IOP PUBLISHING and INTERNATIONAL ATOMIC ENERGY AGENCY NUCLEAR FUSION

Nucl. Fusion 47 (2007) S18–S127 doi:10.1088/0029-5515/47/6/S02

Chapter 2: Plasma confinement and

transport
E.J. Doyle1 (Chair Transport Physics), W.A. Houlberg2,a (Chair Confinement Database and
Modelling), Y. Kamada3 (Chair Pedestal and Edge), V. Mukhovatov4 (co-Chair Transport Physics),
T.H. Osborne5 (co-Chair Pedestal and Edge), A. Polevoi4 (co-Chair Confinement Database and
Modelling), G. Bateman6 , J.W. Connor7 , J.G. Cordey7 (retired), T. Fujita3 , X. Garbet8 , T.S. Hahm9 ,
L.D. Horton10 , A.E. Hubbard11 , F. Imbeaux8 , F. Jenko10 , J.E. Kinsey5 , Y. Kishimoto12 , J. Li13 ,
T.C. Luce5 , Y. Martin14 , M. Ossipenko15 , V. Parail7 , A. Peeters10 , T.L. Rhodes1 , J.E. Rice11 ,
C.M. Roach7 , V. Rozhansky16 , F. Ryter10 , G. Saibene17 , R. Sartori17 , A.C.C. Sips10 , J.A. Snipes11 ,
M. Sugihara4 , E.J. Synakowski18 , H. Takenaga3 , T. Takizuka3 , K. Thomsen17 , M.R. Wade5 ,
H.R. Wilson19 , ITPA Transport Physics Topical Group, ITPA Confinement Database and
Modelling Topical Group and ITPA Pedestal and Edge Topical Group
1
Department of Electrical Engineering and PSTI, University of California, Los Angeles, CA, USA
2
Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA
3
Japan Atomic Energy Agency, Naka, Ibaraki-ken 311-0193, Japan
4
ITER Organization, 13108 Saint-Paul-Lez-Durance, France
5
General Atomics, San Diego, CA 92186-5608, USA
6
Physics Department, Lehigh University, Bethlehem, PA 18015, USA
7
EURATOM/UKAEA Fusion Association, Culham Science Centre, Abingdon, Oxfordshire OX14 3DB, UK
8
Association Euratom-CEA, Centre d’ Etudes de Cadarache, Saint-Paul-Lez-Durance, France
9
Princeton Plasma Physics Laboratory, PO Box 451, Princeton, NJ 08543, USA
10
Max-Planck-Institut für Plasmaphysik, EURATOM Association, D-85748 Garching, Germany
11
Massachusetts Institute of Technology, Plasma Science and Fusion Center, Cambridge, MA 02139, USA
12
Graduate School of Energy Science, Kyoto University, Gokasho, Uji, Japan
13
Southwestern Institute of Physics, Chengdu, People’s Republic of China
14
Centre de Recherches en Physique des Plasmas, Association Euratom-Confédération Suisse, Ecole Polytechnique Fédérale de Lausanne,
CH-1015 Lausanne, Switzerland
15
Nuclear Fusion Institute, Russian Research Centre Kurchatov Institute, Moscow, 123182 Russia
16
St Petersburg State Polytechnical University, St Petersburg 195251, Russian Federation
17
European Fusion Development Agreement Close Support Unit, D-85748 Garching, Germany
18
Lawrence Livermore National Laboratory, Livermore, CA 94551, USA
19
University of York, Heslington, York, YO10 5DD, UK
E-mail: [email protected]

Received 12 June 2006, accepted for publication 12 April 2007

Published 1 June 2007
Online at stacks.iop.org/NF/47/S18
Abstract
The understanding and predictive capability of transport physics and plasma confinement is reviewed from the
perspective of achieving reactor-scale burning plasmas in the ITER tokamak, for both core and edge plasma
regions. Very considerable progress has been made in understanding, controlling and predicting tokamak transport
across a wide variety of plasma conditions and regimes since the publication of the ITER Physics Basis (IPB)
document (1999 Nucl. Fusion 39 2137–2664). Major areas of progress considered here follow. (1) Substantial
improvement in the physics content, capability and reliability of transport simulation and modelling codes, leading
to much increased theory/experiment interaction as these codes are increasingly used to interpret and predict
experiment. (2) Remarkable progress has been made in developing and understanding regimes of improved core
confinement. Internal transport barriers and other forms of reduced core transport are now routinely obtained in
all the leading tokamak devices worldwide. (3) The importance of controlling the H-mode edge pedestal is now
generally recognized. Substantial progress has been made in extending high confinement H-mode operation to the
Greenwald density, the demonstration of Type I ELM mitigation and control techniques and systematic explanation

a Author to whom any correspondence should be addressed.

0029-5515/07/060018+110$30.00 © 2007 IAEA, Vienna Printed in the UK S18

 Chapter 2: Plasma confinement and transport

of Type I ELM stability. Theory-based predictive capability has also shown progress by integrating the plasma and
neutral transport with MHD stability. (4) Transport projections to ITER are now made using three complementary
approaches: empirical or global scaling, theory-based transport modelling and dimensionless parameter scaling
(previously, empirical scaling was the dominant approach). For the ITER base case or the reference scenario of
conventional ELMy H-mode operation, all three techniques predict that ITER will have sufficient confinement to
meet its design target of Q = 10 operation, within similar uncertainties.

PACS numbers: 28.52.−s, 52.55.Fa, 52.25.Fi

(Some figures in this article are in colour only in the electronic version)

Contents

1. Introduction
2. Fundamental transport processes
2.1. Theory of turbulent transport
2.2. Turbulence simulation
2.3. Turbulence measurements in tokamaks and comparison with theory/simulation
2.4. Neoclassical transport
2.5. Summary of issues
3. Core transport
3.1. Enhanced core confinement regimes
3.2. Ion thermal confinement
3.3. Electron thermal transport
3.4. Particle and impurity transport
3.5. Toroidal momentum transport and spontaneous rotation
3.6. Dimensionless parameter scaling experiments
3.7. Improved core confinement regimes for advanced operation scenarios
3.8. Summary and outstanding issues
4. Pedestal transport and dynamics
4.1. Regimes of improved H-mode confinement at high density and operational limits
4.2. Pedestal characteristics and structure
4.3. L–H transitions
4.4. Pedestal transport theory and modelling
4.5. Modelling the pedestal structure
4.6. Type I ELM structure, effects on the pedestal profiles and mitigation techniques
4.7. Alternatives to Type I ELMy H-mode regime
4.8. Pedestal stability
4.9. Possible pedestal control scenarios
4.10. Summary of pedestal structure and transport
5. Predictive capability and projections for ITER
5.1. Improved database resources for modelling and scaling studies
5.2. Pedestal and edge characteristics
5.3. Global scaling
5.4. Non-dimensional scaling
5.5. Modelling codes, including edge modelling capability
5.6. Summary of progress and remaining issues
6. Summary

1. Introduction is governed by highly non-linear turbulence processes, with

multiple turbulence drives and suppression mechanisms,
A fundamental feature of tokamak plasmas is that measured occurring on multiple scales. Due to this inherent complexity,
energy transport rates typically exceed those calculated understanding transport in fusion plasmas is generally regarded
for binary collisions in a toroidal plasma, i.e. transport as a scientific ‘grand challenge.’ Despite this intrinsic
is anomalously higher than that predicted by ‘classical’ complexity, very considerable progress has been made since
(cylindrical geometry) or ‘neoclassical’ (toroidal geometry, the publication of the ITER Physics Basis (IPB) document [1]
including drift orbit effects) theory. The anomalous transport in understanding, controlling and predicting tokamak transport

S19
E.J. Doyle et al

across a wide variety of plasma conditions and regimes. It is global scaling and profile modelling studies, the status of
the goal of this chapter to describe this progress in theory, in predictive capabilities for the plasma edge and pedestal, the
numerical simulation/modelling and in experiment, covering latest global scaling, non-dimensional scaling and modelling
both plasma core and edge, and to present the current transport results and projections for ITER and the status of modelling
projections for ITER. capabilities in general. The chapter concludes with an overall
The structure and content of this chapter reflects both summary (section 6), and in addition each main section
substantial changes in emphasis within the field since the concludes with an individual summary and a list of outstanding
IPB was published and also a desire to address the major issues.
outstanding issues that it identified. Major areas of change and
progress since the IPB follow. (1) Substantial improvement 2. Fundamental transport processes
in the physics content, capability and reliability of transport
simulation and modelling codes, leading to much increased The purpose of this section is to provide a description of
theory/experiment interaction as these codes are increasingly both the current level of understanding and the outstanding
used to interpret and predict experiment. (2) Remarkable issues with regard to fundamental transport processes in
progress has been made in developing and understanding tokamak plasmas. This section is structured into subsections
regimes of improved core confinement, which is relevant as follows. Section 2.1 provides an introduction to turbulent
for the steady-state and hybrid scenarios that were not transport theory, covering issues of current concerns such as
emphasized in the ITER design at the time of the IPB. turbulence correlation time and length, non-linear turbulence
Internal transport barriers and other forms of reduced core self-regulation via zonal flows, streamers, multiple-scale
transport are now routinely obtained in all the leading tokamak spatio-temporal transport, probabilistic transport and Bohm
devices worldwide. Reduced transport has been achieved versus gyro-Bohm transport. In section 2.2, we detail
in all four transport channels (ion and electron thermal, the substantial progress made with turbulence simulation
particle and momentum transport channels), sometimes codes. Here, progress has been such that quantitative
simultaneously. (3) The importance of understanding and comparisons with experiment are now possible, and gyro-
controlling the H-mode edge pedestal is now generally kinetic simulations of electron as well as ion transport are
recognized, especially with regard to maximizing pedestal being performed. The succeeding section 2.3 provides a review
parameters while mitigating or eliminating Type I edge of quantitative experimental tests of theory and simulation,
localized modes (ELMs). (4) Transport projections to illustrating the substantial progress that has been made in
ITER are now made using three complementary approaches: experimentally testing and validating the standard theory of
empirical or global scaling, theory-based transport modelling turbulent transport. Neoclassical transport theory, which
and dimensionless parameter scaling (previously, empirical describes drift and Coulomb collision-driven transport, is
scaling was the dominant approach). Projections to ITER described in section 2.4. The routine achievement of regimes
using these three approaches show general agreement, though with reduced turbulent transport means that neoclassical ion
detailed differences remain between and within the three transport is now often observed, and neoclassical theory is
techniques. also widely used in bootstrap current and poloidal rotation
Given these areas of emphasis within the current transport calculations. Consequently, the validity and limits of
research, the remainder of this chapter is structured as neoclassical theory are being tested more rigorously than
follows: section 2 presents the progress in understanding previously. Finally, section 2.5 provides a summary and a
fundamental transport processes, including analytic theory, list of outstanding issues.
numerical turbulence simulations and neoclassical transport
theory, as well as progress in efforts to benchmark theory via
2.1. Theory of turbulent transport
direct comparisons with turbulence measurements. Section 3
presents the current understanding of tokamak core plasma Understanding turbulence-driven (anomalous) transport is one
transport, covering ion and electron thermal, particle and of the most important issues in present magnetized plasmas
momentum transport, as well as dimensionless parameter and future fusion reactors. As realized in previous years,
scaling experiments and the transport properties of enhanced plasma turbulence is driven by different free energy sources
core confinement regimes. Progress in experiment and for micro-instabilities, mainly the inhomogeneity of plasma
modelling are integrated in the descriptions and discussion of profiles including plasma temperature and density as well as
this section, reflecting the close and still evolving coupling the equilibrium magnetic field. The main instabilities that
between both. The transport and structure of the edge H-mode may contribute to anomalous transport in tokamaks were
pedestal are considered in section 4. Topics covered here briefly reviewed in Chapter 2 of the ITER Physics Basis
include obtaining high density with high confinement, pedestal document [2]. There has been considerable progress since
characteristics and structure, L–H transitions, pedestal theory then in understanding and modelling turbulent transport
and modelling, ELM structure, ELM mitigation techniques in tokamaks. This has been greatly assisted by large
and alternatives to operating in the Type I ELM regime and increases in computational power, giving the ability to simulate
pedestal stability and control. turbulence and transport with ever more realistic plasma
Confinement projections for ITER and the status of model equations, as discussed in section 2.2, enabling the
predictive transport capabilities are presented in section 5. construction of physics-based transport models. This has
Subsections here include a description of the improved advanced our understanding of the complexity of plasma
database resources, both scalar and profile, now available for turbulence, of the formation of large-scale structures such

S20
 Chapter 2: Plasma confinement and transport

as zonal flows, streamers and their non-linear self-regulation understanding of turbulent transport has become evident. It
processes and of the physical mechanisms of internal/edge emphasizes the dynamic non-linear interaction between the
transport barrier (ETB/ITB) formation as well as the L–H different scales and may reveal new channels of energy as
transition. Furthermore, new trends have become clear well as particle and momentum transport, so that the plasma
from the emphasis on some important issues such as the confinement can be experimentally controlled.
multiple-scale interaction between turbulence and structures,
the spatial transport of turbulence and the coupling between 2.1.1. Physical mechanisms and description of turbulent
core and edge, the underlying mechanism of electron transport, transport. In a plasma, binary collisions are the basic
a new transport phenomenon in alpha particle heated burning mechanisms for the cross-field particle and heat transport based
plasmas such as ITER and transport feedback control. on the classic random walk model. It is generally described by
Regulation of turbulent transport and the resultant a diffusion coefficient, χ:
confinement improvement in tokamak plasmas have been
shown to be related to complex structures and dynamics [3]. χ ∼ λ2 /τc , (1)
Although plasma discharge conditions evolve gradually in
space and time, such as the safety factor profile or the auxiliary where the characteristic walking time τc and the step length λ
heating power, distinct confinement improvements may are given by the collision time (i.e. inverse collision frequency)
abruptly occur and have prominent structures such as an ITB and the gyro-radius of the ions or the electrons, for ion or
or/and an ETB. It is recognized that the formation of these high electron heat transport, respectively, in a cylindrical geometry.
confinement modes, including potentially the L–H transition, In a toroidal plasma this collisional transport is enhanced
may closely relate to the suppression of turbulence by sheared by particle drift orbit effects due to the inhomogeneous



E × B or turbulence-driven zonal flows [4]. A change in equilibrium magnetic field. This process is called neoclassical
magnetic topology, such as the formation of magnetic islands, transport. For a typical tokamak plasma, neoclassical ion
is another plausible mechanism [5, 6]. Transport barriers thermal conductivity is of the order √ χi ∼ 0.1 m2 s−1 and
can lead to higher confinement in an ITER burning plasma, the electron counterpart is smaller by me /mi (with me /mi
although generating a core-sheared flow to trigger and sustain being the electron to ion mass ratio) [8]. However, the ion
ITBs is a challenge for an ITER-scale device. The key neoclassical transport level is much lower than the observed
questions are in regard to the mechanisms responsible for perpendicular transport in most tokamak plasmas, and the
triggering and controlling transport barrier formation in the electron neoclassical transport level is never achieved. The
various channels, their strength, width and locations, impact on higher observed transport level has traditionally been referred
confinement, their dynamics and sustainment, and the roles of to as anomalous transport, which we now understand from



E × B flow shear and magnetic shear as well as the Shafranov theory, modelling and experimental analyses is caused by
shift in these processes [7]. Transport in tokamak plasmas various turbulent fluctuations.
is a rather complex process not only due to the highly non-
Description of turbulent transport and its properties.
linear dynamics of turbulence but also due to the dependence
Turbulent transport in a tokamak plasma is mainly produced
of the magnetic geometry and the device size. The shape of
by micro-scale drift-type turbulence, which

is driven


by the
the plasma cross-section and the
divertor design may influence

gradients of temperature or density. The E × B convection
the generation of sheared E × B or zonal flows in the central
due to the turbulent electric field rotates the plasma element
and edge regions. Synthesizing all these effects is a challenging
along the electric potential contours across the magnetic field.
task, but it may predict a transport scaling with Bohm or gyro-
Once the phase difference between potential and density
Bohm dependence for ITER or a future DEMO, which is still
perturbations is established, this process largely determines the
to be resolved.
radial particle and heat loss from a high temperature plasma.
Electron transport is a particularly important issue for At the most general level, one can investigate what physical
ITER. Electron heating by the fusion-generated alpha particles mechanisms may be involved in turbulent transport coefficients
will be dominant in ITER and future reactors, but energy based on the dimensional analysis or the scaling properties
transfer will lead to Ti ≈ Te . However, from the point of the governing equations. However, a specific form for
of view of theory and modelling, electron heat transport

these coefficients may be much preferred [9]. The analytical
has, so far, been less studied than ions. Sheared E × B description of the net convective particle and the thermal fluxes
flows with equilibrium scale or weak zonal flows are less across a given surface S, which are represented by i,e and qi,e ,
effective for the short wavelength electron turbulence. Hence, are given by [10]:
the dynamics of other large-scale structures such as radially

1
dni,e dTi,e
elongated eddies, for e.g. streamers, has attracted attention. i,e = ni,e vE · d a = −D11 − D12 , (2)
Furthermore, fluctuations from the long wavelength ion scale S S dx dx
to the intermediate skin depth scale may drive electron

1
dni,e dTi,e
transport through various non-adiabatic responses by the qi,e = ni,e Ti,e vE · d a = −D21 − D22 , (3)
S S dx dx
electrons, such as trapped electrons and renormalized current



diffusion. where vE is the E × B convection velocity, a is an area vector
Based on the realization of various linear and non-linear normal to the flux surface and ni,e and Ti,e are the equilibrium
instability free energy sources which cover a wide spatio- density and the temperature of ions and electrons, respectively.
temporal range from the ion to the electron gyro-radius, a These equations represent flux (particle or heat) versus gradient
new research direction that is essential for a comprehensive (density or temperature) relations. The matrices Dαβ provide

S21
E.J. Doyle et al

follows:


1 dni,e
i,e = ni,e v · da = −Di,e

eff
, (4)
S S dx


1 dTi,e
qi,e = ni,e Ti,e v · da = −χi,e

eff
. (5)
S S dx
Calculating these diffusion coefficients is the objective of
turbulent transport theory. They are determined by the
properties of various micro-instabilities. The main instabilities
that are possible underlying mechanisms for plasma turbulence
in tokamaks are summarized in [2].
In a toroidal plasma, linear stability analyses predict the
existence of a temperature gradient threshold for the ion or
the electron temperature gradient driven instabilities, referred
Figure 1. Three types of electron thermal flux versus electron
to as ITG or ETG modes, respectively (or ηi (ηe ) modes),
temperature gradient. Type I exhibits a threshold, Type II represents
a purely diffusive model where curvature is associated with a which belong to Type I. Here ηi = |Ln /LTi |(ηe = |Ln /LTe |)
non-linear dependence on the temperature gradient, and Type III has denotes the ratio of the scale length between the density and
contributions to the flux from other gradients. Reprinted with the ion (electron) temperature. Micro-tearing and current
permission from [10]. © 2003 Institute of Physics. diffusive ballooning modes, where dissipative processes such
as collisional resistivity or anomalous current diffusion play an
essential role, may be categorized as Type III. TEM and TIM
physical insight into the important unstable modes. A are sometimes driven by density gradients as well and can have
phase shift between the fluctuating electrostatic potential, φ̃, behaviours ranging from Type I to Type III, depending on the
and the fluctuating density, δ ñ, or the temperature, δ T̃ , can value of the density gradient and the collisionality. At the edge,
produce the particle or the heat transport. Different phase the drift-Alfvén mode and pressure (or resistivity and current)
shift leads to the off-diagonal terms in the above equations. gradient driven modes may become more important.
It may induce an inward heat flux contribution that gives a
critical gradient above which the heat flux rises to a high level. Methods for describing diffusion. Generally speaking,
Generally, different signs and parameter dependences of the turbulent transport is determined by two factors. One is the
off-diagonal terms D12 and D21 may lead to different types fluctuation level, i.e. the saturation amplitude of turbulence.
of flux-gradient relations that occur in plasmas. Figure 1 Another is the phase relation of the turbulent structures. It
illustrates three different types, in which the electron heat is frequently described by an estimate using a random walk
transport is taken as an example with the flux qe = χe ∇Te [10]. model as described in equation (1). There are two approaches
Type I represents a critical temperature gradient model above to measure the turbulent diffusivity coefficients. The most
widely invoked quasi-linear expression estimates the saturation
which the transport process starts. A rapid increase in transport


amplitude by balancing the E × B non-linearity against the
when the critical gradient is exceeded anchors the profiles near
drift wave frequency, ω∗ , leading to the familiar mixing length
the critical value; this property is referred to as profile stiffness.
estimate,
Type II describes a simplified turbulent transport model, where
D⊥ ∼ (γL /k⊥ 2
)max k⊥ , (6)
the flux vanishes when the gradient vanishes, although the
flux can have a non-linear dependence on the gradient that where γL is the linear growth rate of the instability and k⊥ is a
produces curvature. Type III occurs when the turbulence characteristic perpendicular wavenumber of the turbulence [9].
is driven by other gradients, such as the ion temperature Equation (6) can be interpreted as balancing the growth rate
or the electron density gradient. Type III is extensively against the turbulent



diffusion at saturation. This is an upper-
applied in transport modelling codes that implement theory- bound: if the E × B non-linearity is balanced against the
based models for predictive analysis [11–13], whereas growth rate instead, then the weak turbulence result follows,
Types I and II are typically employed in interpretive
D⊥ ∼ (γL /ω∗ )(γL /k⊥
2
). (7)
analyses.
Due to the complexity of the plasma turbulence, a When the linear or the non-linear growth of waves at
comprehensive theoretical description of anomalous transport longer wavelengths is balanced by the energy transfer to the
is still being developed. Turbulent transport models can short wavelength region by some dissipative processes such
generally predict a complicated relation between particle and as viscosity, the growth rate and perpendicular wavenumber in
heat transport and show the importance of the off-diagonal equation (6) should be replaced by the inverse correlation time,
terms. However, one would prefer a simple model for τC , and the correlation length, LC , respectively, in a quasi-
the turbulent transport, in which the particle and energy steady turbulent plasma. Equation (6) is then expressed as
fluxes have a form that is analogous to the classical or the follows:
neoclassical ones [14]. Thus, equations (2) and (3) for total D⊥ ∼ L2C /τC , (8)
particle and heat fluxes are expressed in terms of effective for the statistical analysis of turbulence, which may be linked
eff eff
particle and thermal diffusivity coefficients, Di,e and χi,e , as to fluctuation diagnostics, i.e. with this formula the transport

S22
 Chapter 2: Plasma confinement and transport

level can be analysed by measuring the statistically averaged

correlation time and length.
An alternative phenomenological point of view for
determining self-consistently the instability saturation and the
turbulent heat transport is that such turbulent diffusivity not
only produces the observed anomalous transport but may
also influence the stability properties of the mode itself and
quench the instability through the non-linear damping rate,
γd . The balance between non-linear and linear processes
may lead to a specific form of turbulent diffusion coefficient
which involves the instability [15]. For example, an electron
heat diffusion coefficient based on the trapped electron mode
is derived in this way to explain the formation of electron
internal transport barriers (ITB) in strongly reversed shear
plasmas [16]. An anomalous transport model has also been Figure 2. An outline summary of drift wave turbulence scales, with
derived by balancing the non-linear growth of the current corresponding turbulence mechanisms, affected transport channels
diffusion ballooning mode (CDBM) against other turbulent and stabilization mechanisms. Reprinted with permission from [22].
transport effects such as viscosity and thermal diffusion [17].
Under this quasi-linear estimate, the analyses of the linear wavelength electron-scale fluctuation, the ETG mode, may
or non-linear growth rates and the mode structures for the produce large electron heat flux. ITG and ETG instabilities
instabilities are necessary and useful. have critical temperature gradient thresholds, which categorize
The above relatively simple estimates are being them as Type I transport models in figure 1. The trapped
superceded by the results of state-of-the-art computational electron mode (TEM) with its intermediate spatial scale (meso-
models of turbulent transport. This approach takes the scale) can cause electron particle transport and also turbulent
fundamental equations describing the plasma and computes the electron heat flux; it is categorized as Type II. Micro-tearing
turbulent fluctuations and consequent transport. A hierarchy and current diffusive ballooning modes, in which dissipative
of plasma models can be considered. The most complete processes such as collisional resistivity or anomalous current
is at the level of the Vlasov or the Fokker–Planck equation, diffusion play essential roles, may be categorized as Type III.
but this can be reduced to the somewhat simpler gyro-kinetic At the edge, the drift-Alfvén mode and the pressure (or
description that averages over the rapid gyro-motion [18]. resistivity and current) gradient driven modes may become
Further approximations result from fluid closures, such as more important.
the gyro-Landau fluid moment equations that capture the
kinetic effects of Landau damping or ultimately the two-fluid Turbulent particle pinch. Particle transport is a central
Braginskii equations, normally only applied in the plasma edge question in burning plasmas, since fusion power increases as
region [19]. In kinetic approaches one writes the distribution the square of the density. Because of this, attention must be
function f = f0 + δf and solves the dynamical equation for paid to the existence and nature of physical processes leading
the turbulent fluctuation δf that responds to the perturbed to density peaking. Recently the theory of turbulent pinches
electromagnetic fields. A radial transport equation can be has made significant progress [23]. Based on equation (2),
obtained for the response of the flux surface averaged f0 to the additive term that is proportional to the logarithmic
sources and the turbulent radial fluxes. It is also possible to temperature gradient associated with thermo-diffusion, D12 ,
construct improved quasi-linear transport models by accurately contributes to the turbulent particle pinch. In addition, the
2
determining the linear characteristics, e.g. γL and k⊥ from thermodynamic forces resulting from the coupling of the
kinetic stability calculations, and benchmarking them against gradients of density and temperature and the magnetic field
computational turbulence models. A leading example of this geometry also play an essential role in the particle pinch. TEM
approach is the GLF23 transport model [20]. Alternatively, turbulence is one of the main mechanisms of particle transport.
one can obtain parametric fits to the outputs of the turbulence It has been shown that the magnetic shear can influence the
calculations; here a seminal example was the IFS/PPPL pinch velocity and that collisionality decreases the density
model [21]. peaking factor. At high collisionality, the pinch velocity is
close to the Ware value, whereas it is larger at low collisionality.
2.1.2. Progress in turbulent transport theory. It also suggests that density profiles in ITER may be more
peaked than planned. However, it is not clear whether the
Micro-instability and turbulent transport. Drift wave passing electrons take part in the particle pinch, which calls
turbulence in tokamak plasmas is a highly complex non-linear for further theory and simulation studies. See section 3.4 for
system involving multiple-scale


turbulence modes, non-linear

further discussion of pinches and density peaking.
self-regulation of sheared E × B or zonal flow and various
stabilizing mechanisms influenced by the magnetic geometry Dynamics of sheared flow in ion-scale turbulence. Except for
and plasma compositions, as schematically shown in an outline the dependence of plasma turbulence on complicated magnetic
summary in figure 2 [22]. The typical turbulent fluctuation geometry and device parameters, remarkable progress has been
in the ion regime is the so-called ITG mode, which mainly made in understanding



the suppression of turbulent transport
responds to ion heat transport. Correspondingly, the short by sheared E × B flows, which is thought to contribute to the

S23
E.J. Doyle et al

formation of internal (or edge) transport

barriers


in tokamak
plasmas [24, 25]. As to the origin of E × B flows, besides
externally and/or neoclassically driven macro-scale flows,
the importance of self-generated zonal flows (with poloidal
wavenumber kpoloidal ∼ = 0), which are convective cell modes
non-linearly generated in turbulent plasmas that self-regulate
the transport level, has been widely recognized. The model
of


turbulence


decorrelation and stabilization by such sheared
E × B and zonal flows [26, 27] has been extensively verified
by large-scale turbulence simulations (see section 2.2) and has
also been tested by different experiments (see section 2.3).
Large-scale ITG turbulence simulations based on gyro-
kinetic or fluid (gyro-fluid) models not only predict the
generation of zonal flows but also show the reduction
Figure 3. Ion thermal conductivity can change scaling as the
of turbulent ion heat transport by such zonal flows (see tokamak minor radius increases with respect to the gyro-radius.
section 2.2). The reduction of transport depends on the Reprinted with permission from [33]. © 2002 American Physical
suppression of the fluctuation amplitude, the dephasing (cross- Society.
phase) of fluctuations or their synergetic changes. The
physical mechanism for the dephasing of fluctuations is are strongly unstable. The ETG mode is another plausible
the randomization of the coherent structure, which is a candidate responsible for the electron thermal transport.
complement to the flow shearing decorrelation of turbulence Besides zonal flows, another convective cell mode structure,
[27]. The transport reduction may be understood as a decrease referred to as a streamer (radial wavenumber kradial ∼= 0), can
in the correlation length of the turbulence in the model be generated in ETG turbulence. Such streamers give rise
expressed by equation (8). to transport in excess of simple mixing length estimates for
ETG modes. Furthermore, the generation of zonal flows in
Transport scaling in tokamak plasmas. Based on the ETG turbulence may be less important than in ITG turbulence,
expression for the anomalous transport coefficient in again leading to larger transport. Gyro-kinetic and gyro-fluid
equation (8), transport scaling is determined by the simulations are being used to identify conditions under which
characteristic correlation length and time. When turbulent streamers may be a relevant factor in electron thermal transport
fluctuations are caused by low frequency drift waves with (see section 2.2 for further discussion). Large transport may
time scale τC−1 ∼ ω∗i,e and spatial scale LC ∼ ρi,e (ion or tend to drive the temperature profile to near marginal stability.
electron gyro-radius), the transport coefficient is described Electron transport in recent experiments seems to exhibit a
by the so-called gyro-Bohm scaling, D⊥ ∼ (ρi,e /LT )T /eB, threshold depending on the electron temperature gradient,
where LT is the temperature gradient scale length. This is the which may explain the so-called profile stiffness or profile
expression expected from local drift wave turbulence theory resilience (see section 3.3). Further, the stabilization of
[28–30]. If the turbulent fluctuations are characterized by the ETG mode [37] and the TEM mode [38] is sometimes
the macroscopic size of the plasma, the transport has the consistent with the conditions required for the formation of
conventional Bohm scaling, D⊥ ∼ T /eB. In addition, the internal transport barriers.
drift wave in toroidal geometry may form a radially extended
non-local structure whose spatial length is approximately given Intermittency in turbulent transport. Although the model
by the geometrical mean between the ion Larmor radius, ρi , given by equation (8) is simple and useful in dimensional
and the equilibrium
 scale length, LT , or the plasma size, a, arguments, it does not describe the underlying physics of the
i.e. LC ∼ ρi LT /ŝ (here ŝ is the magnetic shear). A Bohm- structure formation or bifurcations such as the L–H transition.
like (or large-scale non-local) scaling may occur in plasmas This is because plasma turbulence can be highly intermittent,
with peaked temperature [31, 32]. Recent ITG simulations and the turbulent transport is then characterized by bursty
show a transition of turbulent ion transport scaling from the behaviour. Theoretical descriptions of turbulence should then
Bohm-like dependence for a small system size to the gyro- be based on a statistical approach rather than on a deterministic
Bohm scaling in larger plasma systems with a/ρi > 300 [33], point of view [39,40]. Intermittent transport often results from
as shown in figure 3. This trend has significant implications, the dynamics associated with the self-organized-criticality [41]
since an accurate description of size scaling of transport is involving large-scale transport events such as avalanches and
critical for the design of fusion reactors. The appearance of large-scale coherent structures such as zonal flows [42–44],
Bohm-like transport in an essentially gyro-Bohm model can streamers [45–47], low-frequency Kelvin–Helmholtz (KH)
result from the stabilizing effect of rotation shear arising from [48–50], generalized KH modes [51, 52], jets [53] or blobs
diamagnetic fluxes, represented by χ ∼ χgB (1 − ρ∗ /ρ∗crit ) in the tokamak scrape-off layer (SOL) [54, 55]. Research on
[34], or from turbulence spreading in which there is radial the rare, large transport events that accompany the coherent
propagation of turbulent fluctuations from unstable to stable structures has proceeded at several levels: (1) generation
regions of plasma [35, 36]. of large-scale coherent structures in turbulent plasmas, (2)
interaction among different scale fluctuations including the
Electron thermal transport. Trapped electrons can produce coherent structures and (3) statistical approaches to turbulence
high anomalous heat flux when either TEM or ITG modes theory.

S24
 Chapter 2: Plasma confinement and transport

Dynamics of zonal flows and multiple-scale interaction.

Typical large-scale coherent structures in tokamak plasma
turbulence are toroidally and poloidally symmetric zonal
flows/fields and radially elongated streamers. The former
can suppress the turbulent fluctuations and transport through
the flow shear decorrelation of turbulence, while the
latter may enhance plasma transport by increasing the
radial correlation length. These structures are generally
excited by a modulational instability in ITG or ETG
turbulence [45, 46, 50, 56–66]. The underlying mechanism can
be basically analysed using a simplified drift wave turbulence
model, namely, the Hasegawa–Mima turbulence model. One
important effect of zonal flows on drift wave turbulence is the
shearing of turbulent eddies. Noting the complex temporal
behaviour of zonal flows, it has been shown that fast time
varying components of zonal flows are less effective in shearing
turbulence eddies [67] than the mean E × B shear. This
is the reason why the geodesic acoustic mode (GAM) [68]
with ωGAM ∼ = v Ti /R does not reduce the ambient turbulence
significantly for typical core parameters [69]. At the edge, Figure 4. Typical spatio-temporal scales of micro-instabilities that
cause anomalous transport: (D)TIM ((dissipative) trapped ion
sharp pressure gradients make the diamagnetic drift frequency mode); (C)TIM ((collisionless) trapped ion mode); ITG (ηi ) (ion
at the relevant long wavelength closer to the GAM frequency. temperature gradient mode); (D)TEM ((dissipative) trapped electron
Therefore, the GAM could possibly affect the edge ambient mode); CDBM (current diffusive ballooning mode); EM-δp
turbulence [70, 71]. Since the GAM can take energy out of (electro-magnetic skin depth mode); (C)TEM ((collisionless)
the lower frequency zonal flow component, due to geodesic trapped electron mode); ES/EM ETG (ηe )
(electrostatic/electromagnetic electron temperature gradient mode).
coupling [71], GAMs can either enhance the shearing or reduce Reprinted with permission from [747].
the shearing overall, depending on parameters. While the zonal
flows are extensively studied especially in ITG turbulence, an electromagnetic two-fluid code, so-called CUTIE, has
the dynamics of streamers, particularly in ETG turbulence, is described quantitatively some processes occurring in some
still under investigation. Further, low-frequency fluid-like KH tokamak plasmas, which may involve the activity of meso-
modes may be excited as one of the saturation mechanisms of scale fluctuations [77–79].
zonal flows or streamers [45, 48, 49]. In addition, a poloidally
non-symmetrical flow, the so-called generalized KH mode, Statistical description of turbulent transport. As a result
in contrast to the zonal flow, may be non-linearly generated of the progress made in the theory and the simulation of
through the modulational instability of drift wave turbulence turbulent transport, a statistical approach describing the role
[52]. It may enhance the anomalous transport. of non-linearity in plasma turbulence has been advanced.
The complexity of plasma turbulence results from the The importance of developing a probabilistic theory of
properties of fluctuations, which are characterized by multiple plasma transport, which focuses on calculating the probability
spatio-temporal scales, as summarized schematically in distribution function (PDF) of the flux, rather than anomalous
figure 4. The interactions among these different spatio- transport coefficients, has been considered in [52, 80]. The
temporal scales has been emphasized in recent years [3]. These statistical nature of the turbulent fluctuations and their
processes are studied by including the interaction between the influence on transport are determined by the non-linear terms
turbulent fluctuations with separate scales [40,72,73], between in the system of equations. These non-linear terms can be
turbulent ITG/ETG fluctuations and flows with anisotropic expressed as a sum of the drag, the drive and the noise with
spectra such as zonal flows/streamers and also among the mean different scales in a renormalized Langevin equation [81]. The
shear flows, zonal flows and generalized KH modes in drift non-Gaussian PDF of the flux includes tails, which have a
wave turbulence [52]. From the theoretical point of view, a contribution from rare large events, such as the formation of
new direction in studying the turbulence interaction and its some coherent structures. These large-scale structures could
role in transport aims to establish a theoretical framework that play a dominant role in determining the averaged transport. A
treats all fluctuations as an interacting dynamical system. On calculation of the tail of the PDF of heat flux H due to toroidal
the other hand, a simplified minimal model that describes ITG modes, based on a non-perturbative method (i.e. instanton
the coupled drift wave turbulence and zonal flows and/or method), is carried out by a given white noise and the tail is
streamers has been proposed [45]. This model has the estimated as P (H ) ∼ exp(−cH 3/2 ) (here c is a constant) [52].
characteristic form of a ‘predator–prey’ system in which the The stretched exponential PDF tail implies that a coherent
population of the drift wave quanta (prey) grows via linear structure enhances heat transport over a Gaussian prediction.
ITG/ETG instability, generates zonal flows or streamers (and Non-linearity in the strong noise leads to more prominent
or generalized KH modes) (predator) via the modulational non-Gaussian tails. A statistical theory of strong turbulence
instability. Meanwhile, effort is being made to numerically in inhomogeneous plasmas has been developed for the cases
simulate the direct or indirect interactions among the multiple- where fluctuations with different scale lengths co-exist [81]. In
scale fluctuations and flows [44, 73–76]. More practically, a dynamical interacting system, there may exist several states

S25
E.J. Doyle et al

equilibrium magnetic field with plasma flows is established.

Non-linear loop (2) Fluctuation and self-organization dynamics: under the
new equilibrium, various fluctuations from macro-scale MHD
modes to micro/meso-scale electrostatic and electromagnetic
modes (as shown in figure 4) are excited. These modes
non-linearly interact in a wide frequency and wavenumber
space, leading to self-organization in the fluctuation spectrum,
including non-linearly generated global structures. These
fluctuations change the plasma profile in various ways, such as
diffusive transport, non-diffusive or intermittent transport and
bursting blobs. It is also pointed out that zonal fields, which
are the magnetic field version of zonal flows, may change
the magnetic structures, specifically in high beta plasmas.
(3) Global linkage as a non-linear complex loop system:
each physical process with its own spatio-temporal scale is
hierarchically linked to others through the evolving plasma
profile and establishes a closed non-linear loop in a self-
consistent manner as shown in figure 5. The closed loop
is coupled to the external heating system, which is used
Figure 5. A schematic flowchart of multi-hierarchical high to control the complex loop dynamics. Once a transport
performance tokamak plasmas shows the interactions among barrier is triggered, the loop reveals a strongly autonomous
different time and spatial scale fluctuations that play an essential nature, so that external control becomes less effective. It is
role in structure formation.
understood that steady-state transport barrier formation may
correspond to one of the fixed solutions of the non-linear loop
of fluctuations due to the complex interactions among different equations. In burning plasmas, the non-linear loop may be
scales. The suppressing (quenching) or enhancing (exciting) internally sustained via nuclear fusion, so that the level of
role of one mode on another mode in one state is determined
the autonomous nature and/or complexity of the loop may
by the relative magnitudes of the driving powers of different
be increased, depending on the operating Q (figure of merit)
scale fluctuations by the global inhomogeneity. This statistical
value.
theory has also been applied to analysing the bifurcation of
Building on the knowledge of the more elementary
the radial electric field so as to describe the L–H transition in
processes, i.e. (1) and (2), that have been significantly
toroidal plasmas with multiple-scale fluctuations. The PDF of
advanced so far, fusion theory will need to step into the new
the radial electric field shows power-law tails, with the power
regime, described by (3), in order to understand the complex
index depending on the global inhomogeneity as well as the
autonomous dynamics of the non-linear loop and to develop
size of volume average. As the global gradient becomes larger,
control methods for burning plasmas.
the tail becomes more prominent.

Concluding remarks. Stimulated by the growing interest in 2.2. Turbulence simulation

ITER burning plasma physics, the theory of plasma turbulence
and transport has been advanced with the aid of significantly It should be noted that the assumption of adiabatic
improved computer simulations. As we discussed in this electron dynamics in the description of ITG-driven transport
section, high performance plasmas with high confinement and implies zero net particle transport, which is an unphysical
of a steady-state nature can be realized by having various simplification. Modern gyro-kinetic codes (GYRO, GS2, for
prominent structures in the plasma. Therefore, understanding example) are capable of treating ITG instabilities including
the underlying physical mechanisms and the selection rules for complete kinetic electron dynamics, while the best theory-
such structure formation and developing corresponding control based drift wave models (GLF23 and MM95) also include
methods are central subjects that will lead the fusion program kinetic electron physics. For serious physics studies, the
to success. It is now widely recognized that such structure GYRO and GS2 codes routinely include kinetic electron
formation exhibits a complex nature when multiple physical physics. Thus, the adiabatic electron model is largely irrelevant
processes are linked to one another. It is difficult to grasp such for modern gyro-kinetic transport studies, e.g. [82] gives
complex process based on the idea of a conventional one-to- detailed gyro-kinetic particle transport results (including ion,
one correspondence between ‘cause’ and ‘result’. electron and ash transport) with full electron dynamics. In
In order to grasp the global picture of self-organized order to make the distinction between the simplified and
plasmas that exhibit various structure formations, a non-linear the more complete models, we refer to the former as ‘ITG
loop system that shows the mutual linkage between different with adiabatic electrons.’ If the model including kinetic
physical processes in tokamaks is illustrated in figure 5. The electrons has dominant linear instabilities which propagate in
loop may be categorized into the following three primary the ion direction, it is customary to call this ‘ITG-dominated
dynamics. (1) Neoclassical dynamics: this leads to the self- transport,’ whereas if the linear modes propagate in the electron
generated bootstrap current and electric field (and thus to direction, one generally refers to them as ‘trapped electron
plasma flow and rotation) from the pressure gradient. A new mode (TEM) instabilties.’

S26
 Chapter 2: Plasma confinement and transport

Table 1. Overview of gyro-kinetic, toroidal turbulence codes presently in use.

Code Main authors Method Simulation domain Electron dynamics
SUMMIT Parker et al [131] PIC Flux tube Kinetic, electromagnetic
GTC Lin [111] PIC Full torus Adiabatic, electrostatic
GT3D Idomura et al [92] PIC Full torus Adiabatic, electrostatic
ORB Villard et al [93] PIC Full torus Adiabatic, electrostatic
GENE Jenko and co-workers [122, 135] Vlasov Flux tube Kinetic, electromagnetic
GS2 Dorland et al [46] Vlasov Flux tube Kinetic, electromagnetic
GYRO Waltz and co-workers [109, 748] Vlasov Full torus Kinetic, electromagnetic

2.2.1. Models and codes. In the case of ordinary fluids, direct for reliable quantitive predictions, a gyro-kinetic treatment
numerical simulation of turbulent flows is based on the Navier– cannot be avoided [86–88]. Nevertheless, gyro-fluid codes
Stokes equation. Here, most attention is usually paid to physics contributed significantly to the progress in plasma micro-
issues concerning the inertial range of spatial scales in which— turbulence research in the 1990s, and they might continue
due to the absence of energy sources and sinks—kinetic energy to be helpful tools if used synergistically with gyro-kinetic
is merely passed on from one set of scales to the next. codes. The most advanced non-linear gyro-fluid codes that are
Unfortunately, the total computational effort that is required presently in use are described in [69, 89–91]. For the rest of
to address these issues scales approximately as the third power this brief overview, we will focus mainly on gyro-kinetic and
of the Reynolds number. Therefore, despite continuous growth gyro-fluid simulation of tokamak core plasmas.
in computer power, studies of high-Reynolds-number fluid
turbulence will have to be based on approximation schemes Computational gyro-kinetics. The employed algorithms for
in the foreseeable future. solving the gyro-kinetic Vlasov–Maxwell equations can be
In contrast to this, direct numerical simulation of plasma roughly divided into two groups according to the Lagrangian
micro-turbulence is gradually reaching a level of maturity and the Eulerian description of phase space dynamics (besides
that allows for quantitative comparisons with experiments. this, there also exist so-called semi-Lagrangian methods which
This might, at first, sound amazing in light of the enormous will not be discussed here).
complexity of the problem (multiple drive and saturation
mechanisms, kinetic effects and cross-scale coupling, to name • Particle-in-cell (PIC) codes follow the trajectories of
just a few). However, it is essential to note that in the case of an ensemble of N marker particles, at the same time
magnetized fusion plasmas, the focus tends to be not on inertial using a real space grid to compute the interaction forces
range physics but on the turbulent transport. The latter is often effectively. While they are relatively easy to implement
dominated by a rather limited number of degrees of freedom and parallelize, one has to deal with the build-up of
that are embedded in a turbulent bath. Although resolution numerical noise which can affect long-time simulations
requirements may vary a lot with plasma parameters, one can as well as the dynamics of zonal flows or kinetic shear
generally say that quite comprehensive numerical simulations Alfvén waves [92–98]. This problem has been a subject of
of plasma micro-turbulence under realistic conditions are increasing interest in recent years, and significant progress
becoming feasible with present-day supercomputers. Despite has been achieved (see below).
this good general prospect, it turns out that much effort needs • A complementary approach is to represent the gyro-centre
to be put into the careful development and testing of turbulence distribution functions on a fixed grid in a five-dimensional
models and codes. phase space. The non-linear gyro-kinetic equations are
then finite-differenced and solved according to techniques
Kinetic and fluid models. It is widely agreed upon that first- borrowed from computational fluid dynamics. Such
principles simulations of low-frequency (with respect to the ion schemes avoid noise problems and their consequences, but
cyclotron frequency) micro-turbulence in weakly collisional they tend to be more difficult to implement and somewhat
plasmas should be based on the gyro-phase-averaged Vlasov– less efficient. Moreover, one has to carefully remove grid-
Maxwell equations, which were first derived in the 1980s scale fluctuations in phase space, an issue well known from
[18]. Here, the challenge is to compute the time history fluid turbulence studies.
of each particle species’ distribution function in a five-
dimensional phase space (the sixth dimension corresponds to At present, a number of non-linear gyro-kinetic codes for
the gyro-phase and is separated from the other five) along studying tokamak micro-turbulence are in use, with a few more
with that of the perturbed electromagnetic fields in real under construction. An overview is given in table 1.
space. Noting that of the order of 100 (or more) velocity
space quantities per grid point/cell are typically involved Noise reduction in gyro-kinetic PIC codes. As mentioned
in gyro-kinetic computations, it has also been attempted to before, PIC simulations are based on sampling the phase space
construct numerically less expensive models on the basis dynamics via superparticles. It can be shown [99,100] that the
of non-standard fluid equations that capture all the relevant statistical error of this method is given by σ/N 1/2 where σ is the
kinetic effects [83–85]. These ‘gyro-fluid’ models have been variance of the estimator and N is the number of particles used
relatively successful in describing many qualitative features of in the computation. In principle, convergence can thus always
plasma micro-turbulence and the associated transport. Over be achieved by increasing N . But, in practice, this brute-force
the last couple of years it has become clear, however, that approach is often numerically prohibitive. One therefore aims

S27
E.J. Doyle et al

at reducing σ . Over the last decade or so, much progress has

been made in this area.
The ‘δf ’ method [101–104] utilizes a split of the total
distribution function f into an equilibrium part f0 and a
fluctuating part f˜ = f − f0 . Only the latter is represented
by particles, while the former is treated analytically. Thus,
a weight is assigned to each particle accounting for the fact
that (in contrast to f ) f˜ is not constant along phase space
trajectories. However, as has been demonstrated recently,
collisionless systems can also do without weights if the exact
equations of motion are solved [99]. In so-called ‘split-
weight’ schemes [105], only the non-adiabatic part of f , i.e.
h̃ = f − f0 (1 − qφ/T ), is kept. A modification of this
method has been applied successfully in the electromagnetic
regime [106].
More recently, an optimized particle loading technique Figure 6. Ion heat diffusivity as a function of the normalized ion
has been proposed to further minimize the variance σ [107]. temperature gradient for various gyro-kinetic and gyro-fluid codes:
Here, the idea is to increase the particle density in regions the critical gradient exhibits a non-linear upshift. Reprinted with
of phase space where |δf | is large. Through this adaptation, permission from [110]. © 2000 American Institute of Physics.
optimal sampling is achieved. The numerical noise is greatly
reduced, along with the degree to which conservation laws by self-generated zonal E× B flows. In some regimes, this
are violated. Using optimized loading, gyro-kinetic PIC may even result in turbulence suppression, associated with
simulations could be extended well into the non-linear regime an effective upshift of (R/LT )crit [110]. This is shown
for the first time [93]. Nevertheless, dealing with particle noise in figure 6. On the other hand, linearly damped modes
may remain a challenge, as has been reemphasized recently in may be non-linearly excited in the course of ‘turbulence
the context of electron-scale turbulence [108]. spreading’ (see [33] and references therein). For reasons
like these, (quasi-) linear theory—although tremendously
2.2.2. Some main physics issues. useful for transport modelling—cannot replace massively
parallel non-linear computations.
Basic ITG turbulence. Over the past decade or so, much
• Although scalings of the ITG mode induced χi with
emphasis has been put on the study of fluctuations driven by
collisionality [42, 117] and system size [32–34] may be
ion temperature gradient (ITG) modes. A basic model for these
found in the literature, these results must be viewed
micro-instabilities may be obtained by taking the electrons to
as preliminary. More comprehensive and realistic
be adiabatic (i.e. they are in the Boltzmann equilibrium at any
simulations, involving non-adiabatic electrons or core-to-
time). This assumption removes complications associated with
edge coupling, are expected to give different answers.
the time scale separation between electron and ion dynamics
and with the co-existence of multiple micro-instabilities. It • Beyond allowing us to gain basic insights into the
allows one to study a simplified but experimentally relevant behaviour of turbulent plasmas in toroidal magnetic
system with reasonable computational effort. Several key devices, the adiabatic ITG model has also served as an
features of basic ITG turbulence have been identified and environment for inter-code comparisons, involving both
investigated. gyro-kinetic and gyro-fluid codes. The classic test case
has become a set of parameters—called the cyclone base
• ITG modes are linearly destabilized if and only if the case—inspired by a particular DIII-D discharge. Many of
inverse scale length of the ion temperature gradient the gyro-kinetic codes listed in table 1 have been able to
exceeds a certain threshold, i.e. for R/LTi ≡ |R∇Ti |/Ti > reach convergence in this test case [110, 118, 119]. In this
(R/LTi )crit where R is the major radius of the toroidal context, see figure 6.
device. Increasing R/LTi beyond its critical value (which
depends on the local plasma parameters), the induced ion
thermal transport rapidly rises to values of the order of Ion-scale turbulence involving non-adiabatic electrons. The
χi ∼ ρi2 vti /LTi and then stays at that level [109, 110]. adiabatic electron approximation automatically implies that
Here, ρi is the ion gyro-radius and vti is the ion thermal both the particle and the electron thermal transport vanish
velocity. From this one can predict that, under typical identically. Furthermore, finite β effects cannot be
plasma conditions, moderate central ion heating is not incorporated since they are caused by parallel currents carried
expected to push the Ti profile much beyond its critical mainly by passing (i.e. non-adiabatic) electrons. Finite β
gradient. This kind of ‘stiffness’ in the ion temperature effects (β denotes the ratio of the plasma pressure and the
profiles has indeed been confirmed experimentally in magnetic pressure) result from the inclusion of magnetic field
various discharges and devices. fluctuations that can affect both the driving micro-instabilities
• As has been shown by many investigators (see, e.g. and the non-linear saturation mechanisms; moreover, they
[21, 66, 111–116]), the dominant non-linear saturation allow for magnetic transport. Therefore, it is clear that this
mechanism for ITG modes—especially near the linear severe restriction must be transcended in order to proceed
threshold—turns out to be the shearing of turbulent eddies towards more realistic turbulence models. Fluid codes have

S28
 Chapter 2: Plasma confinement and transport

pioneered this area of plasma micro-turbulence simulation, but • Electromagnetic effects. Gyro-fluid models have also
gyro-kinetic computations are also becoming available. been extended to include passing electrons in a small
• Trapped electron effects. Gyro-fluid turbulence codes mass ratio expansion (applicable to the plasma core,
involving the dynamics of bounce-averaged trapped but not to the edge) [89, 130]. Here, trapped electrons
electrons were developed in the mid-1990’s [120]. were neglected. It was found that far below the ideal
Here, passing electrons and finiteβ effects were neglected. MHD ballooning limit, the scaling of ITG turbulence
Trapped electrons can contribute to the drive of ITG with β basically follows the (quasi-)linear expectations,
modes, but they may also be the source of trapped electron which in this case predict a substantial drop in χi .
modes (TEMs). An attempt to predict thermal and particle However, as the critical β is reached from below,
diffusivities based on non-linear gyro-fluid computations the (electrostatic) transport rises sharply. A likely
in flux-tube geometry can be found in [121]. On the explanation of this finding is that at the turning point,
other hand, a systematic study of collisionless TEM the dominant drive of the turbulence transitions from
turbulence in the framework of non-linear gyro-kinetics ITG modes to kinetic ballooning modes. Recent gyro-
is presented in [119, 122]. Therein, it is shown that kinetic simulations have confirmed that for ITG/TEM
(1) even in the absence of a linear threshold, the non- systems well below the ideal MHD ballooning threshold,
linear system typically exhibits an effective threshold in the transport tends to be mainly electrostatic, following the
the normalized electron temperature gradient, (2) zonal (quasi-)linear expectations [131,132]. However, there are
flows tend to be relatively weak away from the effective also indications that the electron heat flux might obtain a
threshold and (3) a modified quasi-linear model is able significant magnetic component as the kinetic ballooning
to capture many features and parameter dependences regime is entered [132,133]. Similar findings are expected
of the TEM induced transport reasonably well. On for cases in which the dominant micro-instability becomes
the other hand, it was observed in the case of purely a micro-tearing mode. This situation might occur, for
density-gradient driven TEM turbulence that there exists example, in spherical tokamaks such as MAST [134].
a non-linear upshift of the critical density gradient that
is analogous to the Dimits shift in ITG turbulence with ETG turbulence and cross-scale coupling. Turbulence
adiabatic electrons [123]. These results indicate that non- suppression in ITBs is generally attributed to the effect of
linear saturation of TEMs seems to be caused by zonal equilibrium scale E × B shear flows, possibly aided by other
flows if and only if one approaches the linear stability mechanisms (such as low magnetic shear ŝ
0 or high
thresholds. Further investigations of TEM saturation are normalized pressure gradient α > 0) which tend to weaken
thus called for. the drive of toroidal ITG modes and trapped electron modes.
• Particle pinch effects. A full torus fluid code involving It is not clear, however, why the electron thermal transport is
both ions and trapped electrons has been used to study often not changed much while the ion channel can be reduced
various aspects of turbulent transport [124]. It is found, down to its neoclassical level. One plausible explanation of this
e.g. that an anomalous particle pinch exists, driven by both experimental finding is the existence of turbulent fluctuations
a curvature effect and thermo-diffusion. The curvature at space-time scales smaller than those of ITG turbulence. This
contribution is exactly the same as the one predicted by is one of the reasons why ETG turbulence has been the centre
‘turbulence equipartition’ (TEP) theory [125, 126]. The of attention in recent years. Some of the key results in this area
amplitude of thermo-diffusion depends on the average of research are the following.
phase velocity of fluctuations. In the regime dominated
• Toroidal ETG turbulence may exhibit streamers (radially
by ITG modes, it is directed inwards. When moving
elongated vortices illustrated in figure 7) that have
to high values of Te /Ti , the trapped electron mode
been identified as residuals of linear modes [46, 135].
branch becomes dominant and the thermo-diffusion term
Due to their large saturation amplitude, the associated
ultimately reverses its sign. Moreover, it has been
demonstrated that the turbulence can generate significant electron heat transport can exceed the small mixing length
amounts of flows in the parallel direction when the E × B expectations by more than an order of magnitude, reaching
shearing rate becomes large (as in ITBs) [127]. Beyond experimentally relevant levels. The observed boost factor
these fluid results, recent gyro-kinetic work revealed primarily depends on R/LTe − (R/LTe )crit and magnetic
that particle transport tends to be described surprisingly shear [47, 135]. In contrast to ITG turbulence, zonal
well by quasi-linear theory [82, 119]. This is largely flows are found to be rather weak in toroidal simulations.
due to the fact that the non-linear cross-phases tend to (It is interesting to note that streamers have also been
resemble the linear ones. As one surprising consequence, observed in fluid simulations of resistive ballooning
one finds that an anomalous particle pinch can also be turbulence [136] and in gyro-kinetic simulations of TEM
carried by passing (not trapped) electrons [119, 128]. turbulence [122]).
This kinetic effect had been overlooked in previous fluid • This surprising result may be understood in the following
studies. It must be stressed, however, that despite way [47]. Depending on the relative importance of
these interesting developments the direct applicability of compression in the parallel and perpendicular dynamics,
turbulent pinch results to experiment is hampered by the one of two Kelvin–Helmholtz-like secondary instabilities
fact that gyro-kinetic simulations generally predict that stop the exponential growth of linear ETG modes.
the pinch disappears for realistic values of collisionality Balancing the growth rates of primary and secondary
[82, 129]. This discrepancy needs to be addressed in modes (the latter is proportional to the amplitude of the
future work. former), one can estimate the saturation amplitude and the

S29
E.J. Doyle et al

Figure 7. Characteristic contour lines of the electrostatic potential in the outboard x–y plane for toroidal ETG turbulence: formation of
streamers (radially elongated structures). Reprinted with permission from [135]. © 2000 American Institute of Physics.

induced transport level. The observed agreement between

this model and gyro-kinetic simulations is quite good [47].
• Recently, the above ETG work has been extended to
include trapped electron effects. Based on results from L-mode
non-local PIC simulations, Lin et al [137] argued that
in this case, the ETG instability saturates via non-linear
toroidal coupling. In such non-linear interactions, the
wave energy at the unstable high toroidal mode number
H-mode
domain cascades towards the more stable lower toroidal
mode number domain via scatterings off the driven low
mode number quasi-modes. The transport level still
exceeded the corresponding ITG level, but not by as
much as in the simulations without trapping by Jenko
and Dorland [47]. However, Nevins and co-workers
showed that the transport at late times is very likely to
be determined by numerical noise [108].
• To address the issues of ITB physics, ETG turbulence has
also been studied by means of radially non-local gyro- Figure 8. Gyro-kinetic simulations of tokamak edge turbulence
kinetic simulations [138]. In this system, streamers are without (‘L-mode’) and with (‘H-mode’) imposed E × B shear
absent, and zonal flows—formed around (but not directly flows: the superposition principle for the electron heat diffusivity is
violated, and ETG turbulence provides a floor in H-mode. Reprinted
at) the qmin (minimum safety factor) surface—tend to with permission from [140].
become more prominent as the magnetic shear goes to
zero. Zonal flow saturation is found to be caused by a
geometry, taking flux surface shaping into account. Instead
Kelvin–Helmholtz-like instability.
of getting the required geometrical information from MHD
• Recently, this line of research triggered interest in the
equilibrium codes, one can also use a local solution of the
non-linear coupling of various kinds of plasma micro-
Grad–Shafranov equation, which generalizes the well-known
turbulence. In [48], it was shown that linear ITG
ŝ − α model to finite aspect ratio, elongation and triangularity
modes can be suppressed by strong, ETG-scale zonal
[141]. Both PIC and Vlasov flux tube and full torus gyro-
flows. Using non-linear gyro-fluid simulations, it could be
kinetic codes have been developed and complement one
demonstrated that a complex interplay of ITG turbulence,
another very nicely (see table 1). To uncover systematic errors
ITG-scale zonal flows and (externally imposed) ETG-
and numerical artefacts in models or codes, it will be vital to
scale zonal flows can lead to a general reduction and a
maintain a healthy variety of turbulence codes. This allows
pronounced burstiness of the ion heat transport [44, 139].
for cross-checking, which will help to reach a consensus on
The first self-consistent numerical simulations of cross-
important issues and to build confidence in the results obtained.
scale coupled turbulence have been presented in [140].
There is much literature on comparisons between linear or
Therein it was demonstrated explicitly that, in general,
quasi-linear theory and experiments. This approach, although
transport on ion and electrons scales does not obey the
limited, has proven to be very useful. For example, one can
superposition principle (see figure 8). In particular, it was
shown that the residual level of electron heat transport in determine the expected presence or absence of certain micro-
a transport barrier can be set by ETG turbulence. instabilities, critical gradients, linear growth rates (which can
then be compared with measured E × B shearing rates),
mixing length or quasi-linear estimates of transport coefficients
2.2.3. Comparisons with experiments. Direct numerical and general trends as the plasma parameters are changed.
simulation of plasma micro-turbulence is gradually reaching However, there are many examples of situations in which
a level of maturity that allows quantitative comparisons (quasi-)linear theory is known to be misleading or even wrong.
with experiments. Supercomputers are becoming ever more For example, it is well established that drift waves in a
powerful, and the codes are getting more comprehensive. sheared slab geometry—although linearly stable—are non-
Several codes now include non-adiabatic electrons (both linearly unstable [142]. Moreover, electron thermal transport
trapped and passing) as well as electromagnetic effects. In induced by ETG modes can exceed quasi-linear estimates
addition, many of them may be run in general tokamak by more than an order of magnitude [47, 135]. And ITG

S30
 Chapter 2: Plasma confinement and transport

turbulence is often controlled or even completely suppressed 4

by zonal flows, again an inherently non-linear effect [110,111]. (a) +10%
To enhance our understanding of transport in fusion plasmas
3 simulation
and to establish a more firm predictive capability, we must base

χi /χGB at r/a=0.6
therefore continue to perform and analyse non-linear gyro-
kinetic (and fluid) simulations.
First quantitative comparisons of turbulence simulations 2 -10%
with experimental data have been published in the last few base
-15%
years. Gyro-fluid simulations [121] for plasma parameters +10%
1 -10%
from a particular L-mode discharge on the DIII-D tokamak
yielded ion and electron thermal fluxes that were within a
factor of two of the experimental results. On the other hand, the -15% experiment
0
particle flux was too small, and the density fluctuation level was 2.0 2.5 3.0 3.5
4–5 times too large. Nevertheless, the wavenumber spectra of -dln(Ti)/dr
the density fluctuations showed good qualitative agreement. 2.5
Only ions and trapped electrons were retained in these flux- (b)
tube simulations. Finite β and finite ρ∗ effects were neglected, simulation +10%
2.0
including equilibrium E × B shear, which was taken into

χe /χGB at r/a=0.6
account only as an a posteriori correction. Unfortunately, base
repeating these computations with a more comprehensive 1.5
gyro-kinetic flux-tube code did not reconcile simulation and -10% base
-15% +10%
experiment [121]. The density fluctuation level and heat fluxes 1.0
were still too large by about a factor of 3. -10%
By means of Vlasov simulations including (1) gyro-kinetic
0.5
ions and drift-kinetic electrons (keeping trapped and passing
-15% experiment
populations in both cases), (2) electron pitch-angle scattering,
0.0
(3) finite β effects, (4) shaped equilibria and (5) finite ρ∗ 2.0 2.5 3.0 3.5
effects (in particular, equilibrium E × B shear), the measured -dln(Ti)/dr
thermal diffusivities in a DIII-D discharge could be matched by
changing the nominal value of R/LTi within the experimental Figure 9. Ion and electron heat transport for a particular DIII-D
error bars [32]. The associated particle transport was not discharge: the experimental values can be matched by the simulation
compared. Neither were fluctuation levels or wavenumber if the normalized ion temperature gradient is varied within the error
bars. Reprinted with permission from [32]. © 2003 American
spectra. So although this is a very encouraging result, it will be
Physical Society.
crucial in future work to try to recover simultaneously as many
turbulent features as possible. Such an approach will help to
narrow the window of uncertainty even though the error bars on of internal transport barriers, as well as many issues in the area
individual plasma parameters may be rather large. The long- of edge turbulence (not covered in this review). Some answers
may hold surprises, possibly redirecting the way we think about
term goal must be to reproduce all transport-relevant aspects in
turbulence and transport in fusion plasmas. Moreover, the
a variety of discharges, thus increasing our insight into the basic
complex issue of self-consistency between profile dynamics
turbulent processes and enhancing our capability to predict the
and turbulent transport will have to be tackled on the road to a
performance of future devices such as ITER.
predictive transport capability. This will force us to expand our
view of turbulent dynamics, also taking into account processes
2.2.4. Conclusions. As is shown in this brief survey, non- traditionally treated in neighbouring sub-disciplines of fusion
linear computations of plasma turbulence are becoming more physics.
and more comprehensive and mature, allowing for serious
comparisons with experimental data (see figure 9). In the
2.3. Turbulence measurements in tokamaks and comparison
1990s, the focus of most work in this area was on ITG
with theory/simulation
turbulence with adiabatic electrons, leading to significant
progress in our understanding of the anomalous ion heat flux As often noted, the energy and particle confinement in
observed in the experiments. Starting around the year 2000, tokamak experiments is generally not as good as predicted
the adiabaticity constraint was transcended in various ways, from transport calculations based on collisional processes
allowing for investigations of the origin of electron heat and alone, and the discrepancy is often attributed to transport
particle fluxes, as well as the role of electromagnetic effects. arising from micro-instabilities driven unstable by various
Despite significant progress over the last few years, these topics free energy sources. These micro-instabilities include ion
will continue to challenge both theory and simulations for the temperature gradient driven modes (ITG: k⊥ ρi
0.5),
foreseeable future. There is a host of unsolved questions—both trapped electron modes (TEM: k⊥ ρi ∼ 0.2–1) and electron
fundamental and practical—which still need to be addressed, temperature gradient driven modes (ETG: k⊥ ρi  2)
including the role of electron-scale turbulence and non-local where k⊥ is the fluctuation wavenumber perpendicular to the
phenomena in various regimes, the non-linear coupling of ion- magnetic field and ρi is the ion gyro-radius. In order to
scale turbulence to both larger and smaller scales, the physics better understand the physics associated with these losses,

S31
E.J. Doyle et al

research at various tokamak facilities has been directed at plasma parameter space of multiple turbulence features, that
detailed comparisons between turbulence measurements and provide confidence in the validity of a given theory/simulation.
turbulence theory/simulations. The goals of such comparisons The general comparisons in this section are generally the most
are to improve the understanding of turbulence and transport, advanced. Nevertheless, much work remains before a set
to test and provide feedback to theory and to obtain the of validated turbulence simulation codes is available. For
ability to predict more confidently the operation of ITER. example, note that Ohmic and L-mode plasmas have been the
Such a predictive ability would potentially also allow us focus of many of the comparisons, while our ultimate goal
to extend the operating regimes of present-day machines, is to understand turbulence well enough to exploit enhanced
which would provide a stronger baseline for planning the confinement regimes more effectively.
ITER experimental program. Observations and simulations One of the most basic and straightforward measurements
of low frequency, long wavelength fluctuations are consistent to compare is the magnitude and frequency distribution
with the picture of these instabilities driving ion thermal (i.e. frequency power spectra) of the density fluctuations.
and particle transport as well as some or all of the electron Measurements of the density fluctuation power spectra
thermal transport [28, 143]. However, detailed validation from reflectometry were compared with results from the
of the simulations is still in progress. The physical DALFTI Landau-fluid model on ASDEX Upgrade [144].
complexity of a three-dimensional tokamak plasma, combined The DALFTI model accounts for drift (Alfvén) wave, slab
with many free energy sources and associated possible and toroidal ITG modes, as well as ideal and resistive
instabilities, has made the construction of complete simulation interchange/ballooning modes [145]. The frequency spectra
models difficult—the required computational times alone place of the density fluctuations from several locations in the
restrictions on the work. These difficulties, combined with the region ρ = 0.85–0.95 (where ρ is roughly equivalent to
complexities measuring relevant quantities measurements (e.g. the radial location r/a) were found to compare favourably
we would like to measure turbulence-induced transport but with the DALFTI results. The authors tentatively concluded
often the best we can do is measure the fluctuating density ñ), that the DALFTI code with realistic flux-tube geometry
make code validation a time and resource consuming effort. appeared to reproduce the structure of the L-mode tokamak
In this sub-section the more recent comparisons between turbulence, although the fluctuation level might have been
tokamak experimental measurements and theory/simulation slightly underestimated.
are reviewed. The scope of this review is restricted to direct A low frequency, low wavenumber density turbulence
or nearly direct comparisons. Regrettably, this limitation feature (f < 400 kHz, k ∼ 2 cm−1 ) was found to appear in
eliminates a great deal of work—both experimental and the far infrared scattering signal on DIII-D as the line-averaged
theoretical—that does not include such direct comparisons. density was increased [146]. The appearance of this feature
First are presented comparisons of general turbulence coincided with the saturation of the energy confinement time
behaviour, followed by comparisons of high-k and magnetic and was consistent with linear (GKS) [147] and non-linear
turbulence and finally a section on experimental-theory (UCAN) [116] gyro-kinetic turbulence simulation codes. The
comparisons in improved confinement regimes. Physics GKS linear stability code predicted that at higher line-average
topics include measurement-simulation comparison relevant density the ITG mode would emerge to be dominant over
to low through high wavenumber turbulence (e.g. ion the TEM, which is consistent with both the appearance of
temperature gradient turbulence, trapped electron mode the low frequency, low wavenumber feature and the change
turbulence, etc), impurity enhanced confinement and core in the propagation direction of the density fluctuations from
and edge transport barriers, including both density and the electron diamagnetic to the ion diamagnetic direction
temperature fluctuations and turbulence-induced fluxes. These (as measured in the lab frame). The radial variation and the
fluxes and their dependence upon fluctuation level are magnitude of radial correlation lengths r from reflectometry
generally inferred (except at the far edge where Langmuir were found to be similar to the trends predicted by ITG and
probes can penetrate) from the correlation between power electron drift wave estimates, as well as some meso-scale
balance transport calculations and simultaneous turbulence type correlation lengths [148]. In a study of dimensionally
measurements. Herein it is sought to convey a description similar discharges, the radial variations of the heat diffusivities
of the measured quantities, the corresponding simulations or were found to scale between Bohm and gyro-Bohm consistent
theory and the conclusions from the work. with GLF23 [20] transport simulations [149]. Experimental
measurements of r in an L-mode plasma were compared with
2.3.1. General turbulence features. General features two UCAN simulations of the same plasma—one simulation
of plasma turbulence include fluctuation levels, fluctuation with and a second simulation without zonal flows. The
wavenumber and frequency spectra and correlation lengths— simulated values of r without zonal flows were very long,
quantities that characterize the turbulent state of the spanning much of the 65 cm minor radius, while the inclusion
plasma. Appropriate calculations for direct comparison of zonal flows decreased the simulated r to near the measured
with experimental measurements include the correct spatial, values in both magnitude and radial behaviour.
wavenumber and frequency ranges of the diagnostic. More difficult to compare are the thermal transport
Additionally, similar or equivalent analysis techniques levels. This is due to the fact that (with the exception of
(e.g. Fourier transforms, time averages) must be used. While the far edge plasma which can be accessed by Langmuir
it is possible to obtain reasonable agreement of a single feature probes) the transport must be inferred from a combination of
(e.g. spectral shape or correlation length), it is the total of such measured temperature profiles, plasma species cross-coupling
comparisons, e.g. agreement over a large radial region and calculations and heating deposition estimates rather than from

S32
 Chapter 2: Plasma confinement and transport

(a)

(a)

(b)
(b)
Figure 11. Comparison of power spectra of (a) density fluctuations
Figure 10. Wavenumber spectra from (a) BES measurements and and (b) particle flux from Langmuir probe data and BOUT L-mode
(b) GRYFFIN simulation showing some similarity in shape. simulations showing reasonable agreement between simulation and
Reprinted with permission from [150]. © 2002 American Institute measurement. Reprinted with permission from [152].
of Physics.
Langmuir probes on DIII-D were compared with BOUT [19]
measurements of the turbulence-induced fluxes themselves. simulations. Reasonable agreement in the shape and the
Thus, the comparison is at least one step removed from what spectral roll-off [152] was observed, as seen in figure 11. A
one would ideally prefer. Such comparisons are nevertheless two-scale falloff in the fluctuation spectrum was observed: f −1
important and useful. For example, comparisons between scaling at low frequency and an f −4 scaling at higher frequency
DIII-D experiments and simulations by Ross et al [150] found (200 kHz). The authors found that the spatial profile of the
that ion thermal transport simulated by the GRYFFIN code [21] turbulence-generated diffusivity was in qualitative agreement
was larger than the experimental value by a factor of 1.5– with that required to produce pedestal profiles. Also, the
2.0, and the density fluctuation level was overestimated by a turbulence-generated electric field profiles (as calculated by
factor of ∼4 with (ñ/n)exp ≈ 0.4%, (ñ/n)Gryffin ≈ 1.6–1.9%. BOUT) across the separatrix were qualitatively consistent with
Figure 10 [150] compares wavenumber spectra from the beam those from H-mode experiments.
emission spectroscopy (BES) with the GRYFFIN simulation, The density fluctuation behaviour as measured using
showing similarity in the peak power location and the spread reflectometry and Langmuir probes on T-10 was consistent
in wavenumber. A following study found that the energy with ITG and TEM in the core and with drift-resistive
fluxes obtained in GS2 [46] simulations were comparable to ballooning (DRB) mode instability in the edge [153]. Good
the GRYFFIN calculations and exceed the experiment by a agreement was found in the radial location where the core ITG
factor of ∼2, while the simulated density fluctuation level and TEM growth rates dominate over the DRB as calculated
(ñ/n) was closer to experiment, but still exceeded it by a factor using local turbulence models [154,155] when compared with
of ∼2 [150]. It should be noted that a recent re-evaluation of the spatial location where the experimental turbulence changed
the BES data indicate an experimental (ñ/n) fluctuation level character. Density fluctuations measured by reflectometry
near 1% which brings the GS2 calculations of 1.3–1.5% into were reported to possess long poloidal correlation lengths
closer agreement [151]. which were consistent with zonal flow activity. Finally, long
As mentioned earlier, measurements of turbulence- radial correlation lengths with zero phase shift similar to
induced fluxes are most easily achieved in the far edge plasma. ‘streamer’-like activity, were observed.
These allow a direct comparison between the measured and Comparison of different fluctuating fields (e.g. density and
the simulated flux value. Measurements of L-mode density temperature) can also be used to test and validate theory and
fluctuation spectra and fluctuation induced particle flux from simulation. In this regard, high frequency charge-exchange

S33
E.J. Doyle et al

recombination spectroscopy (HF-CHERS) was utilized to They concluded that electromagnetic drift wave turbulence
measure simultaneously the carbon ion temperature and driven by electron temperature gradient modes provides an
density fluctuations (T̃i and ñC ) on TFTR in order to provide explanation of the measured confinement properties. They
a test of ITG theory [156]. Using BES measurements of the further found that this conclusion is consistent in broad
main ion density fluctuations ñi , it was found that the spectra of terms with the measurement of internal magnetic fluctuations
all three were similar across the radial range r/a ∼ 0.58–0.99. (measured using cross-polarization scattering) and their role
The ratio of relative fluctuation levels (T̃i /Ti )/(ñi /ni ) was ∼2, in electron heat transport [157].
consistent with the general expectations of ITG theory, but it A new mm-wave backscatter technique was employed at
is also consistent with the effects of convective eddies on the DIII-D to probe wavenumbers in the range k⊥ ρi ∼ 2–10,
fluctuations. Additionally, they found the results consistent which is applicable to ETG type modes [161]. The existence
with the ion temperature gradient being the free energy driving of short wavelength density turbulence was confirmed by
source in both the edge and core, while the damping terms utilizing several validity tests based on the physics of the
might be different in the edge and core allowing different scattering process. Additionally, multiple diagnostics were
instabilities to dominate. utilized to obtain density fluctuations measurements over a
broad wavenumber range (k ∼ 1–40 cm−1 or k⊥ ρi ∼ 0.2–10).
2.3.2. High wavenumber and magnetic turbulence. This diagnostic set included an FIR scattering system, the
The higher wavenumber range can be designated as the already mentioned mm-wave backscatter, beam emission
range k⊥ ρi > 1 and is typically thought to contain spectroscopy system, fluctuation and correlation reflectometry
wavenumbers applicable to electron temperature gradient and phase contrast imaging system. The data clearly indicated
driven (ETG) modes, trapped electron (TEM) modes and turbulence activity extending from low to high k with an
high k ion temperature gradient driven (ITG) modes. observed trend of increasing frequency width with increasing
Experimental measurements of high wavenumber turbulence wavenumber. Calculations using the GKS linear stability
and comparisons with theoretical predictions have been code showed that these discharges were unstable to a wide
generally less common than lower wavenumber comparisons. range of instabilities: ETG, ITG and TEM. It was found that
Fortunately, this situation has recently begun to change the plasma was theoretically unstable over a broad range in
with comparisons between measurements and theory in this wavenumber and spatial extent consistent with experimental
wavenumber range becoming available. On the other hand, observations. The predicted frequencies were smaller than
while measurement of internal magnetic fluctuations has those seen experimentally but were in qualitative agreement
undergone increased effort (e.g. see [157, 158]), to date with the measured spectra.
there has been a lack of direct comparison with simulation.
Nevertheless, this is an area of great importance as it directly 2.3.3. Turbulence features in improved confinement regimes.
addresses the knowledge and understanding of transport in Here we consider improved confinement regimes as well as
regimes where transport due to other instabilities is thought to regimes leading to the L–H transition. In a multi-machine
be low. Thus, there is a definite need to expand comparisons  B velocity shear suppression
review paper, multiple tests of E×
in this area. Below are presented comparisons between of turbulence and turbulence-driven transport in both the edge
high k measurements and simulation, with some reference to and core of tokamak plasmas were summarized by Burrell
magnetic turbulence from the Tore Supra results. The work [162]. Externally biased and naturally occurring H-modes,
from DIII-D reviewed below covers a range of wavenumbers, VH-modes, core negative and low magnetic shear plasmas
from high to low, but is included here (rather than in from DIII-D, TEXTOR, CCT, TFTR, JT-60U, JFT-2M, TCV,
section 2.3.1) due to the high wavenumber results. TEXT were presented, but the review focused on tests
Density fluctuation measurements at higher wavenumbers of causality in DIII-D, TEXTOR and TFTR. The author
(8.9 cm−1 ) were performed on TFTR using micro-wave concluded that the model of E × B velocity shear suppression
scattering and compared with local calculations of electron of turbulence-induced transport has the universality needed to
diffusivity and the current diffusive ballooning mode (CDBM) explain the observations for a wide range of plasma conditions
model [159]. The fluctuations were found to propagate in in both the plasma edge and core. Furthermore, measured
the ion diamagnetic drift direction in the plasma reference changes in the E × B shear coincident with measured changes
frame at about 8% of the ion diamagnetic drift frequency. The in turbulence and turbulent transport were consistent with both
fluctuation amplitude (at 8.9 cm−1 ) was seen to correlate well qualitative and quantitative predictions of the model. Based
with local calculations of the electron diffusivity. The authors upon the multiple causality tests covered in the paper, it was
report that in these enhanced reversed shear plasmas (ERS) concluded that there was significant confidence that the basic
the global energy confinement time correlates with fluctuation model is correct.
amplitude, but only for fluctuations with wavenumbers greater Utilizing scanning Langmuir probes in the TEXTOR-94
than 5 cm−1 . The fluctuation frequency, wavelength and tokamak, the variation of the plasma turbulence, the
propagation direction of the 8.9 cm−1 data were found to agree turbulence-induced particle flux and the density-potential
qualitatively with CDBM theory [160]. The authors note that cross-phase with externally induced E × B rotation was
CDBM theory should be extended to the shorter wavelength compared with several theoretical predictions [163]. Scaling of
regime before a quantitative comparison could be made. these quantities with local E × B shear flow was reported to be
A comparison between thermal transport in hot electron similar to theoretical predictions by [164, 165]. This similarity
plasmas (Te > 2Ti ) and theoretical turbulence based transport is shown in figure 12 [163], which shows reasonable agreement
predictions was carried out by Horton et al on Tore Supra [12]. between measured values and the analytic predictions leading

S34
 Chapter 2: Plasma confinement and transport

× × × ×

(a)

Figure 13. Comparison of a GAM E × B oscillation observed in a

BOUT simulation of experimental discharge (109644) to the
turbulence poloidal velocity spectrum derived from BES data.
Reprinted with permission from [174]. © 2003 American Institute
of Physics.

× × × × thermal diffusivities consistent with GKS calculations of

(b) destabilization of both long and short wavelength modes [169].
Impurity gas injection into negative central magnetic shear
Figure 12. Scaling of (a) normalized density fluctuations DIII-D discharges produced a reduction in density fluctuation
comparing the Langmuir probe measurements and predictions of the levels (as measured by BES) as well as reduced transport
Zhang-Mahajan theory [165] leading up to the L–H transition, and
levels [170]. This was in qualitative agreement with both linear
(b) normalized particle flux comparing the Langmuir probe
measurements and predictions of the Ware–Terry theory [164] GKS simulations that showed a reduction of linear growth rates
leading up to L–H transition. Reprinted with permission from [163]. as well as with non-linear UCAN simulation that showed a
reduction of non-linear turbulence levels. GLF23 transport
simulations indicated that while the growth rates of the low
to the L–H transition. Further, the authors found that the
wavenumber fluctuations were reduced with neon injection, it
scaling of the cross-phase term with E × B shear was as strong
was the E × B velocity shear suppression—resulting from
as that found for the scaling of the turbulence, revealing that the
improved angular momentum confinement and consequent
cross-phase is an important component in the shear suppression
higher rotation velocity—that was the dominant factor in
of turbulent transport.
the reduced turbulence and transport levels [171]. This
A study of micro-turbulence growth rates and flow shear suggests that a positive feedback mechanism exists between
induced turbulence suppression during JET high performance the fluctuations and momentum transport: increased rotation
ITB plasmas was reported by Budny et al [166]. Comparison velocity increases the radial electric field and the E × B flow
of linear growth rates from the GS2 code with VE×B shearing shear. Similar results were found when comparing impurity
rates found that the VE×B shearing rate was comparable to the injection on DIII-D and JET [172].
growth rate of the fastest growing toroidal mode number. This Time-delay-estimation techniques applied to two-
occurred where the ITB was formed and, importantly, tracked dimensional BES measurements of density fluctuations
its subsequent outward movement. Measurements of density revealed a coherent oscillation in the poloidal flow field on
fluctuation levels—from a single reflectometry channel on the DIII-D [173, 174] which was consistent with a high frequency
same discharge—showed a reduction in fluctuations outside branch of zonal flows identified as the geodesic acoustic mode
the region where the ITB formed. Additionally, reflectometry (GAM) [70]. Turbulence simulations using the BOUT code
data from a different discharge showed reduction in density [175] show an oscillation at the experimentally observed fre-
fluctuations at wavelengths and frequencies in the ITG range quency consistent with the identification of the mode as a
in the region where the ITB was formed [167]. The authors GAM. Figure 13 shows reasonable agreement between the
of [166] conclude that the reduction of ion transport correlates experimental measurement and the BOUT code simulation.
with increases in the ratio of the VE×B shearing rate relative to The enhanced Dα H-mode (EDA H-mode) on Alcator
the ITG linear growth rate, (γE×B /γITG ). C-Mod was found to be accompanied by a continuous,
High performance ITB plasmas on DIII-D show a decrease high frequency, quasi-coherent fluctuation (QC mode). This
in density turbulence, as measured by BES and FIR scattering, QC mode is believed to limit the impurity build-up during
along with a decrease in thermal diffusivities consistent with H-mode while allowing good energy confinement [176].
the paradigm of E × B flow shear stabilization of turbulence The authors report that measurements of this mode by
[168]. Work on understanding the electron transport channel various diagnostics (including phase contrast imaging) indicate
in ITB plasmas utilizing direct electron heating (either reasonable agreement with BOUT simulations with respect
ECH or FW) resulted in increased ion and electron to the mode structure, the fluctuation amplitude and the

S35
E.J. Doyle et al

frequency. Based upon this they have identified this quasi- this paper. Since the IPB [1] was published, (1) neoclassical
coherent mode as the resistive X-point mode that is observed ion thermal conductivity has been established as a firmer
in the BOUT simulations. Measured poloidal correlation basis for transport when turbulence is suppressed, (2) initial
lengths and poloidal wavenumber spectra (from localized measurements of the bootstrap current in the H-mode pedestal
gas puff imaging techniques) and the time-averaged particle have, somewhat surprisingly, been found to agree with
flux were found to agree within a factor of 2 with results standard neoclassical expressions, (3) initial measurements
from the NLET turbulence simulation code [177]. Similar of the relative toroidal rotation of different ion species
agreement was found by the authors when comparing with the appear to agree with neoclassical theory and (4) neoclassical
results of the BOUT turbulence simulation code. Based upon ion temperature gradient screening of impurities has been
these observations and comparisons, the authors tentatively observed, as discussed further in section 3.4. The issue of
concluded that the turbulence in the edge of Alcator C-mod is whether poloidal rotation is neoclassical is still open. The
due to the resistive ballooning instability. confirmation that neoclassical ion thermal conductivity is
attainable under enhanced confinement conditions could have
2.3.4. Summary and conclusions. A range of experiment to a very positive impact on ITER performance projections.
theory/simulation comparisons has been conducted on various The transport of particles, energy, momentum and currents
tokamaks. Far edge and scrape-off layer plasmas, as well is all treated self-consistently in standard neoclassical theory—
as deep core plasmas, are being examined and compared within the limitations of some ordering assumptions. One of
with theory. Many different theories and simulations have the key assumptions is the ordering of scale lengths: drift orbit
been addressed using a variety of fluctuation diagnostics and (banana) width is assumed small compared with the gradient
techniques. The emerging picture is one of fairly good scale lengths of density, the temperature and the radial electric
agreement between experiment and theory for various specific field and small compared with the proximity to the magnetic
comparison points. These points include the predictions axis or the plasma boundary. These and other assumptions
of ITG/TEM simulation and theory of turbulent correlation (e.g. standard neoclassical theory does not include high plasma
lengths, changes in confinement with impurity injection rotation or non-axisymmetric effects due to the presence of
and gradient scale lengths and relative temperature/density toroidal field ripple or magnetic islands) can be broken; they
fluctuation levels. The correlation (from a number may have important implications for burning plasmas and
of machines) of core turbulence reduction, confinement therefore need to be evaluated. Neoclassical aspects of particle
improvement and changes in the growth and damping rates of transport are described in section 3.4. Here we focus on other
the various instabilities (ITG, TEM, ETG, etc) have improved areas where significant new information on the validity of
confidence in these basic instability models. The identification neoclassical transport has been obtained since the IPB report:
of the unstable mode(s) in the edge of some machines, as well ion thermal transport, bootstrap current, the relative toroidal
as zonal flow activity (GAMs), is encouraging. Finally, the rotation velocity of ion species for constructing the radial
 B velocity shear suppression of turbulence and
evidence of E× electric field and poloidal rotation. We also identify areas
turbulent transport, and its detailed effect upon the fluctuation where theory is addressing limitations of the standard model,
parameters, has provided confidence in the correctness of that but because that literature is fairly extensive, we only cover the
model/paradigm. From the reviewed work it is apparent that general issues here.
these comparisons are only in their early stages. Considerable
work remains before a set of validated turbulence simulation
2.4.1. Ion thermal transport. Ion thermal transport appears
codes suitable for predictive purposes is available.
to be governed by long wavelength turbulence under most
plasma conditions as described in other sections of this
2.4. Neoclassical transport chapter. However, the emphasis in recent years on attaining
The theoretical groundwork for neoclassical transport became plasma conditions with suppressed turbulence has led to some
well established in the 1970s for axisymmetric toroidal conditions where the underlying neoclassical ion thermal
plasmas. Neoclassical theory describes the process by which transport might be tested.
particles, momentum and energy flow within and across the The most notable reductions in long wavelength
magnetic flux surfaces due to drifts and Coulomb collisions turbulence are in transport barrier regions—either at the plasma
between plasma species. Two early reviews of the original edge (ETBs) or internally (ITBs). In both ETBs and narrow
work still serve today as primary references [178, 179]. ITBs, the ion banana width is comparable to the barrier width,
Although the theoretical basis has been long established, many which violates the ordering in standard neoclassical theory.
of the features of neoclassical theory and its limits are still Experimental evidence indicates that ion thermal conductivity
undergoing validation. Advanced diagnostics, the attainment is often comparable to the standard neoclassical value in ITBs
of experimental conditions with low collisionality and reduced [25]. Various theoretical extensions to standard neoclassical
turbulence and the incorporation of more rigorous models for theory have been developed to examine the effects of large
the analysis of experimental data are facilitating this validation. ion orbits in steep gradient regions and the effects of steep
In addition to reviewing the progress in evaluating neoclassical gradients on the orbits themselves. For example, ion orbit
transport processes in the bulk plasma, we also address the squeezing, which is the reduction of the banana width due to
extent to which neoclassical characteristics are observed in a gradient in the radial electric field [180], generally reduces
barriers (i.e. ion heat transport and bootstrap current), although the banana width to less than the gradient scale lengths and
barrier physics is discussed more extensively elsewhere in might explain a moderate reduction in ion thermal conduction

S36
 Chapter 2: Plasma confinement and transport

Figure 14. Single fluid thermal diffusivity in NSTX indicates a Figure 15. Calculation of edge current density from LIBEAM
broad region in the core where the ion thermal conductivity may be pitch-angle profile measurements for shot 115114 during the
less than the standard neoclassical value. The magnetic axis is L-mode (grey) and late ELM-free H-mode (black) phase just before
located at ∼103 cm, and the outer separatrix surface is at ∼148 cm. the collapse of the pedestal pressure, showing a large current peak in
Reprinted with permission from [182]. © 2003 American Institute the pedestal region. Also shown for comparison is the toroidal
of Physics. current density calculated from a bootstrap constrained fit (dashed
curve) for the H-mode phase. The last closed flux surface from EFIT
in narrow ITBs. But incorporating the squeezing effect self- on the LIBEAM trajectory is indicated by the dotted line. Reprinted
consistently with the profile of the radial electric field is only with permission from [184]. © 2004 American Physical Society.
approximate because it requires incorporating non-local effects
in the otherwise local assumptions of neoclassical theory. 2.4.2. Bootstrap current in the edge pedestal and near the
There have been several attempts to develop kinetic theories axis. The bootstrap current near the plasma boundary likely
that relax or reverse the orbit size ordering, but usually in plays a crucial role in edge physics—through ELM dynamics,
the extreme limit such that the gradient is much narrower beta limits and access to second stability. Both dynamic
than the orbit size. It is fair to conclude that neither the [183] and static [184] analyses of the bootstrap current in the
simplified orbit squeezing model nor other models have yet pedestal of DIII-D during a long ELM-free period following
received sufficient experimental validation and acceptance to the L–H transition show agreement with standard neoclassical
be used as reliable predictive models for the ion thermal theory. Results of the static analysis by Thomas et al [184],
transport in narrow ITBs. However, narrow ITBs generally which utilizes the high spatial resolution of the Li ion beam
are not considered attractive for high performance either in diagnostic, are shown in figure 15. It can be seen that the
present plasmas or in future burning plasmas because of bootstrap current generates a local peak in the current density
their limitations on MHD stability. Analysis of ion thermal profile in the ETB region, while it is absent in the L-mode.
transport in the edge pedestal is complicated by additional The work of Wade et al [183] shows that the NCLASS [185]
effects, such as the interaction with neutrals and ELM activity. and Sauter et al [186, 187] models yield similar agreement
Recent work on JT-60U has shown that the heat diffusivity in modelling the dynamic response, which exhibits a back-
between ELMs is reduced to the level of ion neoclassical EMF in response to the sudden turn-on of the bootstrap
transport in the ELMing pedestal region at low ν∗ [181]. current. (Although NCLASS and the Sauter models both
But increased ELM losses at low ν∗ make it difficult to take incorporate full geometry effects, they have different additional
advantage of the high confinement offered by the neoclassical strengths: NCLASS includes a more complete model for
transport. Nonetheless, this provides encouraging support impurities, while the Sauter model has been fitted to kinetic
for the use of standard neoclassical transport as a reasonable calculations that use a more comprehensive model for the
lower bound for the transport between ELMs in predictive viscosity. Differences between them are generally less than
simulations. ∼15%, which is comparable to the fundamental uncertainty
A greater interest lies in broad ITBs or other conditions in the basis theory.) The general agreement with standard
where turbulence is reduced across most of the plasma. Such neoclassical theory indicates that even though the ion orbit size
conditions have been observed in NSTX [182], where the is comparable to the barrier width, the standard neoclassical
low toroidal field and large trapped particle fraction also model produces reasonable bootstrap current results. This
contribute to the expected dominance of neoclassical ion may be due to the fact that the electron banana orbit widths
thermal conductivity. However, as shown in figure 14, still satisfy neoclassical assumptions, while the corresponding
transport analysis of some discharges indicates that the assumption on ion orbits is only marginally broken and both
effective ion thermal conductivity is possibly even lower contribute to the bootstrap current.
than neoclassical over the central half of the plasma. The bootstrap current vanishes at the magnetic axis in
This has also been observed in conventional aspect ratio standard neoclassical theory because the trapped particle
tokamaks over more limited regions. Uncertainties in the fraction vanishes under the usual ordering, but this may also
heating source and temperature measurements as well as be modified by finite orbits [188]. The presence of a finite
modifications to standard neoclassical theory by the relatively bootstrap current and/or stability of a region near the axis
large ion orbits, low aspect ratio and unbalanced neutral beam with a very low current density (‘current hole’), as observed
injection are still being investigated as possible sources of the in some experiments [189, 190], could reduce the need for a
difference. seed current. The interesting phenomena of current holes are

S37
E.J. Doyle et al

500 the neoclassical electric field. The neoclassical model also

C explains the reduction in rotational shear for reversed magnetic
400 He calc fields, which leads to an increase in the L–H transition power
D calc threshold.
300 However, some measurements in the core of low
collisionality plasmas appear to show much stronger impurity
Vtor (km/s)

200 poloidal rotation than can be explained by standard

neoclassical theory [195], but these measurements are still
100
undergoing validation. Charge-exchange recombination
spectroscopy is the primary method used to measure poloidal
0
rotation in the core, and it is unfortunately restricted to species
-100 DIII-D 105882 with Z > 1. This means that the poloidal rotation of main
3.500s hydrogenic ions, which are expected to have much larger
-200 poloidal rotation because of their higher pressure gradients,
0.0 0.2 0.4 0.6 0.8 1.0 cannot be measured directly. Experiments with helium
ρ plasmas may help resolve the poloidal rotation issue for main
Figure 16. Profiles of the calculated helium (red dots) and ions. In parallel with the improvements in diagnostics and
deuterium (green dashes) toroidal rotation velocities compared with analysis, theoretical models for the effects of strong auxiliary
the measured helium (red triangles) and carbon (solid blue) toroidal heating are being developed to evaluate their effect on poloidal
rotation profiles. Errors in the measured C and He velocities are rotation.
smaller than the difference between them. Reprinted with Spikes in the poloidal rotation velocity associated with
permission from [191]. © 2004 American Institute of Physics.
bifurcations (not under quasi-static conditions, e.g. see [196])
are not explained by standard neoclassical theory, which is
discussed further in section 3.7. But here again, self-consistent only applicable for time scales longer than the ion–ion collision
validated models for the current, equilibrium and stability do timescale. The TFTR observations of a spike in the poloidal
not yet exist for such extreme conditions. rotation during the formation of ITBs in TFTR [196] may be an
illustration of ambipolarity breaking by loss of ripple-trapped
fast ions from neutral beam injection followed by relaxation
2.4.3. Relative toroidal rotation of ion species. The relative to a new state as the poloidal rotation is damped to the new
toroidal rotation of ion species has long expected to be state [197], which is analogous to the ion orbit loss model for
governed by the strong parallel friction of neoclassical theory. ETB formation [198].
Recent simultaneous measurements of the toroidal rotation of
two species on DIII-D [191] have confirmed the neoclassical
model, as shown in figure 16. Although these observations 2.4.5. Conclusions. Significant progress has been made in
were made under conditions where the neoclassical impurity validating several aspects of standard neoclassical transport:
poloidal rotation is negligible, they provide a firmer basis for (1) neoclassical ion thermal transport is a reasonable lower
reconstructing the radial electric field and the shear damping limit when turbulence is suppressed, even in barriers, (2) the
from impurity rotation measurements. Further refinements in bootstrap current appears to be reasonably valid in the pedestal
the measurements and their interpretations are still needed, according to initial measurements and analysis, (3) the relative
toroidal rotation of ion species obeys neoclassical theory
particularly in regard to the impact of poloidal rotation, which
and can be used to construct the radial electric field and
is discussed next. The bulk toroidal rotation, however, is
(4) the poloidal rotation in the collisional regime appears to be
not explained by standard neoclassical theory and can be
neoclassical. All these observations strengthen our ability to
significant even in Ohmic and RF discharges as discussed in
model ITER performance. Unresolved issues, all of which are
section 3.5. However, extensions to neoclassical theory that
likely to have only a small impact on ITER projections, include
incorporate the symmetry breaking around magnetic islands
(1) ion thermal transport near the axis of reversed shear plasmas
may contribute to the toroidal damping along with radial
where large ion orbits exist, (2) the bootstrap current near the
transport driven by turbulence.
axis of reversed shear plasmas and (3) poloidal rotation in low
collisionality plasmas with strong auxiliary heating. However,
2.4.4. Poloidal rotation. Poloidal rotation in the bulk if poloidal rotation can be generated which is much greater than
plasma, under quasi-static conditions, is often assumed to expected from neoclassical theory in the absence of torques
be governed by neoclassical theory because of the strong (e.g. from turbulence) it could affect the radial electric field,
poloidal damping provided by the neoclassical viscosity. This the E × B shear damping and the overall confinement. More
has been confirmed in collisional Ohmic plasmas [192]. accurate diagnostics for measuring the poloidal rotation are
Modelling of the collisional plasma edge (Pfirsch–Schlüter beginning operation and are expected to clarify the issue.
regime) in ASDEX Upgrade [193, 194] also indicates that
the radial electric field is of the order of the neoclassical
2.5. Summary of issues
field, which implies that any anomalous poloidal rotation
contribution is small. Furthermore, its dependence on the With regard to understanding fundamental transport processes
local ion temperature, the density, the poloidal and toroidal several important new trends and results can be identified
magnetic fields and the average toroidal rotation is similar to since the publication of the IPB. (1) Ion thermal transport

S38
 Chapter 2: Plasma confinement and transport

is believed to be reasonably understood in terms of the follows: section 3.1 provides an introduction to the need
dynamics of multiple-scale (wavenumber range and spatial for and benefits from reduced core transport in burning
distribution) turbulence, but predominantly long wavelength, plasmas such as ITER, as well as an overview of the physical
ITG-type turbulence, interacting with self-driven zonal flows basis for obtaining such improved core confinement regimes.
and driven E × B flows, and also depending on other factors This is followed by four sections (3.2–3.5), detailing both
such as magnetic shear and plasma shaping. However, experimental and modelling progress in understanding core
electron thermal transport is not as well understood. A major ion and electron thermal, particle and impurity and momentum
open question for electron transport is whether short-scale- transport, respectively. Dimensionless parameter scaling
type, ETG or TEM, turbulence can generate sufficiently large experiments provide an important additional technique for
transport to account for experimentally observed transport transport understanding and projection—progress in this area
rates. In particular, whether the radially extended structures is described in section 3.6. This is followed by section 3.7,
(‘streamers’) observed in some simulations exist and can which provides a summary of the transport properties of the
account for observed electron transport rates is still unresolved. improved core confinement regimes on current devices that
(2) Coupling between theory, numerical simulations, transport are relevant to advanced operation on ITER, organized by
modelling and experiment has dramatically increased. Gyro- differences in the q profile. Finally, section 3.8 provides a
kinetic turbulence simulation codes have been used to directly summary and a list of outstanding issues. The material in this
investigate transport properties of tokamak plasmas, as well section reflects the current focus of core transport research,
as to benchmark the leading theory-based transport models. which lies in understanding the physical processes governing
Both simulations (limited) and modelling (more extensively) the triggering and dynamics of regimes with reduced turbulent
make quantitative transport predictions and are now routinely transport. Reduced transport has been achieved in all four
used to compare and interpret experiments, while theories of transport channels (ion and electron thermal, particle and
turbulence regulation and suppression now inform all areas of momentum transport channels), sometimes simultaneously.
work on improved confinement and transport. (3) Confidence Rapid progress has also been made in improving the accuracy
that the current ‘standard model’ of turbulence (as already and physics capabilities of transport models, such that
described) is generally correct has been substantially increased experimental and modelling results are integrated in the
by direct, quantitative comparison of turbulence measurements descriptions, rather than treated separately.
with simulations, e.g. zonal flow/GAM features have now
been directly observed in plasma turbulence measurements, 3.1. Enhanced core confinement regimes
etc. (4) There has been a resurgence of interest in neoclassical
transport theory and in refining neoclassical predictions, based Since the publication of the ITER Physics Basis (IPB)
on the now widespread ability to obtain neoclassical levels of document [1], the focus of attention in transport studies has
ion thermal transport. Overall, this progress in fundamental transitioned from conventional ELMy H-mode operation (the
theory, non-linear simulations, and validation of turbulence base or reference scenario in the IPB) to more ‘advanced’
models has substantially enhanced the ability to predict modes of plasma operation with reduced core turbulent
transport, particularly ion transport, on ITER and increased transport and enhanced confinement. As described in
confidence in these predictions. subsequent subsections, internal transport barriers and other
However, many issues still remain; other important issues forms of reduced core transport are now routinely obtained in
not already mentioned include the extent to which recent all the leading tokamak devices [24, 25]. Rapid worldwide
areas of theoretical emphasis, such as turbulence spreading, progress in the development of enhanced confinement modes
intermittency and statistical descriptions, will prove necessary of operation has contributed to an increased emphasis on
in describing tokamak transport processes. With regard developing both ‘hybrid’ and steady-state operation modes
to numerical simulations, the trade-offs between different for ITER, as described in detail in chapter 6 of this issue
numerical implementations with regard to computational [199]. Given this new focus on improved core transport, it
efficiency, numerical noise and adherence to conservations is appropriate to begin this section with a consideration of
laws are currently an area of active investigation. Most the need for and benefits of, enhanced confinement, as well
importantly, there is still no conclusion as to whether short as an overview of the achieved improved transport regimes
wavelength modes (ETG, TEM) and cross-scale coupling and the physical mechanisms responsible for the transport
can explain anomalous electron thermal (heat) and particle improvement.
transport, especially in the presence of reduced or neoclassical
ion transport. Also, while zonal flow and GAM turbulence 3.1.1. Need for and benefits of improved confinement and
features have now been observed on multiple devices, it has transport control. Optimizing the tokamak as an energy
not yet been shown that these features have either the ubiquity producing system involves the simultaneous consideration of
or importance in controlling transport predicted by theory. the following features: (1) high fusion power density, ∝β 2 ,
(2) high fusion gain, Q ∝ βτ and (3) high bootstrap fraction,
3. Core transport fBS ∝ q 2 β. From this, it can be seen that a simultaneous
optimization of plasma confinement, a beta and current profile
The purpose of this section is to provide a description of is required in order to optimize overall tokamak burning plasma
both the current level of understanding and the outstanding performance. Commonly used figures of merit to evaluate
issues with regard to core transport in tokamak plasmas. success in this optimization include βN HL89 , which combines
To this end, the section is structured into subsections as the stability and confinement figures of merit, and a machine

S39
E.J. Doyle et al

Figure 17. Ideal n = 1 stability calculations, using DIII-D parameters, indicate that the maximum achievable βN (and hence plasma
performance) increases with ITB radius, (a), and width, (b) [201].

size independent fusion gain parameter, G ≡ (βN HL89 )/q95 2 plasma state. (2) Optimization of core profiles, e.g. ITB
(see section 6.2), where HL89 ≡ τ/τITER-89P and τITER-89P radius and strength, is required in order to maximize plasma
is a confinement scaling expression for L-mode plasmas [1]. confinement while avoiding stability limits. Too low a local
To provide some quantitative context, conventional ELMy transport coefficient, as can occur with some ITBs, leads
H-mode discharges have βN HL89 ∼ 4–5 with fBS ∼ 30– to pressure gradients that are too steep, thereby lowering
40%, hybrid scenario discharges have achieved βN HL89 ∼ 7 MHD stability margins. Modelling predicts that increased
with fBS ∼ 50%, while target conditions that would enable fusion performance, stability limits and well-aligned bootstrap
steady-state tokamak operation are βN HL89  10, with fBS  current fraction are all simultaneously favoured by having
70%. Specific use of reduced transport/improved confinement moderate profile gradients (weak ITBs) at large plasma radius
include (1) for fixed Q, reduced machine size and/or maximum [200]. An example of such modelling, for DIII-D, is shown in
Bt , hence reduced machine cost, (2) for fixed machine size and figure 17, indicating that the maximum achievable βN (and
plasma current, Ip , increased Q, and (3) for fixed machine size hence plasma performance) increases with the ITB radius
and Q, reduced Ip . This last option forms the basis of much and the width [201]. Experimental data support this picture;
of the hybrid and steady-state scenario development for ITER multi-machine data for high performance regimes show a clear
and is highly attractive as reduced current operation leads to correlation between maximum achievable βN and pressure
both longer discharge pulse lengths and more reduced machine peaking [202,203]. (3) In regimes, with ‘stiff’ plasma profiles,
disruption loadings (see chapter 6 of this issue [199]). both core profiles and global confinement are determined by
It is important to realize that the desired goal is the edge pedestal values (see section 4 of this chapter for
transport optimization and control, rather than confinement more details). Thus, in a burning plasma with stiff profiles,
maximization per se. The benefits of an ability to modify and Q will be determined by edge pedestal temperature values
control core transport are multifold and include the following: (see section 5.5 of this chapter), high values of which are
both difficult to achieve and are a challenge for divertor
1. facilitation of ITB formation and of achieving enhanced
operation. In contrast, regimes with non-stiff core profiles,
confinement regimes,
e.g. ITBs, allow global confinement to vary independently of
2. optimization of confinement while avoiding stability
pedestal temperatures, reducing the required pedestal values
limits,
for a given Q. (4) As advanced tokamak research progresses
3. generation of non-stiff core profiles breaks the link
and bootstrap current fractions increasingly dominate the total
between core profiles and pedestal values that exists with
plasma current, the pressure profile, which is determined by
stiff profiles; with stiff core profiles, performance is
plasma transport, will determine the current profile, which
strongly governed by the achieved pedestal temperatures,
in turn couples back on the plasma transport and stability
4. local transport control is the only possibility for
properties, i.e. advanced tokamak plasmas will be a highly
controlling pressure and bootstrap current profiles in self-
coupled, self-organized non-linear system. A schematic
sustained magnetic plasma configurations and
illustration of multiple non-linear feedback mechanisms and
5. transport control is required in order to vary the relative
couplings that can influence tokamak transport is given in
transport in different channels, e.g. in order to prevent
figure 18 [204]. Local transport control is the only possibility
preferential high-Z impurity accumulation.
for controlling pressure and bootstrap current profiles in such
These issues are now briefly introduced, using the same self-sustained magnetic plasma configurations. (5) Transport
numbering as in the list above. (1) Core transport control tools control is needed to vary the relative transport in different
such as E × B shear flow and magnetic shear modification channels or of particular particle species. For example, highly
(via current profile modification) are actively used to favour peaked density profiles and transport barriers can combine
the creation of ITBs and reduced transport modes [24, 25]. to generate an undesirable central accumulation of high-Z
Use of such techniques is envisaged to aid in obtaining the impurities in JET [205], ASDEX Upgrade [206], DIII-D [207]
initial increase in Q in a reactor so as to enter a burning and JT-60U [208]. As described in section 3.4, techniques

S40
 Chapter 2: Plasma confinement and transport

reversed and positive magnetic shear [210]. However, with

reverse magnetic shear, some sudden transitions to enhanced
confinement are observed that recall the dynamics of H-modes
transitions, e.g. TFTR enhanced reverse shear (ERS) plasmas
[196]. The analogue to the H-mode extends deeper than
this bifurcating quality and extends to the observation of the
formation of sheared flow layers preceding the bifurcation
in some cases, as in TFTR [196], and for some H-modes in
DIII-D [162]. While a consensus view regarding the physics
underlying this shear layer formation is not in hand at this
time, some speculation points to the possibility of transient
turbulence-induced Reynolds stress-driven flows that decay
after the bifurcation event.
As already mentioned, E × B flow shear is believed to be
an element responsible for a wide range of plasma dynamics
Figure 18. Illustration of the many non-linear feedback loops and observed in many of these experiments. Indeed, driven flows
couplings which govern tokamak transport. Reprinted with and the accompanying variations in E × B flow shear have
permission from [204]. been found to have a profound impact on transport barrier
dynamics and even their existence. For example, variations
such as central RF heating and real-time ITB control have of the E × B flow shear with constant heating power on TFTR
been successfully used on several devices to preferentially yielded reverse shear plasmas that spanned the confinement
flatten the density profile and reduce or eliminate such impurity spectrum in a single discharge [211]. Such a discharge began
accumulation [206–208]. with a bifurcation induced by poloidal flows to initiate the entry
into the ERS state. Strong co-rotation yielded a reduction in
3.1.2. Physics basis for improved core confinement regimes. the E × B flow shear that had been dominated by the plasma
It was appreciated in the last decade that a physics basis for pressure, yielding a subsequent collapse of this confinement
core transport should be able to account for the wide range state to the L-mode state. This was followed by another
of transport regimes that have been identified. Indeed, the increase in the E × B flow shear that was dominated by rotation
dialogue in the community is shifting from one wherein plasma and the (re)obtaining of neoclassical ion thermal transport
conditions are labelled as a separate regime to one in which a rates. JT-60U has observed differing characteristics of the
common physics basis is invoked to explain different plasma internal transport barrier characteristics, including variations
characteristics. A successful physics basis ultimately must between ‘box-type’ and ‘peaked’ temperature profiles that are
have at its heart a description of a long wavelength turbulence correlated with the flow induced by various admixtures of co-
set by the scale size of the ion gyro-radius and the means and counter-injecting neutral beams [212].
for moderating or eliminating it and quite possibly a shorter In addition to magnetic shear and flow shear effects,
wavelength turbulence set by the scale size of the electron impurity seeding in TEXTOR [213], DIII-D [171, 214] and
gyro-radius and any means for influencing this as well. JET [215] has been found to induce reduced ion thermal
The richness of confinement characteristics observed in transport in some cases. Turbulence measurements have been
today’s tokamak experiments is quite remarkable and cannot compared with theory calculations for DIII-D and indicate that
be simply described as a linear function of global confinement the impurity content can influence the linear stability of the
quality. For example, all tokamak devices have access to long wavelength modes [170, 171, 214].
L-mode confinement, which is typically characterized by ion- The range of electron thermal transport characteristics
scale turbulence. Many have access to neoclassical transport observed is even harder to characterize than ion thermal
in the ion channel and the means to achieve this vary. Reversal transport reductions. What is true is that reductions in electron
of the magnetic shear profile is found to facilitate entry to thermal transport do not always accompany reductions in
enhanced ion confinement regimes, but shear reversal is also ion thermal transport. Also, strong electron heating has
found to be neither a necessary nor a sufficient condition for generally been found to be accompanied by electron thermal
this entry. transport reductions in reverse shear plasmas (see references
The wide variety of dynamics observed helps fuel the in [24, 25]), even in cases where the ion channel is unaffected.
strong desire to identify a unifying physics picture. Such Understanding the cause of electron thermal transport and
variety includes the following. With reverse shear, the the variety of observations regarding its reduction remains
entry to enhanced confinement in the ion channel has been an outstanding theoretical and experimental challenge. For
found to be continuous and slow compared with an energy a burning plasma the question is critical, as alpha heating will
confinement time, e.g. DIII-D negative central shear (NCS) be predominantly in the electron thermal channel.
plasmas and strongly co-rotating TFTR reverse shear (RS) The leading picture that has the most promise for capturing
plasmas [209]. This continuous character was also found the range of dynamics observed includes the following
on TFTR with monotonic q profiles, with reductions in ion characteristics.
thermal transport being found as the density profiles peaked in 1. Turbulence suppression mechanisms are the key to
going from the L-mode to the supershot regimes [209]. Similar obtaining reduced transport and transport control. Using
results have been obtained in power scans on JT-60U with linear theory, turbulence is expected to be suppressed

S41
E.J. Doyle et al

if the turbulence shearing rate ωE×B exceeds the It should be emphasized that not all parts of the above
maximum linear growth rate γmax of the dominant picture are equally well understood or validated. In general,
unstable mode [20, 27]. However, the various turbulence ion thermal transport is better understood than electron
stabilization mechanisms differ with regard to the thermal, momentum or particle transport. In addition, there
turbulence wavelength they affect, such that obtaining are experimental observations that are not well understood
reduced transport is not equally possible with all in the context of the picture outlined above. Most notable
suppression mechanisms (see section 2.1.2). in this regard is that the onset of enhanced core confinement
2. Shear in the E × B flow velocity likely plays a role with reverse shear is often found to be correlated with the
in governing ion dynamics in many core enhanced location and the timing of the appearance of low order rational
confinement regimes. The multiplicity of feedback values of q in the plasma, most notably q = 2 (ASDEX
loops and accompanying dynamics, including variations Upgrade, DIII-D, JET, JT-60U, TFTR, RTP and others, see
in time scales, is qualitatively captured by this model references in [24, 25]. As the power threshold for ITB
[24, 25, 162, 216]. Sheared E × B flows affect mainly formation can be substantially reduced at such rational q
long wavelength (low-k) turbulence and as such can values, this route to ITB formation is a standard method to
govern ITG-type fluctuations that are believed to control access reduced core transport in machines with limited power
transport in the ion thermal and angular momentum density, such as JET, and is potentially ITER relevant [6].
channels. However, shorter wavelength (medium- and Experiments and modelling aimed at further understanding
high-k) turbulence is not expected to respond to the E × B this effect are actively underway. In this regard, spectroscopic
shear due to smaller spatial scales and larger growth rates. observations from the LHD stellarator point to the existence
Quantitative modelling of profile evolution in the presence of large sheared flows at the boundaries of magnetic islands,
of E × B shear effects on turbulence is still in its early suggesting a possible mechanism for the development of
stages but has managed to replicate key experimental transport bifurcations in tokamaks upon the appearance of
observations (see section 3.2). Today’s plasmas, with low order rational surfaces [219]. In any case, these LHD
their wide range of dynamics, provide the opportunity for observations suggest that a plasma flow induced by external
developing such models and for assessing the value of means (e.g. ion Bernstein wave heating) may provide a means
flow shear control tools that may be used for controlling of initiating enhanced core confinement behaviour and for
the turbulence and thus the pressure and bootstrap current manipulating the pressure profile.
profiles.
3. Magnetic shear reversal and/or α-stabilization/Shafranov 3.1.3. Conclusions. Confinement optimization is an
shift also often play a key role in facilitating entry into essential component of tokamak performance optimization,
enhanced confinement. Theoretically, negative (reversed) and transport control is desirable for a wide variety of reasons.
or low magnetic shear, ŝ, and high α both act to reduce Tokamak transport is a highly coupled, self-organized non-
turbulence growth rates [20], thus facilitating turbulence linear system, and a common physics basis is essential to
reduction. In addition, reversed magnetic shear can also understand the multiple confinement regimes made possible
enable access to second stable core pressure gradients by these non-linear interactions. Over the last decade
[217]. However, a key difference, when compared with such an understanding has begun to emerge, based on the
the E × B shear, that affects only low-k turbulence, is interaction of turbulence suppression mechanisms, such as
that both these mechanisms can affect a wide range of E × B sheared flows, negative or weak magnetic shear
turbulence wavenumbers, i.e. high-k as well as low-k. and Shafranov shift stabilization, interacting with multiple
potential turbulence modes (ITG, TEM and ETG), spanning a
As such, these mechanisms can affect both electron and
broad wavenumber range. These non-linear coupled feedback
particle transports, which are thought to be governed by
mechanisms represent both an opportunity and a challenge
high-k ETG- and medium-k TEM-type turbulence. In
for transport modification and control in burning plasmas.
accordance with this picture, these mechanisms appear to
Indeed, it is only when burning plasmas are realized that
play a key role in the experimental generation of states
possible additional feedback mechanisms, involving alpha
of reduced electron transport [24, 25]. In general, the
heating effects, will be encountered for the first time. For this
interplay between magnetic shear and flow shear effects
reason, a full understanding of the transport characteristics of
has been identified as an essential component in ITB
burning plasma requires the ITER experiment.
formation [216, 218].
4. Impurity seeding is now understood to reduce the
linear instability of long wavelength turbulence in some 3.2. Ion thermal confinement
circumstances, leading to the entry of radiation-improved Significant advances in our understanding of ion thermal con-
confinement (RI) modes as seen on TEXTOR [213], finement have been accomplished through close collaboration
DIII-D [171, 214] and JET [215]. between theory and experiment. Progress has been made in
5. Reduction of edge working gas influxes often improves our ability to carry out detailed experiments and in our ability
the prospects for enhanced core confinement access by to calculate the underlying processes governing confinement.
increasing density peaking. Leading models of enhanced Overall progress in our understanding of transport has been fos-
core confinement point to the benefits of more centralized tered by an increased emphasis on modelling being used both
particle and heat deposition that may result from such as an interpretive tool for current experiments and as a predic-
reductions, yielding increases in pressure and velocity tive tool. While physics questions remain, analysis and predic-
gradients and accompanying increases in E × B shear. tive approaches have culminated in drift wave turbulence being

S42
 Chapter 2: Plasma confinement and transport

commonly accepted as the standard model for cross-field turbu-

lent transport in tokamak plasmas. This is, of course, neglect-
ing the neoclassical and MHD effects that can also influence
confinement. The mechanisms of drift wave transport include
ion temperature gradient (ITG) modes, trapped electron modes
(TEM) and electron temperature gradient (ETG) modes. The
dominant mechanism for ion thermal transport is typically pre-
dicted to be the ITG mode. Numerous experiments on various
tokamaks have been carefully performed to specifically isolate
various dependences of the ion heat flux on quantities such
as the normalized gyro-radius [220–224], plasma beta [224–
226], collisionality [223, 227–230], safety factor [231], E × B
shear [232] and ion to electron temperature ratio [233] in plas-
mas with L- and H-mode edges. Analyses of many these ex-
periments have yielded results that are consistent with the pre-
dictions of drift wave theory.
Theoretically derived transport models have been
developed and systematically tested using standardized profile
data stored in the ITER profile database [234–236]. Three
Figure 19. Ion heat (solid circles) and electron heat (hollow circles)
theory-based drift wave models that are now commonly
diffusivities versus γE /γmax(0) with (dashed lines) and without (solid
used by the fusion community are the IFS/PPPL [237], lines) the destabilizing effect of parallel velocity shear γp . The
MMM95 [238, 239] and GLF23 [20] models. While these points are the GYRO results and the lines are curve fits. Reprinted
models may not robustly yield accurate quantitative agreement with permission from [241]. © 2005 American Institute of Physics.
with experimental data, the models typically predict many
of the correct trends. Here, we will briefly discuss some value of αE needed to quench the turbulence in the absence
recent successes as well as some outstanding physics issues in of parallel velocity shear drive was believed to be in the range
predicting the transport in L- and H-mode tokamak discharges 0.6–0.75 [34, 240]. Recent ITG/TEM non-linear simulations
including enhanced confinement regimes. (including kinetic electrons) using the GYRO code [32] show
that αE is robustly at 0.5 (see figure 19) including cases where
3.2.1. Effect of E × B shear stabilization. A commonly all the modes in the spectrum are in the electron direction [241].
accepted paradigm associated with drift wave theory that has When parallel velocity shear is included, the transport may not
proven to be an essential ingredient in obtaining many core be quenched by any level of E × B shear. This is especially
and edge transport barriers is E × B shear physics [114]. important for higher q values and thus relevant to advanced
It has been widely predicted to be an important effect on tokamak (AT) scenarios.
long wavelength drift wave turbulence and has been validated
experimentally on a wide variety of devices. A prime example 3.2.2. Effects of magnetic shear on internal transport barrier
of an experiment validating the E × B shear paradigm is a formation. Internal transport barriers (ITBs), particularly
DIII-D experiment by Petty et al where co- and counter-NBI in the ion channel, are now commonly observed in all
were applied to a plasma with an ELMing H-mode edge [220]. major tokamaks [242–251]. ITBs are regions within the
Here, the normalized ion gyro-radius was varied while other core plasma that can be characterized as having quenched
dimensionless parameters, except the Mach number, were held or suppressed turbulent transport levels. ITBs provide
constant. Analyses of the experiment show that while the improved performance and scenarios for breaking the profile
electron energy transport did not depend on the direction of stiffness. While the importance of the E × B shear in ITB
the toroidally driven rotation, the ion transport was affected by plasmas has long been realized [162, 218], particularly for ion
the plasma rotation (see section 3.6). The ion thermal transport thermal confinement, the effects of negative central magnetic
exhibited a Bohm-like behaviour for the counter-NBI pair of shear and Shafranov shift stabilization have recently been
discharges, compared with gyro-Bohm-like behaviour for the clarified [252–254]. Recent evidence from JET, ASDEX
co-NBI pair. The electron thermal transport followed a gyro- Upgrade, and JT-60U shows that the magnetic shear directly
Bohm scaling for both pairs. influences the ion heat transport when going from weakly
The differences in the observed thermal transport scaling positive to negative magnetic shear [210, 249]. Experiments
were attributed to differences in the E × B shear between the on JET have shown that the target q profile influences the
co- and counter-rotating plasmas. While the Waltz E × B ITB triggering power [255, 256]. While negative magnetic
shear quench rule appears to work in many cases, it is worth shear does not appear to be necessary for ITB formation, data
noting that it assumes a reduction in transport that can be from several devices suggest that electron thermal ITBs require
approximated by prescribing a net growth rate γnet = γ −αE γE it [257–259]. Furthermore, strong negative shear allows ITBs
that enters the quasi-linear mixing length rule [20]. Here, γ is to form where the turbulent transport is suppressed in all
the linear drift-ballooning mode growth rate in the absence of channels, as demonstrated on DIII-D [218, 260]. This is in
E × B shear, γE is the E × B shearing rate and αE is the E × B  B shear driven ITBs which typically only
contrast to purely E×
shear coefficient. This prescription was derived based on non- result in ion thermal barriers. ITB discharges with strongly
linear ITG simulations with adiabatic electrons. The critical reversed q profiles with little to no central poloidal field are now

S43
E.J. Doyle et al

Figure 20. Temperature profiles for ions (solid lines) and electrons (dashed lines) along with the safety factor profile for ITB discharges
from JT-60U, DIII-D and JET with strongly reversed magnetic shear [234].

routinely produced and very sharply defined ion temperature neoclassical to being strongly turbulent. DIII-D has been
profiles have been produced. Figure 20 shows examples of successful in demonstrating sustained H-mode discharges with
ITB discharges from JT-60U, DIII-D and JET with strongly reversed magnetic shear and weak or ‘leaky’ ITBs even in the
reversed q profiles from the ITPA profile database [234] (see presence of Type I ELMs [264, 265].
also section 5.1).
In some strong negative shear discharges, the central ion 3.2.3. Temperature ratio effects. Another factor that can
temperature profiles are very flat within the reversed shear play a key role in suppressing ITG/TEM transport is the ion
region (see figure 20(a)). An example of this is the box-type to electron temperature ratio. The Ti /Te ratio is of particular
ITBs observed in JT-60U [259]. It remains unclear whether interest when projecting the favourable transport from existing
this central flattening is due to MHD activity, a breakdown hot ion plasmas to proposed burning plasma experiments where
of MHD equilibrium, enhanced neoclassical and/or turbulent Ti /Te is expected to be less than or equal to unity. A recent
transport or a combination of effects. The optimum reversal DIII-D H-mode similarity experiment found that increasing
of the q profile for a given operating scenario has yet to be Ti /Te reduced the ion and electron energy transport and the
realized. In any case, negative central magnetic shear operation helium particle transport for a 20% scan in Ti /Te at fixed
is now commonplace and understanding the relative roles of plasma β [233]. GLF23 modelling showed that the sensitivity
magnetic shear and Shafranov shift stabilization in core barrier of the energy transport to changes in Ti /Te is likely explained
formation is an active area of research. Furthermore, since by changes in the critical ion temperature gradient close to the
toroidally driven E × B shear flows are likely to be small in a threshold.
reactor such as ITER, negative magnetic shear and Shafranov Temperature ratio effects can also impact the temporal
shift stabilization are both more likely to be achievable and response of turbulent transport. Recent DIII-D modulated
impact all channels of transport. ECH experiments, including ion measurements, have shown
Emphasis on the so-called AT operational regime has evidence of an ion heat pulse and its rapid propagation in
aimed at achieving steady-state, long pulse, high performance response to an electron heat pulse which are consistent with
H-mode discharges that utilize weak or slightly reversed ITG transport [266, 267]. Electron heating at ρ ≈ 0.45
central magnetic shear profiles at high beta with a central produced a phase shift resulting in a cooling of the ions at
safety factor greater than unity [261]. While AT discharges the plasma centre. Modelling showed that the drop in the ion
tend to have larger amounts of Shafranov shift stabilization, temperature was the result of a decrease in Ti /Te , which then
parallel velocity shear [γp = (Rq/r)γE ] destabilization can caused an increase in the ion thermal diffusivity [267].
prevent E × B shear (γE ) quenching of the transport, which In perturbative experiments with cold edge pulses, the
results in the E × B shear only suppressing the transport observed fast radial propagation of the pulses from the plasma
rather than eliminating it [262]. Another regime is also edge to the core has been a challenge to be explained by local
being developed that demonstrates high beta operation with diffusive transport models. The ITG threshold has the unique
q(0) ∼ 1, a broad region of low magnetic shear and the signature that it increases when the Ti /Te ratio increases and,
absence of sawtooth activity. This regime has been labelled therefore, offers a unique explanation of cold edge pulses
the ‘hybrid’ regime (see chapter 6 of this issue [199]) by propagating inwards and producing a core temperature rise
working groups of the ITPA [263] and offers the potential of as shown in TEXT [268], TFTR [269], ASDEX Upgrade
achieving many of the performance goals of ITER including [270] and RTP [271] experiments. At the plasma edge, a
high fusion gain [264]. At low q, the hybrid scenario tends reduction in Te raises Ti /Te , which then increases the ITG
to have less Shafranov shift stabilization than the higher q AT threshold and produces a pulse of better confinement that
 B
scenario, but non-linear simulations at low q suggest that E× appears as heating. Time-dependent transport modelling
stabilization can win over γp destabilization and the transport studies [267, 270] demonstrated that the IFS/PPPL, GLF23
can be quenched to neoclassical levels [241]. Unlike L-mode and MMM95 models (which have an ITG threshold) have this
discharges with ITBs, high performance AT and hybrid H- phenomenon but single fluid models such as the CDBM model
mode discharges usually do not demonstrate sharp breaks in do not. While simulations using stiff critical gradient ITG
the profiles where the transport changes dramatically from models, such as the IFS/PPPL and GLF23, have produced the

S44
 Chapter 2: Plasma confinement and transport

same qualitative behaviour observed in cold pulse experiments,

quantitative agreement is still lacking.

3.2.4. Improved confinement through impurity seeding.

Another method of stabilizing drift wave turbulence that is
potentially applicable to reactors is through impurity seeding,
called the radiative improved RI-mode of operation. A deeper
understanding of mechanisms that suppress ion temperature
gradient (ITG) turbulence has been achieved by deliberate
puffing of impurities, e.g. the RI-mode in TEXTOR and
impurity seeded discharges in DIII-D and JET [171,213,215].
Such plasmas are characterized by an increased level of the
effective ion charge, Zeff , and electron density gradient. It
has been demonstrated that both these factors are important
in reducing the ITG transport via increases in the critical
temperature gradient and the E × B shear suppression. An
initial increase in Zeff with impurity seeding leads to a
−1/2 Figure 21. Ion thermal diffusivity versus normalized ion
reduction of ITG particle diffusivity, which scales as Zeff . As temperature gradient from GYRO non-linear simulations (points),
a result, a particle pinch generated mainly by trapped electrons original GLF23 model (dashed line), and retuned GLF23 model
leads to density peaking. This further reduces the growth rate (solid line). Reprinted with permission from [533]. © 2005
of ITG modes and finally leads to a bifurcation of an RI-mode American Institute of Physics.
state with suppressed ITG transport. Gyro-kinetic analysis of
a DIII-D L-mode experiment with neon injection showed that closer to that of the MMM95 model. Figure 21 shows the
improvements in the ion thermal transport were attributable predicted ion thermal diffusivity from GLF23 plotted against
to an impurity driven enhancement of the E × B shearing the GYRO non-linear gyro-kinetic results for the standard case
rate along with a direct reduction of ITG turbulence [171]. described in the 1996 GLF23 paper [20]. At the reference
Under H-mode conditions, impurity seeding can lead to a normalized ion temperature gradient of a/LTi = 3.0 (vertical
widening of the edge transport barrier where ITG is suppressed dashed line), there is nearly a factor of three reduction in the
by both the density gradient and the radial electric field. Such transport as a result of renormalizing the model against the
a barrier broadening was observed by argon injection into GYRO non-linear gyro-kinetic results.
JT-60 U [272]. Despite differing levels of stiffnesses, the gyro-fluid based
(GLF23, IFS/PPPL) and fluid based (MMM95, Weiland)
3.2.5. Outstanding issues. While significant advances have models have exhibited the same level of agreement when
been made in our predictive understanding of ion thermal compared with the experimental profile data. In the
transport, there remain important outstanding physics issues. standardized testing carried out by the ITER Modelling
While there are outstanding theoretical issues in neoclassical Working Group, it was found that it was difficult to distinguish
ion thermal transport, much of the focus is on turbulent statistically between various models based on how well they
transport issues. A prime example involves the degree of predict the temperature profiles in L- and H-mode discharges
profile stiffness. Here, we define profile stiffness as the rate that do not display significant core transport barriers [234,235].
of increase in the ion heat flux in response to an increase in However, the models lead to very different predictions in
the ion temperature gradient above the threshold value. If the burning plasma performance when applied to future devices
core temperature profiles are stiff, they are then unresponsive such as ITER [236, 279]. While the required pedestal
to changes in auxiliary power. A stiff core plasma is also temperatures Tped from the MMM95, IFS/PPPL and GLF23
sensitive to changes in the plasma edge conditions. There is models needed to obtain a specific fusion gain can differ
ample experimental evidence of temperature profile stiffness significantly, it is worth noting that they are all within the
from a variety of devices such as Alcator C-Mod [229], current range of possible Tped projections.
ASDEX Upgrade [273–275], DIII-D [276] and JT-60U [277]. Another example of an unresolved physics issue pertains
For H-mode operation, the core performance is predicted to to robust experimental observations in L-mode whereby the
depend sensitively on the height of the pedestal. However, analysed ion thermal transport appears to follow a Bohm-like
quantifying the level of stiffness in the plasma core has proven scaling. This result has been a challenge to explain the given
difficult. The MMM95 model [238, 239], which is based on intrinsic gyro-Bohm scaling of drift wave models. Transport
the Weiland/Chalmers ITG/TEM drift wave model [278], can modelling [239, 280] and turbulence simulations [32, 34, 281]
be categorized as weak to moderately stiff. Models based on of particular cases have been able to produce apparent Bohm-
gyro-fluid simulations, such as the IFS/PPPL [237] and GLF23 like scalings of the diffusivities. Figure 22 shows the results
[20] models, are comparatively more stiff. For a given ion of GYRO non-linear simulations with various physical effects
temperature gradient, the predicted ion heat fluxes from gyro- and their influence on the gyro-radius scaling in comparison
fluid simulations are significantly larger than those obtained with the experimental analysis of a recent DIII-D L-mode gyro-
from gyro-kinetic simulations [110]. Recently, the GLF23 radius scan [32]. While the gyro-Bohm scaling can be broken
model was renormalized to non-linear gyro-kinetic simulations close to the threshold temperature gradient, isolating a single
making it less stiff and yielding fusion projections somewhat piece of physics that robustly yields non-gyro-Bohm behaviour

S45
E.J. Doyle et al

also typically operate at a much higher β. Recent analyses of

NBI-heated NSTX discharges indicate that within the spatial
region from r/a = 0.2–0.65, where the heat transport is
confidently known, the electrons dominate the heat loss with
χi χe . This is seen in figure 24, where the ion thermal
diffusivities at r/a = 0.4 for a collection of both L- and
H-mode neutral beam heated discharges are up to an order
of magnitude, on average, less than the electron thermal
diffusivity. Furthermore, χi is at or above the level predicted
by neoclassical theory (as determined from the NCLASS
model [185]).

3.3. Electron thermal transport

For more than two decades the electron temperature profiles
in conventional tokamak plasmas, L- or H-modes without
ITBs, are observed to react weakly to changes in the auxiliary
Figure 22. Impact of selected physical effects on the predicted ion heating power deposition profile, as described in [285–291]
energy transport using the GYRO code in comparison with
experimental estimates. Reprinted with permission from [32].
and more recently in [11, 274, 292–295]. This property has
© 2003 American Physical Society. been named ‘profile consistency’ [296], ‘profile resilience’
or more recently ‘profile stiffness’. In experiments with
in the L-mode for the ion channel (but not the electron channel) central heating the normalized electron temperature gradient,
remains elusive. R/LTe = R∇Te /Te , exhibits in all tokamaks a rather constant
Transient transport phenomena have also posed a value around 10 [297]. With off-axis heating the electron
significant challenge to be explained by the theory and temperature profile generally remains peaked and R/LTe keeps
modelling community. DIII-D modulated ECH experiments a finite value well above zero. Electron heat transport has
have shown clear evidence that electron perturbations create been extensively investigated, in particular supported by the
changes in the ion heat response which are consistent with ITG installation of electron cyclotron heating systems in several
transport [266]. Figure 23 shows the amplitude and phase of devices. This heating method provides narrow on-axis or
the electron and ion heat pulses for a DIII-D modulated ECH off-axis power deposition profiles with pure electron heating.
experiment accompanied by the predictions of the IFS/PPPL, These studies yield experimental results strongly suggestive
GLF23, MMM95 and Itoh–Itoh–Fukuyama (IIF) models. This of electron heat transport that is governed by turbulence
provides a clear example of an ion heat pulse and its rapid increasing above a threshold in normalized gradient, R/LTe,crit .
propagation in response to an electron heat pulse. Here, the This means that the heat flux is low below the threshold and
IFS/PPPL and GLF23 models demonstrated the best level of increases above it. The rate of increase above the threshold
agreement with the ion behaviour. can be characterized by the ‘stiffness factor’. Some of the
While the modulated ECH experiments provide an results addressing this hypothesis are summarized in [297]
additional test for transport models, clear discrimination and discussed for individual devices in [298–304]. It will be
between the models in terms of their temporal response has shown later that in general the value of R/LTe deduced from
proven difficult to achieve [266, 267, 282]. No single model the experimental temperature profiles is significantly larger
has been able to consistently perform better than other models than R/LTe,crit : the electron temperature profiles in present-day
in reproducing the observed phase and amplitudes of both tokamaks are not infinitely stiff.
the electron and the ion temperature perturbations as well The theory of heat transport presented in section 2
as the equilibrium profiles. One can, however, discriminate indicates that electron heat transport may be driven by
against single fluid models based on these results. In other turbulence excited by TEM and/or ITG modes [20, 278, 305]
perturbative experiments with cold edge pulses, the observed and/or by ETG modes [306, 307]. It must be stressed that,
fast radial propagation of the pulses from the edge into the in general in conventional scenarios, due to its very small
plasma core has been difficult to explain by local diffusive (i.e. scale ETG driven turbulence can only cause a relevant electron
drift wave) transport models. Simulations using stiff critical heat flux if larger cells, so-called streamers, develop [47, 307].
gradient ITG models, such as the IFS/PPPL and GLF23, have These three different instabilities have respective thresholds
produced the same qualitative behaviour observed in cold pulse in ion or electron normalized temperature gradients, above
experiments, but quantitative agreement is still lacking and, which turbulence and the corresponding transport increase.
like the modulated ECH experiments, no single model has been Formulae have been derived for the threshold of TEM [308]
able to satisfactorily reproduce all of the observed temporal and ETG [47], which can be easily compared with the
behaviour [267, 283]. experiment. Several plasma parameters play a role, in
One final example of an outstanding physics issue pertains particular Te /Ti , the density gradient, the magnetic shear and
to understanding turbulent transport in low aspect ratio the safety factor. In addition TEM are predicted to be stabilized
tokamaks. Unlike standard higher aspect ratio tokamaks, low by collisionality due to detrapping of electrons. Note that,
aspect ratio devices such as NSTX [284] and MAST appear as already indicated in section 3.2 of this chapter, the ion
to be dominated by high-k rather than low-k turbulence. They temperature profiles also exhibit profile resilience which can be

S46
 Chapter 2: Plasma confinement and transport

Figure 23. Fourier analysis of the phase and amplitude for δTe and δTi for measured data (solid circles), the IFS/PPPL model (dashed lines),
the IIF model (dotted lines), the MMM95 model (chained lines) and the original GLF23 model (solid grey lines) for the case of ρECH = 0.3.
Reprinted with permission from [266].

due to Te > Ti the calculated threshold is clearly higher

than the measured R/LTe . This simple and clear situation
where the TEM contribution dominates can be investigated
experimentally under good conditions as described below.
In contrast, comparable electron and ion heat fluxes lead
to similar Te and Ti values and all the three modes can be
expected to contribute to electron heat transport. This is a
complicated experimental situation that will be discussed later
in this section. Finally, if the turbulence can be stabilized
in a region of the plasma, an electron ITB can be created as
discussed at the end of this section. In this work we report the
essential recent results on electron heat transport in tokamaks
and do not intend to provide a complete overview of this
topic.
It must be underlined that electron heat transport can also
Figure 24. Ion thermal diffusivity plotted against the electron
be driven by global MHD instabilities. This topic is out of the
thermal diffusivity for L-mode (blue) and H-mode (red) NSTX scope of this section and we concentrate on cases where their
plasmas. effects are negligible.

attributed to the same reasons applying to ion heat transport by 3.3.1. Electron heat transport with dominant electron heating.
ITG [275, 309].
In the cases with dominant electron heating, with low heat
A simple empirical model based on the existence of a
flux in the ion channel and high heat flux in the electron
threshold R/LTe,crit has been successfully tested on ASDEX
channel, Te > Ti and R/LTe > R/LTi , the electron heat flux
Upgrade ECRH heated plasmas [310]. It includes the increase
can be considered in general as mainly driven by the TEM
in transport above the threshold and the T 3/2 gyro-Bohm
instability [308]. Such a situation can be achieved in plasmas
dependence. Based on this initial study, this transport model
with strong electron heating only, for instance ECRH or ICRF,
has been extended with a q 3/2 dependence, which yields
and at moderate and low collisionality. Collisionality plays
the required radial dependence of transport as well as its
dependence with plasma current [23]. The q dependence takes an important role in two ways. Firstly, low collisionality
into account the shift of the k spectrum of the modes to lower allows the separation of the ion and electron temperature
values with q [20, 122]. profiles and heat fluxes. Secondly, increasing collisionality
The relative contributions of these three different gradually stabilizes the TEM modes as discussed later. In the
instabilities to electron heat transport vary depending on the cases where electrons and ions remain separated, the study is
plasma conditions. In the cases with dominant electron heating simplified and electron heat transport can be studied in practice
and Te > Ti , for instance with ECRH, the electron heat independently of the ions. The experimental investigations
flux is large compared with the ion heat flux and transport and the comparison with theory are clearer, providing an
is dominated by the TEM. The ITG contribution is small unambiguous check for the TEM theory of electron heat
because of the low ion heat flux. The ETG modes are stable: transport. Even if such situations are not relevant of the

S47
E.J. Doyle et al

2.0
5.0 ASDEX Upgrade q=1 barrier RTP
13556 -13558 ECH
1 300kW
180 kW
100 kW
Te [keV]

Ohmic

T e [keV]
1.0 0.5

0.5
Ohmic
0.8 MW ECH
1.6 MW ECH
0.1 0.1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
ρ ρ
1
3.0 10
FTU 18282 TCV #18224
ECH 2.70 MW ECH
0.8 MW
1 Ohmic T [keV] 1.35 MW

0
T e [keV]

10
Ohmic
0.5
e
0.39 MW
0.4 MW

–1
10

0.1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
ρ ρ
vol
10 10
10 MW ICRF JET Tore Supra
H-modes 5.0
ICRF D(H) + NBI 2.3 MW
5.0 (NBI = 14 MW) 3.5 MW
7.4 MW
Te [keV]
Te [keV]

1.0
8 MW ICRF

6 MW ICRF
1.0 0.2
3.0 3.2 3.4 3.6 3.8 0 0.2 0.4 0.6 0.8 1
R [m] r/a

Figure 25. Electron temperature profiles from different tokamaks on logarithmic scale showing the region with constant normalized
temperature gradient. Reprinted with permission from [297]. © 2001 Institute of Physics.

conditions which will be found in burning plasmas, they are on the edge or pedestal temperature. The same slope means
essential to assess our understanding of the underlying physics. that R/LTe is the same, or equivalently, the core temperature
In the following we summarize the experimental results is proportional to the edge temperature [275]. Only a large
obtained in plasmas with dominant electron heating. These are variation of the electron heating power (up to 7 MW) deposited
L-modes with rather low coupling between ion and electron centrally, as done in Tore Supra, could produce a measurable
and with a low heat flux in the ion channel compared with that variation of R/LTe [300]. In such cases only, the extrapolation
transported by the electrons. to zero power may yield the actual threshold R/LTe,crit . Under
With central heating, the electron temperature profiles this assumption the Tore Supra results suggest a threshold with
plotted on a logarithmic scale have a very similar shape [297]. a dependence on s/q, the ratio of magnetic shear to safety
Examples are shown in figure 25. The profiles exhibit the factor. In the cases with a smaller range in heating power,
same slope and are shifted with respect to each other depending the measured R/LTe remains about constant independently of

S48
 Chapter 2: Plasma confinement and transport

4 8 modes. In particular, in ASDEX Upgrade linear calculations

Te PB of turbulence stability were carried out with the gyro-kinetic
χe 7
code GS2 for these discharges and compared in detail with the
3 6 experimental results [308]. They indicate that TEM indeed
dominate electron heat transport in these conditions. Due to
Te [keV]

χe
the large Te /Ti ratio, the ETG turbulence is not excited in these

PB
2 4 cases. The threshold in R/LTe provided by GS2 analyses
agrees with the experimental value R/LTe,crit = 4 and the

[m /s]
3

2
increase of heat flux with R/LTe above the threshold also has
1 2 the same dependence as in the experiment.
1 These similar results from ASDEX Upgrade, DIII-D and
ECH
TCV underline the important point that in general, in the region
0 0
0.0 0.2 0.4 0.6 0.8 1.0 of the plasma where the electron heat flux is large, the value
ρ of R/LTe is clearly above the threshold by a factor up to 2
tor
or 3. To some extent this is also visible in experiments with
4.0 a very large variation in central heating power [300]. Hence
T
3.0 e the Te profiles are intrinsically not very stiff: a change in heat
flux can in principle cause a significant variation in R/LTe .
Te [keV] (log scale)

The fact that varying the central heating power in general does
not cause a large change in R/LTe is due to the three effects.
1.0 Firstly, the Te profiles cannot easily drop below the R/LTe
0.8 threshold: even with Ohmic heating only they remain above
0.6 the threshold. Secondly, the edge temperature increases with
0.4
heating power, which allows the core electron temperature to
ECH
increase without changing R/LTe significantly. Thirdly, due
3/2
to the Te gyro-Bohm dependence of heat transport driven
0.2
0.0 0.2 0.4 0.6 0.8 1.0 by micro-turbulence, the actual profile stiffness increases with
ρ temperature: at high temperature the profiles tend to be closer
tor
to the threshold, preventing any strong change of gradient with
Figure 26. Top: Temperature profile (linear scale) and power heat flux. The empirical transport model reproduces well the
balance heat diffusivity for moderately off-axis ECRH. Bottom: general behaviour of the temperature profiles and is in rough
Electron temperature profile on log scale which indicates the change
in slope (∇Te /Te ) at the ECRH deposition, indicated by the dashed agreement with the results from global confinement scaling
lines. Reprinted with permission from [311]. laws [23]. This study describes in detail the effect of the edge
temperature, stiffness characteristics and power deposition on
heating power, but this value is believed to be significantly the properties of the electron temperature profiles.
above the threshold, as discussed below. Power modulation and the analysis of the resulting
Localized electron heating deposited off-axis is a powerful temperature oscillations yield the so-called heat pulse
tool to investigate properties of electron heat transport and diffusivity χeHP [315]. This quantity is the derivative of the
the possible existence of a threshold. In fact, when off-axis electron heat flux versus electron temperature gradient at the
heating is applied the temperature profile, plotted on linear time-averaged working point. It yields precious information
scale, remains peaked and seems to react weakly to the on the stiffness properties of the electron heat transport. In
large difference in heat flux on each side of the deposition particular, if a threshold exists the change of slope in qe , from
[290, 298, 303]. However, plotted on a logarithmic scale, the flat to steep, induces a jump in χeHP as pointed out in [310].
Te profile indicates indeed a low value of R/LTe inside of This provides a direct experimental indicator of a threshold.
the power deposition compared with the outer part [311]. An Power modulation and transients have been studied for a long
example with off-axis ECRH is illustrated in figure 26 showing time as reported for instance in the review papers [315–317]
that the Te profile appears quite peaked on a linear scale, but and references therein. However, at that time, the experimental
less peaked in the core in term of normalized gradient. It is information could not be ordered to a consistent physics picture
believed that in the plasma region inside of the off-axis power of electron heat transport.
deposition R/LTe is just above but close to the threshold [312]. Recent studies using modulation techniques have been
Experiments were made in ASDEX Upgrade in which guided by the possible existence of a threshold in R/LTe
R/LTe has been varied gradually by using ECRH at two and could indeed support this hypothesis in ASDEX Upgrade
different radial positions, keeping the edge temperature [298, 304, 312]; DIII-D [313] and FTU [303, 318]. In these
constant. The results strongly support the existence of a devices the propagation of the heat pulses reacts strongly
threshold: the electron heat diffusivity goes to zero for a clearly to the off-axis heating. In the ASDEX Upgrade and DIII-
finite value of ∇Te corresponding to R/LTe [312]. Similar D common experiments described above [312, 313], where
experiments carried out more recently in DIII-D [313] and R/LTe has been varied keeping the edge temperature constant,
TCV [314] yield very comparable results. power modulation was also applied. The results from the
The analysis of turbulence stability for these three two devices for χeHP are shown in figure 27 and support the
experiments indicates that the TEMs are the most unstable possible existence of a threshold. Indeed, as shown in [312],

S49
E.J. Doyle et al

collisionality when in addition TEM modes are stabilized. So

far no evidence for ETG driven heat transport has been found
in present machines. However, this is a crucial issue for ITER
because ETG can be expected to be unstable in its core region.

3.3.2. Electron heat transport with comparable electron and

ion heating. In the cases with comparable ion and electron
heating (e.g. with NBI heating) the Te profiles also exhibit
profile stiffness described above. They have about the same
value of R/LTe as indicated above and it is also widely
observed that this value does not vary in power or density
scans, [274, 275, 292–294]. In plasmas heated by NBI the ion
heat flux may be larger than the electron heat flux. However,
the latter is never negligible and reaches at least 20% of
the total heat flux, in general more. Consequently, even in
experiments using in addition off-axis ECRH, the electron heat
flux inside of the ECRH deposition cannot be very small and
Figure 27. Data from ASDEX Upgrade and DIII-D from common
it is significantly above the residual Ohmic heat flux. This
experiments where ∇Te /Te has been varied. This shows rather good
agreement for the power balance, including an offset in ∇Te /Te . strongly limits the experimental possibilities. In contrast to
The values for χeHP are not inconsistent. Reprinted with permission the plasmas described in the previous section, the temperature
from [313]. profiles exhibit R/LTi larger or close to R/LTe . The TEM and
ITG modes are both contributing to electron heat flux and the
the empirical model with critical R/LTe agrees with both the resulting electron heat transport depends on both the ion and
χePB (heat diffusivity defined from the static power balance) and electron temperature profiles. The investigation of these cases
χeHP data for a threshold value just below the range covered by is complicated by the coupling between channels. In addition,
the data in R/LTe . However, in these experiments the values of due to the different profiles of R/LTe and R/LTi the dominant
R/LTe could not be reduced enough to drop below R/LTe,crit . modes might depend on the radial position [309]. Moreover,
Therefore these studies have been extended very recently in in such situations Te /Ti is generally close to or below unity
ASDEX Upgrade experiments in a new variation of R/LTe at and the threshold formula for ETG given in [47] indicates that
low current to reduce the residual Ohmic power in the off-axis these may be unstable. Nevertheless, dedicated experiments
case. This allowed R/LTe to be indeed scanned across the provide important elements to be compared with theory.
threshold which was clearly revealed by the expected jump in The specific aspect of electron transport in the frame
χeHP [304]. of ITG/TEM physics has been addressed in two dedicated
Collisions are expected to stabilize the TEM. This property series of experiments in DIII-D [233, 266]. The ratio Te /Ti
can explain experimental observation made in TCV [314] and is predicted by theory to be essential for the stability of these
ASDEX Upgrade [129, 304]. Indeed a transition from TEM modes. The first set of experiments presents steady-state
to ITG-driven electron heat transport has been observed. In investigations in which the ratio Te /Ti has been varied keeping
such case the ETG can also be unstable according to theory, either Te or Ti constant. They indicate a strong dependence
but no experimental evidence has been found so far [129]. It of electron heat transport which suggests χe ∝ Ti3 . This is
must be stressed again that ETG turbulence drives significant in agreement with the sensitivity of the ITG/TEM stability
transport if the cells are large enough, streamers are required. to Te /Ti . The second set of experiments [266] uses ECRH
The stability criterion is not a sufficient condition for heat to power modulation in NBI-heated L-modes. A modulation in
be driven by ETG. In FTU experiments using pellet injection both the electron and ion temperatures is observed. As the
have carried out leading to quite high collisionality with large energy exchange time is long compared with the modulation
gradients. The stability analyses show that the TEM modes are period, this indicates a dependence of χe upon Ti or Te /Ti . The
expected to be stable and that electron transport may be driven transport models based on ITG/TEM physics indeed produce
by ITG and ETG [319]. They also indicate that the increase in a modulation in both Te and Ti . The quantitative agreement,
the density gradient caused by the pellets leads to a stabilization which was rather poor in reference [266], could be improved
of these modes which explains the observed improvement of recently by increasing the stiffness of electron transport [320].
confinement. In ASDEX Upgrade, experiments using steady-state and
Summarizing, several experiments and analyses under modulated ECRH were performed in low density H-modes to
quite different conditions and devices strongly support the ensure strong ion heat flux and low coupling with the electron
hypothesis that in plasmas with dominant electron heating a [321]. These results also show that ion transport is strongly
threshold in R/LTe exists and leads to the observed stiffness. In affected by changes in the electron temperature when ECRH
the majority of the cases, the most unstable modes are found to is applied. The analysis of modulation data also suggests the
be the TEM which are believed to be the main cause for electron existence of a threshold R/LTe = 6 outside the very central
heat transport in plasmas with dominant electron heating. The part of the plasma (ρ > 0.25). This is in agreement with the
threshold of the ETG modes decreases with decreasing values expected values within the experimental uncertainties. The
of Te /Ti and this turbulence might dominate electron heat results also indicate an increase in stiffness towards the edge as
transport in some cases described in [300] and [318] and at high experimental results from DIII-D [322]. Turbulence stability

S50
 Chapter 2: Plasma confinement and transport

analyses indicate that in these discharges all the three modes

ITG, TEM and ETG may be unstable but a clear statement on
their respective contributions to electron heat transport cannot
be made so far due to the experimental uncertainties.
The scaling experiments made in DIII-D and mentioned
in section 3.2 for the ions also yield important information
on electron heat transport in NBI-heated L- and H-
modes [225, 227, 231–233]. There are three main conclusions.
Firstly, the ρ∗ scaling experiments indicate that electron
transport is always gyro-Bohm whereas ions do not show an
unambiguous behaviour. Secondly, the beta dependence is
very weak for the electrons, which probably rules out magnetic
flutter as a main contributor. Thirdly, electron heat transport
seems to be weakly sensitive to E × B shear stabilization.
In summary, the understanding of electron heat transport
in experiments with comparable ion and electron heating is Figure 28. Electron temperature profiles in TCV showing that
more complicated and more difficult to assess than in the central counter-ECCD (solid line Ptot = 2.25 MW; dashed–dotted
cases with dominant electron heating. It is highly probable line Ptot = 1.8 MW) creates a strong electron ITB whereas ECRH
that ITG/TEM turbulence is the main contributor to electron (dashed line Ptot = 2.25 MW) does not. Reprinted with permission
from [337]. © 2001 American Institute of Physics.
heat transport, as ITG modes are the main players for ion heat
transport (section 3.2). However, the quantitative comparison
is not precise and the ETG modes might also play an important [25, 323] and can be found for the different tokamaks
role, not quantified at all yet. as follows: ASDEX Upgrade [324–326]; FTU [327];
In burning plasma experiments one expects a strong JET [328–330]; JT-60U [331–333]; Tore Supra [334, 335];
centrally peaked electron heating and a weaker ion heating. TCV [336–339] and T-10 [340].
The weak ion-electron coupling is compensated by the large Electrons ITBs, generally obtained at very low densities,
size of the machine and by the long energy confinement time, lead to central electron temperatures that can reach up to
such that electrons and ions are weakly coupled in the very core 20 keV [326] and steep gradients with R/LTe up to 20 [323].
but increasingly coupled towards the plasma edge. Therefore, This latter value is 2–3 times larger than generally measured
in conventional plasmas and indicates a clear barrier. The
electron heat transport in the very core could be dominated by
radial position of the barrier foot is related to the position
TEM modes, similar to the description given at the beginning of
of the minimum q value; however, due to the uncertainties
this section. Due to the very low collisionality these modes will
of that quantity it is not possible to specify if it is right
be highly unstable and will probably dominate electron heat
at this position or slightly inside. The change in gradient
transport. Further outside, the ITG contribution will increase
which determines the foot of the barrier is generally very clear
and electron heat transport is expected to be close to what was
on the temperature profile, see for instance [328, 332, 337].
addressed in the above paragraphs, with the difference being
An example of electron ITB created with central counter-
that collisionality will be lower and ρ∗ as well. The effect
ECCD is shown in figure 28. In all the studies the creation
of this latter point is an important issue for extrapolations.
and sustainment of electron ITBs are found to be related to
In a fusion reactor the temperatures will be higher than in
3/2 magnetic shear. In T-10 flat or even slightly positive shear
present tokamaks and due to the Te dependence of transport
seems to be sufficient to create an electron ITB [340]. In
the temperature profiles will be kept closer to their respective the other tokamaks electron ITBs with dominant electron
thresholds. It is therefore essential to assess the sensitivity of heating require negative magnetic shear [325]. Experiments
the ITG/TEM stability to Te /Ti and more experimental results in which the depth of the reversed q profile could be varied
on this topic would be very useful. Finally, the ETG may indicate that the strength of the barrier increases with the
well contribute to electron heat transport but and experimental reversed character of the q profile [326, 333, 339, 341]. In
results on this issue are highly desirable. JET, turbulence calculations indicate that the formation of an
electron ITB is related to the stabilization of the TEM for a
3.3.3. Electron internal transport barriers. Internal transport reversed shear such that s
−0.5 [330]. As the ITB develops
barriers characterized by large temperature gradients are able in time after it has started, the velocity shearing rate can further
to yield simultaneously a large fraction of non-inductive contribute to turbulence reduction. An additional contribution
bootstrap current and high performance due to the high to stabilization can be provided by the gradient of β [254].
temperature which can be reached. Scenarios exploiting These mechanisms provide a positive feedback loop as higher
the dominant ion heating by NBI have been extensively pressure and steeper gradients can further stabilize. In a reactor
investigated, as described in the previous sections. They such a loop may be supported by the increase in fusion power
produce strong ITBs in the ion channels but generally weak with increasing pressure. As discussed above, one may expect
or no ITBs in the electron channels. Strong electron ITBs that TEM driven turbulence is responsible for electron heat
have been created with dominant electron heating, generally transport in most of the low density plasmas heated with pure
in conjunction with current drive, using RF scenarios such as electron heating. This is particularly true for these plasmas
LH, ECRH and to a lesser extent ICRF. The main experimental before the ITB formation. Therefore, one may assume that
studies and results on this subject have been reviewed in the stabilization of these modes causes the ITB formation, as

S51
E.J. Doyle et al

Figure 29. Results from JT-60U showing that the electron ITB decreases with time which corresponds to an increase of NBI power and ion
temperature. Reprinted with permission from [333]. © 2004 Institute of Physics.

indicated by the JET analysis. This is supported by turbulence 3.4. Particle and impurity transport
measurement as shown in [334, 340, 342]. It is however not
Density control is essential for burn control in a fusion
excluded that in fully developed electron ITBs the large value
reactor, where understanding of particle confinement and
of R/LTe either destabilizes them again or destabilizes the
ETG modes, which therefore may limit the maximum gradient. transport is indispensable. Studies of particle confinement and
This has been analysed for electron ITBs obtained in Tore transport have been performed from the earliest investigations
Supra [343]. This latter possibility is supported by simulations of confinement in fusion relevant plasmas. However, this
suggesting that ETG driven turbulence is required to keep the understanding is still limited compared with that of energy
gradient close to the experimental value [20]. Transport studies confinement and transport due to conditions peculiar to particle
indicate that the electron heat diffusivity χe drops by almost confinement and transport, such as the existence of two kinds
an order of magnitude across an electron ITB. Typical values of dominant particle source (central and edge sources), and
for χe are 1–4 m2 s−1 outside of the barrier and 0.5–0.1 m2 s−1 the important role of a convection in addition to diffusion.
inside of the barrier [327, 328, 334–336]. This value is low but Recently, understanding of particle confinement and transport
remains at least one order of magnitude higher than the electron has been strongly enhanced by establishing experimental
neoclassical value. Therefore turbulent transport is not fully and analytical methods in conjunction with theory-based
suppressed, in contrast to what is observed in ion ITBs. modelling (see also sections 2.1 and 2.2). An anomalous
In electron ITBs with dominant electron heating the ion particle pinch at low collisionality has been observed in a
temperature remains much lower than the electron values number of devices and explained in the framework of the
and Te /Ti can be as high as 30. The Ti profiles do not ITG/TEM transport theory. In improved confinement plasmas
exhibit signs of an ITB. Attempts to increase Ti adding NBI with internal transport barriers (ITBs), impurity accumulation
heating to plasmas with an established electron ITB lead to due to reduced turbulent transport and a strong neoclassical
the decrease in the strength of the electron ITB [333], as inward pinch driven by the large density gradient is recognized
illustrated in figure 29. With sufficient ion heating an ion ITB as one of the largest concerns for applying ITB plasmas to a
is formed whereas the electron ITB seems to disappear. In fusion reactor. Suppression of impurity accumulation has been
one single case, clear electron and ion ITBs could be produced developed using central electron heating, and it is understood
simultaneously [324]. In general, plasma density also leads to a in terms of the combined effects of neoclassical impurity
decrease in the electron temperature and eventually the electron transport and turbulence-driven transport.
ITB disappears. These effects are not well documented and
incompletely understood so far. 3.4.1. Particle confinement times with consideration of central
Electron ITBs can be sustained in steady-state with an and edge particle sources. As noted above, the understanding
adequate shaping of the current profile as demonstrated on of particle confinement has lagged behind the understanding of
one hand, with LHCD in JET [328], JT-60U [344] and Tore energy confinement chiefly due to existence of two dominant
Supra [335], and on the other hand, with fully non-inductive particle sources, i.e. central fuelling by NBI and edge fuelling
current drive based on a combination of off-axis co-ECCD and by gas puffing and recycling. In order to improve this
central co-ECCD or counter-ECCD in TCV [338, 339]. situation, a scaling law for the total number of ions in the main
The plasmas with electron ITBs do not yield performances plasma of JT-60U ELMy H-mode plasmas was proposed that
particularly relevant for future burning plasmas, but they uses separately defined confinement times for central fuelling
demonstrate the possibility for steady-state and control of and edge fuelling [345]. The confinement time increases
barriers. In addition, they yield physics understanding which with density for core fuelling and decreases with density for
will be useful to assess theory. They show that strong rotation edge fuelling. However, the particle confinement exhibited
is not required and that current shaping is essential. Further a different dependence in reversed magnetic shear plasmas
useful issues deal with investigations on the cause of a barrier with an internal transport barrier. The scaling also enabled
(change in threshold or real turbulence stabilization), cause the discussion of density controllability by considering the
of the residual transport in the barrier and effect of Te /Ti on different effects on the particle balance of fuelling and divertor
achievement of electron ITBs. Understanding the physics pumping. The same method was applied to DIII-D ELMy
reasons may be useful in obtaining ion and electron ITBs H-mode plasmas, where a stronger density dependence was
simultaneously, which is a key issue for advanced tokamak obtained compared with the JT-60U plasmas [346]. For a more
scenarios, particularly in burning plasma with strong electron systematic understanding of particle confinement, a database
heating and no momentum source. should be accumulated that includes many machines.

S52
 Chapter 2: Plasma confinement and transport

Figure 30. Tritium diffusion and convection coefficients measured in the JET trace tritium experiments in comparison with the neoclassical
predictions for the high-density ELMy H-mode (left column) and for a discharge with strong internal transport barrier (right column).
Shaded areas indicate the confidence limits for the fitted quantities. Reprinted with permission from [350].

3.4.2. Estimation of local particle transport coefficients. shear, no sawteeth), increasing triangularity and plasma current
Local particle transport has been analysed to understand the increased particle confinement time, which can be explained
physical mechanisms responsible for the shape of the density in terms of a reduction in DT , but transport remained in
profile. In the analysis of many cases, an inward pinch is excess of neoclassical values. Comparing different regimes
introduced to explain a peaked density profile in the absence (ELMy H-mode, ITB plasma, and hybrid scenarios) outside
of a central particle source. The particle flux is expressed the central plasma region (0.65 < r/a < 0.80), the tritium
as  = −D∇n + nV , where D is the particle diffusivity diffusion coefficient (DT /Bt ) scaled in a manner closer to
and V is the convection velocity. A negative value of V gyro-Bohm (∼(ρθ∗ )3 , where ρθ∗ = qρ∗ ), but with an added
indicates an inward particle pinch. The existence of the inverse β dependence. In contrast, for ELMy H-mode
convection term makes estimation of the particle diffusivity discharge pairs with all dimensionless parameters except ρ∗
difficult, because perturbative techniques, such as modulated kept constant, gyro-Bohm behaviour was confined to the inner
gas puffing and density profile evolution, are then necessary part of the plasma (r/a < 0.4), and the outer plasma behaved
to estimate separately the D and V . In DIII-D, particle more like Bohm (∼ρ∗2 ). Similar dimensionless parameter
transport coefficients were estimated for different operating scans established contrasting trends for particle confinement
modes by analysing the temporal evolution of the electron (increases with ν∗ and β) and energy confinement (decreases
density profile with modulated gas puffing and immediately with ν∗ and independent of β) resulting in strong variation of
after the L–H transition [347]. The value of D increased the DT /χeff ratio, i.e. between 0.3 (at high density and low
with radius in the L-mode and ELMy H-mode plasmas and q95 ) and 2.0 (at low density and high q95 ). This result seems
D at the edge for the L-mode plasma was about twice that contrary to results obtained in ASDEX Upgrade [206,353] and
for the ELMy H-mode plasmas. In the ELM-free H-mode DIII-D [354]. In ASDEX Upgrade, strong linkage between
plasmas, D decreased rapidly outside a normalized radius of particle and energy transport as D ∝ χ (typically D/χ =
about 0.8 and an inward pinch velocity was observed. The 0.15–0.25) was observed. In DIII-D, it was experimentally
trace tritium transport experiments performed in TFTR [348] observed that the particle diffusivity and the thermal diffusivity
and JET [349] also provided valuable data for estimating D do not differ greatly and have roughly the same radial
and V . In these machines, an inward pinch velocity was also dependence. In JT-60U weak positive shear and reversed
observed. In TFTR, the tritium diffusivity in the central region shear plasmas, the electron effective diffusivity, defined only
of enhanced reversed shear plasma was smaller than that in the considering the diffusion term, was well correlated with the
central region of reversed shear plasmas. However, the inward ion thermal diffusivity in the ITB region [208].
pinch velocity was almost the same for both plasmas.
Recent JET trace tritium experiments provided thermal 3.4.3. Mechanisms of inward pinch. Neoclassical transport
tritium particle transport coefficients (DT , VT ) and their predicts a particle pinch driven by the toroidal electric field (i.e.
dependence on dimensionless parameters in the wide variety Ware pinch). In ASDEX Upgrade high-density plasmas, the
of plasma operating regimes [350–352]. The values of behaviour of the density profiles was described well with the
DT and VT were found to substantially exceed neoclassical assumption of D ∝ χ (typically D/χeff turb
= 0.15–0.25, where
exp exp
values in all regimes except in ELMy H-modes at high χeff = χe + χi − χi ) and an inward pinch of the order
turb neo

density, and in the region of ITBs in RS plasmas as shown of the Ware pinch [206, 353]. In JET, a long timescale density
in figure 30. In hybrid scenarios (qmin ∼ 1, low positive peaking was observed, leading to plasmas with densities

S53
E.J. Doyle et al

(a)

(b)
Figure 32. Density peaking, defined as ne (ρ = 0.4)/ne (ρ = 0.8)
Figure 31. 1D simulation of discharge No 30428 with LHCD in
versus the line-average density (in 1020 m−3 ), panel (a), and versus
Tore Supra, at t = 30 s: (a) density profile (line: simulation; circles:
νeff , panel (b), for the subset of stationary plasmas in the ASDEX
reflectometry measurements) and particle source profile; (b) pinch
Upgrade H-mode database, with total NBI heating power of 5 MW.
velocity (squares) and diffusion coefficient (diamonds) used to
The effective collisionality, νeff , is defined as νeff ≡ νei /ωDe where
reproduce measured density profile, Vneo given by NCLASS
νei is the electron-ion collision frequency and ωDe is the curvature
(triangles), Vneo when assuming Zeff = 6 instead of 2 and iron
drift frequency, ωDe ≡ 2k⊥ ρs cs /R . Reprinted with permission
impurity only (dashed line/triangles). Reprinted with permission
from [360]. © 2003 American Institute of Physics.
from [357]. © 2003 American Physical Society.

exceeding the Greenwald density, nG (1020 m−3 ) = Ip /π a 2 ratio of V /D is independent of collisionality [362]. The
(MA,m). The value of D ∼ = 0.25χeff and the Ware pinch transport analysis in JET also indicated that the anomalous
gave an acceptable fit to the measured density profile [355]. pinch was of the order of the Ware pinch, and the discharge
While other analyses from JET indicated that an anomalous with lower collisionality showed flatter density profiles in H-
pinch seems to be necessary, at least for L-mode plasmas [356]. mode plasmas, while the effect of the anomalous pinch tended
Clear experimental evidence for the existence of an anomalous to increase at low collisionality in L-mode plasmas [356]. The
inward pinch was shown in Tore Supra [357] and TCV [358], dependence of the anomalous inward pinch on dimensionless
where peaked density profiles without central fuelling were parameters should be understood systematically as discussed
observed with zero loop voltage. Figure 31 shows a density above for the particle diffusivity in the future work.
profile in a stationary, fully relaxed discharge in Tore Supra
with current fully driven by the LH waves. To explain the
shape of the experimental density profile, a particle pinch 3.4.4. Model of anomalous particle transport. Theory-
velocity of 2 orders of magnitude above the neoclassical value based transport models have been developed to explain the
is required [357]. experimental observations (also see sections 2.1 and 2.2).
One of the candidate explanations to the puzzle of what The experimental observation that the density peaking factor
conditions lead to the anomalous pinch was proposed in increases with decreasing collisionality was explained with a
analysis of ASDEX Upgrade: collisionality plays a relevant theory-based fluid transport model for ion temperature gradient
role in determining the density peaking [359,360]. The density and trapped electron modes, GLF23 [20]. It was shown that the
peaking measured in ASDEX Upgrade H-mode plasmas was anomalous particle inward pinch decreases with collisionality
shown to decrease with increasing collisionality as shown and the relative role of the Ware pinch becomes important at
in figure 32. Analysis of an extensive database of JET H- high collisionality. Other models have also been proposed
mode plasma density profiles showed that the density peaking to explain the anomalous particle inward pinch. Numerical
factor increases as the plasma collisionality decreases [361], simulations have confirmed that a turbulent particle pinch
which confirms the observation in ASDEX Upgrade. However, exists, which is mainly driven by magnetic field curvature and
the transport analysis in ASDEX Upgrade indicated that the thermodiffusion [124]. In TCV, the best overall agreement

S54
 Chapter 2: Plasma confinement and transport

with a database was obtained with models combining an

anomalous pinch mechanism, such as the curvature pinch,
which is proportional to the gradient of the safety factor, with
the Ware pinch [358]. The multi-mode model, in which the ion
temperature gradient modes, trapped electron modes and drift-
resistive ballooning modes as well as smaller contributions
from kinetic ballooning modes are included, provided a good
match to density profiles in TFTR plasmas [238]. A canonical
profile depending on the ratio of the diffusion coefficients for
passing and trapped particles was compared favourably with
the experiments in DIII-D [363]. When the sources can be
neglected, a crucial parameter determining the shape of the
density profile was the relative transport between the passing
and trapped particles. In Alcator C-Mod plasmas without a
central particle source, the outflow was driven by the TEM
turbulence in the internal transport barrier and it was balanced
by the inward Ware pinch, leading to steady-state [123]. Many
models have been proposed as described above, but a clear
answer is still missing for mechanisms driving the anomalous
transport.

3.4.5. Effects of density peaking on confinement. One of the

main physics issues for ITER standard operation is sustaining
good H-mode confinement in high density close to Greenwald
density. Several methods have been discovered as described in
section 2.7 of chapter 4 of this issue [364]. Related to particle
transport, good H-mode confinement has been obtained with
peaked density profiles in the high-density region. Figure 33
shows a DIII-D discharge with a good H-mode confinement,
HL89 ≈ 2 (i.e. defined relative L-mode scaling), where a Figure 33. (a) Time behaviour of the plasma stored energy W ,
continuous rise in the line-averaged density up to n̄e ≈ 1.4nG normalized energy confinement time relative to ITERL-89P scaling,
was obtained at low constant gas puff rate [365]. The pedestal gas puff rate, line-averaged and pedestal density normalized to the
density in this discharge quickly saturated at nped ≈ 0.8nG , and Greenwald density, and amplitude of a 3/2 MHD mode, as measured
by Mirnov coils, in high-density DIII-D discharge. (b) Density
further n̄e increase was entirely due to density peaking. Similar profiles at t = 2 s (just after the onset of gas puffing) and at t = 4.5 s
discharges have been obtained in ASDEX Upgrade [353] and (near the end of discharge). Reprinted with permission from [365].
JET [355]. The slow density peaking considered as a main
contributor to improved confinement at high density seems to
of the ITG mode growth rate due to a Zeff increase and
be a result of the neoclassical particle pinch described above.
a peaking of the density profile caused by an anomalous
In JT-60U weak and reversed shear plasmas, high confinement
particle pinch driven by the dissipative trapped electron (DTE)
was achieved at density above the Greenwald density by
mode in combination with the effect of the E × B shearing
tailoring the density ITB [366]. Stable H-mode operation
rate [213, 371].
beyond the Greenwald density was obtained with high field
side pellet injection in ASDEX Upgrade [367]. Deeper fuel
penetration was achieved with high field side injection due to a 3.4.6. Prediction of ITER density profile. The importance
rapid movement of the ablatant towards the outer major radius of the inward pinch increases in a fusion reactor due to low
attributed to a vertical curvature and ∇B drift current induced central particle fuelling. In the present ITER design, a flat
inside the ionized ablated material. The deep pellet fuelling density profile is usually assumed [372]. However, as shown
allows higher central densities to be achieved compared with in figure 34 [373], moderately peaked density profiles due
those with strong gas puffing and similar edge densities. See to the anomalous inward pinch can be expected in ITER
section 2.7.3 for more discussion of pellet fuelling and its according to predictions of the GLF23 transport model and the
impact on divertor operation. theory of turbulent equipartition (TEP) [374, 375]. If a peaked
Density peaking is effective for achieving high average density profile is obtained in ITER due to an inward pinch
density with relatively low edge density. However, its even with edge particle fuelling, higher fusion gain will be
direct effects on confinement improvement are not clear in achieved. However, strongly peaked density profiles in ITER
the above discharges. The clearest effects were found in should be avoided since they may provoke an early onset of
impurity seeded plasmas. Enhanced confinement regimes with the neoclassical tearing modes (NTMs) or undesirable central
impurity seeding such as the RI-mode in TEXTOR [368] have accumulation of high Z impurities.
been demonstrated more recently in high-density discharges
in DIII-D [369] and JET [370]. The effect of impurity 3.4.7. Density profile control. Density profile control is
seeding on plasma confinement is explained by reduction important not only for burn control but also for suppression

S55
E.J. Doyle et al

along with strong ELM mixing was needed to simulate the

impurity peofiles. He, C, Ar and Ne in DIII-D H-mode plasmas
also showed no tendency to accumulate in any local region
when the electron density is relatively flat [381].
In more detailed studies, the temporal evolution of
impurity transport has been used to separately determine the
diffusivity and pinch contributions to the impurity fluxes as
a function of the charge. In ASDEX Upgrade [382] the
temporal evolution of impurity profiles between sawtooth
crashes showed radially increasing particle diffusivites for Ne,
Ar, Kr and Xe. The diffusivities and convective velocities
were close to neoclassical for the lower Z species (Ne and
Ar), while both the diffusivities and inward drift velocites were
Figure 34. Steady-state distributions of the plasma density, ne as a much stronger than neoclassical for the higher Z species. High
function of the minor radius in the reference ITER inductive toroidal rotation was not taken into account in the neoclassical
scenario (15 MA, Q = 10). Full curve shows the results of the calculations and may have influenced the behaviour. In
GLF23 model with the boundary conditions DIII-D [183], the transition from H-mode to VH-mode led to
ne,ped = 0.79 × 1020 m−3 and Te,ped = 5 keV at the pedestal top.
Results of the scaling-based model are shown as the broken line.
a transient response from peaked to hollow impurity profiles
The dotted line shows the prediction of the turbulent equipartition where the diffusivities and convective veolcites of various
(TEP) theory, where the shear profile is defined by the GLF23 impurites could be determined [383]. The flat electron
modelling. Reprinted with permission from [373]. density and strongly peaked ion temperature profiles in the low
turbulence VH-mode plasmas allowed a confirmation of the
of impurity accumulation as will be described later. In neoclassical ion temperature gradient screening contribution
ASDEX Upgrade [206], a strong linkage between D and χ (including its increasing strength with charge) to the outward
was observed in the work described above. The variations convective velocity, even though the diffusivity was strongly
in the heat flux profile have a strong effect on the χ-profile governed by turbulence. The temperature gradient screening
as well as on the D-profile due to the generally observed effect should have a favourable effect on higher Z impurity
self-similarity of the temperature profile. Consequently, on- profiles in low collisionality plasmas such as ITER even when
and off-axis heating led to flat and peaked density profiles, turbulent diffusivity dominates over neoclassical [384].
respectively. Particle depletion from the plasma core was The transport of impurities in enhanced confinement
observed with centrally deposited ECH and ECCD in TCV regimes, especially those with internal transport barriers
[376]. The correlation of density pumpout with the loss of (ITBs), is a critical issue in burning plasma regimes due to the
axisymmetry suggested that neoclassical transport processes possibility of excessive fuel dilution caused by accumulation
involving locally trapped particles near the helically displaced of impurities, including helium ash. Because the turbulence-
magnetic axis might account for the phenomenon. The driven transport in these regimes is relatively small, the
flattening of the density profile by ECH is commonly observed effect of collision-driven transport (i.e. neoclassical transport)
not only in tokamaks but also in helical devices, so studies becomes increasingly important and in some cases may
that develop a common physics basis for transport in toroidal dominate the transport properties. The effect of neoclassical
plasmas might help improve the understanding of mechanisms transport is particularly acute in regimes that have peaked
responsible for the density profile flattening by ECH. A similar density profiles due to the strong dependence of the impurity
particle depletion was observed in DIII-D quiescent double- convection on the background density gradient. To first order,
barrier (QDB) plasmas [377] and JT-60U weak positive shear neoclassical theory predicts that VZneo /DZneo to be strongly
plasmas with Ar accumulation [208] by applying central ECH, dependent on the ion density gradient and weakly dependent on
and in Alcator C-Mod [378] by applying central ICRF, which the ion temperature gradient: VZneo /DZneo ≈ gnD→Z ∇nD /nD +
were accompanied by impurity exhaust from the inside of the gTi ∇Ti /Ti wheregnD→Z and gTi are complex functions of the
ITB as discussed below. In T-10, a peaked density profile was impurity charge Z, plasma collisionality ν∗ , and magnetic
obtained after pellet injection even with ECH [379]. geometry. Generally, gnD→Z > 1 and gTi < 1 in the banana
regime of transport (i.e. ν∗ < 1) and |gnD→Z | |gTi |, with
3.4.8. Impurity transport in ELMy H-mode and enhanced both increasing in magnitude strongly with Z. There are
confinement regimes. Impurity transport in standard ELMy several ramifications apparent from this formula. First, in the
H-mode plasmas exhibits a variety of behaviours, which is cases with strongly peaked density profiles (and low levels
strongly influenced by the main ion density and temperature of turbulence-driven transport), preferential accumulation of
profiles and sometimes a function of the impurity charge, impurities in the core region would be expected regardless
plasma rotation and other parameters as well. Intrinsic carbon of the magnitude of the ion temperature gradient. Second,
and neon in JET ELMy H-mode plasmas [380] were found the degree of accumulation would be strongly Z dependent
to have similar transport properties and both had hollow with the higher Z impurities showing stronger preferential
profiles. Strong outward convective velocities in the core were accumulation. Third, in the cases with flat density profiles
consistent with neoclassical temperature gradient screening in but a strong ion temperature gradient, hollow impurity density
these plasmas where the electron density was relatively flat. profiles would result due to the outward transport associated
In the edge region, r/a > 0.8, an (anomalous) impurity pinch with ∇Ti . Finally, preferential accumulation of helium ash

S56
 Chapter 2: Plasma confinement and transport

in an even more strongly peaked impurity density profile. The

helium density has not been observed to be strongly peaked in
any of the experiments, because it is a low Z impurity. The
helium diffusivity and convection velocity estimated in JT-60U
indicated the possibility of sufficient helium exhaust even in
ITB plasmas [386]. For more discussion of He exhaust see
section 2.3 of chapter 4 of this issue [364].

3.4.9. Impurity control. The application of ECH inside

the ITB region has been shown to be effective in reducing
the level of impurity accumulation in experiments on JT-60U
[208] and ASDEX Upgrade [206]. The application of on-axis
ICRF heating is also shown to be effective in reducing the
density and impurity peaking in Alcator C-Mod [378]. In
both cases, the reduced impurity accumulation is attributed
to a reduction in the background density gradient, leading
to a marked decrease in the inward convection of impurities.
Figure 36 shows JT-60U case [389], where both electron and
argon density profiles become flatter during ECH injection.
The flattening of the argon density profile is consistent with
the reduction of the neoclassical inward convection velocity
Figure 35. (Upper figures) profiles of Ti , Te and q, (middle figures)
due to the reduction of background plasma density gradient. In
profiles of ni , ne , nHe and nC , (bottom figures) profiles of ne and nAr contrast, observations in enhanced confinement regimes with
in (a) reversed shear plasma (IP = 1.3 MA, BT = 3.7 T, flat density profiles indicate that neoclassical transport can play
q95 = 4.9–5.2 and H Hy2 ∼ 1.6) and (b) weak positive shear plasma a beneficial role in screening impurities from the core. In
(IP = 1.0 MA, BT = 2.0–3.8 T, q95 = 3.7–6.3 and H Hy2 ∼ 1.0). In particular, experiments in DIII-D VH-mode plasmas [383,390]
the bottom figures, ne is normalized at r/a = 0 and nAr is adjusted
to ne outside the ITB. Reprinted with permission from [208].
confirmed the existence of the ‘temperature screening’ effect
of neoclassical impurity transport as shown in figure 37,
including the predicted Z dependence. The observed outward
due to neoclassical transport is expected to be weak (but still convection led to extremely hollow carbon and neon density
present) due to its low Z. Note that this does not preclude profiles in these plasmas. In the enhanced confinement
accumulation of helium ash due to differences in the sources regime in JET impurity seeding plasmas [370] with high
of helium and the fuel mix. triangularity configuration and continuous D2 puffing, an
Experiments tend to support the predictions of neoclassi- outward impurity convection velocity was also observed,
cal impurity transport in regimes with enhanced confinement. which features flat or slightly hollow impurity profiles and
However, there are few instances in which both the measured high radiation from a narrow region at the edge [391]. In
particle diffusivity and convective velocity are found to be con- DIII-D argon seeded plasmas, a hollow Zeff profile was
sistent with neoclassical predictions. The earliest of the stud- observed [369].
ies on ITB plasmas was done on TFTR using perturbative gas
injection to measure DZ and VZ simultaneously [348, 385].
3.4.10. Impurity transport in burning plasmas. Observations
The inferred diffusivities for deuterium, helium and carbon
to date are consistent with a transport model that combines
were found to be consistent with neoclassical predictions,
turbulence-driven and collisional transport in a linear fashion:
within the uncertainties of the measurement and theoretical
prediction. Subsequent studies in JT-60U RS plasmas [386], Z = −(DZturb + DZneo )∇nZ + nZ (VZturb + VZneo ). (9)
JET strong ITB plasmas [205, 387] and DIII-D QDB plas-
mas [388] showed the strong Z dependence of VZ predicted by Theoretically, this is valid since the particle transport due
neoclassical theory even at relatively low Z (helium to neon). to fluctuations results from collective mechanisms that have
In the JT-60U (He) and JET cases, the measured DZ is found to structures much larger than the characteristic scale length for
be consistent with neoclassical predictions in the ITB region collisional momentum exchange between particles, allowing
while in the DIII-D and JT-60U (C and Ar) cases the mea- one to treat the processes separately. Because of the extremely
sured DZ is significantly higher than the neoclassical predic- low value of DZneo in most cases, DZturb is expected to dominate
tion. In all the cases, either strong accumulation or long con- the total diffusivity. However, the strong Z dependence of
finement times of higher Z impurities (argon in JT-60U, nickel VZneo can make it such that collisional effects dominate impurity
in JET and nickel and calcium in DIII-D) are observed. Typi- convection in certain situations. Hence, it is possible to observe
cal profiles of electron and ion temperatures, safety factor and effects of neoclassical transport even in plasmas in which
densities of electron, helium, carbon and argon in JT-60U are DZturb > DZneo and also to have impurity density profiles that are
shown in figure 35 for (a) reversed shear and (b) weak positive significantly different from the profile expected purely from
shear plasmas with ITB. Note that all of these regimes have either turbulence-driven or neoclassical transport. This has
strongly peaked electron (and, hence, deuterium ion) density been shown to be the case in analysis of experiments on DIII-
profiles, which according to neoclassical theory should result D [383] and JET [205]. Applying this type of model to burning

S57
E.J. Doyle et al

Figure 36. Argon exhaust with EC heating in a high βp mode plasma. Profiles of (a) electron density ne , (b) argon density nAr , (c) electron
temperature Te and (d) ion temperature Ti before and during EC heating. Profiles before EC heating are denoted by solid lines (with open
symbols), while those during EC heating are denoted by dotted lines (with closed symbols). Reprinted with permission from [389].

NCS with L—mode Edge VH—mode plasma conditions, it has been shown that the combination
20 (a) ne
20
(b) of reduced turbulence-driven transport and peaked density
profiles (typical of ITB plasmas) lead to stringent constraints
(× 1019 m–3, keV)

(× 1019 m–3, keV)

Ti
on the allowable level of low Z impurities for ignition [390].
10 In contrast, flat density profiles are found to be beneficial,
10
allowing a significantly higher impurity level than their ITB
counterparts.

0 0
3.4.11. Summary. Significant progress has been achieved in
0.6 (c) 0.4
Helium (d) experimental studies and theory-based modelling of particle
Carbon and impurity transport. Since density profile has a large
(× 1019 m–3)

(× 1019 m–3)

0.4 Neon
impact on the plasma performance in a fusion reactor, further
0.2 systematic understanding of particle confinement and transport
0.2 is required. The dependence of particle transport coefficients
on dimensionless parameters should be investigated for
0.0 0.0 systematically understanding collisionality dependence of the
0.6 0.4 density peaking. Recent studies indicate that the density profile
(e) Helium (f) could be peaked even with low central fuelling for ITER
Carbon
standard operation due to the turbulent driven inward pinch.
(× 1019 m–3)
(× 1019 m–3)

0.4 Neon
On the other hand, in ITER steady-state operation with reduced
0.2 turbulent transport, it is important to investigate whether a
0.2 peaked density profile and a density ITB can be obtained
under the reactor-relevant conditions of low central fuelling.
0.0 0.0 Furthermore, the optimum density profile for achieving high
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 fusion gain without strong impurity accumulation should be
Normalized Radius Normalized Radius investigated together with the establishment of control methods
Figure 37. (a)–(b) Measured electron density and ion temperature for the density profile and impurity accumulation in the
profiles, (c)–(d) measured helium, carbon and neon density future work.
profiles (normalized ) and (e)–(f) computed helium, carbon
and neon density profiles using the transport model in an NCS 3.5. Toroidal momentum transport and spontaneous rotation
and VH-mode discharge in DIII-D. Reprinted with permission
from [390]. Experiments and simulations of toroidal momentum transport
in tokamaks have demonstrated that transport of toroidal

S58
 Chapter 2: Plasma confinement and transport

rotation is determined by mechanisms similar to those driving

particle and energy transport [392, 393]. Toroidal momentum 10 6467 6468 experiment
transport is generally found to be anomalous, with χφ much 6467 6468 code
5

Toroidal velocity (km/s)

larger than neoclassical values. Moreover, the viscosity 6468 code, small puff
coefficient, which is responsible for the radial transport
of toroidal rotation, is reduced inside ITB similar to the 0
diffusion and thermal heat conductivity coefficients. This
-5
was demonstrated by many simulations using various transport
models for DIII-D [394, 395], JET [396, 397], Alcator C-Mod
-10
[398], MAST [393, 399] and other tokamaks. We briefly
discuss the response to an applied torque, and then address the
-15
issue of rotation in the absence of a strong torque in more detail
because of its greater relevance to burning plasmas like ITER. -20 core separatrix SOL
The behaviour of toroidal rotation in the vicinity of an ITB
is of particular interest because of its influence on triggering 1.24 1.28 1.32 1.36 1.40 1.44 1.48
and/or sustaining the barrier. Radial coordinate Y, (m)
Figure 38. Toroidal velocity in MAST at the outer midplane for
3.5.1. Transport in the presence of strong applied torque. shot Nos 6467 (outboard puff) and Nos 6468 (inboard puff).
The toroidal rotation induced by NBI in machines such Simulations were performed with the B2SOLPS5.0 transport code.
as JET, DIII-D, JT-60U and ASDEX Upgrade to a large Reprinted with permission from [414].
extent determines the radial electric field in the core, see
e.g [333, 397]. The toroidal rotation profile hence is an In [414] a mechanism for toroidal rotation generation in
important element in ITB formation. It may influence transport the edge plasma by inboard/outboard gas puff was suggested.
coefficients through creation of shear in the E × B drift and The inboard gas puff leads to more counter-current directed
suppression of turbulence levels. This is similar to the situation toroidal rotation at the outer midplane. It is demonstrated that
that one might expect for burning plasma conditions. Indeed, counter-current toroidal torque can be generated by inboard gas
even without NBI heating one might obtain strong enough puff by creating inboard/outboard parallel fluxes associated
toroidal rotation in the core to increase the radial electric field. with the ionization source on the closed field lines. These
fluxes are transported by the vertical ∇B ion drift providing
counter-current acceleration. Simulations performed for
3.5.2. Spontaneous rotation. Toroidal rotation in the absence
MAST by B2SOLPS5.0 transport code are consistent with
of NBI (in Ohmic or ICRF heated plasmas) has been observed
experimental observations, as shown in figure 38. The positive
in Alcator C-Mod [398, 400–402], MAST [403, 404], JET
direction here corresponds to the counter-current direction, so
[405, 406], Tore Supra [407, 408] and other machines. It
indeed for the case of inboard gas puff the toroidal rotation is
is especially large in the H-mode. The toroidal rotation is
more positive than for the case of outboard gas puff.
co-current directed, and according to [398] the change of
In [415] the effect of ITG modes on the generation of
toroidal rotation after the L–H transition is Vφ = kW/Ip ,
toroidal rotation has been considered. It was shown that a
where k ∼ 0.1 , Vφ is in km s−1 , the stored energy W is
specific feature of the ITG modes, i.e. the dependence of
in kJ, plasma current Ip is in MA. For Tore Supra [407] the
the frequency on the radial toroidal velocity gradient, causes
coefficient k is a factor of 2 less, possibly related to the relative
inward transport of toroidal momentum at the edge plasma.
machine size. On MAST the direction of the toroidal rotation It is not, however, quite clear whether the ITG modes are
changed from the co-current to the counter-current direction excited in the edge region, since it is possible that the parameter
when switching from inboard to outboard gas puffing. Strong ηi = d ln Ti /d ln n is not large enough within the edge transport
counter-current rotation has been observed in DIII-D plasmas barrier. An inward pinch of the toroidal momentum was
with ECH [409], which further demonstrates the possibilities also predicted in [416]. The convective term was calculated
of rotation drive in the absence of external momentum input. using a combination of the quasi-linear and neoclassical
The rotation is counter-current in the centre and co-current approaches for the banana regime. This convective term
near the plasma edge, and depending upon the ECH resonance exists independently of the type of the electrostatic turbulence.
location, the velocity shear profile can be controlled. Since both models [415] and [416] predict an inward pinch of
Several mechanisms might be responsible for the the toroidal momentum, the resulting steady-state profile of
generation of the toroidal rotation in the absence of external the toroidal velocity should depend on the steady-state density
NBI torque, and several theoretical models have been put and temperature profiles, i.e. on the edge values of the density,
forward. In [410, 411] the neoclassical torque in the temperatures and toroidal rotation.
Pfirsch–Schlüter regime has been calculated. The role of Another attempt to explain the co-current rotation
neutral viscosity in the framework of neoclassical theory observed in ICRF heated H-mode discharges is through the
has been considered in [412, 413]. However, since from toroidal torque provided by the radial electric field arising
many experiments it is known that the toroidal momentum due to orbit shifts of high energy ions generated by ICRF
is transported in the radial direction with transport coefficients waves [417–420]. A particular prediction of some of these
similar to those for particles and energy, the pure neoclassical theories [418, 420] is that the rotation should switch direction
approach might be not sufficient. to counter-current with the ICRF resonance located on the

S59
E.J. Doyle et al

Figure 39. The central toroidal rotation velocity in Alcator C-Mod

during the L-mode portion of LSN (green dots), DN (purple Figure 40. Er and its components 0.6 s before (a) and 0.6 s after (b)
diamonds) and USN (red asterisks) 0.8 MA, 5.4 T discharges with the ITB formation. The main component Er,φ is proportional to the
nc = 1.4 × 1020 m−3 as a function of distance between the toroidal rotation velocity. The footpoint of the ITB is shown by the
separatrices SSEP is shown in the top frame. In the bottom is the vertical dashed line in (b). Reprinted with permission from [397]. ©
minimum ICRF power required to induce L–H transition as a 2001 Institute of Physics.
function of SSEP. Reprinted with permission from [424].

high magnetic field side, which was not observed in the

experiments [398].
Modelling of the toroidal rotation profile in ICRF and
Ohmic H-mode discharges was performed for Alcator C-Mod
without momentum input in [421, 422]. A simple model was
used
 
∂nmi Vφ 1 ∂ ∂ Vc r
+ r −Dφ (nmi Vφ ) − nmi Vφ = 0.
∂t r ∂r ∂r a
(10)
The toroidal momentum diffusivity Dφ was ∼0.2 m2 s−1 in
the L-mode and ∼0.07 m2 s−1 in the H-mode. In ELM-free
H-mode the profiles were consistent with an inward convection
velocity of the order of 10 m s−1 . The values of the transport
coefficients were significantly larger than the neoclassical
predictions.
In the Alcator C-Mod experiments [421,423,424] the role
of the edge plasma in the formation of the central toroidal
Figure 41. The plasma stored energy, impurity toroidal rotation
rotation in the absence of the toroidal momentum torque velocity at three radii and the edge Dα brightness for an off-axis
has been emphasized. When switching from lower single ICRF heated ITB discharge. Reprinted with permission from [421].
null to the upper single null divertor configuration (with
the ion B × ∇B
 drift downward), the edge toroidal rotation
changed sign from co-current to the counter-current direction JET [397]. One can see the rise of the E × B shear in the
[424]. This is illustrated by figure 39, where the central barrier vicinity. Without NBI the toroidal rotation profile was
toroidal rotation in the L-mode of Alcator C-Mod is plotted measured on Alcator C-Mod, figure 41 [421]. The hollow
versus the distance between two separatrixes. This seems to profile in the centre indicates the presence of a negative electric
play a role in the higher H-mode power threshold for upper field in the vicinity of the barrier.
single null discharges. Similar indications were observed on There is concern for ITER that the possibility of no
MAST [403, 404]. external torque from neutral beam injection will preclude any
of the beneficial effects desired from rotation and velocity
3.5.3. Toroidal rotation in the vicinity of an ITB. In the shear: ITB formation and resistive wall mode suppression.
presence of an ITB a drop in toroidal rotation is usually While the mechanism responsible for the generation of
observed [396,425]. An example of the toroidal rotation profile spontaneous rotation observed on many devices is not well
with an ITB in the presence of NBI is shown in figure 40 for understood, it is promising that ICRH and ECH may provide

S60
 Chapter 2: Plasma confinement and transport

control knobs for rotation and velocity shear production, and in terms of ratios of like-dimensional quantities such as aspect
allow ITB formation without NBI. ratio or elongation. The plasma safety factor also belongs to
this latter category. Finally, since the plasma is composed of
3.5.4. Conclusions. As presented above there are many multiple species of particles, the ratios of the quantities among
effects in plasma rotation generation and transport that cannot the species such as mass and charge can appear, as can ratios of
be explained with existing theoretical models. Therefore we the various moments of the distribution between species, such
can conclude that our present understanding of the transport as the ratio of the electron temperature to the ion temperature.
of toroidal momentum is incomplete and the mechanisms of It is not possible to obtain plasmas in two different
generation of toroidal rotation in the absence of NBI are not tokamaks in which all of these dimensionless parameters are
completely understood. Further investigations are necessary, fixed. However, the parameter ND is typically 109 larger than
but there is promise for velocity (and shear) generation in the other dimensionless parameters. Therefore, it is assumed
that mismatches in this variable are ignorable for transport
future devices without external momentum input. Scaling and
physics, which is equivalent to ignoring very fast or very small
parameter studies of spontaneous rotation are ongoing on a
scale effects. To validate this assumption, so-called ‘identity’
variety of devices with the goal of extrapolation to ITER.
discharges have been made between pairs of tokamaks. These
discharges match the remaining dimensionless parameters
3.6. Dimensionless parameter scaling experiments (other than ND ) and then test whether the measured transport
Most experimental investigations of transport seek to isolate scales in a manner predicted by the dimensional analysis, given
the dependence of the transport on one or more of the the choice of dimensionless parameters.
control (or ‘engineering’) variables that can be set by the The approach outlined above has now been validated in
experimenter, such as the toroidal field or the plasma current. both L-mode and H-mode experiments [429]. An example
However, the underlying equations believed to govern the for H-mode plasmas in DIII-D and JET is shown in figure 42.
Over the region where the dimensionless parameters are well
plasma behaviour are sensitive only to algebraic combinations
matched, 0.35 < ρ < 0.85, the measured energy transport
of these engineering variables that can be cast in dimensionless
scales as predicted by the dimensional analysis. The same type
form [426]. Therefore, posing experiments where these
of match was obtained in L-mode plasmas between Alcator
dimensionless parameters are varied individually may have
C-Mod and DIII-D [429]. It must be emphasized that all
significant benefits in both understanding and projection to
the engineering parameters are different in these experiments,
future devices. For example, different models for plasma
yet the dimensionless parameters which describe the intrinsic
turbulence yield significantly different scalings with respect
properties of the plasmas are held fixed. The mere fact
to dimensionless parameters [426]. Therefore, measurement
that the profiles can be made to agree, with control only
of these scalings may eliminate large classes of potential
over global quantities, is a substantial confirmation that the
candidates to explain anomalous plasma transport. Another
variables chosen are an appropriate set to describe the plasma.
advantage of this approach is that the projection to burning The match of the properly normalized diffusivities indicates
plasma experiments from present-day transport experiments that plasma energy transport can be described by means of
can be reduced to a one-parameter extrapolation, compared algebraic combinations of these variables. Furthermore, any
with the standard engineering variable approach with five or quantities of importance that were not considered are either
more variables [427]. In this section, recent experimental matched fortuitously or the dependence of the transport on
results and interpretation will be discussed. The use of the these quantities is sufficiently weak that the mismatch can be
dimensionless scaling approach to prediction of transport in ignored. For example, the mismatches in toroidal rotation and
future devices will be discussed in section 5.4. Comparison Zeff , shown in figure 42, are either insignificant or lead to
of statistical analysis of the international global energy coincidental offsetting errors.
confinement database with the experimental determinations of A case in which the chosen variables do not describe the
confinement scaling using dimensionless parameters will be plasma behaviour has been published recently [430]. Many
discussed in section 5.3. authors have quoted confinement quality degradation as a
The choice of dimensionless parameters is clearly a function of the proximity to the empirical density limit (nG ≡
crucial step in this approach to understanding transport. The Ip /π a 2 ), where nG is the line-averaged density in units of
formal theory of dimensional analysis [428] tells how many 1020 m−3 , Ip is the total plasma current in MA and a is the
parameters are required and sets the algebraic form they must midplane minor radius in m. It has been proposed that n/nG ,
take. However, any linearly independent combination is also along with ρ∗ and β, might be a better dimensionless set than
suitable. The approach adopted in most experiments is to the one described above with ν∗ . In order to test this hypothesis,
choose variables that have physical significance [426]. The identity discharges between JET and DIII-D were found. The
standard set of variables includes the particle gyro-radius first observation was that it was not possible to match the
normalized to the minor radius of the plasma (ρ∗ ), the ratio profiles of the normalized plasma parameters with this choice
of the kinetic energy density to the magnetic energy density of variables (see figure 43) [430]. Given this mismatch, it is
(β), the collision time normalized to the particle transit time not surprising that the global energy confinement did not match
(ν∗ ), the plasma flow velocity normalized to the ion sound the identity constraints. Because the proposed variable n/nG
wave velocity (M) and the number of particles in the Debye contains integral quantities rather than purely local quantities,
sphere (ND ). These parameters uniquely specify the plasma this failure to realize a match may not be too surprising. It
conditions. The device geometry also may be important for seems inappropriate, however, to reverse the argument in order
transport, but the parameters defining the geometry are written to refute that the density limit is related to transport [431].

S61
E.J. Doyle et al

Figure 42. Comparison of the fitted or inferred scaled parameters for identical ELMing H-mode discharges in DIII-D (solid lines) and JET
(dashed or shaded lines) as a function of normalized radius: (a) scaled density (b) scaled temperature, (c) scaled toroidal rotation, (d) Zeff
and (e) scaled one-fluid diffusivity (with uncertainties). Reprinted with permission from [429].

Significant new work on the scaling of energy transport the ideal limit, a sharp increase in transport would be expected
with β has been published. Earlier work on DIII-D [225] and due to the onset of resistive instabilities below the ideal limit. In
JET [228] indicated that energy confinement was practically addition, the increasing magnetic well with increasing β (due
independent of β in both L-mode and H-mode plasmas. to the Shafranov shift) would also be stabilizing. Therefore,
Analysis of the international database of global energy one might expect a favourable β dependence on transport up
confinement in L-mode and H-mode plasmas consistently to a significant fraction of the ideal well limit, followed by a
yielded a strong degradation of energy confinement with β. sharp increase in transport close to the ideal limit. Clearly, a
The β scaling does not affect the projection to ITER using more complete understanding of these effects on the plasma
the dimensionless scaling method, since β in the two cases turbulence causing the energy transport is needed.
should be the same. However, the β scaling has a significant Recent experiments have also shed light on the role of
impact on the optimization of tokamak fusion performance shear in the bulk E × B rotation. As mentioned above,
[432]. This is discussed further in section 5.4. In order to the toroidal Mach number of the ion flow, M, is one of the
resolve the discrepancy between the database analysis and the dimensionless parameters used to characterize the plasma.
experiments, confinement scaling scans in H-mode have been Experiments in TFTR showed that M does not have a
extended to a larger range in β. The independence of transport significant effect on transport [435]. However, experiments
from β over the range from just above the L–H threshold up and theory have indicated that the shear in the bulk E × B
to 90% of the ideal ballooning limit has been verified with rotation can have a significant effect on the transport [34,245].
multiple point scans [432, 433]. The experimental database is From radial force balance, the E × B velocity has a pressure
summarized in figure 44, along with the β scaling implicit in gradient term, which scales as 1/ρ∗ , and a rotation term which
the IPB98y, 2 confinement scaling [2]. Recent work on the scales like M [220]. In DIII-D, two ρ∗ scans were performed—
method of analysis of the global database [434] has improved one with counter-NBI and the other with co-NBI. The ρ∗
the agreement of the database result with the experimental data scaling obtained is different in the two cases, with the main
(see section 5.3). difference in the measured profiles being the quality of the
The complete independence of confinement on β match of M in each case (figure 45). Using the GLF23
is somewhat surprising from theoretical considerations. model including E × B velocity shear by a linear, no-threshold
Increasing β increases the coupling between drift waves and model, the difference in scaling is shown to be consistent with
Alfvén waves. For β well below the ideal ballooning limit, this the lack of match in M. Furthermore, the effect of E × B
increased coupling should be stabilizing to drift modes, since velocity shear only affects the ion transport scaling and not the
the electromagnetic wave is robustly stable. As β approaches electron scaling. This has serious implications for theoretical

S62
 Chapter 2: Plasma confinement and transport

Figure 43. Comparison of the fitted or inferred scaled parameters for identical ELMing H-mode discharges in DIII-D (solid lines) and JET
(dashed or shaded lines) as a function of normalized radius. The collisionality has been replaced with the proximity to the empirical density
limit in the dimensionless parameter set: (a) scaled density (b) scaled electron temperature, (c) scaled ion temperature, (d) Zeff and (e) scaled
one-fluid diffusivity (with uncertainties). Reprinted with permission from [430]. © 2004 Institute of Physics.

relation as that governing the ion transport, the turbulence

dominating the electron transport must be at wavelengths that
are not affected by the E × B velocity shear. In this DIII-D
experiment (and most others in present-day tokamaks), the
E × B velocity shear is dominated by the terms proportional
to M. However, in ITER, it is expected that the pressure
gradient term will become important. This change does not
affect the applicability of dimensionless scaling results from
present-day experiments as long as care is taken to match Mor
the scaling results are obtained in regimes where the effects of
E × B shearing are not significant.
Previous experiments from DIII-D, JET, and ASDEX
Upgrade indicated that the ρ∗ scaling of energy transport
in low-q, ELMy H-mode plasmas (like the ITER baseline
scenario) is locally gyro-Bohm [2]. More recent experiments
in JET with Type I and Type III ELMs find gyro-Bohm scaling
in both cases [351]. Experiments in JT-60U [230] find the
Figure 44. Scaling of thermal confinement time with β (defined by ρ∗ scaling in H-mode plasmas is weaker than gyro-Bohm;
Bτth ∼ β −α ) versus the normalized thermal β (βNth ) to indicate the however, there is a systematic mismatch in β in the scans that
proximity to the β limit. The effective β scaling of the IPB98y,2 may move the scaling closer to gyro-Bohm.
scaling is also shown. Reprinted with permission from [432]. ©
2004 American Institute of Physics. Three other dimensionless scaling studies warrant
discussion here. First, experiments on DIII-D have shown
that both the electron and the ion energy transport have strong
models. It is normally assumed that the turbulence governing dependences on the temperature ratio (Ti /Te ) [233]. This
ion transport also dominates the electron transport. Evidently, implies that simple 0D extrapolations to burning plasmas from
this is not the case, or at a minimum, if the electron transport discharges with Ti > Te in present-day machines may be
is dominated by turbulence with the same linear dispersion optimistic in plasmas where Ti ∼= Te . Modelling indicated that

S63
E.J. Doyle et al

dependence is offset by an unfavourable q dependence, i.e. the

shape and q dependences largely cancel. However, at fixed q,
the shape dependence shows itself quite strongly. Significant
modelling is still required to understand how q, magnetic shear,
and cross-section shape affect the turbulence that drives the
transport.
In the ITER Physics Basis [1] there was extensive discus-
sion of potential systematic errors that could affect the dimen-
sionless parameter scaling technique, such as the mismatch of
E × B rotational shear addressed above. The effect of ‘stiff’
transport or threshold behaviour of the turbulence has been
invoked from the beginning of dimensionless scaling experi-
ments [305] as a potential problem. In order to analyse such
a situation, a specific model for such a dependence will be
used [310].
 β  
|∇T | |∇T |
χ = χo + CT α −k H −k . (11)
T T
Here, the threshold behaviour occurs when the temperature
scale length decreases below 1/k. In a perfect ρ∗ scan, the
temperature scale length is fixed, so all the discharges in the ρ∗
scan are equally near to the threshold, unless k is dependent on
ρ∗ . Both full numerical simulations and parameterized models
such as the IFS/PPPL model exhibit gyro-Bohm scaling,
which indicates these theoretical models have k independent of
ρ∗ . Therefore, threshold-type phenomena do not necessarily
invalidate the dimensionless parameter scaling approach.
In general, there are no direct limitations to the
dimensionless scaling technique. However, there are potential
systematic errors, in addition to the obvious problem
of choosing the correct set of dimensionless parameters,
discussed above. If the plasma behaviour had a strong
non-linear dependence on a physical parameter, then slight
mismatches in that parameter in scans where it should be held
fixed would affect the inferred scaling. Measuring individually
the scaling of transport with each of the dimensionless
parameters allows one to identify which parameters could
be so sensitive. The strongest dependence found is the ρ∗
scaling, which is the one of most interest. This would indicate
that mismatches in the other parameters have only a slight
effect on the ρ∗ scaling. Quantities such as the fusion cross-
Figure 45. (a) Comparison of the shear in the bulk E × B rotation section and atomic transition rates do not depend only on the
velocity for co-NBI and counter-NBI discharges with otherwise
dimensionless parameters from plasma physics, and therefore,
similar parameters in DIII-D. (b) Ratio of ion diffusivities for
co-NBI and counter-NBI ρ∗ scans. (c) Ratio of electron diffusivities these are not scalable using only these parameters. However,
for co-NBI and counter-NBI ρ∗ scans. Reprinted with permission for burning plasma applications, the effects of the fusion
from [220]. © 2002 American Institute of Physics. reactions and radiation on the transport of energy in the core
can be added separately. The goal of the dimensionless scaling
is to determine the loss power required to support the scaled
this strong dependence may arise from the dependence of the
profiles against turbulent transport, and the contribution of
ITG threshold on the temperature ratio. Scaling on transport
self-heating and radiation to this loss power can be estimated
with hydrogenic ion mass has been evaluated in H-mode
directly without resort to scaling. A difficulty arises if the
discharges in JET [436]. There is an apparent favourable scaling on the boundary layer at the edge is set by atomic
dependence of global confinement on the ion mass that cancels physics, for example, through fuelling and radiation. Because
out the unfavourable mass scaling from the increase in ρ∗ . H-mode plasmas have a significant fraction of the total stored
However, the local transport analysis does not clearly show energy confined by the edge transport barrier, divergence of
this. A detailed discussion is beyond the scope of this review. the scaling of the pedestal from the core scaling could result
Finally, the dependence of energy transport on cross-sectional in a global scaling different from that measured in the core.
shape has been studied in DIII-D discharges in both L-mode The experimental data gathered thus far indicate that the ρ∗
and H-mode [437]. In H-mode, there appears to be a strong scaling of plasmas most like the ITER baseline scenario (low-q,
favourable dependence with increased elongation. If the H-mode) have the same scaling for global and local transport.
shape change is done at fixed plasma current, this favourable This points to the necessity for machines at a variety of ρ∗

S64
 Chapter 2: Plasma confinement and transport

values to gain confidence in this scaling method. Another

potential systematic error that can occur is that the relative
proximity to the L–H threshold condition can change during
a scan. Previous experiments on JET clearly showed that the
power required to obtain fixed β can be dominated by the power
requirement to stay in H mode, rather than the loss power
to transport. In general, the dimensionless scaling method
assumes that the system does not undergo a qualitative ‘change
of state’ during the scan. (The specific case of threshold
phenomena was addressed above.) It should be noted that it is
a new extensive review of dimensionless scaling experiments
and techniques where these issues are addressed [438].

3.7. Improved core confinement regimes for advanced

operation scenarios
The advanced operation scenarios for next step devices such as
ITER should have a larger fraction of self-generated bootstrap
current than the standard ELMy H-mode operation to extend
the pulse length. These include ‘steady-state operation’ and
‘hybrid operation’ as discussed in chapter 6 of this issue
[199]; the former means the operation where the whole of the Figure 46. Radial profiles of ion temperature Ti , electron
plasma current is non-inductively driven, while in the latter a temperature Te , safety factor q and current density j , in a JT-60U
eq
substantial fraction of the plasma current is non-inductively current hole discharge in which QDT = 1.2 has been achieved.
driven to enable extended operation under the available flux Ip = 2.6 MA, BT = 4.3 T, q95 = 3.3, ne (0) = 9.9 × 1019 m−3 .
swing of the transformer. These operations require a high Reprinted with permission from [389].
poloidal beta with a reduced plasma current to enhance the
low due to a localized large pressure gradient. One of the
bootstrap current fraction fBS . To achieve a high fusion power
mechanisms for strong ITB formation seems to be an enhanced
density and a high fusion gain with reduced plasma current, a
effect of the Shafranov shift (α stabilization) in a high q
high normalized beta (βN ) and high confinement enhancement
regime [439], though parallel velocity shear destabilization
(H-factor) are also needed. The latter is achieved by transport
can prevent E × B shear quenching of the turbulent transport
reduction and enhanced confinement in the plasma core
[262]. Since the ITB foot is usually located near the qmin
including formation of an ITB. In the ITER Physics Basis [2],
radius, expanding the negative shear region by off-axis current
the pellet enhanced performance (PEP) mode and optimized
drive is useful to enlarge the volume enclosed by the ITB and
shear (OS) mode in JET, supershot in TFTR, high βp mode
improve the performance [440]. The ITB foot, however, can
and reversed shear plasmas in JT-60U and negative central
expand into the positive shear region for weaker ITBs with
shear (NCS) mode in DIII-D are described as examples of this
reduced temperature gradient [212], and the formation of ITBs
regime. In recent years, development of this regime using weak
in both positive and negative shear regions (double ITBs) is also
or negative shear plasmas has been pursued as a major objective
observed [441].
in many devices. Here, results in leading tokamaks with NBI
In this regime, ITBs were obtained in ASDEX
heating, namely JET, JT-60U, DIII-D and ASDEX Upgrade
Upgrade with NBI (+EC) heating during the current ramp
will be mainly shown. Experiments in other devices are found
[324, 442, 443], in addition to DIII-D, JET and JT-60U.
in recent review papers [24, 25]. Because avoidance of large
In JT-60U, optimization of reversed shear plasmas with an
sawtooth crashes is believed to be necessary to achieve this
L-mode edge was continued in the modified W-shaped divertor,
regime, operation with q(r)  1 is employed. In the following, eq
resulting in the record of DT-equivalent fusion gain QDT =
discharges will be classified in order of their values of magnetic
1.25 [444]. In JET, improved coupling of LH waves resulted
shear (q profiles) and their q(0) values. As shown below,
in stronger reversed shear than previous OS plasmas, and it
tailoring and controlling the q profile is closely related to the
was found that lower power was required to form ITBs and
improved core confinement and stability, and therefore is a key
to achieve enhanced confinement [256]. The ITB emergence
to developing these scenarios. A schematic graph of a desired
occurred preferentially when the minimum q reached an
q profile can be seen in figure 1 in chapter 6 of this issue [199].
integer value, and two ITBs developed and followed two
integer surfaces on each side of qmin [6, 441]. A rarefaction of
3.7.1. Strong reversed shear regime. In the strong reversed the rational surfaces around the low-order ones may be related
shear regime, namely smin < −1 (smin denotes the minimum to the ITB formation [13, 330].
∼
value of magnetic shear s) or q(0)/qmin  1.5, strong ITBs Existence of a central region with nearly zero toroidal
with a clear change in the slope in ion temperature (Ti ), current, a ‘current hole’, was observed in an extreme case
electron temperature (Te ) and electron density (ne ) profiles in this regime on JET [190] and JT-60U [189]. Though the
are observed in many machines, and high bootstrap current temperature gradient was nearly flat in the current hole, high
fractions (fBS ) are expected. The strong ITBs can result temperature plasmas exceeding 10 keV were confined with
in very high confinement, though the beta limit tends to be ITBs around the current hole as shown in figure 46, which

S65
E.J. Doyle et al
eq
resulted in high QDT = 1.2 in JT-60U [389]. The current hole
formation is attributed to generation of off-axis non-inductive
current; the toroidal electric field at the centre and hence the
central toroidal current density decreases with rising off-axis
non-inductive current, and the current hole in the central region
is finally formed. The decrease of central current density stops
(‘current clamp’) when it drops to nearly zero and the current
hole structure is maintained by some mechanism. Though
some simulation studies indicate that a resistive kink instability
works to maintain the current hole [445,446], no corresponding
signals have been observed in experiments so far [447] and the
mechanism is not fully understood yet. Flat temperature and
density profiles inside a steep gradient ITB layer, ‘box-type
ITBs’, are often observed in strong reversed shear plasmas
even without a current hole [37, 447, 448]. This suggests that
some mechanism, other than high q values, causes flat profiles.
Collapses were observed with a relatively low beta in this
regime, and extending the duration of the ITB was one of
the major issues. This was accomplished by off-axis current
drive (by LHCD, bootstrap current and ECCD) to maintain
the inverted q profile as described in section 3.3 of chapter
6 of this issue [199] and pressure profile control (decrease
in the pressure gradient in the ITB layer) with an H-mode
edge [444]. It was found that the box-type ITB with strong
negative shear can be sustained, resulting in high confinement,
HH 98(y,2) = 2.2, and a large bootstrap current fraction, fBS >
80%, with βN = 1.9–2.2 in JT-60U [449] without collapses
if the q profile was maintained and the pressure gradient was
not so large. It should be noted that q95 was high (>8) and
the toroidal beta was low in these discharges to enhance fBS .
Lower q95 is possible with strong off-axis current drive that Figure 47. (a) Ion temperature profile along a major radius in a JET
replaces the bootstrap current [440]. In DIII-D, strong negative optimized shear D–T discharge. Ip was continuously increased up to
shear was established by off-axis ECCD during the high beta 3.4 MA and BT = 3.8 T. (b) Radial profile of ion thermal diffusivity
in D–T and D–D JET optimized shear discharges. Reprinted with
phase, which increased the Ti gradient and sustained the high permission from [247]. © 1998 American Physical Society.
beta conditions; βN ∼ 2.8, βt ∼ 2.9% and HL89 ∼ 2.3 were
maintained with qmin > 2 for nearly 2 s [450]. It was noted were produced in D–T plasmas with similar additional heating
that the Te gradient was weak in these discharges in spite of power levels and similar current profiles to those in D–D.
strong negative shear, which is in contrast to results in JT-60U Central ion temperatures of approaching 40 keV as shown
and JET. The extended duration made it easier to investigate in figure 47(a) and a triple product of 1.1 × 1021 m−3 keV s
the transport properties in these discharges, including impurity were achieved, leading to 8.2 MW of fusion power [247, 453].
accumulation discussed in section 3.4. The values of χi were similar to those in D–D discharges
as shown in figure 47(b). Improvement in the coupling
of the LHRF system made it possible to employ off-axis
3.7.2. Weak shear regime with q(0) > 1. The weak
LHCD during the high heating power phase to sustain the
negative or positive shear regime, namely with −1 < s < 1
∼ ∼ inverted q profile, and the ITB was maintained up to 11 s
in the core region or with 1 < q(0) < 1.5qmin , is believed using feedback control of pressure profiles [454]. Steeper
∼
the most promising candidate for steady-state operation in temperature gradients were observed in these discharges with
ITER since enhance core confinement, high bootstrap current LHCD and weak negative shear than those without LHCD and
fraction, high MHD stability and good confinement of high with weak positive shear [455].
energy particles are expected. In JET, OS plasmas belong The high βp mode in JT-60U also belongs to this
to this regime. It was found that the ITBs were formed in regime. Though beta collapses were observed in L-mode
the vicinity of low-order rational q surfaces with q = 2 or edge discharges, quasi-steady sustainment was obtained in
q = 3 [246, 451]. It is considered that the destabilization of ELMy H-mode edge discharges (high βp ELMy H-mode), in
the MHD mode at q = 2 or q = 3 by the coupling to the particular with high plasma triangularity, δ [456]. A clear Te
surface mode could provide a locally enhanced shear in the ITB, in addition to the Ti and ne ITBs, was observed, which was
plasma flow and act as a trigger for ITB formation [6, 451]. not seen in previous high βp mode with an L-mode edge and
The highest D-D neutron emission rate of 5.6 × 1016 n s−1 , low δ. Erosion of the ITB by Type I ELMs was not observed.
which was equal to the records in TFTR and JT-60U, was The sustainable value of βN HL89 increased with δ, which
achieved in a discharge with an L-mode edge [452]. In seems to be caused by higher edge stability for high δ. High
1998, D–T operation was performed in the OS plasmas. ITBs performance with βN ∼ 2.5, HH 98(y,2) ∼ 1.4, fBS ∼ 50% and

S66
 Chapter 2: Plasma confinement and transport

Figure 48. Transport properties of a DIII-D weak negative shear discharge. Ip = 1.2 MA, BT = 1.6 T, q95 = 5.5, qmin ∼ 1.5 and q(0) ∼ 2.
The qmin surface is located at ρ ∼ 0.35. (a) χeff compared with χeff for a low q95 discharge with the same shape, size and toroidal field,
(b) χi , χe and χineo ; (c) measured electron (squares) and ion (circles) temperature profiles and drift wave model predictions for ion (solid
line) and electron (dashed line) temperatures and (d) a comparison of the normalized linear growth rate and the normalized E × B shearing
rate for the drift wave calculation. Reprinted with permission from [459]. © 2001 American Institute of Physics.

full non-inductive current drive was achieved with negative modes (TEM) and electron temperature gradient modes (ETG),
neutral beam (N-NB) injection into high δ, high βp H-mode  B shear on the mode spectrum [461].
including the effect of E×
discharge [457]. Weak, parabolic-type ITBs were obtained The q profile continues to evolve and reaches qmin ∼ q(0) ∼ 1.
with a monotonically increasing q profile with q(0) ∼ 1 and Hence this kind of discharge is connected to the following
q95 = 4.7. class of regime (‘hybrid scenario’) continuously in DIII-D.
In DIII-D weak negative shear (WNS) plasmas belong Full non-inductive operation was recently achieved with strong
to this regime. In this kind of discharge, early NBI heating off-axis ECCD [462, 463], in which βN ∼ 3.5, β ∼ 3.6%
is employed during the current ramp as in NCS plasmas. and qmin > 1.5 were maintained for 1 s. The profiles of the
The formation of strong ITBs was avoided by triggering the toroidal rotation, in addition to Ti and Te , were well reproduced
H-mode transition during the current ramp [458]. βN ∼ 2.9 in a discharge in this regime with the transport modelling code
and HH 98(y,2) ∼ 1.4 were obtained with weak ITBs whose GLF23 [463].
foot was located at ρ ∼ 0.5–0.6. The normalized radius In JET and DIII-D, the ITB can be degraded or destroyed
of the qmin surface was less than 0.4 and the ITB foot was by giant ELMs, though it is maintained in plasmas with giant
located in the positive shear region. The sustainable beta was ELMs on JT-60U. In JET, introducing high Z impurities (e.g.
limited by neoclassical tearing modes (NTMs). βN HL89 ∼ 10 krypton) was employed to enhance the edge radiation and to
(βN ∼ 3.8) was maintained for 0.8 s and βN HL89 ∼ 9 for reduce the edge pedestal pressure and the ELM amplitudes, and
2.0 s (16τE ) [459,460]. In the first case, βN exceeded the ideal the ITB was maintained in H-mode edge plasmas [464, 465].
no-wall limit. The results of transport analysis are shown in In DIII-D, a sustainable combination of ITB and edge pedestal
figure 48 together with results for a lower q95 , sawtoothing was achieved with counter-NBI and was called the QDB
ELMy discharge. The effective one-fluid heat diffusivity, mode [242, 466]. The q profile with weak negative shear
χeff , is compared for two discharges in figure 48(a). The continued to evolve, but slowly possibly thanks to the effect
two discharges have similar χeff in the core, while the high q of counter NBCD. The χi is close to the neoclassical theory
discharge has significantly higher χeff in the outer 40% of the prediction in the negative shear region [200]. In the double-
plasma. The ratio of χeff is smaller than the q 2 scaling observed null configuration with higher triangularity (δx ∼ 0.8), the
in dimensionless scaling experiments in DIII-D [231]. In the plasma pressure has been increased throughout the discharge
high q discharge, χi in the central region is significantly smaller volume due to the improvement in the edge stability [467].
than χe and within a factor two of neoclassical prediction
(figure 48(b)); however, there does not appear to be the 3.7.3. Regime with central flat q profile with q(0) ∼1. The
formation of a clear ion ITB since only modest gradients are last class of regime can be defined as plasmas with q(0) ∼ 1
observed in the measured Ti profile as shown in figure 48(c). without large sawteeth and with a relatively large region of
This figure also shows that the observed Ti and Te profiles flat q around the axis. This regime was developed in DIII-D
are well matched by GLF23 transport model simulation. The [265, 458, 459] and ASDEX Upgrade [442, 468] intensively.
GLF23 model is a gyro-fluid representation of the transport The key point is that q(0) does not fall below unity by
due to ion temperature gradient (ITG) modes, trapped electron some mechanism, like small fishbone (ASDEX Upgrade) or

S67
E.J. Doyle et al

NTM (DIII-D), though the classical diffusion of poloidal flux

predicts that q(0) should fall below unity. Higher βN and
higher confinement are expected than the standard H-mode
because of absence of a large sawtooth that triggers NTMs.
This class of discharge is believed to be the best candidate
for the so-called hybrid scenario in ITER, in which extended
pulse length than the standard operation is required to gain the
neutron fluence. The development of this class of regime is
discussed in chapter 6 of this issue [199] in detail. In both
DIII-D and ASDEX Upgrade, early NBI heating during the
current ramp was employed to enter this regime.
In DIII-D, high performance with βN ∼ 3.8 and
HH 98(y,2) ∼ 2 was obtained with infrequent ELMs, but was
terminated by the appearance of a 2/1 NTM [458]. In
these discharges, large E × B shear was obtained and χi
was reduced over most of the discharge to within 2–3 times
the neoclassical prediction, though the temperature profile
was smoothly peaked and no strong ITB structure was seen.
Using feedback control for the density and beta, βN ∼ 2.7,
HL89  2.5 and βN HL89 ∼ 7 were maintained for 6.3 s
or ∼34τE or >3τR , where τR denotes the current relaxation
time [459, 460]. The reference value for βN was chosen
just below (95% of) the beta limit for m/n = 2/1 tearing
modes. A small m/n = 3/2 mode appeared and remained
throughout the discharge with little impact on confinement;
the effect on τE is estimated to be less than 10%. The q
profile reached a stationary state after less than 2 s at high
βN and q(0) was kept just above unity without sawteeth or
fishbones. The m/n = 3/2 tearing mode seems to have a role
in maintaining this q profile. Predictive transport calculations
utilizing the GLF23 code demonstrate that the predicted Te
and Ti profiles agree well with the experimental results. The
GLF23 results indicate that the reduction in transport is due
to a combination of Ti /Te > 1 and sufficient E × B shear,
although χi is significantly larger than the neoclassical value
[265]. In discharges where higher rotation was obtained with
proper error field correction applied, higher confinement was
observed [469].
The extent of the existence domain of operating regime
was investigated by scanning q95 and the density [469, 470].
The q95 was scanned from 4.9 to 3.2. For discharges
with q95 < 4, sawteeth were observed, which reduced the
operational βN limit slightly. The presence of sawteeth did
not affect strongly the confinement and H89 remained almost
constant. βN = 2.7 and HL89 = 2.3 were maintained for Figure 49. The behaviour of the ion temperature, the electron
9.5 s in a discharge with q95 = 3.2. The density range was temperature and the density of the selected discharges in ASDEX
Upgrade. (a) Ti (0) as a function of Ti (0.8), (b) Te (0) as a function of
30–70% of nG , and they found that βN increases while HL89 Te (0.8) and (c) ne (0)/ne (0.8) as a function of ne (0.8). The data
decreases with the density. At the highest density, βN = 3.2, from the improved H-mode and the data from the standard H-mode
HH 98(y,2) = 1.2, Ti /Te = 1.3 were obtained [469]. are shown by open squares and closed circles, respectively. Stars
In ASDEX Upgrade, the stationarity of q profiles is denote discharges with pre-heating and gas puff for higher density.
Triangles denote discharges with additional EC heating. Reprinted
explained by magnetic reconnection driven by strong m/n =
with permission from [472].
1/1 fishbones, which, in the absence of sawteeth, also expel
energy and impurities [442, 471]. The highest fusion triple
product in ASDEX Upgrade, 0.9 × 1020 keV s m−3 (βN = 2.4, are shown in figure 49 [472]. The data from the improved
HL89 = 3.0) was obtained in this regime. Though this regime, H-mode is shown by open squares, whereas the data from the
called the ‘improved H-mode’, was originally considered with standard H-mode, in which no NBI heating during Ip ramp
ITBs, it was found that the temperature profiles were also ‘stiff’ was employed, is shown by closed circles. The density was
in this regime hence no ITBs were formed. The behaviour scanned for 0.3 < n̄e /nG < 0.6. The Ti profiles of the
of the Ti , Te and ne profiles of selected discharges with the dataset are observed to be stiff, as shown in figure 49(a), i.e.
same plasma current, NBI heating power and plasma shape one can use a single multiplication factor to normalize the

S68
 Chapter 2: Plasma confinement and transport

profiles after which they have a unique shape. The standard The core transport in hybrid regime is not fully understood
H-mode discharges and improved H-mode discharges follow yet. No ITB structures are recognized, and the temperature
the same scaling, though the improved H-mode has higher profiles are reported to remain stiff in some devices. The
edge and core temperatures. On the other hand, the central confinement is, however, better (typically 20% and up to 50%)
Te falls below the linear scaling in the high edge temperature than the prediction by the ELMy H-mode scaling and the E×  B
regime, as shown in figure 49(b). This may be explained shear seems to play a role in the transport reduction. The
by the fact that NBI predominantly heats ions at these high experimental temperature profiles are reproduced only when
electron temperatures. Indeed, when EC heating is applied the E × B shearing stabilization is included in the GLF23
at the centre, Te (0) increases and reaches values which are simulation [262]. This suggests that the turbulent transport
close to the linear scaling, as shown by open triangles in is somewhat regulated, though not fully suppressed, in this
figure 49(b). The ne profiles are more peaked at lower densities regime. This process, however, may also exist in the standard
even in standard H-modes as shown in figure 49(c). In the ELMy H-mode, and the higher confinement observed for the
improved H-mode, a stronger increase in the density peaking hybrid regime could be related to the negative β dependence
is observed inside ρ = 0.4. The density profile peaking of the IPB98y2 scaling [478]. The improved performance on
seems to be a key to improving the energy confinement with sustainable beta and confinement in this regime is attributed to
stiff temperature profiles. In a double-null configuration with stabilization of NTMs by eliminating large sawtooth activities.
higher triangularity, δ = 0.43, in a modified divertor geometry, Since NTM can be triggered without sawteeth and the threshold
higher βN = 3.5 was maintained over a range of q95 = 3.0–4.5 beta of NTM onset depends on ρ∗ or ν∗ , extrapolation of the
in steady-state [468,473]. The pedestal density was raised with operational domain to lower ρ∗ and ν∗ is required, and is
increasing triangurality and reached 1×1020 m−3 . HH 98(y,2) = actually in progress, to address the feasibility of this mode
1.1–1.2 with Type II ELMs was obtained at the high density to ITER.
n̄e /nG = 0.85. Higher confinement up to HH 98(y,2) = 1.5
was obtained in a lower density regime. The simulation of the 3.7.4. Prospect towards reactor-relevant conditions. Most
temperature profiles with the Weiland model are in agreement of the experimental results so far were obtained with dominant
with experimental observations within the error bars of the NBI heating with energies < 120 keV, which implies dominant
∼
measurements [474]. The existence domain was documented ion heating, Te < Ti , as well as substantial input of toroidal
for 3.2 < q95 < 4.5 [475]. momentum and particles into the plasma core. In ITER and
In JET, similar discharges to those hybrid scenario future fusion reactors, the heating will be dominated by high
discharges in DIII-D and ASDEX Upgrade were obtained with energy α particles, MeV-range NBI and RF, all of which heat
NBI heating or LHCD during current ramp. The fishbone electrons dominantly and input minimal toroidal momentum
instabilities were observed, and q(0) ∼ 1 was maintained or particles. Therefore, it is of great concern if improved
without sawtooth instabilities. At Ip = 2.8 MA, Bt = 2.6 T core confinement regimes can also be maintained in such
and q95 = 3.2, βN = 1.4 and HL89 = 2.0 were obtained. conditions.
A weak Ti ITB was observed though the confinement was The high values in Te /Ti are predicted to enhance the
moderate [476]. Experiments matching the plasma shape ITG instability [483] that is believed to dominate the ion
q profile and ρ∗ of ASDEX Upgrade was attempted, and heat transport, and therefore it is a concern whether the
βN = 2.8 and HH 98(y,2) = 1.4 were achieved without ITBs at improved ion confinement is maintained even with Te > Ti .
Ip = 1.4 MA and Bt = 1.7 T [477,478]. Stationary conditions In fact, a tendency was found that the H factor increased
were obtained with small NTM and fishbone activity in with Ti /Te in the database of advanced tokamak discharges
the core. as show in figure 50 [203], suggesting a role of Ti /Te in
In JT-60U, βN = 2.7 was sustained for 7.4 s in a high βp confinement improvement. In strong reversed shear plasmas,
ELMy H-mode discharge with Ip = 1.0 MA, Bt = 1.8 T, it was observed that the ion ITB was maintained with Te >
q95 = 3.3 and δ ∼ 0.45 [479]. The m/n = 3/2 NTM Ti [324, 332, 344]. In JT-60U, it was also observed that
was observed continuously in this discharge and q(0) was the formation of ion ITB was possible with the existence
maintained above unity. No strong ITBs were observed. These of the electron ITB, namely with Te > Ti conditions. On
features are similar to those of hybrid scenario discharges the other hand, ITBs in positive shear plasmas were found
in DIII-D and ASDEX Upgrade, though the confinement to degrade with electron heating using ECRF in DIII-D and
was moderate (HL89
1.8, HH 98(y,2)
0.9). After the JT-60U [169, 332]. In JT-60U experiments (figure 51), the
modification of NBI systems for extended pulse length, longer ion ITB degradation took place 0.5 s after the start of ECH,
sustainment of high βN was achieved; βN = 2.5, HL89 ∼ 2.0 and Te /Ti was saturated while the gradient of Er continued
at q95 = 3.4 for 16.5 s and βN = 2.3 for 22.3 s [480, 481]. to decrease. This suggests that the degradation of ion ITB
It should be noted that no sawteeth were observed even in may not be caused by a change in Te /Ti but decrease in the
a discharge without 3/2 NTMs. No fishbone instabilities Er gradient caused by ECH. The mechanism of Er gradient
were observed either. Analysis remains to be performed to reduction by ECH is not well understood yet. In the hybrid
determine if the current diffusion in this kind of discharge is regime, Te ∼ Ti was achieved with ICRF heating in ASDEX
classical. In a low-q regime with q95 = 2.2–2.7, βN = 3.0 was Upgrade and no change in confinement was observed [475].
maintained for 6.2 s without large sawteeth or NTMs [482]. The large E × B shear is believed to suppress the ITG
The absence of sawteeth in a low-q regime seem to be mode [114, 218] and is effective to establish and maintain the
related to broad temperature profiles due to off-axis NBI ion ITB. A significant fraction of E × B shear is generated
heating. through the toroidal momentum input from tangential NBI,

S69
E.J. Doyle et al

without NBI heating. Limited experiments have been made

for ion heating without NBI. In JET, ICRF heating in a 3 He
minority scheme was employed to heat the bulk ions without
any external momentum input in a reversed shear plasma
produced and maintained by LHCD [484]. As a result, an ion
ITB was formed in a nearly zero shear region, though its radius
was small. The toroidal rotation and the E × B shear were
about one order magnitude smaller than those in typical JET
ITB plasmas. In hybrid regime, the E × B shearing rate seems
to have a role in reducing the transport though no clear ITB
structure is recognized and hence the small momentum input
experiment is important as well as in ITB plasmas. Strong
ICRF heating, comparable to NBI heating power has been
attempted in JET and ASDEX Upgrade [475, 484, 485]. In
ASDEX Upgrade, no confinement degradation was observed.
In JET, HL89 = 2 was achieved at Ti (0)/Te (0) ∼ 0.8 though
βN was moderate, up to 1.55, and the pedestal height was
low compared with the NBI-dominated hybrid discharges with
similar heating power. Transport properties in these discharges
Figure 50. Confinement enhancement factor H89 versus the ratio of are not analysed yet.
the central ion (Ti0 ) and electron (Te0 ) temperatures in the ITPA Some discharges for advanced operation scenarios have
database for advanced tokamak plasmas. Transient (open symbols) lower densities in terms of the density normalized by the
and stationary (closed symbols) results are given. Reprinted with
Greenwald density (nG ) required for ITER and future reactors.
permission from [203].
This is partly because the ITB formation seems to be easy in
a low density regime [486], where the intense central heating
with NBI is possible and a larger E × B shear tends to be
obtained through larger toroidal rotation with the same injected
torque. In strong ITB plasmas with reversed shear, high central
densities were obtained with NBI fuelling and pellet fueling
inside the ITB layer [366, 440, 487], and we will have no
problems in viewpoint of core transport for high densities.
However, the edge or pedestal densities were relatively low
in previous experiments. In ITER and fusion reactors, central
NBI fuelling should be small and fuelling with pellet inside
the ITB will be difficult. The strong density ITB should be
avoided to suppress impurity accumulation as discussed in
section 3.4. Hence a moderately peaked density profile is
expected and high pedestal density is required. A reduced
density gradient may result in enhanced transport since a steep
density gradient or a small value in ηi = Lne /LTi , where Lne
and LTi are scale lengths for ne and Ti , profiles, respectively,
is believed to work to suppress the ITG mode. In a JT-60U
reversed shear plasma, about half of NBI heating power was
Figure 51. Degradation of an ion ITB with ECRF heating observed replaced by ICRF, and a strong Ti ITB was maintained with a
in a JT-60U weak positive shear plasma. (a) Profiles of toroidal reduced density gradient and an enhanced Te ITB [488]. This
rotation of carbon ions. (b) Profiles of radial electric field Er . indicates that a steep density gradient is not necessary for Ti
(c) Temporal changes in the effective Er gradient (dEr /dr)eff and ITB sustainment. On the other hand, a high pedestal density
the maximum in the normalized ion temperature gradient, Rp /LTi will enhance the power for off-axis current drive to maintain
profile. Here (dEr /dr)eff ≡ (|dEr /dr|max + |dEr /dr|min )/2, which
is closely related to transport reduction in the ITB in JT-60U.
the reversed shear profile. In ITB plasmas with weak magnetic
Reprinted with permission from [332]. shear, few attempts were made to raise the density near nG . In
JT-60U high βp H-mode, the achievable density was limited
to 70% of nG with gas puffing and pellet injection. Injection
and the ITB control was demonstrated by changing the NBI of Ar was required to reach 92% of nG with HH 98(y,2) = 0.96
directions [200, 212]. Though strong ITBs are observed with [366]. The temperature ITBs seems to get weaker, though the
balanced injection of NBI in JT-60U [212], a difference in density ITB remained, in these high-density regime, but the
orbits of co-injected and counter-injected beam ions may reasons are not well understood. In hybrid regime, density scan
produce local toroidal rotation shear. In strong reversed was intensively performed in DIII-D and ASDEX Upgrade as
shear plasmas, effects of negative magnetic shear and of the described in section 3.7.3. Though in both devices, it was
Shafranov shift, the latter is enhanced due to larger q values, found that the confinement was slightly degraded with the
seem to suppress anomalous transport even without the E × B density, the ASDEX Upgrade results, HH 98(y,2) > 1.1 at 85%
shear. In fact, strong Te ITBs are observed in many devices of the Greenwald density [468, 473], is promising.

S70
 Chapter 2: Plasma confinement and transport

3.8. Summary and outstanding issues Particle transport (section 3.4) is also relatively less well
understood than ion thermal transport, due to the existence of
Remarkable progress has been made in developing and
both edge and core particle sources, and the importance of
understanding regimes of improved core confinement since
both convective and diffusive transport. However, substantial
the publication of the IPB. Internal transport barriers and other
progress has been made since the publication of the IPB. High
forms of reduced core transport are now routinely obtained in
confinement has been obtained with peaked density at high
all the leading tokamak devices worldwide. Reduced transport
density (at and above the Greenwald density). An inward
has been achieved in all four transport channels (ion and
particle pinch has been observed on multiple devices, and may
electron thermal, particle, and momentum transport channels),
be explained in the framework of ITG/TEM turbulent transport.
sometimes simultaneously. This rapid worldwide progress in
Evidence for neoclassical inward particle transport driven by
the development of enhanced confinement modes of operation
the density gradient has also been found in ITB discharges on
has contributed to an increased emphasis on developing both
multiple devices. However, this inward neoclassical particle
hybrid and steady-state operation modes for ITER.
transport is also consistent with the observation of high-Z
Ion thermal transport (section 3.2) is now relatively
impurity accumulation in many ITB discharges. To counter
well understood and is believed to be regulated by ITG-
this latter effect, control of the density peaking has been
type turbulence. Theory-based modelling can qualitatively
developed for such ITB discharges, utilizing on-/off-axis RF
replicate ion transport across a wide range of operating
heating. Remaining issues in this area include: determining
regimes (RI-mode, ITB discharges with varying levels of
magnetic shear, hybrid, non-ITB AT discharges, etc), with the optimum level of density peaking in ITER, as there
quantitative agreement in many cases. The predictive ability is competition between improved confinement with central
of these ion transport models has progressed to the point density peaking, and confinement reduction due to high-Z
that modelling is now regularly used in experimental design. impurity accumulation, which also increases with density
These successes have increased confidence in projections to peaking; determination of what density profile can be obtained
ITER, as discussed further in section 5. Important outstanding in ITER with low central fuelling; and establishing whether
issues include: quantifying the level of profile stiffness in the current density profile control techniques are consistent with
plasma core; identifying the physical mechanism for the robust reactor conditions.
observation of Bohm-like ion transport scaling in L-mode For a number of reasons, momentum transport and
(drift wave models are intrinsically gyro-Bohm); replicating plasma rotation are less well understood than the other
both the amplitude and phase response of modulated transport plasma transport channels (section 3.5). These include
experiments in a single transport model, and understanding an incomplete knowledge of momentum sources (e.g. the
the fast radial propagation of heat pulse experiments, which is source of ‘spontaneous’ plasma toroidal rotation), and sinks
difficult to explain using local diffusive models. (e.g. resonant and non-resonant interaction with error fields).
Electron thermal transport (section 3.3) is relatively less The applicability of neoclassical theory in determining
understood than ion thermal transport. A larger number of poloidal rotation is also currently an open question. Thus,
turbulent modes, covering a broader wavenumber range, can while scaling studies and similarity experiments have been
contribute to electron transport, namely ITG, TEM, as well initiated, an ability to predict momentum transport and rotation
as ETG type turbulence (the latter short wavelength ETG for ITER is currently lacking.
modes only being capable of causing significant transport With regard to the dimensionless parameter scaling
if larger-cells, so-called ‘streamers,’ can be formed). This techniques (section 3.6), considerable progress has been made
larger range of turbulence activity and interactions is perhaps since the IPB, and several of the issues outstanding at that
responsible for the more diverse range of experimental results time have been addressed. This approach has now been
and theoretical interpretation seen in this area, as compared successfully applied to transport scaling in both L- and H-mode
with ion transport. A feature of electron transport are ‘profile plasmas, and the matching of plasma parameters in the identity
stiffness’, with extensive evidence for the existence of a experiments have been expanded to include E×  B shear effects.
threshold in the normalized gradient above which turbulent Significant new work on the scaling of energy transport with β
transport increases dramatically, potentially explaining the has been performed, showing a weak or null dependence of the
observed stiffness. Electron transport barriers can be achieved, transport on beta, in contrast to both theoretical expectations
usually in association with reversed magnetic shear profiles. and global database scalings (see section 5.4). Other issues
Current evidence is that TEM and ITG turbulence usually are that dimension scaling studies indicate that both electron
dominate electron transport, but that ETG modes may play and ion energy transport have strong dependences on the
a limiting role when electron ITBs are formed. Scaling ion/electron temperature ratio, implying that 0-D extrapolation
experiments indicate that electron transport is gyro-Bohm, to reactor conditions (Ti ∼ Te ) from hot ion plasmas may
in accord with theoretical expectations for drift wave-like be optimistic. Such dependences are, however, in agreement
turbulence. Outstanding issues in this area include: providing with theoretical and modelling expectations for ITG dominated
an unambiguous resolution of whether a threshold exists in the transport.
normalized temperature gradient; resolving the existence and Finally, the transport properties of improved confinement
relative role of TEM and ETG turbulence, and including their regimes relevant to advanced operation on ITER (steady-state
behaviour accurately in transport models; and more extensive and hybrid scenarios) were presented in section 3.7. A major
experimental studies of transport behaviour with dominant feature of the results is that improved confinement regimes can
electron heating, especially in enhanced core confinement be routinely obtained on all major devices with a broad range of
regimes. q profiles (with strongly reversed through weak shear, and with

S71
E.J. Doyle et al

For MHD stability at high βN , we need to increase the pedestal

pressure so as to achieve a broad pressure profile, since a
moderate pressure gradient in the core region is needed to
stabilize global ideal low n modes and neoclassical tearing
modes. In order to sustain the current profile with an optimum
shape for high confinement and stability, distribution of the
edge bootstrap current has to be controlled at an optimum
level. The pedestal temperature, density and ELMs determine
the shielding and pumping of impurities, thus the pedestal
characteristics are a key to particle control. In addition, the
ELM heat load on the divertor plates should be reduced to a
tolerable level. The data show that the energy loss in a Type I
Figure 52. Schematic radial profile of plasma pressure. ELM crash tends to increase with increasing pedestal pressure
(or both the energy loss and the pedestal pressure increase with
Confinement normalized decreasing collisionality). We have to develop techniques for
Improvement density Type I ELM mitigation, or access to small/no-ELM regimes,
HHy2 by keeping a favourable pedestal pressure at an ITER-relevant
normalized n e /n GW
1.3 low collisionality.
pressure βN 0.83
2.56 P rad /P heat From the physics point of view, the H-mode pedestal is
bootstrap 0.56 radiation
a complex system. The pedestal structure and its dynamics
fraction fBS are determined by both the plasma physics (transport, MHD
fraction
0.5
stability) and the atomic physics. The scale length of the
0.77
non-inductive 1 plasma pedestal parameters is of the same order as the ion orbits and
current fraction fCD fDT purity the penetration depth of the neutrals. Furthermore, the spatial
distribution of the pedestal parameters and their temporal
Figure 53. Integrated plasma performance required for a
behaviour show two or three-dimensional structures. The time
steady-state operation in ITER.
scales of the leading phenomena span from the growth time
of ideal MHD/turbulence to the wall saturation time, with the
a range of qmin values). With regard to extrapolation to reactor parallel and perpendicular transport times and the edge current
conditions, progress has been made with regard to obtaining growth time in between.
reduced core transport with Ti ∼ Te , low momentum input and The main research issues were raised and categorized
at high density, though in all the cases further investigation in the excellent review reports [490–493]. Since then, by
and experimental demonstration of robust high confinement revealing the above mentioned physics through experimental,
operation are desirable. theoretical and numerical approaches, recent pedestal research
has shown significant progress, particularly in achieving
4. Pedestal transport and dynamics high confinement at high density, Type I ELM mitigation
and development of small/no-ELM regimes, and predictive
This section reviews recent progress in the research on the capability for burning plasmas. In the following subsections,
H-mode edge and pedestal physics. The edge pedestal we will review recent progress and remaining issues for
parameters are essentially important for burning plasma pedestal research: enhanced confinement at high density,
performance because they determine the boundary conditions the pedestal structure, turbulence suppression and the L–H
for the core plasma as well as the source of heat and transition, pedestal transport theories, integrated modelling,
particle flows into the SOL and divertor plasmas (figure 52). Type I ELMs and their mitigation, small/no-ELM regimes,
For example, projections for ITER show that its fusion pedestal MHD stability theories and pedestal control schemes.
gain depends strongly on the pedestal ion temperature.
Furthermore, the H-mode pedestal plays a central role in 4.1. Regimes of improved H-mode confinement at high
achieving integrated plasma performance: in order to sustain density and operational limits
burning plasmas in tokamak power plants, we need to sustain
high values of the energy confinement improvement factor One of the main physics issues identified in the ITER Physics
(the H-factor), normalized beta (βN ), bootstrap and non- Basis document [1] was an observation that it is difficult
inductively driven current fractions, plasma density, fuel to maintain H-mode confinement in tokamak experiments at
purity and radiation power simultaneously (figure 53) [489]. high plasma density with gas puff fuelling. While regression
The roles of the pedestal in each of these are as follows. analysis of the Type I ELMy H-mode confinement database
Because of the stiffness in the temperature profile, the energy was indicative of τE increasing with plasma density, τE ∝ n̄0.41
e
confinement and the fusion gain are strongly affected by the [1], the density scans with gas puff fuelling in individual
pedestal temperature. The pedestal temperature decreases with devices revealed a saturation or even decrease of τE , as initially
increasing pedestal density, because the pedestal pressure is shown in JET [494] and later reproduced in other devices.
limited by ELMs to be a nearly constant value. Therefore, Figure 54(a) shows that the HH 98(y,2) factor decreases with
in order to achieve high confinement enhancement at a high increasing density as reported from JT-60U [294]. In the Type
density, improved pedestal pressure is essentially important. I ELMy H-mode, the pedestal pressure is limited by ELMs.

S72
 Chapter 2: Plasma confinement and transport

1.4 2
(a) (b)
δ =0.45

Te-ped [keV]
1.2 1.5
1

HH y2
1
0.8
δ =0.16 0.5 1.8 MA, 3.0 T
0.6 δx = 0.16-0.18
Pressure = const.
0.4 0
0.2 0.4 0.6 0.8 0 1 2 3 4
n e / nGW n e -ped [1019m-3 ]

15 2
(d) δ =0.16 (e)
Ti (0) , Te (0) (keV)

(c) 10
1.5 CORE δ = 0.45
δ =0.16
H H-ONL
10

T e (keV)
1
Ti
5
0.5 1 δ = 0.20
Te PEDESTAL
0 0
0 1 2 3 4 0 1 2 3 0 1
T i-ped , T e-ped (keV) T -ped [keV] r /a
i

Figure 54. Type I ELMy H-mode in JT-60U [294]. (a) Density dependence of the H Hy2 -factor for the Type I ELMy H-mode discharges
(low-δ(= 0.16), high-δ(= 0.45)). (b) Pedestal electron temperature as a function of pedestal electron density (δ = 0.16). (c) Relationship
between pedestal and central values of Te and Ti . (d) Pedestal ion temperature dependence of confinement enhancement factors based on the
offset non-linear scaling. (e) Comparison of the electron temperature profile for high and low δ cases with the same pedestal density.

Under this limitation, when the pedestal density increases, to the H-mode confinement scalings [1] decrease with n̄e /nG
the pedestal temperature decreases (figure 54(b)) and the core at constant δ and increase with δ at fixed n̄e /nG . In highly
temperature also decreases due to the profile stiffness effect, i.e. shaped plasmas (δ = 0.35–0.45), the HH 98(y,2) factors reach
Tcore ∝ Tped as seen in figure 54(c) (see also section 3.3). As a 1 at n̄e /nG ≈ 1.1 in both devices. In JET at high triangularity
result of this linkage, the stored energy does not increase with (δ > 0.4), a good confinement (HH 98(y,2) = 0.9–1.0) at high
increasing density and the confinement enhancement factor density (n̄e /nG  1) is linked with access to the mixed Type I/II
decreases because the scaling has a positive dependence of ELMy regime characterized by higher pedestal pressure at high
the confinement time on density. Figure 54(d) shows that density than with Type I ELMs (figure 56) [500].
the core confinement enhancement factor relative to the offset Another favourable effect of high triangularity was
non-linear scaling [495] decreases with decreasing pedestal reported from JT-60U [501]. Figure 57(a) (δ = 0.44–
ion temperature. Therefore, the basic factors determining the 0.48) shows that the pedestal pressure (∼ne,ped Te,ped ) remains
confinement degradation at high density is the limitation of the roughly constant for the standard ELMy H-mode with Type I
pedestal pressure by ELMs and the profile stiffness. ELMs (open circles). While in the high βp ELMy H-mode
Based on this knowledge, in recent years several methods (small closed circles), the pedestal pressure can be higher
have been discovered that allow H-mode operation with a than that of the standard H-mode. This enhanced pedestal
good confinement at densities close to and even higher than pressure is due to improved edge stability enhanced by high
so-called the Greenwald density. These methods include the βp values: figure 57(b) shows the pedestal βp (βp,ped ) increases
increase of triangularity in the plasma cross-section [494], deep with increasing total βp at high δ = 0.44–0.48 (circles). This
pellet fuelling, strong gas puff at plasma midplane combined relationship between βp,ped and the total βp seems independent
with intense pumping from the divertor [365], low steady gas of existence of the ITBs (open symbols: without ITB, closed
puff allowing density peaking [355] and controlled impurity symbols: with ITB), which means that this relation does not
come from the profile stiffness. On the other hand, βp,ped is
injection [370].
almost constant at low δ (squares). One candidate explanation
for this improvement is that a larger Shafranov shift improves
4.1.1. Effects of plasma shape on global and pedestal the pedestal stability at high triangularity.
confinement. At high triangularity, δ, a good confinement A range of plasma parameters (δ  0.4, βN ∼ 3, q95  3.6
enhancement factor can be kept at higher density [496, 497] and q(0) ∼ 1 at a near double-null configuration) has been
because of the improved stability against ELMs [498] (see found in ASDEX Upgrade, where the ‘improved’ H-mode with
section 4.8). At high triangularity, the edge pressure is HH 98(y,2) = 1.2–1.3 and small (Type II) ELMs is observed at
higher than that in low δ discharges and thus the pedestal densities up to n̄e /nG ≈ (0.8–0.9) [468]. Similar regimes
and core temperatures are higher at a given pedestal density with somewhat different characteristics have been obtained
(figure 54(e)), and the confinement enhancement factor in DIII-D [469, 470], JT-60U [366, 479] and JET [478, 485].
becomes higher as shown in figure 54(a) [499]. Figure 55 The physics of these favourable regimes and prospects for an
illustrates the behaviour observed in JET [370] and ASDEX ‘improved hybrid’ scenario in ITER based on such regimes
Upgrade [248] at different, fixed values of the plasma are discussed in section 3.8 of this chapter and section 3.3 of
triangularity δ. The confinement improvement factors relative chapter 6 of this issue [199].

S73
E.J. Doyle et al

Figure 55. Confinement enhancement factors relative to the empirical H-mode confinement scalings as a function of n̄e /nG at different
triangularities δ. (Left) JET. Reprinted with permission from [370]. (Right) ASDEX Upgrade. Reprinted with permission from [248].

work is required. High density discharges with good H-mode

confinement sustained for long times with density peaking and
with impurity seeding are discussed in section 3.4.5.

4.1.3. Understanding of density-limiting processes. There

are several density limits in tokamaks [1, 431]. Two of them,
i.e. the H-mode density limit associated with a back transition
from H- to L-mode and the ultimate L-mode density limit
related to plasma disruption are the most important for reactor
like devices. The figure of merit for the L-mode density limit
is the Greenwald density [1, 431],
Ip Bt
nG = ≡ 1.59g (1020 m−3 , MA, T, m), (12)
π a2 q95 R
where g = q95 /qcyl is the plasma shaping factor with qcyl =
Figure 56. Confinement enhancement factors for the density scans 5a 2 Bt /(RIp ). Typically, at operation in the Type I ELMy
in JET at q95 = 3, with Ip /Bt of 2.5 MA/2.25 T, 3 MA/2.7 T and H-mode with gas puff fuelling, an increase in density above
3.5 MA/3.2 T. Reprinted with permission from [500].
some limit leads to a transition from Type I to Type III
ELMs accompanied with reduction of the stored plasma energy
4.1.2. High density discharges with pellet fuelling. Figure 58
by 15–40% [503]. A further increase in the gas fuelling
(left column) shows a JET discharge demonstrating the
rate leads to a back transition to the L-mode, correlated
potential of high field side (HFS) pellet injection for fuelling
with complete divertor detachment and/or divertor/X-point
the plasma to high density with simultaneous high confinement
MARFE formation [1]. At even higher fuelling rates,
in a large tokamak [502]. In this discharge with a tailored
the L-mode density limit disruption occurs, terminating the
injection rate, a density above the Greenwald density has
been achieved with a large peaking factor, ne (0)/nped ≈ 2, discharge.
HH 98(y,2) ≈ 0.82 and βN ≈ 1.8, although further work is
needed to optimize the stationarity. The right column in Models for the H-mode density limit. A number of semi-
figure 58 shows a flat density profile in ASDEX Upgrade empirical and theoretical models for the H-mode density
near the H-mode density limit with strong gas puffing, and limit have been suggested. The Borrass model identifies
strongly peaked density profiles with high field side pellet the density limit with a complete divertor detachment. The
injection [367]. One can see that pellet fuelling achieves higher limit on the separatrix density that follows from the ‘two-
central densities compared with strong gas puffing at similar point’ SOL model has been obtained in the form [504]
5/16
edge densities. Evolution of the density profile after injection nsep ∝ q⊥x Bt /(q95 R)11/16−x , where x = (10 − ξ )/[16(1 +
of a single pellet is also shown. ξ )], ξ is an arbitrary constant, q⊥ = (Pheat − Prad tot
)/Ssep ,
HFS pellet injection is presently considered as a Prad = Prad + Prad and Ssep is the separatrix surface area.
tot core div

main fuelling technique compatible with reactor conditions. This relation, in combination with an empirical scaling for
However, this technique is not yet sufficiently developed to be nped /nsep , normalized to a typical JET discharge with a
accepted as a convenient tool in current experiments. Further strong gas fuelling and flat density profile gave the Borrass

S74
 Chapter 2: Plasma confinement and transport

(a) 2 limit similar to equation (13) [510]. There are data, however,
nT= const. pellet which contradict to assumption that the divertor detachment
E37413
is a main reason for the H–L transition. Results of studies
1.5 of high-density discharges in JT-60U [511] suggest that the
T e - ped [keV]

ultimate cause of the H-mode density limit is deterioration of

high βp
ELMy H-mode
confinement. It was observed that density increase under low
1 radiation conditions was limited by either H–L transition or
transition from a Type I to a Type III ELMy regime without
ELMy E32398
H-mode puff any sign of divertor detachment, although the latter always
0.5 Ip=1.0 MA occurred at densities approaching the density limit under high
δ x = 0.44-0.48 radiating conditions. Also, in deriving the Borrass scaling, a
type I ELM type III ELM Bohm-like radial transport in the SOL was assumed, although
0
0 1 2 3 4 much higher absolute values of D⊥ (a factor of 50–100 in
19 -3 DIII-D case [431, 512]) are needed in model calculations for
n e - ped [10 m ]
quantitative agreement with experimental data. Therefore, the
relation (13) should be treated as empirical rather than theory-
δ =0.44-0.48 δ =0.14 - 0.20
(b) with ITB based scaling.
(E3235
8) w/o ITB The scaling (13) predicts a relatively low H-mode density
0.6 limit for ITER, nBLS (I T ER) ≈ 0.45nG , suggesting that
pellet high ne
it could be difficult to obtain the required plasma density,
0.5 n̄e  0.85nG , in this device with gas puff alone. On the other
puff high ne
E32398 hand, Mahdavi et al [513] using a model for divertor power
δ=0.44 E37413
0.4 -0.48 evolution
balance combined with an analytical gas fuelling model, have
β pPED

shown that the maximum achievable pedestal density in ITER

with ITB could be close to or exceed nG with gas fuelling alone, if a
0.3 DIII-D type open divertor configuration with pumping from the
divertor is utilized. Although the basic ideas of this approach
0.2 δ=0.14 - 0.20 are reasonable, the model itself is over-simplified and can
hardly be expected to be relevant for extrapolation to ITER.
w/o ITB
0.1 In particular, the model for power balance in the SOL neglects
δ =0.44-0.48
the impurity radiation and the volumetric recombination of the
1MA, giant ELMs
divertor plasma.
0
0 1 2 3 Modelling of the edge plasma in ITER using the
βp B2-EIRENE code revealed a saturation of the separatrix
density at the level of (3–4) × 1019 m−3 under gas fuelled
Figure 57. JT-60U pedestal conditions. (a) ne,ped versus Te,ped for conditions, in good agreement with the Borrass scaling
ELMy discharges at 1 MA and δ = 0.44–0.48 (open circles: predictions [514]. To ensure a broad operational space in ITER
standard ELMy H-mode, small closed circles: Type I ELMy high βp with a closed divertor, substantial core fuelling, e.g. injection
H-mode including pellet injected discharges). (b) Dependence of
of pellets penetrating beyond the H-mode pedestal has been
the pedestal poloidal beta, βp,ped , for the Type I ELMy discharges at
1 MA (circles: high δ = 0.44–0.48, squares: low δ = 0.14–0.2, adopted [515].
open: without ITB, closed: with ITB). Reprinted with permission Integrated core–pedestal-SOL (ICPS) modelling with the
from [501]. multi-mode transport model (MMM) applied for the plasma
core with boundary conditions on the separatrix obtained from
scaling [505], B2-EIRENE calculations shows that a reasonable operation
space with n̄e = (0.7–1.3)nG and Q = 6–30 would exist in
q⊥0.094 Bt0.53
nBLS = 4.14 (1020 m−3 , MW m−2 , T, m), (13) ITER at a strong core fuelling rate 15–60 Pa m3 s−1 combined
(q95 R)0.88 with an edge gas puff fuelling rate 0–70 Pa m3 s−1 , which
where nBLS = nped ≈ n̄e . This scaling with the same satisfies acceptable steady-state power loads (
10 MW m−2 )
normalization constant fits also the ASDEX Upgrade data on the divertor target [516]. In these simulations, a modified
[506]. Simulations of SOL and divertor plasma in JET with MMM transport model is applied at the plasma edge to produce
the B2-EIRENE code [507, 508] have revealed a divertor an edge pedestal. The ICPS model has been calibrated against
detachment and a saturation of the upstream separatrix density an H-mode discharge in JET and fitted well the edge parameters
at plasma parameters similar to those at the H-mode density in ASDEX Upgrade density scan experiments.
limit in experiments [504]. The H-mode density limits
observed in JET and ASDEX Upgrade with gas fuelling agree Models for the L-mode density limit. In most experiments
well with the Borrass scaling (figure 59(a)) and are, in average, with gas puff fuelling and flat density profiles, the H–L
∼20% below the Greenwald scaling although some points transition occurs rather close to the L-mode disruptive limit
are close to nG (figure 59(b)) [509]. A log-linear regression (DL), nH –L /nDL = 0.7–0.9 [431, 509]. Therefore, both the
analysis of the JET database using Bt , R, q⊥ , q95 and δ as H- and L-mode density limits in present machines can be
variables yields an empirical scaling for the H-mode density reasonably well characterized by the Borrass and Greenwald

S75
E.J. Doyle et al

Figure 58. (Left) Time evolution of the normalized electron density ne /nG and total plasma energy content Wdia in JET discharge #53212
with high field side pellet injection at Ip = 2.5 MA, B = 2.4 T, δU /δL = 0.34/0.3 and κ = 1.67. Dashed lines correspond to a reference
discharge without pellet injection and gas puff. Reprinted with permission from [502]. © 2002 Institute of Physics. (Right): (a) density
profiles in ASDEX Upgrade when ramping up from the initial profiles to the H-mode density limit by strong gas puff (dashed curve) or to
stable H-mode operation beyond the Greenwald density by pellet refuelling with turbomolecular pumps only (solid curve) or with
cryopumping (chain curve). (b) Density profiles before and after single pellet injection and the pellet emission profile. Reprinted with
permission from [367].

scalings. However, due to stronger Bt dependence (at constant Three-dimensional non-linear gyro-fluid simulations of the
q95 ) in the Greenwald scaling, it predicts a significantly higher Braginskii equations in a simple geometry have revealed very
(by a factor of 2) density limit for ITER compared with the high turbulent transport (∼50 m2 s−1 ) driven by the resistive
Borrass scaling. In this regard, an accurate comparison of the ballooning instability at plasma parameters typical for the SOL
density limits predicted by the two scalings with the disruption at the density limit in present experiments [520]. Simulations
limit in Alcator C-Mod at high Bt , close to that in ITER, would of the edge plasma turbulence using the BOUT code in a
be of interest. realistic divertor geometry [177,521] have shown that turbulent
It is generally accepted that the disruptive density limit transport at the plasma edge increases strongly with density
is associated with cooling of the plasma edge that causes the
(by factors 10–100) near the Greenwald density and can lead
current profile to contract, destabilizing the m/n = 2/1 MHD
to an X-point MARFE formation when impurity radiation is
mode [1]. Two cause of edge cooling, i.e. increase in radiative
included. Tokar [522] has demonstrated the importance of the
power loss and transport enhancement, are considered [431].
Radiation models describe satisfactorily a number of synergy between anomalous transport and impurity radiation
aspects of the L-mode density limit, such as radiation collapse for density limit phenomena using an analytic model for the
at Prad ≈ Pheat , MARFE formation and divertor detachment drift-resistive ballooning instability in a plasma with a circular
[1]. However, they predict a strong dependence of the limit on cross-section.
heating power and impurity content that are not seen in some The ratio n̄e /nG is not dimensionally correct. Experiments
experiments [431]. on DIII-D and JET have shown that, when the ratio n̄e /nG was
Recent experiments support earlier indications of a kept fixed with other dimensionless parameters kept constant,
possible link between the cross-field particle transport in the the normalized energy confinement time on DIII-D was 20%
edge plasma and the density limit [431]. General trends smaller than on JET, while at fixed ν∗ the normalized energy
observed in Alcator C-Mod [517] and DIII-D [518, 519] confinement times on JET and DIII-D agreed within ∼5%.
are as follows. As n̄e /nG is increased from low values to These results suggest that scaling transport properties from
values approaching ∼1, a benign cross-field heat convection present devices to ITER at fixed n̄e /nG would result in incorrect
increases to a level that impacts the SOL power loss predictions [430].
channels and reduces the separatrix electron temperature.
And finally, non-diffusive, intermittent transport phenomena Using the standard non-dimensional scaling approach,
known as ‘blobs’ (normally associated with a far SOL) i.e. taking βN and ν∗ the same as those in JET shot #53212
invade the closed flux surface region and carry a convective (Bt = 2.4 T, I = 2.5 MA, a = 0.96 m, βN = 1.8, n̄e /nG = 1,
power loss that impacts the power balance of the discharge. and pellet fuelling, as shown in figure 56) [370] one obtains
Shortly before a disruption, the radiation power loss also    1/3  2/3  
increases. n̄e Bt,ITER aITER n̄e
= ≈2
Strong enhancement of turbulent transport in the SOL with nG ITER Bt,JET aJET nG JET
increasing density has been found in theory-based modelling. (14)

S76
 Chapter 2: Plasma confinement and transport

significantly higher collisionality ν∗ than ITER at the same βN

and n̄e /nG ∼ 1. Recent experimental and theoretical studies
confirm the hypothesis that increase in anomalous particle
transport at the plasma edge plays a role in the L-mode density
limitation. Due to higher temperature and lower ν∗ expected
at n̄e /nG ∼ 1 in ITER, the resistive ballooning modes near
the separatrix could be stable and an X-point MARFE may
not penetrate inside the separatrix, making density limitations
less severe compared with present experiments. Definitive
experimental tests of this statement can only be made in
reactor-scale devices.
The H-mode density limits in present experiments with gas
puff fuelling and flat density profiles are described reasonably
well by the Borrass scaling. The density limit predicted by
this scaling for ITER is about 2 times below the Greenwald
density. Experiments show that achievable averaged plasma
density in the H-mode increases with density profile peaking.
Some theoretical models for turbulent transport, supported by
limited experimental data, predict moderate density peaking
in ITER in the absence of core plasma fuelling. However, to
ensure achieving high core density in ITER, substantial core
fuelling, e.g. via pellet injection in addition to gas puff, would
be needed.
To predict achievable densities in ITER with a good
accuracy, first principle theories for the density limit are
required. Such theories should include detailed SOL
and pedestal physics taking into account transport, atomic
and MHD processes. For testing the theories, accurate
measurements of transport coefficients and radiation in the
SOL and H-mode pedestal with a good temporal and space
resolution are necessary.
Figure 59. Experimental H-mode density limits, nDL , versus
Borrass (a) and Greenwald (b) scaling predictions for a set of data
from JET and ASDEX Upgrade discharges with gas puff fuelling. 4.2. Pedestal characteristics and structure
Experimental data are from JET Mark-I and Mark-II (upward
triangles), the gas box divertor configuration (bullets) and from the 4.2.1. The importance of H-mode pedestal characteristics.
ASDEX Upgrade Divertor II configuration (downward triangles). Successful ITER operation will require both high pressure
Reprinted with permission from [509]. at the inner edge of the H-mode transport barrier and small
ELMs. ITER will operate near the density limit in order to
and achieve the highest fusion power output, Pfus ∝ n2 , and to
 2/3  1/3 maintain the power above the value required for transition to
Bt,ITER aITER
T ITER = T JET ≈ 2 T JET H-mode, Pfus /PLH ∝ nx , where 1 < x < 1.4. The relatively
Bt,JET aJET
flat density profile expected in ITER H-mode discharges
(15)
implies a pedestal density also near the density limit. To
for the ITER non-dimensionally similar discharge. At achieve high efficiency, Q = Pfus /Paux , requires good energy
n̄e /nG ∼ 1 in ITER, the plasma temperature should be y
confinement, Pfus /Ploss ∝ nx HL89 , where −1 < x < 1
four times higher and ν∗  eight times lower than those in and 2 < y < 3 [2], where HL89 is the ratio of H-mode
the JET discharge. For coupled core and edge temperatures to L-mode energy confinement time. HL89 is expected to
(e.g. ‘stiff’ profiles) this suggests that radiative instabilities improve with pedestal temperature at a given pedestal density
(MARFEs) and resistive ballooning mode instabilities would as seen experimentally, for example in figure 54, and expected
be less pronounced in ITER than in the JET discharge. This theoretically from turbulent transport models [523, 524] that
means that the Greenwald density may be exceeded in ITER. predict core temperatures to be highly dependent on the value
of the temperature at the top of the H-mode pedestal. Estimates
4.1.4. Summary. Experiments on major tokamaks of the pedestal temperature required for Q = 10 at Pfus =
demonstrate that increasing the triangularity of the plasma 400 MW with nped ∼ 0.7 × 1020 m−3 vary from Tped ∼ 4 keV
cross-section permits the attainment of high energy for GLF23 [236] to Tped ∼ 2.5 keV with the less stiff multi-
confinement in Type I ELMy H-mode at high plasma mode model [236,525]. Thus high pedestal pressure is needed
densities, close to the Greenwald density where ITER is to simultaneously meet the goals of high fusion power output
planned to operate. However, these favourable results cannot and high efficiency. The largest allowable energy loss across
be extrapolated directly to ITER using the dimensionless the separatrix during an ELM is set by the maximum acceptable
scaling techniques since plasma in these discharges has peak heat flux to the divertor plates as described in chapter 4 of

S77
E.J. Doyle et al

this issue [364] and section 4.6 of this chapter. Unfortunately, PED
the ELM energy loss generally increases with the pedestal ∆Ti,Te /av
pressure. Although several factors, including the duration of 0.15 JT60U
the ELM energy loss, can influence the divertor peak heat flux JT60U–New Data
from ELMs (chapter 4, this issue [364]), the acceptable ELM D3D
energy loss for the Tped range above is roughly 5–10% of the CMOD
energy in the pedestal, Wped = (3/2)pped V . Since this ELM 0.10 ITER–FEAT
energy loss as a fraction of pedestal energy is typical for Type I requirement
ELMs in present-day tokamaks, ELM mitigation techniques based on
(section 4.6.3) or alternatives to the Type I ELM regime which GLF23
maintain high pedestal pressure (section 4.7) may be required
0.05
for ITER.

4.2.2. Requirements and projections for H-mode transport

barrier width. As described in section 4.8, the peeling–
0.00
ballooning mode model for the onset of the ELM instability is 0.00 0.02 0.04 0.06 0.08 0.10
well supported by experiments. Since the peeling–ballooning 1/2
mode limits the pressure gradient in the H-mode pedestal 2(6ε/(1+ε)) ρPi /av
region, the pedestal pressure is set by a combination of
Figure 60. Normalized transport barrier width is correlated with ion
this limit and the radial extent of the steep gradient region poloidal gyro-radius, ρpi . Older JT-60U data at low triangularity is
that corresponds to the edge transport barrier. Using the in agreement with scaling based on turbulence suppression from an
predictions from peeling–ballooning theory for ITER shown edge radial electric field generated by loss of ions on banana orbits
in figure 88, a ratio of transport barrier width to minor radius crossing the separatrix, as described in section 4.2.2. Reprinted with
of 1 < /a < 2.5% would be needed to reach pedestal permission from [529].
temperatures in the required 2–4 keV. Transport barrier widths
in this range are typical in present-day tokamaks [526–528], Some authors have suggested a scaling for the transport
however a well-validated means of predicting the H-mode barrier width based on a process that might create the velocity
transport barrier width in ITER is still lacking and remains a shear. Shaing [535] and others have argued that the velocity
major topic of research. Attempts to project a barrier width for shear that suppresses turbulence in the H-mode pedestal is
ITER through physics based or empirical scaling from multi- driven by the J × B force associated with currents that must
tokamak databases are described below [529]. This approach flow in response to the loss of ions on orbits that cross the
has not yielded a result that is consistent with the experience on separatrix. The range of this force is then roughly the width of
all tokamaks, perhaps due to the wide variation of parameters the banana orbit for a barely trapped ion
in such databases and the difficulty of separating the edge √
ft vti 2ε/(1 + ε)3kTi /mi 
stability constraint. A more controlled set of inter-machine ≈2 ≈2 = 2 6ε/(1 + ε)ρpi ,
pi eBp /mi
comparison experiments has been, or will be, carried out which
use the dimensionless parameter approach to derive a scaling (16)
for the transport barrier width [530–532]. These experiments, where pi = Bp /mi is√the ion gyro-frequency in the poloidal
some of which are described below, have so far not provided a field, Bp , and ρpi = kTi /mi / pi is the ion poloidal gyro-
complete scaling relation for the barrier width but do provide radius. The radial electric
√ field acts to squeeze the orbits
some insight into the important parameters. Also, work is giving [198]  ∝ ρpi / Q, where Q = |1 − Er /Bp pi |.
underway to extend turbulent transport models through the Since the ∇p term typically dominates the Er determined from
pedestal region [533]. This approach may ultimately provide radial force balance, we obtain Q = 1 + (ρpi /)2 . Itoh [536]
a firm basis for understanding the pedestal transport barrier noted that viscosity, µ, would drive the shear flow region
structure. inward from the region where the force was present giving

Below we describe three physics-based arguments as to
a scaling for the transport barrier width,  ∝ ρpi2 + µ/νi ,
what might set the transport barrier width and examine how
well they fit the available data and what they would predict where νi is the ion collisionality. The measured barrier width
for ITER. Scaling laws derived for the H-mode transport for the three machines is compared with equation (16) in
barrier width are compared with experimental results in figure 60 using poloidally averaged values for the poloidal
databases available to the international community through field, minor radius and width: Bp =µ0 Ip / , where  is the
the ITPA. PDB3V2 [534] includes pedestal information for plasma circumference, and av = V /2π 2 R, where V and
Alcator C-Mod, ASDEX Upgrade, DIII-D, JET and JT- R are the plasma volume and major radius. Previous work
60U, including both electron and ion pedestal parameters; on JT-60U [526] gave Ti ∼ = 2.3ρpi for ELM-free discharges,
however, the data is mostly averaged over ELMs. D3DPED where this relation held well over a range of safety factors,
[529] is comprised entirely of DIII-D data with the Thomson 2 < q < 7, and to some extent was independent of plasma
scattering measurements of the electron pedestal parameters shape over a range of triangularity, 0.05 < δ < 0.4, and
taken roughly every 10 ms in 173 discharges. At present there elongation, 1.4 < κ < 1.7. This is close to the value predicted
are transport barrier width measurements available for JT-60U, by equation (16) for JT-60U,  = 2.2ρpi or  = 2.0ρpi ,
DIII-D and Alcator C-Mod. with orbit squeezing. More recent JT-60U data for ELMing

S78
 Chapter 2: Plasma confinement and transport

In another approach to H-mode transport barrier width

∆Te,Ti /a (Midplane) scaling, the process which creates Er is not directly considered,
0.15
rather it is assumed that any Er profile consistent with the
radial force balance can be obtained by adjusting the particle
and power fluxes. Taking transport coefficients that were
0.10
reduced continuously with increasing E × B velocity shear,
Hinton and Staebler [537] demonstrated a transport bifurcation
and derived a barrier width scaling set primarily by the edge
0.05 localized neutral particle source
D3D
JT60U
 ≈ [2λLsep
n n(csep Qsep )]
1/2
, (17)
(a)
0.00
sep
0.00 0.05 0.10 0.15 where λ = νn /ne σ v is the neutral mean free path, Ln
∆/a Neutral Penetration Model is the density-gradient scale length at the separatrix, and
sep , Qsep are the particle and heat fluxes at the separatrix
0.03 on which the width depends only logarithmically. Work on
∆/a Neutral
Penetration Model DIII-D [513, 538] has shown that the shape of the density
profile in the transport barrier is consistent with what would
0.02 be expected for neutral penetration. In this model, when
the density is sufficiently high, neutrals cannot cross the
SOL to reach the main plasma without undergoing charge
0.01 exchange or ionization (in which case they are recycled to the
divertor). Thus, at high density, neutrals crossing the separatrix
(b)  acquired the velocity of the ions in this region, vn ≈
have
sep
0.00 2Ti /π mi . This type of process generally characterizes
0 1 2 3 4
the DIII-D discharges. At low density, neutrals at the Frank–
TiSEP – ITER Condon velocity can cross the separatrix and reach regions
of higher ion temperature. Since the charge-exchange rate
Figure 61. (a) Comparison between the transport barrier width at
the midplane and the width of the steep density-gradient region is somewhat larger than the ionization rate, a large fraction
inside the separatrix from neutral penetration. (b) Transport barrier of the Frank–Condon neutrals can acquire velocities more
width prediction for ITER as a function of the separatrix ion characteristic of the H-mode pedestal temperature. This effect
temperature; the 3% width required by stability is reached only at is further enhanced by the fact that above 300 eV the ionization
ped
very high Ti . Reprinted with permission from [529]. rate decreases while the charge-exchange rate continues to
increase. The JT-60U discharges in the ITPA database with
discharges, most of which is at higher triangularity, departs large widths are generally in the Frank-Condon dominated
from this scaling but a strong correlation with ρp is still clear. regime. Figure 61 shows a comparison between this neutral
The DIII-D data lie near the predicted value, however, there penetration model [538] and transport barrier widths. The
is no clear trend with ρpi . The Alcator C-Mod data is also Alcator C-Mod results are not included in the figure since
near the predicted value, but other data from Alcator C-Mod almost all the density profile is predicted to be outside the
show no correlation with ρpi [527]. This scaling would give separatrix making it difficult to relate the density profile to
a pedestal width in ITER of /a = 0.004 (v /av = 0.01), the particle source inside the separatrix. In this analysis
which would give an expected pedestal temperature of only the separatrix ion temperature for DIII-D is taken to be the
1 keV for ITER based on the stability calculations of figure 88. temperature at 95% of the poloidal flux, which may account
(Note that the required value for ITER based on GLF23 from for some of the large predicted values. Because of the
figure 88 is mapped from the midplane value to av in figure 60.) high densities required in ITER for a high level of fusion
Experiments on DIII-D, however, in which divertor pumping power, the ITER case is similar to DIII-D. A prediction of
and gas puffing were used to vary the pedestal temperature over the ITER pedestal width based on this model as a function
a wide range indicated that a simple function of temperature of separatrix ion temperature is shown in figure 61(b). This
could not account for both the variation of the width between model suggests that, at least in the case of gas puff fuelling,
ELMs and the overall behaviour as the density was varied it would be difficult to obtain a large enough width based
[528]. The JT-60U discharges are in a low collisionality on stability considerations (figure 88) unless the separatrix
regime where the viscosity term may be expected to play a temperature was so high that it would be in conflict with the
role and result in a different scaling. In addition, DIII-D ITER divertor requirements. However, specific inter-machine
discharges where the ∇B  drift direction was reversed did not comparison experiments [530–532] were carried out to address
show a significant change in transport barrier width although the question of whether the neutral penetration depth is setting
a significant change in orbit loss would be expected. These the H-mode transport barrier width. No change in /a was
results suggest that a more detailed study of the possibility observed in comparison with the Alcator C-Mod, DIII-D and
that the ion orbit loss region sets the transport barrier width is JET tokamaks with ν∗ , ρ∗ , β, q and the plasma shape fixed at
needed before it can be applied with confidence to a prediction the top pedestal under conditions where the neutral penetration
for ITER. would be expected to vary significantly.

S79
E.J. Doyle et al

Another approach to determining the barrier width is based

∆ TPED /a
on the idea that the inner edge of the barrier is defined by 0.15 i,T e
the point at which the velocity shear is sufficient to quench
the turbulence [523]. There is computational and theoretical
support for the principle that the turbulence is quenched when 0.10
the velocity shearing rate in the absence of turbulence exceeds
the linear growth rate of the instabilities [523]. ITG modes JT-60U
still may dominate the region near the inner boundary of the DIII-D
0.05 C-Mod
transport barrier and we take as a guide the results from core
transport simulations [20,523]. Typically the pressure gradient (a)
term dominates the radial electric field derived from ion radial 0.00
force balance so that 0 10–3 2×10–3 3×10–3 4×10–3
ρ*s
cs2 (cs a)(a/)ζ1 (Te /Ti )ζ2
ωE×B ∼ > γL ∼ . (18)
2 ci f (s, α)g(Zeff )h(ν∗s )
The destabilizing effect of the temperature gradient is
given by a/ taking the temperature gradient scale length as
the transport barrier width. The exponent ζ1 is a function of
the type of instability, e.g. slab versus toroidal ITG; for shifted
circles GLF23 simulations give ζ1 ≈ 3/2 [20]. The Te /Ti
term is the stabilizing effect of high ion relative to electron
temperature, for shifted circles ζ2 ≈ 3/2 [20]. f (s, α) is a (b)
function that has a similar form to the s–α diagram of the
ideal MHD ballooning mode, reflecting the similar effects
of curvature and shear on ITG modes [20]. As with ideal
MHD, at high magnetic shear increasing shear is stabilizing Figure 62. (a) The normalized transport barrier width on the outer
and increasing pressure gradient is destabilizing, while the midplane is not well correlated with normalized gyro-radius,
opposite is true at low shear where something like a second suggesting that the more complex terms in equation (19) may be
stable regime occurs. We would expect f (s, α) also to be important. Reprinted with permission from [529]. (b) The
a function of plasma shape as is the case for ideal MHD. DIII-D/JET pedestal comparison experiments show no strong
ped ped
dependence of the pedestal width on ρ∗ when ν∗ , β ped , q and
g(Zeff ) represents the stabilizing effect at high Zeff . h(ν∗s )
plasma shape were held fixed [532].
represents the stabilizing effect of collisionality where the
collision frequency in this case is normalized to the sound
transit time characteristic of the eddy turnover time; for DIII-D, barrier width for about a factor of 2 change in ρ∗s as shown in
JT-60U and ITER, ν∗s 1. Recent work [253] has shown that figure 62(b) [532]. These considerations indicate that a more
the effectiveness of the velocity shear is expected to be reduced complete theoretical understanding of the behaviour of the ITG
at high elongation through essentially replacing the field in ci growth rates in the pedestal region in real geometry is required
by Beff ≈ κBt . Solving equation (18) for ∗ = /a gives before this type of argument can be applied to a prediction of
the barrier width.
 1/2
Te Finally we consider an empirical scaling approach guided
∗ = ρ∗s
1/2 f (s, α)g(Zeff )h(ν∗s ). (19)
to some extent by the discussion above [529]. The rather
Ti
clear dependence of the JT-60U data on ρp suggest this
The normalized width, ∗ , is plotted against ρ∗s = cs / ci type of term should be included. The DIII-D data however
for data from JT-60U, DIII-D and Alcator C-Mod in figure 62, suggest a mixture of density and temperature dependence.
where all the values are measured at the top of the pedestal. It Previous empirical scaling work on DIII-D gave pe ∝
is clear that this quantity does not organize the data well; the ped
(βp )0.4 . Following Hatae [539], separate temperature
small trend between DIII-D and JT-60U comes mostly from
and density dependences are allowed for through ρ∗ and
the κ term in the effective field. Adding the temperature ratio
a dimensionally correct form for the density normalized to
term only further increases the scatter in such a plot. It should
the Greenwald density, n∗G = 0.1n(1020 /m3 )a 1.75 /Ip (MA).
be noted that the scale length for the variation of the shear,
Terms reflecting the effect of the plasma cross-sectional shape
and also possibly of the temperature ratio and Zeff , may be
on the velocity shear and ITG growth rates are also included.
comparable to the transport barrier width that introduces an
Fits were done to both a dimensionless form and a form that
implicit dependence of the terms on the right-hand side of
allows for explicit size dependence. For the DIII-D data the
equation (19) on ∗ . For example, strong shear dependence
larger D3PED data were included. The size dependent fit gives
could make the width more a function of the shear profile
−0.5
and only weakly dependent on ρ∗s . The strong dependence ∗ = 0.12ρ∗s
0.39
(Bt /Bpave )0.31 n0.19
∗G κ (1 + δ)0.57 R 1.6 a −0.5 .
of the barrier width on ρ∗s given in equation (19) is also (20)
inconsistent with experiments on DIII-D in which ρ∗s at the top
of the pedestal was varied while keeping other dimensionless This form predicts a pedestal width for ITER well above
parameters fixed. These experiments showed no change in the value required by stability. Fitting to a dimensionless

S80
 Chapter 2: Plasma confinement and transport

expression gives Tped= 5.3 keV

ITER O
∗ = 0.38
0.044ρ∗s (Bt /Bpave )0.27 n0.19 −0.98
+ δ)0.52 1.6 −2.3 100 nped= 7x1019 m-3
∗G κ (1 R ε .
(21)
This form does not organize the different machines as well as

Pped exp (kPa)

equation (20), but it also predicts that the required pedestal 10
width for ITER is met.

4.2.3. H-mode pedestal pressure scaling. A scaling law can ASDEX-U

be derived directly for the pedestal pressure from the PDB3V2 1 DIII-D
database that was described above [540]. For this scaling JET
JT-60U
the pedestal pressure gradient is taken as the ideal infinite n
RMSE=25.8%
ballooning mode stability criteria,
1 10 100
   2  
dp 1 RBp 2 Ppedmodel (kPa)
= (22)
dr 0 2µ0 R a 1 + κ2
Figure 63. Comparison of pedestal pressure values in the PDB3V2
modified by a factor to take into account a Shafranov shift database with the model given in equation (25). This model predicts
dependence and the stronger shape dependence of lower n Tped = 5.3 keV for ITER at nped = 0.7 × 1020 m−3 . Reprinted with
peeling–ballooning modes, permission from [540].
   
dp dp
= F (shape, βp ), (23) the large ELMs associated with the large pedestal energy can
dr crit dr 0 be mitigated or replaced by a more benign form of H-mode
pedestal transport as discussed in sections 4.6 and 4.7 of this
F (shape, βp ) ∝ κ C1 (1 + δ)C2 AC3 (Ptot /PLH )C4 , (24) chapter, as well as in chapter 4 of this issue [364].
where PLH = 2.84M −1 B 0.82 n̄0.58e Ra
0.81
is the L- to H-mode
transition critical power, A is the aspect ratio and M is the 4.3. L–H transitions
atomic mass number. Expressing the βp dependence in terms
of power also allows for an explicit variation of the pressure The transition from the L-mode to the H-mode (L–H transition)
with power. The H-mode transport barrier width is taken as is a subject that has been under constant investigation for a

∗ ∝ ρp /a [541]; this form is suggested by equation (19) long time because of its complexity. This section presents
where the strong variation of the magnetic shear, s, over the new experimental results, while theoretical developments
barrier introduces a ∗ term to the right-hand side. The are presented in the following section. Joint experimental
resulting scaling, and theoretical studies have led to the implementation of a
 1/3 4/3 2 number of models that capture the experimentally observed
M R Ip,MA
pped = 2.41 × 103 dependence of the threshold power on plasma density,
nped,20 a 4  1 + κ 2 7/3 magnetic field and ion ∇B  drift direction at the L–H transition.
2 Experiments have also been performed expressly to validate
 0.06 the theoretical models. For example, a recent analytic model
Ptot
×(1 + δ)3.2 κ 3.62 A−2.94 (25) for the finite beta stabilization of drift-resistive ballooning
PLH modes involving zonal flow generation [542] has shown good
is shown in figure 63. This scaling predicts a Tped = 5.3 keV agreement with data on DIII-D and Alcator C-Mod [543,544] .
at nped = 0.7 × 1020 m−3 , which exceeds the requirement for Also, new edge diagnostics or diagnostics with better temporal
ITER to achieve Q = 10 even for the stiff GLF23 model. or spatial resolution at the plasma edge have been developed
and have contributed to advances on the experimental side of
4.2.4. Conclusions. There remains large uncertainty in the L–H transition studies.
prediction of the ITER pedestal parameters. The peeling– In COMPASS-D, the H-mode is initiated when all
ballooning model fits the variation of the edge pressure turbulence mechanisms are stabilized [545]. The stabilization
gradient with shape and other parameters well (section 4.8), but of one of these mechanisms cannot be considered as a criterion
understanding of the H-mode transport barrier width scaling to reach the H-mode. The plasma edge potential, and therefore
is still lacking. The transport barrier width relative to the the radial electric field, in JFT-2M changes rapidly at the L–H
minor radius required for ITER is at least not larger than transition [546,547]. This synchronization is clearly evidenced
is common in existing tokamaks. Furthermore, although in discharges with the injected power slightly above the
arguments like those leading to equation (19) would suggest threshold power. For a given plasma configuration and global
that the transport barrier width would be reduced at the small parameters, the transport coefficients in Alcator C-Mod show
ρ∗ values expected in ITER, no such strong dependence has a bifurcation when the temperature or temperature gradient at
been observed. Scaling of the pedestal pressure, although not the plasma edge rises above a threshold value [548, 549]. The
strongly based in physics, also indicates that the ITER pedestal analysis of JET H-mode discharges shows that the electron
temperature should more than meet the required value. The edge temperature or the dimensionless parameter ρ∗ must
main concern for the ITER pedestal may then be whether exceed a threshold value to obtain an L–H transition [550].

S81
E.J. Doyle et al

The same behaviour was also shown with the ion temperature known as occurring at an excessive threshold power, such
[551]. However, the comparison between experimental data as transitions obtained below q95 = 2.5, were deselected
and models in JET only shows partial agreement. Therefore, [559]. This operation alone reduced the RMSE value of the
in general, further model development is still required to cover fits in the plasma density ne , magnetic field Bt and plasma
more precisely all situations encountered in all devices. outer surface area S from 27.9% to 27.1%. The RMSE
The search for critical parameters in L–H transition value decreased to 21.4% when a reduced set of data was
experiments also continues in order to shed light on the used. This reduced set consists of more homogeneous data
underlying processes. In most cases, the variation in the regarding future devices and normalization, device per device,
threshold power constitutes the measure of the influence of was applied with known influencing parameters. The resulting
the different parameters. For instance, the ion ∇B  drift scaling has the following form:
direction relative to the X-point position changes the threshold
PLH = 0.042n0.73
20 Bt
0.74 0.98
S (MW). (26)
power in DIII-D by a factor of two [552]. The potential
underlying process responsible for the L–H transition must In ITER, it is planned to operate the L–H transition at ne =
have: (a) different values at the same power for the two ∇B 
0.5 × 1020 m−3 . With Bt = 5.3 T and S = 680 m2 , the
drift directions and (b) equal values at the L–H transition. The threshold power in ITER will be 52 MW.
shear in the poloidal group velocity of density fluctuations at Recently, the roles of aspect ratio (A = R/a), absolute
the plasma edge fits these conditions. This is in contrast with magnetic field at the outer equatorial mid plane (Bt,out ) and
JET plasmas where the ion ∇B  drift direction has less influence effective ion charge (Zeff ) in the threshold power estimation
on the threshold power, but induces a significant difference in were studied [560]. The contribution from MAST and NSTX
the ion temperature at the plasma edge [551]. allows the incorporation of the aspect ratio in the scaling. The
The plasma shape and divertor configuration have influence of Zeff is extracted from a subset of a data since the
significant influence on the threshold power. In JET, the value is missing for many time slices. The threshold power
threshold power decreases with decreasing X-point height can then be expressed as
above the divertor floor [553]. The removal of the divertor
septum, however, had a reduced effect on the threshold PLH = 0.072n0.7 0.7 0.9 0.7 γ
20 Bt S (Zeff /2) F (A) , (27)
power. Increasing the lower plasma triangularity decreases the
where F (A) = 0.1A/f (A), f (A) = 1 − [2/(1 + A)]0.5 and
threshold power when the edge density exceeds 1.7×1019 m−3 .
γ = 0.5. Using this expression, for Zeff = 2.0 the prediction
The H-mode accessibility in TCV Ohmic plasmas strongly
for ITER is PLH = 40–50 MW, which should be reliable.
depends on the plasma configuration. The minimum threshold
Recent progress in the analysis of the threshold database
power is obtained when L–H transitions occur during the
has not led to significant changes in the scaling coefficients.
formation of the diverted configuration [554]. The plasma
Therefore, these scalings can be used for predicting the
configuration has an even stronger influence on the H-mode
threshold power for access to the H-mode regime in future
access in the new large spherical tokamaks (STs). In MAST,
devices.
the threshold power is substantially reduced when the plasma is
in double-null configuration [555], while in NSTX, the single-
null configuration has a lower threshold [556]. 4.4. Pedestal transport theory and modelling
The underlying effect of the plasma or divertor Among all theories excellently reviewed by Connor and
configuration on the threshold power is often related to the Wilson [561], most experimental evidence supported those that
influence of the neutral density at the plasma boundary. This involved the paradigm of sheared electric field suppression
is confirmed by the reduction of the threshold power when of turbulence. However, the underlying mechanism that
the gas is injected from the high field side instead of the generates the radial electric field (Er ) and the corresponding
low field side as observed in NSTX [556], MAST [555] and ‘trigger’ (the Reynolds stress, ion orbit loss, the Stringer spin-
COMPASS-D [555]. In all the devices, the L–H transitions up, effect of neutrals) is still open: the variety of observations
are synchronized with the sawteeth [547, 557], indicating that cannot be encompassed by any single theory. This may imply
the energy conveyed to the plasma edge following a sawtooth that in different experiments different triggers can be involved.
crash can also play a role in the L–H transition process. The most elaborated family of theories can be categorized
All these effects contribute to the observed variation in as phase-transition models. First of all, these models have
the threshold power around the value estimated from the a simple structure that allows additional effects to be included,
simple scaling that takes into account the plasma density, and secondly these models can be readily incorporated into
magnetic field and plasma size. For example, the repetition transport codes. Recently, a minimal model of the L–H
of an identical ASDEX Upgrade discharge on a day-to-day transition was formulated by Kim and Diamond [52], which
basis leads to a residual scattering of about 10%. This recognizes that in addition to the mean flows turbulence
variation cannot be attributed to the vessel conditioning after suppression may be due to zonal flows that are self-generated
an opening since these transitory phases were removed from by turbulence via the Reynolds stress. Zonal flows are very
the analysis [558]. effective in suppressing turbulence because their radial scale
The reduction in the scattering in the international is the same as the radial scale of turbulence, and their energy
threshold database (13 tokamaks, approximately 700 selected source comes from the non-linear coupling with turbulence.
time slices) is one of the major goals of the ITPA Confinement Zonal flows and turbulence together conserve total energy.
Database and Modelling Topical Group. Several contributions Thus, suppression of turbulence by zonal flow means non-
from different tokamaks were revised and L–H transitions linear energy transfer from fluctuations to axisymmetric modes

S82
 Chapter 2: Plasma confinement and transport

that do not contribute to heat transport. Suppression of steeper and b increases and thereby increases the growth
turbulence leads to a weakening of zonal flow generation rate of zonal flow. Maximizing the growth rate over the
that can result in bursty temporal behaviour (as in predator- radial mode numbers gives the threshold condition. In terms
prey models). Zonal flows may trigger the L–H transition of measurable plasma parameters the threshold condition is
1/2
by damping turbulence until the shearing of the mean flows  > c , where  = Te /Ln , Te is the electron temperature,
becomes strong enough to damp both turbulence and zonal Ln is the density scale length at the steepest part of the density
2/3 2/3
flows. Taking into account zonal flows reduces the L–H gradient and c = 0.45Bt Zeff /(RAi )1/6 . This threshold
transition power threshold. shows very good agreement with edge data from discharges
For predictive modelling a ‘flow-turbulence’ relationship undergoing the L–H transition in DIII-D [543]. The critical
must be incorporated into a transport code. To approximate the parameter is the same for the L–H transition in discharges
theory, a detailed model of the temporal evolution through the with oppositely directed ∇B drifts. The difference in power
L–H transition (rather than only stationary profiles in the L- thresholds (edge temperature) is due to a difference in density
and H-modes) of the Te , Ti , n, Er profiles and turbulent fluxes scale length for the two cases. Pellet-induced H-modes, which
should be included. This demands that Te , Ti and n must all have a reduced power threshold, occur at the same value of
be included in the ‘minimal model’ mentioned above. This the critical parameter. This is an impressive example of how
was done in [562], where the four-field edge turbulent layer seemingly different L–H transition mechanisms can be unified
model (ETL-model) was formulated to describe the non-linear by identifying the appropriate trigger parameter.
dynamics of zonal flows and convective cells driven by the So what should we expect from the comprehensive theory-
drift-resistive interchange instability at the plasma periphery. based model of the L–H transition?
It was shown that the Braginskii hydrodynamic equations for
four fluctuating fields—electron and ion temperatures, density 1. Instead of comparing models with stationary experimental
and the electrostatic potential—can be reduced to the three profiles in the L- and H-modes, the temporal evolution
Lorenz-like equations coupled through the equation for the of Te , Ti , n, Er profiles and turbulent fluxes through
kinetic energy of the fluctuations. It was shown that the control the L–H transition should be reproduced. The causal
parameter for the L–H transition is the edge pressure gradient relationship between the sheared E × B flow generation,
and that its critical value is much lower in the four-field (Te , the corresponding turbulent transport suppression and the
Ti , n, ϕ) model than in a two-field (T , ϕ) model. The ETL- steep temperature and density-gradient formation in a
model was then used as a boundary condition of the third kind narrow layer inside the separatrix should be explained.
(matching of fluxes) for the ASTRA transport code [155]. The Note that the time scale of ETB formation is much faster
ETL-model simulations pointed out that the L–H transition than the energy confinement time.
power threshold depends on the values of the temperature 2. The critical parameter for the onset of ETB formation
and density of plasma and neutrals at the separatrix, which should be identified, which seems to be related somehow
can be ‘hidden’ parameters influencing scalings for the power to the edge temperature. The critical temperature scales
threshold and width of the edge transport barrier (ETB). For positively with Bt and negatively with density.
example, the U-shape dependence of the L–H transition power 3. The simulations must give the correct parametric
threshold, Pth , on averaged density in DIII-D, dependences for the power threshold: Pth increases with
density, magnetic field and major radius, and Pth can show
Pth ≈ 1.7 n19  + 15.4/ n19  − 7.9, (28) hysteresis.
4. The model must include neutral physics, because in
where Pth is in MW and n19  is the density in units of 1019 m−3 , DIII-D a good correlation between the width of the
can be simulated by a change in the neutral density at the density ETB and the neutral penetration length was
separatrix, which mainly influences the transition through the found, although the ETB width does not show any clear
charge-exchange friction term in the shear velocity equation dependence on plasma parameters.
(see figure 5 of [155]). Recent pedestal studies on DIII-D
5. A challenging test for an L–H transition model is
confirmed a strong correlation between the width of the
simulation of the core-edge interaction: ETB formation
H-mode density barrier and the neutral penetration length
influences global confinement on a time scale much faster
[563]. These results are obtained by comparing experimental
than the energy confinement time.
ne profiles with the predictions of an analytic model based on
6. The underlying physical mechanism of an L–H transition
the coupled continuity equations for the electrons and neutrals.
model must be robust because the H-mode is universal
So the pedestal width depends on both the fuelling source and
with respect to heating methods and devices.
plasma transport.
Theory based simulations of zonal flow generation by Existing models are able to reproduce key features of ETB
finite β drift waves have been carried out by Guzdar et al formation, and the relevant physical mechanisms appear to
[543]. It was shown that the growth rate of zonal flows, have been identified, i.e. the ‘bricks’ are ready to construct
which are E × B flows at zero frequency driven by a the castle of a unified L–H transition model. A good
modulational instability of finite β drift waves, is determined example of implementing theory-based models in a boundary
by the dimensionless parameter b = β(qR/Ln )2 and has turbulence code is BOUT (for BOUndary Turbulence), which
a minimum at b = bc . bc is a threshold for the onset of extends the simpler models by including the full divertor
the L–H transition because when b > bc positive feedback geometry with a magnetic separatrix, an X-point and scrape-
takes place: zonal flow growth suppresses fluctuations and off layer (SOL) physics [564]. L–H transition simulations
decreases turbulent transport, then the density profile becomes were performed using a three-dimensional (3D) nonlocal

S83
E.J. Doyle et al

neutral’s physics

Figure 65. Sketch of the L–H transition physics: the pedestal zone
plays the role of a tap for the energy flow from the core to the SOL
plasma. So the core energy content is determined by edge pedestal
Figure 64. (a) Time history of electron heat flux with sources, physics, which includes the self-consistent turbulence-flow
(b) potential k spectrum before the transition and (c) potential k generation driven by the pressure gradient and is strongly influenced
spectrum after the transition. Reprinted with permission from [564]. by neutral physics and SOL transport.

electromagnetic turbulence code, which models the boundary to suppress the turbulence. After the transition, large Er shear
plasma using fluid equations for plasma vorticity, density, flow is sustained by the steep pressure gradients and the plasma
electron and ion temperatures and parallel momenta. With stays in the H-mode.
sources added in the core-edge region and sinks in the Although these results are very suggestive and encourag-
SOL, the code follows the self-consistent profile evolution ing, for a detailed quantitative comparison of simulations with
together with turbulence. Under DIII-D tokamak L-mode experiment, some additional physics must be incorporated. For
conditions, the dominant source of turbulence is pressure- example, it was shown analytically that the influence of an
gradient-driven resistive X-point modes. These modes are X-point on the L–H transition power threshold appears to be
electromagnetic and curvature-driven at the outside midplane important [565]. This conclusion was drawn in the context
region but become electrostatic near X-points due to magnetic of the Alfvén-drift paradigm for the L–H transition. It took
shear and collisionality. Classical resistive ballooning modes into account that difference in the up–down temperature due
at high toroidal mode number, n, co-exist with these modes to curvature drift in tokamaks with a separatrix can be of or-
but are subdominant. Results indicate that, as the power is der unity depending on the ion drift direction. A numerical
increased, these modes are stabilized by increased turbulence- test of the L–H transition power threshold dependence on the
generated velocity shear, resulting in an abrupt suppression of change in local beta near the X-point, as noted in [565], would
high-n turbulence and the formation of a pedestal in density be valuable.
and temperature, as is characteristic of the L–H transition. The
time history of the radial electron heat flux at the separatrix 4.5. Modelling the pedestal structure
is shown in figure 64(a). At ωci t = 3200, the electron and
ion heating power (each of 1 MW) are turned on near the The major purpose of modelling the pedestal structure is
inner boundary and plasma background profiles are allowed to understand and predict the temporal variation of the
to evolve. After a period of adjustment, the electron heat pedestal width and height for the plasma density and
flux and fluctuating electron temperature are suppressed by temperatures. Several codes have been developed for this
more than one order of magnitude. Radial profiles of the purpose [516, 533, 566–570]. This modelling could be thought
poloidal fluctuation spectrum before (figure 64(b)) and after of as providing the boundary conditions for the core plasma
(figure 64(c)) the transition indicate the suppression of high- and SOL/divertor plasma. On the other hand, the pedestal
n modes, which yields the reduction of the related transport. structure is affected by the core plasma (e.g. α-stabilization
These simulations indicate the following scenario for the due to large Shafranov shift), and divertor conditions also
bifurcation transition as shown in figure 65: as the plasma affect the separatrix conditions for the pedestal. Furthermore,
is heated in the core, the ion and electron temperature at there is little separation of spatial scales in this region, since
the core-edge boundary increase and their gradient scale the ion poloidal gyro-radius is comparable to the plasma
lengths get smaller, especially at the edge due to edge-SOL gradient scale lengths. Therefore, the modelling of pedestal
boundary conditions. The steep pressure profiles drive strong structure inevitably requires the full integration with the
instabilities and turbulence, which provides large cross-field core and SOL/divertor plasmas. Such a full integration is
transport. The strong turbulence also drives strong E × B shear the final target of the model development, while at present
flow due to Reynolds stress. Before the transition, this flow most codes use the separatrix surface as an interfacing
dominates over the diamagnetic flow. However, as the pressure surface with other sophisticated SOL/divertor code results.
profiles steepen, the E × B shear flow reaches a sufficient level One of the difficulties in modelling the pedestal structure

S84
 Chapter 2: Plasma confinement and transport

is that there exist various physics mechanisms with very estimate for the growth rate in the absence of stabilization;
different time scales. The global pedestal structure evolves ωE×B is the E × B shearing rate; and s is the magnetic shear
on a transport time scale, but during this evolution, MHD with threshold parameter t.
phenomena with very short time scales occur, as from Type I Figure 66 shows the application of the model to JET.
ELMs. Thus, most of the modelling codes are based on a The electric field shear term (first term) alone produces a
transport code with some description of the enhanced transport pedestal appreciably lower than that experimentally obtained
from microturbulence and MHD events. MHD events are in JET (figure 66 left, upward triangles). The additional
often modelled only by evaluating analytic stability criteria magnetic shear term (second ‘s’ term above) is therefore
calibrated with an MHD code (e.g. MISHKA) [567, 568] or included, and a threshold formulation is used because the major
the pedestal is treated in a time-averaged way with the critical part of the profile (low shear) is well represented by MMM
pressure gradient determined by a simple analytic formula for transport. Smaller numerical values for G and t give stronger
ballooning modes [516, 566]. The occurrence of an MHD stabilization. The profiles obtained for two combinations of
event (ELM) is characterized by an increase in the turbulent these parameters (t = 0.5, G = 0.5 and t = 0.5, G = 1.0) are
cross-field transport or an average value of the enhanced shown in figure 66. The former combination gives a better fit
transport. Modelling of the transport barrier width is not to the temperature profile; and the density profile is well fitted
yet well developed. Some codes employ an ad hoc model for both combinations. This best fit (t = 0.5, G = 0.5) also
of the transport barrier width by introducing an appropriate agrees well with the ASDEX Upgrade (AUG) database shown
 B shearing rate together
transport suppression factor due to E× in figure 67.
with the reduction of turbulence growth rate by magnetic Another code example is the dynamic evolution of electron
shear [516, 566, 568]. More fundamental treatment of the temperature, density and current density after an ELM crash
turbulence and its suppression is another interesting approach calculated by the JETTO code [567], which is shown in
when integrated with a transport code [569, 570]. Detailed figure 68. In this calculation, the pedestal width is prescribed
treatment of neutrals penetrating from the SOL is another and all transport coefficients within the pedestal are reduced to
important ingredient for the modelling of pedestal structure, the level of ion neoclassical thermal conductivity. When the
since the density pedestal may be dominated by the localization ELM stability criterion for either ballooning or kink modes is
of ion source from neutrals [569, 570]. Table 2 summarizes violated, transport within the barrier is temporally increased.
the existing codes and their employed physics models for the The critical pressure gradient and edge current are variable
modelling of pedestal structure. Somewhat more detailed numerical parameters, which are checked against the results of
descriptions and calculation results with some of the codes the MHD stability codes IDBALL and MISHKA and adjusted
as examples are provided below. accordingly. Both the pressure gradient and the edge current
The integrated ICPS model presently consists of two parts: drop as a result, so the plasma returns to the pre-ELM state and
the core, from the centre of the plasma to the separatrix the cycle repeats [576].
and including the pedestal, is modelled with the 1-1/2 D
code ASTRA, and the scrape-off layer and divertor plasma 4.6. Type I ELM structure, effects on the pedestal profiles and
are modelled using a parametrization of separatrix plasma mitigation techniques
conditions obtained from the coupled plasma fluid–neutral
Monte Carlo code B2-EIRENE for ITER conditions. A variant Type I ELMs are spontaneous periodic relaxations of the
of the model, which uses a simpler two-point model for the pedestal of H-mode plasmas, occurring in a repetitive manner
scrape-off and divertor plasma, is used to fit results from once the applied power exceeds ∼1.5–2 times the L–H
present-day experiments [516, 566]. B2-EIRENE simulations threshold power [577, 578]. Type I ELMs have similar
revealed the existence of two regimes, one at moderate divertor characteristics in all devices, and their crash dynamics, MHD
pressure and a second at higher (saturated) divertor pressure signature and stability are known better than for any other
[571]. Scaling relations were developed in each regime ELM type. Type I ELMs are replaced, at low input power
[571,572], and these were then used as boundary conditions for (compared with the L–H threshold power) or at high density,
the core plasma modelling [516, 566, 573–575]. The transport by Type III ELMs. This section deals mainly with Type I
model for the core–pedestal part is neoclassical plus a modified ELMs but reference is made to Type III ELMs when significant
multi-mode model (MMM95) for plasma turbulence. The differences are observed or specific experimental information
particle transport coefficient is taken as 0.1(χe + χi ), where is available. This section first describes the ELM cycle and the
the χ’s are electron and ion thermal diffusivities, and the effect of ELMs on the pedestal profiles. Then, the influence of
Ware pinch is also included. Sawteeth are approximated by plasma parameters, such as q95 and plasma shape on ELMs, is
increasing the transport coefficients by a factor of four inside discussed. And last, recent progress in Type I ELM mitigation
the q = 1 surface. The effect of ELMs is represented in a techniques is briefly reviewed.
time-averaged sense by limiting the pressure gradient to the
ballooning limit. The turbulent transport is assumed to be 4.6.1. The ELM cycle—build-up and collapse of the pedestal.
reduced by the radial electric field shear and magnetic shear, Figure 69 shows the time evolution of the Dα emission from the
according to divertor, plasma stored energy and pedestal temperature and
density in a JET Type I ELMy H-mode [497]. After the fast
χ = χMMM {(1 + (ωE×B /(Gγ0 ))2 ) · max(1, (s − t)2 )}, (29)
collapse of the pedestal, both ne and Te start to build up again,
where γ0 = γITG  is the volume averaged growth rate for ITG until the next ELM occurs. The ELM crash occurs on very fast
mode turbulence for the normalized radius ρ
0.9 and is an timescales, of the order of 100–300 µs [579–582], with very

S85
E.J. Doyle et al
Table 2. Comparison of physics models incorporated in existing codes for modelling of pedestal structure.

Code

LEHIGH [568]
JETTO [567] (BALDUR, JETTO, ASTRA
EDGE2D together with HELENA and
Model ICPS [516, 566] NIMBUS MISHKA) BOUT-UEDGE [569] XPTOR [533]

1. Transport models Originally; modified Bohm/gyro-Bohm MMM95 [236, 238, 239] Self-consistent GLF23 [20] has
for IFSPPL model for Weiland GLF23 [20] turbulent fluxes been used to
turbulent ions and modified MMM95 Mixed-Bohm from 3D resistive model the ion
plasma RLW transport for /gyro-Bohm ballooning modes, particle,
electrons including X-point electron and
Updated to modified and EM ion thermal,
multi-mode model effects and momentum transport
(MMM95 within the pedestal
[236, 238, 239]) region
2. Mechanisms of E × B + magnetic E × B + magnetic E × B shear E × B and magnetic shear E × B shear
turbulence shear for shear or included in + magnetic shear
suppression in growth rate simply magnetic shear turbulence + shafranov shift
pedestal region stabilization
3. Transport model in Neoclassical Local neoclassical NCLASS + residual Residual NCLASS
the pedestal region (NCLASS) + fraction of anomalous turbulence +
after suppression of anomalous or non-local transport used in ASTRA code estimate from simple
turbulent transport neoclassical on Ion thermal neoclassical
the top of neoclassical expression
barrier transport at top of pedestal
used throughout
pedestal in
JETTO
4. Model for pedestal Self-consistently Fixed or given Self-consistently Edge profile Self-consistently
width determined by formulae (ASTRA) evolved determined
with transport Fixed or given self-consistently from with transport
suppression by formulae fluid neutral suppression
model (JETTO) model source model
Models for pedestal and residual
width in turbulence
[721] (BALDUR)
5. Model for Analytical formula Ballooning criteria Ballooning criteria Pressure and MHD constraints not
critical for ideal for αcrit for αcrit or current terms presently employed
pressure gradient and ballooning mode or analytical analytical peeling (ballooning/peeling) that Transport is
ELM trigger peeling criteria criteria for drive ELMs assumed to be
for current current (JETTO, ASTRA) are in the electrostatic
MISHKA is used Whichever mode simulation equations, within GLF23
routinely to becomes similar to
check the validity of unstable first ELITE
analytical causes the ELM
criteria crash
6. Model for Not explicitly Gaussian shape for Dynamic model ELM crash Presently not
transport treated Time χI with for ELMs and full ELM being treated
during ELM cycle averaged pedestal width from MISHKA (JETTO, ASTRA) cycle under
structure Amplitude and Each ELM crash development
is calculated duration can produces a
be prescribed or rapid change
taken from in the edge
a solution profiles for T ,
of corresponding n, and j ,
time evolving equation followed by
a rebuilding of
the pedestal
7. Model for MHD Not solved ELM crash and full ELM Presently not
‘equilibrium’ during cycle under development treated
ELM
8. Model for particle 0.1(χi + χe ) Bohm/gyro-Bohm Self-consistent GLF23 + NCLASS
transport and Ware pinch Weiland and MMM from fluid for ions Electron density from
(including turbulence quasi-neutrality using
off-diagonal predicted main ion
elements) profile along with
Ware pinch experimental impurity
NCLASS pinches and fast ion
density profiles
9. Model for Astra routine based on Diffusive equation for Monte Carlo (JETTO, Various models Particle source
neutrals Dnestrovski and neutrals in JETTO BALDUR) including taken from
Kostomarov (FRANTIC) fluid, analytic experimental analysis
For ITER, direct Monte Carlo or Monte (TRANSP or ONETWO)
core fuelling for EDGE2D Carlo
with profile (NIMBUS) (seldom used and
calculated from not yet with turbulence
simple NBI –transport coupling)
model (without
energy input); results
insensitive to
penetration
depth for
penetration inside
pedestal radius
10. Boundary condition Scaling relations Density and Prescribed (JETTO, ASTRA) Extends from Boundary conditions
for separatrix values calculated by temperature at the Boundary taken at top of pedestal to wall, enforced inside
B2-EIRENE separatrix are usually pedestal from thus including separatrix
prescribed, sometimes taken pedestal model separatrix location using
from EDGE2D (BALDUR) experimental values
S86
 Chapter 2: Plasma confinement and transport

0.8 4.6.2. Effect of ELM crash on n and T profiles. The success

4 0.7 of the peeling–ballooning model for the prediction of Type I
0.6 ELM onset (see section 4.8 and references therein) has brought

n [10 m ]
3

-3
0.5 about a hypothesis linking the change in mode number and
T [keV]

20
2
0.4 spatial extent of the most unstable mode to the ELM size. The
e

0.3 correlation between the plasma volume affected by the ELM

e
1 0.2 crash and ELM size has been analysed in detail using fast
0.1 electron temperature profile measurements in JET [580, 591].
0 0
0 0.4 0.8 0 0.4 0.8 An example of such a study is shown in figure 70, for a density
r
mid
[m] r
mid
[m] scan at fixed plasma parameters: the plasma depth affected by
bdry
an ELM is determined by subtracting the post-ELM Te profile
t=0.5 G=0.5 t=0.5 G=1.0
expt. no thresh G=1.0 CG model
from the pre-ELM Te profile (the two profiles are separated in
time by ∼300 µs, and the time resolution of the measurement
Figure 66. Profiles for JET: experimental, MMM model results for is ∼10 µs), and the perturbed profiles are then normalized to
the two parameter sets and without shear stabilization and CG their maximum. While the energy loss per ELM (WELM )
(critical gradient) model results. Left—electron temperature.
Right—electron density. Reprinted with permission from [516].
decreases by a factor 2 to 3 as the density increases, the plasma
depth affected by the ELM does not change, as long as Type
I ELMs are present. Similar results are found for giant ELMs
little variation observed as function of plasma parameters and in JT-60U [592], as well as in DIII-D (figure 9 in [593]).
machine size. After the crash, the recovery of the pedestal top Further analysis of the Te and ne profile perturbations at the
values occurs on a much longer timescale, accompanied by ELM crash reveals that the change in ELM size with density is
the rebuilding of the pedestal profiles. The ELM crash causes due to a strong reduction in the amplitude of the temperature
a widening of the pedestal profiles, which then progressively drop at the ELM, while the magnitude of the density crash
narrow down and become steeper until the next ELM occurs stays approximately constant with density [579, 593]. At high
(see for instance [583] for a detailed study of the dynamics pedestal density, Type I ELMs can occur without any change in
of the density profiles in ASDEX Upgrade and [580] for the the pedestal temperature, i.e. with the ELM loss channel being
behaviour of Te pedestal profiles on the ELM timescale in purely convective [580, 593]. In the case of MAST [594],
JET). The comparison of the ELM density collapse at low all ELMs appear to be purely convective, although in that
and high field sides [583–585] indicates that the ELM crash case the ELM classification (Type I or Type III) is not clear.
occurs first in the low field side of the tokamak, with the density Data from NSTX [595] also indicate that ELMs are prevalently
perturbation propagating to the inboard side at approximately convective, in this case for Type I ELMs.
the speed of ion sound. This observation is consistent with The beneficial effect of triangularity δ in achieving a
the MHD interpretation of ELM events as ballooning-type high density at high confinement [494, 497] has stimulated
instabilities. intensive studies of the pedestal and ELMs at the high δ, near
ELM MHD precursors have been studied in most that foreseen for ITER. In general, increasing δ corresponds
tokamaks (a comprehensive review is found in [586]). Clear to an increase in the pedestal pressure and the absolute
precursors correlated to the onset of the ELM have been ELM size, although in most cases [579, 580], the relative
identified in ASDEX Upgrade for both Type I and III ELMs ELM losses (i.e. the energy loss per ELM normalized to the
pedestal energy content) depend weakly on the triangularity
[587] and in JT-60U for Type I ELMs [581,582]. This is not the
(see chapter 4 of this issue [364]). JET [580] and DIII-
case for JET and DIII-D, where a cause–effect between MHD
D [596] data indicate that the plasma depth affected by
precursors and the Type I ELM crash has not been clearly
the ELM decreases with increasing δ, although the absolute
established [588, 589].
size of the ELM increases. In contrast, [597] reports a
As also shown in the example of figure 69, it is commonly triangularity dependence of the relative ELM losses and
observed that dTe /dt between ELMs is not constant and that Te ELM-affected depth in ASDEX Upgrade. The role of
may reach saturation well before the ELM crash. A possible plasma shape (in particular of triangularity and squareness)
correlation between the inter-ELM transport (or the rate at in the MHD stability of the pedestal and ELMs has been
which the pedestal profiles recover after an ELM and ultimately clearly demonstrated by the analysis of DIII-D shape scans
the ELM frequency) and MHD activity is investigated in [590] [598, 599], where shape variations, pedestal pressure and
in JET ELMy H-mode plasmas. A correlation is found between access to second stability have been convincingly linked. In
the intensity of washboard modes (band(s) of fluctuating NSTX [595], both the frequency and size of Type I ELM
magnetic activity rotating in the electron diamagnetic drift vary considerably between SN and DN plasmas, although a
direction) and the rate of rise of the pedestal temperature link between ELM characteristics and the distance between
between ELMs, which slows down or even saturates for the the first and second separatrix has not been identified
increasing intensity of the WB mode activity. The correlation so far.
between these MHD modes (possibly of resistive origin) and Recent JET experiments have found a dependence of the
the build-up of the electron temperature give strong indications pedestal pressure [600] and ELM size [592] on the safety factor.
that WB may be responsible for an enhanced transport of In particular, it appears that increasing q95 from ∼3.6 to 4.6
energy across the separatrix, although this enhanced transport causes a reduction in relative ELM losses of more than a factor
is not sufficient to saturate the density increase between ELMs. of 2, due to a decrease in the amplitude of the temperature drop

S87
E.J. Doyle et al

1.6 2.5 1
1.4
2 0.8
1.2

ne [1020 m-3]
1 0.6
Te [keV]

1.5

Ti [keV]
0.8
1 0.4
0.6
0.4 0.5 0.2
0.2
0 0 0
0 0.4 0.8 0 0.4 0.8 0 0.4 0.8
<ne > [1020 m-3] <n e > [1020 m-3] <ne > [1020 m-3]

exp at 0.95a exp at edge t=0.5 G=0.5

Figure 67. Simulated and experimental edge parameters for a density scan for ASDEX Upgrade (t = 0.5, G = 0.5). Reprinted with
permission from [516].

Figure 68. Evolution of electron temperature, density and current density after ELM crash from the JETTO code.

at the ELM, at a constant perturbation radius. Comparison of without Type I ELMs are described in section 4.7. These
these findings with data from other devices is a matter of high techniques are of interest for ELM studies because of the
priority for future work. insight they can provide on the physics of ELMs, as well as for
their possible applications to ITER. The extrapolation of these
4.6.3. Type I ELM mitigation techniques. The concern ELM control techniques will require a detailed assessment
about the lifetime of the divertor target plates in ITER, due to quantify how much of the decrease in ELM energy loss
to erosion and/or sublimation under the high transient power comes from the method itself and how much comes from
loads associated with Type I ELMs [601], has stimulated reducing the pedestal pressure, and therefore the plasma’s
research of control techniques and plasma regimes that couple global performance.
the required global plasma confinement properties to tolerable Experiments where the plasma edge current is changed
ELM loads of plasma facing components. This section reviews by external means were carried out in COMPASS-D [602]
ELM control techniques relying on external actuators, namely and JET [603]. In both the cases, an increase in the edge
control of edge current, edge magnetic field ergodization in H- current density Jedge was associated with a modification of the
mode plasmas and pellet-induced ELMs, while plasma regimes ELM behaviour: in COMPASS-D this increase is associated

S88
 Chapter 2: Plasma confinement and transport

Pulse No: 53767

10
Divertor H
8
a.u.

6
4
2
total stored energy
5.3
W (MJ)

5.2
5.1

1100 Tped
Te (eV)

1000 end of ELM cycle

900
beginning of ELM cycle
4.4 nped
ne (1019 m-3)

minimum density
4.0 Figure 71. Example of Type I ELM suppression by application of

JG02.41.26c
an edge resonant magnetic perturbation by the ‘I coils’ in the
3.6 DIII-D: comparison of discharge115468 (I coils off, black) and
23.27 23.28 23.29 23.30 23.31 115467 (I coils on, red). In order from top to bottom: Dα recycling
Time (s) at lower divertor for I coils off (first box) and on (second box);
plasma density and gas fuelling (box 3); plasma total stored energy
Figure 69. Example of a typical ELM cycle (Type I, JET). The time (box 4) and electron pressure at the top of the pedestal (last box).
evolution of the Dα emission from the divertor, the plasma stored The shaded region indicates the time when the ‘I coils’ are pulsed on
energy W , pedestal electron temperature Te and density ne are shown in discharge 115467 [749].
in boxes 1 to 4. The fast collapse of Dα , Te and ne at the ELM crash
are highlighted by the arrow. Reprinted with permission from [497].
ELMy H-modes, with the frequency of the perturbation ∼2
times the natural ELM frequency of the plasma under study.
The resulting up/down movement of the plasma in an up/down
asymmetric field results in periodic variations of Jedge with
good correlation between the perturbation induced by the coil
(when in the direction of increasing Jedge ) and the generation
of an ELM, with the average ELM frequency increasing by
upto a factor of two, compared with a control case. Further
work is required, for assessing the technical viability of this
technique for ITER, to quantify the achievable reduction in
the ELM size as well as the impact of such a control method
on plasma confinement. This technique has been applied in
the ASDEX Upgrade tokamak, and ELMs have indeed been
generated by plasma ‘jolts’, as described in [605].
The application of a magnetic perturbation resonant at the
edge of the plasma is another method that has been studied for
ELM control. Experiments in COMPASS-D [602] showed
that the application of a resonant field could increase the
frequency of Type III ELMs, as well as induce ELMs in
Figure 70. Perturbation of the electron temperature profile caused ELM-free H-modes. The same conceptual method has been
by ELMs (normalized to its maximum) versus normalized major recently used in DIII-D [606], where the suppression of Type I
radius, for a density scan at fixed plasma parameters in JET. With ELMs at constant plasma confinement has been demonstrated
Type I ELMs, the ELM-affected volume is constant for a variation
using in-vessel magnetic field coils (figure 71). Further
of WELM ∼ 2–3, while it decreases with Type III ELMs. Reprinted
with permission from [591]. experimental work is required to extend these promising results
to ITER. In particular, Type I moderation or suppression should
be extended to a range of relevant plasma parameters (and
with bursts of ELM activity (Type III ELMs), while in JET possibly different devices), as well as to demonstrate the
an increase in Jedge is invoked to explain the suppression of compatibility of this control scheme with the overall plasma
Type I ELMs (in favour of Type III ELMs) in plasmas with MHD stability.
a non-monotonic q profile. In both the cases, the effects The possibility to control ELM frequency and size by
are claimed to be consistent with the destabilization of edge pellet injection was put forward in [607], observing that
peeling modes. A possible control scheme, exploiting the pellets trigger ELMs and that those pellet-triggered ELMs may
effect of edge currents on the MHD stability of the pedestal be associated with a reduced power load onto the divertor
and on ELM frequency (and size), has been investigated in plates. More recently, ELM control by a repetitive pellet
TCV [604]. In these experiments, square voltage perturbations injection was demonstrated in ASDEX Upgrade [608] (see
are applied to the poloidal field coils, on single null, Type III figure 89 in section 4.9). Pellet size and velocity were adjusted

S89
E.J. Doyle et al

(within the technical capability of the system) to trigger ELMs

without contributing significantly to core plasma fuelling. It is
observed that ELM triggering occurs only ∼200 µs after the
injection, suggesting that only a fraction of the pellet mass is
sufficient to destabilize locally the pedestal and cause an ELM.
It is found that MHD signature and divertor power loads of
pellet-triggered ELMs are similar to those of ‘natural’ Type I
ELMs of the same frequency (the frequency is adjusted by
means of an external gas fuelling in a control discharge with
all parameters as in the pellet case). Although, to establish a
basis for extrapolation, this technique has to be extended to a
wider range of plasma parameters, in particular towards high
Te pedestals, as well as a rigorous assessment of the impact
on confinement, code simulations [609] indicate that shallow Figure 72. Electron temperature versus electron average pressure
pellet injection has the potential for Type I ELM mitigation gradient. Type III ELMs occurs only below a certain edge
in ITER. temperature limit. Reprinted with permission from [274].
Other techniques, such as the direct current drive in
the plasma periphery for edge current control (and ELM
performance projections than is now available for Type I
suppression), are being investigated at the moment in several
ELMs. Inter-machine comparisons have proven particularly
devices, and first results should be available in the near future.
useful in regard to understanding the conditions for transferring
and extrapolating the regimes. Issues requiring further study
4.7. Alternatives to Type I ELMy H-mode regime are discussed in section 4.7.6.
H-modes with Type I ELMs are by far the most common
high confinement regimes in today’s tokamaks, and their 4.7.1. Type III ELMs. Type III ELMy H-modes are
global performance and pedestal parameters have been studied commonly observed in all divertor tokamaks [274, 614–616].
extensively. Accordingly, this regime has been chosen as Compared with the Type I ELMy regime, H-modes with Type
the reference operating scenario for ITER and other divertor III ELMs have higher ELM frequency, smaller energy loss per
burning plasma experiments. There is good confidence that ELM, WELM , but reduced energy confinement (by 10–30%).
the regime can be accessed and that plasma confinement will In terms of reproducibility and robustness of the regime, Type
be sufficient to meet the fusion performance targets. The III ELMy H-modes are comparable to H-modes with Type
periodic energy losses due to ELM crashes are included in I ELMs. The Type III ELMy regime limits the operational
these projections and they do not have a major impact on space for Type I ELMs. In fact, Type III ELMs are observed
this performance. There is a concern, however, as discussed at low power above the L–H threshold power, so the transition
in section 4.6 of this chapter as well as in chapter 4 of this to Type I ELMs requires a minimum power. Moreover, the
issue [364], about the impact of these losses on the surfaces transition to Type III ELMs limits the density achievable in
of the divertor and the first wall. If too large, repeated the Type I ELMy H-mode regime (i.e. the density achievable
heat pulses could cause surface erosion and/or melting, with good confinement). In contrast to Type I ELMs, Type III
leading to a reduced divertor lifetime and increased impurity ELMs are commonly observed also in plasmas with an internal
influxes. Another issue, mainly arising in advanced scenarios transport barrier, where the benefit of a Type III ELMy edge
as discussed in section 3.7, is many experiments find that can be combined with the reduced core transport [611]. The
large ELMs can disrupt or terminate internal transport barriers, study of the physics mechanisms driving Type III ELMs, and
through coupling to core MHD and/or transient modification of the scaling of the operational space of Type III ELMs with
of profiles [377, 603, 610–613]. It is recognized that some machine size and plasma parameters, is therefore important to
mechanism is needed to increase particle transport at the be able to control or avoid this regime in ITER.
edge of an H-mode plasma, since otherwise particles and Type III ELMs are observed below a critical pedestal
impurities tend to accumulate. To avoid these potentially temperature, Tcrit , and at pedestal pressures at or below that
serious drawbacks, it would obviously be preferable to have of Type I ELMs, as shown in figure 72 [274]. Tcrit increases
a regime which gave similar plasma performance to the with decreasing density (Tcrit ∝ 1/n) at low density [614,616].
Type I ELMy regime, which has controlled density and Type III ELMs of standard ELMy H-modes and of plasma with
impurities but in which heat loss is more continuous and evenly an ITB belong to this low collisionality branch of the pedestal
distributed, either with smaller periodic ELMs or a continuous ne –Te diagram. In this region of ne –Te , consistent with the
edge fluctuation. This has motivated increasing attention decrease in Tcrit with density, it is found that increasing density
to alternative regimes in the years since the ITER Physics decreases the power required for the transition from Type III
Basis [1]. Several candidate regimes have been identified and to Type I ELMs. At high density, in contrast, Tcrit is low
there are active research programs on all the world’s divertor and almost independent of density. This high collisionality
tokamaks. The key experimental results on each regime are branch of ne –Te includes the high-density ELMy H-mode with
summarized below. It must be recognized that, since most of or without impurity seeding. ASDEX Upgrade [293] and
these regimes are comparatively new, there is a less complete Alcator C-Mod [615] have only observed Type III ELMs at
physics picture of their accessibility, pedestal parameters and high collisionality, while in DIII-D [614] and JET [616] they

S90
 Chapter 2: Plasma confinement and transport

3500 [620]. Although the uncertainties in the extrapolation are large,

Type III Elms, Bt = 3.4T at low q95 (= 2.7) and with H98(y,2) = 0.75–0.8 at n = nG , this
Type III Elms, Bt = 2.4T scenario has a combination of plasma parameters that just falls
3000 & Type III Boundary (predicted)
within the predicted operation domain for the Q = 10 ITER
inductive operation at 17 MA.
2500 Although fewer data exist for Type III than for Type I
ELMy H-modes, they indicate that the fraction of pedestal
energy loss at the ELM, WELM /Wped , decreases with
Te,ped (eV)

2000
increasing pedestal collisionality [620], similar to the trend
found with Type I ELMs. At the same collisionality, there is
1500
no difference between the WELM /Wped of impurity seeded
and non-seeded Type III ELMy H-modes, and this appears to
1000 be lower than for Type I ELMs. This result is consistent with
the fraction of plasma volume affected by the ELM temperature
500 crash being smaller for Type III ELMs than for Type I ELMs

JG99.383/8c
[591].
0
0 2 4 6 8 4.7.2. Enhanced D-alpha and similar regimes. The
ne,ped (1019 m-3) ‘Enhanced Dα ’ (EDA) H-mode regime was the first regime
observed to provide control of particles and impurities without
Figure 73. Pedestal ne –Te diagram for Type III ELMs in JET and any periodic ELMs. It was first observed on Alcator C-Mod
comparison with the model. Reprinted with permission from [616]. in 1996 [621–623]. Global confinement can be as good as in
Type I ELMy regimes, with HITER89P up to 2, while density
are found also at low collisionality [616, 617]. Tcrit tends to and radiated power are steady in contrast to usual ELM-
increase with toroidal field in both high and low ne branches free H-modes. The regime is favoured by higher q95 and
[615, 616]. triangularity and, for deuterium plasmas, is obtained most
A model for the Type III ELM instability [618] that reliably at q95 > 3.5 and δ > 0.35 [624]; some examples
considers the resistive interchange instability with magnetic at lower q and q have since been observed and the q95 limit
flutter (RIF) reproduces the density dependence of Tcrit for in hydrogen plasmas is 2.5 or lower [625]. ELM-free and
Alcator C-Mod, ASDEX Upgrade, DIII-D and JET. In JET, this EDA discharges are best separated in edge operational space
model also describes the toroidal field dependence of Tcrit , as by the collisionality and normalized pressure gradient, with
shown in figure 73. Nevertheless, the JET results also show that ν∗95  1.5 at the highest αped [527].
this model does not predict correctly the experimental variation The salient feature of this regime is a continuous,
of Tcrit with isotopic mass and q95 [616]. localized fluctuation referred to as a quasi-coherent (QC)
Whether Type III ELMs at low and high collisionality mode. Detailed multi-diagnostic measurements have been
are due to the same physics mechanism is however still done in recent years of fluctuations in density, electrostatic
an open question, since there are results that point to a potential and magnetic field [176, 625–629]. The mode is
different instability for the two collisionality branches. The localized to a region of a few mm near the bottom of the
result that a plasma current ramp down can stabilize low density pedestal, has a peak frequency in the range 50–120 kHz
collisionality Type III ELMs suggests current driven peeling and is of fairly short wavelength (i.e. high m, n), with k⊥ ∼
modes as the driving instability for those ELMs [612,616]. At 1.2–4 cm−1 . Accordingly, the magnetic perturbation falls
high collisionality, a model based on the resistive ballooning rapidly with distance from the mode surface. Figure 74
instability [619] reproduces correctly the JET data of the shows magnetic and density signatures on Alcator C-Mod.
critical density for the transition from Type I to Type III as The mode amplitude is variable and correlates well with
ne,crit = Bt /q 5/4 . the effective particle transport; with a weak amplitude ne
The energy confinement enhancement factor, H98(y,2) , for radiation can still rise, while a very strong mode can start to
Type III ELMs decreases with density and increases with increase energy as well as particle transport. Target density
triangularity in a similar manner as for Type I ELMs, indicating and neutral pressure have been shown to correlate with the
that those trends are not specifically related to the ELM type. existence and strength of the mode. The observed scalings of
Although the transition from Type I to Type III ELMs results the fluctuation characteristics and their regime of occurrence
in a quantitatively similar loss of confinement both at low and suggest that the QC mode may be some type of resistive drift-
high collisionality, the observed decrease in pedestal pressure ballooning mode. Features similar to the QC mode have
at the transition is due to cooling of the pedestal at high been observed in electromagnetic edge turbulence simulations
collisionality, while it is due to loss of pedestal density (at [627,630]. Simulations of an Alcator C-Mod experiment with
constant or increased temperature) at low collisionality [616]. the boundary-plasma turbulence code BOUT show a resistive
The increase in confinement at high density for Type III X-point mode in good agreement with observations [176].
ELMy H-modes at high triangularity has been exploited in JET Studies of pedestal profiles and stability have shown that
to demonstrate a radiating Type III ELMy H-mode scenario there is not a marked difference in the width of ne and Te
(radiative power fraction > 70%) where low inter-ELM heat pedestals between ELM-free and EDA plasmas [631]. Widths
flux to the divertor and low ELM energy losses are combined vary from 2 to 8 mm [527,625,632] and show little systematic

S91
E.J. Doyle et al

EDA appears at lower edge density and higher edge temperature. It

is found that the ELMy/HRS operational boundary occurs at
the normalized electron collisionality of νe∗ ∼ 1 in the plasma
edge region, depending slightly on q95 (figure 76(d)).
A key feature of the HRS H-mode is the presence of
coherent magnetic fluctuations in the frequency range of
the order of 10–100 kHz. It is suggested that the edge
MHD activity may keep the edge pressure below a certain
level needed to induce a large ELM. Detailed fluctuation
measurements with magnetics, reflectometry and probes
[638, 639] show that there are both a high frequency mode
of f  200 kHz with n ∼ 7 and a lower frequency mode of
f ∼ 50–120 kHz with n = 1.
Recent inter-machine comparisons between Alcator C-
Mod and JFT-2M, at matched plasma shape, show very similar
access conditions, suggesting that the EDA and HRS H-modes
Figure 74. Measurements of the quasi-coherent mode in an are in fact the same regime though there are differences in
enhanced Dα H-mode on Alcator C-Mod, made by a magnetic probe
close to the LCFS (top) and a reflectometer channel probing the
fluctuation details [640]. Both the EDA and HRS regimes have
density pedestal (middle). Reprinted with permission from [629]. been successfully combined with internal transport barriers for
long durations [378, 398, 636, 641–643], giving steady double-
variation with ρi,pol or other plasma parameters. Widths of the barrier regimes.
soft x-ray emission do vary and are narrower and shifted to
smaller midplane radii in ELM-free plasmas. This reflects 4.7.3. ‘Quiescent’ H-mode. The ‘quiescent H-mode’ (QH
the differences in the impurity pinch and diffusivity [633]. regime) was first observed on DIII-D in 1999 [644] and has
Discharges with a QC mode, but no ELMs, are found to be ideal been reproduced most robustly on ASDEX Upgrade [645,646].
MHD stable despite steep gradients of up to dp/dψ ∼ 1.2 × Like the EDA regime, it offers good confinement, similar to
107 Pa Wb−1 when edge bootstrap currents are consistently that of the Type I ELMy regime, and has particle transport
included. As discussed further below and in section 4.8, at sufficient to stop the accumulation of ne and impurities.
higher Tped (typically for Tped > 400 eV) and/or ∇p, the However, the details of the responsible mode, and its regime
QC mode broadens and small ELMs, tentatively identified of accessibility, are quite different [466]. An edge fluctuation
as Type II can appear on top of a generally enhanced Dα termed the ‘edge-harmonic mode’ is observed which has a
level [634]; the two regimes appear closely linked in their much longer wavelength than the quasi-coherent mode. It
operational space. is typically seen by magnetics diagnostics and on ne and Te
Experiments aimed at reproducing the EDA H-mode diagnostics and is localized at the foot of the pedestal. The
have been carried out on DIII-D, JET and ASDEX Upgrade oscillation is coherent and periodic, with a base frequency
[530, 617, 635]. On DIII-D and ASDEX Upgrade, when of ∼ 6–10 kHz, but not sinusoidal, giving a variable mix of
edge dimensionless parameters β, ν∗ and ρ∗ were matched toroidal wavenumbers, e.g. n = 1, 2, 3, and/or 4, etc and
to values on Alcator C-Mod, an edge fluctuation similar to has a poloidal wavelength of ∼1 m. An example is shown
the QC mode appeared, indicating that the access regime in figure 77. Up to 11 harmonics have been seen on ASDEX
is well described by such variables. Figure 75 shows an Upgrade [646]. On ASDEX Upgrade, in addition to the low-
example of such an experiment on Alcator C-Mod and DIII-D, f EHO, a much higher frequency magnetic mode (named the
in both an ‘EDA’ (left) and ‘small ELM’ (right) regime. HFO) is also observed. This has f ∼ 350–500 kHz and its
The operating window varies between machines, partly due modulation envelope corresponds closely to the waveform of
to differences in L–H thresholds. It has proven difficult, the EHO, as shown in figure 78. This raises the possibility that
however, to produce a steady-state ELM-free regime on these the EHO is a manifestation of the higher frequency instability.
experiments; edge densities tend to rise and periodic ELMs Due to diagnostic resolution it is not certain whether the HFO
are observed, indicating that the mode strength is insufficient also exists on DIII-D.
to control particle transport. A regime similar to EDA was The ELM-free QH regime has only been observed on
recently seen on JET in an experiment which matched the shape DIII-D with counter-NBI, though the EHO is sometimes seen
of Type II ELMy discharges on ASDEX Upgrade [600]. on ELMy H-modes with co-NBI [647]. Rotation is always in
A regime very similar to the EDA has been achieved on the direction of neutral beams. A large plasma–wall distance
the JFT-2M tokamak and named the ‘high recycling steady’ at the LFS of the tokamak is also required, suggesting the
(HRS) H-mode [636]. It is also characterized by steady role of fast ions. On ASDEX Upgrade, the QH regime has
conditions without large ELMs (figure 76(c)). The regime also been obtained with near-radial NBI and is facilitated
is again favoured by higher q95 , higher density and neutral by the additional application of ICRH, which would increase
pressure. Recent experiments in the JFT-2M tokamak have the fast particle population, though RF is not required [645].
concentrated on the studies of the access conditions for the On JT-60U, ELM frequency and size both clearly decrease
HRS H-mode regime in terms of the pedestal parameters [637]. as beams are shifted from co to counter-injection [648]. A
The HRS regime is more likely at higher edge density and lower quiescent, ELM-free regime occurs with full counter-injection.
edge temperature, while the ELMy H-mode having large ELMs ELM-free phases and fluctuations similar to the EHO have

S92
 Chapter 2: Plasma confinement and transport

Figure 75. Example of a matching experiment between Alcator C-Mod and DIII-D. DIII-D pedestal parameters (blue) are scaled to the
machine size of Alcator C-Mod (red). When parameters are matched, similar fluctuation and global behaviour is seen in both the EDA (left,
0.820–0.92 s) and small ELM regime (right, 0.92–1.25 s). Reprinted with permission from [530].

3 2
(a) #99937 (ELMy)
pe, edge [kPa]

Dα [a.u.]

2
1
1

0 0 ELMy (large ELMs)

650 700 750 Mix. (grassy ELMs)
3 2 HRS (small ELMs)
(b) #99940 (Mix) 6
p e, edge [kPa]

(d)
Dα [a.u.]

2
1
5
1
q95

0 0 4
600 650 700
3 2
(c) #99941 (HRS) 3
[kPa]

Dα [a.u.]

2
1
2
e, edge

1 0.1 1 10
νe*
p

0 0
600 650 700
Time [ms]
edge
Figure 76. Time history of the edge electron pressure pe (solid) and Dα emission (dotted). (a) ‘ELMy’ (#99937), (b) ‘Mixture’ (#99940)
and (c) ‘HRS’ (#99941). (d) Plot of ELMy (blue), Mix. (green) and HRS (red) operational regimes in safety factor at the 95% flux surface
q95 versus edge normalized electron collisionality ν∗e . Reprinted with permission from [637].

also been seen during some counter-injection experiments have shown that with stronger shaping [650, 651] the pedestal
on JET, though a steady-state QH regime has not yet been density and pressure can be increased; n̄e is in the range (1.7–
accessed. The mechanism and conditions for the EHO, and its 7.4)×1019 m−3 and q95 = 3.4–5.8, raising the electron and ion
relation to the HFO and fast particles, remain open questions. collisionalities to 0.5 and 1.4, respectively. This presumably
Pedestal pressures and gradients in the QH regime are reflects changes in ELM stability. The limits of the operational
comparable to those in ELMy H-mode [388, 649]. In contrast space in this regime are thus not fully known. QH-mode
to the EDA and HRS regimes, access to the QH regime also has been combined with internal transport barriers, and
is favoured by relatively low edge densities, in some cases the resulting QDB regime has been extensively studied on
requiring wall pumping and high pedestal temperatures; most DIII-D [200, 242, 466]. This high performance regime can
discharges have low pedestal collisionality, with typical νe∗ ∼ be maintained for the duration of the plasma discharge. An
0.04 and νi∗ ∼ 0.14. However, recent DIII-D experiments issue for extrapolation to burning plasmas is that Zeff tends to

S93
E.J. Doyle et al

ELMs have also been identified in high power H-mode plasmas

in Alcator C-Mod, with βN > 1.2, in discharges that exhibit
an EDA edge at lower input powers [624, 634].
The interest in Type II ELMy H-modes originates from the
attractive global characteristics of these plasmas, combining
a drastic reduction in ELM transient power loads to plasma
facing components with steady-state high density (upto
∼0.85–0.95nG ), good global confinement (H98(y,2) ∼ 0.95,
a reduction from standard ELMy H-mode confinement with
Type I ELMs of < 10%) and no impurity accumulation [653].
Type II ELMs do not cause a measurable perturbation of the
pedestal temperature or density profiles and therefore are, in
principle, compatible with ITBs.
Type II ELMs are characterized by a specific enhancement
of MHD fluctuations in the plasma edge region. In ASDEX
Upgrade, as well as in JET and Alcator C-Mod, Type II
ELM activity is associated with an increase in broadband,
low frequency magnetic fluctuations (30–50 kHz region).
In the case of JET, these fluctuations are attributed to an
increase in the intensity of so-called washboard modes [590],
always observed between ELMs in standard ELMy H-modes.
Enhanced density fluctuations in the plasma edge region
are also observed both in the ASDEX Upgrade [653] and
in JET [600, 654]. It is hypothesized that these enhanced
fluctuations are responsible for the quasi-continuous power
and particle losses associated with Type II ELMs, although
direct measurements of turbulent fluxes are not yet available.
The gradual nature of the transition from Type I to
Type II ELMs, and the fact that these two types of ELMs
can co-exist, complicates somewhat the definition of exact
access conditions. Analysis of ASDEX Upgrade results
shows that the key element required to access a pure Type II
ELMy H-mode is the proximity to quasi-double-null (QDN)
configuration (the distance between the first and second
Figure 77. Frequency spectra of the edge-harmonic oscillation separatrix at the midplane DX needs to be < 1 cm) and plasma
(EHO) as measured by magnetics and phase contrast imaging densities ∼0.85nG (or nped ∼ 0.65nG ) [473, 653, 655]. In the
diagnostics, in a QH discharge on DIII-D. Reprinted with case of medium β plasmas (βp ∼ 1.1–1.3 and βN ∼ 1.9–
permission from [466].
2.2), access to pure Type II ELMy H-modes requires high
triangularity, as well as q95  4.2–4.5. If the plasma δ is
be higher in QH than in Type I ELMy plasmas, typically in reduced, mixed Type I–II ELMs are also observed if the QDN
the range ∼2–4 in both the DIII-D and ASDEX Upgrade QH configuration is approached, indicating that both the shaping
discharges. Average Zeff values do decrease as expected at and the proximity to double null play a role in the onset of
higher operating density. Type II ELMs. Ideal MHD analysis of these discharges [656]
indeed shows that the high δand q95 tend to make the peeling
4.7.4. Type II ELMs. The first observation of a transition modes more localized and stable, and the QDN configuration
from Type I ELMs to small ELMs at high plasma confinement further improves the peeling mode stability (see section 4.8).
was reported by DIII-D in 1990 [652]. This transition was H-modes with Type II ELMy edge have been obtained
observed in a single-null discharge by increasing the plasma in ASDEX Upgrade also at very high β (βN ∼ 3–3.5 and
elongation at very high δ (κ  1.8 and δ ∼ 0.5) and βp ∼ 1.7–2), in the so-called high βN or hybrid scenario (these
q95 ∼ 7, and was attributed to a change of MHD stability high values of β are obtained because q0 > 1 and fishbones
in the plasma edge. The main characteristic of this regime replace sawteeth, reducing the trigger for NTM destabilization,
was the co-existence of high pedestal pressure and high see section 2 of chapter 3 of this issue [657]) [473]. The onset
frequency/irregular, very small ELMs. of pure Type II ELM edge in this regime occurs for pedestal
More recently [653], Type II ELMy H-modes have conditions and plasma geometry very similar to that of the
been identified in the ASDEX Upgrade tokamak and access mediumβ, standard H-modes (high density, high shaping and
conditions and plasma properties thoroughly characterized. proximity to double null).
Type II access has also been investigated in JET [497], and The comparison of access conditions to Type II ELMs for
mixed Type I–II ELMs have been reproducibly obtained, high and medium β plasmas in ASDEX Upgrade suggests a
although so far H-modes with a pure Type II ELMs have not possible trade-off between β and q95 . In particular, for the
yet been achieved [600]. Periods of small, apparently Type II highest β obtained (βN ∼ 3.5, βp ∼ 2, with Pin /PLH ∼ 5), the

S94
 Chapter 2: Plasma confinement and transport

(a) Radial magnetic field pickup coil (f > 100 kHz)

2
ASDEX Upgrade #16112
1

[a.u.]
0
-1
-2

(b) Radial magnetic field at EHO frequency and harmonics

40
20
[a.u.]

0
-20
-40

(c) Soft X-ray intensity (peripheral chord)

150
[W / m ]
2

140
130
120

(d) D intensity (outer divertor)

0.16
[a.u.]

0.14
0.12
0.10
2.0810 2.0812 2.0814 2.0816 2.0818 2.0820 2.0822
time [s]

Figure 78. Measurements of (a) the high frequency HFO and the EHO in measurements of (b) Br , (c) soft x-rays and (d) Dα , in a quiescent
discharge on ASDEX Upgrade. Reprinted with permission from [646].

Figure 79. Heat load on the inner and outer lower divertor plates in ASDEX Upgrade for three plasma discharges in advanced scenarios,
with increasing shaping, β and density ((c) to (a) in the figure). (a) #14521, δ = 0.43, n/nGr ∼ 0.88, q95 = 3.6, βN = 3.5 (b) #15486,
δ = 0.43, n/nGr ∼ 0.83, q95 = 3.7, βN = 3.2 and (c), #15524 δ = 0.33, n/nGr ∼ 0.50, q95 = 4.4, βN = 2.3. The short, high power loads
measured by IR thermography are caused by Type I ELM crashes. For the highest density and β (case (a)), no such events are detected,
while at intermediate β and densities (case (b)), mixed Type I–II Elms are observed. At low β/shape and density (case (c)) the plasma
exhibits Type I ELMs, although q95 > 4. These plasmas have DX ∼1 cm. Reprinted with permission from [468].

minimum q95 at which pure Type II are obtained is reduced II) ∼ Tcrit (Type I→III transition)), with collisionality ν∗ ∼ 1
to ∼3.6 from q95 ∼ 4.3–4.5 required at βN ∼ 2. The role of at the top of the pedestal at low β, reduced to ν∗ ∼ 0.5 at
plasma shape, density andβ in the onset of Type II ELMs is the highest β. Investigation of Type II ELM access in high-
illustrated by the example in figure 79 (from [468]). density/low collisionality conditions is a high priority for the
The extrapolation of Type II ELMy H-modes to burning validation of this regime for ITER.
plasma conditions and to ITER in particular is not yet clear. A confident extrapolation of Type II ELMs to ITER
Although this ELM regime occurs in the right range of plasma conditions (HH 98(y,2) ∼ 1, n/nG  0.85 at βN  1.8–2),
densities, the typical pedestal temperatures are low (Tped (Type as well as the full understanding of the underlying physics,

S95
E.J. Doyle et al

would strongly benefit by the extension of the Type II ELM

2 Grassy

D α div
regime to other existing tokamaks. In JET, access to Type

(a.u.)
II ELM edge in high performance ELMy H-modes has been
0 3
extensively investigated [497, 600]. Mixed Type I–II ELMy

(10 19 m -3 )
3 ne pellet

H-modes with the required global confinement and plasma

βp
ne
βp
E39459, 1MA, 2.05T, q 95 =3.8, δ=0.58, κ=1.35
β have been achieved routinely, in single-null plasmas with 0 0
0.35
δ
0.5, κ > 1.75 and 3
q95
4, with 2 Mixture
0.7
nped /nG
1 (ν∗ ∼ 0.5). In contrast to ASDEX

D α div
(a.u.)
Upgrade, QDN plasma configurations and increasing q95 have 0 pellet
pellet 3
not produced, so far, suppression of Type I ELMs but rather

(10 19 m -3 )
3 ne

βp
ne
an early transition (in terms of density) to Type III ELMs βp
E39505, 1MA, 1.95T, q 95 =3.7, δ=0.56, κ=1.37
and reduced plasma confinement. Predictive MHD stability 0 0
calculations indicate that the appearance of Type II ELMs in 2 Giant

D α div
(a.u.)
JET could be caused by a transition of a narrow region of
the pedestal next to the separatrix from the second to first 0 ne
pellet pellet
3

(10 19 m -3 )
3
stability [658].

ne

βp
A small ELM regime has recently been observed in βp E39511, 1MA, 1.97T, q 95 =3.6, δ=0.45, κ=1.41
the National Spherical Torus Experiment (NSTX), which 0 0
a) 4.5 time (s) 5.0
is characterized by a low Bt (<0.6 T) and aspect ratio
(R/a  1.26). This regime is also characterized by small, 0.6
Grassy
rapid ELMs, occurring at higher pedestal pressures than Type triangularity δ 0.5

III ELMs. The ELMs, observed on several edge diagnostics, 0.4

mixture
have been designated as ‘Type V’ [595, 659]. They are 0.3 Giant
characterized by an intermittent n = 1 mode rotating counter 0.2
to the plasma current, which slightly precedes each ELM 0.1 1MA (2T, 3T, 3.6T)
58m3
crash and vanishes between ELMs. In contrast to the EDA 0
3 4 5 6 7
or QH regimes, there does not appear to be a persistent edge b) q95 c)
fluctuation controlling the pressure. Densities tend to rise
slowly during Type V H-mode regimes, as is typical of other Figure 80. (a) The change of ELM behaviour from giant
(#E39511), to mixed giant-grassy ELMS (# E39505) to pure grassy
NSTX H-modes, and energy confinement is modestly reduced edge (# E39459) is obtained by gradual increase of plasma shaping,
compared with ELM-free H-modes. NSTX H-modes to date q95 (at constant Ip ) and βp . (b) q95 –δ existence diagram for the
have a pedestal collisionality ν∗ > 1. Further experiments grassy ELM regime: all plasmas with pure grassy ELM edge have
will be required to understand the physical mechanism of these βp > 1.6. (c): JT-60U equilibrium with δ = 0.6. Reprinted with
ELMs and how they may be related to Type II or other small permission from [501].
ELMs. These should help to understand the extrapolability
of the regime to other conditions and to a higher aspect illustrated in figure 80(a). Figure 80(b) shows the ‘trade-off’
ratio. between triangularity and q95 to obtain grassy ELMs edge; in
particular, increasing δ to ∼0.6 allows the reduction of the
4.7.5. ‘Grassy’ ELMs. Small ELMs at high plasma confine- minimum q95 to less than 4 [501]. Even with this very high
ment are also achieved in JT-60U [660, 661]. These ELMs triangularity, the second X–point remains outside the vacuum
are obtained in high triangularity, high βp plasmas with a vessel (figure 80(c)).
weak reversed shear (i.e. in plasmas that have both an ITB MHD stability analysis of the high βp JT-60U plasmas
and an ETB). The pedestal pressure (Tped ) with ‘grassy’ ELMs (at high q95 ) [662] indicates that the transition from giant
is higher than in comparable standard Type I ELMy H-modes. to grassy ELMs is due to improved pedestal stability, with
As in the case of Type II ELMs, grassy ELMs do not perturb the pedestal accessing the second region of stability for
in a measurable way the pedestal profiles and cause very low ideal ballooning modes. A stability change in the edge
power/continuous losses from the plasma. could also be promoted by the strong Shafranov shift at
The transition from a Type I ELMy edge to grassy ELMs high βp .
H-modes is gradual, and similarly to the case of Type II Very small ELMs, reminiscent of the grassy ELMs of
ELMs, grassy and giant ELMs can co-exist, with transitions JT-60U, have been obtained in high βp H-modes in JET [663].
from one ELM type to another at (apparently) constant pedestal The transition from Type I to grassy ELMs occurs at βp ∼ 1.6–
parameters. The disappearance of giant (Type I) ELMs and the 1.7, very similar to the critical threshold reported by JT-60U.
onset of grassy ELMs occur when the plasma triangularity, q95 So far, grassy ELMs in JET have been obtained only in H-
and βp are high enough. Detailed experimental investigation modes with conventional q profile, very high edge q (q95 ∼ 7)
of access conditions in JT-60U has found that βp is the critical and in quasi-double-null configuration (δ ∼ 0.5).
parameter to obtain grassy ELMs, and a threshold value of Some features of the JT-60U high βp regime with grassy
βp > 1.6–1.7 has been established, although high δ and q are ELMs are of clear relevance for ITER and burning plasmas,
also required. such as the demonstrated compatibility with ITBs, full non-
The transition from giant to grassy ELMs in a series inductive operation, low edge collisionality (0.1 < ν∗ < 0.3),
of JT-60U discharges with increasing βp , q95 and δ is good impurity transport and, as shown in figure 80(a), possible

S96
 Chapter 2: Plasma confinement and transport

compatibility with pellet injection. On the other hand, linked to plasma confinement and fusion performance. To
the extreme shape required for obtaining grassy ELMs at be attractive as an operating scenario, a regime should have
reasonable values of q95 , as well as the very high threshold in βp energy confinement at least as great as that of the Type I
(this issue is common to the hybrid/high βN plasmas in ASDEX ELMy H-mode. Lower confinement is the primary drawback
Upgrade and to JET) make a simple extrapolation to ITER of the Type III ELM regime. For the other small ELM
conditions impossible. As with other small ELM regimes, the regimes, high H-factors appear to be achievable in the limited
high βp grassy ELMs regime needs to be investigated further, parameter range of single devices. However, because each of
to gain more understanding of the basic physics mechanisms the regimes described has only recently been reproduced on
causing the transition in the pedestal behaviour, so as to learn more than one or two experiments, there are not yet robust
how to obtain this very promising plasma regime in other multi-machine scalings of pedestal widths and heights, or
devices and in ITER. energy confinement, in H-mode regimes which do not have
Type I ELMs. Other important parameters such as Zeff , which
4.7.6. Summary and issues for extrapolation to burning tend to be higher in the low density, low ν∗ regimes, also need to
plasmas. It is evident from the large number of recent be documented. This remains an important task in the coming
experiments and publications that significant progress has years and will require coordinated experiments among devices
been made since the ITER Physics Basis [1] in obtaining with a range of sizes and plasma parameters. The issues of
and documenting H-mode regimes which offer an alternative accessibility and performance are closely linked; for example
to Type I ELMs. As examples, the EDA regime has been if it is necessary to raise q, and lower Ip , to achieve Type II or
extended from Alcator C-Mod to several other tokamaks, and other ELMs, confinement would need to be correspondingly
the responsible quasi-coherent mode has been measured in higher to compensate. Lower pedestal pressure with small
detail and reproduced in simulations. The quiescent H-mode ELMs could in some operation scenarios be offset by improved
(QH) regime has been discovered on DIII-D and reproduced on core confinement. Given the promise of alternative regimes
other experiments. The operational regime of Type II ELMs and the present uncertainties, they need to be actively pursued
has been expanded to n/nG ∼ 0.85 on ASDEX Upgrade and in parallel with the further studies of Type I ELMs discussed
that of grassy ELMs to lower q95 and high βp on JT-60U, in sections 4.2 and 4.6 of this chapter and in chapter 4 of this
as well as being reproduced in JET at high q95 . Each of issue [364]. Such experiments, together with an increased and
the regimes has been successfully sustained for many τE coordinated effort to model possible instabilities responsible
and combined with internal transport barriers, making them for the edge transport in each regime, offer the prospect of good
attractive for advanced scenarios. It is also clear from inter- physical understanding required for confident extrapolation.
machine experiments that there is not yet the completeness of
understanding which is required to confidently extrapolate any 4.8. Pedestal stability
regime to a burning plasma experiment, though several routes
look promising. Simply transferring an operational ‘recipe’ to It is now generally accepted that MHD stability has a large
a machine of different size often fails to reproduce the same influence on the pedestal characteristics. Ideal MHD modes
behaviour (e.g. EDA and Type II ELM experiments on JET). provide a limit to the maximum achievable pressure gradient
On the other hand, if shape and edge dimensionless parameters and so, for a given pedestal width, determine the maximum
are matched, similar fluctuations are generally seen. Each height of the pressure pedestal. For stiff transport models,
regime has accessibility issues which need to be further the pressure pedestal is predicted to have a significant impact
explored. The EDA H-mode tends to evolve to a Type II ELMy on the confinement, and therefore fusion performance, of any
regime at higher edge temperatures and pressures, though burning plasma tokamak designed to operate in H-mode. Thus,
there are indications of a quasi-coherent mode present between understanding the pedestal stability constraints is crucial for
ELMs. There are limits to pedestal density in the quiescent performance predictions of future tokamaks.
H-mode which apparently depend on shaping. All present QH The ELMs associated with the pedestal region are now also
experiments use counter-injected NBI, and the role of fast ions widely believed to be a consequence of MHD instabilities. To
versus rotation is not fully understood. Any burning plasma develop a predictive capability for their resulting heat loads,
regime needs to be compatible with predominant heating, we therefore need to understand not only the onset criteria for
by alpha particles and RF, of electrons. Most regimes are these instabilities but also their non-linear evolution. This is
presently found at somewhat higher q95 , and with stronger a challenge, but significant progress has been made in recent
shaping and/or nearness to double null, than is envisioned for years, as described in this section.
example in ITER reference scenarios. An exception is the To understand fully the ELM dynamics and pedestal
Type III ELMy regime, which has been observed robustly height constraints, it is necessary to integrate the stability
on all divertor tokamaks and whose access conditions are calculations with transport codes. Such integrated modelling
now well documented and at least partially explained by is covered in section 5.5 and will not be repeated here.
theory. Continued study and extrapolation to burning plasmas
is important since Type III ELMs tend to occur at power close to 4.8.1. MHD instabilities in the pedestal. There are two
the L–H threshold and at low pedestal temperature or pressure; ideal MHD instabilities associated with the pedestal region.
with limited power this regime may occur whether or not it is High toroidal mode number, n, ballooning modes can be
desired. driven unstable if the pressure gradient exceeds a certain
Just as important as the access requirements for alternative critical value. However, with sufficient shaping, and in the
H-mode regimes is that of pedestal scaling, which is closely presence of sufficient current density, these ballooning modes

S97
E.J. Doyle et al

Current density
III

I II
Stable

Pressure gradient

Figure 81. (a) Stability calculation for a JET discharge using the MISHKA code [670]. Shaded areas are unstable, numbers indicate the
most unstable toroidal mode number calculated and the two curves indicate the n = ∞ ballooning stability boundary at two flux surfaces. α
is the normalized pressure gradient and jped , j0 are the current density in the pedestal region and centre, respectively. The grey area indicates
the peeling or kink unstable region. The unshaded region is stable. (b) A sketch of the marginal stability curve (full curve), together with
possible interpretations of large (I) and small (II and III) ELM cycles [666, 669].

can be stabilized, providing access to higher pressure gradients: to n ∼ 20–30 but becomes increasingly computationally
the so-called second stability. The destabilizing effect of the demanding at the higher n; it also has the capability to
pressure gradient is a consequence of the fact that the mode explore diamagnetic effects [668]. KINX can cope with a
localizes on the outboard side, where the curvature is bad. similar range of toroidal mode numbers but has the unique
The second ideal MHD instability is the peeling mode, capability to treat the separatrix geometry [671, 672]. ELITE
which can be thought of as an edge-localized kink mode has been developed specifically to treat intermediate to high n
[664]. The instability is driven by the current density (or its modes (typically accurate for n > 5) efficiently at the plasma
gradient) in the pedestal region. However, because it is edge [673]; it has been used for extensive parameter scans.
approximately poloidally symmetric, it experiences the good As well as these ‘purpose-developed’ codes, the advances in
average curvature of the tokamak so that pressure gradient is computing capability have meant that the low n codes, such as
stabilizing for this mode. The stabilizing effect of the pressure GATO/ERATO [674], can now access toroidal mode numbers
gradient is less effective when the shaping is weak, and then as high as n ∼ 8. Extracting the fast variation through a phase
one finds that the current density required to stabilize the factor permits even higher mode numbers to be explored [675].
ballooning modes destabilizes the peeling mode. The result As an example of the use of these codes, we show a stability
is that there is no access to the second stability regime diagram for a JET discharge, obtained using MISHKA in
referred to above. Increasing the shaping both reduces the figure 81. Results from KINX show that when a separatrix
current density required to stabilize the ballooning mode and is included, the results are broadly similar [672], but a higher
increases the current density required to destabilize the peeling current density is typically required to trigger a peeling mode.
mode, so access to the second stability can then be achieved.
Nevertheless, the pressure gradient is ultimately limited by
ideal MHD modes with an intermediate n, typically n ∼ 6–12, 4.8.2. Links of MHD stability calculations to ELM types.
where both the pressure gradient and current density play a The proximity of the pedestal parameters to the ideal stability
role in destabilizing the modes. In these situations, where boundaries in the cases where Type I and Type II ELMs are
both drives are operative, the modes have become known as observed has led to the suggestion that these ELMs are a
the (coupled) peeling–ballooning modes [665–669]. consequence of the ideal MHD instabilities (figure 81(b)).
In summary, the three factors that have most influence on However, Type III ELMs typically occur at a pressure gradient
the ideal MHD stability of the pedestal are current density, significantly below the critical value for ballooning instability.
pressure gradient and shaping. These can therefore all While it has been suggested that peeling modes could be
influence the pedestal characteristics. responsible in some situations, it is unlikely that all Type III
The importance of the current density is particularly ELMs can be interpreted in this way. One possibility is that
interesting. For modern tokamaks, where the collisionality resistive ballooning modes play a role here [676]. A challenge
is relatively low, the bootstrap current is the dominant to this idea is to explain why such instabilities would lead to
contribution in the pedestal, and this depends sensitively a large transient heat flux, rather than simply contribute to the
on the individual density and temperature profiles, not just steady turbulent transport.
the pressure profile. Thus, the ideal MHD stability can be It is now generally accepted that large Type I ELMs
significantly influenced by varying the plasma density. are triggered by intermediate n peeling–ballooning modes.
Because of the importance of their role in determining the Figure 82 shows an example of a DIII-D discharge which is
pedestal characteristics, significant effort has been invested found to be stable to ideal MHD modes throughout the ELM-
in developing computer codes for quantitative ideal MHD free period, but just prior to the ELM the plasma edge becomes
stability analyses. MISHKA [670] can in principle deal unstable to an n = 10 mode [669]. Allowing for the effects of
with a range of toroidal mode numbers from n = 1 up diamagnetism, the coincidence between the time of instability

S98
 Chapter 2: Plasma confinement and transport

a.u.

(a)

–0.05

Figure 82. The upper trace shows the onset of the first ELM in a
DIII-D discharge as a spike in the Dα emission. The lower trace
shows the results of a sequence of stability analyses. The pedestal
pressure gradient and current density rise continuously up to
2180 ms, when an n = 10 ideal MHD mode is predicted to go
unstable just prior to the first ELM. Reprinted with permission
from [669].

(when the growth rate γ > ω∗ /2, where ω∗ is the diamagnetic

frequency) and the ELM onset is remarkable. (b)
It is natural to ask why we see such a range of different
ELM types. One explanation that has been proposed is that Figure 83. (a) Temperature perturbation built up from a sequence of
the ELM size is related to the width of the linear eigenmode. ELMs in a DIII-D discharge. (b) The structure of the most unstable
eigenmode, n = 10, calculated using ELITE. Reprinted with
Thus, a more extended eigenmode might be expected to affect permission from [669].
more of the plasma, and hence result in a larger ELM. While
such a model has little theoretical basis, and there is clearly
more to the story [580], there is some experimental evidence Dα (EDA) mode observed in Alcator C-Mod [624]. Stability
that the ELM-affected area is indeed related to the linear analyses (see figure 86) suggest that the EDA mode plasmas
eigenmode width. Figure 83 shows the radial distribution of the are stable to ideal MHD modes, and this could explain the
temperature perturbation due to ELMs, built up from a series of absence of ELMs [634]. Nevertheless, a coherent mode is
ELMs in a particular DIII-D discharge [669]. Also shown in observed during the EDA, and it is believed that this provides
that figure is the eigenmode structure for the most unstable the density control. Calculations with the BOUT code [19]
toroidal mode number, from which it can be seen that the (a non-linear code, based on a Braginskii fluid description of
radial structure is similar to that observed for the temperature the plasma, treating the full divertor geometry) have led to the
perturbation. Such a model would suggest that Type II ELM speculation that this coherent mode may be associated with a
regimes originate from situations when the linear eigenmode resistive ballooning mode [176].
is restricted to a narrow radial region. A study of this was The QH-mode [646, 647], which was first observed on
performed using GATO for a range of discharges characteristic DIII-D, is a second regime which provides density control,
of ASDEX Upgrade [656]. It was found that a combination but without ELMs. The mode requires counter-neutral beam
of high edge safety factor, q95 , and high triangularity leads to injection and low plasma density. ELMs disappear and
a reduction in the radial mode width (figure 84). This could an edge-harmonic oscillation appears to control the density
provide a possible explanation for the observed transition to instead. Precisely what the edge-harmonic oscillation is and
Type II ELMs in this regime. A similar result was obtained why the ELMs disappear are still uncertain, but the latter may
from JT-60U in their ‘grassy’ ELM regime (figure 85) [662] be related to the toroidal momentum input [648].
with q95 = 3.4 for large ELMs and q95 = 6 for small ELMs.
The poloidal extent of the mode could also influence the 4.8.3. Non-linear MHD analyses. In order to model the
resulting ELM size, and both JT-60U [584] and MAST [594] crash phase of ELMs, it is clearly important to understand the
do indicate that ELMs predominantly affect the pressure on the non-linear evolution of the MHD instabilities. A non-linear
outboard side. theory for the ballooning mode predicts that the mode evolves
As well as small ELM regimes, there are examples of into a flux tube, which narrows and twists to allow it to erupt
regimes with no ELMs, in which density control is provided by explosively out into the scrape-off layer without reconnection
a more benign feature observed in the pedestal region. There [677, 678]. This helps to explain the rapid growth time of the
are two such modes of operation. The first is the enhanced ELM. It is then envisaged that the flux tube provides a direct

S99
E.J. Doyle et al

Figure 84. Eigenmode structures for a sequence of discharges with ASDEX Upgrade parameters, calculated using the GATO code [674].
High δ = 0.45 and q95 = 5 lead to narrower radial mode widths than low δ = 0.15 and q95 = 4.3, providing a possible interpretation of
Type II ELMs. Reprinted with permission from [656].

Figure 85. Eigenmode structures for the most unstable modes in

two JT-60U discharges, showing a correlation between ELM size
and radial eigenmode width [662].
Figure 87. Filamentary structures observed in MAST during an
700 ELM, consistent with the predictions of flux-tube eruptions from
EDA stable (ELITE) non-linear ballooning mode theory. Reprinted with permission
ELMs unstable (ELITE) from [555].
600
4.8.4. Implications for ITER. Clearly the understanding of
500 ELM dynamics and associated phenomena have progressed
Teped (eV)

significantly since the ITER Physics Basis. Nevertheless, a

400 fully quantitative model for ELM size remains elusive. There
are clear trends which are observed experimentally and which
300 have an interpretation in terms of the linear mode width. Thus
we would expect smaller ELM sizes in discharges with higher
200 triangularity and higher edge q95 . Flexibility to access these
regimes on ITER is clearly of importance. The role of the
100 flux-tube eruptions predicted by theory and observed in MAST
1.0 1.5 2.0 2.5 3.0 3.5 4.0 needs to be understood for tokamaks with close-fitting walls
αMHD to quantify the implications for ITER. These flux tubes would
carry a high power density if they connect directly to the core
Figure 86. Observations of EDA and small ‘grassy’ ELMs in plasma, so they could have significant implications if they
Alcator C-Mod, showing EDA occurs at low pedestal temperature were to strike the vessel wall. On the other hand, the presence
(open circles). Stability analyses (squares and diamonds) are
broadly consistent with the EDA mode plasmas being stable to ideal
of the wall itself may help to limit the radial extent of the
MHD modes. Reprinted with permission from [634]. flux-tube eruption.
Of the regimes where density control is possible without
ELMs, it seems unlikely that the EDA will be a viable operating
route for the hot core plasma to escape into the scrape-off layer, regime for ITER because of the low temperature that is
though the precise mechanism for this remains unclear. Direct apparently required. Nevertheless, the quasi-coherent mode
experimental evidence for such structures has been obtained should continue to be studied to understand whether it has a
from MAST (figure 87) [555, 679], and there are also signs of role for density control in ITER-like conditions. The QH-mode
such structures in BOUT simulations [680]. offers more promise, having now been observed on a number of

S100
 Chapter 2: Plasma confinement and transport

Tped limits for ITER, nped=7.1×10 13 cm-3

12.00
2nd stable
DIII-D data
Pedestal Electron Temperature (keV)

Stability Calculation

Maximum Stable Tped (keV)

8.00

4.00
n=30
n=20
n=15
n=10
0.00
0.00 0.02 0.04 0.06 0.08
19
Pedestal Density (10 m-3 ) pedestal width/minor radius (∆/a)

Figure 88. Comparison of the predicted temperature pedestal height as a function of density (left) from stability calculations (curve) and
from a Type I ELMy H-mode DIII-D data set (with other parameters constrained to a specified narrow range). The figure on the right shows
predictions of the maximum stable temperature pedestal for ITER as a function of pedestal width from the ideal MHD stability calculations
for a range of n. The relevant n at any particular pedestal width is that for which the maximum stable Tped is lowest. Reprinted with
permission from [681].

tokamaks, but extrapolation to ITER is very uncertain without We close with a few comments regarding the remaining
a reliable physics-based model for the phenomenon. The high areas of uncertainty in the theory of MHD in the pedestal and
impurity content of the QH-mode plasmas is also a concern. predictions for ITER. The importance of diamagnetic effects
Based on the qualitative understanding of ELM dynamics (including shear in the diamagnetic frequency) and also fast
described here, it is also possible to suggest means for ELM particle effects on the linear stability theory should be assessed,
control. Ramping the current up or down can trigger ELMs to complete the understanding of the trigger mechanism. The
or suppress them as found experimentally and as expected non-linear evolution should address how the pressure and
theoretically from our picture of the role of peeling modes. current density evolve through the crash phase, which will
Thus, some form of current profile control in the pedestal require a model beyond ideal MHD, including transport effects.
region would provide a control parameter for ELMs, but it Without such a model it is difficult to predict the size of ELMs
remains to be seen how effective such a technique might be with accuracy. The role of the plasma filaments needs to be
(if feasible). A means to degrade the pedestal confinement studied further, in particular whether or not they have any
and so limit the pressure gradient a little below the ballooning implications for the vessel wall.
stability boundary may also help to reduce the ELM size (the In summary, there has been much progress in the area of
radial mode width would be more narrow). pedestal MHD, but there is more work to be done to have a
A more quantitative use of the MHD stability results completely predictive model for ITER.
is to calculate the maximum achievable pressure pedestal in
ITER. These calculations assume that the edge current density 4.9. Possible pedestal control scenarios
profile has reached steady state and is given by the bootstrap
current. With this assumption, the pressure pedestal height A key feature of the control of burning and steady-state
can be calculated. This has been done for a range of DIII-D high beta plasmas is that the object plasma is a highly self-
discharges and, provided only those discharges in a specified, sustaining system. In this system, our goal is to maintain
narrow range of parameter space are chosen (including the the integration of multiple performance criteria required for
pedestal width), a good agreement can be obtained, as shown the reactor core. Since the pedestal area stands at the pivot
in figure 88 [681]. Note that this comparison has no free of the multiple criteria discussed in the introduction to this
fitting parameters. The predicted pedestal height for ITER section, we need to develop and evaluate pedestal control
as a function of pedestal width can be determined [498]. Note scenarios from the viewpoint of optimizing the whole plasma
that a significant pedestal height can be achieved even down at system. For ‘pedestal control’, which includes control of the
quite low values of the pedestal width. This is presumably structure/profile of the pedestal parameters and of the ELM
associated with the finite n stabilizing corrections for the activity, the possible approaches are based on modifying the
ballooning modes, which are stronger when the pedestal is plasma transport, the neutral distribution and MHD stability.
narrow [682]. An important conclusion is that estimates of the
pedestal height for ITER based on a constant critical pressure 4.9.1. Control of pedestal height and structure. In burning
gradient are inaccurate at a low pedestal width, predicting a plasma experiments, the basic requirement is to establish
lower height than we derive here. a favourable steady-state solution of the pedestal structure

S101
E.J. Doyle et al

without additional power to control the pedestal in order

to maximize QDT . Therefore, the optimum choice of
equilibrium shape, in combination with heating power and
fuelling/pumping, is essentially important. This optimization
should be conducted in a burning plasma experiment as the
first step. According to present knowledge, high elongation
with high triangularity seems to be a favourable equilibrium
shape. The heating and current drive power should be applied
to the core region (not to the pedestal directly) in order to
maximize QDT . As for fuelling, the pellet injection seems to
be favourable for high confinement at high density by reducing
the neutrals surrounding the plasma and for minimizing the
fuel circulation. In addition, some amount of gas puffing to
the divertor area may be required to maintain the detached
condition. Pumping from the inner strike point will be
optimized under a trade-off with the required triangularity.
Once a steady-state operational point is achieved, real-
time feedback control of the pedestal structure can be applied
to a burning plasma. However, there are practical limitations
to be considered. As for the shape control, the ac loss in the
superconducting poloidal field coils has to be kept within a
allowable level. Edge plasma current control is theoretically
effective, but inductive drive is limited by the allowable ac loss.
For non-inductive edge current drive, the problem is that the
pedestal temperature is low. ECCD has a favourably localized
driven current profile, however it requires a much better current
drive efficiency, ηCD . LHCD has the highest ηCD , however the
launcher–plasma interaction issue has to be solved. Another
candidate is edge rotation/radial electric field drive. The edge
transport barrier ELMs, the internal transport barrier and high
beta MHD stability (such as due to resistive wall modes) can
be controlled by rotation drive. The quantitative evaluation of
the required rotation drive power using NBI and RF is quite an
important issue to decide in order to determine whether these
are both effective and practical in burning plasma experiments
and reactors.
Figure 89. ELM control by pellet injection demonstrated in
4.9.2. ELM control. In ITER, real-time ELM control ASDEX Upgrade: (a) density, diamagnetic energy, pellet monitor
scenarios are needed for ELM mitigation (for both reduction of and Dα signals during external gas puff without pellets
(fELM = 29 Hz), small pellets and no puff (fELM = 68 Hz) and with
the divertor heat load and compatibility with the ITB), impurity external puff only (fELM = 68 Hz). (b): diamagnetic energy and
control and fuel control while keeping Wped high enough to line-averaged electron density dependence on ELM frequency
maintain burning plasma performance. See section 2.2 of without gas (squares), gas puffed (triangles) and pellet phases
chapter 4 of this issue [364] for discussion of ELM transport (circles). Increasing fpel imposes slight refuelling and confinement
−0.16
in the SOL. In addition, any negative impact on other key degradation. A fit to data with pellets: WMHD ∼ fELM .
Experimental scaling with gas puff in ASDEX Upgrade:
performance characteristics has to be minimized. At present, −0.6
WMHD ∼ fELM . Reprinted with permission from [684].
as discussed in the previous sections (e.g. see section 4.6.3),
two means of ELM control are apparent. The first is to
change the mode number or eigenfunction distribution of the section 2.7.3 in chapter 4 of this issue [364] for more discussion
instability. The second is to enhance other energy losses. And of the effects of pellets on ELMs. Although the ELM
in some cases, such as in the small ELM regimes, these two trigger mechanism by pellets is a remaining issue, a study of
seem to be mixed. In this context, this section summarizes the projection to ITER [609] determined that the minimum
active control schemes demonstrated in recent experiments required pellet frequency, fpel , is 3 Hz to restrict the allowable
[683]. ELM energy loss to WELM < 7 MJ with Ploss ≈ 100 MW.
And high field side injection of moderate sized (dpel < 7 mm)
Pellet injection. In ASDEX Upgrade, the Type I ELM pellets at 4 Hz would reduce WELM to the acceptable level
frequency was controlled (fELM /fpel = 1) by injecting small of ∼5 MJ.
pellets (cubic (1.4 mm)3 ∼ 1 × 1019 D-atoms, at 1 km s−1 )
as shown in figure 89 [684, 685]. The ELM characteristics Edge current control. Changing the edge current can modify
were similar to the intrinsic Type I ELMs with the same the stability boundary and ELM characteristics as predicted
frequency, and confinement degradation was very mild. See by the peeling–ballooning mode theory. In COMPASS-D

S102
 Chapter 2: Plasma confinement and transport

[602], Type I ELMs were triggered when the edge current

was increased and vice versa. In TCV [604], using edge
current drive with vertical plasma movement, ELMs were
triggered when a positive edge current was driven (upwards
for LSN). For application to ITER, however, we need to
evaluate the practical effects within the ac loss and power
supply limitations.

Application of external field. In JFT-2M, an ELM-free

H-mode was controlled to be a steady ELMy H-mode
by applying an ergodic magnetic field [686]. Testing
many configurations with three sets of ergodic coils, they
concluded that the helical field component of n = 4 is
effective in producing ELMs [687]. COMPASS-D [602] also
demonstrated the transition from ELM-free to ELMy H-mode
when a radial magnetic perturbation was applied. On the other
hand, in DIII-D [606], Type I ELMs were suppressed almost
completely with an external helical magnetic perturbation
(mainly n = 3 resonant at the plasma edge of q95 = 3–4) while
keeping the confinement performance constant. In order to
apply an edge ergodization scheme to ITER, however, further
evaluation is needed not only for the effects on the pedestal but
also for the stability of the low-n global mode at high beta.

Edge rotation control. As discussed in section 4.7, counter-

rotation produces the QH-mode. In addition, by utilizing
co-, counter- and perpendicular NBI, JT-60U demonstrated
that the Type I ELM frequency increases with increasing
counter toroidal rotation velocity, and the ELMs change from
Type I to grassy while keeping almost constant pedestal
pressure (figure 90) [661]. In addition to this ELM control
scheme, rotation drive can expand the controllability of the
burning plasma in many other aspects. However, for burning
plasmas the required power has to be carefully evaluated
Figure 90. ELM control by rotation demonstrated in JT-60U:
quantitatively. (a) Toroidal rotation profiles mapped into outer midplane measured
with charge-exchange recombination spectroscopy. The shaded
4.10. Summary of pedestal structure and transport region shows the area at the top of the Ti pedestal. (b)–(e) Time
evolution of the Dα signal during a plasma toroidal rotation scan
The edge pedestal is a key area determining the integrated (q95 ∼ 4.9 and δ ∼ 0.59). Plasma rotation profiles were changed by
performance required for burning steady-state plasmas. In using different beam combinations: (b) 2CO+2perp+2N−NB,
the pedestal physics research field, the main issues identified (c) 2CO+3perp+1N−NB, (d) 2CO+5perp and (e)
1CO+1CTR+5perp. Reprinted with permission from [661].
in the ITER Physics Basis [1] were: (i) extension of the
high confinement regime to high plasma density (∼nG );
(ii) establishment of the predictive capability of the pedestal and no-ELM regimes have been developed and some of them
pressure and its radial structure for burning plasmas; have been demonstrated in the low collisionality regime. High
(iii) mitigation of the giant (Type I) ELM amplitude; and triangularity shape, safety factor, edge rotation, collisionality
(iv) development of control techniques for the pedestal and harmonic oscillations have been found to play key roles in
pressure and ELMs. It was also emphasized that identification these regimes. Theory-based predictive capability has shown
of the edge transport and stability processes and their interplay good progress by integrating plasma transport (in particular,
determining the pedestal structure and dynamics is essentially turbulence suppression models based on zonal flows), neutral
important. transport and MHD stability including the core, pedestal,
Towards this goal, remarkable progress has been obtained SOL and divertor regions. These approaches reproduce most
in the intervening years. H-mode confinement has been aspects of the L–H transition. Also, on the experimental
maintained up to the Greenwald density with high triangularity. front, high spatial and temporal resolution diagnostics have
The peeling–ballooning mode theory explains the pedestal revealed the rapid change in the radial electric field and
pressure gradient in the Type I ELM regime and its variation the threshold in temperature or its gradient for the L–H
with plasma shape. Experimentally, the detailed structure transition.
of the ELM crash (poloidal asymmetry, radial distribution, Finally, we list the remaining important issues. The
three-dimensional propagation in the SOL, etc) has been most important is the uncertainty of the edge transport
clarified. Type I ELM control techniques (e.g. pellet pace barrier width. Although establishing the inter-machine
making) have also been demonstrated. A variety of small pedestal database has improved our understanding of the

S103
E.J. Doyle et al

pedestal structure, the width scaling has not been established. of machines and physics as well as unification of structures
Recent efforts on inter-machine comparison experiments will and analysis tools. This is followed by a summary of
clarify this issue. For the density profile shape, the effects our expectations for the plasma edge and H-mode pedestal
of neutrals have been modelled but need to be verified conditions in ITER, then the performance projections from
systematically. As for the high-density operation near the each of the three approaches described above. The pedestal
Greenwald density, extension to the low collisionality regime conditions are used in global confinement projections as well
is needed. For understanding the ELM crash dynamics, non- as in profile modelling, which improves the confidence in our
linear evolution of the MHD instabilities has to be clarified. extrapolations to ITER over those used in the IPB report. But
Also, for understanding the ELM crash dynamics and the ELM limitations still exist and are described to the best of our present
period, time evolution of the edge current profile across an knowledge.
ELM crash and the inter-ELM transport has to be analysed
together with the edge oscillations. These are also important 5.1. Improved database resources for modelling and scaling
for extending the small and no-ELM regimes to reactor- studies
relevant conditions. In order to solve these issues, advanced
diagnostics for the edge current density profile measurement The results of the confinement modelling and scaling studies
with high spatial and temporal resolution is essentially presented in this paper are based upon six international multi-
important. machine databases that are currently maintained by ITPA
working groups. These are the following.
5. Predictive capability and projections for ITER 1. Global L-mode confinement database [688–694].
2. H-mode power threshold database [559, 560, 689,
Three primary approaches are used to project the performance 695–698].
capabilities of ITER. The first is scaling from present machines 3. Global H-mode confinement database [2, 689, 690,
using global parameters that encompass the collective effects 698–709].
of all the physics and machine conditions: the interactions 4. H-mode pedestal database [529, 534, 541, 702, 706, 707].
between plasma confinement, MHD equilibrium and stability, 5. ITB database [25, 202, 263, 486, 710–716].
particle, momentum, energy and current sources, as well as 6. Profile database [2, 234].
interactions between the plasma and the physical boundaries.
The energy confinement time given by the ITERH-98(y,2) The references included in this list refer to work related
scaling for an inductive scenario with a plasma current of to the databases published after (or not included in) the
15 MA and a plasma density 15% below the Greenwald density publication of the ITER Physics Basis report [2]. The L-mode,
is 3.7 s with one estimated technical standard deviation of 14%. threshold and H-mode databases are purely scalar whereas the
For levels of helium removal and impurity concentration that pedestal and ITB databases have also a profile part that are kept
are expected to be attainable, this translates to Q ∼ 6–15 with in the same format as the profile database.
an auxiliary heating power, Paux = 40 MW, and Q ∼ 6–30 Table 3 summarizes which devices have contributed data
at the minimum heating power satisfying a good confinement to the databases and table 4 lists where the databases can be
ELMy H-mode [279]. The scatter in such projections from found and who manages them. In the following six sections
present machines is in part due to a variety of conditions that more details will be given about each of the databases in turn.
are not all directly relevant to ITER, which is not captured
by the relatively few physics parameters in the fits. So 5.1.1. ITPA global L-mode confinement database. The
the global studies are augmented by a second approach, global L-mode confinement database contains 8749 entries
dimensionless scaling experiments, that more closely resemble from various tokamaks (see table 3), corresponding to Ohmic
the ITER configuration and operating conditions. The third (6067 entries) and L-mode discharges (i.e. with additional
approach is the simulation of time-dependent plasma profiles. heating and L-mode edge). Most of the discharges do not
This requires enough understanding of the underlying physics feature enhanced core confinement, except those specially
to enable the construction of models for validation against marked in the database as corresponding to improved phases
existing data and extrapolation to ITER conditions. Over the (189 entries). This database is however important for the
past thirty years, the time-dependent simulation of plasma operation of ITER, which will start with Ohmic and L-mode
profiles has progressed from being dominated by empirical regimes.
models to being dominated by models that are more closely The present version of the database is v2.9. Though a large
connected to fundamental theoretical descriptions. Each of number of new entries have been contributed since the previous
the three approaches has its own strengths and limitations. major publication of this database, the L-mode scaling law
Collectively, they are used to find the scope of the potential derived in [688] has not yet been revisited and is still used for
operating domain within which we can explore the new physics extrapolation to burning plasmas. However, it has been found
and dynamics of burning plasmas. that the same definition of elongation as in the IPB98(y,2)
We begin this section by reviewing the status of scaling expression [2] should be used and that this leads to a
the confinement-related databases that are developed and stronger inverse aspect ratio dependence [691]. But there is
maintained by the ITPA. These are absolutely necessary for still a significant difference in the aspect ratio dependence of
validating models over a wide range of plasma conditions. L-mode and H-mode scaling expressions that may explain why
They have undergone major changes since the publication of the difference in confinement is very small between the L-mode
the ITER Physics Basis (IPB), including expanded coverage and H-mode in high aspect ratio tokamaks such as T-10 [694].

S104
 Chapter 2: Plasma confinement and transport

Table 3. Summary of tokamaks that have contributed data to the different databases.
Database/Version
L-mode/ Threshold/ H-mode/ Pedestal/ ITB/ Profile
Tokamaks Ldb2v9 Thdb4v4 Hdb3v13 Pdb3v3 v1.8

Alcator C ×
Alcator C-mod × × × × ×
ASDEX × × ×
ASDEX Upgrade × × × × ×
COMPASS-D × × ×
DIII ×
DIII-D × × × × × ×
FT ×
FTU × × ×
HL-1M ×
JET × × × × × ×
JFT-2M × × × ×
JT-60 ×
JT-60U × × × × ×
MAST × × ×
NSTX × ×
PBX-M × × ×
PDX × ×
RTP × × ×
START × ×
TCV × × ×
TdeV × ×
TEXTOR × × ×
TFTR × × × ×
Tore Supra × × ×
TUMAN-3M × ×
T-10 × × × ×

No. of devices 20 13 19 8 10 11

Table 4. Database web access and managers.

ITPA DB Online at http:// Manager E-mail
L-mode www-itpa0d.cea.fr/ F. Imbeaux [email protected]
Threshold efdasql.ipp.mpg.de/threshold/ Y.R. Martin [email protected]
H-mode efdasql.ipp.mpg.de/igd/ K. Thomsen [email protected]
Pedestal efdasql.ipp.mpg.de/peddb/ L.D. Horton [email protected]
ITB Global www-itpa0d.cea.fr/ F. Imbeaux [email protected]
ITB Profile tokamak-profiledb.ukaea.org.uk/ F. Imbeaux [email protected]
Profile tokamak-profiledb.ukaea.org.uk/ C.M. Roach [email protected]

The L-mode and H-mode scaling expressions also differ in convenient for downloading the database. This website
their dependence upon normalized Larmor radius [692], with also features a simplified SQL request editor, which allows
the L-mode closer to a Bohm-like dependence whereas the data filtering and carrying out some preliminary analysis on
H-mode is more like a gyro-Bohm dependence [689]. Finally, the database. Further details about the ITPA 0D database
a two-term scaling expression has been established to predict servers and data submission can be obtained by e-mail to:
the confinement of high magnetic field devices [693]. The [email protected].
L-mode database also contains a significant amount of Ohmic
data [717, 718] that could be used to establish the size scaling 5.1.2. ITPA H-mode power threshold database. The H-mode
of an Ohmic confinement scaling expression [690]. power threshold database contains at present 7673 entries
The public versions of the L-mode database are accessible from 13 tokamaks: ASDEX (600), ASDEX Upgrade (636),
using a web browser at: http://www-itpa0d.cea.fr (common Alcator C-Mod (1227), COMPASS-D (46), DIII-D (752), JET
public website for ITPA 0D L-mode and ITB Global (3111), JFT-2M (1013), JT-60U (109), MAST (20), NSTX (8),
Databases). This site can be used to browse the database PBX-M (5), TCV (131) and TUMAN-3M (15). The current
and to download it in various formats: U-File, CSV and version of the database is labelled IGDBTH4v4.
SQL. Downloading the SQL dump file allows regenerating While H-mode threshold power scaling relations are based
the database on the destination computer in the SQL format. on data taken at the L–H transition only, it is worthwhile
The CSV format is recognized by most spreadsheet editors to have access to L-mode and H-mode data for comparison
(such as Excel, OpenOffice, etc) and for this reason is very and analysis with different statistical analyses such as the

S105
E.J. Doyle et al

discriminant analysis. Therefore the database contains 2660 by a significant error bias. New regression studies using regres-
time slices taken at the L–H transition, 2345 L-mode time sion techniques that include measurements errors have shown
slices, 2261 H-mode time slices and 404 time slices taken at the that these indeed could be responsible for the observed dis-
H–L transition. A series of selection criteria has been applied crepancies in the beta and ν∗ dependences [708, 709].
to this collection of data in order to make the comparison The various versions of the database are available on
between tokamaks reliable. For instance, only single-null the EFDA SQL server (efdasql.ipp.mpg.de) and can be
discharges with the ion ∇B drift directed towards the X-point downloaded via a web interface [705].
and operated in deuterium are kept. With these restrictions,
the number of L–H transition data used in the calculation of 5.1.4. ITPA H-mode pedestal scalar database. The ITPA
the scaling is 1302: ASDEX (43), ASDEX Upgrade (232), pedestal scalar database was established with the goal of
Alcator C-Mod (184), COMPASS-D (21), DIII-D (58), JET facilitating multi-machine comparisons of high resolution,
(585), JFT-2M (53), JT-60U (58), MAST (5), NSTX (6) and plasma edge measurements. The profile information in the
TCV (7). database is parametrized using either a hyperbolic tangent fit
The threshold power is still found to scale, predominantly, or a linear fit. Numerical tests have shown that the pedestal
with the plasma density, the magnetic field and the size of width deduced by the linear fit is approximately twice that
the device [559, 697, 698]. However, the potential effects of of a tanh fit to the same data. For use in confinement
other parameters are currently being investigated, aiming at studies, pedestal-top data are also stored for cases where profile
a reduction in the scatter between the actual threshold power
information is not available. Data are presently available from
and the estimation based on the scaling. For instance, the
the ASDEX Upgrade, Alcator C-Mod, COMPASS-D, DIII-D,
effect of the aspect ratio has now been addressed thanks to the
JET, JFT-2M, JT-60U and MAST tokamaks.
recent contributions of MAST and NSTX. The roles of the
With the increasing evidence that the pressure gradient in
effective Z (Zeff ), of the plasma shape and of low density
the pedestal region is limited by MHD stability, the focus of
have recently been studied as well [560]. Although these
multi-machine analysis has been on the physics determining
analyses resulted in a better estimation of the fit coefficients,
the pedestal width. Using the database, width scaling has been
the RMSE value of the fits remains larger than 20%. This
tested against models based on ion loss, neutral penetration and
implies a still larger uncertainty in the ITER H-mode threshold
velocity shear turbulence stabilization due to a combination
power.
of magnetic and E × B shear [529, 534, 541]. Although
A public version (IGDBTH3v2) of the database is
some correlation is found between the data and the models,
available on the EFDA SQL server (efdasql.ipp.mpg.de). The
significant scatter remains in each of these theory-based fits
database can be downloaded using any software equipped with
a decent SQL request interface. to the database. Some of this scatter may be due to the
simplified peeling–ballooning model, which is incorporated
into the global scalings. In addition, the difficulty of making
5.1.3. ITPA global H-mode confinement database. The
precise measurements leads to large scatter when a wide range
H-mode database contains at present 10382 entries from 19
of data is required. For these reasons, the Pedestal Group is
tokamaks (see table 3). The current version of the database
presently constructing a profile database which will contain
is labelled DB3v13. The IPB98(y,2) scaling expression
high quality data for a few discharges which can then be tested
[2] is still recommended for the prediction of the ELMy
in detail against the various width models in parallel with the
H-mode confinement. The confidence interval of the ITER
best available MHD stability calculations.
prediction using this scaling has been studied in detail
An additional goal of the pedestal scalar database is to
[279, 690, 701, 703, 704]. See section 5.3.4 for a summary
determine the scaling of the pedestal energy confinement and
of ITER projections using global data. Improvements to the
its contribution to the global confinement. For this reason,
log-linear scaling expression by adding interaction terms as
the pedestal database was developed in close cooperation with
well as other effects such as density peaking, high density and
the global confinement group and in such a way that the two
triangularity have been studied [697,704]. The effects of these
improvements are included in the above confidence interval. databases can be easily combined. By separately fitting the
ASDEX Upgrade, Alcator C-Mod, DIII-D, JET, JFT-2M, core and pedestal energy, it has been possible to develop a
JT-60U and MAST also provide data to the H-mode Pedestal two-term model for the scaling of energy confinement [707].
Database and a significant overlap between the two databases Two models for the pedestal confinement were tested: a thermal
has been achieved that has facilitated two-term scaling studies conduction model and an MHD stability limit model. When
to be performed [698, 704, 706, 707]. See also the pedestal combined with a standard regression to the remaining core
database section below. energy, the models result in similar predictions for the energy
There is now further experimental evidence confinement in ITER, both within the 95% confidence limit
[225, 228, 430, 432, 433] that the beta and ν∗ dependences of of the IPB98(y,2) scaling [2]. The precision of the fit is
the H-mode confinement scaling expressions are not correct limited by the conditioning of the database. More data from
whereas the ρ∗ dependence is confirmed. The strong beta smaller machines and from different shapes and aspect ratios
degradation of the global scaling expressions has been a con- are required to improve the quality of the extrapolation to
cern for a long time [700] and various causes have been sug- next generation machines and to differentiate between the
gested [699]. Recently it was recognized after measurement conduction and MHD models for the pedestal confinement.
errors had been collected from all data providers that the small-
est principal components have large measurement errors asso- 5.1.5. ITPA ITB database. The purpose of the ITB database
ciated with them so that the log-linear regressions are affected is to provide a comprehensive multi-machine dataset for

S106
 Chapter 2: Plasma confinement and transport

the analysis of enhanced core confinement discharges. The MDSplus system, which has been widely adopted by the fusion
database content corresponds mostly to discharges featuring community. MDSplus simplifies interactions with the data and
internal transport barriers, but has been recently extended is interfaced to many standard tools. The website hosts the data
to the ‘hybrid’ regime, which also features enhanced core itself, a range of analysis tools and documentation covering:
confinement. data variable definitions and formats and descriptions of ac-
The global (0D) database part contains 1777 entries from cess and submission procedures. The tools add considerable
various tokamaks: ASDEX Upgrade, DIII-D, FT-U, JET, value to the database and these include a powerful relational
JT-60U, RTP, T10, TCV, TFTR and Tore Supra (public release database search facility to help users find discharges of interest,
v1.8). This global database is used to investigate the origin of data display tools and codes to interface the profile database
the enhanced core confinement, by multi-machine analysis of to gyro-kinetic microstability codes (KINEZERO [719] and
local quantities such as the magnetic shear at the foot of the GS2 [147] see also http://gs2.sourceforge.net/ ).
barrier [263, 486, 710] and by deducing a possible threshold in The profile database has continued to grow under the ITPA
the heating power required to form an ITB [486]. Also, the framework, and more than 100 discharges have been submitted
0D database has been used for mapping the performances and since June 2001, when the database was moved to Culham. The
operational space of enhanced core confinement discharges in additional discharges have included electron heated plasmas,
terms of dimensionless parameters [202]. The ITB databases high performance ELMy H-modes with scans in density and
feature an extended set of 128 0D variables, which also shaping, high performing DT plasmas and discharges with
characterize 1D physical quantities such as safety factor at pellet injection. There is a particular interest in testing models
given positions and power deposition profiles. in high performance reference scenarios, as these discharges
The ITB profile database part is used for predictive mod- may be the most useful for extrapolating to next step devices.
elling of enhanced core confinement discharges and studies In addition, many machines have developed high performance
that require profile analysis [711, 713–716]. Thanks to this scenarios with internal transport barriers (ITBs). Profile
database, transport model predictions and turbulence micro- effects are known to be especially crucial in achieving these
stability calculations can be compared with experimental data conditions, and such discharges may hold valuable clues to
from various tokamaks, which help in determining the common the basic mechanisms that underlie anomalous transport in
mechanisms underlying enhanced core confinement. Though tokamaks. Within the framework of the ITB profile database,
the working versions of the ITB profile database and the gen- profile data has been collected for ITB discharges from a
eral profile database are managed independently, their contents number of machines, following closely the framework adopted
have been merged into a common database for public release. for the profile data that was released in 1998.
The 0D ITB database part can be browsed and The value of a coherent profile database, for testing
downloaded in various formats (U-File, CSV, SQL) at the physics models against experimental data from a wide variety
URL: http://www-itpa0d.cea.fr (common public website for of machines, is being increasingly appreciated by the fusion
ITPA 0D L-mode and ITB databases). The ITB profile community. Fusion scientists, who are studying a wide variety
database part can be browsed and downloaded at the URL: of tokamak phenomena (e.g. internal transport barriers, the
http://tokamak-profiledb.ukaea.org.uk (common website for edge pedestal, steady-state scenarios, pellet injection, etc),
ITPA profile databases). need access to profile data so as to develop their models. There
is great benefit in having uniform methods for accessing such
5.1.6. ITPA profile database. The international multi- data from a variety of machines. The ITPA profile database is
tokamak profile database was developed in the late 1990s broadening its scope to accommodate the type of profile data
for the testing of local models of anomalous heat transport that is required in a wider range of tokamak physics analyses.
in tokamaks, and approximately two hundred tokamak As a first step in the broadening of the profile database, the
discharges were collected from a large number of major ITB discharges with profile data are being incorporated, and
devices including ASDEX Upgrade, Alcator C-Mod, DIII- when this task is completed, it is intended that the integrated
D, FTU, JET, JT-60U, RTP, T10, TFTR, Tore Supra database will be made publicly available.
and TEXTOR. These discharges cover a wide range of
confinement modes, heating schemes and plasma parameters, 5.2. Pedestal and edge characteristics
and a substantial subset was used in the testing of local
transport models. This model validation exercise improved For the prediction of ITER performance, the prediction of
confidence in applying transport models to predict plasma pedestal pressure and/or temperature is one of the essential
performance in future devices such as ITER and placed ingredients due to the strong link between the pedestal and
such calculations in context. The transport model validation the core through stiffness in the temperature profiles. In this
work has been reported in [2, 234] and in 1998 the resulting section, various approaches to predicting the pedestal pressure
database was made publicly available to the wider fusion and/or temperature developed in the past several years are
community. summarized according to the details presented in section 4.
The profile database has evolved considerably since Emphasis is placed on the Type I ELMy regime, which is the
then, both through the addition of discharges and reference operation mode in ITER.
through a number of major improvements to the database
infrastructure. The database is available online at: 5.2.1. MHD approach. Recently, it has become widely
http://tokamak-profiledb.ukaea.org.uk. Data is stored both in accepted that the critical pressure gradient and trigger of
the original ASCII U-File format and in the more convenient Type I ELMs are governed by an ideal MHD stability. Both

S107
E.J. Doyle et al

t=0.5 G=0.5 C t=0.5 G=0.5 C

25
100
1.2
20
1

n [1020m-3]
15 0.8
T [keV]

10

Q
0.6
10
0.4
5 0.2
0 1
0 0 1 2 0 0.4 0.8 1.2
0 1 2
(b) rmid (m) (c) 20 -3
(a) rmid (m) <ne > [10 m ]
a Te Ti a n *10
α t=0.5 G=0.5 no C
n n *10 t=0.5 G=0.5 C
e
C

Figure 91. Predicted ITER pedestal structure and core, (a) temperature and (b) density, with the ICPS model in the ASTRA 1.5-D transport
code, which models the critical pressure gradient by analytic formula derived from MHD calculation and the E × B shearing stabilization of
the turbulence in the pedestal region. Fusion gain Q is also shown in (c) for the case with and without carbon impurities. Reprinted with
permission from [516].

the pressure gradient and current density in the pedestal separately in most of the codes. Thus, the condition for
region play an essential role in determining the maximum the critical gradient could be less accurate than the detailed
achievable pressure gradient. The gradient is ultimately MHD stand-alone calculations, whereas transport codes can
limited by ideal MHD modes with an intermediate toroidal calculate various aspects of the pedestal characteristics and
mode number n, typically n ∼ 6–12, which are known as their transient behaviours, e.g. the dynamic behaviour of
coupled peeling–ballooning modes. Detailed MHD codes ELMs, the integrated nature of the pedestal and core and
have been developed to calculate these intermediate n modes, separate treatment of temperature and density pedestals. For
which include MISHKA [670], KINX [671] and ELITE [673]. the prediction of the pedestal pressure/temperature with these
MISHKA can in principle deal with a range from n = 1 up to transport codes, a model of the pedestal width is also necessary
n ∼ 20–30. It also has the capability to explore diamagnetic as in the MHD approach. An example of this model is
effects. KINX can cope with a similar range of toroidal mode to employ turbulence suppression by the E × B shearing
numbers but has the unique capability to treat the separatrix rate together with the magnetic shear [516]. With this
geometry. ELITE has been developed specifically to treat model, the pedestal structure is self-organizationally formed
intermediate to high n modes (typically accurate for n > 5) and accordingly the pedestal width and height are naturally
efficiently at the plasma edge; it has been used for extensive determined. The ITER prediction by this code is shown in
parameter scans. In addition to these codes, the low n codes, figure 91, in which temperature (a) and density (b) profiles
such as GATO/ERATO [674, 675], can now access modes as for the average density of ≈1020 m−3 are calculated for the
high as n ∼ 8. Once the critical pressure gradient is calculated entire plasma region inside the separatrix. Here, the adjusting
with these codes, the pedestal pressure/temperature can be parameters t and G are calibrated with ASDEX upgrade
evaluated by an appropriate assumption of the pedestal width. and JET experiments as was explained in section 4.6. In
Such a predictive calculation of the pedestal temperature figure 91(c), the fusion gain Q calculated for various average
(pedestal density is assumed as 7.1 × 1019 m−3 ) for ITER is densities are also shown with and without carbon impurities.
performed with the ELITE code and is shown in figure 88 In some codes [576], the pedestal width is artificially
(in section 4). In this figure, the maximum stable pedestal fixed. The emphasis of this method is placed on calculating
temperature is shown as a function of pedestal width for a range the detailed behaviour of ELM dynamics instead of predicting
of toroidal mode numbers, n. The governing n that determines the pedestal pressure/temperature accurately. In another
the maximum stable pedestal temperature differs from n = 30 code [568], the pedestal temperature is specified by scaling
for a narrow pedestal width (/a
3%) to n = 15 for a wide formulations derived from various models of the pedestal width
pedestal width (/a
5.5%). If /a ≈ 4–5% is assumed, to evaluate the core plasma performance, and thus this method
the pedestal temperature for ITER is expected to be 6–7 keV. can be categorized as a scaling approach discussed next.
The remaining key point in this approach is an appropriate
assumption of the pedestal width. 5.2.3. Scaling approach. In this approach, scaling
formulations of the pedestal pressure are derived based on
5.2.2. Transport modelling approach. In this approach, the critical pressure gradient by an MHD model together
a 1.5D transport code is used to predict the pedestal with various models for the pedestal width. Unknown
pressure/temperature. Representative codes developed for coefficients and exponents are determined by the least square
this purpose are summarized in table 2 (in section 4). error method using experimental data for the pedestal pressure
Since detailed MHD calculation for the critical pressure or temperature. One of these approaches is described in
gradient is time consuming, it is calculated with an analytical section 4.2. In this method, a simple analytic expression for
formula derived from the MHD code calculation performed the critical pressure gradient is assumed to be modified by the

S108
 Chapter 2: Plasma confinement and transport

plasma shaping and beta, which is suggested by the coupled

peeling–ballooning mode theory. Using the pedestal width
√
model,  ∝ ρpol a [720], and fitting with the experimental
data archived in the ITER pedestal database [534], the resulting
scaling expression is given in equation (22) [540]. Fitting

(y,2)
of the data and ITER prediction are shown in figure 63
(section 4). The predicted pedestal temperature for ITER is
5.3 keV, assuming a pedestal density of 7 × 1019 m−3 .
A similar investigation has evaluated the pedestal
temperature for a variety of pedestal width models using an
analytic formula for the critical pressure gradient against the
ballooning mode instability [721]. Numerical coefficients
for each pedestal width model are determined by the least
square error method using the ITER pedestal database [534].
Unfortunately, all these models provide rather large RMSE
Figure 92. H H factor (≡τth /τth,98y2 ) versus n/nG . Reprinted with
(typically 30–40%), and therefore the best model cannot be permission from [698].
identified based on the present database. Although the pedestal
temperatures predicted by some ofthe models are in a similar
the first approximation τth is in agreement with values given by
range, e.g. 2.9 keV for  ∝ R βp [236] or 2.7 keV for
the above scaling (MAST [723]), although τth can reach values
 ∝ ρtor S 2 [722], further systematic improvement of the of more than 20% greater than those given by IPB98(y,2) and
model of pedestal width is essential. indicates a non-linear behaviour of the power degradation on
Although the predictive capability for the pedestal plasma current (NSTX [724]). The detailed dependence on ε
pressure/temperature in the Type I ELMy regime has and on other parameters is currently being assessed.
progressed significantly over the past several years, models for
the pedestal width still have large uncertainty. Examinations
5.3.2. High-density H-mode. As for the density dependence
of various models for the pedestal width are described in
of τth , the favourable dependence, τth ∼ n0.4 19 , is lost
section 4.2 in detail. Further development of the model and its
when the average density approaches the Greenwald density,
validation with a more solid database can greatly promote the
nG (1020 m−3 ) = Ip /π a 2 (MA,m), and the confinement
predictive capability for all of the approaches listed above.
enhancement factor, HH 98(y,2) ≡ τth /τth,98y2 , decreases below
unity. This degrading nature has been widely observed in
5.3. Global scaling various tokamaks. It has also been found that strong shaping of
Global scaling expressions for the energy confinement time the plasma cross-section, such as increasing triangularity, can
(τE ), or the stored energy (W ), are powerful tools for predicting mitigate this degradation at high n/nG values [273, 494, 725].
the confinement performance of burning plasmas. These Figure 92 shows the behaviour of HH 98(y,2) against n/nG
expressions are described using engineering parameters, such for the ITPA global energy confinement database, where the
as the major radius (R), minor radius (a) or inverse aspect degradation is seen for n/nG > 0.8 [698]. Although the
ratio (ε = a/R), elongation (κ), toroidal magnetic field (Bt ), averaged density becomes high, the peripheral density stays
plasma current (Ip ), electron density (ne ), heating power (P ) low and the high confinement can be maintained. Spontaneous
or loss power (PL ≡ P − dW/dt) and ion mass number (M). peaking of the density profile with simultaneous gas fuelling
One of the most reliable scaling expressions since 1998 for the and good divertor pumping in DIII-D [365], or with reduced
ELMy H-mode thermal energy confinement time (τth ) is the gas puffing and allowing for longer timescale in JET [355], was
so-called IPB98(y,2) scaling [2]: found to be important for achieving good confinement at high
−0.69 1.97 0.58 0.78 0.19
density. Introducing a density peaking factor, γn = (n0 /n̄ +
τth,98y2 = 0.0562Ip0.93 Bt0.15 n0.41
19 PL R ε κa M n̄/ n)/2 (n0 is the central density, n̄ the line-averaged density
(30) and n the volume-averaged density) [690], the enhancement
19 −3 factor was roughly given by HH 98(y,2) = 1 − 0.07(n̄/nG ) +
(in s, MA, T, 10 m , MW, m). The effective elongation
0.17γn [698]. Methods to control the density peaking have
is defined as κa = Sc /πa 2 , where Sc is the plasma cross-
not yet been fully established. The good confinement can
sectional area. The interval estimation of τth in the ITER
be achieved even when the loss power, PL , is near the L–H
FDR with the use of such scaling expressions was studied in
transition threshold, PLH . The enhancement factor was found
detail [2, 701]. Later estimation for the present ITER design
to lie around unity for a wide range of PL /PLH values [698].
using the extended database ITERH.DB3 showed a smaller
interval of a 95% log-linear uncertainty (+14%/ − 13%) than
that for ITER FDR (+25%/ − 20%) [704]. 5.3.3. Two-term scaling and beta dependence of ELMy
H-mode confinement. The expression of IPB98(y,2) is
converted to a physics form [2]:
5.3.1. H-mode in low aspect ratio tokamaks. The IPB98(y,2)
scaling was obtained from the data of standard tokamaks with τH 98(y,2) ∝ τB ρ∗−0.7 β −0.9 ν∗−0.01 , (31)
0.15 < ε < 0.45. The later H-mode experiments in tight
aspect ratio tokamaks expanded the database significantly to where τB ∝ a 2 Bt /T is the Bohm confinement time (T is the
0.65 < ε < 0.8 and towards higher toroidal beta values. To plasma temperature), ρ∗ ∝ T 0.5 /aBt is the normalized Larmor

S109
E.J. Doyle et al

radius, β ∝ nT /Bt2 is the beta value and ν∗ ∝ na/T 2 is scaling based on the database ITERH.DB2.8 [2]. Data from
the collisionality. Strong degradation of τth with increasing β ASDEX, ASDEX Upgrade, Alcator C-Mod, COMPASS-D,
was at first considered a result of the MHD limitation in the DIII-D, JET, JFT-2M, JT-60U, MAST, NSTX, PBX-M, PDX,
pedestal stored energy, Wped , in ELMy H-mode plasmas. If START, TCV, TdeV and TFTR were used. In this analysis a
W = τcore PL +Wped and Wped (or βped ) is assumed independent weighting factor inversely proportional to the square root of
of PL , τE (≡ W/PL ) = τcore /(1 − βped /β) deteriorates with the number of observations from each tokamak was applied.
β even when the core energy confinement time, τcore , is A straightforward regression gives [709]
independent of β. According to such a consideration two- τth,04(1) = 0.0596Ip0.86 Bt0.21 n0.40 −0.65 2.00 0.69 0.84 0.08
19 PL R ε κa M .
term models or offset non-linear models have been developed.
An early-proposed expression for the MHD-limited Wped (34)
is given as Wped(1) = 0.082Ip Bt R 2 εκ(Bt R 1.25 )−0.1 (MJ), The physics expression is τth,04(1) ∝ τB ρ∗−0.8 β −0.66 ν∗−0.09 ,
and the core thermal energy confinement time τcore,th(1) = which has a β-degradation nature, a little weaker than τH 98(y,2) .
−0.4 2.3
0.043Ip0.6 n0.6
19 PL R ε(Bt R 1.25 )−0.15 [2, 495]. This was The condition of the dataset was examined in detail to study
obtained from the H-mode global confinement database the accuracy of τth,04(1) and τH 98(y,2) . When eight variables are
without using the pedestal database. An international pedestal used, as in equations (30) and (34), there is strong collinearity
database has been assembled [534], as described in section 5.2, between variables. As a result, three principal components
and various scalings for Wped have been developed from among eight do not have enough spread, being less than
it. It has been found obviously from H-mode experiments the spread of experimental measurement errors. To exclude
and the database that Wped increases with strong shaping these three smallest principal components, an ITER-like subset
of the plasma cross-section, such as high elongation and of the data was selected by windowing on the parameters
high triangularity. In order to include this shaping effect, 1.8 < M < 2.2, 1.6 < qcyl < 2.8 and 1.4 < κa < 1.9, and
a shaping factor Fq ≡ q95 /qcyl was introduced, where q95 only five variables (Ip , n, PL , R, ε) were used for the regression
is the safety factor at 95% poloidal flux surface and qcyl = analysis. Now all five principal components have enough
2πBt κa 2 /µ0 Ip R is the cylindrical safety factor [690, 726]. spread. A scaling expression for this ITER-like subset is given
One of the scaling expressions of the MHD-limited Wped as [709]
including this Fq variable is given as [706]
−0.55 1.73 0.56
τth,04(2) = 0.095Ip1.00 n0.37
19 PL R ε . (35)
Wped(2) = 0.01Ip2 Rε −1.68 (M/n19 R 2 )0.13 Fq1.28 (32)
The physics form of this scaling is τth,04(2) ∝ τB ρ∗−0.78 β −0.2
and
ν∗−0.2 . A weaker β and stronger ν∗ dependence agrees fairly
−0.6 2.55 1.87 0.88 0.2
τcore,th(2) = 0.065Ip0.45 Bt0.35 n0.6
19 PL R ε κa M . well with the non-dimensional transport experiments reviewed
(33) in sections 3.6 and 5.4 of this chapter.
This τcore,th(2) ∝ τB /ρ∗ is just a pure gyro-Bohm confinement
time, while τcore,th(1) ∝ τB /ρ∗0.78 . 5.3.4. Prediction for ITER. The predicted values of τth in
It has been known from H-mode experiments that the the ITER standard operation (Ip = 15 MA, Bt = 5.3 T, n19 =
pedestal stored energy and the core stored energy link with each 10 × 1019 m−3 , PL = 87 MW, R = 6.2 m, ε = 0.32, κa = 1.7,
other in many cases [2]. This means that Wped is not always M = 2.5, Fq = 1.5) are summarized here: τH 98(y,2) = 3.6 s,
independent of PL . Several expressions for Wped dependent on τtht−t (2) = 3.5 s (0.9 s+2.6 s), τtht−t (4) = 3.9 s (1.9 s+2.0 s),
PL as well as on Fq have been proposed; Wped(3) ∼ PL0.2 Fq1.67 τth,04(1) = 3.4 s and τth,04(2) = 3.5 s. The global thermal
[690, 706] and Wped(4) ∼ PL042 Fq2.09 [707]. For both the confinement time for two-term models is defined as τth,t−t =
expressions, the weak density dependence of Wped is a distinct Wped /PL + τth,core . In contrast to the above coincidence, the
feature from the global W . Since Wped increases with PL , prediction of τth for higher beta operation (or higher heating
this type of Wped expression was called a ‘thermal conduction power operation) in ITER is rather uncertain. Superiority
model’ [707]. The physics mechanism of the PL -dependent among the above scaling expressions is not defined at present.
Wped , however, is not so simple. The enhancement of the edge Consequently, the recommended scaling for ITER operation
MHD stability in high-βp plasmas is also a cause for the explicit remains the IPB98 scaling law, while this issue is further
PL dependence of Wped [501]. When these Wped scalings are investigated. More comprehensive studies in experiments and
adopted in the analysis of the core confinement, the stronger analyses are required to clarify the beta dependence of ELMy
power degradation of τcore,th compared with τcore,th(1,2) is found, H-mode confinement characteristics.
i.e. τcore,th(3) ∼ PL−0.65 [690,706] and τcore,th(4) ∼ PL−0.75 [707].
It is noted that τcore,th(3) is close to a gyro-Bohm one, τcore,th(3) ∝ 5.3.5. Scaling for higher confinement plasmas with ITBs. A
τB ρ∗−0.8 β −0.11 ν∗0.17 , while τcore,th(4) is a weak gyro-Bohm one scaling study for higher confinement plasmas with internal
with strong β degradation, τcore,th(4) ∝ τB ρ∗−0.4 β −1.3 . The transport barriers (ITBs) has been started. Relations between
latter result of the two-term modelling does not agree with the H -factor (HL89 ≡ τE /τITER−89P ) and plasma parameters
the result of non-dimensional transport experiments. One of were surveyed based on the international ITB database [202].
the causes for this difference comes from the different data Explicit scaling expressions for the global energy confinement
sets between pedestal and global databases. Matching of the based on this database have not been obtained so far (before
pedestal and global databases are planned for future work. 2004). With the use of a confinement database of reversed-
Recent analysis based on the ITPA H-mode database shear plasmas with box-type ITB in JT-60U, a scaling
DB3v12 [709] brought a similar scaling to the IPB98(y,2) expression of the core stored energy inside ITB region was

S110
 Chapter 2: Plasma confinement and transport

Table 5. ITER confinement projections scaled from DIII-D and JET Table 6. The ITER parameters are taken as: Ip = 15 MA,
results using non-dimensional parameters. BT = 5.3 T, n = 1020 m−3 , R = 6.2 m, a = 2 m, κ = κa = 1.75,
M = 2.5 AMU and PL = 87 MW for βNth = 1.6 and PL = 348 MW
Tokamak βNth ρ∗ /ρ∗ITER α δα τthITER (s) δτ (σ ) (S) for the βNth = 2.5 case.
DIII-D 2.1 5.5 1.1 ±0.4 12 ±6 βNth τth,98y2 (s) τthEGB (s) τthES (s)
JET 2.2 4.2 0.7 ±0.3 3.3 ±1.8
1.6 3.67 4.57 3.75
2.5 1.38 3.46 2.25
shown [727]. The core stored energy Wcore does not simply
increase with PL . A scaling of Wcore was given as Wscale = JG04.445-2c

Cεf−1 Bpf
2
Vcore , where εf is the inverse aspect ratio at the ITB
foot, Bpf is the poloidal magnetic field at the outer midplane 1.2
1.4
ITB foot and Vcore is the core volume inside the ITB foot.
This scaling is equivalent to the condition for the core poloidal
beta εf βp,core = C1 with C1 ≈ 1/4. It becomes clear that 1.2
5 10 5 1.0
the confinement scaling of ITB plasmas cannot be described

ne (1020 m-3)
only with external engineering parameters. Development of 1.0
0.8

ne / nGDL
the scaling of τE,core for these plasmas is left for future work. 3.0

0.8
5.4. Non-dimensional scaling 0.6

0.6 2.0
From ρ∗ scans in present devices, in which the other
dimensionless parameters β, ν∗ and q are fixed at their values 0.4
in ITER, it is possible to predict the confinement time in ITER 0.4
1.3
by projecting to the value of ρ∗ in ITER. This technique is (a) IPB98(y,2) 1.0
1.0
0.2 0.2
described in more detail in [2, 728]. Using the scaling relation
2 4 6 8 10 12 14
for the dimensionless confinement time <Te> (keV)

BT τth = ρ∗−(2+α) F (ν∗ , q, . . .)

JG04.473-14c
(36) 1.6
50
and taking into account the error in the determination of α 1.4 1.2
from the errors in the measurement of the thermal energy
confinement time, τth , one can derive the ITER confinement
1.2 1.0
time using the DIII-D and JET ρ∗ scans. The results for the 20
present version of ITER are given in table 5. The reason that
ne (1020 m-3)

the errors in τth are so large is the fact that the range in ρ∗ 1.0

ne / nGDL
0.8
is very small in the experiments (for toroidal magnetic field 3.0
10
values of 1 and 2 T in DIII-D: ρ∗1T /ρ∗2T = 1.6 and for 1 and 0.8
2.6 T in JET: ρ∗1T /ρ∗2.6T = 1.9). 0.6
The results from the dimensionless scaling experiments 0.6 2.0
have also been used in conjunction with the ELMy H-mode 5
0.4
database to predict the confinement time in ITER. In 0.4
this technique the regression is constrained to give the (b) Electrostatic 1.0 1.3 1.0
dimensionless scaling form of the single scan experiments. 0.2 0.2
For example Petty et al [729] constrained the fit such that it 2 4 6 8 10 12 14
satisfied the electrostatic gyro-Bohm constraints <Te> (keV)

BT τthEGB ∝ ρ∗−3 β 0 F (ν∗ , q, . . .). (37) Figure 93. POPCON plots of the ITER operation space in volume
averaged density and temperature showing contours of fusion gain
The expression had the form Q (red solid lines), normalized βNth (blue dashed and dotted line) and
power relative to the L–H threshold (green dotted lines). Energy
−0.55 1.81 0.3 1.75 0.14 confinement is assumed to scale as (a) IPB98(y,2) and (b) the
τthEGB = 0.028Ip0.83 BT0.07 n0.49
19 PL R a κa M (38)
electrostatic model of equation (40). nGDL designates the Greenwald
density in this case.
in engineering variables.
Similarly, McDonald et al [351] have derived an
For the ITER standard operating condition (thermal βNth = 1.6),
expression which satisfied the dimensionless electrostatic
equations (38) and (40) give a slightly higher value for τth in
constraint only:
ITER than the standard IPB98(y,2) scaling as can be seen in
BT τthES ∝ ρ∗−2.8 β 0 ν∗−0.09 . (39) table 6. However by operating at higher βNth and fusion power
output, a substantial improvement in the predicted τth is given
In engineering units this expression has the form by equations (38) and (40), as also shown in table 6.
In figure 93 POPCON plots comparing the operational
−0.55 2.14 0.74 0.78 0.1
τthES = 0.0487Ip0.72 BT0.09 n0.51
19 PL R κa ε M . (40) range for ITER with the conventional IPB98(y, 2) scaling and

S111
E.J. Doyle et al

that of equation (40), the scaling with zero β dependence, much more comprehensive than its predecessors in covering
are presented. The main difference is that high Qs are now many transport regimes. Some closed form expressions were
accessible at high temperatures with the β independent scaling. used for some diffusivities, but others required concurrent
The optimum operational point for Q = 10 is at higher values solutions of simplified dispersion relations followed by
of βNth (∼ 2.5) and average temperature (∼10 keV). This means quasilinear substitution and mixing length rules. The IFS-
that the steady-state operation with full current drive would be PPPL model [237] used a linear gyro-kinetic stability code to
more feasible with this type of scaling. fit a complicated parametric formula for the ITG threshold and
the early gyro-fluid ITG simulations to fit a heuristic model
5.5. Modelling codes, including edge modelling capability for the diffusivity. Taking nothing from experimental fits,
this was the first example of a ‘first principles’ model. The
The higher dimensional treatments (1.5D and 2D) in mod- GLF23 [20] used many of the methods from MMM95 and IFS-
elling codes introduce the next level of complexity over global PPPL to develop a comprehensive model covering ITG/TEM
analyses and are used to predict more details of plasma per- as well as ETG fit solely to ITG (with adiabatic electrons)
formance. Such modelling ultimately allows the prediction and ETG (with adiabatic ions) simulations but checked against
of the distribution of all relevant plasma parameters across ITG/TEM gyro-fluid simulations. All these models are local
closed magnetic surfaces in the plasma core and both across transport models (diffusivities dependent only on local plasma
and along open magnetic field lines in the SOL. A complete gradients) and have local gyro-Bohm scaling strictly accurate
picture of the plasma evolution and access to quasi-steady so- only at vanishingly small ρ∗ .
lutions requires modelling of electron and ion temperatures, all As noted earlier, several core transport models yielded
hydrogenic and impurity ion densities and neutrals, sources, comparable agreement with the profile database at the time
sinks, MHD equilibria, as well as the current distribution the IPB was published. It was suspected that much of the
and plasma rotation. Such a fully self-consistent simulation remaining scatter was due to E × B rotational shear damping,
is an enormous challenge and is far from being fully im- geometry and other effects that were being explored in more
plemented. Nonetheless, significant progress in understand- comprehensive kinetic models but not adequately expressed in
ing the physics of the underlying transport processes and its the core transport models used in 1.5D simulations. Since then
implementation in integrated core-edge transport codes has there has been a continuing effort to upgrade these models. One
been reported in the past few years, which improves our com- of the more sophisticated and widely used of these is the GLF23
prehension of plasma dynamics and advances our predictive model; its development history is summarized here briefly to
capabilities. illustrate the improvements being made in core models.
At the time the IPB was published, several core The GLF23 model has been retuned recently [236] in
confinement models were able to model the temperature order to extend its validity to negative magnetic shear and
profiles of a series of H-mode and L-mode plasmas in pedestal parameters for low to moderate values of MHD α.
the profile database [234] with comparable accuracy when The retuning yields improved agreement with the linear gyro-
the boundary conditions and density profiles were given kinetic growth rates for reversed magnetic shear and H-mode
by experimental values. Reference ITER confinement pedestal parameters, particularly for large values of MHD α.
projections from these models were then primarily based The retuned model has been tested against a variety of L-
on the assumption of flat density profiles with boundary and H-mode discharges with and without negative magnetic
conditions for densities and temperatures based on rough shear [533]. Modelling results using the XPTOR code have
approximations to the expected pedestal height. But sensitivity yielded good agreement with the temperature profiles for
studies also showed that ITER performance was very sensitive several DIII-D ITB discharges with strongly reversed magnetic
to the boundary conditions and density profile. In the shear (see figure 94). Simulations have also demonstrated
following four subsections we review the progress on the that the model is successful in predicting the density profile
implementation and validation of some of the most advanced in discharges without ITBs. In discharges with ITBs, it was
physics components for improving those projections: core found that some additional background particle diffusivity is
transport, the H-mode pedestal, ELMs and the scrape-off layer. needed in order to reproduce the measured density profiles
Then we summarize analyses that use at least a subset of these within the barrier region where the ITG and TEM transport has
more advanced models to evaluate ITER performance in the been quenched by rotational shear stabilization. The saturation
areas of energy confinement, density peaking and impurity levels in the original GLF23 model [20] were normalized to
profiles. Applications to steady-state and hybrid scenarios are non-linear gyro-fluid simulations. The new version of GLF23
reported in chapter 6 of this issue [199]. has been renormalized using non-linear GYRO gyro-kinetic
simulations (see figure 21). As a result, the stiffness of the
5.5.1. Developments in core transport models. Core model has been reduced.
transport code models necessarily started with simplified It is worth noting here that although self-consistent
theoretically motivated models with several coefficients modelling of all main plasma parameters, including ion
adjusted to best fit the experimentally measured profiles, given density and toroidal rotation, is an ultimate goal of any
the calculated sources. The multi-mode model (MMM95) predictive modelling, its role is paramount in a self-consistent
[238] is a good example. Various strength coefficients are description of the radial electric field. The latter determines
assigned to the diffusivities of various component modes: ITG the shearing rate of plasma rotation, seen as one of the
and TEM from the Weiland model [154] plus kinetic Alfvén main factors in suppressing plasma turbulence and developing
and resistive ballooning mode contributions. MMM95 was both edge and internal transport barriers (ETBs and ITBs).

S112
 Chapter 2: Plasma confinement and transport

through the ETB) are used as boundary conditions. This is the

outermost flux surface, where drift wave turbulence still makes
a significant contribution to anomalous transport. Another
group [566] uses a combination of shear in plasma rotation
and strong magnetic shear as two stabilizing factors to induce
the ETB. The MMM transport model [236,238,239], combined
with the NCLASS neoclassical model [185], was used in the
simulations. It is worth noting that the presence of a drift-
resistive ballooning mode in MMM95 (or drift Alfvén waves
in its later realization [145]) allows setting boundary conditions
just inside the separatrix. Finally, two other groups [576, 733]
prescribe the ETB width using a number of theory-motivated
scalings [721]. Different assumptions about transport within
the ETB are used, but the most common is that the long
wavelength turbulence (responsible for ion transport) is fully
suppressed so that ion heat flux is controlled by a collisional
transport only.
Figure 94. Experimental profiles for a strong NCS DIII-D plasmas
with an L-mode edge are compared with a simulation using the
retuned GLF23 transport model. The points represent the measured 5.5.3. Modelling the effect of ELMs. Simulation of ELMs is
temperatures and the lines represent the mode predictions. another area of active research in predictive modelling, with the
Reprinted with permission from [533]. various types and effort to model them described in section 4.
Simulation of ELMs introduces two additional ingredients,
Fully self-consistent modelling of the radial electric field which were not previously used in predictive modelling. The
(including toroidal rotation) and its role in the formation first one is an incorporation of MHD stability analysis of the
of an ITB has been recently reported [253, 716]. Three plasma within the ETB into the simulation. The simplest
pairs of shots from JET, JT-60U and DIII-D with ITBs (one approach, used by all participating parties [566, 567] is to use
shot with and another without negative magnetic shear) were analytical formulae for ballooning and peeling mode stability.
simulated, using either the mixed Bohm/gyro-Bohm [397,730] If one of the stability criteria is violated, either the transport
or Weiland [731] models. One of the main results of this coefficients within the barrier are temporarily increased to
study was the recognition of high sensitivity of the ITB simulate an ELM [567, 658, 734] or the plasma parameters
formation and its further evolution with respect to a wide within the ETB are reconstructed to take into account the
range of plasma parameters. Many researchers use a somewhat edge losses during the ELM [735]. Another option, used
simplified, not fully self-consistent approach to ITB modelling in [566, 574] is to use a pressure gradient that corresponds to
[253, 714, 715, 732]. Figure 95 shows an example of the the ideal ballooning stability limit as a boundary condition for
modelling of a JET ITB plasma from [714]. The Bohm term the plasma pressure at the top of the ETB. Whatever approach
was assumed to have a threshold represented by a Heaviside is used, it is very important to check the validity of simple
step function, H (0.5−s −CωE×B /γ ), where s is the magnetic analytical MHD stability criteria with comprehensive MHD
shear, C an adjustable factor of order unity, γ the growth rate stability codes. This requires an interface between transport
and ωE×B the rotational shearing rate. Only electron and ion and MHD stability codes. Because of a strong disparity in
temperatures were allowed to evolve in this simulation. The CPU time needed to evolve transport equations (a few minutes)
rest of the plasma parameters (density of ions and impurities and to perform a comprehensive MHD stability analysis (a
and toroidal rotation) were taken from experiment. Although few hours for peeling and finite ballooning modes), these
these studies indicate that progress has been made in modelling two simulations are not used in a feedback loop at present.
some aspects of ITBs, further model development is necessary Instead, results of transport simulation taken shortly before
before a fully predictive capability exists for ITER advanced the ELM crash are stored in a special file and transferred to
scenarios. MHD stability codes. This file contains information about
simulated pressure and current profiles as well as details of the
5.5.2. Transport simulations using models for the H-mode plasma equilibrium. A high quality equilibrium solver in the
pedestal. Modelling of ELMy H-mode has advanced much transport code is therefore needed in order to have a meaningful
during the past few years as well. First of all, several comparison between the simplified stability analysis, used
groups now explicitly include the edge transport barrier by the transport code, and the one generated by a dedicated
in their simulations [193, 516, 533, 566] with a range of MHD stability code. Comparison between transport and MHD
assumptions about the mechanism of the barrier formation and codes in terms of the MHD stability is then used to re-adjust
about transport coefficients within the ETB. One approach analytical MHD stability formulae implemented in transport
[533] calculates the radial electric field near the separatrix codes.
(using a neoclassical approximation for poloidal rotation and
neglecting toroidal rotation) that suppresses ITG and TEM 5.5.4. Modelling the connection between the ETB and the
turbulence in the GLF23 model if the shearing rate exceeds SOL. The edge transport barrier is a narrow region that
the corresponding growth rates. Experimental temperatures separates the plasma core from the scrape-off layer and both
and densities at normalized minor radius ρ = 0.9625 (halfway areas influence the ETB structure and behaviour as a result.

S113
E.J. Doyle et al

(a) JET #46664 1.2 (b) JET #46664 5 (c) JET #46664
3.0 T=6 s, Ti T=6 s, Te T=6 s, q
JETTO predict 1.0 JETTO predict JETTO predict
4
0.8 EXP

104 eV
2.0
C=1.0 C=1.0
C=1.5 C=1.5
0.6
EXP 3
1.0 0.4

0.2
2
0.0 0.0
10
1.4 (d) JET #53521 (e) JET #53521 (f) JET #53521
8.0
1.2
T=6 s, Ti T=6 s, Te T=6 s, q
JETTO predict JETTO predict 8 JETTO predict
1.0 6.0
103 eV
0.8 C=1.0 C=1.0
C=1.5 C=1.5 6
0.6 4.0

0.4 EXP 4
2.0
0.2
EXP
0.0 2
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Figure 95. Predicted temperatures using the JETTO model compared against the experimental profiles for JET discharges #46664
(optimized shear) and #53521 (reversed shear). Reprinted with permission from [714].

Figure 96. Expanded profiles of total electron and ion transport coefficients and the E × B shear reduction factor near the edge in ITER
(without symbols at 100 s, with symbols at 400 s) using boundary conditions implemented as a set of scaling relations derived from
B2-Eirene modelling of the divertor. Reprinted with permission from [574].

Modelling of the pedestal structure is discussed in more detail The next important finding of integrated core–edge
in section 4.5. Here we will try to highlight the role of the modelling comes from the realization that gas puffing controls
SOL in the ETB formation and its evolution, as well as the the density within the ETB rather than in the core. And that the
importance of an integrated approach towards modelling of radial distribution of the ion density within the ETB can vary in
ETBs. The radial distribution of cold neutrals is probably accordance with the gas puffing even if the ion density inside
the most important ingredient of predictive modelling that is the ETB does not change. It was shown that the de-coupling
influenced by the SOL. It has been recognized that penetration between core and edge density influences MHD stability of the
of cold neutrals through the SOL may influence the structure ETB and dynamics of ELMs [734]. The somewhat simplified
of the ETB [538] as well as ELM dynamics [736]. Therefore integrated core–pedestal SOL (ICPS) model has been used
self-consistent predictive modelling of ELMy H-mode plasmas in the ASTRA code for ITER studies [574]. In ICPS the
requires 2D modelling of the SOL as well as the plasma conditions at the separatrix are parametrized for different ITER
core. Few attempts have been reported, first of all because of scenarios based on an extensive modelling of the SOL plasma
the complexity of the interface between a 1D core transport with the 2D SOL transport code B2/EIRENE. The suppression
code and 2D SOL codes [514, 734, 737]. Probably the of transport coefficients near the plasma edge by E × B shear
most important finding has been that the penetration of cold damping is shown in figure 96.
neutrals from external gas puffing through the separatrix is a
very non-linear function of plasma density and machine size 5.5.5. Confinement projections for ITER. Here we focus
[737]. Predictive modelling of the SOL plasma for an ITER on a few aspects of modelling confinement in ITER standard
configuration shows that gas puffing cannot provide adequate inductive H-mode plasmas that have been influenced by
core fuelling of ITER independently of the level of gas progress or new information since the IPB. (1) Although
puffing [514]. the most commonly exercised physics-based models for core

S114
 Chapter 2: Plasma confinement and transport

Figure 97. Fusion gain Q predicted for ITER by the MMM95

(MM) [238], IFS/PPPL [237] and GLF23 [20] transport models at
the same input parameters using the ASTRA code [738]. Also
shown are predictions of the Weiland model at similar input
parameters [739]. Along the horizontal axis are indicated the ranges
of expected pedestal temperatures from various models (a) and
(b) [707], (c) [541], (d) [722], (e) [540]. Reprinted with permission
from [750]. © 2003 Institute of Physics.

transport generally exhibit less stiffness than the range of Figure 98. Steady-state distributions of the plasma density, ne , and
models considered in the IPB, there is still a large uncertainty electron and ion temperatures, Te and Ti , respectively, as functions
in the projections due to the uncertainty in the pedestal. (2) of the minor radius a in the midplane. Full curves show the results
of the GLF23 model with the boundary conditions
New evidence for an anomalous pinch that generally increases ne,ped = 7.8 × 1020 m−3 and Te,ped = Ti,ped = 5 keV at the pedestal
with decreasing collisionality influences many aspects of top. Results of the ITER reference model are shown as broken lines.
performance: the fusion production, edge density at a given The dotted line shows the prediction of the turbulent equipartition
power and turbulence growth rates. (3) Transport of low Z (TEP) theory where the shear profile is defined by the GLF23
impurities, especially He ash, is dominated by turbulence and modelling. Reprinted with permission from [373].
impurities are not expected to accumulate to very high levels,
and although higher Z impurities are increasingly influenced because it is now less stiff. This illustrates that they yield
by neoclassical effects, the neoclassical ion temperature very similar projections for a given pedestal temperature. The
gradient screening and turbulence-driven diffusion may help uncertainty in the pedestal temperature appears to be more
reduce the possibility of their accumulation in the core. critical than differences between the models.
Projections for steady-state and hybrid scenarios are discussed In the ITER studies by Pacher et al [574] using the ICPS
in section 6 of chapter 6 of this issue [199]. model discussed above, Q values were found to range between
5 and 20 depending on pumping speed, heat flow across the
Energy confinement and sensitivity to the pedestal. Drift separatrix, average plasma density and allowable peak heat
wave transport models have demonstrated success in predicting power on the divertor plate. So it is very clear that core, edge
the experimental profiles in a variety of tokamaks when using and SOL simulations need to be integrated to reduce the overall
experimental data for the temperatures, densities and rotation uncertainty in ITER performance.
near the plasma boundary. But improved models for the
ETB are clearly needed in order to increase confidence in Density peaking. As described in section 3.4.6, recent
predicting the performance of burning plasmas. Simulations studies provide evidence of density peaking at low
using various core transport models in combination with collisionality that is apparently governed by turbulence. This
various models for the pedestal illustrate the sensitivity of may impact ITER performance in many ways, including
ITER performance to uncertainties in the projected pedestal energy confinement (density-gradient-dependent turbulence),
characteristics. Figure 97 shows the resulting fusion Q as pedestal densities, fusion power production and impurity
a function of pedestal temperature for the MMM95 [238], accumulation. Pereverzev et al [373] have recently employed
IFS/PPPL [237] and GLF23 [20] transport models at the same the GLF23 model as well as the turbulent equipartition (TEP)
input parameters using the ASTRA code [738]. Also shown are theory [374] in a study of their effects on ITER density and
predictions of the Weiland model at similar input parameters temperature profiles as well as performance. They compare
[739]. Along the horizontal axis are indicated the predicted the profiles with the results of an often-used ITER reference
ranges of several pedestal models discussed in section 5.2. The model [372], which uses a fixed diffusivity profile shape with
retuned GLF23 model would be expected to yield somewhat amplitude normalized to match the ITERH-98P(y,2) global
higher Q values than indicated for the earlier version of GLF23 confinement scaling. The differences in the profiles are

S115
E.J. Doyle et al

shown in figure 98. The density profiles for these cases should improve our understanding of particle transport and
were discussed in section 3.4 and shown in figure 34, but here how peaked we can expect the density and impurity profiles to
we have added the temperature profiles to show that there are be in ITER.
corresponding changes in the temperature profiles, i.e. they
are not independent. It is seen that the GLF23 and TEP models 5.6. Summary of progress and remaining issues
both show more peaked density profiles and flatter temperature
profiles than the scaling model and consequently yield ∼30% Our various approaches to projecting ITER confinement
more fusion power when the impurity profiles were kept fixed. capabilities have progressed well since the IPB, supported
But, as we discuss next, peaking of the impurity profile may by improvements in diagnostic measurements, theoretical
cancel this benefit. understanding and computational models. The six interna-
tional multi-machine databases developed and maintained by
Impurities. Important impurity issues to be examined with the ITPA confinement-related topical groups have been a vital
core–edge integrated modelling include the profiles and resource for developing improved qualitative and quantitative
amounts of intrinsic or seeded impurities and He ash in the understanding of confinement over a broad range of parame-
core. Both processes involve successive impurity transport ters. We expect these to continue to be an important bridge
through the core and the SOL and, as such, require a link between the present experimental program and burning plas-
between 1D and 2D transport codes. A number of stand- mas in ITER—by providing a basis for validating more com-
alone simulations for either core or SOL have been reported prehensive physics models and for designing the experimental
previously [737, 740–743]. An attempt at self-consistent program.
modelling of ITER-EDA using coupling between the 1.5D Some of our greatest progress has been in an improved
core transport code JETTO/SANCO and the 2D SOL code qualitative understanding of the physics of the plasma edge.
EDGE2D/NIMBUS has been reported [742]. As expected The interactions between MHD and transport processes in
from such complicated simulations, they observed a strong the edge are very complex. The dynamics takes place over
sensitivity of the predicted impurity and He confinement to a wide range of timescales from the fast ELM crashes and
assumptions about anomalous transport both in the core and bifurcations of L–H transitions to the slower evolution of the
the SOL. bootstrap current during the quiescent periods in between. But,
More recently, Pacher et al [744] examined intrinsic we still need to develop more reliable physics-based models for
carbon in ITER with the ICPS model in ASTRA. Using the L–H transition, the pedestal width and for the full range
the MMM95 model for the anomalous thermal conductivity of ELM types. Such knowledge would give us not only a
with E × B shear stabilization fitted to JET and ASDEX better predictive capability but would allow us also to identify
Upgrade experimental results plus a simplified model for opportunities to intervene with external control techniques for
neoclassical transport, the authors found that the increased controlling the pedestal and ELM characteristics consistent
peaking of the carbon density essentially cancelled out any with optimal SOL and divertor performance.
increases in performance that might be expected as improved Global scaling, identity experiments and detailed
core confinement (there was no anomalous particle pinch). modelling all show general agreement on ITER’s performance
Leonov and Zhogolev [745] examined higher Z impurities and have similar uncertainties. Global scaling studies
(tungsten and argon) with a much more sophisticated model for have historically provided the first step in evaluating
the impurities in ASTRA: the multi-species dynamic impurity performance characteristics of machines during their design
code ZIMPUR was used to model the ionization states of up to phases and generally have proven to be very reliable for
three impurity fluxes simultaneously and NCLASS [185] was moderate projections beyond the existing experimental physics
used for the full neoclassical transport of each charge state. base. More recently, these have been augmented with
Even in the cases of high anomalous transport, both the Ar and dimensionless scaling experiments that constrain the geometry
W densities were peaked towards the edge where they radiated and a few other parameters to be more closely aligned with the
a large fraction of the power without significant contamination target design. The dimensionless scaling experiments allow us
of the core. This was because the neoclassical ion temperature to identify trends that may be more critical in the direction of
gradient screening effect provided a stronger outward flux than extrapolation than in the broader database. Several issues that
the inward flux driven by the very flat fuel density profiles, as have been identified through these comparisons are: (1) how
described in section 3.4.8, and because it was assumed that global confinement scales with beta, (2) whether collisionality
there was no anomalous pinch. Polevoi et al [746] performed or proxmity to the Greenwald density is more important in
a survey of the sensitivity of ITER performance to variations governing confinement degradation at higher densities and (3)
in the particle diffusivity and pinch while using high field the effects of strong shaping (e.g. elongation and triangularity)
side pellet launch for deeper fuelling and shallow pellets for on confinement. A better understanding of these trends will
ELM control. The opposing consequences of fuel ion and reduce the uncertainty in projecting ITER performance.
impurity peaking resulted in very little net change in plasma Time-dependent modelling of the plasma profiles has
performance over the range of parameters considered. often used assumed profiles for transport coefficients with
These studies highlight the many competing effects amplitudes constrained by global fits. Although use of these
to consider in modelling the density and impurity profiles models allowed the investigation of issues of dynamics, access
and their effects on performance. Recent emphasis in the to interesting regimes and superposition of expected new
experimental program to identify the parameters that govern physics, they did not add substantially new perspective on
density peaking, even in the extreme conditions of non- the overall expected confinement. The success of theory-
inductive operation (no Ware pinch) and no central sources, based turbulence models in simulating experimental plasmas,

S116
 Chapter 2: Plasma confinement and transport

which was highlighted in the IPB, has continued. They are momentum transport to match that which exists for ion thermal
now viewed as a very strong complement to the global scaling transport.
models and identity experiments in projecting to ITER. Since Remarkable progress has also been made in developing
the IPB, the theory-based models have converged somewhat and understanding regimes of improved core confinement since
on their stiffness and thereby yield a narrower range of the publication of the IPB. Internal transport barriers and other
projections when using a common set of boundary conditions forms of reduced core transport are now routinely obtained in
at the top of the pedestal. Narrowing the uncertainty in all the leading tokamak devices worldwide. Reduced transport
the overall confinement projections will be heavily based on has been achieved in all four transport channels (ion and
integrating with improved models for the edge. Experimental electron thermal, particle and momentum transport channels),
observations and theory also tell us to be cautious about the sometimes simultaneously, and ion thermal transport is often
present extrapolations with these models until we strengthen reduced to neoclassical levels. A wide range of transport
them in the areas where they require the most extrapolation dynamics in these reduced transport modes can be understood
to ITER. These include: (1) electron transport in plasmas in terms of the interplay between turbulence drive and
with predominant electron heating, (2) density peaking at low suppression mechanisms, the latter including E × B sheared
collisionality in plasmas with small or negligible sources in the flows, magnetic shear reversal, α-stabilization and impurity
core, (3) rotation in low torque plasmas and (4) the formation effects on turbulence growth rates. These mechanisms can
and sustainment of ITBs. The ITB issue is of particular successfully explain regimes such as the radiation-improved
relevance for the enhanced confinement that will be needed confinement (RI) mode, which were poorly understood at the
for the steady-state and hybrid scenarios discussed in chapter time of the IPB. This rapid progress in the development of
6 of this issue [199]. Although they appear to have many enhanced confinement modes of operation has contributed
characteristics in common with ETBs, their onset, location to an increased emphasis on developing both hybrid and
and strength is much more variable than ETBs, and they steady-state operation modes for ITER. However, a concern
therefore present an even greater theoretical and computational about these operating regimes on present devices is that
challenge. they are typically hot ion modes, at moderate density,
with high plasma rotation rates due to external momentum
6. Summary input, i.e. they typically operate under non-reactor conditions.
Progress has been made with regard to extrapolation to
Transport in tokamak plasmas is primarily governed by non- reactor conditions, though in all cases further investigation
linear turbulence processes, with multiple turbulence drives and experimental demonstration of robust high confinement
and suppression mechanisms, occurring on multiple scales. operation are desirable. Ion thermal transport is now relatively
Despite this intrinsic complexity, very considerable progress well understood and is believed to be regulated by ITG-
has been made in understanding, controlling and predicting type turbulence. Theory-based modelling can qualitatively
tokamak transport across a wide variety of plasma conditions replicate ion transport across a wide range of operating
and regimes since the publication of the ITER Physics Basis regimes with quantitative agreement in many cases. However,
(IPB) document [1]. transport in the other channels (electron thermal, particle
The understanding of fundamental transport processes and momentum) is relatively not as well understood as
has been considerably advanced by analytic theory, numerical ion transport, either experimentally or theoretically. In
turbulence simulations and neoclassical transport theory, as particular, momentum transport and plasma rotation are less
well as by efforts to benchmark theory via direct comparisons well understood than the other plasma transport channels;
to turbulence measurements. A major step forward here an ability to predict momentum transport and rotation for
is in the capability of non-linear, gyro-kinetic turbulence ITER is currently lacking. Using non-dimensional scaling
simulation codes. These codes can now treat large fractions techniques, significant new work on the scaling of energy
of the plasma radius, a, on relevant turbulence time scales transport with β has been performed, showing a weak or
(100s of a/cs , where cs is the sound speed) and with null dependence of the transport on beta, in contrast to both
expanded physics content, such as non-adiabatic electrons theoretical expectations and global database scalings. This
and electromagnetic effects. Consequently, such simulation discrepancy is an outstanding issue.
codes are now being used to model and interpret experimental The critical importance of the edge H-mode pedestal
discharges, as well as to benchmark transport models which to overall plasma performance is increasingly recognized,
contain simplified physics content. With regard to turbulence and H-mode access is of course essential for ITER. Theory
regulation and suppression, the critical role of E × B sheared and modelling can now reproduce many aspects of the L–H
flows and zonal flows, as well as the effect of magnetic shear, transition but have not yet produced a reliable quantitative
is now generally accepted. Significant progress has been prediction for the power threshold. Consequently, global
made in quantitative comparisons of turbulence measurements scaling techniques employing fits to experimental multi-
of fluctuation amplitude, spectra and correlation lengths machine data are still employed to project the L–H transition
to theory and simulation predictions, showing reasonable threshold power. Using an improved and expanded database
agreement, and essential features of the turbulence theory the latest projection for ITER is a threshold power in
such as zonal flows have been identified in multiple devices. the range ∼40–50 MW, within the capability of the ITER
However, multiple outstanding issues still remain with regard heating system (70 MW). Once in H-mode, global fusion
to fundamental transport understanding, such as obtaining a performance is strongly influenced by the height of the edge
successful physics description of electron thermal, particle and temperature and density pedestals; with relatively stiff central

S117
E.J. Doyle et al

profiles the pedestal parameters have a large impact on global the beta scaling of transport, β 0 versus β −0.9 , respectively. At
performance. In predicting pedestal parameters, the most the ITER design operating point of βN = 1.6 this different
important unresolved issue is determining the pedestal width, scaling has little effect on confinement projections, but this
i.e. the width of the edge transport barrier. Multiple models would become important for advanced operation at higher
for the width, which have divergent predictions for ITER, have beta, with the non-dimensional result being more optimistic.
been tested versus data from present devices, but this testing This discrepancy is not currently understood and is under
has failed to discriminate between the models. It is hoped that active investigation. Other areas of current research with
further inter-machine comparisons will clarify this issue. A regard to scaling projections include the effect of shaping
successful, quantitative theoretical model has been developed on confinement and whether collisionality or proximity to
for the pressure limit in Type I ELMs, in terms of the stability the Greenwald density is more important in determining
of peeling–ballooning modes. However, a full understanding confinement at high density. With regard to projections by
of the ELM crash dynamics requires further clarification of the transport models, a comparison of four leading models
the non-linear evolution of the MHD instabilities. Several (MM, IFS/PPPL, Weiland and GLF23) shows ITER achieving
successful methods have been demonstrated to either mitigate Q = 10 operation for edge pedestal temperatures in the
or eliminate Type I ELMs, in order to avoid unacceptable range 3.5–5 keV, depending on the model. In general, the
erosion of the first wall material (ELMs generate transient predictions of the transport models have converged since
pulsed heat and particle fluxes to the wall and divertor). the publication of the IPB but still differ in detail due to
These mitigated and no-ELM regimes need to be extended varying levels of ‘stiffness’ in the models. Also, these models
to reactor-relevant conditions. Also with regard to reactor are more accurate in replicating ion thermal as compared
compatibility, high confinement at high density has now been with electron thermal, particle or momentum transport, all
robustly demonstrated, as required for the ITER operating of which are critical to ITER performance, so caution is still
point. However, extension of these results to lower, ITER-like required in using these projections. In further extending the
collisionality is desired. applicability of these models the current focus is on developing
Predictive capabilities for transport and confinement and integrating models of the edge pedestal with the core
projections to ITER have improved. Multi-machine plasma.
experimental scalar and profile databases for constructing With regard to operation of ITER in regimes with
global confinement scalings and local transport modelling have enhanced core confinement, such as with ITBs, none of the
been expanded and are now administered under the auspices predictive techniques are as yet in a position to make reliable
of the ITPA. The three primary predictive techniques—global projections. For the global scaling approach the limitation may
scaling, transport modelling and non-dimensional scaling— be intrinsic, in that the development and sustainment of ITBs
have each been refined and extended in scope since the depends on local plasma parameters (i.e. on detailed plasma
publication of the ITB. Empirical scaling, using parameter profiles), which are not captured in scalar databases. For the
fits to global experimental transport data from present transport models, while progress has been made in replicating
devices, was previously the primary predictive approach, ITB formation and sustainment, further work is required before
but it is now extensively complemented by theory-based projections can be made with confidence.
transport modelling. The capability and reliability of In an overall summary, it may be stated that both
transport models has been substantially increased, and the experimental and transport modelling/simulation indicate that
models are benchmarked via comparison to both present ITER will meet its baseline design confinement requirements.
experiments and to more physically comprehensive transport Advanced operation on ITER with enhanced core confinement
simulation codes. The non-dimensional scaling technique, is becoming an increasingly realistic and attractive prospect,
in which plasma parameters are matched in dimensionless but a major experimental emphasis is required to demonstrate
fashion across devices, leaving only a single parameter that such an advanced operation is compatible with reactor
extrapolation to be made to ITER, e.g. in ρ∗ , has also operating conditions. Substantial advances have been made
been refined, especially with regard to the beta scaling of in improving the physics content and reliability of transport
transport. modelling and simulation codes, but a fully consistent and
For the ITER base case or reference scenario (conventional integrated (core and edge) predictive capability which can
ELMy H-mode operation), all three predictive techniques accurately describe all transport channels is still some way
show that ITER will have sufficient confinement to meet its in the future.
design target of Q = 10 operation, within similar ranges of
uncertainties. Specifically, while several new global scalings Reference
have been developed, the recommended scaling for ITER
remains the IPB98(y,2) scaling law [2]. For the ITER design [1] ITER Physics Basis Editors et al 1999 Nucl. Fusion
operating point this scaling predicts τth = 3.6 s. However, 39 2137
[2] ITER Physics Expert Groups on Confinement and
the estimated 95% log-linear uncertainty interval for the ITER Transport and Confinement Modelling and Database,
projection has been significantly reduced; using a revised, ITER Physics Basis Editors and ITER EDA 1999 Nucl.
expanded database the interval is now (+14%/ − 13%), as Fusion 39 2175
compared with (+25%/ − 20%) for the earlier ITER design [3] Itoh K. 2003 Nucl. Fusion 43 1710
[1]. The non-dimensional scaling projection for ITER is very [4] Diamond P.H. et al 2004 Fusion Energy 2004: Proc. 20th Int.
Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
similar, with a minimum τth = 3.3 s. However, the non- OV/2-1 and http://www-naweb.iaea.org/napc/physics/
dimensional and global scaling results differ with regard to fec/fec2004/datasets/index.html

S118
 Chapter 2: Plasma confinement and transport

[5] Shaing K.C. et al 2003 Nucl. Fusion 43 258 [59] Lashmore-Davies C.N., McCarthy D.R. and Thyagaraja A.
[6] Joffrin E. et al 2003 Nucl. Fusion 43 1167 2001 Phys. Plasmas 8 5121
[7] Connor J.W. et al 2005 Plasma Phys. Control. Fusion 47 941 [60] Smolyakov A.I., Diamond P.H. and Malkov M. 2000 Phys.
[8] Terry P.W. 2000 Rev. Mod. Phys. 72 109 Rev. Lett. 84 491
[9] Connor J.W. and Pogutse O.P. 2001 Plasma Phys. Control. [61] Champeaux S. and Diamond P.H. 2001 Phys. Lett. 288 214
Fusion 43 155 [62] Holland C. and Diamond P.H. 2002 Phys. Plasmas 9 3857
[10] Horton W. et al 2003 New J. Phys. 5 14.1 [63] Holland C. et al 2003 Nucl. Fusion 43 761
[11] Parail V.V. 2002 Plasma Phys. Control. Fusion 44 A63 [64] Li J. and Kishimoto Y. 2002 Phys. Plasmas 9 1241
[12] Horton W. et al 2000 Phys. Plasmas 7 1494 [65] Manfredi G., Roach C.M. and Dendy R.O. 2001 Plasma
[13] Garbet X. et al 2003 Nucl. Fusion 43 975 Phys. Control. Fusion 43 825
[14] Baker D.R. et al 2001 Phys. Plasmas 8 1565 [66] Diamond P.H. et al 2005 Plasma Phys. Control. Fusion
[15] Kaw P.K. 1982 Phys. Rev. A 90A 290 47 R35
[16] Li J. and Kishimoto Y. 2002 Plasma Phys. Control. Fusion [67] Hahm T.S. et al 1999 Phys. Plasmas 6 922
44 A479 [68] Winsor N., Johnson J.L. and Dawson J.M. 1968 Phys. Fluids
[17] Itoh K., Itoh S.-I. and Fukuyama A. 1999 Transport and 11 2448
Structural Formation in Plasmas (Bristol: Institute of [69] Miyato N., Kishimoto Y. and Li J. 2004 Phys. Plasmas
Physics Publishing) 11 5557
[18] Hahm T.S., Lee W.W. and Brizard A. 1988 Phys. Fluids [70] Hallatschek K. and Biskamp D. 2001 Phys. Rev. Lett. 86 1223
31 1940 [71] Scott B. 2003 Phys. Lett. A 320 53
[19] Xu X.Q. and Cohen R.H. 1998 Contrib. Plasma Phys. 38 158 [72] Krommes J.A. 2002 Phys. Rep. 360 1
[20] Waltz R.E. et al 1997 Phys. Plasmas 4 2482 [73] Yagi M. et al 2002 Fusion Energy 2002: Proc. 19th Int. Conf.
[21] Dorland W. and Hammett G.W. 1993 Phys. Fluids B 5 812 (Lyon, 2002) (Vienna: IAEA) CD-ROM file TH/1-4 and
[22] Doyle E.J. et al 2000 Fusion Energy 2000: Proc. 18th Int. http://www.iaea.org/programmes/ripc/physics/fec2002/
Conf. (Sorrento, 2000) (Vienna: IAEA) CD-ROM file html/fec2002.htm
EX6/2 and http://www.iaea.org/programmes/ripc/ [74] Idomura Y., Tokuda S. and Kishimoto Y. 2002 Fusion Energy
physics/fec2000/html/fec2000.htm 2002: Proc. 19th Int. Conf. (Lyon, 2002) (Vienna: IAEA)
[23] Garbet X. et al 2004 Plasma Phys. Control. Fusion 46 1351 CD-ROM file TH/P1-08 and http://www.iaea.org/
[24] Wolf R.C. 2003 Plasma Phys. Control. Fusion 45 R1 programmes/ripc/physics/fec2002/html/fec2002.htm
[25] Connor J.W. et al 2004 Nucl. Fusion 44 R1 [75] Jenko F. et al 2002 Fusion Energy 2002: Proc. 19th Int. Conf.
[26] Biglari H., Diamond P.H. and Terry P.W. 1990 Phys. Fluid B (Lyon, 2002) (Vienna: IAEA) CD-ROM file TH/1-2 and
21 http://www.iaea.org/programmes/ripc/physics/
[27] Hahm T.S. and Burrell K.H. 1995 Phys. Plasmas 2 1648 fec2002/html/fec2002.htm
[28] Horton W. 1999 Rev. Mod. Phys. 71 735 [76] Roach C.M. et al 2005 Plasma Phys. Control. Fusion
[29] Ottaviani M. and Manfredi G. 1999 Phys. Plasmas 6 3267 47 B323
[30] Labit B. and Ottaviani M. 2003 Phys. Plasmas 10 126 [77] Thyagaraja A. 2000 Plasma Phys. Control. Fusion
[31] Kishimoto Y. et al 1996 Phys. Plasmas 3 1289 42 B255
[32] Candy J. and Waltz R.E. 2003 Phys. Rev. Lett. 91 045001 [78] McClements K.G. et al 2002 Nucl. Fusion 42 1155
[33] Lin Z. et al 2002 Phys. Rev. Lett. 88 195004 [79] deBaar M.R. et al 2005 Phys. Rev. Lett. 94 035002
[34] Waltz R.E., Candy J.M. and Rosenbluth M.N. 2002 Phys. [80] Itoh S.-I. et al 2002 Phys. Plasmas 9 1947
Plasmas 9 1938 [81] Itoh S.-I. and Itoh K. 2001 Plasma Phys. Control. Fusion
[35] Kim E. et al 2003 Nucl. Fusion 43 961 43 1055
[36] Hahm T.S. et al 2004 Plasma Phys. Control. Fusion 46 A323 [82] Estrada-Mila C., Candy J. and Waltz R.E. 2005 Phys.
[37] Stallard B.W. et al 1999 Phys. Plasmas 6 1978 Plasmas 12 022305
[38] Roach C.M., Connor J.W. and Janjuat S. 1995 Plasma Phys. [83] Hammett G.W. and Perkins F.W. 1990 Phys. Rev. Lett.
Control. Fusion 37 679 64 3019
[39] Krommes J.A. and Kim C.B. 2000 Phys. Rev. E 62 8508 [84] Hammett G.W., Dorland W. and Perkins F.W. 1992 Phys.
[40] Itoh S.-I. et al 2002 Plasma Phys. Control. Fusion 44 1311 Fluids B 4 2052
[41] Diamond P.H. and Hahm T.S. 1995 Phys. Plasmas 2 3640 [85] Waltz R.E., Dominguez R.R. and Hammett G.W. 1992 Phys.
[42] Lin Z. et al 1999 Phys. Rev. Lett. 83 3645 Fluids B 4 3138
[43] Malkov M.A., Diamond P.H. and Rosenbluth M.N. 2001 [86] Rosenbluth M.N. and Hinton F.L. 1998 Phys. Rev. Lett.
Phys. Plasmas 8 5073 80 724
[44] Li J.Q. and Kishimoto Y. 2002 Phys. Rev. Lett. 89 115002 [87] Dastgeer S., Mahajan S. and Weiland J. 2002 Phys. Plasmas
[45] Diamond P.H. et al 2001 Nucl. Fusion 41 1067 9 4911
[46] Dorland W. et al 2000 Phys. Rev. Lett. 85 5579 [88] Sugama H., Watanabe T.H. and Horton W. 2003 Phys.
[47] Jenko F. and Dorland W. 2002 Phys. Rev. Lett. 89 225001 Plasmas 10 726
[48] Idomura Y. et al 2001 Nucl. Fusion 41 437 [89] Snyder P.B. and Hammett G.W. 2001 Phys. Plasmas 8 3199
[49] Kishimoto Y. et al 2002 Fusion Energy 2002: Proc. 19th Int. [90] Beer M.A. and Hammett G.W. 1996 Phys. Plasmas 3 4046
Conf. (Lyon, 2002) (Vienna: IAEA) CD-ROM file TH/1-5 [91] Scott B.D. 2005 Phys. Plasmas 12 102307
and http://www.iaea.org/programmes/ripc/physics/ [92] Idomura Y., Tokuda S. and Kishimoto Y. 2003 Nucl. Fusion
fec2002/html/fec2002.htm 43 234
[50] Rogers B.N., Dorland W. and Kotschenreuther M. 2000 Phys. [93] Villard L. et al 2004 Nucl. Fusion 44 172
Rev. Lett. 85 5336 [94] Lin Z. and Chen L. 2001 Phys. Plasmas 8 1447
[51] Kim E. and Diamond P.H. 2003 Phys. Rev. Lett. 90 185006 [95] Cohen B.I. et al 2002 Phys. Plasmas 9 1915
[52] Kim E. and Diamond P.H. 2003 Phys. Plasmas 10 1698 [96] Cohen B.I. et al 2002 Phys. Plasmas 9 251
[53] Staebler G.M. 2000 Phys. Rev. Lett. 84 3610 [97] Chen Y. and Parker S.E. 2003 J. Comput. Phys. 189 463
[54] D’Ippolito D.A. and Myra J.R. 2003 Phys. Plasmas 10 4029 [98] Chen Y. et al 2003 Nucl. Fusion 43 1121
[55] Yu G.Q. and Krasheninnikov S.I. 2003 Phys. Plasmas [99] Allfrey S.J. and Hatzky R. 2003 Comput. Phys. Commun.
10 4413 154 98
[56] Chen L., Lin Z. and White R. 2000 Phys. Plasmas 7 3129 [100] Aydemir A.Y. 1994 Phys. Plasmas 1 822
[57] Guzdar P.N., Kleva R.G. and Chen L. 2002 Phys. Plasmas 8 [101] Denton R.E. and Kotschenreuther M. 1995 J. Comput. Phys.
459 119 283
[58] Guzdar P.N. et al 2002 Phys. Plasmas 8 3907 [102] Kotschenreuther M. 1988 Bull. Am. Phys. Soc. 33 2107

S119
E.J. Doyle et al

[103] Dimits A.M. and Lee W.W. 1993 J. Comput. Phys. 107 309 [161] Rhodes T.L. et al 2004 Fusion Energy 2004: Proc. 20th Int.
[104] Parker S.E. and Lee W.W. 1993 Phys. Fluids B 5 77 Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
[105] Manuilskiy I. and Lee W.W. 2000 Phys. Plasmas 7 1381 EX/P6-23 and http://www-naweb.iaea.org/napc/physics/
[106] Lee W.W. et al 2001 Phys. Plasmas 8 4435 fec/fec2004/datasets/index.html
[107] Hatzky R. et al 2002 Phys. Plasmas 9 898 [162] Burrell K.H. 1999 Phys. Plasmas 6 4418
[108] Nevins W.M. et al 2005 Phys. Plasmas 12 122305 [163] Boedo J.A. et al 2002 Nucl. Fusion 42 117
[109] Candy J. and Waltz R.E. 2003 J. Comput. Phys. 186 545 [164] Ware A.S. et al 1998 Phys. Plasmas 5 173
[110] Dimits A.M. et al 2000 Phys. Plasmas 7 969 [165] Zhang Y.Z. and Mahajan S.M. 1992 Phys. Fluids B 4 1385
[111] Lin Z. et al 1998 Science 281 1835 [166] Budny R.V. et al 2002 Plasma Phys. Control. Fusion 44 1215
[112] Hammett G.W. et al 1993 Plasma Phys. Control. Fusion [167] Conway G.D. et al 2000 Phys. Rev. Lett. 84 1463
35 973 [168] Greenfield C.M. et al 1997 Phys. Plasmas 4 1596
[113] Beer M.A. 1995 Gyrofluid models of turbulent transport in [169] Greenfield C.M. et al 1999 Nucl. Fusion 39 1723
tokamaks, PhD Thesis, Princeton University [170] McKee G.R. et al 2000 Phys. Plasmas 7 1870
[114] Waltz R.E., Kerbel G.D. and Milovich J. 1994 Phys. Plasmas [171] Murakami M. et al 2001 Nucl. Fusion 41 317
1 2229 [172] Jackson G.L. et al 2002 Plasma Phys. Control. Fusion
[115] Dimits A.M. et al 1996 Phys. Rev. Lett. 77 71 44 1893
[116] Sydora R.D., Decyk V.K. and Dawson J.M. 1996 Plasma [173] Jakubowski M., Fonck R.J. and McKee G.R. 2002 Phys. Rev.
Phys. Control. Fusion 38 A281 Lett. 89 265003
[117] Lin Z. et al 2000 Phys. Plasmas 7 1857 [174] McKee G.R. et al 2003 Phys. Plasmas 10 1712
[118] Candy J., Waltz R.E. and Dorland W. 2004 Phys. Plasmas [175] Xu X.Q. et al 2002 New J. Phys. 4 53.1
11 L25 [176] Mazurenko A. et al 2002 Phys. Rev. Lett. 89 225004
[119] Jenko F., Dannert T. and Angioni C. 2005 Plasma Phys. [177] Terry J.L. et al 2003 Phys. Plasmas 10 1739
Control. Fusion 47 B195 [178] Hinton F.L. and Hazeltine R.D. 1976 Rev. Mod. Phys. 48 239
[120] Beer M.A. and Hammett G.W. 1996 Phys. Plasmas 3 4018 [179] Hirshman S.P. and Sigmar D.J. 1981 Nucl. Fusion 21 1079
[121] Ross D.W. and Dorland W. 2002 Phys. Plasmas 9 5031 [180] Shaing K.C., Hsu C.T. and Hazeltine R.D. 1994 Phys.
[122] Dannert T. and Jenko F. 2005 Phys. Plasmas 12 072309 Plasmas 1 3365
[123] Ernst D.R. et al 2004 Phys. Plasmas 11 2637 [181] Urano H. et al 2005 Phys. Rev. Lett. 95 035003
[124] Garbet X. et al 2003 Phys. Rev. Lett. 91 035001 [182] Gates D. and NSTX National Research Team 2003 Phys.
[125] Yankov V.V. 1995 Plasma Phys. Rep. 21 719 Plasmas 10 1659
[126] Isichenko M.B. et al 1996 Phys. Plasmas 3 1916 [183] Wade M.R., Murakami M. and Politzer P.A. 2004 Phys. Rev.
[127] Garbet X. et al 2002 Phys. Plasmas 9 3893 Lett. 92 235005
[128] Hallatschek K. and Dorland W. 2005 Phys. Rev. Lett. [184] Thomas D.M. et al 2004 Phys. Rev. Lett. 93 065003
95 055002 [185] Houlberg W.A. et al 1997 Phys. Plasmas 4 3230
[129] Angioni C. et al 2005 Phys. Plasmas 12 040701 [186] Sauter O., Angioni C. and Lin-Liu Y.R. 1999 Phys. Plasmas
[130] Snyder P.B. and Hammett G.W. 2001 Phys. Plasmas 8 744 6 2834
[131] Parker S.E. et al 2004 Phys. Plasmas 11 2594 [187] Sauter O., Angioni C. and Lin-Liu Y.R. 2002 Phys. Plasmas
[132] Candy J. 2005 Phys. Plasmas 2005 072307 9 5140
[133] Jenko F. and Dorland W. 2001 Plasma Phys. Control. Fusion [188] Shaing K.C., Hazeltine R.D. and Zarnstorff M.C. 1997 Phys.
43 A141 Plasmas 4 1375
[134] Applegate D.J. et al 2004 Phys. Plasmas 12 5085 [189] Fujita T. et al 2001 Phys. Rev. Lett. 87 245001
[135] Jenko F. et al 2000 Phys. Plasmas 7 1904 [190] Hawkes N.C. et al 2001 Phys. Rev. Lett. 87 115001
[136] Beyer P. et al 2000 Phys. Rev. Lett. 85 4892 [191] Baylor L.R. et al 2004 Phys. Plasmas 11 3100
[137] Lin Z., Chen L. and Zonca F. 2005 Phys. Plasmas 12 056125 [192] Severo J.H.F. et al 2004 Phys. Plasmas 11 846
[138] Idomura Y., Wakatani M. and Tokuda S. 2000 Phys. Plasmas [193] Rozhansky V. 2004 Plasma Phys. Control. Fusion 46 A1
7 3551 [194] Rozhansky V. et al 2002 Nucl. Fusion 42 1110
[139] Li J.Q. and Kishimoto Y. 2003 Phys. Plasmas 10 683 [195] Solomon W.M. et al 2004 Fusion Energy 2004: Proc. 20th
[140] Jenko F. 2004 J. Plasma Fusion Res. Series 6 11 Int. Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM
[141] Waltz R.E. and Miller R.L. 1999 Phys. Plasmas 6 4265 file EX/P4-10 and http://www-naweb.iaea.org/
[142] Scott B.D. 2002 New J. Phys. 4 52.1 napc/physics/fec/fec2004/datasets/index.html
[143] Carreras B.A. 1997 IEEE Trans. Plasma Sci. 25 1281 [196] Bell R.E. et al 1998 Phys. Rev. Lett. 81 1429
[144] Conway G.D. et al 2002 Plasma Phys. Control. Fusion [197] Shaing K.C. et al 1999 Phys. Rev. Lett. 83 3840
44 451 [198] Shaing K.C. 1992 Phys. Fluids B 4 290
[145] Scott B. 1998 Plasma Phys. Control. Fusion 40 823 [199] Gormezano C. et al 2007 Progress in the ITER Physics Basis
[146] Rettig C.L. et al 2001 Phys. Plasmas 8 2232 Nucl. Fusion 47 S285–S336
[147] Kotschenreuther M., Rewoldt G. and Tang W.M. 1995 [200] Doyle E.J. et al 2002 Nucl. Fusion 42 333
Comput. Phys. Commun. 88 128 [201] Lao L.L. et al 1999 Bull. Am. Phys. Soc. 44 77
[148] Rhodes T.L. et al 2002 Phys. Plasmas 9 2142 [202] Litaudon X. et al 2004 Plasma Phys. Control. Fusion
[149] McKee G.R. et al 2001 Nucl. Fusion 41 1235 46 A19
[150] Ross D.W. et al 2002 Phys. Plasmas 9 177 [203] Sips A.C.C. et al 2004 Fusion Energy 2004: Proc. 20th Int.
[151] McKee G.R. 2004 private communication, General Atomics Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
[152] Xu X.Q. et al 1999 J. Nucl. Mater. 266–269 993 IT/P3-36 and http://www-naweb.iaea.org/napc/physics/
[153] Vershkov V.A. et al 2002 Fusion Energy 2002: Proc. 19th fec/fec2004/datasets/index.html
Int. Conf. (Lyon, 2002) (Vienna: IAEA) CD-ROM file [204] Politzer P.A. et al 2004 Fusion Energy 2004: Proc. 20th Int.
EX/P3-04 and http://www.iaea.org/programmes/ripc/ Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
physics/fec2002/html/fec2002.htm EX/P2-7 and http://www-naweb.iaea.org/napc/physics/
[154] Weiland J. and Nordman H. 1991 Nucl. Fusion 31 390 fec/fec2004/datasets/index.html
[155] Ossipenko M.V. and Tsaun S.V. 2001 Plasma Phys. Rep. 27 1 [205] Dux R. et al 2004 Nucl. Fusion 44 260
[156] Evensen H.T. et al 1998 Nucl. Fusion 38 237 [206] Stober J. et al 2003 Nucl. Fusion 43 1265
[157] Colas L. et al 1998 Nucl. Fusion 38 903 [207] Doyle E.J. et al 2002 Fusion Energy 2002: Proc. 19th Int.
[158] Entrop I. et al 2000 Phys. Rev. Lett. 84 3606 Conf. (Lyon, 2002) (Vienna: IAEA) CD-ROM file
[159] Wong K.-L. et al 1997 Phys. Lett. 236 339 EX/C3-2 and http://www.iaea.org/programmes/
[160] Itoh K. et al 1994 Plasma Phys. Control. Fusion 36 279 ripc/physics/fec2002/html/fec2002.htm

S120
 Chapter 2: Plasma confinement and transport

[208] Takenaga H. et al 2003 Nucl. Fusion 43 1235 [259] Koide Y. 1997 Phys. Plasmas 4 1623
[209] Synakowski E.J. et al 1999 Nucl. Fusion 39 1733 [260] Doyle E.J. et al 2000 Plasma Phys. Control. Fusion 42 A237
[210] Sakamoto Y. et al 2004 Nucl. Fusion 44 876 [261] Burrell K.H. and DIII-D Team 2003 Nucl. Fusion 43 1555
[211] Synakowski E.J. et al 1997 Phys. Rev. Lett. 78 2972 [262] Kinsey J.E. et al 2005 Nucl. Fusion 45 450
[212] Sakamoto Y. et al 2001 Nucl. Fusion 41 865 [263] Fukuda T. and International ITB Database Working Group
[213] Tokar M.Z. et al 2000 Phys. Rev. Lett. 84 895 2001 Europhysics Conference Abstracts: Proc. 28th EPS
[214] McKee G. et al 2000 Phys. Rev. Lett. 84 1922 Conf. on Controlled Fusion and Plasma Physics (Madeira,
[215] Tokar M.Z. et al 2002 Plasma Phys. Control. Fusion 44 1903 2001) (Geneva: European Physical Society) vol 25A
[216] Synakowski E.J. 1998 Plasma Phys. Control. Fusion 40 581 (ECA) p 1349
[217] Turnbull A.D. et al 1995 Phys. Rev. Lett. 74 718 [264] Wade M.R. et al 2003 Nucl. Fusion 43 634
[218] Burrell K.H. 1997 Phys. Plasmas 4 1499 [265] Luce T.C. et al 2003 Nucl. Fusion 43 321
[219] Ida K. et al 2002 Phys. Rev. Lett. 88 015002 [266] DeBoo J.C. et al 1999 Nucl. Fusion 39 1935
[220] Petty C.C. et al 2002 Phys. Plasmas 9 128 [267] Kinsey J.E., Waltz R.E. and DeBoo J.C. 1999 Phys. Plasmas
[221] Balet B. et al 1995 Europhysics Conference Abstracts: Proc. 6 1865
22nd EPS Conf. on Controlled Fusion and Plasma Physics [268] Gentle K.W. et al 1995 Phys. Rev. Lett. 74 3620
(Bournemouth, 1995) (Geneva: European Physical [269] Kissick M.W. et al 1996 Nucl. Fusion 36 1691
Society) vol 19C part I (ECA) p 9 [270] Ryter F. et al 2000 Nucl. Fusion 40 1917
[222] Cordey J.G. et al 1996 Plasma Phys. Control. Fusion 38 A67 [271] Galli P. et al 1999 Nucl. Fusion 39 1355
[223] Perkins F.W. et al 1993 Phys. Fluids B 5 477 [272] Kubo H. et al 2001 Nucl. Fusion 41 227
[224] Scott S.D. et al 1993 Plasma Physics and Controlled Nuclear [273] Stober J. et al 2000 Plasma Phys. Control. Fusion 42 A211
Fusion Research 1992: Proc. 14th Int. Conf. (Würzburg, [274] Suttrop W. et al 1997 Plasma Phys. Control. Fusion 39 2051
1992) (Vienna: IAEA) vol 3 p 427 [275] Tardini G. et al 2002 Nucl. Fusion 42 258
[225] Petty C.C. et al 1998 Nucl. Fusion 38 1183 [276] Baker D.R. et al 2001 Phys. Plasmas 8 4128
[226] Christiansen J.P. et al 1997 Europhysics Conference [277] Mikkelsen D.R. et al 2003 Nucl. Fusion 43 30
Abstracts: Proc. 24th EPS Conf. on Controlled Fusion and [278] Nordman H., Weiland J. and Jarmen A. 1990 Nucl. Fusion 30
Plasma Physics (Berchtesgaden, 1997) (Geneva: European 983
Physical Society) vol 21A part I (ECA) p 89 [279] Mukhovatov V. et al 2003 Nucl. Fusion 43 942
[227] Petty C.C. and Luce T.C. 1999 Phys. Plasmas 6 909 [280] Onjun T. et al 2001 Phys. Plasmas 8 975
[228] JET Team (presented by J.G. Cordey) 1997 Fusion Energy [281] Garbet X. and Waltz R.E. 1996 Phys. Plasmas 3 1898
1996: Proc. 16th Int. Conf. (Montreal, 1996) (Vienna: [282] Neu R. et al 2000 Fusion Energy 2000: Proc. 18th Int. Conf.
IAEA) vol 1 p 603 (Sorrento, 2000) (Vienna: IAEA) CD-ROM file EXP5/33
[229] Greenwald M. et al 1998 Plasma Phys. Control. Fusion and http://www.iaea.org/programmes/ripc/physics/
40 789 fec2000/html/fec2000.htm
[230] Shirai H. et al 2000 Plasma Phys. Control. Fusion 42 1193 [283] Mantica P. et al 2002 Plasma Phys. Control. Fusion 44 2185
[231] Petty C.C. et al 1998 Phys. Plasmas 5 1695 [284] Synakowski E.J. et al 2003 Nucl. Fusion 43 1653
[232] Petty C.C. et al 2000 Plasma Phys. Control. Fusion 42 B75 [285] Goldston R.J. et al 1987 Plasma Physics and Controlled
[233] Petty C.C. et al 1999 Phys. Rev. Lett. 83 3661 Nuclear Fusion Research: Proc. 11th Int. Conf. (Kyoto,
[234] ITER 1D Modelling Working Group: et al 2000 Nucl. Fusion 1986) (Vienna: IAEA) vol 1 p 75
40 1955 [286] Alikaev V.V. et al 1987 Plasma Physics and Controlled
[235] Mikkelsen D.R. et al 1999 Fusion Energy 1998: Proc. 17th Nuclear Fusion Research 1986: Proc. 11th Int. Conf.
Int. Conf. (Yokohama, 1998) (Vienna: IAEA) CD-ROM (Kyoto, 1986) (Vienna: IAEA) vol 1 p 111
file IAEA-F1-CN-69/ITERP1/08 and www.iaea.org/ [287] Wagner F. et al 1986 Phys. Rev. Lett. 56 2187
programmes/ripc/physics/start.htm [288] Callen J.D. et al 1987 Nucl. Fusion 27 1857
[236] Kinsey J.E. et al 2003 Nucl. Fusion 43 1845 [289] Taylor G. et al 1989 Nucl. Fusion 29 3
[237] Kotschenreuther M. et al 1995 Phys. Plasmas 2 2381 [290] Luce T.C., Petty C.C. and de Haas J.C.M. 1992 Phys. Rev.
[238] Bateman G. et al 1998 Phys. Plasmas 5 1793 Lett. 68 52
[239] Kinsey J.E. and Bateman G. 1996 Phys. Plasmas 3 3344 [291] Petty C.C. and Luce T.C. 1994 Nucl. Fusion 34 121
[240] Dimits A.M. et al 2000 Fusion Energy 2000: Proc. 18th Int. [292] Gohil P., Burrell K.H. and Osborne T.H. 1998 Nucl. Fusion
Conf. (Sorrento, 2000) (Vienna: IAEA) CD-ROM file 38 425
THP1/03 and http://www.iaea.org/programmes/ [293] Horton L.D. et al 1999 Plasma Phys. Control. Fusion
ripc/physics/fec2000/html/fec2000.htm 41 B329
[241] Kinsey J.E., Waltz R.E. and Candy J. 2005 Phys. Plasmas [294] Urano H. et al 2002 Nucl. Fusion 42 76
12 062302 [295] Watkins M.L. 2002 Plasma Phys. Control. Fusion
[242] Greenfield C.M. et al 2001 Phys. Rev. Lett. 86 4544 44 B173
[243] Strait E.J. et al 1995 Phys. Rev. Lett. 75 4421 [296] Coppi B. 1980 Comm. on Plasma Phys. Control. Fusion 5
[244] Levinton F.M. et al 1995 Phys. Rev. Lett. 75 4417 261
[245] Synakowski E.J. et al 1997 Phys. Plasmas 4 1736 [297] Ryter F. et al 2001 Plasma Phys. Control. Fusion 43 A323
[246] Challis C.D. et al 2001 Plasma Phys. Control. Fusion 43 861 [298] Ryter F. et al 2001 Phys. Rev. Lett. 86 2325
[247] Gormezano C. et al 1998 Phys. Rev. Lett. 80 5544 [299] Ryter F. et al 2001 Phys. Rev. Lett. 86 5498
[248] Gruber O. et al 2001 Nucl. Fusion 41 1369 [300] Hoang G.T. et al 2001 Phys. Rev. Lett. 87 125001
[249] Wolf R.C. et al 2003 Plasma Phys. Control. Fusion 45 1757 [301] Suttrop W. et al 2001 Europhysics Conference Abstracts:
[250] Koide Y. et al 1994 Phys. Rev. Lett. 72 3662 Proc. 28th EPS Conf. on Controlled Fusion and Plasma
[251] Shirai H. et al 1999 Nucl. Fusion 39 1713 Physics (Madeira, 2001) (Geneva: European Physical
[252] Beer M. et al 1997 Phys. Plasmas 4 1792 Society) vol 25A (ECA) p 989
[253] Kinsey J.E., Staebler G.M. and Waltz R.E. 2002 Phys. [302] Jacchia A. et al 2002 Nucl. Fusion 42 1116
Plasmas 9 1676 [303] Cirant S. et al 2003 Nucl. Fusion 43 1384
[254] Bourdelle C. et al 2003 Phys. Plasmas 10 2881 [304] Ryter F.R. et al 2005 Phys. Rev. Lett. 95 085001
[255] Gormezano C. 1999 Plasma Phys. Control. Fusion 41 B367 [305] Waltz R.E., DeBoo J.C. and Osborne T.H. 1992 Nucl. Fusion
[256] Challis C.D. et al 2002 Plasma Phys. Control. Fusion 32 1051
44 1031 [306] Jenko F. 2000 Phys. Plasmas 7 514
[257] Fujita T. et al 1997 Phys. Rev. Lett. 78 2377 [307] Jenko F., Dorland W. and Hammett G.W. 2001 Phys. Plasmas
[258] Litaudon X. et al 1996 Plasma Phys. Control. Fusion 38 1603 8 4096

S121
E.J. Doyle et al

[308] Peeters A.G. et al 2005 Phys. Plasmas 12 022505 [352] Zastrow K.-D. et al 2004 Plasma Phys. Control. Fusion
[309] Baker D.R. et al 2003 Phys. Plasmas 10 4419 46 B255
[310] Imbeaux F., Ryter F. and Garbet X. 2001 Plasma Phys. [353] Stober J. et al 2002 Plasma Phys. Control. Fusion 44 A159
Control. Fusion 43 1503 [354] Baker D.R. et al 2000 Nucl. Fusion 40 1003
[311] Leuterer F. et al 2003 Nucl. Fusion 43 1329 [355] Valovic M. et al 2002 Plasma Phys. Control. Fusion 44 1911
[312] Ryter F. et al 2003 Nucl. Fusion 43 1396 [356] Garzotti L. et al 2003 Nucl. Fusion 43 1829
[313] DeBoo J.C. et al 2005 Nucl. Fusion 45 494 [357] Hoang G.T. et al 2003 Phys. Rev. Lett. 90 155002
[314] Pochelon A. et al 2004 Fusion Energy 2004: Proc. 20th Int. [358] Zabolotsky A., Weisen H. and TCV Team 2003 Plasma Phys.
Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file Control. Fusion 45 735
EX/9-1 and http://www-naweb.iaea.org/napc/physics/ [359] Angioni C. et al 2003 Phys. Rev. Lett. 90 205003
fec/fec2004/datasets/index.html [360] Angioni C. et al 2003 Phys. Plasmas 10 3225
[315] Lopes Cardozo N.J. 1995 Plasma Phys. Control. Fusion [361] Weisen H. et al 2005 Nucl. Fusion 45 L1
37 799 [362] Becker G. 2004 Nucl. Fusion 44 933
[316] Callen J.D. and Kissick M.W. 1997 Plasma Phys. Control. [363] Baker D.R. and Rosenbluth M.N. 1998 Phys. Plasmas 5 2936
Fusion 39 B173 [364] Lorate A. et al 2007 Progress in the ITER Physics Basis 47
[317] Lopes Cardozo N.J. et al 1990 Plasma Phys. Control. Fusion S203–S263
32 983 [365] Mahdavi M.A. et al 2002 Nucl. Fusion 42 52
[318] Jacchia A. et al 2003 Europhysics Conference Abstracts: [366] Takenaga H. et al 2004 Fusion Energy 2004: Proc. 20th Int.
Proc. 30th EPS Conf. on Controlled Fusion and Plasma Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
Physics (St Petersburg, 2003) (Geneva: European Physical EX/6-1 and http://www-naweb.iaea.org/napc/physics/
Society) vol 27A (ECA) paper P-1.117 fec/fec2004/datasets/index.html
[319] Romanelli M., Bourdelle C. and Dorland W. 2004 Phys. [367] Lang P.T. et al 2000 Nucl. Fusion 40 245
Plasmas 11 3845 [368] Messiaen A. et al 1996 Phys. Rev. Lett. 77 2487
[320] Kinsey J.E. et al 2002 Fusion Energy 2002: Proc. 19th Int. [369] Jackson G.L. et al 2002 Nucl. Fusion 42 28
Conf. (Lyon, 2002) (Vienna: IAEA) CD-ROM file [370] Ongena J. et al 2004 Nucl. Fusion 44 124
TH/P1-09 and http://www.iaea.org/programmes/ [371] Tokar M.Z. 2003 Plasma Phys. Control. Fusion 45 1323
ripc/physics/fec2002/html/fec2002.htm [372] Polevoi A.R. et al 2002 J. Plasma Fusion Res. Ser. 5 82
[321] Manini A. et al 2004 Plasma Phys. Control. Fusion 46 1723 [373] Pereverzev G.V. et al 2005 Nucl. Fusion 45 221
[322] Gentle K.W., Austin M.E. and Phillips P.E. 2003 Phys. Rev. [374] Yankov V.V. 1994 JETP Lett. 60 171
Lett. 91 255001 [375] Isichenko M.B., Gruzinov A.V. and Diamond P.H. 1996
[323] Barbato E. 2001 Plasma Phys. Control. Fusion 43 A287 Phys. Rev. Lett. 74 4436
[324] Günter S. et al 2000 Phys. Rev. Lett. 84 3097 [376] Weisen H., Furno I. and TCV Team 2001 Nucl. Fusion
[325] Wolf R.C. et al 2001 Nucl. Fusion 41 1259 41 1227
[326] Leuterer F. et al 2003 Nucl. Fusion 43 744 [377] Gohil P. et al 2003 Plasma Phys. Control. Fusion 45 601
[327] Pericoli Ridolfini V. et al 2003 Nucl. Fusion 43 469 [378] Rice J.E. et al 2002 Nucl. Fusion 42 510
[328] Hogeweij G.M.D. et al 2002 Plasma Phys. Control. Fusion [379] Ossipenko M.V. and T-10 team 2003 Nucl. Fusion 43 1641
44 1155 [380] Chen H. et al 2001 Plasma Phys. Control. Fusion 43 1
[329] Mailloux J. et al 2002 Phys. Plasmas 9 2156 [381] Baker D.R. et al 2000 Fusion Energy 2000: Proc. 18th Int.
[330] Baranov Y.F. et al 2004 Plasma Phys. Control. Fusion Conf. (Sorrento, 2000) (Vienna: IAEA) CD-ROM file
46 1181 EXP5/03 and http://www.iaea.org/programmes/
[331] Ide S. et al 2000 Nucl. Fusion 40 445 ripc/physics/fec2000/html/fec2000.htm
[332] Ide S. et al 2004 Nucl. Fusion 44 87 [382] Dux R. et al 1999 Nucl. Fusion 39 1509
[333] Fujita T. et al 2004 Plasma Phys. Control. Fusion 46 A35 [383] Wade M.R., Houlberg W.A. and Baylor L.R. 2000 Phys. Rev.
[334] Hoang G.T. et al 2000 Nucl. Fusion 40 913 Lett. 84 282
[335] Litaudon X. et al 2001 Plasma Phys. Control. Fusion 43 677 [384] Houlberg W.A. and Baylor L.R. 1998 Fusion Technol. 34 591
[336] Pietrzyk Z.A. et al 2000 Phys. Plasmas 7 2909 [385] Efthimion P.C. et al 1999 Nucl. Fusion 39 1905
[337] Sauter O. et al 2001 Phys. Plasmas 8 2199 [386] Takenaga H. and JT-60 Team 2001 Phys. Plasmas 8 2217
[338] Henderson M.A. et al 2003 Phys. Plasmas 10 1796 [387] Chen H. et al 2001 Nucl. Fusion 41 31
[339] Henderson M.A. et al 2004 Plasma Phys. Control. Fusion [388] West W.P. et al 2002 Phys. Plasmas 9 1970
46 A275 [389] Fujita T. and JT-60 Team 2003 Nucl. Fusion 43 1527
[340] Razumova K.A. et al 2000 Plasma Phys. Control. Fusion [390] Wade M.R. et al 2001 J. Nucl. Mater. 290–293 773
42 973 [391] Puiatti M.E. et al 2002 Plasma Phys. Control. Fusion 44 1863
[341] Sauter O. et al 2005 Phys. Rev. Lett. 94 105002 [392] Scott S.D. et al 1990 Phys. Rev. Lett. 64 531
[342] Conway G.D. et al 2002 Plasma Phys. Control. Fusion [393] Nishijima D. et al 2005 Plasma Phys. Control. Fusion 47 89
44 1167 [394] Staebler G.M., Waltz R.E. and Wiley J.C. 1997 Nucl. Fusion
[343] Fourment C. et al 2003 Plasma Phys. Control. Fusion 45 233 37 287
[344] Ide S. et al 2002 Plasma Phys. Control. Fusion 44 A137 [395] Staebler G.M., Kinsey J.E. and Waltz R.E. 2004 Plasma
[345] Takenaga H. et al 1999 Nucl. Fusion 39 1917 Phys. Control. Fusion 46 A265
[346] Takenaga H., Mahdavi M.A. and Baker D.R. 2001 Phys. [396] Parail V.V. et al 1999 Nucl. Fusion 39 429
Plasmas 8 1607 [397] Tala T.J.J. et al 2001 Plasma Phys. Control. Fusion 43 507
[347] Baker D.R. et al 1998 Nucl. Fusion 38 485 [398] Rice J.E. et al 2001 Nucl. Fusion 41 277
[348] Efthimion P.C. et al 1998 Phys. Plasmas 5 1832 [399] Meyer H. et al 2004 Plasma Phys. Control. Fusion 46 A291
[349] JET Team (prepared by K.-D. Zastrow) 1999 Nucl. Fusion [400] Hutchinson I.H. 2000 Phys. Rev. Lett. 84 3330
39 1891 [401] Rice J.E. et al 1998 Nucl. Fusion 38 75
[350] Stork D. et al 2004 Fusion Energy 2004: Proc. 20th Int. [402] Rice J.E. et al 1999 Nucl. Fusion 39 1175
Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file [403] Field A.R. et al 2004 Plasma Phys. Control. Fusion 46 981
OV/4-1 and http://www-naweb.iaea.org/napc/physics/ [404] Field A.R. et al 2003 Europhysics Conference Abstracts:
fec/fec2004/datasets/index.html Proc. 30th EPS Conf. on Controlled Fusion and Plasma
[351] McDonald D.C. et al 2004 Fusion Energy 2004: Proc. 20th Physics (St Petersburg, 2003) (Geneva: European Physical
Int. Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM Society) vol 27A (ECA) paper P-2.148
file EX/6-6 and http://www-naweb.iaea.org/napc/physics/ [405] Eriksson L.-G., Righi E. and Zastrow K.-D. 1997 Plasma
fec/fec2004/datasets/index.html Phys. Control. Fusion 39 27

S122
 Chapter 2: Plasma confinement and transport

[406] Noterdaeme J.-M. et al 2003 Nucl. Fusion 43 274 [455] Litaudon X. et al 2002 Plasma Phys. Control. Fusion 44 1057
[407] Assas S. et al 2003 Europhysics Conference Abstracts: Proc. [456] Kamada Y. et al 1999 Nucl. Fusion 39 1845
30th EPS Conf. on Controlled Fusion and Plasma Physics [457] Isayama A. et al 2001 Nucl. Fusion 41 761
(St Petersburg, 2003) (Geneva: European Physical [458] Rice B.W. et al 1999 Nucl. Fusion 39 1855
Society) vol 27A (ECA) paper P-1.138 [459] Wade M.R. et al 2001 Phys. Plasmas 8 2208
[408] Eriksson L.-G., Hoang G.T. and Bergeaud V. 2001 Nucl. [460] Luce T.C. et al 2001 Nucl. Fusion 41 1585
Fusion 41 91 [461] Waltz R.E., Dewar R.L. and Garbet X. 1998 Phys. Plasmas
[409] DeGrassie J. et al 2004 Phys. Plasmas 11 4323 5 1784
[410] Claassen H.A. et al 2000 Phys. Plasmas 7 3699 [462] Greenfield C.M. et al 2004 Plasma Phys. Control. Fusion
[411] Rogister A.L. 1999 Phys. Plasmas 6 200 46 B213
[412] Fülöp T., Helander P. and Catto P.J. 2002 Phys. Rev. Lett. [463] Murakami M. et al 2004 Proc. 20th Int. Conf. on Fusion
89 225003 Energy 2004: (Vilamoura, 2004) (Vienna: IAEA)
[413] Helander P., Fülöp T. and Catto P.J. 2003 Phys. Plasmas CD-ROM file EX/1-2 and http://www-naweb.iaea.org/
10 4396 napc/physics/fec/fec2004/datasets/index.html
[414] Rozhansky V. et al 2005 J. Nucl. Mater. 337–339 291 [464] JET Team (prepared by Gormezano C.) 1999 Nucl. Fusion
[415] Coppi B. 2002 Nucl. Fusion 42 1 39 1875
[416] Shaing K.C. 2001 Phys. Rev. Lett. 86 640 [465] JET Team (prepared by Söldner F.X.) 1999 Nucl. Fusion
[417] Chang C.S. et al 1999 Phys. Plasmas 6 1969 39 1883
[418] Perkins F.W. et al 2001 Phys. Plasmas 8 2181 [466] Burrell K.H. et al 2001 Phys. Plasmas 8 2153
[419] Chan V.S., Chiu S.C. and Omelchenko Y.A. 2002 Phys. [467] Burrell K.H. et al 2005 Phys. Plasmas 12 056121
Plasmas 9 501 [468] Sips A.C.C. et al 2002 Plasma Phys. Control. Fusion 44 B69
[420] Eriksson L.-G. and Porcelli F. 2002 Nucl. Fusion 42 959 [469] Wade M.R. et al 2004 Proc.20th Int. Conf. on Fusion Energy
[421] Rice J.E. et al 2004 Nucl. Fusion 44 379 2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
[422] Lee W.D. et al 2003 Phys. Rev. Lett. 91 205003 EX/4-1 and http://www-naweb.iaea.org/napc/physics/
[423] LaBombard B. et al 2004 Nucl. Fusion 44 1047 fec/fec2004/datasets/index.html
[424] Rice J.E. et al 2005 Nucl. Fusion 45 251 [470] Luce T.C. et al 2004 Phys. Plasmas 11 2627
[425] Rice J.E. et al 2003 Nucl. Fusion 43 781 [471] Gruber O. et al 1999 Phys. Rev. Lett. 83 1787
[426] Connor J.W. 1988 Plasma Phys. Control. Fusion 30 619 [472] Peeters A.G. et al 2002 Nucl. Fusion 42 1376
[427] Waltz R.E., DeBoo J.C. and Rosenbluth M.N. 1990 Phys. [473] Sips A.C.C. et al 2002 Plasma Phys. Control. Fusion
Rev. Lett. 65 2390 44 A151
[428] Buckingham E. 1914 Phys. Rev. 4 345 [474] Na Y.-S. et al 2002 Plasma Phys. Control. Fusion 44 1285
[429] Luce T.C. et al 2002 Nucl. Fusion 42 1193 [475] Staebler A. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
[430] Petty C.C. et al 2004 Plasma Phys. Control. Fusion 46 A207 2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
[431] Greenwald M. 2002 Plasma Phys. Control. Fusion 44 R27 EX/4-5 and http://www-naweb.iaea.org/napc/physics/
[432] Petty C.C. et al 2004 Phys. Plasmas 11 2514 fec/fec2004/datasets/index.html
[433] McDonald D.C. et al 2004 Plasma Phys. Control. Fusion [476] Joffrin E. et al 2002 Plasma Phys. Control. Fusion 44 1203
46 A215 [477] Sips A.C.C. et al 2003 Europhysics Conference Abstracts:
[434] Cordey J.G. et al 2005 Nucl. Fusion 45 1078 Proc. 30th EPS Conf. on Controlled Fusion and Plasma
[435] Batha S.H. et al 1997 Europhysics Conference Abstracts: Physics (St Petersburg, 2003) (Geneva: European Physical
Proc. 24th EPS Conf. on Controlled Fusion and Plasma Society) vol 27A (ECA) paper O-1.3A
Physics (Berchtesgaden, 1997) (Geneva: European [478] Joffrin E.H. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
Physical Society) vol 21A part III (ECA) p 1057 2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
[436] Cordey J.G. et al 2000 Plasma Phys. Control. Fusion EX/4-2 and http://www-naweb.iaea.org/napc/physics/
42 A127 fec/fec2004/datasets/index.html
[437] Luce T.C., Petty C.C. and Kinsey J.E. 2001 Europhysics [479] Isayama A. et al 2003 Nucl. Fusion 43 1272
Conference Abstracts: Proc. 28th EPS Conf. on Controlled [480] Ide S. and JT-60 Team 2004 Proc. 20th Int. Conf. on Fusion
Fusion and Plasma Physics (Madeira, 2001) (Geneva: Energy 2004: (Vilamoura, 2004) (Vienna: IAEA)
European Physical Society) vol 25A (ECA) p 1377 CD-ROM file OV/1-1 and http://www-naweb.iaea.org/
[438] Luce T.C., Petty C.C. and Cordey J.G. 2007 Application of napc/physics/fec/fec2004/datasets/index.html
dimensionless parameter scaling techniques to the design [481] Isayama A. and JT-60 Team 2005 Phys. Plasmas 12 056117
and interpretation of magnetic fusion experiments Plasma [482] Suzuki T. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
Phys. Control. Fusion submitted 2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
[439] Bourdelle C. et al 2005 Nucl. Fusion 45 110 EX/1-3 and http://www-naweb.iaea.org/napc/physics/
[440] Ide S. et al 2002 Plasma Phys. Control. Fusion 44 L63 fec/fec2004/datasets/index.html
[441] Joffrin E. et al 2002 Plasma Phys. Control. Fusion 44 1739 [483] Guo S.C. and Romanelli F. 1993 Phys. Fluids B 5 520
[442] Wolf R.C. et al 1999 Plasma Phys. Control.Fusion 41 B93 [484] Crisanti F. et al 2004 Fusion Energy 2004: Proc. 20th Int.
[443] Quigley E.D. et al 2004 Nucl. Fusion 44 1189 Conf. (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
[444] Fujita T. et al 1999 Nucl. Fusion 39 1627 EX/P2-1 and http://www-naweb.iaea.org/napc/physics/
[445] Huysmans G.T.A. et al 2001 Phys. Rev. Lett. 87 245002 fec/fec2004/datasets/index.html
[446] Breslau J.A., Jardin S.C. and Park W. 2003 Phys. Plasmas [485] Gormezano C. et al 2004 Plasma Phys. Control. Fusion
10 1665 46 B435
[447] Fujita T. et al 2004 Proc. 20th Int. Conf. on Fusion Energy [486] Sips A.C.C. et al 2002 Plasma Phys. Control. Fusion
2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file 44 A391
EX/P4-3 and http://www-naweb.iaea.org/napc/physics/ [487] Tuccillo A.A. et al 2004 Proc. 20th Int. Conf. on Fusion
fec/fec2004/datasets/index.html Energy 2004: (Vilamoura, 2004) (Vienna: IAEA)
[448] Fujita T. 2002 Plasma Phys. Control. Fusion 44 A19 CD-ROM file EX/1-1 and http://www-naweb.iaea.org/
[449] Fujita T. et al 2001 Phys. Rev. Lett. 87 085001 napc/physics/fec/fec2004/datasets/index.html
[450] Murakami M. et al 2003 Phys. Rev. Lett. 90 255001 [488] Fujita T. and JT-60 Team 1997 Plasma Phys. Control. Fusion
[451] Joffrin E. et al 2002 Nucl. Fusion 42 235 39 B75
[452] Sips A.C.C. et al 1998 Plasma Phys. Control.Fusion 40 1171 [489] Kamada Y. and JT-60 Team 2001 Nucl. Fusion 41 1311
[453] Keilhacker M. et al 1999 Nucl. Fusion 39 209 [490] Suttrop W. 2000 Plasma Phys. Control. Fusion 42 A1
[454] Crisanti F. et al 2002 Phys. Rev. Lett. 88 145004 [491] Hubbard A.E. 2000 Plasma Phys. Control. Fusion 42 A15

S123
E.J. Doyle et al

[492] Horton L.D. 2000 Plasma Phys. Control. Fusion 42 A37 [530] Mossessian D.A. et al 2003 Phys. Plasmas 10 689
[493] Lao L.L. 2000 Plasma Phys. Control. Fusion 42 A51 [531] Mossessian D.A. et al 2003 Europhysics Conference
[494] Saibene G. et al 1999 Nucl. Fusion 39 1133 Abstracts: Proc. 30th EPS Conf. on Controlled Fusion and
[495] Takizuka T. 1998 Plasma Phys. Control. Fusion 40 851 Plasma Physics (St Petersburg, 2003) (Geneva: European
[496] Kamada Y. et al 1997 Proc. 16th Int. Conf. on Fusion Energy Physical Society) vol 27A (ECA) paper P-3.182
1996: (Montreal, 1996) (Vienna: IAEA) vol 1 p 247 [532] Fenstermacher M.E. et al 2005 Nucl. Fusion 45 1493
[497] Saibene G. et al 2002 Plasma Phys. Control. Fusion 44 1769 [533] Kinsey J.E., Staebler G.M. and Waltz R.E. 2005 Phys.
[498] Snyder P.B. and Wilson H.R. 2003 Plasma Phys. Control. Plasmas 12 052503
Fusion 45 1671 [534] Hatae T. et al 2001 Nucl. Fusion 41 285
[499] Urano H. et al 2002 Plasma Phys. Control. Fusion 44 11 [535] Shaing K.C. and Crume E.C. 1989 Phys. Rev. Lett. 63 2369
[500] Sartori R. et al 2004 Proc. 20th Int. Conf. on Fusion Energy [536] Itoh S.-I. et al 1994 Phys. Rev. Lett. 72 1200
2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file [537] Hinton F.L. and Staebler G.M. 1993 Phys. Fluids B 5 1281
EX/6-3 and http://www-naweb.iaea.org/napc/physics/ [538] Groebner R.J. et al 2002 Phys. Plasmas 9 2134
fec/fec2004/datasets/index.html [539] Hatae T. et al 2000 Plasma Phys. Control. Fusion 42 A283
[501] Kamada Y. et al 2002 Plasma Phys. Control. Fusion 44 A279 [540] Sugihara M. et al 2003 Plasma Phys. Control. Fusion 45 L55
[502] Lang P.T. et al 2002 Plasma Phys. Control. Fusion 44 1919 [541] Sugihara M. et al 2000 Nucl. Fusion 40 1743
[503] Sartori R. et al 2002 Plasma Phys. Control. Fusion [542] Guzdar P.N. et al 2001 Phys. Rev. Lett. 87 015001
44 1801 [543] Guzdar P.N. et al 2002 Phys. Rev. Lett. 89 265004
[504] Borrass K., Schneider R. and Farengo R. 1997 Nucl. Fusion [544] Hubbard A.E. et al 2004 Plasma Phys. Control. Fusion
37 523 46 A95
[505] Borrass K., Lingertat J. and Schneider R. 1998 Contrib. [545] Fielding S.J. et al 2001 Nucl. Fusion 41 909
Plasma Phys. 38 130 [546] Ido T. et al 2000 Plasma Phys. Control. Fusion 42 A309
[506] Coster D.P. et al 1999 J. Nucl. Mater. 266–269 804 [547] Miura Y. et al 2001 Nucl. Fusion 41 973
[507] Reiter D. 1992 J. Nucl. Mater. 196–198 80 [548] Hubbard A.E. et al 1998 Plasma Phys. Control. Fusion
[508] Schneider R. et al 1992 J. Nucl. Mater. 196–198 810 40 689
[509] Borrass K. et al 2001 Europhysics Conference Abstracts: [549] Hubbard A.E. et al 2002 Plasma Phys. Control. Fusion
Proc. 28th EPS Conf. on Controlled Fusion and Plasma 44 A359
Physics (Madeira, 2001) (Geneva: European Physical [550] Righi E. et al 2000 Plasma Phys. Control. Fusion 42 A199
Society) vol 25A (ECA) p 501 [551] Andrew Y. et al 2004 Plasma Phys. Control. Fusion 46 337
[510] Borrass K. et al 2004 Nucl. Fusion 44 752 [552] Carlstrom T.N. et al 2002 Plasma Phys. Control. Fusion
[511] Chankin A.V., Itami K. and Asakura N. 2002 Plasma Phys. 44 A333
Control. Fusion 44 A399 [553] Andrew Y. et al 2004 Plasma Phys. Control. Fusion 46 A87
[512] Maingi R. et al 1999 J. Nucl. Mater. 266–269 598 [554] Martin Y.R. and TCV Team 2002 Plasma Phys. Control.
[513] Mahdavi M.A. et al 2003 Phys. Plasmas 10 3984 Fusion 44 A143
[514] Kukushkin A.S. et al 2001 Europhysics Conference [555] Akers R.J. et al 2003 Plasma Phys. Control. Fusion 45 A175
Abstracts: Proc. 28th EPS Conf. on Controlled Fusion and [556] Bush C.E. et al 2003 Phys. Plasmas 10 1755
Plasma Physics (Madeira, 2001) (Geneva: European [557] Martin Y.R. and TCV team 2004 Plasma Phys. Control.
Physical Society) vol 25A (ECA) p 2113 Fusion 46 A77
[515] Kukushkin A.S. and Pacher H.D. 2002 Plasma Phys. Control. [558] Ryter F. et al 2002 Plasma Phys. Control. Fusion
Fusion 44 931 44 A407
[516] Pacher G.W. et al 2004 Plasma Phys. Control. Fusion [559] Ryter F. and H-mode Threshold Database Group 2002
46 A257 Plasma Phys. Control. Fusion 44 A415
[517] LaBombard B. et al 2002 Proc. 19th Int. Conf. on Fusion [560] ITPA H-mode Power Threshold Database Working Group
Energy 2002: (Lyon, 2002) (Vienna: IAEA) CD-ROM file (presented by T. Takizuka) 2004 Plasma Phys. Control.
EX/D2-1 and http://www.iaea.org/programmes/ Fusion 46 A227
ripc/physics/fec2002/html/fec2002.htm [561] Connor J.W. and Wilson H.R. 2000 Plasma Phys. Control.
[518] Myra J.R. et al 2002 Europhysics Conference Abstracts: Fusion 42 R1
Proc. 29th EPS Conf. on Controlled Fusion and Plasma [562] Ossipenko M.V. and Tsaun S.V. 2000 Plasma Phys. Rep.
Physics (Montreux, 2002) (Geneva: European Physical 26 465
Society) vol 26B (ECA) paper O-3.23 [563] Groebner R.J. et al 2002 Proc. 19th Int. Conf. on Fusion
[519] Rudakov D.L. et al 2002 Plasma Phys. Control. Fusion Energy 2002: (Lyon, 2002) (Vienna: IAEA) CD-ROM file
44 717 EX/C2-3 and http://www.iaea.org/programmes/ripc/
[520] Rogers B.N. and Drake J.F. 1998 Phys. Rev. Lett. 81 4396 physics/fec2002/html/fec2002.htm
[521] Russell D.A. et al 2004 Phys. Rev. Lett. 93 265001 [564] Xu X.Q. et al 2000 Phys. Plasmas 7 1951
[522] Tokar M.Z. 2003 Phys. Rev. Lett. 91 095001 [565] Connor J.W. and Pogutse O.P. 2001 Plasma Phys. Control.
[523] Kotschenreuther M. et al 1997 Proc. 16th Int. Conf. on Fusion 43 281
Fusion Energy 1996: (Montreal, 1996) (Vienna: IAEA) [566] Janeschitz G. et al 2002 Plasma Phys. Control. Fusion
vol 2 p 371 44 A459
[524] Waltz R.E. et al 1997 Proc. 16th Int. Conf. on Fusion Energy [567] Parail V. et al 2002 Proc. 19th Int. Conf. on Fusion Energy
1996: (Montreal, 1996) (Vienna: IAEA) vol 2 p 385 2002: (Lyon, 2002) (Vienna: IAEA) CD-ROM file
[525] Kinsey J.E., Waltz R.E. and Schissel D.P. 1997 Europhysics TH/P3-08 and http://www.iaea.org/programmes/
Conference Abstracts: Proc. 24th EPS Conf. on Controlled ripc/physics/fec2002/html/fec2002.htm
Fusion and Plasma Physics (Berchtesgaden, 1997) [568] Bateman G., Onjun T. and Kritz A.H. 2003 Plasma Phys.
(Geneva: European Physical Society) vol 21A part III Control. Fusion 45 1939
(ECA) p 1081 [569] Rognlien T.D. et al 2004 Contrib. Plasma Phys. 44 188
[526] Hatae T. et al 1998 Plasma Phys. Control. Fusion 40 1073 [570] Xu X.Q. et al 2004 Contrib. Plasma Phys. 44 105
[527] Hughes J.W. et al 2002 Phys. Plasmas 9 3019 [571] Kukushkin A.S. et al 2003 Nucl. Fusion 43 716
[528] Groebner R.J. and Osborne T.H. 1998 Phys. Plasmas 5 1800 [572] Pacher H.D. et al 2003 J. Nucl. Mater. 313-316 657
[529] Osborne T.H. et al 2002 Proc. 19th Int. Conf. on Fusion [573] Pacher G.W. et al 2001 Europhysics Conference Abstracts:
Energy 2002: (Lyon, 2002) (Vienna: IAEA) CD-ROM file Proc. 28th EPS Conf. on Controlled Fusion and Plasma
CT/3 and http://www.iaea.org/programmes/ripc/physics/ Physics (Madiera, 2001) (Geneva: European Physical
fec2002/html/fec2002.htm Society) vol 25A (ECA) p 625

S124
 Chapter 2: Plasma confinement and transport

[574] Pacher G.W. et al 2003 Nucl. Fusion 43 188 [615] Hubbard A.E. et al 1998 Phys. Plasmas 5 1744
[575] Pacher G.W. et al 2003 Europhysics Conference Abstracts: [616] Sartori R. et al 2004 Plasma Phys. Control. Fusion 46 723
Proc. 30th EPS Conf. on Controlled Fusion and Plasma [617] Suttrop W. et al 2002 Proc. 19th Int. Conf. on Fusion Energy
Physics (St Petersburg, 2003) (Geneva: European Physical 2002: (Lyon, 2002) (Vienna: IAEA) CD-ROM file
Society) vol 27A (ECA) paper P-3.139 EX/P5-07 and http://www.iaea.org/programmes/
[576] Parail V.V. et al 2003 Plasma Phys. Rep. 29 539 ripc/physics/fec2002/html/fec2002.htm
[577] Sartori R. et al 1998 Plasma Phys. Control. Fusion 40 757 [618] Igitkhanov Y. et al 2000 Contrib. Plasma Phys. 40 368
[578] Sartori R. et al 1999 Europhysics Conference Abstracts: [619] Chankin A.V. and Saibene G. 1999 Plasma Phys. Control.
Proc. 26th EPS Conf. on Controlled Fusion and Plasma Fusion 41 913
Physics (Maastricht, 1999) (Geneva: European Physical [620] Rapp J. et al 2004 Nucl. Fusion 44 312
Society) vol 23J (ECA) p 197 [621] Takase Y. et al 1997 Proc. 16th Int. Conf. on Fusion Energy
[579] Leonard A.W. et al 2001 J. Nucl. Mater. 290-293 1097 1996: (Montreal, 1996) (Vienna: IAEA) vol 1 p 475
[580] Loarte A. et al 2002 Plasma Phys. Control. Fusion 44 1815 [622] Greenwald M. et al 1997 Nucl. Fusion 37 793
[581] Oyama N. et al 2001 Plasma Phys. Control. Fusion 43 717 [623] Takase Y. et al 1997 Phys. Plasmas 4 1647
[582] Oyama N. et al 2004 Nucl. Fusion 44 582 [624] Greenwald M. et al 1999 Phys. Plasmas 6 1943
[583] Nunes I. et al 2003 Europhysics Conference Abstracts: Proc. [625] Hubbard A.E. et al 2001 Phys. Plasmas 8 2033
30th EPS Conf. on Controlled Fusion and Plasma Physics [626] Lin Y. et al 1999 Rev. Sci. Instrum. 70 1078
(St Petersburg, 2003) (Geneva: European Physical [627] Greenwald M. et al 2000 Plasma Phys. Control. Fusion
Society) vol 27A (ECA) paper P-1.157 42 A263
[584] Oyama N. et al 2003 Nucl. Fusion 43 1250 [628] Marmar E.S. et al 2000 Proc. 18th Int. Conf. on Fusion
[585] Kirk A. et al 2004 Plasma Phys. Control. Fusion 46 551 Energy 2000: (Sorrento, 2000) (Vienna: IAEA) CD-ROM
[586] Connor J.W. 1998 Plasma Phys. Control. Fusion 40 531 file EX2/5 and http://www.iaea.org/programmes/
[587] Kass T. et al 1998 Nucl. Fusion 38 111 ripc/physics/fec2000/html/fec2000.htm
[588] Zohm H. et al 1995 Nucl. Fusion 35 543 [629] Snipes J.A. et al 2001 Plasma Phys. Control. Fusion 43 L23
[589] Perez C.P. et al 2004 Nucl. Fusion 44 609 [630] Rogers B. and Drake J.F. 1999 Phys. Plasmas 6 2797
[590] Perez C.P. et al 2004 Plasma Phys. Control. Fusion 46 61 [631] Mossessian D.A. et al 2002 Plasma Phys. Control. Fusion
[591] Loarte A. et al 2004 Phys. Plasmas 11 2668 44 423
[592] Loarte A. et al 2003 Plasma Phys. Control. Fusion 45 1549 [632] Mossessian D. et al 2000 Plasma Phys. Control. Fusion
[593] Leonard A.W. et al 2002 Plasma Phys. Control.Fusion 44 42 A255
945 [633] Sunn Pedersen T. et al 2000 Nucl. Fusion 40 1795
[594] Counsell G.F. et al 2003 Nucl. Fusion 43 1197 [634] Mossessian D.A. et al 2003 Phys. Plasmas 10 1720
[595] Maingi R. et al 2005 J. Nucl. Mater. 337–339 727 [635] Maddison G. et al 2002 Plasma Phys. Control. Fusion
[596] Saibene G. et al 2002 Transport in fusion plasmas: transport 44 1937
near operational limits Proc. 9th EU-US Transport Task [636] Kamiya K. et al 2003 Nucl. Fusion 43 1214
Force Workshop (Córdoba, 2002) [637] Kamiya K. et al 2004 Plasma Phys. Control. Fusion 46 1745
http://www-fusion.ciemat.es/ ttf2002/ [638] Kamiya K. et al 2004 Plasma Phys. Control. Fusion 46 A157
[597] Urano H. et al 2003 Plasma Phys. Control. Fusion 45 1571 [639] Nagashima N. et al 2004 Plasma Phys. Control. Fusion
[598] Lao L.L. et al 1999 Nucl. Fusion 39 1785 46 A381
[599] Ferron J.R. et al 2000 Nucl. Fusion 40 1411 [640] Hubbard A.E. et al 2006 Plasma Phys. Control. Fusion A121
[600] Saibene G. et al 2003 Europhysics Conference Abstracts: [641] Fiore C.L. et al 2004 Phys. Plasmas 11 2480
Proc. 30th EPS Conf. on Controlled Fusion and Plasma [642] Fiore C.L. et al 2001 Phys. Plasmas 8 2023
Physics (St Petersburg, 2003) (European Physical Society) [643] Wukitch S.J. et al 2002 Phys. Plasmas 9 2149
vol 27A (ECA) paper P-1.192 [644] Burrell K.H. et al 1999 Bull. Am. Phys. Soc. 44 127
[601] Federici G., Loarte A. and Strohmayer G. 2003 Plasma Phys. [645] Suttrop W. et al 2004 Plasma Phys. Control. Fusion 46 A151
Control. Fusion 45 1523 [646] Suttrop W. et al 2003 Plasma Phys. Control. Fusion 45 1399
[602] Fielding S.J. et al 2001 Europhysics Conference Abstracts: [647] Burrell K.H. et al 2002 Plasma Phys. Control. Fusion
Proc. 28th EPS Conf. on Controlled Fusion and Plasma 44 A253
Physics (Madeira, 2001) (Geneva: European Physical [648] Sakamoto Y. et al 2004 Plasma Phys. Control. Fusion
Society) vol 25A (ECA) p 1825 46 A299
[603] Becoulet M. et al 2003 Plasma Phys. Control. Fusion [649] Groebner R.J. et al 2001 Nucl. Fusion 41 1789
45 A93 [650] Burrell K.H. et al 2004 Plasma Phys. Control. Fusion
[604] Degeling A.W. et al 2003 Plasma Phys. Control. Fusion 46 A165
45 1637 [651] West W.P. et al 2004 Plasma Phys. Control. Fusion 46 A179
[605] Horton L. et al 2004 Plasma Phys. Control. Fusion 46 B511 [652] Ozeki T. et al 1990 Nucl. Fusion 30 1425
[606] Evans T.E. et al 2004 Phys. Rev. Lett. 92 235003 [653] Stober J. et al 2001 Nucl. Fusion 41 1123
[607] Kupschus P. et al 1997 Europhysics Conference Abstracts: [654] Koslowski H.R. et al 2003 Europhysics Conference
Proc. 24th EPS Conf. on Controlled Fusion and Plasma Abstracts: Proc. 30th EPS Conf. on Controlled Fusion and
Physics (Berchtesgaden, 1997) (Geneva: European Plasma Physics (St Petersburg, 2003) (Geneva: European
Physical Society) vol 21A, part I (ECA) p 45 Physical Society) vol 27A (ECA) paper P-1.102
[608] Lang P.T. et al 2003 Nucl. Fusion 43 1110 [655] Stober J. et al 2001 Plasma Phys. Control. Fusion 43 A39
[609] Polevoi A.R. et al 2003 Nucl. Fusion 43 1072 [656] Saarelma S. et al 2003 Nucl. Fusion 43 262
[610] Turnbull A.D. et al 2003 Plasma Phys. Control. Fusion [657] Hender T.C. et al 2007 Progress in the ITER Physics Basis
45 1845 Nucl. Fusion 47 S128–S202
[611] JET Team (prepared by A.C.C. Sips) 2001 Nucl. Fusion [658] Lönnroth J.-S. et al 2004 Plasma Phys. Control. Fusion
41 1559 46 A249
[612] Sarazin Y. et al 2002 Plasma Phys. Control. Fusion 44 2445 [659] Maingi R. et al 2004 Europhysics Conference Abstracts:
[613] Becoulet M. et al 2002 Plasma Phys. Control. Fusion Proc. 31st EPS Conf. on Plasma Physics (London, 2004)
44 A103 (Geneva: European Physical Society) vol 28G (ECA)
[614] Osborne T.H. et al 1997 Europhysics Conference Abstracts: paper P-2.189
Proc. 24th EPS Conf. on Controlled Fusion and Plasma [660] Kamada Y. et al 2000 Plasma Phys. Control. Fusion 42 A247
Physics (Berchtesgaden, 1997) (Geneva: European [661] Oyama N. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
Physical Society) vol 21A part III (ECA) p 1101 2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file

S125
E.J. Doyle et al

EX/2-1 and http://www-naweb.iaea.org/napc/physics/fec/ [697] Snipes J.A. and International H-mode Threshold Database
fec2004/datasets/index.html Working Group 2000 Plasma Phys. Control. Fusion
[662] Lao L.L. et al 2001 Nucl. Fusion 41 295 42 A299
[663] Saibene G. et al 2005 Nucl. Fusion 45 297 [698] Snipes J.A. et al 2002 Proc. 19th Int. Conf. on Fusion Energy
[664] Manickam J. 1992 Phys. Fluids B 4 1901 2002: (Lyon, 2002) (Vienna: IAEA) CD-ROM file
[665] Hegna C.C. et al 1996 Phys. Plasmas 3 584 CT/P-04 and http://www.iaea.org/programmes/
[666] Connor J.W. et al 1998 Phys. Plasmas 5 2687 ripc/physics/fec2002/html/fec2002.htm
[667] Wilson H.R. et al 2000 Nucl. Fusion 40 713 [699] Christiansen J.P. and Cordey J.G. 1998 Nucl. Fusion
[668] Huysmans G.T.A. et al 2001 Phys. Plasmas 8 4292 38 1757
[669] Snyder P.B. et al 2002 Phys. Plasmas 9 2037 [700] Valovic M. and ITER H-mode Confinement Database
[670] Mikhailovskii A.B. et al 1997 Plasma Phys. Rep. 23 844 Working Group 1998 Europhysics Conference Abstracts:
[671] Degtyarev L. et al 1997 Comput. Phys. Commun. 103 10 Proc. 1998 International Congress on Plasma Physics
[672] Medvedev S.Y. et al 2003 Europhysics Conference Abstracts: combined with the 25th EPS Conf. on Controlled Fusion
Proc. 30th EPS Conf. on Controlled Fusion and Plasma and Plasma Physics (Praha, 1998) (Geneva: European
Physics (St Petersburg, 2003) (Geneva: European Physical Physical Society) vol 22C (ECA) 675
Society) vol 27A (ECA) paper P-3.129 [701] Kardaun O.J.W.F. 1999 Plasma Phys. Control. Fusion
[673] Wilson H.R. et al 2002 Phys. Plasmas 9 1277 41 429
[674] Gruber R. et al 1981 Comput. Phys. Commun. 21 323 [702] Cordey J.G. et al 2002 Proc. 19th Int. Conf. on Fusion
[675] Gruber R. et al 1981 Comput. Phys. Commun. 24 363 Energy 2002: (Lyon, 2002) (Vienna: IAA) CD-ROM file
[676] Igitkhanov Y. et al 1998 Plasma Phys. Control. Fusion CT/P-02 and http://www.iaea.org/programmes/
40 837 ripc/physics/fec2002/html/fec2002.htm
[677] Cowley S.C. et al 2003 Plasma Phys. Control. Fusion 45 A31 [703] Kardaun O.J.W.F. 2002 Nucl. Fusion 42 841
[678] Wilson H.R. and Cowley S.C. 2004 Phys. Rev. Lett. [704] Kardaun O.J.W.F. 2002 Garching Report IPP-IR 2002/5 1.1,
92 175006 Max Planck Institut fuer Plasmaphysik
[679] Kirk A. et al 2004 Phys. Rev. Lett. 92 245002 [705] Thomsen K. and H-mode Database Working Group 2002
[680] Snyder P.B. et al 2004 Plasma Phys. Control. Fusion Fusion Eng. Des. 60 347
46 A131 [706] Thomsen K. et al 2002 Plasma Phys. Control. Fusion
[681] Snyder P.B. et al 2004 Nucl. Fusion 44 320 44 A429
[682] Connor J.W., Hastie R.J. and Taylor J.B. 1979 Proc. R. Soc. [707] Cordey J.G., ITPA H-Mode Database Working Group and
Lond., Ser. A (Math. Phys. Sci.) A365 1 ITPA Pedestal Database Working Group 2003 Nucl.
[683] Becoulet M. et al 2005 J. Nucl. Mater. 337–339 677 Fusion 43 670
[684] Lang P.T. et al 2003 Europhysics Conference Abstracts: [708] Thomsen K. et al 2004 Europhysics Conference Abstracts:
Proc. 30th EPS Conf. on Controlled Fusion and Plasma Proc. 31st EPS Conf. on Plasma Physics (London, 2004)
Physics (St Petersburg, 2003) (Geneva: European Physical (Geneva: European Physical Society) vol 28G (ECA)
Society) vol 27A (ECA) paper P-1.129 paper P-5.145
[685] Urano H. et al 2004 Plasma Phys. Control. Fusion 46 A315 [709] Cordey J.G. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
[686] Miura Y. et al 1991 Proc. 13th Int. Conf. on Plasma Physics 2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
and Controlled Nuclear Fusion Research (Washington, IT/P3-32 and http://www-naweb.iaea.org/napc/physics/
D.C., 1990) (Vienna: IAEA) vol 1 p 325 fec/fec2004/datasets/index.html
[687] Mori M. et al 1993 Proc. 14th Int. Conf. on Plasma Physics [710] Baranov Y. and International ITB Database Working Group
and Controlled Nuclear Fusion Research 1992 (Würzburg, 2001 Bull. Am. Phys. Soc. 46 117
1992) (Vienna: IAEA) vol 2 p 567 [711] Fukuda T. and JT-60 Team 2002 Plasma Phys. Control.
[688] Kaye S.M. et al 1997 Nucl. Fusion 37 1303 Fusion 44 B39
[689] Thomsen K. et al 1999 Proc. 17th Int. Conf. on Fusion [712] Hoang G.T. et al 2002 Europhysics Conference Abstracts:
Energy 1998: (Yokohama, 1998) (Vienna: IAEA) Proc. 29th EPS Conf. on Controlled Fusion and Plasma
CD-ROM file IAEA-F1-CN-69/ITERP1/07 and Physics (Montreux, 2002) (Geneva: European Physical
http://www.iaea.org/programmes/ripc/physics/ start.htm Society) vol 26B (ECA) paper P-4.068
[690] Kardaun O.J.W.F. et al 2000 Proc. 18th Int. Conf. on Fusion [713] Fujita T. et al 2003 Europhysics Conference Abstracts: Proc.
Energy 2000: (Sorrento, 2000) (Vienna: IAEA) CD-ROM 30th EPS Conf. on Controlled Fusion and Plasma Physics
file ITERP/04 and http://www.iaea.org/programmes/ (St Petersburg, 2003) (Geneva: European Physical
ripc/physics/fec2000/html/fec2000.htm Society) vol 27A (ECA) paper P-2.131
[691] Bracco G. 2001 Proc. 1st ITPA Confinement Database and [714] Gohil P. et al 2003 Nucl. Fusion 43 708
Modelling Group Meeting (Toki, 2001) [715] Imbeaux F. et al 2004 Europhysics Conference Abstracts:
http://efdasql.ipp.mpg.de/igd/PublicH-Mode/ Proc. 31st EPS Conf. on Plasma Physics (London, 2004)
ExpertMeetings/Toki2001/Toki2001.htm (Geneva: European Physical Society) vol 28G (ECA)
[692] Kardaun O.J.W.F. 2002 Proc. 3rd ITPA Confinement paper P-4.143
Database and Modelling Meeting (Cadarache, 2002) [716] Tala T.J.J. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
http://efdasql.ipp.mpg.de/igd/PublicH-Mode/ 2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
ExpertMeetings/Cadarache2002/Cadarache2002.htm TH/P2-9 and http://www-naweb.iaea.org/napc/physics/
[693] Sugiyama L.E. 2002 Cambridge Report PTP-02/02, fec/fec2004/datasets/index.html
Massachusetts Institute of Technology [717] Bracco G. and Thomsen K. 1997 Nucl. Fusion 37 759
[694] Leonov V.M. and Chudnovskiy A.N. 2003 Plasma Phys. Rep. [718] Bracco G. and Thomsen K. 1997 Abingdon Report
29 97 JET-R(97)17, JET Joint Undertaking
[695] Martin Y.R. and ITER H-mode Threshold Database Working [719] Bourdelle C. et al 2002 Nucl. Fusion 42 892
Group 1998 Europhysics Conference Abstracts: Proc. [720] Sugihara M., Takizuka T. and International H-Mode Edge
1998 International Congress on Plasma Physics combined Pedestal Expert Group 2002 Plasma Phys. Control. Fusion
with the 25th EPS Conf. on Controlled Fusion and Plasma 44 A299
Physics (Praha, 1998) (Geneva: European Physical [721] Onjun T. et al 2002 Phys. Plasmas 9 5018
Society) vol 22C (ECA) 667 [722] Kritz A.H. et al 2002 Europhysics Conference Abstracts:
[696] ITER H-mode Threshold Database Working Group Proc. 29th EPS Conf. on Controlled Fusion and Plasma
presented by E. Righi 1998 Plasma Phys. Control. Fusion Physics (Montreux, 2002) (Geneva: European Physical
40 857 Society) vol 26B (ECA) paper D-5.001

S126
 Chapter 2: Plasma confinement and transport

[723] Lloyd B. et al 2002 Proc. 19th Int. Conf. on Fusion Energy [737] Taroni A. et al 1998 Contrib. Plasma Phys. 38 37
2002: (Lyon, 2002) (Vienna: IAEA) CD-ROM file OV/2-3 [738] Pereverzev G.V. and Zolotukhin O.V. 2003 Europhysics
and http://www.iaea.org/programmes/ripc/physics/ Conference Abstracts: Proc. 30th EPS Conf. on
fec2002/html/fec2002.htm Controlled Fusion and Plasma Physics (St Petersburg,
[724] LeBlanc B.P. et al 2004 Nucl. Fusion 44 513 2003) (European Physical Society) vol 27A (ECA)
[725] Osborne T.H. et al 2000 Plasma Phys. Control. Fusion paper P-3.138
42 A175 [739] Weiland J. 2001 Europhysics Conference Abstracts: Proc.
[726] Horton L.D. et al 2002 Plasma Phys. Control. Fusion 28th EPS Conf. on Controlled Fusion and Plasma Physics
44 A273 (Madiera, 2001) (Geneva: European Physical Society) vol
[727] Takizuka T. et al 2002 Plasma Phys. Control. Fusion 25A (ECA) p 633
44 A423 [740] Belo P. et al 2004 Plasma Phys. Control. Fusion
[728] ITER Confinement Database & Modelling Group and Cordey 46 1299
J.G.1997 Plasma Phys. Control. Fusion 39 B115 [741] Kukushkin A.S. et al 1998 Contrib. Plasma Phys.
[729] Petty C.C. et al 2003 Fusion Sci. Technol. 43 1 38 20
[730] Erba M. et al 1997 Plasma Phys. Control. Fusion 39 261 [742] Fichtmüller M. et al 1998 Abingdon Report JET-P(98)69,
[731] Weiland J. 2000 Collective Modes in Inhomogeneous JET Joint Undertaking
Plasmas (Bristol: Institute of Physics Publishing) [743] Hogan J.T. and Hillis D.L. 1996 Nucl. Fusion 36 1079
[732] Zhu P. et al 2000 Phys. Plasmas 7 2898 [744] Pacher G.W. et al 2005 Nucl. Fusion 45 581
[733] Bateman G. et al 1999 Proc. 17th Int. Conf. on Fusion [745] Leonov V.M. and Zhogolev V.E. 2005 Plasma Phys. Control.
Energy 1998: (Yokohama, 1998) (Vienna: IAEA) Fusion 47 903
CD-ROM file IAEA-F1-CN-69/THP2/19 and [746] Polevoi A.R. et al 2005 Nucl. Fusion 45 1451
http://www.iaea.org/programmes/ripc/ physics/start.htm [747] Kishimoto Y. 2000 J. Plasma Fusion Res. 76 1280
[734] Lönnroth J.-S. et al 2003 Plasma Phys. Control. Fusion [748] Waltz R.E. et al 2005 Nucl. Fusion 45 741
45 1689 [749] Evans T.E. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
[735] Pankin A.Y. et al 2005 Plasma Phys. Control. Fusion 47 483 2004: (Vilamoura, 2004) (Vienna: IAEA) CD-ROM file
[736] Saibene G. et al 2001 Europhysics Conference Abstracts: EX/2-5Ra and http://www-naweb.iaea.org/napc/physics/
Proc. 28th EPS Conf. on Controlled Fusion and Plasma fec/fec2004/datasets/index.html
Physics (Madeira, 2001) (Geneva: European Physical [750] Mukhovatov V. et al 2003 Plasma Phys. Control. Fusion
Society) vol 25A (ECA) p 1749 45 A235

S127