer valid. Instead of the non-linear approach using the Rutherford equation, we
should use a linear approach.
We could now ask ourselves whether in TEXTOR the maximum island width Wmax =
2 )/(|ω|τ ) 1/3 given by the non-linear theory is large enough to have ‘no

1.939 (mrs Wvac R
slip’. In other words: Is the non-linear theory given above valid?
The non-linear theory is valid when the maximum island width Wmax is larger than
1/3 1/6 1/6
the linear layer width. This linear layer width is defined as δlayer = 2.56 τH /(τR τV ),
with τV the viscous time constant, which is in order of the energy confinement time, τH
the local hydromagnetic timescale, which is inversely proportional to the Alfvèn speed,
and τR the resistive diffusion time [28, 29]. For TEXTOR τV ≈ 0.03 s, τH ≈ 1.5 · 10−7 s
and τR ≈ 1.5 s are typical values. This results in a δlayer of 6.4 mm. The maximum island
width in figure 7.5 is below this value!

The expression δlayer = Wmax gives us a line in the (Wvac , ω)-plane. Below this line
δlayer < Wmax and the non-linear theory is valid, above this line δlayer > Wmax and the
linear approach has to be used.
The vacuum island width√in TEXTOR is proportional to the square root of the DED
current: Wvac [mm] ≈ 22.8 · IDED [kA]. This means that the line dividing the (Wvac , ω)-
plane can be converted in a line dividing the (IDED , ω)-plane into a linear and a non-linear
region, as is shown in figure 7.7.
With |ω| 0 and with the DED current increasing from zero, we thus start in the
linear region of the (IDED , ω)-plane. In TEXTOR we should therefore at least start with
the linear theory and not with the non-linear expressions (7.31) and (7.33) that are given
above. The dashed line in figure 7.7 gives the evolution of ω as a function of IDED up to
the point of the mode excitation threshold (see later). It shows that during the whole
DED current ramp, up to the moment of mode excitation, the linear theory is valid. We
will therefore us the linear theory to describe the evolution of the perturbed flux |ψ| and
the EM torque, not the non-linear theory.
4
x 10
2
Wmax < δlayer ⇒ linear approach

1.5
ω (IDED)
ω (rad/s)
0.5
Wmax > δlayer ⇒ non−linear approach

0
0 0.5 1 1.5 2
IDED (kA)
Figure 7.7 : Depending on ω and IDED linear or non-linear theory has to be used. Before the
DED is switched on, IDED = 0 and |ω| 0. When IDED is ramped up, we therefore start in
the linear region. ω will reduce according to the linear theory and remains during the whole
current ramp in the linear region. To describe the evolution of the perturbed flux and the
electromagnetic torque in TEXTOR, we will therefore use the linear theory.
Linear theory We have come to the conclusion that we have to use a linear approach
with small islands that slip through the electron fluid. The fact that the islands slip
through the electron fluid means that the flux ψ, seen from within the moving elec-
tron fluid, has not a stationary but a time-dependent phase. We assume that the time-
dependence of ψ is the same as that of the perturbation field. Moving with the electron
fluid this means: ψ(r, t) = |ψ(r)| exp (i(ϑ + ωt)) and ψvac (r, t) = |ψvac (r)| exp (i(ϑc + ωt)),
where ϑ and ϑc are the stationary phases and ωt is the time-dependent phase. The phase
difference between ψ and ψvac is then constant in time (∆ϑ = ϑc − ϑ), leading to a
stationary EM torque given by equation (7.28).

The expression for ψ that follows from linear theory is [28, 29]:
2m
ψ(rs ) = ψvac (rs ), (7.34)
−rs ∆0 + rs ∆layer (ω)
where ∆0 = ∆0free < 0, because the plasma is intrinsically stable with respect to tearing
modes, and ∆layer (ω) is a complex function of the slip frequency that depends on inertia,
viscosity and resistivity.
Most tokamaks are in the so-called visco-resistive regime, where the influence of vis-
cosity and resistivity are of the same order, whereas inertia plays an unimportant role:
1/3 2/3 2/3
|ω| τV /τH τR . Using the values of τV , τH and τR given above, TEXTOR is in the
visco-resistive regime as long as |ω| 104 rad/s. The slip frequency |ω0 | before the DED
is applied is usually in the order of or slightly higher than 104 rad/s. During the DED
ramp up |ω| will however reduce. We will therefore assume that the visco-resistive regime
can be applied in TEXTOR.
In the visco-resistive regime the complex function ∆layer takes following form:
1/3 5/6
!
2.1036 τH τR
∆layer (ω) = 1/6
ω exp(iπ/2) = i|∆0 | τrec ω, (7.35)
rs τV
1/3 5/6

2.1036 τH τR
where τrec = |rs ∆0 | 1/6 is the reconnection timescale; for TEXTOR τrec ≈ 3 ms if
τV
|rs ∆0 | = 10. Using (7.35) the expression for ψ becomes:
1/2
2m 1
ψ= |ψvac | exp [i(arctan(ωτrec ) + ϑc + ωt)] (7.36)
|rs ∆0 | 1 + ω 2 τrec
2
Substituting (7.36) into the expression for the EM torque (7.28), where ∆ϑ =
arctan(ωτrec ), yields:
2 m2 n ωτrec
Tφ,EM = 4π 2 R0 |ψvac (rs )|2 (7.37)
µ0 |rs ∆0 | 1 + ω 2 τrec
2
Mode penetration We saw that both the non-linear theory (7.33) and the linear theory
(7.37) resulted in an electromagnetic torque that:
– tries to reduce |ω|
– increases with increasing |ψvac |
– increases with decreasing |ω|.
For TEXTOR we found that the linear expression (7.37) of the EM torque has to be
used.
The EM torque is balanced by the viscous torque. This viscous torque is proportional
to ∆ω = ω − ω0 . Equation (7.37) shows that for |ω| 1/τrec , the EM torque can be
reduced to Tφ,EM ∝ |ψvac |2 /ω. For TEXTOR the condition |ω| 1/τrec = 333 rad/s is

easily achieved, such that the balance between the EM torque an the viscous torque can
be written as:
|ψvac |2
+ A∆ω = 0
ω
|ψvac |2 = −Aω(ω − ω0 ) (7.38)
√ The right hand side of (7.38) has √ a maximum for ω = ω0 /2. The corresponding |ψvac | is
A/2 |ω0 |. If |ψvac | increases above A/2 |ω0 |, the EM torque can no longer be balanced
by the viscous torque. At that point ω drops to zero. We saw in the previous section that
once ω = 0 large islands develop in the plasma. The transition from ω 0 to ω = 0 is
therefore called mode excitation or mode penetration. The value of |ψvac | at which this
jump to ω = 0 happens is called the excitation or penetration threshold; hence:
√
A
|ψthreshold | = |ω0 |. (7.39)
2
This means that the torque balance between the EM torque and the viscous torque results
in a mode penetration threshold that is proportional to the slip frequency |ω0 |.
With the expression for the threshold known, we can write the proportionality factor
A as a function of |ψthreshold |. This leads to an expression for the change in ω as a function
of |ψvac |:  s 
2
ω0  |ψvac |
ω= 1+ 1−  (7.40)
2 |ψthreshold |
To conclude we summarise the process of mode penetration (see figure 7.8).

• At |ψvac | = 0 the slip frequency |ω| = |ω0 | 0. There is also no island in the plasma,
thus |ψ| = 0.
• When |ψvac | is increased, the slip frequency reduces, following equation (7.40). The
flux in the plasma |ψ| is according to equation (7.36) proportional with |ψvac |, but
also inversely proportional with 1 + ω 2 τrec
2 . Because |ω| remains large, |ψ| is very
low and thus the islands are suppressed.

• |ω| reduces until, at |ψvac | = |ψthreshold |, it reaches ω = ω0 /2. At this threshold the
viscous torque can no longer balance the EM torque and ω drops to zero. Once
ω = 0, |ψ| is only proportional to |ψvac |. Large islands suddenly appear in the
plasma. For |rs ∆0 | < 2m, |ψ| > |ψvac | and hence the islands in the plasma will be
larger than the vacuum islands.
Locking to the wall and mode coupling In the final paragraph of this section we
shortly mention two other processes, that are related to mode penetration:
• In a toroidal geometry the total perturbed flux of a (m,n) island has, apart from
the main (m,n) Fourier component, (m ± 1, n) components as well [16, 30]. If these
components are sufficiently large, modes at the q = m±1 n surfaces can lock to the
(m,n) island. This process is called ‘mode coupling’ and is similar to the non-linear
theory of mode penetration.

Suppressed island Excited island
1 1
0 0.8 0.8
0.6 0.6
ω/ω
|ψ|
0.4 0.4
0.2 0.2
0 0
0 0.2 0.4 0.6 0.8 1
|ψ |
vac
Figure 7.8 : The slip frequency ω (solid line) and the amplitude of the perturbed flux |ψ| (dashed
line) are drawn as a function of the vacuum flux |ψvac |.
We assume |ω| = |ω0 | 0 before the perturbation field is applied. When the perturbation
field is increased, |ω| reduces towards |ω0 |/2. The amplitude of the flux |ψ| is inversely
proportional to 1 + ω 2 τrec
2
. Because |ω| remains above |ω0 |/2, |ψ| is small and the islands
are suppressed: they can not be observed by diagnostics.
Once |ψvac | has increased to |ψthreshold |: the value where |ω| = |ω0 |/2, the threshold for mode
penetration is reached. At this point the viscous torque can no longer balance the EM torque
and ω drops to zero.
With ω = 0 the flux |ψ| is no longer suppressed. On the contrary: usually |ψ| > |ψvac |. In
the plasma large islands can be observed that are locked to the perturbation field.
• Suppose there is an island present in the plasma. In general the flux of this island
will not be zero at the wall of the tokamak |ψ(rwall , t)| > 0. If the wall is conducting
– and usually a tokamak wall is – currents will be induced to keep the ‘island field’
out. These shielding currents will give rise to an EM torque that works on the wall,
but also on the island. Again, depending on the force balance between viscous torque
and EM torque, the island will slow down and finally it could lock to the wall. The
physics of locking to a conducting wall is very similar to the linear theory of mode
penetration [28].
• The process of mode locking is not only important for magnetic islands. Ideal kink
modes, a deformation of the flux surfaces without reconnection, become so-called
resistive wall modes (RWM) when the tokamak wall is resistive. Just as in the case
of tearing modes, a RWM will cause a non-zero flux at the wall of the tokamak
|ψ(rwall , t)| > 0, hence it will induce currents in the tokamak wall. When the RWM
rotates (i.e. when the plasma rotates), the wall currents will stabilise the RWM: it
will be suppressed. The interaction between the suppressed RWM and the stabilising

currents will however also slow down the RWM. If the rotation frequency of the RWM
becomes too low, it will lock to the wall and be destabilised, causing a disruption.
The process of RWM locking, tearing mode locking and tearing mode excitation can
all be described within the same philosophy. Investigating tearing mode excitation
can therefore give us insight in RWM locking, which is an important limiting factor
in high β machines like ITER.
7.3 The effect of an external perturbation in the plasma

edge
7.3.1 Introduction
In the previous section the reason why the mode penetration threshold is proportional
to |ω0 | and the reason why |ω| monotonically reduces towards |ω0 |/2 and then locks,
was clarified. It was the result of a torque balance between a EM torque at the rational
q surface and a viscous torque that was proportional to ∆ω. One could expect that
the threshold and the evolution of ω would change if an extra torque is introduced into
the torque balance: a torque that changes with |ψvac | but is not located at a rational q
surface. The question is: Does such a torque exists?
7.3.2 The stochastic torque in the plasma edge

So far we focussed on the plasma core. Let us now have a look more towards the edge of
the plasma. What is the effect of an external perturbation field there?
• As mentioned above the (m,n) component of the external field is shielded by a parallel
current at the corresponding rational q surface. Outside the q surface however the
external field increases towards the edge, so that at the position of the perturbation
coil (rc ) ≈ ψvac (rc ) (see figure 7.3). Therefore the total residual field Ψres =
P ψcoilm,n
m,n ψcoil (r, t) is non-zero, be it non-resonant.
• Very close to the plasma edge, the field lines connect to the wall and no parallel,
shielding currents can flow. In this laminar zone the perturbation field is not shielded
at all.
• Finally, at rational q surfaces close to the edge the resistivity will be higher, thus
τrec lower. On top of that, because the edge is closer to the perturbation coils, |ψvac |
will be high at these q surfaces. At rational q surfaces close to the edge islands are
therefore possibly excited at very low |ψvac |. If these excited islands start to overlap
they create a so-called ergodic zone where the field lines fill a volume rather than
lying on a flux surface.
The above three considerations indicate that near the plasma boundary the magnetic
field, in presence of an external perturbation field, will differ significantly from the equilib-
rium field. This is especially the case when the perturbation field consists of a superposition
of many different Fourier components. In general one can say that in the outer region of
the plasma the perturbed field within the plasma is similar to the vacuum field as seen in
Section 7.3 - The effect of an external perturbation in the plasma edge 111
figure 3.5: with a laminar zone that has open field lines and an ergodic zone where the
excited tearing modes overlap [35].
Both in the laminar as in the ergodic zone the field lines fill a volume, rather than
lying on a flux surface; i.e. the field lines have a radial component. The region where field
lines fill a volume will be called the ‘stochastic’ region from here onwards; without making
the distinction between ergodic and laminar.
Parallel transport along the field lines is larger for electrons than it is for ions, due
to their mass difference. Because the field lines in the stochastic region have a radial
component, the difference in parallel ion and electron velocity results in a net radial current
k
jr . This radial current cannot be stationary: radial electric fields build up resulting in a
perpendicular current. The radial component of this perpendicular current (jr⊥ ) balances
that of the parallel current, such that the total radial current in steady-state is zero
k
(jr = jr + jr⊥ = 0). Parallel and perpendicular currents, and their radial components are
sketched in figure 7.9.
Figure 7.9 : In the stochastic region a parallel electron loss current exists, due to the higher
electron mobility compared to ion mobility and the radial component of the field lines. This
leads to charge separation and the build up of a radial electric field. This radial electric field
will drive a perpendicular return current. The sum of the radial components of the parallel
loss current and the radial return current is zero, such that the ambipolarity constraint is
fulfilled. [11].
The parallel current will of course not cause any j × B torque on the plasma, but the
perpendicular current will. This torque can written as:
Tφ,ST = R jr⊥ Bθ (7.41)
The radial electron loss current is directed inwards, which means that the radial compo-
nent of the perpendicular return current is directed outwards. When we take into account
that the direction of the poloidal field Bθ is linked with the plasma current, we find that
the stochastic torque Tφ,ST is always directed in the co-current direction.
k
The value of jr⊥ = −jr depends on the collisionality in the plasma edge, given by the
mean free path λmf p , and on the amount of stochasticity, given by a characteristic length
Lchar . This characteristic length is the Kolmogorov length LK – the 1/e-length of two
exponentially diverging field lines – in the ergodic zone and the connection length Lc –
the length of a field line that goes from wall to wall – in the laminar zone. The expression
k
for jr is [45]:
112 Section 7.3 - The effect of an external perturbation in the plasma edge

Df l Te d ln ne 1.71 dTe
λmf p Lchar : jrk = σk Er + +
Lchar e dr e dr
r
k 2 2 Te d ln ne 0.5 dTe
λmf p Lchar : jr = iσ Df l ne e Er + + , (7.42)
me T e π e dr e dr
where σk is the parallel resistivity, Df l is the field line diffusion and iσ is function of parallel
and perpendicular resistivity. The edge region in TEXTOR has a fairly high collisionality
(plateau regime), so the first expression in (7.42) is applicable.
In (7.42) Df l , Lchar , Te (r) and ne (r) all depend on |ψvac |. From vacuum field calcula-
tions it follows that: Df l /Lc ∼ |ψvac |8/3 [1, 35]. The dependence of Te (r) and ne (r) is not
easily derived from theory and therefore has to be determined from experiments.
The radial electric field Er in equation (7.42) is however unknown. It is, through
Ohm’s law, related to the plasma rotation velocity. The plasma rotation velocity in turn
depends on the torque (7.41).
Thus, in contrast to the EM torque that was discussed in section 7.2, the stochastic
torque is not a simple function of |ψvac |. A solution for Tφ,ST can only be found by solving
equation (7.42), Ohm’s law, the toroidal (2.37) and the poloidal (2.39) force balance in a
self-consistent way.
7.3.3 Conclusion
In the edge of the plasma the magnetic field can be stochastic. This is usually the case if
the perturbation field has several different Fourier components.
In this stochastic region parallel transport causes a loss of electrons to the wall. This
results in the build up of an electric field that drives a perpendicular return current, that
assures ambipolarity. The return current gives rise to a j × B-torque in the stochastic
region.
The stochastic torque is always directed in the co-current direction. A simple expres-
sion for this torque is however not available: the expression for the loss current, Ohm’s
law and the force balance equations have to be solved simultaneously. The lack of a simple
expression makes it difficult to include the stochastic torque directly in the torque balance
for mode penetration.
7.4 Conclusion
The application of an external perturbation field has two main effects on a plasma. First
of all shielding currents will flow on the rational q surfaces that shield the corresponding
Fourier components of the external field. These shielding currents will change the stability
of the plasma with respect to modes: they add an extra term in the Rutherford equation
that describes the evolution of modes. If the slip frequency ω between modes in the
plasma and the perturbation field is large, then the shielding currents are responsible for
a suppression of the modes or islands. The islands are undetectable. If the slip frequency
becomes zero however, the islands grow and become even larger than vacuum islands of
the perturbation field.

The shielding currents will also cause an electromagnetic torque onto the plasma at
the position of the rational q surface. This torque will try to bring the slip frequency ω to
zero. The EM torque is balanced by a viscous torque that is proportional to the difference
between the slip frequency ω0 , in absence of an external perturbation, and the actual slip
frequency ω.
If only these two torques are present ω will go from ω0 to ω0 /2 when the perturbation
field is increased. At ω0 /2 the viscous torque can no longer balance the EM torque and
ω drops to zero. This is called mode excitation, because the plasma goes from a phase
with suppressed islands into a phase with large islands. Because the threshold occurs at
ω = ω0 /2, the threshold is proportional to ω0 .
A second effect of the external perturbation field lies in the plasma edge. There a
stochastic region is created. Along the field lines in the stochastic region there is a parallel
loss of electrons that results in a stochastic torque. This stochastic torque is always
directed in the co-current direction. The strength of the stochastic torque depends on
the amount of stochasticity in the plasma edge. There is however no straightforward
expression for the stochastic torque. It can only be found by solving the expression for
the loss current, Ohm’s law and the force balance equations simultaneously. The DED
in TEXTOR was designed to significantly disturb the plasma edge, thus the stochastic
torque is expected to be significant and can not be neglected.
When describing mode penetration at TEXTOR we have to include three torques in the
torque balance: the EM torque at the rational q surface, the viscous torque that opposes
changes from the natural slip frequency ω0 and a stochastic torque that is always directed
in the co-current direction. Because there is no simple expression for the stochastic torque,
it is difficult to include it in the torque balance. In the next chapter we will therefore try
to derive a straightforward expression for the stochastic torque at TEXTOR.

Chapter 8
Momentum input by a stochastic

edge field
8.1 Introduction
Two torques can be attributed to the DED: a resonant electromagnetic torque at the
rational q surface and a stochastic torque in the edge region of the plasma due to electron
loss current. The first tries to bring the slip frequency ω to zero, thus causing mode
excitation. The latter is always directed in the co-current direction. If ‘bringing ω to zero’
requires a change of the rotation in the co-direction, the stochastic torque will facilitate
the mode excitation; if it requires a change in the counter-direction then the stochastic
torque opposes mode excitation.
The DED in TEXTOR is designed to significantly disturb the plasma edge. As a
consequence the stochastic torque can not be ignored. The torque balance that describes
the mode excitation has to include the EM torque, the viscous torque and the stochastic
torque.
We saw in the previous chapter that there is no simple expression for the stochastic
torque as a function of the perturbation field. In order to find the stochastic torque, the
expression for the electron loss current, Ohm’s law and the toroidal and poloidal force
balance equations have to be solved in a self-consistent way. The lack of a straightforward
expression of the stochastic torque makes it difficult to include it into the torque balance
with the EM and the viscous torque.
The aim of this chapter is (a) to get the value of the stochastic torque and the change in
rotation by solving the set of equations; and (b) try to find an straightforward expression
for the stochastic torque that only depends on the perturbation strength and measured
quantities like temperature and density, but does not depend on the rotation itself.
8.2 12/4 DED operation

Because we want to investigate the stochastic torque, we would like to have experimental
conditions where we can neglect all other torques, like e.g. the electromagnetic torques at
the rational q surfaces.
m,n 2
In the previous chapter we saw that the EM torque is proportional to m2 |ψvac | , with
m,n
|ψvac | the resonant component of the vacuum flux at rational the q surface. This is due
115
Chapter 8 - Momentum input by a stochastic edge field
m,n
to the fact that the EM torque is caused by the shielding current. If |ψvac | is large, than
a high current is needed to shield the vacuum field and consequently the EM torque is
large.
The vacuum flux as a function of radius is given by:
m
m,n m,n r
ψvac (r, t) = ψc (t) (0 < r < rc− ), (8.1)
rc
which means that the vacuum flux drops more quickly for higher m-numbers. The DED
has several modes of operation with principle mode numbers m/n: 3/1, 6/2 and 12/4.
The Fourier components of the DED field will have the highest m-numbers in 12/4 mode
– centred around m = 12 (see figure 3.3) – and the lowest in 3/1 mode – centred around
m = 3 (see figure 3.4). This suggests that, in order to reduce influence of the EM torques
at the main rational q surfaces (q = 1, q = 2 and q = 3), we should use the DED in 12/4
mode.
The vacuum flux may be lower at the rational q surfaces in 12/4 mode than in 3/1
mode, but in order to neglect the EM torque in 12/4 mode the vacuum flux must be
significantly lower. We therefore need to compare the resonant components of the vacuum
1,1 2,1
flux in 3/1 and 12/4 mode. For 3/1 these resonant components are |ψvac | at q = 1, |ψvac |
3,1 4,4 8,4
at q = 2 and |ψvac | at q = 3. For 12/4 the resonant components are |ψvac | at q = 1, |ψvac |
12,4
at q = 2 and |ψvac | at q = 3.
−3
x 10
3.5
q=1 q=2 q=3 2
3
2.5
m/n = 1/1
|ψvac| (Tm)
2
m/n = 2/1
q
1.5 1
m/n = 3/1
1 m/n = 12/4
0.5
m/n = 8/4
0 m/n = 4/4
0 0.1 0.2 0.3 0.4 0.5
r (m)
Figure 8.1 : The vacuum flux of field components resonant at q = 1, q = 2 and q = 3 are plotted
in blue for 3/1 DED operation, in red for 12/4 operation. One clearly sees that the resonant
components of the vacuum flux at the rational q surfaces are a lot stronger in 3/1 operation
than in 12/4 operation. This means that the electromagnetic torque, that depends on the
vacuum flux, is in 12/4 DED operation negligible at the q = 1, q = 2 and q = 3 surfaces.
In figures 3.3 and 3.4 the Fourier components of perturbation field δBrm,n at
the q = 3 surface are given. The plasma and DED conditions are comparable:
3,1 12,4
Ip = 400 kA, Bφ = 1.9 T, IDED = 1.5 kA, IDED = 7.5 kA1 . Using equations (7.3) and
1 3,1 12,4
IDED = 1.5 kA and IDED = 7.5 kA are comparable because in 3/1 mode 4 coils are grouped together
116 Section 8.2 - 12/4 DED operation

m,n
(8.1) we can derive the resonant |ψvac (r)| components in 3/1 and 12/4 operation. The
resonant components in 3/1 mode are given by the blue lines in figure 8.1, the resonant
components in 12/4 mode are plotted in red. We immediately see that the vacuum flux at
the rational q surfaces is significantly lower in 12/4 mode, compared to 3/1 mode. At the
0
whichis the most sensitive to the EM torque due to a low |rs ∆ |, the ratio
q = 2 surface,
8,4 2,1
8 |ψvac | / 2 |ψvac | = 0.04 1. This means that the EM torque (Tφ,EM ∝ m2 |ψvac |2 )
at q = 2 is 600 times smaller in 12/4 mode than it is in 3/1 mode. This is small enough
to neglect it.
In the plasma edge however a stochastic zone is still present. In the case of 12/4 DED
this stochastic region is dominated by the laminar zone [56]. In the stochastic region
an electron loss current is responsible for the stochastic torque. We will discussed this
stochastic torque in the next section.
8.3 Analysis of discharge #95592

In this section the effect of the stochastic torque will be discussed. This work is a con-
tinuation of the work presented in [11]. Where the focus in [11] was mainly on the effect
on poloidal rotation and this for two situations: no DED and full DED perturbation;
the analysis here will concentrate on toroidal rotation and this during the whole DED
pulse: ramp up, flat top, ramp down. We will use the same TEXTOR discharge as in
[11]: #95592. For this discharge a large data set was available, especially concerning the
poloidal rotation and the edge density and temperature.
Self-consistent solution The stochastic torque is a result of the parallel loss current
due to the electrons in the stochastic region. In order to fulfill ambipolarity this loss
current is balanced by a perpendicular return current jr⊥ , leading to a stochastic j ⊥r × B-
⊥
force. The expression of the perpendicular return current jr depends on the radial electric
field Er . The radial electric field depends on the poloidal vθ and toroidal vφ velocity. And
the poloidal and toroidal velocity depend on the stochastic torque j ⊥ r × B.
We therefore have to solve a set of four equations with 4 unknowns (jr , Er , vφ and vθ ):

jr⊥ = −σk Er + + (8.2)
Lc e dr e dr
1 dpi
Er = vφ Bθ − vθ Bφ + (8.3)
ne e dr

R0 d d
− rD (nmΩφ ) = jr⊥ Bθ − nmνcx vφ + FN BI (8.4)
r dr dr
α(vθ − vθneo ) = −jr⊥ Bφ − (1 + q 2 )nmνcx vθ (8.5)
The first equation describes the radial component of the perpendicular return current.
This is equal and opposite to the electron loss current. In the edge of TEXTOR the
collisionality is rather high (i.e. plateau regime), such that the collisional expression for
the return current is used (see equation (7.42)).
In the stochastic region, the vacuum field is used as an approximation of the actual
8/3
field. From the vacuum field calculations it follows that: Df l /Lc ∝ |ψvac |8/3 ∝ IDED
Section 8.3 - Analysis of discharge #95592 117

[1, 35]. The electron temperature and density in equation (8.2) is measured, and for the
parallel conductivity σk the Spitzer conductivity – with neoclassical corrections – is used.
The second equation is Ohm’s law (2.10), where the terms − mee ∂j
∂t , −∇ · Πe and Re
are neglected.
Equation (8.4) is the toroidal force balance (see also (2.37)). The left hand side
describes the momentum transport. The right hand side contains the toroidal forces.
These forces come from the neutral beam, friction at the edge and of course the stochastic
force.
The last equation is the poloidal force balance as it is given in equation (2.39). The
left hand side describes the damping towards the neoclassical value, the right hand side
contains the poloidal friction and stochastic force. The neoclassical damping factor α and
the neoclassical rotation vθneo in the plateau regime are given by:
0.5 ∂Ti
vθneo = − (8.6)
eBφ ∂r
√
π qvth
α = mi ni [11] (8.7)
2 R
The above set of equations consists of three algebraic and one differential equation.
The algebraic equations (8.2), (8.3) and (8.5), are combined to find jr⊥ as a function of
Ωφ :
jr⊥ = c1(r) Ωφ + c0(r), (8.8)
where c1(r) and c0(r) only depend on known or measured quantities. Subsequently (8.8)
is substituted in equation (8.4), which then can be rewritten as:
d2 Ωφ dΩφ
+ A(r) + B(r) = 0 (8.9)
dr2 dr
This second order differential equation can be solved with a standard numerical ODE
solver. Once a solution for Ωφ is found, working backwards through equations (8.2), (8.3)
and (8.5) results for vθ , Er and jr⊥ are retrieved. The results of this calculation are shown
in figures 8.2 to 8.6.
For this calculation the measured Ti (CXRS), Te (ECE) and ne (interferometer) were
used as an input. For σk neoclassical Spitzer conductivity was used [88]. The Spitzer
resistivity, when we assume a constant Zef f , relates the temperature to the current density:
3/2
j ∝ Te . With the total plasma current measured, Te gives us the current density j, and
thus the poloidal field Bθ and the q profile.
Before the DED is switched on, the poloidal rotation should be equal to the neoclassical
poloidal rotation given by equation (8.6). In figure 8.4 we see however that there is a
rather large discrepancy between the calculated vθneo (black, dashed line) and the measured
poloidal rotation before DED (blue line). We therefore did not used the calculated vθneo .
Instead the neoclassical rotation vθneo in equation (8.5) was replaced by the measured
poloidal rotation in absence of the DED.
The charge exchange frequency in the friction term is given by: νcx = hσvicx n0 , with
the hσvicx the rate coefficient for charge exchange with cold neutrals and n0 the neutral
118 Section 8.3 - Analysis of discharge #95592

Measured toroidal rotation Calculated toroidal rotation

4 4
x 10 x 10
8 8
6 6
Ωφ (rad/s)
Ωφ (rad/s)
4 4
2 2
0 0
0 0
0.2 4 0.2 4
3 3
0.4 2 0.4 2
r (m) 1 time (s) r (m) 1 time (s)
Figure 8.2 : The measured and calculated time evolution of the toroidal rotation of shot #95592.
The solid line indicates the DED current.
density in the edge [65]. For TEXTOR hσvicx = 4.7 · 10−14 m3 /s. Typical values of the
neutral density are n0 (rwall ) = 5 · 1017 m3 at the wall, decaying to zero over a neutral
penetration length, which depends on Te and ne in the plasma edge and is approximately
8 cm. This means that the friction due to charge exchange reactions with edge neutrals
occurs over the scrape off layer and the stochastic region.
The force exerted by the neutral beam is given by PNBI /vNBI , where vNBI is the velocity
of the beam particles and PNBI is the flux surface averaged power deposition profile. The
power profile of the neutral beam is a bi-Gaussian profile with respect to the beam axis.
The FWHM was 20 cm in the equatorial plane and 10 cm in the vertical direction. From
that the flux surface averaged power deposition profile was approximated by a Gaussian
with a FWHM of 15 cm.
8/3
As mentioned before the ratio Df l /Lc is proportional to IDED . For a full DED current
of 12 kA Df l /Lc varies over the stochastic region between 10−8 and 10−7 , with an average
of 5 · 10−8 [11]. To get an agreement between the measured poloidal and toroidal rotation
and the calculated rotation we found that a slightly lower ratio Df l /Lc is needed. For
the calculation shown here, we set Df l /Lc at 12kA equal to 1 · 10−8 . A possible reason
why the ratio Df l /Lc is too high, might be the fact that the Df l /Lc is calculated from
the vacuum field and not from the actual field inside the plasma. It is possible that the
stochasticity of the plasma field is less than that of the vacuum field due to the shielding
of the resonant Fourier components.
The last input needed is the anomalous diffusion coefficient D. For this calculation D =
2
1 m /s constant over the plasma and over time was assumed. This value corresponds with
the measured energy confinement time τE = 30 ms. This anomalous diffusion coefficient
is responsible for the momentum transport over the plasma. Due to its rather high value
a force exerted near the edge of the plasma – like the stochastic torque – has an influence
over the whole plasma core, as can be seen in figure 8.3.
All input parameters of the calculation are summarised in table 8.1.

4
x 10
8
t = 1.45 s, IDED = 0 kA
t = 2.45 s, I = 10.9 kA
DED
6
Ωφ (rad/s)
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
F (N/m3)
10 Neutral beam Stochastic Friction

5
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
r (m)
Figure 8.3 : The toroidal rotation profiles before (blue) and during (red) 12/4 DED operation.
The full lines are the result of a self-consisted solution of the toroidal balance equation, the
individual points are the measurements done with CXRS. The lower plot shows the different
toroidal forces working on the plasma during DED.
8000
6000
4000
Vθ (m/s)
2000
−2000
vneo
θ
−4000 t = 1.45 s, IDED = 0 kA
−6000 t = 2.45 s, IDED = 10.9 kA
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

r (m)
Figure 8.4 : The calculated neoclassical poloidal rotation (black), the poloidal rotation profiles
before (blue) and during (red) 12/4 DED operation. For the calculation of the effect that
stochasticity has on rotation, we replaced the calculated neoclassical, poloidal rotation by
the measured poloidal rotation before DED was applied. The full lines are the result the
calculation, the individual points are measurements of the RUDI charge exchange system
[11].

The result of the calculation is an increase in toroidal rotation of ∆Ωφ = 0.6·104 rad/s
over the whole plasma, a local increase of the poloidal rotation of ∆vθ = 8 km/s in the
stochastic region, a radial current in the stochastic region of about jr⊥ = 25A/m2 and
a change in the electric field of ∆Er = 15kV /m. The local stochastic force density in
toroidal direction over the stochastic zone is :Fφ,ST = jr⊥ Bθ = 5 N/m3 . Integrating this
over the volume of the stochastic region, located at rst = 0.33 m and wst = 3 cm wide,
yields the total toroidal, stochastic force:
total
Fφ,ST = 4π 2 R0 rst wst jr⊥ Bθ = 3.5 N (8.10)
Table 8.1 : Input parameters stochastic force calculation
Ti CXRS measurement Ti (0) = 1.5 keV

Te ECE measurement Te (0) = 1.5 keV
ni = ne HCN measurement ne (0) = 5· 1019 m−3
vneo RUDI measurement vneo (rst ) = −3 · 103 m/s
hσvicx CX rate coefficient for cold neutrals hσvicx = 4.7 · 10−14 m3 /s
n0 neutral density n0 (rwall ) = 5 · 1017 m−3
FN BI Gaussian, FWHM = 15 cm FN BI (0) = 5 N/m3
Beam power/Beam velocity Ftotal = 0.7 N
8/3
Df l /Lc Vacuum field calculation Df l /Lc = 1.5 · 10−11 IDED (kA)
D anomalous diffusion D = 1 m2 /s
A straightforward expression for the stochastic torque We found that the solu-
tion of the above set of four equations, is in good agreement with the measured rotation.
The method we used is, however, quite cumbersome: we start with an unknown stochastic
torque, electric field, poloidal and toroidal velocity, and after some lengthy calculation
we find simultaneously a solution for the stochastic torque, the electric field, the poloidal
and the toroidal velocity. For practical use we would like to have an expression of the
stochastic torque that looks like:
Tφ,ST = F (IDED ). (8.11)
With a given IDED , we would then start with a known stochastic torque, apply the toroidal
force balance equation and find the toroidal rotation velocity. A much easier approach.
In order to find such a straightforward expression we will look at the local change in
rotation over the stochastic region. We approximate the stochastic region as thin, localised
layer at the position rst and with a width of wst . The change in toroidal rotation occurs
over this layer. This is sketched in figure 8.5.
During DED operation the increase in toroidal rotation over this layer, is given by
∆Ωφ . This means that Ωφ (rst,+ ) = Ωφ,0 (rst ) and Ω(rst,− ) = Ωφ,0 (rst ) + ∆Ωφ , where Ωφ,0
is the toroidal rotation in absence of a DED field. Also [dΩφ /dr]rst,+ = [dΩφ /dr]rst,− =
[dΩφ,0 /dr]rst + ∆Ωφ /wst . All other plasma parameters, including the radial return current

wst
Ωϕ
Figure 8.5 : To find a more straightforward ex-
∆Ωϕ pression for the stochastic torque, we assume
that it works only on a thin stochastic layer.
In this layer the rotation increases from Ωφ,0
to Ωφ,0 + ∆Ωφ .
rst a
jr⊥ , are assumed constant over this layer. Due to the FWHM of 15 cm FN BI (rst ) = 0. We
now integrate the toroidal force balance (8.4) over the volume of this stochastic layer:
Z rst,+ Z 2π Z 2π
R0 d d
− rD (nmΩφ ) rR0 drdθdφ
rst,− 0 0 r dr dr
Z rst,+ Z 2π Z 2π
= [jr,⊥ Bθ ] rR0 drdθdφ
rst,− 0 0
Z rst,+ Z 2π Z 2π
− [nmνcx vφ ] rR0 drdθdφ
rst,− 0 0
Z rst,+ Z 2π Z 2π
+ [FN BI ] rR0 drdθdφ
rst,− 0 0
rst,+
2 d
−4π R02 rD (nmΩφ )
dr rst,−

∆Ωφ
= 4π 2
R0 jr⊥ Bθ (rst )rst wst 2
− 4π R0 n(rst )mνcx R Ωφ,0 + rst wst
2

dn
−R0 Dm ∆Ωφ
dr r=rst

∆Ωφ
= −jr⊥ Bθ (rst )wst + n(rst )mνcx R Ωφ,0 + wst (8.12)
2
The total set of equations at r = rst now is:

jr⊥= −σk Er + + (8.13)
Lc e dr e dr

∆Ωφ 1 dpi
Er = R Ωφ,0 + Bθ − vθ Bφ + (8.14)
2 ne e dr
!
R0 mD dn nmνc xR
+ ∆Ωφ = jr⊥ Bθ − nmνcx RΩφ,0 (8.15)
wst dr rst 2
α + (1 + q 2 )nmνcx vθ = αvθneo − jr⊥ Bφ

(8.16)
This is a set of algebraic equations with as unknowns jr⊥ , Er , vθ and ∆Ωφ . Taking into

8/3
account that Df l /Lc ∝ IDED , we can rewrite the above set to:
8/3 8/3
jr⊥ = β IDED Er + γ IDED (8.17)
Er = δ ∆Ωφ + vθ + ζ (8.18)
∆Ωφ = η jr⊥ + κ (8.19)
vθ = λ jr⊥ + µ, (8.20)
where β, γ, . . . depend on the measured temperature, density, Ωφ,0 , Bθ , et cetera at r = rst .

The equations can easily be solved towards jr⊥ :
8/3
f (ne , Te , Ti , Ωφ,0 ) IDED
jr⊥ = 8/3
(8.21)
1 + g(ne , Te , Ti ) IDED
30
0.17 I8/3 [kA]
[A/m2]
DED
25 jr = −3 8/3
1 + 6.15 . 10 IDED [kA]
20
jr (A/m2)
15
10
0
0 2 4 6 8 10 12
IDED (kA)
Figure 8.6 : The return current jr⊥ is plotted as a function of the DED current. The individual
points are the result of the global approach where the differential equation for the toroidal force
balance was solved. The dashed lines come from the straightforward expression (8.21). The
result of the simple expression is in good agreement with the results of the global calculation.
The scatter in the results of the global calculation is due to the fact that for each time-
frame the measured temperature and density was used. The coefficients f and g in the
straightforward expression of jr⊥ were calculated using the average density and temperature
profiles in the DED phase.
The result of this local calculation of jr⊥ is shown in figure 8.6. The assumption of
a thin, stochastic layer is in good agreement with the global calculation. We have thus
found our straightforward expression for the stochastic torque:
8/3
f (ne , Te , Ti , Ωφ,0 ) IDED
Tφ,ST = R Bθ 8/3
(8.22)
1 + g(ne , Te , Ti ) IDED
This is again the local torque density. To get the total torque, (8.22) has to be integrated
over the volume of the stochastic region. In practice this means multiplied by 4π 2 R0 rst wst .

The temperature and density at the position of the stochastic region depend on IDED .
This has however little influence on the functions f (ne , Te , Ti , Ωφ,0 ) and g(ne , Te , Ti ). Equa-
tion (8.22) reveals that for low DED currents the stochastic torque will increase rather
8/3
quickly Tφ,ST ≈ R Bθ f IDED . For high DED currents however the stochastic torque will
saturate Tφ,ST ≈ R Bθ f /g.
8.4 Conclusion
We investigated the stochastic torque in 12/4 DED operation. In this mode of operation
the resonant EM torque at the rational q surface can be neglected, so that the only
influence on rotation during DED operation comes from the stochastic torque.
To find the values of the loss current, the poloidal rotation, the toroidal rotation and
the radial electric field a set of four equations had to be solved. The solutions of poloidal
and toroidal rotation are in good agreement with the measurements. However, a lower
ratio Df l /Lc than that retrieved form vacuum field calculations had to be used. This is
possibly due shielding of the vacuum field: the stochasticity of the field inside the plasma
might be less than that of the vacuum field.
Solving a set of four equations is not very practical. We could simplify the equations by
representing the stochastic region as a thin layer over which the change in rotation occurs.
As a result we found an expression for the loss current, and hence the stochastic torque
(8.22), that only depended on the measured density and temperatures, the rotation before
application of the DED and the DED current. This expression shows that for small DED
currents the stochastic torque rapidly increases. For higher DED currents the stochastic
torques saturates.
The simple expression for the stochastic torque can be used in the more complex
situation of 3/1 DED operation where the interplay between the stochastic torque, the
EM torque at the q = 2 surface and the viscous torque determines the mode penetration
process.

Chapter 9
Mode penetration
9.1 Introduction
In chapter 6 we saw that the threshold for mode excitation in TEXTOR was not symmetric
around ω0 = 0. Also ω did not change monotonically towards ω = 0. This is not in
agreement with the literature, where we find that the threshold is proportional with |ω0 |,
hence symmetric around ω0 = 0, and that ω goes monotonically towards zero. In chapter 7
we found that the reason for Ithreshold ∝ |ω0 | and for the monotonic change in ω is the
force balance between two torques: the electromagnetic torque at the rational q surface –
for TEXTOR this is q = 2 – and the viscous torque that is proportional to ∆ω.
We also found that the presence of a stochastic region in the plasma edge due to the
external perturbation field, results in a torque which is directed in the co-current direction.
This is due to the parallel loss of electrons. In the previous chapter we looked more closely
into this stochastic torque. We found that (a) it is quite substantial at TEXTOR and thus
can not be ignored, and (b) that the torque increases with the DED current and saturates
at high DED currents.
In this chapter we bring everything together. Instead of the balance between two
torques, we use the balance between three torques – EM torque, viscous torque and
stochastic torque – in order to predict the mode penetration threshold at TEXTOR and
the change in rotation.
9.2 Force balance model

9.2.1 Electromagnetic Torque
The two torques that are related to the DED are the resonant electromagnetic torque at
q = 2 and the stochastic torque. The latter has been discussed in the previous chapter. In
this section we translate the theoretical expression for the electromagnetic torque to the
specific situation at TEXTOR.
2,1
As seen in chapter 7, the total EM torque caused by the resonant |ψvac | flux at the
q = 2 surface is (7.37):
8 ωτrec
Tφ,EM = −4π 2 R0 0
2,1
|ψvac (rs )|2 (9.1)
µ0 |rs ∆ | 1 + ω 2 τrec
2
2,1
The resonant |ψvac | flux at q = 2 is, in the TEXTOR case, proportional to the effective
125
Chapter 9 - Mode penetration
2,1
DED current: |ψvac (rq=2 )| = 2.8·10−4 IDED . The slip frequency is ω = Ωφ −2πfe∗ −2πfDED .
A typical reconnection time for TEXTOR is τrec = 3 ms, which corresponds with a
|rs ∆0 | = 10. When we use these values we can rewrite equation (9.1) as a function of IDED
and Ωφ :
Ωφ − 2πfe∗ − 2πfDED
TEM = −1.15 · 103 2
IDED , (9.2)
104 + (Ωφ − 2πfe∗ − 2πfDED )2
with Ω in rad/s, fe∗ and fDED in Hz and IDED in kA. Figure 9.1 shows the EM torque as a
function of Ωφ for two values of IDED , and as a function of IDED for two values of Ωφ . One
clearly sees a resonance at Ωφ = 2π(fDED + fe∗ ).
8 2.5
6
2
4
2
TEM (Nm)
(Nm) 1.5
0
EM
1
T
−2
−4
0.5
−6
−8 0
−2 −1 0 1 2 0 1 2 3
Ωφ (rad/s) 4
x 10 IDED (kA)
Figure 9.1 : The left plot shows the electromagnetic torque as a function of Ωφ for IDED = 1 kA
(solid line) and IDED = 2 kA (dashed line). We clearly see the resonance at Ωφ = 2π(fe∗ +
fDED ), which corresponds with ω = 0.
The plot on the right shows the EM torque as a function of IDED for Ωφ = −1 104 rad/s
(solid line) and Ωφ = −1 103 rad/s (dashed line). It reveals the quadratic dependence of
the EM torque on the DED current.
For both plots the DED frequency is fDED = −1 kHz and the diamagnetic frequency is
fe∗ = 1.65 kHz.
When deriving equation (9.1) we assumed that all shielding current is positioned at the
rational q surface. In other words: we assume that the resistivity for the parallel shielding
currents is zero at the rational q surface. In reality resistivity for the parallel shielding
currents is lower at the rational q surface than outside the rational q surface, but not zero.
A more careful approach will result in a profile of the shielding current density over the
plasma. In [48] the shielding current density for the 2/1 component of the DED field was
calculated using a linearised four-field model. This resulted in a very sharp current density
profile centred around the q = 2 surface. The resulting electromagnetic force is given in
figure 9.21 . It is clear that assumption that all shielding current flows on the rational q
surface is justified.
In both the equation (9.1) and [48] the difference between plasma rotation and the
tearing mode rotation described by fe∗ was taken into account as a known effect. In [39]
a kinetic approach is used to calculate the EM torque. The position of the resonance
1
Here the electron diamagnetic direction was defined as ‘positive’ and the force is plotted as a function
of fDED for a fixed Ωφ . This explains why figure 9.2 is upsidedown in comparison with figure 9.1
126 Section 9.2 - Force balance model

Figure 9.2 : Electromagnetic force as a function of the DED frequency, for

Ωφ = 0 rad/s, calculated using a linearised four-field model [48]. The
result of such a more advanced model is in good agreement with the sim-
plified picture presented in section 7.2.5 of chapter 7 and figure 9.1.
at Ωφ = 2π(fDED + fe∗ ) now follows from finite Larmor radius effects. The result of this
kinetic approach is again in agreement with the simplified model that is used here.
9.2.2 Force balance at q = 2

During DED operation several torques work on the plasma:
T N BI The momentum input by the neutral beams. This torque is

constant during the discharge. It is located near the centre of
the plasma.
T CX (Ωφ ) The momentum loss due to friction at the wall of the tokamak
(see section 2.3.2 in chapter 2). This friction is caused by charge
exchange reactions with cold neutral particles. The torque is
proportional to the plasma velocity and located in the scrape
off layer.
T ST (IDED ) The stochastic torque that was discussed in section 7.3 of chap-
ter 7 and in chapter 8. This torque is located in the stochas-
tic region near the plasma edge and is always directed in the
co-direction. It depends only on IDED and is given by equa-
tion (8.22).
T EM (Ωφ , IDED ) The electromagnetic torque at q = 2. This torque is a quadratic

function of IDED and inversely proportional to the slip frequency
ω, which in turn is a function of Ωφ . The expression for the
EM torque at q = 2 in TEXTOR is given by equation (9.2).
Section 9.2 - Force balance model 127

These four torques are shown in figure 9.3, for a situation where Ωφ (q = 2) 2π(fDED +
fe∗ ).
Φ q=2 stochastic zone wall
TNBI
co-direction TST
r
cnt-direction TEM
TCX
R0 d  
2
d ρΩ φ
 rD 
r d r 
−
dr 
Figure 9.3 : Four torques work on the plasma during DED. Two are non-DED-related: the neu-
tral beam torque and the friction. They will lead that the viscous torque that is proportional
to ∆Ωφ . The two other torques are DED-related: the resonant EM torque at the q = 2 sur-
face and the stochastic torque near the plasma edge. They enter separately into the torque
balance. The force balance then is TEM + TST + TV S = 0.
Looking at different plasma rotation frequencies Ωφ,0 at q = 2 before the DED is

applied, we can now understand the rotational behaviour of the plasma during DED.
Ωφ,0 ≈ 2π (fe∗ + fDED ) In this case, when the DED current is ramped up, a very strong
electromagnetic torque rises at the q=2 surface, due to the reso-
nance at Ωφ = 2π (fe∗ + fDED ). The transition to the locked island
phase is therefore almost instantaneous. It explains why the minima
in the thresholds in figure 6.14 lie at Ωφ,0 = 2π (fe∗ + fDED ).
Ωφ,0 2π (fe∗ + fDED ) If the plasma is strongly rotating in the co-direction, then we are
far from the resonance of the electromagnetic torque when switching
on the DED. The electromagnetic torque TEM is in the counter-
direction and is rather small. At that point the stochastic torque
TST , that is always in co-direction and does not depend on Ωφ , is
larger than TEM . Hence, at low IDED there is a net torque in the
co-direction and an increase in co-rotation. As the DED current
increases however, TST reaches saturation, while TEM increasingly
grows. A turning point will be reached where the net torque changes
from co to counter. Once this occurs the plasma rotation will brake
and finally the transition to the locked phase is made.
For Ωφ,0 2π (fe∗ + fDED ) the stochastic torque TST is opposite to
the electromagnetic torque TEM . The threshold for mode penetra-
tion is therefore higher than one would expect in the case where no

TST is present, which explains the steep slopes in figure 6.14 right
of the minima.
Ωφ,0 2π (fe∗ + fDED ) A third possibility is a strongly counter-rotating plasma. Just as in

the co-rotating situation the electromagnetic torque TEM is rather
weak at first. In contrast to the co-rotating case TEM is now directed
in the co-direction; just like TST is. The plasma rotation will brake,
i.e. change in the co-direction, until mode penetration is reached.
The stochastic torque TST is now pointed the same direction as the
electromagnetic torque TEM . The threshold for mode penetration
is therefore lower than one would expect in the case where no TST
is present. This explains the less steep slopes in figure 6.14 left of
the minima.
The actual force balance equation One could try to make the qualitative description
above more quantitatively. We therefore project the torques outside the q = 2 surface onto
the q = 2 surface and get as a result the torque balance at q = 2:
TEM + TST + TN BI + TCX = 0. (9.3)
Before the DED is applied this is:
TN BI + TCX = 0, (9.4)
where TCX is proportional to the plasma rotation in absence of the DED: TCX = κΩφ,0 .
During DED operation TCX = κΩφ , which allows us to rewrite (9.3) into:
TEM + TST + κ (Ωφ − Ωφ,0 ) = 0 (9.5)
The neutral beam torque and the friction give rise to a term in the torque balance that
is proportional to ∆Ω = (Ωφ − Ωφ,0 ). This means the sum of neutral beam torque and
friction gives us the viscous torque TV S that is proportional to the change in velocity. The
parameter κ can be calculated from the neutral beam torque and the friction before the
DED is switched on. This gives κ = −4.6 · 10−5 N m/(rads−1 ).
We finally complete the torque balance by filling in the expressions for the EM torque
(9.2) and the stochastic torque (8.22):
8/3
Ωφ − 2πfe∗ − 2πfDED αIDED
− 1.15 · 103 2
IDED + − 4.6 · 10−5 (Ωφ − Ωφ,0 ) = 0
104 + (Ωφ − 2πfe∗ − 2πfDED )2 8/3
1 + βIDED
(9.6)
As seen in chapter 8 the factors α and β depend properties that can be measured, like
the density and the temperature in the stochastic region. For the discharges used in
the experiments that are presented further on, we found α = 1.84 N m/kA8/3 and β =
4 kA−8/3 .

4
x 10
1.5
(2)
A B
1
(1) C D
0.5
Ωφ = 2 π (f*e + fDED)
Ωφ (rad/s)
0
B
−0.5
A (3)
−1
−1.5
0 0.5 1 1.5 2
IDED (kA)
Figure 9.4 : Solutions of equation (9.6) are given for different initial rotation frequencies Ωφ,0 .
For Ωφ,0 close to 2π (fe∗ + fDED ) – line (1) – the transition from the suppressed (|ω| 0) to
excited island state (ω = 0) is continuous.
If Ωφ,0 2π (fe∗ + fDED ) – line (2) and (3) – a bifurcation of the rotational states is
present; when IDED is increased the rotation velocity follows the suppressed branch from A to
B. If IDED is increased over the certain threshold – indicated by B – the rotation jumps from
the suppressed to the excited branch and goes further towards D. The dashed line indicates
Ωφ = 2π (fe∗ + fDED )
Solving the force balance equation The force balance equation (9.6) can be solved,
yielding Ωφ as a function of IDED . The solution for three different initial rotation frequen-
cies Ωφ,0 is given in figure 9.4.
Line (1) starts from Ωφ,0 ≈ 2π (fe∗ + fDED ). The transition from |ω| > 0 – the sup-
pressed island state – to ω = 0 – the excited island state – is continuous. We can not
really define a threshold for mode penetration.
For Ωφ,0 2π (fe∗ + fDED ) and Ωφ,0 2π (fe∗ + fDED ) – lines (2) and (3) in figure 9.4
– equation (9.6) yields three branches of solutions: A-B, C-D and C-B.
When IDED is increased from zero, the plasma rotation Ωφ will follow the branch that
starts in A and goes to B. This is the suppressed branch: on this branch |ω| 0 and no
islands are visible in the plasma: they are suppressed.
When the plasma rotation Ωφ reaches the point B, it can no longer follow the sup-
pressed branch when IDED is further increased. The plasma rotation jumps to the branch
C-D. This is the excited branch: Ωφ = 2π (fe∗ + fDED ), this is ω = 0, and large 2/1 islands
are excited. In the specific case of TEXTOR, the large 2/1 islands will immediately couple
with a 1/1 internal kink mode, resulting in a flat rotation profile.
The third branch of solutions – C-B – is just an artifact of mathematics. Physically
it can never be accessed.
One sees in figure 9.4 that when Ωφ,0 2π (fe∗ + fDED ), the plasma rotation initially
increases in the co-direction, i.e. moves away from Ωφ,0 = 2π (fe∗ + fDED ), and only later
slows down and locks. This initial increase in co-rotation is due to the stochastic torque

that is in the co-direction. If Ωφ,0 2π (fe∗ + fDED ) the rotation changes monotonically
towards from Ωφ,0 = 2π (fe∗ + fDED ), because the stochastic torque now works with the
electromagnetic torque, instead of against it. As a result the thresholds for mode pene-
tration are lower for Ωφ,0 2π (fe∗ + fDED ) than for Ωφ,0 2π (fe∗ + fDED ).
0.6
T
EM
T
0.4 ST
T
VS
0.2
T (Nm)
−0.2
−0.4
0 0.5 1 1.5
IDED (kA)
Figure 9.5 : The EM torque (TEM ), the stochastic torque (TST ) and the
viscous torque (TV S ) are plotted as a function of the effective DED cur-
rent, for the solution on the suppressed branch. The initial rotation is
Ωφ,0 = 104 rad/s 2π (fe∗ + fDED ). The stochastic torque is positive,
i.e. in the co-direction, and is balanced by the EM torque and the viscous
torque that both work in the counter direction.
In figure 9.5 the different terms of equation (9.6) are plotted against
the effective DED current for the solutions on the suppressed branch with
Ωφ,0 = 104 rad/s 2π (fe∗ + fDED ). Both the electromagnetic torque TEM and
the viscous torque TV S are in the counter-direction. Their sum balances the stochastic
torque TST that is in the co-direction. One sees that TST saturates, while TEM increas-
ingly grows. Because the viscous torque is proportional to the change in rotation, the
point where |TV S | reaches a maximum is the turning point where the plasma rotation
starts braking.
Apart from the plasma rotation Ωφ as a function of the effective DED current, solving
the force balance equation (9.6) also gives us the threshold for mode excitation as a function
of Ωφ,0 : the points B in figure 9.4.
In figure 9.6 the measured thresholds for mode penetration that where given in fig-
ure 6.14 are compared with the threshold following from the force balance model. We find
that there is quite a good agreement between the model and the measured data.
The threshold is of course only one point in the Ωφ (IDED ) curve that is predicted by
the model. A stronger argument for the validity of the model is the agreement between
the calculated and the measured Ωφ (IDED ), before mode penetration. This is shown in
figure 9.7 for (some of) the DC rotation scan discharges of figure 9.6. Figure 9.8 shows
the result for the discharges of the rotation scan with AC+ DED operation. The solution
Ωφ (IDED ) of the force balance equation (9.6) on the suppressed branch is plotted as solid

1.5
Effective IDED (kA)
0.5
0
−1 −0.5 0 0.5 1 1.5 2
Ωφ,0(q=2) (rad/s) x 10
4
Figure 9.6 : Threshold for mode excitation plotted against the rotation at
q = 2 before the DED is applied. The density (1.5 1019 m−3 ) and the
power input (∼ β) were kept constant. The data for DC DED operation
is given in blue, the data for AC+ DED (−1 kHz) is given in red. The
dotted lines are the result of the force balance model.
lines with the same colour as the corresponding experimental data. We find that also
here the model is in good agreement with the measurement.
The model predicts that for Ωφ,0 = 2π (fe∗ + fDED ) the threshold for mode excitation
is zero. The measurements of the threshold do show a minimal threshold for Ωφ,0 =
2π (fe∗ + fDED ), but this threshold is clearly larger than zero.
The reason why the our model gives a zero threshold when ω0 = 0 is the definition
of the threshold itself: the threshold is defined as the moment when ω jumps to zero. If
ω0 = 0 this is immediate. However, ω = 0 does not necessarily mean that we can observe
islands in the plasma. In section 7.2.4 we found that the island width for locked modes is
given by: r
m √
W ≈ W0 + f (Wvac ) cos ∆ϑ Wvac , (9.7)
2 |rs ∆0free (0)|
where f (Wvac ) is approximately one and only weakly depends on Wvac . If no islands are
present before the DED is applied, then W0 = 0, so that the island width W becomes:
r
m √
W ≈ cos ∆ϑ Wvac . (9.8)
2 |rs ∆0free |
This means that for low perturbation levels√(Wvac is small), a large phase difference
between the vacuum and the plasma island ( cos ∆ϑ is small) and a plasma that is
naturally stable against tearing modes (rs ∆0free 0), the island width W will be small.
W could be too small, so that the magnetic islands cannot be observed. The parallel
transport from one side of the island to the other along the separatrix is then of the same
order as the cross-field transport over the island.

4
x 10
2
1.5
1
Ω (rad/s)
0.5
0
φ
−0.5
−1
−1.5
0 0.5 1 1.5 2
IDED (kA)
Figure 9.7 : The measured plasma rotation (data points) and the rotation following from the
force balance model (solid lines) plotted against the effective DED current, for 6 discharges
out of the rotation scan with DC DED operation given in figure 9.6. The dotted line indicates
Ωφ = 2πfe∗ .
4
x 10
2
1.5
1
Ωφ (rad/s)
0.5
−0.5
−1
−1.5
0 0.5 1 1.5 2
IDED (kA)
Figure 9.8 : The measured plasma rotation (data points) and the rotation following from the
force balance model (solid lines) plotted against the effective DED current, for the discharges
out of the rotation scan with AC+ DED operation given in figure 9.6. The dotted line
indicates Ωφ = 2π(fe∗ − 1 kHz).

The measured threshold is determined in two ways:

• by looking at a change in plasma rotation, which is clearly seen when the 1/1 internal
kink mode couples to the 2/1 island.
• by looking at drop in the density signal, which indicates a decrease in plasma con-
finement.
Both the coupling of the 1/1 and 2/1 mode and the decrease in confinement require a
rather large 2/1 island. The deviation from the model around ω0 = 0, can hence possibly
be explained by the fact that the mode is locked, but too small to be observed.
An analytical solution of the force balance equation The results shown above were
obtained by numerically solving the force balance equation (9.6). An analytical solution,
however, usually gives more physical insight. Let us take another look at (9.6):
8/3
ω 1.84 IDED
− 1.15 · 103 4 2
2
IDED + 8/3
− 4.6 · 10−5 (ω − ω0 ) = 0. (9.9)
10 + ω 1 + 4 IDED
We now specifically look at the situation: |ω| 100. This means that the electro-
magnetic term can be approximated by 1.15 · 103 IDED
2 /ω. If |ω| 100, it is of course also
a lot greater than zero. Consequently we expect the threshold in the DED current to be
high. We saw in chapter 8 that the stochastic force saturates for high DED currents. The
stochastic term in (9.9) can thus be approximated by the constant 1.84/4 = 0.46. We can
now rewrite (9.9) as:
2
IDED = A (ω0 − ω + cST ) ω, (9.10)
where for TEXTOR A = 4 · 10−8 and cST = 104 .
The right hand side of equation (9.10) has a maximum at ω = (ω0 + cST )/2. The value
of IDED for ω = (ω0 + cST )/2 is the mode penetration threshold:
√
A
Ithreshold = |ω0 + cST |. (9.11)
2
If cST = 0, i.e. when there is no stochastic region, this is the scaling law for mode
penetration that is commonly found in the literature. The introduction of a stochastic
region causes a shift in this scaling law, resulting in a virtual minimum threshold for
ω0 = −cST . This is a virtual minimum, because the real minimum threshold still lies at
ω0 = 0; equation (9.11) is only valid for ω0 0! The transition from equation (9.11)
for |ω0 | 0 to the minimum at ω0 = 0, leads to the observed steeper gradient in the
threshold plot for ω0 > 0 in the vicinity of ω0 = 0 than for ω0 < 0.
9.3 Conclusion
When the DED is operated in 3/1 mode mode excitation will occur. This is due to
the resonant electromagnetic torque at q = 2 that will try to bring the slip frequency
ω = Ωφ − 2πfe∗ − 2πfDED to zero. As long as |ω| 0 islands in the plasma are suppressed,
once ω = 0 large 2/1 islands are excited at the q = 2 surface.
The EM torque is not the only torque in the plasma. A viscous torque, proportional
to the change in rotation, and a stochastic torque, due to parallel electron loss in the

perturbed plasma edge, will balance the EM torque. The stochastic torque is directed in
the co-current direction. It is stronger than the EM torque for low DED currents, but
saturates at high DED currents.
There are now two possible situations. If the plasma rotation is such that ω0 > 0
before the DED is applied, the rotation will first increase in the co-direction due to the
high stochastic torque. At higher DED currents the stochastic torque will saturate and the
EM torque takes over: the plasma rotation brakes and finally mode penetration occurs.
Because the stochastic torque opposes the EM torque the mode penetration threshold will
be high.
In a second situation ω0 < 0. Now both the stochastic and the EM torque will try to
bring ω to zero, resulting in a low mode penetration threshold.
For |ω0 | 0 the stochastic force will be completely saturated when the mode excitation
threshold is reached. This results in a threshold that is proportional to Ithreshold ∝ |ω0 +
cST |, where cST depends on the amount of stochasticity that the perturbation field induces.
This means that the scaling law found in the literature – Ithreshold ∝ |ω0 | – shifts by −cST
when a stochastic region is present. This proportionality to |ω0 +cST | is however only valid
for |ω0 | 0, the minimal threshold is still at ω0 = 0. Because Ithreshold ∝ |ω0 + cST | for
|ω0 | 0 and the minimal threshold lies at ω0 = 0, there is an asymmetry in Ithreshold (ω0 )
around zero.

Chapter 10
Changing the plasma rotation by

EM waves
10.1 Introduction
The previous chapters showed that fast rotating plasmas are less sensitive to error field
mode penetration. Apart from increasing the stability with respect to tearing modes,
fast rotation also increases the confinement, through the suppression of turbulence. In
the current generation of tokamaks fast plasma rotation is achieved by the momentum
input of tangential neutral beams. In the next generation of fusion reactors neutral beam
injection will not be able to induce fast plasma rotation: the plasma mass, that scales
with R0 a2 ne , will be a lot higher and the neutral beam power has its limits. The search
for other methods of momentum input therefore is a hot topic in the fusion community.
In many tokamaks electromagnetic (EM) or radio frequent (RF) waves are used to
heat the plasma. The momentum of these waves is very small and therefore they are not
expected to act as a momentum source. It is however observed in many machines that the
injection of RF power does influence the plasma rotation.
In TEXTOR two types of EM waves can be injected into the plasma: electron cyclotron
waves and ion cyclotron waves. The influence that these waves have on the plasma rotation
will be discussed in this chapter.
10.2 Changing the momentum by ion cyclotron radiation

As said above the direct momentum input by RF waves can be neglected. One can say
that the RF waves adds up to the total thermal energy source, but not to the momentum
source. As mentioned in chapter 2 the ion energy confinement time τEion is strongly coupled
with the toroidal momentum confinement time τφ . It is also known that increasing the
power input leads to a confinement degradation; thus to a lower τEion and τφ .
Consider a plasma where the toroidal rotation is driven by tangential neutral beam
injection. When ion cyclotron resonant heating (ICRH) is applied in such a plasma, the
power input is increased – leading to a lower τφ – while the momentum input remains
unchanged. The lower τφ causes the momentum input of the NBI to be lost more quickly
during ICRH than without ICRH, resulting in a smaller toroidal rotation velocity. This
has been seen on several machines [18, 19, 59, 66]. Also on TEXTOR a reduction of
137
Chapter 10 - Changing the plasma rotation by EM waves
toroidal rotation is seen when ICRH is applied in co-rotating plasmas (see figure 10.1 (a)).
However, if confinement degradation would be the only mechanism at work, we would
expect the reduction of rotation to be equally large in co- and counter-rotating plasmas.
Figure 10.1 shows that this is not the case. The co-rotating plasma (fig. 10.1 (a)) has a
very strong reduction – almost to zero – of the toroidal rotation, whereas the counter-
rotating plasma has no reduction, but even a small increase in counter-rotation (fig. 10.1
(b)). In [66] similar asymmetries between co- and counter-rotating plasmas have been
observed. A possible explanation is that the assumption that ICRH waves do not change
the total momentum input is wrong.
Another argument that suggests momentum input by ICRH is the observation on
several tokamaks that the toroidal rotation increases significantly over the level of Ohmic
rotation (section 2.4.3 of chapter 2) when ICRH is applied in a plasma without external
momentum input [21, 23, 25, 70].
4
x 10
4
(a)
3
Ω (rad/s)
1
φ
0
1MW ICRH
−1
4
x 10
1
(b)
0
Ω (rad/s)
−1
−2
φ
−3
1MW ICRH
−4
1.5 2 2.5 3 3.5
time (s)
Figure 10.1 : Time traces of the rotation in TEXTOR at R = 1.85 m

during the application of ICRH. For a beam-driven, co-rotating plasma
(a) the rotation drops almost to zero. For a counter-rotating plasma
(b) the same amount of ICRH power leads to a slightly more negative
rotation (i.e. faster counter-rotation). A reduction of the momentum
confinement time τφ , due to the increased power input, can not explain
the asymmetry. This means that ICRH is not only a source of energy,
but also a source of momentum.
The rotation due to ICRH in plasmas without external momentum input has been
observed to be both in co- as in counter direction: on one hand high density plasmas have
a tendency to rotate in the co-current direction when ICRH power is applied, while in low
density discharges counter-rotation is more likely to occur. This suggests that there might
138 Section 10.2 - Changing the momentum by ion cyclotron radiation

be several competing processes at work [22].

In the literature several mechanisms have been proposed to explain the ICRH generated
rotation:
• A first model takes into account ripple-induced losses of the fast ions created by
ICRH. This fast ion loss causes an inwards pointing radial electric field and corre-
sponding counter-rotation [21].
• A second model looks at trapped ions. Each time such a trapped ion travels through
the resonance layer of the ICRH deposition it increases its perpendicular velocity,
and hence also its Larmor radius ρ = v⊥ /ωc increases. Because the width of the
banana orbit is proportional to the Larmor radius (see section 2.4.1 in chapter 2),
the banana width increases as well. As a result there is a net outward drift of the
averaged centre of the trapped ion’s orbit. This outward displacement current is
balanced by a inward return current jr , leading to a jr × B force in the counter
direction [12].
• For near-axis absorbed ICRH, a portion of the trapped fast ions will be scattered into
co-passing orbits, leading to co-rotation, while another portion will be scattered into
counter-passing orbits, with a resulting counter-rotation. This at first sight means no
net rotation remains. However, it turns out that the counter-passing orbits lie more
towards the edge of the plasma than the co-passing orbits. The counter rotation is
therefore more damped than the co-rotation, leading to a net co-rotation [68, 22].
• In the case of off-axis ICRH absorption, there is a radial diffusion of fast ions from
their origin point at the resonant layer. The radial current due to this diffusion is – for
ambipolarity reasons – compensated by a return current. In between the resonance
layer and the plasma centre this return current is pointed outwards, leading to a
co-current jr × B force. Between the resonance layer and the plasma edge jr causes
a counter-current torque. As it is located more towards the edge the counter-force
is damped more strongly than the co-torque resulting in a net co-rotation [68, 22].
• As has been said before the toroidal momentum confinement is strongly coupled
to the ion energy confinement. Moreover there seems to be a profile consistency
between toroidal rotation and ion temperature: Ωφ (r)/Ti (r) ≈ constant [86, 18]. In
the case of ICRH induced co-rotation the toroidal velocity also scales with the ion
pressure [21, 23, 25, 70].
This suggests that the plasma has natural tendency to rotate in the co-direction
with a rotation profile proportional to the ion temperature profile. When the ion
confinement is improved by ICRH, the toroidal rotation consequently increases as
well. The underlaying physics linking Ti and Ωφ is related to anomalous transport
[78].
In order to explain the data presented in figure 10.1 it seems necessary to include
both confinement degradation and the input of counter-momentum. In a counter-rotating
plasma (fig. 10.1 (b)) the loss of momentum confinement τφ would then be compensated,
and even a bit overcompensated, by an increased momentum input in the counter-direction
Ttotal = TN BI + TICRH . In the co-rotating case (fig. 10.1 (a)) there is a loss of momentum
confinement τφ and the total momentum input in the co-direction is decreased Ttotal =
TN BI − TICRH , leading to a strong reduction of the co-rotation.
Section 10.3 - Changing the momentum by ion cyclotron radiation 139

10.3 Influence of ECRH on rotation

At TEXTOR a 1 M W gyrotron is installed that injects EM wave with a frequency that
is a higher harmonic of the electron cyclotron frequency. These EC waves will increase
the thermal energy of the electrons, and the method is therefore called electron cyclotron
resonant heating (ECRH). Just like in the ICRH case the EC waves have little momentum.
Nevertheless it was observed that ICRH, despite its negligible direct momentum
input, significantly changes the plasma rotation. This means that in principle ECRH
could have an effect on plasma rotation as well.
The main mechanisms for changing the plasma rotation in the ICRH case were:
• Changing the plasma confinement. This can also be done by ECRH.
• Radial displacement currents and fast ion losses. In the case of ECRH, radial dis-
placement currents and fast electron transport or losses are also possible.
• Pitch-angle scattering of trapped fast ions into co-passing orbits. Although trapped
fast electrons can also be scattered into co-passing orbits, this would – due to their
small mass – not have any significant influence on the plasma rotation.
The above indicates that for the mechanisms that are responsible for a change in plasma
rotation during ICRH, similar mechanisms exist during the application of ECRH. We
therefore expect to observe an effect of ECRH on the plasma rotation.
In figures 10.2 and 10.3 the toroidal plasma rotation before and during on- and off-axis
ECRH is plotted, for beam driven plasmas. The application of ECRH has a clear effect
on the rotation in both cases.
4
x 10 co−rotation, no ECRH
co−rotation, 700 kW on−axis ECRH
2 cnt−rotation, no ECRH
cnt−rotation, 700 kW on−axis ECRH
1.5
0.5
Ωφ (rad/s)
−0.5
−1
−1.5
−2
1.75 1.8 1.85 1.9 1.95 2 2.05

R (m)
Figure 10.2 : Toroidal rotation profiles before and during the application of on-axis ECRH. Both
co- and counter rotation reduces when 700 kW of ECRH power is deposited in the vicinity
of the magnetic axis.
140 Section 10.3 - Influence of ECRH on rotation

For on-axis ECRH the rotation reduces, both for co- and counter-rotating plasmas.
Assuming on-axis ECRH is responsible for larger energy transport, this would mean –
due to the coupling between energy confinement and momentum confinement – a larger
momentum transport. The momentum input by the neutral beam in the plasma centre
would be transported faster to the wall in the case of ECRH than it would in a plasma
without ECRH. As a result the total angular momentum of the plasma will be lower, as is
seen in figure 10.2. Confinement degradation due to ECRH injection is therefore a good
candidate to explain the observed change in rotation.
4
x 10
4
q=1 no ECRH
700 kW off−axis ECRH
3.5
2.5
Ωφ (rad/s)
2
3.1 time (s) 3.2
1.5
1
2.7 time (s) 2.8
0.5
0
1.75 1.8 1.85 1.9 1.95 2 2.05
R (m)
Figure 10.3 : The toroidal rotation profiles during (2.75 s) and after (3.15 s) the application of
off-axis ECRH (Rdep = 1.9 m). The insets are ECE measurements at R = 1.79 m during
ECRH (2.7 − 2.8 s) and after ECRH (3.1 − 3.2 s). During the ECRH the rotation profile
is peaked, with a fast central rotation and lower rotation near the edge. At the same time
the ECE measurement shows no sawteeth over a period of at least 100 ms. After ECRH is
switched of the rotation profile gets broader. In the ECE time trace sawteeth with a period
of about 15 ms.
The data plotted in figure 10.3 was obtained during an experiment where the sawteeth
were suppressed using off-axis ECRH. The sawteeth mentioned are periodic oscillations
of the central pressure: the pressure rises slowly until it reaches a certain threshold, after
which it rapidly drops, causing a flat pressure profile within the q = 1 surface [61, 88]. The
presence of sawteeth limits the central pressure. When applying ECRH power outside the
q = 1 surface – in this particular case at R = 1.86 m – the sawtooth period is increased,
allowing the central pressure to reach higher values [61]. This means that the central
confinement is increased. An improvement of the local energy confinement would also
mean a higher momentum confinement time τφ inside the q = 1 surface. Consequently the
rotation driven by the neutral beam is expected to be higher inside q = 1 – as can be seen
in figure 10.3. Outside q = 1 the ECRH is expected to increase transport, resulting in a
reduced momentum confinement τφ . This leads to a lower driven rotation outside q = 1
during ECRH, which is also observed in figure 10.3.
The data in figures 10.2 and 10.3 indicates that ECRH indeed has an influence on
Section 10.3 - Influence of ECRH on rotation 141

rotation. Not only at TEXTOR, but also on other tokamaks an effect on rotation by
ECRH has been observed [18]. These first experiments indicate that the change in rotation
is probably mainly caused by a change in the (local) energy confinement.
10.4 Conclusion
First experiments have been done at TEXTOR, investigating the influence of ICRH and
ECRH on plasma rotation. The experiments with ICRH show that confinement degra-
dation due to the higher momentum input can not explain the observed rotation. Apart
from reducing of the momentum confinement, ICRH also acts as a momentum source in
the counter-current direction.
The experiments with ECRH show that possibly the change in energy transport caused
by ECRH is also reflected in the momentum transport. Not only globally, but – as the
sawtooth suppression experiment shows – also locally. This opens possibilities for shaping
the rotation profile. Changing the local momentum transport, the velocity shear can be
locally increased, resulting in a local reduction of the turbulence.
The experiments that were carried out give a first qualitative look on the influence
that ICRH and ECRH have on the rotation in TEXTOR. These results are promising and
open the possibility for further and more dedicated experiments on this topic.

Chapter 11
Discussion, conclusions and

outlook towards future devices
11.1 Exciting tearing modes with the DED in TEXTOR

How to avoid the excitation of error field modes in tokamak plasmas, was the main question
of this thesis. In order to study this problem, we reversed the question: how do we excite
error field modes? Because if you know how error field modes are excited, you also know
what measures have to be taken to avoid the excitation. The most important parameter
in mode excitation is the plasma rotation. The aim of this thesis was to perform dedicated
experiments on the interplay between mode excitation and plasma rotation.
The experiments were conducted at the TEXTOR tokamak. TEXTOR is equipped
with two tangential neutral beam injectors that allow us to set up a wide variety of
plasma rotation profiles. Furthermore a set of perturbation coils is installed at TEXTOR:
the Dynamic Ergodic Divertor or DED. The DED can be used to set up a perturba-
tion field with a predefined topology, strength and rotation. The tangential neutral
beams and the DED allowed us to change both the plasma rotation and the pertur-
bation field, and hence create the desired conditions for the mode penetration experiments.
From these experiments we found that in 3/1 operation of the DED, large magnetic
islands where excited at the q = 2 surface. The threshold for this mode excitation is
a function of the plasma rotation Ωφ,0 before the perturbation field was applied. Both
for high co- and high counter-rotation the threshold is high. The minimal threshold is
however not located at zero rotation and depends on the rotation frequency of the DED
field.
So far these observations are in agreement with the theory on mode penetration as
it is described in [28, 29] and [31]. That theory states that the plasma response on ex-
ternal perturbation field depends strongly on the slip frequency ω. This slip frequency is
the difference between the rotation frequency of the perturbation field and the rotation
frequency of MHD modes in the plasma. Because MHD modes are frozen in the electron
fluid and the plasma fluid rotation is carried mainly by the ions, the MHD rotation fre-
quency has an offset with respect to the plasma rotation frequency which is given by fe∗ .
In TEXTOR, where the plasma current and toroidal magnetic field are anti-parallel and
where perturbation field is generated by the DED, the expression for the slip frequency of
143
Chapter 11 - Discussion, conclusions and outlook towards future devices
a 2/1 mode is given by:

ω = Ωφ − 2πfe∗ − 2πfDED . (11.1)
The mode penetration theory predicts suppressed islands as long as |ω| 0 and the
excitation of large islands when ω = 0. If ω0 is the slip frequency before the penetration
field is applied, then an immediate excitation of the mode is expected once the perturbation
field is switched on, if ω0 = 0. In other words: the minimum in excitation threshold is
found where ω0 = 0. Equation (11.1) explains why the observed minimal threshold is not
located at Ωφ,0 = 0 and why it depends on the frequency of the DED.
When |ω0 | 0 the plasma rotation first has to change so that ω becomes zero, before
the mode excitation occurs. The torque that reduces ω is a j × B torque, coming from the
cross product of the induced parallel shielding currents and the magnetic perturbation at
the resonant q surface. For a large |ω0 |, a large change in rotation, thus a strong j × B
force, is required before ω reaches zero. As a result fast rotating plasmas have a high
threshold for the mode excitation.
The result of the mode penetration theory of [28, 29] and [31] is a mode excitation
threshold that is proportional to |ω0 |. It also predicts that plasmas with ω0 > 0 will slow
down, while plasmas with ω0 < 0 will spin up when the external perturbation is applied.
This is summarised in the blue lines of figure 11.1.
The measurements of plasma rotation and excitation threshold do not fully agree with
the predictions of the mode penetration theory. It is for example observed that plasmas
with ω0 > 0 initial spin up when the perturbation field is ramped up, and only at a later
stage start to slow down towards ω = 0. Also the threshold versus rotation plot is not
symmetric around ω0 = 0, as would be expected from the model given in [28, 29] and [31].
It was therefore necessary to investigate whether the perturbation field exerts another
torque onto the plasma, apart from the j × B torque caused by the shielding currents
at the rational q surfaces. Near the edge of the plasma the external perturbation field
is not completely shielded: the Fourier components of the external field are peeled off at
the corresponding resonant q surfaces, not at the plasma edge. Close to the plasma edge
the magnetic field therefore becomes stochastic when the perturbation field is increased.
In [45] it is described that in this stochastic region, a parallel electron loss current exists,
which is balanced by a perpendicular return current. The j × B torque due to this
perpendicular return current is toroidally always directed in the co-current direction.
For 12/4 DED operation the vacuum field decays fast with increasing distance from
the coils, due to the high m-numbers. The shielding currents at the rational q surfaces
will be low and the corresponding j × B torques at the rational q surfaces can be
neglected. Any changes in rotation during 12/4 operation of the DED can therefore be
attributed to the stochastic torque. For a correct treatment of the stochastic torque
the expression for the return current, the toroidal and poloidal force balance and the
generalised Ohm’s law have to be solved in a self-consistent way. As a result we found
that a stochastic torque exists that increases rapidly for low perturbation strengths and
saturates for higher perturbation strengths.
When we combined the stochastic torque in the edge, the torque at the rational q
surface, the momentum input by the neutral beams and the friction with neutrals in the
plasma edge, we came to a force balance equation that predicts the change in plasma veloc-
144 Section 11.1 - Exciting tearing modes with the DED in TEXTOR
4
x 10
2
1.8 1
1.6
ω = Ω (q=2) − 2 π f* (rad/s)
1.4 0.5
φ
1.2
1 0
0.8
e
0.6 −0.5
0.4
0.2 −1
0
−2 −1 0 1 0 0.5 1 1.5 2
ω0 = Ωφ,0(q=2) − 2 π f*e (rad/s) IDED (kA)
4
x 10
Figure 11.1 : At the left the threshold for mode excitation by a static DED field in TEXTOR
is plotted. At the right the rotation during a ramp up of the DED current is plotted for two
initial rotations ω0 . The black and gray data points are measurements.
The blue lines are the predictions of the mode penetration model that is described in [28, 29]
and [31]. This model predicts the threshold to be symmetric around the minimum at ω0 = 0.
It also predicts that the rotation will change monotonically towards ω = 0. It is seen that
these predictions are not in agreement with the measurements.
The red lines are the predictions of the mode penetration model where we included a stochastic
torque at the edge of the plasma. This torque is always in the positive (co-current) direction.
For large |ω0 | the threshold curve is shifted, with a virtual minimum at ω0 = −6.103 rad/s.
For lower |ω0 | the curve bends, because the real minimum still has to be at ω0 = 0. As a
result the threshold curve is less steep for ω0 < 0 than it is for ω0 > 0. Due to the stochastic
torque a plasma with an initial velocity of ω0 > 0 first spins up and only later decreases
towards ω = 0. A plasma with ω0 < 0 will goes faster to ω = 0 than the model without a
stochastic torque predicts. The model where a stochastic torque is introduced, is clearly in
better agreement with the measurements than the model without stochastic torque.
ity before mode excitation and the value of the mode excitation threshold. The predictions
we derived where in good agreement with the measurements of both the thresholds and
the change in rotation when the DED strength is increased.
The stochastic torque always tries to spin up the plasma rotation. For ω0 < 0 the
stochastic torque helps the j × B torque at the resonant q surface to bring ω towards zero,
for ω0 > 0 the stochastic torque counteracts the resonant torque. At large perturbation
strengths stochastic force saturates. This means that for |ω0 | 0, where a large threshold
is expected, the stochastic force can be treated as a constant. As a result we derived
following scaling law:
Ithreshold ∝ |ω0 + cST |, for |ω0 | 0 (11.2)
This means that the inclusion of an edge effect due to the perturbation field leads to an
offset in the scaling law. Due to the fact that the stochastic torque is always directed in
the co-direction the offset cST is positive. The exact value of cST depends on the transport
Section 11.1 - Exciting tearing modes with the DED in TEXTOR 145
properties in the stochastic region that is created by the perturbation field.

For lower perturbation strengths the stochastic force rapidly decreases. This means
that for |ω0 | closer to zero, where the threshold is lower, the effect of the stochastic torque
is lower. The minimal threshold therefore still lies at ω0 = 0 and not at ω0 = −cST , like
equation (11.2) would suggest. The transition from equation (11.2) for |ω0 | 0 to the
minimum at ω0 = 0, leads to a steeper gradient for ω0 > 0 in the vicinity of ω0 = 0 than
for ω0 < 0.
The effect that the inclusion of a stochastic torque in the mode penetration model has
on the rotation and the threshold is shown by the red lines in figure 11.1. This extended
model agrees much better with the measurements. The fact that the minimal, measured
threshold at ω0 = 0 is not zero, as is predicted by the model, could however not be resolved.
The measurements of the mode penetration threshold and the change in rotation at
TEXTOR could be explained by introducing a stochastic torque into the existing mode
penetration models. This stochastic torque is a result of the creation of a stochastic
zone near the plasma edge. In this stochastic region a radial electric field builds up due
to parallel electron loss. This electric field results in an E × B velocity in the co-current
direction. Due to the introduction of this stochastic term, the dependence of the excitation
threshold on the rotation changes. For ω0 > 0 the threshold gets higher, for ω0 < 0 the
threshold is lower than in the case without stochastic field. The minimal threshold however
remains at ω0 = 0.
As an answer to the question ‘How do we avoid error field modes’, we found out that a
high ω0 and thus a high rotation Ωφ,0 is important. Because the minimum of the excitation
threshold lies at ω0 = 0 and not at Ωφ,0 = 0 the direction of the plasma rotation, relative
towards the direction of the electron diamagnetic frequency fe∗ is important: A plasma
rotation velocity in the direction of the electron diamagnetic drift direction has a higher
threshold than the same plasma rotation velocity in the ion diamagnetic drift direction.
If a stochastic region is present in the plasma edge, plasmas with ω0 rotating in the co-
current direction have a higher threshold for mode excitation than those plasmas that are
rotating with ω0 in the counter-current direction. The best way to avoid mode excitation
therefore is to create a fast rotating plasma in the co-current direction in a tokamak where
the electron diamagnetic drift direction is also in the co-current direction, i.e. where the
plasma current and the toroidal magnetic field are parallel.
11.2 Influence of ICRH and ECRH on rotation in TEXTOR

As an answer to our question how to avoid error field modes, we found out that a fast
rotating plasma is crucial. Also the direction of the rotation is important. In current
day devices plasma rotation is mainly driven by tangential neutral beam injection. This
method might not be too efficient in future devices. It is therefore advisable to look into
other possible techniques to transfer rotational momentum to a plasma.
Ion cyclotron resonant heating (ICRH) is often mentioned as a promising source of
momentum. In several tokamaks a change in plasma rotation is observed when ICRH is
applied. The mechanism that creates this rotation is however not very well understood.
Some models predict a counter-rotation, e.g. through the loss of fast ions. Other models
predict co-rotation, e.g. through a scaling of spontaneous co-rotation with the thermal
energy of the plasma. Measurements confirm that both co- and counter-rotation are
146 Section 11.2 - Influence of ICRH and ECRH on rotation in TEXTOR

possible, depending on specific plasma conditions.

Some preliminary experiments were carried out to investigate the influence of ICRH on
the rotation in TEXTOR. It was found that co-rotating, beam-driven plasmas show a large
decrease in plasma rotation during ICRH, while in counter-rotating plasmas the rotation
slightly increased during ICRH. These measurements suggest that the mechanisms at work
in TEXTOR are
• a reduction of the momentum confinement time due to the extra
power input by ICRH.
• a momentum input in the counter-direction, presumably through
the loss of fast ions.
These experiments show that the investigation of ICRH generated rotation is an interest-
ing topic for future research at TEXTOR.
If ICRH influences plasma rotation, it is only a small step to the question whether
ECRH will have an effect on rotation. Also in this field we did some first experiments.
Again it is observed that ECRH changes the plasma rotation. At first sight it seems that
local changes in energy transport by ECRH also effect the local transport of momentum.
However further and more dedicated experiments are needed to fully investigate this topic.
11.3 Rotation in ITER

Creating a fast rotating plasma is essential for avoiding MHD modes that are triggered
by the error field of a tokamak. Also a large rotation velocity, or more precisely a large
velocity shear, suppresses turbulent transport and thus increases the energy confinement.
In current day devices high rotation velocities are achieved by momentum input with
tangential neutral beams. Apart from neutral beam injection, rotation can be created with
ICRH. Even without external momentum input spontaneous plasma rotation is observed
in several tokamaks.
ITER, the ‘next step device’, has to prove that fusion energy is scientifically, technically
and economically feasible. It therefore has to operate at maximum performance; this
means – among other things – high confinement and high stability. Fast rotating plasmas
in ITER, in order to achieve the confinement and stability requirements, are a must.
Sceptics of the ITER project have pointed out that rotation might be a weak point in
ITER [36]: high turbulence would lead to a poor confinement if it is not suppressed by
rotational shear [52]. We therefore have a closer look into the expected rotation at ITER.
Co- and counter rotation in ITER Let us first point out the directions in ITER.
Looking on top of the machine the plasma current and toroidal magnetic field are clockwise.
Also the heating neutral beams inject momentum in the clockwise direction. We define
positive toroidal rotation as the rotation in the clockwise direction, hence in the direction
of the current and the magnetic field. Momentum input by the neutral beams causes a
positive rotation. Because the magnetic field and plasma current are in the same direction
the electron diamagnetic drift is in the co-direction – hence fe∗ is positive – while the ion
diamagnetic drift is in the counter-direction.
To have an idea about the plasma in ITER, table 11.1 gives some typical ITER pa-
rameters.
Section 11.3 - Rotation in ITER 147

Table 11.1 : Parameters for ITER
Major radius (R) 6.2 m

Minor radius (a) 2.0 m
Elongation 1.8
Plasma current (Ip ) 15 M A
Toroidal magnetic field (Bφ ) 5.3 T
Edge safety factor (q95 ) 3
Averaged density (n¯e ) 1 · 1020 m−3
Averaged ion temperature (T̄i ) 8.1 keV
Averaged electron temperature (T̄e ) 8.9 keV
External heating power 40 M W
Normalised plasma energy (βN ) 1.8
NBI driven rotation in ITER In ITER the neutral beams will provide toroidal mo-
mentum input as well. ITER is however a much bigger device than the current generation
of tokamaks. To accelerate the larger plasma volume a lot of torque, and therefore beam
power, is needed. On top of that the injection geometry of the neutral beams is not opti-
mal for toroidal momentum transfer, due to the limited space between the toroidal field
coils. In [3] and [62] the maximum central toroidal velocity in ITER, driven by the neutral
beams, is estimated to be 70 km/s. This is lower than the maximum central toroidal
velocity in TEXTOR (100 km/s) where the beam power is only 1.5 M W compared to
the 33 M W in ITER. For error field mode excitation it is not the rotation velocity that
is important, but the rotation frequency. Due to the large major radius of ITER the
comparison between the rotation frequencies at TEXTOR and ITER is even worse:
Ωφ,ITER = 1.2 · 104 rad/s

Ωφ,TEXTOR = 6.0 · 104 rad/s
Spontaneous and ICRH driven rotation in ITER Above we saw that the neutral
beam driven rotation in ITER will be low compared to present day machines. Whereas a
high rotation is desirable for a good confinement and stability.
Neutral beams are however not the only sources of rotation. It is observed in many
present day machines that even without external momentum input a plasma does rotate.
The amount of this rotation and even its direction does differ quite a bit and there is, up
to now, no theoretical model that can fully explain this rotation. What is known, however,
is the following:
• For L-mode plasmas both spontaneous rotation in the co- as the counter-direction is
observed, but for H-mode plasmas the plasma rotation is in the co-current direction.
• The profile of ion temperature Ti and toroidal velocity Ωφ are similar [18, 86]. This
means the rotation scales with the thermal energy of the plasma.
148 Section 11.3 - Rotation in ITER

• Several theoretical models, e.g. [17] and [78], are in qualitative agreement with the
above observations.
Because ITER is supposed to operate in H-mode a spontaneous co-rotation is expected.

This rotation is expected to scale with the thermal energy.
To increase the spontaneous rotation, one can try to increase the thermal energy of the
plasma, e.g. with ICRH. Some caution is necessary here, as ICRH can also be responsible
for a loss of fast ions resulting in a counter-momentum input and an effective reduction of
the spontaneous co-rotation.
In [41] and [62] the scaling of rotation with thermal energy is applied to many machines
and an extrapolation is made towards ITER. The best scaling is given by M = (0.19 ±
0.11)βN , where M = vφ /vth is the Mach number and βN represents the plasma energy.
Due to the large scatter in the data points, the error is this scaling is huge. Extrapolation
to ITER, with an expected βN of 1.8, yields a spontaneous rotation velocity of vφ =
350 ± 200 km/s or Ωφ = 5.8 ± 3.3 · 104 rad/s. In the worst case this is still a higher
rotation than the beam driven one, in the best case the rotation in ITER is in the order
of the rotation we have in current day tokamaks, but then without the need for neutral
beams.
Error field penetration and resistive wall modes in ITER When taking sponta-
neous rotation into account it might be expected that there is a reasonable rotation in
the co-direction in ITER. The question remains: Given the expected plasma rotation in
ITER, which error field, e.g. due to coil misalignment, is allowed in ITER before modes
are excited?
Above we mentioned that the threshold for error field mode excitation, when the
stochastic term is neglected, is proportional to |ω0 |, where ω0 = nΩ0 + 2πfe∗ . This was
the result of a calculation in the so-called visco-resistive regime. In a large machine like
ITER the resistivity will be low, and the viscous, instead of visco-resistive, approach has
to be used. This leads to a threshold for a m/n tearing mode given by [54, 31]:
m,n 4/3
Br,crit

ω0 R
∝ n̄e2/3 (11.3)
Bφ Bφ
The minimal threshold still lies at ω0 = 0 thus Ω0 = −2πfe∗ /n. As said above for
ITER fe∗ > 0. This means that any plasma rotation in the co-direction leads you away
from the minimal threshold. This in contrast to the TEXTOR case were fe∗ < 0 and the
minimum lies at a specific co-rotation.
If we now take the values from table 11.1, fe∗ = 0.2 kHz. For a non-rotating ITER
plasma, this means that ω0 = 1.2 · 103 rad/s for a 2/1 mode. The proportionality factor
for equation (11.3) can be found in [54] and is 2 · 10−21 . As a result the threshold for the
2,1
excitation of a 2/1 mode in ITER, when the plasma is not rotating is: Br,crit /Bφ = 5·10−5 .
2,1
For TEXTOR this was measured1 to be: Br,crit /Bφ = 4 · 10−4 .
ITER will thus be more sensitive to mode excitation by error fields than e.g. TEX-
2,1
TOR. The value Br,crit /Bφ = 5 · 10−5 is taken as a crucial parameter in the construction
of the ITER coil system: the alignment of the coils must be accurate enough to have
Br2,1 /Bφ < 5 · 10−5 . Of course for rotating ITER plasmas the threshold will increase. For
1
The conversion from IDED uses in this thesis to Br2,1 is: Br2,1 [T ] = 10−5 IDED [kA]
Section 11.3 - Rotation in ITER 149

a rotation of Ωφ,0 = 3 · 103 rad/s at the q = 2 surface the threshold is already six times
2,1
bigger: Br,crit /Bφ = 3.0 · 10−4 .
Although important, tearing modes are not the main concern in ITER; resistive wall
modes are (RWM). These RWM’s occur in plasmas with a high plasma energy (β), like in
ITER. An ideal kink mode at the plasma edge, a deformation of the flux surfaces without
reconnection, takes the form of a resistive wall mode, when the tokamak wall is resistive.
As long as a RWM rotates fast enough with respect to the wall, it will be suppressed. The
interaction between the suppressed RWM and the resistive tokamak wall will, however,
slow down the RWM rotation. This can lead to a locking of the RWM, which causes a
disruption.
Although RWM’s have a different topology than tearing modes, the theory describing
RWM locking is very similar to the theory that describes the locking of tearing modes to
the wall and the excitation of tearing mode by an external error field. This theory predicts
that a plasma rotation velocity of about 2% of the Alfvèn velocity should be sufficient to
stabilise RWM’s in a typical ITER discharge [62]. The Alfvèn velocity in ITER will be
around 105 km/s, which means that a plasma rotation of 200 km/s is needed [62]. The
NBI driven rotation can not induce this kind of rotation velocity; as we saw above the NBI
driven rotation velocity in ITER goes up to 70 km/s. However, although our knowledge
on spontaneous rotation is limited, the extrapolation shown [41] and [62] indicates that
the spontaneous rotation velocity could reach high enough values (150 to 550 km/s) to
assure RWM stabilisation in ITER.
Influence of a stochastic boundary In this thesis we found that a stochastic plasma

edge has an influence on the mode penetration as well. It will increase the threshold for
plasmas with ω0 > 0 and decrease the threshold for plasmas with ω0 < 0. Thanks to the
fact that spontaneous rotation and beam driven rotation is in the co-direction in ITER
and that the electron diamagnetic drift is in the co-direction as well, the ITER stability
would benefit from a stochastic edge.
It was also seen that for ω0 > 0 an initial increase in rotation occurs. Such an increased
rotation is not only beneficial for stability, but can also attribute to the suppression of
turbulence.
In TEXTOR the DED is designed to create a strong perturbation of the magnetic
field in the edge. In ITER this is not the case. The intrinsic error field is not expected to
create a large stochastic region in the plasma edge, thus the effects of due to a stochastic
boundary will be negligible. It may be difficult to implement perturbation coils that create
a stochastic region in a complex machine like ITER, but the possibility of a better stability,
a higher rotation and the suppression of unwanted ELM’s during H-mode – not discussed
in this thesis, but observed in the DIII-D tokamak [24] – make it at least worthwhile to
look into this matter.
150 Section 11.3 - Rotation in ITER

Summary
Fusion energy is an attractive candidate for future energy production. At very high tem-
peratures the nuclei of two hydrogen isotopes, deuterium and tritium, fuse, producing
helium, a neutron and a very large amount of energy.
The high temperatures that are needed to make fusion reactions possible can be reached
in a plasma that is magnetically confined. The most successful concept for the magnetic
confinement of a plasma is the tokamak. However, despite the success of the tokamak,
several challenges still lie ahead. One of these challenges is the magnetic stability.
If we want to assure the magnetic stability of a tokamak, it is necessary to know what
causes magnetic instabilities. An important class of magnetohydrodynamic instabilities
are so-called tearing modes or magnetic islands. These tearing modes can be triggered
by an external magnetic error field, which can be due to misalignment of field coils,
asymmetries in the tokamak vessel et cetera. It has been observed on several tokamaks
that the excitation of these error field induced modes depends strongly on the plasma
rotation. Also theories have been developed to describe this dependency.
The TEXTOR tokamak is very well suited to investigate the relation between plasma
rotation and the excitation of tearing modes. Two tangential neutral beam injectors give
us complete control over the plasma rotation. A set of perturbation coils, called the
dynamic ergodic divertor (DED), provide a fully adjustable perturbation field. With the
tangential neutral beams and the DED, dedicated experiments were carried out to test
the theory on mode excitation.
When an external perturbation field is applied, the plasma rotation is expected change
until the rotation velocity is so that tearing modes would be at rest in the frame of the
perturbation field. Once this rotation velocity is reached, large tearing modes will develop
in the plasma. This is called mode excitation or mode penetration. A too low perturbation
field may not induce enough change in plasma rotation to reach the rotation velocity for
which the tearing modes are at rest in the frame of the perturbation field. This means
that the perturbation level needs to reach a certain threshold in order to excite tearing
modes.
The theory found in the literature predicts a monotonic change in the plasma rotation.
It also predicts that the threshold for mode excitation increases with the plasma rotation
velocity. The direction of the rotation, within the frame of the perturbation field, is not
important, hence the threshold as a function of the rotation velocity is symmetric around
the minimal threshold.
The results of our experiments deviate from the predictions made by the theory. In-
stead of a monotonic change in the plasma rotation, we found that the plasma has an
initial tendency to rotate in the co-current direction when the DED is applied. For a
plasma that rotates faster in the co-current direction than the DED field, this means that
151
the plasma first accelerates and only at a high perturbation strength starts to slow down.
Not only the measured change in plasma rotation was different than expected. Also
the relation between the plasma rotation and the threshold for mode excitation was not
in agreement with the theory. Instead of the predicted symmetry around the minimal
threshold, we measured that the threshold increases faster with increasing plasma co-
rotation than with increasing counter-rotation.
The reason for this discrepancy between theory and experiment was found in the
plasma edge. An external perturbation field creates a stochastic zone in the plasma edge.
Fast electron transport parallel to the stochastic field lines leads to a radial electron loss
current. Ambipolarity requires this loss current to be balanced by a perpendicular return
current, that in turn exerts a j × B-force onto the plasma. This stochastic torque is
always in the co-current direction and saturates for a high perturbation strength. In
most tokamaks the stochastic region will be very small and the stochastic force can be
ignored. In TEXTOR however, the stochastic force is quite large, because one of the main
objectives of the DED is to significantly perturb the plasma edge.
The conventional theory of mode excitation does not take this stochastic force into
account. The change in plasma rotation and the threshold as a function of plasma rotation
is found by balancing an electromagnetic torque at the position of the mode with a viscous
torque that opposes any change in the plasma rotation. Because the stochastic torque
can not be ignored, in TEXTOR the balance of the EM torque, the viscous torque and
the stochastic torque has to be taken. With the stochastic force included in the mode
excitation model, the measurements of the plasma rotation and the rotation dependence
of the mode excitation could be explained.
For co-rotating plasmas the EM torque tries to slow the plasma down, while the
stochastic torque tries to increase the plasma velocity. In an initial stage the stochas-
tic torque is dominant and the plasma rotation increases. When the perturbation field is
increased the EM torque takes over, the plasma slows down and mode excitation occurs.
The stochastic torque opposes the EM torque, resulting in a high the excitation thresh-
old. For counter-rotating plasmas, both the stochastic and the EM torque slow down the
plasma. As a result the threshold for mode excitation is not as high for counter-rotating
plasmas as it is for plasmas rotating with the same speed in the co-direction.
Two conclusions can be drawn from this work: (a) In absence of a stochastic field the
conventional mode excitation theory is valid. (b) The introduction of a stochastic edge
field is beneficial for co-rotating plasmas: the threshold for mode excitation increases and
– for moderate perturbations – the rotation itself also increases.
ITER is expected to rotate in the co-direction. It may be therefore worthwhile to
look into the possibility of a stochastic edge. Especially because, apart from a higher
mode threshold and a faster rotation, stochasticity also suppresses unfavourable ELM’s
in H-mode.
Both with and without a stochastic edge region, the fact remains that a fast rotating
plasma is less sensitive to tearing modes than a slowly rotating plasma. In present day
tokamaks neutral beam injectors provide the necessary high rotation speed. For the next
generation of fusion reactors the neutral beams will not be able to deliver enough momen-
tum to a plasma. Other sources of momentum input are therefore needed. A promising
source of momentum input is ion cyclotron resonance heating (ICRH). The effect of ICRH
on plasma rotation is, however, still not very well understood. Depending on the plasma
152
conditions ICRH can induce both co- or counter-rotation. In TEXTOR it was found that
ICRH acts as a source of momentum in the counter-current direction.
If ICRH changes the plasma rotation, we may assume that also electron cyclotron
resonance heating (ECRH) has an influence on the plasma rotation. In TEXTOR it was
found that the changes in rotation during ECRH were related to changes in the local
transport. This opens perspectives for local shaping of the rotation profile. Especially
the possibility to locally increase the velocity gradient is interesting, because a sheared
velocity suppresses turbulence and thus increases the confinement.
153
Samenvatting
Voor de toekomstige energieproductie is fusie een veelbelovende kandidaat. Bij hoge tem-
peraturen kunnen de kernen van de waterstofisotopen deuterium en tritium fuseren, waar-
bij helium, een neutron en veel energie vrijkomt.
De hoge temperaturen die nodig zijn om de fusiereacties mogelijk te maken kunnen
bereikt worden in een plasma dat magnetisch wordt opgesloten. Het meest succesvolle
concept voor magnetische opsluiting is de tokamak. Ondanks het succes van de tokamak
blijven er nog heel wat uitdagingen over in het fusieonderzoek. Een van die uitdagingen
is de magnetische stabiliteit in een tokamak.
Wanneer we de magnetische stabiliteit willen verzekeren, moeten we weten hoe we
magnetische instabiliteiten kunnen vermijden. Een belangrijke klasse van magnetohydro-
dynamische instabiliteiten zijn de zogenaamde ‘tearing modes’ of magnetische eilanden.
Deze modes kunnen worden aangeslagen door een imperfectie in het magnetisch veld;
bijvoorbeeld door een foute uitlijning van magnetische spoelen, door het feit dat een toka-
mak niet perfect symmetrisch is, etc. Bij verscheidene tokamaks is opgemerkt dat de
excitatie van modes sterk afhangt van de plasmarotatie. Theoriën zijn ontwikkeld om
deze afhankelijkheid te beschrijven.
De TEXTOR tokamak is erg geschikt om de relatie tussen plasmarotatie en de excitatie
van modes te onderzoeken. Twee tangentiële neutrale bundels verzekeren een volledige
controle over de plasmarotatie. Een set perturbatiespoelen, de dynamisch-ergodische di-
vertor (DED), voorziet in een aanpasbaar perturbatieveld. Met de neutrale bundels en de
DED werden experimenten uitgevoerd die de theorie over mode-excitatie kunnen testen.
Wanneer een extern perturbatieveld wordt aangeschakeld, verwacht men dat de plas-
marotatie verandert tot de rotatiesnelheid zo is dat magnetische eilanden in rust zijn ten
opzichte van het perturbatieveld. Op het moment dat deze rotatiesneltheid bereikt is,
zullen grote magnetisch eilanden zich onwikkelen in het plasma. Dit wordt mode-excitatie
of mode-penetratie genoemd. Een zwak perturbatieveld zou te weinig verandering in ro-
tatie kunnen veroorzaken, zodat de rotatiesnelheid, waarbij de magnetische eilanden in
rust zijn ten opzichte van het perturbatieveld, niet gehaald wordt. Dit betekent dat de
perturbatie een zekere drempelwaarde moet bereiken, alvorens magnetische eilanden wor-
den aangeslagen.
De theorie, die men kan terugvinden in de literatuur, voorspelt een monotone veran-
dering in de rotatiesnelheid. Ze voorspelt ook dat de drempelwaarde voor mode-excitatie
toeneemt met toenemende rotatiesnelheid. De richting van de rotatie, ten opzichte van het
perturbatieveld, speelt echter geen rol, hetgeen betekent dat drempelwaarde als functie
van de rotatiesnelheid symmetrisch is rondom de minimale drempelwaarde.
De resultaten van de uitgevoerde experimenten wijken af van de theoretische voor-
spellingen. In plaats van een monotone verandering van de plasmarotatie in de richting
155
van de rotatiesnelheid van het perturbatieveld, vonden we dat de plasmarotatie aanvanke-
lijk de neiging heeft om te versnellen in de richting van de plasmastroom (co-rotatie). Voor
een plasma dat roteert in de co-richting met een rotatiesnelheid hoger dan de snelheid van
het perturbatieveld, betekent dit dat rotatie bij het verhogen van het perturbatieveld
eerst toeneemt en pas later – bij een sterk perturbatieveld – afneemt in de richting van de
perturbatiesnelheid.
Niet enkel de gemeten rotatie is anders dan de verwachting. Ook de relatie tussen
de plasmarotatie en de drempelwaarde voor mode-excitatie komt niet overeen met de
theoretische voorspelling. In plaats van een symmetrie rond de minimale drempelwaarde
vinden we een drempelwaarde die sneller stijgt met stijgende co-rotatie dan met stijgende
counter-rotatie.
De reden voor het verschil tussen de theorie en de experimenten werd gevonden in
de plasmarand. Een extern perturbatieveld creëert daar een stochastische zone. Het
snelle transport van elektronen, parallel aan de stochastische veldlijnen, veroorzaakt een
radiële verliesstroom aan elektronen. Ambipolariteit zorgt ervoor dat deze verliesstroom
gecompenseerd wordt door een loodrechte terugkeerstroom, die op zijn beurt een j × B-
kracht uitoefent op het plasma. Deze stochastische kracht is steeds gericht in de co-richting
en verzadigt bij een sterk perturbatieveld. In de meeste tokamaks is de stochastische
zone zeer klein en kan de stochastische kracht worden verwaarloosd. In TEXTOR is de
stochastische kracht behoorlijk sterk, aangezien het net één van de objectieven van de
DED is de plasmarand sterk te verstoren.
De conventionele theorie over mode-excitatie neemt deze stochastische kracht niet in
aanmerking. De verandering van de plasmarotatie en de drempelwaarde voor excitatie als
functie van de plasmarotatie wordt gevonden door een elektromagnetische kracht op de
positie van de mode te balanceren met een visceuze kracht die elke verandering in plas-
marotatie tegengaat. Aangezien de stochastische kracht in TEXTOR niet verwaarloosd
kan worden, moet in TEXTOR the elektromagnetische kracht gebalanceerd worden door
de visceuze én de stochastische kracht. Wanneer de stochastische kracht in het mode-
excitatie model wordt ingevoerd, dan kunnen de metingen van de plasmarotatie en de
drempelwaarde voor mode-excitatie worden verklaard.
Voor co-roterende plasmas tracht de EM-kracht het plasma af te remmen, terwijl
de stochastische kracht de rotatie wil laten toenemen. Aanvankelijk is de stochastische
kracht dominant en versnelt het plasma. Wanneer het perturbatieveld toeneemt, neemt
de EM-kracht het over. De plasmarotatie remt af naar de rotatiesnelheid van het pertur-
batieveld en uiteindelijk wordt de mode geëxciteerd. Aangezien de stochastische kracht de
EM-kracht tegenwerkt, is de drempelwaarde voor excitatie hoog. Voor counter-roterende
plasma’s remmen zowel de EM-kracht als de stochastische kracht de plasmarotatie af
in de richting van de perturbatiesnelheid. Het gevolg van de samenwerking tussen EM-
en stochastische kracht is een lagere drempelwaarde voor mode-excitatie voor counter-
roterende plasma’s dan voor co-roterende plasma’s met dezelfde snelheid.
Twee conclusies kunnen getrokken worden uit dit werk: (a) In afwezigheid van een
stochastische zone is de bestaande mode-excitatie theorie correct. (b) De introductie van
een stochastische plasmarand is voordelig voor co-roterende plasmas: de drempelwaarde
voor mode-excitatie stijgt en de rotatie zelf stijgt eveneens.
De plasma’s in ITER zullen roteren in de co-richting. Een studie naar de mogelijkheid
van het implementeren van een stochatische rand in ITER is de moeite waard. Vooral
omdat, naast een hogere drempelwaarde voor mode-excitatie en een snellere rotatie,
156
stochasticiteit ook zorgt voor de onderdrukking van nefaste ELM’s in H-mode plasma’s.
Zowel met als zonder stochastische rand zijn snel roterende plasma’s minder gevoelig
voor mode-excitatie dan traag roterende plasma’s. In de huidige generatie tokamaks zorgen
tangentiële neutrale bundels voor een snelle rotatie. In de komende generatie fusiereac-
toren zullen neutrale bundels niet in staat zijn om voldoende momentum te genereren.
Andere bronnen van rotatie zijn daarom noodzakelijk. Een interessante rotatiebron is
ionen cyclotron verhitting (ICRH). Het effect dat ICRH heeft op de plasmarotatie wordt
echter nog niet goed begrepen. Afhankelijk van de plasmacondities kan ICRH het plasma
laten roteren in de co- of in de counter-richting. In TEXTOR gedraagt ICRH zich als een
momentumbron in de counter-richting.
Als ICRH de plasmarotatie verandert, dan kunnen we ook aannemen dat elektron
cyclotron verhitting (ECRH) een invloed heeft op de plasmarotatie. In TEXTOR kon-
den we waarnemen dat de verandering in rotatie tijdens ECRH gerelateerd is aan het
lokale transport van energie. Dit opent perspectieven voor het lokaal vormgeven van het
rotatieprofiel. Vooral het veroorzaken van een lokale gradiënt in de rotatie is daarbij
interessant, aangezien een snelheidsgradiënt turbulentie onderdrukt en op die manier de
opsluiting van het plasma verbetert.
157
Curriculum Vitae
I was born on December 27, 1979 in the Belgian city Sint-Niklaas. I attended sec-
ondary school there and in 1997 I graduated from high school with a certificate in Latin-
Mathematics. In September 1997 I went to Ghent University in Belgium to study applied
physics. As a part of my study I attended a traineeship at the European Organisation for
Nuclear Research (CERN), Geneva, in the summer of 2001. In July 2002 I graduated as
a M.Sc. in Applied Physics at Ghent University.
One month later, in August 2002, I started to work as a research student at the FOM-
Institute for Plasma Physics “Rijnhuizen” (Nieuwegein, The Netherlands). I investigated
the interplay between external, magnetic perturbation fields and the rotation of a plasma
in a tokamak. This work, which is reported in this thesis, was carried out at the TEXTOR
tokamak of the Institut für Plasmaphysik, Forschungszentrum Jülich, Germany.
159
Acknowledgements
I would like to express my gratitude to all those people that supported me and helped me
with my work during the course of the last four and a half years. First of all I would like
to thank Niek Lopes Cardozo, Chris Schüller and Roger Jaspers, who as my promotors
and co-promotor guided me through the jungle of plasma physics. The discussions with
Niek, on a wide range of scientific topics, gave direction to this work: what are the main
questions we should pose, what do we want to achieve with this investigation? Through the
everyday discussions with Roger, my daily supervisor, I got familiar with tokamak physics,
TEXTOR and the CXRS system. His enthusiasm was a constant source of motivation
and a counterweight for my scepticism. Without their help this thesis would not be lying
in front of you today.
These last couple of years I worked at the TEXTOR tokamak of the Institut für Plasma
Physik of the Forschungszentrum Jülich. I would like to thank the whole TEXTOR team
and the ‘Jülich Team’ of the FOM Institute for Plasma Physics “Rijnhuizen”, because
without their help the experiments that were carried out for this thesis would not have
been possible.
I also would like to thank Manfred von Hellermann. Whenever I got stuck with the
analysis, I could always rely on his expertise in charge exchange recombination spec-
troscopy.
The discussions on MHD ‘things’ with Ivo Classen, my roommate for most of my time
in Jülich, were a great help to improve my understanding of plasma physics. Or, when we
both didn’t have a clue as to what our measurements meant, it was at least a comfort to
find that I was not the only one who did get puzzled by all this physics.
For the harsh job of proofreading a thesis on plasma physics as a non-physicist, I
would like to thank Fieke Van der Gucht. Her help allowed me to minimise the number
of violations against the English language.
There is, however, more to life than just physics. So I would like to thank a lot of
people – Cor, Ben, Geert-Willem, Tony, Egbert, Serge, Fenton, Jurrian, Cor, Ephrem and
all other OIO’s and FOMMER’s – for the many conversations during the coffee breaks
and the Tuesday-night-beers. Those conversations, that usually had nothing to do with
physics at all, more or less brought me back from physics to the real world.
Finally I would like to thank my family and friends in Belgium, for putting up with me
when I was talking ‘strange’ again, trying to explain them the benefits of nuclear fusion.
They always gave me a warm welcome when I returned to Belgium for the weekend after
a week of physics in Germany.
161
Appendix A
List of the discharges used in this

thesis
figure 1.6 #98369, #98370, #98371, #98372, #98373, #98374, #98441,

#98442, #98443, #98450, #98453, #98457, #98458, #98518,
#98523, #98524
figure 5.1 #94091, #95022, #95028, #96619, #97613,
#98451, #101059, #101060
figure 5.2 #101398
figure 5.3
figure 5.4 #96619
figure 6.1 #98451, #98452, #98453
figure 6.2 #98452
figure 6.3 #98453
figure 6.4 #98451
figure 6.5 #94091, #94092, #94093, #97613, #97617, #97621, #98451,
figure 6.6 #98452, #98453, #95022, #95024, #95025, #95026, #95027,
#95028
figure 6.7 #94087, #94088, #94089, #94090, #94091, #94092, #94093
figure 6.9 #98455, #98519
figure 6.10 #94091, #94092, #94262, #94464, #94466, #94469, #94471,
#94584, #94586, #94993, #95022, #98451, #98453
163
Appendix A - List of the discharges used in this thesis
figure 6.11 #95202, #98182

figure 6.12 #94727
figure 6.13
figure 6.14 #98369, #98370, #98371, #98372, #98373, #98374, #98441,
#98442, #98443, #98450, #98453, #98457, #98458, #98518,
#98523, #98524 (DC scan)
#98451, #98455, #98519, #98520, #98521, #98532 (AC+ scan)
figure 6.15 #98369, #98370, #98371, #98372, #98373, #98374, #98441,
#98442, #98443, #98450, #98453, #98457, #98458, #98518,
#98523, #98524
figure 6.16 #98369, #98370, #98371, #98372, #98373, #98374, #98441,
#98442, #98443, #98450, #98453, #98457, #98458, #98518,
#98523, #98524 (threshold plot)
#98455, #98519 (rotation plot)
figure 8.2 #95592
figure 8.3
figure 8.4
figure 8.6
figure 9.6 #98369, #98370, #98371, #98372, #98373, #98374, #98441,
#98442, #98443, #98450, #98453, #98457, #98458, #98518,
#98523, #98524 (DC scan)
#98451, #98455, #98519, #98520, #98521, #98532 (AC+ scan)
figure 9.7 #98369, #98373, #98374, #98443, #98450, #98453
figure 9.8 #98451, #98455, #98519, #98520, #98521, #98532
figure 10.1 #101059, #101060
figure 10.2 #96507, #96508
figure 10.3 #98213
figure 11.1 #98369, #98370, #98371, #98372, #98373, #98374, #98441,
#98442, #98443, #98450, #98453, #98457, #98458, #98518,
#98523, #98524 (threshold plot)
#98455, #98519 (rotation plot)
164
Bibliography
[1] S. S. Abdullaev, K. Finken and K. H. Spatschek. “Asymptotical and mapping meth-

ods in study of ergodic divertor magnetic field in a toroidal system.” Physics of
Plasmas, vol. 6 (1) pp. 153–174 (1999). Cited at pages 40, 113, 118
[2] R. Balescu. Transport processes in plasmas: Neoclassical transport. North Holland

(1988). Cited at pages 11, 24
[3] I. P. basis. “Chapter 2: Plasma confinement and transport.” Nuclear Fusion, vol. 39
pp. 2175–2249 (1999). Cited at pages 148
[4] S. H. Batha, S. D. Scott, D. R. Mikkelsen et al. “Effect of Radial Electric Field Shear
on Tokamak Transport: Flow Shear and Magnetic Field Scaling.” In ECA, ed., “24th
EPS Conference on Controlled Fusion and Plasma Physics,” (1997). Cited at pages
20
[5] L. R. Baylor, K. H. Burrell, R. J. Groebner et al. “Comparison of toroidal rotation

velocities of different impurity ions in the DIII-D tokamak.” Physics of Plasmas,
vol. 11 (6) pp. 3100–3105 (2004). Cited at pages 30
[6] J. M. Berger, W. A. Newcomb, J. M. Dawson et al. “Heating of a Confined Plasma by

Oscillating Electromagnetic Fields.” Physics of Fluids, vol. 1 (4) pp. 301–307 (1958).
Cited at pages 26
[7] H. Biglari, P. H. Diamond and P. W. Terry. “Influence of sheared poloidal rotation

on edge turbulence.” Physics of Fluids B, vol. 2 (1) pp. 1–4 (1990). Cited at pages 5
[8] J. W. S. Blokland, R. Keppens, J. P. Goedbloed et al. “MHD equilibrium reconstruc-

tion for TEXTOR: from static to stationary equilibria.” In ECA, ed., “32nd EPS
Conference on Controlled Fusion and Plasma Physics,” (2005). Cited at pages 16, 17
[9] S. I. Braginskii. Transport processes in a plasma, Reviews of plasma physics, vol. 1.

Consultants Bureau, New York, m. a. leontovich edn. (1965). Cited at pages 13, 21
[10] S. Brezinsek, A. Huber, S. Jachmich et al. “Plasma edge diagnostics for TEXTOR.”
Fusion Science and Technology, vol. 47 (2) pp. 209–219 (2005). Cited at pages 44
[11] C. Busch. Spektroskopische Untersuchung der poloidalen Plasmarotation unter dem

Einfluss statischer und dynamischer Ergodisierung am Tokamak TEXTOR. Ph.D.
thesis, Heinrich-Heine-Universität Düsseldorf (2006). Cited at pages 27, 81, 82, 83,
112, 117, 118, 119, 120
165
[12] C. S. Chang, C. K. Phillips, R. White et al. “Generation of plasma rotation by
ion cyclotron resonance heating in tokamaks.” Physics of Plasmas, vol. 6 (5) pp.
1969–1977 (1999). Cited at pages 139
[13] F. F. Chen. Plasma Physics and Controlled Fusion, vol. 1. Pl. Press N.Y. (1993).
Cited at pages 27
[14] H. A. Claassen, H. Gerhauser, A. Rogister et al. “Neoclassical theory of rotation and

electric field in high collisionality plasmas with steep gradients.” Physics of Plasmas,
vol. 7 (9) pp. 3699–3706 (2000). Cited at pages 28, 69, 70, 74
[15] J. W. Connor, S. C. Cowley, R. J. Hastie et al. “Toroidal rotation and momentum

transport.” Plasma Physics and Controlled Fusion, vol. 29 (7) pp. 919–931 (1987).
Cited at pages 22
[16] J. W. Connor, S. C. Cowley, R. J. Hastie et al. “Tearing modes in toroidal geometry.”

Physics of Fluids, vol. 31 (3) pp. 577–590 (1988). Cited at pages 109
[17] B. Coppi. “Accretion theory of ‘spontaneous’ rotation in toroidal plasmas.” Nuclear

Fusion, vol. 42 pp. 1–4 (2002). Cited at pages 71, 149
[18] J. S. deGrassie. “NBI-driven Momentum Confinement in DIII-D (and some

JET results).” 10th meeting of ITPA Transport Physics TG, Apr. 24-27, 2006,
PPPL, Princeton, USA (2006). URL: http://itpa06.pppl.gov/Transport/Tuesday AM
(RMTWG)/deGrassie DIII-D JET.pdf. Cited at pages 20, 137, 139, 142, 148
[19] J. S. deGrassie, D. R. Baker, K. H. Burrell et al. “Toroidal rotation in neutral

beam heated discharges in DIII-D.” Nuclear Fusion, vol. 43 pp. 142–156 (2003).
Cited at pages 19, 20, 137
[20] A. J. H. Donné, M. F. M. de Bock, I. G. J. Classen et al. “Overview of core diagnos-

tics for TEXTOR.” Fusion Science and Technology, vol. 47 (2) pp. 220–245 (2005).
Cited at pages 44
[21] L.-G. Eriksson, G. Hoang and V. Bergeaud. “On the role of ion heating in ICRF
heated discharges in Tore Supra.” Nuclear Fusion, vol. 41 (1) pp. 91–97 (2001).
Cited at pages 138, 139
[22] L.-G. Eriksson and F. Porcelli. “Toroidal plasma rotation induced by fast ions without
external momentum injection in tokamaks.” Nuclear Fusion, vol. 42 pp. 959–971
(2002). Cited at pages 139
[23] L.-G. Eriksson, E. Righi and K.-D. Zastrow. “Toroidal rotation in ICRF-heated H-
modes on JET.” Plasma Physics and Controlled Fusion, vol. 39 pp. 27–42 (1997).
[24] T. E. Evans, R. A. Moyer, P. R. Thomas et al. “Suppression of Large Edge-Localized

Modes in High-Confinement DIII-D plasmas with a stochastic magnetic boundary.”
Physical Review Letters, vol. 92 (23) pp. 235003–1–235003–4 (2004). Cited at pages
150
166
[25] C. Fenzi-Bonizec. “Overview of Plasma Rotation with ICRH and / or LHCD in

Tore Supra.” 10th meeting of ITPA Transport Physics TG, Apr. 24-27, 2006,
PPPL, Princeton, USA (2006). URL: http://itpa06.pppl.gov/Transport/Tuesday AM
(RMTWG)/ITPA Fenzi.ppt. Cited at pages 138, 139
[26] K. H. Finken. “Special Issue: Dynamic Ergodic Divertor.” Fusion Engineering and
Design, vol. 37 (3) pp. 335–448 (1997). Cited at pages 35, 39
[27] K. H. Finken, S. S. Abdullaev, M. Jakubowski et al. The structure of magnetic field

in the TEXTOR-DED. Textor report, IPP (2005). Cited at pages 40, 41
[28] R. Fitzpatrick. “Interaction of tearing modes with external structures in cylindrical

geometry.” Nuclear Fusion, vol. 33 (7) pp. 1049–1084 (1993). Cited at pages 75, 83,
89, 92, 93, 95, 96, 97, 98, 106, 108, 110, 143, 144, 145
[29] R. Fitzpatrick. “Bifurcated states of a rotating tokamak plasma in the presence of a

static error-field.” Physics of Plasmas, vol. 5 (9) pp. 3325–3341 (1998). Cited at pages
96, 98, 105, 106, 108, 143, 144, 145
[30] R. Fitzpatrick, R. J. Hastie, T. J. Martin et al. “Stability of coupled tearing modes

in tokamaks.” Nuclear Fusion, vol. 33 (10) pp. 1533–1576 (1993). Cited at pages 109
[31] R. Fitzpatrick and T. C. Hender. “The interaction of resonant magnetic perturbations

with rotating plasmas.” Physics of Fluids, vol. 3 pp. 644–673 (1991). Cited at pages
96, 143, 144, 145, 149
[32] J. P. Freidberg. Ideal Magnetohydrodynamics. Pl. Press N.Y. (1987). Cited at pages
11, 15
[33] H. P. Furth, J. Killeen and M. N. Rosenbluth. “Finite-Resistivity Instabilities of a

Sheet Pinch.” Physics of Fluids, vol. 6 pp. 459–484 (1963). Cited at pages 98
[34] J. F. M. van Gelder. Electron Cyclotron Wave Absorption Experiments in Tokamak

Plasmas. Ph.D. thesis, Universiteit Eindhoven (the Netherlands), FOM Institute for
Plasma Physics (the Netherlands) (1996). Cited at pages 31, 32
[35] P. Ghendrih, A. Grosman and H. Capes. “Theoretical and experimental investiga-

tions of stochastic boundaries in tokamaks.” Plasma Physics and Controlled Fusion,
vol. 38 (10) pp. 1653–1724 (1996). Cited at pages 112, 113, 118
[36] J. Glanz. “Turbulence may sink titanic reactor.” Science, vol. 274 pp. 1600–1601
(1996). News & Comments. Cited at pages 147
[37] R. D. Hazeltine. “Rotation of a toroidally confined, collisional plasma.” Physics of

Fluids, vol. 17 (5) pp. 961–968 (1974). Cited at pages 28
[38] M. von Hellermann, P. Breger, J. Frieling et al. “Analytical approximation of cross-

section effects on charge exchange spectra observed in hot fusion plasmas.” Plasma
Physics and Controlled Fusion, vol. 37 pp. 71–94 (1995). Cited at pages 49, 54, 55
[39] M. Heyn, I. Ivanov, S. Kasilov et al. “Kinetic modelling of the interaction of rotating
magnetic fields with a radially inhomogeneous plasma.” Nuclear Fusion, vol. 46 pp.
S159–S169 (2006). Cited at pages 126
167
[40] S. P. Hirshman and D. J. Sigmar. Nuclear Fusion, vol. 21 pp. 1079–1230 (1981).
Cited at pages 29
[41] A. Ince-Cushman, J. E. Rice, Y. Podpaly et al. “Multi Machine Scaling of Sponta-
neous Rotation.” 10th meeting of ITPA Transport Physics TG, Apr. 24-27, 2006,
PPPL, Princeton, USA (2006). URL: http://itpa06.pppl.gov/Transport/Wednesday
AM (RMTWG-ITER, Theory&Modeling))/ITPA-Spontaneous Rotation Scaling.ppt.
Cited at pages 71, 149, 150
[42] R. C. Isler. “An overview of charge-exchange spectroscopy as a plasma diagnostic.”
Plasma Physics and Controlled Fusion, vol. 36 pp. 171–208 (1994). Cited at pages
49, 56
[43] S. Jachmich, Y. Xu, M. Jakubowski et al. “Effects of edge ergodization induced by
DED on turbulence and particle transport in TEXTOR.” In ECA, ed., “32nd EPS
Conference on Controlled Fusion and Plasma Physics,” vol. 29C, pp. P–2.033. EPS
[44] B. B. Kadomtsev. “Disruptive instability in Tokamaks.” Soviet Journal of Plasma
Physics, vol. 1 pp. 389–391 (1975). Cited at pages 90
[45] I. Kaganovich and V. Rozhansky. “Transverse conductivity in a braided magnetic
field.” Physics of Plasmas, vol. 5 (11) pp. 3901–3909 (1998). Cited at pages 95, 112,
144
[46] A. Kallenbach, H.-M. Mayer, G. . Fussmann et al. “Characterization of the angular
momentum transport in ASDEX.” Plasma Physics and Controlled Fusion, vol. 33 (6)
pp. 595–605 (1991). Cited at pages 19, 20
[47] F. A. Karelse, M. de Bruijne, C. J. Barth et al. “Measurements of the current density
profile with tangential Thomson scattering in RTP.” Plasma Physics and Controlled
Fusion, vol. 43 pp. 443–468 (2001). Cited at pages 30
[48] Y. Kikuchi, M. de Bock, K. Finken et al. “Forced Magnetic Reconnection and Field
Penetration of an Externally Applied Rotating Helical Magnetic Field in the TEX-
TOR Tokamak.” Physical Review Letters, vol. 97 pp. 085003–1–085003–4 (2006).
[49] Y. B. Kim, P. H. Diamond and R. J. Groebner. “Neoclassical poloidal and toroidal
rotation in tokamaks.” Physics of Fluids B, vol. 3 (8) pp. 2050–2060 (1991).
Cited at pages 28, 29, 68, 69, 70, 74
[50] R. Koch, P. Dumortier, F. Durodié et al. “Ion Cyclotron Resonance Heating on TEX-
TOR.” Fusion Science and Technology, vol. 47 (2) pp. 97–107 (2005). Cited at pages
36
[51] H. Koslowski, E. Westerhof, M. de Bock et al. “Tearing mode physics studies applying
the Dynamic Ergodic Divertor on TEXTOR.” Plasma Physics and Controlled Fusion,
vol. 48 pp. B53–B61 (2006). Cited at pages 90
[52] M. Kotschenreuther, W. Dorland, M. A. Beer et al. “Quantitative predictions of
tokamak energy confinement from first-principles simulations with kinetic effects.”
Physics of Plasmas, vol. 2 (6) pp. 2381–2389 (1995). Cited at pages 147
168
[53] A. Krämer-Flecken, S. Soldatov, C. Busch et al. “Edge turbulence studies at TEX-

TOR during Dynamic Ergodic Divertor opetraion by means of reflectometry.” In
ECA, ed., “32nd EPS Conference on Controlled Fusion and Plasma Physics,” vol.
29C, pp. P–2.031. EPS (2005). Cited at pages 81
[54] R. J. La Haye, R. Fitzpatrick, T. C. Hender et al. “Critical error fields for locked
mode instability in tokamaks.” Physics of Fluids B, vol. 4 (7) pp. 2098–2103 (1992).
Cited at pages 149
[55] E. Lazzaro, R. J. Buttery, T. C. Hender et al. “Error field locked modes thresholds
in rotating plasmas, anomalous braking and spin-up.” Physics of Plasmas, vol. 9 (9)
pp. 3906–3918 (2002). Cited at pages 75, 83, 84, 89, 92, 93
[56] M. Lehnen, S. Abdullaev, W. Biel et al. “Transport and divertor properties of the
dynamic ergodic divertor.” Plasma Physics and Controlled Fusion, vol. 47 pp. B237–
B248 (2005). Cited at pages 117
[57] Y. Liang, H. Koslowski, A. Krämer-Flecken et al. “Influence of anisotropic pressure on
the locking of 2/1 tearing modes in TEXTOR.” In ECA, ed., “32nd EPS Conference
on Controlled Fusion and Plasma Physics,” vol. 29. EPS (2005). Cited at pages 87
[58] Z. Lin, T. S. Hahm, W. W. Lee et al. “Turbulent transport reduction by zonal flows:
Massively parallel simulations.” Science, vol. 281 pp. 1835–1837 (1998). Cited at pages
5
[59] M. Mattioli, J. Ramette, B. Saoutic et al. “Impurity ion temperature and toroidal
rotation velocity in JET from high-resolution x-ray and XUV spectroscopy.” Journal
of Applied Physics, vol. 64 (7) pp. 3345–3352 (1988). Cited at pages 137
[60] N. Mattor and P. H. Diamond. “Momentum and thermal transport in neutral-beam-
heated tokamaks.” Physics of Fluids, vol. 31 (5) pp. 1180–1189 (1988). Cited at pages
19, 22
[61] O. Merkulov. Control of plasma profiles and stability through localised Electron Cy-
clotron Current Drive. Ph.D. thesis, University of Utrecht (2006). Cited at pages 90,
141
[62] V. Mukhovatov. “Estimates of Plasma Rotation Rate in ITER.” 10th meet-
ing of ITPA Transport Physics TG, Apr. 24-27, 2006, PPPL, Princeton, USA
(2006). URL: http://itpa06.pppl.gov/Transport/Wednesday AM (RMTWG-ITER,
Theory&Modeling))/Plasma Rotation in ITER.ppt. Cited at pages 71, 148, 149, 150
[63] National Institute for Standards and Technology. “NIST Atomic Spectra Database.”
(). URL: http://physics.nist.gov/PhysRefData/ASD/index.html. Cited at pages 51,
56, 57
[64] O. Neubauer, G. Czymek, B. Giesen et al. “Design features of the tokamak TEX-
TOR.” Fusion Science and Technology, vol. 47 (2) pp. 76–86 (2005). Cited at pages
36
[65] A. Nicolai, U. Daybelge and C. Yarim. “Modelling of plasma rotation accounting
for neutral beam injection and perturbation coils in JET and TEXTOR.” Nuclear
Fusion, vol. 44 (6) pp. S93–S107 (2004). Cited at pages 119
169
[66] D. Nishijima, A. Kallenbach, S. Günter et al. “Experimental studies of toroidal
momentum transport in ASDEX Upgrade.” Plasma Physics and Controlled Fusion,
vol. 47 pp. 89–115 (2005). Cited at pages 137, 138
[67] J. Ongena and A. M. Messiaen. “Heating, confinement and extrapolation to reactors.”
Fusion Science and Technology, vol. 45 (2T) pp. 453–466 (2004). Cited at pages 20
[68] F. W. Perkins, R. B. White, P. T. Bonoli et al. “Generation of plasma rotation in
a tokamak by ion-cyclotron absorption of fast Alfvén waves.” Physics of Plasmas,
[69] J. Rice, E. Marmar, F. Bombarda et al. “X-Ray observations of central toroidal
rotation in ohmic Alcator C-MOD plasmas.” Nuclear Fusion, vol. 37 (3) pp. 421–426
[70] J. E. Rice, P. T. Bonoli, J. A. Goetz et al. “Central impurity toroidal rotation in
ICRF heated Alcator C-Mod plasmas.” Nuclear Fusion, vol. 39 (9) pp. 1175–1186
(1999). Cited at pages 138, 139
[71] V. Rozhansky and M. Tendler. Reviews of Plasma Physics, chap. Plasma rotation in
tokamaks, p. 147. Consultants Bureau, New York – London (1996). Cited at pages
27
[72] P. H. Rutherford. “Nonlinear growth of the tearing mode.” Physics of Fluids,
[73] U. Samm and the TEXTOR Team. “Special Issue: TEXTOR.” Fusion Science and
Technology, vol. 47 (2) pp. 73–273 (2005). Cited at pages 35
[74] M. van Schoor, S. Jachmich and R. R. Weynants. “An experimental and theoretical
study on the formation of electric field induced flow shear in the tokamak edge.”
Journal of Nuclear Materials, vol. 313–316 pp. 1326–1330 (2003). Cited at pages 22,
23
[75] R. Schorn, E. Wolfrum, F. Aumayr et al. “Radial temperature distributions of C6+
ions in the TEXTOR edge plasma meausered with lithium beam activated charge ex-
change spectroscopy.” Nuclear Fusion, vol. 32 (3) pp. 351–359 (1992). Cited at pages
56
[76] F. C. Schüller. “Profile consistency as a result of coupling between the radial profile
functions of pressure and current density.” In ECA, ed., “18th EPS Conference on
Controlled Fusion and Plasma Physics,” (1991). Cited at pages 31, 32
[77] V. D. Shafranov. “On magnetohydrodynamical equilibrium configurations.” Sov.
Phys. J. Exper. Theor. Phys., vol. 6 (33) pp. 545–545 (1958). Cited at pages 15
[78] K. C. Shaing. “Toroidal momentum confinement in neoclassical quasilinear theory in
tokamaks.” Physics of Plasmas, vol. 8 (1) pp. 193–198 (2001). Cited at pages 139,
149
[79] W. M. Stacey. “Gyroviscous Momentum Transfer and Toroidal Rotation in Tokamaks
with Unbalanced NBI.” Plasma Physics and Controlled Fusion, vol. 31 (9) pp. 1451–
1468 (1989). Cited at pages 22
170
[80] W. M. Stacey and D. J. Sigmar. “Viscous effects in a collisional tokamak plasma with
strong rotation.” Physics of Fluids, vol. 28 (9) pp. 2800–2807 (1985). Cited at pages
13, 21, 22
[81] T. H. Stix. “Decay of poloidal rotation in tokamak plasma.” Physics of Fluids,

[82] D. Testa, C. Giroud, A. Fasoli et al. “On the measurement of toroidal rotation for
the impurity and the main ion species on the Joint European Torus.” Physics of
Plasmas, vol. 9 (1) pp. 243–250 (2002). Cited at pages 29, 30
[83] R. Uhlemann and J. Ongena. “Variation of injected neutral beam power at constant
particle energy by changing the beam target aperture of the TEXTOR neutral beam
injectors.” Fusion Technology, vol. 35 pp. 42–52 (1999). Cited at pages 38
[84] S. Varshney. Dynamics of mesoscale structures in tokamak plasmas. Ph.D. the-

sis, Technische Universiteit Eindhoven (the Netherlands), FOM Institute for Plasma
Physics (the Netherlands) (2006). Cited at pages 87
[85] P. de Vries. Magnetic islands in tokamak plasmas. Ph.D. thesis, University of Utrecht
(1997). Cited at pages 31, 73, 74, 90
[86] P. de Vries. “Momentum Transport Studies at JET.” 10th meeting of

ITPA Transport Physics TG, Apr. 24-27, 2006, PPPL, Princeton, USA (2006).
URL: http://itpa06.pppl.gov/Transport/Tuesday AM (RMTWG)/deVries JET-
Momentum-Transp.pdf. Cited at pages 19, 20, 67, 139, 148
[87] F. Wagner, G. Becker, K. Behringer et al. “Regime of Improved Confinement and

High Beta in Neutral-Beam-Heated Divertor Discharges of the ASDEX Tokamak.”
Physical Review Letters, vol. 49 (19) pp. 1408–1412 (1982). Cited at pages 5
[88] J. Wesson. Tokamaks. Oxford University Press, Great Clarendon Street, Oxford OX2
6DP, 2nd edn. (1997). Cited at pages 11, 13, 15, 19, 21, 22, 59, 90, 97, 118, 141
[89] E. Westerhof, J. A. Hoekzema, G. M. D. Hogeweij et al. “Electron Cyclotron Reso-

nance Heating on TEXTOR.” Fusion Science and Technology, vol. 47 (2) pp. 108–118
[90] H. L. M. Widdershoven and R. Jaspers. “The Influence of Angular Momentum on

Energy Confinement on the TEXTOR-94 Tokamak.” In ECA, ed., “26th EPS Con-
ference on Controlled Fusion and Plasma Physics,” vol. 23J, pp. 609–612. EPS (1999).
Cited at pages 20
[91] K.-D. Zastrow, W. G. F. Core, L.-G. Eriksson et al. “Transfer rates of toroidal angular
momentum during neutral beam injection.” Nuclear Fusion, vol. 38 (2) pp. 257–263
[92] K.-D. Zastrow, K. Crombé, S. E. Sharapov et al. “Special task force meeting on
poloidal rotation.” JET combined TF meeting, 2nd March, 2006 (). Cited at pages
32
171

Understanding and Controlling Plasma Rotation in Tokamaks: Proefschrift

Uploaded by

Copyright:

Available Formats

Understanding and Controlling Plasma Rotation in Tokamaks: Proefschrift

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Understanding and Controlling Plasma Rotation in Tokamaks: Proefschrift

Uploaded by

Copyright:

Available Formats

Understanding and controlling

plasma rotation in tokamaks

ter verkrijging van de graad van doctor aan

prof.dr. N.J. Lopes Cardozo

CIP-data library Technische Universiteit Eindhoven

Trefwoorden: plasmafysica / kernfusie / fusieplasma’s / plasmarotatie / magnetische

That’s orbiting at 19 miles a second,

The sun and you and me,

In an outer spiral arm,

2 Theory of plasma rotation 11

3 TEXTOR, DED and diagnostics 35

4 Charge Exchange Recombination Spectroscopy 49

5 Plasma rotation at TEXTOR 67

6 Measurements of plasma rotation during DED operation 75

7 Theory of perturbation fields 95

8 Momentum input by a stochastic edge field 115

8.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

9 Mode penetration 125

10 Changing the plasma rotation by EM waves 137

11 Discussion, conclusions and outlook towards future devices 143

Curriculum Vitae 159

A List of the discharges used in this thesis 163

1.1 Thermonuclear Fusion

1.1.2 Fusion reactions

n T τE > 5 · 1021 keV s/m3 (1.2)

2 Section 1.1 - Thermonuclear Fusion

Figure 1.2 : Charged particles gyrate closely around

Section 1.3 - Tokamaks 3

1.3 Importance of plasma rotation

4 Section 1.3 - Importance of plasma rotation

Section 1.3 - Importance of plasma rotation 5

6 Section 1.4 - Importance of plasma rotation

Mode threshold (IDED / kA)

1.4 This thesis

Section 1.4 - This thesis 7

To answer this question we can reverse it:

How can we excite modes with an external perturbation field?

How does the threshold for mode excitation by an external

with as specific sub-questions:

- Why does the minimal threshold occur at a finite rotation

How does the external perturbation field change the plasma

8 Section 1.4 - This thesis

1.5 List of publications

Section 1.5 - List of publications 9

10 Section 1.5 - List of publications

Theory of plasma rotation

We start with the general description of a plasma as a collection of individual particles,

to rotate spontaneously. This intrinsic or spontaneous rotation could be very helpful in

2.2 From single particle motion to plasma rotation

12 Section 2.2 - From single particle motion to plasma rotation

vα = hui the fluid velocity

In equations (2.6)-(2.7) the pressure tensor P = P e + P i is split up in the scalar

Section 2.2 - From single particle motion to plasma rotation 13

Usually the assumption of an adiabatic fluid

2.2.2 Tokamak equilibrium

14 Section 2.2 - From single particle motion to plasma rotation

Figure 2.1 : Magnetic flux surfaces in a tokamak

Grad-Shafranov equation To calculate the magnetic R configuration in a tokamak it is

Section 2.2 - From single particle motion to plasma rotation 15