INTERNATIONAL ATOMIC ENERGY AGENCY

VIENNA
ISBN 9789201304100
!
@
Edited by:
Mitsuru Kikuchi
Karl Lackner
Minh Quang Tran
E
d
i
t
e
d

b
y
:
M
i
t
s
u
r
u

K
i
k
u
c
h
i
K
a
r
l

L
a
c
k
n
e
r
M
i
n
h

Q
u
a
n
g

T
r
a
n
This publication is a comprehensive reference book for graduate
students and an invaluable guide for more experienced researchers.
It provides an introduction to nuclear fusion and its status and
prospects, and features specialized chapters written by leaders in
the feld, presenting the main research and development concepts
in fusion physics. It starts with an introduction to the case for
the development of fusion as an energy source. Magnetic and
inertial confnement are addressed. Dedicated chapters focus
on the physics of confnement, the equilibrium and stability of
tokamaks, diagnostics, heating and current drive by neutral beam
and radiofrequency waves, and plasmawall interactions. While
the tokamak is a leading concept for the realization of fusion,
other concepts (helical confnement and, in a broader sense, other
magnetic and inertial confgurations) are also addressed in the
book. At over 1100 pages, this publication provides an unparalleled
resource for fusion physicists and engineers.
FUSION PHYSICS
The following States are Members of the International Atomic Energy Agency:
The Agencys Statute was approved on 23 October 1956 by the Conference on the Statute of the
IAEA held at United Nations Headquarters, New York; it entered into force on 29 July 1957. The
Headquarters of the Agency are situated in Vienna. Its principal objective is to accelerate and enlarge the
contribution of atomic energy to peace, health and prosperity throughout the world.
AFGHANISTAN
ALBANIA
ALGERIA
ANGOLA
ARGENTINA
ARMENIA
AUSTRALIA
AUSTRIA
AZERBAIJAN
BAHRAIN
BANGLADESH
BELARUS
BELGIUM
BELIZE
BENIN
BOLIVIA
BOSNIA AND HERZEGOVINA
BOTSWANA
BRAZIL
BULGARIA
BURKINA FASO
BURUNDI
CAMBODIA
CAMEROON
CANADA
CENTRAL AFRICAN
REPUBLIC
CHAD
CHILE
CHINA
COLOMBIA
CONGO
COSTA RICA
CTE DIVOIRE
CROATIA
CUBA
CYPRUS
CZECH REPUBLIC
DEMOCRATIC REPUBLIC
OF THE CONGO
DENMARK
DOMINICA
DOMINICAN REPUBLIC
ECUADOR
EGYPT
EL SALVADOR
ERITREA
ESTONIA
ETHIOPIA
FINLAND
FRANCE
GABON
GEORGIA
GERMANY
GHANA
GREECE
GUATEMALA
HAITI
HOLY SEE
HONDURAS
HUNGARY
ICELAND
INDIA
INDONESIA
IRAN, ISLAMIC REPUBLIC OF
IRAQ
IRELAND
ISRAEL
ITALY
JAMAICA
JAPAN
JORDAN
KAZAKHSTAN
KENYA
KOREA, REPUBLIC OF
KUWAIT
KYRGYZSTAN
LAO PEOPLES DEMOCRATIC
REPUBLIC
LATVIA
LEBANON
LESOTHO
LIBERIA
LIBYA
LIECHTENSTEIN
LITHUANIA
LUXEMBOURG
MADAGASCAR
MALAWI
MALAYSIA
MALI
MALTA
MARSHALL ISLANDS
MAURITANIA
MAURITIUS
MEXICO
MONACO
MONGOLIA
MONTENEGRO
MOROCCO
MOZAMBIQUE
MYANMAR
NAMIBIA
NEPAL
NETHERLANDS
NEW ZEALAND
NICARAGUA
NIGER
NIGERIA
NORWAY
OMAN
PAKISTAN
PALAU
PANAMA
PAPUA NEW GUINEA
PARAGUAY
PERU
PHILIPPINES
POLAND
PORTUGAL
QATAR
REPUBLIC OF MOLDOVA
ROMANIA
RUSSIAN FEDERATION
RWANDA
SAUDI ARABIA
SENEGAL
SERBIA
SEYCHELLES
SIERRA LEONE
SINGAPORE
SLOVAKIA
SLOVENIA
SOUTH AFRICA
SPAIN
SRI LANKA
SUDAN
SWEDEN
SWITZERLAND
SYRIAN ARAB REPUBLIC
TAJIKISTAN
THAILAND
THE FORMER YUGOSLAV
REPUBLIC OF MACEDONIA
TUNISIA
TURKEY
UGANDA
UKRAINE
UNITED ARAB EMIRATES
UNITED KINGDOM OF
GREAT BRITAIN AND
NORTHERN IRELAND
UNITED REPUBLIC
OF TANZANIA
UNITED STATES OF AMERICA
URUGUAY
UZBEKISTAN
VENEZUELA
VIETNAM
YEMEN
ZAMBIA
ZIMBABWE
FUSION PHYSICS
EDITED BY:
MITSURU KIKUCHI
KARL LACKNER
MINH QUANG TRAN
INTERNATIONAL ATOMIC ENERGY AGENCY
VIENNA, 2012
IAEA, 2012
Printed by the IAEA in Austria
September 2012
STI/PUB/1562
IAEA Library Cataloguing in Publication Data
Fusion physics. Vienna : International Atomic Energy Agency, 2012.
p. ; 24 cm.
STI/PUB/1562
ISBN 9789201304100
Includes bibliographical references.
1. Nuclear fusion. 2. International Thermonuclear Experimental Reactor
(Project). 3. Controlled fusion International cooperation. 4. Tokamaks
International cooperation. I. International Atomic Energy Agency.
IAEAL 1200774
COPYRIGHT NOTICE
All IAEA scientific and technical publications are protected by the terms of
the Universal Copyright Convention as adopted in 1952 (Berne) and as revised
in 1972 (Paris). The copyright has since been extended by the World Intellectual
Property Organization (Geneva) to include electronic and virtual intellectual
property. Permission to use whole or parts of texts contained in IAEA publications
in printed or electronic form must be obtained and is usually subject to royalty
agreements. Proposals for non-commercial reproductions and translations are
welcomed and considered on a case-by-case basis. Enquiries should be addressed
to the IAEA Publishing Section at:
Marketing and Sales Unit, Publishing Section
International Atomic Energy Agency
Vienna International Centre
PO Box 100
1400 Vienna, Austria
fax: +43 1 2600 29302
tel.: +43 1 2600 22417
email: [email protected]
http://www.iaea.org/books
FOREWORD
Recreating the energy production process of the Sun nuclear fusion
on Earth in a controlled fashion is one of the greatest challenges of this century. If
achieved at affordable costs, energy supply security would be greatly enhanced and
environmental degradation from fossil fuels greatly diminished. Fusion Physics
describes the last fifty years or so of physics and research in innovative technologies
to achieve controlled thermonuclear fusion for energy production.
The International Atomic Energy Agency (IAEA) has been involved
since its establishment in 1957 in fusion research. It has been the driving force
behind the biennial conferences on Plasma Physics and Controlled Thermonuclear
Fusion, today known as the Fusion Energy Conference. Hosted by several Member
States, this biennial conference provides a global forum for exchange of the latest
achievements in fusion research against the backdrop of the requirements for a net
energy producing fusion device and, eventually, a fusion power plant. The scientific
and technological knowledge compiled during this series of conferences, as well as
by the IAEA Nuclear Fusion journal, is immense and will surely continue to grow
in the future. It has led to the establishment of the International Thermonuclear
Experimental Reactor (ITER), which represents the biggest experiment in energy
production ever envisaged by humankind.
The IAEA also would like to thank the editors of the book, M. Kikuchi,
K. Lackner and Minh Quang Tran, for preparing this comprehensive manuscript
on fusion, including magnetic and inertial fusion concepts. They have selected a
prominent group of contributors, many of whom have provided seminal scientific
contributions to important developments in the field. The IAEA also conveys its
gratitude to the authors for their long standing cooperation. Their work is highly
appreciated, and this present compendium will help to raise awareness of the
opportunities offered by fusion and the path towards a demonstration fusion power
plant.
PREFACE
In 1958, during the second Conference on Peaceful Uses of Nuclear
Energy in Geneva, nuclear fusion research was declassified. At this time the
basis of nuclear fusion science and technology was confined to a few books and
monographs written by pioneers.
After this event, the tradition to periodically exchange the latest discoveries
in fusion research development was established by the International Atomic
Energy Agency (IAEA) through its series of Fusion Energy Conferences (IAEA
FECs). It was natural that in 2008 the 22nd IAEA FEC came back to the same
location in Geneva, the Palais des Nations, to celebrate the fiftieth anniversary
of the declassification.
The progress over the past half century has been immense. We are now
in the building phase of the International Thermonuclear Experimental Reactor
(ITER) with the prospect of having 500 MW of fusion power in the second half of
the 2020s and starting studies for the step beyond ITER, a demonstration reactor
(usually referred to as the DEMO power plant). A new generation of scientists
and engineers is needed to build and exploit ITER and accompanying fusion
devices, and to prepare the next step beyond ITER. Master and PhD programmes
have been set up in many universities worldwide to train what is usually referred to
as the ITER Generation of Scientists. Compared to 1958, the growth of the field
of nuclear fusion has led to a multiplicity of specialized subfields, each having its
own textbooks.
The occasion of the 2008 IAEA FEC prompted us to propose to the IAEA
International Fusion Research Council (IFRC) to sponsor a tutorial book for
post-graduate students. Our aim is to provide an introduction to nuclear fusion,
its status and perspectives. Specialized chapters are devoted to the main concepts
under R&D (magnetic and inertial conferment) together with the physics as
well as the technology basis. With the strong support and under the guidance of
the IFRC, we have invited international experts to contribute to the project. Our
vision of the book was shared by all contacted colleagues, who enthusiastically
accepted this difficult tutorial task.
It is our hope that the material presented will allow post-graduate level
students to become familiar with the topics of their studies. More advanced
researchers will also find materials on topics adjacent to their field of
specialization. The progress in nuclear fusion research is such that it has become
impossible to cover in detail all the key issues: this book is not intended to
replace specialized monographs or review articles.
The book starts with an introduction to the case for the development of
fusion as an energy source, followed by chapters on the physics of confinement,
equilibrium and stability of tokamaks. Diagnostics, heating and current drive
by neutral beams and radiofrequency waves, and plasmawall interactions
are described in detail. While the tokamak is currently the leading concept
for the realization of nuclear fusion, it is important to note that other concepts
(helical confinement concepts and, in a broader sense, other magnetic configu-
rations) have also received wide interest worldwide. Last but not least, inertial
confinement fusion is one of the important lines of research, which naturally finds
its place in the book. The later part of the book is oriented towards ITER and
fusion technology.
The realization of this book would not have been possible without the
enthusiastic commitment of all authors, who took upon themselves the task
of sharing their vast knowledge with the ITER generation in parallel with
their research duties. We would like to wholeheartedly thank them for their
dedication. Our responsibility has also included careful reading of the contributed
manuscripts. That was done with the help of a few colleagues, whose contribution
is gratefully acknowledged.
Last but not least, our appreciation also goes to the IAEA and its staff,
which provided an unfailing support and encouragement. We would like to
particularly thank G. Mank, R. Kamendje, R. Kaiser and T. Desai, whose support
throughout this endeavour has rendered the publication of this volume possible.
We also would like to acknowledge the very useful contribution of B. Gulejova,
whose professional expertise has helped solve a multitude of editorial issues.
Mitsuru Kikuchi Karl Lackner Minh Quang Tran
December 2011
EDITORIAL NOTE
Although great care has been taken to maintain the accuracy of information contained
in this publication, neither the IAEA nor its Member States assume any responsibility for
consequences which may arise from its use.
The mention of names of specific companies or products (whether or not indicated
as registered) does not imply any intention to infringe proprietary rights, nor should it be
construed as an endorsement or recommendation on the part of the IAEA.
The authors are responsible for having obtained the necessary permission for the IAEA
to reproduce, translate or use material from sources already protected by copyrights.
Material prepared by authors who are in contractual relation with governments is
copyrighted by the IAEA, as publisher, only to the extent permitted by the appropriate national
regulations.
This publication has been prepared from the original material as submitted by the
authors. The views expressed do not necessarily reflect those of the IAEA, the governments of
the nominating Member States or the nominating organizations.
The IAEA has no responsibility for the persistence or accuracy of URLs for external or
third party Internet web sites referred to in this book and does not guarantee that any content
on such web sites is, or will remain, accurate or appropriate.
CONTENTS
CHAPTER 1
THE CASE FOR FUSION
1.1. INTRODUCTION.......................................................................................................... 1
1.2. ENERGYSCENARIOS.............................................................................................. 3
1.2.1. Neartermenergyscenario............................................................................... 5
1.2.2. Longtermenergyscenarioandtheroleoffusion..................................... 7
1.3. FUSIONBASICS........................................................................................................... 14
1.3.1. Whatisfusion?.................................................................................................. 14
1.3.2. FusionpowergainQ........................................................................................ 17
1.3.3. Fusionreactions................................................................................................. 20
1.3.4. Fusionfuels........................................................................................................ 23
1.3.5. Directconversiontoelectricity....................................................................... 26
1.4. APPROACHESTOFUSION.................................................................................... 26
1.4.1. Magneticconfnementfusion.......................................................................... 27
1.4.1.1. Progressintokamakbasedmagneticconfnementfusion
research............................................................................................ 32
1.4.2. Inertialconfnementfusion............................................................................. 35
1.4.2.1. Progressininertialconfnementfusionresearch..................... 38
1.5. SOCIOECONOMICPERSPECTIVES................................................................ 40
1.5.1. Environment,safetyandnon-proliferation.................................................. 40
1.5.1.1. Emissionsinnormaloperation................................................... 40
1.5.1.2. Possibleaccidents.......................................................................... 41
1.5.1.3. Waste................................................................................................ 41
1.5.2. Costcomparisonwithothersourcesofenergy............................................ 42
1.5.2.1. Directcostsoffusionpowerproduction................................... 42
1.5.2.2. Externalcostsoffusionpowerproduction............................... 45
1.5.3. Publicacceptanceoffusion............................................................................ 46
1.5.4. Spin-offsoffusionresearch............................................................................ 47
1.6. CONCLUSION............................................................................................................... 51
CHAPTER 2
PHYSICS OF CONFINEMENT
2.1. INTRODUCTIONANDOVERVIEW................................................................. 59
2.2. NEOCLASSICALTRANSPORT............................................................................ 63
2.2.1. Introduction........................................................................................................ 63
2.2.2. Toroidalmagneticconfguration.................................................................... 65
2.2.3. Guidingcentremotion...................................................................................... 66
2.2.4. Thedriftkineticequation............................................................................... 71
2.2.5. Bootstrapcurrent.............................................................................................. 75
2.2.6. Magneticfeldripple........................................................................................ 78
2.2.7. Transportinastochasticfeld......................................................................... 82
2.3. TURBULENTTRANSPORT.................................................................................... 83
2.3.1. Introduction........................................................................................................ 83
2.3.2. Examplesofbasicmicroinstabilities............................................................ 86
2.3.2.1. Electrondriftinstabilities........................................................... 86
2.3.2.2. Iontemperaturegradientinstabilities....................................... 89
2.3.3. Non-lineargyrokineticequationsfortokamakturbulence....................... 91
2.3.4. Kineticdescriptionofmicroinstabilities...................................................... 95
2.3.4.1. Linearonsetconditionforiontemperaturegradientmode.. 96
2.3.4.2. Electrontemperaturegradientinstability.................................. 99
2.3.4.3. Trappedparticleinstabilities...................................................... 99
2.3.5. Spatialstructureofmicroturbulence.............................................................. 103
2.3.5.1. Roleofmagneticshear:Shearedslabgeometry.................... 104
2.3.5.2. Roleoftoroidalgeometry:Ballooningrepresentation.......... 106
2.3.5.3. Roleofzonalfowshear............................................................. 108
2.3.6. Differentchannelsofturbulencetransport................................................... 110
2.3.6.1. Ionthermaltransport................................................................... 111
2.3.6.2. Electronthermaltransport.......................................................... 113
2.3.6.3. Particletransport........................................................................... 114
2.3.6.4. Momentumtransport................................................................... 116
2.3.7. Physicsoftransportbarriers............................................................................ 118
2.4. GLOBALENERGYCONFINEMENTSCALINGSTUDIES.................... 123
2.4.1. Introduction........................................................................................................ 123
2.4.2. Energyconfnementscalings:Dimensionalparameters............................ 124
2.4.2.1. OhmicandL-modeplasmaconfnementtrends...................... 124
2.4.2.2. H-modeconfnementtrendsandscalings................................. 128
2.4.2.3. Advancedstatisticalmethodsanderroranalysis..................... 133
2.4.3. Dimensionlessanalysis.................................................................................... 136
2.4.3.1. Basics.............................................................................................. 136
2.4.3.2. Scalings........................................................................................... 139
2.4.4. L-Hthresholdscalings..................................................................................... 140
2.4.5. Implicationsrelativetotheoreticalmodels.................................................. 144
2.5. LOCALTRANSPORT................................................................................................. 145
2.5.1. Basics................................................................................................................... 145
2.5.2. Models................................................................................................................. 147
2.5.2.1. Neutralbeammodels.................................................................... 147
2.5.2.2. Neutraltransport............................................................................ 149
2.5.2.3. RFheatingandcurrentdrivemodels......................................... 150
2.5.2.4. Currentdrivemodels.................................................................... 152
2.5.3. Perturbativetransport....................................................................................... 155
2.5.4. Proflestiffness.................................................................................................. 160
2.5.5. Transportbarriers.............................................................................................. 163
2.5.5.1. Core.................................................................................................. 163
2.5.5.2. Edge................................................................................................. 169
2.5.6. Comparisonwiththeoreticalmodels............................................................. 173
2.6. TuRbulENCEMEasuREMENTs................................................................... 176
2.6.1. Measurementtechniques.................................................................................. 177
2.6.1.1. Probetechniques............................................................................ 177
2.6.1.2. Electromagneticwavescattering................................................ 178
2.6.1.3. Electromagneticwaverefectometry......................................... 179
2.6.1.4. ElectromagneticwaveDopplerbackscattering(Dbs)........... 180
2.6.1.5. Phasecontrastimaging(PCI)...................................................... 181
2.6.1.6. Heavyionbeamprobe(HIbP)................................................... 181
2.6.1.7. beamemissionspectroscopy(bEs).......................................... 182
2.6.1.8. Highfrequencychargeexchangespectroscopy
(HFCHERs).................................................................................. 183
2.6.1.9. Correlatedelectroncyclotronemission(CECE)..................... 183
2.6.1.10. analysistechniques...................................................................... 184
2.6.2. Experimentalresults......................................................................................... 185
2.6.2.1. Wavenumberspectra..................................................................... 185
2.6.2.2. Frequencyspectra.......................................................................... 187
2.6.2.3. Radialprofles................................................................................ 189
2.6.3.3. Ionandelectrontemperaturefuctuations................................. 191
2.6.3.4. Zonalfowsandgeodesicacousticmodes................................ 192
2.6.3.5. Coreturbulencesuppression....................................................... 194
2.6.3.6. Evidencefornon-localeffects.................................................... 195
CHAPTER 3
EQUILIBRIUM AND MACROSCOPIC STABILITY
OF TOKAMAKS
3.1. INTRODuCTION.......................................................................................................... 225
3.1.1. basictokamakconfguration........................................................................... 227
3.1.2. Timescalesoftokamakdynamics.................................................................. 228
3.2. TOKaMaKEQuIlIbRIuM.................................................................................... 230
3.2.1. Thestraightcylinder......................................................................................... 231
3.2.2. Toroidalplasmaequilibrium........................................................................... 232
3.2.2.1. Thelargeaspectratioapproximation........................................ 234
3.2.2.2. FormulationsoftheGradShafranovequation...................... 236
3.2.2.3. Dimensionlessparameterscharacterizing
theplasmaequilibrium................................................................. 239
3.2.2.4. Shapingofplasmacross-sectionand
thetokamakpoloidalfeldsystem.............................................. 242
3.2.2.5. Axisymmetricstabilityoftokamakequilibria......................... 244
3.3. THEORYOFMACROSCOPICINSTABILITIESINTOKAMAKS....... 246
3.3.1. One-fuidMHDstabilitytheory..................................................................... 246
3.3.1.1. IdealMHDstability...................................................................... 246
3.3.1.2. Effectsofdissipationonlinearstabilityofglobalmodes...... 269
3.3.1.3. Finiteamplitudeperturbations.................................................... 281
3.3.2. ExtensionstotheMHDmodel....................................................................... 287
3.3.2.1. Introduction.................................................................................... 287
3.3.2.2. Derivationexample:reduced(lowfrequency)MHD............. 291
3.3.2.3. Derivationexample:two-fuidequations.................................. 296
3.3.2.4. Relationtokineticmodels........................................................... 299
3.3.2.5. Qualitativetwo-fuideffectsonMHDinstabilities................. 300
3.3.3. EnergeticparticlephysicsandkineticMHD.............................................. 302
3.3.3.1. Introduction.................................................................................... 302
3.3.3.2. Models............................................................................................. 303
3.3.3.3. Globalmodesintoroidalgeometry............................................ 306
3.4. ExPERIMENTALOBSERVATIONSOFMHDMODES........................... 316
3.4.1. Internalkinkmodes.......................................................................................... 317
3.4.2. Classicaltearingmodes.................................................................................... 322
3.4.3. Neoclassicaltearingmodes(NTMs)............................................................. 325
3.4.4. Edgelocalizedmodes(ELMs)........................................................................ 327
3.4.5. Idealpressurelimitingmodes......................................................................... 331
3.4.6. WalleffectsonMHDmodes........................................................................... 334
3.4.7. Alfvnandenergeticparticlemodes(EPMs).............................................. 337
3.5. DISRUPTIVEINSTABILITIES............................................................................ 342
3.5.1. Introduction....................................................................................................... 342
3.5.2. Classifcationbycauses.................................................................................. 343
3.5.2.1. Densitylimitdisruptions.............................................................. 343
3.5.2.2. Betalimit......................................................................................... 344
3.5.2.3. Limitonq......................................................................................... 344
3.5.3. Phasesofdisruptions........................................................................................ 345
3.5.3.1. Thermalquench............................................................................. 345
3.5.3.2. Currentquench............................................................................... 347
3.5.4. Damagepotential............................................................................................... 349
3.5.4.1. Heatpulse....................................................................................... 349
3.5.4.2. Electromagneticforces................................................................. 350
3.5.4.3. Runawaygeneration..................................................................... 352
3.5.5. Mitigationmethods........................................................................................... 353
CHAPTER 4
PLASMA DIAGNOSTICS
4.1. INTRODUCTION.......................................................................................................... 360
4.2. PASSIVEDIAGNOSTICMETHODS.................................................................. 362
4.2.1. Magneticmeasurements................................................................................. 362
4.2.2. Electricalprobes(Langmuirprobes)............................................................. 368
4.2.2.1. Introductiontoelectricalprobes................................................. 368
4.2.2.2. Langmuirprobes............................................................................ 368
4.2.3. VisibleandUVspectroscopy.......................................................................... 375
4.2.4. Bolometry........................................................................................................... 378
4.2.4.1. Thermaldevices............................................................................ 378
4.2.4.2. AxUVdetectors............................................................................ 382
4.2.5. Methodsbasedonelectroncyclotronemissionmeasurements................ 384
4.2.6. Passiveneutralparticleanalysis..................................................................... 393
4.2.7. Methodsbasedonxrayradiation................................................................. 400
4.2.7.1. Softxrayradiation...................................................................... 400
4.2.7.2. Highresolutionxrayspectroscopy........................................... 404
4.2.8. Experimentalnuclearphysicsmethods......................................................... 412
4.2.8.1. Neutrons,chargedfusionproducts,hardxrays..................... 412
4.2.8.2. Gammarayspectrometry............................................................. 421
4.3. ACTIVEDIAGNOSTICMETHODS.................................................................... 426
4.3.1. Probingbylaserbeams.................................................................................... 426
4.3.1.1. Thomsonscattering....................................................................... 426
4.3.1.2. Laserinducedfuorescencetechnique....................................... 432
4.3.2. Probingbyparticles.......................................................................................... 435
4.3.2.1. Diagnosticneutralbeamsforplasmastudiesin
magneticfusiondevices............................................................... 435
4.3.2.2. Activechargeexchangerecombinationspectroscopy
diagnostic........................................................................................ 438
4.3.2.3. MotionalStarkeffect................................................................... 443
4.3.2.4. ActivechargeexchangeandRutherfordscattering................ 445
4.3.2.5. Heavyionbeamprobediagnostics............................................ 449
4.3.3. Probingbymicrowavesandlaserbeam:
interferometry,polarimetry,refectometry................................................... 452
4.3.3.1. Introduction.................................................................................... 452
4.3.3.2. Interferometryandpolarimetry................................................... 454
4.3.3.3. Plasmarefectometryforelectronconcentrationprofle
measurements................................................................................. 473
4.3.3.4. Applicationofrefectometrytodensityfuctuation
measurements................................................................................. 482
4.4. ENGINEERINGPROBLEMS.................................................................................. 485
4.4.1. Changingofmechanicalproperties............................................................... 486
4.4.2. Changinginelectricalproperties.................................................................... 486
4.4.3. Changesinopticalproperties.......................................................................... 487
4.5. CONCLUSION............................................................................................................... 489
CHAPTER 5
PLASMA HEATING AND CURRENT DRIVE BY NEUTRAL
BEAM AND LPHA PARTICLES
5.1. HEATINGANDCURRENTDRIVEPHYSICSBYNEUTRAL
BEAMANDALPHAPARTICLES....................................................................... 535
5.1.1. Basicprocessesofneutralbeaminjection.................................................... 535
5.1.2. Physicsofionizationofinjectedneutralbeam........................................... 537
5.1.3. Multi-stepionizationandLorenzionization................................................ 540
5.1.4. Energytransfertoelectronsandionsbyneutralbeaminjection............. 543
5.1.5. Energeticparticleorbitsontheaxisymmetricmagneticsurfaces........... 546
5.1.6. FastionbehaviourandhightemperatureproductionwithNBinjection 549
5.1.7. Physicsofneutralbeamcurrentdrive:fastiondistributionfunction..... 551
5.1.7.1. RayleighRitzmethod.................................................................. 553
5.1.8. Physicsofneutralbeamcurrentdrive:shieldingcurrentand
NBCDeffciency............................................................................................... 554
5.1.9. Experimentalobservationofbeam-drivencurrent..................................... 558
5.1.10. Physicsofripplelossoffastions:bananadriftandrippletrapped
losses.................................................................................................................... 560
5.1.11. Physicsofparticletrajectoriesinnon-axisymmetricfelds....................... 563
5.1.12. Alphaheating..................................................................................................... 565
5.1.13. D-Texperimentsinlargetokamaks(TFTRandJET)................................ 568
5.2. NEUTRALBEAMHEATING.................................................................................. 571
5.2.1. Introduction........................................................................................................ 571
5.2.2. Ionsource............................................................................................................ 575
5.2.2.1. Plasmagenerator........................................................................... 575
5.2.2.2. Multicuspsource........................................................................... 576
5.2.2.3. Radiofrequencyplasmasource................................................... 578
5.2.3. Extractionandacceleration............................................................................. 579
5.2.3.1. ChildLangmuirLaw................................................................... 579
5.2.3.2. Multi-stageacceleration............................................................... 581
5.2.3.3. Beamletsteering............................................................................ 582
5.2.4. Positive-ion-basedNBI.................................................................................... 584
5.2.5. Negativeionsource.......................................................................................... 587
5.2.5.1. Negativeionproduction............................................................... 588
5.2.5.2. Advantagesofvolumeproduction............................................. 590
5.2.5.3. Caesiumseeding............................................................................ 592
5.2.6. Negativeionextraction/acceleration............................................................. 594
5.2.7. Negative-ion-basedNBI.................................................................................. 595
5.2.7.1. N-NBIforJT-60............................................................................ 595
5.2.7.2. N-NBIforLHD............................................................................. 596
5.2.7.3. N-NBIforITER............................................................................. 597
5.2.8. MVpowersupply.............................................................................................. 599
5.2.9. RemarksonfutureNBItechnology............................................................... 600
CHAPTER 6
RADIOFREQUENCY WAVES, HEATING AND CURRENT
DRIVE IN MAGNETICALLY CONFINED PLASMAS
6.1. INTRODUCTION.......................................................................................................... 609
6.2. THEORYOFRFWAVEPROPAGATIONINAMAGNETIZED
PLASMA........................................................................................................................... 611
6.2.1. Electroncyclotronwaves................................................................................. 620
6.2.2. Lowerhybridwavepropagationandaccessibility...................................... 624
6.2.3. Ioncyclotronwavepropagationandaccessibility...................................... 630
6.2.4. ICRFwaveabsorptioninahotplasma......................................................... 633
6.2.4.1. Absorptiononelectrons............................................................... 634
6.2.4.2. AbsorptionofICRFwavesonioncyclotronharmonics....... 635
6.2.4.3. Minorityioncyclotronabsorption............................................. 637
6.2.5. Quasi-linearabsorptionofICRFwavesonions......................................... 638
6.2.6. Quasi-linearabsorptiononelectrons............................................................. 642
6.2.6.1. Lowerhybridcurrentdrive.......................................................... 642
6.2.6.2. Electroncyclotroncurrentdrive................................................. 645
6.2.6.3. CurrentdrivebythefastICRFwaves....................................... 647
6.2.7. Wavepropagationintoroidalgeometry,geometricoptics........................ 648
6.2.7.1. Geometricopticsandtherayequations.................................... 649
6.2.7.2. Modifcationstowaveaccessibilityandabsorption
intoroidalgeometry...................................................................... 651
6.2.7.3. Numericalsimulationsoflowerhybridcurrentdrive
usingcoupledFokkerPlanckandraytracingcalculations.. 652
6.2.8. Wavepropagationintoroidalgeometry,full-wavetreatment................. 653
6.2.8.1. Full-wavesimulationsofminorityheatingintheICRF........ 653
6.2.8.2. Full-wavesimulationsofmodeconversionintheICRF....... 656
6.2.8.3. Full-wavesimulationsoflowerhybridwaves........................ 657
6.3. ExPERIMENTALRESULTSONRFHEATING
ANDCURRENTDRIVE............................................................................................ 658
6.3.1. ICRFheatingandcurrentdriveexperiments............................................... 659
6.3.2. ICRFantennaandtransmissionlinedesign................................................. 666
6.3.2.1. ICRFtransmissionlinearchitecture.......................................... 666
6.3.2.2. ICRFsourcetechnology............................................................... 667
6.3.2.3. ICRFtransmissionlinedesign.................................................... 669
6.3.2.4. ICRFwavelaunchers.................................................................. 671
6.3.3. Lowerhybridheatingandcurrentdriveexperiments................................ 677
6.3.4. Lowerhybridwavelaunchers......................................................................... 682
6.3.4.1. RFcouplingtheory........................................................................ 685
6.3.4.2. Numericalcouplingcodes........................................................... 687
6.3.4.3. EvolutionofLHlaunchersintokamaks................................... 690
6.3.5. Applicationsofelectroncyclotronheatingandcurrentdrive................... 694
6.3.5.1. Wavepropagationandabsorptionexperiments....................... 695
6.3.5.2. Electroncyclotroncurrentdrive(ECCD)experiments.......... 703
6.3.6. Electroncyclotrontransmissionlineandantennadesign.......................... 710
6.4. GYROTRONSFORECRHEATINGANDCURRENTDRIVE............... 714
6.4.1. Introductiontogyrotrons................................................................................. 714
6.4.2. Physicalprinciplesofthegyrotron................................................................ 716
6.4.3. Overviewofgyrotrontheory........................................................................... 717
6.4.4. Engineeringfeaturesofthegyrotron............................................................. 722
6.4.4.1. Electrongunandbeamtunnel..................................................... 722
6.4.4.2. Gyrotroncavity.............................................................................. 723
6.4.4.3. Internalmodeconverter(IMC)................................................... 723
6.4.4.4. Phasecorrectingmirrorsandoutputwindow.......................... 724
6.4.4.5. Depressedcollector....................................................................... 724
6.4.4.6. Auxiliarycomponents.................................................................. 726
6.4.5. Stateoftheartgyrotrons.................................................................................. 726
6.4.6. Prospectsandfuturedirections....................................................................... 730
6.4.6.1. Multi-megawattgyrotrons........................................................... 730
6.4.6.2. Frequencytuneablegyrotrons..................................................... 731
6.4.6.3. Improvementofgyrotroneffciency.......................................... 732
6.4.6.4. GyrotronsforDEMO.................................................................... 732
CHAPTER 7
PLASMAWALL INTERACTIONS
7.1. INTRODUCTION.......................................................................................................... 756
7.2. BASICPHYSICALPROCESSESANDUNDERLYINGTHEORY........ 758
7.2.1. Basicconcepts.................................................................................................... 758
7.2.1.1. Boundarylayer............................................................................... 758
7.2.1.2. Theplasmasheath......................................................................... 760
7.2.1.3. Thescrape-offlayer...................................................................... 762
7.2.2. Recycling............................................................................................................ 764
7.2.3. Atomicandmolecularprocesses.................................................................... 767
7.2.4. Erosionprocesses.............................................................................................. 772
7.2.4.1. Physicalsputtering........................................................................ 772
7.2.4.2. Chemicalsputtering..................................................................... 774
7.2.4.3. Arcing.............................................................................................. 777
7.3. DEVELOPMENTOFPLASMAFACINGMATERIALS............................. 779
7.3.1. Carboncontainingmaterials.......................................................................... 779
7.3.2. Beryllium........................................................................................................... 782
7.3.3. High-Zmaterials.............................................................................................. 784
7.3.4. Mixedmaterials................................................................................................. 784
7.4. PRESENTPROGRESSOFPLASMAWALLINTERACTIONSIN
TOKAMAKS.................................................................................................................. 786
7.4.1. Wallconditioning.............................................................................................. 786
7.4.1.1. Surfacecleaning........................................................................... 787
7.4.1.2. Plasmaassistedcoatingofthinflms......................................... 789
7.4.2. Impuritiesanddusts.......................................................................................... 791
7.4.2.1. Impurities........................................................................................ 791
7.4.2.2. Dusts................................................................................................ 793
7.4.3. Erosionandre-deposition................................................................................ 794
7.4.3.1. Divertorerosionandre-deposition............................................ 794
7.4.3.2. Limitererosionandre-deposition.............................................. 795
7.4.3.3. Mainchamberwallerosionandre-deposition......................... 795
7.4.4. Hydogenisotoperetentionandremoval....................................................... 796
7.5. CONTROLOFPLASMAWALLINTERACTIONS..................................... 798
7.5.1. Divertors............................................................................................................. 798
7.5.1.1. Introduction.................................................................................... 798
7.5.1.2. Divertoroperationregimes.......................................................... 799
7.5.1.3. Effectofdivertorgeometry......................................................... 806
7.5.2. ParticletransportinthedivertorSOL........................................................... 810
7.5.2.1. Plasmafowanddrifteffects....................................................... 810
7.5.2.2. Recentadvancesoncross-feldparticletransport................... 819
7.5.3. Energydeposition.............................................................................................. 823
7.5.3.1. Steadystatepowerloadonthedivertor.................................... 823
7.5.3.2. TransientenergydepositionduringELMs............................... 827
7.5.3.3. Activecontrolofpeakheatfuxes.............................................. 831
7.6. ISSUESOFPLASMAWALLINTERACTIONSFORNExTSTEP
TOKAMAKS................................................................................................................... 831
7.6.1. Powerhandlingandparticleexhaust............................................................. 832
7.6.2. Transientheatloads.......................................................................................... 833
7.6.3. Materialmigration............................................................................................. 835
7.6.4. Controlofin-vesseltritiuminventory........................................................... 835
7.6.5. IntegratedPWIissuesforsteadystateoperation........................................ 836
CHAPTER 8
HELICAL CONFINEMENT CONCEPTS
8.1. INTRODUCTION.......................................................................................................... 847
8.2. HELICALCONFINEMENTCONCEPTS.......................................................... 849
8.2.1. Productionof3-Dtoroidalfuxsurfaces....................................................... 849
8.2.2. Helicalconfnementdevices........................................................................... 852
8.2.2.1. Classicalstellarators..................................................................... 853
8.2.2.2. Torsatrons/Heliotrons................................................................... 855
8.2.2.3. Heliacs............................................................................................. 858
8.2.2.4. Modularstellarators...................................................................... 859
8.3. THEPHYSICSOFHELICALSYSTEMS.......................................................... 861
8.3.1. Heatingandparticlewaveinteraction.......................................................... 861
8.3.1.1. Electroncyclotronheatingandcurrentdrive........................... 861
8.3.1.2. ElectronBernsteinwaveheatinginstellarators...................... 866
8.3.1.3. Ioncyclotronheatinginhelicalsystems................................... 868
8.3.2. Plasmaequilibrium........................................................................................... 870
8.3.2.1. Basicequilibriumproperties....................................................... 870
8.3.2.2. Optimizedstellarators.................................................................. 871
8.3.2.3. Computationof3-Dequilibria.................................................... 873
8.3.2.4. Experimentalequilibriumidentifcation................................... 874
8.3.3. Plasmastability.................................................................................................. 877
8.3.3.1. Introductionandtheoreticalbackground.................................. 877
8.3.3.2. Ballooningmodes.......................................................................... 881
8.3.3.3. Globalpressuredrivenmodes..................................................... 882
8.3.3.4. EnergeticparticledrivenAlfvninstabilities........................... 885
8.3.4. Bounce-averagedparticleorbits.................................................................... 894
8.3.5. Neoclassicaltransportmodels....................................................................... 896
8.3.5.1. Neoclassicalmodelling................................................................ 896
8.3.5.2. Classicalandadvancedstellaratorconfgurations................... 898
8.3.5.3. Radialelectricfeld....................................................................... 899
8.3.5.4. Bootstrapcurrent........................................................................... 900
8.3.5.5. Parallelresistivity.......................................................................... 900
8.3.5.6. Optimizationofneoclassicaltransport...................................... 901
8.3.6. 3-Dfeaturesoftransportandconfnement.................................................. 903
8.3.6.1. Roleofneoclassicaltransport.................................................... 904
8.3.6.2. Characterizationofturbulenttransport.................................... 906
8.3.6.3. Particletransport............................................................................ 909
8.3.6.4. Electrontemperatureevolution................................................... 911
8.3.6.5. Isotopiceffect................................................................................ 912
8.3.6.6. Magneticconfgurationeffectsonconfnement...................... 913
8.3.6.7. Improvedconfnementregimes.................................................. 915
8.3.6.8. Impuritytransport.......................................................................... 918
8.3.7. Boundarylayeranddivertorphysics............................................................. 923
8.3.7.1. Islanddivertorconcepts............................................................... 923
8.3.7.2. TransportfeaturesofhelicalSOLs............................................ 924
8.3.7.3. Detachmentregime....................................................................... 928
8.4. OPERATIONALLIMITS........................................................................................... 930
8.4.1. Densitylimit...................................................................................................... 930
8.4.2. Betalimit............................................................................................................ 933
8.5. STELLARATOROPTIMIZATION........................................................................ 936
8.5.1. Equilibrium......................................................................................................... 936
8.5.2. Particledrift........................................................................................................ 936
8.5.3. Stability............................................................................................................... 938
8.5.4. Integratedoptimizationconcepts................................................................... 938
8.6. HELICALREACTORCONCEPTS....................................................................... 941
8.6.1. Heliotronreactors............................................................................................. 941
8.6.2. Modularstellaratorreactors........................................................................... 942
8.6.2.1. UScompactstellaratorpowerplantstudy(SPPS).................. 943
8.6.2.2. TheHeliasreactor......................................................................... 944
8.6.3. Concludingremarks.......................................................................................... 945
CHAPTER 9
THE BROADER SPECTRUM OF MAGNETIC
CONFIGURATIONS FOR FUSION
9.1. INTRODUCTION.......................................................................................................... 958
9.2. REVERSEDFIELDPINCH:TOROIDALCONFINEMENTAT
WEAKMAGNETICFIELD..................................................................................... 959
9.2.1. Theconfguration:idealMHDequilibriumandstability.......................... 961
9.2.1.1. Currentdriveninstability............................................................. 964
9.2.1.2. Pressuredriveninstability............................................................ 964
9.2.2. Reversalofmagneticfeld:minimumenergystateandreconnection.... 965
9.2.2.1. Minimumenergystate.................................................................. 965
9.2.2.2. StandardmodelofRFPdynamics:non-lineartearing
modesandmagneticreconnection............................................. 967
9.2.3. Confnement:effectsofmagneticstochasticity........................................... 972
9.2.4. Improvedconfnement.................................................................................... 976
9.2.4.1. Eliminatingthefreeenergysource............................................ 976
9.2.4.2. Singlehelicitystates..................................................................... 979
9.2.5. BetalimitsintheRFP..................................................................................... 982
9.2.6. Resistivewallinstabilitiesandtheircontrol................................................ 984
9.2.6.1. CharacteristicsofresistivewallinstabilitiesintheRFP........ 984
9.2.6.2. Feedbacksuppressionofresistivewallinstabilities............... 985
9.2.7. Sustainingtheplasma...................................................................................... 986
9.3. COMPACTTORI:ELIMINATINGTHEHOLEINTHETORUS............ 990
9.3.1. Spheromaks........................................................................................................ 990
9.3.1.1. Magneticsurfaces.......................................................................... 990
9.3.1.2. Spheromakformation................................................................... 991
9.3.1.3. Sustainmentphaseandhelicityinjection.................................. 994
9.3.1.4. Currentamplifcation.................................................................... 997
9.3.1.5. Regularversusstochasticmagneticfeld.................................. 998
9.3.1.6. Spheromak:afexibleconfguration.......................................... 998
9.3.2. Fieldreversedconfguration(FRC)............................................................... 1000
9.3.2.1. Fieldgeometry............................................................................... 1000
9.3.2.2. FRCformation............................................................................... 1002
9.3.2.3. FRCequilibrium............................................................................ 1003
9.3.2.4. Plasmarotationandstability....................................................... 1005
9.3.2.5. SteadystateFRCs.......................................................................... 1006
9.4. OPENCONFINEMENTSYSTEMS...................................................................... 1008
9.4.1. Confningaplasmaontheopenfeldlines................................................. 1008
9.4.2. Mirrorconfnement........................................................................................... 1008
9.4.3. Roleofioncollisions........................................................................................ 1010
9.4.4. Electronaxialconfnement.............................................................................. 1011
9.4.5. Velocity-spacemicroinstabilities.................................................................... 1012
9.4.6. MHDstability.................................................................................................... 1014
9.4.7. Suppressionofendlosses............................................................................... 1015
9.4.7.1. Tandemmirrors.............................................................................. 1015
9.4.7.2. Plasmarotation.............................................................................. 1018
9.4.7.3. Gas-dynamictrap.......................................................................... 1019
9.5. OTHERFUSIONCONFIGURATIONS:ABROADRANGEOF
IDEAS................................................................................................................................. 1020
9.5.1. Levitateddipoles............................................................................................... 1020
9.5.2. Cusps................................................................................................................... 1022
9.5.3. Magneto-electrostaticconfnementandelectrostaticconfnement.......... 1023
9.5.4. Pinches................................................................................................................ 1023
9.5.5. Magnetizedtargetfusion................................................................................. 1024
9.5.6. Otherconcepts................................................................................................... 1025
9.6. SUMMARY...................................................................................................................... 1026
CHAPTER 10
INERTIAL FUSION ENERGY
10.1. BRIEFHISTORY........................................................................................................... 1043
10.2. INTRODUCTION.......................................................................................................... 1044
10.3. BASICCONCEPTOFINERTIALFUSIONENERGY................................ 1044
10.4. REQUIREMENTONFUSIONGAIN.................................................................. 1046
10.5. IMPLOSIONCONCEPT............................................................................................ 1048
10.5.1. Whycompression?............................................................................................ 1048
10.5.2. Gainscalingofimplosionignition................................................................. 1050
10.6. IMPLOSIONPHYSICS.............................................................................................. 1051
10.6.1. Ablativeaccelerationandlaserthroughoutplasmainteractionin
directdriveimplosion....................................................................................... 1053
10.6.2. Classicalabsorption.......................................................................................... 1054
10.6.3. Collectivelaserplasmainteractions.............................................................. 1055
10.6.3.1. Resonantmodeconversion......................................................... 1056
10.6.3.2. Densityproflemodifcations...................................................... 1058
10.6.3.3. Electroniondecayandoscillatingtwostreaminstabilities 1058
10.6.3.4. Two-plasmondecayinstability.................................................. 1059
10.6.3.5. StimulatedRamanscattering....................................................... 1060
10.6.3.6. StimulatedBrillouinscattering.................................................. 1061
10.6.3.7. Filamentationinstability:self-focusing................................... 1062
10.7. ABLATIONPRESSUREANDDIRECTDRIVEIMPLOSION
HYDRODYNAMICS................................................................................................... 1063
10.7.1. Ablationpressure............................................................................................... 1063
10.7.2. Rocketmodelandhydrodynamiceffciency................................................ 1065
10.7.3. Implosionnon-uniformityandstability........................................................ 1065
10.7.4. Implosioninstabilities...................................................................................... 1068
10.8. FASTIGNITION:PRESENTSTATUSANDFUTUREPROSPECTS... 1069
10.8.1. Whyfastignition?............................................................................................. 1069
10.8.2. Ignitionconditions............................................................................................ 1070
10.8.3. Fastignitionintegratedexperiments............................................................ 1074
10.8.3.1. Directplasmaheating................................................................... 1074
10.8.3.2. Cone-shelltarget............................................................................ 1076
10.8.4. Fastignitionsimulation.................................................................................... 1077
10.8.4.1. Integratedsimulation.................................................................... 1077
10.8.4.2. Radiationhydrodynamicsimulation.......................................... 1078
10.8.5. Relativisticelectrongeneration,transportandheatingphysics............... 1080
10.8.6. PetawattlaserandfutureFastIgnitionFacility.......................................... 1082
10.9. INDIRECTDRIVEFUSION................................................................................... 1085
10.9.1. Conceptofindirectdriveimplosion.............................................................. 1085
10.9.2. Statusofindirectdriveimplosionexperiments........................................... 1088
10.9.3. Recentprogressinindirectdriveexperiments............................................. 1090
10.9.4. Otherdriversforindirectdriveimplosion.................................................... 1093
10.10. IFEPOWERPLANTDEVELOPMENT............................................................. 1095
10.10.1. IFEpowerplantsystems................................................................................. 1095
10.10.2. Driverdevelopment......................................................................................... 1097
10.10.3. Reactionchamber............................................................................................. 1099
10.10.4. Fuelpellet.......................................................................................................... 1101
10.10.5. Recentprogressininertialfusiontechnologies........................................... 1102
10.10.5.1. LIFEprojectatLLNL.................................................................. 1102
10.10.5.2. HiPERprojectintheEU.............................................................. 1103
10.10.5.3. Recentprogressinlasertechnology......................................... 1104
10.10.5.4. Recentprogressoftargetfabricationandinjection
technology....................................................................................... 1105
INDEX ........................................................................................................................................ 1119
1
CHAPTER 1
THE CASE FOR FUSION
P.K. Kaw, I. Bandyopadhyay
Institute for Plasma Research,
Bhat, Gandhinagar, Gujarat, India
1.1. INTRODUCTION
Humans do not live by bread alone. Physically we are puny creatures with
limited prowess, but with unlimited dreams. We see a mountain and want to
move it to carve out a path for ourselves. We see a river and want to tame it so
that it irrigates our fields. We see a star and want to fly to its planets to secure a
future for our progeny. For all this, we need a genie who will do our bidding at
a flip of our fingers. Energy is such a genie. Modern humans need energy and
lots of it to live a life of comfort. In fact, the quality of life in different regions of
the world can be directly correlated with the per capita use of energy [1.11.5].
In this regard, the human development index (HDI) of various countries based
on various reports by the United Nations Development Programme (UNDP)
[1.6] (Fig. 1.1), which is a parameter measuring the quality of life in a given
part of the world, is directly determined by the amount of per capita electricity
consumption. Most of the developing world (~5 billion people) is crawling up the
UN curve of HDI versus per capita electricity consumption, from abysmally low
values of today towards the average of the whole world and eventually towards
the average of the developed world. This translates into a massive energy hunger
for the globe as a whole. It has been estimated that by the year 2050, the global
electricity demand will go up by a factor of up to 3 in a high growth scenario
[1.71.9]. The requirements beyond 2050 go up even higher.
How is humankind going to produce the vast amount of energy it needs? In
the absence of any new developments, the workhorse is likely to be based on fossil
fuels. On the other hand, the use of fossil fuels as the major source of energy over
the last century has led to significant global warming through the emission of
greenhouse gases. Also, the fossil fuel reserves have started depleting with only
coal, the largest emitter of CO
2
, having the potential to last a few hundred years.
Oil and gas reserves have already started dwindling and would last a few tens
of years by various estimates; political and military conflicts for control of oil
and gas have already dominated the world energy scenario over the last decade.
Deployment of conventional fission based nuclear power, on the other hand,
has faced serious public opposition due to concerns of proliferation, radioactive
hazardous wastes, the potential for catastrophic Chernobyl like disasters, etc., all
2
KAW and BANDYOPADHYAY
politically and technically soluble problems and yet a matter of concern even
today. Alternative clean energy resources such as solar and wind, though having
the potential to become a major source of energy, yet have to significantly address
issues of energy density (which makes them unsuitable for large urban industrial
complexes), efficiency and cost of production before they can become a viable
alternative. Thus the human race is at a critical juncture today, when we need
to quickly develop a viable alternative source of clean energy with easy global
accessibility which can lead to sustainable development.
FIG. 1.1. United Nations human development index as a function of per capita energy use
in kWh (60 countries, 1997). Electricity consumption increases with human development.
Courtesy of Ref. [1.1].
It takes considerable time to develop new energy technologies and even
more time for them to be established as an alternate energy source in a cost
effective, safe and environmentally friendly way. Thus it is not too early to start
now. In this chapter we shall present a case for rapid development and deployment
of energy produced by nuclear fusion, an advanced nuclear technology with none
of the concerns of proliferation or accident scenarios of conventional fission
reactors and minimal radioactive wastes. Nuclear fusion research has seen a
remarkable progress since it started in a major way about 50 years ago and it has
culminated in the start of construction of the first experimental fusion reactor
called ITER [1.10] in a wonderful cooperation involving in effect more than
3
THE CASE FOR FUSION
half of the worlds humanity. This is particularly gratifying because finally the
world is coming to realize that global problems have to be solved by burying
national differences and working together to find technical solutions to difficult
problems. The world of today is a highly interconnected web where the engine of
industrial growth in one region is fuelled by investments from another region and
the economic ills and stagnation of one region directly influence the prosperity in
another region. Hence this cooperation in nuclear fusion research may become a
model for the technical solution of other global problems in the future.
In Section 1.2 we give a brief description of the energy needs of the
developing and the developed world to bring out with clarity the magnitudes of
energy requirements in the near term (say up to the year 2050) and in the longer
term (for the rest of the century), the methods which are likely to be employed
in the absence of any new technologies, the consequences for our environment
and how we might benefit in the long term by the use of non-fossil fuel sources
of energy. We also argue that for fulfilling the demand of centralized industrial
and urban centres, it will be necessary to promote the growth of CO
2
free nuclear
energy sources to about 40% of the total demand. Among advanced nuclear
technologies, a special place is filled by nuclear fusion because of its merits
such as easy, universal and almost unlimited access to the basic fuel, reduced
and more benign wastes, better safety features and the promise suggested by
recent technical developments in the field. Many of these features are detailed in
the subsequent sections. Thus in Section 1.3 we discuss the basic fusion process
itself, including the merits of fusion, the fusion reactions likely to be exploited
and the possibility of using advanced fuels. In Section 1.4 the basic approaches
to fusion have been summarized, with the bulk of the discussion on magnetic
confinement and the tokamak concept. A discussion is also provided for the
inertial fusion energy concept and the two large experiments, namely the US
based National Ignition Facility (NIF) [1.11] and the Laser Mega Joule (LMJ)
[1.12] experiment of France. In Section 1.5 we discuss the various socioeconomic
issues of relevance to the public acceptability of fusion, such as the costs, safety
issues and spin-offs, and in the final section we conclude our discussion.
1.2. ENERGY SCENARIOS
The main energy resource for the world in the past few centuries has been
fossil fuels. Analysis carried out by the World Energy Council (WEC) [1.13], the
International Energy Agency (IEA) [1.14] and other international organizations
in 1996 estimated that the fossil fuel resource lives based on reserves of coal,
natural gas and petroleum (using present technologies) were 231, 63 and 44 years
respectively [1.15]. Furthermore, the resource lives based on total resource base
including non-conventional oil resources such as oil shale, tar sand and heavy
oil, or gas resources such as shale gas from the Devonian period, tar sand gas,
4
KAW and BANDYOPADHYAY
underground aquifers, coal bed methane, methane hydrate and deep layer gas
(using evolving technologies) may be a few hundred years. Thus if we so desire,
fossil fuel resources may last us a hundred years or so. However, it is worth
noting that the fossil fuel wealth which has been created by millions of years of
geological evolution on the Earth, and which may be needed for sundry reasons
(other than burning) by our grandchildren and their progeny, will have been
squandered away in about 300 years of modern human civilization.
Furthermore, the massive use of fossil fuels for energy production during
the past century has started seriously degrading our environment. Greenhouse
gases are causing a significant warming of the planet with all the associated
consequences. Massive use of coal has also made phenomena like smog and
acid rain a common predicament of our major urban centres. If we increase the
consumption of fossil fuels by a factor of three, the consequences for the global
environment are likely to be staggering. In view of the above discussion, it is
unlikely that the use of fossil fuels for energy production will have an unfettered
growth in spite of the needs.
On the other hand, present day fission based nuclear energy, although it
plays an important supportive role (or even dominant role in countries such
as France), has associated safety, radioactive waste disposal and proliferation
issues. Similarly, power plants based on renewable resources such as wind or
solar, while already growing at a significant pace, will continue to play only a
supportive role because of the low energy density and lack of suitability to power
urban industrial complexes.
Thus, although we may be able to somehow satisfy our energy needs in
the short term, what is the remedy for long term energy needs and who is going
to take up the challenge? Fortunately, of late there is a growing recognition
among governments around the world of the possible disastrous consequences of
uncontrolled global warming [1.16] and the need for supporting the development
of new energy technologies. Many governments have set a target of not allowing a
temperature rise of more than 2
o
C. This would need reduction of carbon emissions
by about 80% over the next half a century [1.17]. It is important to note that there
are starkly contrasting requirements on the energy chain, with the ever increasing
global demand on the one hand and the environmental and social concerns on
the other. Furthermore, energy systems leading to sustainable development of the
entire human race must address the needs of the present without compromising
those of future generations. Hence we must look for alternative energy resources
which neither stress the ecosystem beyond the present level, nor totally exhaust
the already dwindling fuel reserves to prevent future generations from their use
for various applications. The new energy resources must be developed well in
time, taking into account the following:
y y They must be based on effcient and clean energy conversion processes with
widespread public acceptance and involvement.
5
THE CASE FOR FUSION
y y They must lead to sustainable development of the entire world, so must be
based on virtually inexhaustible resources, available globally without having
to depend on resources from politically unstable regions.
y y They must be suffciently energy dense, capable of driving large scale
industries, without requiring very large scale, widely distributed installations.
1.2.1. Near term energy scenario
In the near term, specifically over the next 2030 years, several of the
already available non-fossil energy sources which are continuously being
improved for better efficiency, wider availability and reduced pollution are likely
to be increasingly used to supplement the fossil fuel energy sources so as to fulfil
the global energy demand as well as possibly address environmental concerns.
Two such sources immediately come to mind. One is solar energy and the other
is nuclear energy.
Solar energy is renewable and is hence an obvious candidate for sustainable
development. Solar energy may be utilized through solar photovoltaic methods
(solar cells) or through solar thermal methods (hot water for residential purposes,
commercial use or electricity production) or through biofuel cultivation. In all
these cases, the basic problem is the low flux of solar energy on the Earths
surface, which makes it difficult to plan massive energy hungry industrialized
urban centres running on solar energy. Nevertheless, solar energy technologies
have seen remarkable development of late with the advent of nanotechnology.
New plastic materials made of specially designed nanoparticles of polymer called
quantum dots can convert the invisible infrared spectrum of the solar energy into
electric energy. Conventional solar panels, including plastic solar cells, use the
visible part of the energy, whereas about 50% of the Suns energy actually lies in
the infrared spectrum [1.18]. Scientists from Spectrolab, a subsidiary of Boeing,
have recently reported [1.19] development of multijunction solar cells with an
efficiency of more than 40%, a new world record for solar photovoltaic cells.
This greatly surpasses todays industry average of 1218% efficiencies and the
best available solar cells with 22% efficiency. The Spectrolab scientists also
predict that concentrator solar cells could achieve efficiencies of more than 45%
or even 50% in the future, with theoretical efficiencies being about 58% in cells
with more than three junctions.
Even though the early use of solar photovoltaic methods was mostly ranging
from small individual appliances such as calculators to powering remote homes
not connected to grids, of late a number of medium sized solar photovoltaic
power plants have been installed, mainly in Europe and the USA. For example,
Spain has several power stations producing tens of megawatts, the largest being
the 60 MW (85 GWh) Parque Fotovoltaico Olmedilla de Alarcn, while the
14 MW (30 GWh) Nellice Airforce base photovoltaic station is the largest in the
6
KAW and BANDYOPADHYAY
USA [1.20]. Several large photovoltaic power stations are also being constructed,
the largest being the 550 MW Topaz Solar Farm in California. Japan, one of the
largest markets of solar energy, intends to increase its residential electric energy
consumption from solar to 50% by 2030 from the present level of a fraction of
one per cent [1.21]. In the USA, the goal is to meet 10% of US peak generation
capacity by 2030, which would be the energy equivalent of about 180 million
barrels of oil at that time [1.22].
Thus, while solar photovoltaic is gradually establishing itself as an
alternative energy source, it is likely to play only a supportive role for several
generations. It still has to address several issues: it lacks the energy density of
other conventional large power plants, for example nuclear reactors, and so
cannot be deployed in large urban industrial complexes (it needs large scale
deployment). The stability of a grid with significant photovoltaic contribution,
especially in winter without enough sunlight, or during night time, has not been
fully studied. The high investment cost and maintainability are still issues to be
resolved, as are the problems with generation of toxic wastes in the manufacturing
of photovoltaic panels and their ultimate disposal.
Biofuel technologies such as the production of ethanol or biodiesel either
from sugar or starch rich vegetation or from biological wastes on the other hand
have also experienced significant development over the years. Biofuels are being
used routinely in many countries now, mixed with conventional fuels such as
petrol or diesel to be used as primary automobile fuels. Newer generation biofuels
such as algae oils or oilgae [1.23], conversion of vegetable oils or biodiesel into
gasoline or genetically engineered plants consuming more carbon than is released
from combustion of the biofuels they produce are also becoming significant.
While biofuels are likely to play a supportive role in primary fuels, they have been
plagued with issues such as altering food prices as crop cultivation is reoriented
from food to fuel production, or adding significantly to greenhouse emission, soil
erosion, deforestation and desertification. In some developing countries the use
of biofuels has contributed to arid lands, expansion of deserts, general losses of
biodiversity and instability in food prices. Hence it is highly unlikely that biofuels
will become a dominant primary energy source for electricity production.
The nuclear energy option based on nuclear fission, on the other hand, is
a valuable one which is already being exploited at ~25% average level in the
developed world. Countries such as France are even using a much higher (~78%)
percentage of nuclear energy. Naturally fissile materials such as
235
U will perhaps
get exhausted in a few hundred years, but as one masters the use of fissile
materials such as
239
Pu and
233
U, which can be bred from fissile materials like
238
U
and
232
Th, fission can supply the world with energy for several thousand years.
Even though nuclear energy plants can readily fulfil the needs of centralized
industrial centres, wider exploitation has been curtailed because of fears of
nuclear proliferation and lack of safety. This has prevented free access to nuclear
7
THE CASE FOR FUSION
technologies. Furthermore, there are worries about the need for great care in the
handling of nuclear waste. Most of these problems have technical solutions but
may have significant economic and other political and social implications. Thus
it is uncertain as to how much of the energy needs can be satisfied by utilization
of nuclear fission power.
Thus over the next 2030 years, the world energy scenario is likely to
remain more or less the same with fossil fuels remaining the workhorse, but oil
and gas gradually becoming more scarce and expensive. Conventional nuclear
reactors will increase their share of energy sources, especially in emerging
economies such as India, while alternative sources like solar photovoltaic and
biofuels are expected to play a more significant supportive role, especially in
developed economies. However, new energy technologies take significant time
to establish and hence the seeds of alternative energy sources to serve the world
in the long term for sustainable development need to be sown now.
1.2.2. Long term energy scenario and the role of fusion
There have been several detailed attempts at developing scenarios for
electricity demand and supply on both a regional as well as a global basis.
Some of the most recent comprehensive ones are the studies being carried out
[1.241.29] by the International Institute for Applied Systems Analysis (IIASA)
[1.30] under the auspices of the World Energy Council (WEC) [1.13] and also
the studies by the Intergovernmental Panel on Climate Change (IPCC) [1.31].
The main objective of these studies has been to estimate the upper range of future
electricity supply to be assured by the existing power generating capacities along
with an estimate of additional power supply capacities generated by prospective
technologies. Another objective of these studies has been to determine a
region-wise breakup of the possible development paths of power generation
systems, especially in a scenario of fast depleting fossil fuel reserves. They
also include estimates of the respective share in total installed capacities and
maximum electricity supply of each technology or fuel type, including advanced
energy technologies, such as thermonuclear fusion.
However, one of the problems with such studies is that they all naturally
have to assume a set of likely prevailing scenarios with underlying assumptions,
for example regarding population, economic and industrial growth, new
technology developments, the availability of primary energy resources and a host
of other factors. As a result, the predictions from these models differ somewhat
depending on the underlying factors in a given scenario. For example, the IIASA
WEC study [1.32] on eleven different world regions describes three alternative
cases of future economic development and energy consumption trends that further
divide into six different scenarios, and quantifies their implications. Figure 1.2
shows the projected world energy consumption until the year 2100. Case A
8
KAW and BANDYOPADHYAY
corresponds to a scenario of remarkable technological improvements leading to
rapid economic growth and consequently resulting in the highest energy demand.
Case B corresponds to a more realistic and less spectacular growth scenario
where technology improvements are also moderate, and consequently results in
lower energy consumption. Case C is driven by ecological considerations for
the future, though it allows for significant technological progress, especially in
areas related to alternative (non-fossil) energy resources, and relies on extensive
international cooperation focused on environmental protection and equitable
economic growth. Consequently, the projected energy consumption in case C
is the lowest among all scenarios.
FIG. 1.2. World Final Energy Consumption until the year 2100 in the IIASAWEC Study
Global Energy Perspectives. Plotted using the data from Ref. [1.32].
On the other hand, Fig. 1.3 shows past usage and projections from the
IIASAWEC analysis [1.30] of the total primary energy consumption in different
regions of the world. Thus we see, as is well documented in many recent studies,
that while the technologically advanced OECD nations show a steady saturation
or small growth over this century (or even a slight decrease in North America in
the later part of this century), the developing nations in South and Central Asia,
such as India and China, or the Middle East nations, show a spectacular growth
(up to tenfold) in primary energy consumption. In fact, it is expected that by the
end of the century final energy consumption in these developing countries will be
more than three times higher than in the industrialized OECD and Former Soviet
9
THE CASE FOR FUSION
Union countries. This projection is very real, as is evident from the economies
of India and China, which are already showing sustained, close to double digit,
growth rates and have the potential to grow even faster throughout the coming
few decades. Such sustained growth in the developing world is going to put an
enormous demand on primary energy resources, which, simply stated, cannot be
sustained by the ever dwindling and potentially environment degrading reserves
of fossil fuels.
FIG. 1.3. Region-wise primary energy consumption in IIASAWEC scenario B. Plotted using
the data from Ref. [1.32].
In spite of differences in the predictions on the global energy scenario,
there is one factor generally common to all of them: they all envisage a gradual
shift away from fossil fuel sources by the end of the 21st century. Figure 1.4
shows the past actual use of various energy resources and the projections till the
end of this century by the IIASAWEC scenario B model [1.30]. Important to
note in this projection is the substantially growing dependence on non-fossil-
fuel resources, for example nuclear and renewables, in the second half of this
century, simultaneous with gradual reduction in the dependence on oil and gas. It
is important to note that, in this model, the contribution of coal remains more or
less constant at the present day value.
10
KAW and BANDYOPADHYAY
FIG. 1.4. Evolution of primary energy shares, historical development from 1850 to 1990 and
projections till 2100, for Case B. Plotted using the data from Ref. [1.32].
However, on the other hand, even though coal will be available in the
world for slightly over 150 years, the prospect of countries such as India and
China being forced to rely on coal for their energy demand would be disastrous
to the environment of the whole Earth, as is clear from the recent discussions
on the mounting evidence of an incipient global warming. The world is already
experiencing a steady rise in temperature (Fig. 1.5) [1.33] and its effects are
already showing in global climate changes caused by changes in flow patterns in
global ocean currents, as well as depleting ice layers in the polar and Himalayan
regions. Even though, as some argue, there may be uncertainties in long time
predictions of climate, it would be foolhardy to assume that the uncertainties
would necessarily lead to a favourable situation.
FIG. 1.5. Global average temperature over the last one and a half centuries showing a more
or less steady increase over the last fifty years or so. The fluctuations and their cycles can be
correlated with various events such as solar cycles [1.33].
11
THE CASE FOR FUSION
However, as we look at nuclear energy as an important primary energy
resource, it is important, in view of the uncertainties involved in a very wide
acceptance of the utilization of fission power, that we do not put all our
nuclear eggs in one basket. We must also look at all advanced nuclear energy
technologies. Fusion is one such option. Thus, if one aims at restricting CO
2

emission levels to within 450 parts per million (ppm) (the pre-industrial level
being about 280 ppm), as the constraint tightens, the use of coal diminishes and
the contribution from other sources increases, with fission and fusion power
together playing a dominant role (see Fig. 1.6).
FIG. 1.6. Projections for CO
2
emission constrained energy scenarios in Western Europe in
2100 with a large role of nuclear fusion [1.34]. (1 exajoule-electric (EJe) ~ 277.7 terawatt-
hours (TWh).)
Let us now consider for example the specific case of India. From 1981 to
2000, its GDP has grown at an average rate of 6% and since 2000, because of
opening up of the economy through policy decisions, at a faster rate of close to
8%. It has the potential to grow at about 5% even as far as 2050. The ratio of its
growth of electricity generation to GDP growth over the past decades has been
a steady factor of about 1.2. The total electricity generated in 2002 was about
638 TWh, of which about 66.7% was through coal and lignite, 19.6% through
oil and gas and only about 3% through nuclear. Projections by the Department
of Atomic Energy [1.35] predict the electricity production to go up to about
8000 TWh by 2050, still 47% of which is likely to be resourced from coal. This
scenario is very challenging and barely sustainable as this would mean the total
carbon emission in India would jump from a level of about 300 metric tonnes
of carbon (MtC) today to about 2100 MtC by 2050 (which is about 30% of the
global carbon emissions in 2000). By a much more conservative estimate using
the ANSWER/MARKAL model [1.36], the total electricity production in India
12
KAW and BANDYOPADHYAY
will increase to about 3000 TWh by 2050 and about 4300 TWh by 2100 (which
still implies a carbon emission from India by 2100 of about a quarter of the global
emission level). A scenario in which the global atmospheric CO
2
concentration is
restricted to be within 550 ppm is shown in Fig. 1.7. By this model, the restricted
emission scenario can be achieved if fusion starts playing a dominant role beyond
2060 with a share of about 10% by the turn of this century, with about 430 TWh
of fusion electricity produced in India from about 67 GW of installed capacity.
FIG. 1.7. Projections of dependence on different primary energy resources in India with global
CO
2
emission restricted to 550 ppm by the year 2100, with India contributing only 7.5% of the
global emissions. Reprinted from Ref. [1.36]. Copyright (2010), with permission from Elsevier.
These models show that fusion power is likely to give us a wonderful
opportunity to provide a viable and credible solution to the long term sustainable
energy needs of the world. It has none of the CO
2
emission problems of fossil
fuels. The amount of CO
2
emission in the entire life cycle of fusion reactors
(through manufacturing processes of some reactor components), as shown in
Fig. 1.8 [1.37], can be up to a factor of 45 times less than that of coal based
reactors of the same capacity. It is somewhat more than the CO
2
emission from
a comparable fission based nuclear reactor. Thus fusion power, in a way similar
to fission power, can alleviate the already deteriorating climatic conditions and
prospects of global warming that the world is facing today. In contrast to the
renewables, it is energy dense and can therefore be used for satisfying the needs
of urban industrial complexes. On the other hand, compared to the conventional
fission based nuclear power plants, it has the prospect of considerably reduced
long lived radioactive emission problems and inherent operational safety. The
biological hazard potential of fission plants (defined in terms of the ratio of the
amount of radioactive material in a reactor to the allowed level of concentration
in the atmosphere) is several orders of magnitude higher than that predicted
for fusion reactors. Some of the common radioactive waste materials in fission
reactors, such as
131
I,
90
Sr or
137
Cs, are highly toxic and hazardous, especially
13
THE CASE FOR FUSION
the latter two, which have half-lives of about 30 years, which is long enough to
require their absolute containment for hundreds of years. On the other hand, the
main radioactive fuel in a fusion reactor, tritium, can easily be discharged from
our bodies through metabolism. It is a weak (18.6 keV maximum energy) beta
emitter whose radiation can easily be absorbed by a thin sheet of paper [1.37]. As
a result, the allowable concentration level of tritium in the atmosphere is 500 times
higher than that for
131
I. However, the main potential problems of radioactivity
in fusion reactors will be from neutron activated reactor components. Fusion
reactors are being designed with carefully chosen low activity materials so as to
require containment of less than one hundred years after decommissioning of the
reactors. Thus fusion power is likely to get much more social acceptance when it
becomes commercial. Moreover, the resources are plentiful to the extent that they
are virtually inexhaustible and easily accessible to the entire cross-section of the
world population.
FIG. 1.8. CO
2
emission level of power reactors based on various fuel resources in their entire
life cycle, showing fusion reactors as the third lowest CO
2
emitter after hydro and fission
reactors [1.37].
It is satisfying to note therefore that the tremendous amount of research
and development over the last about 50 years in the field of fusion science and
technology has reached a critical stage today. Scientists from some of the worlds
major nations, namely, China, the whole of the European Union, India, Japan, the
Republic of Korea, the Russian Federation and the USA, which together account
for more than half of the worlds population, have come together to build ITER,
the first experimental thermonuclear reactor which will produce energy ten times
greater than the input auxiliary heating power. Various countries have national
programmes to build demonstration reactors or DEMO (for example in the EU
[1.38, 1.39], Japan [1.40], the USA [1.41] and India [1.42]) some as early as
14
KAW and BANDYOPADHYAY
in 2030, but most likely by 2050 which will actually supply electricity to the
grid.
1

In conclusion then, in the short term of the next 3040 years, the world
energy scenario is likely to be still dominated by gradually depleting fossil fuels,
with nuclear fission and renewables taking a gradually increasing share. During
this time fusion energy will establish itself through experiments like ITER and
demonstration power plants like DEMO. In the long term, towards the end of
this century, however, fusion power is likely to become commercial and play an
increasingly dominant role in the world energy scenario.
1.3. FUSION BASICS
1.3.1. What is fusion?
The brightest example of fusion around us is provided by the stars and the
Sun, which have been burning brilliantly for billions of years using this option. In
contrast to nuclear fission, where heavy nuclei like uranium are fragmented and
release energy, in fusion one starts with light elements and brings them together
so that they may fuse to form heavier elements. The resulting heavier elements
have slightly less mass than the fusing elements and this mass difference results
in the release of energy. As an example, when deuterium and tritium nuclei
(which are the two heavier isotopes of hydrogen with mass numbers 2 and 3
respectively) are brought together, they fuse and form a helium nucleus and a
neutron; the mass difference is released as 17.6 MeV of energy. Energy comes
out in the form of the kinetic energies of the product nuclei, from which it may
be trapped and used for electricity production. For fusion to occur, one has to
bring the protons or heavier reactant nuclei (which are positively charged and
naturally repulsive) close enough to overcome the electrostatic repulsion, so that
the nuclear strong force (a very short range force), which binds the nucleons
together in a nucleus, helps them fuse by a quantum mechanical tunnelling
process. This is possible when the nuclei are heated to very high thermonuclear
temperatures, when the kinetic energy of the thermal particles is enough to help
1
However, the 14 MeV neutron fluence in ITER at about 0.5 dpa/a (displacement per
atom per year) will be much less than that of DEMO (~20 dpa/a) or other future fusions reactors.
Because of this, ITER will not be able to test reactor relevant first wall materials with high
neutron irradiation doses. Furthermore, neutron fluxes of pressurized water fission reactors are
about 100 times lower than they would be in fusion reactors, and have lower neutron energies. So
fission reactors also cannot be used to fully test fusion reactor materials. To specifically address
the materials issue, therefore, the International Fusion Materials Irradiation Facility (IFMIF)
[1.43] is being launched in a collaboration between Japan, the EU, the Russian Federation
and the USA. IFMIF will have an accelerator based D-Li neutron source with fluence of up to
20 dpa/a.
15
THE CASE FOR FUSION
them overcome the electrostatic repulsion and come close enough to fuse. For the
case of deuterium and tritium nuclei, this requires that a mixture of D-T nuclei
be heated to a temperature of the order of ~10 keV (i.e. ~100 million C); the
resulting process is known as thermonuclear fusion.
The energy released through a fusion reaction is much larger than that
in chemical reactions, because the binding energy that holds the nucleons in a
nucleus together is far greater than the energy that binds atoms and molecules
together through electronic linkages. For example, the ionization energy of
a hydrogen atom (energy required to strip the single electron from a hydrogen
atom) is 13.6 eV less than one millionth of the 17.6 MeV released in the D-T
reaction. Fusion reactions also have an energy density (energy released per unit
mass of the reactants) much larger than fission reactions, even though individual
fission reactions involving very heavy nuclei are generally much more energetic
than individual fusion reactions. Only in direct conversion of mass into energy,
for example through matterantimatter collisions, could more energy per unit of
mass than in the fusion reactions be released.
A slow thermonuclear fusion of protons is what mainly powers the Sun and
the stars, the tremendous gravitational energy due to the very large mass holding
the fusing protons together against the de-confining tendency due to thermal
expansion. Uncontrolled fusion reactions in the form of thermonuclear explosions
using deuterium and tritium, where the plasma is unconfined, have already been
achieved on Earth. Thus there is no doubt that thermonuclear fusion works and
produces lots of energy. As it is impossible to produce astronomical masses on
the surface of the Earth, the biggest challenge for scientists is achieving fusion in
a controlled manner in a confined plasma, which can then be used for electricity
production or any other useful application. Controlled thermonuclear fusion
research, aimed at converting fusion energy into electric energy, has been carried
out around the world for more than 50 years. In spite of the tremendous scientific
and technological needs of this field, fusion research has seen a remarkable and
steady progress, comparable to some of the fastest growing technical challenges
in modern science and technology, such as developments in semiconductor chip
manufacture and accelerator research. Tokamak based fusion experiments based
on the magnetic confinement fusion concept have also doubled the characteristic
confinement parameter, the so called triple product (which we shall describe in
detail later), once every couple of years and have already achieved the so called
break-even condition where the output energy produced as a result of the fusion
reaction is equal to the input energy spent in achieving the reaction. This has
culminated in commencement of construction of the ITER device which will
produce about ten times more power than the input heating power; ITER is slated
to start operations in about a decade from now.
What are the merits of fusion energy, should we succeed in exploiting it?
First of all its fuel is plentiful. Deuterium is found mixed up (1 part in 6000) as
16
KAW and BANDYOPADHYAY
heavy water (D
2
O) in natural water. It is virtually inexhaustible. Tritium does not
exist naturally because it has too many neutrons and is therefore unstable to beta
decay, with a half-life of about 12.6 years. It has to be bred from lithium, which is
widely available in the Earths crust and also in the oceans. Pure deuterium fusion
is also possible, although the conditions required for its successful exploitation
are more stringent (because of weaker fusion cross-sections) and are therefore
likely to be met only in second generation fusion reactors. The fuel is therefore
virtually inexhaustible and is likely to last several tens of thousands of years.
2

Secondly, fusion fuel is readily accessible from everywhere. This gives
tremendous energy security to all nations and is to be contrasted with the
politically inflammable uneven distribution of fossil fuels such as oil (6% of the
nations own more than 66% of the oil wealth of the world).
Thirdly, there are limited radioactive waste problems. None of the fusion
reaction products are radioactive in the first place. Fusion neutrons can induce
radioactivity but that can be minimized by a clever choice of structural and other
materials so that only short lived radioactivity is produced. There is also the
prospect of using advanced fuels such as protonboron which produce essentially
neutron-less fusion and hence no radioactivity.
Fourthly, the fusion reaction is inherently safe. There is no danger of
runaway reactions, criticality or a meltdown. At a given instant of time the total
inventory of fusion materials in the device is just enough to produce the power for
a few seconds. This is to be contrasted with fission reactors, where the inventory
stored at a given instant is enough to cause a major explosion.
Lastly, there are no dangers of proliferation or of a terrorist group or fringe
group running away with key materials which may be put together to form a
crude device. The fusion reaction is so difficult to initiate that it needs a major
technical establishment such as a magnetic fusion reactor or a laser fusion reactor
or an atomic fission device to create conditions under which fusion may be
initiated. It is thus completely free from such misuse.
2
It is worth mentioning here that a D-T based 1 GW(e) fusion reactor will burn about
37 kg of D and 56 kg of T annually. It is to be noted that while the D inventory is not an issue,
tritium has to be bred in the fusion reactor itself from lithium and cannot be stored indefinitely
as it is radioactive with a half-life of 12.6 years. It is to be noted that the accumulated yield of
T by all of CANDUs fission reactors will be only about 30 kg in 2025 after more than 40 years
of operation. However, with stringent inventory control, and by achieving even a modest net
tritium breeding ratio of 1.01 in a fusion reactor, it is possible to achieve a doubling time (the
time when one fusion reactor can produce enough excess T to support a second reactor) of less
than 5 years [1.441.46].
17
THE CASE FOR FUSION
1.3.2. Fusion power gain Q
Now let us examine the conditions which a reactor successfully giving
fusion energy output must satisfy. We first note that we have to invest energy in
raising the temperature of the D-T mixture to about 100 million C. This converts
the gas mixture into a plasma, which radiates mainly through bremsstrahlung
radiation because of electronion and electronelectron encounters. Secondly, if
the D-T plasma is diluted by some impurities which are not fully ionized, there
is some radiative power loss associated with impurity radiations which cools
the electrons and through them the D-T mixture because the electrons and ions
are in near thermal equilibrium. In addition to this, if the plasma is confined
by magnetic fields, it radiates by synchrotron radiation. The power required to
maintain the plasma at 100 million C is thus related to the power required to
sustain the plasma temperature against thermal conduction/convection losses
plus the radiative power losses of the above three varieties.
The net power output of a fusion power reactor can be measured in terms of
the steady state fusion power gain or the Q factor defined as the ratio of the fusion
power output to the input power, i.e. the auxiliary power supplied from outside
to sustain the reaction: / /
output input fusion aux
Q P P P P . Thus, for fusion power
to be successful, the minimum criterion for a fusion power plant is 1 Q > . The
state 1 Q is known as the break-even condition, when fusion output power just
equals the auxiliary input power. On the other hand, the thermonuclear fusion
plasma can also be confined in an ignited state when 0
aux
P or Q ~ , which
happens for the D-T fusion reaction, for example, when the output alpha particles
from the fusion reaction lose all of their energy in keeping the thermonuclear
plasma hot and thus the alpha power accounts for the transport and radiation
losses. In such a scenario, the fusion reaction is completely self-sustained by
the alpha power and no external heating power is required. Now, the net heating
power in the plasma can be obtained through the power balance:
/
heat aux Br trans p
P P P P P dW dt
o
+ + (1.1)
where P
o
and
Br
P are the power in the alpha particles and bremsstrahlung losses
respectively and are given by:
2
v V v V / 4
D T p e p
P n n E n E
o o o
o o (1.2)
1/ 2 2
V
Br B e p
P C T n (1.3)
where it is assumed that
D T e
n n n for a 50/50 D-T mix, v o is the average
collision cross-section of the reactants, E
o
is the energy carried by the alpha
particles (3.5 MeV) and V
p
is the plasma volume.
trans
P is the total power
18
KAW and BANDYOPADHYAY
lost through various transport processes and /
p
dW dt is the power required to
increase the internal (kinetic) energy of the plasma, with 3 V
p e B e p
W n k T ,
B
k
being the Boltzmann constant. The last term /
p
dW dt in Eq. (1.1) is applicable
only in transient cases. The transport losses are related to the total energy content
through the energy confinement time
E
t , defined as:
/ ( / )
E p heat p
W P dW dt t (1.4)
Thus the power balance equation Eq. (1.1) can be written in terms of the
energy confinement time
E
t and the fusion gain Q (note that 5
fusion
P P
o
) as:
( )
2 1/ 2 2
3
5
v 3
4
e B e
e B e e e B e
E
n k T
Q d
n E C T n n k T
Q dt
o
o
t
| ` +
< > +

. }
(1.5)
Thus for a steady state condition, i.e. / 0 d dt , Eq. (1.5) yields for the
very important product of density times energy confinement time:
1/ 2
3
v ( 5)
4
B e
e E
B e
k T
n
E Q
C T
Q
o
t
o

< > +

(1.6)
Thus the break-even condition 1 Q

leads to the critical minimum condition
for a fusion plasma, usually known as the Lawson criterion [1.47]. Equation (1.6)
gets slightly modified as follows if the conversion efficiencies of input power
sources (
in
n ) and that of output thermal energy to electric energy (
out
n ) are
considered [1.48]:
( )
( )
1/ 2
3 1 .
v ( 5)
. 1 .
4
in out B e
e E
in out B in out e
k T
n
E Q
C T
Q
o
n n
t
o
n n n n

< > +

(1.7)
Typically one can assume .
in out
n n ~ 1/3 and then the optimum temperature
for break-even for a 50/50 D-T plasma occurs at T
e
~ 15 keV; thereby the Lawson
criterion for the break-even condition becomes:
20 3
10 sm
e E
n t (1.8)
If one wants to consider the fusion reaction for a range of temperatures, one
has to use the variation of the fusion cross-section with temperature. Roughly
speaking, the fusion rate v o varies with temperature as
2
e
T and hence the
Lawson criterion for a D-T plasma at T
e
= 15 keV may be rewritten in the form:

21 3
1.5 10 keVsm
e e E
n T t > (1.9)
19
THE CASE FOR FUSION
For the D-D reaction, the cross-section for fusion is smaller than that for
D-T and the fusion energy yield per reaction is also less, so that the critical
condition is a temperature of 400 million C and an
e E
n t product given by
22 3
10 sm
e E
n t > .
Lawson also gave the minimum temperature required for a fusion reactor,
which is the threshold temperature at which the fusion power density just exceeds
the bremsstrahlung radiation. Thus, we see from Fig. 1.9 that the minimum
temperature for D-T reactions is 3keV T > ,

and for D-D reactions, 20keV T > .
FIG. 1.9. Power density for a density of n
D
= n
T
= 5 10
20
m
3
(plotted with data from
Ref. [1.49], p. 45).
We can also see from Eq. (1.6) that the ignition criterion (Q ~) becomes:
1/ 2
3
v / 4
B e
e E
B e
k T
n
E C T
o
t
o

< >
(1.10)
When the plasma attains the ignited state, no external heating is required
and the fusion reaction is self-sustained through the energy of the fusion products
till the plasma burns up all the fuel as long as the necessary condition for ignition
is maintained. For ignition, the optimum temperature for a 50/50 D-T fuel at
20 3
10 m
e
n

is 30 keV
e
T
21 3
8.1 10 keVsm
e e E
n T t > (1.11)
20
KAW and BANDYOPADHYAY
Here of course we have assumed that most of the radiated energy is in
the form of bremsstrahlung radiation (true for fusion plasmas at ~1020 keV)
and thus
,min
P
rad Br
P ; otherwise there are terms like
2
/
cy e e
C B T n also for
cyclotron radiation in the denominator in Eqs (1.6) and (1.10) which we have
neglected here.
1.3.3. Fusion reactions
The most convenient, though not necessarily most efficient fusion reaction
to carry out in present day magnetic fusion devices like tokamaks is that between
deuterium (D) and tritium (T) (Fig. 1.10), both isotopes of hydrogen, one with
one neutron and the other with two neutrons in their nucleus respectively (normal
hydrogen has only one proton and no neutron in its nucleus). The D-T fusion
is most conveniently achievable because the collision cross-section of the D-T
fusion reactions is the highest and occurs at the lowest temperature.
FIG. 1.10. Schematic of a D-T fusion reaction.
The reasons for this can be visualized in the following way. In a nucleus, the
protons and the neutrons are bound together by the nuclear strong force, which
is of very short range; as a result the binding is primarily between immediately
neighbouring nucleons. Thus the nucleons which are completely surrounded by
neighbouring nucleons, for example those in the interior of a nucleus, are more
strongly bound than those on the surface. Thus the nuclear binding energy is the
energy required to separate the nucleons. Its magnitude is given by the Einstein
formula
2
E mc A A , where m A is the small mass difference between the
nucleus and the sum total of constituent nucleon masses and c is the speed of light.
Since for the smaller nuclei, the ratio of their volume to surface area is small, for
them the binding energy per nucleon due to the strong force is generally smaller
than that of the large sized nuclei. This is because for the large sized nuclei, the
interior nucleons are more completely surrounded by neighbouring nucleons and
the binding energy per nucleon approaches a limiting value corresponding to that
21
THE CASE FOR FUSION
of a fully surrounded nucleon. On the other hand, the electrostatic force between
protons is long range and follows the inverse square law. Thus to add a proton
to a nucleus, one has to overcome the so called Coulomb barrier experienced
by the approaching proton due to the repulsive electrostatic force from all other
protons already within the nucleus. As these two forces oppose each other in a
nucleus, the net resulting binding energy per nucleon generally increases with the
size of the nucleus, up to the nuclei of the elements iron and nickel. Beyond these
nuclei, the binding energy per nucleon starts decreasing for heavier nuclei, until it
becomes negative for very heavy ones, which are then naturally unstable, fissile
and radioactive. Thus
56
Fe, the most common isotope of iron with 26 protons
and 30 neutrons, is one of the most stable nuclei with a binding energy of
8.790 MeV/nucleon and is formed as the end product of most fusion and fission
chain reactions. This is the reason why
56
Fe is the most abundant element in the
universe and is found at the core of the Sun and the stars.
3
As the deuterium and the tritium nuclei have the lowest binding energy per
nucleon, the kinetic energy required for them to have the highest probability for
fusion to occur is the lowest. This is represented by the collision cross-section
curves, which are plotted in Fig. 1.11 for the D-T reaction as well as for some
of the other common fusion reactions. In this figure, the kinetic energy of the
reacting particles increases from left to right, while the vertical axis (in barns)
represents the probability of a collision between two particles with that kinetic
energy resulting in a fusion reaction. Thus the D-T reaction has the largest
cross-section at about 100 keV, while for the other reactions, the peak probability
occurs at much higher temperature as also they have much lower probability; for
example, the p-
11
B reactions have the largest cross-section, of about 1.2 barns
(about 5 times less than the peak D-T cross-section) at 642 keV.
3
Although technically
62
Ni is slightly more stable with a binding energy of
8.794 MeV/nucleon it is less abundant than
56
Fe as it is not the natural end product of the fusion
and fission chain reactions.
22
KAW and BANDYOPADHYAY
FIG. 1.11. Fusion cross-sections of various fusion reactions as a function of kinetic energy of
an incident D or p on a stationary target. The data for the D-D, D-T, D-He
3
and p-
6
Li curves
are taken from the ENDF B-VII database [1.50] for incident deuterium/proton, while that for
p-
11
B is taken from Ref. [1.51]. The curve for D-D represents a sum over the cross-sections of
the reaction branches.
TABLE 1.1. LIST OF THE MOST FAVOURABLE FUSION REACTIONS
1. D + T

4
He (3.5 MeV) + n (14.1 MeV)
2. D + D

T (1.01 MeV) + p (3.02 MeV) (50%)

3
He (0.82 MeV) + n (2.45 MeV) (50%)
3. D +
3
He

4
He (3.6 MeV) + p (14.7 MeV)
4. T +T

4
He + 2 n + 11.3 MeV
5.
3
He +
3
He

4
He + 2 p
6.
3
He + T

4
He + p + n + 12.1 MeV (51%)

4
He (4.8 MeV) + D (9.5 MeV) (43%)

4
He (0.5 MeV) + n (1.9 MeV) + p (11.9 MeV) (6%)
7. D +
6
Li

2
4
He + 22.4 MeV
8. p +
6
Li

4
He (1.7 MeV) +
3
He ( 2.3 MeV)
9.
3
He +
6
Li

2
4
He + p + 16.9 MeV
10. p +
11
B

3
4
He + 8.7 MeV
23
THE CASE FOR FUSION
1.3.4. Fusion fuels
The D-T reaction is not necessarily the most desirable fusion reaction
as one of its major disadvantages is the production of a 14.1 MeV neutron in
each reaction. The neutrons get absorbed in the surrounding fusion chamber
(for example in tokamaks in the surrounding mostly metallic structures such as
blankets and vessels). The effect of neutron activation on metals is roughly that
they become hardened, brittle, as well as radioactive. This has two important
effects: first, it becomes necessary to use low activation materials inside the
vacuum chambers of fusion devices, and second, the lifespan of the machine
components surrounding the fusion plasma gets seriously limited. Thus, even
though the first generation of fusion devices will depend on D-T reactions
because of the ease of achieving them, future devices may use more efficient
reactions resulting in less neutron activation. Table 1.1 contains a list of fusion
reactions that are most favourable.
Any of the above reactions can in principle be the basis of future fusion
reactors for power production. However, the level of difficulty in achieving them
depends on several factors. In the first generation tokamak based fusion reactors
like ITER, a mixture of D-T particles is confined in a magnetic trap and heated to
high energies (about 1020 keV mean), where the particles have a distribution of
kinetic energy (called a Maxwellian distribution), with a small fraction of them
attaining 40 keV and fusing.
In addition to the temperature and cross-section of the reactions, one must
also consider the total energy of the fusion products
fus
E , the energy carried by the
resultant charged particles
ch
E and the atomic number Z of the non-hydrogenic
reactant. One interesting case is the D-D reaction, which has two branches, one
resulting in a T and a proton, the other in a
3
He and a neutron, both branches
having a 50/50 probability. The T can further undergo a D-T reaction and the
3
He can undergo a D
3
He reaction. However, while the T in fact gets completely
burnt up in a deuterium plasma and adds its energy further to the reaction chain,
the optimum temperature for D
3
He reactions is much higher than that of D-D
reactions and is not expected to contribute to the overall fusion energy. Thus one
can calculate the overall D-D fusion energy as 12.5 MeV
fus
E and the energy
in charged particles as 4.2 MeV
ch
E .
Another important parameter in the fusion reactions is the neutronicity of
the reaction, which is measured as
( )
/
fus ch fus
E E E , the fraction of the fusion
energy released through neutrons. Neutronicity is an important indicator of the
magnitude of the problems associated with neutrons, such as radiation damage,
biological shielding, remote handling and safety. The higher the neutronicity, the
more complex are the problems associated with neutron generation. In Table 1.2
some of the important parameters for the four most important fusion reactions
are tabulated. For the D-D and D-T reactions the neutronicity is easy to calculate,
24
KAW and BANDYOPADHYAY
while for the last two reactions, where this calculation would give zero, the
values quoted are rough estimates based on side reactions that produce neutrons
in a plasma in thermal equilibrium. Sometimes, even though the main reaction
may not produce neutrons, one of the chain reactions involving one or more
constituents of the fuel or resulting ions may produce neutrons. A fusion reaction
is considered aneutronic if the overall neutron energy from all possible chains is
less than 1% of the fusion energy.
TABLE 1.2. SOME IMPORTANT PARAMETERS OF THE FOUR MOST
IMPORTANT FUSION REACTIONS
A unique aspect of the D-D reaction is that there is only one reactant, which
is an important factor when calculating the reaction rate. In any of these four
reactions, the reactants should also be mixed in the optimal proportions. This is
the case when each reactant ion plus its associated electrons accounts for half the
pressure. Assuming that the total pressure is fixed, this means that the density
of the non-hydrogenic ion is smaller than that of the hydrogenic ion by a factor
2/(Z+1). Therefore the rate for these reactions is reduced by the same factor, on top
of the differences in the values of
2
v /
e
T o . On the other hand, because the D-D
reaction has only one reactant, the rate is twice as high. Thus for non-hydrogenic
fuels, due to the fact that they require more electrons, which account for nearly
half the pressure without having any role in the fusion reaction, the reaction
rates are lower by a factor of 2/(Z+1). It is usually a good assumption that the
electron temperature will be nearly equal to the ion temperature. However, in
certain modes of operations, for example in tokamaks, there is a possibility that
the electrons could be maintained at a substantially lower temperature than the
ions. In such a case, known as the so called hot ion mode, there is no reduction
due to the charge number. Thus for D-D reactions, as any particle can react with
any other particle, there is a net gain by a factor of 2.
25
THE CASE FOR FUSION
Thus in Table 1.2 the values for reactivity are found by dividing 1.24 10
24

by the product of
2
v /
e
T o and 2/(Z+1). The reactivity is indicative of the factor
by which the other reactions occur more slowly than the D-T reaction under
comparable conditions. The Lawson criterion values are the reactivity weighted
with E
ch
and give an indication of how much more difficult it is to achieve
ignition with these reactions, relative to the difficulty for the D-T reaction.
The last column in Table 1.2 gives the fusion power density achieved in the
different reactions. Thus we see that even though the D-T reaction has the highest
fusion power density as also it is the easiest to achieve, it also has the highest
neutronicity. Thus D-T reactions need the maximum amount of neutron shielding,
have to necessarily use low activation materials and suffer from lifetime issues
for critical machine components.
On the other hand, the so called advanced fuels like D-
3
He or p-
11
B have
the great advantage of being very low in neutronicity. In the p-
11
B reactions,
for example, small amounts of neutrons and gamma radiations are indeed
generated through intermediate chains in the p-
11
B fusion as also x rays through
brems strahlung radiation. However, with careful design and selection of fuel
concentrations, it would be possible to limit the overall neutron energy to less
than 0.2% and both neutron as well as x ray and gamma doses to acceptable
levels of occupational doses through moderate amounts of shielding.
Similarly, the D-
3
He reactions also directly produce no neutrons; however,
as the optimum temperature for these reactions is more than 50 keV, the ensuing
D-D reactions (half of them) produce a triton and a proton, with the other half
producing a
3
He and a 2.45 MeV neutron. Further, the resulting D-T reactions
produce the 14.1 MeV neutrons. Although the D-D reactions can be reduced by
optimally selecting the D fuel concentration, overall neutron energy in D-
3
He
reactions cannot be reduced to below several per cent, because of which D-
3
He
reactions cannot be considered as aneutronic.
However, the major problem for advanced fuels is that one needs to heat
the fusion plasma to much higher | (thus higher n and T) values compared to
the D-T plasmas, as also much higher confinement times so as to achieve a high
enough Lawson factor. This may pose the major road block for fusion reactors
based on advanced fuels, unless novel concepts and designs for stable config-
urations emerge to make that possible. Moreover, since most of the energy in
D-T fusion is extracted from the neutrons that are captured in the surrounding
blankets, it has the advantage that one has to extract heat volumetrically from
the reactor blankets. On the contrary, to have reactors based on the low (or zero)
neutronicity fusion reactions, one has to extract very large amounts of surface
heat heat that is conducted to the divertor targets by the plasma particles. Of
course at the high temperatures of these fusion reactions, a significant amount
of the plasma energy is also radiated out through synchrotron radiation in the
microwave range and it should be possible to develop technologies to trap this
26
KAW and BANDYOPADHYAY
microwave radiation and convert it into electricity. The x rays emitted through
bremsstrahlung radiation, on the other hand, can be absorbed in layers of metal
foils. Nevertheless, the divertor targets for the aneutronic reactions would have
to bear an unprecedented amount of heat load, and great strides in ultra-high heat
flux material development and very significant R&D in other related technologies
will be necessary. However, not enough research has gone into exploring these
avenues to totally rule out advanced fuel based fusion devices in the future. Thus,
even though initial fusion reactors will be based on D-T fusion because it is easily
achieved, future fusion devices may use more innovative fuels.
1.3.5. Direct conversion to electricity
Since many of the fusion reactions give energy to charged particles, it is
possible to convert fusion energy directly to electricity through various methods
such as induction effects or through electrostatic effects decelerating charged
particles in an electric field. It is true that in the first generation of D-T reaction
based fusion reactors, as 80% of the fusion energy goes into the neutrons,
electricity generation will have to be mostly through conventional steam turbines
using the heat generated by the neutron absorption. However, in future fusion
devices based mainly on aneutronic fusion like p-
11
B, it would be possible to
convert about 80% of the charged particle energy directly into electricity. The
fraction of fusion energy which goes into the charged particles also gets released
in the form of microwave radiation through synchrotron radiation and x rays
through bremsstrahlung radiation. The microwaves can be used either to drive a
current in the plasma itself (as required in the case of tokamaks) [1.52, 1.53] or
suitable technology can be evolved to capture the microwave energy and convert
it into electricity. The x ray energy can also in principle be converted into
electricity through photoelectric effects by making the rays pass through arrays
of conducting foils. However, due to deep penetration of the x rays, this might
need many layers of metallic foils to absorb all the energy and would require
clever reactor designs [1.54].
1.4. APPROACHES TO FUSION
To achieve fusion on Earth, one has to create a plasma of the fusion
reactants of sufficiently high temperature and density, and also hold it confined
for a sufficiently long time away from any surrounding material walls. There
are two main approaches for achieving this: magnetic confinement and inertial
confinement. Magnetic confinement fusion is the more developed of the two
approaches and is presently the most promising path to developing future fusion
reactors based on this concept, especially as they can confine the plasma in a
steady state for long durations. The ITER device, presently under construction at
27
THE CASE FOR FUSION
Cadarache in France, will be the first fusion reactor producing 500 MW of fusion
power, 10 times more than the input auxiliary heating power. It would also have
plasma discharges of 3000 s duration in the non-inductive phase of operations, a
first step to steady state operations.
The inertial confinement fusion (ICF) approach, on the other hand, works
primarily in a pulsed fashion, achieving thermonuclear fusion through micro-
explosions of reactant targets induced by high power laser or particle beams at
a high repetition rate. The two largest ICF experiments in the world presently
are the National Ignition Facility (NIF) at the Lawrence Livermore National
Laboratory in the USA, which was commissioned in March 2009, and the Laser
Megajoule (LMJ) facility in France, scheduled for construction completion in
2014. Both these facilities are designed to achieve ignition conditions. However,
given the rather limited driver conversion efficiency (~10%) of the lasers and the
conversion efficiency of ~3540% of fusion to electric power needed for power
production, ICF facilities would need a fusion gain of about 100200 for net
power production. Hence, ICF reactor systems for power generation seem to be
more challenging.
1.4.1. Magnetic confnement fusion
In magnetic confinement fusion, one utilizes the electromagnetic properties
of the charged plasma particles to trap them away from material walls in a
specifically designed magnetic field configuration called a magnetic bottle. This
exploits the ability of a steady magnetic field to dramatically restrain the motion
of the charged particles in a plasma across the magnetic lines of force, while
allowing them to move freely along them. Research and development in the
area of magnetic confinement fusion has given rise to a variety of confinement
configuration concepts.
The most successful way of trapping the plasma particles along the
magnetic field lines has been achieved by magnetic confinement fusion based on
the tokamak concept. In a tokamak, the plasma is formed in the shape of a torus
or a doughnut, through a specially designed magnetic field configuration. Such a
configuration can be formed, for example, as shown schematically in Fig. 1.12,
by turning around a solenoidal set of ring-like coils, in the form of a torus, thus
forming a toroidal magnetic field. However, a purely toroidal magnetic field
cannot confine the plasma particles as the curvature of the field lines produces
opposite particle drifts for the ions and the electrons, leading to a charge
separation; the consequent electric field leads to a rapid loss of the plasma to the
walls. To prevent the charge separation, one needs an additional component of
the magnetic field turning around on the minor cross-section of the torus the
poloidal magnetic field. The poloidal magnetic field in tokamaks is formed by
passing a toroidal current in the plasma itself. The resultants of the toroidal and
28
KAW and BANDYOPADHYAY
poloidal magnetic field lines spiral infinitely around the torus in nested toroidal
surfaces forming globally closed magnetic flux surfaces. The pitch of the field
lines is defined as the safety factor q, so called because of its immense importance
in determining MHD plasma stability. Such a toroidal plasma configuration under
toroidal and poloidal magnetic fields, however, experiences a hoop force, trying
to expand the plasma torus radially outward. To achieve radial equilibrium, one
has to provide vertical magnetic fields through vertical field coils, which are
typical current carrying rings concentric with the torus.
The combination of the toroidal and poloidal magnetic fields leads to the
concept of toroidal magnetic bottles in which one can confine the hot plasma
far away from material walls in a stable way for long durations. In these nested
magnetic surfaces, the hot plasma in the core is thermally insulated from the
colder plasma near the edge. The charged particles can move across the magnetic
surfaces only by collisions with other charged particles or through turbulence.
However, they typically cover a distance millions of times the length of the vessel
before they reach the material walls and are lost. Detailed reviews of tokamaks
as well as other toroidal magnetic confinement configurations can be found in
several publications [1.551.59].
FIG. 1.12. Schematic of a tokamak.
The plasma current in most tokamaks is driven inductively through
transformer action. The primary of a central solenoid is placed concentrically
29
THE CASE FOR FUSION
at the centre of the torus, with the plasma itself (being made of charged particles)
acting as the secondary. However, inductive effects being inherently transient
in nature, such tokamak plasma current drive is also transient. Steady state
tokamak discharges can be achieved by driving plasma current non-inductively
by several means. Injection of high power energetic neutral particle beams
and/or of radiofrequency waves at characteristic resonant frequencies, imparting
momentum selectively to the ions or electrons, has so far proved very successful
in driving currents in tokamak-like devices. A significant part of the plasma
current (theoretically 100%) can also be driven by the self-generated bootstrap
current the name given to a current produced by radial diffusion in the
plasma itself in the presence of density gradients through momentum exchange
between trapped and passing particles. Since the plasma pressure profile peaks
around the tokamak centre and the bootstrap current is proportional to the radial
gradient of pressure, the bootstrap current is also necessarily off-axis and zero
at the centre. Hence some additional current drive is still required to drive the
current at the plasma centre in largely bootstrap current driven devices. Future
advanced tokamaks are being designed in which most of the plasma current will
be bootstrap driven, so as to minimize the energy spent in auxiliary current drive.
In magnetic confinement configurations, the magnetic forces have
to act against the plasma pressure to confine the plasma. The maximum
magnetic forces that one can apply today in such configurations are limited by
technological constraints: the maximum magnetic field strength achieved today
in cryogenically cooled superconducting coils is about 16 T (on the surface of
the conductor), although in future, development of high Tc superconductors
holds the promise to exceed this limit. The plasma pressure, on the other hand,
is proportional to the product of the density and the temperature. The magnetic
field applies a magnetic pressure on the plasma and the ratio of the plasma
pressure to the magnetic pressure is generally referred to as the plasma beta
(
( )
2
0
/ / 2 nT B , B being the magnetic field strength, n and T being the
plasma density and temperature). The plasma has great importance regarding
the stability of magnetically confined plasmas. The maximum values
achieved in magnetic fusion plasmas are around a few per cent, implying that the
plasma pressure is always much smaller than the magnetic field pressure. Since
the temperature required for achieving fusion reaction is fixed (T ~ 10 keV), and
the plasma pressure must be low, the plasma density must fall in a certain range
which is less than one billionth of solid density. The Lawson criterion, on the
other hand, requires a confinement time of the order of a few seconds. (On the
timescale of single particle motion in the confining magnetic field, one second is
a very long time!)
The magnitude of achieved is essentially limited by the efficiency of
the means by which one heats the plasma to high temperatures and/or by the
efficiency of the magnetic bottle how well it is able to retain the thermal energy
30
KAW and BANDYOPADHYAY
that the plasma is fed with. The plasma can be heated by Joule heating by driving
internal plasma currents (ohmic heating as in tokamaks, reversed field pinches,
etc.), which is, however, limited to temperatures of up to about 1 keV. This is
because, as the plasma resistivity decreases with increasing temperature, ohmic
heating becomes ineffective. Further heating to fusion relevant temperatures of
tens of keV is possible by either injecting high energy neutral particle beams
of isotopes of hydrogen (NBI) across magnetic fields, or by injecting high
power radiofrequency (RF) waves with frequencies which resonate with natural
plasma frequencies such as the ion cyclotron (ICRF), electron cyclotron (ECRF)
or lower hybrid (LHRF) frequencies. However, even though all these auxiliary
techniques have been successful in heating the plasma to fusion relevant
temperatures, auxiliary heating has also led to significant degradation of the
energy confinement. This happens because the confinement is largely determined
by turbulent processes in plasmas, which are significantly amplified by the
heating processes. The energy confinement time, which is dependent on several
machine and plasma parameters, crucially increases with increase in the plasma
major radius, i.e. the machine size. This has necessitated the construction of very
large devices like ITER to achieve 1 Q > operation, almost twice as large as any
of the existing tokamaks.
However, over the past few decades, rapid strides have been made in the
improvement of plasma confinement in tokamaks by either shaping of the plasma
poloidal cross-section or operating the tokamaks in novel regimes of parameter
space. For example, with the aim of achieving higher

and higher plasma
current in tokamaks, it has been found that the optimum configuration is to have a
non-circular plasma configuration, typically a D-shaped one, characterized by an
elongation and a triangularity. Such a D-shaped plasma cross-section is provided
by suitably shaping the poloidal magnetic field configuration by an additional set
of poloidal field (PF) coils, which are also concentric with the torus. Moreover, it
has been found that the energy confinement time critically depends on the plasma
properties near the edge region of the tokamak plasma, and hence the edge
region in tokamaks has to be handled with care. In the next subsection we will
see evidence of the tremendous improvement in the confinement properties of
tokamak plasmas by a careful control of the plasma edge. In the first generation
of tokamaks, the plasma edge used to be limited where the outermost boundary
is restricted by a material limiter schematically shown in Fig. 1.13, with which
the plasma is in contact. However, in such configurations, a considerable amount
of impurity is released into the plasma due to erosion of the material limiter
caused by bombardment of the hot plasma particles on it, which leads to large
impurity radiation losses. A much better configuration in terms of energy and
particle exhaust was achieved when the plasma cross-sections became diverted,
where the outermost magnetic field flux lines are opened up to make them
strike on a chosen divertor target. In this configuration, the divertor bears most
31
THE CASE FOR FUSION
of the thermal loads carried by the plasma particles transporting across the last
closed magnetic flux surface, which then flow along the open field lines on to
the divertor target. One can reduce the energy of the impinging particles by
suitably maintaining the density in the divertor region, i.e. the region close to the
divertor target, shown schematically in Fig. 1.13. Even though modern tokamaks
use divertor targets, the plasma is still initiated from a limited configuration and
then gradually brought to a divertor configuration through plasma shaping by
changing appropriately the poloidal field coil currents. Strong pumping of the
exhaust from the divertor region can prevent the impurities from going back to
the plasma. Thus the deleterious effects of the impurities in plasma confinement
could be reduced in such configurations. Divertors are also very efficient in
removing the helium ash from the plasma.
FIG. 1.13. Vertical cut section of the tokamak schematic shown in Fig. 1.12, showing the major
tokamak components and parameters.
However, tokamak plasmas are subject to various kinds of instabilities and
perform reliably only when operated within certain boundaries in the parameter
space, for example below certain density or current limits. Close to or above
these limits, the plasma current generally suffers sudden disruption due to
internal plasma instabilities. This leads to large induced currents and undesirable
electromagnetic forces on the surrounding machine components. However, many
methods have now been found to increase the domain of reliable operation of
tokamaks, avoiding plasma disruptions.
In this context, it is worth mentioning the significant progress in research
with the other major toroidal confinement concept, namely the stellarator. In
stellarators, depicted schematically in Fig. 1.14, the poloidal magnetic field is
32
KAW and BANDYOPADHYAY
produced by careful profiling of the magnetic field topology through complex
non-planar toroidal field coils. The stellarator is toroidally non-axisymmetric
(helical), and closed toroidal magnetic surfaces are formed in three dimensions
by fields entirely produced by the externally wound coils. Thus there is no need to
drive plasma currents in stellarators, which eliminates the danger of macroscopic
disruptive instabilities and makes them very attractive for reactor concepts. Work
on large stellarators such as LHD in Japan [1.60] and Wendelstein-7x [1.61]
in Germany and on innovative stellarator concepts in the USA are especially
encouraging since stellarators form the most advanced toroidal confinement
concept after tokamaks. Some of the major shortcomings of stellarators are the
complex technology of large coils producing 3-D magnetic configurations, poorer
(than tokamak) particle confinement properties due to large deviations of particle
orbits from 3-D flux surfaces, the associated intense neo-classical transport of the
plasma and alpha particles and the as yet relatively insufficient database on the
turbulent flux of heat and particles.
FIG. 1.14. Schematic of stellarator magnetic field coils and plasma configuration (courtesy of
Ref. [1.62]).
1.4.1.1. Progress in tokamak based magnetic confnement fusion research
Overall, tokamak based magnetic confinement fusion has registered
impressive progress in key areas of reactor relevance in the last 1015 years.
Significant improvement of the transport properties, operating density and
stability of the tokamak plasmas have been achieved in this period. A number
of improved confinement regimes were discovered; as a result, the fusion triple
product
E
nTt and
DT
Q equivalent have increased to about 1.5 10
21
keVsm
3
and about 1.25 respectively in separate plasma discharges in JT-60U. Many D-T
experiments have been conducted on JET and TFTR, which has produced fusion
33
THE CASE FOR FUSION
powers of 16 MW and 11 MW respectively in these two machines (see Fig. 1.15),
with the JET device achieving
4

DT
Q ~ 1 in 50/50 D-T operations.
FIG. 1.15. Fusion power produced in JET and TFTR experiments [1.63].
FIG. 1.16. Progress in achieved central ion temperature and fusion triple product.
4
The JET 16MW discharge actually achieved a maximum
DT
Q

of 0.64. However,
this was in a transient discharge, which in steady state conditions would indeed amount
to about unity if one takes into account the factor of / dW dt in the fusion gain calculation
( / ( / )
DT fusion aux
Q P P dW dt .
34
KAW and BANDYOPADHYAY
FIG. 1.17. Development of tokamak based fusion research measured in terms of the fusion
product of fuel density times energy confinement time (nt) as compared to semiconductor chip
development (Moores Law) as well as that of accelerator research. Courtesy of J.B. Lister
(CRPP-EPFL, Lausanne, Switzerland), and also Ref. [1.64].
It is worth considering here the progress of magnetic confinement fusion
research over the last about half a century depicted in Fig. 1.16 in terms of the
fusion triple product achieved in various tokamak devices. Since the start of
tokamak experiments in the 1960s, this product has increased by more than three
orders of magnitude. In fact the progress of magnetic confinement fusion research
based on the tokamak concept has been faster than the Moores Law (Fig. 1.17)
which is traditionally applied as a measure of development of semiconductor
chips.
Thus, tokamaks are today the most advanced of the toroidal confinement
systems, which makes them the most promising candidate for the first generation
of fusion reactors. This has culminated in the start of construction of the ITER
device (Fig. 1.18) based on the tokamak concept. Scientists and engineers from
the European Union, China, India, Japan, the Republic of Korea, the Russian
Federation and the USA have come together in an unprecedented international
collaboration to build the ITER device at Cadarache, in the south of France.
ITER is aimed at demonstrating the scientific and technical feasibility of fusion
reactors and intends to operate largely at Q = 10 conditions, while the possibility
of achieving controlled ignition operating conditions is also foreseen. It is to
be noted here that it is not essential for fusion reactors to operate at ignition
conditions. Considering that reactors would have an overall energy efficiency
(ratio of final electric energy pumped to the grid to the fusion energy produced
in the devices) of about 35%, operation with Q between 20 and 40 would be
sufficient for fusion reactors to be a viable energy option.
35
THE CASE FOR FUSION
FIG. 1.18. Schematic of the ITER tokamak [1.10].
1.4.2. Inertial confnement fusion
Inertial fusion is a completely different approach to a fusion reactor. If a
pellet of solid fusion fuel is compressed to very high densities and a portion of
it is heated to the temperatures required for ignition, it is possible to generate
net fusion power before the pellet blows itself apart in a micro-explosion. In
this approach confinement is not achieved by external fields, but by the inertia
of the hot fuel that keeps it together for a finite time (hence the name inertial
confinement). Compression is accomplished by heating just the surface of a
spherical hollow shell target uniformly from all sides, whereby the intense
heating of the surface creates an inward implosion of the fuel as the surface layer
ablates and explodes outward. The driver of the compression process can be
either a giant laser pulse or a pulsed heavy ion beam from a particle accelerator.
Let us first consider here a few elementary aspects of inertial fusion; details
may be obtained in Refs [1.65, 1.66]. In order to reuse the containment chamber
of the micro-explosion again and again, the typical fusion output energy must
be about 100 MJ; if this is repeated with a frequency of a few Hz, we have a
fusion reactor of several hundreds of MW fusion power. Working out the size of
a typical D-T fusion pellet at cryogenic (solid) densities and fusion temperatures
(~10 keV) which would produce this much fusion energy by complete burning,
we get a pellet radius of about 1 mm. The typical disassembly time of the pellet is
the time it takes a rarefaction sound wave to travel from the surface of the pellet
to its interior and is of the order of a fraction of a nanosecond in a D-T plasma at
fusion temperatures. The gain factor
f
c b
in
E
G f
E
n
| `


. }
(1.12)
36
KAW and BANDYOPADHYAY
depends on the coupling efficiency (
c
n ), the fractional burn ratio (
b
f ) and the ratio
of the maximum fusion energy output (
f
E ) to the input energy content (
in
E ) and
can be maximized by optimizing each of these ratios. Its desirable value is given
by the condition of low circulating electric power (~25%) in the reactor, giving
10
d
G n > (or G = 70200 for a 515% driver efficiency
d
n ). The fractional
burn ratio of the pellet is determined by the ratio of fusion reaction time and the
disassembly time and is given by
( )
b
f
R
f
H R
p
p

+
(1.13)
where p is the fuel density and (8 )/ v
f f s
H m c o < > , where
s
c is the sound
speed given by 2 /
s B f
c k T m and
f
m is the average mass of the fuel ions
(for an equimolar D-T plasma, 2.5
f p
m m ,
p
m being the proton mass) [1.66].
Thus the factor R p of the pellet determines the fractional burn ratio. For inertial
fusion, the basic ignition condition is attained by heating a fuel with a R p of
about 4 kgm
2
to a minimum temperature of about 5 keV, corresponding to a
fractional burn of about 3.5%. However, ignition alone is inadequate for inertial
fusion, which must additionally overcome a factor of ~10 in hydrodynamic
coupling efficiency and another factor of ~10 in driver efficiency. To overcome
these inefficiencies, ICF capsule implosions will require a global fuel R p of
approximately 30 kgm
2
, resulting in a fractional burn of about 30%. This R p
criterion in inertial fusion is like the Lawson criterion in magnetic fusion. It can
obviously be achieved by having either a low density, large radius configuration,
or a high density, small radius configuration. Simple estimates can be made that
show that the required driver energy scales approximately as the inverse of the
square of the fuel density. Therefore, in order to minimize the driver size required
for inertial fusion, compression of the D-T fuel to very high densities (about
200400 g.cm
3
of D-T density for a mm size pellet, which is about 10002000
times the normal solid density) will be required. It also turns out that energetically
it takes an order of magnitude less energy to compress the fuel to the required
densities (where matter is degenerate) than it takes to bring the whole assembly
to fusion ignition temperatures. The strategy which is therefore adopted is one
where the entire matter is compressed to required densities but only a small hot
spot is heated to ignition temperatures. The size of the hot spot is chosen in such
a manner that the alpha particles produced in the hot spot can come out and
heat neighbouring layers of the fuel, leading to the propagation of a burn wave
producing fusion reactions throughout the whole fuel.
Schematically the four important phases associated with a pellet implosion
are as follows: as the driver energy impinges on the initially solid matter of
the capsule, this material is transformed into a plasma with a temperature of
37
THE CASE FOR FUSION
the order of a few keV, the capsule is ablated and the heated plasma expands
outward with a velocity of about 1001000 kms
1
. As a reaction, the remaining
portion of the spherical capsule is accelerated to about 300500 kms
1
towards
the centre, compressing and heating (via mechanical work) the thermonuclear
fuel. If, at the culmination of the implosion, specific conditions associated with
the compressed fuel configuration are attained, thermonuclear ignition at the
centre and subsequent propagation of a thermonuclear burn wave to the whole
fuel will occur, leading to large burnup. The final density achieved by the fuel
during a capsule implosion depends on: (1) the external driving pressure, which
must reach 10
8
atm; (2) the resistance of the target matter to compression (low
isentrope should be kept); and (3) the development of hydrodynamic instabilities
and asymmetry during the implosion process. The external pressure is determined
by the ablated material and the incident driver energy flux. Although compression
is energetically attractive and reduces the driver size required for efficient fuel
burnup, high gain also requires a high temperature region (referred to as the hot
spot) for ignition to occur. The hot spots form during the final stages of the inertial
confinement fusion capsule implosion, provided that hydrodynamic instabilities
do not mix them with the surrounding cold fuel. For D-T fuel, the capsule can
be considered as transparent to the 14.6 MeV neutrons produced in the fusion
reaction (especially near the time of ignition). Therefore, the self-heating of the
hot spot is done predominantly by the alpha particles released from the D-T
reactions and by the work by the implosion. For effective self-heating, the areal
density of the hot spot must be comparable to the alpha particle range, which
turns out to be 0.3 gcm
2
at a temperature of 10 keV. The formation of a hot
spot within the cold main fuel is the key scientific issue for inertial fusion. A
crucial requirement for the generation of the central hot spot is a high degree
of symmetry of the spherical implosion, which demands uniform irradiation and
control of certain hydrodynamic instabilities.
Another concept in inertial fusion systems, which has been quite active
during the last decade or so, is the heavy ion accelerator driven inertial fusion
system. In this concept the laser driver is replaced by a heavy ion beam, which
can be produced with a much higher electrical efficiency and is also excellent
at heating solid density plasmas by classical Coulomb collisions. It also tries to
eliminate a major concern about neutron contamination of final focusing optics
in laser driven configurations by the use of magnets in accelerator driven config-
urations. Earlier concerns were related to transport of the beams over significant
distances, problems due to space charge effects, etc. The recent focus of these
experiments is towards reduction of target chamber size and hence the cost of the
experiments, for example by the use of liquid protected target designs allowing
highly compact target chambers. The compact target chambers allow focusing
magnets to be placed closer to the target, resulting in reduced focal spot size and
hence a smaller driver, which finally results in a large cost reduction.
38
KAW and BANDYOPADHYAY
It is instructive to make a comparison between fusion plasmas in magnetic
confinement systems and those encountered in inertial fusion energy systems.
Inertial fusion plasmas produce energy in a pulsed manner whereas magnetic
confinement plasmas are essentially steady state. This is why the fusion gain
in inertial fusion is expressed in terms of ratio of total fusion energy release to
input consumed energy (and not as a ratio of output to input power). In magnetic
confinement fusion, the particle densities are very small, ~5 10
20
m
3
in fusion
grade devices. In comparison, air density at atmospheric pressure is about
5 10
25
m
3
. Thus, the plasma in a magnetic confinement device is like a vacuum
for all practical purposes! This is understandable. In magnetic fusion the plasma
pressure is contained by a ten times larger magnetic pressure and since the
containment fields are limited to a few tesla by the present technology limits of
superconducting magnets, the plasma pressure is limited to 10 atm and the plasma
density to near high vacuum densities of the order of 5 10
20
m
3
. On the other
hand, the ICF plasma produces enormous pressures of the order of 10
12
atm and
particle densities of the order of 5 10
31
m
3
. In magnetic confinement, particle
and energy confinement times have to be in the range of 110 s, which is indeed
a very long time compared to single particle gyration times in a magnetic field.
It is determined by many complicated factors such as geometry and turbulent
processes in the plasma and is often impossible to predict or even determine
empirically. On the contrary in inertial fusion, the confinement times required
are very small, of the order of a fraction of a nanosecond, and are essentially
guaranteed by virtue of inertia.
1.4.2.1. Progress in inertial confnement fusion research
The results of ICF experiments performed over the last two decades have
formed the basis for the design of the ignition experiments National Ignition
Facility (NIF) and Laser Megajoule (LMJ).
NIF, currently in operation at the University of Californias Lawrence
Livermore National Laboratory, is a stadium sized facility containing a 192 beam,
1.8 MJ, 500 TW, 351 nm laser system together with a 10 m diameter target
chamber with nearly 100 experimental diagnostics. NIF is being built by the
Department of Energy, United States National Nuclear Security Administration,
and when completed will be the worlds largest laser experimental system,
providing a national centre to study inertial confinement fusion and the physics
of matter at extreme energy densities and pressures. NIFs 192 laser beams will
compress fusion targets to conditions where they will ignite and burn, liberating
more energy than required to initiate the fusion reactions. NIF experiments will
allow the study of physical processes at temperatures approaching 100 million K
and 100 billion times atmospheric pressure.
39
THE CASE FOR FUSION
NIF was fully commissioned in March 2009 when its 192 laser beams
delivered a record of 1.1 MJ of laser energy in the target chamber. By
December 2008 the NIF team had already delivered 1.005 MJ of 1u infrared
and 626 kJ of 3u ultraviolet energy to the target chamber. (NIFs main laser
produces laser light in the infrared range of frequencies, which is commonly
known as 1u. The final optics assembly of NIF converts this laser frequency
to green light (2u) and finally to ultraviolet (3u). The ultraviolet laser light
has the maximum efficiency in compressing NIFs fuel capsules.) In these
commissioning activities, 540 kJ of 3u energy was produced in 96 beams
exceeding the project completion energy criterion. During this commissioning
process, 207 TW of laser beam power has been injected into the Target Chamber
Center in a 2 ns pulse in 96 beams, meeting the 200 TW peak power requirement.
The LMJ facility under construction in France is the largest inertial
confinement fusion facility being built outside the USA and is designed to deliver
on target 1.8 MJ and 550 TW of 350 nm ultraviolet light. Completion of the LMJ
facility is expected in 2014 (originally 2008) and the system will be able to deliver
600 shots per year. The target chamber will be 10 m in diameter and a (cryogenic)
high precision target holder is being developed. A prototype experiment (Ligne
dIntegration Laser, LIL) consists of a group of 8 laser beamlines (out of 240 for
LMJ) and a 4 m diameter target chamber. It started to operate in 2002. First results
have been obtained on one beamline of LIL with pulses of 18 kJ at 1053 nm and
7 kJ of ultraviolet, shaped as a 20 ns prepulse followed by a 3 ns main pulse, all
in accordance with the expectations. The LMJ experiment is complemented by
vigorous efforts in numerical simulation, increasing the present computational
capability by a factor of 20.
However, as mentioned before, due to the poor driver efficiency
(conversion from electric to laser energy), conventional ICF experiments would
need very large fusion gains (~100) to be attractive for power generation.
Nevertheless, a more recent development of the fast ignition concept has the
promise to dramatically reduce the incident laser power requirement, which has
again made ICF experiments interesting for power generation. In this approach,
the heating and compression phases of the implosion are decoupled the target
is first compressed in the conventional way using a driver laser system, and then
when the imploding target reaches maximum density, a second ultra-short pulse
ultra-high power petawatt (PW) laser delivers a single pulse focused on one side
of the core, starting the ignition process. A number of experiments are currently
under way to explore the fast ignition concept, the most prominent being the
GEKKO xII facility in Japan and the proposed HiPER facility of the EU. If the
fast ignition approach becomes successful, it could dramatically reduce the total
energy required to be delivered at the target. For example, whereas NIF uses
192 ultraviolet laser beams of a total energy of 1.8 MJ, the driver of the proposed
40
KAW and BANDYOPADHYAY
HiPER facility would be 200 kJ and the heater 70 kJ, yet the predicted fusion
gain in HiPER is even higher than in NIF.
1.5. SOCIOECONOMIC PERSPECTIVES
We now turn to certain socioeconomic issues which are relevant to the
acceptability of developing thermonuclear fusion as an alternative energy source
of the future. These issues are related to public perception of the environmental
impact of the normally operating reactor as well as a reactor subject to potential
accident scenarios, cost comparison with other energy alternatives and spin-offs
to other fields of interest.
1.5.1. Environment, safety and non-proliferation
1.5.1.1. Emissions in normal operation
The safe confinement of fusion reactor inventories and the minimization
of releases during normal operation, possible accidents, decommissioning and
storage of waste, are major objectives in fusion power plant design. Besides
tritium, the other source of radioactivity in the plant is the intense flux of fusion
neutrons penetrating into the material surrounding the plasma and causing
activation. Fusion power stations will have an extremely low density of fuel
inventory inside the chamber holding the fusing plasma; hence their power
production stops a few seconds after fuelling is stopped. Fusion reactors are not
associated with any greenhouse gases and they would have low levels of residual
power density arising from the decay of neutron activated materials in their
structure after the termination of burn. These favourable generic features lead
to substantial safety and environmental advantages, but the full exploitation of
these advantages will depend upon the details of design and materials selection.
A number of reactor studies have been carried out, especially in the EU, the USA
and Japan. They generally differ substantially in their gross power, major radii,
aspect ratio and power density, etc., although detailed safety and environmental
analyses have been performed for most of them. The foundations of all the
analyses of safety and environmental impacts were based on detailed and
comprehensive calculations of neutronics, activation and derived quantities.
Three confinement barriers are foreseen: vacuum vessel, cryostat and outer
building. Small fractions of the radioactive materials are released during normal
operation. The amounts depend strongly on design characteristics such as cooling
medium, choice of structural materials and blanket design. Several studies on
normal operation doses to the general public have been performed. According
to one such study, presented in detail in Ref. [1.67], the releases during normal
operation for two different plant models are summarized in Table 1.3. The
41
THE CASE FOR FUSION
expected doses to the most exposed individuals of the public would stay well
below the internationally recommended limits [1.42] and amount to less than 1%
of the naturally occurring dose rate.
TABLE 1.3. DOSES TO THE PUBLIC DUE TO NORMAL OPERATION
EFFLUENTS FOR TWO DIFFERENT FUSION PLANT MODELS
1.5.1.2. Possible accidents
Detailed accident scenarios have been analysed in which different methods
(bottom-up and top-down) are applied to guarantee a complete list of the
accident sequences. Reactivity excursions are for several reasons not possible
in a fusion power plant. Therefore, the most severe accidents are typically
related to failures of the cooling system or confinement systems. Even without
any cooling, temperatures in the structural materials will stay well below the
melting temperature and keep the confinement barriers intact. During the most
severe internal accident in a fusion plant the expected dose to the most exposed
individuals of the public would stay in the range of the annually occurring natural
radiation (~1 Sv) [1.671.70].
As a worst case scenario it was assumed that the complete vulnerable
tritium inventory (roughly 1 kg) of the fusion plant is released at ground level.
The initiator of such an accident could only be very energetic outside events
such as an airplane crash on the plant. Even if the worst weather conditions are
assumed, only a very small area, most likely within the perimeter of the site,
would have to be evacuated [1.34, 1.68, 1.69].
1.5.1.3. Waste
All the radioactive material produced in a fusion plant is neutron induced.
Detailed analysis of the amount and composition of the fusion power waste has
been performed [1.15, 1.71, 1.72]. The time evolution of the radiotoxicity of
the waste is shown in Fig. 1.19. The plant model assumed is based on available
materials. The figure shows a rapid decrease in radiotoxicity once the plant is
42
KAW and BANDYOPADHYAY
shut down. The time evolution of the fusion waste is compared with the time
evolution of the waste from a PWR fission plant and with the radiotoxicity of ash
in a coalfired power plant. The radiotoxicity of the waste of fission plants hardly
changes on the timescale of a few hundred years.
FIG. 1.19. Development of radiotoxicity for a fusion plant, a fission plant and the ash of a coal
plant. It is assumed that all the plants produce the same quantity of electricity. The volume of
coal ash is of course 23 orders higher [1.71].
1.5.2. Cost comparison with other sources of energy
Power production technologies based on fusion belong to the class with
low external costs, comparable with solar and wind based technologies. In the
last decade, several models have been developed to estimate the cost of power
produced by fusion based technology. Like any other conventional source of
energy, a fusion power source would involve two kinds of costs.
1.5.2.1. Direct costs of fusion power production
The direct costs comprise the contributions from constructing, fuelling,
operating and disposing of the power plants. The estimation of direct costs
43
THE CASE FOR FUSION
requires a deep understanding of fusion physics and technology, hence this work
is generally performed by fusion experts. However, the economic methodology
and calculations applied by them are standard, as recommended by several
international agencies, and have been verified by non-fusion experts. On the other
hand, the external costs have mainly been estimated by non-fusion experts. One
such estimate has been carried out by EFDA-SERF [1.69], derived by a computer
model, dividing the plant into about eighty systems that are individually costed.
This costing model was benchmarked against the costs of the proposed large
fusion experiment ITER, which have been derived in very detailed studies,
mainly conducted by industry, and in independent, parallel efforts by the ITER
Parties. The total capital cost, including that of interest rates during construction,
was combined with estimated replacement costs, other operating costs, payments
into a decommissioning fund and assumptions on availability, to obtain the
cost of electricity. This was done in a standard manner, the levelized cost
methodology, which is used for example in OECD and IAEA studies. Figure 1.20
shows the breakdown of the cost of electricity resulting from a typical set of
assumptions. The prediction of the absolute cost of fusion generated electricity in
more than fifty years from now requires in particular a realistic assessment of the
uncertainties. A large number of calculations were therefore carried out, covering
ranges of assumptions for the key physics and technology parameters affecting
the economic performance of a power plant.
FIG. 1.20. Composition of the cost of electricity from a fusion power plant. Fuel costs, which
are very small, are included under operation and maintenance. Reprinted from Ref. [1.69].
Copyright (2010), with permission from Elsevier.
There have been several estimates of the cost of electricity (CoE) of fusion
electricity. The major cost of a fusion power plant is the cost of construction as
44
KAW and BANDYOPADHYAY
the fuel cost would be negligible in comparison. The target cost of construction
according to these estimates is about 22003750 euro/kW(e) (kW(e)) = kilowatts
electrical), while the target normalized CoE (ratio of cost of fusion electricity to
the cost of electricity of a coal plant of the same power capacity) is about 0.71.5
[1.37]. Thus even though the expected maximum target direct cost of fusion
electricity is about 50% greater than those of coal (without emission abatement
costs) or fission, it becomes competitive with clean coal when emission
abatement costs are included. Renewables vary greatly in their cost prospects,
their range, and their capability to contribute to baseload power production.
Fusion electricity is found to be competitive in production costs with typical
renewables, but can provide, in contrast to most of the latter, steady baseload
power without additional energy storage costs. However, it should be noted that
current projections may not be adequately dealing with certain realities which
influence real costs; these are: likely availability of fusion power plants affecting
real costs, likely political decision making influencing real schedules, likely
economic and regulatory environments strongly affecting introduction of fusion
energy, etc.
FIG. 1.21. Learning curve for fusion power (for 1 GW(e) plants). (Figure adapted from
Ref. [1.71] assuming first commercial fusion power plant around 2045.)
Moreover, the collective construction and operation experience is expected
to lead to considerable cost reduction owing to accumulated learning processes
[1.34, 1.71]. Learning curves describe the correlation between the cost reductions
and the cumulated installed capacity. The slope of the curve the so-called
progress ratio gives the cost reduction for a doubling of the capacity. A
45
THE CASE FOR FUSION
progress ratio of 0.8 is assumed for the novel components in the fusion core. This
ratio is well within the values generally experienced in industry; possible physics
progress is also included. Figure 1.21 shows the expected cost development with
time [1.71, 1.73]. Thus, the cost evolution of the total fusion plant suggests that
the initial cost of fusion powered electricity would be relatively high, mainly
because of the high cost of construction of the reactors. However, with continued
effort to develop more cost effective technologies and accumulative learning,
it is likely to come down in the early part of this century to within the target
cost, making the cost of fusion electricity close to those from other conventional
sources and hence affordable.
1.5.2.2. External costs of fusion power production
The direct costs of electricity generation, calculated as described above, do
not include costs such as those associated with environmental damage or adverse
impacts upon health. In the case of some present sources of electricity, these
external costs are substantial.
A methodology for evaluating, in a standardized way, the external costs
of electricity generation by different fuel cycles had been previously developed
for the European Commission in the framwork of the ExternE project, which
involved more than 40 different European institutes from nine countries [1.74].
The ExternE methodology is a bottom-up methodology, with a sitespecific
approach, considering the effects of an additional fuel cycle located in a
specific place. It assesses the entire life, fuel cycle and closure of a power plant.
This includes materials manufacturing, construction, operation of the plant,
dismantling, site restoration and disposal of waste. At each stage, factors such
as hazardous chemical or radioactive emissions, road accidents, occupational
accidents, accidents at the plant exposing the public to risks, and occupational
exposure to hazards are considered. Even effects such as the destruction of visual
amenity are evaluated. The adverse effects are quantified in monetary terms
and summed to produce an estimate of the total external costs. This implies
that goods which are usually neither traded nor priced need to be valued. The
approach involves, evidently, significant uncertainties in such issues as the impact
of CO
2
emissions on climate and the doseresponse relationships, and raises
controversies in the economic valuation of adverse health effects and deaths. It
also poses the question of how to cost, at the moment of power generation, health
consequences arising in the relatively distant future. For the present studies, such
costs were attributed in full as if arising at the moment of power generation,
without any discounting, taking a social, rather than a commercial, viewpoint.
Under the Joule III programme of the European Union, the ExternE project
had generated a large set of comparable and validated results, covering more
than 60 individual power plant cases, for 15 countries and 12 different fuel
46
KAW and BANDYOPADHYAY
cycles. A wide range of technologies had been analysed, including fossil fuels,
fission power and renewables. Under the SERF1 and SERF2 programmes this
methodology was extended to include fusion, evaluating the externalities for a
hypothetical power plant situated at Lauffen, a city in the vicinity of Stuttgart,
Germany. The fusion specific data were initially taken from the results of the
European Safety and Environmental Assessment of Fusion Power (SEAFP),
for the two power station concepts studied in SEAFP [1.70], and later updated
to include a further power plant model of favourable characteristics. The main
conclusions of the studies [1.34, 1.70] are summarized in Fig. 1.22, which
shows the calculated external costs, per kilowatt-hour of electricity generated,
of electricity from coal, gas, fission, biomass, photovoltaics, wind and fusion
[1.9]. All the numbers shown are subject to significant uncertainty. Because the
external costs associated with global climate change are particularly uncertain,
these are displayed separately.
FIG. 1.22. External costs per kWh of energy production. Taken from Ref. [1.34], drawn using
the data from Ref. [1.75].
1.5.3. Public acceptance of fusion
Sociological studies within SERF also investigated Fusion and the public
opinion. With large scale commercial applications some 50 years in the future,
fusion energy production and the associated benefits and risks are not yet a
matter of public debate. Empirical sociological studies were therefore either
restricted to groups with above-average prior knowledge of fusion (structured
interviews with science journalists, environmental journalists, fusion and fission
47
THE CASE FOR FUSION
experts, members of environmental movements), or worked with focus groups,
which received introductory briefings from both fusion advocates and critics.
Both approaches showed differing expectations for the future energy mix (with
everybody wishing for an increasing role for renewables), but also a rather broad
agreement for a continuation of research into the fusion option.
What industries will most pay attention to in the future is being conscious
about the global environment. In this context, fusion waste management can be
one of the most important aspects in accordance with this trend. Recent studies
in this field have focused on a minimization of the radiotoxicity and a reduction
of the volume of the radioactive components, attempting to find a solution by
introducing the concept of clearance and recycling under regulatory control and
by improving the purity of materials. A study indicates that reinforced neutron
shielding can clear massive structural components located outside the shield
from regulatory control, leading to a dramatic reduction of radioactive waste as
shown in Fig. 1.23 [1.76]. This kind of concept can improve public acceptance of
fusion in a practical way.
FIG. 1.23. Comparison of waste weight per 1 GW(e) output for fusion reactors and the impact
of reinforced neutron shielding (courtesy of Ref. [1.76]).
1.5.4. Spin-offs of fusion research
One of the important aspects of investment in fusion science and
technologies is the role of R&D spin-offs. The role, proper analysis and projection
of spin-offs and estimation of the rate of societal return involved in a particular
R&D programme is a very crucial deciding factor in national R&D policies and
public funding. It is generally true about R&D programmes and the organizations
undertaking them that they cannot appropriate the integral returns of their
investment unless the advances in knowledge spin-off to other disciplines and
48
KAW and BANDYOPADHYAY
their consumers are properly accounted for. Accordingly, the total social payoffs
of any R&D activity are usually higher than the direct returns, especially true in
the case of basic research. For example, the pressurized light water based nuclear
reactor concept was originally developed for nuclear submarine propulsion at the
Oak Ridge National Laboratory in the USA, but also resulted in great additional
spin-offs in the areas of satellite communication, radiotherapy, etc., leading to
substantial social and economic benefits and returns. Similar spin-offs have
resulted in all other major science and technology R&D areas such as space
research and high energy physics.
There are several types of spin-offs associated with fusion research. The
foremost ones are the technology areas directly supporting fusion, for example
superconducting magnets, neutral beam systems, cryogenics and vacuum
technologies. A second category involves non-fusion plasma technologies and
associated diagnostics that arise primarily in association with fusion plasma
research, for example plasma processing of materials and a host of other industrial
plasma applications. Yet another type of spin-off, which is uniquely dependent
on fusion plasmas, are those that utilize the fusion neutrons and protons, like
waste and material treatments, production of rare but useful isotopes of different
elements and fusion based space propulsion systems.
Above all, there has been a great amount of knowledge spin-off. The
ongoing activities in basic science and technology R&D in the areas of plasma
and fusion technology have resulted in and continue to generate a great amount
of knowledge spin-off through the vast number of publications, patents, codes
and standards, highly trained workforces and social networks. This great wealth
of knowledge base forms the basis for future advancement of applied R&D
activities and is expected to increase over time with the construction of large
scale experimental facilities like the ITER, IFMIF and DEMO reactors. The
predominantly open and public nature of fusion R&D funding, unprecedented
international collaborations among major R&D labs across all continents, the
technological complexities and significant number of researchers and institutions
involved in fusion R&D programmes emphasize the knowledge spin-off effects
in fusion science and technology.
Probably the greatest beneficiaries of technology spin-off of fusion
research are the technology areas that directly contribute to fusion science and
technology. These include developments in superconducting magnets, high
heat flux materials, remote handling technologies, neutral beam systems and
various diagnostic systems, which have wide ranging applications in the fields of
material and waste processing, medical sciences, space research and so on. For
example, tremendous developments in the field of superconducting (SC) magnet
technologies, especially NbTi magnet strands, driven primarily by large volume
high field magnet requirements in magnetic confinement fusion devices, have
contributed to SC magnets used in magnetic resonance imaging (MRI) devices to
49
THE CASE FOR FUSION
scan body tissues in hospitals, as also to other R&D areas like the development of
the SC magnets in the Large Hadron Collider (LHC) in CERN.
The development of high heat flux components like carbon fibre composite
(CFC) to metal (e.g. CuCrZr) bondings developed for limiter, divertor and first
wall components in magnetic confinement devices has contributed to applications
in heavy duty electrical switchgears and high heat flux components in space
shuttles. Optical laser diagnostics, developed for measuring plasma parameters
such as velocity fields in fusion devices, has also led to the development of
laser anemometers for detecting wind velocities in front of wind turbine blades.
Developments in the field of high power radiofrequency generators such as
gyrotrons, used widely for heating and current drive in magnetic confinement
fusion devices like tokamaks, have led to their application in the field of material
processing in the form of intense instantaneous, volumetric and homogeneous
heating sources. They are now being used for industrial processing of materials
with low thermal conductivity, such as powders, powder compacts, polymers,
glass or composites. Heating with microwaves leads to a considerable reduction
in processing time and energy consumption compared to conventional resistive
heating or gas firing. Very comprehensive, robust and computationally
exhaustive numerical codes and tools developed for modelling plasma and
material properties have found wide applications in various other fields. For
example, detailed computer codes have been developed to analyse the effects of
fast plasma ions impinging on the material walls surrounding the hot plasma in
a fusion device, aimed at simulating the trajectories of ions penetrating the wall
with different energies and angles of incidence, getting implanted and causing
material defects and damages. These codes are now being widely applied to
study semiconductor properties where ion implantation changes the electrical
conductivity of the materials.
One of the most remarkable examples of fusion spin-off is the development
of a host of technologies developed for producing and manipulating low
temperature plasmas in various industrial applications. As discussed in a recent
report of the International Fusion Research Council (IFRC) [1.77], the all
pervasive influence of plasma technology can be seen in practically all walks
of life, starting from high efficiency fluorescent lamps and plasma displays to
advanced plasma based systems for manufacturing computer chips, sterilization
in medicine and the food industry, surface and exhaust gas cleaning, etc. The
IFRC Report further goes on to state: In the time period in which advances in
fusion research have taken place, fusion science has established its own identity.
Initially indebted to all other fields of physics, plasma physics, the basis of fusion,
has made contributions back to astrophysics, space physics, applied mathematics,
materials science and others. The design, construction and operation of fusion
machines have contributed to progress in most engineering disciplines, and, more
50
KAW and BANDYOPADHYAY
generally, a whole family of practical applications of plasma technologies has
emerged.
The ongoing R&D on fusion technologies has a signifcant potential spin-
off application to a host of crucial areas [1.24, 1.77, 1.78] involving high energy
neutrons, thermal neutrons, high energy protons, electromagnetic radiation
(microwaves to x rays to gamma rays), high energy electrons coupled with
photons providing ultrahigh heat fuxes, etc.
One of the fundamental roles of fusion reactors would be production
of tritium both for its own use as fuel as well as for industrial, medical and
military applications. Calculations show that fusion reactors, specially designed
to generate high neutron fux, for example a 300 MW spherical tokamak based
volume neutron source (VNS), would be capable of producing an excess of tritium
of 2 kg annually.
TABLE 1.4. USEFUL APPLICATIONS OF FUSION BY-PRODUCTS
Fusion reactor blankets can also be used for the production of critical
radioisotopes in high demand in medical applications, for example
32
P,
56
Mn,
60
Co
and
99
Mo. Especially
60
Co is an intensely radioactive material with primary use
in sterilizing medical products and secondary applications in irradiating cancer
51
THE CASE FOR FUSION
cells and food items. Present worldwide demand for replenishing
60
Co is about
11 megacuries (MCi) per year, most of which is produced at the CANDU reactors
in Canada. However, in future high demand scenarios, commercial fusion power
reactors could produce hundreds of MCi of
60
Co per year with a blanket optimized
for
60
Co production.
Another very important application of fusion neutrons is in processing of
long lived toxic radioactive hazardous wastes produced in fission reactors. At
present, fission reactor wastes have to be buried deep underground or underneath
the ocean beds, but decay heat accumulated over a long time duration can
potentially trigger the release of these radioactive materials to the environment
long after they are buried. The neutrons from a fusion reactor could be used to
change radioactive fission wastes into shorter lived materials that would decay
more quickly and hence need to be buried over shorter periods. Table 1.4 shows
the various useful applications of fusion by-products.
Another very significant spin-off of fusion research is the potential
for application of fusion based highly efficient rocket propulsion systems in
interplanetary voyages, which can make the entire solar system or even the space
beyond accessible for future human explorations. Present day rocket engines
based on chemical propulsion work for near-Earth space explorations like moon
voyages, but are totally inadequate for interplanetary voyages because they lack
the required thrust. The performance of various propulsion systems is measured
in terms of their specific impulse (exhaust velocity divided by acceleration due to
Earths gravity at sea level); thus high fuel efficiency is equivalent to high specific
impulse. Propulsive thrust for a particular fuel type is directly proportional to
specific impulse, whereas the power required to generate thrust is proportional to
the square of specific impulse. Today, only fusion based propulsion systems (or
perhaps more exotic propulsion systems based on matterantimatter annihilation)
could provide the required thrust for interplanetary spaceships. Thus, manned
spaceships to Mars and beyond would require a fusion reactor that produces
about 10100 GW of power. Table 1.5 provides a comparison of specific impulse
comparison and thrust-to-weight ratio of various propulsion systems.
1.6. CONCLUSION
Nuclear fusion energy is an energy source which has the potential to satisfy
humanitys hunger for a sustainable pollution-free and universally accessible
option for tens of thousands of years. The research programme in magnetically
confined and inertial confinement approaches is at a critical juncture today and
is likely to lead to successful fusion ignition experiments within a decade or so
and commercially viable fusion reactors generating electricity by the decade
20402050.
52
KAW and BANDYOPADHYAY
TABLE 1.5. SPECIFIC IMPULSE AND THRUST-TO-WEIGHT RATIOS OF
DIFFERENT PROPULSION SYSTEMS
In this chapter it has been shown that fusion has many inherent merits to
be the workhorse for sustainable human development. Fusion power is based
on a virtually inexhaustible and universally accessible fuel. Fusion power plants
would have inherent operational safety because of the low inventory of fuels in
the reactor, low radioactivity related problems through proper choice of materials
and negligible proliferation and terrorism concerns. Very high energy density
makes fusion an appropriate energy source for economically profitable urban
industrial complexes. Fusion power would have acceptable costs, greater socio-
economic acceptability and significant spin-off applications. Fusion research
has made tremendous strides over the last fifty years, which has culminated
in the unprecedented international collaboration to start construction of the
experimental thermonuclear reactor called ITER. A successful ITER experiment
will lead to development of DEMO reactors which will supply electricity to the
grid. Beyond successful DEMOs we can see commercial exploitation of fusion
power. For reasons mentioned above, fusion power is likely to get greater public
acceptability than competing energy sources of sustainable development.
Thus the potential merits of fusion as an energy source are so compelling
that the stakes in this development are very large. It is also very heartening to note
that in magnetic confinement fusion research, there is a very broad international
cooperation among diverse nations of the world and the beneficiaries of this
technology will involve countries such as China and India which, because of
their population size and development needs, are likely to be among the most
energy hungry nations of the globe. The biggest problem for this enterprise is its
ability to attract and retain appropriate human resources for great lengths of time
which are essential for the complete development of this challenging technology.
It is hoped that a wide dissemination of the promise and potential of this energy
resource worldwide will go a long way in attracting appropriate scientific and
technical leadership for this effort.
53
THE CASE FOR FUSION
REFERENCES
[1.1] PASTERNAK, A., Global Energy Futures and Human Development: A
Framework for Analysis, US Department of Energy Report UCRL-ID-
140773, Lawrence Livermore National Laboratory, Livermore, CA
(2000) 38.
[1.2] GOLDEMBERG, J., Energy and Human Well-being, United Nations
Development Programme Report, New York (2001).
[1.3] NASERI, H., The relationship between energy and human
development, IAOS Conf. on Statistics, Development and Human
Rights, Session C-Pa 6e, Montreux, 2000.
[1.4] SMIL, V., Energy at the Crossroads, MIT Press, Cambridge, MA
(2003).
[1.5] DIAS, R.A., MATTOS, C.A., BALESTIERI, J.A.P., The limits of
human development and the use of energy and natural resources,
Energy Policy 34 (9) (2006) 10261031.
[1.6] UNITED NATIONS DEVELOPMENT PROGRAMME,
http://hdr.undp.org/
[1.7] UNDP Report on Energy Scenarios:
http://stone.undp.org/undpweb/seed/wea/pdfs/chapter9.pdf
[1.8] FELLS, I., The need of energy, Europhys. News 29 (1998) 193.
[1.9] BARABASHI, S., et al, Fusion Programme Evaluation 1996. UR
17521, Office for Official Publications of the EU, Luxembourg.
[1.10] ITER:
http://www.iter.org/
[1.11] NATIONAL IGNITION FACILITY, NIF:
https://lasers.llnl.gov/programs/nif/
[1.12] LASER MEGAJOULE:
http://www-lmj.cea.fr/html/cea.htm
[1.13] WORLD ENERGY COUNCIL (WEC):
http://www.worldenergy.org/
[1.14] INTERNATIONAL ENERGY AGENCY:
http://www.iea.org/
54
KAW and BANDYOPADHYAY
[1.15] SUBCOMMITTEE OF THE FUSION COUNCIL FOR FUSION
DEVELOPMENT STRATEGY, Report on the Technical Feasibility
of Fusion Energy and Extension of the Fusion Program and Basic
Supporting Researches, May 17, 2000 (in Japanese; English translation
is available at
http://www.naka.jaeri.go.jp/SCI/).
[1.16] UNITED NATIONS FRAMEWORK CONVENTION ON CLIMATE
CHANGE: The Kyoto Protocol,
http://unfccc.int/kyoto_protocol/items/2830.php
[1.17] MOOMAW, W., Renewable Energy and Climate Change An
Overview, Proc. IPCC Scoping Meeting on Renewable Energy
Sources,
http://www.ipcc.ch/pdf/special-reports/srren.pdf
[1.18] NATIONAL GEOGRAPHIC:
http://news.nationalgeographic.com/news/2005/01/0114_050114_
solarplastic.html
[1.19] KING, R.R., et al., Appl. Phys. Letters 90 (2007) 183516.
[1.20] Large Scale Photovoltaic Power Plants:
http://www.pvresources.com/download/AnnualReport2007.pdf
[1.21] YAMAGUCHI, M., et al., Sol. Energy 80 (2006) 104.
[1.22] US photovoltaic roadmap:
http://photovoltaics.sandia.gov/docs/PDF/PV_Road_Map.pdf
[1.23] OILGAE website:
http://www.oilgae.com
[1.24] GNANSOUNOU, E., BEDNYAGIN, D., Multi-regional long-term
electricity supply scenarios with fusion, Fusion Sci. Technol. 52(3)
(2007) 388.
[1.25] NAKICENOVIC, N., et al., UNDP report on Energy Scenarios
(Chapter 9).
[1.26] EHERER, C., BAUMANN, M., DWEKE, J., HAMACHER, T.,
Nuclear fusion as new energy option in a global single-regional
energy system model, Proc. 20th IAEA Fusion Energy Conf., 2004,
Vilamoura, Portugal, paper SE/P3-39;.
http://www-naweb.iaea.org/napc/physics/fec/fec2004/papers/se_p3-39.
pdf
55
THE CASE FOR FUSION
[1.27] EHERER, C., BAUMANN, M., Long-Term Technology Based
Modelling with TIMES, Interim Report, EFDA/SERF sub task
TW3-TWE-FESA/A, Garching, June 2004.
[1.28] LECHON, Y., et al, A global energy model with fusion, Fusion Eng.
Des. 75-79 (2005) 1141.
[1.29] BAUMANN, M., EHERER, Ch., Recent long-term energy scenario
results with the TUGIPP global single-regional energy model: ETL in
the TUG-IPP model, IEA Workshop on Socio-Economics of Fusion,
Culham 2005.
http://fusion.org.uk/socioecon/Eherer.pdf
[1.30] INTERNATIONAL INSTITUTE FOR APPLIED SYSTEMS
ANALYSIS:
http://www.iiasa.ac.at
[1.31] INTERGOVERNMENTAL PANEL ON CLIMATE CHANGE (IPCC):
http://www.ipcc.ch
[1.32] IIASA-WEC, Global Energy Perspectives, (NAKICENOVIC, N.,
GRUBLER, A., McDONALD, A., Eds), Cambridge University Press,
ISBN 0-521-64569-7. Online version of the report also available at
http://www.iiasa.ac.at/cgi-bin/ecs/book_dyn/bookcnt.py
[1.33] Global Warming page on Wikipedia:
http://en.wikipedia.org/wiki/Global_warming
[1.34] COOK, I., MILLER, R.L., WARD, D.J., Prospects for Economic
Fusion Electricity Fusion Eng. Des. 63-64 (2002) 25.
[1.35] Department of Atomic Energy (DAE), Government of India report:
http://www.dae.gov.in/publ/doc10/
[1.36] HAMACHER, T., SHUKLA, P.R., SEEBREGTS, T., The possible role
of fusion in the Indian energy system of the future, Fusion Eng. Des. 69
(2003) 733.
[1.37] KIKUCHI, M., INNOUE, N., Role of fusion energy for the 21
century energy market and development strategy with international
thermonuclear experimental reactor, Proc. 18th World Energy Conf.,
Buenos Aires, 2001, DS-6, No 01-06-09
(http://fire.pppl.gov/energy_ja_wec01.pdf).
[1.38] MAISONNIER, D., et al., DEMO and fusion power plant conceptual
studies in Europe, Fusion Eng. Des. 81 (2006) 11231130.
56
KAW and BANDYOPADHYAY
[1.39] MAISONNIER, D., et al., The European power plant conceptual study,
Fusion Eng. Des. 7579 (2005), 1173-1179.
[1.40] TOBITA, K. et al., Design study of fusion DEMO plant at JAERI,
Fusion Eng. Des. 81 (2006) 11511158.
[1.41] NAJMABADI, F., et al., The ARIES-AT advanced tokamak advanced
technology fusion power plant, Fusion Eng. Des. 80 (2006) 323.
[1.42] SRINIVASAN, R., et al., Strategy for the Indian DEMO design, Fusion
Eng. Des. 83 (2008) 889.
[1.43] IFMIF website:
http://www.frascati.enea.it/ifmif/
[1.44] ABDOU, M.A., Fusion Technol. 9 (1986) 250; ibid 28 (1995) 5.
[1.45] KUAN, W., ABDOU, M.A., A new approach for assessing the required
tritium breeding ratio and startup inventory in future fusion reactors,
Fusion Technol. 35 (1999) 309.
[1.46] SWAN, M., ABDOU, M.A., Physics and technology conditions for
attaining tritium self-sufficiency for the DT fuel cycle, Fusion Eng.
Des. 81 (814) (2006) 1131.
[1.47] LAWSON, J.D., Some Criteria for a Useful Thermonuclear Reactor,
Atomic Energy Research Establishment, 1955 Culham Report No:
A.E.R.E. GP/R 1807.
[1.48] HARMS, A.A., KINGDON, D.R., SCHOEPF, K.F., MILEY, G.H.,
Principles of Fusion Energy, World Scientific Publications, ISBN
978-981-238-033-3.
[1.49] NRL Plasma Formulary Edn 2007, page 45.
[1.50] ENDF database at the IAEA:
http://www-nds.iaea.org/exfor/e4explorer.htm
[1.51] NEVINS, W.M., SWAIN, R., The thermonuclear fusion rate coefficient
for p-
11
B reactions, Nucl. Fusion 40 (2000) 865.
[1.52] DAWSON, J.M., KAW, P.K., Current maintenance in tokamaks by used
of synchrotron radiation, Phys. Rev. Letters 48 (1982) 1730.
[1.53] ROMANELLI, F., GIRUZZI, G., A D-
3
He fusion reactor with an edge
radiation barrier, Nucl. Fusion 38 (1998) 103.
[1.54] ROSENBLUTH, M.N., HINTON, F., Generic issues for direct
conversation of fusion energy from alternative fuels, Plasma Phys.
Control. Fusion 36 (1994) 1255.
57
THE CASE FOR FUSION
[1.55] WESSON, J., Tokamaks, Oxford University Press, New York (1987)
ISBN 0-19-856328-0.
[1.56] ARTISIMOVICH, L.A., Controlled Thermonuclear Reactions, ISBN
067720020x (1964).
[1.57] MYAMOTO, K., Plasma Physics for Nuclear Fusion, MIT Press,
Cambridge, MA (1976).
[1.58] DOLAN, T.J., Fusion Research, Pergamon Press, New York (1982).
[1.59] WATAKANI, M., Stellarator and Heliotron Devices, Oxford University.
Press, New York, (1998), ISBN-0-19-507831-4.
[1.60] LHD website:
http://www.lhd.nifs.ac.jp/en/
[1.61] Wendelstein-7x:
http://www.ipp.mpg.de/ippcms/eng/pr/forschung/w7x/index.html
[1.62] IPP:
http://www.ipp.mpg.de/
[1.63] EFDA-JET:
http://www.jet.efda.org/
[1.64] HOANG, G.T., JACQUINOT, J., Controlled fusion: the next step,
Physics World (10.1.2004) 21.
[1.65] LINDL, J.D., Inertial Confinement Fusion: The Quest for Ignition and
Energy Gain Using Indirect Drive, Springer-Verlag, New York (1998),
ISBN: 1-56396-662-x.
[1.66] ATZENI, S., MEYER-TER-VEHN, J., The Physics of Inertial Fusion,
Beam Plasma Interaction, Hydrodynamics, Hot Dense Matter, Oxford
University Press, Oxford (2004), ISBN 0 19 856264 0.
[1.67] RAEDER, J., et al., Safety and Environmental Assessment of Fusion
Power (SEAFP), Report of the SEAFP Project, European Commission
DGxII, Fusion Programme, EURFUBRU xII-217/95, Brussels (June
1995).
[1.68] ICRP Publication 60, Recommendations of the International
Commission on Radiological Protection, 1990, Pergamon Press.
[1.69] COOK, I., et al., Results, conclusions and implications of the SEAFP-2
programme, Fusion Eng Des. 5152 (2000) 409.
[1.70] COOK, I., et al., Safety and Environmental Impact of Fusion, EFDA
report, 2001, EUR (01) CCE-FU / FTC 8/5.
58
KAW and BANDYOPADHYAY
[1.71] HAMACHER, T., BRADSHAW, A.M., Fusion as a future power
source-recent achievements and prospects, Proc. 18th Congress of
World Energy Council, Buenos Aires, October 2001
(http://fire.pppl.gov/energy_eu_wec01.pdf).
[1.72] BORELLI, G., et al., Socio-economic Research on Fusion Summary
of EU Research 1997-2000:
http://www.efda.org/eu_fusion_programme/downloads/scientific_and_
technical_publications/SERF_final.pdf
[1.73] ARGOTE, L., EPPLE, D., Learning curves in manufacturing, Science
247 (1990) 920.
[1.74] ExternE homepage:
http://externe.jrc.es/, http://www.externe.info/
[1.75] SAEZ, R., et al., Socio-economic Research in Fusion SERF 1997-98:
Externalities of the Fusion Fuel Cycle. Final Report. Coleccion
Documentos CIEMAT (1999).
[1.76] TOBITA, K., et al., Fusion reactor design towards radwaste minimum
with advanced shield material, J. Plasma Fusion Res. 77 (2001) 1035.
[1.77] Internation Fusion Research Council (IFRC), Status Report on Fusion
Research, Nucl. Fusion 52 (2005) A1A28.
[1.78] McCARTHY, K., et al., Non-electric applications of fusion, J. Fusion
Energy 21 (2002) 121.
59
CHAPTER 2
PHYSICS OF CONFINEMENT
R.J. Goldston, R.B. White, T.S. Hahm, S. Kaye
Princeton Plasma Physics Laboratory,
Princeton, United States of America
2.1. INTRODUCTION AND OVERVIEW
The measured radial fluxes of particles, ion and electron heat, and toroidal
angular momentum in tokamaks are usually greater than given by neoclassical
transport theory, described in Section 2.2. In this collisional theory, appropriate
for a quiescent plasma, the random-walk step size is generally set by the width
of closed banana trajectories in the steady state magnetic fields, and the step
period is set by the collisional decorrelation time. At low collisionalities relevant
to fusioning plasmas, detailed calculations give radial transport coefficients of
the form
2
c
G p v , where
c
G is a dimensionless geometrical factor, p is the
gyroradius of the species under consideration and v is the collision frequency.
The observation of enhanced, or anomalous, cross-field transport greater
than the predicted neoclassical level is qualitatively consistent with theoretical
considerations and large computer simulations, discussed in Section 2.3, which
indicate that tokamak plasmas should be subject to strong dominantly electrostatic
gyroBohm turbulence. In such turbulence the correlation length,
c
p , is a modest
multiple of the ion sound gyroradius,
s
p , and the turbulence amplitude is of the
order of the mixing length estimate (
1
/ ( )
n
n n k L o

1
) in which the gradient of
the perturbed density approximately equals the gradient of the ambient density.
n
L is the density scale length
1
( / )
n
L n dn dr

and k
1
is the wavenumber
perpendicular to the magnetic field. In electrostatic turbulence, one estimates
/ / e T n n o o 4 , where 4is the electrostatic potential. Note that temperature T is
expressed everywhere in energy units eV suppressing Boltzmanns constant. Using
the mixing length estimate for / n n o , the turbulent velocity can be estimated at
v k /B o o
1
4 , and so a random-walk transport coefficient, v
c
p o ,

takes the form
/ ( )( / )
T s n
G T eB L p , where
T
G is a dimensionless factor that requires detailed
calculation and can depend on such parameters as collisionality, and q , but
not further factors of p normalized by the system size.
In typical existing fusion plasma confinement systems, and especially in
ITER, gyroBohm transport is a great deal weaker than the worst-case estimate of
turbulent transport, the Bohm level, which obtains for extremely strong turbulence
with / 1 n n o and 1
c
k p
1
, and so scales as / T eB by the arguments
above. However, consistent with experiment, gyroBohm transport dominates
60
GOLDSTON et al.
neoclassical except in some cases where ion neoclassical thermal transport can
be large (e.g. at high collisionality), or where turbulence is suppressed.
A quantitative, predictive understanding of gryoBohm turbulence, and full
understanding of the roles of other potential sources of turbulent transport, such as
turbulence driven on the electron gyroradius scale or electromagnetic turbulence,
remain outstanding challenges. It is clear that full understanding will require:
(1) Detailed measurements of transport;
(2) Detailed measurements of turbulence;
(3) Detailed comparisons of both with non-linear turbulence simulations.
All of these need to be undertaken over a sufficiently wide range of
conditions and operating modes to establish that the basic understanding evolved
from these studies, coupled in some cases with transport due to large scale
instabilities discussed in Chapter 3, explains the observed scalings of global
confinement with overall plasma parameters such as plasma current, major radius
and heating power, described in Section 2.4. The scaling studies discussed there,
particularly for energy confinement, have been very successful in correlating
data from many experiments, with RMS errors typically in the range of 1015%.
They call out for full explication based on detailed measurements and simulation.
Also described in Section 2.4 are global confinement studies to estimate at a
global level the scaling of
T
G with dimensionless parameters, and to examine
any non-gyroBohm scaling with p.
As discussed in Section 2.5, local transport of particles, heat and momentum
have been measured in many conditions, and comparisons have been made
with theory over a more limited range of conditions, but most often only with
linear stability calculations or quasi-linear estimates, not with fully non-linear
simulations. Studies of the variation of local transport with dimensionless
parameters have been made as well, providing further data on the variation of
T
G , but not enough has been done to provide a satisfactory link to the observed
global scalings.
The powerful technique of studying transport response to perturbations has
been exploited. This and careful studies of temperature gradients versus power
flows have confirmed a basic feature of the theoretical gyroBohm turbulence
model for tokamaks discussed in Section 2.3, that turbulent transport, as measured
against gyroBohm scaling, is low below a certain threshold, typically set by a
critical logarithmic derivative (
1
/ T dT dr

) of the ion or electron temperature,

and when the temperature gradient exceeds this threshold, transport rises very
quickly. This gives rise to the long-observed stiffness, particularly in the radial
profile shape of the electron temperature, since the logarithmic temperature
gradient tends to be stuck close to its critical value, independent of the level or
profile of heating power.
61
PHYSICS OF CONFINEMENT
GyroBohm transport theory has a further distinctive feature that has also
been observed in experiments. The turbulent decorrelation rate,
1
( / v)
c
p o

, is
calculated to be of the order of the linear growth rate, having a scaling of the
form, /
s n
c L . If natural turbulent structures are torn apart by some mechanism
at a rate much faster than the linear growth rate, or equivalently, faster than
the non-linear decorrelation rate, one should expect strong suppression of the
turbulence and its associated transport. A possible mechanism for destruction of
turbulent structures is their breakup due to shear in ambient flows. The velocity
at which the components of a structure move apart in a sheared flow is given
by its scale size multiplied by v / d dr. The time to shear apart the structure is
given by its scale size divided by this velocity with the consequence that the
shearing decorrelation rate is just v / d dr , independent of structure size. Thus
theory predicts, in rough terms, that if / ( / ) d dr E B exceeds the linear growth
rate of an instability one should see turbulence suppression. This effect is indeed
observed both close to the plasma edge (in H-mode) and also in the plasma core,
with strong shear flow and especially with reversed magnetic shear ( / 0 dq dr < )
such that the linear instability growth rates for both ion and electron turbulence
are reduced, as described in Sections 2.3 and 2.6.
As described in Section 2.6, powerful techniques have been developed
to measure turbulence within tokamak plasmas, but they are limited to date in
their application and in their comparison with simulation. They are also limited
in the turbulence properties they can detect. Full comparison between theory
and experiment demands excellent spatial resolution as well as resolution in
frequency and wavenumber, and reliable absolute calibration. Furthermore
it is very valuable to measure more than one fluctuating quantity, and even to
measure the phases between these quantities to determine turbulent transport
rates, for example v
r
n o o I . In any given case, it is possible to obtain only
some aspects of the desired data. Nonetheless, results to date generally support
the gyroBohm model. The wavelengths, and correlation times and distances,
of the measured turbulence are consistent with the gyroBohm model, as is its
rough magnitude. Confinement tends to decrease with increasing turbulence; for
example, turbulence is generally seen to increase with increasing input power,
which leads to decreased energy confinement. While almost all measurements
have been of density fluctuations, ion and electron temperature fluctuations have
been measured in some cases, and are found to be roughly as predicted.
Interestingly, analytic theory and computer models predict that localized
sheared flows in the plasma, generated by the turbulence itself, provide for
self-regulation of the turbulence level, and indeed break up meso-scale structures
that could give rise to Bohm-like scaling. These flows have been observed in
some experimental conditions. Furthermore, under conditions where theory
predicts turbulence suppression due to sheared flows and/or reduced linear
growth rates, turbulence has also been measured to be suppressed dramatically.
62
GOLDSTON et al.
Overall, however, it should be recognized that the gyroBohm model of
turbulence dominated by fluctuations at the ion scale, while highly successful,
is not fully established; indeed a range of anomalies should be appreciated. In
the L-mode, global scalings and local transport variations are more consistent
with an overall Bohm-like scaling than with the gyroBohm variation as
/
s n
L p . However, the turbulence driven self-regulating flows that should give
rise to gyroBohm scaling have been detected in the L-mode, but fade below
detectability in the H-mode. Furthermore, at the H-mode transition rapid changes
have been observed in core transport and turbulence, at least suggesting a
non-local effect. (Non-local effects, called turbulence spreading, are discussed
in Section 2.3.) Turbulence and transport grow rapidly towards the plasma
edge, and initial studies of this are inconsistent with simulations. Significant
transport is also observed even when the electron temperature profile is very
flat near the plasma centre, for example within a transport barrier established
by magnetic shear reversal, but gyroBohm turbulence is calculated to be stable
in this region. These two effects could also be due to non-local phenomena, in
the one case perhaps edge turbulence leaking into the confinement region, and
in the other confinement region turbulence leaking into the central region. It
has been observed, as well, that some small scale instabilities can be driven by
energetic beam ions, and these could be responsible for enhanced transport in the
plasma core. In some circumstance electromagnetic instabilities are calculated
to be unstable and potentially important. Finally, high resolution measurements
have recently been made of turbulence at high wavenumber k, corresponding to
instabilities driven by the electron temperature gradient. This turbulence grows
and is even suppressed by reversed shear as predicted by theory, and may play a
role in electron thermal transport.
Returning to the triad of requirements for measurements of transport and
of turbulence, and for comparison of large scale simulations with experiment
over a broad range of conditions, it is clear that with the advent of massively
parallel computing, non-linear turbulence simulations have become much more
detailed and complete, but they remain sufficiently costly and difficult that they
have not been broadly compared with experiment. Experimental run time is also
at a premium and as a consequence three-way correlations of measured transport,
measured turbulence and non-linear simulation, across a broad enough range of
conditions to make comparisons with global scalings, do not exist. Thus this most
basic test of whether the gyroBohm model explains the observed phenomenology
of tokamak confinement remains as a challenge.
While the overall picture that gyroBohm turbulence plays a key role in
tokamak transport is reasonably secure, and together with global scaling results
gives an acceptable basis for projection to future devices such as ITER, it is clear
that a great deal of exciting research remains to be done in the area of plasma
turbulence and transport.
63
PHYSICS OF CONFINEMENT
2.2. NEOCLASSICAL TRANSPORT
2.2.1. Introduction
A plasma is completely described by a distribution function v f(x, ,t)
. .
giving
the time dependent histories of all the particles of which it is comprised. Particle
positions evolve in time under the Lorentz force ( / )( v ) e m E B +
. .
.
, with E
.

and
B
.
the local electric and magnetic fields. The Hamiltonian nature of the particle
trajectories implies phase space conservation, or
v
v ( / )( v ) 0
f
f e m E B f
t
o
+ V + + V
o
. .
. .
(2.1)
with
v
V the velocity space gradient operator. Solution of this Vlasov equation
in general is equivalent to solving the classical electromagnetic many-body
problem, an impossible task. However, the distribution function contains
vastly more information than is needed or can be possible measured, so it is
essential to simplify the description of the plasma. To smooth the discreteness
of the distribution function f , an ensemble average is carried out, resulting in
an additional term from the correlations due to close encounters, the collision
term. The long range effect of particles on each other is accounted for by the
self-consistent electric and magnetic fields of the plasma. Only very close
interactions of particles, at distances smaller than a Debye length, are not treated
by the Vlasov equation.
Approximations are made in two directions. First the equations of motion
are simplified by expanding, in the available small parameters describing the
system, the ratio of the cyclotron radius to the scale length of the fields, the ratio
of the collision frequency to typical timescales of plasma response, and the ratio
of electron to ion mass. Neglecting collisions and to lowest order in the expansion
of the equations of motion in cyclotron radius the particles simply follow the field
lines. The guiding centre approximation [2.12.4] includes the particle drifts, one
order higher in this expansion. Gyrokinetic equations are obtained by carrying
out the analysis to still one higher order in the gyroradius expansion [2.52.7].
Provided that the resulting approximate equations are Hamiltonian in nature, if
collisions are neglected phase space conservation results in a Vlasov equation at
each level.
Secondly, the magnetic fields are simplified by making use of the fact that
they describe, at least very closely, a toroidal magnetohydrodynamic (MHD)
equilibrium, allowing long time plasma confinement. Perturbations of the
equilibrium are assumed to consist of small amplitude waves with scale lengths
much larger than the cyclotron radius and frequencies small compared to the
cyclotron frequency.
64
GOLDSTON et al.
Finally, using the distribution function for each species, ( , v) f x
. .
, one
can construct fluid theory functions depending only on position and time by
taking velocity moments, such as density,
3
( ) v n x d f
l
.
, particle flux density
3
( ) v v nu x d f
l
. . .
(where u
.
is a fluid velocity and v
.
is particle velocity), stress
tensor
3
,
( ) v v v P x d fm
o o

l
.
(where o, label components), and energy
flux density,
3 2
( ) v v v / 2 Q x d fm
l
.
. .
. Taking moments of the ensemble averaged
kinetic equation then results in fluid equations relating these quantities, but each
quantity is coupled to higher order moments and one must make approximations
to give closure to these equations [2.82.10]. The first few moments give, for
a single species, and evaluated in the rest frame, the Braginskii equations for
density, momentum and pressure
0
dn
n u
dt
+ V
.

(2.2)
( )
du
mn p en E u B F
dt
r + V + V +
.
. . .
. .

(2.3)
,
3 5
2 2
dp
p u u Q W
dt
o o
r + V + o + V
. .
(2.4)
where / x
o o
o o o ,
, ,
P p
o o o
o r + , with p the pressure and r the
viscosity tensor,
, o
o is the Kronecker delta function and F
.
and W are source
terms. The coefficients of the terms describing cross-field particle and energy
flux, momentum diffusion, bootstrap current and resistivity are dependent on the
nature of the particle orbits in the system considered. For toroidal confinement
systems these are called neoclassical transport coefficients [2.11, 2.12]. Normally
their values are calculated using the guiding centre approximation for the particle
trajectories. Other moments will be discussed in Section 2.5.5. If there exist
fluctuating magnetic or electric fields, due to MHD or kinetic instabilities, this
will produce additional anomalous transport.
In the following we first consider the simplification of the fields resulting
from toroidal equilibrium, Section 2.2.2, then guiding centre motion in such an
equilibrium, Section 2.2.3, and the resulting drift kinetic equation resulting from
these two approximations in Section 2.2.4, used to calculate some neoclassical
transport coefficients. Section 2.2.5 treats bootstrap current, and Section 2.2.6 the
modification of transport due to field ripple. Section 2.2.7 discusses transport in
a stochastic field.
65
PHYSICS OF CONFINEMENT
2.2.2. Toroidal magnetic confguration
Because of the very high particle velocity along magnetic field lines, the
field must, at least approximately, possess toroidal magnetic surfaces, to prevent
rapid particle transport from the inside of the torus to the exterior, and thus
allow the existence of pressure and temperature gradients. We introduce general
coordinates
p
v , labelling the flux surfaces, u, a poloidal angle, and c , a toroidal
angle. Since B
.
is orthogonal to
p
v V , use the Clebsch representation
p
B G v V V
.
(2.5)
and choose the function ( ) ( , , )
p p
G q v u c i v u c + where
u

and c

are arbitrary poloidal and toroidal angles, respectively. We choose
2
p
rv

to be the poloidal flux inside the magnetic surface. We then
find that i is periodic in c . We then have the contravariant represen-
tation for B
.
,

( ) (1 )
p p
B q
.
u c
i v u i c v + V V + V V , with primes
indicating derivatives with respect to the subscripted variable. Now choose
0
2 2 ( )
p
p p
q d
v
rv r v v
l
to be the toroidal flux inside the surface. We then find
that i is periodic in

u and / ( )
p p
d d q v v v , the flux surfaces can be labelled
by either
v
or p
v
.
Straight field line coordinates are often used for equilibrium and stability
studies. Replace c i with c in Eq. (2.5), equivalent to setting i equal to
zero, possible because of the periodicity of i . The contravariant representation
for B
.
then becomes
( )
p p p
B q
.
v v u c v V V + V V (2.6)
and ( )
p
q v is the local helicity of the field line, / d d u c following the field. It is
referred to as the safety factor in tokamak research since if 1 q < the equilibrium
is unstable. The derivative of q gives the local shear of the equilibrium field. It is
also useful to note the vector potential associated with B
.
, with B A V
. .
,
p
A
.
v u v c V V (2.7)
Now we make use of the MHD equilibrium property of B
.
. We assume a
scalar pressure equilibrium, ( ) p p v . From the equilibrium condition j B p V
. .

with ( ) p p v and j
.
the plasma current density, it follows that 0 j
.
v V . Write
j W
.
v V V , and choose ( ) ( ) ( , ) W I g v u v c o v u + + , where / I dI dv
and / g dg dv . Then ( ( ) ) ( ) j I g
.
u
v o v u v v c + V V + V V , where
/
u
o o u o o . Choose 2 I r to be the toroidal current inside v , and 2 g r to
be the poloidal current outside v . Then, as in the case with i, o is periodic in
u , c . Then B j V
. .
gives the covariant representation for B
.
B g I
.
c u o v V + V + V (2.8)
66
GOLDSTON et al.
with g g < > , I I < > = and the brackets indicate averaging over u and c . In the
case of axisymmetry ( ) g g v , and also the poloidal coordinate can be chosen
so that ( ) I I v . In a stellarator equilibrium g, I , o are all functions of all three
variables. With the coordinate system chosen as in Fig. 2.1, g, q and I

are all
positive, the plasma current flows clockwise as seen from above, and the field
helicity is positive.
FIG. 2.1. General coordinates.
The ratio of the large toroidal axis R to the small toroidal axis a is
called the aspect ratio. If / 1 R a >> , the toroidal plasma cross-section circular,
and the plasma pressure small compared to the magnetic field pressure,
plasma beta
2
2 / 1 p B << , then the equilibrium is very simple, given by
0
g B R ,
2
0
/ 2 B r v ,
2 2
0
/ ( ( )) I B r R q r , 0 o , and the field strength
is
0
(1 cos / ) B B r R u with
0
B being the magnetic field on the axis, and
u and c are normal poloidal and toroidal coordinates and the flux surfaces
coincide with the minor radius r. The helicity function ( ) q r is determined by the
plasma current, hence the relation between q and I. Although most tokamaks
are non-circular, and can be a few per cent, the large aspect ratio quantities
give good intuitive approximations to many properties. For high and strongly
shaped equilibria one must find the equilibrium field by solving the Grad
Shafranov equation [2.13, 2.14], but expansion in the inverse aspect ratio is also
useful [2.15].
2.2.3. Guiding centre motion
The Larmor radius of charged particles in a reactor scale plasma is
small compared to the plasma dimensions and the radius of curvature of the
67
PHYSICS OF CONFINEMENT
magnetic field. Thus charged particle motion is well described using the drift
approximation. The guiding centre motion is determined by the cyclotron radius
and the field strength and configuration. It is very convenient to use units that
make the equations of the problem being considered simple. Introducing units
of time given by
1
0
u

, where
0
/ ( ) eB mc u is the (signed) on-axis cyclotron
frequency, and units of distance given by the major radius R , the basic unit of
energy becomes
2 2
0
m R u , which can also be written as
2 2 2
( v / 2)(2 / ) m R p ,
the gyroradius is v / 1 B p << , and the magnetic moment
2
v / (2 ) B
1
is of
order
2
p . Particle motion both along and across the field lines is of order p
but to leading order the cross-field motion is the periodic cyclotron motion, and
cross-field drift is of order
2
p . It is trivial to restore convential MKS units to
final results, and this will be illustrated in the following. See for example the next
to last paragraph of this section.
Begin with the non-relativistic Lagrangian for the motion of a charged
particle in an electromagnetic field [2.16],
( , ) v (v, ) L A x t x H x
]
+
]
.
. . . .

(2.9)
with B A V
. .
, and the Hamiltonian
2
v / 2 ( , ) H x t + 4
. .
, where ( , ) x t 4
.
is the
electric potential. Putting the Lagrangian in the form
i i
p q H Z we recognize
the canonical momentum for a charged particle in an electromagnetic field
to be v p A +
.
. .
. Now explicitly separate the motion along the field from the
cross-field motion through
||

v v w b c
.
+ with v
|

being velocity parallel to
the magnetic field,

/ b B B
.
, and
1 2
sin cos c e e c c with
1
e and
2
e unit
vectors orthogonal to B
.
and to each other chosen so that
1 2

e e b , and c the
gyrophase, and w the magnitude of the perpendicular velocity v
1
.
. Define the
particle gyrocentre through w / x X a B
.
.
+ with
1 2
cos sin a e e c c , so that
c a b
.
. .
, and all quantities on the right are evaluated at the guiding centre X
.
.
The Lagrangian then becomes
||
w

( , ) v w
d a
L A x t b c X H
dt B
. .
. ]
| `
]
+ + +
]
]
. }
]
(2.10)
Now make an expansion in the small parameter w . Assume that all spatial
scales are large enough compared to w so that the time variations of the field
viewed by the particle, as well as all explicit time variations of the field, are slow
compared to the gyration frequency / d dt c . Add perfect derivatives to L and
average over the fast time scale, giving
||
( ) v L A B H p c + +
. .
.

(2.11)
with
2 2
||
/ 2 H B B p + + 4 the Hamiltonian and 4 the electric potential, and
v / B p
| |
.
68
GOLDSTON et al.
We now use the contravariant and covariant representations for
B
.
. Substitute Eq. (2.7) for A
.
and Eq. (2.8) for B
.
into the Lagrangian, giving
|| || ||
( ) ( )
p p
L I g H v p u p v c c op v + + + +


. Lagranges equations give
[2.12.3]
||
2
|| ||
(1 )
( )
D
g
B
B
p
p p
u u

o o4
]
+ +
]
o o
]

||
2
||
( )
( )
q I
B
B
D
p
p
c c
+
] o o4
+ +
]
o o
]
(2.12)
2 2
|| ||
( ) ( )
p
g B I B
B B
D D

v p p
u u c c
] o o4 o o4
]
+ + + + +
]
]
o o o o
]
]
(2.13)
2
||
2
|| ||
(1 ) ( )
p p
B
g B
g B
D D

p
u p p
v v
]
o o4
+ + + ]
o o
]
]
(2.14)
2
||
2
|| ||
( ) ( )
p
p p
B
I B
q I B
D D
v u
p
c p o p
v v
]
o o4
]
+ + + + ]
]
]
o o
]
]

+
2
||
( )
B
B
D
o
p
u u
o o4
]
+ +
]
o o
]
(2.15)
with
||
( )
p p
D gq I gI Ig g I
v v u c
p o o + + + .
Note that the function o modifies
p
v , u and
||
p only through a
renormalization of the time interval, through the denominator D. Thus the
projection of the orbit onto the poloidal plane is independent of this function.
The second order drift of a trapped particle over one transit of the orbit in
u is
dt d
c
c c u
u
A
l l

(2.16)
Then from Eqs (2.14, 2.15), to first order in
||
p and with no potential,
0 4 ,
||
( )
p
d q I g q
v u
c u p o
]
A + +
]
]
l

2
||
2
||
( )
( )
p
B
B B
d I gq
B
p
o
u o
v ou p
] +
o
+ ]
o
]
]
l
(2.17)
69
PHYSICS OF CONFINEMENT
Using energy conservation with
2 2
||
/ 2 E B B p + to find
2 2 2
|| || || ||
( ) B B B B B
u u u
p p p p + o o o the terms in o can be written as
d

u
l
0
| u
op
]
o
]
. Thus neglecting o produces only a non-secular change
in the guiding centre motion. Dropping the term in o by a redefinition of the
guiding centre, the Lagrangian takes the form
|| ||
( ) ( )
p
L I g H

v p u p v c c + + + (2.18)
and the canonical coordinates and momenta are immediately identified to be
u , c and P
u
, P
c
, with
|| p
P g
c
p v ,
||
P I
u
v p + . These expressions are
immediately recognizable as the covariant components of the sum of particle
momentum and vector potential, just as in classical electrodynamics, Eq. (2.9)
[2.17].
From the guiding centre equations we have the important fact that the
particle trajectories depend only on the magnitude of B , not on its vector
components. Thus it is not geometrical symmetry that determines an associated
integral of the motion, but only symmetry of the field magnitude, referred to as
quasi-symmetry. This property is exploited in the design of stellarators, which
do not possess geometric symmetry, but nevertheless can approximate quasi-
symmetry. With no electric field the particle energy takes the form
2
2
2
( , )
2
p
P P P B
W B
g
c c u
v

]
+
]
+ (2.19)
In an axisymmetric system

( , ) c o \ v u + , with
o
the cylindrical
toroidal coordinate and \ the difference between general and cylindrical
toroidal coordinate. Thus

/ / c o o o o o
with v and

u
fixed and thus

c
is
also an ignorable coordinate. From the equations of motion we see that

P
c
is
constant if in the system

c is an ignorable coordinate, i.e. if

B is independent
of

c . This integral of the motion considerably simplifies the particle orbits. In
this case,

W is a function of

||
P I
u
v p + and

u
. But
1/ 2
||
(2 2 ) / W B B p
is also a function of

v , u , so W constant defines closed curves in the v ,
u

plane. Thus the particle motion is restricted to a two dimensional drift surface
characterized by the values of W ,

, regardless of the shape of the flux surface
cross-section.
To first order in p a particle follows the field lines and to next order it
drifts across the flux surfaces. The drift motion in the
p
v , u plane is given by
the terms of second order in p
2
||
( )
p
g B
B
D

v p
u
o
+
o
,
2
||
( )
p
g B
B
D

u p
v
o
+
o
(2.20)
70
GOLDSTON et al.
where we neglect the (normally small)
2
||
g p
term and consider the case of
zero potential. The motion across the flux surfaces is usually odd in u since
B increases in the direction of decreasing major radius, the distance from the
axis of symmetry, so 0
p

v < for 0 u r < < and 0

p

v > for 0 r u < < , i.e.

inward drift in the upper half of the torus and outward drift in the lower half. As
it moves around u towards decreasing major radius it will normally encounter
increasing field magnitude, and may encounter a mirror point, where
||
0 p
because ( , ) /
p b
B E v u ,
b
u being the bounce angle. If such a point exists
the particle is called trapped, otherwise it is called passing. Since drifts across
v are proportional to
2
p the orbit of a trapped particle has the form of a
banana with width ~ p , as shown in Fig. 2.2. To calculate the drift from the
flux surface
p
v use conservation of P
c
and neglect variation of g along the
orbit, which gives a higher order correction in p , to find
|| p
g v p A A . For a
trapped particle
||
p is zero near the banana tip so
1/ 2
||
(2 2 ) / E B B p . Using
large aspect ratio expressions we find the width to be
1/ 2
2 ( / ) r q R r p A with r
the minor radius, and from Eq. (2.20) the drift is vertically downward for ions,
vertically upward for electrons. Here down is defined by the current direction:
in the configuration shown the toroidal field and plasma current are clockwise
from above; if one looks at the device from below they appear counterclockwise.
Clearly for confinement the minor radius must be large compared to the banana
width. This gives a condition on the poloidal field. For confinement of 3.5 MeV
alpha particles we find a minimum plasma current of
6
2 10 / r R [A].

FIG. 2.2. A trapped particle drift orbit in ITER. On the outward part of the orbit an ion moves
in the direction of the plasma current. Electrons do the opposite.
71
PHYSICS OF CONFINEMENT
The toroidal drift for a trapped particle becomes /
d
T u c A , where
/ T du u
l

is the time for one bounce period, giving

2
||
2
||
( )
1
( )
d
p
B
B
d I gq
T B

p
u u
v p
+
o
+
o
l
(2.21)
and
2
||
/ ( ) T d D B u p
l
. This precession is proportional to the particle
energy. Deeply trapped particles spend their time near the outer midplane
where for a low beta equilibrium /
p
B v o o is negative, so the precession is
in the direction of the toroidal current. For a deeply trapped particle we have
/
d
Eq r u , which upon restoring the units for energy
2 2
0
( ) m R u , distance ( ) R
and time
1
0
( ) u

becomes
0
/ ( )
d
Eq mRr u u . A convenient way to express this
is
2 2
0
( / ) / ( )
d
E mc qc rR u u , recalling that
2
mc 931 MeV per proton and
that
3
0
9.58 10 Z / B m u with Z the charge in proton units, B in Gauss and
m in proton masses. In this expression, if
0
u is regarded as a signed quantity,
the correct direction of precession is obtained also for electrons.
Toroidal precession of high energy particles is particularly important
because of the possibility of a resonant interaction of deeply trapped high energy
particles with otherwise stable MHD modes. The fishbone mode [2.18, 2.19] is a
prime example of this.
2.2.4. The drift kinetic equation
The collisional Vlasov or Boltzmann equation for guiding centre motion is
v ( )
f
f C f
t
o
+ V
o
.
(2.22)
where f , the distribution function, is a function of
p
v , u , c , E , and
||
(v / v) sign o , ( ) C f is the modification of f due to collisions, and v
.
is
the guiding centre drift velocity. For an axisymmetric equilibrium any function
of E , , P
c
is a solution of Eq. (2.22) with ( ) 0 C f . The collisions can be
approximated as being local and instantaneous, producing a flow of particles
in velocity space only. Further, for a Maxwellian distribution, , ( ) 0
M M
f C f ,
since the Maxwellian is the state of highest entropy. For ions and electrons the
collision term for the ion distribution f has the form ( ) ( , )
ii i ie i e
C f C f f + .
Analogous to the Braginskii equations, velocity space and flux surface
averages of this equation multiplied by powers of the velocity give rise to
equations coupling scalar quantities. Even if the system is in magnetohydro-
dynamic equilibrium, if one passes from a fluid to a kinetic description it is not
in thermodynamic equilibrium, so the interest is in the rate at which it approaches
the equilibrium state of zero density. The quantities of particular interest are the
72
GOLDSTON et al.
fluxes of ions and electrons across the equilibrium surfaces, as well as the rate
of momentum and energy flow, describing the quality of the confinement, and
the induced currents, important because they modify the equilibrium. Many of
these fluxes are due to toroidal geometry, and were first found by considering the
special nature of particle orbits in a toroidal system, given by Eqs (2.122.15).
There is an inward motion of trapped particles called the Ware pinch, induced
toroidal current called the bootstrap current, and different regimes of cross-field
transport depending on the magnitude of the collision frequency, known as
banana, plateau and PfirschSchlter domains. Resistivity is modified by toroidal
geometry, as well as the transport of momentum.
Many of these effects were calculated in detail by Rosenbluth et al. [2.20]
in 1972. A 1995 review by Kikuchi and Azumi [2.21] gives expressions for
velocity moments of the drift kinetic equation of first and third order showing
the parallel friction force, viscous and electric forces, and momentum and heat
sources. The equations take the form
|| ||
U LU M V S E
]

]
(2.23)
where L and M are friction and viscosity matrices and
||
U , V , S and E
are vectors of parallel heat flow, thermodynamic forces, momentum and heat
sources, and the electric field, respectively. Matrix inversion of this equation
gives a generalized Ohms Law.
Some of these neoclassical expressions can be estimated by simple heuristic
arguments, and we will do this in the following. The derivations assume a static
axisymmetric equilibrium, so while some of the derived expressions are observed
experimentally, there are also significant deviations from them due to fluctuation
spectra of unstable plasma modes as well as deviations from axisymmetry
caused by the discrete toroidal field coils. In particular, small scale fluctuations
produce electron heat transport an order of magnitude larger than the neoclassical
estimates, and also ion loss is typically a few times the neoclassical value.
Toroidal magnetic field ripple due to discrete field coils above a low stochastic
threshold value causes a very rapid loss of high energy trapped particles, and
if fusion alpha particles are to be confined this places a severe restriction on
machine design.
The collisions conserve the number of particles of each species; in a quasi-
symmetric system the sum over all particles of the canonical toroidal momentum
P
c
is conserved. These conditions are local, i.e. they hold on each flux surface.
The canonical momentum is
|| p
P g
c
p v . The time derivative of the total
canonical momentum is
3 3
v v ( )
s s
d df
P d P fd P C f
dt dt
c c c


l
(2.24)
73
PHYSICS OF CONFINEMENT
where
s

indicates a sum over species. The

p
v terms in P
c
do not contribute
to this expression, because for each species
3
v ( ) 0 d C f
l
. This is simply the
statement that collisions conserve the total number of particles of each species.
Thus the conservation of total canonical momentum, following from symmetry,
gives
3
||
0 v )
s
d C(f p

l
.
In a uniform magnetic field, classical cross-field particle diffusion occurs
because of a particles gyroradius. Collisions causing a 90 deflection in particle
velocity result in a displacement of the particle gyrocentre by one gyroradius.
A random walk approximation then gives for the classical diffusion coefficient
2
/ 2 D vp , where
ei
v v is the electronion collision frequency for a 90
collision and p the electron gyroradius, and similarly for ions.
In toroidal geometry the class of trapped particles (in large aspect
ratio having
1/ 2
||
v ( / ) v r R
1
< ) have a bounce frequency less than
1/ 2
( / ) (v / )
b th
r R Rq u , with v
th
the thermal velocity. For collision frequencies
larger than
b
u so that banana orbits do not exist, there still exist passing particle
orbits with excursions from the flux surface of r qp A . Collisions produce a
random walk with a step of this magnitude, and we obtain
2 2
/ 2
PS ei
D q v p
for the PfirschSchlter diffusion coefficient.
If the collision time is long enough so that banana orbits can be completed,
i.e.
ei b
v u < , the drift kinetic formalism is applicable. This collisionality domain
is called the banana regime. Assuming toroidal symmetry with nested magnetic
surfaces, the particle flux across a surface
p
v
is given by
3
v
v
p
p
f
d
.
v
v
V
I
V
l
(2.25)
Substituting the expression for v
.
from Section 2.2.3 we find
|| ||
v ( / ) /
p
q
.
v p p u V o o where we have used the large aspect ratio expression
( / )
p
r q r v V . This drift term has the opposite sign for electrons. Substituting,
we find for the average flux over a magnetic surface:
2
||
3
v
2 2
d
d f
p
u
r u
| `
o
I

o
. }
l
(2.26)
with the corresponding to the sign of the charge. Integrating by parts and
using
3
v ( ) 0 d C f
l
, and
3
||
v ) 0
s
d C(f p

l
, we find for the average flux
through surface r for a single species:
3
|| v
( )
v
4
q r
d d C(f ) u p
r
I
l l
(2.27)
74
GOLDSTON et al.
Thus the flux due to collisions among particles of a single species is zero.
In a two-component plasma, the ion and electron particle fluxes are equal, i.e.
collisional diffusion is ambipolar. Changing integration variables through
3
||
v 2 / d dEd
o
r p

(with o referring to the sign of the pitch) where the

summation is over the two signs of
||
p and substituting a Lorentz collision term
|| ||
( ) ( )
u u
C f f vp p o o the modification of the distribution function can be found
explicitly [2.22]. Substitution then gives for the average particle flux:
1/ 2
1/ 2
1.46
r dp
R dr B
u
v
| `
I

. }
(2.28)
as originally found by Galeev and Sagdeev [2.23, 2.24], where
B
u
is the
poloidal component of magnetic field. The diffusion rate in the banana
regime can be obtained heuristically using large aspect ratio expressions. The
diffusion is dominated by the trapped particles, and during a collision the step
size becomes the banana width,
1/ 2
2 ( / ) r q R r p A . In addition, since trapped
particles must have
1/ 2
||
v / v ( / ) r R
1
< , a corresponding smaller collisional
modification of the velocity direction is necessary to change the orbit. For
multiple small collisions
( )
2
t u v A , so the effective collision frequency for a
trapped particle is ( / )
eff ei
R r v v . In addition, only a fraction ( )
1/ 2
/ r R of the
particle population is trapped, giving for the diffusion coefficient in the banana
regime
2 2 3/ 2
4 ( / )
ei
D q R r v p , in agreement with Eq. (2.28) except for the
numerical constant. Plotted versus the effective collision frequency normalized
to bounce frequency, /
eff b
v v u

, the neoclassical diffusion coefficient then

has the approximate form [2.24] shown in Fig. 2.3. The transition region between
banana and PfirschSchlter domains is called the plateau domain. However,
a clear distinction between these domains is a result of the large aspect ratio
approximation. In a real tokamak the domains smoothly merge, and with an
aspect ratio of 4 the plateau domain is barely discernible.
Note that the diffusion scales as
2
p v , with v the effective collision
frequency for 90 scattering. This quantity thus scales with mass as
1/ 2
m since
for fixed energy ~ m p and ~1/ m v . In an axisymmetric configuration the
ion and electron rates must be equal, as required by Eq. (2.27). A more careful
analysis shows that they should approximately equal the slower electron value
[2.22]. In fact they are experimentally found to be a few times the ion neoclassical
value, indicating that the electron diffusion is not neoclassical.
By operating in a regime of reversed shear, stable to confinement
destroying fluctuations, several tokamaks achieved the theoretical minimum
neoclassical level of transport [2.25]. The stabilizing influence of very small
or even negative shear has led to the exploration of configurations with large
q on-axis, referred to as an advanced tokamak configuration. However, there
are associated restrictions due to the strong q dependence of stochastic ripple
75
PHYSICS OF CONFINEMENT
loss (Section 2.2.6), and difficulty of thermal equilibrium associated with these
configurations. The momentum equation involves the collision dependent
viscosity term. The contributing effects are termed parallel-, perpendicular- and
gyroviscosity, arising from collision induced random walks along and across the
field, and gyroviscosity is similar to bootstrap current (Section 2.2.5) insofar that
it is induced by collisions but its value is independent of collision frequency.
Experiments performed on TFTR [2.26] and Alcator C-Mod [2.27] show however
that momentum transport is not neoclassical, but much larger [2.25, 2.28]. Ion
momentum and thermal diffusivities are found to be comparable in magnitude
and vary similarly with plasma current and minor radius. This correlation is
consistent with the anomalous transport driven by collisionless electrostatic
instabilities including ion temperature gradient driven modes and collisionless
trapped electron modes.
FIG. 2.3. Neoclassical diffusion.
2.2.5. Bootstrap current
The collisional trapping and detrapping of particles in the banana regime
gives rise to a net toroidal current which is called the bootstrap current [2.29,
2.30]. It results from a small perturbation of the particle distribution from that
of uniform particle pitch ( v / v i
|
), and is due to toroidal curvature and the
76
GOLDSTON et al.
existence of banana orbits. It can be calculated by using the perturbed distribution
function of Section 2.2.4, but it can also be arrived at directly.
FIG. 2.4. Banana orbits originating at 0 = 0 on the same flux surface with opposite pitches.
Consider a flux surface v , and the two associated ion banana orbits
with opposite velocities at 0 u on the flux surface of their common origin,
as shown in Fig. 2.4. To be trapped they must have pitch / / r R r R i < <
at the midplane. The inside banana is co-moving at its initial point, and the
outside banana is counter-moving. Because of the density gradient, on this flux
surface the density of co-moving trapped particles will be larger than the counter-
moving by a factor ( / ) / 2
b
n dn dr W o , with
b
W the banana width, as shown
in Fig. 2.5. The bootstrap current is due to the associated asymmetry in the co-
and counter-moving passing particles necessary to maintain steady state through
small pitch angle scattering. Pitch angle scattering causes a diffusion in , and
the constant source of trapped particles leads to a uniform pitch distribution of
passing particles, with a density of 1 n o + for co-moving, and 1 n o for counter-
moving, as shown in Fig. 2.5. Substituting
1/ 2
( / )
b
W R r qp , multiplying by the
average velocity of the passing particles, / 2 p , and using the expression for the
pressure,
2
/ 2 p np , we find for the bootstrap current density
( ) / 1 r R <<
1/ 2
1
BS
r p
j
R B r
u
o
| `


. }
o
(2.29)
77
PHYSICS OF CONFINEMENT
This expression is for a given pressure, independent of mass. Electrons
produce a flow in the opposite direction, and thus a current also given by
Eq. (2.29). For an evaluation of the exact numerical coefficient for this expression
see Ref. [2.20]. Note that this current, although dependent on the collisions for the
equilibration of trapped and passing populations, is independent of the collision
frequency. Bootstrap current is not free, it is driven by the diffusion of particles
down the density gradient associated with the pitch angle scattering.

FIG. 2.5. Particle density versus pitch at the midplane. Particles with pitch / / r R r R i < <
are trapped. There is a greater density of co-moving trapped particles than counter-moving.
Pitch angle scattering then produces a uniform passing distribution in i, as shown by the
dashed lines.
In a stellarator the bootstrap current is more complex because of the
possibility of resonances, and important because it may be the only source of
current in the equilibrium. Particle motion in any equilibrium depends only
on the magnitude of the magnetic field, which can be expanded in Fourier
harmonics with respect to the poloidal and toroidal variables u and c ,
,
,
cos( )
m n
m n
B B m n u c

, where m, n are poloidal and toroidal mode numbers

respectively. The
, m n
B are perturbations of the magnitude of B
.
only, and do not
produce magnetic islands in the equilibrium. The analytic expression for the
bootstrap current for a monoenergetic distribution is /
BS BS t
j G f Edn dv ,
with
BS
G a geometric factor and
t
f the trapped distribution [2.31, 2.32] which
may contain resonant denominators.
78
GOLDSTON et al.
Trapped particles on orbits which take longer to complete than the time for
them to collisionally scatter into passing orbits do not contribute to the bootstrap
current. In the axisymmetric case (a tokamak) the correction has been analytically
calculated [2.33] by finding the collisionally modified distribution function in
the boundary layer between the trapped and untrapped particle regions of phase
space. The result is that in the banana regime the bootstrap current is decreased
by a function of the dimensionless collision frequency v

. Clearly, if the collision
frequency is so large that banana orbits do not exist, there is no bootstrap current.
Bootstrap current is especially important because of the possibility of
producing almost all of the current profile for a tokamak. This is critical because
the ohmic current is produced by pulsing the toroidal field coils, and thus it
cannot be maintained in steady state. The bootstrap current makes it possible to
extend the pulse length significantly. It was first observed in an octopole device
[2.34], where both the bootstrap current and the PfirschSchlter current, which
maintains charge neutrality in the presence of the diamagnetic current, were
measured. First observation of the bootstrap current in a tokamak was performed
on TFTR [2.35, 2.36] by comparing the measured surface voltage with simulations
with and without bootstrap current. It was also observed in a high density
discharge in the JET tokamak [2.37, 2.38], and later a large bootstrap fraction of
80% of the current was observed in the JT-60 tokamak [2.39, 2.40]. Knowledge
of the bootstrap current is necessary in order to confirm the generalized Ohms
Law for current in a tokamak, given by
|| ||
( )
BS
E j j n , with E
|
the parallel
electric field and
||
n the neoclassical resistivity. The validity of neoclassical
resistivity was confirmed on JET [2.38], JT-60 [2.39] and TFTR [2.41]. The 1995
review by Kikuchi and Azumi [2.21] gives experimental evidence for bootstrap
current and generalized Ohms Law.
2.2.6. Magnetic feld ripple
The toroidal field in a tokamak is not exactly axisymmetric due to the
finite number of toroidal field coils. A simple model for the field magnitude
including this ripple in a large aspect ratio circular device is given by
0
(1 cos cos ) B B r N u o o + , with r in units of the major radius R and o
the toroidal angle, N the number of coils and o the ripple magnitude. The
magnitude of the field along a field line is shown in Fig. 2.6. If the perturbation
is strong enough, magnetic wells are formed. Along a field line q o u , r
constant so ripple wells exist if sin r Nq u o < . To be trapped in one of these
wells (a superbanana), a particle must have
||
v / v o < . If trapped, the loss is
very rapid, since the particle simply drifts down directly to the wall, shown also
in Fig. 2.7: the particle executed a number of bounces in a poloidally trapped
orbit before encountering a ripple well with a phase allowing it to be trapped in
the well, after which it rapidly drifts to the wall.
79
PHYSICS OF CONFINEMENT

FIG. 2.6. Magnetic ripple along a particle trajectory.
FIG. 2.7. Loss orbit for a particle trapped in a ripple well.
Even if ripple wells do not exist, the phase space of banana tip trajectories
can become stochastic due to ripple. Because of particle drift, when the ripple
causes a banana orbit to bounce either early or late it produces a small change
in the radial position of the banana tip. Shown in Fig. 2.8 is the trajectory of a
trapped particle in resonance with the field coils, shown at the left with an .
Here 2
b b
q o u is the toroidal motion along the field and
p
o is the precessional
drift motion in one bounce. For small ripple the plot of radial banana tip position
80
GOLDSTON et al.
versus toroidal angle is periodic, but there are resonances producing islands at
radii where the precessional motion of the tip moves one coil spacing in one
bounce, as shown in Fig. 2.9. Typically there can be one hundred such resonances
across the minor radius. If the ripple is sufficiently large the islands due to these
resonances overlap and the banana tip motion becomes stochastic. Because of
conservation of and energy, the tip is confined to a surface of constant B , but
this means it can move vertically until the particle is lost, as shown in Fig. 2.10.
FIG. 2.8. Trapped particle orbit showing resonance with field coils.
b
u

and
b
u
+

are the positive
and negative bounce points, respectively.
FIG. 2.9. Poincar plot of banana tip motion showing ripple resonances.
81
PHYSICS OF CONFINEMENT
FIG. 2.10. Trajectory of stochastic ripple loss banana orbit.
The change of the bounce point due to the radial drift can be calculated
using 2
b b
q o u and B at the bounce points, ( , ) /
b b
B E v u giving
/ 2 / 2 /
p b
b b b p
d d dq d q B B
v u
o v u v o o . The motion becomes stochastic
approximately [2.42, 2.43] when the modification of the toroidal bounce point
A is as large as a ripple phase / 1
b
N d dr o A > . This loss produces an important
restriction on ripple in a reactor
[ ]
3/ 2
/ ( ) / ( ) r RNq q o r p s (2.30)
important only for high energies, particularly relevant for fusion product alpha
particles.
The first experiments to give a confirmation of stochastic ripple loss were
carried out on TFTR [2.44] measuring the loss of D-D fusion products. Two
clearly separated loss peaks in toroidal pitch angle were observed, the first
agreeing with the distribution of prompt loss particles and the second with those
due to stochastic ripple loss. Experiments were carried out on JET [2.45], where
the ripple magnitude can be adjusted by changing current in alternate coils. The
plasma response to increased ripple is quite obvious and dramatic. The most
obvious effect of increasing ripple is the large plasma density reduction in the
H-mode phase. Both the plasma average density and the edge density show a
pronounced pump-out. At the same time, the plasma electron temperature
remains, on average, approximately constant as the ripple is increased. Increasing
the ripple also results in a marked change of the plasma toroidal rotation. Since
banana orbits exit the plasma while moving in the co-current direction, there is
a transfer of momentum to the wall, and substantial plasma braking is observed,
with the reduction of the positive (co-current) plasma rotation measured across
the whole profile, the rotation becoming counter-current for ripple of 1%.
Increased values of the toroidal field ripple also affect the edge localized mode
82
GOLDSTON et al.
(ELM) behaviour. When the ripple is increased from 0.08% to 0.5%, the type
I ELM frequency almost doubles, going from ~12 Hz to ~20 Hz. When the ripple
is increased further to 0.7% and finally to 1%, ELMs become irregular, with
Type I, Type III and long ELM-free phases, in spite of the fact that power across
the separatrix remains approximately constant.
2.2.7. Transport in a stochastic feld
Magnetic perturbations with a field component across the equilibrium
flux surfaces can lead to a change in topology, with magnetic islands forming at
surfaces where the perturbation is resonant with the local q value. These islands
essentially short-circuit transport. Furthermore, if nearby islands overlap, field
lines can wander between them [2.46], producing a broad region in which the
field is stochastic, and in which a density gradient cannot be supported. Assuming
that the perturbation amplitudes are well above the threshold for stochasticity, a
random phase approximation may be used and the resulting diffusive motion of
particles was calculated by Rechester and Rosenbluth [2.47].
For times long compared to the collision time, the particles move along
the field line diffusively, i.e.
2
||
( ) t D o A , with
||
D the parallel collisional
diffusion. If the field lines diffuse across the equilibrium surfaces with diffusion
m
D , the particles will then move radially with
1/ 2
~
s
D t

giving zero for large

time; there is no cross-field diffusion due to diffusive motion along the field line.
However, another effect enters, namely the shift of the particle to a different field
line during a collision, due to the gyroradius. If the field is stochastic, nearby
field lines on the average diverge exponentially, with the distance between lines
given by
0
hz
d d e , z being the distance along the line and h the Kolmogorov
entropy. Then in a time
2
||
/
h h
L D t the particle will be carried a distance o
1

from the initial line, with (1/ ) ln( )
h
L h o
1
and o
1
the wavelength across the
unperturbed field, defining the decorrelation distance of the initial field.
The particle diffusion is then
||
v
m
m
h h
D D
D D
L L
v
i | `


. }
(2.31)
with
v
i the mean free path at velocity v.
A perfect testbed for the understanding of transport in stochastic magnetic
fields is provided by the reversed field pinch, which has a well known steady state
spectrum of saturated tearing modes. These modes produce a field which is well
above stochastic threshold, but nevertheless still possesses a great deal of structure,
including streamers formed by the partially destroyed island remnants. The field
is sufficiently close to stochastic threshold that a random phase approximation
for the particle motion is not valid, and thus RechesterRosenbluth transport is
83
PHYSICS OF CONFINEMENT
not operative. Transport in this system has been shown [2.48] to be subdiffusive
and non-local. Passing particles explore the stochastic field, participating in Levy
flights [2.49] with flight time and radial displacement distributions with long tails
that do not possess either a mean time or a mean distance. It is very possible that
electron transport in tokamaks is also non-local and non-diffusive, increasing the
complexity of understanding electron confinement.
2.3. TURBULENT TRANSPORT
2.3.1. Introduction
Confining hot and dense plasmas in the interior of fusion devices while
keeping boundaries from being damaged by excessive heat necessarily creates
an inhomogeneous state. Such magnetically confined high temperature plasmas
are not only unstable to various linear instabilities with different wavelengths and
complex frequencies, but often non-linearly self-organize themselves into more
energetically favourable states.
A large scale MHD instability, when it is violently unstable, can terminate a
tokamak plasma by throwing it into the wall, and this is called a disruption. Even
in the absence of large scale MHD instabilities, tokamak transport is usually
anomalous, which means that the transport rate is much higher than predictions
from theories based on Coulomb collisions. This anomalous transport is caused
by small scale collective instabilities which are typically driven by gradients in
temperature, density and so on. These microinstabilities saturate at low amplitude
due to non-linear mechanisms. As a consequence, particles E B
. .
drift radially
in a random manner, due to fluctuating electric fields. This is how plasmas escape
from the interior of a tokamak.
The amplitude of tokamak microturbulence is very low. The relative
fluctuation amplitude,
0
/ n n o , is typically less than 1% in the core. Towards
the edge, it can be greater than 10%. These general trends were confirmed
in different machines, using different diagnostics as described in detail in
Section 2.6. In summary, the properties of tokamak core microturbulence are
as follows:
0
/ ~1% n n o , ~ ~ 0.1~ 0.2
r i i
k k
u
p p ,
||
k 1/ qR k
1
<< , and
~ ~
E
k u
.
.
u u u

A , where
E
u
.
is the mean E B
. .
velocity. Here, n o and
0
n
are the fluctuating and mean parts of density respectively. v ( / )
i
i T i
M c eB p is
the thermal ion gyroradius, where v
i
T
and c are ion thermal speed and speed of
light respectively, and
r
k , k
u
,
||
k and k
1
are the radial, poloidal, parallel and
perpendicular (to equilibrium magnetic field) components of the k

vector of the
fluctuations respectively. q is the safety factor (see Section 2.3.5) and R is the
major radius. In addition, u is the mode frequency, and u A is the measured
broadening thereof. u

is the diamagnetic drift frequency. The broadband

frequency ( ~ u u A ) indicates strong turbulence. Sometimes the Doppler shift
84
GOLDSTON et al.
dominates in rotating plasmas, and this should be taken into account for the mode
identification.
FIG. 2.11. Massively parallel gyrokinetic simulation of ITG turbulence using the GTS code
[2.50] exhibits fluctuation spatial structures in 3-D and the effect of self-generated zonal flows
on them. From Powerful Beyond Imagination, SC06 International Conference for High
Performance Computing, Networking, Storage and Analysis, 1117 November 2006, Tampa,
Florida (http://sc06.supercomputing.org/video.php) [2.54]. Courtesy of W. Wang, Princeton
Plasma Physics Laboratory, and S. Klasky, Oak Ridge National Laboratory.
These properties of tokamak microturbulence are now being reproduced
from large scale gyrokinetic simulations. For instance, Fig. 2.11 is from an ion
temperature gradient (ITG) driven turbulence simulation using the Gyrokinetic
Tokamak Simulation (GTS) code [2.50] with shaping capability. Density
fluctuation contours in the presence of self-generated zonal flows [2.51] indeed
show that the average turbulence eddy size is around several ion gyroradii
across the magnetic field, and the fluctuation is aligned almost parallel to the
magnetic field, with parallel wavelength of the order of the connection length.
Microturbulence in a tokamak plasma consists of many modes or eddies
non-linearly interacting with each other. The small scale linear instabilities
(often called microinstabilities) are driven by the expansion of the free energy
source stored in n V ,
i
T V or
e
T V . To make collective waves unstable, these free
energies should be accessible for waves to tap. This is provided by dissipation
85
PHYSICS OF CONFINEMENT
including waveparticle resonant interaction or reactive mechanisms including
bad magnetic curvature and negative compressibility. Microinstabilities can be
classified into various modes according to the variety of the free energy source,
accessibility mechanism and magnetic geometry of a system. In this section, we
first illustrate a few basic examples, and discuss candidates which are currently
believed to be of experimental relevance in present day and future fusion
devices. Due to severe space limitation, we limit our discussion to electrostatic
fluctuations. A more detailed zoology of these microinstabilities can be found in
review articles [2.52, 2.53]. A partial list is given in Table 2.1.
A much more detailed description of physical kinetics of turbulent plasmas,
in particular on the subjects of spectral transfer, weak turbulence and strong
turbulence theories which cannot be covered here, can be found in Ref. [2.55].
TABLE 2.1. LIST OF ELECTROSTATIC MICROINSTABILITIES
( /
e s s n
k c L
u
u p

is the electron diamagnetic drift frequency defined with the

density gradient scale length
n
L , /
pe s s p
k c L
u
u p

is the electron diamagnetic

drift frequency defined with the pressure gradient scale length
p
L ; subscripts e
and i stand for electrons and ions respectively)
Classification:
Free energy
Spatio-temporal scales
(wavelength,
frequency direction,
rough magnitude)
Accesibility mechanism
for instability
Trapped ion mode
n ,
e
T or
i
T
u
p
e
u

Trapped ion precession

resonance (collisionless)
Collisions between trapped
and passing ions (dissipative)
Ion temperature gradient
mode
i
T
i
p >
pi
u

<
Bad curvature or
negative compressibility
Trapped electron mode
n or
e
T
i
p
e
u

<
Trapped electron precession
resonance (collisionless)
Collisions between trapped
and passing electrons
(dissipative)
Electron temperature
gradient mode
e
T
e
p >
pe
u

<
Bad curvature or
negative compressibility
86
GOLDSTON et al.
2.3.2. Examples of basic microinstabilities
In this section, we consider a slab geometry with a uniform magnetic
field
z
B Be
.
.
(where
z
e
.
stands for the unit vector in the z direction) and
n ,
i
T and
e
T varying only in the radial ( x ) direction. Fourier decomposition
of a fluctuating quantity in the y and z directions is understood, i.e.
( )
,
,
i k x t
k
k
e
u
u
u
o o

4 4

.
.
.
. .
2.3.2.1. Electron drift instabilities
Historically, electron drift waves (which propagate in the electron
diamagnetic drift direction, driven by n V ) were theoretically investigated first.
For these electron drift waves, the ion dynamics in a uniform magnetic field, in
its simplest non-trivial context (
||
( / ) v
i
T
k u >> , 1
i
k p
1
<< , and
i e
T T << ) can be
described by the fluid equations:
||i || i
d
m u e
dt
o o V 4 (2.32)
2
2

i
E pol
M c
b
u u u c
B t eB
o
o o o o
1 1
V 4 o
+ + V 4
o
.
. . .
(2.33)
Here, c and v
i
T
are the ion sound and ion thermal speed respectively,
/
s e i i
T T p p ,
n
L is the density gradient scale length, o4 is the fluctuating
electrostatic potential,
i
u o
|
is the fluctuating ion fluid velocity along B
.
, u o
1
.
is the perpendicular fluid velocity and
pol
u o
.
is the polarization drift velocity. By
using the linearized ion density continuity equation,
0
( ) 0
t i
n n u
.
o o o + V , we
obtain the ion density response:
2 2
2 2
2
0
s
i e
s
e
k c
n e
k
n T
o u o
p
u u

1
| `
4
+

. }
|
(2.34)
Electrons, on the other hand, can adjust quickly to potential variations, and
equilibrate along the magnetic field owing to their high velocity ( v /
e
T
k u
|

with v
e
T
being the electron thermal velocity), i.e.
0
0
e e
e n T n o o V 4 V
| |
leads
to the Boltzmann response for the electron density:
0
e
e
n e
n T
o o4
(2.35)
This is often called an adiabatic response, referring to the relatively slow
timescale of the wave compared to the electron transit motion ( v
e
T
k u
|
< ).
Since the Debye length
De
i is much shorter than the microinstabilitys spatial
87
PHYSICS OF CONFINEMENT
scales, the Poisson equation can be approximated by a quasi-neutrality condition:
e i
n n o o . Then, by using Eqs (2.34) and (2.35), we obtain the linear dispersion
relation for the electron drift wave:
2 2
2 2
2
1 0
s
e
s
k c
k
|
u
p
u u

1
+ (2.36)
Note that Eq. (2.36) describes a wave which oscillates in time, not an
instability which grows in time.
The radial flux of electron density carried by the electron drift wave is
given by v
e e r
n o o I . Here, denotes an ensemble average which is
replaced by an average over an ignorable coordinate (e.g. the toroidal direction
in an axisymmetric tokamak, or the poloidal direction in simple slab geometry)
or a time average. Since

v ( / )
r x
c B b e o o V 4
. .
. .
(where
x
e
.
is the unit
vector in the radial direction x) and
e
n o o ~ 4, we can easily observe that the
transporter v
r
o and the transportee
e
n o are 90 degrees out of phase and
0
e
I . Therefore, for non-vanishing particle flux carried by an electron drift
wave, a phase shift between
e
n o and o4 is required. This can be provided by
electron dissipation and the electron drift wave can be driven unstable by it. To
be more precise, this is dissipation on electron particles such that waves gain
energy from electrons. So unstable electron drift waves can drive a net outward
flux of electron density. We emphasize net here, since the total flux consists of
a diffusive part (always outward for normal, centrally peaked density profiles)
and a (convective) pinch part which could be either inward or outward. This will
be discussed in Section 2.3.6 in detail.
Depending on the electron dissipation mechanism, electron drift waves
can be classified into the collisionless drift wave which is excited by the inverse
Landau damping on electrons, and the collisional drift wave which becomes
unstable due to Coulomb collisions of electrons on either electrons or ions. The
collisionless electron drift instability has been sometimes called the universal
instability, since it can occur with a density gradient which must exist in
magnetically confined plasmas, even in the absence of collisions.
Since the ion Landau damping (when kinetic effects are included in the ion
response in Eq. (2.34)) has a stabilizing effect, the linear stability of this universal
instability in a sheared magnetic field was a subject of the highest theoretical
interest until the late 1970s [2.56]. While the final word was stability in a
sheared slab geometry [2.57], now it is believed that collisionless drift waves
can become unstable in toroidal geometry with trapped electron precession-wave
resonance as a more effective destabilizing mechanism, replacing the inverse
Landau damping of passing electrons.
88
GOLDSTON et al.
From the linearized drift kinetic equation which will be discussed in
Section 2.3.3, the perturbed electron distribution function
e
f o satisfies
0 0
v
v
e
E
e
e
f u F F
t m
o o o
o o
| `
+ V V V 4

. }
o o
.
| | |
|
.
(2.37)
where
0
F is the equilibrium disctribution function. Note that the left hand side
(LHS) contains the waveparticle resonance (i.e. the Landau damping) and
the right hand side (RHS) describes the competition between the destabilizing
influence of density gradient relaxation and the stabilizing influence of
electron heating. After subtracting the adiabatic electron response by defining
0
( / )
e e e
h f e T F o o o 4 , and taking the velocity moments for / v
e
T
k u
|
< ,
we get
3
0
v 1
2 v
e
e e
e
e T e
n e e
d h i
n T k T
o u u o o r
o


4 4

+ +
'


l
|
.

(2.38)
where only the dominant resonant part of
e
h o (a contribution from a pole
at v / k u
| |
) has been kept. From Eqs (2.36) and (2.38), we conclude that
the aforementioned phaseshift between
e
n o and o4 indeed destabilizes the
electron drift wave if
e
u u

< . This downshift of the eigenfrequency required

for excitation of the electron drift wave is easily provided by the polarization
drift effect contained in Eqs (2.33, 2.34) and Eq. (2.36). In the simple limit of
2 2 2
s
k c
|
<u ,
2 2
/ (1 )
e s
k u u p
1
e + , where
2 2
s
k p
1
comes from the polarization
drift. It is also straightforward to include the effect of an electron temperature
gradient in the first term on the RHS of Eq. (2.37). Then, one can show that
e
T V (for normal centrally peaked
e
T profiles) is stabilizing. At this juncture, we
warn the reader that, in a different system (e.g. in toroidal geometry with trapped
electrons),
e
T V can be destabilizing. Furthermore,
e
T is the main free energy
source for the hyperfine (
e
) scale electron temperature gradient (ETG) mode,
as the name suggests.
Another subset of electron drift waves is the collisional drift wave.
Since collisional here implies that the electrons, rather than the ions (recall
( / )
ee ei i e ii
m m v v v with m
i
, m
e
the ion and electron mass respectively) are
in a high collisionality regime, we can still use Eq. (2.34) for the ion density
response. For electrons, we assume v v /
e e
ei T T
k qR v
|
, such that typical
electrons suffer a collision before circulating around the torus over a connection
length, qR . We still look for a drift wave (such as in the case of a collisionless
drift wave) satisfying / v
e
T
k u
|
< . Then, we can use Braginskii fluid equations
[2.8]:
0 0
0
e E e
n u n n u
t
o o o
o
+ V + V
o
| |
.
.
(2.39)
89
PHYSICS OF CONFINEMENT
0 0
(1 ) 0
e e t e
e n T n n T o o o o V 4+ V + + V
| | |
(2.40)
and
2
2
0 0 0
v
3

2
e
T
e E e e e e e
e
n T u T n T u n T
t
o o o , o
v
o
| `
+ V + V V

. }
o
| | |
.
.
(2.41)
where
e
u
|
is an electron fluid velocity along the magnetic field, and
0.71
t
o and

3.16
e
, are the coefficients defined in Ref. [2.8]. For
2 2
v
e
T ei
k uv
|
, electrons diffuse and thermalize along the equilibrium magnetic
field on a timescale faster than one wave period. In this so-called semi-collisional
regime, the electron density response is given by
2 2
0
0.51 3
1 (1 ) 1

2 v
e
e e e e
t e
e e T
n e
i
n T k
o uv u u o
o n
, u u

| `
4
| `
+
'

. }
. } |
(2.42)
where
0
ln / ln
e e
d T d n n is the ratio between density gradient scale length
and electron temperature gradient scale length. As in the collisionless case
(Eq. (2.38)), instability occurs when u is downshifted below
e
u

(by
2 2
s
k p
1
), when 0
e
n . Then, by using the quasi-neutrality condition, we can
obtain a dispersion relation from Eq. (2.34) and Eqs (2.392.41). If we further
assume that
2 2
/ (1 )
e s
k u u p
1
+ , with
2 2 2
s
k c
|
<u , 0
e
n ,
2 2
2 2
( )
v
e
e e
s
e T
k
k
u v
y
p
u

~
|
(2.43)
where y is the linear growth rate. This is similar to the results in Refs [2.58, 2.59].
Positive
e
n (or
e
T V ) has a stabilizing influence [2.59]. While this collisional
drift instability is no longer relevant in the core (interior) of high temperature
plasmas in present day large tokamaks, it is applicable to the relatively cold
plasma edge, in particular for high collisionality plasmas as in C-Mod [2.60], and
previously in TExT [2.61].
2.3.2.2. Ion temperature gradient instabilities
Since a high central ion temperature is required for magnetic fusion,
the ion temperature gradient (ITG) driven instabilities are of primary interest
in magnetic confinement, and are now believed to be the major contributor to
anomalous ion thermal transport. In this subsection, simple illustrations are given
for both acoustic [2.62] and magnetic curvature driven [2.63] versions of ITG
modes, using a fluid description. Unlike electron drift waves, ITG modes can be
driven unstable even with electrons satisfying the Boltzmann response. In this
limiting case, no particle flux is produced by ITG modes.
90
GOLDSTON et al.
First, we consider the acoustic version of ITG modes in a straight magnetic
field with a radial gradient in ion temperature. Once again, we consider a simple
slab geometry with

z
B Be
.
.
, and gradients in the x direction. Ion dynamics is
described by the following set of linearized hydrodynamic (fluid) equations:
0 0
0
E
n u n n u
t
| |
.
o o o
o
+ V + V
o
(2.44)
0
1
i i
m u e p
t n
o o o
o
V 4 V
o
| | |
(2.45)
0 0
0
i E
p u p p u
t
| |
.
o o o
o
+ V + I V
o
(2.46)
where
i
p o is the perturbed ion pressure and
0
p is the equilibrium ion
pressure. Note that the E B
. .
flow
E
u
.
o is incompressible in a straight system.
Equations (2.44) and (2.45) contain the ion sound (acoustic) wave dynamics.
Assuming a flat density profile (i.e.
0
0 n V ) for simplicity, we can obtain a
dispersion relation from Eqs (2.442.46) and the quasi-neutrality condition with
Boltzmann electron response (Eq. (2.35)). This procedure yields a cubic equation
in u ,
2 2
1 1 0
i
T s
k c
|
u
u t u

I | `
+
'

. }


(2.47)
Here ( / )v
i i i
T i T T
k L
u
u p

is the ion temperature gradient diamagnetic

frequency,
i
T
L is the ion temperature gradient length, and /
e i
T T t . Equation
(2.47) describes the acoustic oscillation modified by the ion temperature gradient.
Since the expression in
{ } involves the adiabatic exponent I which quantifies
the compressibility, we can regard that expression to be the effective compress-
ibility. Equation (2.47) yields three roots in the limit
i
T s
k c
|
u u

:
2
1/3
2/3
3
( )
i
N
i
T s
k c e
|
r
u u

e (2.48)
with N

= 0, 1, 2. For N

= 1, we obtain an instability ( Im( ) 0 u > ) which propagates
in the ion diamagnetic drift direction with an acoustic character. For this root, the
effective compressibility (its real part) is negative! This negative compressibility
is a key physical mechanism of ITG instability in a straight system. Note that
while the instability gets stronger as k
|
is increased, this trend will be limited by
the ion Landau damping as v
i
T
k
|
gets closer to u . This important kinetic effect
will be discussed in Section 2.3.4.1.
Another important limiting case of the ITG mode is now discussed in a
toroidal geometry with a curved magnetic field. Once again, we only consider
the radial gradient of the ion temperature (with a flat density profile). Neglecting
91
PHYSICS OF CONFINEMENT
dynamics along
0
B
.
, the essence of ion dynamics in a curved magnetic field in
the long wavelength limit can be approximated by the following equations:
0
0
i
di
i
T
n
t n T
o
o
u
o
+ e
o
(2.49)
0
0
i E
T u T
t
.
o o
o
+ V e
o
(2.50)
Equation (2.49) describes a coupling between density fluctuations and
temperature fluctuations which comes from the particle velocity dependence of the
B V and curvature drifts in toroidal geometry. Here,
{ }
( ) / ( )
di i y
cT eBR k u is
the ion drift frequency evaluated at the low field side midplane. Equation (2.50)
simply describes the relaxation of the temperature gradient via perturbed E B
. .

motion leading to a temperature fluctuation. With a quasineutrality condition for
Boltzmann electron response, we obtain a linear dispersion relation
2
i i
e e
di T di T
i i
T T
T T
u u u u u

(2.51)
Since 0
di
u < at the bad curvature low field side, and 0
i
T
u

< for normal

temperature profiles, Eq. (2.51) clearly predicts the existence of a purely growing
reactive instability in this simple limit. A pictorial description of the feedback
loop leading to this instability can be found in Fig. 1.1 of Ref. [2.64]. In reality,
ITG modes in tokamaks contain both toroidal and acoustic characteristics. While
their amplitudes peak at the bad curvature low field side, such mode structures
are only possible by having non-vanishing k
|
, not kept in Eqs (2.49, 2.50).
Further discussion of the mode structure can be found in Section 2.3.5.2.
2.3.3. Non-linear gyrokinetic equations for tokamak turbulence
For more detailed discussions of microinstabilities and turbulence
driven transport in the coming sections, we present the basics of the non-linear
gyrokinetic equation. Despite the tremendous increase of computational
power in recent years, the direct simulation of actual size fusion plasmas in
realistic geometry using the primitive non-linear plasma equations (such as the
Klimontovich [2.65] or Vlasov equations) is still far beyond the computational
capability of even the foreseeable future. Thus, reduced equations have been
employed to simplify the basic dynamical equations. In this section, we discuss
basic procedures involved in derivations from the most fundamental, Vlasov
equation, to the non-linear gyrokinetic equations which are now widely used
in various turbulence research in magnetically confined plasmas. We focus our
discussion on collisionless plasmas.
92
GOLDSTON et al.
Evolution of a collisionless plasma can be described by the Vlasov equation
for the particle distribution function for each particle species in a six-dimensional
phase space,
1
v v 0
v
f f q f
E B
t x m c
o o o
| `
+ + +

. }
o o o
. .
. .
. .
(2.52)
where E
.
and B
.
are the electric and magnetic fields produced by the particle
motion which satisfy Maxwells equations. This system of equations can describe
various phenomena over a wide range of spatio-temporal scales. For instance,
accurate descriptions of particle dynamics in the presence of high frequency
electromagnetic waves are needed for studying plasma heating by waves [2.66].
For turbulence and transport problems in magnetically confined plasmas, the
timescales of collective electromagnetic fluctuations of interest are much longer
than the period of a charged particles cyclotron motion (gyromotion), while
the wavelengths and correlation lengths of such fluctuations are much smaller
than the device size or the scale length of the magnetic field inhomogeneity. In
these circumstances, details of the charged particles gyration motion are not
of physical interest, and it is desirable to develop a reduced set of dynamical
equations which still captures the essential features of the low frequency
phenomena of practical relevance. By decoupling the gyromotion, one can
derive the gyrokinetic equation which describes the spatio-temporal evolution of
the gyrocentre distribution function, which is independent of the gyrophase, c,
defined over a five-dimensional phase space ( , v , R
|
.
). In simulating strongly
magnetized plasmas, one can, thus, save enormous amounts of computing time
by having a time step greater than the gyroperiod, and by reducing the number of
dynamical variables. We note that in the gyrokinetic approach, the gyrophase is
an ignorable coordinate and the magnitude of the perpendicular velocity enters as
a parameter in terms of an adiabatic invariant (
2
v / 2B
1
). This is the leading
term of an adiabatic invariant which does not change for each particle.
Due to the broad scope of this book and space limitations, it is not
appropriate to describe here the rigorous derivations of the non-linear gyrokinetic
equations in toroidal geometry. Details can be found in the literature, which
includes:
(i) a conventional perturbative derivation with introduction of a standard
non-linear gyrokinetic ordering [2.67],
(ii) a derivation mainly for the purpose of particle-in-cell simulation [2.68],
(iii) the Hamiltonian derivations of non-linear gyrokinetic equations for
electrostatic fuctuations [2.69], and electromagnetic fuctuations [2.70]
in a uniform magnetic feld, and fnally,
(iv) generalizations to arbitrary geometry using phase-space Lagrangian
derivations (which are the most transparent and effcient to date
93
PHYSICS OF CONFINEMENT
in the authors opinion) for electrostatic fuctuations [2.71], and
electromagnetic fuctuations [2.72]. There is a theoretically oriented
review ot the subject in Ref. [2.5].
Both formulations and simulations traditionally focused on tokamak core
turbulence in which the fluctuation amplitude is relatively small, i.e. the relative
density fluctuation amplitude is less than 1% and the gradients in macroscopic
parameters such as pressure are relatively mild with characteristic lengths of
the order of a fraction of the minor radius. In the non-linear gyrokinetic theory,
there exist three expansion parameters: /
ci u
r u O with
ci
O the ion cyclotron
frequency, /
B i B
L o p with
B
L the scale length of the equilibrium magnetic
field, and
0
/ /
e
f F e T r o o
4
4 , where
u
r ,
B
o and r
4
characterize the slow
timescale associated with turbulence, weak inhomogenity of the equilibrium
magnetic field and low relative amplitude of fluctuations respectively. The
conventional gyrokinetic theory [2.67] assumes that all three parameters are
comparable in formal ordering, although this choice needs to be modified in
some cases including plasmas with transport barriers [2.73]. In addition, in the
non-linear gyrokinetic ordering, the perpendicular wavelength can be comparable
to the gyroradius, while the parallel wavelength of the fluctuation is comparable
to the connection length of the system qR . Thus, the ordering is consistent with
the spatial structure of the fluctuation which aligns with the equilibrium magnetic
field (see Fig. 2.11).
Here, we simply sketch a heuristic derivation procedure in a uniform
magnetic field to illustrate some key features in the non-linear gyrokinetic
equations. This is similar to one employed in Ref. [2.68]. For more details,
readers should consult the aforementioned literature.
Transforming from the particle variables ( , v x
. .
) to the guiding centre variables
( , , v , R c
|
.
) [2.74], R x
.
. .
p ,
2
v / 2B
1
,
1 2
v v v ( cos sin ) b e e c c
1
+ +
|
.
. .
,
we can write Eq. (2.52) as

v
v
c q
f b f E b f E f
t B m
R R
o o o o
+ + +
o o
o o
| |
|
. . .
. .

2
v v
0
v
ci E
ci ci
q
f f
f
mB c c c
1
1
O
o o4 o o
O + + O
o o o o
. .
(2.53)
where

v ( / )
E
c B E b
. .
.
is the E B
. .
velocity. For
ci
< u O , the lowest order
part of Eq. (2.53) is / 0
ci
f c O o o . Writing
a
f f f + with
a
f f in
which indicates a gyrophase average, and
a
f is the gyrophase dependent
part, the next order equations:
94
GOLDSTON et al.

v
v
c q
f b f E b f E f
t B m
R R
o o o o
+ + +
o o
o o
| |
|
. . .
. .

0
ci a
q
f f
mB c
| ` o o
O 4

o . o }
(2.54)
Here, we have ignored the last term in Eq. (2.55) which corresponds to a
small local correction to the gyrofrequency
ci
O due to the perturbed E
.
field
which is not of our primary interest. Gyrophase-averaging Eq. (2.54), we obtain
the desired gyrokinetic Vlasov equation in a uniform magnetic field,

v 0
v
c q
f b f E b f E f
t B m
R R
o o o o
+ + +
o o
o o
| |
|
. . .
. . (2.55)
We emphasize that Eq. (2.55) is in the reduced gyrocentre coordinates
( , , v R
|
.
), i.e. without c , rather than in the particle coordinates ( , v x
. .
).
To complete the gyrokinetic description of the influence of the particle
motion on the electromagnetic field, one needs to express the charge density and
the current density in Maxwells equations in terms of the gyrocentre distribution
function ( , , v , ) f R t
|
.
. This requires the so-called pull-back transformation
to the particle coordinates [2.75]. Subtracting Eq. (2.55) from Eq. (2.54), and
keeping the lowest order term only, we obtain
( )
a
f
q
f
mB
o
4 4
o
(2.56)
Despite a widespread misconception that the gyrokinetic equation keeps
only the gyrophase independent part (and throwing away the gyrophase dependent
information), the gyrophase dependent part of the distribution function
a
f in
Eq. (2.56) has been kept in the gyrokinetic theories [2.682.70], and has been
shown to play a crucial role. It can be shown that by performing a phase-space
integral (equivalently a pull-back transformation), the ion polarization density
( )
pol
n x
.
, which accounts for the difference between the actual ion particle
density and the gyroaveraged gyrocentre density, is
3
( ) v ( )
a
n x d Rd d d B R x f
u
c o p +
l
|
. .
. . .
=
0
0
(1 ( ))
ik x
i
k
k
i
en
b e
T
o
1

I 4

.
.
.
. (2.57)
for Maxwellian f , and
ik x
k
k
e o

4 4

.
.
.
. with
k
o4 being the k component
of 4. Here,
2 2
i i
b k p
1
,
0 0
( ) ( )
i
b
i i
b I b e

I , and
0
I is the modified Bessel
function. This is the ion polarization density. In the gyrokinetic quasi-neutrality
condition, it is the sum of the gyrocentre density at the particle position and this
95
PHYSICS OF CONFINEMENT
polarization density, which should be equal to the electron density. We reiterate
that the effect of the polarization drift appears in the gyrokinetic Poisson equation
as the polarization density, rather than in the guiding centre drift in the Vlasov
equation. That representation of the polarization drift as a shielding term has
been found to be very useful computationally [2.68].
By taking the small Larmor radius limit ( 0
i
k p
1
) of the gyrokinetic
equation, we can recover the drift kinetic equation. However, the drift kinetic
equation can be derived without assuming a small relative fluctuation amplitude
[2.76]. Therefore, the drift kinetic equation is not a subset of the gyrokinetic
equation in a strict sense.
Another example of the reduced non-linear kinetic equations is the bounce-
averaged kinetic equation. When one is interested in low frequency fluctuations
such that the characteristic mode frequency u is lower than the bounce
frequency of charged particles of species j,
bj
u u < , the bounce action (or the
second adiabatic invariant) v J dl
l
|

is conserved where dl
l
is taken along
the bouncing particle orbit. Then one can decouple the bounce motion from
lower frequency dynamics related to the wave and the slower charged particle
precession motion in the toroidal direction (see Section 2.2.3). Consequently
the bounce angle becomes an ignorable coordinate, and a non-linear kinetic
equation can be written for a distribution function ( , , , ) F J o v , where o
and designate the equilibrium magnetic field, and / v d dl v
|
is an angle
variable for the motion along the field. There exists a conventional derivation
for application to trapped-particle-driven microturbulence [2.77], as well as a
modern derivation based on the phase-space Lagrangian method [2.78].
The bounce-averaged kinetic equation is useful for applications to the
trapped electron mode (TEM) and the trapped ion mode (TIM), as will be
discussed in Section 2.3.4.3. It is possible to calculate the real trapped particle
density via the pull-back transformation from the bounce-averaged banana centre
distribution function ( , , , ) F J o v in the reduced phase space. From this, one
can identify the neoclassical polarization density [2.78], which is closely related
to the neoclassical polarization shielding [2.79].
2.3.4. Kinetic description of microinstabilities
In Section 2.3.2.2, basic microinstabilites in a strong magnetic field have
been discussed in the context of a fluid picture. While the fluid treatment offers
a simpler physical picture and can be a good approximation when instability is
very strong, a kinetic approach leads to more accurate results in most cases. In
this section, we discuss kinetic effects, in particular the waveparticle resonant
interaction (Landau damping). This section is organized in the order of the linear
stability criterion for ITG modes, properties of ETG modes and finally trapped
particle instabilities.
96
GOLDSTON et al.
2.3.4.1. Linear onset condition for ion temperature gradient mode
In Section 3.2.2.2, we discussed the fluid treatment of ITG instabilities. For
the acoustic version, v
i
T
k u
|
was assumed ignoring the ion Landau damping
which occurs for v / k u
| |
. As a consequence, the resulting dispersion relation,
(2 /3) 1/3 2/3
| | ( )
i
i
T s
e k c
r
u u
|
indicates an instability even for an arbitrarily
weak ion temperature gradient. This is not intuitively plausible. Indeed, as
i
T V
becomes weaker,
i
T
u

will decrease, making u smaller. At some point, the fluid

approximation v
i
T
k u
|
cannot be satisfied anymore, and one should pursue
a kinetic theory which captures a proper behaviour for v
i
T
k u
|
. Furthermore,
one needs to relax the assumption of a flat density profile which was employed in
Section 2.3.2.2 for a simple illustration.
Starting from the electrostatic gyrokinetic equation in a straight magnetic
field, Eq. (2.55), we can make a linear approximation ignoring non-linear terms
involving o4 and f o , after expanding
0
F F f o + , and assuming that o4
and f o are infinitesimally small. Then, the non-adiabatic part of the perturbed
gyrocentre distribution function g o satisfies:
( )
0

v 0
t
c
g b F
t B
o o o
o
| `
o + V + 4 + V 4 V

. }
o
| |
.
(2.58)
Here,
( )
0
/
i
g f e T F o o o + 4 , and
0
( )
i
J k o p o
1
4 4 is the
gyrophase-averaged perturbed potential, where
0
J is the Bessel function. With an
adiabatic electron density response, the quasi-neutrality condition yields:
3
0
v( )
e i
e e
d J g
T T
o o
o
4 4
+
l
.
(2.59)
After evaluating
0
F V
.
for a local Maxwellian
0
F with non-uniform density
and temperature profiles, g o is given by
2
2
0 0
v 3
1
2 2v
v
i
i i
T
i
e
g J F
k T
u n u
o
o
u

| `

+ +
'

. } | ` 4

. }
| |
(2.60)
where
0
ln / ln
i i
d T d n n is the density gradient length scale divided by the
ion temperature length scale and
( )
/ v /
i
i i e e i T n
T T k L
u
u u p

. Then,
Eq. (2.59) becomes
97
PHYSICS OF CONFINEMENT
2 2
2 2
v / 2v
2
0
2
3/ 2 3
0
v v
3
1 ( )
2 2v
2
v v v
v (2 ) v
T
i
i
i
i
e i i
e T
T
T
J k e
T
d d
k
u u n p
r
u r
1
1
~ ~
1 1
~

| `
+

+ +
'

. }

l l
|
|
| |
1
i
e
T
T
+

(2.61)
By taking a long perpendicular wavelength limit ( 0
i
k p
1
), and the
fluid approximation ( v k u
| |
) in the denominator, we can recover the acoustic
version of the fluid ITG dispersion relation (Eq. (2.47) from Eq. (2.59) for 0 I
and 0 n V ).
In the kinetic regime with strong ion Landau damping, / k u
|

v
i
T
, we can determine the linear threshold condition for ITG modes by noting
that stability is determined by the imaginary part of Eq. (2.59). Note that
{ } L.H.S. / ( v ) 1
i
e T
O k u

]
|
meanwhile,
{ } { } Re L.H.S. Im R.H.S. < and
R.H.S. (1) O . Formally the imaginary part dominates. Since ion heating (due
to the usual ion Landau damping in a homogeneous plasma) is negligible
(
e
u u

< ), the linear onset condition corresponds to near zero relaxation of

the velocity dependent free energy. It is determined by
2 2
2 2
2
0
2
v / 2v
0 0
v v
3
1 ( )
2 2v
lim Im v v v
v
i
T
i
i
e i i
e T
T
J k
T
d d e
k
u
u n p
u
1
1
~ ~

1 1
~

| `
+

+
'

. }

l l
|
|
| |
0 (2.62)
Tracing back the physics origin of the numerator (the radial relaxation of
free energy), it is
2 2
0
2
v v
3
1 v ln
2 2v
i
i
i e
i T i
e T
T
k F
T r
u
u
n p
1

| `
+
o
+
'
o
. }

|
(2.63)
By interchanging the order of
0
v d
~
1
l
and / r o o , we obtain
2 2
v / 2v
0 2
0
3/ 2
0 0
1
lim Im v v v ( )
v
T
i
i
i
n
d d e J k
k r T
u
p
u
~ ~

1 1 1
~
]
| `
o
]

o
. }
]
]
l l
|
| |
2 2
( / ) / 2v
0
0
1/ 2
0
( )
lim ( ) 0
( )
T
i
k
i
i
n r
e b
r T r
k
u
u
r

]
o
I
]
o
]
]
|
|
(2.64)
where
2 2
i i
b k p
1
. Therefore, for / v
i
T
k u
|
< ,
98
GOLDSTON et al.
0
0
1/ 2
( )
( ) 0
( )
i
i
n r
b
r T r
]
o
I
]
o
]
]
(2.65)
defines the linear threshold of ITG instability. Note that given a wavelength
k
1
and density profile
0
n ,
1/ 2
0 0
( ( ) / ) ( )
i i
n r T b I gives a critical temperature
profile ( )
i
T r . The onset condition for ITG instability can be written as
1
0
2
( )
1 2 1
( )
i
i
i
i
I b
b
I b
n >
| `
+

. }
(2.66)
where
0
I and
1
I are modified Bessel functions. Equation (2.66) is the well
known result by Kadomtsev and Pogutse [2.80]. This shows that an instability
occurs if
i
n exceeds a certain value of the order of unity, roughly 1~2. Due to
this reason, ITG modes are sometimes called the
i
n modes. Since the threshold
is lowest for ~1
i
k p
1
, short wavelength modes with 1
i
k p
1
are more easily
excited by the ion temperature gradient. However, ITG mode is a better name
since there exists a threshold in terms of
i
T V , even for a flat density profile such
that
i
n ~ . An illuminating analytic formula for the instability condition in a
sheared magnetic field (see Section 2.3.5.1) is given in Ref. [2.81]
1.9 1
i
s i
T e
L T
L T
| `
> +

. }
(2.67)
where /
s
L qR s is the magnetic shear length discussed in Section 2.3.5.1 and
ln / ln s d q d r is the magnetic shear. This indicates the favourable roles of
high /
i e
T T and high magnetic shear for ITG stability.
In toroidal geometry, one should consider the B V and curvature
drift (
di
u ) in the gyrokinetic equation, and perform a similar but much
more complicated kinetic calculation including a resonance given
by (v , v ) v 0
di
k u u
1

| | |
. The 0 k
|
limit has been considered (while
0 k
|
and ballooning mode structure strongly peaked at the low B
.

field and bad
curvature side are not totally compatible) for which some analytic progress has
been made [2.82]. Furthermore, a powerful Nyquist diagram technique has been
applied to this problem and appeared in a textbook [2.83]. However, keeping
both k
|
and
di
u requires a numerical calculation [2.84]. For flat density, the
linear instability condition given by Ref. [2.82] is
4
1
3
i
i
T e
T
R
L T
| `
> +

. }
(2.68)
99
PHYSICS OF CONFINEMENT
2.3.4.2. Electron temperature gradient instability
The electron temperature gradient instability [2.85, 2.86] is a very short
wavelength fluctuation with 1
i
k p
1
. Since the ions with Larmor radius greater
than the perpendicular wavelength of fluctuations feel the gyroaveraged electric
field which is much smaller than the local electric field at one position, the ion
dynamics becomes almost unmagnetized. Formally speaking, as 1
i
k p
1
, the
gyroaveraged reduction of potential (
0
( ) 0
i
J k o p o
1
4 4 ) leads the ion
response to be adiabatic (since 0 g o ), i.e.
0
/ /
i i
n n e T o o o 4 . On the other
hand, with a general ordering 1
e
k p
1
with
e
p being the electron gyroradius,
now the electron dynamics is described by the gyrokinetic equation. In this sense,
the roles of ions and electrons are reversed for the ETG mode, compared to those
for the ITG mode where the electron response was assumed to be adiabatic for a
different reason. Therefore, in the simplest context, there exists an isomorphism
between ITG and ETG. For instance, simplified onset conditions in Eqs (2.67,
2.68) apply to the ETG modes as well, with subscripts i and e exchanged. Now,
high /
e i
T T is favourable for ETG stability.
Because the ion response is near adiabatic, the ETG mode produces
negligible particle flux, ion heat flux and momentum flux. Since the radial
width of ETG modes is related to the electron gyroradius,
e
p , it cannot produce
significant electron thermal transport unless there is a mechanism to enhance
its radial width. Some gyrokinetic simulations report the existence of radially
elongated fluctuations, so-called streamers [2.87, 2.88]. We discuss this further in
Section 2.3.5.2. The relation between ITG and ETG is summarized in Table 2.2.
2.3.4.3. Trapped particle instabilities
Electron drift waves destabilized by magnetically trapped electrons are
called trapped electron modes (TEM). In present day tokamak plasmas with low
collisionality, / 1
e ei be
v v ru

< (where
0
/ r R r is the inverse aspect ratio
and
be
u the bounce frequency of trapped electrons) the trapped electron drive
is typically stronger than that from the passing electrons which were discussed
in Section 2.3.2.1. The destabilizing influence of passing electrons scales with
/ ( v )
e
e T
k u
|
for collisionless electron drift waves, and with
2
/ ( v )
e
e ei T
k u v
|

for collisional drift waves.
As can be seen from the bounce averaged drift kinetic equation for trapped
electrons, after a decoupling of the relatively fast bounce motion along the B
.

field, the trapped electron dynamics across the B
.
field has a hydrodynamic
character describing E B
. .
motion and toroidal precession. The expansion-
free energy in n V or
e
T V is released by the radial component of E B
. .

motion. When the energy is transferred from trapped electrons to waves via
wave-precession-drift resonance, the mode is called the collisionless trapped
100
GOLDSTON et al.
electron mode (CTEM), while if the energy transfer occurs via trapped
electron collisions, it is called the dissipative trapped electron mode (DTEM).
Both trapped electron instabilities require that most trapped electrons be in
the banana collisionality regime 1
e
v

< , such that they can execute a bounce

motion before getting detrapped by collisions on average. Furthermore, both
instabilities also require
be
u u < , such that the second adiabatic invariant
v J dl
l
|

is conserved. There also exist trapped ion modes with extremely low
frequency /
bi
u u (where
bi
u is the bounce frequency of trapped ions) and long
wavelength. We discuss those in Section 2.3.4.3.3.
2.3.4.3.1. Collisionless trapped electron mode
The collisionless trapped electron mode (CTEM) can occur if collisions
are rare enough that a trapped electron can also precession drift around the torus
more than once in the toroidal direction before it gets detrapped by collisions.
For the CTEM, the collisionless bounce-averaged drift kinetic equation [2.77]
provides a very good description of trapped electron dynamics:
TABLE 2.2. ITGETG ISOMORPHISM [2.51]
(
o
, ,
i
, ,
e
, and
J
, are diffusivity for toroidal momentum, ion heat, electron
heat and toroidal current respectively;
e
n and
i
n are perturbed density for
electrons and ions respectively)
Key issue ITG ETG
Linear response in the
electrostatic limit
i
n
From gyrokinetic equation /
i
e T 4 ; pure adiabatic
e
n
( )
/
e
e T 4 4 ;
adiabatic with zonal flow
From gyrokinetic equation
Disparity in transport
channels caused by
particular turbulence
,
i e
D
o
, , , > , ,
e J i
D
o
, , , , >
Zonal flow strength in
non-linear regime
Typically strong Typically weaker
Radial correlation
length of ambient
turbulence at non-linear
saturation
Several
i
p
Uncertain current research
Isomorphism breaker Zonal flow
Residual magnetization of ion
response
Electromagnetic effect
Debye shielding
101
PHYSICS OF CONFINEMENT
0

v 0
T
de e
c
h b F
t t B
o o o
o o
| ` | `
+ V 4+ V 4 V

. } . }
o o
.
(2.69)
where
T
e
h is the non-adiabatic part of the perturbed trapped electron distribution
function, o4 is the bounce-averaged o4, and v
de
.
is the precession velocity. In
Fourier space, Eq. (2.69) becomes
0
( ) ( ( ) )
e
T
de e T
e
e
i h i E
T F
o
u u o u u

4
(2.70)
where
{ }
( ) 1 ( / 3 / 2)
e
T e e e
E E T u u n

+ . As described in Section 2.3.2.1,
the imaginary part of the non-adiabatic electron density response determines the
stability of the electron drift wave. Therefore, for the CTEM [2.89],
3
0
0
1
Im( / ) Im v
NA T
e
n n d f
n
o o
l
3/ 2
/ *
3
2 1 1
2
de
G e
e
de de e
e
e
G G T
u u
u o u u
rr n
u u u

]

| ` | ` 4

+ ]
'

. } . }
]
]
(2.71)
where
NA
n is the non-adiabatic density response, and T in
T
e
h stands for
trapped. Note that an instability can occur (with 0
NA
n o < ) for
e
u u

< in the
absence of the electron temperature gradient. For the CTEM,
e
T V (or positive
e
n ) is destabilizing, in contrast to electron drift instabilities driven by passing
electrons. Of course, this contribution comes from the
de
u u resonance. Since
( / )
de de e
G E T u u with / ( )
de e
cT eBR u , mostly high energy electrons
with / / /
e e de n
E T G R GL u u

resonate with electron drift waves. Here, a

dimensionless parameter ( , ) G s x captures the pitch angle x and the magnetic
shear s

dependence of the trapped particle precession motion [2.89].

Furthermore,
de
u can change its sign 0 G < for barely trapped electrons.
These precession-reversed electrons cannot resonate with the electron drift wave
and have a stabilizing influence on it. For reversed magnetic shear plasmas, a
significant fraction of trapped electrons can precess in the favourable direction
[2.90] (opposite to
e
u

) and the CTEM can be stabilized.

2.3.4.3.2. Dissipative trapped electron mode
This instability is relevant for an intermediate collisionality regime
satisfying /
be ei e
u v r u

> > . So on average, typical trapped electrons get

detrapped by collisions during one wave period, but not during a shorter bounce
motion period. Then the bounce-averaged kinetic equation can still be used, but
with collisions included. While the conservation properties of the collisional
operator are important for a quantitatively accurate derivation, here we simply
102
GOLDSTON et al.
illustrate the basic trends of parametric dependences. If the effective collisional
frequency is higher than u and
e
u

, the simplest form of non-adiabatic trapped

electron response is given by
3
0
3
1
2
/ v
/
e e
NA T
e e
Tr
ei e
e
n n d f i
T
u u n
o
o o r
v r

| `
+

4 . }

l
(2.72)
For 0
e
n , instability occurs when
e
u u

< . 0
e
n > is destabilizing as in
the case of CTEM. From this, the linear growth rate of the DTEM scales like
( )
3/ 2
/
e ei e
y n r v u

~ [2.91].
In reality, for given plasma parameters, one should solve a very complicated
linear dispersion relation including all the effects discussed in Sections 2.3.4 and
2.3.5. Each mode we have discussed in each subsection is a very idealized
limiting case. For instance, the adiabatic electron response for ITG was used
just to show that the electron dissipation is not required for ITG excitation.
Even with ITG dominated turbulence, significant electron thermal and particle
transport can occur. However, this simplified classification is still useful since it
gives an understanding of the global picture of the multidimensional parameter
space ( /
i
T
R L , /
n
R L , collisionality, ...). On the other hand, even the most
comprehensive, realistic and large scale simulation can be only done for one set
of parameters at a time.
2.3.4.3.3. Trapped ion mode
While most of the drift waves discussed in this chapter are characterized
by perpendicular wavelengths which scale with the ion gyroradius, there can be
much longer wavelength electrostatic instabilities. Since the frequency of drift
waves scales linearly with k
o
, for 1
i
k p
1
< , long wavelength modes have
low frequencies. If the mode frequency is smaller than the bounce frequency
of trapped thermal ions, i.e. v / ( )
i
bi T
qR u u r < , most passing ions can
make a transit motion around the torus at least a few times during one wave
period. Therefore the passing ion dynamics can be approximated by an adiabatic
response, while the trapped ion dynamics should be described by the bounce-
averaged kinetic equation. This is almost identical to the treatment of the
electron dynamics for TEMs.
The trapped ion modes (TIMs) can occur for
bi
u u < , corresponding to
toroidal mode number 10 n < in major tokamak core plasmas, and for ions in
the banana collisionality regime 1
i
v

< . As TEMs, the trapped ion instabilities

can fall into two distinct classes depending upon whether they propagate
in the electron or the ion diamagnetic direction. The dissipative version
(
,
/ ,
i eff ii e
v v r u u

> ) of the first category is usually called the dissipative

trapped ion mode [2.80, 2.92], while the dissipative ion temperature gradient
103
PHYSICS OF CONFINEMENT
driven trapped ion mode [2.93] which occurs for 2 / 3
i
n > , falls into the second.
There also exists a collisionless trapped ion mode (with
, di i eff
u v > ) [2.90]. The
trapped ion instabilities were considered to be potentially very dangerous to
confinement due to their large radial scales. Indeed, rather persistent trends of
confinement scalings closer to Bohm rather than gyroBohm reported from major
tokamak experiments justify attention to this class of large scale instabilities
which were also observed in a comprehensive two-dimensional kinetic
calculation [2.94].
2.3.5. Spatial structure of microturbulence
From the spatio-temporal scales of tokamak microturbulence, one can
make a very rough estimate of the transport coefficient
turb
D using dimensional
analysis based on a random walk argument,
2
2 2 2
( )
i i
turb
r r r i
k cT
r
D
t L eB k k k
u
p u y
p

| `
A
~

A
. }
(2.73)
where r A is the radial step size of the random walk and L is a macroscopic scale
length of the order of the minor radius. While this elementary estimate reveals
neither dynamical insight on the detailed transport process by random E B
. .

motion nor the significant role played by fluctuation amplitude in determining
transport scalings, it illustrates the basic relation between the spatio-temporal
scales of fluctuations and transport scaling. More detailed heuristic discussions
have been given in Refs [2.53] and [2.95]. If it is assumed that
1
~
r i
k k
u
p

as
observed in gyrokinetic simulations with self-generated zonal flows and in some
experiments [2.96, 2.97], then
~
i i
turb
cT
D
L eB
p
(2.74)
This is called gyroBohm scaling because the Bohm scaling ( /
i
cT eB ) is
reduced by a factor / 1
i
L p < , the ratio of gyroradius to a macroscopic length
scale. This scaling is expected when local physics dominates the transport
mechanism. It can be modified due to a variety of mesoscale phenomena such as
avalanches [2.98] and turbulence spreading [2.992.101]. Some early turbulence
simulations for small system size without self-consistently including zonal flow
exhibited Bohm-like transport scaling in the presence of radially elongated
eddies which can span a fraction of the minor radius. This can be understood
by considering the worst strong turbulence case described in Section 2.1
of this chapter, i.e. by taking
1
r L k
u

~ ~ . in Eq. (2.73). If we allow for a

spectral anisotropy in the plane perpendicular to the equilibrium magnetic field,
104
GOLDSTON et al.
somewhat weaker turbulence with mesoscale eddies also leads to Bohm scaling
( ~ /
turb i
D cT eB ) from Eq. (2.73). In the latter case,
i
r Lp ~ . determines
the fluctuation amplitude, while
1
i
k
u
p

~ , alongside the potential fluctuation

amplitude, determines the turbulent radial E B
. .
velocity. While the distinction
between Bohm scaling and gyroBohm scaling may seem obvious, it can be
complicated by many subtle issues.
Since the radial scale of microinstabilities plays a crucial role in
determining the transport scaling, we discuss the important elements which
affect those. In small size devices, quantization inside a device determines the
radial wavelength of instabilities. For relatively large scale experiments in which
( 1)
i
a p p

< < , the magnetic shear, toroidicity, and E B

. .
flow shear are
believed to be important effects among others. The equilibrium magnetic field B
.

in toroidal geometry is given by

B B e B e
o o u u
+
.
. .
, where the subscripts o and
u denote toroidal and poloidal angles, respectively. For circular cross-section,
high aspect ratio ( / 1 R a ) plasmas,
1
0
(1 ( / ) cos ) B B B r R u

+
.
.
2.3.5.1. Role of magnetic shear: Sheared slab geometry
When one considers a perturbation with a given single set of mode
numbers (i.e. a single helicity perturbation), it is convenient to introduce a
sheared slab geometry to study the effect of magnetic shear on its stability and
mode localization. Here, we deal with an electrostatic potential perturbation:
( )
( , , ) ( )
i m n
r r e
u o
o u o o

4 4 (2.75)
with toroidal mode number n and poloidal mode number m. The magnetic safety
factor q(r) is approximately given by ( / )( / ) B B r R
o u
. We are interested in a
set of (n, m) such that ( ) /
s
q r m n is satisfied somewhere inside the plasma.
At
s
r r , the pitch of the magnetic field described by q(r) and the pitch of the
perturbation (or helicity) are identical. We call this the mode rational surface.
Since we have learned that the stability of various modes depends on the value
of k
|
, we are motivated to find a model geometry and a coordinate system
which simplifies the expression for k
|
. Noting that / k n R
o
, and / k m r
u
,
( ( ))
B
B B
k B m n
k m nq r
r B R B rB
B
o
u u

|
. .
.

(2.76)
Therefore, 0 k
|
at the mode rational surface ( ) /
s
q r m n . For non-zero
/ dq dr , the pitch of the perturbation deviates from that of the magnetic field
as one moves away from the mode rational surface radially. This trend is most
easily captured in the helical coordinate system corresponding to (n, m).
105
PHYSICS OF CONFINEMENT

y
n r
e e e
m R
u o

. . .
(2.77)

z
n r
e e e
m R
o u
+
. . .
(2.78)
Then,
( )
m n
k B q r B
rB m
u u
| `


. }
|
(2.79)

y
m n r m
B B e B
rB m R rB
u o
| `


. }
.
.
(2.80)
Expanding ( ) ( ) ( ) /
s s
q r q r r r q r + o o , ignoring
2 2
/ q r o o ,
( ) ( )

1 1
( ) ( )
y y
s
nq r r q r
B e B B B
m q r R q r
u o
| `
| `



. }
. }
.
. 1
( )
( )
s
s
r q
B r r
q r R q r
o
o

o
( )
s
B
s r r
qR
o
(2.81)
Therefore, one can model the magnetic field near the rational surface as
0

z y
s
x
B B e e
L
| `
+

. }
.
. .
(2.82)
with magnetic shear ( / )( / ) s r q dq dr , and magnetic shear length
/
s
L qR s .
Now, we can discuss the effects of magnetic shear on the radial
mode width and damping of drift waves in a sheared slab geometry.
We closely follow a discussion in Ref. [2.102]. In a sheared slab,
( )
( , ) ( )
y
i k y t
x t x e
u
o o

4 4
.
, ( / )
y s
k k L x
|
(since 0
z
k , from Eq. (2.78)),
and
2 2 2 2
/
y
k x
1
V + o o . Therefore, the electron drift wave linear dispersion
relation in the cold ion limit (Eq. (2.36)) becomes an eigenmode equation, which
is similar to the Schrdinger equation,
2
2
2 2 2
2
1 ( ) 0
y
e s
s y s
s
k
c
k x x
L x
u
p p o
u u

| `
| `
o
+ + 4

o
. }
. }

(2.83)
where
s
c

is the ion sound speed. Since the effective potential
( ) { }
2
( ) / ( / )
y s s
V x k L c x u
has an anti-well or potential hill structure for
106
GOLDSTON et al.
Re( ) Im( ) u u , we should ensure that an appropriate boundary condition
satisfies the causality requirement; i.e.
lim ( ) 0
x
x o
~
4 for Im( ) 0 u > (2.84)
which describes the fluctuation amplitude growing in time at the centre 0 x e
first and localized in x, such that it should decay in space as x ~ at a given
time.
It is straightforward to show from the large x asymptotic behaviour of the
WKB solution,
2
( ) ~ exp ( ) / (2 )
y s s s
x i k c x L u p
]
4
]
, that it is equivalent to
requiring that the radial group velocity v
gx
satisfies the following:
lim v lim 0
gx
x x
x
k
u
~ ~
o
>
o
for 0 x > (2.85)
and
lim v lim 0
gx
x x
x
k
u
~ ~
o
<
o
for 0 x < (2.86)
i.e. the outgoing wave boundary condition. Physically, it describes the convective
wave energy leakage from the centre to the x ~ region, and is called the
magnetic shear damping of drift waves [2.56].
Indeed, the eigenvalue u now contains a sheared-induced damping term:
2 2 2 2
1 (2 1)
1 1
n
s e y s y s
L
i n
L k k
u
u p p

+

+ +
(2.87)
where n = 0, 1, ... is the radial mode (quantum) number.
Also, from the WKB solution
2
exp( ( ) / (2 ) )
y s s s
i k c L x e , the radial
width of an eigenmode scales like:
~ ~
s e s s
s
y s n
L L
x
k c L
u p
p

A (2.88)
Therefore, the stronger the magnetic shear, the shorter the radial width
of the drift wave. Also, using a very simple random walk argument, one can
speculate that the drift wave in a sheared slab geometry will result in a diffusion
coefficient with gyroBohm scaling, since
s
x p A ~ . A similar scaling argument
can be applied to the acoustic version of the ITG mode in a sheared slab [2.62].
2.3.5.2. Role of toroidal geometry: Ballooning representation
The magnetic curvature is unfavourable on the low field side
of the tokamak, so that most instabilities can be more easily excited at
107
PHYSICS OF CONFINEMENT
that location, resulting in a ballooning (at the outside) structure of the
fluctuations. Obviously the usual Fourier representation of perturbations, i.e.
( )
,
,
( , , ) ( )
i m n
n m
n m
r r e
u o
o u o o

4 4

is not an efficient representation of the

ballooning-like fluctuations in toroidal geometry. In particular, since the poloidal
mode number m is not a good quantum number in the torus, any eigenmode in the
torus consists of many different poloidal harmonics which couple to each other.
Fortunately if ( )
s
nq r is high enough for the fluctuations under consideration,
there exists a more efficient way of representing perturbations in toroidal
geometry. This takes advantage of the fact that for these high n fluctuations the
radial characteristic length of the equilibrium (say
p
L ) is much longer than that
of the distance between two adjacent mode rational surfaces where each poloidal
harmonic is located. Then, we can treat each poloidal harmonic as almost
equivalent. Mathematically, we can express this as
{ }
0
( )
0
0, 1, ...

( , , )
i m j n
j
s
j
r r
r e
r
u o
o u o o
+

| `
4 4

A
. }

(2.89)
where
0
1 m ,
0 0
( ) nq r m , and
0
1/ ( )
s
r nqs r A is the distance
between two adjacent rational surfaces for a given n [2.103]. Defining the
dimensionless radial variable
0
( ) /
s
s r r r , we assume that each

( )
j
s o4
has the same radial structure. Then, we can write

( ) ( )
jj
s s j o o 4 4 and
this is called quasi-translational invariance (valid for
0
j m < ). Then,
we can check that the expression in Eq. (2.89) satisfies both periodicity
(

( , , ) ( , 2 , 2 ) r r m n o u o o u r o r 4 4 + + ) for non-zero magnetic shear s (and

finite
s
r A ). The

( , ) s j o4 are called quasi-modes, and for each quasi-mode

( ) / ( ) k s j qR
|
. A real toroidal eigenmode is a radial envelope consisting of
a superposition of many poloidal harmonics as shown in Fig. 2.12.
FIG. 2.12. Quasi-translational invariance of poloidal harmonics.
In theory [2.104, 2.105], its radial extent can be as large as a fraction of
system size (determined by the shape of the equilibrium profiles), and at that
scale, the quasi-translational invariance is broken. For instance, if r A (the radial
108
GOLDSTON et al.
extent of a toroidal eigenmode)
1/ 2
~ ( )
i
ap , from the random walk argument, one
can obtain a Bohm-like scaling,
2
~ / ~ ( ) ~
turb i
cT
D r t a
eB
p u

A A (2.90)
for ~1
i
k
u
p . In reality, the radial extent of the non-linearly saturated, finite
amplitude fluctuations is the relevant quantity. The radially elongated non-linear
structures are called streamers. Some theories suggest that the non-linear
generation mechanism, rather than the linear toroidal coupling of different
poloidal harmonics, plays a more essential role in the formation of streamers,
while others emphasize the visual resemblance of streamers and the toroidal
linear eigenmode. Implications of the existence of streamers include not only
a consequent Bohm scaling of confinement due to the ion gyroradius scale
fluctuations which has been just discussed, but also a possible relevance
of ETG turbulence as a dominant electron thermal transport mechanism in
some confinement regimes in spherical tori and tokamaks. As discussed in
Section 2.3.4.2, the radial scale of the ETG mode is very short, and scales with
the electron gyroradius. Unless the radially elongated streamers are dominant
characteristics of ETG turbulence, the electron thermal transport driven by
ETG turbulence is too small ( ~ /
ETG ITG
e e i i
m M , , , roughly, due to the mass
dependence of gyroradii) to be relevant to experimental results.
Now, the mode structure along the field line can be represented by a
coordinate which is a Fourier conjugate to the radial variable z s j ;

( ) ( )
i z
dze z
n
c n c
~
~

l
(2.91)
The extended poloidal (or ballooning) angle n is defined for
n ~ < < ~ [2.106, 2.107]. Various eigenmode equations for microinsta-
bilities in a tokamak (including the effect of magnetic curvature and B V
drift:
( ) / ( )(cos sin )
di i
cT k eBR s
u
u n n n + ) can be written in a relatively
simple form in this ballooning coordinate. One useful example of this is a clear
demonstration that the magnetic shear induced damping of drift waves discussed
in the previous section becomes ineffective due to formation of a toroidicity
induced electron drift wave [2.108].
2.3.5.3. Role of zonal fow shear
While the physics of the mean E B
. .
flow is relatively well established
[2.109], study of turbulence driven zonal flows is still an active area of current
research in both the theory and experimental communities even after publication
109
PHYSICS OF CONFINEMENT
of an extensive review [2.51]. Zonal flows are E B
. .
flows associated with
azimuthally (n = 0) and poloidally (m = 0) symmetric, but radially varying
electric potential fluctuations. Therefore, zonal flows are not directly responsible
for radial transport and, unlike various instabilities in confined plasmas, cannot
grow at the expense of the expansion-free energy associated with the radial
gradient of either temperature or density. Thus, zonal flows are linearly stable,
and can only grow at the expense of fluctuation energy through a non-linear
interaction. Zonal flows coexist with, and are excited by, ambient turbulence
developed from a variety of collective instabilities, and regulate transport by
shearing the ambient turbulence. As seen in Fig. 2.11, the spatial structure of
turbulence eddies is modified by the presence of zonal flows. The main effect
of E B
. .
shear associated with zonal flows is to reduce the radial size of eddies
(shearing) [2.1102.112] and turbulence amplitude and transport. Unlike the
mean E B
. .
flow which can be driven externally (for instance, by neutral
beam injection) and can exist in the absence of turbulence, zonal flows are
spontaneously generated by turbulence [2.1132.116].
Zonal flows are of practical importance since they reduce transport, and
shift the effective threshold condition for the onset of significant transport. In
the magnetic fusion community, the possible importance of self-generated
zonal flows in regulating turbulence and transport was first recognized from
fluid simulations of drift wave turbulence [2.117]. Following the observation
from an early non-linear gyrofluid simulation [2.118], there were ensuing
non-linear gyrokinetic simulations [2.119] with more realistic kinetic dynamics
and a proper treatment of the undamped zonal flow component in collisionless
toroidal geometry [2.79]. It is now widely recognized that understanding zonal
flow dynamics is essential in predicting confinement in future devices [2.120].
Advances in theory and simulation of zonal flows have influenced the
experimental community as described in a recent review [2.51]. In particular,
characterization of the experimentally measurable features of zonal flow
properties from non-linear gyrokinetic simulations [2.121] has motivated
experimental measurements. In toroidal plasmas, GAMs (geodesic acoustic
modes) consisting of the (n = 0, m = 1) updown asymmetric sideband pressure
perturbation linearly coupled to the (n = 0, m = 0) electric field potential
perturbation via geodesic curvature [2.122] can also be non-linearly excited, in
addition to the low frequency (near zero) zonal flow (see Fig. 2.13). The low
frequency component of zonal flows has been shown to be more important in
regulating core turbulence [2.123].
Remaining outstanding issues in zonal flow physics which need to be
addressed in the near future include the following [2.124]. More direct links
between the zonal flow amplitude and enhancement of confinement needs to
be demonstrated. While there have been some systematic, but limited, scans
110
GOLDSTON et al.
on the collisional damping of zonal flows [2.125], the q dependence of zonal
flow properties [2.126, 2.127] and their effect on confinement, in most cases
the effects of zonal flows have been exhibited in case by case simulations
[2.64, 2.119, 2.120, 2.128] or in the context of simple non-linear models [2.115,
2.116]. A rigorous theory based quantitative prediction of zonal flow induced
confinement improvement with enough physics realism is still non-existent.
Without theoretical understanding of saturation of zonal flows in collisionless
plasmas, such a predictive capability remains elusive.
FIG. 2.13. ITG core turbulence simulation using GTC code exhibits
r
k

and frequency spectra
of self-generated zonal flows, consisting of zero-frequency zonal flow and GAM sideband
[2.121].
2.3.6. Different channels of turbulence transport
One of the major goals in tokamak fusion research is to achieve high ion
temperature. Experimentalists usually plot the central ion temperature as a
function of the heating power (Fig. 2.14). On the other hand, theoreticians prefer
to think about the plasma confinement problem in terms of the flux-gradient
relation which is familiar from Ficks law in thermodynamics. For instance, the
ion heat flux can be plotted as a function of temperature gradient (see Fig. 2.14).
In confined plasmas, there exist many sources of free energy for
instabilities and transport. These include the radial gradients of ion temperature,
electron temperature, density and toroidal rotation. One can formally write a
generalized flux-gradient-relation in matrix form. We note that the radial flux of
each quantity can not only be driven by the radial gradient of that quantity itself,
but also by the gradients of other quantities. The contribution of the diagonal
term in the transport matrix is called the diffusive term. The contributions from
the off-diagonal terms are called the non-diffusive flux. Some of them are
casually called the pinch term or the convective flux.
111
PHYSICS OF CONFINEMENT
The ion thermal transport is discussed in Section 2.3.6.1; Sections 2.3.6.2
and 2.3.6.3 cover ion and electron thermal transport and particle transport
respectively. Finally, momentum transport is discussed in Section 2.3.6.4.
FIG. 2.14. Ion temperature T
i
as a function of heating power P
in
, and heat flux Q
i
as a function
of ion temperature gradient
i
T V .These diagrams illustrate that the central temperature of
a tokamak can be increased by applying more heating power, and that radial heat flux can
increase as the temperature gradient steepens.
2.3.6.1. Ion thermal transport
Ion thermal transport is currently better understood compared to other
channels of transport. In auxiliary heated plasmas, an anomalous level of ion
thermal transport has been observed which is much greater than the prediction
from neoclassical collisional transport theory. Ion temperature gradient driven
turbulence is the leading candidate for the anomalous ion thermal transport.
The linear threshold of ITG instability is also of high practical interest in
tokamak experiments. This is because strongly unstable ITG modes can drive
large amounts of radial ion heat flux, resulting in a reduction of the local ion
temperature gradient. In combination with a competing effect due to ion
heat deposition (via neutral beam injection (NBI) or ion cyclotron resonance
heating (ICRH), for instance), this results in ion temperature profiles which
do not deviate significantly from the onset condition. This onset condition can
be calculated accurately using various gyrokinetic codes. A parametrization
of various dependences can be found in Ref. [2.129], for instance. There is
theoretical evidence that the onset condition is further modified by the bursty
nature of a driven system near marginality [2.98] and the E B
. .
zonal flows
[2.120]. Theoretical predictions from different gyrokinetic codes begin to
converge for simple cases, not only in numbers, but also in underlying non-linear
physics mechanisms. An outstanding example is the zonal flow induced upshift
of the ITG onset condition [2.120].
112
GOLDSTON et al.
In the last decade, the recognition of the importance of zonal flows has
brought a paradigm shift in the tokamak turbulence theory community, from
the drift wave turbulence problem to the drift wavezonal flow system problem
[2.51]. Zonal flows exist in nature, for instance on Jupiter and the Sun, and have
been observed in many gyrokinetic simulations. Many theoreticians believe they
should exist in ITER plasmas as well. Zonal flow is a toroidally symmetric,
poloidally symmetric and radially varying E B
. .
flow. The interaction of
zonal flows with drift wave turbulence is schematically illustrated in Fig. 33
of Ref. [2.51]. Zonal flows can be damped by the collisional friction between
trapped and passing ions [2.130]. Also, zonal flows regulate turbulence via
E B
. .
shearing [2.1102.112].
Next, some examples are presented in which the zonal flows play crucial
roles in determining the transport scalings with respect to key dimensionless
parameters. Transport scaling with respect to the machine size, or /
i
a p p

, is
very important, since we want to know by how much the future larger machines
will perform better than the present ones. As our understanding of ITG turbulence
got deeper, our predictions were modified. Theoretical predictions of toroidal
ITG instabilities in the early 1980s [2.63] and of their global radial structures in
the early 1990s [2.104, 2.105] seemed to indicate the possibility of Bohm scaling.
Since the mid-1990s, extensive studies of zonal flows indicated that the radially
elongated global ITG eigenmodes get broken up by zonal flows into smaller
eddies [2.118, 2.119, 2.128]. Naturally, most people expected gyroBohm scaling
from the eddy size scaling. More recently, it has been indicated that, from global
gyrokinetic simulations, for a moderate system size, non-negligible deviations
from gyroBohm scaling can exist [2.131]. This can occur due to turbulence
spreading into the linearly stable zone [2.100]. While the studies on turbulence
spreading go back at least to the mid-1990s [2.99], it was the importance of its
relation to the p

scaling which revived wide interest in the theory community
[2.100, 2.101, 2.1322.137].
Plasma current scaling of confinement, as described in Section 2.4, is one
of the remaining puzzles of ion thermal transport. It is possible to obtain the q
dependence of the transport from zonal flow characteristics which depend on the
q value. Due to their Landau damping, geodesic acoustic modes (GAMs) can
exist only in a high-q region. Therefore, in a low-q region, only stationary zonal
flows will persist. Since GAMs oscillate with a relatively high frequency, they
are not very effective in reducing turbulence [2.123]. So we expect a lower level
of turbulence in the low-q region where static zonal flow dominates. As shown
from a recent non-linear gyrofluid simulation [2.126], the ion heat flux from ITG
turbulence is lower for the case with lower q, since the stationary zonal flows
dominate. This is confirmed by an independent gyrokinetic simulation [2.127].
However, the relevance of this mechanism to experiment still remains to be
demonstrated.
113
PHYSICS OF CONFINEMENT
The collisionality dependence of ion heat transport observed in tokamak
experiments is a puzzle too, since the transport level was much higher than the
predictions from collisional transport theory, while ITG turbulence in the core of
a tokamak is essentially collisionless. Recognizing that the zonal flows regulate
the turbulence level, near ITG marginality, collisional damping of zonal flows
controls the transport, and the transport level can increase with an increasing ion
collisionality as shown from gyrokinetic simulations [2.125] and confirmed by
an independent gyrofluid simulation [2.138]. From a theoretical point of view,
it is important to note that this trend can be reversed if one includes the trapped
electron response, which decreases with electron collisionality [2.139]. While
some experiments report transport increases with collisionality, further parameter
scans are required to identify the underlying physics mechanisms.
2.3.6.2. Electron thermal transport
It has been well known for many years that all operational modes of
tokamaks and spherical tori have exhibited anomalous electron thermal transport.
There is evidence that electron temperature profiles are stiff in tokamaks.
Perturbative transport experiments, including heat pulse propagation, and the need
for an inward heat pinch term in transport analysis of auxiliary heated plasmas,
all indicate this possibility. Possible theoretical candidates for electron thermal
transport include the trapped electron mode (TEM) and the ITG mode with
trapped electrons, the electron temperature gradient (ETG) mode, and magnetic
flutter driven transport.
Starting with the TEM, it can produce an electron thermal diffusivity of
experimental relevance when strongly turbulent. It is interesting to note that
DTEM based non-linear theory [2.140] can produce a scaling of electron thermal
diffusivity
e
, close to the Neo-Alcator scaling which has been widely observed in
many ohmically heated tokamak plasmas. One supporting result for TEM comes
from the ASDEx-Upgrade ECH experiment. The transport shows a threshold
behaviour in agreement with the TEM theory prediction. Furthermore, the experi-
mentally measured heat flux and the calculated TEM linear growth rate show a
similar trend as functions of the electron temperature gradient [2.141]. On the
other hand, evidence for TEM-like fluctuations from measurements is relatively
rare, as discussed in Section 2.6.2.
Another candidate is ETG turbulence. In some spherical torus plasmas such
as NSTx, electron transport is highly anomalous even when low-k
.
fluctuations
are believed to be linearly stable. In NSTx neutral beam heated plasmas [2.142],
most of the heating power from the energetic particles goes to electrons, and this is
similar to a situation expected for ITER. The linear stability analysis shows that the
low- k
.
fluctuations are expected to be stable due to the high E B
. .
shearing rate,
leaving only high- k
.
ETG modes unstable. Note that the ion thermal diffusivity is
114
GOLDSTON et al.
down to a neoclassical level, qualitatively consistent with the stabilization of low-
k
.
ITG/TEM modes. Due to its low magnetic field and consequent large electron
gyroradius, NSTx is an ideal place to look for ETG-like fluctuations with good
spatial resolution. Recent progress in experimental measurements using high-k
.

tangential scattering indeed supports the existence of density fluctuations in the
ETG spectral range [2.143]. More details can be found in Section 2.6.2.
To produce electron heat transport at a level of experimental relevance in
tokamaks, the radial correlation length of ETG turbulence has to be significantly
enhanced over the electron gyroradius. Radially elongated streamers have been
observed in many non-linear gyrokinetic simulations [2.87, 2.88]. However, from
a different simulation, the correlation between the electron heat flux and the radial
size of streamers has been reported to be weak [2.144]. As of now, there is no
consensus in the theory community regarding whether the ETG-driven electron
heat transport is high enough to explain experimental observations.
2.3.6.3. Particle transport
Particle transport is also anomalous in most cases, while its associated
diffusion coefficient is typically smaller than that associated with ion thermal
or electron thermal transport. In many experiments, electron density profiles are
observed to be centrally peaked, even when the particle source exists only at the
edge. This has motivated theoretical research on an inward convective flux (pinch)
of particles. The particle flux can be written as
v
pinch
n
D n
r
o
I +
o
(2.92)
i.e. as a sum of diffusion (the first term) and a convective flux proportional to a
pinch velocity v
pinch
. As described in Section 2.2, an inward pinch of particles
can occur through a collisional process involving magnetically trapped particles.
This Ware pinch can account for density peaking in some tokamak cores [2.145],
but there seems to be no shortage of counter examples to this, including cases
without an inductive electric field [2.146, 2.147] which is required for the Ware
pinch. From the form of Eq. (2.92), it is clear that an electron particle pinch can be
driven by either
e
T V or B V
.
, i.e. magnetic field inhomogeneity. The first class of
pinch is called the thermoelectric pinch, for the obvious reason, while the second
class is called the turbulent equipartition (TEP) pinch [2.148], for a reason that we
explain shortly.
For the ITG dominated case (but with non-adiabatic electron response,
contributing to turbulent particle flux), it has been known for a few decades that
inward particle flux can occur when the collisionality of the plasma is either very
high or very low. The first possibility is called the ion mixing mode [2.58], which
occurs for high collisionality satisfying 1
e
v

> , and v ,
e
e T e
k v u u

> >
|
. The ion
115
PHYSICS OF CONFINEMENT
mixing mode can be of very high pedagogical value since the instability is driven
by
i
T V , while n V and
e
T V drive diffusive and convective particle fluxes associated
with it, respectively. Electron drift wave stability in this regime has been discussed
in Section 2.3.2.1. The second possibility, which is more relevant to present day and
future tokamak core plasmas, occurs when 1
e
v

< , such that the trapped electron

response becomes collisionless enough that
e de
u u

> /
eff e
v v r . Since its
first discovery [2.149], there has been gradual theoretical progress in refining
the calculations, including more realistic effects [2.1502.152]. Comparisons to
experiments are also under way, as discussed in Ref. [2.153], for instance.
Another class of particle pinch can occur due to the inhomogeneity of
the equilibrium magnetic field. The fundamental concept behind the turbulent
equipartition (TEP) pinch can be elucidated by considering a 2-D system with
straight but inhomogeneous magnetic field ( , ) B B x y z
.
, in which density is
mixed by E B
. .
flow [2.154]. In 2-D, the density continuity equation is
( ) 0
E
n nu
t
o
+ V
o
.
.
(2.93)
with ( / )
E
u c B z V4
.
.
. Now,
E
u
.
is no longer incompressible due to the
inhomogeneity in B
.
. However, we can write Eq. (2.93) as
0
E
n n
u
t B B
o
| ` | `
+ V

. } . }
o
.
.
(2.94)
using the fact that 0 z V V4
. .
. Then a magnetically weighted density / n B
is advected by the E B
. .
flow. Thus, / n B is locally conserved and eventually
mixed or homogenized by the turbulent E B
. .
flow. Then, the spatial profile
of the mean / n B would relax according to
/ /
/ /
n B n B
D n B x I o o so that
/
/ / 0
n B
n t x o o + oI o .
According to the conventional classification given in Eq. (2.92),
/ /
2
1
v
n B n B n pinch
n
D n B D n n
B x x x B
| `
o o o
I e +

o o o . }
(2.95)
where
/
( ) /
n n B
D D B ,
2
/
v ( / ) /
pinch n B
D B B x o o , and / / B n n B B .
Relaxation of gradients of the locally advected, magnetically weighted quantity
/ n B led to a pinch
2
/
v ( / ) /
pinch n B
D B B x o o of an observed physical
quantity n . This is the simplest example of the TEP pinch. It is called the TEP
pinch since time-asymptotically it will lead to a complete relaxation, mixing, or
equipartition of / n B . Note that unlike
e
T V , B V
.
is not a thermodynamic force,
and TEP is of a different sort from a thermoelectric pinch.
In toroidal geometry, most of the TEP particle pinch comes from
magnetically trapped electrons [2.155, 2.156]. This leads to a simple formula from
116
GOLDSTON et al.
0
2 1 2
v /
4 3
TEP
pinch
s
D
R
| `
+

. }
(2.96)
with
0
R being the major radius at the axis. Some degree of agreement with
experiments exist for L-mode plasmas, in particular in its relation to q profiles
(through the s dependence) [2.146, 2.157].
2.3.6.4. Momentum transport
Now, it is well appreciated that plasma rotation plays a central role in
magnetic confinement. E B
. .
shear can reduce turbulence and transport, and in
turn rotation can be generated by microturbulence via Reynolds stress [2.158].
It can stabilize MHD instabilities, such as resistive wall modes, but can be
damped via non-axisymmetric magnetic fields produced by MHD instabilities.
In addition, the radial electric field interacts with energetic particles. Therefore,
understanding momentum transport is important. In particular, for successful
operation of ITER, we want to know whether ITER will have enough wave and
neutral beam power to generate sufficient rotation for turbulence and transport
reduction as well as stabilization of resistive wall modes.
Historically, momentum diffusivity has been predicted [2.159], and
observed [2.26] to be very close to ion thermal diffusivity. But, more recently,
many tokamak experiments indicated that the ratio between these two quantities
can deviate from unity significantly. This may suggest the existence of an
off-diagonal contribution to the momentum flux as demonstrated in detailed
experimental studies [2.160]. Indeed, a momentum pinch has been indicated
by modulation experiments [2.161, 2.162]. It is also important to note that
fluctuation measurements indicate that plasma rotation is related to the Reynolds
stress [2.163].
Intrinsic toroidal rotation has been observed in the absence of an external
torque input from neutral beams in nearly all tokamaks [2.164, 2.165]. The
behaviour of L-mode plasmas remains complex, and is closely tied to scrape-off
layer flows [2.166]. On the other hand, Hmode plasmas demonstrate simple and
clear empirical trends, summarized as the Rice scaling [2.167]. Toroidal rotation
in H-mode plasmas is in the co-current direction, increases with incremental
stored energy and decreases with plasma current. These observations appear to
be consistent with rotation originating at the edge with the transition and building
inward from there. There are also experimental observations that cannot be
explained by diffusion and a pinch only [2.168].
The radial flux of momentum can be semi-theoretically written as a sum
of the diffusive flux, the convective flux proportional to the momentum pinch
velocity
,
v
mom pinch
and the residual stress,
117
PHYSICS OF CONFINEMENT
,
v
mom pinch RS
u u
r
o o o o
,
o
I + + H
o
(2.97)
This can be viewed as a quasi-local expression resulting from a Taylor
expansion of a general expression for
o
I in / u r
o
o o and u
o
. There is no a
priori reason to rule out the residual stress
RS
H which is independent of toroidal
flow and toroidal flow shear. It can be proportional to the gradient of pressure or
density. In the presence of residual stress, the resulting flow evolution equation
governed by the divergence of
o
I should satisfy the basic coordinate transfor-
mation properties, such as Galilean invariance in a simple geometry. Now, we
discuss diffusive flux, the momentum pinch and finally the residual stress.
Fluid theory [2.159] predicted that the intrinsic Prandtl number (the ratio
of momentum diffusion to ion heat diffusion), defined with only the diffusive
momentum flux, should be near unity. From the gyrokinetic theory point of
view, momentum and ion thermal transport involve different velocity moments
respectively. Therefore, if a certain particular set of particles (for instance, those
resonant with waves) contribute dominantly to transport, the intrinsic Prandtl
number should deviate from unity [2.169].
A pinch can come from various physical mechanisms [2.170] including
waveparticle resonance. In this section, we discuss the curvature driven
toroidal momentum pinch, which has two parts. The first part is the turbulent
equipartition (TEP) pinch which is mode independent, originating from the fact
that the E B
. .
flow is no longer incompressible in a torus [2.171]. This is the
simple and robust part. The second part is the thermoelectric pinch which is mode
dependent, and sensitive to turbulence characteristics. The physics of the TEP
pinch is well illustrated by the particle pinch in a slab with an inhomogeneous
B field in the previous section. Mixing of the magnetically weighted, locally
conserved quantity n / B leads to an effective pinch in the observed physical
quantity n. Then, how about for momentum? Starting from angular momentum
density conservation, and using the E B
. .
flow compressibility, one can show
that the magnetically weighted angular momentum
2
/ nu R B
o
is convected by
the turbulent E B
. .
flow [2.171]. As a consequence of mixing and diffusion of
the magnetically weighted quantity
2
/ nu R B
o
, we can obtain an inward pinch
in the observed quantity nu
o
by performing a quasi-linear calculation [2.172,
2.173]. By separating the radial gradients of this expression into two parts, we
can identify the TEP pinch term. The resulting pinch velocity, normalized to
the momentum diffusivity, is -4/R, if we define the momentum diffusivity with
respect to the angular rotation frequency. These results can also be derived
from the conservative gyrokinetic equation [2.171]. For this toroidal pinch,
the symmetry breaking comes from ballooning fluctuations which have more
intensity on the unfavourable curvature, low field side.
118
GOLDSTON et al.
The thermoelectric pinch is a thermodynamic part of the Reynolds stress
related to the ion temperature fluctuations. It is dependent on mode character-
istics, in particular, the phase angle between
i
T o and o4. It is also sensitive to
the proximity to marginality, the gyrofluid approximation, and the dispersion
relation used in the derivation [2.174]. An independent formulation in the rotating
frame attributes the aforementioned curvature driven momentum pinch to Coriolis
pinch force related terms [2.175].
In contrast to the particle transport problem where particle number is
conserved, and the pinch is the only off-diagonal flux, momentum can be
transferred between particle and waves. This can lead to the residual stress, the last
term in Eq. (2.97) [2.174]. Since the residual stress can make a net contribution
after a radial integration from the plasma boundary, it can accelerate the resting
plasma. Therefore, the residual stress acting with an appropriate boundary
condition can generate an intrinsic rotation [2.176]. To get a finite residual
stress, we need a broken symmetry in the turbulence. The key point about the
residual stress is that E B
. .
shear converts poloidal rotation shear into toroidal
rotation shear via asymmetry in the wave to particle momentum deposition. This
asymmetry can be provided by having a finite E B
. .
shear and generic drift-
acoustic coupling, which will shift the eigenmode in the radial direction, resulting
in a preferred sign of k
|
, as noted by many authors [2.177, 2.178]. The underlying
physics mechanism behind the residual stress is a directional imbalance in the
acoustic wave populations and in the profile of momentum deposition by ion
Landau damping [2.169].
2.3.7. Physics of transport barriers
Over recent years, the performance in tokamak plasmas has been improved
considerably by exploiting enhanced confinement regimes. While there are
differences between the operational techniques as well as the initial locations for
a wide variety of transport barriers, a considerable amount of similarity in the
transition dynamics appears to exist.
In this section, we discuss the common physics elements which constitute
the conceptual building blocks of many transport barrier theories and simulations
[2.179]. A more quantitative, theoretically oriented review also exists [2.180].
Significant insight into turbulence suppression by the sheared mean E B
. .

flow has been gained via extensive experimental investigation in the 1990s of
the physics of transport barrier formation [2.181, 2.182]. On the theory side,
development of the E B
. .
shear driven non-linear decorrelation theory in a simple
geometry [2.110], and its extension to general toroidal geometry [2.112, 2.183],
as well as non-linear gyrofluid simulations [2.128], have led to an approximate
criterion for ion thermal transport barrier formation: namely the E B
. .
shearing
rate [2.112] versus the maximum linear growth rate of microinstabilities in the
119
PHYSICS OF CONFINEMENT
absence of flow. While this is an oversimplified rule of thumb, rather than a
final statement of the E B
. .
shear suppression, it has been widely used in the
experimental community as a rough measure for the assessment of the role of
E B
. .
shear in transport barrier formation. Now, there is a significant effort
towards self-consistent, non-linear dynamic simulations of transport barrier
formation.
The strength of the E B
. .
shear effect is characterized by the shearing rate
E
u in general toroidal geometry [2.112]:
( )
2
0 0 r r
E
RB
r E E
RB l B RB
u
u u
v
u
o v v
1
| ` | ` A A
o o


A o A o
. } . }
(2.98)
Here, v is the poloidal flux representing the radial coordinate via
d RB dr
u
v ,
0 0
/ r RB
u
v A A is the radial correlation length and R o A is the
toroidal correlation length, and / l RB B
u
o
1
A A is the correlation length of the
ambient turbulence in the direction perpendicular to the field line, but in the flux
surface. Other notations are standard.
( )( )
/ /
r
E RB
u
v o o is a flux function in
most cases.
The radial force balance relation for a plasma ion species i reads
/
r i i i i i
E p Z en u B u B
o u u o
V + (2.99)
where
r
E is the mean radial electric field, and
i
p ,
i
u
o

and
i
u
u

are ion pressure, ion
toroidal velocity and ion poloidal velocity respectively. Therefore, /
r
E RB
u
can
come from the poloidal flow, the toroidal flow and the pressure gradient. This is
one of the reasons that this paradigm has been widely applicable to various cases
including:
(i) ITB formation in TFTR [2.184] and JET [2.185] plasmas, and ETB
formation [2.182, 2.186, 2.187] in high (H)-mode plasmas [2.188],
where the poloidal fow plays a crucial role in the transition, while the
pressure gradient is signifcant after the transition [2.186, 2.189];
(ii) plasmas with strong toroidal rotation such as negative central shear
(NCS) [2.190] and very high (VH)-mode [2.191] plasmas in DIII-D,
optimized shear (OS) plasmas in JET [2.192], and in many other
tokamaks [2.1932.195] and in spherical torii [2.142]; here
r
E is
typically positive for co-rotating plasmas and negative for counter-
rotating plasmas.
Now, we discuss the role of q profiles and magnetic shear in forming
transport barriers. From a theoretical point of view, the q profile can affect
turbulence in many ways:
120
GOLDSTON et al.
(i) Each poloidal harmonic of a toroidal eigenmode is centred on a rational
q surface, as illustrated in Fig. 2.12.
(ii) Its radial extent is often limited by linear Landau damping and resistive
dissipation, which increase away from the rational surface at a rate
which depends on the magnetic shear (ln ) / (ln ) s d q d r .
(iii) The radial overlap of these poloidal harmonics strongly depends on
the density of the rational surfaces, which in turn depends on the q
profle shape. Radial overlap gets easier as higher mode numbers are
considered. Low mode number fuctuations can exhibit discontinuities
for low s values, especially near the radial location where q becomes a
minimum in reversed shear (RS) plasmas, due to a rarefaction of rational
q surfaces [2.105].
(iv) Reversed shear has a stabilizing infuence on curvature driven
instabilities.
The discovery of the H-mode in a diverted tokamak [2.188] motivated
an early theory of the H-mode [2.196], which relied on mechanism (ii), which
becomes pronounced when s becomes very high at the edge, either just inside
the last closed flux surface of the diverted tokamak plasma or due to current
rampdown [2.197]. There have been numerous linear stability analyses (for
instance Refs [2.90, 2.1982.204]) on various microinstabilities addressing effect
(iv). Non-linear simulations in flux tube geometry [2.87, 2.128, 2.201, 2.205]
take into account effects (i), (ii) and (iv). However, the ballooning represen-
tation is closely related to the quasi-translational lattice symmetry as discussed
in Section 2.3.5.1, which is not applicable to low mode number fluctuations or
near
min
q of reversed shear plasmas. Consequently, these simulations find that
reversed magnetic shear (via mechanism (iv)) is more favourable for transport
reduction compared to low magnetic shear (via mechanism (iii)) which should be
more pronounced for low mode number fluctuations.
Global gyrokinetic simulations keeping relatively low toroidal mode
number fluctuations observe transport reduction near
min
q due to a lack of radial
overlap (discontinuity) of poloidal harmonics [2.206]. However, this effect gets
weaker and transport tunnels through the
min
q surface as more modes are added
in an independent global gyrokinetic simulation [2.207]. A study based on global
fluid simulations of ITG turbulence [2.208] addressed all the aforementioned
effects associated with q profiles, although sub-ion-gyroradius-scale high mode
number fluctuations are not kept. The overall conclusion is that an ITB is most
easily formed near
min
q of RS plasmas, especially when
min
q is near a low order
rational value due to a rarefaction of rational q surfaces. However, ITB formation
near
min
q when its value is near a low order rational number does not necessarily
imply that mechanism (iii) is responsible for it. More recent gyrokinetic
simulations exhibited turbulence driven zonal flows being preferentially generated
121
PHYSICS OF CONFINEMENT
near gaps in the density of rational surfaces localized in the vicinity of
min
q
[2.209]. It has also been pointed out from analytic theory [2.210] that low mode
number electrostatic convective cells may play a key role in the mean profile
flattening localized to the low-q surface as a transition precursor as observed in
experiment [2.211]. These effects could play a role in the type II ERS transitions
in TFTR [2.212] which occur near
min
q when its value reaches near 2, and in
internal transport barriers in RS plasmas in DIII-D [2.211]. Other theoretical
possibilities supporting the important role of low order rational q surfaces include
the generation of localized velocity via MHD instability as proposed by the JET
team [2.213] and via a loss of fast ions as proposed by the ASDEx-Upgrade team
[2.214].
On the other hand, recent findings from NSTx experiments [2.215] seem to
support the hypothesis that while ion thermal transport barriers are formed in the
region where the E B
. .
shearing rate is a maximum, the electron thermal transport
barrier location correlates well with the (negative) magnetic shear minimum
location, i.e. where the negative magnetic shear is strongest. The appearance of an
electron barrier for negative magnetic shear has previously been reported in many
tokamaks including JET [2.216]. Finally, an analysis of the Tore-Supra plasmas
has yielded a puzzling result that the time of internal barrier formation shows little
correlation with either the value of
min
q or the value of (the negative) magnetic
shear [2.217]. It correlates with the central q value, (0) q , crossing a low order
rational value in time. In summary, there exist a variety of both experimental
and theoretical findings regarding the role of the q profile in transport barrier
formation. Further study on this subject is required.
It is useful to discuss the basic physics mechanism associated with transport
barrier formation and evolution in the context of the transport flux landscape
[2.218] which describes transport flux in the radial position-density gradient (or
temperature gradient, or pressure gradient) space. Usually, the density gradient
is considered for simplicity. At each radius within the barrier zone, the flux I is
typically an inverted S-shaped non-monotonic single valued function [2.219] of a
local gradient of density. We focus our discussion on this case, which is called a
(soft) first order transition. Discussions of the hard transition when the flux
is a multivalued function of a local gradient or of a second order transition
when the flux is a monotonic function with a change of slope, can be found in
Refs [2.180, 2.220]. Most theoretical studies based on the E B
. .
shear and the
reversed magnetic shear mechanisms use a transport coefficient D in the form:
E
turb neo
lin
D D f D
u
y
| `
+

. }
(2.100)
where
2
1
( / ) 1/ (1 ( / ) )
c
E lin E lin
f c u y u y + , with
lin
y

the characteristic linear
growth rate of instability and
1
c ,
2
c constants of the order of unity, for instance
122
GOLDSTON et al.
Refs [2.219, 2.221], characterize the E B
. .
shear induced reduction of the
anomalous diffusivity
A
D . While various authors have tried different functional
forms of the monotonic f(x) with f(0) = 1 and f(x) = 0 for 1 x , we confine our
discussions to the generic features of this class of models [2.222]. We also note
that the favourable role of reversed magnetic shear can be contained in the radial
variations of
A
D and
lin
y . Since
0 E
u is related to the density gradient through
radial force balance, Eq. (2.100) will produce a non-monotonic single valued
S-shaped curve for the particle flux ( / ) D n r I o o as a function of local density
gradient ( / ) n r o o (Fig. 2.15), and a different S-shaped curve will apply at each
minor radius. Correspondingly, for a given density gradient, the flux landscape
depicts the familiar I curve which increases with minor radius. Until recently, it
had been thought [2.219] that for each S-shaped curve, the local maximum and
the local minimum (at a higher density gradient) define the threshold particle
flux for the forward (RS to ERS, or L to H) transition and the backward (ERS to
RS, or H to L) transition respectively. Therefore, the situation exhibits hysteresis.
However, there is a new theoretical development which indicates that the
aforementioned simple picture overpredicts the hysteresis [2.223]. Indeed, some
tokamaks report a lack of strong hysteresis in LH and HL transitions, unlike
DIII-D. It is encouraging to note that data from C-Mod Hmode plasmas [2.224]
can be characterized by the aforementioned S-curve based theoretical picture.
Finally, identification of practical ways to generate sheared flows [2.225, 2.226]
in reactor relevant future devices will be, of course, highly desirable.
FIG. 2.15. Particle flux as a function of density gradient.A corresponds to the L-mode root
with a high slope given by ( )
turb neo
D D , while B corresponds to the H-mode root with
a low slope given by
neo
D , with
turb
D significantly reduced by the E B
. .
shear associated with
a strong density gradient.
123
PHYSICS OF CONFINEMENT
Acknowledgments
T.S. Hahm gratefully acknowledges E.S. Yoon for help in preparing the
manuscript. This work was supported by the US Department of Energy Contract
No. DE-AC02-09CH11466 (PPPL).
2.4. GLOBAL ENERGY CONFINEMENT SCALING STUDIES
2.4.1. Introduction
Global confinement scaling studies serve to provide a high level perspective
on the behaviour of the plasma energy confinement as discharge parameters are
independently varied. Scalings have been developed that describe quantitatively
how the energy confinement time
E
t and L-mode to H-mode threshold power
varies as other parameters vary across a wide range of plasma conditions and
energy states (L-mode, H-mode, etc.), and thus they can provide insights into
and tests on the validity of theories believed to drive local energy transport.
The parameters to which most attention is paid are the so-called engineering
parameters such as plasma current, toroidal magnetic field, density, heating
power, that can be controlled externally. Other, more physics based parameters,
such as the plasma beta , collisionality
*
v and normalized gyroradius
*
p ,
depend on the plasma response to external control, and are not as easily imposed.
One of the usual misnomers is to refer to global scalings as laws. They
cannot, in any sense, be viewed as on a par with the fundamental laws of physics.
They are derived descriptions of the behaviour of data, and as such, and as will
be seen, each scaling expression has considerable uncertainty associated with
its quantitative formulation. Their usefulness is to serve as guides for theory;
ultimately any validated transport theory, or combination of theories that profess
to explain transport in tokamaks, must end up giving rise to the global trends
observed in experiment.
There are simple approaches to deriving the global scaling expressions
(GSEs), such as ordinary least squares regression, to more sophisticated
approaches such as error-in-variables or maximum-likelihood and Bayesian. As
will be seen, the different approaches lead to expressions that have quantitative
differences. Because the GSEs are descriptive and are not based on fundamental
physics, there is danger in using these expressions for extrapolations, especially
if the extrapolations extend well beyond the range of independent parameters
being studied, without understanding the systematic and random uncertainties
inherent in the expressions and their predictions.
At the heart of any global or local study is the determination of the
confinement and transport properties of the plasma, and to do this, good
measurements of the plasma characteristics are necessary. At the most basic
124
GOLDSTON et al.
level, equilibrium reconstructions based on magnetics measurements can, in
certain circumstances [2.227], give reasonable estimates of the total stored
energy of the thermal plasma and any fast ion energy component created by
auxiliary heating methods such as neutral beams, radiofrequency waves, etc.
More detailed determinations and separation of these components can be
obtained through measurements of the thermal ion and electron energy content
directly. This can be done by measuring the electron and ion temperature and
density. The electron temperature and density can be measured directly. The
main plasma species ion temperature and density is presently covered through
measurements of these profiles for the main plasma impurity (e.g. carbon) and
then corrections due to collisions between the impurity and main ion species.
The capability to measure the main ion species temperature and density directly
is presently being developed. Since both low-Z and high-Z impurities typically
are observed in plasmas, the thermal ion density is less than that of the electrons.
Determination of the fast ion or electron component of the plasma is coupled to
the use of modelling codes, and these will be described in Section 2.5.
Plasma profiles are generally peaked in the centre of the plasma, and
decrease with increasing radial distance towards the plasma edge. In low-
confinement, or L-mode, plasmas, the density and temperatures are usually near
zero at the plasma separatrix. Highconfinement, or Hmode, plasmas are typified
by temperature pedestals of up to several hundred eV and density pedestals of
up to several 10
19
m
3
within (inside) a few cm of the plasma separatrix. These
pedestals lead to strong pressure gradients in the plasma edge, which can lead to
intermittent destabilization of edge localized modes, instabilities driven near the
plasma edge, resulting in a temporary decrease in the edge pressure gradient and
thus confinement. Other, more core localized instabilities such as sawteeth and
tearing modes, can cause degradation of confinement in that region.
In this section, we will describe the background leading to the development
of GSEs, the approaches towards their development and the results of these
efforts. Reviews and overviews of the results can be found in Refs [2.2282.231].
2.4.2. Energy confnement scalings: Dimensional parameters
2.4.2.1. Ohmic and L-mode plasma confnement trends
Ohmic plasmas from a number of experiments [2.232] provided the first
basis for developing energy confinement time scalings. At lower densities, the
confinement time in ohmic plasmas was found to depend on plasma density, size
and q, leading to the Alcator (or INTOR) scaling
2 1/ 2
E e
n a q t where
e
n is
the line averaged plasma density, a the plasma minor radius and q the inverse
cylindrical rotational transform. A modification to this scaling was made by
Pfeiffer and Waltz [2.233], with
2
E e
n aR t , and, with an additional q
1/2

125
PHYSICS OF CONFINEMENT
factor, was subsequently called neo-Alcator scaling. The confinement time in
this linear (with density) ohmic confinement (LOC) regime depended neither
on the plasma current nor the toroidal field. At higher densities, however, the
confinement time showed no further increase with density (saturated ohmic
confinement, or SOC regime), but did show a dependence on plasma current.
It was quickly realized, however, that in order for significant fusion power
to be produced, temperatures in excess of those produced by ohmic heating alone
were necessary. Consequently, the mid to late 1970s and early 1980s saw an
increasing number of experiments incorporating neutral beam injection to heat
the plasmas. The earliest neutral injection results came from small tokamaks,
such as CLEO [2.234], ATC [2.235], ORMAK [2.236] and TFR [2.237], but
while the neutral beam powers were effective in increasing the ion temperature to
values ~1.8 keV, well in excess of the ohmic value, this auxiliary power level was
still small compared to the ohmic heating power, and thus showed only limited
heating of the electrons. A more than doubling of the central electron temperature
was measured, however, when the auxiliary heating power level was increased to
twice the ohmic level [2.235]. With 2.4 MW of neutral beam power injected into
the PLT tokamak [2.238], ion and electron temperatures increased up to 7 and
4 keV respectively.
The neutral beam heated discharges from a number of tokamaks exhibited
energy confinement trends that were quite different from those observed in
ohmic plasmas in the IOC regime. Approximately 16 tokamaks, with minor radii,
a, ranging from 0.13 to 0.44 m, major radii R = 0.70 to 1.47 m, plasma current
p
I = 0.08 to 1 MA, toroidal magnetic field B
o
= 1.2 to 3.5 T, line averaged
density
e
n = 2.5 to 13 10
19
m
3
and neutral beam injection powers
inj
P = 0.05
to 8 MW comprised the basis for establishing the fundamental trends that, to
this day, describe confinement in low confinement, or L-mode discharges. These
experiments were carried out primarily in tokamaks whose plasmas were limited
by either poloidal rail or toroidal bumper limiters, or through the use of magnetic
divertors that allowed unimpeded transport of neutral particles from the divertor
region to the plasma main chamber. A summary of these experiments and their
results is detailed in Ref. [2.228].
One of the key observations, which was observed in all devices, was that
for injected powers greater than approximately twice the ohmic power, the energy
confinement time decreased at the onset of neutral beam injection (NBI) by a
factor of 1.4 to 4 from that observed during the ohmic phase (see Ref. [2.228] and
references therein). A typical L-mode discharge evolution from PDx is shown in
Fig. 2.16; note the drop in confinement time when the neutral beams are injected
at 0.4 s. Increasing neutral beam power resulted in decreasing confinement
time, although the degradation of
E
t was less than linear, with
E
t going as
1/3 2/3 to
heat
P

, where
heat
P is the total heating power. Furthermore, the energy
confinement time, unlike in ohmic discharges, did not increase with increasing q,
126
GOLDSTON et al.
but rather showed a nearly linear increase with plasma current,
p
I , similar to that
observed in the SOC regime. The global confinement time in L-mode discharges
showed little if any dependence on B
o
or n
e
, but experiments on D-III [2.239]
did indicate an increase in
E
t with increasing plasma elongation, x . Increased
plasma shaping allowed more current to be carried by the plasma, but when the
current was fixed, it was found that shaping still played an important role, with
1/ 2
~
E
t x . NBI experiments in ASDEx [2.240], DIII-D [2.241] and JFT-2M
[2.242] indicated that plasma isotope, or mass, was an important consideration,
with
1/ 2
~ ( / )
E i p
m m t , where
i
m is the ionic mass and
p
m is the proton mass.
The ASDEx results were obtained from experiments comparing performance
in H
0
NBI injection into H
+
and H
0
injection into D
+
plasmas (see Ref. [2.228]
and references therein). The direction of injection, either co- or counter- to the
plasma current, was also found to influence confinement. Beam orbit losses
with counter-injection on PDx and ISx-B led to a loss of approximately 40%
of the input power, as compared to only ~10% with co-injection. Despite this,
little difference in stored energy was observed between co- and counter- on the
respective devices, leading to an increase in energy confinement time of between
30 and 50%.
FIG. 2.16. Output from a 0-D, between discharge analysis programme for a typical PDX
L-mode discharge. The energy confinement time (lower right) is calculated from measured
magnetics information, an assumed current profile shape, and the total (NBI + ohmic) heating
power. Reprinted from Ref. [2.228], Copyright (2010), American Institute of Physics.
What was lacking from the results of these individual experiments was
just how the confinement time scaled from device to device, i.e. with plasma
and machine size. Because the majority of these L-mode experiments exhibited
127
PHYSICS OF CONFINEMENT
similar parametric trends, it was possible to pull together these data and, from
the differences in average confinement time from device to device, determine
how the confinement time varied with plasma minor and major radius if all other
parameters are fixed. This approach was adopted by Goldston [2.243], who
started the analysis by noting that the density and current scaling of high density
ohmic plasmas was similar to that in NBI L-mode plasmas, thus enabling a
coupling between the two regimes (ohmic and auxiliary heated L-mode plasmas).
He proposed an energy confinement scaling of the form:
1/ 2
2 2
, ,
1 1
E
E OH E AUX
t
t t
]
+ ]
]
]
(2.101)
where the ohmic scaling,
, E OH
t , was obtained from surveys of low density ohmic
scaling experiments, and the non-size portion for the auxiliary heating scaling
portion
, E AUX
t was obtained from averaging the parametric dependences of
confinement on
p
I , x and
,
( )
heat heat NBI OH
P P P + , where
, heat NBI
P is the
absorbed beam heating power and
OH
P is the ohmic heating power, from ISx-B,
ASDEx, D-III and PDx L-mode discharges. The size scaling portion of
, E AUX
t
was obtained from a regression of the device-to-device confinement variation
about some reference
p
I
, x and
heat
P . The expressions that resulted from this
technique are:
22 1.00 3 1.04 2.04 0.50
,
[s] 7.1 10 [cm ] [cm] [cm]
e
E OH
n a R q t

(2.102)
and
8 0.50 1.00 0.50 0.37 1.75
,
[s] 6.4 10 [A] [W] [cm] [cm]
E AUX p heat
I P a R t x

(2.103)
Although this scaling expression was developed on the basis of data
from small to medium-sized devices, it did a remarkably good job predicting
the L-mode confinement trends and actual energy confinement times in the
generation of large machines (TFTR and JET) that would soon, at that time, be
operational [2.244, 2.245].
To extend this study, Kaye [2.246] embarked on assembling a much broader
spectrum of data from a larger number of devices in order to better parameterize,
in a statistical manner, the scaling of L-mode confinement. Neutral beam heated
data were used to assemble the first global confinement database, which consisted
of 677 representative L-mode discharges from ISx-B, DITE, TFR, D-III, PDx,
PLT and ASDEx. The dataset consisted of mixed D
+
and H
+
plasmas, so no
isotope scaling could be derived. The analysis that was performed assumed that
the energy confinement time depended on the independent engineering variables
in a power law fashion, and using a two-step weighted regression procedure, first
to develop the non-size variable scaling and then the size variable scaling, found
that the energy confinement time scaled as:
128
GOLDSTON et al.
8 0.28 1.24 0.09 0.26 13 3
[s] 2.77 10 [kA] [T] [10 cm ]
E p e
I B n
o
t x


0.58 0.49 1.65
[MW] [cm] [cm]
heat
P a R

(2.104)
The fit of the scaling to the full dataset was good, with a coefficient of
determination
2
R = 0.94.
Yushmanov et al. [2.247] used an expanded database, which included data
from the devices used in Kaye-Goldston as well as from JT-60, TFTR, JET,
JFT-2M, DIII-D and T-10 to produce an updated scaling known as ITER89P,
which included a dependence on isotopic mass, ( / )
i p
M m m , given by
0.5 0.5 0.85 0.2 0.1 20 3
[s] 0.048 [MA] [T] [10 m ]
E p e
M I B n
o
t x
0.5 0.3 1.2
[MW] [m] [m]
heat
P a R

(2.105)
Finally, Kaye [2.248] updated the L-mode database, for the final time to
date, by including additional data from the devices already in the database as
well as data from Alcator C-MOD, TExTOR and TORE-SUPRA. This database
consisted of 1881 L-mode time slices that included radiofrequency as well as NBI
heated discharges. A number of variations of the power law scaling expressions
can be found in Ref. [2.229].
Global scalings for L-mode plasmas were also cast in the form of an
offset-linear expression, which attempted to separate the confinement behaviour
in the core from that farther out, near the edge. The offset-linear expression by
RebutLallia [2.249] did not fit the data as well as the power law scalings. The
RebutLallia expression, however, was meant to describe electron confinement
only, with the total plasma energy assumed to be twice that of the electrons. As
such, the model does well in describing the confinement scaling in many electron
dominated plasmas.
2.4.2.2. H-mode confnement trends and scalings
The L-mode represents only just one confinement state in plasmas, and,
in the absence of severe MHD activity, appears to give the miminum energy
for a set of given engineering parameters. Other confinement states exist, ones
in which transport barriers (i.e. regions of reduced transport resulting in large
pressure gradients) occur both within or near the plasma core (ITBs or internal
transport barriers) or near the plasma edge (ETBs or edge transport barriers).
The H-mode is an example of an ETB plasma. The H-mode was initially seen
in experiments with closed divertor operation (impeded neutral transport from
the divertor to the main chamber regions), and it was first reported on ASDEx
[2.188], shortly followed by reports from PDx [2.250] and later on from D-III
129
PHYSICS OF CONFINEMENT
FIG. 2.17. Time dependence of various plasma parameters of the L-mode (left column) and
H-mode (right column): (a) line-averaged density, (b) external gas flux in millibar-litre per
second, (c) atom flux reflected from divertor plate, (d) central electron temperature, (e) beta
poloidal
u
. Reprinted from [2.188], Copyright (2010), the American Physical Society
1
.
1
[2.251]. The H-mode was due to a spontaneous transition of the plasma from the
L-mode confinement state to a new state in which the energy confinement could
be a factor of approximately 2 higher for similar engineering parameters. The
transition to the H-mode was seen as a sudden drop in the D
o
emissivity, an
indicator of recycling at the plasma edge, an increase in the time rate of increase
of the plasma density, indicating possibly a higher particle confinement, and the
development of pedestals in the edge electron temperature and density profiles,
reflecting the development of an edge transport barrier. A comparison of the time
evolution of an L-mode (left column) and H-mode (right column) early ASDEx
discharge (from Ref. [2.188]) is shown in Fig. 2.17. While the earliest H-modes
were seen only in NBI heated discharges, this confinement regime was later on
seen in plasmas with RF heating, in those with open divertors or even in limiter
1
For material reproduced with permission from the American Physical Society, readers
may view, browse, and/or download material for temporary copying purposes only, provided
these uses are for non-commercial personal purposes. Except as provided by law, this material
may not be further reproduced, distributed, transmitted, modified, adapted, performed, dis-
played, published or sold, in whole or part, without prior written permission from the American
Physical Society.
130
GOLDSTON et al.
configurations, and on stellarators [2.252]. The H-mode has proved itself to be
a robust feature throughout the range of magnetic confinement configurations.
Because of the large gradients near the plasma edge, a class of instabilities,
called edge localized modes or ELMs was found to be a distinctive and unique
feature of the H-mode. The ELMs, seen in the right panel (c) in Fig. 2.17, are
driven unstable by the edge pressure and/or current gradients [2.253, 2.254],
periodically lower the edge confinement in the H-mode. The size and frequency
of these ELMs vary, but can cause significant instantaneous deposition of power
on the divertor plates.
The H-mode confinement time showed trends qualitatively similar to those
in L-mode plasmas. The trends are a degradation of confinement with heating
power, and an increase of confinement nearly linearly with
p
I and approximately
as
1/ 2
x . As in the L-mode, virtually no dependence on toroidal field or plasma
density was seen for the total, including beam particles, confinement.
In order to determine the confinement trends more quantitatively and
comprehensively, an effort was undertaken starting in 19881989 to assemble
an H-mode confinement database similar to that for the L-mode database.
Representatives from ASDEx, D-III, JET, JFT-2M, PBx-M and PDx met
regularly for approximately two years to define the elements of the database and
to ensure that the data submitted by each experiment were defined similarly by all.
This database went farther than the L-mode database in that the thermal plasma
confinement time was defined and input as well as the total confinement time
(plasma plus fast ion content). A total of 3466 observations were assembled in the
first H-mode confinement database [2.255]. The group went a step farther, and
that was to define a set of selection criteria that could focus analysis on a subset
of discharges. The selection criteria eliminated those discharges from the analysis
that had high radiated power or fast ion fractions, that were too transient, and that
had too much MHD activity either through beta being too high or q being too
low. After applying these criteria, a total of 1239 observations formed the subset
to be analysed. The first multiple linear regression result for this standard
H-mode data set yielded a scaling, for total energy confinement time, given by
0.5 0.68 0.75 0.60 0.16 19 3
[s] 0.025 [MA] [T] [10 m ]
E p e
M I B n
o
t x

0.55 0.05 2.04

[MW] [m] [m]
L
P a R

(2.106)
with a root mean square error (RMSE) of 13.8%. The RMSE is a metric of the
deviation of the fit to the experimental points, and it will be defined later in this
chapter. In the above expression,
L
P is defined as the total heating power less the
time rate of change of the stored energy ( /
heat
P dW dt ). It was important to
determine whether this, or any other scaling, was representative of each device,
or whether it represented an average of all. This could be done by computing
the uncertainties in each scaling parameter coefficient, and it was found that the
131
PHYSICS OF CONFINEMENT
above expression did indeed represent the trends across all machines, except
for the B
o
scaling, which is artificially strong in the above expression due to
interdependences among some of the scaling parameters. How to handle these
interdependences is discussed in the next section. Various other scalings were
derived from this dataset, further constrained on whether the discharges were
ELM-ing or ELM-free.
Updates to the H-mode confinement database, selection criteria for standard
subsets, and, thus, confinement scaling have continually taken place since the
first results were reported. These were reported in Refs [2.230, 2.256, 2.257]. The
updates to the database involved adding data from existing machines as well as
data from additional devices (ultimately with data included from Alcator C-MOD,
ASDEx-Upgrade, COMPASS-D, DIII-D, JT-60U, MAST, NSTx, START, T-10,
TCV, TdeV, TExTOR, TFTR and TUMAN-3M in addition to the data from
the original six devices). Various additional constraints were applied on the full
dataset to develop a number of standard subsets; the additional constraints
(not all applied simultaneously) included (but were not limited to) those on ion
to electron temperature ratio, an upper limit to the internal inductance, isotopic
mass, ELM activity and type, density evolution and ITER-specific criteria
narrowing the range of x ,
cyl
q and isotopic mass, where q
cyl
is the cylindrical
equivalent q. The full, and present, H-mode database consists of a total of 10 382
data entries from twenty devices. Only 3093 of these discharges qualify for the
most basic standard dataset.
The most widely used power law scaling, called ITER98y,2 [2.230] is
based on the parametric trends of the thermal confinement time, and given by the
expression
0.19 0.78 0.93 0.15 0.41 19 3
[s] 0.0562 [MA] [T] [10 m ]
E a p e
M I B n
o
t x

0.69 0.58 1.39

[MW] [m] [m]
L
P a R

(2.107)
In this expression, the elongation
a
x is defined as (
2
/ Area a r ) instead of
the usual ratio of vertical/horizontal extent.
The power law scaling expression determined from the most up-to-date
standard dataset is given by
0.08 0.84 0.86 0.21 0.39 19 3
[s] 0.0593 [MA] [T] [10 m ]
E a p e
M I B n
o
t x

0.65 0.68 1.31

[MW] [m] [m]
L
P a R

(2.108)
and the resulting fit to the data used to determine the scaling, taken from
Ref. [2.257], is shown in Fig. 2.18.
132
GOLDSTON et al.
FIG. 2.18. Comparison of the experimentally observed energy confinement time with a scaling
expression derived from an ordinary least squares fit to the standard dataset. Taken from Fig. 6
in Ref. [2.257].
In Ref. [2.258], Kaye et al. specifically took into account the confinement
trends introduced by the low aspect ratio tokamaks, NSTx and MAST. These
tokamaks, whose aspect ratio / 1.3 R a , as compared to conventional aspect
ratio tokamaks, where Ref.
/ 2.5 3 R a
, exhibit confinement trends that are
different. Specifically, as reported in Refs [2.142, 2.259, 2.260], the confinement
time varies almost linearly with toroidal magnetic field and only as the square
root of the plasma current. These trends were confirmed in recent experiments on
MAST, another low aspect ratio tokamak [2.261]. The magnitude of the power
degradation of confinement depends on whether the plasma is ELM-ing or not. It
can be near zero (no ELMs) or quite high (large ELMs at high power). Previous
scaling expressions tended to overpredict the confinement time at low aspect
ratio. Including these specific aspect ratio effects did change the scalings to some
extent, increasing the dependence on B
o
, reducing the dependence on
p
I , and
changing the aspect ratio dependence to compensate for the overprediction. A
scaling resulting from this analysis, and obtained by ordinary least squares
regression, is
10 0.80 0.32 0.39 3
[s] 6.25 10 [MA] [T] [m ]
E p e
I B n
o
t

0.66 0.95 1.17

[MW] [m] [m]
L
P a R

(2.109)
133
PHYSICS OF CONFINEMENT
for deuterium only discharges (M = 1.6 to 2.4), and with x in the range from 1.6
to 2.4.
Two term H-mode scalings, separating the core and the pedestal
confinement, have also been developed [2.257, 2.262], where the core and
pedestal stored energies are given by
0.23 0.24 0.88 0.11 0.49 19 3
[MJ] 0.103 [MA] [T] [10 m ]
core a p e
W M I B n
o
x

0.25 1.22 0.80

[MW] [m] [m]
L
P a R (2.110)
and
4 0.20 1.81 1.58 0.06 0.08 19 3
[ ] 6.43 10 [MA] [T] [10 m ]
ped a p e
W MJ M I B n
o
x

0.42 2.13 3.21 2.09

[MW] [m] [m]
L q
P a R F

(2.111)
where
95
( / )
q cyl
F q q and
E
( ) /
core ped L
W W P t + . Here, q
95
is the q-value at
the 95% flux surface. The fit to the two-term model is not as tight as that to the
better power law expressions, with about a 60% increase in the RMSE.
2.4.2.3. Advanced statistical methods and error analysis
The GSEs discussed to this point were all determined using ordinary
least squares multiple linear regression (OLSR). This technique [2.263] is
a direct method of calculating fit coefficients from a linear equation. The
assumption that the confinement time can be cast in the form of a power law,
1
( )
j NVAR
E j j
c X
o
t

H , where the
j
X are the independent variables and NVAR is
the number of independent variables, lends itself to linearization by taking the
logs of the independent (predictor) and dependent (response) variables, so that
0
1
log log
NVAR
E j j
j
X t o o

+

(2.112)
where
0 1
, o o are the scaling coefficients.
For a set of N observations and NVAR independent variables, the least
squares solution can be set up as a minimization problem, where
2
0
1 1
log log
N NVAR
E j j
i j
S X t o o

| `



. }

(2.113)
and the
j
o scaling coefficients are determined by solving the matrix given by
/ 0
j
S o o o . In the applications described in the previous sections a weighted
OLSR technique was used in which the observations from each device were
134
GOLDSTON et al.
weighted by a value w
j
inversely proportional to the number of observations
from that device, using either
1/ 2
w 1/
j j
N or
1/ 2
w 1/ (2 / 4)
j j
N + .
The precision of the fit given by OLSR can be defined by a number of
metrics, one of which is the root mean square error, defined as
RMSE =
( )
1/ 2
2
exp
1
RMSE=
2
N
fit
i
y y
N

]
]

]
]

(2.114)
where
exp
y is the log of the experimental value,
fit
y is the log of
the corresponding value from the scaling and N is the number of
observations. Another metric is the coefficient of determination, defined as
( ) ( )
2 2
2
, ,exp
/
i fit avg i avg
i i
R y y y y

, where
avg
y is the average value
of the dependent variable. In addition, variances, or uncertainties, in each scaling
coefficient can be readily calculated [2.263].
The simple OLSR technique by its nature does not consider errors in either
the independent or dependent variables. There are several methods that have
been used [2.230, 2.248, 2.255, 2.257, 2.258, 2.2642.266] that can take into
account the measurement errors, which can be significant. The first such method
is to perform a regression not on the variables themselves, but the principal
components (PCs) of the variables. The PCs are a set of orthogonal eigenvectors
whose variance is related to the eigenvalue. A PC with a small eigenvalue is one in
which the variation of data in that direction is small. The regression is performed
only on the most important PCs, which are taken to be those whose variance,
pc
i , is significantly greater than the measurement errors,
e
i , in the variables
that make up that PC. Typically, the measurement error has been considered to be
negligible relative to the variance in the PC if 4
pc e
i i > .
An offshoot of this technique is the principal component error-in-
variable (PCEIV) technique, also called EIV or EVOR (error-in-variables
orthogonal regression). In this approach the PCs of the logs of the predictor
and response variables, normalized by their individual errors, are computed. A
PC with a zero eigenvalue gives an exact linear relation among the variables
(i.e. 0
j ij i
i
w z o

, where the summation is over the number of predictor

variables
i
z ). In practice, the minimum eigenvalue is not zero, but it is small and
it is related to the scatter of the data about the PC. For this PC, the eigenvalue
is assumed to be zero, and the normalized variables are expressed in their
non-normalized form to give an expression for the
E
t . Differences between the
OLSR and PCEIV solutions can be readily seen in Refs [2.257, 2.258].
The PCEIV method is particularly powerful for being able to quantify
the uncertainties in the scaling coefficients. This can be done by varying
the experimental uncertainties of the dependent and independent variables.
Figure 2.19 is taken from Ref. [2.265], and it shows the scaling coefficients for
135
PHYSICS OF CONFINEMENT
power P and density n as a function of the relative uncertainty in P for three
different uncertainties in thermal plasma stored energy
th
W . For the dataset
used in this analysis, the average relative uncertainties in these variables are
12%
th
W o , 12% P o and 5.9% n o . These variables have the largest
uncertainties of all the variables used in the analysis. As can be seen in
Fig. 2.19, large variations in both exponents can result as the uncertainties are
varied, leading to large uncertainties in extrapolations of the resulting scaling
expressions.
FIG. 2.19. Scaling coefficients with power a
p
and density a
n
versus the error in power in per
cent for three different errors in W
th
[2.231].
Other statistical methods that have been used include two closely related
ones, maximum-likelihood and a Bayesian approach. These methods have
differing philosophies regarding the approach to probabilistic determinations, but
both use expressions describing the probability density for the measured data that
incorporates both the data and the errors. In applications of both methods for
the confinement studies, the value of the independent variables is assumed to
come from a normal distribution with a known standard deviation. In maximum-
likelihood, the best fit to the probability distributions gives rise to the parametric
scaling (see Ref. [2.266] and references therein).
136
GOLDSTON et al.
The Bayesian approach, while similar, assumes also a distribution for the
dependent variable that is based on prior knowledge of its value. In the application
in Ref. [2.258], the value of the dependent variable was also modelled as coming
from a normal distribution, but one with a prior known value but unknown error
(large variance). In this application, the Bayesian approach actually gave rise to
the scaling with the lowest RMSE (best fit), although the data in this, and the
maximum-likelihood, analysis in Refs [2.257, 2.258] were not weighted, and thus
the result was biased towards the devices with the largest number of datapoints.
2.4.3. Dimensionless analysis
2.4.3.1. Basics
The use of engineering variables to describe energy confinement trends
is convenient in that it allows experimenters to control confinement by
defining the set of operational parameters for which the experiment is to be
run. Engineering parameters, however, do not reflect the basic plasma physics
processes that truly control confinement. To take into account constraints imposed
by plasma theory, a different set of parameters that describe confinement must be
chosen. These parameters are dimensionless and represent fundamental internal
plasma characteristics. The set of these include
2
/ nT B ~ ,
0.5
/ T aB p

~
and
2.5 2
/ anq T v r

~ , and these, coupled with other dimensionless variables

( / a R r , x , q ) form a set that can describe plasma confinement:
( )
3 1 2
, , , , , , ,... ...
E
B F M q
o o o
t p v r x p v

- (2.115)
where
1 2
, ,... o o are the scaling coefficients.
Note that because these dimensionless (or physics) variables
can be expressed in terms of the dimensional (engineering) ones (e.g.
1 2 3 4 x x x x
E
B T B n a t , with T being related to P, n,
E
t and geometry and
x
1
,x
2
,x
3
being the scaling coefficients), the scalings derived for one set can be
expressed in terms of the other set by the above transformations using a simple
matrix inversion technique. As will be seen, this is important for assessing which
physics models most closely match the global scaling expressions.
In 1977, Connor and Taylor used the principle of scale invariance to
constrain the number of permissible scaling expressions [2.2672.269]. Starting
with the Vlasov equation, the equation for charge neutrality in the electrostatic
limit, and including Maxwells equations (Faradays and Amperes Laws) in the
electromagnetic limit, scale invariance implies that there is a constraint imposed
on the scaling coefficients,
i
x , so that the scaling of
E
Bt is invariant under a
linear transformation of the parameters that define the above equations. Applying
physics constraints imposed by various theoretical plasma models (e.g. ideal and
137
PHYSICS OF CONFINEMENT
resistive MHD, collisional and collisionless drift waves), Connor and Taylor
[2.267] and Connor [2.268] developed generalized scaling expressions based
on the functional forms with , p

and v

among the fundamental scaling

parameters. The relations among the scaling coefficients of these parameters
determined how well the confinement trends could be described by the various
physics models.
Because of the importance of the physics variable scalings in reflecting
underlying physics models controlling confinement, a large number of carefully
planned experiments studied the confinement variation with these parameters.
In these experiments, one dimensionless parameter was varied while the others
were held as constant as possible across their entire profile. An excellent recent
review of these experiments and their results is given in Ref. [2.269]. We will
follow closely the format of Ref. [2.269] in describing the highlights of the
dimensionless single parameter scaling experiments and we will focus on the
results of the H-mode experiments.
All devices found that the normalized confinement time increased with
increasing collisionality, with
E
B
v
o
t v

, where the scaling coefficient for v ,

v
o varied from 0.2 to ~0.8. The values of
v
o , however, were found to depend on
v itself, with
v
o increasing with increasing v (Fig. 2.20), indicating that the v
dependence was not a simple one. The overall scaling is consistent with physics
models indicating increased damping of zonal flows that suppress turbulence, or
decreased resistive ballooning modes, with decreasing collisionality.
FIG. 2.20. Collisionality scaling exponent of thermal energy confinement for H-mode plasmas
as a function of the ion collision frequency normalized to the bounce frequency. Taken from
Fig. 2 in Ref. [2.269].
138
GOLDSTON et al.
Single parameter scans with well controlled ELM activity in DIII-D
[2.270], JET [2.271] and NSTx [2.142] showed that there was little scaling
of normalized confinement with . In particular, a large range of toroidal
beta
o
was accessible in the low aspect ratio, low B
o
NSTx device. Such
results could imply that electromagnetic effects are not important. Results from
ASDEx-U [2.272] and JT60-U [2.273], however, indicate a strong degradation
of confinement on . The difference between these two sets of results is that
the strong degradation of confinement on was observed in plasmas that were
more weakly shaped (x , o , where o is triangularity). Future experiments will
attempt to identify if plasma shape is controlling, and why. Similar to Fig. 2.20,
Fig. 2.21 shows the range of scaling coefficients,

o , for the ~
E
B

o
t

results.
The situation for the scaling of confinement with normalized gyroradius is
not so clear. This scaling is particularly important in that it determines whether
the diffusion step-size is of the order of the ion gyroradius (
3
~
E
Bt p

) or the
device size (
2
~
E
Bt p

). The former is called gyroBohm scaling, while the

latter is referred to as Bohm scaling. GyroBohm scaling gives a much more
optimistic picture of confinement in future, lower p

devices. While H-mode
plasmas overall were found to scale in the gyroBohm regime, the transport was
found to have a local dependence, with the core transport scaling as gyroBohm,
but the transport in the outer region scaling as Bohm. Reference [2.269] goes on
to examine scaling trends with other dimensionless parameters, and the reader is
encouraged to read that article.
FIG. 2.21. -scaling exponents of the thermal energy confinement time of L-mode plasmas
(blue) and H-mode plasmas (red and green). Reprinted from Fig. 7 in Ref. [2.269]. Copyright
(2010), American Institute of Physics.
139
PHYSICS OF CONFINEMENT
2.4.3.2. Scalings
Full dimensional scaling expressions can be determined in two ways. One
way is to use the dimensionless variables as predictor (independent) variables
and perform direct statistical analysis as was done for the engineering variables.
The second is to cast the engineering variable scalings in terms of the physics
variables through the transformation between the two. Use of the first approach
is complicated by the fact that the physics variables can have high correlations
among themselves, due to the fact that they are, in part, defined by the same
parameters (e.g. T appears in , v

and p

, n appears in

and v

). Correlations
were found even in a reduced set of physics variables, with common factors
normalized out [2.258]. Also, correlations are found among some of the variables
that are not even this direct. For instance, a strong correlation was found between
r and or p

since the low aspect ratio devices NSTx and MAST, with high
/ a R r , operate at low B
o

and thus at high

and p

. Consequently, due to the

correlations among the variables and their errors, the fits based on these physics
parameters tend to be poor (in terms of the fit metrics) and uncertain.
It has, then, been more common to develop scalings based on the
engineering variables and then transform to physics variables. For instance, the
ITER98y,2 scaling transforms to, in physics variables,
2.70 0.90 0.01 3.00 3.30 0.73 0.96
~
E a
B q M t p v x r


(2.116)
The expression from Ref. [2.258], discussed previously, which takes into
account data from low aspect ratio devices but which was constrained to a narrow
range of elongation x and isotopic mass M, transforms to
2.86 0.70 0.09 2.26 0.62
~
E
B q t p v r


(2.117)
Both expressions show a near gyroBohm scaling, but both show a strong
degradation of the normalized confinement on , in contradiction to the single
parameter scans. Cordey et al. [2.265] showed that the uncertainties inherent in the
engineering parameters, which translate into uncertainties in the dimensionless
physics variables, as well as the relatively small range in these parameters that is
accessible experimentally [2.274], can lead to a large uncertainty in the -scaling
coefficient, with

o ranging from 0 to 0.6. The later work by Kaye et al. [2.258]

also showed a large range in the -scaling coefficient (

o ~ 0 to 0.8), and
further showed how this range related to the range of r -scaling coefficients
due to the correlation between these two variables. The latter result is shown
in Fig. 2.22 [2.258]. These results indicate the issues with a statistical approach
to determining the physics variable scaling in light of the magnitude of the
uncertainties, their interdependences and the limited accessible experimental
140
GOLDSTON et al.
range in these parameters. Future experiments must be carefully thought out to
avoid these pitfalls.
FIG. 2.22. Range of exponents for power (red) and b (blue) as a function of the exponent of e
from error analysis based on the PCEIV method. Taken from Fig. 4 in Ref. [2.258].
2.4.4. L-H threshold scalings
Operation in H-mode is critical to achieving high performance, and thus
the performance goals, in present and future devices. The design and operating
scenarios of these devices have to be developed based on confidence that such a
regime can be obtained. In an approach similar to that for the global confinement
scalings, controlled experiments have been conducted to study regimes in which
the H-mode is accessible, and these device-specific studies have been coupled to a
statistical approach in order to understand what variables are key to obtaining the
H-mode. These studies have been mostly empirical ones, and an understanding of
all the underlying physics that causes the L-H transition has been elusive.
The H-mode is achieved when a certain threshold power is surpassed. This
threshold power, like the energy confinement, depends on plasma parameters in
both gross as well as subtle ways. It is the case that the transition occurs at the
plasma edge, setting up temperature and density pedestals (i.e. transport barriers)
that can be confined to just inside the separatrix to extending farther into the
plasma on timescales of less than 100 s. The L-H transition is most easily
141
PHYSICS OF CONFINEMENT
obtained on devices using magnetic divertors, although transitions have been
observed both in limiter plasmas [2.275] and on stellarators [2.252].
The most important dependences of the L-H threshold are summarized in
Refs [2.230, 2.231, 2.276]. Early attempts to connect the L-H phenomenology
to theory are discussed in Refs [2.277, 2.278]. Studies to understand the L-H
transition in the framework of turbulence reduction due to a buildup of E B
shear at the plasma edge, and the relation to zonal flow generation, can be found
in Refs [2.277, 2.279]. While some attempts to connect experimental results
to the theories have been successful [2.280, 2.281], others have not [2.282,
2.283], underscoring the present lack of complete understanding of the transition
dynamics.
The key characteristics of the threshold dependences from early studies are
as follows. The threshold power is lower for single null plasmas with the ion
B V drift towards the dominant x-point than for single null configurations where
the drift is in the opposite direction or in a double null configuration that is not
perfectly balanced. Good wall conditioning and a divertor configuration leading
to good retention of neutrals in the divertor region, and a reduced neutral density
in the main plasma chamber, also lead to lower power thresholds. The threshold
power, based on single parameter scans across a range of devices, is found to
depend in almost a linear fashion on plasma density (above some critical density)
and toroidal field. Finally, the power threshold was found to be approximately a
factor of two greater for hydrogen than for deuterium plasmas.
Based on these results, a global database was assembled with the aim of
establishing empirical scalings of the L-H threshold. The database focused on
parameters that included global variables ( B
o
, I
p
, n
e
, plasma size and shape, etc.)
as well as other information found to influence the threshold power (plasma
wall gaps, ion B V drift direction, divertor configuration and some limited edge
plasma information). The edge information, however, is not complete owing to
difficulties in measuring some of these parameters accurately. A full description
of the L-H threshold database can be found in Ref. [2.276].
Analysis of this combined dataset at first attempted to determine the
dependence of the threshold power, P
LH
, on the variables such as B
o
and n
e
,
which were the most important dependences found in individual device scans.
P
LH
was defined to be the total power through the separatrix without core
radiation subtracted. This is a good assumption for present day devices, where
core radiation is small, but the assumption can be violated in future burning
plasma devices where core radiation from bremsstrahlung can be large. Statistical
analysis found dependences consistent with those found in the individual scans.
Dependence on device size could be found only through analysis of the combined
database, with the initial results revealing a strong dependence on device size,
with
2.5
~
LH e
P n B R
o
or
0.75
e
n B S
o
, where S is the surface area of the plasma.
These scalings are consistent with P
LH
being expressed in a dimensionally correct
142
GOLDSTON et al.
form ~ v P nT S , with T is temperature and v is velocity, taken to be the thermal
velocity (T/M)
1/2
([2.284, 2.285] and references therein).
As more data were added to the database, the analyses evolved but still
remained focused on the few global variables and reached similar conclusions.
The P
LH
scalings, however, had larger uncertainties in the scaling coefficients
than did the
E
t scalings, and the P
LH
scalings led to wider ranges in values due
to these uncertainties when extrapolated to future devices [2.286]. This was
found also even when different statistical techniques were applied to analyse the
threshold database [2.287]. Factors of almost two were found in predictions to
ITER, depending on which scaling expression was used.
These results suggested strongly that sufficient information was not
contained in the parametric descriptions, and that more advanced statistical
techniques could not overcome the lack of information. The consideration
of several other important factors was reported in Refs [2.286, 2.288]. In
Ref. [2.286], the role of aspect ratio was considered by including a parameter
( ) / ( )
mm
R R a R a + in the analysis, and this inclusion led to a small, but
statistically significant improvement in the fit (see Eq. 7 in Ref. [2.287]). The
consideration of aspect ratio was necessary owing to data from MAST and
NSTx, two low aspect ratio tokamaks, being included in the updated database.
In Ref. [2.288], aspect ratio was taken into account in two fashions. The first was
by using
2 2 1/ 2
{ } B B B
o u
+ instead of just B
o
. Here, B
u
is the poloidal magnetic
field. This recognizes that B B
o u
in the outer midplane for low aspect ratio
devices ( B B
o
at higher aspect ratio). The second was based on an assumption
that the details of the particle distribution, namely the trapped to untrapped
fraction, play a role in the transition dynamics. Thus, a parameter F(A)=A/f(A),
where A=R/a and f(A) is the fraction of trapped particles { }
1/ 2
1 2 / (1 ) A - + , was
included in the scaling. This work further considered the effect of Z
eff
, which
was clearly seen in the JT60-U data, as well as on other devices where this
information was available (although with greater scatter). The scaling emerging
from this analysis was
0.7
0.7 0.9 0.7
( )
LH e eff
P B n S Z F A
y
~ (2.118)
where y

= 0.5 0.5 (highly uncertain possibly due to non-linear dependences).
This scaling resulted in a better fit than for scalings using only the standard
global parameters. Not included in this scaling was an I
p
dependence of P
LH
that
was observed on the low aspect ratio NSTx device [2.289], but not at higher
aspect ratio.
Recent work has confirmed the roles, or has shown the importance, of other
variables in determining the threshold power. The P
LH
was found to be the same
for pure D
+
and pure He
++
plasmas heated by ECRH in ASDEx-U (Fig. 2.23,
taken from Ref. [2.290]), although most other experiments indicate that P
LH
(He)
143
PHYSICS OF CONFINEMENT
is about 20 to 40% higher than P
LH
(D). The threshold for H
+
plasmas was found
to be two times greater than that for D
+
with NBI heating [2.291]. Further, it
was shown also in Ref. [2.291] that the minimum threshold power for otherwise
fixed conditions occurred at low rotation, where the neutral beam torque was
actually less than zero (Fig. 2.24, taken from Ref. [2.291]). Other studies [2.286]
have shown that there exists a minimum in threshold power that is controlled
by plasma density. This minimum threshold density is a common feature in
a large number of tokamaks, and it occurs in the 2 to 4 10
19
m
3
range with
larger machines generally having smaller minimum threshold densities. This can
be seen in the ASDEx-U data in Fig. 2.23. Figure 2.25 shows this parameter
plotted as a function of toroidal field. For ASDEx-U, C-Mod and JFT-2M, the
value is seen to increase strongly with B
o
while for JET and JT-60U, the two
larger devices, the dependence is weaker (left panel). When normalized to the
Greenwald density value,
2
/
p
I a r [2.292], a slight decrease with B
o
for all
devices is seen (right panel).
0
0.4
0.8
1.2
1.6
2
2.4
2.8
2 3 4 5 6 7 8 9
ECH
H-NBI
ECH
D-NBI
P
t
h
r

[
M
W
]
n
e
[10
19
m
-3
]
4
He plasmas D plasmas
P
thr,scal08
FIG. 2.23. Threshold power versus line-averaged density from ASDEX-U. Taken from Fig. 4
in Ref. [2.290].
FIG. 2.24. Power threshold for D
+
and H
+
plasmas on DIII-D as a function of injected neutral
beam (NB) torque. Adapted from Fig. 3 in Ref. [2.291].
144
GOLDSTON et al.
FIG. 2.25. (a) Minimum threshold density plotted as a function of B
o
; (b) density normalized
to the Greenwald parameter versus B
o
[2.286].
An important physics issue having to do with these threshold scaling studies
is the confinement quality for heating powers very close to the threshold power.
Related to this is whether there is a hysteresis effect; that is, will a discharge
remain in H-mode as it evolves post-transition to a state where P
heat
<P
LH
(i.e.
at fixed heating power but higher density)? Studies [2.286] have shown that the
H-mode confinement enhancement is generally in the range of
98 ,2
0.6
y
H to
1.4, and it does not depend on the power ratio P
heat
/P
LH
.
2.4.5. Implications relative to theoretical models
Trying to relate the results of global scaling analyses to detailed physics
theories can be quite difficult. For one thing, theoretical underpinnings are most
often expressed in terms of local parameters that are not readily transferable
to global variables. Further, validation of these theories often takes place at a
lower level (e.g. local turbulence and transport properties), not at the global level.
Comparisons to specific theories, then, are best made at the local level, and this
will be covered in another subsection.
The global scalings can nevertheless be related to general classes of theories,
and this has been discussed to some extent in the section on dimensionless
scalings. For instance, when the scalings based on engineering parameters are
converted to those based on physics variables, it was generally found that L-mode
plasmas were more Bohm-like ( ~ a A ) rather than gyroBohm-like ( ~
i
p A )
[2.269]. This result is not that straightforward, however, in that while the ions
appeared to be Bohm-like or worse, the electrons were in the gyroBohm regime.
H-mode scalings typically appeared to be more in the gyroBohm regime, but even
here there were subtleties, with the core being in the gyroBohm regime but the
edge behaving more Bohm-like in some cases. There were, however, variations
from device to device that depended on the level of heating power relative to
145
PHYSICS OF CONFINEMENT
P
LH
and E B
. .
shearing rate ([2.269] and references therein). Other dependences
that directly re