FUSION PHYSICS

- introduction to
the physics behind fusion energy

Jan Scheffel
Per Brunsell

8th English edition

February, 2023

Electromagnetic Engineering and Fusion Science

KTH
1
 PREFACE
Fusion power is one of very few pathways to sustainable future large scale energy pro-
duction. Yet further, large efforts in the field is required to lead the development to a
power-producing reactor.
The realization of fusion power has turned out to be one of the biggest technological
challenges of mankind. The course in Energy and Fusion Research, given KTH, Stock-
holm, will hopefully motivate this development and provide understanding for the physics
behind fusion energy.
In this book, we shall first discuss the problems related to global energy supply, and
why fusion research is motivated. Then, we shall discuss the fusion process, moving on to
learning some fusion physics. We will subsequently return to the experimental situation
of today and discuss some promising fusion configurations.
In the last chapter, we will study how future reactors may be designed, and investigate
the advantages of fusion as compared to other energy sources with respect to safety and
environments.
This book has resulted from several years of teaching the course. It is, as far as we
can see, the first book to cover not only the various aspects of fusion physics but also
the contextual energy background. Our intended reader has a minimum background of
a master’s student on the advanced level, but also other readers may find contents of
interest. Basic knowledge of electromagnetic theory and vector analysis is helpful. To
ease the studies, the text has been divided into levels, corresponding to a taxonomy
of understanding. Sections marked with ”*” provide more detailed information for the
interested reader, but detailed understanding is not required.
A short note on nomenclature: Temperature T is expressed in electron volt (eV), so
that kB T (K)=eT (eV), where kB is Boltzmann’s constant, and e is the elementary charge.
Consequently a temperature of 1 eV corresponds to around 11 600 K (rule of thought
≈ 104 K). More information about temperature and energy units are collected in an
appendix to the book.
In this eight edition, a major update has been made of the first chapter of the book
dedicated to the global energy system and the environment.
Any comments on the text are gratefully received.
Finally we would like to express our deepest gratitude to Miro Sedlacek, who has
helped with the layout of this book, and to Per Sundelin, who has performed the arduous
task of translating the text from Swedish.

KTH, Stockholm in February 2023

Jan Scheffel and Per Brunsell

I
II
Contents

1 FUSION and the ENERGY PROBLEMS 3

1.1 The energy situation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Global energy consumption . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Energy and environment . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Fossil fuels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.1 Carbon capture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Renewable energy sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.1 Solar energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.2 Wind power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.3 Wave energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.4 Hydropower . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.5 Bioenergy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3.6 Geothermal power . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Energy storage systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4.1 Batteries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4.2 Thermal energy storage . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.3 Mechanical energy storage . . . . . . . . . . . . . . . . . . . . . . . 17
1.4.4 Hydrogen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4.5 Pumped hydroelectric storage . . . . . . . . . . . . . . . . . . . . . 19
1.5 Nuclear power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5.1 Basic concepts in nuclear physics . . . . . . . . . . . . . . . . . . . 20
1.5.2 Nuclear fission process . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.5.3 Fission reactor technologies . . . . . . . . . . . . . . . . . . . . . . 25
1.5.4 Public concerns about fission reactors . . . . . . . . . . . . . . . . . 27
1.6 Fusion energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.6.1 Fusion reactions in the Sun . . . . . . . . . . . . . . . . . . . . . . 29
1.6.2 The promise of fusion energy . . . . . . . . . . . . . . . . . . . . . . 30
1.6.3 Possible ways to realize fusion energy . . . . . . . . . . . . . . . . . 31
1.6.4 Fusion reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.6.5 Fusion in a thermal plasma . . . . . . . . . . . . . . . . . . . . . . 35
1.6.6 Requirements for break-even . . . . . . . . . . . . . . . . . . . . . . 36
1.6.7 The Lawson criterion for a fusion reactor . . . . . . . . . . . . . . . 37
1.6.8 Ignition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.6.9 Fusion research today . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.6.10 Muon-catalyzed fusion∗ . . . . . . . . . . . . . . . . . . . . . . . . . 41

III
2 THE FUSION PLASMA 43
2.1 Parameters of a fusion plasma . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.2 Plasma models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.3 Particle dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.3.1 Homogeneous magnetic field . . . . . . . . . . . . . . . . . . . . . . 46
2.3.2 Guiding center drift . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.3.3 Vertical drift in a toroidal magnetic field . . . . . . . . . . . . . . . 51
2.4 The extension of the particle fields . . . . . . . . . . . . . . . . . . . . . . 52
2.5 Kinetic plasma model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.5.1 The Vlasov equation∗ . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.5.2 The Fokker-Planck equation∗ . . . . . . . . . . . . . . . . . . . . . 54
2.6 Coulomb collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.6.1 Mechanics of the Coulomb interaction . . . . . . . . . . . . . . . . . 55
2.6.2 Coulomb collision times . . . . . . . . . . . . . . . . . . . . . . . . 57
2.6.3 Plasma resistivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.7 Fluid models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.7.1 Derivation of the fluid model equations∗ . . . . . . . . . . . . . . . 60
2.7.2 Magnetohydrodynamic model (MHD) . . . . . . . . . . . . . . . . . 61
2.7.3 Advanced fluid models∗ . . . . . . . . . . . . . . . . . . . . . . . . . 63

3 EQUILIBRIUM 64
3.1 Cylindrical equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.2 Beta value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3 Toroidal equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3.1 Equilibrium vertical field . . . . . . . . . . . . . . . . . . . . . . . . 67
3.4 Magnetic flux surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4.1 Flux functions∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.4.2 Grad-Shafranov equation∗ . . . . . . . . . . . . . . . . . . . . . . . 73

4 INSTABILITIES AND WAVES 75

4.1 Stabilizing magnetic fields . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2 The safety factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.3 How can we analyze instabilities? . . . . . . . . . . . . . . . . . . . . . . . 78
4.4 Waves in an MHD plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4.1 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4.2 Alfvén waves∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5 Ideal MHD stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.5.1 The energy principle . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.5.2 Extended form of the energy principle∗ . . . . . . . . . . . . . . . . 88
4.5.3 Eigenmodes∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.6 Tokamak stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.6.1 Instability in early pinches . . . . . . . . . . . . . . . . . . . . . . . 89
4.6.2 External kink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.6.3 Internal kink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.6.4 Resistive instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.6.5 Magnetic islands . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.6.6 Pressure driven instabilities . . . . . . . . . . . . . . . . . . . . . . 100
4.6.7 Disruptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.6.8 Beta limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

IV
 4.6.9 Further instabilities influencing confinement∗ . . . . . . . . . . . . . 102

5 TRANSPORT 105
5.1 Classical transport theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.1.1 The random walk process . . . . . . . . . . . . . . . . . . . . . . . 106
5.1.2 Particle model for classical diffusion . . . . . . . . . . . . . . . . . . 107
5.1.3 Fluid model for classical diffusion . . . . . . . . . . . . . . . . . . . 108
5.1.4 Relation between confinement time and diffusion coefficient . . . . . 109
5.1.5 Classical heat conduction . . . . . . . . . . . . . . . . . . . . . . . . 110
5.1.6 Energy confinement time for ohmically heated plasma∗ . . . . . . . 111
5.2 Neoclassical transport theory . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.2.1 Guiding-centre orbits in a tokamak . . . . . . . . . . . . . . . . . . 112
5.2.2 Particle model for Pfirsch-Schlüter diffusion . . . . . . . . . . . . . 114
5.2.3 Trapped particles in a tokamak . . . . . . . . . . . . . . . . . . . . 115
5.2.4 Banana orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.2.5 Banana orbit diffusion . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.2.6 Summary of neoclassical transport . . . . . . . . . . . . . . . . . . 120
5.2.7 Fluid model for Pfirsch-Schlüter diffusion∗ . . . . . . . . . . . . . . 121
5.3 Experimentally measured transport . . . . . . . . . . . . . . . . . . . . . . 121
5.3.1 Confinement scaling . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.4 Models for anomalous transport . . . . . . . . . . . . . . . . . . . . . . . . 123
5.4.1 Particle flux from electrostatic fluctuations . . . . . . . . . . . . . . 124
5.4.2 Transport caused by drift waves . . . . . . . . . . . . . . . . . . . . 125
5.4.3 Stochastic magnetic field∗ . . . . . . . . . . . . . . . . . . . . . . . 127
5.4.4 Transport in a stochastic magnetic field∗ . . . . . . . . . . . . . . . 128

6 RADIATION, BOUNDARY LAYER, HEATING 131

6.1 Emission of electromagnetic radiation . . . . . . . . . . . . . . . . . . . . . 131
6.1.1 Bremsstrahlung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.1.2 Cyclotron radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.1.3 Line radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.2 The plasma boundary layer . . . . . . . . . . . . . . . . . . . . . . . . . . 136
6.2.1 Recycling of neutral gas . . . . . . . . . . . . . . . . . . . . . . . . 137
6.2.2 Atomic and molecular processes in the boundary layer . . . . . . . 138
6.2.3 Desorption of impurities from the wall surface . . . . . . . . . . . . 138
6.2.4 Sputtering of wall material . . . . . . . . . . . . . . . . . . . . . . . 139
6.2.5 Arc discharges to the wall surface . . . . . . . . . . . . . . . . . . . 140
6.2.6 Evaporation of the wall material . . . . . . . . . . . . . . . . . . . . 141
6.2.7 Limiters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.2.8 Magnetic divertors . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.3 Ohmic heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.4 Neutral beam heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.4.1 Injection of energetic neutral beams . . . . . . . . . . . . . . . . . . 145
6.4.2 A neutral beam injection system . . . . . . . . . . . . . . . . . . . 148
6.5 Radio-frequency heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.5.1 Plasma dispersion relation . . . . . . . . . . . . . . . . . . . . . . . 149
6.5.2 Reflection and resonance . . . . . . . . . . . . . . . . . . . . . . . . 152
6.5.3 Absorption of waves in the plasma . . . . . . . . . . . . . . . . . . 153
6.5.4 Different RF heating methods . . . . . . . . . . . . . . . . . . . . . 154

V
7 PLASMA DIAGNOSTICS 156
7.1 Rogowski coil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.2 Interferometric measurement of electron density . . . . . . . . . . . . . . . 158
7.2.1 Principle of interferometry measurement . . . . . . . . . . . . . . . 159
7.2.2 A laser interferometer system . . . . . . . . . . . . . . . . . . . . . 160
7.3 Electron temperature measurement by Thomson scattering . . . . . . . . . 161
7.3.1 Principle of Thomson scattering . . . . . . . . . . . . . . . . . . . . 161
7.3.2 Thomson scattering diagnostic system . . . . . . . . . . . . . . . . 164
7.4 Measurement of electron cyclotron radiation . . . . . . . . . . . . . . . . . 165
7.5 Measurement of bremsstrahlung spectrum . . . . . . . . . . . . . . . . . . 167
7.6 Neutral particle analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.6.1 Neutral particle analyser system . . . . . . . . . . . . . . . . . . . . 170

8 THE TOKAMAK AND THE ALTERNATIVES 172

8.1 Tokamak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.1.1 ITER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.2 Alternative configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
8.2.1 High-field tokamak . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
8.2.2 Spherical tokamak . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
8.2.3 Stellarator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
8.2.4 Reversed-field pinch . . . . . . . . . . . . . . . . . . . . . . . . . . 177
8.3 Inertial confinement fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
8.4 The fusion-fission hybrid reactor∗ . . . . . . . . . . . . . . . . . . . . . . . 182

9 THE FUSION REACTOR 183

9.1 Reactor components and function . . . . . . . . . . . . . . . . . . . . . . . 185
9.1.1 Divertor / limiter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
9.1.2 First wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
9.1.3 Breeding blanket . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
9.1.4 Shield . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
9.1.5 Magnetic field coils . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
9.1.6 Critical disturbances . . . . . . . . . . . . . . . . . . . . . . . . . . 189
9.2 Safety issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
9.2.1 Tritium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
9.3 Activated wall material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

A Useful formulas and constants 192

A.1 Physics constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
A.2 Energy and temperature units . . . . . . . . . . . . . . . . . . . . . . . . . 192
A.3 Thermal velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
A.4 Vector formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
A.5 Cylindrical coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
A.6 Electromagnetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

1
Figure 1: Professor Lyman Spitzer next to an early stellarator. (Figure credit: FUSION EXPO-
EUROfusion)

Prologue
The idea to copy the energy producing process of the sun arose already in the 1920s.
Researchers found that at very high temperatures, hydrogen nuclei in the solar core are
joined together to form heavier ions, releasing energy as they subsequently split into
particles with lower total mass. This process has become known as thermonuclear fusion.
In 1937, the exile German physicists Hans Bethe and Leo Szilard, discussed the possi-
bility of using deuterium instead of hydrogen. The German Houtermans, the Australian
Thonemann and the Englishman Thomson were thinking along similar lines. Shortly after
World War II, secret experimental programs were initiated both in England and in the
Soviet Union. The military potential of the fusion process was clear. However, the real
starting point for fusion energy research took place in 1951 when the dictator of Argentina,
Juan Peron, declared that a German-Argentinian physicist had demonstrated thermonu-
clear reactions under controlled conditions. A year later the claim was disproved but in
the USA, Lyman Spitzer of Princeton University was selected to lead a project to inves-
tigate the possibility of thermonuclear fusion and in the Soviet Union Lev Artsimovich
became leader of another team with the same mission. In figure 1, Spitzer is posing next
to the magnetic confinement device he invented and named stellarator.
When fusion research was finally declassified at the Geneva conference in 1958, it
turned out that the UK, the Soviet Union and the USA had been focusing their efforts
on different configurations. The UK had been experimenting with so called pinches,
which unfortunately proved unstable. The Russians had realized the strong stabilizing
effect of axial magnetic fields, a discovery that led them to the successful tokamak. The
American investigation of the stellarator did not lead to any significant achievements;
plasma diffusion losses exceeded classical predictions by far. When open conferences on
fusion now started, the researcher’s hopes of quickly overcoming the problems related
to controlled fusion were however dashed. Instabilities and energy losses were tough
problems, coming to dominate research for decades to come.

2
Chapter 1

FUSION and the ENERGY

PROBLEMS

In this chapter, we give the background and motivation to fusion research; the closely
related problems of global energy supply and climate change. Then follows a brief overview
of currently available energy sources and related technologies. We end the chapter with
an overview of fusion energy and current fusion research. This overview serves as an
introduction to the more in-depth presentation of the physics behind fusion energy, which
follows in the remaining chapters of the book.

1.1 The energy situation

1.1.1 Global energy consumption
A widely used source of information on the world’s energy consumption is the annual pub-
lication ”Statistical Review of World Energy” 1 published by the BP company (formerly
known as the British Petroleum company). Another important source of information on
the world’s demographic change, energy and environment is produced by ”Our World in
Data”2 , a project of the ”Global Change Data Lab”, a non-profit organization based in
the United Kingdom, in which the researchers at the University of Oxford are the scien-
tific contributors to the website content. The data on the global energy situation in this
book chapter is largely based on these sources.
Today (year 2021) the global yearly energy consumption is 176 000 TWh, according to
data from the BP Statistical Review of World Energy. This number may not mean much
to us. We are more familiar with measuring the energy in kWh (kilowatt-hour). One
kWh is 103 Wh (watt-hour) and one TWh is 1012 Wh or one billion kWh. We can obtain
some intuitive feeling for this number by putting it in relation to the power production
of one of the largest power plants in Sweden, the Forsmark nuclear power plant, in which
three reactors produce at total of 20-25 TWh in a normal year 3 . The world’s energy
production thus corresponds to that of around 8000 Forsmark power plants.
Primary energy refers to energy in the raw form, before it has been converted into
other forms such as electricity, heat or transport fuels. It is coal, oil or gas before we burn
them, or solar or wind energy before we convert them to electricity. There are two ways
1
https://www.bp.com/en/global/corporate/energy-economics/statistical-review-of-world-energy.html
2
https://ourworldindata.org/
3
https://group.vattenfall.com/se/var-verksamhet/forsmark/forsmark-nuclear-power-plant

3
Figure 1.1: Global primary energy consumption by source calculated with the substitution
method, re-calculating the non-fossil fuel inputs to take into account the fossil fuel energy
conversion inefficiency. Data provided by the BP Statistical Review of World Energy (Figure
credit: Our World in Data)

in which primary energy can be calculated, the ”direct” method and the ”substitution”
method. The direct method does not take into account the energy lost in the conversion
of fossil fuels to usable energy. The substitution method attempts to correct for this loss.
When we burn fossil fuel in a thermal power plant most of the energy we put in is lost in
the form of heat. Most fossil fuel power plants have an efficiency of 33 % to 40 %. When
we measure electricity generation from renewables, we are measuring the direct output,
with no losses.
In order to compare the different energy sources in a way that take into account the
low efficiency of fossil fuel energy conversion, the primary energy consumption in the BP
Statistical Review is calculated using the substitution method, in which the renewable,
non-fossil fuel primary energy consumption values are converted to primary energy values
by dividing them with an efficiency factor. The efficiency factor used is increasing from
36 % to 40 % over time, reflecting the gradual improvement of the fossil fuel power plant
efficiency. The direct method would clearly show a larger share of the fossil fuels in
the world consumption. However, the substitution method is considered to give a more
accurate understanding of how low-carbon energy sources are competing with fossil fuels.
The primary world energy consumption has increased roughly linearly from the 1950
level of 28 000 TWh to reach the current level of 176 000 TWh, according to BP Statistical
Review of World Energy data shown in figure 1.1. During the period from 1950 to 2021
the global energy consumption has grown by a factor of 6.3. How can we explain that large
increase? First we consider the world’s population growth. The world population in 1950
was 2.5 billion, which has grown to 8.0 billion in 2022, an increase by a factor of 3.2. This
population growth factor is about half of that of the energy consumption growth, which

4
indicates that the average energy consumption per capita has approximately doubled over
this period.
The trends of the energy consumption in various regions in the world are quite different,
as can be seen in figure 1.2. The energy consumption in the Asia-Pacific region, has had
a strong increase over the last 50 years, and the pace of the increase is even faster in the
last 20 years. The trend in the western, developed parts of the world such as Europe and
North America, on the other hand is flat, or slightly decreasing over the same period.
From the figure, it is also clear that the energy consumption in the Asia-Pacific region
surpassed that in Europe in the 1990’s, and that in North America a few years later,
and now in 2021 is larger that the two regions (Europe and North America) combined.
Clearly, the Asia-Pacific region is largely responsible for the overall increasing trend of
the world’s energy consumption. This is of course closely linked to the population growth
in the region, but also to the rapid industrial and economical development, which results
in an increase of the energy consumption per capita. The current main sources of energy

Figure 1.2: Trends of energy consumption in different regions of the world. Data from BP
Statistical Review of World Energy. (Figure credit: Our World in Data)

globally are the fossil fuels:

• coal
• oil
• natural gas
The negative effects on the environment of burning fossil fuels, and the limited resources
of the fossil fuels necessitates a shift of the energy system to non-fossil energy sources:
• solar energy
• wind power
• wave power
• hydropower
• bioenergy

5
 • geothermal energy
• nuclear power
These are the major alternatives available with the technological knowledge of today. It
is among these limited options we must seek the solution to the energy problems for the
world’s increasing population with higher and higher demand for energy. We must find a
long term solution for energy production, which provides the possibility of a society with
clean environment, stable climate and without the risk for energy related disasters. On
the long term, in the second half of this century, new nuclear technologies such as fusion
energy, is expected to contribute to the energy production and be a part of the mix of
sources in the energy system.

1.1.2 Energy and environment

Let us consider the climatological effects of the use of fossil fuels. Climate change is
real and human activities are the main cause. The concentration of greenhouse gases
in the earth’s atmosphere is directly linked to the average global temperature on Earth.
The most abundant greenhouse gas, accounting for about two-thirds of greenhouse gases,
carbon dioxide (CO2 ), is largely the product of burning fossil fuels. In the last century,
an increase of the atmosphere’s average temperature of about 0.85 ◦ C has been measured.
This correlates to the increase in the level of carbon dioxide observed in recent years.
Seen over a very long time period of 800 000 years there have been consistent fluctuations
in the atmospheric CO2 level typically between 200 and 280 ppm (parts-per-million),
caused be the changes in the Earth’s orbit around the sun. Over this long time period,
the atmospheric concentration of CO2 has never exceeded 300 ppm. This situation has
changed with the industrial revolution and the rise in emission caused by the burning of
fossil fuel. The concentration today is well over 400 ppm. The increase in the atmospheric
CO2 level correlates well with the increase in CO2 emissions seen over the last century,
figure 1.3. The figure shows the global CO2 emission from burning of fossil fuels coal,
oil and gas, from the controlled flaring (burning) of natural gas in oil extraction, and
industry processes such as cement production. The total emission has increased from
around 5 billion tonnes in the year 1950 to around 35 billion tonnes in 2021, and increase
by factor of 7. This factor is similar to the growth factor of 6.3 for the global energy
consumption over the same period.
The International Energy Agency (IEA) publishes every year an objective analysis and
projection of the world’s energy market, the annual ”World Energy Outlook” (WEO)4 .
The projections are generated by the IEA’s Global Energy and Climate (GEC) Model,
a large-scale simulation tool, developed by IEA, which builds on over 20 years of mod-
elling experience. The IEA uses a scenario-based approach to highlight key choices and
consequences that lie ahead, to illustrate how the course of the energy system might be
affected by energy policies adopted by governments around the world.
Reducing global carbon dioxide emission to zero by 2050 is in this outlook considered
necessary in order to limit the long term increase in average global temperature to 1.5 ◦ C
by the end of the century. This requires a complete transformation of how to produce,
transport and consume energy. The pledges that have been made to date by a number of
countries to achieve net-zero emission is consistent with a temperature rise around 2.1 ◦ C
by the year 2100. The governments need to significantly strengthen their energy and
climate policies in order to reach the 1.5 ◦ C goal.
4
https://www.iea.org/topics/world-energy-outlook

6
Figure 1.3: Global CO2 emissions by fuel or industry type. Data are shown for emissions due
to the fossil fuels coal, oil, and natural gas, the controlled flaring process of natural gas in oil
extraction, and industry processes such as cement production. Data from Our World in Data
Global Carbon Project (Figure credit: Our World in Data)

In the IEA scenario ”Net Zero by 2050”, the full range of energy issues, including
oil, gas and coal supply and demand, renewable energy technologies, electricity markets,
energy efficiency are examined, with the aim to provide an approach to the goal of net
zero carbon dioxide emission by 2050. According to this scenario, in the year 2050
• the world economy is more than twice as big today,
• the population has increased with 2 billion more people,
• but the global energy demand is 8 % smaller than today, and
• instead of fossil fuels, the energy sector is based largely on renewable energy,
• two-thirds of total energy supply is from wind power, solar energy, bioenergy,
geothermal energy and hydro power,
• solar becomes the largest source, accounting for 20 % of energy supplies,
• solar photo-voltaic capacity increases by a factor of 20,
• wind power increases by a factor of 10 from today,
• use of fossil fuels declines from almost 80 % of total energy supply today to slightly
over 20 %, and
• fossil fuels that remain are used in facilities fitted with carbon capture,
• electricity accounts for almost 50 % of total energy consumption,
• electricity plays a key role in all sectors - from transport and buildings to industry
- and is essential to produce low-emission fuels such as hydrogen,
• total electricity generation increases by a factor of 2.5, and
• almost 90 % of electricity generation comes from renewable sources, with wind and
solar photo-voltaic together accounting for nearly 70 %, and
• the remaining 10 % of electricity comes largely from nuclear power.

7
Table 1.1: Fossil fuel consumption and reserves in the year 2020. Data from BP Statistical
Review of World Energy.
Fuel Yearly consumption (TWh) Fuel reserves left (years)
Coal 44 000 139
Oil 51 000 54
Gas 39 000 49

1.2 Fossil fuels

For millions of years, solar energy has been concentrated through photosynthesis in de-
composing flora and fauna, and it is this fossil energy we are now rapidly consuming.
A fossil fuel is a hydrocarbon-containing material formed naturally in the Earth’s crust
from the remains of dead plants and animals over long time periods of millions of years.
Fossil fuels may be burned to provide heat or to power internal combustion engines, or to
generate electricity. The large scale burning of fossil fuels is the main source of greenhouse
gas emissions. Over 80 % of the carbon dioxide is generated by the burning of fossil fuels.
Another environmental problem is the local air pollution due to combustion of fossil fuels
in the form of sulphuric and nitric acids, which fall to Earth in the form of acid rain. The
fossil fuels are dominantly of three different kinds; coal, oil and natural gas. World-wide
there are in the year 2022 over 2400 coal-fired power stations. Total capacity world-wide
is over 2000 GW, and they correspond to about 20 % of world-wide CO2 emissions, the
single largest source of climate change. When comparing the emission with the energy
production from different fossil fuels, it is apparent that coal produces relatively more
emission than natural gas. The use of natural gas instead of coal and oil would improve
the situation. There has recently been shift in Europe from use of coal fuel to natural
gas, and this shift has contributed to the reduction of CO2 emission from fossil fuels in
this region during the last 20 years.
Another problem connected with the fossil fuels is the limited remaining reserves. The
fossil fuel reserves will diminish rapidly. The supply of coal is expected to last more than
a hundred years, but gas and oil will only last for a few decades more. The world’s current
fossil fuel consumption and reserves as estimated in the BP Statistical Review of World
Energy are shown in the table 1.1. Together, the fossil fuel consumption corresponds to a
total of 134 000 TWh /year, equivalent to a share of the total global energy consumption
of 76 %. The fossil fuel reserves left are reported as the reserves-to-production (R/P)
ratio, which measures the numbers of years of production left based on known reserves
and present (year 2020) annual production levels. These values can change with time
based on the discovery of new reserves, and changes in annual production.
In the IEA Net Zero Emission by 2050 Scenario (NZE) the use of fossil fuels would
be dramatically reduced by 2050 and the CO2 emission due to burning of fossil fuels
and industrial processes, which in 2020 us 34 billion tonnes/year is strongly reduced to 8
billion tonnes/year in 2050.

1.2.1 Carbon capture

Carbon capture and storage (CCS) refers to technologies that can capture carbon dioxide
(CO2 ) from large point sources, such as power plants or industrial facilities that use either
fossil fuels or biomass as fuel. The CO2 can also be captured directly from the atmosphere.
If the captured CO2 is not utilized in an industrial process, the carbon dioxide gas can

8
be compressed and transported to storage sites such as deep geological formations. These
formations can be petroleum or natural gas fields or deep underground water-filled porous
rock reservoirs known as saline aquifers. A version of carbon capture that is economically
attractive is CO2 Enhanced Oil Recovery (CO2 -EOR). In this method, CO2 is injected
into petroleum fields that are in their final stage of production, in order to extend the
life-time of the field by raising the underground reservoir pressure, which then enables
more petroleum to be extracted from the field.
There are currently around 35 commercial facilities applying carbon capture to in-
dustrial processes, fuel transformation, and power generation with a total annual capture
capacity of 44 million tonnes of CO2 . One example is the Norwegian Sleipner natural
gas field in the North Sea5 . Since the mid 1990s, one million tonnes of carbon dioxide is
annually stored in an underground reservoir. The carbon dioxide captured in the Sleipner
field is the result of purification of the natural gas extracted from the field. The captured
carbon dioxide is compressed to 65 bar and injected into the reservoir at some 1000 meters
depth below the ocean floor in the Utsira aquifer. The storage is considered safe and the
potential is huge. It is estimated that the Utsira aquifer, which is 200-300 meters thick
and covers an area of 26 000 square kilometers, has storage potential corresponding to
the carbon dioxide emission in waste gas from all European power plants for hundreds of
years.
The captured CO2 can also be used in industrial processes. World wide, around 230
million tonnes of CO2 are currently used each year, of which 130 million tonnes is used
in fertilizer production and 90 million tonnes is used in enhanced oil recovery.
In the IEA Net Zero Emission by 2050 scenario, the world-wide CCS capacity is
envisioned to increase from today’s capacity of 44 million tonnes to 8 billion tonnes in
2050. In the NZE scenario, the use of fossil fuels is dramatically reduced by 2050 so that
this capture capacity would be enough to capture largely all emission from burning of
fossil fuel and thus achieve the goal of net zero CO2 emission.

1.3 Renewable energy sources

Renewable energy stems indirectly from the continuous solar radiation. Among the re-
newable energy technologies for electricity generation we find photovoltaic solar power,
wind power, wave power, hydro power, and biomass. Renewable technologies for heat gen-
eration include geothermal energy, although being of another origin, and solar thermal
technologies.
Fossil fuels dominate the global power supply because until very recently electricity
from fossil fuels was the cheapest. This has changed dramatically. To make compar-
isons on a consistent basis, energy prices are expressed in ”levelized cost-of-electricity”
(LCOE)6 . LCOE captures the cost of building the power plant itself as well as the ongoing
costs for fuel and operating the power plant over its lifetime. However, the LCOE does
not take into account costs and benefits at the energy system level; price reductions due
to low-carbon generation or higher systemic costs when storage or backup is needed for
variable renewable sources. The trend of the LCOE for renewables are shown in figure
1.4, comparing the cost to that for fossil fuels. If we want to transition to renewables, it
is their price relative to fossil fuels that matter. It is evident that within the last decade,
renewables such as solar and wind have seen a dramatic price decline, and are now in the
5
https://www.equinor.com/energy/sleipner
6
https://ourworldindata.org/cheap-renewables-growth

9
Figure 1.4: Levelized cost-of-electricity (LCOE) for different renewable energy technolo-
gies. Levelized cost estimates the average cost per unit of energy generated across the
life-time of a new power plant. Source: International Renewable Energy Agency (IRENA),
Figure credit: Our world in data

range 0.05 - 0.18 USD/kWh, a price range characteristic of fossil fuels. In most places,
power from new renewable power plants is now cheaper than from new fossil fuel plants.
A striking example is the price per Watt for solar modules, which has declined by
99.6 % from 106 USD/Watt in the mid 1970’s to 0.38 USD/Watt in 2020. The fundamental
driver of this change is that renewable energy technologies follow learning curves, which
means that with each doubling of the cumulative installed capacity, their price declines
by some fraction. For example, the price of solar PV modules has dropped by 20 % for
each doubling of the installed capacity. This price reduction factor is called the learning
rate, which then for solar PV modules is 20 %. Wind power also follows a learning curve,
which for on-shore wind power is 23 %. Off-shore wind power has a learning rate of 10 %,
and is still relatively expensive.

1.3.1 Solar energy

The sunlight, or the solar irradiation incident on Earth amounts to a power of about
178 000 TW in total or 1.3 kW/m2 . This is more than 10 000 times needed for our
total consumption of energy. There are three available technologies for extracting energy
from sunlight. Photovoltaics (solar cells) and concentrating solar power plants generate
electricity, whereas solar thermal collectors are used for heating and cooling. Solar cell
prices have dropped dramatically the last few years, and the installed capacity increases
strongly. Most of the world’s population lives in areas with solar irradiation of 150-300
W/m2 . Commercial photovoltaic panels convert sunlight to electricity with conversion
efficiency of 15-20 %. Large-scale solar farms require unused land areas, but small instal-

10
lations can be integrated into buildings in urban areas. Solar energy is a variable energy
source, which preferably is combined with energy storage systems or hydrogen production.
Concentrating solar power (CSP) plants, using mirrors to track the path of the sun,
concentrate the solar rays to a solar collector where a heat transfer fluid circulates. The
heat can be used for a variety of applications: electricity generation, water desalination,
or chemical production. Heat can be stored in molten salt during day for electricity
production at night-time. The Noor power station in Morocco is in 2021 the world’s
largest concentrated solar power plant with more than half a million mirrors distributed
over the 25 km2 site area.7 The plant uses molten salt storage to produce electricity
at night. Total capacity is 510 MW electricity generation, and annual production is 1.5
TWh.
Solar photovoltaic (PV) cells are made from semiconductor materials that eject elec-
trons when light strikes the surface, producing an electric current. PV cells are small,
and produce a few watts of direct current (DC) electricity. The cells are combined into
PV modules that typically comprise a rectangular grid of 60 to 72 cells.
World-wide the solar PV generation increased by a record 179 TWh ( 22 %) in 2021, to
reach a total of 1000 TWh per year. It is the second largest growing renewable technology
after wind power in 2021. In order to follow the IEA Net Zero Emission by 2050 Scenario,
the annual growth of solar PV generation by 25 % is needed in the period 2022-2030,
and the total solar PV generation in 2030 should reach 7400 TWh. Utility-scale solar
PV is the least costly option for new electricity generation in a majority of countries.
Distributed, small scale solar PV, such as rooftop solar on buildings is expected to grow
faster as a result of higher electricity prices.

1.3.2 Wind power

About 0.25 % of the total solar energy incident on Earth is converted to wind motion.
Wind is a variable energy source, and wind power generation is preferably combined with
energy storage systems, or hydrogen production. The maximum wind turbine power, for
an idealized situation, where all kinetic energy in the wind is utilized follows a simple
formula 8 . Wind energy is the kinetic energy of the moving air. The kinetic energy of a
mass m with the velocity v is
1
E = mv 2 (1.1)
2
The air mass m passing the turbine rotor is determined from the air density ρ and volume
V as m = ρV . Consider a small time interval ∆t in which the air particles travels a
distance s = v∆t. Multiplying the distance with the area swept over by the rotating
rotor blades A = πr2 , where r is the rotor radius, the volume of air passing the rotor is
V = Av∆t. The maximum turbine power is the kinetic energy of the air passing the rotor
divided by the time interval ∆t
E 1 π
P = = ρAv 3 = ρd2 v 3 (1.2)
∆t 2 8
Here, P is the power in Watt, ρ is the air density, typically 1.225 kg/m3 at sea level, d
is the rotor diameter in m, and v is the wind speed in m/s. Clearly, the power increases
quadratically with the rotor diameter, explaining the trend of increasing turbine sizes.
7
https://worldsteel.org/steel-stories/infrastructure/worlds-largest-solar-power-plant-delivers-24-hour-
energy/
8
https://home.uni-leipzig.de/energy/energy-fundamentals/15.htm

11
Table 1.2: Size evolution of wind turbines over time. Source: Environmental Geology by
Steve Earle, 2021, Thompson Rivers University
Year Maximum power (MW) Rotor diameter (m)
1990 0.5 40
2000 2 80
2010 7 126
2020 14 200

Note the cubic dependence on wind speed, which shows the importance of locating wind
power plants to areas with favourable wind conditions. Wind power technology has devel-
oped significantly during recent years, with bigger and bigger wind power turbines being
constructed, see table 1.2.
In 2021, wind electricity generation increased by a record 273 TWh (17 %), to reach
a total production of almost 1 900 TWh per year. This is the highest annual growth
among all renewable power technologies. In order to get on track with the Net Zero
Emissions by 2050 scenario, which has 7 900 TWh annual wind power generation in 2030,
it is necessary to raise average annual capacity additions to 250 GW, more than double
the 2020 growth. In the IEA analysis of the technical potential of offshore wind, the
study showed that close-to-shore wind sites globally could provide almost 36 000 TWh of
electricity per year, which is close to the global electricity demand projected for 2040.
The cost-of-electricity (COE) for on-shore wind power is at reasonable levels, with low
investment costs as compared to several other energy sources.
Currently, Europe’s largest wind power installation is the Markbygden wind farm
located in northern Sweden 9 . The wind farm is comprised of 329 wind turbines in
operation with a total capacity of 1.12 GW (year 2021), and 307 wind turbines under
construction to be completed in 2022-23 with a total capacity of 1.44 GW. Once completed
the total planned capacity is 4 GW and production is estimated to 8-12 TWh per year.

1.3.3 Wave energy

The lunar gravity and the weather’s influence on ocean waves generate vast amounts of
energy. Wave energy is the capture of the kinetic energy in the waves at the sea surface
for electricity generation. Wave energy is typically produced by floating turbine platforms
or buoys that rise and fall with the wave motion, so called wind-to-electricity converter
(WEC) devices. The theoretical potential for wave energy is huge, and the total wave
power incident on the ocean-facing coastlines of the world has been estimated to 2 TW
10
. The extractable wind power with an array of WECs is estimated to be around 100
GW.
Wave power farms are silent and do not disturb the view from the shore, since they
are mainly situated below the surface. Wave farms can complement off-shore wind farms,
sharing electrical infrastructure.
Wave energy is still in an early stage of development, with a number of prototype
systems being tested. The main challenge with this technology is to design enough robust
WEC devices that can withstand very harsh weather conditions at sea. The Swedish
company CorPower Ocean has developed a WEC device that is designed to fulfil this
9
https://svevind.se/en/projects/
10
https://www.sciencedirect.com/science/article/pii/S0960148112001310

12
requirement.11 CorPower’s WEC devices generate up to 300 kW each, and are planned to
be combined into clusters, with 10-20 MW capacity. These clusters are in turn planned to
be laid out side-by-side to form larger wave farms that can have a utility-scale production
capacity. Prototype tests of the WEC devices will start in 2022 off the coast of Portugal.

1.3.4 Hydropower
Hydropower, or hydroelectric power, is one of the oldest and largest resources of renewable
energy. It uses the natural flow of moving water to generate electricity. Water reservoirs
next to the plant stores the energy. Hydropower technologies generate power by using
the elevation difference, created by the reservoir dam. It generates electricity in a turbine
driven by water flowing in on one side of the dam, and flowing out on the other. The
power generated from a hydropower plant is proportional to the water flow and the height,
and can be calculated by a simple formula 12 . The potential energy of the water at the
top of the dam is
W = mgh (1.3)
where m is the mass of water in kg, g = 9.81 m/s2 is the acceleration due to gravity,
and h is the height difference between the two sides of the dam in m. The mass of the
water is obtained as m = ρV , where ρ = 1000 kg/m3 is the density of water, and V is
the volume of the water passing the turbine. Consider a small time interval ∆t in which
the water flows though the turbine. The volume of water passing the turbine is given by
the volumetric flow rate V̇ as V = V̇ ∆t. The maximum turbine power is obtained as
the kinetic energy of the water passing the turbine divided by the time interval ∆t. This
kinetic energy is equivalent to the corresponding change in the potential energy of the
water, and can be written as
W
P = = ρV̇ gh (1.4)
∆t
where P is the power in Watt. For example, a reservoir dam height of 100 meter and a
volumetric flow rate of 100 m3 /s produces 98 MW, which must be multiplied with the
efficiency of the plant, generally 80–90 %, when determining the electrical power.
The benefits of hydropower have been recognized and harnessed for thousands of years,
and it is the leading renewable energy source world wide. There are virtually no emissions
during the production phase, and it can be used both as baseload power and as balancing
power, compensating for variability in other renewable energy sources.
Currently (2022), the world’s largest hydropower plant is the Three-Gorges Dam in
China, which has 32 turbines and a total capacity of 22.5 GW of electricity power. The
annual production is 112 TWh.13
Hydropower is still the largest renewable energy source - excluding traditional biomass
- and accounts for more than 60 % of renewable electricity generation. It is expected to
remain the world’s largest source of renewable electricity generation. The global annual
hydropower generation in 2021 is 4 300 TWh. Global hydropower capacity (2021) is
1 360 GW. In the IEA Net Zero Emission by 2050 scenario, hydropower maintains an
average annual growth rate of about 3 % in 2022-2030, to provide approximately 5 700
TWh of electricity in 2030. Pumped storage hydropower plants represents 30 % of the
net hydropower additions in this forecast.
11
https://corpowerocean.com/wave-energy-technology/
12
https://en.wikipedia.org/wiki/Hydropower
13
https://www.power-technology.com/projects/gorges/

13
 In summary, hydropower is an economical, continuous and reliable source of energy,
but in many countries, it has already been developed as far as concerns for environment
and population allows. The largest remaining potentials for hydropower expansion are
found in Africa, South America and Asia.

1.3.5 Bioenergy
Bioenergy is created in plants when the photosynthesis transforms water, carbon dioxide
and minerals to biomass using energy from sunlight. Bioenergy does not contribute much
to the greenhouse effect, since the carbon dioxide released during combustion was earlier
captured by the plants themselves. Inevitably, transports for harvesting and fertilizing as
well as processing require energy, which in turn should be sustainably produced.
Traditional use of biomass - burning of charcoal, organic wastes, and crop residues
- was an important energy source for a long period in human history. It remains an
important source in the developing countries today. However, high-quality estimates of
energy consumption from these sources is difficult to find. The BP Statistical Review of
World Energy only publishes data on commercially-traded energy, so traditional biomass
is not included. However, modern biofuels are included in the data. Bioethanol and
biodiesel - fuel made from crops such as corn, sugarcane, hemp and cassava, are now a
key transport fuel in many countries.
Bioenergy is a leading energy source in Sweden, and biomass has a dominant position
in the Swedish heat market, to a large part as fuel in district heating. 14
During recent years, the global biofuel energy consumption has increased from around
100 TWh in 1990, to 1000 Twh in 2021, a ten-fold growth. In the IEA Net Zero Emission
by 2050 scenario, an increase in the global bioenergy supply by 60 % is required from
2020 to 2030. However, the increase must not lead to negative effects, such as expan-
sion of cropland for bioenergy or conversion of existing forested land into bioenergy crop
production.

1.3.6 Geothermal power

Geothermal power uses the heat which is generated by radioactive decay and stored inside
the crust of the Earth. This internal thermal energy of the Earth flows to the surface at a
rate of 44 TW, more than double the power needed for the world’s energy consumption.
The upward heat flux is on average 0.1 MW/km2 .15
This source of energy has been used directly for thousands of years in hot springs
and geysers, but it is unfortunately only available in this form in a few locations. The
geothermal temperature gradient in the Earth’s crust is 25-30◦ C per kilometer depth.
Deep geothermal energy is heat energy in the form of hot fluid or steam from reservoirs
at a depth of the order of 1000-2000 meters below the Earth’s surface. In a combined
heat and power plant (CHP), a mixture of steam and fluid is obtained from the wells and
separated into the two phases in the plant:
• The steam is used in a turbine to generate electricity
• The fluid is used in a heat exchanger to produce hot water for heating
The Nesjavellir geothermal plant in Iceland started operation in 1990. 16 The plant
uses 25 boreholes with depths of 1000-2200 meters, and produces 300 MW of geothermal
14
https://www.svebio.se/en/about-bioenergy/
15
https://en.wikipedia.org/wiki/Geothermal energy
16
https://nat.is/nesjavellir-geothermal-power-plant/

14
energy, of which 1600 litre/second hot water for district heating and 120 MW of electricity.
Ground source heat pumps are widely used in many countries to provide renewable
local heating and cooling of buildings. In these systems fluid circulates in a pipe in the
ground a modest depths below the Earth’s surface. The fluid absorbs heat from the
warmer soil, rock or ground water around it. A heat exchanger transfers the heat to
the building heating/cooling system, and the fluid then returns to the ground at a lower
temperature.
In 2021, the global geothermal power production was nearly 100 TWh.

1.4 Energy storage systems

The practical cost for renewable energy systems increases substantially when considering
the need for energy storage. Whereas hydropower is continuously controllable, solar and
wind energy are variable and may not be directly accessible when it is needed. Energy
storage systems have become widely accepted as efficient ways to complement variable
renewable energy sources, such as solar and wind. Different solutions for energy storage
exist 17 :
• batteries
• thermal energy storage
• mechanical storage
• hydrogen
• pumped hydroelectric storage

Grid-scale storage plays an important role in the IEA Net Zero Emission by 2050 Sce-
nario, providing system services for grid stability and long term energy storage. Pumped
hydroelectric storage is the most widely used storage technology and it has significant
potential in several regions. Batteries are the most scalable type of grid-scale storage and
has seen a strong growth in recent years. Hydrogen is an emerging technology that has
potential for long term seasonal storage of renewable energy. Thermal and mechanical
energy storage play a relatively small role in current power systems.

1.4.1 Batteries
Battery energy storage systems (BESS) are one of the fastest growing technologies in the
sustainable energy industry. On its most basic level, a battery is a device consisting of one
or more electrochemical cells that convert stored chemical energy into electrical energy.
Each cell contains a positive terminal, or cathode, and a negative terminal, or anode.
Electrolytes allow ions to move between electrodes and terminals, which allows current to
flow out of the battery. Advances in technology and materials have greatly increased the
reliability, output, and density of modern battery systems, and economics of scale have
dramatically reduced the cost.
• Lithium-ion (Li-Ion) batteries. Many companies are developing larger-format
cells for use in energy-storage applications. Many also expect there be synergies
with the electric vehicles (EV) powered by Li-ion batteries. Li-ion batteries have
been deployed in multi-megawatt batteries for the grid services. Lithium-ion refers
to electrochemical cell which is characterized by the transfer of lithium ions between
the electrodes during the charge and discharge reactions.
17
https://energystorage.org/why-energy-storage/technologies/

15
 • Lead-acid batteries. Lead batteries are the most extensively used rechargeable
battery technology in the world. They have an unrivalled track record for reliability,
which makes them the dominant battery in terms of MWh of production. Lead
batteries are widely deployed to support PV installations both in commercial and
domestic premises. Lead batteries have lower cost than other chemistries.
• Redox flow batteries (RFB). Redox flow batteries uses chemical reduction and
oxidation reactions to store energy in liquid electrolyte solutions which flow through
the battery during charge and discharge. The RFB provides a separation of power
and energy. The system energy is stored in the volume of electrolyte, which can
be in the range of kWh to MWh. The power of the system is determined by the
stack of electrochemical cells. The amount of electrolyte flowing in the stack is
rarely more than a few percent of the total, and flow can easily be stopped during a
fault condition. This is in contrast with packaged batteries such as Li-ion and lead
batteries.
• Sodium-sulphur (NaS) batteries. The system operates at a high temperature
300-350 ◦ C. The active materials in a NaS battery is molten sulphur as the positive
electrode, and molten sodium as the negative. The electrodes are separated by a
solid ceramic, sodium alumina, which also serves as the electrolyte. NaS battery
technology has been demonstrated at over 190 sites in Japan. More than 270 MW
capacity for 6 hours of daily peak shaving have been installed. In Abu Dhabi, 15
NaS systems provide 108 MW/648 MWh. The efficiency is high, typically 89 %.
• Zinc-bromine (Zn/Br) flow batteries. The zinc-bromine battery is a hybrid
redox battery, because much of the total energy is stored by plating zinc metal as
a solid onto the anode plates in the electrochemical stack during charge. During
discharge, the zinc metal is oxidized and dissolved in the aqueous anolyte. The net
DC-DC efficiency of this battery is in the range 65-75 %. Integrated Zn/Br energy
storage systems have been tested up to 1 MW/3 MWh for utility-scale applications.

The Pillswood BESS in UK with power capacity of 98 MW and capacity to store 196 MWh
of energy is currently (2022) Europe’s biggest battery storage system. 18 It is based on
Li-ion battery units developed by Tesla. The storage system will be connected to Dogger
Bank, the world’s largest offshore wind farm in 2023, under development outside the
North East coast of England.
The capacity of grid-scale battery energy storage systems are currently far smaller
than pumped hydroelectric storage, but grid-scale batteries are projected to account for
the majority of storage growth worldwide. Batteries are typically employed for sub-hour,
hourly and daily balancing. Total installed gird scale battery storage capacity is 16 GW
in 2021. In the grid-scale battery technology mix, lithium-ion battery storage is the most
widely used, making up the majority of all new capacity installed. Besides lithium-ion
batteries, flow batteries emerge as a breakthrough technology for stationary storage. In
July 2022, the world’s largest vanadium redox flow battery was commissioned in China,
with a power capacity of 100 MW and a storage volume of 400 MWh.

1.4.2 Thermal energy storage

Thermal energy storage technologies allow us to temporarily reserve energy produced in
the form of heat or cold for use at a different time. Modern solar thermal power plants
18
https://www.harmonyenergy.co.uk/general/harmony-energy-income-trust-brings-europes-largest-
battery-energy-storage-system-online/

16
produce all of their energy when the sun is shining during the day. The excess energy
produced during peak sunlight is often stored in these facilities - in the form of molten salt
or other materials - and can be used into the evening to generate steam to drive a turbine
to produce electricity. Alternatively, a facility can use off-peak electricity rates which are
lower at night to produce ice, which can be incorporated into a building’s cooling system
to lower demand for energy during the day.
• Pumped Heat Electrical Storage (PHES). In Pumped Heat Electrical Storage,
electricity is used to drive a storage engine connected to two large thermal stores.
To store electricity, the electrical energy drives a heat pump, which pumps heat from
the ”cold store” to the ”hot store”, similar to a refrigerator. To recover energy, the
heat pump is reversed to become a heat engine. The engine takes heat from the hot
store, delivers waste heat to the cold store, and produces mechanical work. When
recovering electricity, the heat engine drives a generator. The expected ”round-trip”
efficiency is 75-80 %.
• Liquid Air Energy Storage (LAES). Liquid Air Energy Storage uses electricity
to cool air until it liquefies, stores the liquid ari in a tank, bring the liquid air back
to a gaseous state by exposure to ambient air or with waste heat from an industrial
process, and uses that gas to a turbine and generate electricity. Liquid Air Energy
Storage is sometimes referred to as Cryogenic Energy Storage (CES). LAES plants
can provide large-scale, long duration energy storage with 100 MW output.

1.4.3 Mechanical energy storage

Mechanical energy storage systems take advantage of kinetic or gravitational forces to
store the energy. While the physics of mechanical systems are often quite simple, the
technologies that enable the efficient use of these forces is advanced, requiring high-tech
materials, or cutting-edge control systems.
• Flywheel Energy Storage Systems (FESS). Flywheel energy storage systems
use electric energy input which is stored in the form of kinetic energy, in this case
the motion of a spinning mass, called a rotor. When power is needed, inertia allows
the rotor to continue spinning while the kinetic is converted to electricity
• Compressed Air Energy Storage (CAES). Compressed Air Energy Storage is a
way to store energy by compression of ambient air, or another gas, and store it under
pressure in an underground cavern or container. When electricity is required, the
pressurized air is heated and expanded in an expansion turbine driving a generator
for power production.

1.4.4 Hydrogen
Hydrogen is a versatile energy carrier, which can help tackle various critical energy chal-
lenges. Hydrogen can be produced from almost all energy resources, though today’s use
of hydrogen in oil refining and chemical production is covered from fossil fuels. Clean
hydrogen, being produced from renewables, nuclear or fossil fuels with carbon capture,
can help to decarbonise a range of sectors, including transport, chemicals, iron and steel
production, where it is difficult to reduce emissions.
Hydrogen can also support the integration of variable renewables in the energy system,
being one of the very few options of long term storage of electricity over days, weeks or
months. Clean hydrogen can be produced from water in electrolysers, stored, and later
generate electricity in fuel cells. The hydrogen storage system is depicted in figure 1.5.

17
Figure 1.5: Schematic of the hydrogen storage system. Electricity generated by wind or solar
is user to power the electrolyser which produces hydrogen gas. The hydrogen gas is compressed
and stored in a tank. The hydrogen can later be used in a fuel cell to produce electricity and
heat. The heat from the fuel cell is stored in a hot water tank, and used for heating of tap water.

Electrolysers. Electrolysers are a critical technology for the production of low-

emission hydrogen. Electrolysis of water, also known as electrochemical water splitting, is
the process of using electricity to decompose water into oxygen and hydrogen by electrol-
ysis. The principle is simple: A DC electrical power source is connected to two electrodes,
or plates that are placed in the water. A minimum voltage difference of 1.23 V needs to
be applied, but in practical conditions 1.5 V is required. Hydrogen appears at the cathode
where electrons enter the water, and oxygen appears at the anode. The efficiency of the
process is increased through the addition of an electrolyte, such as a salt, acid, or base.
In the electrolyser, the overall chemical reaction is

2H2 O → 2H2 + O2 (1.5)

As of 2022, commercial electrolyser require around 53 kWh of electric energy to pro-

duce 1 kg of hydrogen. The chemical energy stored in 1 kg of hydrogen is equivalent
to the heat released if the hydrogen is burnt as fuel, which is nominally 33.6 kWh.19 If
the hydrogen would be used as a fuel, then the efficiency of the process would be of the
order of 60 %. There are more efficient uses of the hydrogen, for example for electricity
generation in a fuel cell. The global capacity of electrolysers for hydrogen production
in 2022 is 1.4 GW, rapidly increasing. Global electrolyser capacity is expected to reach
60 GW/year in 2030, with over 700 GW capacity installed by 2030.

Use and storage. Hydrogen has a high energy content by weight, but not by volume.
The hydrogen density is 0.089 kg/m3 at the atmospheric pressure of 1 bar and room
temperature. The energy content of hydrogen is 33.6 kWh/kg, or around 3 kWh/m3 .
This is a rather low energy density, and hydrogen must be stored physically as a high
19
https://en.wikipedia.org/wiki/Electrolysis of water

18
pressure gas or as a liquid. When the hydrogen is compressed to 700 bar corresponding
to a density of 42 kg/m3 , the energy content is 1 400 kWh/m3 . This energy density is
compatible with fuel tanks of motor vehicles. Hydrogen has a wide flammability range
compared to other fuels, corresponding to hydrogen-to-air ratios between 4 % and 75 %.
Hydrogen is the smallest known molecule, and is prone to leakage. In a confined space,
leaking hydrogen can accumulate and reach flammable concentrations. Proper ventilation
and the use of detection sensors can mitigate this hazard. Several car manufacturers
today use compressed hydrogen tanks which are capable of 350 or 700 bars, depending
on the automotive type. Compressed hydrogen fuel tanks are now made of carbon fibre
composites or carbon fiber and metal alloys and composites. The inner liner of the tank is
a high-molecular weight polymer that serves as a hydrogen gas permeation barrier. The
application of hydrogen in the transport sector is
• fuel cell electrical vehicles
• hydrogen combustion engines
• hydrogen-methane mixtures for combustion engines

Fuel cells. A fuel cell is an electrochemical cell that converts chemical energy of the
hydrogen fuel and an oxidizing agent, such as oxygen in air, into electricity. The fuel cell
is composed of an anode, a cathode and an electrolyte membrane. Hydrogen is supplied
at the anode and air is supplied at the cathode. The hydrogen atoms are stripped of their
electron at the anode, and positively charged protons pass the membrane to the cathode,
and the electron flows as electric current in the circuit. The electrons combine with the
protons and the oxygen from the air at the cathode, to generate water and heat. The
overall chemical reaction is the reverse of that in the electrolyser:
2H2 + O2 → 2H2 O (1.6)
A typical fuel cell produces a voltage of 0.6-0.7 V. To deliver the desired power, the fuel
cells can be combined in series to yield higher voltage, and in parallel to allow higher
current in a fuel cell stack. Typical fuel cell efficiency is 40-60 % 20 Assuming 60 %
efficiency in the electrolyser electricity-to-hydrogen production and 50 % efficiency in the
fuel cell hydrogen-to-electricity conversion, the ”round-trip” efficiency of the hydrogen
storage is of the order of 30 %.

1.4.5 Pumped hydroelectric storage

A hydroelectric dam relies on water flowing through a turbine to create electricity. In
order to store energy for use at later time, there are a number of different plants that
use pumps to elevate water into a retained pool behind the dam, creating an on-demand
source that can be unleashed rapidly. Because of the immense scale achieved through
these applications, this is the most common type of grid-level energy storage installed
today. Pumped storage stations are unlike traditional hydroelectric stations in that they
are a net consumer of electricity, due to the hydraulic and electric losses in the cycle
of pumping from lower to upper reservoirs. However, these plants are typically highly
efficient, with ”round-trip” efficiency greater than 80 %.
Pumped hydroelectric storage is the most widely deployed grid-scale storage technol-
ogy today. Total installed power capacity is 160 GW in 2021 and global energy storage
capability was around 8 500 GWh in 2020, accounting for over 90 % of total global elec-
tricity storage. The world’s largest capacity is found in the United States. The majority
20
https://www.energy.gov/eere/fuelcells/articles/fuel-cells-fact-sheet

19
 Binding energy per nucleon (MeV)
8 84
4 56 Kr
He Ni U
54
7
Li Fe
6 6
Li

4
3
He
2 2
D

0
0 40 80 120 160 200 240
Atomic mass number (A)

Figure 1.6: The average nuclear binding energy per nucleon ε = Eb /A as a function of
atomic mass number.

of the plants in operation today are used to provide daily balancing. The world’s most
significant investment in new pumped hydroelectric storage capacity is in China where
currently (2022) 50 GW of capacity is under construction.

1.5 Nuclear power

Nuclear power contributes to the base load fossil-free electricity production. Currently
(2022) there are 422 nuclear power reactors in operation in 32 countries around the world.
New small modular reactors (SMR) are being developed, with advantages of lower invest-
ment costs and faster deployment. The global total current operational capacity of nuclear
power is 413 GW, and the annual global electricity production is 2700 TWh, which cor-
responds to a share of around 10 %.

1.5.1 Basic concepts in nuclear physics

Let us first refresh our memory about atoms and nuclear reactions. The nucleus of an atom
consists of protons and neutrons, known collectively as nucleons. A specific configuration
of nucleons is referred to as a nuclide. An element in the periodic table is characterized
primarily by the charge number (Z), which is equal to the number of positively charged
protons in the nucleus. The total number of nucleons in the nucleus is called the atomic
mass number (A). Elements with the same charge number, but different atomic mass
number are referred to as isotopes.
Nuclear binding energy
The nucleons in the nucleus are held together by the so called strong nuclear force. The
minimum work, or energy that has to be supplied to a nucleus in order to oppose this
force and disassemble the nucleus into free nucleons is called the binding energy. The
binding energy per nucleon ε = Eb /A is different for different elements in the periodic
table, and it varies with the atomic mass number as shown in figure 1.6. In Einstein’s
famous theory of relativity, the central equation is
E = mc2 . (1.7)

20
 Figure 1.7: Illustration of the concept of nuclear cross section

It equates energy and mass, and states that the energy E of a nucleon is equal to m,
the mass of the nucleon, times the square of c, the velocity of light. The energy of the
nucleus is thus the sum of the energies of the individual constituent nucleons calculated
with Einstein’s equation minus the binding energy Eb of the nucleus. Since the energy of
the nucleus is less than the sum of the constituent nucleon energies, it is equivalent to a
reduction in mass of the nucleus relative to the sum of the masses of the nucleons. The
difference is referred to as the mass defect.

Nuclear reactions
A nuclear reaction is a process in which two nuclei collide and produce one or more new
nuclides. The process can be viewed as occurring in two stages: In the first stage, energy
equivalent to the sum of the binding energies of the reacting nuclides is supplied, and the
nuclei are disassembled into free nucleons. In the second stage, the free nucleons assemble
into the final nuclides, and energy equivalent to the sum of the binding energies of the
product nuclides is recovered. The reaction must obey the law of energy conservation,
thus, the difference between the recovered and supplied energies the so called reaction
energy, must be released/supplied as kinetic energy. In the case reaction energy is released,
the reaction products born in the nuclear reaction receive high initial velocities, and this
kinetic energy is then transformed to heat in a nuclear reactor as nuclear energy.
From the nuclear binding energy curve in figure 1.6, it can be seen that the elements
with very low, or very high atomic mass number have the smallest binding energies, and
elements with medium atomic mass number (e.g. Fe, Ni, Cr) have the largest. Nuclear
reactions which transforms nuclides with small binding energies to nuclides with large
binding energies release energy. Clearly, nuclear energy can be released in nuclear reactions
of two types:
• splitting (fission) of a heavy nuclide
• merging (fusion) of two light nuclides

Reaction probability
The nuclear cross section σ is used to describe the probability of a nuclear reaction. The
cross section concept originates from the idea of a projectile hitting a target with a cross
section area σ, as depicted in figure 1.7. The larger the cross section area is, the higher
is the probability that the reaction will occur. Consider a thin layer of the target volume
with area A and thickness dx. The nuclei in the volume has density n, and thus the
number of nuclei in the layer is N = nAdx. Assume that the probability for a reaction
between an incident particle and a target nucleus is equal to the probability of the incident

21
particle colliding with a physical target having the cross section area σ. The probability
P of a reaction in the layer with thickness dx is then equal to the fraction of the layer
that is blocked for the incident particle:

P = N σ/A = nσdx (1.8)

It is here assumed that the layer is very thin so that P is less than unity. An incident par-
ticle moves an average distance λ between collisions in the target and the layer thickness
is a fraction of this distance, which is given by the probability as dx = P λ. The mean
free path is then expressed as
λ = dx/P = 1/nσ (1.9)
An incident particle with velocity v traverses the layer in a time interval dt = dx/v. The
probability for a collision in the time interval dt is then P = nσdt. The average collision
time interval in the target of an incident particle, the collision time is τ , and the layer
traversing time dt is a fraction of this time, which is given by the probability as dt = P τ .
The collision time is then expressed as

τ = dt/P = 1/nσv (1.10)

and the reaction frequency is

ν = 1/τ = nσv (1.11)
The nuclear cross section is measured in unit barn (b):

1 barn = 10−28 m2 (1.12)

1.5.2 Nuclear fission process

It turns out that the uranium isotope 235 92 U (Z=92, A=235) is fissile, meaning that it is
easy to split by an incident slow, low energy neutron. This isotope is utilized as fuel in the
fission reactor. Different fission reactions are possible, such as the following, which splits
uranium into barium and krypton, releasing three secondary neutrons and a total of 200
MeV of reaction energy, distributed mainly to the two nuclides (Ba and Kr) produced in
the fission reaction.
235 144
92 U +n→ 56 Ba +92
36 Kr + 3n + 200 MeV (1.13)

The nuclear energy released in the reaction is measured in unit ”MeV” or million
electron volts. One electron volt is equivalent to the work done on an electron with the
charge −q = e = 1.6 × 10−19 C as it is moved across a potential difference of one Volt.
The conversion from eV to Joule is simply 1 eV = 1.6 × 10−19 J. The energy 200 MeV
released in one fission reaction is thus equal to 200 × 1.6 × 10−13 = 3.2 × 10−11 J.
We can compare the energy released in nuclear reactions with that released in chemical
reactions as follows: The weight of one uranium atom is 235 × 1.67 × 10−27 = 3.92 × 10−25
kg. One kilogram of pure uranium-235 contains 2.54 × 1024 atoms, and if all undergo
fission reactions, the nuclear energy released is 2.54 × 1024 × 3.2 × 10−11 = 8.13 × 1013 J =
22.6 GWh. The chemical energy released in combustion of 1 kg of methane is about 15
kWh. The nuclear energy density is thus of the order of one million times higher than
that of chemical.

Fission chain reaction

In the fission of U-235 there are 2-3 secondary neutrons released. These neutrons are

22
Figure 1.8: Fission cross section for some nuclides in the uranium-plutonium fuel cycle in
unit barn as function of the incident neutron energy in MeV. 1 barn is equal to 10−28 m2

highly energetic particles, receiving on average an energy of 1.5 MeV when they are born.
Some of these secondary neutrons split other fissile nuclei, while others are captured in
collisions with non-fissile neutron-absorbing elements in the fuel, such as U-238. When the
reaction becomes critical, the number of neutrons produced from fission is larger than the
number lost to capture, and a chain reaction is sustained. In a power producing reactor,
neutron absorbing components are inserted, in order to keep the reactor in a marginally
critical state. By varying the amount of neutron absorbing material in the reactor core,
the neutron multiplication factor is adjusted to be close to one, and the reaction can be
maintained self-sustained in a steady state.

Fission cross sections

The likelihood of a neutron incident on a nucleus resulting in fission, the fission cross
section, is shown in figure 1.8 for different nuclides that appear in the so called uranium-
plutonium fuel cycle. The cross section as function of the incident neutron energy in unit
MeV is plotted. The unit of the cross section is barn (b). One barn is equal to 10−28 m2 .
Natural uranium contains a tiny 0.7 % of U-235, while the rest, or over 99.2 % consists of
the U-238 isotope. From the cross-sections, it is apparent that the fission of U-235 is much
more probable than fission of U-238 at neutron energies below 1 MeV. In spite of the small
fraction of fissile U-235 in the fuel, it is the more important isotope for sustaining the
chain reaction, also at high energies. In order to increase the number of fission reactions,
the uranium fuel is usually enriched, raising the U-235 content in the fuel to 3-5 %.
In order to achieve a sustained chain reaction, more abundant fission reactions are
needed. It is clear from figure 1.8 that the fission cross section at low neutron energy
is three order of magnitude larger than the cross section in the 1 MeV range. One
possibility to achieve more fission reactions is to slow down the energetic neutrons to
thermal energies, corresponding to energies that are a fraction of an eV. This is achieved
by adding a neutron moderator in the reactor core. Materials used for moderation need
specific properties. The moderator should slow down, or scatter the neutrons by elastic
collisions, but it should not absorb neutrons by inelastic collisions. Since the mass of

23
Figure 1.9: Cross sections for neutron capture in U-238 (blue) and neutron fission of Pu-
239 (red). Cross sections are shown in unit barn as function of the incident neutron energy
in eV. Cross sections are comparable for higher neutrons energies, which is essential for
raising the plutonium breeding ratio to the order of unity in the fast reactor.

the hydrogen atom is almost identical to the neutron mass it works well as a moderator.
Lighter elements are in general more efficient as moderators. Typical moderator materials
are light (regular) water (H2 O), heavy water (D2 O), and graphite (C). A reactor that is
using a moderator to slow down neutrons is referred to as a thermal, or slow-neutron
reactor.

Uranium-plutonium fission process

Nuclear fission is not the only process occurring in the uranium fuel. Neutrons incident
on U-238 can be captured, causing the nuclide to form the heavier isotope U-239. This
is a short-lived radioactive nuclide, with a half-life of around 1 hour. The U-239 decays
spontaneously to the neptunium isotope Np-239 via beta decay. The nuclide Np-239 is
also radioactive, with a half-life of around 2 days. The Np-239 decays spontaneously,
again via beta decay to the plutonium isotope Pu-239. Beta decay is the term used
for electron emission in nuclear physics. Each beta decay results in the emission of one
electron, increasing the charge number Z of the nuclide by one in each step. The overall
process is
238 239 239 − 239 −
92 U + n → 92 U → 93 Np + e → 94 Pu + e (1.14)

From the graph of the fission cross sections, figure 1.8, it is clear that Pu-239 is also
a fissile nuclide, with a fission probability similar to U-235. Several fission reactions are
possible, such as the one resulting in the nuclides xenon and zirconium, also producing
three secondary neutrons, as shown below
239 134
94 Pu +n→ 54 Xe +103
40 Zr + 3n + 200 MeV (1.15)

The fission of Pu-239 produces nuclear energy, which is added to the energy produced
by fission of U-235. In fact, about one-third of the total energy produced in the common

24
commercial reactors today is due to fission of plutonium. 21
The level of Pu-239 in the reactor is determined by the balance of Pu-239 production
and loss processes. Several reactions are possible, such as the production of Pu-239 due
to neutron capture in U-238, and the loss due to the neutron induced fission of Pu-239.
The U-238 isotope is referred to as a fertile isotope due to it’s ability to convert into
another fissile isotope. The cross-sections for the two processes are shown in figure 1.9. In
a conventional thermal fission reactor operating with slow neutrons, the cross-section for
Pu-239 fission is much higher than that for U-238 capture. This means that the Pu-239 is
lost almost instantly after it is produced, and as a result, the fraction of Pu-239 in the fuel
remains small. The sustainment of the chain reaction would in this case still be mainly
driven by the fission of the U-235 isotope. When the fissile U-235 in the fuel has been
consumed, the reactor fuel has to be replaced. In this time period, much of the U-238 is
still present in the spent fuel, and remains in the fuel that is removed from the reactor.
However, it can be seen in figure 1.9 that, for higher neutron energies, the U-238
capture and the Pu-239 fission cross-sections become comparable, with the consequence
that the production and loss of Pu-239 can be balanced. When the so called breeding
ratio is close to one, the fissile Pu-239 is replenished from U-238 at the same pace as it is
consumed. In this case, there can be a significant amount of Pu-239 present in the fuel,
contributing to the sustainment of the chain reaction. The level of Pu-239 in the fuel
remain in this case constant, while the U-238 is constantly consumed in order to replenish
Pu-239. As a result, the uranium fuel can be used for longer time, and the U-238 isotope
is utilized in the reactor, instead of being left unused in the spent fuel. A reactor that
operates without a moderator, in order to use fast neutrons in the fission process, is
referred to as a fast reactor. Another possibility, to utilize the plutonium fission reaction,
is to use reprocessed reactor fuel, so called mixed-oxide (MOX) fuel, containing a mix of
uranium and plutonium.

1.5.3 Fission reactor technologies

Thermal reactors
The main fission reactor technology used today is the thermal Water Cooled Reactor
(WCR). The WCR is referred to as a thermal reactor, since it incorporates a moderator,
which functions to slow down the neutrons produced in the fission reactions, reducing
their initial energies in the MeV range to low, thermal energies below 1 eV. The uranium
fuel pellets are stacked in thin, long cylinders placed vertically in the reactor core, so
called fuel rods. The water circulates around the fuel rods, and functions both as coolant,
removing the heat produced in the fission process, and as moderator, slowing down the
neutrons. The reactor is kept at a marginally critical state by adjusting the neutron
multiplication factor to a value close to one by inserting neutron absorbing material.
Typical materials used are boron, cadmium, silver, hafnium or indium. The material is
placed in long cylinders, so called control rods placed vertically, mixed with the fuel rods.
By moving the control rods in (out) of the core, the amount of absorbing material in the
reactor is increased (decreased), adjusting the neutron multiplication factor down( up).
The water can be regular water (H2 O), or heavy water (D2 O), where the regular hydrogen
isotope 11 H is replaced by a heavier hydrogen isotope deuterium 21 H, often more simply
denoted by the letter ”D”. The reason for using heavy water, which has to be produced
21
https://world-nuclear.org/information-library/current-and-future-generation/fast-neutron-
reactors.aspx

25
from regular water at some cost, is that is a better material as a moderator. Heavy water
is also effective in slowing down neutrons in elastic scattering collisions, while it absorbs
significantly less neutrons in inelastic collisions. This characteristic of the moderator is
important, in order for the reactor to sustain the fission chain reaction, and it enables
the use of natural uranium with a low percentage (0.7 %) of fissile U-235 as fuel, a main
advantage of the Heavy Water Reactor (HWR).
The water cooled reactor technology, dominating today, is the Light Water Reactor
(LWR), which uses regular water. It requires that the uranium fuel is enriched to 3-5 %
of fissile U-235, in order to sustain chain reaction. Plants for enrichment of uranium are
capital intensive, and there are presently only a relatively operating facilities worldwide,
in France, Germany, Netherlands, UK, USA and China. 22 . The commercial process
employed for the enrichment is based on the mass difference between the U-235 and
U-238 isotopes, and involves spinning gaseous uranium fuel in fast rotating centrifuge
devices. Since the enrichment processes requires uranium in a gaseous form, uranium
oxide from the mine is converted to uranium hexafluoride in a preliminary process.
The water cooled reactors are of two types. In the Boiling Water Reactor (BWR), the
water circulates in a single circuit. The water is heated to the boiling temperature, and
the steam produced drives a turbine, which is connected to a generator that produces
electricity. In the Pressurized Water Reactor (PWR), the water circulates in two separate
circuits, which are connected by a heat exchanger. In the primary circuit, the water is
kept at a high pressure of 150 bar, allowing the water to reach temperatures of over 300 ◦ C
without boiling. An advantage of the PWR is that the irradiated water flowing through
the reactor remains separate from the water in the secondary, outer circuit.

Fast reactors
Fast reactors are characterized by operation without a moderator, and as a consequence
the fission process is relying on highly energetic, fast neutrons. At higher neutron energies,
the production of the Pu-239 plutonium isotope due to neutron capture in the uranium
U-238 isotope can balance the loss of Pu-239 due to neutron induced fission, at a PU-239
breeding ratio close to one.
In a thermal reactor, the chain reaction is sustained by fission of the U-235 isotope,
and the reactor fuel has to be replaced when the U-235 content in the fuel is consumed.
At this time, most of the U-238 isotope in the fuel remains, and is left in the spent fuel
removed from the reactor. In a fast reactor, the fraction of plutonium in the reactor
fuel can be at a significant level, and the fission chain reaction can be sustained by the
plutonium reactions. The plutonium is continuously replenished from U-238 isotope,
which is consumed in the process. As a result, the uranium fuel rods can be used for
much longer time before replacement, and the U-238 isotope is utilized during the reactor
operation, instead of being left unused in the spent fuel. A major advantage of the fast
reactor is that the reactor, by using the main constituent of the uranium fuel in the
breeding process, uses the uranium fuel much more efficiently than a thermal reactor.
This is important for the long-term usage of fission reactors, since it enables them to
utilize the world’s limited uranium reserves more efficiently.
The fast reactor can not use water a coolant, since it would act as a moderator. Fast
reactors typically use liquid metals as coolants, such as sodium or lead at elevated tem-
peratures. Due to their high thermal conductivity, metal coolants remove heat efficiently,
enabling high power density in the reactor. This makes them attractive in applications
22
https://world-nuclear.org/information-library/nuclear-fuel-cycle/conversion-enrichment-and-
fabrication/uranium-enrichment.aspx

26
where size and weight must be limited, such nuclear reactors for powering naval ships
or submarines. A problem with sodium is it’s chemical reactivity, which requires special
precautions to prevent fires. If sodium comes into contact with water, it reacts to produce
sodium hydro-oxide and hydrogen, and the hydrogen burns in contact with air. Lead has
a relatively high melting point of 328 ◦ C, which makes refuelling difficult. The melting
point can be lowered to 123 ◦ C by forming an alloy with bismuth, but a problem is then
that the lead-bismuth alloy is highly corrosive to most reactor structural materials.

Trends in nuclear fission technology

Nuclear reactor technologies are commonly categorized in ”generations”. 23 Generation
I refers to prototype and early reactors in the 1950s and 1960s. Generation II refers to a
class of commerical reactors designed to be economical and reliable, with typical lifetimes
of 40 years. These include the PWR and BWR reactors. Generation III reactors are
essentially Generation II reactors, with evolutionary improvements, included in reactor
design in the years from 1990s and onwards. The improvement is particularly in the area
of the safety systems, introducing the use of passive, rather than active safety systems.
Examples of safety improvements are double concrete walls enclosing the reactor, passive
system for circulation of coolant, and meltdown collectors built-in underneath the reactor
core.
The current research and development in nuclear fission is embodied in the design
of the next Generation IV reactors. The Generation IV International Forum (GIF) has
defined six reactor technologies as candidates for the next Generation IV reactors 24 . Most
of them are fast reactors:
• Gas-cooled fast reactor (GFR) Helium coolant at 800-850 ◦ C
• Lead-cooled fast reactor (LFR) Liquid lead coolant at atmospheric pressure and
temperature 550-800 ◦ C
• Molten salt fast reactor (MSFR) Thorium fuel cycle, fissile material dissolved in
circulating fuel salt
• Advanced high-temperature reactor (AHTR) Molten salt coolant
• Sodium-cooled fast reactor (SFR) Liquid sodium coolant at 500-550 ◦ C
• Supercritical water-cooled reactor (SCWR) Very high pressure water coolant at 250
bar and temperature 510-550 ◦ C
• Very high temperature gas reactor (VHTR) Graphite moderated helium gas cooled
reactor at 900-1000 ◦ C

1.5.4 Public concerns about fission reactors

The development of nuclear power in recent years has in many countries been adversely
affected by public concerns. The main issues with nuclear fission reactors put forward are
connected with radioactivity and the health hazards of ionizing radiation: 25
• Accidents. In the event of a major accident, nuclear power plants can eject dangerous
radioactive material in the atmosphere and surrounding region.
• Nuclear waste. The waste generated by nuclear reactors remains radioactive for
up to 100 000 years. There are still no operational long-term storage solutions for
radioactive waste, and most of it is stored in temporary on-site facilities
23
https://www.amacad.org/sites/default/files/academy/pdfs/nuclearReactors.pdf
24
https://www.gen-4.org/gif/
25
https://www.greenamerica.org/

27
 • Nuclear proliferation. There is a concern that the development of nuclear energy
increases the likelihood of proliferation of nuclear weapons.

Accidents. There have been only two major accidents at nuclear power plants, and
their impacts have been less severe than was feared.26 The most serious nuclear accident
took place in 1986 at the Chernobyl nuclear power plant in Ukraine. A failed safety test
caused two explosions, a fire that lasted over a week, and the release of large amounts
of radioactive material. In 2011, an earthquake outside the coast of Japan triggered a
massive tsunami, and three of the reactor cores at the Fukushima Daiichi nuclear power
plant melted, and there were several hydrogen explosions and release of nuclear material
into the environment.
The radioactive materials and ionizing radiation released in a nuclear accident is a
potential health risk for people. Much of what is known about cancer caused by radi-
ation exposure from nuclear accidents comes from research on the Chernobyl accident.
27
Approximately 600 persons who worked at the plant during the emergency received
very high doses of radiation and suffered from Acute Radiation Syndome (ARS), also
known as radiation sickness or radiation poisoning. Among those, the 28 workers who
received the highest doses died. In the hundred of thousands of people worked as part of
the clean-up crews in the years after the accident an increased risk of leukaemia has been
found. Among 5 million residents near Chernobyl it is found that children have got an
increased risk of developing thyroid cancer.

Nuclear waste. When the nucleus splits, most of the energy is released immediately and
carried off by the coolant. However, energy continues to be released for thousands of years
after the atom splits. The afterglow heat is what makes nuclear waste hazardous. This
delayed energetic emission means that the nuclear waste is highly radioactive. In practice
the spent fuel is kept underwater for 5-8 years in spent fuel pools until the radiation has
decayed to a level where active cooling is no longer needed. Then, in today’s nuclear
power plants, the nuclear waste is either recycled in a fuel recycling plant, or kept on site
as a temporary solution.
Currently, the main option for the long-term solution to the waste problem is per-
manent storage in stable underground geological formations. The Onkalo deep geological
repository in Finland will be the world’s first commercial site to come into operation, serv-
ing the Olkilouto power plant. 28 Deep in the bedrock in southwest Finland, 450 meters
below the Earth’s surface, fuel rods containing spent uranium pellets will be inserted into
copper canisters that rest in water-absorbing bentonite clay. This disposal concept has
been developed by the Swedish nuclear fuel and waste management company SKB, and
a similar storage site is being planned in Sweden at the Forsmark nuclear power plant.
Another approach to the nuclear waste problem that has been suggested is trans-
mutation, in which the waste is transformed to short-lived, or even stable isotopes by
irradiation with neutrons. This can be done in two ways, either by using the neutron flux
in a fission reactor, or with an accelerator-powered system.

Nuclear proliferation. Nuclear weapons are the most dangerous weapons on Earth.
One weapon can destroy a whole city, potentially killing millions of people, and jeopar-
26
https://world-nuclear.org/nuclear-essentials/what-are-the-effects-of-nuclear-accidents.aspx
27
https://www.cancer.gov/about-cancer/causes-prevention/risk/radiation/nuclear-accidents-fact-
sheet
28
https://psmag.com/ideas/the-hiding-place-inside-the-worlds-first-long-term-storage-facility-for-
highly-radioactive-nuclear-waste

28
dizing the natural environment and lives for future generations. The dangers from such
weapons arise from their very existence. Although nuclear weapons have only been used
twice in warface - in the bombings of Hiroshima and Nagasaki in 1945 - many still exist
in our world today.
Nuclear proliferation is the spread of nuclear weapons, nuclear technology, fissionable
material or nuclear weapons-making information to countries that do not posses these.
The United Nations has already in 1946, in the first resolution by the UN General As-
sembly established a commission to ensure that atomic energy is only used for peaceful
purposes. The International Atomic Energy Agency (IAEA) is known as the world’s
”Atoms for Peace and Development” organization within the United Nations.
The IAEA is the international centre for cooperations in the nuclear field, and it has an
important role to verify the fulfilment of obligations assumed under the Non-proliferation
of Nuclear Weapons Treaty (NPT) that came into force in 1970. 29 The treaty has
recognized the countries that have nuclear weapons as ”Nuclear Weapon States” NWS)
and those that do not possess nuclear weapons as ”Non-nuclear Weapon States” (NNWS).
Nuclear proliferation takes place when the NWS pass their nuclear weapons or technology
to the NNWS. Currently, there are 5 NWS (US, Russia, UK, China and France), and 190
NNWS. Four other states are believed to possess nuclear weapons, but are not recognized
as NWS (India, Pakistan, North Korea, and Israel). In 1993, North Korea became the
only state to withdraw from the treaty.

1.6 Fusion energy

Let us now consider the left, low mass number side of the binding energy curve (figure
1.6). Interestingly, merging of two light atomic nuclei releases nuclear energy since they
increases the binding energy. This is called fusion. Fusion reactions between light nuclei
is the most important energy source in the Universe. Our Sun, like all other stars, is a
gigantic fusion reactor. The physics of the Sun was unknown to man until in the 1920’s.
At that time, the idea was born to copy the mechanism of the Sun and construct a new
energy source on Earth; the fusion reactor.

1.6.1 Fusion reactions in the Sun

There are many possible fusion reactions. In the centre of the stars, where the temperature
is around 15 million degrees, protons are fused in a long series of reactions ending with
helium, as seen in figure 1.10. This process releases immense amounts of energy.
The fusion reactions in the centre of the sun are actually a chain of reactions where vast
amounts of energy is released. The chain begins with ordinary hydrogen nuclei (protons;
p) and ends with helium. The reactions can be described by three steps:

1. p + p + e → D +ν + γ+ 1.44 MeV
2. D + p → 3 He+γ+ 5.49 MeV
3. 3 He + 3 He → 4 He + p + p + γ + 12.86 MeV.

Here, e denotes electron, ν neutrino and γ gamma rays. The reaction velocity is very
low at the 15 million degrees, which is the temperature of the Sun. But, since the volume
is huge, 4 million tons of material is transformed into energy every second. This energy,
29
https://www.iaea.org/topics/non-proliferation-treaty

29
Figure 1.10: The fusion processes in the sun. See also section 1.2.1. (Figure credit:
http://FusEdWeb.llnl.gov/CPEP/)

mainly in the form of gamma radiation, is transformed into visible light during a long
(millions of years!) and complicated route through the core of the Sun.

1.6.2 The promise of fusion energy

In 1920, Sir Arthur Eddington gave a fascinating lecture to the British Association for
the Advancement of Science, from which the following excerpt is taken. Scientists were
beginning to unveil the mysteries of fusion, so far unaware of the dedicated effort to fusion
development that would follow in the decades to come.
”A star is drawing on some vast reservoir of energy by means unknown to us. This
reservoir can scarcely be other than the sub-atomic energy which, it is known, exists
abundantly in all matter; we sometimes dream that man will one day learn how to release
it and use it for his service. The store is well-nigh inexhaustible, if only it could be
tapped. There is sufficient in the Sun to maintain its output of heat for 15 billion years.
The nucleus of the helium atom, consists of 4 hydrogen atoms bound with 2 electrons.
But Aston has further shown conclusively that the mass of the helium atom is less than
the sum of the masses of the 4 hydrogen atoms which enter into it. There is a loss of
mass in the synthesis amounting to about 1 part in 120, the atomic weight of hydrogen
being 1.008 and that of helium just 4. We can therefore at once calculate the quantity
of energy liberated when helium is made out of hydrogen. If 5 per cent of a star’s mass
consists initially of hydrogen atoms, which are gradually being combined to form more
complex elements, the total heat liberated will more than suffice for our demands, and
we need look no further for the source of a star’s energy.”
Fusion energy is the energy released in controlled nuclear fusion reactions involving
light elements. To produce energy in this way requires a technology which is very different
from that in the fission reactor, and it has turned out to be much more difficult to realize.
We have, so far, discussed energy sources that are more or less well developed, and with
rather well known potential, regarding both economy and resources. Fusion energy is,
in comparison with these technologies in an early stage of development, and not even a
full scale prototype reactor exists today. In order to realize fusion energy, the field of
fusion science has to be further developed, to a level where reliable predictions of the
performance of fusion devices can be made. In addition, further developments in the field
of fusion technology and fusion reactor materials are required, in order to confidently

30
design a fusion reactor. Fusion research is indeed a long-term effort, which has engaged
a large number of scientists all around the world for many years, and continue to do so
today.
The enormous potential advantages of fusion energy, which has motivated the resources
and human efforts devoted to fusion research world wide over many years, are mainly the
following:
• Abundant energy. In a fusion reactor, energy would be produced in nuclear fusion
reactions involving the two hydrogen isotopes deuterium (21 H) and tritium (31 H).
Deuterium can be extracted from sea-water, while tritium will have to be generated
in the fusion reactor from nuclear reactions in lithium. Terrestrial reserves of lithium
would permit the operation of fusion power plants for more than 1000 years, while
sea-based reserves of lithium would fulfil needs for millions of years.

• No carbon dioxide emission. As with other kinds of nuclear energy, fusion reactions
do not produce greenhouse gas emissions. The by-product of fusion is helium, an
inert, non-toxic gas.

• No long-lived radioactive waste. Nuclear fusion reactors will not produce high ac-
tivity, long-lived nuclear waste. The activation of components in a fusion reactor is
anticipated to be low enough for the materials to be recycled or reused within 100
years.

• Limited risk of nuclear proliferation. Fusion reactors do not employ fissile materials
like uranium and plutonium. There are no materials in a fusion reactor that could
be used to make nuclear weapons.

• No risk of serious nuclear accidents. A serious nuclear accident, equivalent to a

core meltdown in a fission reactor, would not be possible in a fusion reactor. The
quantity of fuel present at any time in the reactor is enough only for a few seconds
of operation, and there is no risk for a chain reaction.
As with other energy sources, fusion electricity will come at a higher cost initially, when
the technology is new, and then gradually become less expensive as the economics of scale
brings the costs down. In a recent report, which examined the costs for first generation
fusion power plants, the levelized-cost-of-electricity (LCOE) for fusion electricity in the
year 2040 is estimated to be in the range 0.06-0.10 $/kWh.30 Capital costs of magnetic
confinement fusion are high, with the core device costs - such as magnets - being almost
50 % of total costs. These costs should go down as more reactors are built, since innovation
in design and manufacture is likely to lead to a price reduction following the learning curve
observed for other new technologies, such as the renewable energy sources, see figure 1.4.

1.6.3 Possible ways to realize fusion energy

If we want to use fusion reactions for energy productions, one problem is immediately
obvious. The electrostatic Coulomb force between the positively charged atomic nuclei
is repelling. Only when the nuclei come very close to each other, about 10−15 m, will
the attracting, strong nuclear force dominate, and pull the nuclei together. In terms
of energy, the kinetic energy of the reacting particles has to be higher than the positive
30
https://www.nucnet.org/news/capital-costs-are-high-but-can-be-reduced-to-economically-
competitive-level-10-4-2021

31
 POTENTIAL
ENERGY
Z1 Z2 e 2
EMAX =
EMAX 4π ε 0 R 0

NUCLEI REPEL
(COULOMB REPULSION)

r
R0 DISTANCE BETWEEN NUCLEI

≈ ≈
NUCLEI ATTRACT
(NUCLEAR FORCE)

Figure 1.11: Potential energy as a function of the distance from the nucleus. Another
incident nucleus requires sufficiently high kinetic energy so that it can pass the positive
Coulomb potential barrier in order come close enough to be pulled in by the negative
potential of strong nuclear force. The peak of the barrier is at a distance of order R0 ∼
10−15 m

potential energy barrier of the Coulomb interaction, see figure ??. It should be mentioned
that quantum mechanical tunnelling allows fusion reactions to take place at lower kinetic
energies.
How do we make the particles come close enough to each other? After some consider-
ation, one might come up with at least three ideas:
1. the particles are confined and heated together (thus forming a plasma) so their
thermal velocity may overcome the necessary Coulomb potential,
2. the particles are brought together at high speed by an accelerator,
3. the positive charge of the nuclei is shielded by negative charges, and the nuclei are
“tricked” to approach each other.
The first idea is thermonuclear fusion, where the fuel atoms are heated to a tempera-
ture of several hundred million degrees. It has so far been dominating, and it has evolved
along two main paths. The first path is some form of magnetic confinement. Most of
this text book will focus on this successful method. The second path is based on the
possibility to ignore any confinement, except that of fuel inertia. If the fuel is compressed
and heated for a few nanoseconds, conditions required for a positive energy exchange may
be achieved. This is called inertial confinement and has, so far, seen a similar progress
as magnetic confinement. The terms laser fusion and ion beam fusion are also used,
depending on which method is used for heating of the fuel.
Possibility 2) might seem tempting. Why not simply use two ion beams of opposite
direction, with a relative energy which optimizes the cross section for fusion? Unfortu-
nately, a simple calculation shows that the cross section for Coulomb collisions is several
orders of magnitude greater, which means the particle beams will simply deflect.
Muons, particles with the same charge as the electron, but with a mass 207 times
greater than the electron’s, have mainly been discussed in connection with possibility 3),
so called muon-catalyzed fusion. By replacing the electrons in heavy hydrogen atoms with
the heavier muons, the nuclei will come sufficiently close to each other so that quantum
mechanical tunnelling will become frequent. Unfortunately, muons have a short lifetime,
which limits the number of fusion reactions each muon can catalyze, and since they are
energy-expensive to produce with accelerators, it is difficult to reach energy break-even.

32
Figure 1.12: Cross sections for different fusion reactions. (Figure credit: By permission of
Oxford University Press)

1.6.4 Fusion reactions

An analytical expression for the fusion cross section σ requires quite sophisticated quan-
tum mechanics and is difficult to calculate. In the 1930s, the physicist Gamow derived
the simple expression which has proved to be a decent approximation. It is derived for the
case that the relative energy E < Emax in figure 1.11. Nowadays, so called semi-empirical
expressions for σ are used. These utilize known relations from quantum mechanics, but
also contain free parameters that are adjusted according to experimental measurements
of σ. One example of a semi-empirical formula is
√
C0 π M Z1 Z2 e2
σ(E) = exp[− √ √ ] (1.16)
E 2ε0 h E
Here C0 is a constant which is determined experimentally. Z1 and Z2 is the charge number
of the nuclei. M = (m1 + m2 )/(m1 m2 ) is the reduced mass.
The cross section of the fusion reaction as a function of the particle’s relative velocity
for different nuclei is shown in figure 1.12.
The reaction with the highest cross section is that between deuterium and tritium,
illustrated in figure 1.13. When the nuclear reaction occurs, the nuclei hold together only
temporarily. The fused state of the nucleus is unstable, and lasts only for about 10−12 s,
after which newly formed elements are released with large quantities of kinetic energy.

D + T → 4 He (3.5 MeV) + n (14.1 MeV)

The released energy in the DT reaction is in the form of the kinetic energy of the α-particle
(used to denote to the 4 He nucleus ) and the neutron. No gamma radiation is released.
The first generation of fusion reactors are expected to use a DT-mixture as fuel. To
get an idea of the energy density of the fusion fuel, the amount of fuel needed every
second in a fusion reactor delivering 1000 MW electric energy can be estimated to be 0.01
g/s, or a little under a kilogram during 24 hours. Deuterium is abundant in nature; one

33
Figure 1.13: Fusion of hydrogen isotopes deuterium and tritium. (Figure credit: EUROfusion)

out of 6700 hydrogen atoms is the the form of deuterium. It exists in heavy water D2 O
molecules which is present in ordinary sea water. Tritium is radioactive, with a half-life of
12.3 years, thus existing only in very small amounts in nature. Therefore, the reactor core
in a DT reactor will be surrounded by a blanket containing lithium, which will produce
tritium by the nuclear fission reactions
6
Li + n → T + 4 He + 4.8 MeV
7
Li + n → T + 4 He + n - 2.5 MeV
The neutrons created in the DT reactions will generate new fuel! To enhance this effect,
some neutron multiplicating material will be included in the blanket. If the produced
deuterium and tritium are combusted at the same rate as it is produced, the reaction is
called fully catalysed, and there is no tritium to take care of.
Figure 1.12 shows the a few other interesting reactions. DD reactions (with the same
probability)
D + D → 3 He + n + 3.27 MeV
D + D → T + p + 4.03 MeV
have the advantage that no tritium is needed. The problem, as we will discover shortly,
is that both higher temperature and higher plasma density is needed in this “advanced
reaction” in order to have the same energy exchange as the DT reaction.
It is also possible to go one step further, to the so called exotic reactions which do not
generate neutrons. Thus, the activation of the reactor walls caused by the fast neutrons
is avoided. As seen in figure 1.12, the reaction
D + 3 He → 4 He + p + 18.3 MeV
has this advantage, and in addition it has a better reaction cross section than the DD
reaction at energies above 100 keV. It is, however, impossible to completely avoid neutron
production from DD reactions in the plasma. Further, 3 He is rare on Earth (but not on
the Moon).
There are other reactions which are completely neutron free. Figure 1.14 shows the
reactivity for some of these reactions. One of the most advantageous seems to be
11
5 B + p → 3 42 He + 8.7 MeV
which, however demands very high temperatures and plasma densities.
It is also important to take in consideration that neutron-free reactions transfer the
fusion energy to kinetic energy in charged particles. Methods for direct transfer of kinetic
energy to electric energy must thus be used. This can for example be achieved by locally
opening the magnetic field lines so that the fast particles can escape into a generator.

34
 10-20
LEGEND
= p - 9Be
= 3He - 6Li
10-21 = p - 6Li
= 3He - 3He
= p - 11B
= p - 7Li
10-22

<σv> (m3/s)
10-23

10-24

10-25

10-26
100 101 102 103
T (keV)

Figure 1.14: The reactivity for some exotic fusion reactions.

1.6.5 Fusion in a thermal plasma

We have thus realized that a plasma must be heated to high temperatures for fusion
reactions to take place. But how high temperatures? And how high heating power is
needed to start the reactions?
If we wish to reach an expression for the number of fusion reactions (and hence the
fusion power) in a sub-volume of the plasma, it is reasonable to look at the frequency of
the fusion reactions. We have studied probabilities, or cross sections, σ, for different fusion
reactions. The fusion reaction frequency for a test particle with velocity v is obtained from
the cross-section and the target particle density as ν = nσv, see section 1.5.1. Let us now
generalize this. Assume that the plasma consists of two different types of ions, with the
velocity distributions f1 (v1 ) and f2 (v2 ). The collision frequency is now dependent on the
occurrence and velocity of both R R these types of particles. Hence, the collision frequency
per unit volume is equal to σ(v)vf1 (v1 )f2 (v2 )dv1 dv2 , where v ≡| v1 − v2 |. This
important expression is called the reaction rate for fusion reactions and is written as

R12 = n1 n2 < σv >12 = n1 n2 ξ12 [m−3 s−1 ] (1.17)

The reactivity is
ξ12 =< σv >12 [m3 s−1 ] (1.18)
which we may regard as the efficiency of a reaction. The reactivities of some reactions as a
function of plasma temperature are shown in figure 1.15. The assumption has been made
that the plasma is thermalized, meaning that the particles are Maxwell distributed. A high
temperature plasma, just as an ordinary gas, relaxes rapidly into a Maxwell distribution
(the time is of the same order of magnitude as the Coulomb collision time). This is for
example relevant when the plasma has been heated by a current flowing through it. But,
as we shall see, other methods of heating, for example ion beams and micro waves, must
also be used to reach fusion temperatures. This disturbs the velocity distribution and a

35
Figure 1.15: Fusion reactivity < σv > for some reactions as function of the plasma temperature
in unit keV. (Figure credit: By permission of Oxford University Press)

high energetic non-Maxwellian component must be taken into account, making it more
complicated to calculate the reactivity.

1.6.6 Requirements for break-even

The term ”break-even” refers to the situation where the generated thermal fusion power
equals the input power to the plasma, used for heating the plasma to the required high
temperature. The parameter Q is often used to express this break-even condition as
Pf us
Q= >1 (1.19)
Pin
The fusion power density from the DT-reaction is
1
Pf us = n2 ξDT eEDT (1.20)
4
where EDT is the energy released in a fusion reaction in unit eV, ξDT =< σv >DT is the
DT reactivity, n = nD + nT is the total plasma ion density, assuming a 50/50 % DT
mix. We note that EDT = 17.6 × 106 eV. In steady state, power input and power loss
are balanced Pin = Ploss . The plasma loss power is quantified by the energy confinement
time τE , defined as
W
τE = (1.21)
Ploss
where
3
W = (ne eTe + ni eTi ) = 3neT (1.22)
2
assuming a plasma with equal electron and ion density n and temperature T , and with
the plasma temperature T in unit eV. The energy break-even requirement is thus
W
Pf us > Pin = Ploss = (1.23)
τE

36
which can be rewritten as
1 2 3neT
n ξDT eEDT > (1.24)
4 τE
The reactivity ξDT is, within 10 % proportional to T 2 for fusion relevant temperatures
T = 10 − 20 keV, which means ξDT /T 2 is nearly constant, and the value is approximately

ξDT
≈ 1.1 × 10−30 m3 s−1 (eV)−2 (1.25)
T2
We can then write the break-even condition as
12
nT τE > (1.26)
EDT (ξDT /T 2 )

Inserting numerical values we get

12
nτE T > −30
= 6.2 × 1023 m−3 s eV (1.27)
17.6 × 106 × 1.1 × 10
The product nτE T is called the triple product and it is often used to quantify the require-
ment for a fusion reactor. We conclude that break-even requires the triple product to be
of order
nτE T ∼ 1021 m−3 s keV (1.28)
when the plasma temperature in expressed in unit keV. We note that this condition must
be regarded as ”scientific” break-down, since it does not take into account of the efficiency
of the conversion of the thermal fusion energy to electricity, or the efficiency of the plasma
heating process.

1.6.7 The Lawson criterion for a fusion reactor

The British engineer J. D. Lawson published an article in 1957, in which he derived a
condition for a fusion reactor to produce as much energy as is required for its heating.
This engineering break-even condition is in a practical situation more strict than the
scientific break-even condition Q > 1, since it considers also the efficiency by which
the produced fusion thermal energy is converted into electricity. We want to balance
the power input with power produced in the reactor. Both fusion reactions and the
power losses produces heat in the reactor material walls, which is harvested to generate
electricity, which subsequently is used to heat the plasma. The ”round-trip” efficiency for
this process is denoted η. Lawson thus requires for net power output of the reactor

Pin < η(Pf us + Ploss ) (1.29)

in a stationary state
Pin = Ploss (1.30)
leading to the Lawson criterion,

Ploss < η(Pf us + Ploss ) (1.31)

or
η
Ploss < Pf us (1.32)
1−η

37
 1021

nτΕ
(m-3 s)

1020

10 100
T (keV)

Figure 1.16: Lawson diagram for the DT reaction, calculated with a reactor efficiency of
η = 1/3. (Figure credit: By permission of Oxford University Press)

We note incidentally, that this criterion gives the condition Q > 1 for a theoretical round
trip efficiency of η = 1/2. However, for a realistic smaller value for η < 1/2, Lawson’s
analysis gives a more strict criterion, as we anticipated.
The fusion power density Pf us for an optimum fuel mixture n1 = n2 = n/2. This gives
us
1
Pf us = n2 ξ12 eE12 (1.33)
4
where E12 is the fusion energy per reaction in unit eV. The electromagnetic radiation loss
power density in form of bremsstrahlung emission is

Prad = Cn2 T 1/2 (1.34)

where C = 1.7 × 10−38 W m3 (eV)−1/2 . Lawson writes the loss power density as

W
Ploss = Prad + . (1.35)
τE
This requires some explanation. In the fusion research literature, the energy confinement
time τE is sometimes used for characterizing the total loss power from the plasma, and
sometimes it is used for characterizing only the non-radiative part. In Lawson’s criterion,
τE refers to the non-radiative losses. The second term is the loss power density due to
all other processes, and it is expressed as the plasma kinetic energy density W = 3neT
divided by the energy confinement time τE .
The Lawson criterion is finally written as
3eT
nτE > √ . (1.36)
η 1
ξ12 eE12 − C T
1−η4
where T and E12 are in unit eV. Observe that the right hand side is function of the plasma
temperature only (for a given value of η). The marginal condition, the break-even case,
is illustrated in figure 1.16 for η = 1/3. The curve has a minimum at T ≈ 25 keV, where
nτE = 6 × 1019 m−3 s. We have seen that the Lawson criterion sets the required values for

38
Figure 1.17: Ignition by heating from α-particles. (Figure credit: By permission of Oxford
University Press)

plasma density, temperature and confinement time if the fusion reactions are to heat the
plasma. The plasma temperature should be about 10-40 keV for the DT and DD reactions
since minima for the product nτE as a function of T is in this range. The Lawson criteria
for the two types of fusion reaction are
DT-reaction nτE > 6 × 1019 m−3 s for T ∼ 25 keV
DD-reaction nτE > 4 × 1021 m−3 s for T ∼ 30 keV.
In a reactor, where we wish to produce for example 1000 MWel in excess of the
electricity used for plasma heating, we need to position ourselves a bit above the curve.
Furthermore, Lawson’s analysis tells us what temperatures, densities and confinement
times we must achieve in the different fusion experiments. In magnetic confinement,
densities should be n = 1020 –1021 m−3 and confinement times τE = 0.1–10 s.

1.6.8 Ignition
Since one of the fusion products always is an ion, it is possible to improve on the Lawson
criterion. If these fast ions can be made to remain in the plasma they will contribute to
the heating by colliding with the plasma particles, transfering their energy directly back
to the plasma particles, thus heating the plasma. This is a much more efficient way of
heating than, as the Lawson criterion assumed, letting the ions hit the wall causing heat
that is transformed to electrical energy. It is even possible that the fast ions will produce
so much heat that all external heating of the plasma can be turned off. This is called
ignition of the plasma.
In the DT case, it is possible to confine the fast α-particle during its deceleration.
This will cause α-particle heating with a power per unit volume of
1
Pα = n2 ξDT eEα (1.37)
4
where Eα = 3.5 × 106 eV is the kinetic energy of the α-particle. If we express the total
power loss density as
3neT
Ploss = , (1.38)
τE
the ignition condition becomes Pα > Ploss , or
12T
nτe > . (1.39)
ξDT Eα

39
where T and Eα both are in unit eV.
This is, of course, a stricter criterion than that of Lawson. But if we take into account
that α-particle heating is becoming significant at 5 keV, ignition may be achieved already
at 10 keV, as illustrated in figure 1.17.
However, a plasma does not have to ignite to be interesting for fusion. A driven
plasma can also be used in a fusion reactor, but, as mentioned above, the economy will
be much better if α-particles heat the plasma directly, instead of giving off heat which,
at a limited efficiency, is transformed to electricity.
An ignited plasma cannot become uncontrolled in the same way as in a fission reactor.
The slightest disturbance of the plasma equilibrium will cause energy losses and cooling.
To control a plasma which has reached ignition, a burning plasma, is however not without
problems.

1.6.9 Fusion research today

In Europe, the strategy for development of new energy technologies has been summarized
by the European Commission in the European ”Strategic Energy Technology Plan”.31
Fusion technology is described in the plan as a high potential attractive long-term, low-
carbon energy solution, and highlights the European involvement in the ITER project.
ITER ( ”the way” in Latin) is one of the most ambitious energy projects in the world
today.32 In southern France, work is under way to build a fusion device that will allow
fusion science to make a huge step forward, and progress to a point where the feasibility
of fusion can be proven. Thousands of engineers and scientists have contributed to the
design of ITER since the idea for the international joint project was launched. The ITER
members - China, EU, India, Japan, Korea, Russia and the US - are now engaged in a
collaboration to build and operate the ITER experimental device.
The European research programme on fusion is based on the European ”Roadmap to
the Realisation of Fusion Energy” 33 , and is funded by the European Commission through
Euratom. It is coordinated by the EUROfusion consortium, comprising members from all
European member states.34 The EUROfusion work programme has two pillars:
• Fusion Science. The main objectives of the Fusion Science programme are the
development of scenarios for ITER experiments and solutions to manage the heat
exhaust in future fusion power plants
• Fusion Technology. The Fusion Technology programme aims at the development
of a design for a future European fusion demonstration power plant, referred to as
”DEMO”

The European programme also involves the development of a fusion materials test facility,
the International Fusion Materials Irradiation Facility - Demo Oriented Neutron Source
(IFMIF-DONES), to be constructed in Spain.35
Currently, the central fusion research facility in Europe is the JET device, at the
Culham Centre for Fusion Energy (CCFE) in UK. It is the largest and most successful
magnetic confinement fusion experiment in the world. JET is collectively used under
EUROfusion management, and over 350 scientists and engineers from all over Europe
31
https://energy.ec.europa.eu/topics/research-and-technology/strategic-energy-technology-plan en
32
https://www.iter.org/
33
https://euro-fusion.org/eurofusion/roadmap/
34
https://euro-fusion.org/
35
https://ifmif-dones.es/

40
Figure 1.18: Schematic drawing of the JET magnetic confinement device in UK. (Figure
credit: EUROfusion-JET).

contributes to the JET programme. In the beginning of 2022, JET scientists announced
a breakthrough result, in which roughly 10 MW of fusion power was produced for a time
period of around 5 seconds.36 The record is considered a clear demonstration of the
potential for fusion energy, and is fully in line with the predictions for ITER.
Later in the same year 2022, the Lawrence Livermore National Laboratory’s National
Ignition Facility (LLNL-NIF) in the US, announced another breakthrough result, reaching
scientific energy break even, using a different approach to fusion energy - inertial confine-
ment. Break even is considered as a very significant milestone in fusion research, meaning
that more energy was produced from the fusion process than the energy put in to drive
it. 37 These recent achievements in magnetic confinement fusion and inertial confinement
fusion research contributes to the current optimism about fusion and the belief that fusion
power plants some day will be realized and deliver electricity to the grid.

1.6.10 Muon-catalyzed fusion∗

If nature would have been slightly more generous, fusion energy would already be reality.
If the conditions for muon-catalyzed fusion had been slightly more advantageous, fusion
reactions at low temperature would have been a net producer of energy. This is, however,
not the case. It all started in 1956 when physicist Luis Alvarez and his colleagues discov-
ered a very unexpected reaction in the bubble chamber they had just invented. They saw
mysterious traces in a mixture of liquid D and T when they injected negatively charged
muons (with a mass 207 times the electron’s). Finally, they came to the conclusion that
36
https://euro-fusion.org/eurofusion-news/european-researchers-achieve-fusion-energy-record/
37
https://www.energy.gov/articles/doe-national-laboratory-makes-history-achieving-fusion-ignition

41
the 1.7 cm long muon traces corresponded to the 5.5 MeV released by fusion of a proton
and a deuterium nucleus,

µ + p + D → 3 He + µ + 5.5 MeV

What happened was that pDµ-molecules were formed, and the muon catalyzed the fusion
reaction by a drastic reduction of the distance between the proton and the deuterium
nucleus, thus allowing quantum mechanical tunnelling. Early hopes about fusion energy
were dashed, however, by calculations (which actually already existed) showing that each
muon on average catalyzed less than one fusion reaction. The reason for this is the short
life time of muons, about 2.2 µs. Another reaction, the DT reaction, was more promising
and the focus of research until the late 1980s. In this case, the muon will catalyze more
than a hundred fusion reactions during its life time.
A major problem with muon catalyzed fusion is the probability for the muon ”sticking”
to one of the produced α-particles. If the muon sticks to the α-particle, it will not catalyze
more fusion reactions, and the so called catalytic chain will be broken. The energy cost for
producing muons by accelerators is very large, in the GeV range, requiring that each muon
catalyze of the order of one thousand fusion reaction in order for muon-catalyzed fusion
to reach energy break-even. The required number of reactions per muon is larger than
what currently is observed in experiments. In order for muon catalyzed fusion to become
a viable fusion energy source, a novel, less energy-expensive method for muon-production
is needed.

42
Chapter 2

THE FUSION PLASMA

Thermonuclear fusion has proved a successful route to fusion. The principle is that a gas
mixture of for example D and T is heated to the plasma state and further to very high
temperatures and is confined in a suitable magnetic bottle. We shall now consider the
requirements on the plasma and confinement that must be fulfilled if reactor conditions
are to be reached. We shall also discuss the plasma models that we need in our further
analysis of the fusion plasma.
Before moving into the large and complex area of plasma modelling. We will need
some common physics constants, such as elementary particle masses and charges. Special
attention has to be paid to the energy and temperature units used in the formulas in this
book. Analysis of a plasma fluid model involves vector algebra, and differentiating vector
fields, using the nabla symbol. We often use the cylindrical coordinate system, which
requires some care when differentiating scalar or vector fields. Further some formulas
from electromagnetics is often employed, for example in order to compute magnetic fields
from currents, and vice versa, in some simple symmetric geometries. We have collected
all this information in an appendix at the end of this book, so it is easy to find.

2.1 Parameters of a fusion plasma

There are two important quality parameters that characterizes the performance of a mag-
netic confinement configuration. The first parameter is the energy confinement time τE .
For a stationary plasma. it is defined as
<W > < 3neT >
τE = Vp = Vp . (2.1)
Ploss Ploss

3
W = (ne eTe + ni eTi ) = 3neT
2
It is here assumed a plasma with equal electron and ion density n and temperature T , and
with the plasma temperature T in unit eV. The volume average < . . . > is taken over the
plasma volume Vp and Ploss is the total power loss from the plasma volume. The energy
confinement time can thus be interpreted as a typical ”energy leakage time” or ”cool off”
time. Let us now assume the plasma to be confined in a tokamak, see figure 2.1. The
importance of τE follows from the Lawson criterion for break-even in a fusion reactor. A
large tokamak such as the JET device has τE ∼ 1 s.

43
Figure 2.1: Principal drawing of plasma and coils in a tokamak . The direction of the ring is
called toroidal and the direction along the edge of the plasma cross section is called poloidal .
(Figure credit: EUROfusion-JET)

The second important quality parameter for the magnetically confined plasma is the β
value, defined as the ratio of plasma kinetic pressure p to magnetic field pressure B 2 /2µ0 :
<p> < 2neT >
β= = . (2.2)
< B 2 /2µ0 > < B 2 /2µ0 >

p = ne eTe + ni eTi = 2neT

Again we assume equal density and temperature for electrons and ions. The importance
of β becomes apparent if we consider the expression for the fusion power density;
 
1 2 2 ξ
Pf us = n ξE ∝ (nT ) E
4 T2
    (2.3)
2 ξ 2 4 ξ
∝p E∝β B E
T2 T2
where ξ = ξ12 is the fusion reactivity, and E = E12 is the energy release per fusion
reaction. The reason for rewriting the expression is the following: The reactivity ξ is,
approximately proportional to T 2 for fusion relevant temperatures, which means ξ/T 2 is
approximately constant. From (2.3), it is clear that β and the magnetic field B are the two
parameters that can be varied to maximize the fusion power. The maximum magnetic
field is limited, and therefore, to have as high β as possible is of outmost importance.
However, we note that for magnetic confinement, the β parameter is naturally limited to
values less than one. In practise it has turned out that β is typically limited to relatively
small values by stability requirements that characterizes each confinement configuration.
In large tokamak, the typical beta value is of the order a few percent.
One question that arises is, what are the typical values of the plasma parameters in
a reactor-class magnetic confinement fusion device? From the condition for the triple

44
product (eq. 1.28), we find the required parameter combination nτE T ∼ 1021 m−3 s keV.
Typical parameters for a large tokamak such as the JET device, with capacity to reach
near break-even condition, are

n ∼ 1020 m−3 , τE ∼ 1 s, T ∼ 10 keV (2.4)

which are line with the triple product required. The stored plasma energy density for
these parameters is

W = 3neT ∼ 3 × 1020 × 1.6 × 10−19 × 104 = 4.8 × 105 Jm−3 (2.5)

The plasma volume of JET is of the order of Vp ∼ 90 m3 , which together with the energy
confinement time τE determines the necessary power input to keep the plasma at this
energy density

W 4.8 × 105
Pin = Vp ∼ × 90 ∼ 4.3 × 107 W ≈ 40 MW (2.6)
τE 1
The plasma pressure is

p = 2neT ∼ 2 × 1020 × 1.6 × 10−19 × 104 = 3.2 × 105 Pa ≈ 3 atm (2.7)

A typical beta value is β ∼ 5 %, and together with the plasma pressure, this determines
the required confinement magnetic field strength. Inserting parameter values we obtain
2µ0 p
B2 = ∼ 2 × 4π × 10−7 × 3.2 × 105 /0.05 = 16 T2 (2.8)
β
giving a required tokamak magnetic field of order B ∼ 4 T.

2.2 Plasma models

A fusion plasma is a very complicated system and is not even close to being completely
understood. The complication is connected to the fact that 1022 ions and electrons si-
multaneously are interacting with each other. The dynamical phenomena are non-linear
and their typical time scales are separated by a factor of 106 or even more. Thus, the
level of complexity is much higher than for an ordinary gas, where only close collisions
and closely connected time scales control the dynamics.
To construct only one model which covers all problems in fusion plasma physics is
impossible at this point, and will remain so for a long time. Particle models for some
109 particles can be calculated by the fastest super computer today, but only for limited
times. Thus, simplified plasma models must be designed. Since these are simplifications,
it is of great importance to know the limitations of the models, so they are only used
for problems where they are valid. The models will contain many aspects of the various
physical problems of the plasma, electromagnetism, equations of motion, reaction kinetics,
kinetic equations, hydrodynamics, atomic physics and surface physics.
We shall now shortly discuss the most common plasma models in fusion research.
The most spectacular of the fusion parameters is perhaps the velocity of the ions. If
we estimate the circumference of a plasma torus to a few tens of meters, every ion will
complete a full turn around it some 100 000 times every second! One gets the impression
that even the most perfect magnetic bottle cannot possibly protect the ion from the wall

45
 y
B
rL

q=-e F=qv×B z θ F=qv×B q=+e

vi
x
ve

Figure 2.2: Gyro-motion for positive and negative charges.

during this long journey. Considering this, it seem like a good idea to investigate how
individual particles move in stationary magnetic and electric fields. In order to have a
convenient way of studying interactions in a system of particles, we subsequently introduce
particle distributions and kinetic theory. To be able to describe complicated geometries
in a reasonable fashion, we will need the so called fluid model .

2.3 Particle dynamics

The motion of individual particles in magnetic and electric fields is discussed below. In this
section, the goal is a general understanding of the motion of individual plasma particles.

2.3.1 Homogeneous magnetic field

Assume that a particle with charge q and mass m is in motion in plane perpendicular to
a homogeneous magnetic field B. The equation of motion becomes
dv
m = qv × B. (2.9)
dt
With the magnetic field in the z direction in a cartesian system, as shown in figure 2.2
we find
dvx
= ωc vy
dt
dvy
= −ωc vx
dt
dvz
=0 (2.10)
dt
where
qB
ωc = (2.11)
m
is the cyclotron angular frequency. It can be shown that the equations describe a gyrating
motion of the particle. We introduce a cylindrical coordinate system (r, θ, z) ( see figure
2.2). The particle gyration velocity is vθ and the corresponding angular frequency is
ω = −ωc . the angular position of the particle is θp (t) = ωt. The particle velocity

46
components are
vx = −vθ sin ωt
vy = vθ cos ωt
vz = const (2.12)
which easily can be verified by insertion in 2.10. It follows that the direction of motion for
positive charges (ions) is in the negative θ-direction, or clock-wise in figure 2.2, whereas
the direction of motion for negative charges (electrons) is in the positive θ-direction, anti-
clockwise. We introduce v⊥ for the thermal velocity in the plane perpendicular to B, and
write (
−v⊥ for positive charges (ions)
vθ = (2.13)
v⊥ for negative charges (electrons)
It is convenient to introduce the following relation between vθ and v⊥
q
vθ = − v⊥ (2.14)
|q|
We integrate eq. 2.12 to get the position of the particle (xp (t), yp (t)) as
xp = rL cos ωt
yp = rL sin ωt (2.15)
The radius of the circular orbit in the xy-plane is
vθ vθ
rL = =− (2.16)
ω ωc
which can be written as
mv⊥
rL = . (2.17)
|q|B
In a thermal plasma, we can set v⊥ = vT , where the thermal velocity vT is obtained as
1
2
mvT2 = eT . The Larmor radius is then
√
2meT
rL = (2.18)
|q|B
Now, how large is the Larmor radius? This is obviously an important question for mag-
netic confinement. Definitely, rL /a  1 must be fulfilled, let’s say rL /a = 0.01 - 0.1.
From (2.17) it is given that the ion Larmor radius is dominating the electron Larmor
radius by a factor of (mi /me )1/2 . For deuterium, we have
√ √
2meT −4 T
rL /a = = 2.0 × 10 (2.19)
eaB aB
which gives rL /a = 2.0 × 10−3 for a = 1 m, T = 10 keV and B0 = 10 T, being comforting.
We also realize that rL /a ≈ 0.07 for the fast α particles created in the DT reaction.
The cyclotron frequency for the thermal deuterons is ω ≈ 5 × 108 s−1 for the above
parameters. Since the ions travel 1 m in about 10−6 s, they gyrate about 500 times around
the field line during this distance! (The magnetic field is not homogeneous in magnetic
bottles, but the purpose of the investigation above is to give understanding of principles
and orders of magnitude).
In conclusion, a particle rotates rapidly at a distance rL from the centre of gyration
of the field line in a homogeneous magnetic field. It is also drifting at constant velocity
along the field line.

47
 z
B

E y
+ -
Vxd
x

Figure 2.3: Guiding center drift due to crossed electric and magnetic fields.

2.3.2 Guiding center drift

E × B drift
When electric fields affect the motion of the particle, it will cause a slow drift of the
orbiting particles guiding center (gyro-center) in a direction perpendicular to the mag-
netic field. Does this mean that confinement is lost? Let us investigate this. Assume
perpendicular homogeneous magnetic and electric fields B = Bz ẑz and E = Ey ŷ
y , as shown
in figure 2.3. The equation of motion then becomes
dv
m = q(E + v × B). (2.20)
dt
Writing out the components
dvx
m = qvy Bz
dt
dvy
m = −qvx Bz + qEy (2.21)
dt
dvz
m =0
dt
and introducing ωc , we get
dvx
= ωc vy
dt
dvy Ey
= −ωc vx + ωc (2.22)
dt Bz
which gives an equation for vy
d2 v y
= −ωc2 vy
dt2
The expression for the velocity components is finally obtained as
Ey
vx = −vθ sin ωt +
Bz
vy = vθ cos ωt (2.23)
vz = const
The guiding center drift velocity vE = Ey /Bz in the x direction is superimposed on the
gyration motion. We can write in more general form the expression for the E × B drift
in vector notation as
E×B
vE = . (2.24)
B2
48
 z
B

Δ
i B
Vxd y
-
x + e
Vxd

Figure 2.4: Illustration of the ∇B drifts for ions and electrons

This guiding center drift is equal for ions and electrons in both absolute value and direc-
tion!

General expression for the guiding center drift

Analogous to the electric force F = qE, any force F that is perpendicular to the magnetic
field, introduced on the right side of eq. (2.20) causes a drift. The general expression for
the guiding drift caused by a force F is written as
F×B
vF = . (2.25)
qB 2
The direction of the drift resulting from this force is always dependent on the charge
unless it is caused by an electric field. We will find in the following sections that this has
large consequences for particle confinement in a magnetic field.
Here, we notice that for example, a gravitational drift would be caused is caused by
the gravitational force F = mg,
mg × B
vg = . (2.26)
qB 2

∇B drift
For a homogeneous magnetic field, the Lorentz force averaged over one gyration period
< F > =< qv × B > is zero. Consider now an inhomogeneous magnetic field in the z
direction with variation of B in the y direction, as is shown in figure 2.4. In this case the
average Lorentz force will not be exactly zero, which will lead to a drift motion, called
∇B drift. In order to calculate this drift, we will consider the drift as a small perturbation
to the gyration motion, and regard the average Lorentz force experienced by the particle
over one gyration period as an external force. This external force is then inserted in the
general expression for the guiding center drift. Assume the magnetic field in the positive
z-direction.
< Fx > = q < vy Bz >
< Fy > = −q < vx Bz >
The particle velocity is approximately same as for homogeneous field:
vx = −vθ sin ωt, vy = vθ cos ωt
Taylor expand the Bz field around the centre of the gyration:
dBz
Bz ≈ B0z + y , y = rL sin ωt
dy

49
 Fcf

VII

i e
Vdz R c Vdz
B
z

Figure 2.5: Illustration of the curvature drift for ions and electrons.

dBz
< Fx >= qvθ rL < cos(ωt) sin(ωt) >= 0
dy
dBz 1 dBz
< Fy >= qvθ rL < sin2 (ωt) >= qvθ rL
dy 2 dy
We note that
2
mv⊥
qvθ rL = −
B
and write the expression for the average Lorentz force
2
1 mv⊥
< F >= − ∇B (2.27)
2 B
Inserting this force in the general expression eq. 2.25 for the guiding-centre drift gives the
∇B drift
2
< F > ×B 1 mv⊥ ∇B × B
v∇B = 2
= − (2.28)
qB 2 qB B2

Curvature drift
When the magnetic field is inhomogeneous, the field lines are generally also curved.
Assume cylinder symmetry, and introduce cylindrical coordinates (R, φ, Z). The magnetic
field is in the azimuthal direction B = Bφ φ̂
φ. Particles are moving at constant speed along
field lines with a velocity v = vk φ̂
φ as shown in figure 2.5. The curvature of the field line
is given by the radial vector Rc = R from the center of curvature to the field line. In
a frame moving with the particle, the particle experiences a centrifugal force opposite to
the inertial force
dv dφ̂
φ
Fc = −m = −mvk (2.29)
dt dt
The particle angular position is
vk
φp (t) = ωt = t (2.30)
R
and we write
dφ̂
φ ∂ φ̂
φ dφp vk
= = − R̂ (2.31)
dt ∂φp dt R
The expression for the centrifugal force is

mvk2
Fc = R̂ (2.32)
R

50
 z Δ
φ B

Bφ
-I
- Bφ + Fcf
-

R
Rc

Figure 2.6: Guiding center drift in a toroidal magnetic field

The drift is obtained by inserting the centrifugal force in the general expression for the
guiding center drift, eq.2.25, which gives
2
Fc × B 1 mvk mvk2
vc = = B R̂ × φ̂
φ= Ẑ (2.33)
qB 2 qB 2 R qRB
The general expression for the curvature drift is

mvk2 Rc × B
vc = (2.34)
qB 2 Rc2

2.3.3 Vertical drift in a toroidal magnetic field

Consider plasma in a toroidal field coil. Introduce a cylindrical coordinate system (R, φ, Z)
as shown in figure 2.6. The toroidal magnetic field inside the coil, is obtained from the
integral form of Ampere’s law, as Bφ = µ0 I/(2πR), where I is the total coil current
linked by the plasma ring. The field varies like 1/R along the major radius direction,
which corresponds to a gradient
∂Bφ Bφ
∇Bφ = R̂ = − R̂
∂R R
The radial variation in magnetic field strength gives a ∇B - drift in the vertical direction:
2 2 2
1 mv⊥ ∇B × B 1 mv⊥ 1 mv⊥
v∇B = − = (R̂ × φ̂
φ ) = Ẑ
2 qB B2 2 qRBφ 2 qRBφ

Particle motion in the toroidal direction along the curved field lines gives a vertical curva-
ture drift. The radius-of-curvature vector of the toroidal field is Rc = R, which is inserted
in the equation for the curvature drift.

mvk2 R × B mvk2 mvk2

vc = = (R̂ × φ̂
φ) = Ẑ
qB 2 R2 qRBφ qRBφ

In a practical situation, the thermal motion of the particles generally has components in
directions both parallel and perpendicular to the field lines, so both types of drift are
present simultaneously. Unfortunately, the gradient drift and the curvature drift are in

51
 z
φ
Bφ
Bθ +
Bθ B
θ
Bφ Bφ
Bθ -I
-φ
+

the same direction, and they add up, giving the total vertical guiding center drift in a
toroidal magnetic field
m 1 2 mvT2 v2
vd = v∇B + vc = (vk2 + v⊥ )Ẑ ≈ Ẑ = T Ẑ (2.35)
qRBφ 2 qRBφ ωc R
Ions and electrons have guiding-centre drift in opposite vertical direction (positive ions
drift upward and negative electrons drift downward for the magnetic field direction shown
in the figure. Since electrons and ions are drifting in the opposite direction a vertical elec-
trical field is built up. From eq. (2.24) it can be seen that an E × B drift in the outward
radial direction results for both ions and electrons! The vertical guiding-centre drift in
a purely toroidal magnetic field prevents confinement. This is the reason why toroidal
plasmas must always have a significant field component in the poloidal direction.

Guiding center drift in a tokamak

In a tokamak the problem has been solved by adding a poloidal magnetic field created
by a toroidal plasma current, as shown in figure 2.3.3. The resulting magnetic field lines
in the tokamak are helical. The radial drift away from the toroidal minor axis is cancelled
when a particle is moving around the torus along the helical field line. Thus, the ions
may (somewhat simplified) drift upward from the magnetic flux surface with radius r to
the surface with radius r + ∆r on the top of the torus, and then, when following the field
line to the bottom of the torus, continue to drift upward, returning them to the original
flux surface at radius r from the magnetic axis. On average, the twisted field lines do not
cause any radial losses.
Spitzer solved the problem in another way in his stellarator. By twisting the torus to
the shape of the number eight (8) he achieved the same effect without adding a poloidal
field.

2.4 The extension of the particle fields

Particle models provide excellent tools when it comes to presenting images of fusion
plasma phenomena, but require that the fields controlling the dynamics are given. Since
the fields caused by the particles have long reach, the plasma is usually dominated by
collective phenomena. Let us estimate the length of this distance.
For this, assume that a charge separation at a given time takes place somewhere in
the plasma; let us displace a few electrons from the surrounding ions, which are assumed
to be stationary on the time scale. The electrostatic energy density caused by this is
ε0 E 2 /2. Poisson’s equation ∇ · E = ρ/ε0 , where ρ is the charge density ne, gives
E ≈ (ne/ε0 )d if d is the electron separation. Since the kinetic energy of the electrons is

52
 - - - PLASMA + +
- - -- - - - + + +
+ +
-
- - + -- -
- - - --
- -
+ +
++ - ++
+ +
λ D -- - -- λD
+
++ +
+

φ
φo
x
φ = φo e
- λ
D

λD

Figure 2.7: Shielding of electrical charges in plasma

given by me ve2 /2 = eT /2 (one degree of freedom) their thermal energy density is ne eTe /2.
Obviously, the separation can only be maintained on a distance less than the distance
that results if the kinetic energy of the electrons is less than the electrostatic energy. If
the two energy densities are equal, we get the so called Debye length
r
ε0 T
λD = . (2.36)
ne
For a fusion plasma, λD ≈ 10−5 – 10−4 m. This can be compared to the average distance
between the particles, which is about n−1/3 ≈ 10−7 m. Thus, every particle interacts with
(103 )3 = 109 other particles within the Debye sphere. Particles outside this volume are in
principle shielded. Not until we reach distances much larger than the Debye length is it
relevant to use fluid models, which do not include effects of charge separation. In these
models, the relation ni = ne is used and the plasma is considered charge neutral.

2.5 Kinetic plasma model

If we want to study collective fusion plasma phenomena for short length and time scales,
some sort of statistical description of the large amount of particles must be introduced.
This is what is done in the kinetic plasma model.

2.5.1 The Vlasov equation∗

In the general case, we want to describe the particle density as a function of time in the
6-dimensional phase space using the distribution functionf (r, v, t). The time evolution
of f is thus given by the initial state of the particles, the forces between them and the
external forces. Let us derive the basic kinetic equation. We use Hamiltonian formalism
by introducing the generalized coordinate and canonical momentum q = r and p = mv,
and write the 6-dimensional vector as w = (∂q/∂t, ∂p/∂t).
Conservation of the number of particles in the phase space gives
∂f
+ ∇ · (f w) = 0 (2.37)
∂t
53
We can now rewrite (2.37):

df ∂f ∂q ∂p
≡ + w · ∇f = −f ∇ · w = −f (∇q · + ∇p · ). (2.38)
dt ∂t ∂t ∂t
On the left hand side, we have
d ∂ ∂ ∂q ∂p
≡ +w·∇= + · ∇q + · ∇p . (2.39)
dt ∂t ∂t ∂t ∂t
The right hand side is equal to zero. This is obvious from Hamilton’s equations
∂q
= ∇p H
∂t
∂p
= −∇q H (2.40)
∂t
which transforms the right hand side to −f (∇q · ∇p H − ∇p · ∇q H) = 0. Thus, what
remains is the very general and very important kinetic equation
∂f ∂q ∂p
+ · ∇q f + · ∇p f = 0. (2.41)
∂t ∂t ∂t
Now, let us rewrite (2.41) for a plasma with low density, which means that collisions
between particles can be neglected on the time scale we are interested in. The equation
of motion becomes
∂p ∂v
=m = q(E + v × B) (2.42)
∂t ∂t
and (2.41) becomes the so called Vlasov equation

∂f q
+ v · ∇f + (E + v × B) · ∇v f = 0. (2.43)
∂t m
This is a very useful relation when collisions between particles can be neglected. The
applications are fast phenomena, instabilities and so on. Note that each kind of particle
species in a plasma is described by its own Vlasov equation!

2.5.2 The Fokker-Planck equation∗

In denser plasmas, or for longer time scales, collision phenomena become important. The
question is how collisions can be added to (2.43). It is not sufficient to, as in statistical
mechanics, add a binary collision term to the right hand side, since plasma collisions
are not binary. Instead, the dominating collision effect is a large amount of small angle
collisions. The right hand side in the so called Fokker-Planck equation
 
∂f q ∂f
+ v · ∇f + (E + v × B) · ∇v f = . (2.44)
∂t m ∂t c

is relatively easy to calculate from statistical Markov chain analysis and will contain con-
tributions from both dynamic friction and diffusion in velocity space such as angular
scattering. However, we shall not give the expression here. As a rule, the Fokker-Planck
equation must be solved numerically. Semi-analytical solutions can be achieved for ex-
ample from calculation of the deceleration of a particle beam in the plasma.

54
 -e, m e
- v χ
r
+ Ze

Figure 2.8: Deflection of an electron due to the Coulomb force from an ion.

An important application of the Fokker-Planck equation is for calculations of char-

acteristic relaxation times in a plasma. We stated earlier that fusion cannot be realized
through injection of colliding beams of fast particles, since these would be slowed down
long before they could cause sufficiently many fusion reactions. Let us now sketch how
deceleration times are calculated.
Assume that a test particle with velocity u is injected into a plasma with Maxwellian
velocity distribution. The slowing down time characterizes the rate of change of the
velocity in the direction of the initial velocity, and is defined as
v v
τs ≡ − =− (2.45)
< dv/dt > < ∆vk >
The deflection time characterized the rate at which the particle velocity is scattered into
the direction perpendicular to its initial velocity, and is defined as
v2 v2
τd ≡ − 2
= − 2
(2.46)
< dv⊥ /dt > < ∆v⊥ >
Using this relation, we determine the relaxation of a distribution of test particles having
different temperature than the background. The time evolution of the test particles is
given by the Fokker-Planck equation (2.44). We may then calculate the time for temper-
ature equilibration between the two species.

2.6 Coulomb collisions

In a fully ionized plasma, the dominating collision process is interactions between par-
ticles by the electric field, the Coulomb force. The Coulomb force works at a distance,
meaning that when a test particle experiences a Coulomb collision, its path is changed by
interaction with many field particles in the plasma. An analysis using the kinetic model
and the Fokker-Planck equation is required to calculate accurate estimates of the collision
times. In this section we will use an instructive, simplified analysis, which results in the
correct scaling with the plasma parameters, although the numerical values of the collision
times are slightly different from those calculated with the Fokker-Planck equation.

2.6.1 Mechanics of the Coulomb interaction

Consider an electron with charge q = −e, mass me , velocity ve , passing by a positive ion
with charge Ze at a distance r, the incidence parameter, see figure 2.8).
Since the interaction takes place at a distance, the process may be described using
classical mechanics. The Coulomb force is a central force field, which means that both
the angular momentum and the energy is constant during the interaction, which gives a
relation between the incidence parameter r and the deflection angle χ;
χ r
cot = (2.47)
2 r0

55
where
Ze2
r0 = (2.48)
4πε0 me ve2
The incidence parameter r = r0 corresponds to a deflection angle π/2. The cross-section
for these ”large-angle” collisions is σ = πr02 , and the corresponding collision frequency in
a plasma with ion density ni is then ν = ni σve = ni πr02 ve .
However, the actual collision frequency is much higher, since the contribution from
”small-angle” collisions at larger distances where r  r0 dominates the summation over
field particles. The deflection angle vary with the incidence parameter as χ ∝ r−1 and at
the same time, the number of field ions N , within a circle of radius r, increase
P as N ∝ r2 .
It follows that adding contributions from small angle collisions gives χ ∝ r and, the
sum diverges.
It has be shown that the electrical field from a charged particle is shielded at a distance
greater than the Debye length λD (eq. 2.21). Thus, only field particles within a sphere
of radius λD around the test particle interact in the collision process. The collision time
is calculated by summing up contributions from field particles within this Debye sphere
under the assumption that the motion of field particles are uncorrelated.
Consider the change of the electron forward momentum p = me ve , as it moves through
a thin layer dx of field ions:
Z λD 
dp = − ni ∆p(r) 2πrdr dx (2.49)
0

where ∆p(r) is the forward momentum lost by interaction with a field ion at the incidence
parameter r. The rate of change with time is obtained if we assume a test electron with
velocity ve passing the layer in a time interval dt = dx/ve , giving
Z λD
dp
= −2πve ni ∆p(r) rdr (2.50)
dt 0

The collision frequency νei is obtained from the relation

dp
= −νei p (2.51)
dt
From eq. 2.47 follows
2r02
∆p = me ve (1 − cos χ) = me ve 2 (2.52)
r + r02
and
λD
2r02
Z  
2 λD
2
νei = 2πni ve rdr = 2πni ve r0 ln 1 + (2.53)
0 r2 + r02 r0
We now introduce Λ = λ/r0 and approximate Λ  1 to obtain the expression for the
collision frequency
νei ≈ 4πni ve r02 ln Λ (2.54)
The collision frequency, including the small-angle collisions, is a factor 4 ln Λ higher than
collision frequency for large-angle collisions. Inserting the expression for r0 from eq. 2.48,
we get the final expression for the Coulomb collision frequency for a test electron incident
on ions as
ni Z 2 e4 ln Λ
νei = (2.55)
4πε20 m2e ve3

56
For electrons and ions in a plasma, the velocity distribution of the electrons has to be
considered, and for a thermal distribution we get scaling of the collision frequency with
plasma parameters as
νei =∝ ni Z 2 ln ΛTe−3/2 (2.56)

2.6.2 Coulomb collision times

The particle collision time characterizes how long time it takes for a test particle to lose
its directed momentum by collisions with plasma ions or electrons. The momentum can
be lost either by being transferred to the transverse direction or by being transferred to
the target particle. In both cases, test particle directed momentum is lost. The electron
collision time, describing scattering of electrons by collisions with ions can be accurately
estimated using kinetic theory and the Fokker-Planck equation as
3/2
Te 11
τe = τei = 3.44 × 10 [s] (2.57)
ni Z 2 ln Λ
The collision time is obtained in unit seconds, and the value inserted for the electron
temperature should be in unit electron volt eV ( or simply volt). The collision time for
electrons scattering on electrons is approximately the same

τee ≈ τei (2.58)

The ion collision time describing scattering of ions by collisions with ions is simply related
to the electron collision time through the ion-to-electron mass ratio
 3/2  1/2
1 Ti mi
τi = τii ≈ τei 2 (2.59)
Z Te me
The collision time for scattering of ions by collisions with electrons is
 3/2  
Ti mi
τie ≈ τei (2.60)
Te me
The equilibration of the electron and ion temperatures is given by the energy exchange
time
 
mi
τex ≈ τei (2.61)
2me
1/2
The Coulomb logarithm ln Λ is weakly dependent on ne and Te (ln Λ ≈ 24 − ln(ne Te−1 )
and we may for a fusion plasma use

ln Λ ≈ 15. (2.62)
From this, it can be seen that in a D plasma, the ion-ion collision time is about 60
times longer than the electron-electron collision time. For a D fusion plasma with density
n = 1020 m−3 and temperature T = 10 keV we get τee = 0.23 ms  τii = 14 ms
 τex = 420 ms. All these times are well below the lifetime of the plasma. In a fusion
experiment with T = 200 eV we get τee = 0.65 µs  τii = 54 µs  τeq = 850 µs instead.

57
2.6.3 Plasma resistivity
Now a brief study of the resistivity of the plasma, which is determined by collisions
between electrons and ions. Let us first consider the simplest model possible. Assume
that an external electric field E generates a constant current through the plasma. The
electrons with charge −e and mass me have reached a drift velocity ve , and a force balance
is achieved between the driving electric field and the friction caused by ion collisions with
characteristic collision frequency νei

−eE = me ve νei . (2.63)

If we further assume Ohm’s law in the simple form E = ηj ≈ −ηne eve , we have
me νei
η= . (2.64)
ne e 2
Inserting the expression for the collision frequency νei , eq. 2.55 we get
ni Z 2 e2 ln Λ
η= (2.65)
4πε20 ne me ve3
For electrons and ions in a plasma, the velocity distribution of the electrons has to be
considered, and for a thermal distribution we get
ni Z 2
η∝ ln Λ Te−3/2 (2.66)
ne
The plasma charge neutrality condition requires that the number of positive and negative
charges in the plasma balances, and it is expressed in term of the densities as ne = Zni .
Using this condition, we obtain
η ∝ Z ln Λ Te−3/2 (2.67)
We now extend this simple analysis to a plasma with multiple ion species. The force
balance equation is in this case written as
X
−eE = me ve νei . (2.68)
i

where the summation is over different ion species enumerated i = 1, 2, 3, . . .. The resis-
tivity for multiple ions is
me X
η= νei . (2.69)
ne e2 i
Inserting eq. 2.55 for the collision frequency we obtain
e2 ln Λ 1 X
η= ni Zi2 (2.70)
4πε20 me ve3 ne i

The charge neutrality condition for a multi-species plasma is expressed as

X
ne = ni Zi (2.71)
i

We introduce Zeff , the effective charge of the plasma as follows

ni Zi2
P
Zeff = Pi (2.72)
i ni Zi

58
which is inserted in the expression for the resistivity

e2 Zeff ln Λ
η= (2.73)
4πε20 me ve3
For electrons and ions in multi-species plasma, the scaling of the resistivity with plasma
parameters is then obtained as

η ∝ Zeff ln Λ Te−3/2 (2.74)

Spitzer performed a careful Fokker-Planck analysis and found, in the case of singly charged
ions, a more accurate expression for the resistivity
me
ηS = 0.51 . (2.75)
ne2 τe
Inserting the expression for the electron collision time, eq. 2.57, the so called classical
Spitzer resistivity is obtained as

ηS = 5.22 × 10−5 Zeff ln Λ Te−3/2 [Ωm] (2.76)

The resistivity is obtained in unit ohm-meter, and the value for the electron temperature
is inserted in unit electron volt eV ( or simply volt).
The resistivity of the plasma is comparable to that of copper at T = 1.4 keV and is one
order of magnitude smaller for thermonuclear temperatures. We will find that this fact
makes it difficult to heat the plasma to thermonuclear temperatures by drawing a current
through it (“ohmic heating”). In a magnetized plasma, the Spitzer resistivity is the
resistivity in the direction along the magnetic field. It is noteworthy that perpendicular
to the magnetic field, the resistivity increases with almost a factor of two; η⊥ = 1.96ηk .

2.7 Fluid models

The kinetic plasma model can be very difficult to handle in realistic geometries. Therefore,
simpler plasma models, so called fluid models, are often preferred. When are fluid models
valid for a plasma? One, sometimes forgotten, fact is that fluid variables express local
relations in space.
The fluid velocity of ions, for example, v(r, t) is, according to (2.86) continuously
controlled by the local electric field E(r, t). If this field were changed, for example if an
external electric field is added, (2.86) must be modified. If particles with large Larmor
radii (rL ≈ a) are incident from areas where the electric field is significantly different from
that in r, these cause a similar effect, which is not included in the fluid equations. The
fluid equations only describe local interactions; variation in kinetic quantities, for example
Larmor radius are lost in the averaging. We should add that some, so called finite Larmor
radius terms, can be kept in the fluid equations. These are valid if the Larmor radius is
much smaller than other characteristic lengths.
If the mean free path of the particles is short, localization is maintained. For hot
fusion plasmas, this is not the case at all. Typical mean free paths for ions and electrons
are λii = τii vT i and λee = τee vT e , respectively. If the density n = 1020 m−3 and the
temperature T = 10 keV these are both 1.3 × 106 m, which, of course, is much larger than
the characteristic dimensions of the plasma. Fortunately, the Larmor radii of the particles
are usually so small, that a localized character remains perpendicular to the field lines.

59
Parallel to the field lines, fluid models are formally not valid! However, kinetic effects can
often damp the deviations from the MHD model, in particular the fast dynamics along
the field lines. But there are still a number of phenomena that the fluid model is unable
to handle.

2.7.1 Derivation of the fluid model equations∗

We shall now derive the plasma fluid model equations from the kinetic model. The kinetic
model describe the plasma in terms of the distribution function f (r, v0 , t). The idea is to
integrate the distribution function over velocity space, removing these three dimensions
of the phase space, and obtain equations for particle density n(r, t), fluid velocity v(r, t),
and pressure p(r, t). There will be one fluid equation for each species (electrons and ions)
in the plasma. The starting point is the Fokker-Planck kinetic equation (2.44):
 
∂f 0 q 0 0 ∂f
+ v · ∇f + (E + v × B) · ∇v f = . (2.77)
∂t m ∂t c
where now the particle velocity is denoted v0 in order to distinguish it from the fluid
velocity v. The equation is multiplied by functions φ(v0 ) and integrated over velocity
space. This procedure is referred to as taking the moment of the kinetic equation. With
φ(v0 ) = 1, v0 , and(v0 − v)(v0 − v), respectively, we obtain equations for
Z
n = f (r, v0 , t)dv0
Z
1
v= f (r, v0 , t)v0 dv0
n
Z
P = m f (r, v0 , t)(v0 − v)(v0 − v)dv0 (2.78)

Here P is the pressure tensor, which can describe also an anisotropic plasma pressure
and non-diagonal contributions from viscosity. For an isotropic distribution function the
tensor reduces to a scalar pressure and is given by
Z
1
p = m f (r, v0 , t)(v0 − v)2 dv0 (2.79)
3
We begin the derivation by integrating eq.2.77 over velocity space. The first term becomes
Z Z
∂f 3 0 ∂ ∂n
dv = f d3 v 0 = (2.80)
∂t ∂t ∂t
where n = n(r, t) is the plasma density. The second term becomes
Z Z
v · ∇f d v = ∇ · v0 f d3 v 0 = ∇ · (nv)
0 3 0
(2.81)

We will now show that the third term of the left side of (2.77) is equal to zero. With
F = q(E + v × B) we have
Z
F
· ∇v0 f d3 v 0 = {∇ · (φA) = φ∇ · A + A · ∇φ} =
m
Z Z
F F
= ∇v0 · ( f )d v − f ∇v0 · d3 v 0 =
3 0
(2.82)
m m
Z I I
F
= { ∇ · AdV = A · dS} = f · dS − 0 = 0
V S m

60
Here, we have used
F q q
∇v 0 · = ∇v0 · (E + v0 × B) = (0 + B · (∇v0 × v0 )) = 0 (2.83)
m m m
and the fact that f → 0 on the surface of the volume in velocity space as it expands to
contain all particles (the surface integral in (2.82)). The result is the so called equation
of continuity
∂n
+ ∇ · (nv) = 0 (2.84)
∂t
The collision term on the right side of (2.77) does not contribute, since collisions cannot
change the number of particles in phase space, and (2.84) describes a conservation, or
continuity, of particles in the phase space. This can be seen if we integrate (2.84) over
space; Z I
∂
ndV = − nv · dS. (2.85)
∂t V S

We may interpret (2.85) as that the time evolution of the total number of particles in a
volume is given only by the particle flow in and out of the surface of the volume. In a
similar way, we can multiply by v0 and integrate. After some calculations, which are not
given here, the result is the equation of motion
∂v
nm( + (v · ∇)v) = −∇ · P + nq(E + v × B) + R (2.86)
∂t
The term R is the change of the momentum caused by collisions with all kinds of particles.
Note that for each kind of particles (different hydrogen ions, electrons, impurities) there
is a pair of equations (2.84) and (2.86). Next we may calculate an energy equation which
describes energy changes in the plasma (heat generation, heat convection and so on) by
multiplying (2.44) by (v0 − v)(v0 − v) and integrating over velocity space.

2.7.2 Magnetohydrodynamic model (MHD)

Considering the problems of enormous complexity facing early fusion researchers, it was
important to develop as simple, but still reliable, plasma models as possible. Already in
the 1950s the so called magnetohydrodynamic (MHD) model was derived. The ideal MHD
model indicates that the resistivity of the plasma is neglected; the plasma is considered an
ideal conductor. We can easily obtain the ideal MHD equations from the fluid relations
we have already discussed.
The first step is to bring the separate electron and ion fluids to a single fluid. We
introduce the one-fluid velocity
mi vi + me ve
v≡ ≈ vi (2.87)
mi + me
√
since me /mi = 5.4 × 10−4 /A and | me ve |  | mi vi | (| vi | / | ve | = 2.3 × 10−2 / A for
thermal velocity), where A is the mass number of the nucleus. Further, we construct the
mass density ρ = nmi (where n = ne = ni ) and for the equation of continuity, we have

∂ρ
+ ∇ · (ρv) = 0 (2.88)
∂t

61
By adding the equations of motion for electrons and ions we obtain the equation of motion
for the fluid
∂v

ρ + (v · ∇)v = j × B − ∇p (2.89)
∂t
Where we have used j = e(ne ve + ni vi ) ≈ enve and assumed isotropic plasma pressure.
Now, we need an equation for p. In principle, this can be found from the equation of
energy discussed earlier. Even early researchers, however, realized the importance of the
special case which is also present in classical thermodynamics, i.e. adiabatic behaviour.
Here, the dissipation is neglected, and the time is assumed too short for the small plasma
elements to be able to exchange heat with their surrounding. This gives the relation (the
adiabatic index Γ = Cp /CV , which is often assumed to be 5/3)
d p

= 0. (2.90)
dt ρΓ
Ohm’s law (here in simplified form) for a plasma is;
E + v × B = ηj. (2.91)
In ideal MHD the resistivity is equal to zero. Equation (2.91) then expresses that the
electric field in a (Lorentz) system which follows the motion always is zero because of the
infinite conductivity of the plasma. To close the system of equations, we add Maxwell’s
equations (the displacement current is neglected):
∇·B=0 (2.92)
∂B
∇×E=− (2.93)
∂t
∇ × B = µ0 j. (2.94)
Let us summarize the ideal MHD equations:
∂ρ
+ ∇ · (ρv) = 0 E+v×B=0
∂t
∂v
∂B
ρ + (v · ∇)v = j × B − ∇p ∇×E=− (2.95)
∂t ∂t
d p

=0 ∇ × B = µ0 j
dt ρΓ
This is a simplified model, but it still cover many important phenomena in a fusion plasma.
Even in this simple model, we have to handle 14 scalar coupled nonlinear time dependent
differential equations in three space dimensions!
We can expand the continuity equation to obtain an equation for the time rate of
change of the density as follows:
∂ρ
+ (v · ∇)ρ = −ρ∇ · v (2.96)
∂t
The left hand side of the equation is the convective time derivative of the density, which
represents the time rate of change of the density of a fluid element that follows the flow of
the fluid. The term on the right hand side represents the time rate change of the density
in that fluid element due to compression or expansion of the fluid. We may more simply
write the continuity equation as follows:
dρ
= −ρ∇ · v (2.97)
dt
62
The equation for the pressure may be expanded as follows
dp Γ dρ
=p (2.98)
dt ρ dt
In the equation for the pressure, the time derivative must be understood as the convective
time derivative. The equation can be combined with the density equation to obtain
dp
= −Γp∇ · v (2.99)
dt
Again the right hand side of the equation represents the time rate of change of the pressure
in a fluid element due to compression or expansion. In an incompressible fluid, the terms
on the right and side of the density and pressure equations would be zero. One important
simplification can, in some cases, be the assumption of incompressibility.
The electric field can be eliminated from the system using Faraday’s law. Combining
Faraday’s law
∂B
∇×E=− (2.100)
∂t
and the ideal-MHD Ohm’s law with zero resistivity,

E+v×B=0 (2.101)

we have
∂B
= ∇ × (v × B). (2.102)
∂t
The ideal MHD equations can then be written in the alternative form
∂ρ
+ (v · ∇)ρ = −ρ∇ · v
∂t
∂v

ρ + (v · ∇)v = j × B − ∇p
∂t
∂p
+ (v · ∇)p = −Γp∇ · v
∂t
µ0 j = ∇ × B
∂B
= ∇ × (v × B) (2.103)
∂t

2.7.3 Advanced fluid models∗

The system of equations (2.95) is, as mentioned, a simplified special case of the general
fluid equations reached from the Fokker-Planck equation. Some of the neglected terms
may, however, be needed. This is for example the case if the energy transport of the
plasma is to be studied. This transport determines the energy confinement time which,
in turn, is a measure of the quality of the fusion plasma.
Already in the early 60’s, Braginskii managed to calculate the elements in the pressure
tensor (ordinary and magnetic viscosity) and gave expressions for heat flux and heat
exchange. He assumed that the plasma particles were Maxwell distributed and that the
local approximation (small characteristic length scales) was valid. We shall, however, not
go into this extensive analysis.
Finite Larmor radius corrections may also be included. These are contributions to the
pressure tensor (magnetic viscosity) and to Ohm’s law (Hall term and diamagnetic effect)
and are, for example, important for studies of certain unstable plasma configurations.

63
Chapter 3

EQUILIBRIUM

3.1 Cylindrical equilibrium

In 1934, Willard Bennett studied a straight, circular cylindrical discharge. In this, a
current is run through a suitable low-pressure gas in a vacuum chamber between two
electrodes as shown in the figure below.

Figure 3.1: The principle of a linear pinch.

While the current heats the plasma through resistive losses, it also counteracts the ex-
pansive force of the plasma column. Here, as elsewhere, we will assume that there is at
least one state available for the plasma, which does not change over time. We shall call
this the equilibrium state. Bennett’s equilibrium equations can be obtained from ideal
MHD if we assume a stationary state, i.e. if all time derivatives are set to zero. If we, like
Bennett, also assume that the equilibrium is static (v = 0) we determine the equilibrium
equation
∇p = j × B. (3.1)
The current j can be eliminated using Ampére’s law

∇ × B = µ0 j. (3.2)

One more scalar equation is needed, and

∇·B=0 (3.3)

does the job. Equations (3.1)-(3.3) describe the equilibrium of a fusion plasma, and their
validity can be shown to be very general. Even in more advanced plasma models, the

64
terms in the system (3.1)-(3.3) are dominating. Thus, there is ample reason to study the
different solutions to the equilibrium equations.
Bennett’s pinch (named after the confining j × B force) has a circular cylindrical
geometry. In cylindrical coordinates, (r, θ, z) we have ∂/∂θ = 0 = ∂/∂z. In equation
(3.1) only the r component remains non-trivial:
dp
= jθ Bz − jz Bθ . (3.4)
dr
Further (3.2) gives
1 dBz
jθ = −
µ0 dr
1 d
jz = (rBθ ) (3.5)
µ0 r dr
Combining (3.4) and (3.5) we obtain the equilibrium equation for a circular cylinder ;
dp Bθ d Bz dBz
=− (rBθ ) − . (3.6)
dr µ0 r dr µ0 dr
Bennett studied a z pinch, in which Bz = 0. By multiplying (3.6) by r2 and integrating
from r = 0 to r = a, assuming p(a) = 0, he obtained the Bennett relation
µ0 Ip2 = 16πN eT̄ (3.7)
where the line density is
Za
N= 2πrndr (3.8)
0
and the Bennett temperature is
Za
1
T̄ = 2πrnT dr. (3.9)
N
0

Here the ideal gas law, combining electrons and ions, p = 2neT has been used. Equation
(3.7) is an integral form of the equilibrium equation, which in a clear way shows the
relation between a plasma current, density and temperature in a z-pinch. We see, for
example, that for N = 1020 m−1 and T = 10 keV the current I = 2.5 MA is required.
Is it difficult to solve the equilibrium equation (3.6)? It depends. If the magnetic
fields are specified as functions of r, only one integration is needed to calculate p(r) and
that would be it. If one of the magnetic fields must be calculated, the analytic integration
usually becomes less straightforward. It is clear, however, that two functions must be
given if the equilibrium equation is to be solved. Now, if (3.6) has several solutions,
which one will the plasma “choose”? The answer will come from a discussion of the
stability of the fusion plasma.
Circular cylindrical fusion configurations are problematic. Since the plasma is “open”
at the ends, the plasma suffers large losses of particles and energy. Further, the electrodes
at the edge usually cause vast amounts of contamination entering the plasma. Finally,
the z pinch and most other 1-D “screw pinches” (when Bz in included, the field lines form
a screw in the axial direction) are unstable. Thus, there is reason to close the ends of the
plasma to form a circular configuration, a torus. Consequently, we must also generalize
the equilibrium equation to two dimensions.

65
3.2 Beta value
The beta parameter can be defined in different ways. The two basic definitions are poloidal
β and toroidal β. The poloidal β is defined as
<p>
βp ≡ (3.10)
Bθ2 (a)/(2µ0 )
R
where the plasma pressure p is the average over the plasma volume; < p >= pdV /V .
For a straight circular cylindrical equilibrium, we can calculate
R Bθ (a) = µ02I/(2πa) and p
can be calculated by integrating eq. 3.6. If we write < p >= 2πrpdr/πa and integrate
by parts, we have
Za
1 dBφ2 2
βp = 1 + 2 r dr. (3.11)
a Bθ (a)2 dr
0

which can be written as

2
Bφa − < Bφ2 >
βp = 1 + 2
(3.12)
Bθa
Here we have introduced Bφ = Bz , and Bφa and Bθa denotes the toroidal and poloidal
fields at the plasma boundary, respectively.
How large is then βp ? As we will see, the plasma equilibrium often involves poloidal
plasma current, which produces a plasma generated toroidal field contribution to the ex-
ternal toroidal field inside the plasma. The plasma generated toroidal field can either
enforce or weaken the externally applied toroidal field. If the plasma equilibrium is para-
magnetic, enforcing the external toroidal field, we have < Bφ2 > > Bφa 2
, and βp < 1. If
the plasma equilibrium is diamagnetic, weakening the external toroidal field, we have
< Bφ2 > < Bφa 2
and βp > 1. It turns out that tokamak equilibria typically have βp > 1,
but is of order unity. In a reversed-field pinch (RFP) equilibria, the magnitude of the
external (reversed) toroidal magnetic field is small Bφa  Bθa , and as result the RFP has
βp < 1. A typical value for an RFP is βp ≈ 0.1. For a z pinch (no axial magnetic field)
we thus have βp = 1, which is also the case when there is a homogeneous axial field.
In a tokamak, the cross section is generally not circular and (3.6) is not directly
applicable. For tokamak equilibria the toroidal beta value
<p>
βt ≡ (3.13)
B02 /(2µ0 )

is instead often used, where B0 is the toroidal field on the magnetic axis r = 0.

3.3 Toroidal equilibrium

Control of the radial position is a fundamental requirement of operation of toroidal devices
such as a tokamak. A toroidal plasma column, like an inflated balloon, tends to extend
by increasing its major radius. The radial force can partly be associated with a net
resulting poloidal magnetic field pressure on the inboard side due to the higher strength
of the inboard field than the outboard, as indicated in figure 3.2. The extension force is
balanced by the force of interaction between the toroidal plasma current and an external
vertical field. The balancing effect can also be explained as a removal of the poloidal
field inboard to outboard asymmetry by the addition of the vertical field, which is in the

66
 8R
F = µo /(ln + � - 1) . (7.221)
l
2 2
FB2 =
h a
The outward force FP exerted by the plasma pressure is The f
F1 =

•
B Balan

111 l)9 µo l/
�

+ =
and t

Fig. 7.10
Poloidal magnetic field due to the combined plasma current and vertical magnetic field. B .1=
Figure 3.2: Plasma expansion
106
force balanced by an external vertical field. (From Plasma
Plasma Equilibria
Physics for Nuclear Fusion, K. Miyamoto, The MIT Press, 1980, p. 186)

�6
I
R

�------

_i}_ ___ +---�

I
Ro

for calculation of radial equilibrium.Fig.(From

Figure 3.3: Geometry used Major 5-3 Fusion of (R, z, 0) and
Relation Plasma
Analysis, W. Stacey, J. Wiley
axis& Sons, 1981, p. 106) toroidal ( r, 0, cf>) coordinate systems.

opposite direction to the and

poloidal field on the inboard side, thus reducing the poloidal
field on the inboard side and increasing it on the o outboard f side. R
B v= _!_ ite = µo (1n 8 o +A-.!.), r>a, (5-59)
When the plasma current and the toroidal z R oRfield is given,
41rR the
a position 2 of the plasma
still depends on the kinetic energy of the plasma and the form of current distribution
whereDuring
over the column cross-section. we haveheating
made useof of
thethe relationships
plasma, both (see
theseFig. 5-3)
characteristics
change, and as a result the plasma column is displaced in the vessel. The fact that the
R=R 0 +rcos0
equilibrium position depends on the plasma parameters introduces the requirement for
control of the position, through the active feedback control z =of
rsinO.
the applied external vertical (5-60)
field.
We now use this model to examine the stability of the plasma equilibrium
to vertical (z direction) and horizontal (R direction) axisymmetric displace­
3.3.1 Equilibrium ments. vertical
If the field
vel)ical field lines(due to the external coils) are slightly concave
The amount of vertical fieldtowards the for
required major axis, the interaction
equilibrium can be derivedof_the_asR-�omponent of the field with
follows: Consider
the current will produce a restoring force F~IX BR that opposes the displace­
a large aspect ratio torus of circular cross section, with minor radius a and major radius
ment. The requirement that the field be concave can be written as a require­
R0 , see figure 3.3. ment upon the field index, n(see Fig. 5-4)
Hoop force R aB 0
The magnetic force from the poloidal magnetic field nacting the>0. (5-61)
=- BvonoR current hoop (ring)
causing it to extend is called the hoop force The magnetic energy associated with the
toroidal plasma current is When the plasma is vertically elongated, instead of circular, the require­
ment for stability against
1 vertical displacement is more complex. For an
W = LI 2 (3.14)
elliptical plasma cross 2 section and a unform current distribution, the height to
where L is the inductancewidth
of theratio, b/a, isring.
current related to the
If the fieldforce
hoop index,Fn, extends
h
in such athe
waymajor
that satisfaction
of Eq. 5-61 requires
radius R0 of the current ring by a small radial increment dR0 , while keeping the current
� -1
a R
2
4
<(!!...) (I
ln SR - .!2),
a 16
(5-62)

67
I constant, the change in magnetic energy is

dW = Fh dR0 (3.15)

The radial hoop force is calculated by differentiating the magnetic energy

  
∂W ∂ 1 2
Fh = = LI (3.16)
∂R0 ∂R0 2 I=const

Here L is the sum of the internal and external self inductance of the current ring:

L = Le + Li (3.17)

The internal self-inductance is obtained from the energy definition:

1
Wi = Li I 2 (3.18)
2
where Wi is the energy in the poloidal magnetic field in the plasma region r < a. By
integration of the poloidal magnetic field energy density Bθ2 /2µ0 over the toroidal plasma
volume (dV = 2πR0 rdθdr), and using Ampere’s law (2πaBθa = µ0 I) the internal self-
inductance is obtained as
1
Li = µ0 R0 li (3.19)
2
where the normalized internal self-inductance is defined as
Z 2π Z a
1
li = 2 2 Bθ2 rdθdr (3.20)
πa Bθa 0 0

It is easily verified that a constant current density corresponds to li = 1/2. The external
self-inductance of a current ring with minor radius a and major radius R0 is obtained by
integrating the poloidal magnetic flux that passes through the center of the ring, crossing
the circular area with radius R = R0 − a
   
8R0
Le = µ0 R0 ln −2 (3.21)
a
and the total inductance is thus
   
8R0 li
L = µ0 R0 ln + −2 (3.22)
a 2
Accordingly, the hoop force is
µ0 I 2
   
8R0 li
Fh = ln + −1 (3.23)
2 a 2

Plasma pressure
There is an additional outward force Fp exerted by the plasma pressure. Consider a thin
sector of the toroidal plasma ring, as shown in figure 3.4: The plasma pressure p on both
sides of the sector will result in a net radial outward force. For a plasma ring with minor
radius a the total force on one side of the sector (in the toroidal direction) is

Fpφ = πa2 < p > (3.24)

68
Figure 3.4: Plasma expansion force balanced by an external vertical field. (From Plasma
Physics for Nuclear Fusion, K. Miyamoto, The MIT Press, 1980, p. 187)

where < p > is the average plasma pressure. If the sector has a toroidal angular extension
dφ, the resultant outward radial component of the forces from both sides of the sector is
 
dφ
dFp = 2Fpφ sin ≈ Fpφ dφ (3.25)
2

The total radial force is then

Z 2π
Fp = Fpφ dφ = 2πFpφ = 2π 2 a2 < p > (3.26)
0

Internal toroidal field

There are tension forces in the toroidal direction on the plasma sector at the poloidal
cross-sections from the toroidal field
< Bφ2 >
FB1φ = −πa2 (3.27)
2µ0

where < Bφ2 > is the average toroidal field in the plasma. This results in an inward total
radial force
< Bφ2 >
FB1 = −2π 2 a2 (3.28)
2µ0

External vacuum toroidal field

Finally the external toroidal magnetic field exerts an inward magnetic pressure on the
plasma sector, producing a inward force on a surface element dA of the sector:
2
Bφv
dFB2r = − dA (3.29)
2µ0
The component of the force in the major radius direction is

dFB2 = dFB2r cos θ (3.30)

The surface element is

dA = aR dφ dθ (3.31)

69
The toroidal external field produced by the toroidal field windings is stronger on the
inboard side than the outboard side, varying with R as
R0
Bφv = Bφv0 (3.32)
R
which results in a net outward force on the plasma sector. By writing R = R0 + a cos θ
(figure 3.3), and using the Taylor expansion
 
1 1 a
≈ 1− cos θ (3.33)
R R0 R0
the outward force on the plasma ring becomes
2
B2
Z 2π
Bφv0
 
a 2 2 φv0
FB2 = − 2π R0 1 − cos θ a cos θ dθ = 2π a (3.34)
2µ0 0 R0 2µ0

Total radial force

The expansion force of the plasma ring due to pressure and the toroidal field becomes
Fβ = Fp + FB1 + FB2 =
2
Bφv < Bφ2 >
 
2 2
= 2π a < p > + −
2µ0 2µ0 (3.35)
2
 
B 1
= 2π 2 a2 θa βp −
µ0 2
where eq. 3.12 for the poloidal beta value βp have been used. The poloidal field at the
boundary is related to the plasma current by Ampere’s law, giving
µ0 I 2
 
1
Fβ = βp − (3.36)
2 2
and the total radial expansion force of the plasma is
µ0 I 2
   
p 8R0 li 3
FR = Fh + Fβ = ln + + βp − (3.37)
2 a 2 2

Equilibrium vertical magnetic field

The plasma outward radial force has to be balanced by an inward radial force from a
external vertical magnetic field
FRv = 2πR0 IBzv (3.38)
which is determined by the force balance condition

FR = FRp + FRv = 0 (3.39)

This gives the required vertical magnetic for equilibrium

   
e µ0 I 8R0 li 3
Bz = − ln + + βp − (3.40)
4πR0 a 2 2

Stability condition
The additional requirement of stability of the plasma equilibrium to vertical and horizontal

70
 Z
B
dFZ =Ip×BR dl
BR Ip
dFZ
dFZ R
Ip
BR

Figure 3.5: Illustration of vertical restoring force from concave vertical equilibrium field.

displacements gives restrictions on the radial variation of the external vertical field. If the
vertical field lines are slightly concave towards the major axis, the interaction of the R
component of the field and the plasma produce a vertical restoring force, see figure 3.5.

Fzv = 2πR0 IBRv (3.41)

The curvature of the external vertical field is described by the field index n, defined as
R ∂Bzv
n=− v (3.42)
Bz ∂R
The requirement that the field be concave for vertical stability gives the condition that
the field index must be positive
n>0 (3.43)
Considering radial displacements around the equilibrium major radius R0 , stability
requires that the net radial force is inward for an outward displacement, and vice versa.
This condition can be written as
∂FR
< 0. (3.44)
∂R0
It can be shown that this condition leads to the requirement
3
n< . (3.45)
2
In summary, the field index of the applied external vertical field has to fulfil
3
0<n< , (3.46)
2
to ensure simultaneous stability to both vertical and radial displacements.

3.4 Magnetic flux surfaces

In a very slim (R/a  1) torus with a circular cross section, the equilibrium may be
described by (3.6), but experience tells us that only a few of these configurations are
interesting. The tokamaks of today have instead been developed as low aspect ratio
machines, with R/a = 1–3, where the cross section usually is highly non-circular. (see
figure 3.2). The two-dimensionality is caused by the fact that flux surfaces no longer have

71
 Figure 3.6: Magnetic flux surfaces.

r = const. In order to formulate a useful equilibrium equation in two dimensions, we need

a new way to describe the flux surfaces. This is achieved with the flux functions.
The force balance (3.1) tells us a great deal about the plasma. The plasma pressure
p is constant both on magnetic flux surfaces and on the surfaces where the current j are
flowing, since

B · ∇p = 0
j · ∇p = 0 (3.47)

Thus, the current is flowing along the magnetic flux surfaces. These are defined by the
magnetic flux due to the poloidal magnetic field Bp , confined from the magnetic axis r = 0
and outwards in the radial direction, calculated as
ZZ ZZ I
Φ= Bp · dS = (∇ × A)p · dS = {Stoke’s law } = A · dl = 2πRAϕ (3.48)
S S Γ

We have introduced coordinates (R, ϕ, z) according to figure 3.3. The magnetic flux
inside each flux surface must be the same everywhere along the surface, and may thus be
characterised using Φ.

3.4.1 Flux functions∗

Generally, the function describing the flux per radian is used instead, ψ ≡ Φ/2π, and we
have  
1 ∂ψ 1 ∂ψ
Bp = (∇ × A)p = ∇ × (0, ψ/R, 0) = − , 0, . (3.49)
R ∂z R ∂R
We see that    
1 ∂ψ 1 ∂ψ ∂ψ ∂ψ
B · ∇ψ = − , Bϕ , · , 0, =0 (3.50)
R ∂z R ∂R ∂R ∂z
which is expected when we have axial symmetry (∂ψ/∂ϕ = 0). Here ψ is a flux function
for the poloidal magnetic field. We will also use a flux function for the poloidal current.
Such a flux function, called χ, must exist due to symmetry between j and B according to
(3.47). Thus, we reach the following ansatz for the poloidal current density:
 
1 ∂χ 1 ∂χ
jp = − , 0, (3.51)
R ∂z R ∂R

72
 z

R
ϕ

Figure 3.7: Coordinate system for 2-D geometry in R, z. Axial symmetry (∂/∂ϕ = 0) is
assumed.

Further, according to Ampére’s law ∇ × B = µ0 j we have

 
1 ∂Bϕ 1 ∂(RBϕ )
jp = − , 0, . (3.52)
µ0 ∂z µ0 R ∂R

since only the magnetic field component Bφ is associated with a poloidal current. Hence,

RBϕ
χ= . (3.53)
µ0
It can be shown that χ, just like p, is a function only of ψ.

3.4.2 Grad-Shafranov equation∗

The flux functions above will now be used to derive the two dimensional equilibrium
equation for an axially symmetric plasma torus. It is named the Grad-Shafranov equation
after its originators. This is one of the most important equations in fusion plasma physics.
The idea is to express jϕ in two different ways, and set the resulting equations equal
to each other. First, we shall use the equilibrium equation (3.1). If we divide it between
poloidal and toroidal direction, it can be written as

jp × eϕ Bϕ + jϕ eϕ × Bp = ∇p (3.54)

where eϕ is the unity vector in the toroidal direction. Then, we use the flux functions to
eliminate Bp and jp ;
 
1 ∂ψ ∂ψ 1
Bp = − , 0, = (∇ψ × eϕ ) (3.55)
R ∂z ∂R R
 
1 ∂χ ∂χ 1
jp = − , 0, = (∇χ × eϕ ) (3.56)
R ∂z ∂R R
Introducing this into (3.54) gives

1 1 Bϕ jϕ
(∇χ × eϕ ) × eϕ Bϕ + jϕ eϕ × (∇ψ × eϕ ) = − ∇χ + ∇ψ = ∇p (3.57)
R R R R

73
since eϕ · ∇ψ = 0 = eϕ · ∇χ. The chain rule then gives
dp
∇p(ψ) = ∇ψ
dψ
dχ
∇χ(ψ) = ∇ψ (3.58)
dψ
Inserting this into (3.57), we can separate jϕ ;
dp dχ dp µ0 dχ
jϕ = R + Bϕ =R + χ . (3.59)
dψ dψ dψ R dψ
We now rewrite jϕ using Ampére’s law (3.2):
∂BpR ∂Bpz 1 ∂ 2ψ ∂ 1 ∂ψ

µ0 jϕ = − =− − . (3.60)
∂z ∂R R ∂z 2 ∂R R ∂R
Finally, combining Eqs.(3.59) and (3.60) we obtain the Grad-Shafranov equation
∂ 2ψ 1 ∂ψ ∂ 2 ψ dp dχ
2
− + 2 = −µ0 R2 − µ20 χ . (3.61)
∂R R ∂R ∂z dψ dψ
This is a so called generalized partial differential equation; ψ is present both as a dependent
variable on the left hand side and as a independent variable on the right hand side! In
order to solve the Grad-Shafranov equation, the functions p = p(ψ) and χ = χ(ψ) are
first specified. Generally, the Grad-Shafranov equation is difficult to solve analytically.
Because of this, a large number of numerical codes have been developed.
An analytical equilibrium solution ψ(R, z) has been found by Solov’ev. By assuming
the following dependencies for the functions p = p(ψ) and χ = χ(ψ);
dp
µ0 = −a
dψ
dχ
µ20 χ = −bR02 (3.62)
dψ
where a, b and R0 are constants, the Grad-Shafranov equation can be solved analytically,
resulting in the following expression for ψ(R, z):
ψ(R, z) = 1
2
(bR02 + cR2 )z 2 + 81 (a − c)(R2 − R02 )2 . (3.63)
Here, R0 is the major radius of the magnetic axis, and c is a constant. Solov’ev’s equilib-
rium is mainly useful for testing equilibrium codes, due to its restricted modelling of the
current density jφ .
Then, why is the Grad-Shafranov equation important? In order to answer that, it is
useful to think about the meaning of magnetic confinement. The toroidal plasma must be
completely balanced by a magnetic field, so it does not drift towards a wall and in a way
which conserves its stability. As we will find in the next chapter, stability problems have
been dominating fusion research over the years. The Grad-Shafranov equation can be
used to calculate the design of the poloidal and toroidal coils balancing the plasma, and
the currents required in these. Just like the 1-D case, two functions (besides geometry
and boundary conditions) must be specified if the equation shall be solved. By studying
different dependencies for p(ψ) and χ(ψ) the stability can be investigated and optimized.
Then, when the plasma current, temperature, density and so on can be measured, more
precise solutions to the Grad-Shafranov equation can be calculated and used for a stability
analysis of the equilibrium.

74
Chapter 4

INSTABILITIES AND WAVES

In the previous chapter, we showed that a confined plasma may, in principle, be in a

state of magnetohydrodynamic equilibrium. A perfect thermodynamic equilibrium can,
however, never be achieved since the plasma is neither homogenous nor isotropic. A
confined plasma must have a density gradient at the boundary. Thus, there is a natural
driving force, constantly working to relax the plasma towards a state with lowest possible
potential energy, and at the same time causing instabilities.
Instabilities are not only caused by density gradients, but also by temperature and
current density gradients. These are generally undesired, since they usually are uncon-
trolled and can, very quickly, force the plasma against the wall or in other ways cause
large power losses. To avoid this, the confining magnetic field must have a stabilizing
effect, and this is our next topic.

4.1 Stabilizing magnetic fields

The force equilibrium equation (3.6) for a 1-D cylinder can be written in vector notation
as
Bθ2 + Bz2 B2
 
∇ p+ = θκ (4.1)
2µ0 µ0
where the magnetic field line curvature vector κ has been introduced. It is defined as
follows
κ ≡ (b · ∇)b. (4.2)
The vector b is a unit vector in the direction of the azimuthal magnetic field, which in
this case is simply the unit vector θ̂θ in the cylindrical coordinate system.
Bθ
b= = θ̂θ (4.3)
| Bθ |
The curvature vector is then obtained as
1 ∂ θ̂θ r̂
κ = (θ̂θ · ∇)θ̂θ = =− (4.4)
r ∂θ r
The curvature vector is pointing radially inwards to the center of curvature of the field
line, and the magnitude is inversely proportional to the field line curvature radius r.
The magnetic field can assert a force on the plasma in two ways. There is a ”pressure”
associated with the magnetic field, which acts on the plasma. If the plasma moves towards
the magnetic field, causing the field to become compressed, the field will create an opposing

75
magnetic pressure reacting back on the plasma. There is also a ”tension” in the magnetic
field, which can create a force on the plasma. If the field lines are stretched out by
the plasma, the field will produce a force that reacts back on the plasma, opposing the
stretching. It is sometimes helpful to think of the magnetic field lines as elastic bands.
The forces acting on the plasma in the cylindrical equilibrium can be described as
follows: The radial variation (gradient) of the combined kinetic and magnetic pressures
produces ”pressure gradient forces” which, in case the plasma pressure is peaked on axis,
and the magnetic pressure is higher on the outside, can provide the required force equi-
librium. One such example is a plasma cylinder in an external axial magnetic field Bz . In
a cylinder with an axial plasma current, there is an azimuthal field Bθ produced by the
current. In this case the plasma and magnetic field pressures are not perfectly balanced,
and there is a net radially outward force. The force equilibrium is in this case main-
tained by a radially inward force produced by the ”tension” in the azimuthal magnetic
field lines. One can imagine the azimuthal magnetic field lines as rubber bands that are
wound around the plasma cylinder and acts to compress it radially.
If the plasma would have moved towards a surrounding wall, repelling currents may
have been generated, which would also be stabilizing. Intuitively, this can be seen as if the
magnetic fields lines between the plasma and the wall are compressed when the plasma is
dislocated towards the wall, thus yielding an opposing magnetic pressure. The magnetic
field may also be stabilizing if it has a convex curvature in the direction of the plasma;
we will discuss this later, in connection to pressure driven instabilities. But let us first
return to the important stabilisation due to field line bending.
For sake of simplicity, we consider a case where the equilibrium magnetic field lines
are straight. If the plasma, for some reason, would be disturbed from the equilibrium
state, and the perturbation would cause a bending of the magnetic field lines, a force
arises due to the stretching of the field, which is stabilizing, acting to straighten the field
lines, returning them back to the original state. Think of the magnetic field lines as
elastic bands, striving to be as straight as possible, and to return to their equilibrium
state after a disturbance. Imagine a plasma, originally at equilibrium, being displaced
somewhat. Consider a small volume of plasma with the equilibrium position r0 . Let
us call the displacement ξ ≡ r − r0 , where r is the new position. It is interesting to
study the change of the magnetic field due to the plasma displacement. We now write
B = B0 + B1 , where index 0 represents the equilibrium field and index 1 represents the
field at the perturbed position. Thus, we have linearized the magnetic field B about the
equilibrium position. In the same manner we may linearize the velocity field about the
static equilibrium position; v = v0 + v1 = v1 . If we introduce these expressions in eq.
2.102, 0:th order terms yield a trivial relation, and first order terms yield
∂B1
= ∇ × (v1 × B0 ). (4.5)
∂t
Since we are interested in B1 , and since v1 = ∂ξξ /∂t, we can integrate eq. 4.5 over time
to obtain
B1 = ∇ × (ξξ × B0 ). (4.6)
In order to illustrate the stabilization effect of field line bending, we can assume a homo-
geneous equilibrium field B0 . The, nabla operator in eg. 4.6 then only operates on the
perturbation ξ , and the equation can be written as

B1 = (B0 · ∇)ξξ − B0 ∇ · ξ (4.7)

76
Then we make a plane wave ansatz, which introduces the wave vector k of the perturba-
tion, pointing in the direction of the wave propagation:

ξ ∼ exp(i(k · r − ωt)) (4.8)

We may then replace ∇ → ik, transforming the partial differential equation to a simpler
algebraic equation:
B1 = i(k · B0 )ξξ − i(k · ξ )B0 . (4.9)
This equation is interesting. It shows that if the direction of the wave propagation is
perpendicular to the equilibrium magnetic field (k · B0 = 0), the perturbed field must
be parallel to the equilibrium field (B1 kB0 ). Hence, the perturbation does not bend the
magnetic field, and there is no reacting magnetic force due to stretching of the field. As
a result, the magnetic field’s defence against instability is weakened! It turns out that
perturbations with k · B0 = 0 are the most dangerous from the stability point of view.

4.2 The safety factor

Let us investigate what k · B0 = 0 means in relation to the equilibrium. Consider an
equilibrium in toroidal geometry. Since both the poloidal and the toroidal directions are
periodic, Fourier expansions are used to express a perturbation quantity Q1 :
X X
Q1 (r, t) = Q1mn (r) exp(i(mθ − nϕ − ωt)) ≡ Q1k (r) exp(i(k · r − ωt)). (4.10)
m,n k

With this notation, we have k ≡ (0, m/r, −n/R) and r ≡ (0, rθ, Rϕ) ; the periodicity
requires 2πr = m(2π/kθ ), or kθ = m/r and similarly for kϕ . Then, we can write
m n
k · B0 = 0 ⇔ Bθ − Bϕ = 0 (4.11)
r R
We now introduce a very important concept, the safety factor q:
rBϕ
q≡ . (4.12)
RBθ
The relation 4.11 tells us where in the plasma localized instabilities might arise; on flux
surfaces where q is a rational number.
m
q= . (4.13)
n
This rational number, m/n, is defined by the rotation of the perturbation along the
magnetic field. Thus, the condition (4.11) can be thought of as a resonance between the
perturbation and the equilibrium magnetic field.
But where in the plasma are the flux surfaces described in (4.13)? It turns out that
in most of the tokamak plasma q > 1, increasing monotonically radially outwards. This
means that the most global perturbation (m, n) = (1, 1) does not satisfy (4.13), which is
very advantageous from a stability point of view. However, close to r = 0, there is a small
region where we may have q < 1, and consequently there is a flux surface where q = 1 and
a resonance for the (m, n) = (1, 1) instability is possible. This instability indeed appears,
and it has a periodic growth and decay manner, which gives it it’s name, the sawtooth
instability.

77
 It can further be concluded that no (1, n) mode, where n ≥ 2, can be resonant in the
tokamak plasma. This leads to good stability, since instabilities with low m, n usually
have the highest energy and thus are the most dangerous (the stabilizing effect of the
magnetic field increases with m and n). We will shortly study the instabilities which are
relevant for the tokamak. First, however, we will study q in a geometric context.
Differential geometry tells us that the equation for a field line can be written as
ds Bθ
= , (4.14)
Rdϕ Bϕ

where ds and Rdϕ are the poloidal and toroidal distances of an infinitesimal segment along
the magnetic field. The total change in ϕ as we make a complete turn in the poloidal
direction is I
Bϕ ds
∆ϕ = , (4.15)
RBθ
which, for a circular cylindrical plasma with R/a  1 is

rBϕ
∆ϕ = 2π = 2πq. (4.16)
RBθ
It is obvious that q = ∆ϕ/(2π) can be interpreted as the number of turns along a magnetic
field line in the toroidal direction (the ”long way”) for each revolution in the poloidal
direction (the ”short way”).
Now it is easier to realize why (4.11) is considered a resonance condition; when the
rotation of the perturbation exactly matches the rotation of the magnetic field, q becomes
a rational number. A physical interpretation of the instability risk associated with the
resonance is that the field lines on the resonance surface close after a few turns (fewer turns
if m and n are small). This means that any electrical field that arises because of some
perturbation is difficult to extinguish; the electrons are tied to field lines which do not
cover the flux surface. The electrical field can then act as a driving force for the unwanted
plasma movement we call instability. If q had been a non-rational real number, the field
lines, and thus the electrons, could have covered the whole flux surface ergodically.
In a general geometry, the safety factor should be calculated from
I
1 Bϕ ds
q= . (4.17)
2π RBθ
Since the integration follows a poloidal turn along a magnetic field line, q is a flux surface
function. A high q is desired from a stability point of view, and can be achieved for
example if the toroidal field is high or the major radius is small.
In some configurations, such as the stellarator, the term rotation transform ι, which
is related to the safety factor by ι ≡ 2π/q is often used.

4.3 How can we analyze instabilities?

The insights above can help us find a way to analyze instabilities in a plasma. Let us
summarize.

• Driving forces for instabilities are density or temperature gradients, or pressure

gradients. Currents and current gradients may also drive instabilities.

78
 • Stabilizing effects come from the plasma itself (magnetic field bending, curvature
and compression) and from a conducting wall.

• Ideal instabilities arise even if the magnetic field is perfectly conducting and not
destroyed anywhere.

• Resistive instabilities (or modes) can arise when the plasma is not perfectly con-
ducting.

• Rational surfaces for q are the flux surfaces which cause most problems for the
stability. This is true both for ideal and resistive modes.

We want to use a plasma model which includes the effects discussed above and at
the same time is simple. The simplest, physically reasonable, theory is the one-fluid
model ideal MHD (and resistive MHD if resistivity is included in Ohm’s law). But the 14
coupled nonlinear differential equations in three space dimensions and one time dimension
are usually far too complicated to handle. In this case, we can make it easy for ourselves
by studying the linear stability.
In linear stability analysis, the plasma is assumed to have attained some equilibrium
state. As shown earlier, a small displacement ξ (r) of the plasma, away from equilibrium,
is performed. In a stable configuration, the plasma returns to equilibrium. In other cases,
the equilibrium is unstable. The different cases in figure 4.1 provide intuitive illustrations
of this.

Figure 4.1: A ball in different stable/unstable situations:(a) stable, (b) unstable, (c)
marginally stable, (d) stable with dissipation (e) unstable with dissipation, (f) linearly
stable, non-linearly unstable, (g) linearly unstable, non-linearly stable

Four methods are used to study the stability. The first three can often be performed
analytically.

1. The energy principle is a way to investigate how the potential energy of the plasma
is affected by a perturbation.

79
Figure 4.2: Unstable motion in a tokamak, calculated from the eigenvalue equations for ideal
MHD. (Figure credit: By permission of Oxford University Press)

2. The eigenvalue problem for a certain perturbation can be expressed and solved using
the MHD equations. This results in eigenfrequencies, which tell us the growth rate
of the instability and also the eigenfunctions. One example of such a solution for
tokamaks can be seen in figure 4.2.

3. Marginal stability can be investigated by assuming that the growth rate is zero for
a given mode (m, n).

4. The initial value problem can be solved by numerical integration of the time devel-
opment after a given perturbation of the plasma.

Finally, it is important to note that ideal MHD stability almost always must be fulfilled,
while fluctuations caused by other instabilities often are present in the plasma.

4.4 Waves in an MHD plasma

A perturbation of a plasma in equilibrium may, as in figure 4.1, cause a number of different
behaviour. In an ideal MHD plasma, the perturbation results in either instability or in
an undamped wave. A damped or growing wave motion cannot be described; in this
case, dissipative effects such as resistivity of viscosity must be included in the model. To
see how this works, we shall make further use of a very important analytical method in
plasma physics; linearization.

4.4.1 Linearization
Let us exemplify linearization by continuing the discussion on charge separation. We
showed earlier that the Debye length λD is the largest distance in a plasma at which
charge separation can be maintained; at larger distances, the thermal energy is too large.
It turns out that the separation corresponds to a motion with a characteristic frequency.
We can derive this so called plasma frequency through linearization.
Let us assume that, for short times, the ions are stationary and the motion of the
electrons are parallel to the magnetic fields described by the fluid equations. We will

80
begin by considering the continuity equation;
∂ne
= −∇ · (ne ve ). (4.18)
∂t
If we now include the first order terms from a Taylor expansion around the equilibrium
state which existed just before the perturbation, we have

ne = ne0 + ne1
ve = 0 + ve1 (4.19)

where we have assumed that the fluid motion of electrons in equilibrium can be neglected.
All perturbed (first order) quantities are now expanded as in (4.10). Since the equations
we will study are linear, we may consider one mode at the time (algebra tells us they are
independent). Further, we see that the gradient and the time derivative of the perturbed
quantities can be replaced according to

∇ → ik
∂
→ −iω (4.20)
∂t
Equation (4.18) then gives one 0:th and one 1:st order equation:

∂ne0
=0
∂t
−iωne1 = −i ne0 k · ve1 (4.21)

The first equation provides no new information. Let us then linearize the electron equation
of motion and Poisson’s equation to obtain a closed system of equations:
dve
me = −eE ⇒ me (−i ω ve1 ) = −eE1
dt
ne e ne1 e
∇·E=− ⇒ ik · E = − (4.22)
ε0 ε0
Combining the algebraic equations (4.21) and (4.22) we find
s
ne0 e2
ω = ωp = . (4.23)
ε0 me

This characteristic frequency for plasma oscillations is named, not surprisingly, the plasma
frequency and has been measured experimentally. The example shows how powerful the
linearization tool can be in combination with Fourier methods. Complicated differential
equations are reduced to easily solvable algebraic expressions.

4.4.2 Alfvén waves∗

Let us now return to the ideal MHD model and study what plasma waves it can describe.
If the plasma pressure is low, we can neglect the ∇p term in the equation of motion.
Further, we assume that the mass density remains constant. We are then left with four
vector fields describing the MHD waves ; fluid velocity v, current density j, magnetic field

81
B and electric field E. There are four differential equations that couple these four vector
fields as follows:
dv
ρ =j×B
dt
j = ∇ × B/µ0
∂B
= −∇ × E
∂t
E = −v × B. (4.24)

The non-perturbed state is an infinite, static (v0 = 0), current-less (j0 = 0) plasma in a
homogeneous magnetic field B0 . We can then make a plane wave ansatz, assuming that
the perturbed quantities vary with space and time as

∼ exp(i(k · r − ωt)) (4.25)

which converts the differential equations to algebraic equations. We linearize, and write
down the 1:st order equations:

−ρiωv1 = j1 × B0
j1 = ik × B1 /µ0
−iωB1 = −ik × E1
E1 = −v1 × B0 . (4.26)

We use a cartesian (x, y, z) coordinate system with the magnetic field B0 in the z-direction,
and assume an Alfvén wave that propagates along the magnetic field direction (k k B0 ).
From the equations we find the directions of the vector fields as follows: v1 ⊥ B0 , j1 ⊥ k,
and v1 ⊥ j1 . We can then orient the system so that v1 is in the y-direction, and j1 is in
the x-direction. Further E1 ⊥ B0 , E1 ⊥ v1 , and B1 ⊥ k, B1 ⊥ E1 , from which follows
that B1 is in the y-direction and E1 is in the x-direction. It turns out that this wave is a
transverse wave, with all perturbation vectors in a plane perpendicular to the wave vector.
The problem is then simplified to four scalar equations for four unknown perturbation
fields vy1 , jx1 , By1 , Ex1 ,

ρiωvy1 = jx1 B0
jx1 = −ikBy1 /µ0
−iωBy1 = −ikEx1
Ex1 = −vy1 B0 . (4.27)

and elimination gives the dispersion relation for the Alfvén wave
B0
ω = kvA , vA = √ . (4.28)
µ0 ρ
Already in 1942 Alfvén described the wave in a short paper in Nature. The discovery
contributed to Alfvén being rewarded the Nobel Price in physics in 1970. This wave is
very important in fusion plasma physics.
If the plasma pressure is included into the equation of motion, two more MHD waves
can be derived using the dispersion relation:
ω2 p
2
= 21 [c2s + va2 ± (c2s + va2 )2 + 4c2s vA2 cos2 θ]. (4.29)
k
82
Figure 4.3: Alfvén wave with kkB0 . (Figure credit: By permission of Oxford University Press)

Figure 4.4: General Alfvén wave. k and B0 are in the plane of the figure, while the velocity
fluctuations are perpendicular to this plane. The wave is propagating in the x direction with
velocity vA . (Figure credit: By permission of Oxford University Press)

Figure 4.5: Magnetoacoustic wave. The velocity fluctuations are in the same plane as k and
B0 . (Figure credit: By permission of Oxford University Press)

p
Here, cs = γp/ρ is the sound velocity and θ is the angle between k and B0 . The two
waves are called the fast and the slow magnetoacoustic wave, respectively. In the case of
the plasma being incompressible, only the Alfvén wave results. Notice the direction of the
three MHD waves in figures 4.3-4.5. The Alfvén wave is propagating by bending the field
lines (B1 ⊥B0 ) while the magnetoacoustic waves are mainly compressing the field lines.
Of course, a large number of other waves can also exist in a plasma, such as electro-
magnetic waves. These cannot be described using the MHD models, however, which only
cover low frequency waves with ω < ωci .
One may wonder why the above perturbations caused waves and not instabilities. The
answer is that there are no driving forces for instabilities in the homogeneous plasma we
discussed.
But how come ideal MHD can describe waves and instabilities, but not damped or
growing oscillations? If we analyze the linearized MHD equations, we realize that the
eigenvalue ω 2 always is real. This is a result of the hermitian character of the system,
which we will not discuss further here.

83
4.5 Ideal MHD stability
The starting point is the ideal MHD equations in the following form, see egn. 2.103;

∂ρ
+ (v · ∇)ρ = −ρ∇ · v
∂t
∂v

ρ + (v · ∇)v = j × B − ∇p
∂t
∂p
+ (v · ∇)p = −Γp∇ · v
∂t
µ0 j = ∇ × B
∂B
= ∇ × (v × B) (4.30)
∂t
Assume a static equilibrium v = 0, and linearize the equations. The density equation
is not needed, since the density does not appear in any of the other equations. The 1:st
order equations can then be written as

∂v1
ρ0 = j1 × B0 + j0 × B1 − ∇p1
∂t
∂p1
+ (v1 · ∇)p0 = −Γp0 ∇ · v1
∂t
µ0 j1 = ∇ × B1
∂B1
= ∇ × (v1 × B0 ) (4.31)
∂t
Here subscript ‘0’ is used for equilibrium and ‘1’ for perturbed quantities. We now intro-
duce the plasma displacement ξ (r) as dξξ /dt = v, and time integrate the equations for the
pressure and the magnetic field:

∂ 2ξ
ρ0 = j1 × B0 + j0 × B1 − ∇p1
∂t2
p1 = −ξξ · ∇p0 − Γp0 ∇ · ξ
µ0 j 1 = ∇ × B1
B1 = ∇ × (ξξ × B0 ) (4.32)

The linearized equations can then be combined into a single equation for ξ :

∂ 2ξ
ρ0 = F(ξξ ) (4.33)
∂t2
where
 
1
F(ξξ ) = (∇ × (∇ × (ξξ × B0 ))) × B0 + (∇ × B0 ) × (∇ × (ξξ × B0 ))
µ0
  (4.34)
+∇ ξ · ∇p0 + Γp0 ∇ · ξ

84
4.5.1 The energy principle
We shall now turn to the energy principle, which is the most important method to study
the stability of the ideal MHD plasma. The method is frequently used, since it is compact
and relatively simple to handle analytically. Sometimes, the result is a so called stability
criterion, which tells us what is required from the equilibrium in order to stabilize a given
perturbation. A disadvantage is that the growth rate of the unstable modes discovered
cannot be calculated; we only know if we have stability or not. For growth rates, we have
to solve the eigenvalue equation, which is usually done numerically. Let us now see how
the energy principle works.
To simplify the notation we introduce

∂ξξ ∂ 2ξ
ξ̇ξ = , ξ̈ξ = (4.35)
∂t ∂t2
and write the equation of motion 4.33 as

ρ0 ξ̈ξ = F(ξξ ) (4.36)

We then perform a scalar multiplication of the equation with ξ̇ξ and integrate over the
plasma volume to obtain Z Z
ρ0 ξ̇ξ · ξ̈ξ dV = ξ̇ξ · F(ξξ )dV (4.37)
V V

The system may be isolated by using using the boundary condition ξ ⊥ = 0. The left hand
side of the equation can be rewritten as
Z Z  Z 
∂ 1 ∂ 1 ˙2
ρ0 ξ̇ξ · ξ̈ξ dV = ρ0 ξ̇ξ · ξ̇ξ dV = ρ0 ξ dV (4.38)
∂t 2 ∂t 2
V V V

It can be shown that this particular form F(ξξ ) is self-adjoint, which means that
Z Z
η · F(ξξ )dV = ξ · F(ηη )dV (4.39)
V V

for any choices of the vector fields η and ξ . We can use this property to show that
Z Z Z
∂

ξ̇ξ · F(ξξ )dV = ξ · F(ξ̇ξ )dV = ξ · F(ξξ ) dV (4.40)
∂t
V V V

The second equality follows from the fact that F is a linear function of ξ . The right hand
side of 4.37 can be rewritten as
Z Z Z  Z 
1 1∂
ξ̇ξ · F(ξξ )dV = ξ̇ξ · F(ξξ )dV + ξ · F(ξ̇ξ )dV = ξ · F(ξξ )dV (4.41)
2 2 ∂t
V V V V

We can now rewrite eq. 4.37, inserting eq. 4.38 and 4.41, as
Z Z 
∂ 1 ˙2 1
ρ0 ξ dV − ξ · F(ξξ )dV = 0 (4.42)
∂t 2 2
V V

85
We interpret this equation as follows: The first term is the perturbed kinetic energy of the
system Z
1 ˙2
δK = ρ0 ξ dV (4.43)
2
V

and the second term is the perturbed potential energy of the system
Z
1
δW = − ξ · F(ξξ )dV (4.44)
2
V

The equation,  
∂
δK + δW = 0 (4.45)
∂t
states that the total (kinetic + potential) perturbed energy is constant in time,

δK + δW = const. (4.46)

It then follows that any perturbation that decreases the potential energy (δW < 0)
produces a corresponding increase in the kinetic energy (δK > 0), which means that the
initial velocity increases, and the system is unstable. Alternatively, if all perturbations
lead to an increase in the potential energy (δW > 0), the system is stable.
Now, the idea is the following. First we define the geometry (e.g. a straight circular
cylinder). Then, we try all possible small displacements ξ of the plasma. If the displace-
ment that minimizes δW gives δW > 0 the plasma is stable, otherwise it is unstable. The
energy principle is a compact and neat way of testing for stability.
Test functions for ξ are very useful when it is possible to show that δW < 0 for a
certain perturbation. But how can we know if a sufficient number of perturbations ξ
have been investigated? There are two good approaches: either it is possible to see how
ξ affects δW in each case, or ξ must be expanded in a complete series of eigenfunctions
(usually truncated after a few terms).
In order to see what affects the stability, we write the perturbed potential energy as
follows:
Z
1

δW = − ξ · j1 × B0 + j0 × B1 − ∇p1 dV (4.47)
2
V

86
The first term in the integral can be rewritten as
Z


ξ · j1 × B0 dV
V
Z
1

= ξ· (∇ × B1 ) × B0 dV
µ0
V
Z  
1


= ∇ · (B1 × B0 ) × ξ − B1 · ∇ × (ξξ × B0 ) dV
µ0
V
I Z (4.48)
1
1
= (B1 × B0 ) × ξ · n̂ dS − B1 · B1 dV
µ0 µ0
S V
I   Z
1 1 2
= (B1 · ξ )(B0 · n̂) − (B1 · B0 )(ξξ · n̂) dS − B dV
µ0 µ0 1
S V
I Z
1
1 2
=− (B1 · B0 )(ξξ · n̂) dS − B dV
µ0 µ0 1
S V

We have here used the Gauss’ theorem to transform part of the the volume integral to a
surface integral over the plasma boundary. Further, we have used B0 · n̂ = 0 since the
equilibrium magnetic field is assumed to be aligned with the plasma boundary. Similarly,
we can rewrite the third term in the integral as
Z


ξ · ∇p1 dV
V
Z  
= ∇ · (p1 ξ ) − p1 (∇ · ξ ) dV
V
I Z (4.49)
= p1 (ξξ · n̂) dS − p1 (∇ · ξ ) dV
S V
I Z  
2
= p1 (ξξ · n̂) dS + (ξξ · ∇p0 )(∇ · ξ ) + Γp0 (∇ · ξ ) dV
S V

We now insert eq. 4.48 and eq. 4.49 in eq. 4.47, giving the following expression for the
perturbed potential energy:
1  B12
Z 
δW = − j0 · (B1 × ξ ) + (ξξ · ∇p0 )(∇ · ξ ) + Γp0 (∇ · ξ )2 dV
2 µ0
V
I (4.50)
1 1

+ B1 · B0 + p1 (ξξ · n̂) dS
2 µ0
S

If there is no displacement perpendicular to the plasma boundary (ξξ ⊥ = ξ · n̂ = 0), then

the surface integral is zero. This is the case when the plasma extends out to a rigid wall.
We can now examine the terms in the volume integral: The first term is a positive,
stabilizing term.
B12
Z
dV
µ0
V

87
This term represents the stabilizing effect that arises from compression and bending of
magnetic field lines. The second term is a potentially destabilizing term.
Z
− j0 · (B1 × ξ )dV
V

It involves the equilibrium current density j0 . An instability that is caused by this term
being negative is referred to as a current-driven instability. The third term is also a
potentially destabilizing term.
Z
(ξξ · ∇p0 )(∇ · ξ )dV
V

It involves the equilibrium pressure gradient ∇p0 . An instability that is caused by this
term being negative is referred to as a pressure-driven instability. The fourth term is a
positive, always stabilizing term.
Z
Γp0 (∇ · ξ )2 dV
V

This term represents the stabilizing effect due to plasma compression.

In summary, the perturbed potential integral contains four terms: There are two stabi-
lizing terms, representing the stabilizing effects of magnetic field compression and bending,
and plasma fluid compression, respectively. In addition, there are two potentially desta-
bilizing terms, representing current-driven and pressure-driven instability, respectively.

4.5.2 Extended form of the energy principle∗

In case the plasma is separated from the rigid wall by a vacuum region, it is possible to de-
rive the so called extended form of the energy principle, taking into account the perturbed
potential energies associated with the discontinuity at the plasma-vacuum interface, and
the perturbed magnetic field in the vacuum region. The total perturbed potential energy
consists of three parts
δW = δWF + δWS + δWV (4.51)
where
1
Z  B2 
1
δWF = − j0 · (B1 × ξ ) + (ξξ · ∇p0 )(∇ · ξ ) + Γp0 (∇ · ξ )2 dV (4.52)
2 µ0
plasma
I  out
1 2 1 2

δWS = (ξξ · n̂) ∇ B0 + p0 · n̂ dS (4.53)
2 2µ0 in
boundary

B12
Z
1
δWV = dV (4.54)
2 µ0
vacuum

The brackets in the expression for WS describes the discontinuous ”jump” in the quantity
inside the bracket across the boundary, calculated as the quantity just outside the bound-
ary minus the quantity just inside the boundary. In the surface integral, n̂ is the normal
unit vector in the outward direction from the plasma’s confining surface. The integra-
tion must be performed over the entire unperturbed plasma volume (F ), over the plasma

88
edge surface (S) and over the outer vacuum volume (V ) between the plasma and the sur-
rounding wall. The transformation between surface and outer volume integrals have been
performed using Gauss’ theorem, and the boundary conditions at the plasma-vacuum
boundary for magnetic fields and pressure.

4.5.3 Eigenmodes∗
Since the linearized equation of motion has constant coefficients in time, it can be writ-
ten as a complex valued eigenvalue equation. The displacement vector is obtained from
the real part of the complex valued eigenfunction ξ (r) and the corresponding complex
eigenvalue ω as
ξ (r, t) = <{ξξ (r) exp(iω)} (4.55)
Inserting this expression in eq. 4.33 gives the following form of the equation of motion

−ρ0 ω 2 ξ = F(ξξ ). (4.56)

By scalar multiplication by ξ ∗ and integration over the entire plasma volume, we obtain
the potential energy change associated with ξ (r). It is better to multiply with the complex
conjugate ξ ∗ rather than ξ since ξ ∗ · ξ is real.
Z Z
− ρ0 ω ξ · ξ dV = ξ ∗ · F(ξξ )dV.
2 ∗
(4.57)
V V

The left hand side is the perturbed kinetic energy δK produced by the eigenfunction, and
the right hand side is the corresponding perturbed potential energy δW . If follows from
the self-adjoint property of F(ξξ ) that ω 2 is real-valued. If we solve for ω 2 from (4.57) we
find;
− 12 V ξ ∗ · F(ξξ )dV
R
2 δW
ω = R 1 2
≡ . (4.58)
ρ |ξξ | dV
V 2 0
K
Since K > 0, the sign of ω 2 is given by δW . If δW < 0 we have ω = ±i|γ|, where γ is
the growth rate of the mode. There will be a growing eigenmode, growing as e|γ|t , as well
a damped eigenmode e−|γ|t , which quickly becomes dominated by the growing mode. For
δW > 0 we get oscillating, undamped eigenmodes(waves) with time dependence e±i|ω|t .

4.6 Tokamak stability

Since the tokamak is the ”pinch-type” magnetic confinement configuration, which has
proved to have the best stability and confinement properties so far, we will study the
stability of this confinement scheme in some detail, using the energy principle.

4.6.1 Instability in early pinches

Both linear and toroidal pinches are unstable, which is illustrated in the photography in
figure 4.6, taken during an experiment in the 1950s’.
It has turned out to be very important to shape the magnetic bottle in a clever way if
instabilities are to be avoided. Some of the most important instabilities are the so called
kink, sausage and Rayleigh-Taylor instabilities which are shown in figure 4.7.

89
Figure 4.6: Toroidal instability in an early fusion experiment. (Figure: EUROfusion-JET)

(a) (b) (c)

Kink instability Sausage instability Rayleigh-Taylor
instability

Figure 4.7: Three important instabilities in a pinch.

90
Figure 4.8: Diagram of unstable modes in a tokamak. (Figure credit: By permission of Oxford
University Press)

It was early realized that a powerful axial magnetic field and a conducting wall close
to the plasma had stabilizing effects, but it would turn out that several other instabilities
can be significant. We will now familiarize ourselves with so called macroscopic insta-
bilities which may be studied in fluid models. There are also a great number of more
or less important microscopic instabilities which are caused by, for example, the charge
distribution or non-Maxwellian velocity distributions. These instabilities must be studied
within kinetic theory.
In principle, we can say that the price paid for the high stability of the tokamak is the
high toroidal magnetic field. This causes the safety factor to assume the value q(0) ≈ 1
increasing steadily to q(a) = 3–4. Because of this, the number of resonant modes is
greatly reduced in the plasma, see figure 4.8. Since q = m/n on the resonance surfaces,
the mode (m, n) = (1, 1) is destabilized close to r = 0. This mode is called internal since
it is resonant in the plasma. Further, it is obvious that no modes (0, n); for all n or (1, n);
n > 1 can be resonant. However, modes like (2,1), (3,1), (3,2) and so on may be. These
are called kinks since the instability is kinking.

4.6.2 External kink

The high toroidal magnetic field in the tokamak results in a toroidal beta value of some-
thing like 5 % causing currents to dominate as driving force for long wavelength kinks.
We will now simplify the energy principle for the case that the tokamak has negligible
plasma pressure (zero beta), circular cross section and negligible toroidal effects. If we
further assume perturbations of the form ξ (r, θ, ϕ) = ξ (r)ei(mθ−nϕ) , the total perturbed
potential energy in (4.51)) can be written as
Za
π 2 Bϕ2  dξ 2  n 1
2
δW = (r ) + (m2 − 1)ξ 2 − rdr+
µ0 R0 dr m q
0

2 n 1
n 1
2  2 2
+ − + (1 + mλ) − a ξa , (4.59)
qa m qa m qa
where Bϕ = B0ϕ , ξ = ξr , λ = [1 + (a/b)2m ]/[1 − (a/b)2m ] > 0. Here, a is the plasma
radius, b is the distance to a perfectly conducting wall and qa = q(a). Let us consider the
case m ≥ 2 (we shall deal with m = 1 later).

91
 We will now demonstrate use of the energy principle. If b = a, i.e. the plasma extends
to the conducting wall, we have ξa = 0. This leaves us with only positive terms in the
integrand in (4.59). Thus, δW > 0 in this case, and the plasma is stable to perturbations.
If the conducting wall is further away, the plasma is still stable to all modes (m, n) where
qa > m/n. With m = 2, n = 1, 2, 3 . . ., we see that the plasma is stable if qa = q(a) > 2,
which is often the case in a tokamak.
Then, which modes are unstable? Let’s assume q(a) = 2.5. Outside the plasma edge
(but inside the conducting wall), q increases as rBϕ /(RBθ ) ∝ (Bϕ /R)[r/(1/r)] which is
∝ r2 since Bϕ ∝ 1/R in a tokamak. Therefore q may equal to 3 between the plasma and
the wall, making the perturbation resonant there and qa < m/n is a possibility. There
is nothing mysterious about this; the energy principle treats the stability of the entire
system plasma + vacuum field. Modes that are resonant between the plasma and the
wall are known as external modes. They are stabilized by, as discussed, having q(a) > 3.
Then, only modes with m > 3 are resonant, and these are generally stable at reasonable
beta values since they cause a considerable bending of the magnetic field.
But, you may ask, does the fact that one term in (4.59) is negative mean that δW < 0?
It does. The advantage with the energy principle is precisely this. We may choose any
perturbation ξ(r) when we minimize δW in search of an instability. We may, for example,
choose ξ(r) very small in the plasma volume, so that the positive contribution of the
integral may be neglected. Then, the negative term after the integration will dominate
close to the resonant surface and we will have an instability.

Figure 4.9: Stability diagram for external kinks in a tokamak. (Figure credit: By permission
of Oxford University Press)

The method for keeping q(a) > 3 is to “peak” the current density profile close to
r = 0. Figure 4.9 shows how the stability against external kinks is improved by increasing
v in the current density profile jϕ = j0 [1 − (r/a)2 ]v , which also increases q(a)/q(0). The
diagram has been calculated using a numerical code, applying the energy principle to the
assumed current profile. We see that the plasma is completely stable to external kinks
for q(a)/q(0) > 3.5, i.e. v ≥ 3.
In conclusion, the tokamak plasma is usually stable to external kinks if q(a) > 3.
This is important since unstable MHD modes grow very fast, on the time scale of one
Alfvén time (less than one µs). Recently, experiments have been carried out in tokamaks
with “reversed shear” where the q profile is permitted to have a minimum in the plasma
and small values on the edge. The significance of having the conducting wall close to

92
 q
1

r
ξ δ
ξ0

rs r

Figure 4.10: Test function ξ(r) for internal (m, n)=(1,1) internal kink instability

the plasma is then increased. The reason for this is that these “advanced” tokamak
configurations allow a big part of the plasma current, called bootstrap current, to be driven
by the plasma itself (more on this in the chapter on heating). Further, this configuration
allows the reduction of the tokamak’s dimensions.

4.6.3 Internal kink

One internally resonant instability is the (m, n) = (1, 1) mode. As we will find, it can
arise even if a conducting wall is close to the plasma. Let ξa = 0, causing the contribution
from (4.59) to disappear. If the plasma is circular cylindrical with R/a  1 we have
Za
π 2 Bϕ2

 dξ 2 2 2
 n 1
2
δW = (r ) + (m − 1)ξ − rdr [1 + O(ε2 )], (4.60)
µ0 R0 dr m q
0

where ε = a/R. Let us investigate if δW < 0 for any perturbation when (m, n) = (1, 1).
It seems like δWmin = 0 if we use the test function ξ(r) = ξ0 , where ξ0 is constant. This
test function is not, however, a possible solution since it does not fulfil ξ(a) = 0 except
for the trivial case ξ0 = 0.
Let us instead use the following test function (see figure 4.10)

ξ0
 0 ≤ r < rs − δ
ξ(r) = ξ0 (rs − r)/δ rs − δ ≤ r ≤ rs (4.61)

0 r > rs


If we use the Taylor expansion around q(r) at distance x = r − rs from the resonance
surface; q ≈ qs + q 0 x = m/n + q 0 x, we have n/m − 1/q = (nq − m)/(mq) ≈ (n2 /m2 )q 0 x.
For m = 1, the second term in the square bracket in the integral is zero and we have
Z0
ξ0
2 0 2 ξ2
δW ∝ − (q x) dx ∝ 02 q 02 δ 3 = O(δ) (4.62)
δ δ
−δ

which approaches zero when δ → 0. So, there are allowed test functions resulting in
δWmin = 0. Thus, the mode is marginally stability to this order, and stability depends
on the terms O(ε2 ). But if terms O(ε2 ) are included we find, after some calculations,
that δWmin < 0 for realistic beta values. This mode, which is independent on an ideally
conducting wall, is called an internal kink .

93
Figure 4.11: Combined stability diagram for external and internal kinks. (Figure credit: By
permission of Oxford University Press)

Experimentally, internal kinks are manifested as sawteeth in measured signals of, for
example the temperature close to r = 0. What happens is this: since the resistivity
is proportional to T 3/2 and the temperature profile decreases outwards from r = 0, the
current density tends to increase in time close to r = 0. This causes a local decrease
of q ∝ rBϕ /(RBθ ) ∝ 1/jϕ . As q approaches 1, the (1,1) mode is destabilized and the
resistivity close to r = 0 increases. This causes a decreasing jϕ and increasing q. The
cycle is then repeated and sawtooth shaped oscillations arise. In this way, q(0) is fixed to
a value close to one in a tokamak.
In general, the (1,1) kinks negatively affect the flux surfaces close to r = 0, causing
flat profiles and poor confinement locally. Resistive effects do have a limiting effect on
the instability, but the best thing would be to eliminate the mode completely.
In the complemented stability diagram in figure 4.11 the area of internal kink modes
has been added.

4.6.4 Resistive instabilities

A plasma always has finite resistivity. This may cause large differences in stability prop-
erties as compared to the ideal MHD model. A perfectly electrically conducting plasma
must retain its magnetic field topology, but if resistive diffusion of field lines is possible,
field lines can be broken and reconfigured. This may in turn change the plasma dynamics
and lead to new instabilities and deteriorated confinement.
An ideally MHD stable plasma may thus be resistively unstable. Fortunately, resistive
instabilities develop much slower than ideal instabilities.
Ever since an early analysis of resistive instabilities was made, these are divided into
tearing modes (driven by the current density gradient), resistive g-modes (driven by mo-
tion along curved field lines) and so called rippling modes (caused by the resistivity gra-
dient). More recent analysis has shown that tearing modes and resistive-g modes are
coupled for finite beta values.
We shall now briefly touch upon the huge field of resistive instabilities in a tokamak.
Let us, for reasons of simplicity, consider “pure” tearing modes. We will outline a common
method for solving the eigenvalue problem. As a starting point we assume a low-beta,

94
high aspect ratio tokamak, which allow as to approximate the plasma geometry with
a circular cylinder of length L = 2πR, and cylindrical co-ordinates (r, θ, ϕ), ϕ = z/R.
The eigenmodes in this geometry are Fourier modes, and the eigenvalues are the inverse
characteristic growth times of the modes, the so called growth rates γ. The ansats for the
eigenmodes are thus

Q(r, θ, ϕ) = <{Q(r) exp(γt + i(mθ − nϕ))} (4.63)

The eigenvalue problem is reduced to solving the one dimensional differential equations for
the complex valued Fourier mode coefficients Q(r) with appropriate boundary conditions.
If a helical eigenmode has the same pitch as the equilibrium magnetic field on some
flux surface, the perturbation is said to be resonant. In a toroidal geometry, this resonance
condition can be expressed using the safety factor as q = m/n. Magnetic field surfaces
where this condition can be fulfilled are called resonant surfaces. It is the region around
the resonant surface that plasma resistivity matters. Outside this reqion, the ideal MHD
model works well. The procedure to calculate the growth rate of the tearing mode is
a type of boundary layer analysis. An internal boundary layer is assumed around the
resonant radius. Outside this boundary layer the resistivity can be neglected, and the
ideal MHD model is used. Further, resistive instabilities grow slowly, so the inertial term
can be neglected in the other, using a perturbed force equilibrium equation. The starting
point of the analysis is thus the force equilibrium equation:

j × B − ∇p = 0 (4.64)

The aim is to obtain a differential equation for the radial magnetic field perturbation
B1r . We start by taking the rotation of the force balance equation, thus eliminating ∇p:

∇ × (j × B) = 0 (4.65)

This equation can be rewritten as

(B · ∇)j − B(∇ · j) + (j · ∇)B − j(∇ · B) = 0 (4.66)

and using ∇ · j = 0 and ∇ · B = 0 the equation simplifies to

(B · ∇)j + (j · ∇)B = 0 (4.67)

Using the low-aspect ratio, tokamak approximation, if follows that B · ∇jϕ  j · ∇Bϕ ,
and, we can write the axial component of the equation as

B · ∇jϕ = 0 (4.68)

and linearizing gives

B0 · ∇j1ϕ + B1 · ∇j0ϕ = 0. (4.69)
Introducing the Fourier mode expansion for the perturbed quantities, we can then with
the help of the linearized ideal MHD equations, rewrite this equation as an equation for
the perturbed radial component of the field B1r as
B0θ nq
 d d
2
 dj0ϕ
1 − (rB 1r − m B1r − B1r = 0. (4.70)
µ0 r 2 m dr dr dr
The first term corresponds to stabilizing field bending, while the second contains the
destabilizing current density gradient. As we approach the resonance surface, the first

95
term approaches zero and (4.70) is no longer valid. We introduce an inner boundary layer
at the resonance surface. In the boundary layer, rs − δ < r < rs + δ, resistivity and inertia
are important, and we have to use the resistive MHD equations, with the inertial term in
the equation of motion included. First we use Ohm’s, Faraday’s, and Ampere’s laws
ηj =E + v × B (4.71)
∂B
− =∇ × E (4.72)
∂t
µ0 j =∇ × B
to write an equation for the time rate of change of the magnetic field, as follows
∂B1 η
− + ∇ × (v1 × B0 ) = ∇ × (∇ × B1 ). (4.73)
∂t µ0
We can rewrite the second term as
∇ × (v1 × B0 ) = (B0 · ∇)v1 − B0 )(∇ · v1 ) (4.74)
and assume incompressible perturbations (∇ · v1 ). We then obtain an equation for the
radial component of the perturbed magnetic field
η
γBr1 − B0 · ∇v1r = ∆Br1 . (4.75)
µ0
We need another equation for the velocity perturbation v1r . The equation of motion,
including the inertial term, but omitting the term 2:nd order in the velocity is used
∂v
ρ = j × B − ∇p. (4.76)
∂t
Taking the rotation gives the following equation
ργ∇ × v1 = ∇ × (j × B) (4.77)
which can be expanded to yield an equation for v1r . We combine the two differential
equations, eqs. 4.75 and 4.77 to obtain a differential equation for B1r in the boundary
layer. The growth rate γ is obtained by matching the solutions for B1r in the outer
ideal MHD region to the solution in the inner resistive boundary layer at the boundaries
between the two regions. The matching procedure introduces the matching parameter ∆0 ,
which has given name to this analysis method which is sometimes called the ”delta-prime
method”. "
0
out #r=rs +δ " 0 in #r=rs +δ
B B1r
∆0 = 1r
= (4.78)
B1r B1r
r=rs −δ r=rs −δ
The bracket indicates the difference of the values taken at the boundaries located at
r = rs + δ and r = rs − δ. We now introduce the characteristic time for magnetic field
diffusion in the plasma, the resistive time constant
µ 0 a2
τR = (4.79)
η
and the inertial time scale, the Alfvén transit time
√
a µ0 ρ
τA = (4.80)
Bϕ

96
Figure 4.12: Cross section of a) normal flux surfaces b) flux surfaces with m=1,2,3 islands

Finally, we obtain the growth rate for the tearing mode as

a aq 0 −3/5 −2/5
 
γ = 0.55 n τR τA (a∆0 )4/5 (4.81)
R q

where n is the toroidal mode number of the eigenmode. The quantities in the expression
for γ are evaluated at the position of the resonant layer r = rs . The characteristic growth
time for the tearing mode is intermediate between the resistive and inertial time scales.
Let us illustrate with a numerical example: Consider a tokamak with magnetic field
Bϕ = 5 T, deuterium plasma ion density n = 1020 m−3 , and plasma temperature T = 1
keV, in a torus with minor radius a=0.5 m. The resistivity η = 2.5 × 10−8 Ωm is estimated
from Spitzer’s formula with these parameters, giving a resistive time constant of τR = 13
s, in the order of seconds. The Alfvén transit time τA = 0.065 µs, which characterize the
growth of ideal MHD modes, is less than microseconds. The tearing mode growth time for
3/5 2/5
these parameter is of the order of milliseconds γ −1 ∼ τR τA ∼ 6 ms. Resistive tearing
modes thus grow considerably slower than ideal MHD modes at high plasma temperature.
For these “pure” tearing modes, γ increases as η 3/5 with plasma resistivity. The growth
rate is approximately inversely proportional to plasma temperature since η ∝ T −3/2 and
γ ∝ η 3/5 ∝ T −9/10 . When realistic beta values are used in the analysis, the dependence of
the growth rate on resistivity can instead become the so called resistive g-mode scaling
η 1/3 .

4.6.5 Magnetic islands

If a magnetic field perturbation with poloidal mode number m and toroidal mode number
n is aligned the equilibrium magnetic field on a rational flux surface, where the safety
factor is q(rs ) = m/n, the perturbation is said to be resonant. In resistive MHD, a
magnetic island can form by reconnection of the magnetic field at resonant surfaces. A
magnetic island can be regarded as a plasma filament with its own magnetic flux surfaces
confining a local helical magnetic axis. (see figure 4.12).
The development of magnetic islands can be described as follows (figure 4.13): Con-
sider a plane model of a toroidal plasma close to a resonant magnetic flux surface (figure
4.13a,b). The poloidal equilibrium magnetic field is assumed to increase with plasma
radius as shown in figure 4.13c. The toroidal equilibrium magnetic field is for simplicity
assumed to be constant. Consider a projection of the magnetic field on a surface perpen-
dicular to a magnetic field line in the resonant flux surface. The equilibrium field B0 close

97
 Figure 4.13: Transformation of the toroidal geometry to cartesian coordinates.

to the resonant surface will then vary as in figure 4.13d. At the resonant surface r = rs
we have B0 = 0. (Alternatively B0 can be seen as the poloidal equilibrium field after
substraction of a field with constant pitch equal to the equilibrium field on the resonant
surface.) Assume that a resonant perturbation with a radial magnetic field component
arises. For simplicity, we assume the perturbation to be sinusoidal in poloidal and toroidal
directions. Some distance away from the resonant surface the effect of the perturbation
will be small, but close to the surface it will be comparable to the equilibrium magnetic
field in this projection. The result is that local flux surfaces, i.e. magnetic islands are
created as in figure 4.13e. The boundary between local flux surfaces in a magnetic is-
land and the surrounding global flux surfaces is called the separatrix . The width of the
separatrix at the widest point on the island is called the island width, denoted W .
An expression for the island width W for a given amplitude of a magnetic perturbation
can be obtained in the following way. Introduce cartesian coordinates (x, y, z) and let
the x direction correspond to radial direction, with x = 0 at r = rs , and B0 is in y
direction (figure 4.13 e). Assume a resonant magnetic field perturbation B1 arises. The
perturbation will be periodic in y direction, and its x component can be written

B1x = B1 sin ky (4.82)

where k = 2π/λ = m/rs is the perturbations wave vector in poloidal direction, λ is the
wave length and m is the poloidal mode number. Close to the resonant surface, the y

98
component of B0 can be approximated by Taylor expansion around x = 0:
B0y ≈ xB00 (4.83)
where  
dB0y
B00 = . (4.84)
dx x=0

In this 2-D geometry it is helpful to introduce a flux function to describe B. The flux
function Ψ is chosen such as
1. Ψ is constant on magnetic flux surfaces: ∇Ψ · B = 0
2. B fulfils the condition: ∇ · B = 0.
These two condition are fulfilled by the following definition for Ψ:
∂Ψ
Bx = −
∂y
∂Ψ
By = (4.85)
∂x
By introducing eq. 5.99 in eq. 5.102 we obtain the perturbation part of the flux function;
B1
Ψ1 = cos ky (4.86)
k
and by inserting eq. 5.100 we get the equilibrium part;
x2 0
Ψ0 = B. (4.87)
2 0
The total flux function is
Ψ = Ψ0 + Ψ1 . (4.88)
The flux function is also constant on the separatrix, i.e. the line separating the area with
local flux surfaces (the magnetic island) from the area with global flux surfaces (the rest
of the plasma). Let the flux function assume the value Ψ = Ψs on the separatrix. An
expression for the island width can be reached as follows: The separatrix passes x = 0 at
ky = p 2π, p = ±0, 1, 2, . . .. Inserting this in eq. 5.105 gives
B1
Ψs = . (4.89)
k
The separatrix also defines the island width. One special case is W = 2x for ky =
π + p 2π, p = ±0, 1, 2, . . ., which are the y values when the separatrix passes the widest
part of the island. Is we insert this in eq. 5.105 we get a second expression for Ψs :
 2
1 W B1
Ψs = B00 − . (4.90)
2 2 k
Combining eq. 5.106 and 5.107 gives us the following expression for the island width W :
s
B1
W =4 . (4.91)
kB00
In cylinder geometry, it can be shown that the expression for the island width W has a
similar form (using the safety factor q):
s  
Br1  q 
W = 4r . (4.92)
mBθ0  rq 0 

99
Figure 4.14: Positive curvature (the inside of the torus) is, on average, stabilizing. (Figure
credit: By permission of Oxford University Press)

4.6.6 Pressure driven instabilities

A reactor requires considerable pressure gradients to be balanced by the magnetic field.
Again, weak points are the resonant surfaces. But in this case, another factor comes
into play; the curvature of the magnetic field towards the plasma. It is energetically
advantageous to compress a magnetic field whose field lines are concave to the magnetic
field than vice versa. Further, more energy is required to compress a magnetic field that
has a strong twist radially outwards than field lines which are parallel radially (as seen
from the outside). This is known as shear .
By including the pressure gradient in δW and minimize using Euler-Lagrange’s method,
Suydam showed in 1958 that the following condition must be fulfilled close to the reso-
nance surfaces in a circular cylindrical, straight plasma:
 2
0 rBz2 q 0
p + > 0. (4.93)
8µ0 q
Here, we user 0 instead of d/dr and the shear factor is q 0 /q, with q = rBz /Bq . Obviously
a higher pressure gradient p 0 requires a greater shear. In a straight tokamak, this condition
would be very difficult to fulfil for interesting plasma pressures.
However, in 1960, Mercier showed that the toroidal curvature of the tokamak field
lines has a beneficial influence. If we bend the plasma to a torus with R/a  1, Suydam’s
criterion takes the following form:
rBϕ2 q 0 2
 
0 2
p (1 − q ) + > 0. (4.94)
8µ0 q
Since q > 1 in the entire plasma (except very close to r = 1, where p 0 is small anyway),
both terms on the left hand side are generally positive, and the plasma is stable against
pressure driven modes. The toroidal shape is thus improving stability. The reason for this
is the field line curvature the plasma sees along a field line; it is composed mainly of good
curvature on the inside of the torus. Here, the poloidal curvature is disadvantageous if
q > 1, but the advantageous toroidal curvature dominates. For lower values of q the field
line is more stretched along the poloidal direction, which gives a negative contribution to
the average curvature: q < 1 ⇒ Bθ /Bϕ > r/R. The pressure profile should be flat close
to r = 0 to ensure good stability.
When very high beta values are desired, localized pressure driven instabilities may
arise on the outside of the torus, where the curvature is bad. This happens for high mode
numbers when
0
Bϕ2 /µ0
p + 2 < 0. (4.95)
q R0

100
 An apt name for these instabilities is ballooning modes. Since these strongly limit the
possible beta values in tokamak, different ways of handling them have been explored. One
interesting possibility has turned out to be the positive effect from the inside of the torus
when the plasma is given a bean like shape. This allows the so called second stability
region to be reached.

4.6.7 Disruptions
An Achilles’ heel of the tokamak is disruptions, and limit operational parameters of the
tokamak such as the maximum current and density. Disruptions are fast events when
plasma current and confinement drastically declines. The sudden termination of the
plasma cause high heat loads and strong mechanical strains. Future tokamak experiments
are designed to withstand a number of smaller disruptions, but in a reactor, it would be
preferred to get rid of them completely. Not all tokamaks suffer from disruptions. At low
aspect ratio, spherical tokamaks, disruptions do not occur.
The cause of disruptions is only partly known. We will now discuss two important
factors; too low qa value and too high density. Tearing modes with high amplitude have
proved to be connected to these cases. When the qa value is too low (≤ 2), the current
density at the edge tends to be steep, which causes strong tearing modes. Thus, it is
important not to have too high plasma current for a given toroidal field.
Disruptions at high density is usually caused by an increase in radiation due to edge
impurities. When the plasma edge thus is cooled, the current density profile it contracted
resulting in a destabilizing gradient. An empirical law for the upper density limit has
been found by Greenwald; n < 1 × 1020 Ip (MA)/(πa2 ). The relation has turned out to
be valid for many experiments. It gives the maximum allowed density 6 × 1020 m−3 in a
reactor with a = 1 m and Ip = 20 MA.

4.6.8 Beta limits

As shown earlier, beta should be maximized in a fusion reactor based on magnetic con-
finement. But since this may cause instabilities, it is relevant to investigate optimal
configurations and beta limits for these. This has been done numerically for tokamaks by
Troyon, who investigated many equilibria for different current and pressure profiles. The
results are consistent with experimental measurements.
We have seen that jϕ (r) must be sufficiently peaked close to the magnetic axis and
that the safety factor qa should be a bit higher than 2.0 to prevent the formation of
external kinks and disruptions. In addition, the safety factor q0 should be at least 1.0
to prevent internal kinks and pressure driven modes. Tearing modes are controlled by
limiting djϕ /dr.
As long as the above requirements are fulfilled, elongation and other plasma shape
parameters may be varied. With these requirements in mind, beta is gradually increased
until ballooning modes occur. The beta limits can then be approximated by the following
analytical expression:
Ip (MA)
βϕ (%) = 2.8 . (4.96)
aBϕ
For most tokamaks, the beta limit corresponds to about 5 %. Note that for low aspect
ratios (R/a < 2.5), eg.(4.96) is no longer valid. In these machines, (spherical tokamaks,
spheromaks) beta values well over 10 % can be reached.

101
Figure 4.15: Comparison between experimental results for beta and equation (4.96). (Figure
credit: By permission of Oxford University Press)

Figure 4.15 shows comparisons with experiments where the Troyon factor 2.8 above
has been replaced by li (the internal inductance of the plasma), which is a measure of
the current profile’s influence on the βϕ limit. The inductance increases when current is
concentrated near the magnetic axis r = 0. The parameter βN denotes the “normalized
beta value”; βN = β(%)/[Ip (MA)/(aBϕ )], where β is the experimental value.

4.6.9 Further instabilities influencing confinement∗

We have earlier discussed some important MHD instabilities in tokamaks. Also other
significant instabilities, limiting confinement, may occur. These usually involve more
physics.

Edge Localised Mode (ELM)

One instability which is regularly seen is called the ELM (Edge Localised Mode). It has the
form of a relaxation oscillation - there is an abrupt instability, removing the destabilising
conditions, and a slow recovery returning the plasma to an unstable condition once again.
This leads to a repetitive instability. The ELM is typically localised to the outer 10 % of
the radius. In its mild form the ELM only has a weak effect on confinement, but it may
also present a serious problem.
The ELM appears when the plasma is in the so-called high confinement (H-mode) con-
figuration. The higher plasma energy in these configurations is due to a ”pedestal”, with
sharp gradient, at the edge of the pressure profile. This pedestal results from pedestals
on both the density and temperature profiles; see a comparison of L-mode and H-mode
temperature profiles in figure 4.16. The L-H transition physics is complicated; we satisfy
ourselves here by noting that it has been found that the shear in the E×B velocity profile
helps to suppress the turbulent transport near the edge, thus enabling better confinement
and pedestals. The transition occurs when the power injected into the plasma exceeds a
certain, empirically determined, threshold. The resulting H-mode confinement is usually
about a factor two higher than L-mode confinement.
While it is obvious that the pedestal is advantageous to achieve a higher confined
energy, the price to be paid is that the inevitable steep gradients at the plasma edge may

102
Figure 4.16: Comparison of L-mode and H-mode temperature profiles showing the temperature
”pedestal” at the edge of the plasma. (Figure credit: EUROfusion-JET)

lead to ELM:s.
ELMs appear in different forms. In their weakest form they occur as frequent small
pulses, producing time traces that look like grass, giving them the name ”grassy” ELMs.
With higher power heating they become more separated in time and larger in size. When
they have become ”giant” ELMs they present a serious problem, the associated sudden
energy release produces a high heat loading on material surfaces, which leads to reactor
technological constraints.

Multi-faceted Asymmetric Radiation from the Edge (MARFE)

A back transition to inferior L-mode confinement can be induced by another instability
called MARFE (multi-faceted asymmetric radiation from the edge), which may appear at
high plasma density. The MARFE region has been cooled by radiation from impurities,
and it is the temperature dependence of this radiation that characterises the instability.
In the temperature range of interest the impurity line radiation increases with decreasing
temperature. As the radiation energy loss causes a fall in temperature, the radiation
is further increased - and an instability results. A further destabilising factor is that,
because the pressure is constant along a field line, the lower temperature in the MARFE
implies an increased density. The radiated power increases with density and this provides
an additional drive. The process is, however, limited. The fall in temperature in the
MARFE produces an increasing temperature difference between the MARFE and the
surrounding plasma. This leads to conduction of heat into the MARFE, and a steady
state is reached when the thermal conduction balances the radiated loss.

Neoclassical Tearing Mode (NTM)

A family of instabilities that limit the plasma pressure are the so-called neoclassical tearing
modes (NTM). They occur when the temperature and pressure of the plasma is high.
Theoreticians have calculated the upper NTM stability limit for the plasma pressure to
be lower, the larger the device is.
Why do they occur? When NTM:s form, the symmetric plasma ring is subject to
magnetic islands. These are perturbations with their own magnetic field structure closed
upon itself. These are triggered as the plasma pressure rises with increasing plasma
temperature. With the emergence of the islands, the magnetic field lines tear and link up

103
with the field lines of neighbouring magnetic surfaces. This gives rise to magnetic short-
circuiting. Since now fast energy exchange can also occur transversally to the surfaces,
the plasma temperature and plasma pressure drop right across the island. This restricts
the plasma pressure attainable.
The formation of such magnetic islands can be hampered, however, by accurately
beaming microwaves into the centre of an emerging island. This generates a local electric
current that makes the island vanish. The magnetic field perturbation is suppressed
and the plasma pressure can rise again. It is presently possible to eradicate an island
completely. One may stabilise a particularly disturbing tearing mode that can lead to
disruption of a discharge; after improvement of the beaming method this could be achieved
with quite low microwave power. For stabilisation, less than ten per cent of the total
heating power applied is sufficient when precisely beamed to the right spot.

104
Chapter 5

TRANSPORT

To reach thermonuclear conditions in a tokamak it is necessary to confine the energy of

the plasma long enough (see chapter 1). The confinement in limited by energy losses
caused by heat conduction and convection across the magnetic field and by energy losses
due to electromagnetic radiation from the plasma. The energy confinement time τE is
defined as the ratio between the energy content of the plasma and its power loss. For a
tokamak with closed magnetic flux surfaces there is always a particle and energy loss, or
transport, across the magnetic field caused by Coulomb collisions. This collision driven
transport is called classical transport in a straight, circular cylindrical plasma column.
In a toroidal plasma the transport is modified and an increased transport level is reached
which is called neoclassical transport.
Experimental measurements of transport in tokamaks show that the heat conduction
for ions is comparable to the neoclassical, while it is two orders of magnitude larger than
neoclassical theory for electrons. The particle transport is also larger then the neoclassical.
The deviation from neoclassical transport is called anomalous transport. There are several
possible explanations of the anomalous transport. Instabilities in the plasma give rise to
local fluctuations in for example plasma potential and plasma density. This so called
electro static turbulence in the plasma can cause particle and energy transport across the
magnetic field. Electro magnetic instabilities can change the structure of the magnetic
field which may cause energy losses along the magnetic field.
There are three main types of transport models:

1. Classical transport theory. Describes transport in a cylindrical symmetrical mag-

netic field driven by Coulomb collisions between charged particles.

2. Neoclassical transport theory. Describes transport in a toroidal magnetic field driven

by Coulomb collisions between charged particles. Guiding-centre drift and confine-
ment of particles in mirror magnetic fields modify transport.

3. Anomalous transport models. Empirical or semi empirical models which describes

transport driven by different kinds of plasma turbulence.

5.1 Classical transport theory

Classical transport theory describes diffusion of particles and energy across a homogeneous
magnetic field due to Coulomb collisions in the plasma. This process is fundamental and
is similar to diffusion in non-ionized gases. An important difference between diffusion in

105
 n

r no

r
ro

Figure 5.1: Schematic radial profile of the plasma density.

t
τ

x
Δx

Figure 5.2: Diagram illustrating the random walk model.

plasma and in neutral gas is that the particles in the plasma do not move in straight
paths between the collisions, but in curved paths given by the gyration in the magnetic
field. Another important difference is that the collisions are not close collisions, but the
result from interactions by many particles over long distances. Despite these differences,
estimations of the order of magnitude of diffusion coefficients can be made with models of
the type used for neutral gases, so called random walk models, which is the starting point
for this description of classical transport theory. The purpose of these estimations is to, in
the easiest possible way, illustrate the principle of the particle transport mechanism. The
analysis can in principle be generalized to include the more important heat diffusivity.
At the end of this section, we will summarize the relations between particle and heat
diffusivity which are useful for an analysis of the plasma energy balance.

5.1.1 The random walk process

A simple model often used to describe fundamental process for diffusion on gases is the
random walk model. Since it can be used to understand the important transport theo-
ries for plasma (classical, neoclassical and anomalous) we will here discuss it in detail.
Consider a plasma cylinder with a radial density profile n(r) (figure 5.1).
The model is based on the following simple assumptions of the collision process:

1. particles colliding with time interval τ

2. particles moving radial distance ∆x

3. particles move in positive and negative direction with equal probability after a col-
lision

Taylor expansion of the density n around r0 :

 
dn
n ≈ n0 + x
dr r0

106
 + ρ

-
Δx

Figure 5.3: The effect of collisions between particles of different charge.

Here, n0 is the density at r = r0 and x = r − r0 . The outward flow is

Z0   !
1 dn
Γ+ = n0 + x dx
2τ dr r0
−∆x

and the inward flow is

Z∆x   !
1 dn
Γ− = n0 + x dx
2τ dr r0
0

The flow of particles through the surface r = r0 is given by integration as

(∆x)2 dn
 
Γ = Γ+ − Γ− = −
2τ dr r0

Fick’s law describe diffusion in the general case;

dn
Γ = −D (5.1)
dr
From which we can identify the diffusion coefficient D as

(∆x)2
D= (5.2)
2τ
This important result, which can be used to estimate the diffusion coefficient, tells us that
the coefficient is proportional to the step length squared divided by the collision time.
This is the diffusion we call random walk process.

5.1.2 Particle model for classical diffusion

Consider a particle in motion in a homogeneous magnetic field. The particle is moving
in a circular path perpendicular to the magnetic field. The parameters describing the
rotational motion are the gyrofrequency ωc and the Larmor radius rL . Positive and
negative particles have different rotational directions.
A simple estimate of the particle diffusion across the magnetic field due to collisions
between charged particles is given by the random walk model: It can be shown (figure
5.3) that the step length at collisions for the gyrating particles is of the same order of
magnitude as the Larmor radius. It can also be shown that diffusion only occurs at

107
 +

Figure 5.4: The effect of collisions between particles of same charge.

z
j
V Δ
p
B

Figure 5.5: Cylindrical geometry when calculating a radial flow.

collisions between particles of different charge (see figure 5.4). The collision time used for
the estimate is thus the one for Coulomb collisions between electrons and ions.
Collisions between particles of different charge cause a movement of the guiding centre
of the particles of the same order of magnitude as the Larmor radius

∆x ≈ rL (5.3)

Inserting this in the expression for the diffusion coefficient in the random walk model
gives
r2
D≈ L (5.4)
2τ
Including the Larmor radius and collision time τe of the electrons, the expression for the
particle diffusion coefficient D becomes
me T
D≈ (5.5)
eB 2 τe
If we insert the Larmor radius and collision time τi of the ions, we find that the
diffusion coefficients for ions and electrons are equal, i.e. De = Di = D. This means that
the classical diffusion is naturally ambipolar , i.e. it conserves the charge neutrality of the
plasma.

5.1.3 Fluid model for classical diffusion

An expression for the particle transport caused by Coulomb collisions in a straight plasma
cylinder can be derived using a fluid model. A resistive MHD model is the simplest model
describing the radial particle flow. (This model is an expansion of the ideal MHD which
includes the term ηj in Ohm’s law, where η is the resistivity of the plasma.) In a fluid
model of the plasma, the Coulomb collisions corresponds to a friction force between the
ion and electron fluids, which is characterized by the resistivity of the plasma. Consider
a plasma cylinder with pressure p, an axial current density j, and an azimuthal magnetic
field B (see figure 5.5). By inserting the force balance equation in Ohm’s law we get an

108
 n
a

a
r

Figure 5.6: Cylindrical model for estimation of confinement time.

expression for the radial fluid velocity perpendicular to the magnetic field:
1 η
v⊥ = 2
E × B − 2 ∇p (5.6)
B B
The pressure can be written as p = 2neT . Now, assume that E = 0 and that T = const.
The radial particle flow is Γ = nv⊥ , or
2neT
Γ = −η ∇n (5.7)
B2
Comparing the above expression with Fick’s law gives an expression for the diffusion
coefficient D:
2neT
D=η 2 (5.8)
B
Inserting the expression for plasma resistivity η gives the diffusion coefficient as
2me T
D= (5.9)
eB 2 τ
It is obvious that both resistive MHD and random walk gives similar expressions for the
classical diffusion coefficient in a straight cylinder.

5.1.4 Relation between confinement time and diffusion coeffi-

cient
A relation between the diffusion coefficient and the particle confinement time is given by
considering a simple model of a plasma cylinder where the diffusion coefficient is assumed
constant over the cross section of the cylinder (figure 5.6) The plasma density n is a
function of the radius r and time t and is given by two partial differential equations: the
continuity equation which can be written
∂n
∇·Γ=− (5.10)
∂t
and Fick’s law which can be written in vector form as
Γ = −D∇n (5.11)
where Γ = nv⊥ . The diffusion equation is
∂n
∇ · (D∇n) = (5.12)
∂t
which is a second order partial differential equation. In this case, when we have a cylindri-
cal symmetry, the equation can be solved by separation of variables. Assume the following
expression for n:  
t
n(r, t) = n(r) exp − (5.13)
τp

109
The constant τp in known as the particle confinement time. In this simple model, τp is
the time constant for the decrease in density at a given radius due to the radial particle
flow. The diffusion equation in cylindrical coordinates is of the form of a Bessel equation:
 
1 ∂ ∂n n
r + =0 (5.14)
r ∂r ∂r τp D

The general solution is the zero’th order Bessel function J0 (r). The condition n(a) = 0
at the plasma edge r = a finally gives
 
 r t
n(r, t) = n0 J0 2.405 exp − (5.15)
a τp
where n0 in the plasma density at the cylinder axis r = 0 at time t = 0 and x = 2.405 is
the first zero of J0 (x). Inserting the solution gives the relation between τp And D:

a2 a2
τp = ≈ 0.17 (5.16)
2.4052 D D
This is a very useful expression for a good approximation of the confinement time when
the diffusion coefficient is given.

5.1.5 Classical heat conduction

The radial particle flow due to a radial density gradient in a plasma cylinder is equal
for ions and electrons according to the classical transport model and is described by the
particle diffusion coefficient D:
dn
Γ = −D (5.17)
dr
A radial temperature gradient in a plasma cylinder will, because of the heat conductivity
of the plasma, drive radial heat fluxes qe and qi described by the heat conductivities Ke
and Ki for electrons and ions, respectively:
dTe
qe = −Ke (5.18)
dr
dTi
qi = −Ki (5.19)
dr
Instead of heat conductivity K, heat diffusivity χ is often used, which has the same
dimension as the diffusion coefficient D. This gives us the expressions for heat flux:
dTe
qe = −nχe (5.20)
dr
dTi
qi = −nχi (5.21)
dr
The following expressions relates D to χe and χi :

χe ≈ D (5.22)
r
mi
χi ≈ D (5.23)
me
The energy confinement times τE,e and τE,i are given from the heat diffusivities χe andχi .

110
5.1.6 Energy confinement time for ohmically heated plasma∗
Consider a tokamak or pinch in which the toroidal current provides both heating and
confinement. The energy confinement time τE in stationary state (Pin = Ploss ) is
3
2
< 2neT >
τE = Vp . (5.24)
Pin

where Vp is the toroidal plasma volume, approximated as Vp = 2π 2 Ra2 . The ohmic heating
power is given by the product of the plasma current Ip and the loop voltage Uloop .

Pin = Ip Uloop (5.25)

Introduce the poloidal beta value

< 2neT >
βθ = . (5.26)
Bθ2 (a)/2µ0

Poloidal beta βθ is exactly equal to unity for a z-pinch. The z-pinch, however, is unstable.
In the more attractive reversed-field pinch, βθ ≈ 0.1. In tokamaks, βθ can reach values of 1-
3 depending on the shape of the plasma cross section. Use Ampere’s law 2πaBθ (a) = µ0 Ip
and rewrite the expression for the energy confinement time as
3 Ip
τE = µ0 Rβθ , (5.27)
8 Uloop

Ohm’s law is written as

2πR
Uloop = Ip fG η (5.28)
πa2
where η is the plasma resistivity. The factor fG takes into consideration the geometry
dependent current path in the plasma. Generally, this is screw shaped making fG = 4-5
for an RFP. The same value is actually valid for a tokamak as well. Here, the shape is
less of a screw, but fG will be of the same size if one takes into account that a certain
portion of the electrons are mirrored by the magnetic field along the current path. The
energy confinement time is then obtained simply as

3βθ µ0 a2 3βθ
τE = = τR (5.29)
16fG η 16fG

where τR = µ0 a2 /η is the plasma resistive time, characterizing the magnetic field diffusion
in the plasma. If we introduce the Spitzer resistivity, eq. 2.76 and combine equations we
get
2.8 × 1014 2 2.5
τE = a βθ (Ip /N )1.5 Ip1.5 (5.30)
fG Zeff ln Λ
The plasma density has been included in the so called line density N = nπa2 . This is the
highest possible energy confinement time for an ohmically heated plasma. The Coulomb
logarithm ln Λ can be set to ≈ 15 for a fusion plasma. With parameters fG = 5, a = 1 m,
Zeff = 1, βθ = 1, n = 1.0×1021 m−3 and Ip = 10 MA we find τE = 21 s. If experiments had
given this value, fusion energy would already be reality. Unfortunately, real life tokamaks
have energy confinement times one or two orders of magnitude smaller. One of the most
important problems in fusion research is to try to explain and diminish this difference.

111
 z
φ
Bφ
Bθ +
Bθ B
θ
Bφ Bφ
Bθ -I
-φ
+

Figure 5.7: Cancellation of vertical guiding-centre drift in a tokamak.

5.2 Neoclassical transport theory

Neoclassical transport theory describes transport driven by Coulomb collisions in a toroidal
geometry. In an inhomogeneous magnetic field, where the field gradient is perpendicular
to the field, the drift of the particle guiding centres will be across the magnetic field lines.
Only in certain special cases will the drift paths form closed orbits within the magnetic
flux surfaces (this is for example true for confinement in the poloidal magnetic field out-
side a ring shaped conductor). Since the guiding-centre orbits in these cases coincide with
the magnetic flux surfaces, this does not affect transport. In the general case, the guiding-
centre orbits will deviate somewhat from the magnetic flux surfaces. This will cause an
increase of transport across the magnetic field. One sort of such neoclassical transport
is called Pfirsch-Schlüter diffusion. Another effect which may influence transport in a
tokamak is the “trapping” of particles in the outer part of the torus caused by mirroring
of the drift motion along the magnetic field lines because of a field gradient parallel to
the magnetic field. This trapping of particles cause a sort of neoclassical transport called
“banana orbit” diffusion, named for the characteristic shape of the guiding-centre orbits
projected on a vertical plane.
In an inhomogeneous magnetic field, charged particles will, in addition to their gy-
rating motion around field lines, have a drift of the centre of gyration in the plane per-
pendicular to the magnetic field. This phenomenon is called guiding-centre drift. In a
toroidal magnetic field this drift will be in vertical direction. The guiding-centre drift is
the sum of drifts caused by the magnetic field gradient and the magnetic field curvature.
It turns out that the two drift motions for a given type of particles (e.g. electrons) are
added. In addition to this, electrons and ions have drift motions in opposite directions
causing a vertical electric charge field. This cause, in turn, a E × B guiding-centre drift
directed radially outwards (along the major radius) for both electrons and ions.
The particle guiding-centre moves along field lines and partly, because of the guiding-
centre drift, across field lines in a resulting guiding-centre orbit. One important conclusion
from this is that the guiding-centre orbits are not confined in a simple toroidal magnetic
field; it requires a more complex magnetic field geometry.

5.2.1 Guiding-centre orbits in a tokamak

The problem with the vertical guiding-centre drift in a purely toroidal magnetic field
has been solved in the tokamak by adding a poloidal magnetic field, and the resulting
magnetic field lines form a ring shaped spiral (figure 5.7). The poloidal magnetic field
component is created by a toroidal electrical current in the plasma. The vertical drift is
approximately cancelled when a particle is moving along the field line one turn around

112
 Driftytor
z Projektion av driftbanor
på vertikalplan
φ
VII Ω
r
B
θ
Rc R
d d
VII
Fluxyta

Figure 5.8: Guiding-centre orbit in a tokamak.

the torus, since the vertical drift changes direction relative to the minor axis making the
average drift motion away from the minor axis small on average. This means that the
particle confinement depends on the particles ability to move freely along the field lines
around the tokamak. Particle collisions, which can be regarded as a friction force slowing
down the motion along the field line, cause increased transport, as we will show in the
next section.
In a tokamak with circular magnetic flux surfaces the projection of the guiding-centre
drift orbits on a vertical plane will be circles with the centre displaced somewhat in
horizontal direction relative the flux surface (figure 5.8). The displacement will be inwards
or outwards depending on the direction of the particle’s velocity along the field line.
The size of the displacement of the guiding-centre orbit relative the flux surface can
be estimated in the following way: Consider a magnetic flux surface with a circular cross
section, its centre at R = Rc and a radius r = r0 . Introduce polar coordinates (r, θ) with
origin at R = Rc to describe the projection of a guiding-centre orbit on a vertical plane
through the torus’ symmetry axis. The guiding-centre drift velocity is split into two parts;
motion along the field line and guiding-centre drift across the field: v = vk + vd .
The parallel velocity has one component in toroidal direction and one in the vertical
plane. When the field line is turning around the minor axis due to the rotation transform,
the velocity component in the vertical plane will be in the poloidal direction; the parallel
velocity vk corresponds to a poloidal rotation with angular frequency Ω;
Bθ
vk,θ = vk = rΩ (5.31)
B
where Bθ in the poloidal component of the magnetic field and B is the absolute value of
the magnetic field. The perpendicular velocity component is caused by the guiding-centre
drift in vertical direction vd = vd Ẑ, given by eq. 2.20:
vT2
vd ≈ (5.32)
ωc R
where R is the major radius and vT is the thermal velocity. The cyclotron angular
frequency have different sign depending on the charge of the particle (positive ion or
negative electron). The perpendicular angular velocity is divided into polar components:

vd,r = vd sin θ
vd,θ = vd cos θ (5.33)

In the vertical plane, the particle position is described by radius r as a function of the
poloidal angle; rp = rp (θp ), where θp = θp (t). The expressions for the radial and poloidal

113
velocities are:
drp
= vd,r
dt
dθp
rp = vk,θ + vd,θ (5.34)
dt

drp
= vd sin θp
dt
dθp vd
= Ω + cos θp (5.35)
dt rp

Assuming |vk |  |vd |, i.e.

dθp
≈Ω (5.36)
dt
We can write
drp drp dθp
= (5.37)
dt dθp dt
drp
≈ d sin θp (5.38)
dθp
vd
d= (5.39)
Ω
Chose the particle paths coinciding with the magnetic flux surface rp = r0 at θp = π/2
and integrate:
rp = r0 − d cos θp (5.40)
The above expression for the drift surface describes a circle with radius r0 and centre at
Rd = R0 − d, displaced in the horizontal direction from the flux surface by distance d.

r0 B vT2
d≈ (5.41)
vk Bθ ωc R

The sign of d depend on the sign of ωc and vk , i.e. the charge of the particle and the di-
rection of its motion parallel to the magnetic field. The expression for d may be simplified
by assuming vk ≈ v⊥ ≈ vT and introducing the Larmor radius rL and the safety factor q

|d| ≈ qrL (5.42)

5.2.2 Particle model for Pfirsch-Schlüter diffusion

One type of collision driven neoclassical transport in tokamaks caused by the fact that
the guiding-centre orbits do not exactly coincide with flux surfaces is called Pfirsch-
Schlüter transport. An approximation of the magnitude of this collision driven diffusion
can be made using the random walk model. The basis for the description of the transport
mechanism in this model is that particles moving along magnetic field lines in opposite
direction have different guiding-centre orbits. A guiding-centre orbit is displaced inwards
and outwards in the horizontal direction relative to a magnetic flux surface depending on
the direction of the particle’s motion along the field. When particles collide, they change
their direction along the magnetic field line, thus changing guiding-centre orbits. The
new drift orbit is close to a magnetic flux surface of greater radius. This means that

114
 z
Magn. fluxytor

Figure 5.9: Schematic illustration of collisions causing Pfirsch-Schlüter diffusion.

+ B

<
θ

<
z

Figure 5.10: Schematic picture of magnetic mirror field.

the particles, at each collision, can move one step length, corresponding to the distance
between the flux surfaces, which is of the same order of magnitude as the displacement d
of the guiding-centre orbit relative the flux surface (figure 5.9).
Inserting the distance d in the random walk model expression for the particle diffusion
coefficient gives the diffusion coefficient for this transport process.
d2
DP S = (5.43)
2τ
The expression for the distance d is

DP S ≈ q 2 DC (5.44)

where DC is the classical diffusion coefficient.

5.2.3 Trapped particles in a tokamak

Another type of neoclassical transport is caused by particles trapped between magnetic
mirror fields in the outer part of the torus where the toroidal magnetic field is weakest.
Consider a rotationally symmetric magnetic field with a gradient along the axis of symme-
try (figure 5.10). A charged particle gyrating around the axis of symmetry is, according
to Lorentz’ force law, influenced by a magnetic force along the axis of symmetry directed
towards the weaker magnetic field. For a particle of charge q in a circular orbit with the
Larmor radius rL the force becomes:

Fz = −qvθ Br (5.45)

Use ∇ · B = 0 to rewrite as
1 ∂Bz
Fz = qvθ rL (5.46)
2 ∂z
115
 Fm Fm
θ
B min z
B

Figure 5.11: Schematic drawing of trapping of a particle between two magnetic mirror
fields

We can rewrite the expression before the derivative as:

1 mv 2
qvθ rL = − ⊥ = −µ (5.47)
2 2B
where the magnetic moment µ has been introduced. Finally, we have the following ex-
pression for the magnetic mirror force Fz
∂Bz
Fz = −µ . (5.48)
∂z
A particle gyrating with a given velocity v⊥ in a homogeneous magnetic field B will have
a constant magnetic moment µ. Since the magnetic field varies very little over space, the
motion will be almost periodic (gyration around a magnetic field line and a slow drift).
If the variation of the magnetic field is negligible over a Larmor radius, the magnetic
moment will remain constant. (The magnetic moment is a so called adiabatic invariant.)
This can be used to construct a condition for reflection of a charged particle against a
magnetic mirror field. Consider a particle with velocity vk0 and v⊥0 in an area with weak
magnetic field Bmin surrounded by two magnetic mirror fields (figure 5.11). The magnetic
momentum µ is assumed to be constant when the particle is moving along the magnetic
field line:
2
v⊥ v2
= ⊥0 . (5.49)
B Bmin
Further, the kinetic energy of the particle is constant, i.e.:
1 1
m(vk2 + v⊥
2 2
) = m(vk0 2
+ v⊥0 ). (5.50)
2 2
At the reflection point, vk = 0 which gives a condition on the magnetic field strength:
2
B 2
v⊥ vk0
= 2 =1+ 2 . (5.51)
Bmin v⊥0 v⊥0

To describe trapping of particles in the outer part of a tokamak, we assume the magnetic
field strength to be equal to the magnitude of the toroidal magnetic field, which is the
dominating magnetic field component in a tokamak (figure 5.12). The toroidal field is
approximately given by the expression
R0 1
B ≈ B0 = B0 (5.52)
R 1 + ε cos θ
where the inverse aspect ratio ε has been introduced:
r
ε= . (5.53)
R

116
 Δ
z F= μ B
φ II

B max
B min
B r θ VII o , VI o

I
Ro R

Δ
F= μ II
B

Figure 5.12: Trapping of particles in the outer part of a tokamak.

The parameter will be of importance for the neoclassical transport of trapped particles
in the tokamak. The variation of the magnetic field strength along field lines decides
whether particles can be trapped:
Bmax
≈ 1 + 2ε. (5.54)
Bmin
The condition for trapping of particles in the outer part of the tokamak is given by
vk0 √
< 2ε. (5.55)
v⊥0
It can be shown that the ratio of trapped particles is

ftrap ∼ ε1/2 . (5.56)

5.2.4 Banana orbits

The name banana orbits is related to the shape of the drift orbits for the particles trapped
in the outer part of the tokamak. The orbits are caused by the vertical guiding-centre
drift being added to an oscillating drift motion along magnetic field lines between two
mirror points. The drift orbits will be displaced relative the the magnetic flux surface.
The displacement will be largest at the center, and its direction (inwards or outwards) will
depend on the direction of the particles motion along the field line. Thus when the particle
oscillates back and forth along the field line the displacement will vary in direction. The
result will be a “banana shaped” drift orbit (figure 5.13).
An expression for the half width wb of the drift orbit (at the widest part in the center)
is given by the following discussion. The drift orbits can be obtained by dividing the
drift velocity into two parts, motion along the field line and guiding-centre drift across
the field. The parallel motion is given by the particles force balance equation along the
magnetic field:
d2 s dB
m 2 = −µ (5.57)
dt ds
where s is a coordinate along the field line. Consider a motion close to the midplane,
where the expression for the magnetic field may be approximated as:
 
1 2
B ≈ B0 1 − ε + ε θ (5.58)
2

117
 z
φ
"Banana orbit"

r θ
θb R

Δr=Wb

Figure 5.13: Projection of a guiding-centre orbit in the vertical plane - a banana orbit

and the relation between s and θ

rθ Bθ
= . (5.59)
s B
We then obtain
d2 s
= −ωb2 s (5.60)
dt2
v2 B 2 ε v⊥02
ωb2 = ⊥0 θ2 = . (5.61)
2RrB 2 R2 q 2
This differential equation describes an oscillatory motion back and forth along the field
line with “ bounce frequency ” ωb . If we use the poloidal angle θ, the expression for the
oscillatory motion becomes
θ = θb sin ωb t. (5.62)
The magnetic field at the turning point is
Bb ε
≈ 1 + θb2 (5.63)
Bmin 2
which gives the expression for the poloidal angle θb of the turning point :
r
2 vk0
θb = . (5.64)
ε v⊥0
The perpendicular drift velocity vd is divided into polar components in the same way as
in section 5.2.3. In the vertical plane, the particle orbit is described by radius rp as a
function of the poloidal angle, rp = rp (θp ) , where θp = θp (t). From the expressions for
radial and poloidal velocity we have
drp
= vd sin θp
dt
dθp vd
= θb ωb cos ωb t + cos θp . (5.65)
dt rp
Assume vk  vd , i.e.
dθp
≈ θb ωb cos ωb t. (5.66)
dt
Further assume θp so small that sin θp ≈ θp . We have
drp
≈ vd θp
dt s  2
dθp θp
≈ θb ωb 1 − . (5.67)
dt θb

118
The particle orbit is given by the differential equation
drp vd θp
≈ s  2 . (5.68)
dθp θb ωb θp
1−
θb

For a particle orbit coinciding with the magnetic flux surface r = r0 at the turning point
θ = θb the expression for the guiding-centre orbit is
 2 !
θp
(rp − r0 )2 = wb2 1 − (5.69)
θb

where
vd θb
wb = . (5.70)
ωb
The expression describes a “banana shaped” drift orbit with half width wb in the midplane
(figure 5.13). Introducing the approximations v⊥0 ≈ vT and vk0 /v⊥0 ≈ ε1/2 , then gives a
simple expression for the banana-orbit half-width:

wb ≈ ε−1/2 qrL . (5.71)

Here, rL is the Larmor radius and q is the safety factor.

5.2.5 Banana orbit diffusion

Banana orbit diffusion denotes the type of neoclassical transport caused by Coulomb
collisions for particles trapped in the outer midplane of the tokamak. In a simplified
way, the diffusion process can be described by considering particles moving in different
directions along magnetic field lines. When colliding, the particles might change their
direction, thus changing drift orbit. This means a motion to a magnetic flux surface at a
larger radius. The particles will step by step move radially corresponding to the distance
between the magnetic flux surfaces. This distance is of the order of the half width of the
banana orbit wb (figure 5.14). As mentioned, this argument is simplified. It can be shown
that the dominating contribution to diffusion actually comes from collisions causing the
particles to become untrapped, or vice versa. Since the half width of the banana orbit wb
for the trapped particles is much larger than the displacement d of the drift orbit of the
untrapped particles, the step length is still approximately equal to wb .
Banana orbit diffusion can be described by the random walk model. In this case it is
the half width wb which is the step length in the expression for the diffusion coefficient.
This expression now becomes a little bit more complicated,

wb2
DB = ftrap . (5.72)
τeff
Here, ftrap is the ratio of trapped particles and τeff is an effective collision time corre-
sponding to the particle confinement time in the magnetic mirror field. It can be shown
that

τeff ≈ ε τ (5.73)

119
 z Magn. flux surfaces
φ
Drift orbit
r

Ro R

Wb

Figure 5.14: Somewhat simplified illustration of the collisions causing banana orbit diffu-
sion.
D 2
q Dc
D PS
D
c
2

/
q
2
3
ε-

rL2
Dc τ
B
D

Classical

3/
2 Vth Vth n
ε qR qR

Figure 5.15: Neoclassical diffusion coefficient as a function of the collision frequency.

where τ is the usual confinement time (defined as the time required for 90◦ deflection
a test particle’s velocity due to the total effect of all collisions). The banana diffusion
coefficient is finally
DB = ε−3/2 q 2 DC (5.74)
where DC is the classical diffusion coefficient.

5.2.6 Summary of neoclassical transport

Which neoclassical diffusion process is dominating for a given case (Pfirsch-Schlüter or
banana orbit) depends on the relation between the collision frequency of the particles and
their bounce frequency (between the mirror fields). If the collision frequency is sufficiently
high, the particles will disappear to the mirror field “loss cone”. Thus, the particles are
not reflected in the mirror field, but will travel around the tokamak along the field lines
and Pfirsch-Schlüter diffusion dominates. In the opposite case, banana orbit diffusion
dominates.
If particles are to be trapped, the collision frequency ν = 1/τ must be sufficiently
small:
νeff < ωb (5.75)
where νeff = ν/ε. Inserting the expression for the bounce frequency with v⊥ ≈ vT , gives
the condition for banana orbit diffusion as
vT
ν < ε3/2 . (5.76)
qR

120
 z
φ Δ
B

B + + +
+ Vion +
B + + V--I =
1
2 Ex B
B
θ

r
R
j ll j ll
Eθ B E θ j ll
- Velec. -
- - -
- -

Figure 5.16: Fluid model for derivation of Pfirsch-Schlüter diffusion.

Pfirsch-Schlüter requires that the particles collide during a poloidal rotation along the
field line. This condition can be written as
vT
ν> (5.77)
qR
for Pfirsch-Schlüter diffusion. The neoclassical diffusion coefficient as a function of the
collision frequency is shown in figure 5.15. There is an area between banana orbit diffusion
and Pfirsch-Schlüter diffusion which is usually called the plateau regime .

5.2.7 Fluid model for Pfirsch-Schlüter diffusion∗

An alternative derivation of the Pfirsch-Schlüter diffusion is based on the fluid model
(figur 5.16). The vertical guiding-centre drift, which has different direction for ions and
electrons causes a vertical electrostatic field E. This electric field is in principle short
circuited by an electric current density j flowing along the magnetic field lines since these
rotate around the minor axis in the tokamak, connecting the upper and lower parts of
the torus. Now, the condition ∇ · j = 0 must be fulfilled, and the plasma has a certain
resistivity. Thus, Ohm’s law will have a component of the electric field perpendicular to
the magnetic field. This will, in turn, give a radial E × B velocity, corresponding to the
Pfirsch-Schlüter diffusion. When we used the fluid model to derive the classical diffusion
coefficient, we neglected the E × B term.

5.3 Experimentally measured transport

To better understand the processes causing energy losses in the plasma, experimental
studies of the radial profile of the energy balance are performed for both particle types.
The plasma energy balance in a fusion experiment can in principle be described as follows
(figure 5.17). Heating takes place in the central plasma (1). Transport processes cause
a radial transport of energy and particles (2). This leads to energy losses and plasma
wall interaction in the plasma edge (3). In addition, electrons suffer a direct energy loss
caused by emission of electromagnetic radiation. Ions interact with neutral gas causing
direct energy losses by charge exchange collisions.

121
 Radiation
3 from edge

Radial
2 transport
Ohmic
1 heating
ELECTRONS Radiation

Additional Energy
1 heating exchange
Charge
IONS exchange

2 Radial
transport

Charge exchange
3 from edge

Figure 5.17: Schematic picture of the plasma energy balance.

60

τE (Neoclassical)
40
τE
(ms)

20

τE (Experimental)
0
0 1 2 3 4 5 6
20
ne (10 m-3 )

Figure 5.18: Comparison between experimental and neoclassical confinement time in the
Alcator tokamak at MIT (1978).

5.3.1 Confinement scaling

The most fundamental empirical measurement of transport in a tokamak is the global
confinement time τE . The value of τE is important to measure, but more important is
the scaling of τE with plasma quantities such as density n, temperature T , minor radius
a, major radius R and safety factor q. Table 5.1 shows examples of some empirically
measured scaling laws for tokamaks with ohmic heating. The difference between the
scaling laws is due to different sets of data being analyzed, and the different choices of
scaling parameters.
The transport measured experimentally is often higher that predicted by neoclassical
theory. This is especially true for electrons while the neoclassical theory works better for
ions. Early tokamak experiments showed this deviation from neoclassical theory (figure
5.18).
Experiments show that the energy confinement time increases linearly with the density
n, as stated by the the empirical scaling laws. This is in strong contrast to the variation

122
 Table 5.1: Scaling laws for τE in an ohmically heated tokamak

Scaling law τE (s)

Intor (1978) 5 × 10−21 n a2
Merezhkin -Mukhovatov (1981) 3.5 × 10−21 n a0.25 R2.75 T −0.5 qa
Goldston (1984) 1.0 × 10−21 n a1.04 R2.04 qa0.5

with density predicted by neoclassical theory.

At the time when JET was being designed there was evidence from the Alcator toka-
mak at the Massachusetts Institute of Technology (MIT) that a fit to empirical data gives
the scaling relation, valid for ohmically heated (L-mode) plasmas, referred to as Intor
(1978) scaling, see Table 5.1. Although this relation was only a tentative attempt to
represent the plasma behaviour it did provide some estimate of what JET might achieve,
at least with ohmic heating. It was not known then at what particle density JET would
operate, but a typical tokamak value of n = 3 × 1019 m−3 , with an average plasma radius
of 1.4 m, gave 0.3 seconds. (Presently more accurate L-mode scaling have been deter-
mined.) A difficulty in the JET design was that tokamak experiments before the design
of JET had low temperatures, typically less than 1 keV, and confinement times measured
in milliseconds. Predictions based on such experiments were therefore clearly unreliable
and, of course, one of the main purposes of JET was to study confinement and reduce the
uncertainties of extrapolation. The early experiments on JET gave encouraging results.
The confinement times in ohmically heated plasmas were up to two times greater than
predicted, a confinement time of 0.6 seconds being achieved.
Since an important parameter to be investigated is the size of the plasma (and the
associated magnitude of the plasma current) it is clearly necessary to use results from a
range of different tokamaks. The following scaling relation has been obtained for H-mode
confinement:
τEH = 0.029I 0.90 B 0.20 P −0.66 n0.40 M 0.20 R1.84 a0.19 κ0.92 . (5.78)
where I is the current in MA, B is the toroidal magnetic field in teslas, P is the heat-
ing power in MW, n is the density in units of 1019 m−3 , M the atomic mass, R and a
are the major and minor radii in metres, and κ is the elongation ratio of the plasma
(height/width). This empirical relation represents the data quite well but is determined
without any concern for the physics involved. This is an unsatisfactory situation and
attempts are made to rectify this weakness in theoretical understanding.

5.4 Models for anomalous transport

The deviation between experiment and theory has motivated studies of models for anoma-
lous transport. In general, the transport processes not described by classical or neoclas-
sical transport theory are called anomalous. The anomalous transport is usually driven
by different collective plasma phenomena, i.e. correlated motion of the plasma particles.
It is the total effect of many correlated particles which cause the transport process. This
is different from the collision driven transport described by neoclassical theory where the
field particles are assumed to be uncorrelated during Coulomb collisions. This means
that fluid models are usually needed for treating anomalous transport, since they take
collective plasma effects into account. The correlated particle motion is often driven by

123
 Eθ
Eθ
z Vr = B
θ B
Vr

Figure 5.19: Cylinder model for calculation of fluctuation-driven radial particle flux.

some kind of plasma instability and is usually observed as, for example, fluctuations in
local plasma density. These fluctuations are called plasma turbulence. The fluctuation
driven transport is usually much larger than the neoclassical. In fusion experiments, it is
desirable to suppress the fluctuations in order to minimize anomalous transport.
The anomalous transport processes can roughly be divided into two groups. Transport
driven by electrostatic fluctuations (1) and transport driven by electromagnetic fluctua-
tions (2). The electrostatic fluctuations can give a moderate increase of transport. This
kind is present in tokamaks and is thought to cause the anomalous transport of electron
energy which can be observed experimentally. If magnetic fluctuations are present they
can cause the magnetic field to become stochastic (this is discussed later), and the result-
ing energy flow along magnetic field lines is one kind of electromagnetic turbulence driven
transport.

5.4.1 Particle flux from electrostatic fluctuations

A general expression for the radial particle flux driven by electrostatic turbulence in a
cylindrical plasma can be derived. An electric field perpendicular to the magnetic field
gives a fluid velocity across the magnetic field according to eq. 5.10. Just as when
treating plasma waves in chapter 4, quantities are divided in an equilibrium part and a
perturbation part which is assumed to be small, e.g. E = E0 + E1 . A fluctuation electric
field E1 then causes a fluctuating radial velocity (figure 5.19):
1
v1 = E1 ×B0 (5.79)
B2
The radial flux is the product of velocity v and density n. For a constant velocity, a
fluctuating velocity causes a fluctuating radial flux, but not necessarily any transport. To
estimate the transport, the time average of the flux is required. The average radial flux
can be written as

Γ =< nv > (5.80)

where < · · · > denotes time average. Assume that the velocity only has one fluctuating
part, i.e.
E1
v = v1 = . (5.81)
B0
The radial flux is then
Γ =< (n0 + n1 )v1 >=< n1 v1 > (5.82)

124
 Δ
y n0 n
+
2π
E1 E1 λ= ky
V1= B -
B
x
z

Figure 5.20: Schematic picture showing the origin of drift waves.

or, alternatively
1
Γ= < n1 E1 > . (5.83)
B0
Transport will only occur if the density fluctuation and the velocity fluctuation are cor-
related and “in phase”, i.e. the time average of the fluctuations must not be zero.

5.4.2 Transport caused by drift waves

Consider a planar model of an inhomogeneous plasma. The plasma has a gradient in x
direction and there is a homogeneous magnetic field B in z direction. Assume a density
perturbation of wavelength λ in y direction (see figure 5.20). The wave vector’s z compo-
nent is assumed small, so that electrons can move along field lines perpendicular to the
wave front. The force balance along field lines requires that the pressure gradient force
caused by the density perturbation is balanced by an electrostatic force. This electrostatic
force is caused by a space charge field arising when the plasma tries to uphold its charge
neutrality ne ≈ ni ≈ n. At the limit mi  me we get a simple correlation between density
n and plasma potential ϕ usually called Boltzmann’s relation:
 
ϕ
n = n0 exp . (5.84)
Te
The electric field in the y direction causes a E × B drift in x direction. This velocity
perturbation, also with period λ in y direction, is 90◦ out of phase with the density
perturbation. This causes a so called drift wave propagating in y direction. Drift waves is
a source of electrostatic fluctuation in the plasma which may cause anomalous transport.
Since the waves are driven by the density gradient in the plasma, these waves are always
possible. Because of this, the drift wave instability is called anuniversal plasma instability.
A simple estimate of the anomalous particle transport caused by drift waves can be
obtained as follows: The perturbation is described with a plane wave anzats where the
perturbation part of all quantities is assumed to vary in space and time as
∼ exp i(ky − ωt) (5.85)
where k = 2π/λ. Perturbation variables n1 , v1 , E1 . . . henceforth denote the complex
amplitude of the perturbation. When ϕ0 = 0 and ϕ1  Te , we can approximate the
exponential function with the first term of the Taylor expansion
n1 ϕ1
≈ (5.86)
n0 Te
The velocity perturbation is
E1 ϕ1
v1 = = −ik . (5.87)
B0 B0

125
The continuity equation provides a correlation between v1 and n1 (after linearisation
where terms of 2nd order are neglected).
dn0
v1 − iωn1 = 0. (5.88)
dx
The phase velocity of the wave can be written as
ω Te dn0
vDe = =− . (5.89)
k B0 n0 dx
The phase velocity for drift waves vDe is equal to the so called diamagnetic drift velocity
for electrons. In this approximative model, the drift wave is oscillating with a constant
amplitude. The density and velocity perturbations are 90◦ out of phase, which means no
radial particle flux. In a realistic plasma model with finite ion mass, the inertia of the ions
causes a phase difference between density and velocity perturbations. E.g., the velocity
perturbation v1 is in the outwards direction when the density perturbation is already
displaced outwards. Thus, a so called drift instability arises. Regard the perturbation
as a plane wave with increasing amplitude. The phase difference between n1 and ϕ1 can
be expressed as a function of the growth rate γ in the following way: All perturbation
quantities are assumed to vary in time and space as

∼ exp (γt + i(ky − ωt)) . (5.90)

In this case, the continuity equation gives

dn0
v1 + (γ − iω) n1 = 0. (5.91)
dx
Extracting n1 gives
ik ϕ1 dn0
n1 = . (5.92)
(γ − iω) B0 dx
When γ/ω  1 we can use a Taylor expansion to approximate the expression as
k γ  ϕ1 dn0
n1 ≈ − 1−i . (5.93)
ω ω B0 dx
Now assume ω/k ≈ vDe . The relation between n1 and ϕ1 is
n1  γ  ϕ1
≈ 1−i . (5.94)
n0 ω Te
The phase difference between n1 and ϕ1 cause a particle flux, i.e. the drift instability
results in transport. The particle flux
1
< Γ >=< n1 (t)v1 (t) >= Re{n1 v1∗ } (5.95)
2
which gives
1 n0 γϕ21 k
< Γ >= . (5.96)
2 B0 Te ω
To estimate the size of the flux we introduce the characteristic length δn for the gradient
of the equilibrium density  
1 1  dn0 
= . (5.97)
δn n0  dx 

126
 n0
n

Figure 5.21: Schematic drawing showing the mixing length approximation.

Introducing the expression for phase velocity then gives

 2
1 ϕ1
< Γ >≈ γn0 δn . (5.98)
2 Te

An alternative is to express the flux in terms of the density perturbation n1

 2
1 n1
< Γ >≈ γn0 δn . (5.99)
2 n0

A diffusion coefficient describing drift wave transport can finally be obtained using Fick’s
law  2
1 2 n1
DD ≈ γδn . (5.100)
2 n0
A very approximate, but simple, estimate of DD is reached if we assume that the amplitude
of the density perturbation is of the same order of magnitude as the gradient of the
equilibrium density
n0
kn1 ≈ . (5.101)
δn
This should be a upper limit for the perturbation amplitude, which in this case cancel
the gradient of the equilibrium density, which is the driving force for the instability, see
figure 5.21. The drift wave diffusion coefficient is finally
γ
DD ≈ . (5.102)
k2
In the literature, this approximation is called the mixing length approximation of the
turbulence driven diffusion coefficient.

5.4.3 Stochastic magnetic field∗

Overlapping of magnetic islands on adjacent magnetic flux surfaces cause the magnetic
field to become stochastic (or ergodic); the flux surfaces cease to exist in the strict sense. If
we follow a field line multiple turns around the torus in this situation, the field line might
move radially in a stochastic manner over a considerable area. Thus the term stochastic
magnetic field. In extreme cases, a field line originating in the centre of the plasma might,
after a number of toroidal turns, end up at the boundary of the plasma. This phenomenon
can be studied using numerical calculations. Figure 5.22 shows the result of a simulation
using a point plot of the magnetic field. The point plot shows a vertical cross section of
the plasma where a magnetic field line is represented by a point for every toroidal turn
when it passes through the surface. In the normal case, without magnetic perturbations,
a magnetic field line will form an unbroken line in the graph, i.e. the cross section of the
magnetic flux surface (figure 5.22a). In the simulation, there are initially small magnetic

127
Figure 5.22: Numerical simulation of gradually increasing magnetic perturbation ampli-
tude.

perturbations at the two surfaces (figure 5.22b), but this does not influence the rest of the
plasma. When the amplitude of the perturbations increases an interesting phenomenon
occurs. Magnetic field lines leave the flux surfaces and are visible as individual points at a
certain distance away from these (figure 5.22c). When the amplitude of the perturbation
is sufficiently large, the magnetic field will look like in figure 5.22d. The magnetic field
lines wander randomly between the resonant surfaces. This is seen in the graph as an
even pattern of points between the surfaces. The magnetic field in this area is stochastic.
Since transport along field lines is much larger than across the magnetic field, plasma
confinement is heavily affected when the magnetic field becomes stochastic. An impor-
tant criterion telling us whether a magnetic field is stochastic or not is if overlapping of
magnetic islands occurs, i. e. if the radial extent (island width W ) of islands on adjacent
magnetic flux surfaces is larger than the radial distance between flux surfaces.

5.4.4 Transport in a stochastic magnetic field∗

To describe how fast a magnetic field line moves in the radial direction when we follow
it in toroidal direction in a stochastic magnetic field, a diffusion coefficient DM can be
defined. Assume that the magnetic field fluctuations have an auto correlation length lc ,
characterizing the distance along the magnetic field over which the magnetic field per-
turbation is correlated. A somewhat simplified interpretation of lc is that the correlation
length is equal to the distance over which the magnetic field perturbation has the same
radial direction. A simple model for the diffusion of the magnetic field can then be ob-
tained if lc is used as the step length in a random walk process. Assume that a magnetic
field line in a tokamak is followed for several turns in toroidal direction. (The magnetic
field in a tokamak is mainly directed in the toroidal direction.) The field line moves a

128
distance δr radially while followed a distance lc in the toroidal direction:
δr δBr
= . (5.103)
lc B
The correlation length in this case has the same meaning as the collision time in the
random walk model of the particle diffusion coefficient:

(δr)2
DM ≈ . (5.104)
2lc
It is obvious that the dimension of DM is meter and it may thus only be regarded as
a diffusion coefficient in a mathematical sense. For an order of magnitude estimate of
DM , δBr can be put equal to the radial amplitude of the magnetic field perturbation
δBr ≈ B1r . The expression for the magnetic field line diffusion coefficient is
 2
B1r
DM ≈ lc . (5.105)
B0

Consider the energy transport for electrons in a stochastic magnetic field. Depending on
the collision frequency of the electrons ν there are two cases. These are characterized by
the relation between the electron mean free path length along the field lines λ and the
correlation length of the magnetic field lc :

1. collision free area: λ > lc

2. collision dominated area λ < lc .

In the collision free area (λ > lc ) the electrons are assumed to move radially the same
distance δr as magnetic field lines when they move along the field lines. The radial heat
diffusivity perpendicular to the magnetic field χe⊥ is estimated in the same way as the
particle diffusion using the random walk model:

(δr)2
χe⊥ ≈ . (5.106)
2τ
In this collision free model, the collision time is replaced by the time interval τ during
which electrons move the distance lc ;
lc
τ= (5.107)
vT
where vT is the thermal velocity of the electrons, and
 2
B1r
χe⊥ ≈ DM vT = lc vT . (5.108)
B0

The correlation length lc can be estimated by numerical simulations of the magnetic

turbulence.
In the collision dominated area λ < lc it is assumed that the electrons move a distance
between collisions given by the mean free path of the electrons along field lines λ:
δBr
δr = λ. (5.109)
B

129
The heat diffusivity along field lines can also be estimated using the random walk model:

λ2
χek ≈ . (5.110)
2τ
Insertion in the random walk estimate of the radial heat diffusivity gives, using the ap-
proximation δBr ≈ B1r :
2 2
λ2 δBr
 
B1r
χe⊥ ≈ = χek . (5.111)
2τ B0 B0

In both cases (collision free and collision dominated) the heat diffusivity is proportional
to the magnetic fluctuation amplitude squared. (Compare to the diffusion coefficient for
electrostatic fluctuations) When the parallel heat transport χek is very high, the radial heat
transport across the magnetic field can be significant already at low levels of fluctuation.

130
Chapter 6

RADIATION, BOUNDARY
LAYER, HEATING

One of the most important processes causing energy losses in a fusion plasma is emission
of electromagnetic radiation. In the central plasma, which is fully ionized, bremsstrahlung
is dominating. In a magnetized plasma there is also cyclotron radiation caused by the
gyrating motion of the electrons in the magnetic field. This kind of radiation is, however,
mainly reabsorbed in the plasma.
In the boundary layer of the plasma there are ions which are not fully ionized. These
cause line radiation in this region. This is also the case for the central plasma if the
temperature is low. Line radiation in the boundary layer is often desirable, since energetic
ions and atoms from the central plasma may transform their kinetic energy to radiation.
The advantage is that they hit the wall with lower energy, which limits the problems
associated with plasma wall interaction. By plasma wall interaction, we mean a number of
processes taking place at the plasma boundary. This often involves energetic ions or atoms
from the plasma hitting the surrounding walls. It can be sputtering or melting of the wall
material, which then cause atoms from the wall to enter the plasma as contaminations.
These contaminations increase the radiation losses from the plasma, and for this reason, it
is desirable to avoid intense plasma wall interactions. Devices designed to control plasma
wall interaction in fusion reactors are so called limiters or magnetic diverters, discussed
later in this chapter.
To reach the temperatures required for a fusion reactor energy must be supplied to the
plasma somehow. There are three main methods used: 1) Resistive, or ohmic heating. 2)
Injection of high energy beams of neutral atoms. 3) Heating with electromagnetic waves
in the radio frequency range.

6.1 Emission of electromagnetic radiation

Emission of electromagnetic radiation is mainly caused by acceleration of electrons in
the plasma. In a fully ionized plasma, the emission from free electrons is dominating
(bremsstrahlung, cyclotron radiation). In a plasma not fully ionized, radiation from bound
electrons is also present. This is known as line radiation.

6.1.1 Bremsstrahlung
Coulomb collisions between electrons and ions cause acceleration of charged particles
which leads to so called bremsstrahlung. Since the electrons are much lighter than the

131
 d
-e, v d
d

+Ze

Figure 6.1: Schematic illustrating the approximation used in the Bremsstrahlung power
estimate.

ions, the electrons will be more accelerated and thus make the dominating contribution
to the bremsstrahlung emission from the plasma.
A simple semi-classical estimate of the emitted bremsstrahlung power can be made
in the following way: The bremsstrahlung power from an electron in a Coulomb collision
is given by a general expression from electromagnetics for the electromagnetic radiated
power emitted from an electron with charge −e and acceleration a.

e2 a2
P = . (6.1)
6πε0 c3
Here c is the velocity of light. The acceleration a is given by the expression for the
Coulomb force
F Ze2
a= = (6.2)
me 4πε0 me r2
where r is the distance between the charged particles. The radiated power increases
rapidly with decreasing distance, P ∼ r−4 , which means that so called close collisions
are dominating. (See the section on Coulomb collisions). Quantum mechanics give the
following estimate of the smallest possible distance d = rmin (see Heisenberg’s uncertainty
relation);
h
d= (6.3)
2πme v
where h is Plank’s constant and v is the electrons velocity. An estimate of the emitted
energy δE from a Coulomb collision can be made using the following simplified model of
the collision (see figure 6.1). Assume that the electron passes on a distance d away from
the ion and that the interaction occurs during a time t corresponding to the time needed
for the electron to travel a distance 2d along it’s path, or
2d
t= . (6.4)
v
Emitted energy per collision is given by eq. 6.1, 6.2 and 6.4 as

Z 2 e6
δE = P · t = . (6.5)
6(2πε0 cd)3 m2e v

The cross section for this close collision is

σ = πd2 (6.6)

132
 2
Pbr
Pα
1

0
0 5 10 15
T(keV)

Figure 6.2: The relation between bremsstrahlung power and alpha particle heating.

and the collision frequency for an electron passing ions of density ni is

ν = ni σv. (6.7)

The total emitted bremsstrahlung power density for electrons of density ne and velocity
v is finally given by eq. 6.5–6.7

ne ni Z 2 e 6 v −38 2
p
PB = ne νδE = = 1.2 × 10 n n
e i Z Te [W/m3 ]. (6.8)
24πε30 c3 hme

The velocity of the electrons is assumed to be v = vT . This assumption is used to obtain

an approximative expression for the bremsstrahlung power as a function of the electron
temperature Te . A more advanced quantum mechanical calculation gives a more accurate
estimated of the bremsstrahlung power per unit volume
p
PB = 1.7 × 10−38 ne ni Z 2 Te [W/m3 ]. (6.9)

In a multi-species plasma, we may use the definition of the effective charge, eq. 2.72, and
write p
PB = 1.7 × 10−38 n2e Zeff Te [W/m3 ]. (6.10)
The temperature is inserted in unit electron volt (or simply volt).
The bremsstrahlung loss provides a criterion for the lowest possible electron temper-
ature in a fusion reactor (approx. 5 keV). Since the heating power from alpha particles
increase faster with higher temperature than the bremsstrahlung loss, a positive net effect
is possible when the electron temperature is higher than this value (figure 6.2).

6.1.2 Cyclotron radiation

The gyrating motion of charged particles in the magnetic field means acceleration and
thus electro magnetic radiation, according to eq. 6.1. Since the electrons have lower mass
than the ions, they will be more accelerated, and thus give the dominating contribution.
This radiation is called cyclotron radiation. If the electrons have non relativistic velocities
and synchrotron radiation when the velocity is relativistic (v ≈ c).
An expression for the power of the electron cyclotron radiation is given by the general
expression for the electron centripetal acceleration
r
2 eB 2eTe
a = ωce rL ≈ ωce vT = . (6.11)
me me

133
Inserting eq. 6.10 in eq. 6.1 gives the following expression for emitted power per unit
volume for a given electron density ne :
e5 B 2 Te
PC ≈ n e = 6.1 × 10−20 ne B 2 Te [W/m3 ]. (6.12)
3πε0 c3 m3e
The cyclotron radiation frequency spectrum shows that the radiation is emitted at
certain frequencies corresponding to the electron cyclotron frequency and its higher har-
monics.
ωce
f =n (6.13)
2π
where n = 1, 2, 3 . . .. In a fusion reactor plasma it is likely that re-absorption of radiation
at these frequencies is very large and the power loss from the plasma surface will be
small. For this reason we often neglect this power loss when analysing the energy balance
of the fusion plasma. Electron cyclotron radiation, however, has great importance as a
diagnostic used to measure the electron temperature in the plasma.

6.1.3 Line radiation

By line radiation we mean the electromagnetic radiation emitted by electrons bound in
so called excited orbits around the nucleus. The radiation is emitted when the electrons
spontaneously “fall down” to their respective ground states. Since the excited states,
according to quantum mechanics, only have certain allowed energy levels, radiation is only
emitted at certain wavelengths. At these wavelengths, the photon energy corresponds to
the energy difference between the excited state and the ground state. These wavelengths
are seen as lines in an experimentally measured wavelength spectrum. Hence the name
line radiation.
In a fusion plasma hydrogen ions are fully ionized, except in the boundary layer close
to the wall. Therefore, line radiation does not cause any significant power loss. Power
losses from line radiation are, however, caused by impurity ions. By impurity ions we
mean all other, unwanted, ions in the plasma. These usually occur in very small amounts.
Impurity ions are divided into two groups, depending on the charge number Z:
1. Low Z impurities, e.g. oxygen (O), carbon (C)
2. High Z impurities, e.g. iron (Fe), tungsten (W)
Low Z impurities like oxygen and carbon are present in the atmosphere and can enter the
plasma via the rest gas in the vacuum vessel. High Z impurities can be formed through
interactions with the walls of the vacuum vessel. These are usually made of stainless
steel containing Fe, Cr and Ni. Impurity ions not fully ionized will emit line radiation.
At a certain electron temperature, a certain impurity will give rise to impurity ions with
different charge number Z, depending on the number of times each ion has been ionized.
The impurity density ni can be written as a sum over the ion densities in the different
charge states.
XZ
ni = ni,z . (6.14)
z=0

The average charge number of the impurity Z i is defined as

Z
1 X
Zi = z ni,z . (6.15)
ni z=0

134
Charge neutrality, which is a fundamental property of a plasma, then gives the electron
density as X
ne = Z i ni . (6.16)
i

The distribution of density in the different charge states is mainly dependent on the
plasma electron temperature.
The distribution of charge state densities can be estimated with a model for the ion-
ization and recombination in the plasma in the following manner: The density in a given
charge state z, for a given ion type i, is described by the following system of ionization
balance equations
dni,z
= ni,z+1 ne αi,z+1 − ni,z ne αi,z + ni,z−1 ne Si,z−1 − ni,z ne Si,z . (6.17)
dt
In equilibrium, when ni,z = const, the density in charge state z is given by the balance
between recombination and ionization to and from adjacent states z − 1 and z + 1. These
two atomic processes are described by the recombination coefficient αi,z = αi,z (Te ) and
the ionization coefficient Si,z = Si,z (Te ). These are mainly dependent on the electron
temperature. Solving the system of equations (eq. 6.15) then gives the charge distribution
ni,z = ni,z (Te ) and average charge Z i = Z i (Te ) expressed in the electron temperature Te .
The density in a given excited state k, for a given charge number z is described by similar
excitation balance equations :
dni,z,k X
= {ni,z,l Ai,z,lk − ni,z,k Ai,z,kl + ni,z,l ne Ci,z,lk − ni,z,k ne Ci,z,kl } . (6.18)
dt l

The density of the excited state in equilibrium is given by the balance between excitation
and de-excitation processes; absorption and emission of radiation and electron collisions.
Here, the constants Ai,z,kl and Ai,z,lk are probabilities for absorption and emission; Ci,z,kl =
Ci,z,kl (Te ) and Ci,z,lk = Ci,z,lk (Te ) are collision coefficients which are functions of the
electron temperature. Under certain conditions (especially low electron density), the
excitation balance is mainly controlled by balance between excitation from the ground
state l = 1, caused by electron collisions and de-excitation by spontaneous emission
of radiation. This approximation is called corona equilibrium. The excitation balance
equation (eq. 6.16) for a state k can then be rewritten as
X
{−ni,z,k Ai,z,kl + ni,z,1 ne Ci,z,1k } = 0 (6.19)
l<k

from which ni,z,k can be extracted

Ci,z,1k
ni,z,k = ni,z,1 ne P . (6.20)
l<k Ai,z,kl

The density and thus the radiated power is proportional to the electron density ne and
the density in the ground state ni,z,1 , which is given by the ionization balance (eq. 6.15).
The total emitted power PR for a given ion type can be calculated as a function of the
electron temperature. The temperature dependence of the radiated power is described by
the radiation parameter Ri = Ri (Te );

PR = ne ni Ri (6.21)

135
 10-30
Tungsten 8 Oxygen
10-31 z
4 Carbon
-32
R 3 10 Iron 0
(Wm ) 60
10-33 z
Oxygen 40 Tungsten
-34
10 20 Iron
Carbon
(a) (b)
10-35 0
1 10 100 1 10 100 1 10 100 1 10 100
eV eV eV keV keV keV eV eV eV keV keV keV
Te Te

Figure 6.3: Radiation parameters R and average charge number Z as a function of electron
temperature

The above expression give the radiated power per unit volume. The radiation parameter
R also include the bremsstrahlung so that PR is the total emitted power for a given
impurity. The left graph in figure 6.3 shows the radiation parameter R for some common
impurity ions as a function of Te . The right graph in figure 6.3 shows the average charge
number Z, also as a function of Te . The figure shows that low Z and high Z impurities
have different effect on the plasma. Low Z impurities cause maximum radiation power at
relatively low temperatures (Te ≈ 10 − 100 eV). At high temperatures (Te > 1 keV), they
are fully ionized and will not cause any line radiation. High Z contaminations have their
maximum radiated power at higher temperatures (Te ≈ 100 eV − 1 keV) and the radiated
power at reactor temperatures (Te ≈ 10 keV) is high, due to the higher bremsstrahlung
(PB ∼ Z 2 ).
For a given impurity, the densities in the different charge states, and thus the emitted
power from the corresponding lines, will vary with the electron temperature. At higher
electron temperatures, the higher charge states will become dominating. This can be
observed by a spectroscopic measurement of the emitted plasma radiation. Figure 6.4
shows an experimental measurement of the radiative power (in relative units) for different
charge states in oxygen (O1+ , O2+ , . . .) as a function of time. In this pulsed experiment
the electron temperature is increasing over time from Te ≈ 0 at time t = 0 to Te ≈ 250 eV
at t = 0.5 ms. Thus, the graph can also be seen as radiation power as a function of
electron temperature and it clearly shows that higher charge states have their maximum
radiative power at higher electron temperatures.

6.2 The plasma boundary layer

The plasma area closest to the surrounding vessel wall is called the boundary layer . The
flux of particles and heat from the plasma to the walls results in different processes in the
boundary layer known as plasma wall interaction. Different processes at the wall surface
can, for example, give a flux of hydrogen atoms and impurity ions back to the plasma.
Recombination at the wall gives recycling of neutral hydrogen which affects the plasma
density and cause atomic and molecular processes in the boundary layer, e.g. ionization,

136
 0.3
0.2

o 1+
0.1

0.0

0.4

o 2+
0.2

0.0

0.6

o 3+
0.4
0.2

0.0

0.4

o 4+
0.2

0.0

0.6

o 5+
0.4

0.2

0.0
6+

0.05
o

0.00
0.0 0.1 0.2 0.3 0.4 0.5 0.6
t(msec) time

Figure 6.4: Experimentally measured radiated power for different charge states in oxygen.
The electron temperature increases from Te ≈ 0 at time t = 0 to Te ≈ 250 eV at t = 0.5 ms.

excitation and charge exchange collisions. The flux of impurities is caused by different
processes releasing atoms and molecules from the wall, e.g. desorption, sputtering, erosion
and vaporization.
The flow of impurities from the wall increases radiation losses from the plasma. To
decrease the negative effects of plasma wall interaction the adsorbed impurity atoms are
removed from the wall. This is done using by conditioning the vacuum vessel through
“baking” and glow discharge cleaning. Other methods used to decrease the effects of
plasma wall interaction in tokamaks are the use of limiters and divertors.

6.2.1 Recycling of neutral gas

The pulse length in a tokamak plasma is usually at least one order of magnitude longer
than the confinement time. Hence, every ion will, on average, hit the wall and return to
the plasma several times during one discharge. This process results in a circulating flux
of particles and energy between the wall and the plasma. This is known as recycling of
neutral gas.
The relation between the recycling flux and the incident flux is described by the
recycling coefficient R The incident flux can also release gas adsorbed on the wall surface,
and R > 1 is possible. The reflection of ions against the wall surface is dependent on ion
type, ion energy and wall material.
The backscattering coefficient Rp (which describes the recycling flux from a “clean”
wall surface) as a function of a so called reduced ion energy ε follows a universal curve
(see figure 6.5).
m2 E
ε = 32.5 . (6.22)
m1 + m2 Z1 Z2 (Z1 + Z22/3 )1/2
2/3

Here, m1 , m2 and Z1 , Z2 are mass number and charge number for the incident ion (1) and
the wall material (2), and E is the ion energy in keV. A corresponding universal curve
also exists for the energy backscattering coefficient RE (figure 6.5).

137
 0
10
Energy

Backscattering coefficient
reflection
-1
10

Particle
-2 reflection
10

-3
10
-3 -2 -1 0 1 2
10 10 10 10 10 10
Reduced energy ε ∝ E

Figure 6.5: Particle backscattering coefficient Rp and energy backscattering coefficient

RE as a function of the reduced ion energy ε (eq. 6.20).

6.2.2 Atomic and molecular processes in the boundary layer

Because of the neutral gas flux from the wall there are neutral hydrogen atoms and
hydrogen molecules as well as not fully ionized impurities in a boundary layer close to
the wall. These atoms are excited and ionized by collisions with ions in the plasma.
Excitation cause emission of radiation and cooling of the edge plasma (which may be an
advantage since it decreases the power flux to the wall). The following atomic processes
are important for the hydrogen flux:

1. ionizationof hydrogen atom by electron collision H + e → H+ + 2e

2. ionization of hydrogen molecule by electron collision H2 + e → H+

2 + 2e

3. ionization of hydrogen atom by proton collision H + H+ → H+ + H+ + e

4. charge exchange collision between hydrogen atom and hydrogen ion H + H+ → H+ + H

5. charge exchange collision between hydrogen molecule and hydrogen ion H2 + H+ → H+ + H2

The rate coefficient (< σv >) for these processes as a function of the particle energy is
shown in figure 6.6.

6.2.3 Desorption of impurities from the wall surface

The wall surface of the vacuum vessel usually contains absorbed atoms or molecules from
the rest gas in the vessel. Desorption means that these particles are released from the
wall and enter the plasma. There are two types of adsorbed gas;

1. physically adsorbed gas

2. chemically adsorbed gas

The physically adsorbed particles are loosely bound to the wall surface, i.e. the so called
binding energy is around 0.3 eV. The chemically adsorbed particles are more tightly

138
 -13 H + Charge exchange on H
10

-14

Rate coeficient (m3 s -1 )

10
H+e H 2 +e
ionization ionization

-15
10
Charge exchange of
+
-16 H 2 on H
10
Proton ionization of H
-17
10
10 0 10 1 10
2
10 3 104 10
5
10 6
Particle energy (eV)

Figure 6.6: The reactivity for different atomic and molecular processes in hydrogen gas.

bound with a binding energy of around 3 eV. The flux of plasma ions to the wall surface
during a plasma discharge causes desorption of absorbed particles and thus a flux of low
Z impurities into the plasma.
To decrease the influx of impurities from the absorbed gas, the gas is removed as much
as possible before the fusion plasma discharge is started in the vacuum vessel. There are
two main methods for this:

1. Heating of the vacuum vessel, so called “baking” of the vessel. This method is used
to remove physically bound adsorbed gas.

2. “Cleaning” of the wall surface using a continuous, weakly ionized plasma discharge
(glow discharge) in the vacuum vessel (known as Glow Discharge Cleaning). Glow
discharges without magnetic field in hydrogen or helium are used. Plasma ions hit
the wall surface and chemically adsorbed gas is released.

6.2.4 Sputtering of wall material

Plasma ions hitting the wall with sufficiently high energy can give rise to a “cascade” of
collisions with the bound atoms of the wall material. As a result of this collision cascade
an atom from the wall material can be released from the wall surface. This process is
called sputtering. Sputtering cause a flux of high Z contaminations into the plasma.
The number of ions released from the wall surface by each incident ion is called the
sputtering yield . The sputtering yield depends on ion type, ion energy and wall material.
Figure 6.7 shows the sputtering yield as a function of ion energy for some common combi-
nations of ions and wall materials. The sputtering yield quickly increases at a certain ion
energy, the so called threshold value for sputtering yield. This value is different depending
on ion type and wall material.
The flux of high Z impurities can to some extent be reduced by a suitable choice of
wall material. It can, however, never completely be avoided as long as there is a flux of
high energy ions to the wall.

139
 Carbon
Stainless steel He
-1 Molybdenum
10
He

Sputtering yield (atoms/ion)

D He
D
-2 H
10 H D

H
-3
10

-4
10
-2 -1
10 10 1 10
Ion energy (keV)

Figure 6.7: Sputtering yield for different combinations of particles and materials.

λD
V

-V s

Figure 6.8: Schematic picture of the space charge sheath at the wall.

6.2.5 Arc discharges to the wall surface

An arc discharge is created between two electrodes in vacuum if the potential between
them is sufficiently high. The electrode material is vaporized and ionized at the negative
electrode (the cathode) causing a current. Arc discharges usually start at a sharp edge
or peak in the material since the electrical field is stronger there. Electrons are emitted
from the surface by field emission . Current flowing from a localized area cause resistive
heating of the material and cathode spots are formed where electrons are emitted by
thermal emission .
A unipolar arc is an arc between the plasma and the wall. When magnetic field lines
pass the wall, charged particles may flow to the wall surface. Since electrons have higher
velocity than ions, they leave the plasma along field lines more easily. These electrons
charge up the wall negatively relative to the plasma. As a result, they create a layer
with positive space charge close to the wall, with a thickness of approximately one Debye
length (figure 6.8). The potential difference over the sheath repels electrons so that ion
and electron fluxes to the wall become equal and the charge neutrality of the plasma is

140
 3
10
Cr

Vapour pressure (Torr)

1 W
Mo
C
-3 Fe
10

-6
Ti
10

-9
Ni
10

1000 2000 4000 6000

Temperature (K)

Figure 6.9: Vapour pressure as a function of temperature for different wall materials.

maintained. The potential difference is given by the expression

   
2πme
Vs = 0.5Te ln −1 . (6.23)
mi
For a hydrogen plasma we have Vs ≈ 3Te . A unipolar arc can occur from the plasma over
the space charge sheath to the wall. Here, the plasma acts as the positive electrode (the
anode) and the wall is the negative electrode (the cathode) in the arc. Evaporation due
to heating at cathode spots cause erosion of the wall material and impurity influx to the
plasma. Arcing can be decreased if sharp corners and edges are avoided in the vacuum
vessel. The problem with unipolar arcing is usually decreased over time since peaks and
edges are gradually melted away.

6.2.6 Evaporation of the wall material

Plasma ions hitting the wall surface cause a heat flux to the wall. When the geometry
of the magnetic field causes the heat flux to be localized to a small area it can strongly
heat the wall surface causing evaporation of the wall material. The evaporation causes a
flux of impurities into the plasma. The amount of vaporized material as a function of the
wall temperature depends on the wall material (figure 6.9). The melting point for each
material is indicated by a point in figure 6.9. Metals that are components of stainless
steel such as iron (Fe) have a melting point of around 2000 K. Special metals like tungsten
(W) and molybdenum (Mo) have higher melting points around 3000–4000 K. Walls made
of graphite (C) can also withstand high temperatures and do not start to vaporize until
the temperature is around 2000 K.

6.2.7 Limiters
A plasma limiter is a material surface extending some distance from the surrounding wall
(figure 6.10). Limiters can have different geometries, either local or extended in poloidal
or toroidal direction.
The purpose of the limiters are mainly to:

141
 Wall
Lc
Boundary layer Γ =nV

Γ =-D dn
Limiter dr
Confined plasma Separatrix
a
B

Figure 6.10: A simple model showing the principle of a limiter.

1. decrease the particle flux to the vessel wall.

2. localize plasma wall interaction to a controlled surface.
The advantage of decreasing the particle flux to the vessel wall is that the vessel, usually
made from stainless steel, does not have optimum properties to withstand the heat flux
from the plasma. (See the section on vaporization) When using a limiter, there are more
options for the material choice. The particle flux to the vessel wall with a limiter can be
estimated using a simple model, see figure 6.10. A circular cylinder plasma that is part
of a torus with minor radius r = a, and major radius R  a is assumed to be surrounded
by poloidal limiters placed at a distance Lc from each other. The main toroidal magnetic
field is directed in axial direction in this model. Inside the limiter (r < a) the field lines
are closed. Outside the limiter (r > a) a boundary layer is formed. Here, the magnetic
field lines pass the side surfaces of the limiter and a parallel particle flux Γk along field
lines to the side surfaces is formed. Because of the parallel flux, the plasma density in
the boundary layer will decrease with increasing radius. Thus, the perpendicular particle
flux Γ⊥ will be small close to the vessel wall.
The plasma density in the boundary layer can be estimated with a simple model (see
figure 6.10). Assume that the boundary layer is thin compared to the plasma radius.
Consider a plane one dimensional model of the boundary layer where the variation of
the density along the field lines is neglected. The particle flux along the field lines Γk is
represented by a loss term described by the confinement time τk in the continuity equation
(eq. 2.45) :
∂n n
+∇·Γ=− . (6.24)
∂t τk
Introduce a cartesian coordinate system with the x coordinate in radial direction and
x = r − a. In equilibrium, eq. 6.22 is reduced to
dΓ⊥ n
=− . (6.25)
dx τk
Fick’s law gives the following expression for Γ⊥ ;
dn
Γ⊥ = −D⊥ . (6.26)
dx
The diffusion coefficient D⊥ is assumed constant in the boundary layer. Inserting eq. 6.23
in eq. 6.24 gives the diffusion equation
d2 n 1
2
= n (6.27)
dx D⊥ τk

142
which have the solution  x
n(x) = n0 exp − (6.28)
λ
where n0 is the density at r = a and the characteristic length for the density variation in
the boundary layer is p
λ = D⊥ τk . (6.29)
The confinement time τk is calculated by dividing the total number of particles in the
boundary layer N = ALc n by the total flow along field lines to the side surfaces of the
limiter 2A
N Lc n
τk = = . (6.30)
2AΓk 2Γk
The flux to the wall surface is controlled by a negative space charge layer at the wall
(figure 6.8) limiting the electron flux so that ion and electron fluxes are equal

Γk = 0.5cs n (6.31)

where r
2eTe
cs = . (6.32)
mi
Inserting eq. 6.29 in eq. 6.28 finally gives
Lc
τk = . (6.33)
cs

6.2.8 Magnetic divertors

The problems with plasma wall interaction may, at least theoretically, be avoided by
using a magnetic divertor . The principle for a magnetic divertor is that the plasma flux
to the wall is diverted to a separate target chamber where the plasma is neutralized. The
neutral gas is pumped away from the target chamber so that impurities do not reenter the
plasma. In tokamaks the divertors are usually toroidally axisymmetric (poloidal field).
One or more conductors parallel to the toroidal plasma current create a magnetic field
which, topologically, has two different areas (see figure 6.11).
In the inner area, which represents most of the tokamak main chamber, the magnetic
field lines are “closed”, just like in a tokamak without divertor. In the outer area, close
to the main chamber’s wall, the magnetic field lines end in a target chamber. These field
lines are “open”; they pass through a material surface where the plasma is neutralized (in
the target chamber). The two areas are limited by a magnetic separatrix (which restricts
the hot plasma in a similar way as a material limiter). In the outer area there will be a
plasma flux along field lines into the target chamber. This reduces the plasma flux to the
walls in the main chamber. The divertor corresponds, in principle, to a ‘magnetic limiter”
with the advantage that it does not introduce as much impurities in the plasma. There
are two main reasons for the decreased impurities:

1. Impurities are pumped from the target chamber and is not brought into the plasma.

2. Lower power flux to the wall with a so called gas target decreases the production of
impurities.

143
 z
φ
Main chamber

Separatrix

Closed field lines

Open field lines

ID Divertor

Figure 6.11: Schematic picture of magnetic divertor in a tokamak.

By increasing the neutral gas density in the target chamber the power released as elec-
tromagnetic radiation from the plasma in the divertor is increased. The power flux to the
wall in the divertor is thus decreased correspondingly and the production of impurities
at the wall surface is reduced. The density gradient along field lines in the outer mag-
netic field area required for this scenario (which is called gas target) can in principle be
maintained by the friction force from the plasma flux into the divertor.

6.3 Ohmic heating

In the tokamak, a toroidal electrical current is needed in the plasma to create the magnetic
field confining the plasma. This current also provides a heating power because of the
plasma resistivity. The power input per unit volume from ohmic heating is

PΩ = ηj 2 (6.34)
where η is the plasma resistivity and j is the plasma electrical current density. The
plasma resistivity depends on Coulomb collisions and decreases with increased plasma
temperature (η ∼ T −3/2 ). This means that the ohmic heating is less efficient at high
plasma temperatures and other heating methods are needed as well.
In a tokamak, the toroidal plasma current is limited by the condition for MHD stability.
MHD stability in the tokamak requires that the safety factor q > 1 in the centre (see
chapter 4). The current density in the centre is in the toroidal direction, i.e. j = jϕ .
Ampére’s law provides a relation between the poloidal magnetic field Bθ and the toroidal
current density in the centre (where jϕ is assumed constant):

1
Bθ = µ0 rjϕ . (6.35)
2
The stability condition q > 1 then gives the following condition for the current density:
2Bϕ
jϕ < (6.36)
µ0 R

144
where R is the major radius and Bϕ is the toroidal magnetic field. Inserting eq. 2.76
gives the following condition for PΩ :
 2
−5 −3/2 2Bϕ
PΩ < 5.22times10 Zeff ln ΛT (6.37)
µ0 R

which, if we assume Zeff = 1 and Λ = 15, gives

 2
9 −3/2 Bϕ
PΩ < 2.0 × 10 T [W/m3 ]. (6.38)
R

The energy confinement time for ohmically heated tokamaks has been experimentally
measured and can, in a first approximation, be described by the following simple empirical
scaling law (see Table 5.1):
τE ≈ 5 × 10−21 na2 [s] (6.39)
where n is the plasma density and a is the minor radius. Since the plasma energy density
is W = 3neT the power loss per unit volume will be
3neT T
Ploss = = 9.6 × 101 2 [W/m3 ]. (6.40)
τE a
The energy balance equation (PΩ = Ploss ), with eq 6.36 and eq. 6.38 inserted, gives the
following condition for the plasma temperature
a 4/5
T < 8.5 × 102 Bϕ [eV]. (6.41)
R
For a tokamak assumed to have aspect ratio R/a = 3 and the magnetic field Bϕ = 10 T
we have that T < 2.2 keV, which indicates that the plasma temperature possible to reach
by ohmic heating in this case is limited to 2–3 keV.

6.4 Neutral beam heating

Neutral beam injection (NBI)) is a method for heating mainly the ions in large tokamaks.
For example, ions in the central plasma has been heated to temperatures of 20–30 keV,
which is more than enough for a fusion reactor. A method sometimes used is injection
of two or more neutral beams tangential to the minor axis of the torus from different
directions. If neutral beams “collide” in this manner, a low density plasma with hot ions
is obtained.

6.4.1 Injection of energetic neutral beams

The heating method is based on injection of beams of energetic neutral atoms into the
plasma. The neutral particles can, because of non-elastic collisions with plasma particles
create energetic ions. Since these are charged, they are trapped in the magnetic field.
The heating is then caused by Coulomb collisions which decelerates the energetic ions.
The following processes cause trapping of the injected neutral atoms:

1. Ionization of a hydrogen atom by electron collision (σe )

Hbeam + e → H+ beam + 2e

145
 Energy of hydrogen beam atoms
-19 10 keV 100 keV
10
Charge
Exchange (σch )
Cross-
section
(m2 ) Ionization
byions (σi )
-20
10

T
e =1
keV
Electron
ionization T =
e 10
keV
-21
10
10 keV 100 keV 1 MeV
Energy of deuterium beam atoms

Figure 6.12: Cross section for some atomic processes in hydrogen gas.

2. Ionization of a hydrogen atom by proton collision (σi )

Hbeam + H+ + +
plas → Hbeam + Hplas + e

3. Charge exchange collision between hydrogen atom and hydrogen ion (σch )
Hbeam + H+ → H+beam + Hplas

The cross sections for these processes as a function of the energy of the injected neutral
atoms are shown in figure 6.12.
An expression for the absorption of the neutral beam due to the mentioned processes
can be obtained in the following way: The intensity of the beam I(x) as a function of the
distance x the beam has passed through the plasma is written as

I(x) = vb Nb (x) (6.42)

where vb is the velocity of the beam particles and Nb (x) is the number of particles per
unit length in the beam. The intensity is the number of beam particles incident on a
surface in the plasma perpendicular to the beam per unit time. Consider a disc of plasma
with area A equal to the cross section of the beam and with thickness dx. The number
of plasma particles in the disc is
Np = nAdx (6.43)
where n is the plasma density. The probability p for collision between an incident beam
particle and a plasma particle in the disc is described by the collision cross section σ:
σ
p= Np = σndx. (6.44)
A
The number of particles in the beam dI absorbed (per unit time) in the disc due to
collisions is then
dI = −pI = −σnIdx (6.45)

146
 1.0 0

Electron heating fraction

Ion heating fraction

Pe
Pi + Pe 0.5 0.5

0 1.0
1 10 100
E b/ Te

Figure 6.13: The relation between heating of electrons and heating of ions.

thus
dI
= −σnI. (6.46)
dx
This differential equation has the solution
 x
I(x) = I0 exp − (6.47)
λ
where
1
λ= (6.48)
σn
is the characteristic penetration depth for the neutral beam in the plasma and I0 is the
incident intensity at the plasma surface. The collision cross section is the sum of the cross
sections for the different processes described above (as shown in figure 6.12), i.e.
1
σ = σch + σi + < ve σe > . (6.49)
vb
In the case of ionization by electron collisions, where the electron velocities are higher than
the neutral atom velocity, an integration (< . . . >) is made over the velocity distribution
of the electrons.
The energetic ions created by absorption of the neutral beam transfer their energy
to the plasma particles when they are decelerated by Coulomb collisions. The fast ions
in the beam, with mass mb and velocity vb , experience a friction force Fbe from electron
collisions
mb vb
Fbe = (6.50)
τie
and a corresponding friction force from ion collisions
mb vb
Fbi = (6.51)
τii
where τie and τii are collision times for electron and ion collisions respectively. The heating
power transferred to each particle type can be written as

Pe = vb Fbe (6.52)

and
mb
Pi = vb Fbi (6.53)
mb + mi

147
 Accelerator

Ion Deflector
source Neutralizer magnet
Neutrals
D+
D+ D0
D0
Plasma
Gas D+
Ions
1 2 3 4

Figure 6.14: Neutral beam injector.

where the energy from ion collisions is divided between the beam ion and the plasma
ion. (In electron-ion collisions, the electron is assumed to absorb all energy, because of its
smaller mass.) The collision time increases with increased relative velocity v between the
colliding particles (τ ∼ v 3 ) (eq. 2.31). This means that, according to eq. 6.48–6.51, at low
neutral beam energies the ions will absorb most of the heating power, and at high neutral
beam energies the electrons will absorb the most. The ratio of the electron heating power
Pe and the ion heating power Pi to the total heating power Pe + Pi as a function of neutral
beam energy Eb is shown in figure 6.13.

6.4.2 A neutral beam injection system

A neutral beam injector has the following components (figure 6.14):

1. Ion source; production of charged particles

2. Accelerator; the particles are accelerated to the injection energy ∼ 50–100 keV
3. Neutralization chamber; the beam pass through a gas and is neutralized
4. Deflection magnet

Charged particles are produced in an ion source. After the acceleration, the charged
particles pass through the neutralization chamber. Since the neutralization is not perfect,
a deflection magnet is used to remove the remaining energetic ions from the beam.
The neutralization chamber contains neutral gas and the charged particles are neu-
tralized by charge exchange collisions with neutral atoms. The energetic ions created
can be ionized again by collisions with neutral atoms. The relation between neutral and
ion densities, the equilibrium neutral fraction, will be determined by the ratio of the
cross-sections for the two inverse processes:
N0 σch
= (6.54)
N+ σion
The neutral fraction as a function of the beam energy is shown in figure 6.15. The figure
also shows how the penetration depth (eq. 6.46) depends on the beam energy.

6.5 Radio-frequency heating

Heating of the plasma using radio-frequency waves is based on the fact that the plasma
has certain “resonances” where electromagnetic waves can be absorbed by the plasma.

148
 1.0

0.8 Neutral
fraction

0.6

0.4

Penetration
0.2 distance(m)

0
10 20 40 60 80 100 200 300
Beam energy (keV)

Figure 6.15: Neutral fraction and penetration depth as a function of beam energy.
z
Bo
k
θ
y
x

Figure 6.16: Diagram showing different directions assumed in the wave model

At these resonances the energy in the electromagnetic wave is transformed to kinetic

energy of ions or electrons through wave-particle interaction . The heating occurs locally
in the plasma in the region where the resonance condition (which can depend on the
magnetic field strength or plasma density) is fulfilled. By changing the frequency of the
wave, absorption can be achieved in different regions. To use this method, antennas are
installed at the plasma boundary. These are transmitting waves which propagate through
the plasma to the resonance area. Wave propagation in plasma is a rather complicated
phenomenon. The usual transverse electromagnetic wave, present in vacuum, will in the
plasma transform into one of several different wave types. The corresponding process
can also occur when the wave, during its propagation through the plasma, experience a
change in plasma parameters - a so called mode conversion can take place, and the wave
is transformed to another type.

6.5.1 Plasma dispersion relation

There is a well developed theory describing wave propagation, mode conversion, reso-
nances and absorption in the plasma. The following section provides a short discussion
on the simplest model for plasma waves describing the basic phenomena of radio-frequency
heating. A wave in a homogeneous plasma is described by solving the plasma dispersion
relation. The theory is briefly outlined below.
Consider a plasma with a homogeneous magnetic field B0 = B0 .ez A plane wave anzats
is used. The perturbation quantities are assumed to depend on space and time as

∼ exp i(k · r − ωt) (6.55)

which describes plane waves with angular frequency ω propagating in a direction given
by the wave vector k (figure 6.16). A wave is described by the relation k(ω), which is
called the dispersion relation. In this section we use E, B, . . . to describe the complex

149
amplitude of a perturbation quantity. Introducing the plane wave anzats in Maxwell’s
equations gives
ik × E = iωB
ik × B = µ0 j − µ0 ε0 iω.E (6.56)
The relation between the vectors j and E in the plasma is described by the conductivity
↔
tensor σ:
↔
j = σ · E. (6.57)
↔
The plasma effect on the wave propagation is contained in the conductivity tensor σ. The
conductivity tensor is calculated from the linearised fluid equations (see chapter 4). In
this model the plasma pressure is assumed small and collision terms are neglected. The
↔
force balance equation and Ohm’s law gives an expression for σ. The refractive index n
is defined as
c
n = k. (6.58)
ω
Eq. 6.54 and 6.55 gives the wave equation
↔
n2 E − n(n · E) − ε · E = 0 (6.59)
↔
where ε is the dielectric tensor ,
i
εkl = σkl + δkl . (6.60)
ωε0
The dielectric tensor has the following form:
 
ε⊥ εxy 0
εkl = −εxy ε⊥ 0  (6.61)
0 0 εk
where
2
X ωpj
ε⊥ = 1 − 2
(6.62)
j
ω 2 − ωcj
2
X ωpj
εk = 1 − (6.63)
j
ω2
2
X ωpj ω 
cj
εxy = i 2
. (6.64)
j
ω 2 − ωcj ω

Summation over j is over particle types (ions, electrons). Here, ωcj is ion and electron
cyclotron frequencies. respectively (eq. 2.9) and ωpj is ion and electron plasma frequencies,
respectively (eq. 4.18). Assume that the wave vector is in the yz plane (see figure 6.16)
where θ is the angle between k and B. Introduce
nk = n cos θ
n⊥ = n sin θ. (6.65)
The wave equation (eq. 6.57) can then be written in matrix form
ε⊥ − n2k − n2⊥
  
εxy 0 Ex
2
 −εxy ε⊥ − n k n k n ⊥   Ey  = 0. (6.66)
0 nk n⊥ εk − n2⊥ Ez

150
 E
z ωt y

Figure 6.17: Schematic picture of a right circular polarized (RCP) wave.

Nontrivial solutions to the homogeneous system of equations (eq. 6.64) are obtained when
the determinant of the matrix is equal to zero. This condition finally gives the plasma
dispersion relation, expressed using n2k and n2⊥ ;
ε⊥ n4⊥ − (ε⊥ − n2k )(ε⊥ + εk ) + ε2xy n2⊥ + εk (ε⊥ − n2k )2 + ε2xy = 0.



(6.67)
The waves propagating in the plasma are described by n2k and n2⊥ that are solutions to
eq. 6.65.
Consider the special case k k B0 , i.e. n⊥ = 0. The following equation is obtained for
2
nk ;
(ε⊥ − n2k )2 + ε2xy = 0. (6.68)
Introduce
R = ε⊥ + iεxy
L = ε⊥ − iεxy (6.69)
which gives two solutions
“R − wave00

R
n2k = (6.70)
L “L − wave00 .
Consider the first component of the wave equation (eq. 6.57), which gives a relation
between the perpendicular components of the wave field E:
(ε⊥ − n2k )Ex + εxy Ey = 0 (6.71)
from which is obtained 
 −i “R − wave00
Ex
= (6.72)
Ey 
i “L − wave00 .
The two perpendicular components of the wave are π/2 out of phase, which means that
the wave is circular polarized, and the E field vector will rotate with angular frequency ω.
If the wave is viewed along z direction, in direction of decreasing z (figure 6.17), the “R”
wave rotation is counter clock-wise, i.e. the “R” wave is right circular polarized (RCP ).
For the “L” wave the rotation is clock-wise, i.e. the “L” wave is left circular polarized
(LCP ).
In the other special case k⊥B0 , i.e. nk = 0, the following equation is obtained for n2⊥ ;
ε⊥ n4⊥ − (ε⊥ εk + RL)n2⊥ + εk RL = 0 (6.73)
which has two solutions
ε “O − wave00

 k

n2⊥ = (6.74)
 RL

“X − wave . 00
ε⊥

151
 Δ
n n 0

Figure 6.18: Reflection of a wave in the plasma.

Δ
n n ∞

Figure 6.19: Absorption of a wave in the in the plasma

6.5.2 Reflection and resonance

In an inhomogeneous plasma the wave propagation is more complicated. In a first ap-
proximation it is assumed that the refractive index n varies slowly in space and that
the wave equation (eq. 6.57) still describes the propagating waves locally. In analogy
to geometrical optics, waves will be refracted when passing through a plasma where the
refractive index has a gradient. Consider a wave incident almost perpendicularly against
surfaces of constant refractive indices. According to the law of refraction (Snell’s law), the
wave will be refracted away from the normal direction of the surface when the refractive
index decreases in the direction of the wave propagation. In the opposite case, when the
refractive index increases in this direction, the wave will be refracted towards the normal.
In the plasma, the refractive index for a given wave can in principle approach zero on
some surface. When the wave is incident on such a surface, it will be reflected back in
the direction of incidence (figure 6.18), and the wave cannot propagate in the area on the
“ other side” of the reflection surface. The phenomenon is called cut-off . In the opposite
limit, when the refractive index approaches infinity at some surface, the wave will be
absorbed. In this case, the surface is called the wave resonance surface (figure 6.19).
The condition for reflection is reached by inserting n2k = 0, n2⊥ = 0 in the dispersion
relation (eq. 6.65):
εk (ε2⊥ + ε2xy ) = εk RL = 0. (6.75)
The refractive index can approach infinity at some surface which means the wave has a
resonance at the surface. In this case the wave will be refracted towards the surface.
When n2 → ∞ the phase velocity of the wave will approach zero vph = c/n → 0 and
the wave is absorbed at the surface. The condition for resonance is obtained by assuming
n2k → ∞, n2⊥ → ∞ in the dispersion relation (eq. 6.65). In this limit, terms of the highest
order (∼ n4 ) will dominate, and

ε⊥ n4⊥ + n2k n2⊥ (ε⊥ + εk ) + εk n4k ≈ 0 (6.76)

152
 V

E
ω
k

Figure 6.20: Schematic illustration of wave-particle interaction in Landau damping.

which, after inserting eq. 6.63 gives the condition for resonance
εk
= − tan2 θ. (6.77)
ε⊥
In the case k k B0 , i.e. θ = 0 there will be resonances according to eq. 6.75 when ε⊥ → ∞.
Eq. 6.60 shows that this occurs when
 2
 ωci “ion cyclotron resonance”
ω2 = (6.78)
 2
ωce “electron cyclotron resonance”

For wave propagation perpendicular to the magnetic field, when k⊥B0 , i.e. θ = π/2,
resonances is given by eq. 6.75 with the condition ε⊥ = 0. Eq. 6.60 gives, with reasonable
2 2 2 2 2 2
approximations ωpi  ωpe , ωci  ωce , ωci  ωpi the following equation for ω 2 ;

ω 4 − ω 2 (ωce
2 2
+ ωpe 2 2
) + ωpi ωce ≈ 0 (6.79)

from which three resonance frequencies are given as

 2 2
2
ωpi ωce
 ωLH ≈ 2 “lower hybrid resonance”


2
ωce + ωpe
ω2 = (6.80)


 2 2 2
ωU H ≈ ωce + ωpe “upper hybrid resonance”

6.5.3 Absorption of waves in the plasma

The fluid model for plasma waves described in the previous section predicts that the energy
of the plasma wave is absorbed in the plasma in regions where the resonance conditions
(eq. 6.76 and 6.78) are fulfilled. The actual absorption process can not be described in
the fluid model. The absorption is caused by wave-particle interaction, which must be
treated using kinetic models.
For a wave, where a component of the wave electrical field and the wave vector are
parallel (E k k), which propagates along the magnetic field ( k k B0 ), absorption of
the wave energy can take place through so called Landau damping. In this process, a
wave-particle resonance arises when a particle moves with velocity v approximately equal
to the wave phase velocity ω/k (figure 6.20). The component of the wave’s electric field
along the wave vector Ek can then accelerate the particle and wave energy is transferred
from the wave to the particle. A condition for efficient transfer of energy is that the
resonance condition v ≈ ω/k is fulfilled. In this case, a particle experiences an electric
field in the same direction for a longer period. If v 6≈ ω/k the particle will be accelerated
successively in opposite directions and the average energy transferred is small.
Another kind of wave particle resonance results if the wave has a component perpen-
dicular to the wave vector (E⊥k). Consider a left circular polarized wave (“L” wave)

153
 ω ci
+

z
k BO y
ωt
E
ω

Figure 6.21: Schematic picture of wave-particle interaction in ion cyclotron resonance.

propagating along the magnetic field (k k B0 ) (figure 6.21). Seen from the positive z axis,
the wave E field vector is rotating clock-wise with angular frequency ω. Ions in the plasma
also gyrate clock-wise around magnetic field lines with the ion cyclotron frequency ωci .
When the wave frequency fulfils the resonance condition ω ≈ ωci , the ion will experience
an electric field mainly perpendicular to the gyration movement during a longer time.
This will accelerate the ion. The acceleration cause energy to be transferred from the
wave to the ion, thus damping the wave. This process is called cyclotron damping. In
this case as well, wave-particle resonance is necessary, since if ω 6≈ ωci the E field will no
longer be in phase with the particle. Thus, the acceleration is successively in different
directions, preventing an effective energy transfer.
Through wave-particle interaction described above (Landau damping and cyclotron
damping) energy will be transferred from the wave to the particles. The wave is damped,
or absorbed. At the same time, the energy transfer increase the particle’s kinetic energy,
thus heating the plasma. This is the effect used for heating of the plasma with radio
frequency waves (RF heating).

6.5.4 Different RF heating methods

Several methods of heating the plasma by absorption of electromagnetic waves in the radio
frequency range are being used. The difference between these is which resonance frequency
is being used. The basic idea is, however, the same for all methods. An antenna close
to the plasma emits an electromagnetic wave which is converted to a plasma wave at the
plasma boundary. The plasma wave propagates through the plasma and can experience
mode conversion – a transition to another type of plasma wave. The wave transports the
wave energy to the resonance region, usually in the center of the plasma, where the energy
is absorbed by ions and electrons. This is where the heating takes place. The following
heating methods are common (The indicated frequencies correspond to those suitable for
a tokamak with a magnetic field B ≈ 5 T);

1. Ion Cyclotron Resonance Heating (ICRH). The typical frequency is f ≈ 40 MHz. A

so called compressible Alfvén wave transports the energy from the plasma boundary
to the center. In the central plasma the wave is mode converted into a ion cyclotron
wave. The absorption is mainly through ion cyclotron damping, heating the ions.

2. Lower Hybrid Resonance Heating (LHRH). The typical frequency is f ≈ 1–5 GHz.
The resonance condition depends, through the plasma frequency, on the plasma
density. This means that the absorption region in the plasma varies with the plasma

154
 density. LHRH heats the ions through damping at the lower hybrid resonance, but
can also heat the electrons by electron-Landau damping.

3. Electron Cyclotron Resonance Heating (ECRH). The typical frequency is f ≈ 100–

200 GHz. Absorption takes place through electron cyclotron damping, which heats
the electrons.

155
Chapter 7

PLASMA DIAGNOSTICS

Plasma diagnostics is the common name for different types of measurement equipment
used to study, i.e. diagnose, fusion experiments. The purpose is to obtain information
about plasma quantities such as temperature, density and so on. A space resolved mea-
surement is usually necessary, since it allows reconstruction of radial profiles of these
quantities. Sometimes dynamic events are studied, and this requires measurements with
high time resolution. These instruments can both be simple, e.g. a magnetic probe mea-
suring the magnetic field outside the plasma, and complicated, e.g. equipment used to
measure the electron temperature by looking at the scattering of electromagnetic waves
in the plasma. A large tokamak experiment usually has several diagnostic systems si-
multaneously measuring important plasma parameters. Quite often, several methods are
used to measure a parameter, e.g. one diagnostic used to measure a parameter locally
in different positions in the plasma at a certain time combined with another measuring
the same parameter in a larger area over time. One such example is the measurement of
the electron temperature, which is done locally at a specific time by scattering of elec-
tromagnetic waves and also continuously by a measurement of bremsstrahlung radiation
emitted from a larger part of the plasma. An example of the use of information from
several plasma diagnostics is the estimation of the confinement time τE in an ohmically
heated tokamak. This requires a simultaneous measurement of electron temperature Te ,
ion temperature Ti , electron density ne , average ion charge (which gives the ion density),
toroidal electric current in the plasma I, toroidal loop voltage V (which gives the ohmic
power input PΩ = V · I).
Methods for plasma diagnostics can be sorted by the physical property of the plasma
on which the measurement is based:

1. Magnetic field. Measurements of the magnetic field inside or outside the plasma
using different types of probes or measurement coils.

2. Plasmas particle flux. Different types of probes in direct contact with the plasma
measuring the flux of plasma particles.

3. Plasma refractive index. Measurements of the plasma refractive index by transmis-

sion of electromagnetic waves through the plasma.

4. Electromagnetic radiation from free electrons . Plasma quantities are estimated by

measuring for example bremsstrahlung or electron cyclotron radiation emitted from
the plasma.

156
 5. Electromagnetic radiation from bound electrons. Spectroscopic measurement of line
radiation emitted from, for example, not fully ionized impurities in the plasma.

6. Scattering of electromagnetic waves. Transmission of electromagnetic waves through

the plasma which, by interactions with plasma particles, give rise to secondary
electromagnetic waves, scattered waves , which are detected.

7. Ion processes. Measurement of phenomena caused by ion processes in the plasma,

e.g. charge exchange collisions between ions and neutral atoms.

Some examples of diagnostic methods and plasma quantities measured are shown in Table
7.1. These methods are only a small selection of those used at large fusion experiments.
Some of the methods are described in more detail in this chapter.

Table 7.1: Examples of plasma diagnostics

method plasma quantity symbol

1. Magnetic field
Rogowski coil plasma current Ip

2. Plasma particle flux

Langmuir probe plasma density, temperature, potential ne , Te , φ

3. Plasma refractive index

interferometry electron density ne

4. Electromagnetic radiation from free electrons

electron cyclotron radiation measurements electron temperature Te
bremsstrahlung measurements electron temperature Te

5. Electromagnetic radiation from bound electrons

visible, UV, VUV spectroscopy effective ion charge, ion temperature Zeff , Ti

6. Scattering of electromagnetic waves

Thomson scattering electron density, electron temperature ne , Te

7. Ion processes
neutral particle analysis neutral density, ion temperature na , Ti
neutron spectroscopy ion temperature Ti

7.1 Rogowski coil

Magnetic diagnostics are used to measure global plasma parameters such as toroidal
electric current I and toroidal loop voltage V , but also for measurements of the local
magnetic field fluctuations B to study MHD instabilities. A common magnetic diagnostic
is the Rogowski coil used to measure the toroidal current (figure 7.1). The coil consists

157
 Figure 7.1: Schematic picture of Rogowski coil used to measure plasma current.

of a long solenoid coil encircling the plasma in the poloidal direction. The coil provides a
measurement of the line integral of the poloidal magnetic field component along poloidal
direction, and Ampére’s law gives the toroidal current passing inside the coil as
I
1
I= B · dl. (7.1)
µ0
Consider a coil with N loops and length L. The voltage from loop i is equal to the time
derivative of the magnetic flux through the loop (figure 7.2).

d  ˆ i = − δA d (Bi · dli ) .

Vi = −δA Bi · ds (7.2)
dt dl dt
The total voltage from the coil is
N N
! I 
X N δA d X N δA d N δA dI
V = Vi = − Bi · dli ≈− B · dl =− µ0 . (7.3)
i=1
L dt i=1
L dt L dt

The plasma current I is finally calculated by time-integrating the output voltage.

dA
ds i
B
L
dl = N

Figure 7.2: Calculation of signal from Rogowski coil.

7.2 Interferometric measurement of electron density

The wavelength of an electromagnetic wave transmitted through a plasma depends on the
plasma refractive index. This, in turn, varies with the electron density in the plasma. A
measurement beam passing through the plasma will thus have a phase shift compared to
a reference beam travelling the same distance in vacuum. This phase shift is proportional
to the electron density. Detection of the phase shift using an interferometer thus gives a
measurement of the plasma electron density.

158
7.2.1 Principle of interferometry measurement
An expression for the phase change of an electromagnetic wave passing through a plasma
can be described in the following way. Consider an electromagnetic wave in vacuum
ω 2π
= k0 = . (7.4)
c λ0
Here, k0 is the wave-number and λ0 is the wavelength in vacuum. When the electromag-
netic wave propagates in the plasma, it is in the form of an “O” wave described by the
plasma dispersion relation (see the section on wave propagation in the plasma) given by
eq. 6.56, 6.61 and 6.72:  2 2
k ωpe
= k = 1 − 2 (7.5)
k0 ω
where k = 2π/λ and λ is the wavelength in the plasma. We have assumed that ω  ωpi .
Since we also have the condition ω  ωpe we have
r
2 2
k ωpe 1 ωpe
= 1− 2 ≈1− . (7.6)
k0 ω 2 ω2
The phase for a measurement beam passing a distance L in the plasma can be written as
Z
φ= k dl. (7.7)
L

Plasma

dl
O

ref. O
Ο

Figure 7.3: The principle for density measurements with an interferometer.

The integral is evaluated along the path of the measurement beam through the plasma.
The phase of a reference beam propagating the corresponding distance in vacuum can be
written as (See figure 7.3) Z
φ0 = k0 dl. (7.8)
L
The phase difference between the measurement beam and the reference beam is
2
1 ωpe
Z Z
∆φ = φ0 − φ = (k0 − k) dl = k0 2 dl. (7.9)
L L 2 ω
Inserting the expression for the electron plasma frequency (eq. 4.18) finally gives
Z
∆φ = λ0 r0 ne dl (7.10)
L
where
e2
r0 = 2
= 2.818 × 10−15 m (7.11)
4π0 me c
is a constant (the classical electron radius). The interferometer thus measures the integral
of the electron density along the measurement beam path through the plasma, the line
density.

159
7.2.2 A laser interferometer system
An interferometer for continuous measurement of the electron density often uses modu-
lation of the incident electromagnetic wave. This is desirable because it helps to avoid
problems when interpreting the measured phase shift. Is some cases it is impossible to
decide if the plasma density is increasing or decreasing when the measured phase shift
changes. This problem can be avoided by modulation. An example of an interferometer

CO 2 laser Bragg Plasma

cell Mirror M1
ωl

1 2 ω
ω
I

ω BC l + ω Light detector
BC
OSC.
40 MHz to quad. unit ω 4
3 BC
Quadrature unit

Figure 7.4: Laser interferometer with modulation and quadrature detection.

with modulation technique for measuring a time dependent plasma density is shown in
figure 7.4. This interferometer uses a carbon dioxide laser as wave source. The interfer-
ometer consists of the following components:

1. Laser of CO2 type. Emits a coherent electromagnetic wave with wavelength λ0 =

10.6µm. The time variation of the wave can be written as I0 ∼ cos ωl t, where
ωl = 2πc/λ0 .

2. Acoustic-optical modulator, a so called Bragg cell . The Bragg cell divides the laser
beam into two parts by deflecting part of the incident beam. The deflected beam
part is frequency shifted ωBC /2π = 40 MHz, which is the carrier frequency of
the Bragg cell. The deflected beam is the interferometer reference beam IR ∼
cos(ωl t + ωBC t). The beam which is not deflected is the interferometer measurement
beam IM . It passes through the plasma, is reflected by mirror M 1 , passes through
the plasma again and is reflected against the exit window of the Bragg cell in the
same direction as the reference beam. The measurement beam that passes through
the plasma will have a time dependent phase shift ∆φ(t) relative the reference beam;
IM ∼ cos(ωl t + ∆φ(t)).

3. Detector. The interference between the measurement beam and the reference beam
takes place in the detector. When the beams are added the result is a signal which,
after low-pass filtering, has a frequency equal to that of the difference of the two
beam frequencies, ID ∼ cos(ωBC t − ∆φ(t)).

4. Quadrature unit. Quadrature detection means that the detector signal is multiplied
(mixed) with the signal from the RF-oscillator driving the Bragg cell. After low-pass
filtering the result is both a cosine signal IC ∼ cos(∆φ(t)) and, by phase shifting
the carrier frequency π/2, a sine signal IS ∼ sin(∆φ(t)). (figure 7.5)

160
 Mixer
Phase shifter LP-filter
cos ωBC t
osc. ωBC o
0 C cos φ
o
90 sin ω t
BC
Digitized
U= ωBC t - φ 0
o
cos U S sin φ
o
0

Figure 7.5: The principle of quadrature detection of the interferometer signal.

Finally, we get the phase shift as a function of time ∆φ(t), and thus the time variation
of the line density by digitizing the cosine and sine signals from the quadrature detector
and calculate ∆φ(t) = arctan(IS /IC ).

7.3 Electron temperature measurement by Thomson

scattering
Scattering of electromagnetic waves against free electrons in the plasma, so called Thom-
son scattering can be used to measure the local electron temperature and electron density
in the plasma.

7.3.1 Principle of Thomson scattering

An electromagnetic wave incident on an electron cause the electron to accelerate and thus
emit a secondary electromagnetic wave. This scattered wave will have a wavelength shift
which is proportional to the electron velocity. This is a form of Doppler effect where the
wavelength shift depends on the electron motion along both the direction of the incident
wave and the direction of observation. If the electrons have a thermal velocity distribution,
the scattered wave will be broadened around the wavelength of the incident wave. From
this broadening, the electron temperature can be estimated.

a R
n.

θ
ikt
s.r
ob

Figure 7.6: Direction definitions for radiation from an accelerated electron.

An electron being accelerated will emit radiation which can be described using the
expression for radiation from a dipole. The radiated power per unit solid angle in a given
direction of observation is
dP e2
= a2 sin2 θ (7.12)
dΩ 16π 2 0 c3
where a is the electron acceleration and θ is the angle between the observation direction
and the acceleration vector a (figure 7.6). Assume that a plane electromagnetic wave is
incident on the plasma

Ei = Ei0 exp i(ki · r − ωi t) . (7.13)

161
The electron is accelerated in the electric field of the incident wave
eEi
a= . (7.14)
me
The average power of the scattered wave is then

dP e2 Ei02
= sin2 θ. (7.15)
dΩ 16π 2 0 c3 m2e

The power of the incident wave is

2
Pi = c0 Ei0 . (7.16)
The differential cross section for Thomson scattering is obtained from eq. 7.11, 7.15 and
7.16 as
dσ
= r02 sin2 θ (7.17)
dΩ
where the expression for the classical electron radius r0 (eq. 7.11) has been used. The
electromagnetic wave generated by the electrons is the scattered wave. The measurement
of the electron temperature is based on the fact that the scattered wave is frequency
shifted relative the incident wave. The frequency shift is a Doppler effect caused by the
electrons velocity component along the wave vector of the incident wave and the electrons
velocity component along the observation direction, i.e. along the scattered wave vector.

ki

Ei

v
ks
vt
R2
Es R1
R0

Figure 7.7: Diagram illustrating the Doppler effect in Thomson scattering.

The Doppler effect can be easily understood from figure 7.7. The scattered wave is
assumed to be generated by an electron at a point given by the vector R1 . The electron is
assumed to have velocity v, and the electron position at time t is given by the expression

R1 = R0 + vt. (7.18)

The scattered wave electric field Es can be written

Es = Es0 exp i(ks · r − ωs t) . (7.19)

162
In the observation point, R2 , the scattered wave is


Es ∼ Ei (R1 , t) exp iks · (R2 − R1 )


∼ exp i(ki · R1 − ωi t + ks · (R2 − R1 ))


∼ exp i(ks · R2 − ωi t − (ks − ki ) · R1 ) . (7.20)
Inserting eq. 7.18 gives


Es ∼ exp i(ks · R2 − ωi t − (ks − ki ) · R0 − (ks − ki ) · vt) . (7.21)
A comparison between eq. 7.19 and eq. 7.21 gives the following relation between ωs and
ωi
ωs = ωi + (ks − ki ) · v. (7.22)
The frequency shift ω and the wave vector k is introduced
ω = ωs − ωi
k = ks − ki (7.23)
which means that eq. 7.22 can be rewritten in the following simple way
ω = k · v. (7.24)
This relation shows that the frequency shift is proportional to the electron velocity com-
ponent in the direction of the wave vector k. Assume that the electrons have a velocity
distribution f (v) along k. A thermal distribution (Maxwell distribution) is described as
me v 2

r
me
f (v)dv = exp − dv. (7.25)
2πeTe 2eTe
It is usually assumed that individual electrons move and radiate independent of each other
(i.e. the electrons are assumed uncorrelated). With this assumption, we get the spectral
function S(ω) for the scattered wave by inserting v = ω/k in eq. 7.25;
me ω 2

r
me 1
S(ω)dω = exp − dω. (7.26)
2πeTe k 2eTe k 2
The radiative power per unit solid angle, angular frequency and volume is finally (for
θ = π/2) :
dP
= ne r02 Pi S(ω). (7.27)
dΩdωdV
A measurement of the electron temperature using Thomson scattering is done by shining
a laser beam through the plasma and measure the spectral function S(λs ) of the scattered
light using spectrometry. Because of this, the spectral function (eq. 7.26) is written as a
function of λs , the wavelength of the scattered light. Assume that the angle between ki
and ks is π/2;
 ω 2 (ω − ω )2 2 2
s i 2 (λi − λs ) 2 λ
= =c ≈c (7.28)
k ks2 + ki2 λ2i + λ2s 2λ2i
where λ = λi − λs has been used. The spectral function S(λ) is given by eq. 7.26 as
 2 !
m e c2 λ
S(λ) ∼ exp − (7.29)
4eTe λi
from which the electron temperature is estimated
√ by measuring the spectral function
half-maximum width This is proportional to Te (figure 7.8).

163
 S(ω)

ω
ω α Te

ωo ωs

Figure 7.8: Schematic picture of the scattered wave spectral function.

Figure 7.9: Thomson diagnostic.

7.3.2 Thomson scattering diagnostic system

The electron temperature measurement using Thomson scattering is complicated by the
small cross section of the Thomson scattering. It is so small that the scattered radiation
easily can be concealed by background radiation from the plasma or radiation scattered
from a vacuum window or other surfaces when the laser beam passes in to the vacuum
vessel. An example of a typical experimental setup for Thomson scattering is shown in
figure 7.9. The measurement system consists of the following main parts:

1. Laser. Usually a pulsed laser with high power, ∼ 1 GW, is used. A common laser
type uses a ruby rod and can give one or several pulses. The beam is focused by
a lens to a selected point in the plasma. To decrease the scattering of light from
the vessel wall, the beam is lead through a number of apertures on its way to the
plasma. A beam dump stops light from being reflected back after passing through
the plasma.

2. Spectrometer. The light scattered an angle π/2 from the incident laser beam is
observed using a lens from a selected region, the scattering volume. The spectrum
of the scattered light is analysed with a spectrometer using filters for the wavelength
intervals measured in the spectrum. The light passing a filter is detected with a
photo multiplier. The measured scattered radiation power (∼ 1 mW) is very small
compared to the incident radiation power.

164
7.4 Measurement of electron cyclotron radiation
The electromagnetic radiation emitted by electrons when they gyrate in the magnetic
field, the electron cyclotron radiation, can be used to measure the plasma electron tem-
perature. In general, if the electrons emit radiation of a given frequency ω, they will also
absorb radiation of the same frequency. The electrons can be assumed to be in thermal
equilibrium, i.e. they have a thermal velocity distribution given by an electron tempera-
ture Te (Maxwell distribution). In this case, the emittance j(ω) has a simple relation to
the absorption coefficient α(ω) (Kirchoff’s law);

j(ω) = α(ω)IB (ω) (7.30)

where IB (ω) is the radiation intensity for a black body . The absorption coefficient is in
the range 0 < α(ω) < 1. A black body is defined as a medium absorbing all radiation
which is incident on it. Its radiation intensity is given by Planck’s radiation law
−1
h̄ω 3
  
h̄ω
IB (ω) = 3 2 exp −1 . (7.31)
8π c eTe

Here, h̄ = h/2π, and h is Planck’s constant. The intensity is defined as radiation power
per unit area, solid angle and angular frequency. The electron cyclotron radiation can in
this case be considered low frequency, h̄ω  eTe , and an approximative expression for IB
can be used
ω 2 eTe
IB (ω) ≈ . (7.32)
8π 3 c2
The radiation intensity observed in a given direction outside the plasma is caused by
emission and absorption of the beam along its path through the plasma.
The beam path can be considered a straight line given by the coordinate s where
s1 < s < s2 . The plasma far boundary is given by s = s1 (where the beam “starts”)
and the plasma near boundary is given as s = s2 (where it “leaves the plasma”). The
intensity of the beam along the path is given by the beam transport equation which gives
the change dI of the intensity when the beam travels a distance ds;

dI = j ds − αI ds. (7.33)

The equation can be solved by a change of variable from s to τ , where τ is defined as

dτ = −α ds. (7.34)

Integrating eq. 7.34 backwards along the beam path

Z τ Z s
0
dτ = − α ds0 (7.35)
τ2 s2

gives (with τ2 = 0); Z s

τ (s) = − α ds0 . (7.36)
s2

The quantity τ (s) is the optical depth for the plasma layer s < s0 < s2 . Inserting eq. 7.30
and 7.34 in eq. 7.33 gives
dI
= −IB + I (7.37)
dτ

165
 a) 3 ω ce 5.0
2 b)
I α
B
ω
4.0
2ω ce
7s
3.0

Te (keV)
3s
I B(I - e-τ )
I
2.0

ω ce
ω 2 Te (R) 1.0

0.5s
0
0 50 100 150 200 250 300 350 3.0 3.5 4.0
Frequency (GHz) Major radius (m)

Figure 7.10: a) Electron cyclotron emission as a function of frequency, b) Radial profile

of the electron temperature from ECE measurement

This differential equation can be solved by multiplication by e−τ ;

d −τ

e I = −IB e−τ . (7.38)
dτ
Assume that IB is constant along the beam. Integration along the entire beam from
τ = τ1 (s = s1 ) till τ = τ2 = 0 (s = s2 ) then gives

I(s2 ) − I(s1 )e−τ1 = IB 1 − e−τ1 .

(7.39)

The intensity incident on the far plasma boundary can be neglected, i.e. I(s1 ) = 0. The
following expression for the intensity observed from a plasma with optical depth τ can
then be obtained from eq. 7.39;

I(ω) = IB (ω) 1 − e−τ (ω) .

(7.40)

Two limiting cases for the emitted radiation intensity, depending on the plasma re-
absorption, exists:

1. Optically thin plasma, τ  1: In this case, the intensity of a beam with length L in
the plasma is equal to I ≈ IB τ ≈ IB αL = jL. The plasma absorption of radiation
is negligible.

2. Optically thick plasma, τ  1: In this case, I ≈ IB , i.e. the plasma radiates with
the same intensity as a black body. This is the highest possible radiation intensity.

In a plasma where the magnetic field varies over space, electron cyclotron radiation can
give information on the local electron temperature, i.e. the plasma temperature profile
Te (R) can be calculated. Emission of electron cyclotron radiation of a given frequency
ω occurs at a resonance layer in the plasma where the resonance condition ω ≈ nωce is
fulfilled. According to Kirchoff’s law absorption of radiation of this frequency will mainly
take place in the resonance layer. Outside the resonance layer we can assume that the
radiation passes with negligible absorption, i.e. the plasma is considered optically thin.
(τ (ω)  1). If the resonance layer is optically thick (τ (ω)  1), the emitted electron
cyclotron radiation is given directly by IB (ω) and thus the local electron temperature Te

166
 Figure 7.11: Example of measured X-ray radiation spectrum.

in the resonance layer. The radiation intensity is in general given by eq. 7.40 using τ (ω),
the optical depth for the resonance layer.
The magnetic field in a tokamak is, in a first approximation, assumed equal to the
vacuum toroidal field, B ∼ 1/R, where R is the major radius. The resonance condition
for electron cyclotron radiation then gives
R0
ω = nωce0 (7.41)
R
where R0 is the plasma major radius in the centre and ωce0 is the electron cyclotron
frequency at R0 . The resonance frequency thus varies with position in the plasma as
ω ∼ 1/R. The observed broadening of the emission line for a given harmonic n, i.e. the
continuous radiation spectrum around ω ≈ nωce0 (figure 7.10a), thus gives the electron
temperature as a function of the frequency

8π 3 c2
Te (ω) = I(ω) (7.42)
eω 2
from which the profile of Te (R) is given if we use the relation between ω and R (eq. 7.41)
(figure 7.10b).

7.5 Measurement of bremsstrahlung spectrum

Electrons emit electromagnetic bremsstrahlung when accelerated by Coulomb collisions in
the plasma. This electromagnetic radiation is in the X-ray range for a hot plasma. When
the electrons have a thermal velocity distribution given by the electron temperature Te
we get the following expression for the energy spectrum of the emitted X-ray photons .

n2e Zeff
 
dN E
E ∼ √ exp − (7.43)
dE eTe eTe

167
 Torus wall Liquid
N 2tank
Aperture and Be
foil box window

Plasma 2
1

Flight Si(Li)
Aperture box tube crystal

3
To multichannel Connection
analyser to electronics
X-ray pulse Amplifier

Figure 7.12: Diagnostic for measuring X-ray radiation.

V
αE

Figure 7.13: Schematic picture of the signal from the silicon detector.

where dN/dE is the total number of photons with energy E emitted per unit energy, time
and volume. The bremsstrahlung gives a continuous spectrum. Measuring the absolute
value of the intensity in a given energy interval can in principle give an estimate of the
effective ion charge Zeff . The quantity E(dN/dE) has a simple exponential dependence
on the photon energy E. This can be used to estimate the electron temperature Te (figure
7.11). Besides the continuous spectrum there is also a discrete spectrum, caused by line
radiation which complicates the measurement. Usually an energy interval ∼ 1–10 keV of
the spectrum is used, i.e. soft X-ray radiation. An example of an experimental setup for
measurement of the soft X-ray spectrum is shown in figure 7.12.
The diagnostic consists of the following components:

1. Absorption foil. A thin disc of a suitable material (e.g. beryllium) absorbs the
low-energy part of the spectrum since it should not be analysed.

2. Detector of lithium-drifted silicon. The detector is cooled with liquid hydrogen

to decrease noise. Thus, the detector can measure single photons. A voltage pulse
proportional to the photon energy is generated every time a photon hits the detector
(figure 7.13).

3. Multichannel pulse height analyser. Pulse height analysis, i.e. counting the num-
ber of voltage pulses in a certain voltage interval, finally gives the photon energy
spectrum.

7.6 Neutral particle analysis

Neutral hydrogen atoms are created when plasma ions recombine at the wall surface.
Low energy (“slow”) neutral hydrogen atoms are transported into the plasma by different
processes. Ionization occur through electron collision, ionization through ion collision and
charge exchange collision with ions. At a charge exchange collision the electron from the

168
 -18
σ 0l -13
10 10

σ 0l υ i (m 3 s -1)
-19 -14
10 10

σ 0l (m 2 )
σ 0l υ i
-20 -15
10 10

-21 -16
10 10
1 10 10
2
10
3
10 4 10 5
1
2m i νi2(eV)

Figure 7.14: Cross section for charge exchange as a function of ion energy.

neutral hydrogen atom is transferred to the ion. Thus, a new neutral atom is formed.
This new neutral particle will have high energy since the ion keeps its kinetic energy when
capturing the electron. In this way, a new population of “fast” neutral particles is created
in the plasma.
+ +
Hslow + Hfast → Hslow + Hfast (7.44)
These high energy neutral particles leave the plasma since they are not confined by the
magnetic field. The “fast” neutral particles will have the same velocity distribution as the
ions. The ions usually have a thermal velocity distribution given by the ion temperature
Ti . The cross section for the charge exchange collision is known (figure 7.14). Thus, an
analysis of the energy spectrum of the flux of fast neutral particles outside the plasma is a
measurement of the ion temperature Ti . Measuring the absolute value of the fast neutral
flux also tells us the density of the “slow” neutral component in the plasma.
An expression for the spectral density of the fast neutral component is obtained in
the following way: Consider the normal case that the electrons have a thermal velocity
distribution. The distribution function f (v)dv gives the number of ions with velocity v
in the interval dv per unit volume in the plasma
3/2
mi v 2
  
mi
f (v)dv = n exp − 4πv 2 dv (7.45)
2πeTi 2eTi
where n is the ion density and mi is the ion mass. The reactivity for charge exchange
collisions for a given velocity v is
ξ(v) =< σch (|v − va |)|v − va | >va (7.46)
where σch is the cross section for charge exchange collisions (which depends on the rel-
ative velocity of the colliding particles), v is the ion velocity, va is the velocity of the
“slow” neutral atoms, and < . . . >va denotes an averaging over the slow neutral velocity
distribution. Since |v| = v  va , the following expression for the reactivity can be used
as an approximation
ξ(v) ≈ σch (v)v (7.47)
The number of fast neutral particles created by charge exchange collisions with velocity
v in an interval dv per unit time and unit volume is
S(v)dv = na ξ(v)f (v)dv = na σch (v)vf (v)dv (7.48)

169
 Neutral
flux
35

Log intensity (arb)

Ti =5.27 keV

30

25
0 10 20 30 40
E 0(keV)
Energy

Figure 7.15: Example of measured distribution of neutral particle energies.

where na is the density of slow neutral particles. Assume that the plasma volume observed
can be considered a cylinder with area A and length L, and that the detector sees a flux
within a solid angle Ωs . The number of particles detected per unit time can then be
written as
Ωs
F (v)dv = A L S(v)dv. (7.49)
4π
Inserting eq. 7.45 and 7.48 in 7.49 gives the following expression for the measured neutral
particle flux;
3/2
mi v 2
  
mi
F (v)dv = Ωs ALnna σch (v) exp − v 3 dv. (7.50)
2πeTi 2eTi
Alternatively, the expression can be written as a function of the ion energy E (figure
7.15);
3/2 
 
1 E E
F (E)dE = Ωs ALnna σch (E) √ exp − dE. (7.51)
πeTi 2mi eTi
The spectral density F (E) divided by σch (E) E has a simple exponential dependence
on the ion energy E  
F E
∼ exp − (7.52)
σch E eTi
which gives the ion temperature Ti .

7.6.1 Neutral particle analyser system

One type of particle analyser is shown in figure 7.16. The system consists of the following
parts:
1. H2 gas cell. Neutral hydrogen atoms from the plasma are ionized in a chamber with
neutral hydrogen gas.
2. Magnetic deflection. The ions pass a perpendicular magnetic field. The curvature
radius of the ions (i.e. their Larmor radius) is rL = v/ωci = mi v/qB, where q is the
ion charge. Ions of different velocity v are deflected to different detectors.

170
 2 Magnetic
pole piece
+
H trajectory B

H2 gas cell Electrostatic

E 3 deflectors
1
o - -
H from plasma + +

Channeltron
detectors

Figure 7.16: Neutral particle analyser.

3. Electrostatic deflection. A perpendicular electric field separates ions with a given

q/mi .

171
Chapter 8

THE TOKAMAK AND THE

ALTERNATIVES

The tokamak is the most developed magnetic confinement configuration. The reason for
this is the good confinement properties and relatively simple geometry of the tokamak.
The resources for tokamak research, in terms of capital investment and man-years, have
been incomparably stronger than those for other confinement schemes.
The diagram in figure 8.1 summarizes results from some of the largest tokamak exper-
iments JET (Europe), TFTR (USA), JT-60U (Japan) and DIII-D (USA). Data from the
compact high-field tokamak Alcator (USA) is also included. By the Q factor we denote
the relation between injected energy and generated fusion energy. Thus, Q = 1 stands for
“break even”. It is clear that enormous progress has been achieved. The triple product
nτE T has increased by a factor 1000, and only a factor of five remains to the region, where
ignition is obtained.
In this chapter we will familiarize ourselves with some of the most successful alter-
natives to the conventional tokamak. This includes both variations of the tokamak, e.g.
the advanced, the spherical and the compact tokamaks, and quite different configurations
like the reversed-field pinch (RFP), the stellarator and the high density pinch. We will
also briefly discuss the principles for inertial confinement. Finally, we will touch upon the
hybrid reactor, which combines characteristics of fusion and fission reactors.

8.1 Tokamak
But let us start with the future development of the conventional tokamak. A schematic
drawing of a conventional tokamak is shown in figure 8.2.
The next step device in the international tokamak research line is the ITER device,
currently under construction in France.

8.1.1 ITER
ITER is latin for ”the way”. It is the name of the next-step device in the tokamak research
line, aimed to ”lead the way to fusion energy”. It is a scaled-up version of present-day
tokamaks. Based on extrapolations of the fusion plasma performance in these devices,
ITER is projected to generate 500 MW of fusion power. It will be the next major step in
fusion research. ITER is a collaboration between EU, Japan, Russia, China, South Korea,
India and USA. Hence nations, representing more than half of the world’s population are
engaged in the project.

172
 Figure 8.1: nτE T diagram (Figure credit: EUROfusion-JET).

Figure 8.2: The principle of a tokamak. (Figure credit: EUROfusion-JET)

173
 ITER is currently (2022) under construction at the Cadarache site in France, and is
expected to come into operation around 2025. ITER is planned to be the final step before a
demonstration reactor aimed at electricity production is constructed. The most important
task for ITER is to generate significant self-heating from nuclear fusion reactions - at the
same time providing a good enough plasma energy confinement - so that the tokamak can
be maintained in a nearly self-sustained ”burning” state. This has never been achieved in
any tokamak, and reaching this milestone would take fusion research a big step forward to
the realisation of fusion reactors. ITER will also serve as a test bench for the technology
of an integrated reactor system. This includes effects of high heat and neutron flux in
addition to studies of the functionality of reactor components.
ITER is a large tokamak with superconducting magnets generating about 500 MW
of fusion power for at least 400 seconds. The plasma volume will be ten times that of
JET and will be close to the size of future commercial reactors. The ITER project will,
for the first time, enable the study of plasmas heated by the hot alpha particles (helium
nuclei) produced in the fusion reaction. ITER will generate ten times more power than the
externally input power required to heat the plasma, equivalent to an energy multiplication
factor of Q ∼ 10.
For more information on ITER, see www.iter.org.

8.2 Alternative configurations

There has been no shortage of ideas concerning principles for magnetic confinement.
Some 50-100 different configurations have been constructed and evaluated, but only a
few remains interesting today. The reason is usually the difficulty to maintain MHD
stability. There used to be a certain optimism that MHD unstable configurations could
be stabilized by non-ideal effects such as large Larmor radius orbits, rotating electron
rings, surrounding high density gases and other mechanisms. This was the case for the
zpinch, which indeed is stabilized by kinetic effects so that instabilities grow more slowly
than those forecasted by ideal MHD, but nevertheless it suffers from unacceptably fast
growing sausage and kink instabilities.
Stability against MHD modes is a very strict requirement, since the Alfvén time in
most configurations is shorter than 10−6 s, and the energy confinement time should be of
the order of 1 s. In inertial fusion, this problem is partly avoided, as we shall see later.
A suitable starting point for a discussion on alternative configurations of magnetic
confinement can be to study the advantages and disadvantages of the tokamak from the
points of view from power industry and society. Consider the table below, where we state
some desirable properties for a fusion reactor in comparison to those of the conventional
tokamak.
PROPERTY Conventional tokamak
Compact size no
High beta value no
Burning plasma yes
Continuous operation ?
Low recirculating power ?
Possibility to use advanced fuels no
Possibility for direct conversion to electricity no
Environmentally superior yes
COE lower than that of fission or fossil fuels no

174
 Figure 8.3: The riggatron.

Several of these are correlated. A high beta value and magnetic field, for example,
results in higher density and smaller dimensions. Let us classify today’s most important
research paths in alternative magnetic confinement by beta value and magnetic field.

High beta value, low magnetic field : Reversed-field pinch (RFP), spherical tokamak.
Low beta value, high magnetic field : Compact tokamak (Ignitor, Riggatron).
Low beta value, low magnetic field: Stellarator.

8.2.1 High-field tokamak

A high-field tokamak utilizes the high stability and confinement of a tokamak in a very
high magnetic field ≈ 15 T. An example of a very high energy density reactor design is
the Riggatron in figure 8.3.
The blanket is placed outside the entire reactor core (FPC, see the next chapter),
which must be exchanged monthly because of the enormous neutron flux. The idea is to
do this using a “plug-in” technique to eliminate the long shut downs normally associated
with exchange of limiter and wall. The toroidal coils are made of copper, and thus, the
machine must be pulsed.
An important problem to solve is that the copper coils limit the neutron flux to the
blanket, where T is formed from Li.

8.2.2 Spherical tokamak

In a spherical tokamak configuration, the toroidal field coils have been eliminated and
replaced by a straight central conductor. This enables a very low aspect ratio R/a ≤ 1.6
and a naturally elongated plasma cross section which is stabilizing and gives a high beta
value. It also provides high safety factor, q, as well as high shear.

175
Figure 8.4: The stellarator Wendelstein VII-AS in Garching, Germany. (Figure credit: EURO-
fusion)

A toroidal beta of 20 % has been measured in the first experimental device START
(Small Tight Aspect Ratio Tokamak) in Culham, England. Some other data for this
experiment are R/a = 0.3 m/0.25 m, Ip ≤ 250 kA, Bϕ ≤ 0.5 T, τ ≈ 50 ms and τE ≈ 2
ms.
The high energy confinement and the high beta value seem very promising and larger
machines like MAST (Culham) and (Princeton, USA) are now evaluating possible plasma
parameters and scalings. The problems associated with spherical tokamaks are mainly
design of blanket and cooling of the central conductor.

8.2.3 Stellarator
One of the earliest principles for magnetic confinement is the stellarator, which was men-
tioned already in the introduction. Since the plasma shape does not have toroidal sym-
metry, it has recently become possible to computationally design the magnetic field coils
with sufficiently low field errors and high confinement. The stellarator is now considered
the most important alternative to the tokamak.
A major advantage is that no toroidal plasma current is needed. This enables the con-
figuration to be driven continuously using negative ion beam heating, possibly combined
with ECH. The modular coils are superconducting. The configuration also features low
orbit losses and low neoclassical diffusion.
Like the tokamak, the geometry is toroidal. The rotational transform is maintained
by a deformation fo the torus (early experiments showed poor confinement) or by having
the external coils form a spiral. Almost the entire confining magnetic field is generated
by external coils. The resulting plasma cross section will not be axisymmetric. The
stellarator is thus a typical three-dimensional configuration (see figure 8.4).
Since the magnetic field in the stellarator is mainly maintained by external coils, the
beta value may not be too high, in which case it will cause internal currents break-
ing the topology. Consequently, a stellarator must be physically large. The dimensions
R/a = 20 m/(1–2) m are realistic, with plasma parameters βϕ = 0.02 och Bϕ = 5 T.

176
 Particle motion in a stellarator is divided into three cathegories:
* circulating orbits, where the particles travel toroidally without being mirrored
* “helically trapped” particles, mirror reflected in the spiraling field
* “toroidally traped” particles, moving in banana orbits as in tokamaks
Depending on how the rotational transform is created the stellarator configurations
may be divided into different sub-types. Examples are the classical stellarator, the tor-
satron, the helias (like the Wendelstein VII-X experiment in Greifswald, Germany) and
the heliotron (like the large active Japanese experiment Large Helical Device; LHD).
Much stellarator research is focused on the possibility of reaching high beta values,
hopefully 5–10 %. Transport studies, however, are difficult to perform because of the 3-D
geometry. Particle and energy transport are dominated by the way particles are helically
or toroidally captured (and by unwanted variations, so called ripples) in the magnetic field
and by the collision frequency. The latter influences how the particles are scattered in
the different orbits, which are characterized by loss frequencies. In axially asymmetrical
systems electrons and ions may diffuse independently of each other.
In general it can be said that the energy confinement time has a similar scaling de-
pendence as for the tokamak, but with a certain deviation in the low collision range,
where the scattering of helically trapped particles enhances the diffusion. The so-called
advanced stellarators aim to suppress this mechanism.
Summarizing, we conclude that the stellarator has certain advantages compared to
the tokamak. The stationary magnetic field simplifies the design of the magnetic coils
(the superconducting coils do not require pulsing, and thus their energy does not need to
be stored). The high aspect ratio in combination with the absence of a transformer core
and the modular design simplifies construction and makes the stellarator interior easily
accessible. The continuous operation allows simpler blanket design (and less wear). The
large wall surface results in less wall load and longer operation time before exchange of
wall material.

8.2.4 Reversed-field pinch

A reversed-field pinch (RFP) is in many ways similar to a tokamak. The plasma is toroidal,
axisymmetric and confined by combined toroidal and poloidal magnetic fields Bϕ and Bθ
(see figure 8.5). One important difference between the two devices is that in the RFP,
Bϕ och Bθ are comparable, while in the tokamak the externally generated toroidal field is
about ten times larger than the poloidal field. Another difference is that Bϕ in the RFP
is mainly generated by currents in the plasma.
The name RFP indicates that the toroidal field is reversed (compared to the direction
of the externally applied toroidal field) in the outer plasma region (see figure 8.6). This
gives a very high stabilizing shear. The reversal can be shown to be a consequence of
the plasma’s efforts to minimize its magnetic energy. During plasma start-up, this is
performed through intense MHD activity, while a number of modes generate a lower
MHD activity, so called dynamo, during the later quasi-stationary phase. In the latter
phase, there is a constant, slow resistive diffusion away from the minimum energy state
followed by a rapid transition back to this state.
In comparison to the tokamak, the RFP has the important advantage that the exter-
nally required magnetic field is low enough to allow for the use of copper coils. Further,
higher plasma currents can, at least in principle, be reached since these are not limited by
the size of the external toroidal field. This would allow for construction of more compact

177
 Figure 8.5: Experimentally measured magnetic fields in an RFP.

Figure 8.6: The toroidal magnetic field in an RFP changes direction in the outer region.

and thus more economical fusion devices. In addition, already early experiments showed
that high beta values could be sustained in the configuration.
The main problems for the conventional RFP is, at present, the quite low energy
confinement and the need for wall stabilization. We will now discuss these issues in some
detail.
MHD instabilities in an RFP can be divided into five main categories:
• m = 0 modes; are resonant at the reversal surface, where the axial field changes
sign.
• Internally resonant m = 1 modes; have n < 0 and are resonant inside the reversal
surface.
• Externally resonant m = 1 modes; have n > 0 and are resonant outside the reversal
surface.
• Internally non-resonant m = 1 modes; have n < 0 and a pitch similar to that of the
equilibrium magnetic field at the magnetic axis.
• Externally non-resonant m = 1 modes; have n > 0 and a pitch similar to that of
the magnetic field just outside the plasma boundary.
In a conventional RFP, the plasma is surrounded by an adjacent conducting wall with an
L/R time (“shell time”) being longer than the lifetime of the plasma pulse. This stabilizes
all modes except those of category 2) above, the internally resonant modes. The latter
modes are also called dynamo modes, since they are responsible for the plasma’s constant
relaxation towards the minimum energy state.
The pulse length in an RFP reactor must be greater than the shell time. Because of
this, it is important to experimentally demonstrate that the RFP configuration can be
maintained with a resistive shell. This is a field of intense experimental and theoretical
research.

178
 Figure 8.7: The RFP Extrap T2R at the Alfvén laboratory, KTH.

One possibility is to rotate the plasma, so that the magnetic field constantly “sees” a
new wall to penetrate. Unfortunately, this requires a rotation velocity close to the Alfvén
velocity, which is unrealistic. So, even if the plasma is rotating, it is unstable to dynamo
modes, as well as to internal and external non-resonant modes (growing on a characteristic
L/R time scale). These modes are called resistive shell modes.
In the Extrap T2R RFP at the Alfvén laboratory at KTH in Stockholm (see figure
8.7) the effect of resistive shell modes on the plasma dynamics is investigated. Previous
experiments at the RFP HBTX-1C in Culham, England, showed that non-resonant, non-
rotating resistive shell modes have a strong, undesired effect on the plasma behaviour.
The Extrap T2R experiment at KTH is characterized by the following parameters:
R/a = 1.24 m/0.183 m, Ip = 100–150 kA, βp = 0.05–0.10 and < n > = 0.5–1.0×1019
m−3 .
From the discussion above it follows that the resistive shell modes must be reduced
through some sort of active feedback. This is similar to the case for the advanced tokamak,
in which the plasma current is continuously generated from an external source in combi-
nation with internal, pressure gradient driven currents. The enhanced current results in
higher beta value and more compact dimensions that those of the conventional tokamak.
In the advanced tokamak, there is only one resistive shell mode, which makes the design
of the feedback coils relatively simple. In the RFP, on the other hand, a large number of
modes with different pitch must be stabilized, which gives a more difficult problem, being
intensively studied at present.
In the conventional RFP, the confinement time is limited due to the magnetic field
fluctuations caused by the dynamo. A stochastization of the magnetic field in the inner
region of the plasma results, with accompanying radial heat transport along the field lines.
At the boundary, heat is transported by perpendicular ion heat conduction along a rela-
tively sharp temperature gradient. This also occurs in the presence of a nearby conducting

179
Figure 8.8: The principle for inertial fusion. (Figure credit: http://physci.llnl.gov/divisions
/fep/ifep/IFEP.html)

wall. Computer simulations have given the relations βp ≈ 3.4 × 10−5 a−0.2 (I 2 /N )−0.40 and
τE ≈ 0.77a1.4 (I 2 /N )0.34 , where N is the line density. These approximatively correspond
to optimized transport in the conventional RFP. The earlier reactor study TITAN uses
the parameters R/a = 3.8 m / 0.6 m, Ip = 18 MA, βp = 0.2 and < n >= 9 × 1020 m−3 ,
resulting in a too optimistic beta value according to the scaling above. In the advanced
RFP, in which the dynamo activity has been reduced by the method of pulsed poloidal
current drive (PPCD), confinement is strongly enhanced. Experiments and computer sim-
ulations show that the energy confinement can be improved several times by this method.
Experimental and theoretical RFP research is mainly carried out in Italy, USA, Japan,
China and Sweden. It is still too early to evaluate the reactor potential of the RFP. Topics
that must be addressed are, as mentioned, resistive wall modes, current profile control
techniques to master high magnetic fluctuations and weak energy confinement, as well as
heating and current drive.

8.3 Inertial confinement fusion

The possibility to generate fusion reactions by inertial confinement has been studied and
developed experimentally for a long time. It has been a successful approach, and the
research yields results comparable to those of magnetic confinement research. Inertial
fusion has very recently (2022) achieved energy break even. Two very large devices are
operating; National Ignition Facility (NIF) in Livermore, CA, USA, and Laser Mega Joule
facility (LMJ) in CESTA, near Bordeaux in France.
We will not in any way do justice to this approach, but only briefly mention the
underlying principles. The idea is that fusion reactions will be achieved by compression
and heating of a frozen grain (a radius of a few mm) of DT. This is done by focusing laser
or ion beams on the grain. The process is very fast; the reaction takes place in a few ns.
During this short time span, the fuel is held together by its own inertia, hence the name
inertial confinement fusion. Thus, the process eliminates the need for external magnetic
fields. The figure below describes the process.
Let us perform a naive estimate of which parameters are involved. Frozen D (or T)
has a density of around 5 × 1028 m−3 . Let us assume that the temperature 10 keV can be
reached during 10−8 s. Then the necessary Lawson criterion nτE > 1020 s m−3 is fulfilled
if the energy confinement time corresponds to the time the grain holds together, say 10

180
ns. Thus τE ≈ τ ≈ R/vT , where R is the grain radius and vT is the thermal velocity. The
radius of the grain is then 0.01 m.
In reality, the DT grains are not made of massive fuel. In the above example a massive
grain would require around 100 MJ to heat up to 10 keV, an energy input which is not
possible with the lasers available today.
The grain is instead made as a hollow sphere. The innermost layer is frozen DT.
Outside is added a so called pusher and the outermost layer is some easily evaporated
material. When the grain is irradiated, the surface is evaporated. The reaction force
compresses the DT fuel to around 1000 times the normal density. The laser energy 1–10
MJ is incident during 1–10 ns, corresponding to an enormous power. The associated
Lawson criterion for inertial fusion is usually written ρR ≥ 10 kg/m2 .
It is not straightforward to carry out the process described above. The major obstacle
has been Rayleigh-Taylor instabilities which easily arise during the compression. By
fabricating very smooth grains and developing homogeneous laser and ion beams, most
stability problems now seem to be solved. The figure below shows a computer simulation
of the time development of Rayleigh-Taylor instabilities.

Figure 8.9: Contours of density and temperature in a computer simulation of inertial fusion.
(Figure credit: FUSION EXPO-EUROfusion)

To achieve homogeneous radiation of the grain, researchers do not only use “direct
drive” as above. The grain may also be surrounded by a gold chamber. When the walls
of the chamber are radiated, they emit high energy bremsstrahlung which compresses and
heats the grain in a process called “indirect drive”. The process is imaged in figure 8.10.
A reactor based on inertial fusion will necessarily be large, because of the large beam
sources. The reactor chamber will have a diameter of about 10 m. The inner walls will
be covered with a fluid Li curtain which captures neutrons and transforms Li to T. The
reactions must occur with a frequency of about 10 Hz. So far, it has not been decided
if direct or indirect radiation will be used, or if laser or ion beams are the final solution.

181
Figure 8.10: The principle for “indirect drive”. (Figure credit: Lawrence Livermore National
Security, LLC, Lawrence Livermore National Laboratory, and the Department of Energy, USA)

(He++ ion beams give higher efficiency in the beam source and better coupling to the
plasma). Reactor studies indicate a rather low COE; 0.05–0.07 Euro/kWh(electricity).

8.4 The fusion-fission hybrid reactor∗

It was early realized that it may be possible to build a reactor which uses neutrons from a
fusion source and generates energy from the resulting fission reactions. The fusion reactor
is thus surrounded by a blanket of fertile or fissile substances. Compared to fission breeder
reactors, an obvious advantages is that the hybrid reactor can operate sub-critical, since
the neutron flux is controlled by the fusion process and not by fission processes in the
fuel. This gives a more controllable and safer reactor. From a fusion point of view, the
fusion-fission hybrid reactor is less demanding to construct than a “pure” fusion reactor.
The main task is to produce large fluxes of neutrons, but the fusion plasma does not need
to reach break-even. Further, the DD reaction can be used, which eliminates the need
for Li and T handling. The hybrid reactor can be developed along two main paths, to be
described briefly.

Heat producing hybrid reactor

Natural U238 or Th232 can form a blanket around an inner fusion reactor which constitutes
a neutron source. The neutrons cause fission of the nuclei and heat production in the
blanket, which is handled in the traditional way. In a reactor scenario this process may
generate 400 MW neutron power, yielding 10 times as much fission power, or 4 000 MW.

Production of nuclear fuel for fission reactors

The uranium supply on Earth is limited. This has led to the development of the so
called fission breeder reactor. In a fission reactor with neutron multiplication using Be, it
possible to breed fissile U233 fuel from Th232 . A hybrid reactor would feature radioactivity
levels only a tenth of those of an ordinary fission reactor while producing fuel for 5–50
fission reactors.

182
Chapter 9

THE FUSION REACTOR

In the European “fast-track schedule” for realization of fusion power, the goal is to have a
demonstration reactor (DEMO), demonstrating electricity to the grid, by the mid-century.
In a reactor, a larger plasma volume than that in today’s tokamaks is required, to provide
the increase in confinement time necessary for power break-even. Further, a reactor must
generate it’s magnetic field for much longer time periods than current tokamaks, which
necessitates the use of superconducting magnetic field coils.
In this chapter we will just scratch the surface of fusion technology. As fusion physics
problems successively become solved or effectively handled, fusion technology issues be-
comes increasingly important in the research and development of fusion. ITER will cer-
tainly provide an important test bench for developing many technological solutions before
we may embark on DEMO and commercial fusion. A new fusion materials test facility
(IFMIF-DONES), will help to optimise the materials that should be used for plasma-
facing components, blanket and the reactor structure.
The next step tokamak device, ITER, will integrate some of the technologies essential
for a fusion reactor, such as superconducting magnets, and test tritium breeding blanket
designs. Many new technologies, from superconducting magnets to novel materials are
being developed for ITER. The ITER device comprises the tokamak itself and associated
systems for heating, fuelling, exhaust of waste and gas, control, and diagnostic measure-
ments. ITER, like JET has a vertical D-shaped plasma and a lower divertor system. The
divertor is a critical component and the main area where the plasma will contact the
material wall. ITER will use low temperature (4 K) superconducting magnets for both
its 18 toroidal and 6 poloidal field coils as well as the central solenoid coil. The toroidal
field coils generate a magnetic field of 5.3 T in the plasma region. The whole ITER device
is enclosed in a 70 K cryostat which helps insulate the superconducting magnetic coils.
External heating systems provide input heating power of about 50 MW using neutral
beam injection and radio-frequency electromagnetic wave heating. The inner surfaces of
the vacuum vessel are covered with blanket modules. These modules will provide shield-
ing from the high-energy neutrons produced by the fusion reactions and will be used for
testing tritium breeding concepts.
The ITER plasma containment vessel will be more than twice as large and 16 times
as heavy as any previously manufactured fusion vessel. The internal structure of a fusion
reactor will become radioactive during operation due to neutron radiation and the presence
of tritium. Remote-handling systems are therefore vital to be able to replace components,
such as the divertor and eventually breeder blanket modules, inside the machine. In a
fusion power plant, cryogenic systems are used to remove impurities from the plasma, cool
the superconducting coils to allow them to operate, separate the waste gases into their

183
 1 5

6
2

7
3

Figure 9.1: ITER. 1) Central solenoid coil, 2) Toroidal field coil, 3) Poloidal field coil, 4)
Diagnostics, 5) Blanket module, 6) Vacuum vessel, 7) Cryostat, 8) External heating sys-
tems 9) Divertor (From ITER - Uniting science today global energy tomorrow, European
Commission, 2007, p. 14)

different individual components for disposal or recycling for fuel, provide the cooling for
radio-frequency heating sources, and control the gas pressure of neutral beam systems.
Large-scale vacuum systems are required to ensure an ultra-high vacuum in the large
reactor vessels and to maintain the vacuum surrounding the superconducting magnets in
the cryostat.
Plasma heating systems are essential for obtaining a high-temperature plasma. For
ITER, the fusion reaction would not continue if the plasma was not heated by an external
source, and for both ITER and future power plant operation it is likely that the heating
systems will be an essential tool to ensure stability and control of the plasma. Initially,
three main types of heating systems will be deployed for ITER:

• Ion Cyclotron Resonance Heating - in this system, ions in the plasma are heated
by electromagnetic waves with a frequency of 30-50 MHz. The main issues concern
how to couple the intense radiation to the plasma and what effect this has on the
performance of the plasma.

• Electron Cyclotron Resonance Heating - here the electrons in the plasma are heated
by electromagnetic waves with a frequency of 100-200 MHz. This radiation has the
advantage that it can be transmitted through air which simplifies the design and
means that the source can be far from the plasma.

• Neutral beam injection - in this system, charged fusion fuel particles are accelerated
to very high speed corresponding to a kinetic energy of 1 MeV and neutralized so

184
 Figure 9.2: The reactor. (Figure credit: EUROfusion-JET)

that high-energy neutral particles can pass through the magnetic field and enter the
fusion plasma. As a result, the plasma is heated by the transfer of kinetic energy.

In ITER, many instruments measuring a variety of parameters are needed to con-

trol the plasma performance, including temperature, density and the type of impurities
present. Diagnostics must be developed to monitor every aspect of the machine. Plasma
diagnostics for ITER fall into three categories; those necessary for machine protection
or basic control; those needed for advanced performance control; and those desirable for
physics studies. There will be about 45 different diagnostic system deployed around the
ITER tokamak and they will use a variety of measurements based for example on mag-
netic, optical and microwave techniques. The most reliable way of measuring temperature
is to shine a very powerful laser into the plasma. The photons in the laser beam scatter
off the energetic plasma electrons and this scattered light can be measured. The Doppler
shift in the wavelength of the scattered light gives a direct measurement of the speed and
hence the temperature of the electrons, whilst the intensity of the reflected light is related
to the density of the plasma.

9.1 Reactor components and function

Figure 9.2 illustrates the main functionality of a future fusion reactor. The time has now
come to discuss the components of the fusion reactor, as well as its function and safety
issues. The main components in the fusion reactor constitutes the so called fusion power
core. They are, from the inside
• divertor / limiter
• first wall
• breeding blanket
• shield
• magnetic field coils

185
 Figure 9.3: Limiter and divertor. (Figure credit: EUROfusion-JET)

We shall now describe these components in turn.

9.1.1 Divertor / limiter

The purpose of the divertor/limiter is to divert impurities and helium ash from the plasma
boundary, and to protect the wall from the plasma. In tokamaks there are sometimes
plasma disruptions, which may seriously damage the wall if it is not protected. The
plasma is separated from the wall by a separatrix, introduced in the magnetic field. From
the separatrix, field lines lead into the divertor which extends toroidally at the bottom of
the plasma vessel.
In order to remove heat, fusion products (helium) and other impurities from the
plasma, the plasma will be allowed to touch its surrounding structure in a controlled
manner. This is achieved by shaping the magnetic field lines in such a way as to enter the
divertor. The divertor consists of two targets designed to withstand that loads of up to 20
MW/m2 . Contact with wall materials elsewhere needs to be minimized as this will erode
the vacuum vessel surface and reduce the lifetime of reactor components. The material
currently used as the target of the divertor is either carbon reinforced with carbon fibre
or tungsten. In addition to this critical part of the divertor design, it is also important
to design components that can withstand the high electromechanical loads experienced
in the reactor chamber, allow high-vacuum pumping to remove helium from the plasma,
and tolerate long exposure to neutron radiation.
Because of this, methods for expanding the scrape-off layer (for example by adding
low Z impurities) at the plasma boundary are developed. This provides a more evenly
distributed load on the divertor. The divertor must be changed more often than other
reactor components and is built into a cassette for easy access.
A limiter is a plate of a suitable material, physically stopping the plasma from direct
contact with the reactor walls. Low Z materials are often used, such as graphite, boron or
beryllium, but molybdenum and tungsten have also been successful. Limiters can extend
both poloidally and toroidally.

186
9.1.2 First wall
The part of the wall being closest to the plasma is called the first wall. It has a protecting
layer of low Z material and inside this some material with very high heat conductivity.
Copper and steel are less suitable materials in a reactor because of their relatively long
lived radioactive isotopes. Preferable materials are low-activity materials such as silicon
carbide, titanium or aluminium. The first wall must be able to withstand high neutron
fluxes. Most materials are destroyed with time, primarily through embrittlement, swelling
and endurance failure (especially due to disruptions). A rule of thumb is that if the
number of dpa (displacements per atom)  1, the crystal structure may become severely
damaged. In a fusion reactor, certain components might be exposed to 200 dpa. Thus,
the neutron flux is a measure of the lifetime of the first wall and the blanket. Stainless
steel is considered to withstand 15 MWy/m2 , but the requirement is somewhat higher.
The conventional tokamak has a relatively large wall surface / plasma volume ratio, which
prolongs the time before a change of the first wall/blanket is needed. In the ARIES-I study,
the calculated neutron load is 2.5 MW/m2 , which corresponds to 6 years of operation using
stainless steel walls. A more compact reactor would require more frequent changes of wall
material. A solution is here to reduce the cost of the FPC to about 10 % of the reactor’s
total cost, making it economically viable to exchange the entire FPC.

9.1.3 Breeding blanket

Just outside the first wall, the breeding blanket is placed.
A reliable and efficient breeder blanket technology is vital for heat transfer and fuel
generation. The energetic neutrons released from fusion reactions do not interact with
the plasma. The role of the blanket, which will surround a reactor is to slow the neu-
trons, recovering their energy as heat, as well as using them to transform lithium into
tritium. The tritium can then be extracted, processed and added to deuterium for refuel-
ing the reactor. By capturing neutrons, the blanket also shields other components, such
as superconducting coils, and thereby protects them from damage. Different concepts are
being explored for breeder blankets. The blanket must breed at least one tritium atom
for every fusion reaction in the plasma. Research in this area is concentrating on the use
of liquid-cooled lithium-lead and helium cooled solid ceramic breeder pebbles.
The blanket must fulfil the following requirements:
• T must be formed and disposed of effectively
• the heat conduction must be high
• it must be reliable and have a long lifetime
• it must provide access for maintenance and disassembly by robots
• it must withstand a heat load of 5–20 MW/m2
• it must have a working temperature of 800–10000 C
• it might have to withstand a pulsed magnetic field (10–20T)
As a rule, the blanket has to have a cooling agent flowing through it.
It is made of a suitable low-activity material and has the purpose of transforming the
energy of the decelerating neutron/charged particle to heat and create new T by the n
+ Li reaction. A few words on tritium formation in the blanket are relevant. Tritium
has a half-life of 12.3 years and is thus rare in nature. Therefore, it is suitable to use the
neutrons from the fusion reactions to create T from the reactions

Li6 + n → T + He4 + 4.8 MeV

Li7 + n → T + He4 + n - 2.5 MeV

187
 The first reaction is exothermal and gives a positive contribution to the energy balance
in the reactor. The neutron balance is also important, of course; each fusion reaction must
result in at least one T atom through these reactions, or the supply of T will eventually
run out.
No experimental device is running yet, but the most popular design solution is to
either let the cooling agent in part or in full be a Li salt, or to use solid Li2 O. The total
inventory of T in a reactor is 1–40 kg, depending on the type of reaction.

9.1.4 Shield
The magnetic field coils, generally superconducting, must be protected from neutrons
and heat. The shield, which is just outside the blanket, takes care of this while also
contributing to T containment.
The combined thickness of the blanket and the shield must be 1.5 m for the DT reaction
and 1.0 m for the DD reaction. The thickness only weakly depends on the material; it is
the same for 316 SS as for V and Ti.
A result of this is that the coils must be placed some distance away from the plasma,
which results in a lower field strength in the plasma. Apparently, the reactor becomes
larger and more costly when we take neutrons into account. Compact reactors, where
water cooled copper coils are inside the blanket have been proposed, but this causes
problems for the neutron balance (absorption in the coils). It is obvious that the use of
advanced fuels can mean a lot for reactor economy.

9.1.5 Magnetic field coils

Strong magnetic fields are required to confine the plasma. If conventional resistive elec-
tromagnets are used, a lot of energy is wasted in the form of heat. The magnetic coils
must be superconducting since large fusion devices require high toroidal fields and the
power consumption in normal conducting coils is too high. The large tokamak experiment
JET, for example, consumes 500–600 MW just for the toroidal coils.
Superconducting magnets have been developed for fusion devices. The most suitable
superconductors are so called type II high temperature super conductors, being discovered
in 1961. NbTi becomes superconducting at 10 K and can withstand fields up to 12 T in
the coil, while NbSn, which is a superconductor from 18 K, can withstand 22 T.
Cooling of these coils is done with the only material being fluid at these temperatures
- helium (boiling point 4.2 K). Operational tokamaks with super conducting coils can be
found, for example, in Cadarache in France (Tore-Supra).
The toroidal field coils and central solenoid in ITER are massive, weighing 290 tonnes
and 840 tonnes respectively, and are made from a superconducting alloy material con-
taining niobium and tin (Nb3 Sn). To achieve superconductivity, the coils must be cooled
to liquid helium temperature (4 K). At this low temperature the resistance of the super-
conducting material falls to zero, thereby greatly reducing the energy required for the
magnet. Nb3 Sn is a brittle material and construction of magnets weighing a few hundred
tonnes is not easy, but the material was chosen because it can support very high magnetic
fields. Each toroidal field coil starts with some 1100 wires about 0.7 mm thick twisted
together inside a 40 mm diameter metal tube to form conductors 820 m long. When
in use, supercritical helium flows within this tube and down a central gap to cool the
Nb3 Sn. The poloidal field coils will be made from a material containing niobium and
titanium (NbTi) which is more commonly used than Nb3 Sn. These coils are located in a

188
region where the strength of the magnetic field is low enough for this material to be used.
However, the position also means that replacing poloidal field coils will be very difficult,
so each coil will be designed with redundant turns so that any faults can be isolated to
ensure that operation of ITER continues unhindered. Once energized, the magnets can
work continuously with very high efficiency. As these magnets run at liquid helium tem-
perature it is necessary to operate them in vacuum to prevent heat in the atmosphere
from boiling off the helium.

9.1.6 Critical disturbances

The effect from some critical disturbances must be limited. One of these are the much
discussed disruptions, which may cause very large electromagnetic forces resulting in
damage of the vacuum vessel or the blanket. Further, the cooling and the superconducting
coils might be disturbed.
Disturbance of cooling is a well known phenomenon from fission power plants. For a
fusion reactor, other coolants are relevant. Liquid lithium is desirable from a breeding
point of view, but difficult to handle because of its high reactivity with air or water
and (since it is a metal) eddy currents and the resulting requirements for high pumping
power. Liquid lithium salts like FLIBE (a fluor-litium-beryllium mix) are electrically non-
conducting, have sufficient lithium content and cause low electrolytic corrosion. Other,
less reactive, coolants that should pass solid lithium oxide (Li2 O) might also be considered.
Very high field energies are stored in the superconducting magnetic windings, in a
tokamak mainly in the toroidal field magnets. If superconductivity would momentarily
be disturbed, the result might be very high heat deposition, that can melt and destroy
the material. Because of this, it is important that the magnetic field is secured and that
no overcritical magnetic fields (which destroy the superconductivity) are formed.

9.2 Safety issues

The first generation of fusion reactors will most probably be based on the DT reaction.
This causes some radiological problems, but these are much easier to handle than those
of a fission reactor. In a fusion reactor, it is possible to choose materials with short
half-life (short storage). Further, a core meltdown is impossible in a fusion reactor. Any
disturbance of the operation would cause an immediate shut-down of the hot plasma.
In the EU there is already a program for safety and environmental issues in fusion
power. The safety of a DT reactor mainly concerns mainly handling of the radioactive
tritium fuel.

9.2.1 Tritium
Tritium beta decays into a neutrino and a neutron (activity: 9600 Ci/g) with a half-life of
12.3 years.. The radiation can cause damage in biological tissue if T enters into the tissue.
The most common form of T when entering tissue is T2 O and THO, which replaces H2 O.
At present, three different reactor options are studied in the EU, based on different
construction materials. The fusion power is 3000 MW. In all these cases, passive safety is
important. No serious accidents are allowed to happen if the safety systems fail for some
reason. The requirement is that the general public should not be exposed to a dose higher

189
than 50mSv/y. Thus, the most serious accident must lead to no more than 50mSv doses,
and the present designs fulfils this requirement.
Since the T inventory (except in DHe3 reactors) is more than 50 g, a special confine-
ment of the entire reactor is needed. The vacuum vessel is separated from the surroundings
by both a vacuum wall and an outer wall. There is also a requirement on a minimum dis-
tance from densely populated areas. It is estimated that tritium confinement is disturbed
10−7 times/y, and this must be handled by the design. Evacuation is not considered to
become necessary.
We should note that also DD and DHe3 reactions generate T (around 0.35 and 0.005
kg/day, respectively), but this is mainly consumed in the plasma.
The handling of T dominates the external costs for fusion, which are low in comparison
with other energy sources. The external costs may be further reduced if the production
of C14 is avoided. This radioactive isotope (half-life 5730 y) stems from reactions with
neutrons and nitrogen, the latter being present in certain types of steel. To avoid the
production of C14 , other materials with less content of natural nitrogen could be used.
Alternatively, the materials could be enriched with respect to the isotope N15 (reduced
activation ferritic steels), resulting in lower levels of C14 .
If air gets into the vacuum vessel, an explosive mix of hydrogen and oxygen can be
formed, and cause release of tritium. The risk is, however, considered very small. Some
designs use an inert gas between the two outer protective walls.

9.3 Activated wall material

For verification of fusion reactor materials, an international fusion reactor material test-
ing facility IFMIF-DONES (International Fusion Material Irradiation Facility - Demo
Oriented Neutron Source) is planned to be constructed in Spain.
Stainless steel 316 SS is a well known material with good metallurgic properties.
Unfortunately, it is unsuitable in a fusion reactor because of its content of Fe, Ni and
Cr which all cause long lived isotopes. Materials like SiC, V and Ti are much preferred.
The first generation fusion reactors will have their walls radiologically activated by the
neutron radiation. The volume of this material corresponds to that from fission reactors,
but the radiotoxity is much lower. The possibility to chose materials will ensure that
fusion waste after 10-100 years will be four or five orders of magnitude less radioactive
than the corresponding fission waste, or about the same as that of coal power plant ash,
which contains small amounts of uranium and thorium.
If we consider the entire fusion device, approximately 35–45 % of the mass is safe to
handle immediately, 50–65 % can be recycled (manually or using robots, see figure 9.4)
and a few percent requires special storage. The latter is around 65 000 tonnes for the
entire life span of the reactor.
Let us now discuss the selection of materials. It is most important to find alloys or
ceramic materials with low activation which do not swell or become brittle, that can
withstand high working temperatures, and are compatible with coolants, hydrogen gas
and powerful magnetic fields. In EU and Japan, research is focused on the following
materials.
Martensitic steel . This material resists swelling caused by radiation, has low thermal
expansion and high thermal conductivity. Radiation tests show that the material retains
its properties at least up to 160 dpa. Ferritic/martensitic (F/M) steel is thus a potential
material for the reactor’s inner components. F/M steel swells less than austenitic steel

190
Figure 9.4: Remote handling robot inside JET’s vacuum vessel. (Figure: EUROfusion-JET)

caused by helium formation during radiation. Disadvantages are that the material can
become brittle at high working temperatures, 550–6500 C is the upper limit, and that it
affects the coil’s magnetic fields. The best alloy at the moment is 9CrWTaV, which also
is resistant to corrosion. From a radiological point of view, F/M steel is not an optimum
material, as mentioned earlier.
Vanadium alloys. These (e.g. V4Ti4Cr) fulfil most of the requirements above, but are
sensitive to reactions with oxygen. Thus, only liquid lithium can be used as coolant, and
this strongly limits the interest for these materials.
Titanium and chromium alloys are mostly studied for their low neutron activation.
Their brittlement limits the possibilities, but models using powdered metal are promising.
Silicon carbide.This composite material (SiC/SiC) is definitely the most promising
when it comes to compromises between heat resistivity (withstands up to 11000 C), dura-
bility and waste. The research is focused on making silicon carbide a better thermal and
electrical conductor, less sensitive to radiation and cheaper.
The case for all these materials is that the testing is difficult since powerful enough
radiation sources are lacking.

191
Appendix A

Useful formulas and constants

A.1 Physics constants

Constant Symbol Value Unit

Atomic mass unit u 1.6605655 × 10−27 kg
Proton mass mp 1.007276 u
Neutron mass mn 1.008665 u
Deuterium nucleus mass md 2.013553 u
Tritium nucleus mass mt 3.015501 u
Helium-3 nucleus mass 3.014933 u
Helium-4 nucleus mass mα 4.001503 u
Lithium-6 nucleus mass 6.013470 u
Lithium-7 nucleus mass 7.014354 u
Electron mass me 9.1096 × 10−31 kg
Elementary charge e 1.602177 × 10−19 As
Velocity of light c 2.9979 × 108 m/s
Planck’s constant h 6.6261 × 10−34 Js
Boltzmann’s constant k 1.380649 × 10−23 J/K
Permeability of free space µ0 4π × 10−7 Vs/Am
Permittivity of free space ε0 8.854 × 10−12 As/Vm

A.2 Energy and temperature units

Energy in the SI-system is measured in Joule (J). However, in this book we follow the
convention in plasma physics literature and use the unit electron volt (eV) for energy.
The unit can be understood as follows: A charge q that moves across a voltage U gains
an energy of W = qU . An energy of 1 eV is the energy that is gained by an electron when
it moves across a voltage of 1 V. We can imagine that the electron is placed between two
parallel plates, starts from the negative plate and is accelerated to the positive plate at
one volt higher potential. The energy gained by the electron is

W = eU = 1.602177 × 10−19 [As] × 1 [V] = 1.602177 × 10−19 [J]

192
The conversion from the energy unit eV to Joule is

1 eV = 1.602177 × 10−19 J

Temperature in the SI-system is measured in Kelvin (K). In the field of plasma physics
we refer to temperatures by their equivalent energies in eV. In one species of a plasma at
a temperature T , for example ions, the ion energy density is (3/2)nkB T , where n is the
ion particle density, and kB is Boltzmann’s constant. We associate an average energy of
(1/2)kB T to each of the three degrees of freedom for a plasma particle. This equivalence
of temperature and energy is used in plasma physics to define the temperature in electron
volts. A temperature in electron volts of TeV = 1 eV is equal to the temperature in Kelvin
(TK ), for which the energy kB TK is 1 eV. The conversion factor is thus obtained from the
relation
eTeV = kB TK
The conversion from the temperature unit eV to Kelvin is

1 eV = (e/kB ) = 1.602177 × 10−19 /1.380649 × 10−23 = 11605 K

In this book we use the symbol T referring to TeV . We have also used the convention to
write out the electron charge e explicitly in the expression eT , which means that T has
the SI-unit Volt (V).

A.3 Thermal velocity

The thermal velocity is a typical velocity of the thermal motion of the particles in the
plasma. Thus, the thermal velocity is a measure of the temperature. Thermal velocity
is not a vector, rather a scalar denoting the speed of the particles. There are different
conventions used in the literature. In this book, we take the most probable speed of the
Maxwellian velocity distribution, in any one direction, and the thermal velocity is thus
defined as r
2eT
vT = (A.1)
m

A.4 Vector formulas

Stoke’s theorem I Z
a · dl = (∇ × a) · n̂
n dS
C S
Gauss’ theorem I Z
a · n̂
n dS = ∇ · a dV
S V
Various vector formulas

a · (b
b × c) = b · (cc × a) = c · (a
a × b)
a × (bb × c) = (a
a · c)bb − (a
a · b)cc
∇ × ∇Ψ = 0
∇ × (∇ × a) = ∇(∇ · a) − ∇2 a
∇ · (aa × b) = b · (∇ × a) − a · (∇ × b)
∇ × (a a × b) = a(∇ · b) − b(∇ · a) + (b b · ∇)a
a − (a
a · ∇)b
b

193
 Figure A.1: Cylindrical coordinate system

A.5 Cylindrical coordinates

Unit vectors in cartesian coordinates

r̂r = (cos θ, sin θ, 0)

θ̂θ = (− sin θ, cos θ, 0)
ẑz = (0, 0, 1)

Line element
dl = drr̂r + rdθθ̂θ + dzẑz
Volume element
dV = rdrdθdz
Gradient, divergence, curl
∂Ψ 1 ∂Ψ ∂Ψ
∇Ψ = r̂r + θ̂θ + ẑz
∂r r ∂θ ∂z
1 ∂ 1 ∂aθ ∂az
∇·a = (rar ) + +
r ∂r r ∂θ ∂z
1 ∂az ∂aθ
∂ar ∂az
1 ∂ 1 ∂ar

∇×a = − r̂r + − θ̂θ + (raθ ) − ẑz
r ∂θ ∂z ∂z ∂r r ∂r r ∂θ

A.6 Electromagnetics
Low frequency approximation of Maxwell’s equations in differential form

∇·B=0
1
∇·E= ρ
ε0
∂B
∇×E=−
∂t
∇ × B = µ0 j

Ampere’s law in integral form is useful for calculating the magnetic field from a current
where there is for example cylinder symmetry. The following form is obtained by applying
Stoke’s theorem I Z
B · dl = µ0 j · n̂
ndS = µ0 Iencl
C S

194
where Iencl is the current enclosed by the curve C. One example is the calculation of the
azimuthal magnetic field from the axial plasma current in the case of circular cylinder
geometry, which then is direcly obtained from Ampere’s law as Bθ = µ0 Iz /(2πr). When
modelling high frequency plasma waves, and electromagnetic waves in vacuum, the exact
form of Ampere’s law is used
1 ∂E
∇ × B = µ0 j +
c2 ∂t

195
Index

absorption, 118 banana orbit, 84

absorption coefficient, 134 banana orbit diffusion, 86
absorption foil, 137 banana-orbit half-width, 85
absorption process, 122 beam dump, 133
accelerator, 117 Bennett, 45
acidification, 4 Bennett pinch, 46
acousto-optical modulator, 129 Bennett relation, 46
activated wall material, 157, 158 beta limit, 67
active feedback, 147 beta value, 31, 50
adsorbed atoms, 107 Bethe, 1
advanced reaction, 22, 27 binding energy, 19
advanced tokamak, 147 bioenergy, 11
air inlet, 157, 158 biomass, 9
Alcator, 141 black body, 134
Alfvén, 57 blanket, 21, 154, 155
Alfvén wave, 57 Boltzmann’s relation, 91
alpha-particle, 21, 26 bounce frequency, 84
alpha-particle heating, 26 bound electron, 100, 126
alternative approach, 141 boundary layer, 105, 111
alternative configuration, 144 Bragg cell, 129
Alvarez, 28 break even, 23, 150
ambipolar, 75 breeder reactor, 152
Ampéreös law, 45 bremsstrahlung, 25, 100
anomalous transport, 71, 89 bremsstrahlung power, 102
antenna, 118 burning plasma, 27
aperture, 133
aquifer, 6 carbon dioxide, 5, 7
arc discharge, 109 carbon dioxide laser, 129
area need, 11 carbon dioxide, increase of, 5
area requirement, 12 carbon dioxide, level of, 5
ARIES-I, 156 cathode spot, 110
aspect ratio, 64 centre of gyration, 36
atmosphere average temperature, 5 characteristic length, 92
auto correlation length, 98 charge exchange collision, 107, 137
average charge number, 103 charge state, 104
average distance, 38 chemical combustion reaction, 15
axially symmetric mode, 66 chromium alloy, 159
circular polarized wave, 120
background radiation, 133 circulating flux, 106
backscattering coefficient, 106 classical electron radius, 129
baking, 108 classical transport, 71
ballooning mode, 66 climatological effect, 5

196
coal, 4 drift wave instability, 91
cold fusion, 21, 29 dynamo, 145
collision cross section, 115 dynamo mode, 146
collision frequency, 23, 86
collision time, 39, 74 earth, 21
compact tokamak, 148 effective charge, 33
conductivity tensor, 119 efficiency, 25
confinement, 68, 70 eigenvalue problem, 54
continuity equation, 41 Einstein, 19
continuous spectrum, 137 electricity, 4
conventional RFP, 145 electromagnetic fluctuation, 89
conventional tokamak, 144 electromagnetic radiation, 100, 125
cooling water, 25 electron cyclotron radiation, 103, 134
corona equilibrium, 104 electron cyclotron resonance, 122
Coulomb collision, 73 electron cyclotron resonance heating, 124
Coulomb force, 19 electron temperature, 130
Coulomb logarithm, 32, 39 electrostatic deflection, 140
critical disturbance, 159 electrostatic fluctuation, 89
cross section, 21 electrostatic turbulence, 90
curvature drift, 37 elongation, 66
cut-off, 121 embrittlement, 156
cyclotron damping, 123 emission, 100
cyclotron frequency, 35 emittance, 134
cyclotron radiation, 102 endurance failure, 156
energy backscattering coefficient, 106
Debye length, 37 energy confinement time, 25, 31, 33
deceleration time, 39 energy consumption, 4
deexcitation, 104 energy density, 21
deflection angle, 74 energy equation, 42
deflection magnet, 117 energy exchange time, 40
DEMO, 143 energy principle, 54, 58
density fluctuation, 90 energy situation, 3, 4
desorption, 107 energy sources, 13, 17
deuterium, 3, 19, 21 energy spectrum, 136
diagnostic method, 126 environment, 5
diamagnetic drift velocity, 92 equation of motion, 42
dielectric tensor, 119 equilibrium, 45
diffusion coefficient, 73 equilibrium equation, 45
direct transfer, 23 evaporation, 110
discharge, 45 excitation, 104
displacement, 52 excitation balance, 104
disruption, 66, 157 exotic fusion reactions, 27
distribution function, 38 exotic reaction, 22, 27
disturbance of cooling, 157, 160 external heating, 33
disturbance of superconducting coils, 157 external kink, 61
divertor, 154, 155 external mode, 62
Doppler effect, 130
fast magnetoacoustic wave, 57
drift, 36
fast neutral particle, 139
drift wave, 91
Fick’s law, 73

197
field emission, 109 hydrogen, 7, 14
first wall, 154, 155 hydropower, 11
fission, 16
fission energy, 19 ideal instability, 54
fission reactor, 4, 15 ideal MHD, 43
flood, 5 ignition, 26
fluid model, 34, 41 impurity ions, 103
fluid velocity, 42 incidence parameter, 74
flux function, 47 incompressibility, 44
flux surface, 47 inertial confinement, 20, 142
Fokker-Planck equation, 39 inertial fusion, 150
Forsmark, 16 initial value problem, 55
fossil fuel, 3, 4 instability, 54
fossil fuel resources, 4 interferometer, 127
fossil gas, 5, 6 intermediate level waste, 15
FPC, 148 internal kink, 63
free electron, 100, 125 internal mode, 61
fuel cell, 14 ion cyclotron resonance, 122
fusion, 19 ion cyclotron resonance heating, 124
fusion cross section, 23 ion process, 126
fusion energy, 18 ion source, 117
fusion power, 18, 50 ion temperature, 139
fusion product, 18 ionization, 104, 107
fusion reaction, 19, 21 ionization coefficient, 104
fusion reactivity, 24 island width, 94
fusion reactor, 4, 19, 154 ITER, 27, 142
fusion-fission, 151 kinetic model, 38
gamma radiation, 21 kinetic theory, 34
Gamow, 24 kink, 61
gas, 4 kink instability, 60
gas target, 112 Kirchoff’s law, 134
geothermal energy, 9
L wave, 120
geothermal power, 12
Landau damping, 122
glow discharge cleaning, 108
Larmor radius, 35
Grad-Shafranov equation, 48, 49
Lawson, 23, 24
gravitational force, 36
Lawson criterion, 24, 31
greenhouse effect, 5
Lawson diagram, 26
Greenwald, 67
left circular polarized wave, 120
growth rate, 59
LHD, 149
guiding-centre drift, 78
limiter, 110, 154, 155
guiding-centre orbit, 78
line density, 33, 129
heat conductivity, 77 line radiation, 103
heat diffusivity, 77 linear stability, 54
heating, 26 linearize, 52
high energy neutral particle, 138 LMJ, 27, 150
high level waste, 15 low level waste, 15
hybrid reactor, 4, 142, 151 lower hybrid resonance, 122
hydrodynamic equation, 42 lower hybrid resonance heating, 124

198
magnetic coils, 154, 157 Ohmös law, 43
magnetic confinement, 20 ohmic heating, 113
magnetic deflection, 140 ohmicly heated plasma, 33
magnetic divertor, 112 Ohms law, 40
magnetic field surface, 47 oil, 4
magnetic island, 93 one-fluid velocity, 43
magnetic mirror field, 82 optical depth, 134
magnetic mirror force, 82 optically thick plasma, 135
magnetic moment, 82 optically thin plasma, 135
major radius, 32 oscillatory motion, 84
marginal stability, 55 overlapping, 97
martensitic steel, 158
mass density, 43 particle, 34
MAST, 148 particle confinement time, 76
Maxwell distribution, 23, 24 particle drift, 36
Maxwellös equations, 43 particle dynamics, 35
mean free path, 23 particle flux, 125
Mercier, 65 Pfirsch-Schlüter transport, 81
MHD wave, 57 phase shift, 127
MHD waves, 56 pinch, 2
mixing length approximation, 93 Planck’s radiation law, 134
mode conversion, 118 plasma, 23
modulation, 129 plasma current, 32
multichannel pulse height analyzer, 137 plasma diagnostics, 125
muon-catalyzed fusion, 21, 28 plasma dispersion relation, 118
plasma electron density, 127
NbSn super conductor, 157 plasma frequency, 56
NbTi super conductor, 157 plasma model, 34
neoclassical diffusion coefficient, 87 plasma oscillation, 56
neoclassical transport, 71, 78 plasma parameter, 34
neptunium, 15 plasma wall interaction, 105
neutral beam heating, 114 plateau regime, 87
neutral beam injection, 114, 117 plutonium, 15
neutral fraction, 117 poloidal β, 50
neutral particle, 138 poloidal beta, 32
neutral particle analysis, 137 poloidal beta value, 31
neutral particle flux, 139 poloidal direction, 31
neutralization chamber, 117 pressure driven instability, 65
neutralize, 117 pressure gradient, 65
neutron, 21 pressure tensor, 42
neutron flux, 156
neutron free reaction, 22 quadrature detection, 129
NIF, 27, 150 quality parameters, 31
NSTX, 148 quantum mechanics, 24
nuclear accidents, 4
R wave, 120
nuclear energy, 15
radiation parameter, 104
nuclear power, 18
radiative power, 132
O-wave, 120 radio-frequency heating, 117
ocean wave, 11 radio-frequency waves, 117

199
radioactivity, 19 scattering volume, 133
random walk model, 72 separatrix, 94, 155
rate coefficient, 107 shear, 65
rational surfaces, 54 shield, 154, 157
Rayleigh-Taylor instability, 60, 151 silicon carbide, 159
reaction kinematics, 23 silicon detector, 137
reaction rate, 23 slow magnetoacoustic wave, 57
reaction velocity, 21 soft X-ray radiation, 137
reactivity, 23 solar cells, 10
recombination, 104 solar energy, 9
recombination coefficient, 104 solar power, 9
recycling, 106 solenoid coil, 126
recycling coefficient, 106 Solov’ev, 49
reduced ion energy, 106 spallation, 16
reference beam, 127 spectral function, 132
reflection, 121 spectral function half-maximum width, 133
refractive index, 119, 125 spectroscopic measurement, 126
relaxation time, 39 spherical tokamak, 147
remote collision, 74 Spitzer, 1
renewable energy, 9, 12 sputtering, 108
renewable energy source, 4, 12 sputtering yield, 108
resistive g-mode, 64 stability, 58
resistive instability, 54, 64 stability diagram, 64
resistive MHD, 75 stabilizing effect, 54
resistive mode, 65 stabilizing magnetic fields, 51
resistive shell mode, 147 stabilizing term, 60
resistivity, 40 stainless steel 316 SS, 156
resonance, 52, 121 star, 19
resonance layer, 135 START, 147
resonance surface, 121 static equilibrium, 45
resonant surface, 93 stationary state, 45
rest gas, 103 stellarator, 1, 2, 37, 142, 148
reversed toroidal field, 145 step length, 73
reversed-field pinch, 50, 142, 144 stochastic magnetic field, 97
RFP, 50, 144 storage of carbon dioxide, 5
Riggatron, 148 strong interaction, 25
right circular polarized, 120 sun, 1, 3, 21
rippling mode, 64 Suydam, 65
Rogowski coil, 126 swelling, 156
rotation transform, 53 Szilard, 1
Rutherford’s scattering law, 74
target chamber, 112
safety factor, 52 tearing mode, 64, 67
safety issue, 157 temperature, I
sausage instability, 60 test function, 59
sawtooth, 53, 63 thermal emission, 109
scattered wave, 126 thermalized plasma, 23
scattering, 126 thermodynamic equilibrium, 51
scattering time, 39 thermonuclear fusion, 1, 20, 31

200
Thomson scattering, 130
titanium alloy, 159
tokamak, 2, 32
tokamak stability, 61
toroidal beta value, 50
toroidal direction, 31
transmission, 126
transmutation, 15
trapped particle, 82
trapping, 114
triple product, 141
tritium, 19, 21, 157
tritium handling, 157
tritium., 3
Troyon factor, 68
turning point, 85

unipolar arc, 109

upper density limit, 67
upper hybrid resonance, 122

vanadium alloys, 159

velocity fluctuation, 90
Vlasov equation, 38

wave energy, 11
wave power plant, 11
wave propagation, 118
wave source, 129
wave-particle interaction, 118
wave-particle resonance, 122
wavelength shift, 130
Wendelstein, 149
wind power, 10
wind power plant, 10

X-ray photon, 136

X-ray radiation, 25
X-wave, 120

z-pinch, 46, 50

201