Progress ITER Physics Basis Chapter 4 2007

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/231016089
Chapter 4: Power and particle control
Article in Nuclear Fusion · June 2007

DOI: 10.1088/0029-5515/47/6/S04
CITATIONS READS
500 315
38 authors, including:
Alberto Loarte Bruce Lipschultz

ITER The University of York
422 PUBLICATIONS 11,813 CITATIONS 456 PUBLICATIONS 8,491 CITATIONS
SEE PROFILE SEE PROFILE
A. S. Kukushkin Nobuyuki Asakura

Kurchatov Institute Japan Atomic Energy Agency
220 PUBLICATIONS 5,546 CITATIONS 310 PUBLICATIONS 5,453 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Preparation of protium deuterium pellets for injection into thermonuclear tokamak fusion plants View project
Greek Comparatives View project
All content following this page was uploaded by Bernard Pegourie on 04 June 2014.
The user has requested enhancement of the downloaded file.

IOP PUBLISHING and INTERNATIONAL ATOMIC ENERGY AGENCY NUCLEAR FUSION
Nucl. Fusion 47 (2007) S203–S263 doi:10.1088/0029-5515/47/6/S04

A. Loarte1,a , B. Lipschultz2 , A.S. Kukushkin3 , G.F. Matthews4 ,
P.C. Stangeby5 , N. Asakura6 , G.F. Counsell4 , G. Federici3 ,
A. Kallenbach7 , K. Krieger7 , A. Mahdavi8 , V. Philipps9 , D. Reiter9 ,
J. Roth7 , J. Strachan10 , D. Whyte11 , R. Doerner12 , T. Eich7 ,
W. Fundamenski4 , A. Herrmann7 , M. Fenstermacher13 , P. Ghendrih14 ,
M. Groth13 , A. Kirschner9 , S. Konoshima6 , B. LaBombard2 , P. Lang7 ,
A.W. Leonard8 , P. Monier-Garbet14 , R. Neu7 , H. Pacher15 ,
B. Pegourie14 , R.A. Pitts16 , S. Takamura17 , J. Terry2 , E. Tsitrone14 and
the ITPA Scrape-off Layer and Divertor Physics Topical Group
1
European Fusion Development Agreement Close Support Unit – Garching, Boltzmannstrasse 2,
D-85748 Garching bei München, Germany
2
Massachusetts Institute of Technology, Plasma Science and Fusion Center, 175 Albany Street,
Cambridge, MA 02139, USA
3
ITER International Team, Boltzmannstrasse 2, D-85748 Garching bei München, Germany
4
EURATOM/UKAEA Fusion Association, Culham Science Centre, Abingdon, Oxfordshire
OX14 3DB, UK
5
University of Toronto Institute for Aerospace Studies, Toronto, M3H 526 Canada
6
Japan Atomic Energy Agency, 801-1 Mukouyama, Naka, Ibaraki-ken 311-0193, Japan
7
Max-Planck-Institut für Plasmaphysik, EURATOM Association, Boltzmannstrasse 2,
D-85748 Garching bei München, Germany
8
DIII-D National Fusion Facility, San Diego, CA 92186-5608, USA
9
Institut für Plasmaphysik, Forschungszentrum Jülich, EURATOM Association, D-52425 Jülich,
Germany
10
Princeton Plasma Physics Laboratory, Princeton University, NJ 08543, USA
11
University of Wisconsin-Madison, 1500 Engineering Drive, Madison WI 53706, USA
12
University of California at San Diego, Center for Energy Research and Department of Mechanical
and Aerospace Engineering, MC 0417, La Jolla, CA 92093–0417, USA
13
Lawrence Livermore National Laboratory, Livermore, CA 94551, USA
14
Association Euratom-CEA, CEA Cadarache, F-13108 St Paul-lez-Durance, France
15
INRS-Energie, Matériaux, et Télécommunciations, Varennes, Québec J3X 1S2, Canada
16
Centre de Recherches en Physique des Plasmas, Association EURATOM–Confédération Suisse,
École Polytechnique Fédérale de Lausanne, CH-1015, Switzerland
17
Department of Energy Engineering and Science, Graduate School of Engineering, Nagoya
University, Nagoya 464-8603, Japan
E-mail: [email protected]
Received 23 August 2005, accepted for publication 2 May 2006

Published 1 June 2007
Online at stacks.iop.org/NF/47/S203
Abstract
Progress, since the ITER Physics Basis publication (ITER Physics Basis Editors et al 1999 Nucl. Fusion 39
2137–2664), in understanding the processes that will determine the properties of the plasma edge and its interaction
with material elements in ITER is described. Experimental areas where significant progress has taken place are
energy transport in the scrape-off layer (SOL) in particular of the anomalous transport scaling, particle transport in
the SOL that plays a major role in the interaction of diverted plasmas with the main-chamber material elements, edge
localized mode (ELM) energy deposition on material elements and the transport mechanism for the ELM energy from
the main plasma to the plasma facing components, the physics of plasma detachment and neutral dynamics including
the edge density profile structure and the control of plasma particle content and He removal, the erosion of low- and
high-Z materials in fusion devices, their transport to the core plasma and their migration at the plasma edge including
the formation of mixed materials, the processes determining the size and location of the retention of tritium in fusion
devices and methods to remove it and the processes determining the efficiency of the various fuelling methods as
well as their development towards the ITER requirements. This experimental progress has been accompanied by the
a Author to whom any correspondence should be addressed.

0029-5515/07/060203+61$30.00 © 2007 IAEA, Vienna Printed in the UK S203
A. Loarte et al
development of modelling tools for the physical processes at the edge plasma and plasma–materials interaction and
the further validation of these models by comparing their predictions with the new experimental results. Progress
in the modelling development and validation has been mostly concentrated in the following areas: refinement in the
predictions for ITER with plasma edge modelling codes by inclusion of detailed geometrical features of the divertor
and the introduction of physical effects, which can play a major role in determining the divertor parameters at the
divertor for ITER conditions such as hydrogen radiation transport and neutral–neutral collisions, modelling of the
ion orbits at the plasma edge, which can play a role in determining power deposition at the divertor target, models
for plasma–materials and plasma dynamics interaction during ELMs and disruptions, models for the transport of
impurities at the plasma edge to describe the core contamination by impurities and the migration of eroded materials
at the edge plasma and its associated tritium retention and models for the turbulent processes that determine the
anomalous transport of energy and particles across the SOL. The implications for the expected performance of the
reference regimes in ITER, the operation of the ITER device and the lifetime of the plasma facing materials are
discussed.
PACS numbers: 52.40.Hf, 52.55.Fa, 52.55.Rk, 52.50.Kj, 52.65.Pp

(Some figures in this article are in colour only in the electronic version)
Contents
1. Introduction
2. Experimental basis
2.1. Steady-state transport in the SOL and its implications for divertor and wall interactions
2.1.1. Steady-state divertor power deposition profiles
2.1.2. Steady-state energy transport in the near SOL
2.1.3. Particle transport in the SOL
2.1.4. SOL flow and classical drifts effects
2.1.5. Use of extrinsic impurity radiation to reduce divertor heat loads
2.2. Transient ELM transport
2.2.1. Materials limits for ELMs and disruptions energy loads on plasma facing components
2.2.2. Transport of ELM energy and particles from the confined plasma to PFCs
2.2.2.1. Temporal evolution of the ELM divertor power deposition
2.2.2.2. Spatial characteristics of the divertor target heat flux profiles during ELMs
2.2.2.3. In/out ELM energy flux asymmetries and total divertor ELM energy flux
2.2.3. Interaction of Type I ELMs with the main-chamber plasma facing components
2.2.4. Mechanism of Type I ELM energy transport to PFCs
2.2.5. ELM energy losses in present experiments and potential physics for scaling to ITER
2.2.6. Ability of radiating plasmas to buffer ELMs
2.3. Neutral processes and implications for the divertor and SOL
2.3.1. Neutral pressure control
2.3.2. Helium exhaust and noble gas impurity enrichment
2.3.3. The role of recombination in divertor processes
2.4. Material migration
2.4.1. Intrinsic impurity sources
2.4.2. Core contamination
2.4.3. Impurity migration
2.4.4. Extrapolation of present results to ITER and open issues
2.5. Divertor and main-chamber materials
2.5.1. Expected requirements of materials
2.5.2. Issues related to use of carbon containing materials
2.5.2.1. Operational experiences with carbon based devices
2.5.2.2. Erosion behaviour of carbon based wall materials
2.5.2.3. Brittle destruction
2.5.2.4. Location and properties of carbon film deposits
2.5.3. Beryllium
2.5.4. High-Z materials
2.5.4.1. Behaviour of high-Z plasma facing components in tokamak experiments
2.5.4.2. Erosion of high-Z materials in a tokamak environment
2.5.4.3. Hydrogen retention and blister/bubble-formation in tungsten
S204
2.5.5. Wall conditioning and particle control

2.5.5.1. Conditioning with permanent magnetic field
2.5.6. Implications on the use of multiple PFC materials
2.5.6.1. Formation of mixed materials
2.5.6.2. Effects of material mixing on erosion
2.5.6.3. Effects on H isotope retention
2.5.6.4. Effects on tritium removal techniques (oxidation or heating)
2.6. Tritium retention
2.6.1. Introduction
2.6.2. Database on fuel retention in present fusion devices
2.6.3. Fuel retention mechanisms in present devices and implications to ITER
2.6.4. Implication of the ITER wall material choice and mixed material effects on tritium retention
2.6.5. Dust in tokamaks (and impact on T retention)
2.6.6. Tritium removal methods
2.7. Plasma fuelling
2.7.1. Separatrix density
2.7.2. Supersonic gas jet and low field side puffing
2.7.3. Pellet fuelling
2.7.4. Compact toroid fuelling
3. Modelling and theory
3.1. Introduction
3.2. Progress in ITER divertor modelling
3.2.1. Model parameters
3.2.2. Modelling results
3.2.3. Future improvements to the physics model
3.3. Development of predictive codes
3.4. Radiation transport
3.4.1. Introduction
3.4.2. Experimental evidence for opacity
3.4.3. Theoretical predictions
3.5. Neutral particle transport
3.5.1. Introduction
3.5.2. Transition to low temperature hydrogen plasma physics
3.5.3. Sensitivity to configurational and physical details
3.5.4. Concluding outlook
3.6. Impurity modelling
3.7. Simulation of cross-field transport of the tokamak SOL
3.7.1. Intermittent transport in the SOL
3.7.2. Present modelling effort of cross-field turbulence
3.7.3. Summary of the outstanding trends
4. Summary and implications for ITER
1. Introduction The steady-state heat fluxes to divertor plates are now a

commonplace measurement. The recognition and analysis of
This chapter outlines the significant progress achieved since the effect of loosely bound layers on the measurement has been
the ITER Physics Basis (IPB) [1] in understanding basic a substantial accomplishment. The continued accumulation
scrape-off layer (SOL) and divertor processes in a tokamak. of data on the magnitude and profile of divertor power flow
The interaction of plasma with first-wall surfaces will have has allowed better cross-tokamak databases of such data
a considerable impact on the performance of fusion plasmas, to be accumulated and analysed. The resulting scalings
the lifetime of plasma facing components and the retention of (section 2.1) have improved along with the comparable
tritium in next step burning plasma experiments (BPXs) (such modelling (section 3.2).
as ITER). At the time of the IPB, it was generally thought that the
Next BPXs, such as ITER, will have larger steady power divertor was the primary area of concern for plasma facing
fluxes onto divertor and main-chamber surfaces. Together with component (PFC) erosion and, hence, lifetime. However, in
the longer pulses, this can cause a significantly larger erosion the intervening period, it has been shown that time-averaged
of these components than in present experiments both during radial transport of particles can compete with parallel transport
steady-state plasma conditions and, particularly, during edge resulting in significant main-chamber ion fluxes in current
localized modes (ELMs) and disruptions. devices, which are linked to impurity sources. Studies of
S205
A. Loarte et al
radial transport, both on macroscopic (time-averaged) and of the field) and local divertor conditions which affect the local
microscopic (turbulence) level (see section 2.1), have been erosion/redeposition balance. This is still a very active field
pursued. In general it is perceived that these steady-state of research, as quantitative extrapolation of present results to
particle fluxes are not a significant concern for first-wall ITER is very complex. Section 2.4 thus summarizes the recent
lifetime, although the database of information is still fairly and ongoing attempts to determine the migration of material
limited. In parallel there has been significant progress in the in tokamaks in a quantitative way, while section 3.6 describes
modelling of the underlying processes (see section 3.7). the improvements in the modelling of impurity transport at
It has been recognized that SOL flows play a central role the plasma edge, which has been carried out to describe the
in a number of transport processes (e.g. material migration and experimental measurements.
divertor asymmetries). In the period since the IPB the range The issue of tritium retention and maintaining high purity
of measurements has been greatly expanded poloidally (see core plasmas over a large range of experimental conditions
section 2.1). The pattern of parallel flow has been shown to be led to the current choice of plasma facing materials covering
similar across many tokamaks, with flow Mach number mag- the PFCs in ITER as described in the IPB. Since this
nitudes in the range 0.5–1 well away from the divertor target. choice was made, a number of new physics phenomena
The models of such flows (See section 3.3) have improved have been identified/quantified (e.g. main-chamber fluxes, co-
significantly, coming much closer to matching experimental deposition). This knowledge has engendered a substantial
measurements. The inclusion of drifts and poloidal variations research effort to explore the implications of these phenomena
in radial transport appear to be the key to this match. for the ITER material choice and on the device operability
Energy deposition during transient events (ELMs and dis- (see section 2.5). Significant progress has been achieved
ruptions) is an even more important threat to the lifetime of PFC in understanding the erosion rates of carbon PFCs in the
components. The magnitude, time and spatial characteristics conditions expected at the ITER divertor plasma, indicating
of the energy flux to the plasma facing components are much that it may be significantly smaller than that observed in
better characterized since the IPB (see section 2.2). The trans- present devices, with the associated decrease in the expected
port mechanisms that deliver energy from the core plasma to tritium retention. The interaction of the plasma with the
the plasma facing components during these events have been main-chamber walls in steady-state and transient conditions,
studied both experimentally (section 2.2) and through mod- which has been found to be significant in present devices,
elling (section 3.3) with models for the process appearing to be has triggered a re-evaluation of the consequences for main-
reasonably successful. In particular, some analogies between chamber PFC lifetime and operability of ITER with a Be wall,
the transient fluxes of energy to the main-chamber walls dur- which is still ongoing. The formation of mixed materials
ing ELMs and that of particles during steady phases (turbulent in ITER (C/Be/W) and the consequences for T retention and
bursts) have been identified, not foreseen in the IPB, which machine operability have been studied in a quantitative way,
are now being modelled with the same sophisticated codes with indication of both positive (Be coverage of the C target,
described in section 3.7. decreasing the retention of tritium) and negative implications
A potentially significant restriction to next step BPXs (retention of tritium in C/O Be layers, etc) for ITER operation.
operability could be caused by the long-term retention All these are described in section 2.5 and remain to be precisely
of tritium inside the vacuum vessel exceeding site limits. quantified for ITER conditions. Finally, there has been a
Summarized in section 2.6 are the results of new studies on this positive experience in the operation of divertor devices with
subject. The retention of tritium has been shown to be linked high-Z PFCs while maintaining an acceptable level of core
with co-deposition of carbon, which occurs preferentially plasma purity. In particular it has been shown that electron
at the inner divertor. The magnitude and location of the cyclotron resonant heating (ECRH) and ion cyclotron resonant
retained tritium is influenced by the mode of operation of heating (ICRH) can be used as control actuators to avoid such
the device (plasma regime, background wall temperature), accumulation under particular conditions (section 2.5).
divertor geometry characteristics and fine details of the divertor Last but not least, the control of neutral species in
target construction (gaps, shadows, etc) and plasma flows. tokamaks and the development of fuelling methods have
Tritium-cleaning techniques (section 2.6) have been developed experienced significant advances since the IPB, as described
based on photonic illumination of the tritium co-deposits, in sections 2.3 and 2.7. The geometry of the divertor has
oxidation of the co-deposits and various radio frequency (RF) proved to be a very effective tool for the control of the neutral
conditioning techniques. pressure at the divertor and the control of the hydrogenic
The problem of tritium retention is, hence, closely linked and helium content of the main plasma. Present estimates
to that of global material migration in tokamaks, which indicate that the requirements for particle control and helium
has been the subject of intense experimental and modelling removal for successful operation can very likely be achieved
research since the IPB in devices both with low-Z and high-Z in ITER. Furthermore, achievement of neutral densities and
plasma facing components. The main wall and, in most collisionalities near that predicted for ITER has allowed the
cases, the outer divertor (for ion B × ∇B drift towards the initial steps to validate edge plasma-neutral modelling codes
dominant divertor) have been identified as zones of erosion (see sections 2.3 and 3.5) with respect to physical phenomena
while net deposition occurs at the inner divertor. Two major such as recombination, Lyα radiation trapping and neutral–
effects have been identified to play a major role in determining neutral collisions. Such processes are expected to play a major
the observed migration pattern: strong SOL flows, which role in determining the neutral and plasma conditions at the
can preferentially direct the main-chamber eroded material ITER divertor. The importance of the SOL opacity to neutral
towards the inner or outer divertor (depending on the direction ionization has been identified as a major player in determining
S206
the efficiency of the fuelling methods in tokamak devices and 18

the value of plasma density at the separatrix (which in turn Outer Target
influences the divertor conditions) as described in section 2.7. 16 Inner Divertor TC
This has led to the development of fuelling methods which can Outer Divertor IR
be effective in the plasma edge conditions expected in ITER, 14 LPx4
of which the injection of frozen pellets from the inner side of
Heat Load (MW/m2)

the torus seems to be the most promising one. These studies 12
have provided the physics basis for carrying out the studies for 2 4
the fuelling required to achieve ITER reference performance, 10 100 eV
which are described in section 3.2. 300 eV
8 1000 eV
Ti ρθi
2. Experimental basis 6
2.1. Steady-state transport in the SOL and its implications 4

for divertor and wall interactions
JG04.547-13c
2 r-rsep (mm-omp)
2.1.1. Steady-state divertor power deposition profiles. A sig-
nificant part of the heating power in divertor tokamaks is lost
0
by conductive and convective heat transport across the sepa- -5 0 5 10 15 20 25
ratrix into the SOL. Having crossed the separatrix, this power Mid-plane radius (mm)
is primarily transported along field lines in the SOL to the
divertor. In present machines, the parallel heat flux in the SOL Figure 1. Comparison of ELM-averaged IR, LP (inter-ELM) and
can reach above 500 MW m−2 [2] and in ITER it is expected to TC (histograms, ELM-averaged) derived heat flux profiles on the
JET outer divertor target for a plasma with 16 MW NBI,
be ∼1 GW m−2 . At the divertor surfaces this must be reduced
2.5 MA/2.4 T and ion B × ∇B towards the X-point. The scale of the
below the technologically feasible, maximum perpendicular electron heat flux (LP) is four times smaller in the high power case.
heat flux for actively cooled structures, typically 10 MW m−2 Also shown are the profiles obtained for a 12 MW Type I H-mode
normal to the surface in steady state or up to 20 MW m−2 during (same field and current), using the shot-by-shot TC method (——).
transients. Several strategies are employed in present machines For comparison the poloidal gyro-radius at the outer mid-plane is
to reduce the peak divertor heat flux; poloidally inclining shown for three values of the ion energy [21].
the divertor tiles, increasing the magnetic flux expansion and
broadening the SOL heat flux width through increased perpen- Scalings for the dependence of the peak heat flux, qmax ,
dicular transport all act to maximize the area over which power and heat flux profile width, λq , with key discharge parameters
is deposited. In addition optimizing the divertor geometry can have been developed on a number of experiments. The depen-
increase the capability of the divertor to radiate power [3, 4] dence on total target heat load, Ptarget , or the closely related
and good tile alignment or tile imbrication (‘fish-scales’) can power to the divertor, Pdiv , are summarized in table 1 for a
minimize toroidal peaking at leading edges. The aim of these number of these scalings. Ptarget (and Pdiv ) will increase by
strategies is to achieve the conditions for partially detached more than an order of magnitude from current experiments
divertor operation [5], the reference regime for ITER, for which to ITER, much larger than the extrapolation in other parame-
modelling predicts tolerable peak heat fluxes [5]. ters such as toroidal magnetic field or density. The scatter in
The dependence of the steady-state divertor heat flux the power dependence from the various studies is large, even
profiles on both global and local parameters has been measured for different studies on a single device, and indeed both posi-
in many tokamaks and for a variety of divertor configurations tive and negative dependences are reported for the λq scalings.
[6, 7]. Heat fluxes are typically derived using data from Several of the scalings are more pessimistic than that assumed
infra-red cameras (IR), Langmuir probe arrays (LP) and in the ITER Physics Basis [1]. Even the most extreme scal-
thermocouples (TC), all of which have interpretation issues ing, however, yields λq for ITER of 3.7 mm ± 1.1 mm [19],
(e.g. IR thermography is subject to the effect of surface not dissimilar to the value of λq ∼ 5 mm obtained from fluid
layers and localized hot-spots [8–11], Langmuir probes only modelling of the ITER SOL [20], which is consistent with the
measure the electron component of the heat flux [12] and currently envisaged ITER operation. The scatter in the scal-
thermocouples rely on slow sweeping of the strike point [13] ings may result from the difficulties in diagnostic interpretation
or shot repetition [14]). A comparison of IR, LP and TC referred to earlier, from poorly constrained data sets (e.g. inclu-
time-averaged heat flux profiles for a typical ELMy H-mode sion of partially detached plasmas or ELM averaging over both
on JET is shown in figure 1. Profiles are usually time- Type I and Type III ELMy periods) or from subtleties in the
averaged in ELMy H-mode (in part because few diagnostics profile shape (e.g. the width is ill defined for a non-exponential
have the necessary time resolution to distinguish ELMs) but fall-off). All that can be strongly concluded from table 1 is that
observations from fast diagnostics indicate that ELMs either do there is a need for improved experimental measurements and
not significantly modify the inter-ELM profile (outer divertor) a theory-oriented approach for making extrapolations for the
[15–17] or dominate it (typical of the inner divertor) [3, 9]. target heat flux in ITER (see section 2.1.2).
The outer target receives the majority of the divertor heat A change in the profile shapes was indicated on JET by
load, concentrated in a narrower profile than at the inner target measurements at low densities. For low ion collisionality,
(typical out:in power asymmetry ∼2.5 : 1 and out : in peak heat νi∗ 5(νi∗ ≡ L /λii , where L is the connection length and λii
flux asymmetry ∼5 : 1). is the ion–ion collisional mean free path), TC measurements of
S207
A. Loarte et al
Table 1. Scalings for dependence of SOL heat flux width and peak target heat flux with total target (or divertor) power load for a number of
machines and multi-machine databases.
λq qmax Comment
0.44±0.04
Multi-machine: JT60-U, DIII-D and Pdiv — Scaling for λq mapped to outer mid-plane
ASDEX-Upgrade (DIVI) [28]
0.35±0.05
Multi-machine: JT60-U, ASDEX-Upgrade [1] Ptarget — —
0.52±0.05 0.5±0.05
ASDEX-Upgrade (DIVI) (IR data) [3] Ptarget Ptarget Type I and III ELMs included and partially detached plasmas.
−0.1 1.1±0.06
ASDEX-Upgrade (DIVII) (IR data) [3] Ptarget Ptarget Type I ELMs and attached plasmas only.
1
DIII-D [30] — Pdiv —
−0.13±0.08 1
JET (IR data) [31] Ptarget Ptarget Inter-ELM. λq from FWHM of profile at outer target.
−0.48±0.09
JET (TC data) [18] Ptarget — ELM-averaged. D (Type I ELMs) and He (Type III ELMs)
discharges included. λq from the integral width of profile
at outer target.
20 profiles on both targets, and also predicts a strong modification

qtot in these profiles with field reversal, which was not observed in
qe D the experiment [19].
qtot He The outer mid-plane heat flux width predicted for ITER
16 qe He
y=1.1157x1.4254 by extrapolation from JET TC analysis is λTq C ∼ 3.7 ±
1.1 mm [18]. Taking the lowest value of ASDEX-Upgrade,
DIII-D, JT-60 data from the multi-machine scaling ( [28],
qt,outerpeak (MW/m2)
12 figure 5) and the ASDEX-Upgrade scaling for DIV II from

IR-measurements, a lower limit of 5 mm is predicted. In
ITER, the target power load must not exceed the allowable
limit: PSOL fdiv /2π Rt λq famp qacc (derived from [12]), where
8 L-mode H-mode
qacc is the maximum acceptable peak heat flux, Rt the major
radius of the outer target, PSOL the power entering the SOL,
fdiv the fraction of PSOL delivered to the outer target, λq
4 the mid-plane heat flux width and famp the ratio of target to
mid-plane heat flux width. The design values for ITER [20]
JG03.549-3c
are qacc = 10 MW m−2 , Rt = 5.5 m, PSOL ∼ 80 MW and

0 famp ∼ 10 (4 from magnetic flux expansion and 2.5 from
2 4 6 target inclination). Then the requirement is fdiv ∼ 0.04λq
Pt,outer(MW) [mm], resulting in fdiv ∼ 0.2 for λq = 5 mm, consistent with
fluid modelling results (which indicate that power loading can
Figure 2. Peak total (TC), qtot and electron (LP), qe heat fluxes to be controlled by the divertor density) [29]. The TC predicted
the outer target for deuterium (D) and helium (He) plasmas as a
function of power entering the SOL, Pt,outer . The ion contribution value of λq = 3.7 mm results in a somewhat more constraining
becomes more pronounced in D at higher power. Also shown is the value of fdiv ∼ 0.15.
peak inter-ELM heat flux to the outer divertor for an H-mode power
scan in JET as measured by IR-measurements (- - - -) [18].
2.1.2. Steady-state energy transport in the near SOL.
qmax at the outer target greatly exceed the inter-ELM LP value The width of the near SOL, within 1–2 λq beyond the
(figures 1 and 2) [18, 21]. Interpreting this as a contribution separatrix, is governed by competition between the poloidal
from hot ions, it was inferred [22,23] that the energy of D+ ions transport, which results from parallel (to the magnetic field
striking the outer target was ∼300 eV at νi∗ ∼ 1, consistent lines) and diamagnetic (tangential to the magnetic surfaces
with CXRS measurement of separatrix ion temperature. The and normal to the field lines) components, and the much
ion heat flux (derived from the difference in the TC and LP slower perpendicular transport (normal to the magnetic flux
profiles) was closely correlated with a narrow peak in the surfaces). Parallel energy transport is determined by classical
outer profile [24]. This narrow peak was absent at the inner conduction and convection, with kinetic corrections to heat
∗
target although the ion component still dominated qmax . This diffusivities χi,e at low (separatrix) collisionalities νi,e , whilst
asymmetry cannot be explained by toroidal geometry and diamagnetic transport is governed by a mix of classical (E ×B,
ballooning-like transport alone. Classical drifts and/or neo- B × ∇B, curvature, diamagnetic, etc) and anomalous drifts
classical ion orbit loss (IOL) may play a role [18, 25, 26]. [12, 25, 32, 33]. Perpendicular, or radial, energy transport is
The dominant role of classical (mainly E × B) drifts in turbulent in nature, with typical heat diffusivities in the range
SOL energy transport was confirmed in analysis of matched χ⊥SOL ∼ 0.1–1 m2 s−1 , well in excess of classical diffusion
forward (fwd-B) and reversed (rev-B) JET discharges (ion (<10−3 m2 s−1 ). The development of a credible theory for
B × ∇B drift towards versus away from the X-point) [19, 27]. χ⊥SOL is key to reliable predictions for near SOL energy
Detailed modelling of direct ion orbit loss on JET using the transport in future devices such as ITER.
ASCOT code (see section 3.3) correctly predicts νi∗ as the Transport coefficients in the near SOL can be
governing parameter and can generally reproduce the fwd-B reconstructed using divertor target profiles of heat flux, density
S208
ne and temperature Te,i (see section 2.1.1) together with 1.0

upstream (e.g. mid-plane) ne and Te profiles from high JET
Te, Ti H-mode, λq,v
spatial resolution Thomson scattering or reciprocating probes. ne
Unfortunately, target profiles can be decoupled from the SOL A(Z)
0.8 q95
transport by recycling neutrals and impurity radiation near the
B
divertor and X-point [34], although these effects are generally
∆ (net and contributions)

less important in the high-power, low-density attached plasma
regimes discussed in section 2.1.1. Given this problem and 0.6
the general scarcity of data, attempts to extract transport
coefficients often assume radially constant χ⊥ values in the
edge, sometimes with a prescribed poloidal variation, in the
0.4
hope of determining scalings for the dependence of χ⊥ on
plasma and device parameters. An interpretive version of
the 2D fluid code B2-SOLPS5.0 [35] has been used ( [36])
to develop a multi-variable power law regression scaling for 0.2
D⊥ and χ⊥ in ASDEX-Upgrade by fitting the inter-ELM
JG03.549-6c
experimental data whilst adjusting constant D⊥ and χ⊥ values
separately. A similar approach was employed for JET data [22]
using the EDGE2D/NIMBUS package [37] to match target A1 A3 A2 N G2 B2 M G3 Q E1 J I O L2 Z1 G1 Z2 H K2 F L1 B1 K1 X D
profiles. In this study, radially piece-wise constant D⊥ and
χ⊥ were assumed; for the outer core, the edge transport barrier Figure 3. RMS error between theoretical and experimental λq
exponents, assuming dominant parallel convection and showing
(ETB) and the SOL, with a radial pinch velocity, being invoked contributions from (Te , Ti ), ne , A(Z), q95 and Bφ exponents. The
(stypically inward in the ETB region and outward in the SOL) dotted–dashed line represents the estimated measurement error,
to match the upstream ne and Te profiles. The results from both while the dashed line is the average over all theories [18].
studies imply a stronger χ⊥ dependence on β than ν ∗ (note that
because of the paucity of the Ti data, χ⊥i = χ⊥e was assumed). mechanisms. Comparison with experiment, figure 3, suggests
Unfortunately, the 2D fluid code approach is computationally that either the Pfirsh-Schlueter-regime ion heat conduction
expensive, limiting the parameter space resolution, and has or collisional ion orbit loss may play an important role in
only been attempted separately for different machines, so no SOL power exhaust. The collisional ion orbit loss heat flux
size scaling has be derived. Thus no extrapolation to ITER has ∗1/2
width λX−ν∗
q = λX q νi , where the direct ion orbit loss
been proposed. X
width, λq , was calculated for JET using the ASCOT guiding
Another, less computationally demanding, approach was centre code.
initially proposed by [38] and expanded by [39]. Analytical Comparing the results of the earlier and more recent
expressions for λq , developed using the two-point SOL model studies suggests that different mechanisms may be involved in
[40] from published theories of SOL radial transport combined electron and ion transport in the SOL, possibly the former being
with limiting cases of parallel transport (pure conduction linked with drift wave turbulence and the latter with collisional
versus pure convection), are evaluated against a wide range heat diffusion. Extrapolation of the results, however, to
of experimental data for λq and the key plasma parameters. a next-step device such as ITER (section 2.1.1) should be
The normalized RMS error of the predicted versus measured taken with care (see also section 3.2). Further experiments
scaling is used as the figure of merit to assess the credibility are clearly required to resolve this issue, including higher-
of the theoretical models for χ⊥ involved. Initial studies resolution diagnostics and advances in transport modelling. In
[39, 41, 42] examined the target heat flux width measured particular, development of diagnostics capable of measuring
by LP (thus dominated by the electron heat flux), λeq from the ion temperature profiles at different locations in the SOL
COMPASS-D (Ohmic and L-mode, νe∗ ∼ 0.2–3), Alcator is a high priority.
C-mod (Ohmic, νe∗ ∼ 20–370) and JET (Ohmic, L- and H-
mode, νe∗ ∼ 10–100) and focused on scalings with respect
2.1.3. Particle transport in the SOL. Particle transport
to ne , PSOL , q95 and Bφ . For νe∗ > 10 (i.e. for both JET
in the edge plasma and SOL plays a key role in the
and Alcator C-mod), five different theories emerged as being
performance and operation of a fusion reactor: setting the
noticeably better than the rest and all these exhibit a χ⊥SOL ∝
width of the SOL density profile and its impurity screening
Te n−1
1/2
e scaling leading to a negative power dependence, characteristics, regulating the energetic particle fluxes onto
−2/5
λq ∝ PSOL . For νe∗ < 3 (COMPASS-D), the three best
e
first-wall components and associated impurity generation rates
theories exhibit the same scaling χ⊥SOL ∝ Te n−1
1/2
e , although and determining the effectiveness of the divertor in receiving
ten other theories, yielding different scalings, could not be particle exhaust and controlling neutral pressures in the main-
statistically rejected. More recent statistical studies [18] of λq chamber. Here we focus on transport physics that occurs
relied solely on JET data for IR- or TC-measured target total in the absence of ELMs, which is based on data from
power profiles, obtained over a wide range of powers, densities, L-mode discharges and H-mode discharges at time intervals
fields and currents (L- and H-mode, νe∗ ∼ 8–15) and including without ELMs.
both deuterium and helium plasmas (see section 2.1.1). The list Observations and analysis of profiles, plasma fluctuation
of the applied theories was extended with collisional (classical phenomena and scalings of particle transport within and across
or neoclassical) and direct ion orbit loss radial transport a number of experimental devices have pointed towards a few
S209
A. Loarte et al
Near/Far SOL and Fluctuations Regression Analysis of Deff,

in Near SOL (ρ = 2 mm)
(1020 m-3)
Density Far 1.0
1.0 SOL Limiter
Near
Shadow
α(λei/L)-1.5
SOL
(m2 s-1)
0.2
0 5 10 15 20 25 0.1
Distance into SOL, ρ (mm)
0.0 1.5
Multiple
Isat/<Isat> 2 Correlation
Coefficient
1
0 0.0 0.5 Time (milliseconds) 1.0 1.5
0.01 = 0.64
0.0 1.5
2
1
0.01 0.1 1.0
0 0.0 1.5
(m2 s-1)
0.0 0.5 1.0 1.5
Time (ms) Scalings in Near SOL (ρ = 2 mm)
1
Deff
(m2s-1)
0.1
Density and Deff Profile Evolution
0.01 0.1 0.2 0.3 0.4 0.5 0.6
1020 Veff
(m-3)
(m s-1)
10 Ip (MA)
1019 ne /nG Density
0.43 1.0
0.27 0.8
-5 0 5 10 15
15 20
20
1.0 0.19 1
(m2s-1)
0.1 0.2 0.3 0.4 0.5
0.5
0.6
0.1 0.1
Deff
0.01
-5 0 5 10 15 20 λei/L
Distance into SOL, Limiter 0.01
Separatrix
ρ (mm) Shadow 0.1 0.2 0.3 0.4 0.5 0.6
ne/nG
Figure 4. SOL profiles and plasma transport characteristics in ohmic L-Mode discharges on Alcator C-Mod (from [51, 78]) which illustrates
key features seen in many devices: (a) near- and far-SOL zones with change in fluctuation character, (b) strong variation in Deff across the
SOL and flattening of the SOL with increasing core density, (c) correlation of near SOL Deff with parallel collisionality and (d) the trend of
increased particle transport coefficients with increased discharge density (collisionality).
key elements which are illustrated in figure 4: (1) the tendency cause was linked to a strong increase in the effective cross-field
for strong variation in the profile gradients and transport particle diffusivity (Deff = −/∇n) with distance into the
parameters (and physics) from the separatrix to the first wall, SOL, a result consistent with earlier heat diffusivity analyses
dividing the SOL into near and far regions, (2) bursty transport [42, 59, 60] and particle transport analysis [43, 45, 46, 61].
dynamics in the far SOL (reminiscent of ‘avalanches’ in a In related work it was shown that divertor neutral
self-organized critical system), carrying significant cross-field bypass leaks and/or open versus closed divertor structures
particle flux towards the first-wall, and (3) the dependence showed little effect on the midplane pressure in Alcator
of near SOL transport on the local collisionality. While a C-mod [59, 62–64], an indication that rapid cross-field particle
first-principles quantitative description of the phenomena is transport was the dominant source of main-chamber neutrals.
not yet in hand, the experimental observations have important These observations paralleled and confirmed those made in
implications in the areas of main-chamber recycling and ASDEX-Upgrade [65, 66] where the main-chamber neutral
impurity sources, threshold conditions for divertor detachment pressure was found to be insensitive to divertor geometry at
and the tokamak density limit. medium and high densities and that the level of main-chamber
Scrape-off layer profiles in many devices are often found recycling and the density in the far-SOL region were strongly
to exhibit a two-zone structure: a steep gradient region in correlated [67]. In DIII-D, the ion flux to the wall is typically of
density and temperature near the separatrix (near SOL) and a the order of the ion flux received at the divertor plate, increases
flatter profile region (far SOL), characterized as a ‘shoulder’, with plasma density and dominates in the detached divertor
extending from approximately one steep-gradient scale length regimes [68]. The contribution of main-chamber recycling
outside the separatrix to the wall (e.g. ASDEX [43, 44], to midplane neutral pressures is not uniformly reported to
ASDEX-Upgrade [45, 46], TEXT-U [47], DIII-D [48, 49], be important, suggesting a sensitivity to divertor/first-wall
C-Mod [50, 51], JT-60U [52], JET [53], TEXTOR [54], geometry and/or operational regime [69]. For example, DIII-D
TCV [55]). The earliest hints that cross-field particle transport [70], JET [71] and JT-60U [52] have reported a reduction in
in the far SOL towards the main-chamber walls could be main-chamber ionization sources and neutral pressures when
important for a tokamak reactor, competing with, or even the divertor was changed to a more closed geometry.
exceeding, the parallel loss into the divertor, came from It is generally accepted that the level of plasma flux that
ASDEX-Upgrade [45, 46]. Subsequently, Alcator C-Mod reaches a given wall surface depends on its wall–plasma gap—
focused attention on this behaviour as it was reported to perhaps explaining some of the variability that is reported
operate predominately in a ‘main-chamber recycling regime’ across different experiments. However, because of practical
with much of the plasma efflux recycling on main-chamber limitations, the main-chamber wall cannot be placed arbitrarily
surfaces rather than on divertor targets [56–58]. The primary far from the plasma. Moreover, the main-chamber wall
S210
includes the neutral baffles at the entrance to the divertor, characteristics appear to be independent of the confinement
structures which are fixed at locations to optimize divertor mode (L- versus H-mode) [85,86], although the time-averaged
neutral retention. These structures, in conjunction with the fluxes tend to be lower in H-mode discharges, consistent with
innermost magnetic flux surface that passes through them, a global increase in particle confinement. It should be noted
may be thought of as defining a ‘window frame’ [68, 72]. that while non-Gaussian PDFs, intermittency and motion of
Since the particle flux crossing this window frame depends coherent structures has been well documented over a number of
only on plasma transport processes, it is largely independent years [88–97], clear connections to a rapid cross-field transport
of the details of the first wall. Thus, while first-wall geometry mechanism in the far SOL and the potential impact on tokamak
and placement may affect the poloidal distribution of main- reactor operation have only recently been widely recognized.
chamber recycling, it is not expected to significantly influence In contrast to the far SOL, fluctuations in the near SOL
the overall level. region are found to exhibit a near-Gaussian PDF. Moreover,
Scaling studies of gradient lengths, effective diffusivities at locations inside the separatrix there is evidence of reversed
and/or time-averaged particle transport velocities have skewness or the formation of plasma ‘holes’ [86]. Folded into
revealed important dependences which serve to constrain this picture are the observations that transport and fluctuation
physics-based descriptions. For example, a number of levels on the high-field side SOL are significantly reduced
experiments have identified a steepening of gradients (inferred relative to the low-field side [85,89,98,99], consistent with the
as a reduction in diffusivity) as the local electron temperature expectation of higher turbulence and transport levels in regions
increased [42, 52, 59, 73, 74] as well as little sensitivity to of ‘bad’ field-line curvature. In summary, fluctuation statistics
the value of the magnetic field strength [39, 42, 59, 75]. combined with data from turbulence imaging ‘movies’ indicate
These observations are clearly incompatible with a Bohm-like that plasma structures intermittently ‘peel away’ from the edge
diffusion model and more in-line with a critical-beta model of the steep-gradient near SOL region and freely propagate
combined with a collisionality dependence [76], perhaps towards the wall, independent of L- or H-mode confinement.
further constrained by a relationship between density and Magnetic curvature and/or variation in |B| appear to play a
temperature gradients, ηe = d(Log Te )/d(Log n) ∼ 2 role. These observations suggest that the overall level of cross-
[67, 77]. Recently, particle diffusion coefficients and cross- field particle transport in the SOL may ultimately be set by
field transport velocities (veff = /n) in the near SOL conditions near the separatrix.
have been correlated with parallel collisionality which also While the appearance of shoulders in the far SOL aided
correlates with discharge density normalized to Greenwald in the discovery of a rapid transport phenomena in this zone,
density (n/nG ) [51, 78]. As collisionality increases, particle balance analysis suggests that the same underlying
fluctuation amplitudes, particle transport and associated heat transport physics is active in discharges without a shoulder.
convection increases dramatically across the SOL, impacting For example, although only a weak shoulder is normally
the SOL/divertor power balance at moderate n/nG (promoting observed in JET, the cross-field convection velocity profiles
divertor detachment) and impacting discharge power balance inferred from particle balance are found to be remarkably
at high n/nG (promoting thermal collapse and density-limit similar to that inferred in discharges from C-Mod [100] where
disruption). A strong variation of cross-field heat diffusivity the dimensionless plasma physics parameters were similar.
with n/nG has also been noted on the ASDEX-Upgrade [79]. Comparisons between DIII-D and C-Mod yield the same
At a basic plasma physics level, the latter observations suggest trends [101]. The opacity of the far SOL to neutral penetration
that SOL turbulence and resultant cross-field transport may set may offer part of the explanation for the appearance of a
a fundamental density limit for tokamaks even in the absence shoulder [101]; if wall-recycled neutrals ionize in the far SOL,
of impurity radiation [80]. In practice, the operational density they are rapidly transported back to the wall as ions. Thus,
limit for a reactor will also involve the limitations of impurity depending on the opacity to neutrals and the underlying cross-
peaking and radiation as well as requirements to remain in an field plasma transport, the far SOL can exhibit a ‘perpendicular
H-mode regime. recycling’ regime with flattened density profiles [58, 101–103].
The underlying physics of cross-field transport involves In tokamak plasmas with the same dimensionless plasma
plasma turbulence. Data from a number of experiments have physics parameters, the SOL would tend to be less opaque to
demonstrated a clear correspondence between the character neutrals as the machine gets larger, consistent with JET–C-Mod
of edge plasma fluctuation phenomena in the near- and far- comparisons [100]. However, the ITER SOL is expected to be
SOL regions and particle transport levels [49, 51, 77, 81–83]. dimensionlessly dissimilar to any current tokamak and opaque
Fluctuations in the far SOL are found to exhibit a ‘bursty’ to neutrals [20], which may lead to the appearance of a shoulder
character, with intermittent ‘transport events’ carrying plasma in the far SOL for gas-fuelled discharges.
towards the main-chamber wall with velocities well exceeding In light of the transport physics in the far SOL, the
100 m s−1 . Turbulence imaging systems record ‘blobs’ or level of plasma interaction with main-chamber surfaces is
plasma ‘filaments’ aligned with respect to the local B-field and a concern. Relative to target-produced impurities, main-
propagating poloidally and radially outwards [84–86]. The chamber impurities are poorly screened by the SOL, can impact
probability distribution functions (PDFs) of the fluctuations the core plasma [104–106] and may affect co-deposition of
exhibit a non-Gaussian tail, such that large-amplitude but carbon and tritium in the divertor [107]. Moreover, the rapid
rare outward-going transport events account for a large transport mechanism sets a minimum level of recycling and
fraction of the total particle flux. A statistical link between neutral pressure surrounding the confined plasma, which is
bursty transport behaviour and the departure from the most largely independent of divertor geometry and neutral bypass
probable gradient has been recently identified [87]. These leakage.
S211
A. Loarte et al
1.0
Bx B JET
Inner
0.5
M
0
Bx B
Outer
-0.5
-1.0
0 20 40 60
Bx B JET Distance to separatrix (mm OMP)
1.0
Inner
Stagnation 1.0
point Bx B JT60-U
Inner
0.5
0.5 Bx B
Bx B
0
M
M
C-Mod
Outer
-0.5 JT60-U Bx B
Outer
-0.5
-1.0
0 5 10 Stagnation -1.0
Distance to separatrix (mm OMP) point Bx B 0 20 40 60
Distance to separatrix (mm OMP)
1.0 1.0
Bx B JT60-U
Inner
Inner
0.5 0.5
Bx B
0 0
M
M
Bx B
Outer
Outer
-0.5 -0.5
JT60-U Bx B
-1.0 -1.0
0 20 40 60 0 20 40 60
Distance to separatrix (mm OMP) Distance to separatrix (mm OMP)
Figure 5. Parallel Mach number measurements are compared for L-mode plasmas in JT-60U [141], JET [115] and C-Mod (inner wall probe
only) [114] (after [142]).
Clearly, more research is required to develop quantita- [7, 51, 59, 108–114]. In figure 5, parallel Mach number (M )
tively accurate models for particle transport in the tokamak results are compared for L-mode plasmas in JT-60U [112],
edge. The development and testing of physics-based descrip- JET [115] and C-Mod (inner wall probe only) [114]. These
tions is particularly important since we need to understand how measurements show that there is a stagnation point in the SOL
the behaviour seen in present devices scales to reactor-size flow between the outer midplane and the lower divertor X-point
machines. Cross-machine comparisons are particularly valu- for the normal field direction (ion grad-B drift direction down).
able to explore the universality of the transport phenomena In the remainder of the SOL there are subsonic flows towards
and the dependence on dimensionless parameters and hope- the inner divertor (M ∼ 0.3–1.0) peaking at the inner
fully allow, at a minimum, an empirical projection to reactor midplane SOL [114] and dropping towards the inner divertor.
devices. The abovementioned comparisons between C-Mod, For the reversed field, the stagnation point appears to be at the
DIII-D and JET shows SOL profiles (including Deff and veff ) top of the plasma for lower single-null plasmas; there are SOL
that are similar in shape and magnitude when scaled accord- flows towards both the inner and outer divertors. The change
ing to dimensionless plasma physics parameters [100,101]. In in flows with reversal of magnetic field is not symmetric about
particular, veff is found to have a weak dependence on these di- zero flow [7, 51, 59, 108, 109, 111, 113].
mensionless parameters. These results give an initial guide and The dependence of SOL flows on main plasma parameters
suggest that the total plasma flux onto main-chamber surfaces has also been investigated. The most common tendency for
may be roughly independent of machine size in dimension- the SOL flow is a reduction in M at the outer midplane
lessly identical discharges, confirming observations in [72]. SOL [7, 59, 115] and M enhancement at the inner midplane
However, the scaling of these results to ITER is complicated SOL [112, 114] with increasing ne . This tendency is common
by the lack of a complete set of matching dimensionless pa- for the two toroidal field directions. The flow in the private flux
rameters in present tokamaks plus the lack of surety regarding region is considered another possible mechanism for producing
the importance of SOL opacity to neutrals. in–out asymmetry in the divertor ion flux (as described below).
The accuracy of flow measurements based on probes [116]
2.1.4. SOL flow and classical drifts effects. SOL mass is the subject of concern given the high Mach numbers inferred,
transport along or parallel to the magnetic field lines (SOL the suspicion that the probe perturbs the measurement [115]
flow) is of direct relevance to divertor pumping efficiency and the difficulty that the modellers have had in matching
and impurity screening properties of the SOL. The primary experimental measurements (covered later in this section).
measurement of such flows in the SOL is by Mach probes Other, less perturbing methods have been attempted to give
S212
further information. One such method is to directly observe formation of a plasma potential and pressure maximum at the
the effect of drifts on the transport of impurities. In Alcator X-point which drives a clockwise circulation (for normal B
C-MOD, 2D measurements, using CCD cameras, of impurity direction) of E × B flux around the X-point [123]. Divertor
plumes (HeII, NII, NIII) from local gas puffs [117, 118] asymmetries in JET have similarly been reproduced by both
from both the HFS and LFS of the machine were analysed. EDGE2D and SOLPS including drift effects [27]. Such
Unfortunately, this method is very complex and gives only a circulation pattern, if of significant strength, could also
qualitative support to probe measurements (e.g. sign but not affect the X-point region and radiation there, thereby affecting
magnitude) with the caveat that it can only be used when aspects of H-modes (transitions, cooling the pedestal).
the ion-impurity collision time of the charge state measured
is shorter than the ionization time. 2.1.5. Use of extrinsic impurity radiation to reduce divertor
Spectroscopic measurements of parallel flow velocities of heat loads. Extrinsic impurity injection is considered
low charge state impurity ions (CII, CIII, BII, HeII) is another necessary to achieve and control the desired radiated power
alternative flow measurement [119, 120]. Such measurements fraction (>50%) for the reduction in the divertor heat load
are roughly consistent with expectations for flow patterns in in ITER to an acceptable level [134]. The challenge is to
the divertor, but direct comparisons with probes have not find a robust regime with sufficiently high radiated power
been made. fraction and acceptable effects on the core (Zeff , cooling of the
There are several contributing physics processes that pedestal). The ability to disperse the short ELM energy pulse
have been implemented in 2D edge transport codes to model using divertor radiation is discussed in section 2.2.6. In this
flows. In neo-classical theory of plasma transport in toroidal section we concentrate on using radiation to reduce the power
geometry, E × B and B × ∇B (or diamagnetic for fluid model) flowing to the divertor in a more general sense, not limited to
drifts play an essential role in determining parallel flows and the divertor or to ELMs.
currents, e.g. the Pfirsch–Schlüter return current is a direct While Type I ELMs are considered the primary
consequence of these effects. In the SOL parallel SOL flow operational regime for ITER it is generally found that
can be produced due to the poloidal asymmetry of Er × Bt and radiation cannot be used to affect the ELM energy pulse itself
∇pi × Bt drifts [12, 121, 122]. (section 2.2.6). Still, averaging over ELM and between-
When implemented in 2D edge transport codes with ELM periods, large radiated fractions have been achieved
realistic magnetic geometries (fluid codes such as UEDGE through addition of Ar. As an example, JET has demonstrated
[33, 123], B2-Eirene [35, 124, 125], EDGE2D [126]), drift high δ discharges with radiated power fraction fR = 0.7,
effects lead to qualitatively similar SOL flow patterns to those HH98(y,2) = 0.9 at βN = 2.1 and Greenwald density fraction,
observed in experiments [108, 123, 126]. In other words, fG = 1.15 [135]. The use of Ar to disperse divertor power flux
they are consistent with the measured SOL flow direction (in has also been applied in JT-60U with initial results similar to
particular, near separatrix) but underestimate the SOL Mach that of JET-0.5% of argon making it possible to create plasmas
numbers by factors of 2–5. having fR = 0.8, fG = 0.65 and HH98(y,2) = 1. Even though
Modelling SOL flows is an active area of research, with half of the radiated power was dissipated in a mantle inside the
several avenues being explored. These include (a) torque separatrix, electron and ion temperatures remained high in the
generation due to surface averaged jr , (b) coupling between centre and in the pedestal also [136–139]. Good confinement
turbulence and SOL flow via Reynold’s stress [87], (c) (HH98(y,2) = 0.96) at high density (n̄e /nGW = 0.92) with high
global circulation of particles due to outward movement of radiation loss fraction (fR ∼ 1) has been demonstrated more
plasma filaments (blobs) and enhanced turbulence on the bad recently in JT-60U high βp H-mode plasmas utilizing high-
curvature (outer) side of the torus [114, 127] and (d) problems field-side pellets and Ar injection. Argon accumulation was
with the Mach probe interpretation itself, e.g. the effect of found to be modest and well within acceptable levels [140].
impurity cooling on the M magnitude [115] or difficulty in Type III ELMs have been observed in all divertor
distinguishing between the parallel and toroidal components tokamaks and are small enough to be easily compatible
of the SOL flow [128]. with the transient power limits for the divertor targets (see
The situation is further complicated in the divertor. This is section 2.2.1), although energy confinement is typically 25%
because of the existence of strong ionization sources as well as lower than with Type I ELMs, due to the loss of pedestal energy.
strong parallel and radial gradients of the plasma temperature Both JET and ASDEX-Upgrade have achieved radiated power
and density near the separatrix. The latter lead to Eθ × Bt and fractions approaching 80% [143, 144]. In ASDEX-Upgrade,
Er × Bt drifts with resultant cross-field and poloidal particle it was found that the reduction in confinement associated
fluxes, respectively [129]. Direct measurements of the private with Type III ELMs could be compensated by core density
flux region Er ∼ 3Te /λq ∼ 5–20 kV m−1 [54, 130, 131] peaking in the so-called completely detached H-mode (CDH)
suggests that E × B fluxes are comparable to the parallel [143]. Unfortunately, in larger tokamaks such as JET,
fluxes under attached conditions but are substantially smaller this compensating enhancement of core confinement has not
following detachment [131]. Enhancement of the divertor been found [145]. However, dependent on certain scaling
parallel flows upon detachment have been both measured assumptions, low q95 (∼ 2.6), nitrogen-seeded, Type III
[7, 130] and simulated using UEDGE and B2 [132, 133]. ELMy H-mode discharges have been demonstrated in JET
Simulations of the divertor flows using the UEDGE code with normalized confinement and density sufficient for Q =
[123, 130, 131] are in qualitative agreement with the above 10 inductive operation in ITER, whilst the steady-state and
findings, suggesting a link between drifts, divertor plasma transient power loads are comfortably within acceptable limits
flows and in–out divertor asymmetries. UEDGE predicts the (figure 6) [144].
S213
A. Loarte et al
temperature is obtained from real-time measurements of the

12 thermoelectric currents flowing between the inner and outer
Power (MW)
Prad
divertor targets [149]. The applicability of such scheme to
8 ITER will thus depend on the feasibility of thermoelectric
current measurements in ITER or their replacement by an
4 alternative control parameter.
Pheat
Accumulation of impurity in the core plasma is a potential
12 GW
ne problem associated with impurity seeding [140, 150]. It has
recently been shown that under specific conditions, the addition
ne (1019 m-3)
8
of central power deposition using wave heating such as ICRH
and ECRH has been identified as an effective tool for core
4 impurity control [151, 152]. Analysis of these discharges
ne
indicates that increased central diffusion, due to additional
3 power flux, counteracts the neo-classical inward pinch for
impurities and/or that flattened central electron density profiles
with increased central heating reduce the inward part of the
neo-classical drift [153, 154]. The α-heating in a burning
Zeff
2
reactor might be expected to provide a centrally peaked power
deposition profile favourable for impurity transport [153].
1.0
2.2. Transient ELM transport
2.2.1. Materials limits for ELMs and disruptions energy
H98(y,2)
0.8
loads on plasma facing components. Erosion during thermal
transient events such as Type I ELMs and disruptions plays
JG03.542-1bw
0.6 a major role in the choice of plasma facing materials for the
ITER divertor and first wall. Since a large number of ELMs
16 18 20 22 24 26 (of the order of several hundred) are expected in each ITER
Time (s) discharge, it is crucial that the surface temperature rise due to an
individual ELM remains below the threshold for sublimation
Figure 6. Overview of JET nitrogen-seeded Type III ELMy or melting (for carbon or metals, respectively). Otherwise
H-mode #59029. Key parameters for this shot are
Ip = 2.5(17) MA, BT = 2(5.3) T, q95 = 2.6(2.6), triangularity significant target plate erosion due to vaporization or loss of
δ = 0.44(0.5), βN = 1.7(1.5), H98 = 0.73(0.75), the melt layer (only for metals) can occur. Details of the
fGDL = 1.0(1.0), frad = 0.8(0.75), Zeff = 2.2(1.7), calculated vaporization (for CFC and W) and melting (only
WELM /WTotal < 0.015(0.03) [144] where the required values for for W) occurring at the ITER divertor under Type I ELM-like
low q95 operation of ITER at Q = 10 are given in brackets. energy loads for a range of assumptions are given in [155].
In contrast to ELMs, disruptions are not very frequent
Radiation dissipation of power efflux has also been applied events (1–10% of discharges are expected to disrupt in ITER)
to internal transport barrier (ITB) plasmas. One of the few but the expected energy fluxes are typically one order of
machines to address this issue has been JT-60U where neon magnitude larger than for ELMs. The major concern in the case
seeding was used in reversed shear plasmas to achieve a of disruptions is the resulting damage to, and erosion of, plasma
radiated power fraction of more than 80% without losing facing components (PFCs), including possible formation of
confinement at high density [146]. However, this strongly melt droplets.
radiating ITB plasma regime remains fragile and there are still Using a simple 1D semi-infinite solid approximation, the
issues related to scaling, accessibility and control, which will various PFC damage √ thresholds for ITER can be roughly
need to be investigated before it can be adopted for ITER. estimated as E/ t ∼ 35–40 MJ m−2 s−1/2 for sublimation
Real-time control of radiating scenarios is essential for of C, ∼45–50 MJ m−2 s−1/2 for sublimation of W and about
maintaining optimal conditions in long pulses. The JET Type ∼40 MJ m−2 s−1/2 for melting of W [3, 16], where E is
I ELM regime described above is an example of simultaneous the energy density deposited in a time interval, t. Recent
feedback control of the H-factor and fraction of the radiated calculations, assuming a short ELM pulse-width, show that a
power using Ar. Good performance was maintained for 5 s large number of ELMs with energy densities much larger than
with constant neutron yield, formation of a radiating belt 1 MJ m−2 would lead to an intolerably short divertor target
and reduction in target plate temperature [135], although the lifetime due to erosion, as shown in figure 7 for a 2 cm thick
resulting core argon concentration in such discharges is higher CFC target in ITER [155].
than that required for ITER. In ASDEX-Upgrade, a new control More precise estimates of the expected lifetime of PFCs
technique has recently been developed in which simultaneous under transient loads and of the effects of the ensuing impurity
feedback on the divertor neutral particle flux and electron fluxes on ITER plasmas require more sophisticated models
temperature are combined with external ELM triggering via [156–160]. These models are being validated at present with
high-field side pellet injection (e.g. [147, 148]), to maintain experiments in ELM simulation facilities [161] and tokamaks
power flux control and to prevent impurity accumulation in order to use them for quantitative estimates of the relevant
caused by ELM-free operation. The ‘effective‘ divertor effects in ITER.
S214
120
10 7
Temperature
10 4
"acceptable" lifetime 50
1st divertor Power 100
Temperature (Celsius)
1300
40 Energy
10 3
Number of pulses
Number of ELMs
Energy (kJ)
80
Power (MW)
∆EELM
10 5
100
0.5 ms 30
0.3 ms 60 1200
1.0 ms
τIR=390µs
0.7 ms 20 40
10
0.1 ms
10 3
10 ∆EELM+Between 20 1100
1
Outer Target JET
1.0 1.5 2.0 2.5 0 0
3000 2000 1000 0 1000 2000 3000
Divertor ELM energy density (MJ/m 2) time ( µ s)
Figure 7. Erosion lifetime (in number of ELMs or corresponding Figure 8. Temporal evolution of the divertor surface temperature,
ITER full power pulses) of an ITER CFC target (2 cm thick) as a deposited ELM power and energy onto the JET outboard divertor
function of the ELM energy density on the divertor target for various target for a typical Type I ELM [162].
timescales of the ELM energy pulses [155].
τIR , with significant power fluxes arriving at the divertor during

Such estimates for ITER are, however, subject to large
time intervals of several ms after the maximum power flux has
uncertainties both in terms of ELM-physics and materials
been reached (i.e. after τIR ).
characteristics: the expected divertor ELM energy load, its
The total amount of deposited energy due to the ELM,
spatial and temporal distribution (subjects of this section), target
WELM , can be estimated by integrating the power during an
characteristics of the redeposition of ELM-eroded material,
behaviour of shallow melt layers (for W), CFC thermo- ELM that exceeds the inter-ELM value, as shown in figure 8.
mechanical behaviour at high temperatures, etc. In addition Using this procedure the divertor ELM energy flux can be
to evaporation and melting, larger scale erosion mechanisms separated into two parts: (a) the ELM energy that reaches
could arise after repetitive high thermal loads. Local stresses the divertor in short timescales (i.e. [0,τIR ]), which is most
in the near surface could lead to deterioration of material important for evaluating divertor erosion by ELMs in ITER
properties and subsequent ejection of chunks/droplets of and (b) the ELM energy which reaches the divertor in longer
material [161]. time scales (i.e. ∼[τIR , ∼ τIR + 3 ms]), which contributes to
the total energy loading of the divertor by ELMs but not to
the peak surface temperature rise and, thus, to the proximity
2.2.2. Transport of ELM energy and particles from the confined
to the threshold for material damage (section 2.2.1). The
plasma to PFCs. A major concern about the proposed Type
ratio of the divertor ELM energy flux in short timescales to
I ELMy H-Mode scenario for ITER is the magnitude of the
the total divertor ELM energy flux estimated in this way is
energy release from the pedestal region towards the divertor
typically in the range 15–40% for a series of JET Type I ELMy
and first wall during Type I ELMs and the resultant surface
H-modes [31, 164], as shown in figure 9. This ratio is seen to
effects described in the previous section. To estimate the ELM
depend on pedestal plasma collisionality, which is correlated
energy arriving at the divertor we rely on the characterization
with the total ELM energy loss from the bulk plasma and the
of several aspects of Type I ELMs: time scales for the ELM
dominant transport mechanism for ELM energy flux out of the
energy flux to divertor PFCs, the area of the divertor subject
confined plasma ( convection versus conduction) as discussed
to ELM transient loads and the asymmetry between the inner
in sections 2.2.4 and 2.2.5.
and outer divertor ELM energy fluxes. These are determined
mostly by infrared measurements with high spatial (∼mm) and The characteristic timescale τIR of the ELM power
time (∼100 µs) resolution of the surface temperature of PFCs deposition at the (outer) divertor target is correlated well with
in present experiments. the transient time of fast ions to flow along the field line from
the pedestal to the divertor target (τconv = 2π Rq95 /cs,ped ,
2.2.2.1. Temporal evolution of the ELM divertor power where cs,ped is the sound speed calculated with pedestal plasma
deposition. An example of the time evolution of the derived parameters). This is shown for Type I ELM measurements
power flux during a Type I ELM at the JET Mk IIGB divertor from JET, ASDEX-Upgrade, JT-60U and MAST in figure 10
is shown in figure 8 [162]. These measurements are typical of [15, 31, 165–167].
experimental results for Type I ELMs in all divertor tokamaks. The good correlation between τIR and τconv is in agreement
τIR is defined as the time required for the surface temperature with PIC-simulations [168, 169], in which the sheath plays
to rise from 10% to 100% of the total increase during the a major role in limiting the energy flux from the pedestal
ELM [162] (390 µs in figure 8) which typically lasts ∼0.1– plasma to the divertor during the ELM power pulse. This
1 ms in the present experiments. A small, pre-ELM, increase physics picture has been further documented by ELM-resolved
in the power flux to the divertor is frequently observed over Langmuir probe measurements in various divertor tokamaks
relatively long time scales (larger than 1–2 ms) in JET and DIII- [3, 163, 170, 171] and by measurements of the soft x-ray
D [162, 163] and not discussed here, being not very relevant emission from the JET divertor target [172], which can resolve
to the issues discussed in section 2.2.1 for ITER. The total the different responses and timescales for the impact of hot
duration of the ELM power flux is indeed much longer than electrons and ions at the divertor target.
S215
A. Loarte et al
DOC-L δ = 0.27 50
out tot
Outer Divertor Target

0.5 45
1.2MA q95 = 3.1 JET,GasBox
ELM
40 AUG,DIV II
0.4 2MA q95 = 3.7
(0< t < τIR)/∆W
2MA q95 = 4.6 35
rel.frequency (%)
3MA q95 = 3.1
0.3 30
25
0.2
20
0.1 15
out
ELM
10
0.0
∆W
5
0.
0
ν*(neo) 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
FWHM /FWHM
ELM InterELM
Figure 9. Proportion of the ELM energy arriving at the outer
divertor in the time interval [0,τIR ] (i.e. the ELM start to the time of Figure 11. Histogram of the full width half maximum (FWHM)
maximum surface temperature) with respect to the total ELM outer ratio of the target heat flux profiles during and between ELMs for a
divertor energy versus pedestal collisionality for a range of Type I range JET and ASDEX-Upgrade Type I ELMs [31].
ELMy H-modes at JET [164].
700 that the ejection of ELM energy from the main plasma is
ITER
toroidally asymmetric. This toroidal asymmetry is only
600 JET retained in the divertor target ELM energy flux profile for areas
ASDEX-Upgrade in the far SOL [175]. The maximum ELM energy flux, near the
JT60U separatrix strike point, remains toroidally symmetric due to the
500
MAST
smearing out of the midplane toroidal asymmetries associated
400 with the long connection length of the field lines close to the
τIR (µ s)
separatrix [175–177]. As a consequence, only a small toroidal

300
peaking factor of the divertor ELM power flux of less than
1.5 has been observed in present experiments [166, 178] and a
similar range is expected in ITER.
200
Despite the large increase in radial transport associated
with the ELM event, the poloidal footprint of the ELM power
100
flux at the outer divertor does not show a significant broadening
when compared with the inter-ELM power flux profile at
0 the divertor [3, 31, 179, 180]. A histogram summarizing
0 50 100 150 200 250 300 350
conv. the results of a statistical analysis of the ELM power width
τ || (µ s)
(characterized by the full width at half maximum) at the outer
Figure 10. Characteristic timescale of the Type I ELM divertor divertor during the ELM is shown in figure 11 for the JET and
power deposition τIR on the (outer) divertor target versus transit ASDEX-Upgrade [31]. Therefore, the large increase in radial
time, τconv , for various divertor tokamak experiments operating in a transport at the ELM is accompanied by a large increase in the
range of conditions, calculated for the pedestal plasma transport along the field line. For attached conditions, the inter-
parameters [31, 166].
ELM width of the divertor power flux is simply correlated with
the SOL power e-folding length at the midplane. Therefore,
Despite the above evidence, it is not possible to rule
the lack of broadening of the power flux during the ELM at
out an alternative explanation to the sheath model for the
the outer divertor can be directly correlated with a lack of
correlation shown in figure 10. The observed correlation
broadening of the ELM power flux at the midplane of the
may be associated with the change in the main plasma ELM
device. For the inner divertor, which is usually detached in
energy loss from being dominated by conduction to being
most experiments, the above description breaks down and
dominated by convection. As discussed in section 2.2.4, with
decreasing temperature and increasing density (and τconv ) the large differences in the shape of the inter-ELM/ELM power
dominant mechanism for the loss of energy from the main deposition profiles are observed. Because the inter-ELM
plasma due to Type I ELMs changes from conduction (decrease power flux width at the divertor changes dramatically with the
in pedestal temperature) to convection (decrease in pedestal state of the divertor (attached/detached), it is best to extrapolate
density) [172–174]. from present experiments to ITER using the ratio of the ELM
SOL power flux width to the inter-ELM SOL power e-folding
2.2.2.2. Spatial characteristics of the divertor target heat flux length. Both can be obtained from divertor measurements
profiles during ELMs. Recent experimental results indicate provided that the divertor plasma is in attached conditions.
S216
0.9 0.8
Dα (a.u.) 114639
0.6
0.8
ELM
0.4
/∆W
0.7 0.2
divertor
1 2 3 4 5 6
0.0
ELM
0.6 3186 3188 3190 3192 3194

Time (ms)
∆W
0.5 4
114639
0.4 3
m )
19 -3
0 0.05 0.1 0.15 0.2
∆WELM ped
/W Bumper Tile
n (10
2
3 1 2 5 Limiter Position
e
Figure 12. Ratio of the ELM energy deposited on the divertor to the 4
energy loss from the main plasma versus normalized main plasma 1
ELM energy loss (to the total pedestal plasma energy) [31]. 6
Separatrix
0
2.2.2.3. In/out ELM energy flux asymmetries and total divertor 2.20 2.25 2.30 2.35 2.40
ELM energy flux. The asymmetry of heat deposition during R (m)
ELMs between the inner and outer divertors is dependent
on a number of plasma characteristics. For double-null- Figure 13. (a) Evolution of the divertor Dα signal during a Type I
ELM in DIII-D and (b) the corresponding profiles of electron density
diverted plasmas the ELM energy predominantly flows to the at the outer midplane. The steep pre-ELM density gradient (curve 1)
outer divertor [181, 182], demonstrating that the ELM energy is lost at the time of peak Dα emission (curve 3) and substantial
preferentially crosses the separatrix at the outer section of the density is observed in the far SOL up to the limiter surfaces [187].
plasma. This is in agreement with the peeling–ballooning
picture for the ELM trigger [183, 184]. On the other hand, determine where the missing ELM energy is deposited, but
for single-null plasmas, most experiments show that Type evidence from other diagnostics, which is reviewed in the next
I ELMs tend to deposit their energy preferentially at the section, indicates that this energy is deposited on plasma facing
inner divertor [178], although variations of this asymmetry components in the main chamber of tokamak devices.
with experimental conditions can be large. For single null
discharges, recent experimental evidence [31] indicates that 2.2.3. Interaction of Type I ELMs with the main-chamber
the in/out ELM energy deposition asymmetry depends on the plasma facing components. Results from fast diagnostics
size of the ELMs and/or divertor plasma conditions, with near the outer midplane of most current tokamak experiments
the asymmetry being largest for the smallest ELMs and the indicate that particles ejected from the pedestal during ELMs
higher densities [3, 31, 185] where the degree of detachment propagate in the SOL to large radial distances from the pre-
of the inner divertor is increased. These measurements thus ELM separatrix location. Such behaviour during Type I ELMs
indicate that the conditions of the divertor between ELMs has been reported by DIII-D [83,187], ASDEX-Upgrade [188,
play an important role in determining the energy flux to each 189], JET [190–192] and JT-60U [193]. Similar behaviour
divertor during ELMs, as expected if the plasma sheath plays a may occur for Type III ELMs in MAST [181] and TCV
major role in determining this flux. However, we cannot again [171]. An example of the SOL density profile evolution
rule out that the change in the in/out ELM energy deposition during Type I ELMs in DIII-D is shown in figure 13 [187].
asymmetry is influenced by the ELM energy loss from the The spatial characteristics of the ELM density perturbation
main plasma changing from being dominated by conduction seem to retain those of the ELM trigger, as determined by the
to convection (see section 2.2.4) or by changes in the peeling– peeling–ballooning ELM model [183, 184]: the perturbation
ballooning nature of the ELM trigger [184, 186]. has a ballooning character as determined by reflectometry
Analysis of divertor ELM energy flux measurements in measurements in ASDEX-Upgrade [189] and it interacts with
JET and ASDEX-Upgrade shows that not all the energy lost by the main wall in a poloidally/toroidally asymmetric way with
the main plasma during ELMs reaches the divertor [3,162]. Up an approximate n-number ∼10–15 [175, 190].
to ∼50% of the main plasma ELM energy loss is transported The typical radial propagation velocity of the plasma ELM
outside the divertor region (or flows to the divertor in time density perturbation in the SOL is of the order of vr ∼ 1 km s−1
scales much longer than ∼3 ms), as shown in figure 12 for in all divertor experiments/ELM types [83,180,191,194]. This
a series of JET Type I ELMy H-modes [31]. For the larger leads to an ELM radial propagation timescale of ∼100 µs,
ELMs, which can lead to a loss of energy from the main plasma which is similar to the timescale for parallel transport in the
that amounts to up to 15% of the pedestal energy, the missing SOL during ELMs, as discussed in section 2.2.2. The observed
ELM energy at the divertor can be up to ∼50% of the total ELM radial propagation velocities are of a size compatible with
main plasma ELM energy loss, with the proportion of missing those expected from a model for the propagation of ‘blobs’
energy decreasing as the normalized ELM size decreases. [195, 196] in the SOL [83, 194], but no detailed quantitative
From these experimental measurements it is not possible to comparisons, with this or other model, or a systematic study
S217
A. Loarte et al
# PDP-16775 ; time - 2.14356 s ; averaged : 0.0 s

400 limiter tile cross section
outer target plate
a) poloidal
toroidal
parallel heat flux /MW/m 2
ELM maximum
50 16775_160_bck.tem.hef
# 16775 @ 2.13 s
b)
parallel heat flux / MW/m 2

300
40
during an ELM
200 30
20
100
10 in between ELMs
in between ELMs 0
0
2.135 2.140 2.145 2.150 0 (R=2.18m) 50 100 150
midplane radius / m distance across limiter / mm
Figure 14. Parallel heat flux profiles on the ASDEX-Upgrade outer divertor target (a) and on an outer midplane limiter tile (b) between and
pre−ELM
during ELMs. The heat flux during ELMs on the limiter at R = 2.18 m (∼2.5λne ) is comparable to that in the far-SOL region of the
pre−ELM
targets (R = 2.145–2.150 m or 0.4–0.6λne ). The peak midplane heat flux during ELMs is within a factor of 3 of the peak target heat
flux [188].
of the dependence of the radial velocity on experimental

conditions have been established so far. a)
Associated with the large particle fluxes to the main
chamber during ELMs, a significant energy flux on the main-
chamber plasma facing components has been measured in
∆
ASDEX-Upgrade [188] during Type I ELMs, as shown in

figure 14. Measurements of power flux to PFCs in the main
chamber and of the electron temperature and density in the far
SOL during ELMs [83, 188, 194, 197] are consistent with the
energy flux onto main-chamber PFCs being associated with
the convective radial transport of ion energy. Pedestal ions
reaching first wall surfaces do not appear to lose much of their
energy either to electrons or to the divertor target [188, 194].
b)
Simple extrapolations of experimental data indicate that
significant ion energy fluxes could be deposited during ELMs δ
in the main-chamber PFCs of ITER, in a similar range to those
seen in present experiments (i.e. 20–50% of the main plasma
∆
ELM energy loss) [188,194]. This result is tentative at present.

However, the implications for the lifetime of main-chamber
components could be considerable. As a consequence, the
experimental determination of the Type I ELM energy and
particle fluxes on the main chamber of divertor tokamaks and
their dependence on plasma parameters are presently areas of
active research in all divertor devices.
2.2.4. Mechanism of Type I ELM energy transport to PFCs. Figure 15. (a) The normalized convected ELM energy as measured
The loss of energy from the main plasma during Type I by Thomson scattering versus the pedestal density normalized by
ELMs can be divided into two components: conductive losses the Greenwald parameter, nped /nGW in DIII-D. (b) The normalized
(due to the decrease in the electron and ion temperatures conducted ELM energy versus the normalized pedestal density in
in the pedestal region) and convective (due to the decrease DIII-D. A dashed line has been added to guide the eye [198].
in the plasma density). These two mechanisms depend on
the characteristics of the pedestal plasma in different ways. transport of energy from the core plasma to the PFCs during
With increasing pedestal plasma density and collisionality, the ELMs.
the normalized conductive losses decrease strongly, while the The phase of a Type I ELM during which the decrease
convective losses remain largely unchanged [172, 198], as of density and temperature in the pedestal region takes place
shown in figure 15. The different transport mechanisms that is characterized by strong broadband MHD activity. During
affect the main plasma energy losses are also involved in the this phase a large flux of electrons with energies typical of
S218
∆tDα
Pulse No. 58809
Soft X-ray Pedestal DOC-L δ=0.27 Open symbols Type III
800
1.2MA q95 = 3.1
1.2MA q95 = 3.1
2.0MA q95 = 2.8
Dαouter divertor 600
2.0MA q95 = 3.7
∆tDα(µs)
2.0MA q95 = 3.7
2.0MA q95 = 4.6
400 3.0MA q95 = 3.1
Soft X-ray hot e- inner div
200 ~ 2.2 x (τII -175)

Dαinner divertor
0
150 200 250 300 350 400
0 0.5 1.0 τII (µs) = L/csped
time (ms)
Figure 16. (a) Measurements with high time resolution (∼4 µs) of the soft x-ray emission pedestal collapse, outer divertor Dα emission,
inner divertor x-ray Bremsstrahlung (from hot electron impact) and inner divertor Dα emission during a Type I ELM in JET showing the
delay between the rise of the Dα emission at both divertors ( tDα ) and between the hot electron pulse and Dα emission at the inner divertor.
(b) Delay between the Dα rise at the inner divertor (with respect to the outer one) versus parallel ion transit time calculated with pedestal
plasma parameters for a large range of H-modes in JET [164].
the pedestal plasma reach the divertor target, as measured by (b) The second phase of the ELM energy flux to the divertor
their Bremsstrahlung emission on impact at the target [192]. is dominated by an increased particle flux, and occurs
This indicates that the pedestal plasma comes into direct over timescales much longer than that of the MHD phase,
contact with the target along the field lines. The flow of extending after the collapse of the pedestal region during
particles to the divertor target takes place on a different time the ELM. During this phase no high energy electrons
scale, dominated by the ion transport, which governs the are observed to reach the divertor target. The transport
convection of energy from the main plasma to the divertor mechanism which dominates this phase is thought to be
target. This separation of timescales and mechanisms for plasma convection associated with the ELM-enhanced
energy transport has been confirmed experimentally in several particle flux to the divertor [163, 164, 166].
divertor tokamak experiments [163, 166, 171, 172, 199]. An
As seen from figures 15 and 16, the relative importance of
example of such timescale separation is shown in figure 16
the above two phases and of the transport mechanisms which
[164] for experiments in JET and is reflected in the lack of
determine the ELM energy flux to the divertor and its time
correlation between the duration of the Type I ELM energy scale seem to be dependent on the pedestal plasma parameters
flux pulse at the divertor target and the duration of the strong before the ELM. As the pedestal density/collisionality is
MHD broadband ELM phase [3, 162, 172]. increased, the energy deposited in the divertor varies from short
Observations in existing experiments are consistent with timescales, dominated by conduction, to longer timescales,
the following transport mechanisms playing a major role dominated by convection [31, 163, 165].
during Type I ELMs, although a quantitative evaluation of their The transport mechanisms carrying the ELM energy to
importance and their overlap in time remains to be done. main-chamber PFCs are more uncertain. Present experimental
(a) In the initial phase, due to the strong MHD activity, the evidence indicates that the dominant mechanism is the
plasma in the pedestal region becomes connected to the convection of ion energy associated with the fast radial
divertor target through some unknown mechanism. Due propagation of the plasma density perturbation caused by the
ELM [194]. In order to confirm this hypothesis, more detailed
to their large velocity, high-energy electrons from the
measurements of the convective ion energy flux are needed.
pedestal arrive almost instantaneously [163, 164, 172].
During this phase electron conduction is expected to play
a major role in the transport of electron energy from 2.2.5. ELM energy losses in present experiments and potential
the pedestal to the divertor target possibly being limited physics for scaling to ITER. The previous sections provide
by the formation of a high-energy sheath [168]. The a description of the temporal and spatial characteristics of the
duration of the large ELM energy flux to the divertor (τIR , ELM energy pulses on PFCs and of the transport mechanisms
section 2.2.2.1) is usually longer than the MHD phase that are involved in carrying the energy from the confined
indicating that either the high-energy sheath is effective plasma to the PFCs during an ELM. This physics basis allows
at limiting the plasma losses along the field [199], or that the extrapolation of these characteristics to ITER, namely the
the ion power flux to the divertor in the initial ELM phase area of the PFCs over which the ELM energy flows and its
is comparable in magnitude to the electron ELM energy temporal waveform, but not of the absolute magnitude of
flux. the ELM energy flux itself. This extrapolation requires the
S219
A. Loarte et al
0.25 JET δ=0.25 3.5MA/3.0T There is some evidence that Type I ELM regimes which
JET δ=0.33 2.5MA/2.7T & 1.9MA/2.0T
JET δ=0.33 bef/@ pellet 2.5MA/2.7T are compatible with a reasonable lifetime of ITER PFCs may
JET δ=0.41 2.5MA/2.7T low L
JET δ=0.47 2.5MA/2.7T be possible. These are the Type I ELMs that exhibit only
0.20 JET δ=0.55 2.5MA/2.7T QDN
JT-60U convective losses and have been denoted ‘minimum energy
DIII-D δ<0.1
DIII-D δ=0.4 and δ=0.29 Type I ELMs’. These minimum energy ELMs are of a size,
ASDEX-U medium δ
0.15 <5%–10% of the pedestal energy, that may be tolerable in an
∆WELM / W ped
ITER-size tokamak but are generally found at too high pedestal

collisionalities for successful ITER QDT = 10 operation.
0.10
More recently, experiments at JET have demonstrated that the
ν* ITER
reduction of ELM energy loss is correlated with the reduction
0.05 of the conductive losses in Type I ELMs at any collisionality by
achieving small convective ELMs in conditions of low pedestal
collisionality as well [164]. Unfortunately, it has not yet been
0.00
0.01 0.1 0 possible to unambiguously determine which parameters are
ν* (neo) responsible for these small ELMs in order to predict their
operational space in a large tokamak such as ITER.
Figure 17. ELM energy normalized to the pedestal energy versus The physics of how the ELM energy reaches the divertor
pedestal collisionality. The multi-machine data collected by the
ITPA Divertor Group include density scans as well as varying
and wall is still too uncertain for predicting the time
configurations [166]. dependence of the heat pulse reaching those surfaces. Further
work is needed for acquiring a detailed model that can follow
the evolution of the ELM instability into its non-linear phase,
comparison of absolute ELM energy losses from the main where transport is expected to occur (e.g. [201]) and thus
plasma in relevant divertor tokamaks in order to identify explain how energy crosses the separatrix. Experiments to
the size scaling of these losses and the physics basis for determine whether it is the plasma physics or the atomic
their extrapolation to ITER. The pedestal requirements (see physics, through recycling, that determines the scaling of the
section 2.5.2), combined with the ELM transport in the SOL pedestal and ELM size should provide significant insight for
and divertor (section 2.2.2) and the material limits of the operation in future tokamaks. If plasma physics sets the
divertor PFCs (section 2.2.1) lead to the maximum ELM size pedestal and ELM size then it is likely that the low collisionality
that a tokamak divertor target can tolerate. For ITER this limit of ITER will result in very large Type I ELMs. However, if
is 5–10% of the pedestal energy [155]. The uncertainty of atomic physics plays a significant role, such as narrowing the
this limit is due to the SOL and divertor ELM transport issues pedestal width, then smaller, tolerable Type I ELMs in ITER
described in the previous sections. may be possible.
Initial studies of ELM energy losses [16] concluded that, in In addition to the Type I ELM regime, several other
order to compare ELM size across devices and configurations operational modes that exhibit a robust edge pedestal with
and to project to future tokamaks, the energy lost at an ELM no, or small, ELMs have been identified in current tokamaks
should be normalized to an estimate of the bulk plasma energy (see chapter 2, section 4.8, this issue [203]). It is unclear if
which is potentially accessible to expulsion by ELMs. This any of these regimes will prove applicable at the parameters
is usually characterized by the pedestal plasma energy, which and constraints envisioned for a future large tokamak such as
is the total pedestal pressure, ions plus electrons, times the ITER. It is important, however, to continue investigating these
plasma volume [16]. These initial studies have been further regimes as an alternative in case the Type I ELM regime proves
pursued by enlarging the original database and including to be incompatible with the ITER divertor.
ITER-relevant Type I ELM energy loss data from a number of
tokamaks [166]. The normalized ELM energy from this data 2.2.6. Ability of radiating plasmas to buffer ELMs. Ideally,
set is plotted versus the pedestal collisionality in figure 17. This radiation can be used to remove energy from an ELM before
figure illustrates the generally observed trend that the ELM it reaches first wall surfaces. Our present understanding of the
size decreases with increasing pedestal density and decreasing so-called ELM ‘buffering’ is that it is fairly limited in its usage,
pedestal temperature [172, 173]. The present database also primarily effective with small, Type III ELMs. For example,
suggests that at the low pedestal collisionality required for recent experiments with nitrogen seeding and deuterium
ITER, current tokamaks would project an ELM size that is puffing in JET [144] have demonstrated full detachment
2–3 times the tolerable limit for the divertor. Although pedestal between Type III ELMs, in agreement with previous JET
collisionality is a good ordering parameter for the ELM energy results [202]. For these conditions, radiative dissipation of
losses across devices, it is not clear at present whether the the ELM energy is observed for radiative power fractions
pedestal collisionality is the relevant parameter to extrapolate above 80%, in agreement with observations in CD-H mode
ELM energy losses to ITER or not, as the detailed physics experiments in ASDEX-Upgrade [204].
mechanisms which lead to these losses have not been precisely Radiative dissipation of the more ITER-relevant, Type
identified. For example, measurements from JT-60U [200], I ELMs, in present experiments has not been as successful.
ASDEX-Upgrade [174] and JET [164] indicate that Type I Nitrogen or argon impurities have been injected into such JET
ELMs with significantly less energy than the trend shown in plasmas, with Wdia up to 500 kJ, achieving radiative fractions
figure 17 and similar pedestal collisionality are seen in present of ∼55% [205]. However, the energy delivered to the target
devices in some experimental conditions. plate Wtarget / Wdia ) is only marginally reduced (by 18–24%)
S220
Mark I Mark IIA 3.5 Mark IIGB

Mark IIA/P
Vertical
Divertor neutral pressure (10-3 mbar)

Divertor
coils 3.0
2.5
2.0 Mark IIA/P

Cryo-pump
Target plates Vertical
Mark IIGB 1.5
Horiz.
Septum
1.0
JG98.583/7c Mark I
Horiz.
0.5
JG98.583/2c
0
2.0 2.5 3.0 3.5 4.0
19 -3
Line averaged density (10 m )
Figure 18. (a) Diagram of the three divertor geometries explored in JET. The divertor closure was increased while going from Mk I → Mk
IIA → Mk II Gas Box. During the exploration of Mk IIA, the conductance of the by-pass leaks was reduced (these divertor experiments
are referred to as Mk IIAP which stands for Mk IIA with plugged divertor leaks). (b) Measured neutral pressure at the divertor cryo-pump
for similar JET L-mode discharges in the divertor geometries in (a) [212].
compared with the non-seeded operation [205]. Plasma edge The understanding and prediction of neutral densities in
modelling of ELM transient radiation is in agreement with the the divertor and main chamber requires a mixture of physics
small effect of buffering the JET Type I ELMs [205]. Previous understanding (e.g. recombination, neutral–neutral collisions)
calculations with B2-Eirene for ITER [206] also predicted that and an understanding of the role of mechanical structures,
a significant effect could only be obtained for very small ELMs whose progress is reviewed in this section. The results herein
in ITER ( Wdia < 0.6 MJ). refer to single-null diverted plasmas unless otherwise noted.
An additional effect providing buffering of Type I ELMs
could take place through the modification of the ionization
2.3.1. Neutral pressure control. Divertor geometries and
processes occurring during the ELM burst. This leads to a
openings have been modified in a number of tokamaks to
decrease of the Dα divertor emission at the ELMs (so-called
enhance divertor closure and thus reduce the leakage of
‘negative’ ELMs) [202, 207]. The effect has been studied
neutrals from the divertor. Divertor closure is influenced by
in linear divertor simulators [208–211]. Unfortunately, no
the divertor plasma parameters through their effects on the
significant energy loss is expected due to this process, both
ionization mean free paths and also by the conductance of
because of the small potential for radiation of the hydrogen
neutrals through and past the divertor structure to the main
atom and because re-ionization of a recombining plasma
plasma chamber [69]. Experiments in many divertor tokamaks
would tend to increase the power flux to the divertor, as the
have shown that increasing the divertor closure has led to larger
energy of hydrogen recombination would be deposited on the
divertor neutral pressures [62, 179, 212, 213, 214], as shown in
divertor target surface rather than being radiated away by the
figures 18(a) and (b) for the JET divertors.
recombination radiation.
In general, the predictions with respect to the behaviour
of deuterium neutral and recycling impurity exhaust by edge
2.3. Neutral processes and implications for the divertor modelling codes have been confirmed by the experiments in
and SOL tokamaks. In particular, the dependences of the neutral particle
The control of neutrals in fusion devices is important for the and impurity exhaust on the details of the divertor geometry
plasma performance in these devices. Significant divertor and on the local magnetic flux surfaces have been confirmed
pumping must be achieved both for control of core fuelling [65, 69, 215–218]. The neutral pressure in the pumping
and of He ash removal. This favours maximizing the divertor plenum can be very sensitive to the exact location of the
pressures for those species. In addition, one wants to minimize divertor strike point [69, 216, 217] when the neutral transport
neutral densities in the main chamber in order to minimize is kinetic (termed ‘ballistic’). For conditions in which the
the erosion of the main-chamber plasma facing components neutral collisional mean free path (C–X and elastic collisions),
(by charge-exchange sputtering, chemical erosion, etc). Such λMFP , is short compared with the divertor dimensions (d)
eroded impurities lead to a reduction of the lifetime of in-vessel and ionization length, λIONIS (λMFP /d < 1, λMFP /λIONIS <
components and to core plasma contamination, as impurities 1), the neutral transport is diffusive and the dependence of
produced in this interaction have a higher probability of divertor pumping on the strike point position is much weaker
penetrating the main plasma than of those generated at the [65,69,215,219]. Next-step tokamaks, such as ITER, will have
divertor (see section 2.4). much smaller λMFP /d compared with current experiments.
S221
A. Loarte et al
Therefore, the research in this field is focused on validating the one shot) [62], as well as by the de-correlation between the
modelling codes with the experiments in the present devices main chamber and divertor neutral pressures once the mechan-
under conditions as close as possible to the ITER divertor ical leakage of the divertor is reduced to sufficiently low values
conditions. The effect of the double-null operation on pumping as seen at JET (Mk IIA → Mk IIA/p → Mk IIGB) [212].
has received much less attention (e.g. [220]). Thus the main-chamber neutral density has some irreducible
The major uncertainty remaining in modelling the neutral minimum determined by anomalous radial transport.
pressure in ITER-like devices concerns processes which occur Another effect that remains to be clarified is the effect
to a much lesser extent in the present experiments but are on the neutral control of the existence of a dome (preventing
likely to dominate the neutral dynamics in ITER. These include neutral escape from the divertor through the private flux
neutral–neutral collisions (viscosity), trapping of Lyman series region) and a septum (preventing direct neutral transport
radiation (of which Lyα is the principal component), and between inner and outer divertors) in the divertor. Modelling
neutral elastic collisions (friction). These processes were all for ASDEX-Upgrade has shown that the presence of the
shown to play an important role in modelling the Alcator dome is essential for explaining the increase of the neutral
C-Mod experimental divertor pressures [221]. The Alcator divertor pressure in the modification from the Div I to the
C-Mod divertor plasma is closest in λMFP /d and ne of all Div II divertors [65]. However, recent experiments at JET
tokamaks to the ITER neutral collisionality and absolute have shown that the effects of the dome and septum on the
plasma density. Including a proper modelling of these effects, divertor neutral pressure distribution and the control of in/out
the modelled pressures were (lower than but) within a factor asymmetries are much smaller than the initial experiments
of 2 of the experimental measurements. Previous modelling, with the dome and septum had anticipated, at least for
which did not include any of the processes important in plasma configurations with the x-point close to the dome
the ITER divertor described above, showed a disagreement [69, 225–227]. The reason for this unexpected result is being
by more than an order of magnitude between the modelled presently studied.
pressure (too low) and experimental measurements [222]. The findings described so far in this section are ‘ELM
Further work is in progress to improve the modelling of averaged’ when they concern ELMy H-modes. ELMs are
neutral interactions and evaluating their effect on ITER divertor seen to modulate the divertor pressure noticeably, particularly
performance [223] and of the radiation transport in these at low plasma densities/low ELM frequencies [228]. At
conditions (including Lyα trapping) [224]. higher plasma densities, which are associated with the higher
Detachment of the divertor leads to a saturation or frequency Type I ELMs, the modulations of the neutral
slower increase of the divertor deuterium neutral pressure pressure at the main chamber and divertor due to ELMs are
with increasing density [202, 215]. This saturation/slower smaller (see figure 14 in [229]) and, thus, the ‘ELM averaged’
increase of the deuterium neutral pressure is associated with neutral densities are adequate to describe the neutral pressure
changes in the neutral source and transport caused by plasma in these conditions. With increasing machine size, the time
detachment [69,215]. For conditions in which the neutral mean constants associated with the size of the divertor and associated
free path is larger/comparable to the dimensions of the divertor pumping plenum become longer. When the ELM frequency
plasma in detached conditions, neutral escape is favoured by becomes high enough (time constant shorter than the divertor
plasma detachment and a saturation of the neutral pressure pumping time constant) it may lead to a smoothing out of the
with detachment is observed [69]. For conditions in which effects of ELMs [228]. Due to divertor lifetime considerations
the neutral mean free path is smaller than the dimensions of (see section 2.2), ELMs in full performance regimes to be
the divertor plasma in detached conditions, such as in Alcator explored in next step devices, such as ITER, must lead to
C-mod, the detachment is associated with a reduced neutral small main plasma losses (in relation to the total energy/particle
mean free path. Therefore, in these conditions, the escape of content of the plasma), which are typically in the lower range
neutrals from the divertor is similar to or smaller than in high of the present experiments [166]. Therefore, the perturbation
recycling conditions and the divertor neutral pressure increases of divertor neutral parameters by ELMs is expected to be lower
with plasma detachment [215]. This last behaviour is the one in the next step device than in the present experiments and the
expected for next step devices because the divertor plasma ‘average’ ELM modelling thus reasonably realistic for those.
will be opaque enough to neutrals even in partially detached
conditions. 2.3.2. Helium exhaust and noble gas impurity enrichment.
The control of the main-chamber neutral pressure by diver- A significant expansion of the database for helium transport
tor geometry remains an area of active research and no quan- and exhaust has been done in recent years for tokamak
titative conclusion on the extrapolation from present results scenarios with improved confinement, i.e. both for H-modes
to ITER has been reached so far. Several experiments have and discharges with internal transport barriers (ITBs).
shown [69, 70, 212, 216] that decreasing the divertor cham- The transport of helium from the core to the pumping
ber conductance to the main chamber can decrease the main- plenum is commonly characterized by two figures of merit:
chamber neutral pressure, while others have found little or no (a) the ratio of the helium particle residence time to the
∗
effect [64, 213]. One explanation, gaining support, is that the energy confinement time, τHe /τE [230] and (b) the ratio of
source of neutrals in the main chamber is a combination of the helium fraction in the exhaust gas to the main plasma
divertor leakage and neutral recycling occurring locally [69], helium fraction or helium enrichment factor, ηHe,exh . The
the latter due to anomalous radial transport (see section 2.1.3). requirements for reliable ITER fusion performance as well
∗
This is supported by Alcator C-Mod experiments varying the as results showing values of τHe /τE within that needed for
∗
conductance between the divertor and main chamber (within the reference ITER ELMy H-mode regime (τHe /τE < 5–10),
S222
are included in the ITER Physics Basis [1]. Recent results The integrated modelling of helium transport in the core,
from several experiments [231–235] for the ITER reference edge and pumping plenum using the COCONUT [246], B2-
H-mode scenario have increased the confidence on the previous EIRENE [247–249], EDGE2D [218] and DIVIMP/NIMBUS
extrapolations of experimental results to ITER. As previously [241, 250] codes has shown qualitative agreement with
reported, the helium exhaust does not appear to be constrained experimental observations, which gives some confidence in
by core transport in ELMy H-modes [236], but is limited by the the modelling predictions for ITER, although further work in
efficiency of helium transport in the SOL and pumping [232]. this area to refine the predictions is on-going.
More recently, the study of the He transport has been expanded Earlier calculations suggested that elastic scattering of
beyond the ELMy H-mode to other scenarios. In JT-60U, helium neutrals by hydrogenic ions could enhance helium
helium beams were injected into reversed shear plasmas with atom transport towards the divertor [251, 252]. Experimental
ITBs, while simultaneously pumping both legs of the divertor measurements have shown that this process mainly heats the
region [234]. In plasmas with strong ITBs and, therefore, helium atoms in the divertor, thus increasing their mean free
good core confinement (HITER−89P ∼ 2.4), the residence time path [253]. This results in two competing effects: more He
of helium inside the ITB is about 1.7 times longer than that atoms can reach the pumping plenum duct and more He atoms
∗
outside the barrier and τHe /τE ∼ 15. Weakening the ITB by can reach the confined plasma. The inclusion of this process
increasing the core density and divertor recycling led to better in the divertor modelling codes has shown that the efficiency
∗
helium exhaust, τHe /τE ∼ 8, at the cost of poorer performance of the helium exhaust can be improved by a factor of 3–5 in
(HITER−89P ∼ 1.2). A similar result was obtained for a weak ITER [254].
∗
ITB at JET, τHe /τE ∼ 5 [232, 235, 237] with ηHe,exh ∼ 0.4–
0.6 [238], well within the ITER requirements [1].
2.3.3. The role of recombination in divertor processes. If
The recent improvement in helium removal rates has
detached plasma conditions are appropriate, recombination
been obtained both by optimization of the divertor plasma
can occur in significant amounts and strongly affect divertor
configuration and by increasing the divertor closure. Helium
operation. The effects are two-fold. (a) Recombination
partial pressures above 0.01 Pa were measured following
reduces plasma flux to the plates. This lowers power
divertor modifications in JT-60U with inner-leg pumping [231]
flux by removing the internal energy (13.6 eV + 2.2 eV
and in diverted corner configurations in the JET Mk II-GB
dissociation potential) of each ion–electron pair, thus further
divertor [239]. Minimizing the gap between the plasma strike
reducing divertor power deposition, beyond the drop in Te
zones and the entrance to the pumping plenum enhances
ballistic scattering of He neutrals into the pump [213,219,240, associated with detachment, and reducing physical sputtering.
241] and minimizes the neutral backflow from the pumping (b) Volume recombination affects divertor performance by
plenum into the divertor [234]. Better closure of divertor strongly changing the ionization balance across the divertor
mechanical leaks enhances compression of deuterium, helium region. Deuterium recombination has been observed on most
[233, 241] and argon [242] but appears to have no effect on diverted tokamaks [207, 255–258].
helium or argon enrichment. Induced SOL flow by the ‘puff The most well-documented recombination process in the
and pump’ technique did not significantly alter the enrichment experiments has been the 3-body recombination, indicative
of helium [213, 241, 243] but enhanced the neon and argon of electron temperatures 1 eV and electron densities greater
enrichment in DIII-D [243] and JT-60U [244], although not than 1020 m−3 (in some cases 1020 m−3 ). For these
in ASDEX-Upgrade [213] or JET [241]. The reasons for this conditions, radiative recombination rates are relatively small.
discrepancy are not yet understood. Another important recombination process, molecular assisted
The different enrichment behaviour of the noble gases recombination (MAR), has also been identified. Although
helium, neon and argon can be understood by the differences difficult to study in a tokamak, MAR continues to be of interest
in impurity neutral ionization mean free path, λIONIS , and the because of its potential in increasing the recombination rate by
effect of friction and thermal forces (impurity ions). Due to orders of magnitude and occurs at higher Te [1]. Estimates
its large ionization potential, helium is more likely to escape of its magnitude, relative to the 3-body rate, range from 20
the divertor as a neutral rather than hydrogenic species or to 30 [259] to ∼2 for ne = 10nH2 [260]. Whereas in linear
other neutrals (e.g. Ne or Ar) leading to a relatively low devices nH2 is usually fixed as an external control parameter,
compression ratio. This effect is enhanced under detached in a divertor it is self-sustained and determined by recycling
divertor conditions and the He/D ratio in the divertor is seen to fluxes and molecular transport. In particular the MAR process
decrease for such conditions [233, 234, 241]. Neon and argon, itself reduces nH2 and, thus, the role of this mechanism on
are ionised much closer to the divertor target and, therefore, divertor dynamics cannot be understood by simply comparing
experience the frictional drag force as ions, driving them the rate coefficients alone. Indeed recent modelling for ITER,
back towards the target. As a consequence, the enrichment including all relevant neutral/molecular processes, has shown
(and compression) increases with increasing noble gas mass that the effect of MAR is negligible compared with the 3-body
and it is usually larger than unity for the heavier species recombination, for typical divertor conditions [223].
in all divertor conditions [241, 243, 245]. The enrichment Experimentally, it appears that the role of recombination
of neon and argon increases with increasing density up to in plasma flux reduction strongly depends on the plasma
detachment [241,243,245] after which it drops. However, such conditions. 3-body recombination has been shown in some
an effect is less likely to occur in next step devices because the cases to account for large fractions (>50%) of the ion sink in
ratio of λIONIS to divertor size will be much smaller than in the divertor [207, 261] and in the MARFE [262]. However,
the present experiments for all regimes (including detached in some detached plasmas [207, 261, 263, 264], in particular
divertor plasmas). those with high SOL heat flux which detach after impurity
S223
A. Loarte et al
puffing [261], recombination appears not to be a significant findings of that review, updated with papers published since
ion sink. This variation in recombination is due to the strong 2001. No present-day experiment contains all the features
dependence of the 3-body recombination is the local plasma (geometry, materials, particle and power fluxes) expected in
density (n3e ). ITER, thus making projections difficult.
Studies of MAR in tokamaks indicate that its effect The following different physical processes occur in the
is small. Measurements in ASDEX-Upgrade show that divertor and the main chamber.
populations in the molecular vibrational levels required for
(a) Divertor processes. In the divertor, the impurity sources
MAR are only a few per cent of the total molecular
are highest near the strike points and can extend along the
density [265, 266] and that molecular influxes can be
target plates. Target plate material is released by deuterium-
comparable to atomic fluxes [267]. The level of MAR in
induced physical sputtering with additional contributions by
detached plasmas has been inferred to be in the range of
chemical sputtering [276, 277] (in the case of carbon) and by
10–50% [258,268] of the overall recombination rate, similarly
impurity-induced physical sputtering (in the case of high-Z
to the modelling done by Reiter [260]. In fact, inclusion
metals [278, 279]). The initial migration process is prompt
of vibrationally excited molecules in a B2-EIRENE model
re-deposition when the ionizing mean free path is of the order
actually reduced the total recombination rate [268].
of the ion gyro radius making the created ions re-impact the
The results of recombination and detachment studies
target material in the vicinity of the initial erosion. Prompt
using ‘divertor simulation plasmas’ of linear devices
re-deposition mitigates erosion rates making the net erosion as
[209, 264, 269, 270] show MAR to be much more important
much as an order of magnitude lower than gross erosion rates.
for such plasmas. The occurrence of MAR coincident
The effect increases with decreasing ionization mean free path,
with detachment has been confirmed, in both its primary
either as a result of increasing plasma density or higher Z
pathways, the so-called ion conversion [269] and dissociative
materials [280]. Impurities migrate for longer distances if they
attachment [271]. In these experiments, which have much
leave the divertor region either as neutral atoms or ions pulled
higher H2 /H and H+2 /H+ ratios than those in tokamak divertors,
out of the divertor by the thermal force. Impurity ion flow
MAR can dominate the radiative and 3-body recombinations.
velocities in the range of 20 km s−1 (out of the divertor) have
These results are, therefore, not necessarily inconsistent with
been observed [120, 281], with the magnitude agreeing with
those quoted above for tokamaks, as the plasma conditions in
UEDGE calculations [282]. These impurities are eventually
linear devices and tokamak divertors are very different.
deposited at some distance from the strike point. A similar
Although there is certainly incontrovertible evidence for
deposition pattern is found for material eroded at the main-
the importance of recombination (3-body) in current detached
chamber plasma facing components, which can be ionized
divertor tokamak plasmas, the influence of this effect on the
already in the periphery of the boundary plasma [283]. Carbon
divertor performance of next step devices is still uncertain.
can be deposited as hydrogenated amorphous films, which
This requires the improvement of the present modelling
are subject to high re-erosion [284]. By subsequent erosion–
codes to include the relevant neutral and molecular processes
deposition steps C can migrate to remote locations, which are
(including recombination) along the lines of the studies in
not in direct or line-of-sight contact with the plasma. The entire
[221,223], together with the transport of Lyman radiation in the
process depends on the geometry, material composition of the
divertor [224], which will eventually determine the ionization
plasma facing components and the plasma parameters in front
balance of the divertor in ITER and its performance.
of the surfaces.
The ITER divertor is designed to retain the emitted carbon
2.4. Material migration near the strike points in a partially detached plasma regime
Material migration studies include erosion of the plasma although this cannot be tested experimentally in the current
facing components (section 2.4.1), core contamination by machines. Further, the ITER material combination has yet
that material (section 2.4.2) and migration to other locations to be used in existing experiments. Therefore projections
(section 2.4.3) where a separate issue is the entrapment of to ITER are uncertain and more dedicated experiments are
tritium in co-deposited layers (section 2.6). The experimental needed [275].
data base on material migration in tokamak devices has been (b) Main-chamber processes. In the main chamber,
extended since the publication of the 1999 ITER Physics Basis impurities are released from the walls and divertor baffles as
study [1]. Major developments include the discoveries of the neutrals and also enter the main chamber as ions from the
following. divertor. The most basic and unavoidable impurity source
in the main chamber is due to the impact of neutral atoms
(1) Re-deposited carbon films with large tritium retention created from the charge exchange of hydrogen neutrals in
(section 2.6) following the 1997 JET deuterium–tritium the core. Another source of impurities can be created by
experiment campaign [272], ion bombardment. Radial transport of ions (section 2.1.3)
(2) strong flows in the plasma boundary, and/or ELMs (section 2.2.3) can be significant. Both enhance
(3) high re-erosion of hydrocarbon deposits and the erosion of the main-chamber walls [285] and protection
(4) initial results of migration studies using tracer methods. limiters in the main chamber [286].
At present, the basic mechanisms behind these Neutral impurities are ionized in the SOL some distance
processes have not been understood sufficiently for confident from the separatrix. Direct impurity ionization in the core
extrapolation to ITER [273, 274]. A comprehensive 2001 plasma is only observed in lower density plasma regimes.
review [275] of plasma–wall interaction physics dealt also with Impurity ions can diffuse in the edge plasma radially inwards
material migration issues. This section outlines the major to the core, or radially outwards to the wall. They experience
S224
parallel transport along the field lines to the inner or outer (2) The DIII-D carbon erosion at the outer strike point
divertor legs as well. Complicating the motion of the impurity was attributed to physical sputtering by WBC/REDEP
ions is the balance of different parallel forces acting on the simulations [295]. However, a 50% difference
ions, which can create large SOL flows (section 2.1.4). between experimental findings and simulation, which
also underestimated the erosion by a factor-of-five at
2.4.1. Intrinsic impurity sources. The dominant distances of 5–10 cm from the strike point [296] remains
impurity source mechanisms depend upon the material used unexplained.
(section 2.5) and on the coatings and material mixes due to re- (3) Recent JET outer strike point erosion rates [297] are orders
deposition (section 2.5.6). Typically, the dominant impurity of magnitude below earlier results (e.g. figure 4 of [296]),
sources originate at the outer divertor target and the main- and the cause of the difference has not been addressed in
chamber walls and limiters. The basic impurity generation the literature.
mechanisms [275] are the following. (4) Current thinking is that the net erosion rate experienced
(1) Sputtering of the main-chamber walls by neutral particle at the outer strike point depends upon the carbon influx
bombardment. These sources can peak in the vicinity from the main chamber. In fact, if the outer strike point is
of gas or neutral beam injection ports [287]. Chemical detached, then erosion rates drop [107] and the area often
sputtering of carbon walls is usually larger than physical becomes a region of net deposition. This phenomenon has
sputtering [276]. not yet been modelled.
(2) Sputtering of the main-chamber walls by steady state
In contrast to sputtering, macroscopic evaporation,
(section 2.1.3) or ELM-related (sections 2.2.3) ion
sublimation and melt layer losses due to heat load
bombardment. These sources may be concentrated
transients have not been well-characterized in present fusion
near the main-chamber components such as protection
experiments. However, such processes are a concern for
limiters [286] and at the baffles near the divertor entrance.
ITER [274] due to the expected larger ELM energy loss
(3) Increased sputtering of RF antennas and protection
(section 2.2).
limiters by energetic ions due to acceleration in RF
rectified sheaths [106, 288].
(4) Divertor targets experience the dominant power loading 2.4.2. Core contamination. Material eroded from the main
and particle flux, which can lead to strong erosion by wall and divertor enters the main-chamber SOL plasma and
sputtering but also to macroscopic erosion processes (melt has a finite probability of being transported into the core
layer losses, evaporation and sublimation due to heat plasma before it is re-deposited. This probability depends on
load transients). For target plates manufactured from parameters such as impurity species, source location, edge and
carbon based materials, chemical sputtering due to D/T, core plasma parameters and cross field transport processes.
oxygen and carbon is an additional important erosion At the present time, the relative contamination caused by
mechanism [289, 290]. the divertor and main-chamber impurity sources can only be
(5) Devices with high-Z target materials often report inferred indirectly from measurements of fuelling efficiency in
impurity-induced physical sputtering as the dominant the two regions. Gross divertor sources can be larger than wall
cause of material erosion [278, 283, 291]. For high-Z sources, but the probability of reaching the core is lower.
materials, self-sputtering can contribute significantly to The core contamination on DIII-D was attributed to the
impurity induced sputtering processes [292]. main chamber as evidenced by the decade long reduction of
(6) Unipolar arcing and ion bombardment can occur on start- the divertor carbon light emission, without a corresponding
up limiters and regions of plasma contact and lead to reduction of the core carbon content [298]. A similar
significant erosion along the arc track [293]. There are no de-correlation of Mo sources and contamination was observed
quantitative estimates on the contribution of this process to in C-Mod [106]. In JET [299] and ASDEX-Upgrade [300],
the overall material migration because there are no reliable when the carbon divertor was replaced by a beryllium and
experimental data on the discharge conditions leading to tungsten divertor, respectively, the change did not affect the
arcing or on the frequency of arc occurrence. carbon core concentration. However, recent spectroscopic
Some results challenge our understanding and, hence, current analysis with improved diagnostic capability at DIII-D
models, for example the following. indicates that additional sources exist at the divertor baffle
(1) Molybdenum erosion at the C-Mod outer strike point regions [301]. Similarly, correlations of the temporal evolution
was attributed to boron-induced physical sputtering, even of the molybdenum core contamination with Mo sources in
though little boron was observed at the erosion region Alcator C-Mod [291] showed that for ICRF-heated cases the
(figure 4 of [279]). Similarly, Tungsten erosion in the main-chamber sources were dominant.
ASDEX-Upgrade divertor I was attributed to carbon-
induced sputtering [278]. In the inner divertor, the Measurements of impurity fuelling efficiency. Several
tungsten became entirely covered by a carbon layer in experiments have quantified impurity fuelling efficiencies
contrast to the outer divertor where the strike point [288, 291, 299, 302, 303]. Typically 5% of the carbon injected
remained metallic tungsten [294]. Models for these as methane at either the JET outer mid-plane or vessel top
synergistic erosion processes can qualitatively explain the reached the separatrix. Slightly smaller values are measured
transition between continuous erosion and formation of by laser ablation for tungsten for source locations both at the
protective layers, however, only for a narrow parameter high- and low-field side of the plasma [304]. For tungsten,
range and limited material combinations. the comparatively small ionization length means that tungsten
S225
A. Loarte et al
ions need to diffuse a longer radial distance across the magnetic the observed erosion asymmetry between divertor plates could
field in order to contaminate the core plasma. also result from differences in the balance of erosion and re-
From the divertor region, a smaller fraction, typically deposition. The inner divertor plasma is often detached and
of the order of 1% reaches the separatrix. For tungsten, generally cooler than the outer divertor. Therefore, the re-
this fraction is smaller due to larger prompt re-deposition erosion of deposited material might be different in the inner
at higher Z [305]. Both Monte-Carlo transport modelling divertor. A similar explanation was reported from EDDY
(DIVIMP [302]) and fluid modelling (UEDGE [306]) show modelling of JT-60U erosion/deposition [315]. A difference
that impurities escape the divertor as ions accelerated by in the inner and outer target temperatures due to the different
the thermal force. Core contamination in both models is power loading induced changes in the chemical sputtering rates
simulated by assuming ad-hoc cross-field diffusion coefficients so that the inner target was a deposition region even though
(section 3.6). the model did not have a strong SOL flow. Measurements of
In ITER, impurities will be ionized further from the tungsten deposition [293] show smaller differences between
separatrix potentially allowing more time to flow to the inner and outer divertor compared with results for carbon [308].
divertor as opposed to diffusing towards the separatrix. For This can be attributed to the small re-erosion rate of
accurate predictions, the scaling of cross-field diffusion and tungsten even in the outer divertor where carbon would be
turbulent intermittent transport in the SOL will need to be fully re-eroded.
characterized. Attempts to establish a total balance of erosion and
re-deposition have been pursued at JET for carbon [316] and
2.4.3. Impurity migration. Eroded wall material will at ASDEX-Upgrade for tungsten [283]. The overall carbon
generally not be redeposited locally but will migrate to deposition in the JET Mk IIA and Mk IIGB configuration
other locations due to long range plasma transport processes. during their divertor operation time of 19 and 16 operation
Formation of redeposited layers due to such migration hours is estimated to be about 1000 g in Mk IIA and 500–600 g
processes is observed in all divertor tokamaks but clear in Mk IIGB. This is in good agreement with the amount eroded
examples of carbon deposition have also been observed in from the first wall derived from spectroscopic data [316, 317].
limiter machines such as in the long duration Tore Supra The carbon deposition rates are similar for both divertor
discharges where co-deposited layers grew with rates of up configurations with consequently similar fuel retention by co-
to 20 nm s−1 [307]. The particular case of hydrocarbons, deposition; however, the co-deposition occurs preferentially
which can locally migrate into castellations, to sides of tiles on different locations, namely the louvre area (Mk IIA) and
and to other remote areas of the divertor is discussed in surfaces in the private flux region excluding the dome (Mk
section 2.5.2.4. IIGB). For beryllium in JET, a similar analysis indicates good
The main chamber is a zone of net erosion [272,279,308], agreement of total erosion and total deposition [317]. Tungsten
while the inner divertor target area is a zone of net deposition. migration from the main-chamber walls to the divertor was
Wall material is observed to preferentially migrate to the investigated in ASDEX-Upgrade. A 10% fraction of the
eroded tungsten reaches the divertor target area [283]. The
inner divertor with examples such as boron in C-Mod [279],
tungsten is mainly eroded during limiter configuration start-up
beryllium in JET [272, 309], 13 C from methane puffing in
and ramp-down discharge phases. DIVIMP modelling showed
JET [310]. However, migration to both the inner and outer
that most of the eroded tungsten is expected to be re-deposited
target is found in ASDEX-Upgrade for tungsten [293] and
at the inner main-chamber wall [292].
for 13 C from methane puffing at the outer mid-plane [311].
13 ELMs could influence material migration in burning
C methane techniques have promise since fixed plasma
plasma experiments. Presently, erosion/deposition and
and divertor conditions can be maintained, and quantitative
material flow measurements are difficult to resolve on an
migration can be determined, thus aiding the modelling.
ELM time scale. Some ELM phenomena have been observed
However, results are just now being analysed and the use of
including the following.
such data to benchmark codes for projections to ITER has not
yet been done. • Increased energy/particle fluxes causing increased tran-
The outer divertor target was a region of net erosion sient erosion of the divertor target and main-chamber
in JT-60U [312], DIII-D [107], ASDEX-Upgrade divertor plasma facing components. The migration of the eroded
I [313] and C-Mod [279] and small net erosion/deposition material has yet to be quantified; however large ELMs on
on JET [297]. The outer strike point became a region of net JET enhanced material deposition to areas not in direct
deposition on DIII-D during the detached operation [107]. In plasma contact [318].
ASDEX-Upgrade the DIV IIb outer divertor baffle is also a • Deuteron (and possibly impurity) flows are modified by
region of erosion [308], but clear experimental evidence does ELMs. Expulsion of impurities from the confined plasma
not exist for the strike point zone. into the SOL occurs at ELMs [319]. Also, flow changes
Directed SOL flows have been observed [111, 114, 314] occur such as flow reversal near the divertor entrance
and are presumed to influence the preferential deposition at following ELMs [320].
the inner target. JET reverse field experiments reduced the
SOL flow, creating a stagnation point at the vessel top [27]. 2.4.4. Extrapolation of present results to ITER and open
Infrared measurements indicated that surface layers were built issues. There are several steps to impurity behaviour—
up on the outer target during reversed field operation pointing erosion, migration and deposition. At each step, attempts
towards increased deposition there, reversing roles with the have been made to characterize current experience and use
inner divertor (forward field). Apart from the SOL flows, that information to predict ITER operation. Modelling
S226
efforts estimated divertor strike point net C erosion rates disruptions. Present knowledge favours metal PFC materials
which vary with values of 6.4 [273], 10 [321], 20 nm s−1 on the former issue and carbon-based PFC materials on
net erosion [274], 30 [322] and 100 nm s−1 [312]. Clearly, the latter, although significant uncertainties remain in the
additional experiments aimed at testing the codes and the base extrapolation from present to next step-devices. As a
assumptions that are incorporated into them are needed. The result, the basis for the choice of the original PFC material
effects of the level of detachment and the mixed materials must distribution in ITER remains unchanged. Furthermore,
also be addressed. recent research has demonstrated the existence of long-range
The prediction of the main-chamber surface erosion is transport of materials in tokamak devices (see section 2.4),
based on charge exchange induced sputtering. The major which causes the formation of mixed material layers (Be,
uncertainties arise from the contribution of ion bombardment C and W in the case of ITER) adding further uncertainties
due to localized or turbulent processes (e.g. ELMs or blobs). to the extrapolability of present results to next step devices,
Corresponding measured erosion rates significantly exceed the which will have different PFC material composition (see
rates due to impact of charge exchange atoms. section 2.5.6).
In the area of transport (material migration), key issues Given the uncertainties in our knowledge of plasma–wall
are the modelling of flows, ion temperature gradient forces interactions for next-step-relevant conditions (in particular, of
and electric fields. Progress has been made in characterizing the interactions between the plasma and the main wall), what
SOL flows. However the magnitude and poloidal variation is cannot be ruled out is that the replacement of major plasma
not understood, so that projection to ITER is difficult. facing components and testing of different combinations of
The situation is similar with respect to deposition. While plasma facing materials will be necessary over the lifetime
a number of C PFC divertor tokamaks observe inner divertor of a device like ITER. This is foreseen in the ITER design
deposition, the detailed pattern in the inner divertor has not both for the divertor target and the main-chamber PFCs.
been simulated. Recent efforts using 13 C tracer studies, The ITER divertor targets are designed to be replaced by
analysis of un-wetted surfaces (e.g. tile sides) and comparison remote handling as part of the routine maintenance of the
with deposition of metallic impurities are aimed at providing device (remote handling shutdown for the replacement of
the needed benchmarking for models. all divertor cassettes of ∼3–6 months duration). On the
In summary, extensive experimental and modelling efforts other hand, the blanket modules, on which the main-chamber
have improved our knowledge of material migration processes. PFCs are mounted, are designed to be replaced by remote
However, the predictions for ITER erosion, optimal material handling in small numbers and only occasionally, following
selection, material migration and resultant tritium retention the initial requirements of the ITER design [324, 325].
are still uncertain. The development of new experimental While not presently a reference scenario, the entire first
measurements (e.g. 13 C and metallic tracers) and progress in wall replacement is, in principle, possible. However, its
modelling (e.g. flows) in both existing experiments and the cost, practical implementation and schedule require further
ITER non-nuclear phase are needed. assessment [326].
2.5. Divertor and main-chamber materials 2.5.2. Issues related to use of carbon containing materials
2.5.1. Expected requirements of materials. Plasma-facing 2.5.2.1. Operational experiences with carbon based devices.
component (PFC) materials used for the design of next-step DT Graphite is the most commonly used PFC material in the
burning plasma devices must meet several demands including present devices because of its good compatibility with a
resistance to damage by large steady-state heat loads, resilience wide variety of plasma operations. The present experience
to transient heating and MHD events, sufficient component with carbon allows for an optimistic extrapolation of its
lifetime against erosion by the plasma, and minimization of compatibility with thermonuclear plasmas in ITER: (a)
tritium retention in the vessel. The demands on PFC materials because of the good thermo-mechanical properties of carbon
in ITER will go far beyond those of the existing machines fibre composites (CFC) and the lack of melting under transient
because of simultaneous large increases in several parameters: power fluxes, the compatibility of these materials with high
power density, plasma stored energy, tritium throughput and power flux plasmas is optimal, (b) despite its relatively large
plasma duration. As a result of these strenuous demands, sputtering yield, high density plasmas in ITER-like regimes
issues arising from plasma–material interactions (PMI) can can be obtained in present experiments with a relatively low
limit the viability of the PFC components and, therefore, the core contamination (Zeff = 1.5–2.0) [327], and (c) due to
operational availability of future devices, as indicated from its low-Z, the core radiative losses associated with C in high
experience in present experiments reviewed in [275]. On the temperature plasmas are low, while the divertor radiative
other hand, significant progress has been made in producing losses by C, which are necessary to maintain semi-detached
actively cooled C or W clad high heat-flux components, which operation in ITER, can decrease the divertor power flux
reliably withstand thousands of thermal cycles up to levels of by more than a factor of ∼2 compared with fully attached
20 MW · m−2 , as required in ITER [323]. conditions [160, 328].
Recent research has identified two critical issues with The most serious issues with carbon PFCs are lifetime
respect to the behaviour of PFC materials in fusion devices: restrictions on net erosion areas due to the large erosion yield
(1) tritium retention rates in plasma-deposited films and of carbon, the associated build-up of carbon deposits and the
(2) lifetime/viability issues for materials undergoing ablation subsequent tritium retention in these layers [160,275], which is
or melting from transient heating events like ELMs and the main mechanism of long-term hydrogenic isotope retention
S227
A. Loarte et al
in carbon clad fusion devices. A tokamak demonstration of a

regime that maximizes the radiation losses at the divertor, while
gross erosion (m-2s-1)

1022
maintaining tritium retention at very low levels, as required in
ITER, remains to be carried out. A target value for the effective
(i.e. after T-removal techniques are applied) T retention for
1020
routine ITER operation is ∼0.1% of the injected T. The
required amount of T in ITER is estimated at ∼50 g/discharge Roth, Tmax ~ 800 oC
for the QDT = 10 reference scenario [329]. This is equivalent Roth, Tmax ~ 1700 oC
to a retention of ∼0.05 g per discharge, which would allow 1018
fixed yield (1%)
∼7000 full performance discharges to be carried out before a
T-removal shutdown is required. -0.2 0 0.2 0.4 0.6
distance along outer target from separatrix (m)
2.5.2.2. Erosion behaviour of carbon based wall materials.
While the mechanisms of physical sputtering are well Figure 19. Calculated chemical gross erosion of the outer ITER
understood, the chemical erosion of carbon, i.e. the release divertor target assuming a fixed erosion yield of 1% [335] or the full
erosion dependence on energy, surface temperature and particle
of volatile hydrocarbon molecules and radicals (e.g. CH3 , flux [277]. Calculations are done for a high and a low surface
CH4 , C2 H4 , etc) caused by ionic and atomic hydrogen temperature case [413] (after [334]).
fluxes, continues to be an active area of research. Chemical
erosion yields depend on a wide variety of conditions: target load and associated tritium retention which makes predictions
temperature (Tsurf ), incident particle energy, flux density i , to ITER uncertain.
plasma impurities, as well as on the surface conditions and
exposure history of the carbon material [275]. (a) The carbon chemical erosion yield depends strongly on
Chemical erosion yield is measured on carbon-clad the target temperature, which is determined by the incident
devices using visible spectroscopy of highly dissociated plasma power flux.
hydrocarbon molecules (CH, C2 ) interpreted through adequate (b) The power flux at the divertor is determined by the
modelling of the molecular processes [330]. The yields radiative losses and the degree of detachment, which
measured for attached (ionizing) divertor plasmas span a range is strongly influenced by the release of carbon into the
from 0.4–10% corresponding to a large range of experimental divertor by chemical erosion.
conditions with Tsurf = 350–600 K, plasma fluxes of i =
On the other hand, changes in the plasma conditions at
0.1–1.0 1023 m−2 s−1 and, possibly, the enhanced erosion of
the divertor from quiescent to transient phases of the discharge
co-deposited films under high power loads [81, 331–333].
(ELMs, disruptions) will most probably contribute more to
DIII-D reports a total suppression of hydrocarbon emission
carbon erosion than the uncertainties in the chemical erosion
in detached plasmas (Te < 2 eV), despite the modelled high
yield described above.
photon efficiencies for CH in these conditions, which suggests
a highly suppressed chemical erosion yield <0.01% [81]. The 2.5.2.3. Brittle destruction. New results have been obtained
very large scatter in experimental measurements indicates that on the erosion of isotropic graphite and carbon fibre
all plasma/surface parameters (such as ion impact energy, composites by explosive emission of small particulates
flux, surface temperature) play a major role in determining occurring under rapid (∼ms) heat loads, the so-called brittle
the magnitude of chemical erosion. The unprecedented high destruction [336], which leads to enhanced erosion and
ion and power flux densities near the ITER divertor strike- surface non-uniformities for carbon materials. This erosion
points, therefore, make the prediction of chemical erosion very mechanism has been observed to set in for graphite above an
complex. energy density threshold >100 MJ m−2 s−1/2 (which is more
A database of chemical erosion yield obtained in different than twice as large as the carbon sublimation limit) and at
tokamaks, linear plasma devices and beam experiments has somewhat lower values for CFC. CFC erosion associated
been recently compiled and used to normalize the yield data with local overheating and matrix-fibre detachment has been
with respect to ion energy, target temperature and ion flux. identified in experiments in electron beams/plasma guns and
The database suggests a weak dependence of the erosion modelling [337]. Therefore, while carbon is chosen for its
yield on ion impact energy between 10 and 200 eV and robustness in transient heating events like ELMs and disruption
a clear tendency for the yield to decrease with increasing in ITER, this resilience is limited, and energy densities from
flux density [277], implying a ∼2.5-fold reduction in yield transient events in ITER must be kept below the sublimation
for the ITER strike-point conditions compared with previous and brittle destruction thresholds for carbon in order to prevent
assumptions (∼1%). Figure 19 illustrates the implications a severe limitation of the ITER divertor target lifetime even
that such a yield reduction (taking into account realistic for CFCs.
surface temperature, impact energy and flux dependences)
has on the calculated gross erosion of the ITER divertor 2.5.2.4. Location and properties of carbon film deposits.
target [334] compared with previous estimates with a ∼1% Recent experiments show that nearly all the carbon films
fixed chemical erosion yield [335]. The reduction of the formed by chemical erosion are deposited in the line-of-
chemical gross erosion yield results in similar improvements sight of high sticking probability Cx Hy fragments [308]. In
in target lifetime and expected tritium retention. This figure general, a strong temperature dependence of the deposition is
illustrates the complex coupling of target erosion and power observed, which is due to the strong temperature dependence
S228
of the re-erosion of deposited layers by atomic hydrogen, influence on oxygen diffusion and H trapping remains to
as demonstrated in laboratory experiments [338]. This has be done. Furthermore, a mechanism which enhances the
important implications for the recovery of the tritium retained erosion of Be (and all metals) at high temperatures has
on deposited hydrocarbon films. Access to these films is been identified [353, 354]. This mechanism is similar to
generally easier if they are deposited on surfaces close to the radiation-enhanced sublimation of carbon and leads to
the region of interaction between the plasma and the material erosion rates which are one order of magnitude larger than
(where the material is subject to both hydrogen atomic fluxes physical sputtering (Eion ∼ 50 eV) for Tsurf ∼ 1200 K. This
and power fluxes and is, therefore, at high temperatures) than enhancement would be particularly important if significant
if they are deposited in remote and/on shadowed areas such energy fluxes reach the ITER first wall during transients such
as tile gaps (see section 2.6 for more details). The amount of as ELMs and make the prediction of the erosion of the ITER
hydrocarbons deposited in very remote areas, such as divertor main wall very uncertain for such conditions.
pump ducts is very small (typically less than 1% of carbon The major issue concerning the use of Be as plasma facing
films) [81, 339–341] while the amount of hydrocarbons that material in next step devices, such as ITER, is related to its
can deposit in shadowed areas close to the divertor can be very relatively low melting temperature (1550 K). Deposition of
large [297]. plasma energy during transients such as ELMs, VDEs and
The structure and hydrogenic content of carbon deposits disruptions can bring the Be temperature to the melting value
is highly dependent on its location. Deposits on plasma- and, thus, cause extended melting of PFCs. Furthermore,
receiving surfaces are hard and brittle, while deposits techniques that mitigate the effect of such events, such as
elsewhere are softer and exhibit columnar structures [342]. massive impurity puffing to mitigate the thermal quench energy
Various deuterium fractions in the carbon deposits can be flux during disruptions [355], cause radiative fluxes leading to a
found. Usually, areas that are subject to large temperature deposition of energy of ∼0.5 MJ m−2 in timescales shorter than
excursions during operation show very low hydrogen 1 ms, causing significant melting of the Be PFCs (20–50 µm).
concentrations (10−3 to 10−2 ) [321, 341, 343] consistent with The dynamics and loss of the layers of molten Be during
hydrogen desorption during these high temperature transients. transients, which are subject to forces associated with plasma
On the contrary, shadowed areas of plasma facing components, impact, Lorentz forces, etc, remain to be studied in detail. The
such as those in TFTR [344] and JET [345], can lead to experience with Be in JET was recently reviewed [356] with
the deposition of hydrocarbons with a large hydrogen content emphasis on the behaviour and influence on the plasma of Be
(much larger than 10−1 ). The contribution of the hydrocarbon melting and molten Be-PFCs. In JET, a single large ELM
deposition in tile gaps to the observed total fuel retention depositing ∼1 MJ in less than 1 ms [357], led to substantial
remains an active field of research of critical importance for melting of the Be JET divertor target and terminated the high
the design of the ITER divertor target. performance ELM-free phase of the discharge. However, the
effects of Be melting on subsequent plasma operations were
2.5.3. Beryllium. Beryllium (Be) has two major advantages not catastrophic, due to the plasma’s high tolerance to the low-
for use as main wall plasma facing material (in areas of low Z Be. Therefore, the precise implications of significant Be
power flux) in fusion devices: (a) it is a low-Z material and, first-wall melting on the lifetime of the wall in ITER, and on
as its sputtering yield is not higher than that of carbon at low the performance of a burning plasma experiment like ITER,
plasma temperatures [346], it leads to low radiative losses from are still not clear and require further study.
the core plasma and (b) it provides gettering for the residual
oxygen from the base vacuum. Both these advantages have 2.5.4. High-Z materials. For general issues on the use
been demonstrated at JET in which both Be evaporation [347] of high-Z plasma facing components in the present fusion
and solid Be plasma facing components [348] have been used. devices and on the development of these materials for fusion
The plasma interaction behaviour of beryllium is affected applications, the reader is referred to recent reviews on these
by the presence of oxygen and other impurities. Oxygen forms topics [358, 359].
a BeO2 layer on PFCs that reduces the physical sputtering
yield with respect to the pure material. On the other hand, the 2.5.4.1. Behaviour of high-Z plasma facing components
presence of oxygen in Be increases the retention of hydrogen, in tokamak experiments. The major focus of research on
which for the pure material is very small [349], to similar high-Z materials in fusion devices has been the achievement
levels of hydrocarbon deposits (H/Be ∼ 0.3) [350, 351]. of low core plasma contamination, due to the low ‘fatal’
However, new results from experiments in PISCES-B have core concentrations (for ignition 10−3 –10−4 for Mo and
produced oxygen-containing Be deposits (O/Be ∼ 0.03– W, respectively) in burning plasma experiments. The
0.3) with low hydrogen retention (H/Be ∼ 0.01–0.03) for core concentration of high-Z impurities is controlled by
surface temperatures of 400–600 K [352], indicating that both the impurity physical sputtering source at PFCs (see
besides oxygen other impurities may be responsible for the section 2.5.4.2) and impurity transport. High-Z materials
large hydrogen levels observed in some oxidized Be samples. have been used for the majority of PFCs in several
Besides this, the temperature for the release of hydrogen from large fusion experiments: FTU with molybdenum toroidal
deposited Be layers is considerably lower than from carbon limiters [360], Alcator C-Mod with a molybdenum first wall
layers. On the basis of these results, retention of hydrogenic and divertor [361], TEXTOR with high-Z components as
fuel in Be is expected to be very small when compared limiters [362] and ASDEX-Upgrade both at the divertor [363]
with carbon, although a complete assessment including the and main chamber [364]. Recent experimental results from
effects on temperature excursions of the PFCs and their these devices show that acceptable core impurity fractions
S229
A. Loarte et al
can be obtained by adequate plasma operation: high plasma further the ELM frequency [364]. This negative feedback loop
density, central heating and magnetic divertor geometry help can be easily broken either by increasing the level of additional
to control the overall core impurity levels and central impurity heating, which increases the ELM frequency, or by triggering
peaking in high performance plasmas. These results have ELMs with pellets [364]. In Alcator C-Mod the molybdenum
increased the confidence for further possible use of high- level is most clearly linked to the confinement mode (ELM-free
Z PFCs in future burning plasma experiments and fusion operation leads to much higher Mo concentrations that EDA
reactors. mode [365]) and the quality of the boronized layer. As the
In general, high-Z core concentrations decrease with boronization layer wears off through operation, the Mo levels
increasing density [362, 365, 366], as expected from the increase [291].
reduction of physical sputtering with edge temperature, and Even with significant surface area coverage by high-Z
are lower (by factors of 2.5–10) in the divertor than in limiter materials, plasmas in present devices are often dominated
geometries [291, 365]. Experience in ASDEX-Upgrade has by the more easily eroded low-Z impurities. These are
shown that it is possible to operate the device in a large range either introduced through conditioning techniques or originate
of conditions while maintaining the core W concentration from other low-Z PFCs in the device. For instance, light
in a range of 10−6 –3 × 10−5 both with W as divertor and intrinsic impurities in ASDEX-Upgrade have not changed
main-chamber material [364]. On the other hand Alcator significantly despite the exchange of 40% of the plasma facing
C-Mod, with complete molybdenum PFCs, shows that the components from graphite to tungsten [364]. The experimental
molybdenum concentrations can reach values as high as 10−3 observations show that this is associated with a highly dynamic
in some operating conditions [365]. The different observations equilibrium of erosion and deposition on the W PFCs of carbon
in ASDEX-Upgrade and Alcator C-Mod may be attributable originating from other components [286]. Similarly, the level
to differences in high-Z first-wall coverage between the two of Mo in the core was shown to be dependent on how recently
devices. the surfaces were boronized [291]. Such results indicate the
Impurity transport, rather than the source, can play a inherent difficulties in the interpretation of experiments with a
dominant role in determining high-Z core concentrations, significant amount of mixed materials and their extrapolation
in agreement with expectations from neo-classical transport
to devices like ITER (see section 2.5.6).
theory (for example, see [367]). Large high-Z impurity
Significant melting of high-Z PFCs under transient loads
accumulation is observed in conditions that are favourable
such as ELMs and disruptions is a major concern for their
for this from neoclassical transport expectations, as seen in
use in next step devices like ITER, but it is not expected in
TEXTOR [368] and FTU [369]. Central high-Z accumulation
present tokamaks due to the lower energy densities that can
leads to strong central radiation, hollow temperature profiles
be achieved. Most of the melting of high-Z PFCs observed in
and can, eventually, lead to the termination of the discharge.
present devices is due to poor thermal contact between high-Z
Studies in ASDEX-Upgrade have shown that the central W
coatings and the substrate, tile misalignments and/or runaway
concentration is dominated by impurity transport in the bulk
plasma with the lowest W concentrations being found for electrons [375]. Energy densities during ITER transients
discharges at high density with flat density profiles and high may cause significant melting of the high-Z PFCs [166, 376].
levels of additional heating [152]. Discharges with peaked The dynamics of the molten layer and the effect of repetitive
density profiles showed, on the contrary, increased core W melting under ELMs and disruptions both on the lifetime of
concentrations and even accumulation [152]. This problem is the PFCs and on ITER plasma operation remain open and they
exacerbated for discharges with reduced core transport, such are, presently, an active field of research [161].
as with ITBs [370].
Because the concentration of high-Z impurities is 2.5.4.2. Erosion of high-Z materials in a tokamak
dominated by transport, it is possible to affect the core levels of environment. Due to the large sputtering thresholds
impurities by applying central heating through the coupling of (∼209 eV for D on W), the erosion of W PFCs by a low
energy and particle transport in many tokamak regimes. This temperature hydrogenic plasma (Te < 20 eV) during quiescent
technique has been demonstrated in TEXTOR [371], Alcator phases of the discharges, particularly for a detached divertor
C-Mod for discharges with ITBs [151] and studied in detail in plasma, will be negligible. No substantial chemical erosion
ASDEX-Upgrade with ICRH and ECRH [372]. The reduction by volatile W-oxides has been measured either, in line with
of the central concentration of W can be explained by the recent molecular dynamic calculations where chemical erosion
changes caused by the central heating in the plasma particle of W by O was found to be negligible in the relevant parameter
transport [372, 373]. On the basis of these physics arguments, range [377].
the present modelling [374] suggests that no W accumulation The major source of sputtering of high-Z materials are
is expected in the ITER reference QDT = 10 scenario. For impurity ions in the background plasma. These may be
regimes with internal transport barriers, no detailed modelling intrinsic or seeded, in order to provide radiative cooling of
has been carried out for ITER conditions but it is expected that the divertor plasma and lead to an ‘effective sputtering yield’
the accumulation of W in the core of these plasmas would lead of W under D-plasma impact (Yeff W
= W / D ), which is
to more serious problems [370]. much larger than that due to the impact of the D ionic
The operating conditions of the device can affect the core flow. Impurities with higher charge states gain additional
impurity content. For ASDEX-Upgrade, which is gradually energy through the sheath potential at targets and dominate
switching to W walls, the clearest influence of W is seen for the effective physical sputtering yield of high-Z materials.
discharges with low levels of input power, in which long ELM- Modelling of Alcator C-Mod shows that the measured Mo
free periods can lead to increased core radiation decreasing influx can only be reproduced if the sputtering by ∼2% B3+
S230
10-1 the melt layer dynamics and prevent its loss under ELM-like
loads [161]. Nevertheless, due to the potential implications for
ITER, the control of ELM and disruption energy fluxes remains
10-2
TEXTOR a high priority research issue for the use of metal PFCs in ITER.
Sputtering yield
Limiter
10-3 2.5.4.3. Hydrogen retention and blister/bubble-formation
in tungsten. In general, the retention of hydrogen in W
is very small [381], although the measured values of the
10-4
retention depend considerably on material grade, fabrication
process, temperature of the W and the presence of carbon and
10-5 oxygen [382]. In TEXTOR, pure W surfaces had D inventories
ASDEX Upgrade
<1019 D m−2 , whereas co-deposited layers of B and C on
Divertor tungsten revealed D-contents higher by more than two orders
10-6
0 20 40 60 80 100 of magnitude [383]. These results indicate that the use of
Te (eV) mixed materials as PFCs in ITER adds significant complexity
to understanding tritium retention with high-Z materials, since
Figure 20. Effective W erosion yield as a function of the local deposit layers are typically dominated by the low-Z plasma
plasma temperature at TEXTOR test limiters and the constituent impurities (see section 2.5.6).
ASDEX-Upgrade outer divertor target (after [358]).
The formation of gaseous blisters/bubbles in high-Z
materials has been studied because of its potential in (a)
(from boronization) and Mo self-sputtering are taken into increasing the amount of T retention by high-Z materials
account [378], while in RF heated discharges localized Mo and (b) in causing surface damage and enhanced erosion of
sputtering near RF antennas was linked to high ion impact these materials. Hydrogen blisters/bubbles are formed in W
energies due to sheath rectification [291]. In ASDEX-Upgrade during large fluence irradiation experiments [384]. This can
(see figure 20), divertor W release is dominated by low- be effectively suppressed either by adequate surface treatment
Z impurity impact and erosion results could be modelled of the plasma facing materials [385, 386] or by adequate
taking into account sputtering and deposition from 1% carbon fabrication of the materials themselves [387]. Helium bubbles
impurities [305]. The higher effective W yields in TEXTOR and holes can also be formed in W under high fluence
(figure 20) at similar plasma temperatures are attributed to He plasma irradiation, even at a relatively high surface
larger fractions of C and O impurity fluxes in these plasmas temperature [388], which could be a critical issue for the
(factors of 2–5) [358] and to higher W self-sputtering due to erosion of W. Both processes can, in principle, increase
higher W concentration in TEXTOR. ASDEX-Upgrade has the hydrogen retention in W and may give rise to an
shown that the measured inner limiter W erosion exceeds initial surface damage (micro cracks), which may then be
the expected values from erosion by charge-exchange (CX) enhanced by overheating under plasma exposure. Despite the
particles [379] by at least an order of magnitude. This is possible enhancement with respect to its base-level retention,
mostly attributed to direct plasma contact during plasma ramp- T retention in W is expected to be very low in ITER, in
up and ramp-down, with an inner-wall limited phase [380], but comparison with that associated with the use of a carbon-based
continuous erosion during the divertor phase due to impurity material, because of the intrinsic low T retention of W (even
ion sputtering contributes significantly as well and, therefore, with blisters/bubbles).
should be taken into account when estimating the erosion of
Although the detailed conditions and mechanisms that
the ITER main chamber.
lead to bubble formation in laboratory experiments have been
Net erosion of high-Z materials is reduced by prompt re-
studied in detail, no bubble formation, either from hydrogen
deposition, which is strong due to the large mass and short
or from helium bombardment, has ever been observed in
ionization mean free path into the plasma [280]. In Alcator
present tokamaks. This is probably associated with the small
C-Mod, the net erosion of Mo outer strike point region is only
integrated fluences on W PFCs that can be achieved in present
∼0.1 nm s−1 [279], which is a factor of 10–100 lower than in
experiments compared with ITER and, possibly, the low He
the C-clad DIII-D divertor [295].
concentrations in these plasmas. Direct extrapolation of the
A major uncertainty in the erosion of high-Z material
results obtained in present tokamak experiments indicates that
in next step devices is their behaviour under transient loads
bubble/blister formation in the ITER W PFCs may not be a
such as ELMs and disruptions, which can cause the high-Z
dominant process for ITER conditions. However, due to the
material to melt [166, 376]. Significant displacement of this
uncertainties with respect to fluences and He concentrations
molten layer and its splashing could cause losses of several
described above, a final quantitative assessment of these effects
micrometres of material per event, which is much larger than
for ITER requires further experimental studies in present
the erosion during quiescent phases of the discharge and can
tokamaks and laboratory experiments for the various W grades
lead to further roughening of the surface, causing the PFCs
foreseen for ITER PFCs.
to be more prone to further melting by the power fluxes
during the steady-state phases of the discharge. The detailed
processes that are involved in these phenomena are complex, 2.5.5. Wall conditioning and particle control. In next-
but initial results in laboratory experiments have demonstrated step superconducting devices like ITER, the tasks of wall
that the detailed PFC design can have a deep influence over conditioning are: (i) commissioning before operation after
S231
A. Loarte et al
major openings, (ii) limiting release of hydrogen and non- control the interaction among the materials and of those with
hydrogenic gas impurities during plasma operation and (iii) the plasma is extremely large and, therefore, our level of
the limitation of the tritium inventory in the plasma facing knowledge in this field and our predictions of the dominant
components. The first task is done with no magnetic field, phenomenology in next-step devices are very primitive at this
and most of the techniques used in present day tokamaks can stage.
be used (see [1]). The last two must be done primarily with
the toroidal field coils energized, which rules out many of the 2.5.6.1. Formation of mixed materials. The carbon–metal
standard conditioning techniques used in today’s experiments. interactions are governed both by carbide formation reactions,
In this section we concentrate on the advances in the last few carbon diffusion into the bulk metal and by the bond type of
years of conditioning techniques for task (ii), while tritium the formed carbide. Transition metals (e.g. W, Ti, Ta, Fe, Ni)
removal techniques are discussed in section 2.6. form metallic carbides, where the formation of carbides occurs
above a material-specific threshold temperature and is followed
2.5.5.1. Conditioning with permanent magnetic field. In by strong carbon diffusion into the metal bulk at elevated
the presence of a permanent magnetic field, RF assisted temperatures. In contrast to these metallic carbides, ionic or
discharge cleaning is the most suitable technique available, covalent carbides (e.g. Be2 C, SiC) lack the diffusive carbon
and experiments have been reported from different tokamaks, loss process. After a temperature threshold, carbide formation
using ECR–DC (electron cyclotron resonance–DC) [389,390] proceeds until the available carbon amount has reacted and
or ICR-DC (ion cyclotron resonance–DC) [391–393] and from the stoichiometric carbide is formed [397]. Ternary systems
the stellarator W7-AS (ICR–DC [394]). These techniques have (i.e. with three material species) pose additional difficulties,
recently been reviewed in [395] where the reader is referred to especially if oxygen is one component. Due to the formation
for more detailed information. Two critical parameters need of volatile species, predictions of the ternary system behaviour,
to be adjusted for the optimization of the conditioning: (a) the e.g. with increasing temperatures, are difficult. For example,
homogeneity of the discharge as achieved in HT-7 [393] and in the ternary system W–C–O, the erosion of carbon films at
(b) for ICR–DC, optimizing the wave heating so that energetic elevated temperatures even depends on the type of oxide [398].
particle bombardment of the first wall occurs, while still Beryllium and tungsten form several alloys, with some
achieving a low density, low temperature plasma to avoid re- alloys exhibiting enriched Be concentrations (e.g. Be12 W) and
ionization and re-deposition of wall-desorbed neutrals before melting temperature similar to Be [399], as reported recently
they are exhausted by the pumping system. Typical values are from experiments in PISCES-B [400, 401]. However, these
1016 m−3 < ne < 1017 m−3 and 8 < Te < 40 eV, but with ion Be-rich phases were not observed when evaporating Be onto
energies of the order of a few keV, allowing isotope exchange W surfaces and investigating the inter-metallic phases using
up to several 100 nms in carbon [391, 393]. The required x-ray photoelectron spectroscopy. Upon heating, a < 2 nm
RF power varies between ten and hundreds of kW [391–393] interfacial layer reacted to Be2 W while for thicker deposits the
in present size devices. The usual sequence consists of first remaining Be layer evaporated [402].
using D2 as the working gas for wall cleaning and isotope It is unclear whether the conditions expected in ITER will
exchange, and then He for deuterium removal [391, 393]. lead to the formation of such alloys between the Be eroded at
The overall efficiency of RF–DC is shown to be comparable the wall that deposits on the divertor and the W divertor PFCs,
with that of GDC (glow discharge cleaning) in the stellarator leading to the W PFCs losing some of their structural and
W7–AS [394]. thermomechanical properties. The formation of such alloys
Very recently, in HT-7 a carbon removal and surface would be more probable for the case of a full W target rather
conditioning campaign was run using He/O–ICRH discharges, than if W PFCs are limited to the baffle and dome region, as
varying the oxygen partial pressure and the ICRH power in the reference ITER start-up PFC material distribution, in
at continuous toroidal field. Carbon removal rates were which the surface temperatures are lower. The formation of
determined by residual mass analysis of the exhaust gas and Be/W alloys thus remains a concern for ITER and is presently
impressive cleaning rates could be extrapolated to ITER [396]. under investigation.
2.5.6. Implications on the use of multiple PFC materials. 2.5.6.2. Effects of material mixing on erosion. The
The present design for ITER foresees the use of three plasma simultaneous impact of plasma ions with impurity ions eroded
facing materials: Be for the first wall (700 m2 ), W for the from other plasma facing materials can lead to both enhanced
divertor and baffle region (100 m2 ) and CFC for the divertor and reduced erosion. As an example, the erosion of W by C/D
strike point tiles (50 m2 ). The choice of the different materials impact was investigated [403]. The erosion of W in areas of
is motivated by the optimization of plasma–wall interaction low plasma temperatures is only possible by sputtering due to
processes for the pure materials. Cross contamination of the impurity ions. The sputtering threshold for D/T ions is of the
material surfaces will be unavoidable. Therefore, the plasma– order of 200 eV, while it is reduced for typical impurity ions
wall processes have to be evaluated for the case of mixed to less than 30 eV (see section 2.5.4.2) [404]. Therefore, the
materials. There is no fusion device to date using the same erosion will depend strongly on the concentration of impurity
material mixture as foreseen for ITER. Limited experience ions in the incident plasma. For low impurity concentration,
exists for the simultaneous use of different materials, e.g. where the incident flux of deuterium keeps the W surface clean,
JET for C and Be and ASDEX-Upgrade for C and W. the erosion rate of W is proportional to the concentration of
Laboratory studies can cover all three possible combinations, carbon ions. For high carbon concentrations the implanted
in principle. However, the number of parameters which carbon atoms are not completely removed and a protective
S232
retains D at similar concentrations as carbon [350, 351], with

desorption starting at 600 K, while in clean Be the retention is
0.01 strongly reduced [349] and the release temperatures are around
[µg/1016]
400 K. As described in section 2.5.3, recent data from PISCES-

B indicates that the Be/D deposition rate and the presence of
0.00 other impurities may be additional controlling parameters to
dW/dΦ in Equilibrium
the D retention in Be with impurities. A C–Be mixed-material

-0.01 layer was formed containing at least an order of magnitude
more deuterium at room temperature than pure Be [400], but
this additional amount of D decreased and vanished at 500 ◦ C.
-0.02
From the examples above, it is clear that the formation
10 of mixed materials have significant effects on the retention
-0.03 8 of hydrogenic isotopes either increasing or decreasing it with
n
tio
ra
6 respect to the pure materials. Determining which will be the
nt
20
ce
Plas 40 4 dominant phenomena in ITER is still far beyond the present
on
ma T 60
emp
-c
eratu understanding of the processes involved and, because of its
C
80 2
re Te
l.
re
[eV] 100 importance in ITER operation, remains an active field of

research.
Figure 21. Dependence of tungsten (W) erosion on plasma electron
temperature and on the carbon concentration (%) in the incident D 2.5.6.4. Effects on tritium removal techniques (oxidation or
plasma [403]. Negative values of dW/d denote net W erosion.
heating). Doped graphites and carbon layers show in many
cases drastically reduced chemical erosion properties, either
layer will form, which suppresses W erosion. The maximum due to their influence on the hydrogen recombinative release
effective erosion yield will be obtained for the C/W system or due to enrichment of the impurity level at the surface, leading
at a carbon concentration in the plasma of about 4% (see to protective surface layers. This can affect not only its erosion
figure 21 [403]), which is a typical concentration of carbon behaviour and its hydrogen retention but also the effectiveness
in tokamaks with carbon PFC’s. At elevated temperatures, of techniques which can be applied successfully to remove
enhanced erosion processes and bulk diffusion may alter the tritium from pure materials. One technique for tritium removal
detailed balance of erosion/deposition [405] leading to higher relies on the chemical erosion of carbon deposits due to
erosion yields. oxidation [411]. Similar to hydrogen, the chemical reactivity
Carbon chemical erosion can be drastically reduced by with oxygen can be reduced by the presence of impurities in C,
impurities on the surface [406] and in the bulk [407]. Surface which makes the removal more difficult. For layers containing
impurities, such as incident Be ions on the ITER C target B due to boronization of the vessel walls in tokamaks, after an
can reduce the chemical erosion of carbon by coverage initial fast oxidative removal of D and C, much lower removal
effects [406,408]. Bulk impurities added on atomic scale, such rate levels were reached, the D being locked up by a stable
as Ti, W or V at concentrations of less than 10 at%, decrease boron layer in the material [412].
the activation energy for hydrogen recombinative release thus Although little is known of the effect of mixed materials
reducing chemical erosion at elevated temperatures [409]. on tritium removal techniques, it is clear that the presence
In general, the introduction of mixed materials in of mixed materials can considerably complicate the practical
fusion devices complicates enormously the erosion/deposition application of these techniques. It is, therefore, necessary
calculations, and the implications for the calculated erosion to proceed to a thorough investigation of tritium removal
rates in devices like ITER are just beginning to be addressed. techniques for the material mixes expected to be formed in
ITER so that a reliable estimate of the viability of such
2.5.6.3. Effects on H isotope retention. The hydrogen techniques can be obtained.
retention in plasma facing materials and in redeposited layers
is very sensitive to surface and bulk contamination of the
2.6. Tritium retention
material. Surface oxide or carbide layers at the surface may act
as diffusion barriers and strongly influence the recombination 2.6.1. Introduction. ITER will routinely use large amounts
necessary for molecular release of hydrogen. Since the of tritium as fuel (e.g. 100 times more in ITER than in present
number of parameters is large (multi-material mixing, material T-experiments), and control of the tritium inventory will be
structure, temperature, ion energy) no general statements can a key operational issue in any next-step experimental burning
be made. Both large increases and reductions of retention are plasma facility both for environmental and safety reasons. The
observed compared with pure materials and reliable estimates working guideline for the maximum in-vessel inventory of
of the situation in next step devices do not exist. Therefore, tritium that can be mobilized in the ITER vacuum vessel is
several examples of experimental results that may be relevant presently ∼350 g. The assumption made for safety analysis
for ITER are summarized here. Superficial oxide or C layers is ∼1 kg of tritium mobilizable in the vacuum vessel [134].
are seen to lead to the increase in hydrogen retention in Independent of safety considerations, tritium is limited in
W [382]. On the other hand, in W samples pre-bombarded with supply and, for this reason alone, as much as possible needs to
C-ions, a transition from retention similar to pure C, to WC be recovered and re-used during the operation of any next-step
and pure W was observed [410]. Similarly, fully oxidized Be device.
S233
A. Loarte et al
Table 2. Parameters of the balance of tritium for the experiments carried out in TFTR and JET [1, 416, 420, 421] and foreseen for ITER.
TFTR JET ITER projections
Parameter T-experiment T-experiment and requirements
Tritium in-vessel inventory limit 2g 20 g site inventory 350 g (guideline value) 1000 g (assumption
for safety evaluation)
T-introduced in the torus 5.2 g 35 g 50 g/pulse (400 s)
Discharge number and duration 708 pulses ≈ 33 min 500 pulses ≈ 250 min —
before inventory limit first
approached
Tritium retained in the torus before cleaning 1.7 g 11.5 g —
Fraction of tritium removed by cleaning 50% 50% (prior to venting) requirement close to 100%
Tritium retained after cleaning 0.85 g 6g
The recycling of hydrogenic fuel and its retention in carbon films show a ratio H/C > 1 at room temperatures,
plasma facing materials considered for next-step devices has which decrease strongly with increasing temperature.
been a traditional research topic in all fusion devices but the Because of its physics nature, the amount of fuel retained
research in this field has become particularly active after the by this mechanism does not saturate in time and it is,
use of considerable amounts of tritium and its incomplete therefore, the mechanism that raises more concern with
recovery in the full D-T campaigns in TFTR [414,415] and JET respect to fuel retention in next-step devices.
[416, 417]. A more detailed description of these observations,
the mechanisms of T retention and the techniques used to 2.6.2. Database on fuel retention in present fusion devices.
remove fuel from PFCs in fusion devices can be found in The majority of the fuel retention data comes from fusion
[275, 381, 418, 419]. Table 2 provides some key quantities devices with carbon PFCs. The amount of fuel retention is
related to tritium balance in experiments carried out in existing determined either by comparing the fuel injected to the exhaust
tokamaks and ITER [1, 416, 420, 421]. (‘gas balance’) or by ex situ analysis of wall tiles. The overall
The number of hydrogenic atoms in the surface of plasma tritium balance in the TFTR and JET tritium experimental
facing components that interact with the plasma exceeds in campaigns showed that 16% (TFTR) and 17% (JET) of the
general the total number of ions in the plasma. This leads injected tritium was still retained in the device after several
to a complex coupled system between the hydrogenic species cleaning procedures [275]. In TFTR, the tritium is mostly
at the wall and the plasma fuel inventory. The recycling and retained at the plasma facing surfaces, with some part on the
retention behaviour at the walls can strongly affect the fuelling vessel wall and a significant retention in the gaps between
of the plasma whilst the plasma changes the composition of the wall tiles. On the contrary, in JET, the vast majority
the wall surface. The mechanisms that cause retention of fuel (> 90%) of this long-term retained tritium was found in carbon
in fusion devices are [275] as follows. layers deposited on water-cooled louvres at the entrance to
(i) Direct implantation of ions and neutrals in a shallow the pumping plenum at the inner divertor [272]. A minor
surface layer and possible diffusion into the bulk. For this proportion of tritium was found on plasma facing surfaces
mechanism, the wall inventory is determined by material and most of it was retained in co-deposited C layers at the
properties, diffusion and recombination coefficients and inner divertor. Only a small fraction of the tritium diffused
the concentration/strength of trapping sites. Depending deeper inside the material along the open porous structure of
on these parameters, the fuel is retained temporarily the CFC material [422]. The JET T retention experience is
(‘dynamic retention’) or quasi-permanently by solution prototypical for carbon PFCs; the retention of T is associated
and/or trapping. The amount of fuel retained by these with the formation of co-deposited C layers while the retention
mechanisms tends to saturate with time and/or plasma associated with direct implantation and diffusion of T into the
fluence, as the surface layer concentrations reach steady- CFC matrix is small.
state values and the trap sites are filled. The quantification of fuel retention by the use of tritium
(ii) Possible production of tritium by transmutation nuclear has the advantage of being very precise but the disadvantage is
reactions. This occurs only for special wall materials (e.g. that these experiments are usually performed under restricted
beryllium) and can cause the build up of a tritium inventory tritium and neutron production limits and are thus, very
within the bulk material by trapping in microscopic defect focused. Therefore, the results obtained may not represent
sites and helium bubbles. the average fuel retention of the fusion device when operated
(iii) Co-deposition of hydrogen with eroded wall material to address a wide research programme. Because of this, it is
forming hydrogen-rich co-deposits inside the fusion important to compare the results above with those obtained
device. The rate of retention by this mechanism depends from the retention of deuterium fuel in present fusion devices.
on: (a) the rate of erosion of the material and (b) the fuel Table 3 summarizes these observations, which are explained
retention capability of the re-deposited material. This in detail below.
retention mechanism is only relevant for few materials Long-term deuterium fuel retention has been determined
(e.g. carbon) in a certain temperature range. For example, at JET by post-mortem surface analysis. These results show
diamond-like amorphous carbon films show retention a large amount of hydrogen-rich carbon deposits at the inner
(H/C ratio) of about 0.4 at room temperature, decreasing divertor [272]. The total amount of carbon deposited at the
significantly only above 1000 K. On the other hand soft inner divertor target are 900 and 500 g for the two divertor
S234
Table 3. Compilation of fuel retention data from various fusion devices.

Average Average Carbon Fuel Fuel
ion flux Carbon fuelling deposition retention retention
(lim/div) deposition rate ratio rate fraction
Device/campaign (D+ –T+ s−1 ) rate (C s−1 ) (D–T s−1 ) C/(D+ –T+ ) (D–T s−1 ) (ret. D-T/inj. fuel)
JET Mk IIA div. 3.8 × 1022 6.5 × 1020 — 0.037 5.8 × 1020 0.17 (in DTE1)1
(D/C = 0.8) 0.11 (in DTE1)2
JET Mk IIGB div. 4.3 × 1022 4.3 × 1020 3.2 × 1021 0.01 1.25 × 1020 0.033
(D/C = 0.3)
TFTRT campaign — — — — — 0.161
AUG — 3.5 × 1020 — — — 0.03533 0.14
TEXTOR 9 × 1021 2.5 × 1020 1.5 × 1020 0.029 1.6 × 1019 0.083
Tore Supra — — 4.6 × 1020 2.5 × 1020(5) 0.55
JT60-U — 2.7 × 1020 — — 5.3 × 1018 —
(D/C = 0.02)
1
T retention after (non-mechanical) T -cleaning.
2
T retention after long term outgassing and mechanical removal of accessible T -deposits.
3
D retention from post mortem analysis.
4
D retention from fuel balance.
5
D retention from fuel balance in dedicated long pulse discharges.
configurations explored, Mk IIA and Mk IIGB divertors, pump limiter with deposition on the remaining part, on
respectively, which accumulated during an integrated time protection limiters and obstacles in the SOL and the pump
of diverted plasma operation of 19 h and 16 h, respectively. limiter neutralizer plates. The averaged carbon deposition rate
The corresponding averaged carbon deposition rates were is about 2.5 × 1020 C s−1 , similar to JET and ASDEX-Upgrade
6.6 × 1020 C s−1 and 4.1 × 1020 C s−1 at the inner divertor values. This carbon deposition leads to a fuel retention of 1–
for the Mk IIA and Mk IIGB divertors, respectively. This 2 × 1019 D s−1 , with a typical D/C ratio in the deposited layers
deposition rate corresponds well with the overall amount in the range 0.05–0.1 [321]. As a consequence, the long-term
of carbon eroded from the first wall, estimated from a deuterium retention in TEXTOR is about 8% of the injected
combination of spectroscopic, carbon screening and Zeff fuel [321].
measurements [142]. The location of the deposits that lead In JT-60U, carbon layers build up at the inner
to the fuel retention changed significantly in the two divertor divertor [312], as in JET and ASDEX-Upgrade and have
configurations. In the Mk IIA divertor, particularly for reached thicknesses up to 100 µm after the operation period
the tritium experiments, the plasma was routinely run on a from 1997–2000 in which ∼8500 shots were performed,
horizontal diverted configuration, which led to the formation leading to a total carbon deposition of 320 g at the divertor,
of large deposits near the divertor corner on the water cooled similar to the JET values [426]. The deposition on shadowed
louvres at the entrance of the pumping plenum. In the Mk IIGB areas (mainly in areas in the private flux region viewing the
divertor, the plasma was routinely run on an ITER-like vertical vertical tiles) is much lower than on the target and only ∼ 9g C
divertor configuration, with pumping through the private flux were deposited in these areas during this operation period. The
region, which led to the largest deposits being formed on the overall carbon deposition at the inner divertor of JT-60U is
horizontal target and on the private flux region components. about 2 × 1020 C s−1 normalized to the total discharge time or
Due to these different configurations and the larger range of 3.3 × 1020 C s−1 normalized to the discharge period in which
experiments carried out in the Mk IIGB divertor compared with
additional heating was applied. Despite this large carbon
the tritium experiments in the Mk IIA divertor configuration,
deposition, the amount of fuel retained in these JT-60U layers
there are significant differences in the amount of long-term
is low. This is due to the D/C ratio in these carbon layers
fuel retained in these two JET campaigns, which is 17% of the
being only about 0.015 on average [427]. This is associated
injected fuel for the Mk IIA tritium campaign and only 3% for
with outgassing of the deuterium from these layers under the
the Mk IIGB divertor deuterium campaign [423].
plasma flux, which is favoured by the relatively high operating
Fuel retention in ASDEX-Upgrade is associated with co-
temperature of 600 K of the JT-60U divertor [427].
deposition of C layers at the divertor with a rate of 3.5 ×
1020 C s−1 , which is similar to the JET values [424]. Most Recently, new attempts have been made to measure
of this deposition occurs at the inner divertor with some the overall fuel retention by particle balance in hydro-
deposition taking place the outer one too, [424, 425] and, at gen/deuterium operation [428–430]. This is one of the pos-
present, cannot be explained only in terms of erosion of main sible methods of measuring the retention of fuel during the
wall carbon PFCs [286, 425]. Retention of fuel in remote initial non-activated phase of ITER operation. In general, these
areas of the divertor, such as pump ducts, is very small in data still suffer from significant uncertainties for a number of
ASDEX-Upgrade. The films deposited in this area are soft and reasons, such as the accuracy of the measuring systems, dif-
have a large deuterium content (D/C = 0.5–1) but the total ficulties in accounting for long term overnight and weekend
amount of carbon deposited in these films is very small (<1% outgassing, release by daily, overnight or weekend wall condi-
of the overall carbon deposition) [339]. tioning, etc but progress in dealing with them is encouraging.
In TEXTOR, global carbon erosion and deposition Time-resolved particle balance during long discharges allows
balance studies show net-erosion on about 2/3 of the toroidal separation between the dynamic and long-term retention in
S235
A. Loarte et al
the discharge and could determine the long-term fuel reten- 2.0x10
19
tion in the device. This approach has been followed in Tore
Carbon deposition (C/cm )

Supra (Twall = 400–500 K) [343, 430] with 250 s discharges
2
at low plasma density (1.5 × 1019 m−3 ). The measurements
19
1.5x10
show a small transient dynamic wall pumping at the begin-
ning of the discharge followed by constant fuel retention at 19
a rate of about 2.5 × 1020 D s−1 , which is about 50% of the

1.0x10
injected fuel. The amount of fuel retained in the analysed

PFCs in Tore Supra is not consistent with this retention [343] 5.0x10
18
and, thus, further research in this area is on-going to resolve

this disagreement. This problem is not only restricted to Tore
Supra, although the disagreement in this device is greatest. In 0.0
general disagreements by factors of 3 or more are found be- 55000 56000 57000 58000 59000 60000 61000 62000 63000
tween post-mortem analysis and fuel retention determination Discharge number

from gas balance of deuterium and hydrogen discharges due Figure 22. Integrated carbon deposition on the QMB detector at the
to the difficulties of the latter technique. The agreement is, JET Mk IIGB inner divertor during the March 2001–January 2004
however, much better for tritium experiments due to the better period, corresponding to a total exposure to divertor plasmas of
diagnostic possibilities of the tritium gas balance compared 26.4 h [435].
with deuterium/hydrogen [416, 417, 420].
flux, causing the hydrocarbons to migrate from their initial
2.6.3. Fuel retention mechanisms in present devices and deposition areas to other areas in which erosion is negligible.
implications to ITER. Since the vast majority of the fuel This process can take place during the quiescent phases of
retention in present carbon-clad devices is by co-deposition the discharges, due to the large erosion yields of hydrocarbon
deposits under plasma flux [321, 433, 434], or on a discharge-
of the fuel with eroded carbon, the understanding of the
to-discharge basis. In this case, hydrocarbon deposits can be
T retention process is mainly the understanding of carbon
formed in areas of the divertor not in contact with plasma
erosion, its migration and finally its co-deposition with the
in some discharges, which are then eroded in discharges for
fuel to form hydrogen-rich layers. Most of these issues are
which such areas are exposed to plasma flux. Eventually,
discussed in section 2.1 (SOL flows), section 2.4 (material
the retained fuel will be localized in areas not accessible
migration) and section 2.5 (divertor and main wall materials)
to the divertor plasma flux. Such a process has been well
and we will, therefore, only concentrate on the aspects relevant
documented in JET experiments with quartz microbalance
to tritium retention here.
monitors (QMB) [435] near the inner divertor corner, whose
All divertor tokamaks show net deposition at the inner
results for the whole experimental campaign in the three years
divertor, while the observations at the outer vary from being
period of operation with the JET Mk IIGB divertor are shown
erosion-dominated to deposition-dominated. In all cases,
in figure 22. Obviously, due to this hydrocarbon migration,
the deposition measured at the divertor cannot be explained
the build up of re-deposited layers, with the corresponding
by local divertor processes and a significant contribution
accumulation of fuel, in various areas of the tokamak is very
from material eroded at the main-chamber wall depositing
dependent on the detailed experimental programme run in the
at the divertor is required to explain the observed divertor devices as well as on the divertor target and magnetic geometry
deposition. Two major processes are known to be involved of the various discharges executed.
in determining the observed carbon deposition at the divertor: Besides it being difficult to predict the expected location of
local divertor plasma conditions, which favour deposition for hydrocarbon deposits in ITER, the size of the deposition itself
low temperature/high density conditions and SOL flows which is also uncertain. This is associated with the uncertainties of the
transport, preferentially, the main-chamber eroded carbon to chemical erosion yield (the source of hydrocarbons in ITER)
the inner divertor. The final location and amount of re- for ITER conditions, as discussed in section 2.5. A simple
deposited hydrocarbon layers, which contain most of the empirical extrapolation of the observed fuel retention in the
retained fuel in carbon-clad devices, depends on local transport present machines, which is dominated by the erosion of the
processes at the inner divertor. Recent measurements at JET main-chamber walls, towards ITER readily demonstrates that
and ASDEX-Upgrade have shown that hydrocarbon deposition T retention would lead to a very short operation time between
at the divertor occurs mostly in areas within the line of sight major clean-up operations if all ITER walls would be made of
of the hydrocarbon source at the divertor target. This is carbon. Present retention fractions of the injected fuel (from
because most of the radicals formed there (>90%) have large table 3) are in the range of 3 to 50%. Simulations of ITER
sticking coefficients [308, 425], which is consistent with the plasmas [329] indicate that up to 50 Pam3 s−1 of tritium will
low amount of redeposited hydrocarbons found in the pump be injected in the ITER reference QDT = 10 scenario, which
ducts of ASDEX-Upgrade and TEXTOR [339, 431]. corresponds to 54 g of tritium for a discharge with 400 s of burn.
On the basis of these observations, most of the fuel The fuel retention if the ITER main wall would be covered by C
retained in ITER could be expected to concentrate in areas would then be in the range 1.6–27 g shot−1 . As a consequence a
in line-of-sight from the carbon sources, with only a small major tritium clean up operation to remove the retained tritium
fraction of carbon travelling longer distances [432]. In reality, would be required every 13–218 reference discharges, if the
the picture of carbon migration is more complex than this and goal of maintaining a maximum in-vessel T inventory of 350 g
significant re-erosion of these layers can occur under plasma has to be fulfilled.
S236
Be eroded from the main wall, which decreases the C surface

concentration and, thus, the production of hydrocarbon layers
and fuel retention in ITER. Although this will be discussed in
more detail in the next section, it is clear that depending on
the mechanism that dominates, the T retention in ITER can
be much smaller or much larger than that estimated for the
inter-ELM conditions shown in figure 23.
Recently, attempts have been undertaken to mitigate the
formation of carbon layers and the associated retention of
fuel by adding certain amounts of nitrogen in the divertor
region of devices with a carbon divertor target, which cause
the formation of volatile radicals. This technique may fulfil
both the requirements for divertor radiation control (nitrogen
Figure 23. Modelled tritium retention in ITER based on T does not affect the chemical erosion of carbon and can replace
co-deposition with a carbon divertor target, showing the expected the carbon as a radiator) and low redeposition probability (low
retention using a fixed chemical erosion yield (1%,
Brooks/Kirschner) and using a more realistic chemical erosion
fuel retention), which are necessary for ITER. The technique is
formula. Different assumptions are included on the sticking well known in the field of plasma processing and a factor of five
probability of hydrocarbon species returning to the divertor surface reduction in hydrocarbon deposition has been demonstrated in
(S = 1, full sticking, S = 0.5, S = 0, zero sticking for tokamaks [442], but more experiments are needed in order to
hydrocarbons). The solid line shows the goal for maximum confirm its effectiveness and applicability to ITER.
in-vessel tritium inventory limit in ITER [334].
A final point to consider is that the divertor PFCs in ITER
will have a significant number of castellations/gaps, which are
The extrapolations above (and others which can be done on necessary for the PFCs to withstand the high cyclic thermal
the fraction of retained fuel normalized to divertor fluxes) are stresses. This creates a specific problem with respect to tritium
purely empirical and, although they provide a sound argument retention for ITER related to the possible retention of fuel
against a full C wall in ITER on the basis of T retention, do in carbon layers that build up in the larger gaps between the
not describe the expected situation in ITER with a full Be wall. divertor cassettes and the narrower gaps between the tiles of the
More realistic predictions can be done by using Monte-Carlo divertor targets and the castellations themselves. A five year
impurity transport modelling codes such as REDEP/WBC or inventory study in TFTR with a total amount of 83 g deuterium
ERO codes [273, 434]. In this modelling, the only source into the vessel, revealed a total of 33.6 g of fuel retained from
of carbon is sputtering (physical and chemical) at the ITER which 47% was on the plasma facing sides, 15% in the gaps of
divertor target and the pathways of eroded species are then tiles and 38% on the vessel wall [275]. During the JET Mk I
followed based on the atomic database, e.g. from [436, 437]. divertor campaign, in which the divertor tiles were ITER-like
Redeposition and sticking of eroded species is computed using castellated, about 30% of the fuel retained in the inner divertor
assumptions for the hydrocarbon radicals (such as 1 or 0) layers was retained in the gaps with the rest deposited on top of
or from MolDyn calculations [438] and TRIM data for ions the inner divertor tiles [443]. At this stage, the comprehensive
and atoms [439, 440]. The resulting modelled tritium co- models of T retention (e.g. the results in figure 23) do not
deposition is shown in figure 23 and varies from ∼0.02 to include these effects and, therefore, no quantitative evaluation
2 g of tritium retained per discharge in ITER [334], depending of the expected T retention in tile gaps at the ITER divertor
on the assumptions for the sticking of hydrocarbons and on exists. Despite this, initial dedicated experiments to address
the sputtering yield. The reduction of the tritium retention this problem have shown that the amount of fuel that can be
in ITER in the new estimates with respect to those of [441] retained in these gaps is large and comparable with that retained
comes mainly from the inclusion of a more realistic model for in the shadowed areas of the plasma facing components [444].
the chemical erosion yield in ITER [334], rather than a fixed
∼1% rate. Besides this, the parameter which influences the 2.6.4. Implication of the ITER wall material choice and mixed
results more is the sticking coefficient of the hydrocarbons material effects on tritium retention. The results described
produced, which can change the calculated retention by a in the two previous sections come from devices dominated
factor of more than 20 when high/low sticking probabilities by carbon PFCs, which will not be the case for ITER. In
are considered. As described above, the effective sticking ITER, carbon PFCs are limited to a narrow zone near the
probability of hydrocarbons is affected by their re-erosion (in- strike points in the lower divertor (50 m2 ) with a large first
shot or shot-to-shot) and it is difficult to predict its value for wall area covered with Be (700 m2 ) and W on the divertor
the ITER divertor conditions. Depending on this value, major baffle and dome. Due to its low-Z and low binding energy,
tritium clean-up operations will be required in ITER for every Be will be eroded from the first wall at a similar rate to
600 high performance discharges (i.e. every few months) or carbon in present experiments, despite the absence of chemical
just once in the operation lifetime of the device, according to erosion. The eroded Be will be transported to the divertor
these new estimates, as shown in figure 23. where it will deposit. Depending on the SOL flow patterns
It is important to note that the above calculations do not and local divertor conditions in ITER, this eroded Be will
include the additional production of carbon associated with deposit predominantly at the inner divertor (as presently is
Type I ELMs (see sections 2.2 and 2.4), which would increase the case at JET) or in a more in/out balanced way. In these
the production of hydrocarbon layers, nor the migration of conditions, it is expected that Be will accumulate on the
S237
A. Loarte et al
detail in section 2.5. Due to concerns of the compatibility
1
1
Mayer
et al. of a large range of plasma operating scenarios with a full
tungsten divertor (including damage under transient loads
Mayer et al. associated with Type I ELMs and disruptions) this is presently
Present
data considered only as a back-up option for ITER, should the
PISCES tritium retention be unacceptable with the reference ITER
D/Be
0.1
0.1
plasma facing material choice.
O/Be
Causey
et al. Causey &
Walsh 2.6.5. Dust in tokamaks (and impact on T retention).
TPE
TPE The subject of dust formation and T retention is reviewed
Causey &
Walsh in [275]. In the past, the production of dust has received
Present
data TPE little attention within the fusion community, mainly because
0.01
0.01
PISCES (a) (b) dust is neither a safety nor an operational problem in existing
tokamaks. However, in a next-step burning plasma device,
400 600 800 400 800 such as ITER, production of dust could have important
Temperature (K) safety implications [447, 448]. The safety issues associated
Figure 24. Retention fraction of D in different re-deposited Be with tokamak dust include radiological hazard (tritium and
layers in various experiments with different oxygen activation products), toxicity and chemical reactivity with
concentrations [352]. steam and air. A variety of products can result from plasma–
wall interactions (e.g. deposited films, flakes, debris), as shown
divertor target thus suppressing the production of hydrocarbons by measurements in present tokamaks [275, 449], and many
at the divertor. This physics hypothesis has been recently of these products can be considered as ‘dust’ hazards. The
demonstrated in experiments at the plasma divertor simulator carbon and beryllium dust limits in ITER stem primarily
PISCES-B, in which Be seeding on a plasma in contact with from concerns related to chemical reactivity, whereas the
a C target has been shown to decrease to negligible levels the tungsten dust limit is determined both by chemical reactivity
chemical erosion yield of carbon even at Be concentrations and activation hazards. There are still large uncertainties
of ∼0.1% [406, 445]. Although the extrapolation of these associated with dust production mechanisms in a Be-W-C
results to produce a quantitative estimate for ITER conditions environment, rates of formation and the determination of the
is uncertain, it is clear that such an effect can have a major dust quantities. The existing database hardly permits the
effect on tritium retention in ITER by reducing the divertor extrapolation from present machines to the next generation
carbon source and hydrocarbon transport in ITER. The size of of tokamaks because of the different material composition
this reduction in ITER will depend on the erosion/redeposition of the wall. Despite some new attempts at improving
balance at the ITER divertors and the expected T retention the experimental situation [450], techniques for reliable
by co-deposition (Be can only deposit on plasma exposed measurement of dust inventory, particularly in hidden areas,
areas, while hydrocarbons migrate into shadowed areas [309]), are not well developed, although they are needed to ensure
provided that these protective Be layers are not removed by the compliance with regulatory limits. Similarly, dust removal
divertor energy pulses associated with Type I ELMs. methods are not developed. Several methods for the removal
If large areas of the divertor are covered by Be layers, of dust-forming co-deposits have been proposed [451], but they
the T retention in these Be layers may become a serious have not been tested and qualified enough with respect to their
issue, in particular if the oxygen content of these layers is application for ITER.
not sufficiently low. The hydrogenic content of mixed Be/C/O
material layers increases with the incorporation of impurities
(in particular of oxygen) increasing the expected T retention 2.6.6. Tritium removal methods. As highlighted before, the
in ITER due to Be [350, 351]. However, recent experiments working guideline for the maximum in-vessel inventory of
from PISCES-B have shown that Be layers formed in realistic tritium that can be mobilized in the ITER vacuum vessel is
ITER conditions [352] can contain much less deuterium than presently ∼350 g. The assumption made for safety analysis
expected from previous measurements [350] so that very small is ∼1 kg of tritium mobilizable in the vacuum vessel [134].
deuterium contents (<0.01 D/Be) are found for temperatures Therefore, operation of ITER will require periodic removal
above about 600–700 K (typical of the ITER divertor target) of tritium captured in in-vessel deposits in order to avoid
even for ratios O/Be ∼0.3, as shown in figure 24. Even if there approaching this limit. Techniques for tritium removal are only
is significant retention of tritium in the Be deposits, removal of partially developed at present and usually not demonstrated
tritium from these deposits should be eased by the fact that the in a tokamak environment. TFTR and JET have intensively
Be layers will only form on the plasma facing sides of PFCs, explored various T removal techniques but the methods suffer
where they can be more easily accessed for cleaning. in general from the fact that they are too slow to cope with the
A major advantage of tungsten as PFC is that its higher duty cycle of a next-step device and that they act mainly
retention of hydrogen is negligibly low at the ITER operation on the plasma facing surfaces [452].
temperatures. This has been measured in various beam The presently foreseen methods to remove tritium from
experiments but also confirmed in the tungsten experiments ITER are reviewed in [416, 418, 420] and can be roughly
in present machines [279, 305, 383, 446] and is discussed in divided into four categories:
S238
(a) Plasma cleaning. On areas in direct contact with the produced by ablative methods and the recovery of plasma
plasma, T retention can be reduced by isotope exchange operation, before a credible tritium clean up schedule and
during deuterium plasma operation or desorption by their consequences for ITER operation can be evaluated in
plasma heating of the material. Plasma heating or isotope a quantitative way.
exchange is restricted to areas of direct plasma impact but
is difficult to apply on remote areas or gaps in between 2.7. Plasma fuelling
tiles.
(b) Surface heating. This can be achieved by heating the 2.7.1. Separatrix density. The separatrix serves as the key
redeposited hydrocarbon by external means such as a laser interface between the pedestal/core density, as requested by
or flash lamp [453], leading to the release of fuel or the core operational and burn criteria, and the divertor particle and
ablation of the film itself. Along these lines, the use of the energy exhaust, which are controlled by the neutral pressure.
end of plasma discharges in ITER for tritium removal by The separatrix density is determined to a large extent by
inducing a radiative collapse similar to that of mitigated the power flux and pressure balance via SOL and divertor
disruptions has been considered [355]. The major problem physics, which affect the ionization of neutrals at the plasma
of the applicability of such a technique in ITER is that edge. Ideally, in a burning plasma the separatrix density
it requires a very high level of reproducibility, to avoid should be regulated at a value corresponding to the maximum
causing frequent large-scale melting of the Be wall, divertor heat dissipation and particle exhaust, while the core
which could impact very negatively on the operability and pedestal plasma densities should be optimized for the
of ITER. Even if successful, the isotropically emitted optimum core performance. However, through anomalous
radiation pulse is unlikely to illuminate the areas in transport, particularly in H-modes [22], and edge ionization
the divertor where the highest deposition is expected to physics, the separatrix density is tightly coupled to the
occur. In addition to the various issues related to their pedestal density [462–464], which constitutes a major fraction
practical application in tokamaks (i.e. accessibility of the of the line average density in H-modes (see section 2.4).
electromagnetic radiation to the redeposits), such methods Therefore, controlling the ratio of separatrix to line average
can cause an uncontrolled production of dust, which is a density to meet the requirements of both core confinement
concern for ITER. and divertor heat load is expected to be a challenging
(c) Oxidative methods. The most promising method in this task and may make it necessary to operate a burning
line is the removal of the C-deposits by oxidation, which plasma device at a density which is not optimum for either
transforms the carbon into CO and CO2 and thereby requirement.
releases the incorporated T, mainly in the form of tritiated Assuming a fixed density peaking factor (nline /nped ),
water [454, 455]. A good database from laboratory knowledge of the following is required to design the optimum
experiment exists on the oxidation of different carbon operating density point of a device.
films by molecular oxygen [456,457] but the experience in
fusion devices is limited. TEXTOR has demonstrated the 1. An acceptable operating range of pedestal densities with
removal of carbon films by oxygen venting of the device sufficiently high pedestal pressure.
at 620 K and successfully recovered plasma operation 2. Core performance as a function of core and pedestal
with standard conditioning (He and D2 GDC) for these density.
conditions [458]. However, the technique has serious 3. A range of separatrix densities that can be obtained for a
drawbacks, such as the required wall temperatures difficult given pedestal density.
to achieve in ITER, which should be above 550 K, the 4. Divertor performance as a function of the separatrix
interaction with other non-carbon in-vessel components density.
(such as Be), the recovery time for normal plasma Of particular importance for item 3 is the physics that
operation and the processing of the resulting tritiated determines the steep density gradient zone up to the pedestal
water. Recently, ozone has been shown to oxidize carbon top. In order to address these points, a new multi-machine
layers at significantly lower temperature compared with database for the separatrix density and related quantities
molecular oxygen [459], but this method has not been yet has been set up for H-mode conditions [465]. It contains
demonstrated in the tokamak. experimental data from Alcator C-Mod, ASDEX-Upgrade,
(d) Conditioning methods. Wall treatment with RF DIII-D, MAST, JET and JT-60U gas fuelled discharges with
conditioning plasmas in the presence of the toroidal NBI and ICRH additional heating in H-mode confinement.
magnetic field have been further explored [396,460,461], The bulk of the measurements consists of electron density and
as described in section 2.5.5. These techniques have been temperature measurements by Thomson scattering, supported
shown to be effective for releasing the retained fuel from by probe measurements in the SOL (JT-60U) and lithium beam
the material elements facing the plasma, which may open measurements and modelling (JET) [22]. A two-point model
options for their use in ITER to release tritium from the based on power balance [77,466] is used to assign the separatrix
surrounding walls by their routine application between location for cases where the uncertainties of the equilibrium
discharges. reconstruction are comparable with the density decay length.
More experiments are necessary to demonstrate the The noisy experimental data are fitted and regularized by a
effectiveness of these techniques, the recovery of released modified hyperbolic-tangent function [67]. The data have
tritium, the interaction with the non-carbon ITER wall been restricted to conditions without strong core density
materials, the handling of the debris that is potentially peaking.
S239
A. Loarte et al
0.8 1.0
AUG a) b)
density widths ratio

C-Mod n e,sep 0.8
n e,sep / n Green
0.6
(inside sep. / full)

= 0.5
DIII-D ne
JET 0.6
n e,sep
0.4 JT60-U = 0.33
ne
0.4
0.2
0.2
0.0 0.0
0.0 0.4 0.8 1.2 0.0 0.1 0.2 0.3 0.4 0.5
n e /nGreen n e,sep / n Green
Figure 25. (a) Separatrix density versus line-averaged density, both normalized to the Greenwald density for H-mode discharges in several
tokamaks. The dashed lines correspond to two different ratios of the separatrix to line averaged density and are drawn to guide the eye. (b)
Fraction of the steep density gradient region inside the separatrix for the discharges in (a) versus separatrix density normalized to the
Greenwald density [465].
The normalized separatrix density, nsep /nGreen , from 1.2

the multi-machine data is shown in figure 25(a) versus the
normalized line-averaged density. The Greenwald density
1
provides excellent normalization of the data, which vary, e.g.
by a factor of 30 in absolute value between C-Mod and JT-60U.
For low Greenwald fractions (fGreen ), the line-averaged density 0.8
DENSITY PROFILE
is a factor of 3 larger than the separatrix density, while this

factor decreases to less than 2 for high values of fGreen .
Figure 25(b) shows a plot of the fraction of the steep density 0.6
gradient region width inside the separatrix (or pedestal density
width), δin , in relation to the full width of the steep gradient 0.4
zone, δtot , versus the normalized separatrix density. The
negative slope indicates an outward shift of the steep gradient nped = 2 1019 m-3
nped = 4 1019 m-3
zone with respect to the separatrix with rising density. Sample 0.2
nped = 6 1019 m-3
profiles can be found in figure 11 of [22]. nped = 8 1019 m-3
An analytic gas fuelling model [467] describes many
0
features of edge density profiles by neutral penetration physics. -3 -2 -1 0 1 2
Figure 26 shows modelled DIII-D density profiles. Consistent R-R S E P(cm)
with data of figures 25(a) and (b), the ratio of the separatrix
to pedestal density increases with increasing pedestal density Figure 26. Modelled DIII-D density profiles, normalized to
while the profile width inside the separatrix becomes narrower. pedestal density. With increasing nped , the width of the density
√ pedestal becomes narrower while the ratio of separatrix to pedestal
According to the model δin tends to scale as Tiped /(nped .E ∗ ), density increases [467].
where E ∗ is a measure of the poloidal distribution of the
neutral source relative to the x point. E ∗ is largest when the
neutral source is concentrated at the x point, the location of
the Greenwald value [128]. This rise in radial transport is
the largest flux expansion. In addition to the direct effect of
the local neutral sources as implemented in the simple model, accompanied by larger wall fluxes and, therefore, affects the
radial particle transport around the separatrix [77] and neutral neutral fuelling in the SOL as well as in the pedestal regions.
screening in the SOL plasma [58] affect the density profile Since at least three different, complicated mechanisms (neutral
shape. penetration, transport in the edge barrier region and SOL
The density width data shown in figure 26(b) are better transport linked to main-chamber recycling) affect the edge
organized when plotted versus the nsep normalized to the density profile, further inter-machine scalings, derived from
Greenwald density. While the outward shift of the steep more complete databases, and detailed comparisons with code
gradient zone of the density profile with increasing nsep can be modelling are necessary, and are in preparation, to disentangle
understood as an effect of the neutral ionization location, the the interplay of these processes.
dependence on the Greenwald density suggests the impact of There are practical operating windows for both the
transport effects. One obvious effect is the dependence of the pedestal and separatrix densities. The lower limit on the
diffusion coefficient on the plasma current. Another possible separatrix density is set by the divertor heat flux. B2-Eirene
player is the radial SOL particle flux caused by intermittent modelling of the ITER edge plasma with 130 MW of heat flux
events [49,195], which has been observed to increase strongly across separatrix shows that a minimum separatrix density
in Alcator C-Mod when the density is increased towards of 4 × 1019 m−3 is needed for an acceptable peak heat flux
S240
[20, 29, 254]. An upper limit on the separatrix density is

set by energy and momentum balance in the SOL [468] and
by volumetric recombination in the divertor plasma [469].
Scaling from DIII-D [467] sets an estimated upper limit of
6×1019 m−3 for ITER, assuming Bohm transport scaling in the
SOL and open geometry. It is conceivable that there are more
restrictive limits on the separatrix density due to its effect on
H–L transition or onset of Type III ELMs. However, a causal
relationship has not, so far, been established between separatrix
density, H–L transition or Type III ELMs. Therefore, further
theoretical studies should be pursued to guide the experimental
and modelling progress in these issues.
Pedestal densities as high as 90% of the Greenwald
limit, with no confinement degradation, have been achieved
in DIII-D [464] and in Type II ELMy discharges in
ASDEX-Upgrade [470]. This has been exceeded in JET
with highly shaped discharges, in which pedestal densities
up to nped /nGreen = 1 have been reached [471]. Because
of the coupling of the separatrix and pedestal densities with Figure 27. Comparison of the plasma particle content behaviour for
standard gas fuelling, a sufficiently high pedestal density 3 fuelling methods (SMBI, GP and low field side PI) for the same
may not be accessible in ITER. Alternate fuelling techniques, number of particles injected on Tore Supra [477].
which provide a higher particle source at the pedestal top, are
discussed in sections 2.7.2–2.7.4.
While a better fuelling efficiency of the SMBI compared to
gas puff was observed, the instantaneous triggering of ELMs
2.7.2. Supersonic gas jet and low field side puffing.
could not be achieved [481], probably due to an insufficiently
Supersonic molecular beam injection (SMBI) offers a
deep neutral penetration.
potentially superior fuelling technique compared with cold
The dependence of the gas fuelling efficiency on the
gas fuelling and is technically simpler in comparison to
fuelling location has been investigated in DIII-D [482]. Inner
pellet injection (PI). The main features of SMBI are a higher
wall gas fuelling did not show any significant increase in
molecular speed and lower divergence compared with gas
fuelling efficiency as observed with pellets [483]. This is
puffing (GP). SMBI has been implemented on the HL-1M
explained by the poorer penetration depth and a lack of
tokamak [472, 473], the HT-7 tokamak [474] and the W7-AS
significant polarization drift effect [484] with gas puffing. The
stellarator [475]. A Laval nozzle is utilized to inject a well
enhanced fuelling efficiency of SMBI is also not attributed to
collimated supersonic jet in the plasma (typically 5 × 1019
a polarization drift, since the beam density is estimated to be
particles per pulse with an injection time ranging from tens to
not high enough to trigger collective effects. Simulations with
hundreds of ms [472]). Gas speeds higher than 1 km s−1 have
a 1D particle transport model taking the plasma cooling into
been obtained [476], to be compared with 300 m s−1 typical of
account have shown good agreement with the experimental
gas puffing.
results [485].
On the Tore Supra tokamak, a repetitive SMBI system
with a fast mechanical valve has been installed, allowing the
use of a shorter injection time (2 ms) and a high injection rate 2.7.3. Pellet fuelling. As discussed above, for efficient core
(1023 particles s−1 or 200 Pam3 s−1 , corresponding to 2 × 1020 fuelling, deep pellet particle deposition is required. The key
particles injected per pulse) with a repetition rate up to parameters for the determination of the pellet penetration depth
10 Hz [477]. Figure 27 shows a comparison of the achieved are plasma density ne , temperature Te , pellet mass mp
plasma particle content for the 3 fuelling methods (SMBI, GP and velocity vp . A multi-machine study [486] has confirmed
and low-field-side PI) with a roughly equivalent number of the neutral gas shielding scaling of the penetration depth
−1/9 −5/9 5/27 1/3
particles injected. As observed in the other experiments, the ∼ n e Te mp vp for outboard launch. The absolute
main results obtained with SMBI are an improved fuelling pellet penetration depth for inboard launch is about twice
efficiency (in the range 30–60% to be compared with 10– that of the outboard launch. The deeper penetration of the
15% for conventional GP and up to 80% for PI) associated high-field-side injection [483, 487] is attributed to the ∇B
with a strong plasma cooling [477]. A reduction in the and outward curvature-drift force on the plasmoid formed
hydrogen wall inventory has been achieved in Tore Supra around the pellet [484, 488, 489]. The observed increase in
long discharges comparing GP and SMBI [478, 479] when penetration depth with inboard launch has been confirmed in
the injected quantity per pulse is significant compared with the several machines [490–493]. Inboard launched pellets must
plasma content (∼50% in the above experiments), inducing a be transferred through a curved guiding system, which limits
strong perturbation of the plasma and a short detachment phase pellet speed. Currently, the fastest pellet accelerators, based
after each injection. This effect is not observed with injection on the multi-stage gas gun principle can launch pellets at up to
at a lower rate. An attempt was made on ASDEX-Upgrade to 4500 m s−1 [494], whereas, the highest pellet speed achieved in
use the perturbation by the SMBI to trigger ELMs in the Type I a centrifuge, which is more suitable for steady-state operations,
ELMy H-mode, as has been done with shallow pellets [480]. is 1200 m s−1 [495]. A repetitive injector, able to operate at
S241
A. Loarte et al
a frequency higher than 10 Hz, has also been developed for a basic paradigm for anomalous transport across field lines,
continuous fuelling [496] and has been successfully tested on which is hard to justify in the edge region, where the gradient
long discharges in Tore Supra [497]. length and the different scales determining turbulence (notably
The planned ITER fuelling system is based on a centrifuge ρi ) approach each other.
injector launching pellets at the high-field side via a transfer Edge code-modelling is instrumental in addressing
system suitable for speeds up to 500 m s−1 [498]. Central a number of ITER edge issues, e.g. the operational
fuelling is questionable with this system, even when the window for acceptable target power, helium exhaust, target
beneficial effects of the high-field side launch are taken into lifetime, impurity migration, tritium retention, core boundary
account. In [499] calculations of the penetration of pellets conditions, etc. Significant progress has been made since the
in ITER with a size d = 6.6 mm and a speed of 500 m s−1 IPB [1] in a number of theory and modelling issues, including
with the integrated ablation/mass relocation model SMART, code developments in neutral/radiation transport, study of fast
which had been benchmarked for pellet injection experiments ion effects, ELM buffering, preliminary drift modelling—
in ASDEX-Upgrade and DIII-D [500], as well as for HFS edge- including replication of power asymmetry scaling in JET for
localized absorption in JT-60U [492] were made. The resulting example, etc. The identification of outstanding problems has
deposition is entirely outside r/a = 0.65; this is not adequate also continued including validated/routine application of drifts,
for central fuelling but very suitable for increasing the pedestal modelling hydrocarbon migration, the issue of how to deal
density. The same system will be used for the triggering of with peripheral turbulent transport and main wall interactions,
ELMs with the aim of reducing ELM amplitude [499]. This ELMs, difficulties in describing the details of detachment,
concept is based on ASDEX-Upgrade observations, which etc. In the remainder of this section some of these matters
show that pellets can trigger high frequency Type I ELMs with are described in detail; however, space limitations prevent
a lower energy loss per ELM relative to the naturally occurring comprehensive coverage and the selection of topics, inevitably
ELMs, with no significant confinement degradation [148]. somewhat arbitrary, aims at touching on a few recent high
points.
2.7.4. Compact toroid fuelling. Compact toroid (CT)
3.2. Progress in ITER divertor modelling
fuelling involves the injection of dense, self-contained
magnetic plasmoids to achieve deep fuelling (see ITER In section 3.2.1 we describe the current status of the models
Physics Basis chapter 3.5.3 [1] and references therein). The used to analyse the ITER divertor performance, emphasizing
possible benefits of CT injection, with respect to other the logics backing the parameter choice. Although some of the
fuelling techniques, are the possibility of deep fuelling and the assumptions may be challenged, these are the models on which
controllability of the fuelling location, which may be essential a number of ITER design solutions are based (section 3.2.2).
for the fuelling of steady-state advanced tokamak plasmas An outlook towards further model development relevant
in burning plasma experiments. CT injection into tokamak to ITER (or other reactor-grade experiments) is given in
plasmas was pioneered in TdeV [501] and was investigated in section 3.2.3.
JFT-2M [502]. CT injection has not matured into a continuous
standard fuelling technique so far. The reasons for this are the 3.2.1. Model parameters. The computational models contain
technical complexity of the injector and the fact that the CT a number of free parameters representing such quantities
length is comparable with the plasma radius in small tokamaks. as anomalous cross-field transport, semi-empirical kinetic
corrections to the parallel transport and boundary conditions,
3. Modelling and theory reflecting the incomplete present state of knowledge of the
physical models describing the transport processes in the edge
3.1. Introduction plasma. In interpretative mode, these free parameters are
adjusted to fit the experimental data; this leads to different
To date, tokamak divertors and wall systems have been values of cross-field diffusivities resulting from fits to different
designed largely on the basis of empirical extrapolation experiments [503] (see also section 2.1). Predictive modelling,
supported by basic theoretical concepts, and this remains such as extrapolation to ITER, then requires a choice of these
the principal means of proceeding to ITER. Computer-based free parameters on the basis of other constraints.
modelling greatly assists this process by identifying the For a given input power, the width of the SOL temperature
potential role of presently known edge effects, through the profile upstream is determined primarily by the ratio of the
construction of self-consistent ‘thought experiments’. Edge cross-field to parallel thermal diffusivity (see also section 2.1),
code modelling also aids in the identification of controlling and the temperature itself is determined by the values of
physics, by pointing to quantities whose measurement these diffusivities. Parallel transport is modelled by classical
in existing tokamaks can expose those effects and their Spitzer–Härm heat conductivity with ‘flux-limit factors’
role. Empirical extrapolation requires sufficient theoretical (FLF) [1], which partially account for kinetic effects (the
understanding to be able to identify scaling factors, and while depletion of the high-energy tails of the particle distribution
present edge code-modelling is not usually capable of detailed functions at low collisionality). From kinetic modelling [504]
prediction, it can be capable of the identification of trends, based on a hybrid, fluid-kinetic model [505], a value of the
as required for scaled extrapolation. Edge code modelling, FLF for electron heat conductivity of 0.2 is selected for the
therefore, continues to play a central role in the design of ITER calculations, whereas for technical reasons, no FLF is
the ITER divertor and wall system. We have to be aware, used for ion heat conductivity in these calculations. The cross-
however, that present codes use a diffusion type ansatz as field transport, which determines the profile, should then be
S242
constrained by considerations of Larmor radius and stability. concerns the neutrals, the Monte-Carlo neutrals are allowed
First, the profile width must be larger than the poloidal Larmor to travel freely in the gap, although in reality there would be
radius of the ions, which is the measure of radial excursion some plasma-neutral interaction there. Both approximations
of the banana particle orbits in the SOL, thus eliminating are justified by the low power flux (several per cent of the
for ITER modelling the lower of the diffusivities fitting the input power) crossing the boundary surface, which renders
experiment [503]. Second, there is no experimental evidence it unlikely that the exact values chosen will have a strong
that radial gradients in the SOL can be steeper than those just impact. However, no sensitivity studies have yet been done for
inside the separatrix [22,506] (see also sections 2.1.1 and 2.7), these parameters, and the model used for the plasma transport
which are determined, apparently, by stability. The gradients boundary conditions lacks a first-principle based theoretical
limiting the plasma stability just inside the separatrix can, justification. Recently, the importance of substantial plasma–
therefore, be taken as constraints also for the radial gradients in wall contact—additional to the plasma-target contact—has
the SOL. These constraints are consistent [507] with a choice been demonstrated experimentally, and model developments
of χ⊥i = χ⊥e = 1 m2 s−1 and D⊥ = 0.3 m2 s−1 for the cross- are proceeding to include this new physics element.
field transport in ITER modelling, resulting in a typical width Generally, the specification of the material properties on
of the power load profile around 5 mm when projected to the the plasma facing components (PFCs) may have a strong
midplane, and in a radial pressure gradient in the midplane influence on the results obtained in the modelling studies.
close to the experimental ballooning limit. Sensitivity studies In long-pulse reactor operation, the state of the surface
in [508] show that reduction of these diffusivities by a factor should reach equilibrium at every point with the impinging
of 2 produces the same effect on the peak power loading as an particle fluxes with regard to erosion and re-deposition, i.e.
increase in the input power by 25%, i.e. less than the size of the boundary conditions evolve in the course of the iterative
the operating window, see section 3.2.2. Intermittent transport solution along with the local plasma parameters. Specifically,
events leading to flat profiles in the far SOL (see section 2.1.3) a pure metallic surface, which would absorb all the incident
may be related to the profile readjustment necessary to maintain carbon, does not retain this property: at low incident DT
the near-SOL profile near the limit, corresponding to the above fluxes (deposition-dominated) a carbon layer builds up and the
specification of the transport; they are not explicitly included surface properties become those of deposited carbon, whereas
in the ITER modelling but may affect the erosion processes at high incident DT fluxes (erosion-dominated) the surface
at the walls. Although a radial pinch may be important to fit properties are those of the metal and all the deposited carbon
some present experiments [126, 509], it is not used in ITER is re-eroded. Interpretative impurity transport modelling has
modelling because the classical Ware pinch [510] is unlikely had only limited quantitative success so far in the cases where
to be important (the loop voltage in ITER is expected to be very chemistry is significantly involved (in particular chemical
low), and present knowledge is not sufficient to extrapolate an sputtering and hydrocarbon chemistry). Nevertheless, based
anomalous pinch velocity to ITER. upon best guesses for the relevant rates, first studies of this
The boundary conditions also require a choice of free effect for ITER conditions ( [513, 514]) have been attempted.
parameters. At the target, the most important region for the They result in an increase of Zeff at the separatrix to a value
solution, the boundary conditions are well-established [1], at ∼1.7, accompanied with a shift of the operational window
least for the conditions investigated so far, in which drifts towards lower densities. In the case of a constant chemical
and electrical currents are assumed to be of minor importance sputter yield, Ych , they predict for ITER a carbon deposition
only [511]. Near the side walls, radially far from the separatrix from plasma ions on the lower outer part of the first wall
(low power levels), the boundary conditions are complicated just above the outer baffle, which disappears if the Ych for
by the fact that the computational grid has to follow the the deposits is an order of magnitude higher [513]. Note that
magnetic surfaces which do not conform to the first wall. the deposition occurs on a time scale of 10 s, i.e. equilibrium
Moreover, a high triangularity of the plasma column, required would be reached within a single shot.
for better performance of the core plasma (see chapters 2 There are very few free parameters in the Monte-Carlo part
and 3), brings the second x-point in ITER slightly inside the of the code, although some free parameters may be inherent in
first wall. Initial modelling studies of double-null divertor the data on particle-wall interaction calculated with the TRIM
configurations for ITER [512] showed the expected target code [439, 440]. A fixed value, Ych = 0.01, is chosen as
loading of the secondary divertor to be below 2 MW m−2 , the standard value in the ITER modelling. This selection
provided that the distance between the separatrices is at least was reasonable given the spread of the experimental data on
2 cm at the outer mid-plane, and this requires no special chemical erosion of carbon materials [1]. Recent data suggest
target design. This allows simplification of the modelling that this modelling may have to be improved to account for the
configuration, and in the recent ITER calculations, the grid flux dependence of the sputtering yield [334] and the possible
is limited radially by the second separatrix, leaving a gap differences in sputtering yields of solid and re-deposited carbon
in the real wall. To account partially for the presence of a surfaces [515].
plasma in this gap and to simulate wall recycling and erosion, Recent development of the neutral transport model for
a gradient scale length of 3 cm is specified at the last magnetic ITER also involves inclusion of neutral–neutral and molecular-
surface boundary for all densities and temperatures (equivalent ion collisions, besides the elastic collisions of helium atoms
to specification of an outward convection with v⊥ = 10 m s−1 with plasma ions, in the calculations [516].
for all ion species, which contributes a constant 43% of the
energy outflow at the grid edge [513]), and a condition of non- 3.2.2. Modelling results. A number of modelling results
slipping (v = 0) is applied for the momentum equations. As on the performance of the ITER divertor have been obtained
S243
A. Loarte et al
in last five years [20, 29, 512–514, 516–518, 254] as new neutral compression near the separatrix strike point and to
experimental results have been obtained, further physical facilitate partial detachment [202, 254, 520]. Secondly, high
effects have been included and a design optimization of ITER gas conductance of the structures supporting the dome was
toward lower cost has been done. The primary aim of found to be essential for reducing the in–out divertor loading
the studies has been the identification of the major control asymmetry [225, 254].
parameters, the exploration of the operational window, the A power law parametrization of the results of 2D
investigation of the effect of divertor geometry and the modelling in terms of the control parameters described above
development of integrated modelling of the whole plasma by and the interface parameters at the separatrix such as ns ,
coupling of the edge and core models at their interface. The DT , cHe , cC and plasma temperatures has been developed
B2-Eirene code package, version SOLPS4.0 [519] with some [20, 513, 518] providing a set of boundary conditions for the
reactor-specific modifications [254], has been used in the ITER core plasma models and thus allowing consistent modelling
divertor modelling—see the previous sub-section for the model of the whole ITER plasma [521–523]. Moreover, the
details. results obtained with different assumptions on the surface
In ITER modelling, fast transients such as ELMs are not composition can be put in a linear order and be then unified
included explicitly. The resulting state, therefore, describes in the same scaling by introducing one more artificial input
a time-averaged base state perturbed by ELMs. Such an parameter [513]. Figure 28 illustrates the quality of the
approach, forced by the lack of a proper ELM model, can be obtained scaling [513]. Besides providing the boundary
justified by noting that in ITER less than 25% of the average conditions for core modelling, the identified scalings of
power is expected to be transferred by ELMs [5], and by the divertor plasma conditions also serve as a framework for
fact that the modelling region in the 2D calculations here does comparison between calculations done with different models,
not include the pedestal. It can of course not cover strongly allowing reliable quantification of the effect of different
non-linear effects arising from large individual ELMs—like modelling assumptions [516].
massive evaporation of target plate material—but these aspects
are dealt with in section 2.2. The results of the time-averaged 3.2.3. Future improvements to the physics model. A
modelling can be briefly summarized as follows. significant future improvement to the physics modelling used
Six major control parameters for the edge plasma are in the ITER modelling studies will be the incorporation of
identified as (1) power entering the SOL, PSOL —strongly drifts and currents in the edge plasma. Whereas a model
affects all the parameters; (2) D/T neutral pressure in the of the drift effects is already included in some versions of
divertor pDT —strongly affects the peak power loading on the B2-Eirene [124, 519, 524], this rather complex extension to
targets, together with the helium and carbon concentrations the code affects the running time and stability of the code,
at the separatrix, cHe and cC , and determines the degree of and its routine use for ITER modelling can be envisaged only
detachment of the plasma from the targets; (3) pumping speed once further tests and verification against experimental data are
and (4) fusion power—both affect cHe ; (5) the distribution carried out. Its incorporation would be expected to have only
of fuelling between gas puffing and core fuelling—affects a moderate effect on target loading in ITER. Indeed, since the
the neutral influx to the core DT , plasma density at the in–out power asymmetry for ITER is already 2:1 (Shafranov
separatrix ns , and cHe and cC and (6) the connection shift) [20], a further increase in this ratio would be similar to
length (differing between inductive and continuous operation an increase in the input power by 20%, i.e. within the operating
scenarios) represented through the q95 value—affects the window margins. A reduction of this asymmetry would lead to
upstream plasma parameters [20,513]. These simulations have a reduction of the power loading of the outer divertor, which is
shown that an operational window in these parameters, with not dangerous, and an increase at the inner divertor target,
acceptable power loading qpk 10 MW m−2 and sufficient which is anyway designed for the same power loading as
helium removal rate, exists and extends to PSOL at least 50% the outer one. The expectation that impurity transport in the
higher than the nominal ITER value of 86 MW. A certain SOL would be affected significantly is not supported by first
margin to cope with uncertainties and flexibility in divertor modelling results for ASDEX-Upgrade [524], although the net
parameters therefore exists. Recent modelling has shown that erosion–deposition pattern can change.
helium removal from the edge plasma is not a severe constraint, Several important items remain to be implemented in
once the elastic collisions of He atoms with D/T ions are taken the neutral transport model. These are Lyman radiation
into account [253, 254, 518]; this relaxes the requirements on transport and hydrocarbon generation and dissociation. The
pumping throughput. Taking into account the neutral–neutral first may have a considerable effect on saturation of the
and molecule–ion collisions shifts the operational window plasma density since excitation of neutrals during absorption
towards higher pDT , 5 to 15 Pa [516], thus alleviating pumping. of radiation (optical pumping) facilitates their ionization and
The modelling predicts saturation of the separatrix ion density changes the ionization–recombination equilibrium, resulting
as the throughput increases, as well as high opacity of the hot in higher ion density. This effect, already included in an
ITER SOL to neutrals, rendering necessary the implementation experimental version of Eirene [525], will soon be included
of core fuelling techniques such as pellet injection [20, 254] in in ITER modelling (see section 3.4). Explicit modelling of
order to maintain the required density in the core in ITER. hydrocarbons can be implemented in the future following
Modelling for ITER has indicated the importance of assessment and validation of the detailed surface interaction
two geometrical features in particular and these have been processes and would then allow improved modelling of
ensured in the design. First, a V-shaped intersection of tritium co-deposition. Another open question, not sufficiently
the targets and the divertor floor is found to enhance the covered by present experiments, is the currently planned Be
S244
qpk [MW/m2] qpk_# ff 0.6 P#-1.26 fw0.37 q95#0.5

5
16
3
8 1
0.8
0.6
0.4
0
0.1 0.2 0.3 0.4 0.5 0.4 0.6 0.8 1
20 -3
ne_sep [10 m ] m = –1.17 µ
Figure 28. Quality of the divertor parameter scaling [513]: the raw data in terms of qpk versus ns (left) and the same data after proper
normalization (right). The individual curves correspond to a density variation for different values of the input power (80 to 130 MW),
different operational modes (inductive versus steady state), different fuelling schemes and different assumptions on the erosion/deposition
−0.87 −0.8 −1 −0.27
properties of the walls. Here µ = pDT# PSOL# ff fw q95# is the normalized neutral pressure in the divertor, ff ∈ [1, 1.2] and
fw ∈ [0.7, 1.1] are the factors quantifying the fuelling scheme and the surface properties and the sub-script # means the normalized value of
the corresponding parameter. Further details can be found in [513].
main-chamber wall and its effect on overall SOL chemistry. between the inner and outer divertor in the private flux region
Beryllium hydride formation and its possible interplay with (PFR) to reduce the asymmetry of the operational conditions in
carbon targets will be addressed by dedicated research the divertors [225,254]. Therefore, whereas a fluid description
projects in the coming years. Initial experiments on PISCES of the neutral transport on the same grid as for the plasma would
[406] indicate that deposition of beryllium may significantly be attractive from the point of view of numerical efficiency, it
reduce the erosion of carbon surfaces, alleviating the tritium is not suitable for a predictive code.
co-deposition problem but raising an issue of necessary As a general requirement, the numerical error (or noise)
impurity seeding to ensure the required level of power radiation introduced by a code must be significantly smaller than the
from the edge. effect to be modelled. The transport equations on which the
In summary, replacement of the free parameters in the modelling codes are based [1] describe the balance between
model (plasma transport, boundary conditions, erosion rates the fluxes and the sources in the volume and on the surfaces,
and so on) by more detailed physical models on the basis of and an estimate of the required code accuracy can be obtained
improved model validation by the physics community against from a comparison of different terms. The major terms in the
experimental data remains a priority. energy equations are the input power, the impurity radiation,
the power spent for recycling and the power delivered to the
3.3. Development of predictive codes targets. In normal conditions these terms are comparable, so
that evaluation of all these power losses or use of the input
Predictive codes are tools that integrate the existing knowledge power as a boundary condition does not impose any special
of various physical processes in the edge plasma in order to requirements on code accuracy. The power transferred to the
project the physics to the range of plasma parameters of future first wall is lower, usually less than 10% of the input power, so
devices. In contrast to the codes normally used in interpretative that modelling of the interaction of the plasma with the first wall
mode to analyse experiments in present-day devices, codes with accuracy better than 10% requires accuracy of the power
used in predictive mode cannot rely on direct input from balance well below one per cent. In the momentum transfer,
experimental data and must, therefore, be self-consistent and unless the plasma is close to a full detachment, the parallel flow
comprehensive as far as possible. In order to avoid non- velocities are mainly determined by conservation of total—
physical solutions, they should use only externally controllable static plus dynamic—pressure [1], modulated to compensate
quantities as input parameters (see also section 3.2). for the drift flow divergence [124,524]. In the particle balance,
Of the codes satisfying the requirements above the most the major terms are the ionization and recombination sources
developed are the three major two-dimensional divertor codes: and the recycling plasma flow to the targets. The source terms
B2-Eirene, EDGE2D and UEDGE [1]. These codes are related to pumping and fuelling, the only quantities which can
widely used in both interpretative and predictive mode. They be controlled, are, e.g. in ITER, three orders of magnitude
incorporate a multi-species fluid description of the charged lower, and correct modelling of these processes thus requires
plasma components coupled to a Monte-Carlo model for that the particle balance be accurate to better than 10−4 . This
neutral particles. The latter is essential once pumping, fuelling is a challenge for a hybrid, fluid plus Monte-Carlo code and
and detailed plasma–material interactions are to be modelled requires modification of the coupling scheme [254].
(see section 3.5). Moreover, both modelling results and The description of the physical processes modelled by the
experimental data indicate the importance of the gas flow major codes has been augmented in recent years particularly
S245
A. Loarte et al
Figure 29. The power loading profile due to fast ions on the JET targets in D (- - - -) and He (——) discharges [531].
by inclusion of the drifts and currents in the edge plasma (a) ASCOT—a kinetic Monte-Carlo code which follows
[33, 124, 126]. This work is in progress, and more effort the fast ion orbits from the pedestal region, taking into account
is needed to validate the models against the experimental the electric fields in the plasma and a broad spectrum of fast
data. The initial results yield parallel particle flows with Mach ion collisions with the background plasma and neutrals [530].
numbers ∼0.2 to 0.5 in the upper part of the SOL [123,124,314] This code reproduces the experimentally observed sharp peaks
(without drifts, these Mach numbers are an order of magnitude of the outer divertor power loads in the H-mode discharges
lower). This flow is directed from the outer to inner divertor for on JET [24, 531, 532], figure 29, and shows that the fast ions
the usual B direction (ion B × ∇B drift towards the divertor), could be responsible for a significant fraction of the in–out
like in experiment, and roughly compensates the E × B drift asymmetry in the divertor power on JET. Initial results exhibit a
in the SOL [526], resulting in a toroidal rotation of the SOL high sensitivity of power load asymmetries to the radial electric
plasma. However, the experiments show factors 2 to 5 higher field in the SOL, taken as a free parameter in this modelling, and
Mach numbers [111, 314, 527], and the effect of the drifts on to the ion temperature gradient just inside the separatrix. Since
divertor asymmetry, especially with reversed magnetic field, the interaction of the fast ions with the background plasma
still remains to be demonstrated in the modelling [27, 528]. is rather weak, this code can be used as a post-processor for
Another important development is the introduction of predictive runs with the major codes, to improve the evaluation
neutral–neutral collisions in the Monte-Carlo description of of the target loading—especially when the model includes a
the neutral particle transport [260] and making it a routine full treatment of drifts and currents including self-consistent
part of the B2-Eirene code package [516]. This makes the electric fields in the edge [124]. Application of this code to
neutral transport model internally consistent for intermediate an ITER case modelled with B2-Eirene [26] shows a minor
(<10%) effect of the fast ions on the target loading once a
collisionality range, where the usual linear approximation of
realistic pedestal width is assumed (the plasma profiles just
molecular flow fails, and improves correspondence between
inside the separatrix determine the source of the fast ions in
the modelling results and experimental measurements of the
the SOL). However, it should be noted that the attempts to
gas pressure in the divertor of Alcator C-Mod [221] (see also
reproduce with ASCOT the distribution of the divertor power
section 3.5).
loading in the experiments with reversed magnetic field on JET
As concerns the numerical aspects of the code were unsuccessful [19]—probably because of using the same
development, beyond the incorporation of the drifts and values of radial electric field as for the standard case.
currents [33, 124, 126] and improvement of the particle (b) HEIGHTS [533, 534] and FOREV-2 [157, 535]—two
balance [254] already mentioned above, the introduction of code packages with similar scope, describing the interaction
the adaptive grids in B2-Eirene [519] has been an important of a strong pulse of heat and particles with the targets,
advance. This method allows better space resolution in the together with the transport of the erosion products and the
regions of interest at reasonable total grid size. A thorough radiation transport. Whereas the HEIGHTS package looks
comparison of the three major 2D codes (B2-Eirene, EDGE2D more developed in terms of detailed description of the target
and UEDGE), which employ different discretization schemes structure, the FOREV-2 code puts more emphasis on a real
and iterative solution methods [1], is still at an early stage [529], magnetic geometry of a tokamak. The codes reproduce
so that the advantages or disadvantages of the different experimental results on the radiation and heat transport from
approaches remain to be evaluated. the Mk-200UG plasma gun for pulses corresponding to the
The inclusion of a certain class of effects, such as disruption loads expected for ITER [534, 535]. For ITER,
propagation of fast ions leaving the core plasma in the SOL or both the codes predict a formation of a dense vapour shield at
interaction of short heat pulses with the targets during ELMs a disruption, with radiation being the main channel of heat
or disruptions, in a major code is impractical because of their deposition on the divertor structures, whereas the efficient
weak interaction with the edge plasma or very different time vapour shield does not form during ELMs [536, 537]. Further
scales. Modelling of these effects requires dedicated codes, validation of the codes against the disruption and ELM data
three of which have been developed recently. from tokamak experiments remains a priority.
S246
In summary, the development of the predictive codes since nA is typically 1013 cm−3 . However, this is not
incorporates progress in the theoretical and experimental the case for very high gas densities (Alcator C-Mod) or
exploration of edge plasma physics. Recent activities, for sufficiently large divertors (JET, ITER). The neutral gas
stimulated by the observation of plasma flows with M ∼ 0.5 temperature and degree of dissociation are also important
to 1 (see section 2.1.4) along the magnetic field in the SOL because only the neutral atomic component can re-absorb this
and of very narrow peaks of divertor power loading, have resonance-radiation. Radiation transport effects lead to an
resulted in the development of the major edge codes to include enhanced degree of ionization through ‘optical pumping’, and
a comprehensive description of the drifts and currents in the consequently, reduced neutral–ion friction. Quantifying them
SOL and in the development of a kinetic code which describes is an essential ingredient for understanding and, ultimately,
the transport of fast ions in the SOL. Current studies of several quantitatively predicting the operational window for divertor
other effects, such as enhanced cross-field transport due to detachment in ITER.
a long-scale turbulence in the outer SOL (see sections 2.1
and 3.7), have not yet reached a state which would permit the 3.4.2. Experimental evidence for opacity. Solid experi-
results to be incorporated directly into the predictive codes. mental proof for the resonance line re-absorption is obtained
Furthermore, a comprehensive and verified model to predict from a modification, with a variation of the neutral density
ELM behaviour is urgently needed, but the underlying physics of line intensity ratios along the same spectrometer line-of-
remains yet to be established. sight, such as between a Lyman line and another line with the
same emitting state (e.g. Lyman-β and Balmer-α). Using this
3.4. Radiation transport method, opacity effects on the resonance Lyman lines have
3.4.1. Introduction. In most quantitative assessments of been observed experimentally, in particular for the very dense
ITER divertor plasmas (see section 3.2) the radiation field has Alcator C-Mod divertor [258] and in MARFEs [538].
been regarded as entirely decoupled from the particle fields, i.e. By a similar method, JET divertor plasma conditions
the plasma is assumed to be optically thin. However, various have been identified as marginally optically thick for Ly-
spectroscopic studies have shown clear evidence of opacity lines [263, 539], whereas the quoted Alcator C-Mod results
(line absorption) effects in the lower Lyman series. In order to show clear evidence of strong line re-absorption, with implied
quantify these effects for the ITER divertor design, ‘radiation values of nA as high as 1020 m−3 [258].
transport code’ development projects have been initiated. Less direct observations, e.g. from line-shape analysis
The opacity τ = L/λmfp (inverse Knudsen-number for or from experimental determination of populations of highly
photons) scales linearly with nA · L, the product of number excited states of atoms and their deviation from collisional
density nA of radiation absorbing particles (in this case, neutral radiative modelling, have also been reported, see [540] for a
hydrogen atoms) and the average chord length, L, of the summary and further references.
absorbing layer. λmfp is the monochromatic photon mean free
path. The photon mean free path at the central wavelength λ0 3.4.3. Theoretical predictions. Opacity effects have been
of a Doppler broadened line is discussed for a long time as possible contributors to overall
divertor dynamics in high density regimes. See [541]
λmfp (λ0 ) = (5.4 × 10−9 · λ0 · fnm · (µ/T )1/2 · nA )−1
for one such early example (INTOR phase), although
∼ 1.3 × 1013 · T 1/2 n−1 −1
A f1m [cm]. (3.1) note that the neutral densities used in the calculations
In the right-hand expression we have inserted the typical were two orders of magnitude larger than are observed in
wavelength for Lyman lines, λ0 ∼ 10−5 cm, and µ = 2, the present experiments. Computational assessments have
the mass of the emitting particle relative to the proton mass. typically been based upon highly idealized, often zero-
Here T is the temperature [eV] of the neutral atoms, nA is dimensional approximations (so-called ‘optical escape factors’
their ground state density [cm−3 ] and fnm is the oscillator in collisional radiative models) [542] or one-dimensional (slab)
strength (f1m = 0.4162, 0.079, 0.029, . . . for the Lyman approximations [543]. Detailed photon transport procedures
series). Radiation transfer is probably the best established are well established in many fields of science, e.g. in
edge plasma transport mechanism, in terms of accuracy of astrophysics, and for high pressure discharges used for lighting
data: the Einstein–Milne coefficients and Doppler broadening purposes. Considerable effort is required to adapt these
for hydrogen are certainly much better known than any other specialized codes to typical fusion edge plasma conditions,
atomic or molecular cross section. On the other hand, the especially configuration details. Some steps have already
photon mean free path (defined here for a fixed wavelength) is been undertaken, e.g. with the CRETIN code [538, 544–546]
an extremely steep function of the photon energy, varying by and with the EIRENE Monte-Carlo Code [224, 525, 540, 547],
many orders of magnitude within only a few Doppler widths which are already in use and well adapted for ITER design
of the line profile. Therefore, the detail of the line profiles studies as a neutral particle module within the B2-EIRENE
is important, with only the near line-centre photons being code package (see section 3.2).
affected in the divertor. This effect, along with differences In the CRETIN code, rather simple geometry (slab) and
in the local conditions of the emitter and absorber, turns plasma models are typically used, and the emphasis is on
modelling of the radiation transport into a rather involved self-consistency between the radiation field, the non-LTE gas
bookkeeping issue often requiring Monte-Carlo methods for conditions and the (1D) plasma. All relevant line broadening
solution. mechanisms are included. In particular, Zeeman splitting in
Equation (3.1) shows that plasmas in most of the current a strong B-field is fully implemented, along with the Stark
divertor tokamaks are optically thin to the Lyman lines effect. Significant changes in the steady-state plasma solution
S247
A. Loarte et al
are reported when the Lyman line re-absorption is taken reaction mechanisms, governed by specified cross sections, are
into account, as compared with the conventional optical thin often involved; sometimes rich chemistry, e.g. hydrocarbons,
assumption from the present edge modelling [548]. must be invoked. In particular the Monte-Carlo method of
In the EIRENE code, the advantage is exploited that the solution is routinely applied for ITER design studies with the
radiation transfer equation is mathematically analogous to EIRENE neutral transport code [547, 549] (see section 3.2).
the linear Boltzmann equation for neutral particles solved by Such ‘recycling models’ can properly account for arbitrary
EIRENE. As a result, a major part of the existing code can be geometrical and physical complexity.
used directly for photon transport problems, with only minor In principle, the factors limiting the predictive quality for
modifications. EIRENE has been run on a set of converged B2- a given host medium (plasma) are the remaining uncertainties
EIRENE simulated ITER divertor plasma conditions, with and in some atomic, molecular and plasma surface interaction data
without line absorption effects taken into account. Although a (see below). In practice, however, it is often quite difficult
self-consistent feedback of the changes in the neutral divertor to anticipate the necessary level of sophistication required (at
pressure caused by radiation trapping has not been studied up the expense of CPU-performance, usually) in any particular
to now, the neutral gas pressure and neutral–ion friction in the study. This makes the necessary pre-selection of tolerable
divertor are always found to be reduced significantly (factors model approximations a rather critical step in predictive neutral
between 2 and 5) [540] as compared with the assumption of gas transport modelling studies. It probably will always have
optical thinness. The dominant line broadening mechanism to be carried out on a case by case basis.
in these studies is Doppler broadening, with the linear Stark
effect and natural line broadening playing only a minor 3.5.2. Transition to low temperature hydrogen plasma physics.
role [525]. However, the opacity effects may have been With the transition of envisaged ITER divertor plasma states
overestimated here because Zeeman line splitting is not yet from the high recycling, ionizing phase (until the early nineties)
implemented, and they might be further reduced in mixed towards a (partially) recombining detached plasma solution, a
D–T cases where the Doppler line profiles would no longer large number of new atomic and molecular processes become
overlap significantly at the 1–2 eV gas temperature observed important, which have been safely ignorable under fully
in these simulations. Benchmarking and validation of the ionizing plasma conditions.
radiation transfer module in EIRENE (excluding non-LTE Molecular hydrogen (and molecular hydrogen ions), in
effects and Zeeman splitting) has recently been accomplished particular vibrational kinetics, and resulting resonant charge
by the application of a coupled photon–plasma code FIDAP– exchange between unlike particles (protons and vibrationally
EIRENE to high intensity discharge plasma lamps (which excited hydrogen molecules) start to play a dominating role
operate at pressures of 20–50 bar and are clearly in LTE) in in the degree of dissociation and the ionization-recombination
a joint effort with the lighting industry. Assessment of the balance in the divertor (see also section 2.3.3, on molecular
code developments through comparison with measurements assisted recombination ‘MAR’).
from the fusion experiments, particularly JET and Alcator The first Fulcher band divertor spectroscopy (i.e. on the
C-Mod, is underway. It, therefore, seems realistic to assume molecular counterpart of atomic Balmer-α emission) carried
that within the next few years ad hoc opacity assumptions can out in the ASDEX-Upgrade divertor [550] had initially led
be eliminated from current divertor codes. to mismatches between the experimentally deduced and B2-
EIRENE simulated molecular densities (a factor of up to 20),
3.5. Neutral particle transport first attributed to a possible role of MAR at that time [550].
Later this discrepancy diminished after replacing the
3.5.1. Introduction. The mutual interaction of neutral gas initial corona assumptions in the experimental Fulcher band
dynamics and the plasma flow provides access to favourable data interpretation by the full collisional radiative model
plasma conditions in divertors, such as the high recycling and for the p-H-H2 -H+2 system [260, 551] used in the EIRENE
the detached plasma regimes. This is similar to other areas molecule transport model. Apparently, not only the atomic
of (computational) fluid dynamics, where there is also often hydrogen (long known regarding high recycling divertors)
a tightly coupled fluid/solid interface. In edge plasmas, this but in detached states also the molecular component requires
interaction not only sets the boundary conditions but dominates detailed non-corona and non-LTE treatment. Corresponding
the exchange of matter, at least for the divertor operating models have been set up [551, 552] and included in the codes
conditions currently envisaged for ITER. [260, 553] and are further being refined currently.
Whereas the basic physical process for the edge plasma, As discussed in section 2.3.3 for the case of MAR,
the Coulomb interaction, is simple, the formulation and vibrational kinetics and hydrogen chemistry is a well-
numerical solution of tractable plasma fluid equations—the established mechanism in low temperature gas discharge
ordering of terms and the identification of a set of consistent plasmas, in which the neutral gas density is set as an external
boundary conditions—is still a very active and rapidly evolving control parameter. In divertors, its role is far less clear
field of edge plasma research. The situation is reversed in the because the molecular pressure is set at a value self-sustained
case of neutral gas transport, which provides this mechanism by the recycling at the targets. As a result, for example,
for exchange of matter, momentum and energy between the in B2-EIRENE simulations of an ASDEX-Upgrade detached
solid, gas and plasma states. This transport mechanism is divertor scenario, the inclusion of the additional vibrational
ruled by the well-known, and often even linear, Boltzmann kinetics (required as precursor for MAR) had led to a strong
equation. Corresponding linear transport theories and their reduction in the self-sustained neutral pressure, an overall
diffusion approximations are well developed. Rather complex reduction in the volume recombination rate (rather than the
S248
enhancement initially predicted from idealized, separated and uncertainties in experimental identification of the weak private
strongly simplified arguments, see [1]) and even a subsequent flux plasma all seem to matter simultaneously. For example,
re-attachment (into a high recycling divertor state) under for the Alcator C-Mod divertor only by a lengthy step-by-
otherwise identical edge model parameters and upstream step model upgrade could an initial mismatch of experimental
boundary conditions [268]. and modelled pumping plenum pressure of a factor 10 be
This modelling result was consistent with the abovemen- reduced to a (still) remaining factor 2 [221,558], see figure 30,
tioned detailed Fulcher band divertor spectroscopy at ASDEX- where the toroidal non-uniformity of the neutral pressure is
Upgrade and indicates that in the current recycling models, illustrated, depending on whether the nearest port is open
without the molecular kinetics, the volume recombination or closed. Note that ignoring the (known) effects of 3D
rate and upstream parameter window may not be underesti- divertor leakage pathways and resonance radiation trapping
mated but instead overestimated. The original choice in ITER gave excellent agreement between the model and experiment,
divertor modelling to neglect them altogether seemed, there- apparently by coincidence. However, this only demonstrates
fore, justified. Indeed, first results of ITER modelling with
that matching experiments with simplified models can be
the molecular kinetics taken into account show no significant
strongly misleading.
effect of MAR [516]. Further quantitative assessments are
currently being carried out. A confirmation of the sensitivity to private flux region
Observations (B2-EIRENE model versus experiment) physics stressing the need for further neutral gas model
confirming this picture of ‘enhanced dissociation’ rather than refinements was also provided by a predictive modelling
‘enhanced recombination’ due to vibrational kinetics have ‘design’ study (again: with B2-EIRENE) on the ASDEX-
been reported from TCV [554, 555]. The initial application of Upgrade divertor modifications. Strong effects on in–out
the B2-EIRENE code to the experimentally observed transition asymmetry in neutral flux densities had been predicted for
to detached states showed that the volumetric recombination the private flux region [247], but no such trends could
rate used in the code model had to be increased artificially by later be found experimentally after completing the hardware
a factor of 5 over the (quite well established) true atomic rates upgrades [152]. Current coordinated research projects on
used in EIRENE for recombination. Vibrational kinetics in H2 low temperature hydrogen plasma chemistry and on transport
transport, and hence ‘MAR’, had not been included in these are also being initiated to improve the predictive quality of
first simulations. However, after activating the vibrational numerical divertor models in this essential aspect.
kinetics, the same trend appeared as in the ASDEX-Upgrade
study, where detachment was even harder to achieve with MAR
than without. 3.5.4. Concluding outlook. Many or perhaps even most of
Therefore, other mechanisms must be active in TCV the relevant processes between neutral particles in the edge
divertor detachment, which have yet to be explicitly identified. plasma in present machines and in ITER, which are relevant
Rather complex hydrocarbon chemistry could play such a for the overall divertor plasma dynamics, are known, even
role [556], and is, amongst other possible contributors (‘blobby quantitatively, and could, in principle, be included in the
transport’), currently being investigated by code simulations modelling codes in great detail. There have been, and still
and further experiments. are, continuing coordinated efforts to improve the atomic and
As a result of the clear signatures of complex hydrogen molecular databases and surface interaction models used in
(and hydrocarbon) chemistry, significant efforts have been neutral gas modelling. Subsets of these databases have been
undertaken to upgrade the corresponding atomic and molecular or are currently being implemented in neutral particle codes
databases used for plasma edge modelling. Most recent (and impurity codes). A sufficiently detailed set of atomic and
compilations of hydrocarbon data are available in [437]
surface physics can and will most probably be implemented
(methane family and their ions) and [330] (ethane and propane
first in the rather laborious but physically transparent
families and their ions). The current hydrogen chemistry data
Monte-Carlo codes currently used for ITER divertor design
status (p-H-H− -H2 -H+2 -H+3 ) has been compiled and critically
studies.
assessed in [557].
However, an ‘a priori’ selection of the relevant processes
to be included for a particular predictive study, in order to
3.5.3. Sensitivity to configurational and physical details.
devise a sufficiently detailed but still ‘simple as possible’
The baffling of the neutral gas in some places, either by
the action of the plasma flow or by divertor hardware, and model, remains a critical task. Amongst these issues, which
the resulting gas conductance between different places in the may matter in some cases but not in all, are proper timescale
divertor (e.g. inner to outer leg) are essential design criteria for separation in collisional radiative approximations, details
particle removal, helium pumping, etc. Present self-consistent of the particle–surface interaction, surface accommodation
plasma–neutral gas models seem to work quite reliably for factors for momentum and energy, 2D versus 3D geometry
those aspects of the divertor dynamics which are dominated for pumping efficiency and divertor leakage studies, main-
by the direct plasma surface contact and recycling from near chamber recycling, recycling at parallel targets (baffles), the
divertor strike points and the outer SOL. If the private flux onset of viscous effects in the gas, radiation trapping (see
plasma is involved in a controlling way, agreement between section 3.4), etc. These issues continue to require major edge
experiment and plasma–neutral gas simulation is much harder physics studies (also related to the recycling and self-sustained
to achieve. Configurational details, low temperature hydrogen neutral gas build-up) in many tokamaks, covering a wide range
plasma chemistry, the onset of neutral gas viscous effects and of densities, temperatures and configurational aspects.
S249
A. Loarte et al
10.8 1.8
8.1 1.3
nD (1020 m-3)
nD (1020 m-3)
5.4 0.9
2
2
2.7 0.4
(a) (b) 0.0

0.0
Figure 30. 3D neutral gas modelling for Alcator C-Mod [221, 558]: toroidal gap and gussets, bypass and poloidal gap are included in order
to calculate toroidal variations in the divertor neutral pressure. The refinement of 3D structure, SOL and PFR plasma solution, atomic and
molecular processes and viscous and radiation trapping effects have led to improved agreement with the measured divertor pressure (factor
of 10 initially, now about a factor of 2). Below left: gas pressure at location of open port, below right: gas pressure at location of closed port,
different colour scale.
3.6. Impurity modelling The modelling of carbon behaviour in tokamaks

continues, generally, to be one of the most pursued—and hard
Modelling of impurity sources and transport in the edge has to explain—areas in edge physics [321, 559–562]. Modelling
always been an important component in edge research and of the JET co-deposition process [272, 273, 441] increases the
this effort has continued across a broad front since the 1999 challenge because of the wide range of edge processes playing
ITER Physics Basis report. In the short space available no a role. Three main steps are evidently involved: (a) the primary
attempt is made at full coverage of this substantial ongoing source of the carbon in the co-deposits appears to be the main-
activity; but only some new developments since 1999—and chamber walls, (b) this carbon is then transported as ions along
which are directly pertinent to ITER—are briefly outlined. the main SOL, and across the private flux zone, to the inner
Impurity transport is modelled in order to understand its effect divertor where it may initially deposit on the inner target, (c)
on tokamak operation and tritium inventory control. The it may then be removed by some process and transported as
impurity level in the core and SOL affects radiation and fuel neutrals to remote, cool regions, out of plasma contact, e.g.
dilution. Co-deposition of impurities with tritium on first-wall the JET louvres. Each of these main steps is characterized by
surfaces can lead to unacceptably high levels of T retained in major questions. Is the source due to carbon eroded by charge-
the vessel. The latter area has been the major emphasis since exchange neutrals or by plasma contact? Does the sputtering
the IPB. occur mainly during ‘normal’ plasma conditions, i.e. the flat-
tops of standard discharges, or during transients such as ramp-
The experience of the JET 1997 DTE campaign shows
up, ramp-down, ELMs, disruptions? Is the wall, more or less
how important C/T co-deposition can be [272] (see also
uniformly, the source or the leading edges of specific edge
section 2.6.3). The co-deposits formed a remarkable pattern:
structures? Although there is now evidence of the existence
the great majority occurred on water-cooled louvres in the of flows that could carry the C to the inner divertor, the basic
pumping recess facing the inner divertor target—thus out of cause has not been clearly identified. This is probably due to
direct plasma contact. Since such an effect potentially provides some combination of drifts [126] and closing of particle ‘loops’
useful opportunities for recovering the T, it is important to driven by radial transport [114]. Finally the process by which
understand how the C and T arrive there and stick. Without neutral carbon is transported from /near the inner target to the
adequate understanding, extrapolation to ITER is uncertain. louvres is not understood.
S250
-130 CD4 density -130 C+ density

-140 -140
ZC (cm)
ZC (cm)
-150 -150
-160 -160
-170 -170
-180 -180
220 230 240 250 260 270 280 290 300 220 230 240 250 260 270 280 290 300
RC (cm) RC (cm)
Figure 31. Simulated CD4 (left) and C+ (right) particle density after chemical erosion of CD4 in the divertor of JET Mk IIA under the
assumption of fully-sticking hydrocarbons [441].
Because of these major uncertainties the modelling to above. In the cold plasma conditions of the detached inner
date has often been characterized by a substantial ad hoc, divertor, chemical sputtering processes have been assumed
prescriptive component. The modelling effort has been mainly to dominate. These codes follow in considerable detail the
on the carbon impurity itself, taking the ‘plasma background’ chemical sputtering processes releasing methane and heavier
as specified. For example, onion-skin modelling (OSM) with a hydrocarbons from the inner target, employing the most recent
heavy prescriptive aspect for the background plasma was used compilations of rate coefficients for the break-up kinetics
to approximately match the measured divertor spectroscopic [565–567]; the transport and break-up of the hydrocarbons,
signals for what was taken to be a representative JET shot [563]; both charged and neutral fragments; the reflection/sticking
in addition, more rigorous modelling, B2-EIRENE/SOLPS of the fragments which strike solid surfaces, based on new
and EDGE2D [564], has also been used. The inner divertor molecular dynamics code simulations [438, 568, 569]; all
is the most important but also most problematical plasma within the detailed spatial structure of the JET inner divertor
region to describe owing to the fact that it is typically and louver region. Figure 31 shows some sample results of
detached and because of the paucity of reliable diagnostic this analysis [441].
data (i.e. 2-D fields of ne , Te , Ti and v ). Of course, this Despite the large amount of empirical input, the
modelling of the background plasma is predicated on the REDEP/ERO-JET modelling using standard assumptions has
questionable assumption that a ‘standard’, as opposed to an not been able to explain the large magnitude of the co-deposits
‘off-normal’, plasma condition is appropriate and has been found on the louvres: calculated tritium co-deposition rates for
identified. With regard to the impurity modelling itself, the the JET Mk IIA divertor, using ‘reference’ chemical erosion
magnitude of the louver deposits shows that the C-throughput yields of order 1%—while higher than previously estimated—
rate can be estimated to be ∼4–6% of the total D+ -flux to are well short (by ∼1/40X) of published data [273]. Because
the inner target integrated over the DTE1 campaign [272]. all of the basic processes involved have their own significant
Accordingly, in the modelling, a wall source of carbon uncertainties, sensitivity tests were performed to see what
release was simply prescribed to give this magnitude without changes from the various ‘reference’ assumptions would be
attempting to actually model the release mechanism. In order required to bring the modelling and the measurements into
to convey the C-ions efficiently to the inner divertor in the agreement. It was found that a reasonable agreement between
OSM modelling, a strong parallel flow was imposed as part the simulation and the experimental observation of the carbon
of the ‘plasma background’, and adjusted—as to speed and deposition at the inner louvres in the Mk IIA JET divertor is
spatial extent—until essentially all the C-ions reached the achieved, if high chemical yields of about 20% together with
inner divertor, in accordance with the measured deposition a negligible effective sticking of hydrocarbons are assumed.
pattern. It turned out, however, that the substantial flow speed Although there are experimental observations indicating such
required to achieve this, i.e. Mach numbers ∼0.5, was in accord high erosion yield, such yields cannot be applied as a stationary
with subsequent measurements by Langmuir Mach probes. value for the effective erosion since it would turn the inner
13
CH4 injected at the top of the JET SOL resulted in 13 C divertor into a net erosion area in contrast to the experimental
deposits in the inner divertor [310], also confirmatory of this findings [441]. Indeed, the chemical erosion yield of the
picture. It is nevertheless the case that the first two parts of the carbon depositing on the inner target may be much higher than
impurity ‘modelling’—source and large-scale ion-transport— that for the base graphite [321].
are, to date, essentially prescriptive and thus not adequate for With regard to tritium–carbon co-deposition, it is evident
predicting future performance (ITER or otherwise). that we are presently far from a predictive capability for ITER:
To date the most ambitious part of the modelling effort the uncertainties about the physical processes involved and the
has been on the 3rd step, the neutral transport to the louvres. experimental database are still too large for reliable predictions
For this analysis work, sophisticated impurity Monte-Carlo of these processes for future devices [273]. The principal
codes have been employed, REDEP [273] and ERO-JET [441], limitation to progress in this type of modelling appears to be
whose field of focus is restricted to the inner divertor and louver a lack of detailed measurements of all the relevant processes
region. These codes have taken as input the prescribed ‘plasma in the divertor region, rather than the modelling tools directly.
background’ and C+ -inflow into the inner divertor, as described For example, establishing the original source of the carbon
S251
A. Loarte et al
Tile length [cm]

0 10 20 30 40 50 60 70 80 90 100 110
18
Z [m]
16 1.0
14
12
W [1015/cm2]
0.5 central column

10 heat shield
8
0
6
4
2 -0.5
0
inner baffle
strike point strike point outer baffle outer baffle
inner baffle area area inner strike outer strike
< 0.5×10 /cm
15 2
point area dome point area
inner divertor outer divertor baffle
dome baffle
Figure 32. Measurements of deposited W in the divertor of ASDEX-Upgrade after the first stage with W tiles only at the lowest heat shield
tiles. One can see a strong deposition at the inner baffle and the inner strike point position. There is also some W deposited at the outer
divertor [292].
in the co-deposits, spatially and temporally, is essentially an of the different charge states at the entrance to the sheath; (c)
experimental issue, requiring much more extensive diagnostic the individual average parallel speeds of the different charge
capability. Interpretation of deposition processes that occurred states at the sheath-entrance. Each of these factors can have a
over an entire campaign is difficult; it may be, for example, large influence on the impact energy, and thus the sputtering
that the louver co-deposits were formed during the cleanup yield. A further complication of such mixed-element situations
procedure when the strike point was intentionally swept along is that the fractional coverage of each different portion of solid
the divertor surfaces to heat them; time-resolved deposition surface needs to be known, something that has to be computed
studies, now in progress, will be invaluable. self-consistently within the impurity modelling itself [571].
The ITER design calls for an extensive use of W at Our present capability to predict ITER impurity behaviour for
the divertor region (initially only at the baffles) and, thus, high-Z impurities is, therefore, even more challenging than for
modelling studies of W transport are being actively pursued. carbon, since the low-Z impurity has to be modelled as well.
An example of modelling of W on wall surfaces in current For helium impurity transport, the importance of elastic
tokamaks is that of ASDEX-Upgrade [292]. In the first phase collisions of He atoms with background ions of hydrogen
of W installation in ASDEX-Upgrade, only the lower portion isotopes has been shown experimentally [253]. Fast
of the heat shield was W-coated, see figure 32. After a one- thermalization of He atoms in the JT-60U divertor found from
year campaign the erosion of the W-coated carbon tiles as well
spectroscopic data is well reproduced in modelling when the
as the migration of W onto the uncoated tiles in the divertor
He elastic collisions are taken into account. These collisions
were measured using ion beam surface analysis. Modelling of
have a beneficial effect on the expected helium removal
the W sputtering and edge transport has been carried out using
efficiency of the ITER divertor [254] (see also section 3.2).
the impurity Monte-Carlo code DIVIMP, extended to allow for
simultaneous plasma contact with the two divertor targets and In contrast to the above-described work on modelling
the heat-shield wall [292]. The ‘plasma background’ required impurity sources and their transport in the SOL there has
as input for DIVIMP was taken from the B2/EIRENE analysis been less effort dedicated to modelling impurity transport
for a well diagnosed discharge—albeit one which might not into the core. This is possibly because the radial transport
be representative of the average plasma exposure conditions. is critically dependent on the parallel transport in the SOL
A constant W-flux density from the wall was assumed. It was which, as discussed above, is in itself already problematic.
found that the calculated W-deposition profile shape was in An example of modelling core impurity levels is based on
reasonable agreement with that measured. JET screening experiments, see [572] which reference also
Modelling the physical sputtering of high-Z elements such provides an extensive set of references for methane injection
as W is particularly challenging since the principal mechanism experiments including studies on core carbon transport, plume
is the impact of light impurity ions [403, 570], such as C and characterization, erosion and re-deposition, molecular break-
Be, which will be present for the current ITER design (the up studies, and evaluation of carbon penetration through
sputtering threshold for D/T-ion impact is so high that this the SOL to the core. The JET work was fairly successful
source can be neglected). It is therefore necessary to first in reproducing the scaling of C impurity levels with core
model the light ion impurity behaviour, and in considerable density and input power. This type of modelling was also
detail. It is necessary to know (a) the mix of charge states applied to Alcator C-Mod with limited success in reproducing
entering the sheath (since the different charges experience experimental results showing poloidal variations in impurity
different sheath accelerations); (b) the individual temperatures screening.
S252
Impurity behaviour is one of the most challenging aspects free streaming particles and collisions. The parallel collisional
of edge physics to understand and model since it is first mean free path Lcoll is then the characteristic transport scale.
necessary to sort out, in some detail, many or most aspects Furthermore, it is assumed that the particle life time is governed
of the ‘plasma background’, i.e. the fuel plasma, including by the parallel transport to the plasma sink at the wall, this leads
such complex phenomena as ELMs and when/where erosion to another characteristic scale, the connection length L . For
& re-deposition occur (e.g. transients versus steady-state the energy transport, we consider collisional diffusion in both
portions of the discharge). Advances in understanding should the parallel and transverse directions. One then readily obtains
benefit from the acquisition of more detailed experimental λq = ρe (L /Lcoll ) where the ratio L /Lcoll is the number
data describing the plasma characteristics as well as that of of collisions during a life time of a particle in the SOL and
the impurities. On such a basis, one has the chance of where the electron Larmor radius ρe is the characteristic step
developing sound understanding with the ultimate objective in the transverse direction. For typical SOL parameters, one
of achieving reliable predictive capability. One technique finds that this estimate of λq is orders of magnitude smaller
used for interpreting such detailed data involves so-called ‘as- than the experimental observations. In the case of electrostatic
simple-as-possible plasma’ experiments [573]. turbulence, we assume that the electric potential eddies have
the same characteristic scale in both transverse directions to
the magnetic field. The density e-folding length is then found
3.7. Simulation of cross-field transport of the tokamak SOL to be a balance between a Bohm-like diffusion and the parallel
3.7.1. Intermittent transport in the SOL. Cross-field life time of particles, λn = gB (L ρs )1/2 . The key parameter
transport is a key physics issue in all critical problems of the here is gB , the ratio between the magnitude of the electric
SOL and divertor plasmas. This physics still remains an open potential fluctuations and Te /e, the electron temperature Te
question although significant knowledge, both experimental divided by the charge e. We have used here the hybrid ion
and theoretical, has been built along the years. In the modelling Larmor radius ρs = cs / i , where cs = (Te /mi )1/2 is the
effort that is crucial for ITER, the cross-field transport is parallel sound velocity and i = eB/mi the ion cyclotron
taken into account by an ad-hoc diffusion coefficient [574]. frequency, mi being the ion mass. For large values of gB ,
Indeed, it was considered that this cross-field transport could gB > 0.1, this estimate of the SOL width is of the order
be addressed as due to a random process. In the limit of a of magnitude reported experimentally. Concerning parallel
very large number of events this approach would lead to a transport, three aspects are highlighted by the analysis. First
there is a strong geometrical dependence that is characterized
diffusive process in the frame of mean field theory. Tied to
by the connection length; second, the large parallel transport
this point of view is the idea that the correlation length, that
will organize transport in such a way that there is little variation
characterizes the steps of the random process, is small, in
along the field lines, leading to the so-called flute effect;
particular much smaller than the SOL width. This point of
third, the collisional mean free path is comparable with the
view is to date the only available description that can be used
connection length so that kinetic features should be accounted
in the extensive modelling effort required for ITER. However,
for. Furthermore, the parallel electron transport will play
this description cannot account for the growing experimental
a strong role in the electric potential fluctuations either via
evidence that transport in the SOL exhibits intermittent events
the parallel resistivity or via plasma boundary effects at the
that are not properly accounted for by a diffusive process. An wall. Finally, in the case of electrostatic turbulence, energy
outstanding experimental aspect is that fluctuations, typically transport will be governed by the same transverse Bohm-like
density fluctuations, are large and comparable with the mean diffusion but with a shorter parallel life time for electrons.
density [575]. This fact directly contradicts the theoretical The electron heat channel is then expected to be (mi /me )1/4
basis that leads to the description of transport in terms of a narrower than the density and ion energy channel. The goal
diffusive process. of electrostatic turbulence simulations is to determine the
Numerous issues in the ITER design involve cross- geometry and magnitude of the electric potential fluctuations
field transport, the most important being the width of the given the various instability mechanisms that can play a role.
energy channel onto the divertor target plates and the long Let us now consider the conceptual difficulty introduced
range particle transport that might govern recycling in the by the fluctuation level of order 1. This signature of SOL
main chamber. These two effects are characterized by cross-field transport questions the understanding in terms of
two scales in the transverse direction to magnetic surfaces, local equilibrium and in particular of detailed balance in the
namely λq the energy channel width and λn , the density master equation governing transport between the magnetic
e-folding length. Although there is significant experimental surfaces. As a consequence, there is no general theoretical
evidence that both energy and particle transport are governed basis for a diffusive process and for scale separation between
by electrostatic turbulence, let us introduce the key physics the transport events and the system size. The very meaning
with a rough estimate of these two SOL widths, either of profiles could also be questioned. This first principle
governed by collisional transport for energy transport or by description of a complex cross-field transport is also backed
electrostatic diffusion for particle transport. The definition by recent experimental evidence of long range transport in the
of the SOL width, namely the volume of open field lines SOL [58, 77, 82, 128], and intermittent transport events in the
where plasma ‘wets’ wall components, introduces several SOL [49, 86–89, 93, 576, 577].
difficulties. Indeed, the SOL appears as a region governed The power loading on the divertor tiles, and, via this
by a balance of two transport processes, parallel transport technological constraint, the operation point of high power
to wall components and cross-field transport. It is assumed devices is usually characterized by the parallel heat flux e-
that parallel transport is classical, hence described in terms of folding length λq . Given our previous conclusions, one readily
S253
A. Loarte et al
understands why there is, to date, no consolidated empirical or important topics are presently investigated. For instance, in
theoretical approach to scale the present experimental results the DIII-D geometry, a pedestal formation is reported due to
to ITER [578]. The question is then to determine the level of the neutral influx, even for L-mode parameters [587]. For
confidence one has in the modelling effort of the ITER plasma– the same DIII-D geometry, the existence of sheared poloidal
wall interaction. flows is analysed numerically and compared with the available
experimental data [588]. A similar effort is carried out for
3.7.2. Present modelling effort of cross-field turbulence. We C-Mod where a particular mode structure is documented [589].
shall now concentrate on the various theoretical analysis of the An original effort has been initiated that couples this turbulence
SOL turbulence. Three instability mechanisms are considered code to the 2D edge simulation tool UEDGE, in order to
for the SOL turbulence: interchange modes coupled to either investigate phenomena that develop on long time scales such
the sheath resistivity [579, 580] or to the plasma resistivity as the change in radiation pattern [590] prior to the density
for the resistive ballooning mode [581, 582] and the negative limit [591]. This effort remains extremely focused on the
sheath resistivity mode due to plasma temperature fluctuations modelling of the DIII-D and C-Mod devices. Relevance of
at the sheath [583]. In most of these models, the electron the results to other devices, and therefore to ITER, remains an
dynamics is embedded in the vorticity equation derived from open question.
the quasi-neutrality constraint so that the electron energy can (iii) A comprehensive investigation of SOL turbulence in
be regarded as a passive scalar. However, in the negative what is called a ‘flux-driven’ regime has been achieved [592].
In this approach, a particle flux is imposed rather than a
sheath resistivity model, the electron temperature fluctuations
mean density gradient. As a consequence, the confinement
are essential. The curvature effects that drive the interchange
properties of the system evolve freely in response to the
mode tend to ‘balloon’ the transport on the low field side
external drive and the turbulent transport [593]. In such a
with little turbulent transport on the high field side. This
regime, the turbulent particle flux is found to be strongly
agrees with several experimental observations [89]. As a
intermittent. Density fronts propagate ballistically [592, 594–
consequence, mechanisms based on the interchange are the
596] and a large fraction of the total particle flux is transported
prevailing instabilities that are used in the simulations. Drift
by rare and large transport events [597]. In these simulations,
waves that are claimed to govern the edge plasma transport can
no scale separation is found between the smallest density
also be considered [584]. Let us review the main theoretical
gradient scale, typically ∼ρs , the mean density gradient ∼55ρs
investigations of SOL turbulence and their implication for the
and the long range ballistic propagation of a given density front
ITER modelling.
∼200ρs . The generation of zonal flows [598] can be viewed as
(i) A phenomenological description of interchange-based the result of an inverse cascade, namely the transfer of energy
cross-field transport [195] has become a major reference in from the large wave vector to the small wave vectors. These
the experimental analysis of edge turbulence. It analytically flows are large scale flows in the poloidal direction that can
characterizes the properties of coherent density structures exhibit fine radial and temporal structures [599, 600]. They
that would propagate ballistically in the SOL. These density are generated by the turbulence and are also regulators of the
structures are termed ‘blobs’. The basic mechanism driving turbulence [601]. The interplay between the zonal flows and
the ballistic motion is in fact the standard interchange mode, the density fronts is a key element in determining the actual
namely the electric drift velocity induced by the charge transport regime in the non-linear phase [602].
separation itself governed by the curvature drift. The finite The scaling of the mean density e-folding length in terms
poloidal extent of the density structure, and, therefore, the of the connection length has been analysed with extensive
existence of a poloidal density gradient is a key element of such simulations that provide an accurate scaling of the SOL width
radial ‘blob’ transport. A flute approach, namely that all fields in terms of the connection length with a power 0.62 [603]. This
are constant along the magnetic field line, is used to reduce scaling is calculated analytically and one finds that the ballistic
the dynamics to a 2D (r, θ) plane [196]. This work provides scaling (power 1), that actually governs density transport [604],
scalings and predictions for these specific events. These results is in fact partly cancelled by the reduced frequency of the
have been incorporated in a 2D modelling of the plasma– intermittent events as the connection length increases so that
wall interaction. Transport is then modelled in terms of two the scaling is rather close to the diffusive scaling [603].
processes, diffusion due to the turbulence and, superimposed, Including the scaling of the curvature effect, one finds that
a radial outward convective flux [585]. Although such a the dependence on the major radius is still weaker so that the
description captures the essence of ballistic flights in the SOL, SOL width is nearly independent of the machine size [605].
its thermodynamical foundation is more difficult to assess No strong conflict thus appears between the extrapolation of
since transport can take place when the system is at rest (no present experimental results performed in the ITER modelling
dependence on a gradient). and the average SOL width scaling obtained in the turbulent
(ii) Another approach is a turbulence simulation of the simulations. Stronger departure from the diffusive modelling
edge plasma in the BOUT code. A flux tube geometry is and the turbulent transport will be found for transient effects
used with a comprehensive set of fluid equations and the where ballistic transport will govern the time scales.
appropriate geometry, in particular the X-point effects [586].
The simulation domain contains the region of both open field 3.7.3. Summary of the outstanding trends. Theoretical and
lines, the SOL, and closed field lines just within the separatrix. simulation results indicate that long range transport events,
This significant effort of edge simulation is well adapted to namely those which exceed significantly the SOL width, must
analyse experimental results from various devices. Several be expected. These transport phenomena depart significantly
S254
from the standard point of view of a diffusive cross- midplane power SOL widths as the scaling parameter for ITER
field transport ruling the plasma–wall interaction. Recent divertor heat load profiles. The presently estimated values
simulations of the scaling of such intermittent transport are similar to those assumed in ITER modelling, although
indicate that the scaling of the SOL width on the plasma major inclusion of kinetic effects in those models remains to be
radius is significantly weaker than one would expect from a implemented. Both empirical and modelling predictions of
ballistic scaling. In fact, it is found to be similar or even in–out divertor heat load asymmetries are less certain although
weaker than expected for a diffusive scaling. These results SOL flows and drifts appear to be the controlling physics.
thus indicate that the ITER modelling of the plasma–wall We expect that continued emphasis on inter-machine physics-
interaction based on diffusive transport capture a reasonable based scaling work will provide more reliable divertor power
description of the transport features in the steady-state regime. profile estimates.
However, transient effects cannot be appropriately modelled Because of their role in heat loads as well as impurity
with diffusive transport. In particular, material born at the wall transport and migration, SOL flows have been the recent
can exhibit a rapid influx in conjunction with an inward ballistic emphasis of both experimental and modelling investigations.
event. The lack of scale separation indicates that turbulent A general understanding is emerging that flows are driven
transport is coupled to macro-scales such as MHD modes. In by two processes: perpendicular transport and drifts. Their
particular, the coupling of turbulent transport to ELM transport relative contribution is still to be sorted out. Models are
must be considered [191]. just beginning to achieve flows that approach those seen in
Present turbulent simulations of the SOL do not provide experiment but considerable progress is needed.
an extensive study of the scaling of the SOL width. Several While most of the energy flows to the divertor during
key issues are still to be addressed. In particular, the geometry steady state this is not the case with particle fluxes; significant
of the system is characterized by several interface problems, particle fluxes are inferred to reach main-chamber PFCs in
jumps in the connection length with secondary limiters and, present experiments. These are associated with turbulence
more importantly, the interface between the edge plasma whereby filaments (along the field) of high density plasma
located on the closed magnetic surfaces and the SOL. Large travel radially at high velocity (∼0.5 km s−1 ), reaching the
changes in the turbulent transport have been reported between first wall before the particles can be lost parallel to the field.
the edge plasma and the SOL plasma [596], in particular, Analytical models and codes handling this effect are still
structures in the SOL are observed to be more flute-like [584]. in development and comparison with experiment is limited.
However, few simulations allow one to address both the Furthermore, it is not clear how models of turbulence will
edge and SOL plasmas. In this case, turbulence spreading be properly interfaced with the present complement of edge
will enhance transport in the most stable region and reduce fluid codes. Initial empirical extrapolations of these fluxes
turbulent transport in the most unstable region. This effect to ITER, based on inter-machine experiments, indicate that
can strongly modify the present analysis. Another issue is such fluxes should not pose a problem for the lifetime of first-
the impact of the large scale flows, both in the radial direction wall components. Given that the present empirical base for
whenever the turbulence eddy size is that of the SOL [606] and such extrapolation is minimal, further measurements and code
in the poloidal and toroidal direction where large scale flows development/validation would be very important to give more
are observed [599, 607]. confidence in this optimistic ITER extrapolation.
A proper modelling of SOL turbulence is important Transient heat loads, in the form of ELMs and disruptions,
not only to model plasma–wall interaction using present are presently the most serious challenge to PFC integrity
experimental data scaled up to ITER but also to model and lifetime. Significant progress has been made in the
ITER operation where SOL cross-field transport will strongly determination of the spatial and temporal characteristics of
influence impurity migration, density build-up, heat flux to the the energy deposition at the divertor target during ELMs. On
divertor and to wall components as well as the performance the basis of these measurements, an empirical extrapolation of
of the ICRH heating system. Furthermore, the appropriate the time-dependence of the ELM power pulse delivered to the
description of such a transport regime will also require ITER divertor plates has been developed. Modelling by kinetic
theoretical effort since the framework imposed by the large SOL codes can describe some of the features identified in the
fluctuation level departs significantly from the local quasi- data, but further detailed validation of these modelling tools
equilibrium that is generally assumed. is required before the predictions for ITER can be considered
quantitative.
4. Summary and implications for ITER The magnitude of the ELM energy loss delivered to
the plates in ITER relevant conditions (high density, high
The significant progress that has been made since the IPB in the confinement H-modes) has been determined across most
area of SOL/divertor research is described in this chapter along experimental devices and a physics-based scaling derived.
with an assessment of the implications for ITER operation. A The scaling of Type I ELM (proposed operating regime for
number of important advances in our understanding of key ITER) energy loads at the ITER divertor gives values that
issues has been made (e.g. ELM power loads) as well as will be close to or above the limits for material damage
uncovering of new concerns (e.g. wall fluxes). (sublimation/melting) and could lead to a reduced lifetime
The study of steady-state heat and particle loads to plasma of the divertor target. The uncertainty in these estimates
facing components (PFCs) has always been of central emphasis is still considerable. Alternative ELM regimes with high
in the divertor/SOL area. The recent better empirical fitting confinement, as well as regimes with smaller Type I ELMs,
of an expanded database has allowed improved use of the have been identified, which, if proved applicable to ITER,
S255
A. Loarte et al
will lead to acceptable heat loads. Further studies of all these operation than had been previously thought. The first issue,
regimes should be pursued together with the development of raised earlier, is the realization that significant heat loads
modelling tools for the transport of ELM energy to all PFCs. of, as yet, poorly quantified nature may result in localized
In the period since the IPB it has been found that a melting of Be PFC surfaces, reducing the wall PFC lifetime.
significant fraction of the ELM energy (up to ∼50% for very JET has shown that it is possible to operate a device with
large ELMs) can be deposited on main-chamber PFCs and not damaged Be without deleterious effects for ITER relevant high
in the divertor. ELMs lead to a fast radial expulsion of ions and density/confinement conditions, but the low density operation
electrons from the core—in essence carrying plasma energy is more restricted. A second materials issue involving Be is its
from the pedestal directly to main-chamber PFCs. The radial effect on the properties of W when deposited on such surfaces.
transit time of the ELM is similar to the ion energy parallel Alloys can be formed which potentially have poorer properties
loss time, such that ELM main-chamber fluxes are of similar than the original W. The use of Be may also lead to positive
magnitude to those reaching the divertor. This raises concerns mixed materials effects. It is expected that Be eroded from
regarding the ITER Be wall and upper divertor lifetime. The main-chamber surfaces will arrive at C divertor surfaces and
characterization of these energy fluxes is made difficult by the assist in suppressing C chemical erosion and the associated
poloidally/toroidally discontinuous character of the fluxes, and T retention. All the above mixed materials questions are
a physics-based inter-machine extrapolation to ITER remains relatively new and experiments, both in the laboratory and in
to be done. Certainly, given the implications for main-chamber tokamaks, should be encouraged. The modelling of the level
PFCs, further experiments and model development/validation of mixing expected in ITER is, of course, dependent on the
are urgently needed. better understanding of impurity transport as described above.
The migration of PFC material around the chamber has The present plans for ITER include tungsten covering the
important implications for first-wall lifetime, core impurity dome and much of the divertor plates. High-Z PFC operation
levels and T retention. For standard ion B × ∇B drift towards in current devices has shown that it is possible to operate ITER
the dominant divertor, the main chamber and outer divertor relevant plasmas in high density/confinement conditions while
are generally found to be net erosion zones with the inner maintaining low core impurity concentrations. This, and the
divertor region being a zone of net deposition. Details of the use of central heating to reduce core levels in certain operating
deposition are linked to divertor operating conditions, specifics regimes, bodes well for ITER, although the uncertainty in
of the divertor geometry and the SOL flows which transport
modelling of W effects on the core requires the development
the eroded material over large distances. Models are still far
of better models of impurity transport in general.
from being able to reproduce present erosion and migration
The plans for ITER call for eventual switching to an
observations and so predictions for ITER are very uncertain.
all W divertor and, potentially, its extensive use elsewhere
Further experiments with C13 and metallic tracers combined
in the machine for reasons of T retention (reducing it), as
with detailed modelling should continue to be pursued.
well as reactor relevance. The main concern over such
The probability of an impurity reaching the core plasma
a change is the melting of W under transient loads (and
launched from a given PFC location (screening) has been
the associated implications for machine operability) and the
investigated in many experiments. In general, it is found that
possible accumulation of W impurities in discharges with
main chamber produced impurities are ∼10 times more likely
to reach the core than those launched from the divertor. The ITBs. Further experiments to control the energy loads during
uncertainty in the modelling of the impurity transport, both ELM and disruption transients, and on W density control
for material migration and core contamination, is linked to during ITBs, are necessary to justify an increase in the W
the uncertainties in SOL flows and radial impurity diffusivity. coverage in ITER from the one presently foreseen.
The former is, as mentioned earlier, the subject of intense We now have a better idea of where T is retained in PFC
experimental and modelling effort and the situation should and other vessel surfaces. Present experience with carbon
improve there. However, our knowledge is very limited PFC tokamaks indicates that T is retained primarily through
regarding the impurity diffusivity (or convective velocity) in co-deposition with C, mainly at the inner divertor although
the SOL and whether it is dependent on the charge state. The a significant fraction (∼15–30%) can be found around the
areas of impurity migration and core contamination are thus chamber in tile gaps. However, the C deposition rate has been
linked in some aspects of their limitations and similarly need shown to vary in location and rate with discharge equilibrium
considerable input to reduce the uncertainties of the models. and divertor geometry; it is a complex multi-step process
The details of material properties as well as that of mixed dependent on C chemical erosion yield, sticking coefficients of
materials have gained more emphasis since the IPB. New hydrocarbons and plasma flux profiles. Models of every step
results on the flux dependence of carbon chemical sputtering of this process need improvement.
and its reduction under detached conditions indicate that The tritium inventory, both in-vessel and on-site, will
steady-state carbon erosion can be expected to be much smaller be a central operational issue for any BPX due to safety
for the ITER divertor than originally expected. On the other considerations and operational limits on the level of T
hand, recent results have raised concerns about the integrity inventory. The ITER specification for the fraction of injected
of CFC carbon under transient loads due to the anisotropies T that can be retained in-vessel is 0.1%. The little tokamak
introduced by the fibres. Whether this will be a problem for tritium experience that exists implies much higher values, even
ITER relevant loads remains to be assessed by experiment and after cleaning efforts. More extensive studies with D under a
modelling in dedicated facilities. variety of divertor geometries and operating conditions have
Although the properties of Be are not in doubt, new resulted in lower levels of D retention, but generally still more
information indicates that its use may add more risk for ITER than an order of magnitude higher than the ITER guideline.
S256
Because of the uncertainties of extrapolating T retention rates Science. Figures 7, 13, 14, 17, 20 and 31 are reproduced in this
from present all-C-wall experiments to a BPX, in which the chapter by kind permission of IOP Publishing Ltd. Figures 9,
surface area of C-clad PFCs will be a small fraction of all 16 and 26 reproduced in this chapter by kind permission of of
PFC surfaces, development of efficient T removal methods the American Institute of Physics.
is required. Removal methods have been developed along
various lines: photonic cleaning, oxidation and RF plasma
conditioning techniques. In general such techniques have been References
demonstrated to remove T, but a T removal rate at the level
[1] ITER Physics Basis Editors 1999 Nucl. Fusion 39 2137
required in ITER remains to be demonstrated. Experiments
[2] Goetz J.A. et al 1990 Phys. Plasmas 6 1899
using these techniques both in well-characterized laboratory [3] Herrmann A. 2002 Plasma Phys. Control. Fusion 44 883
conditions and in tokamak devices are highly desirable. Such [4] Takenaga H. et al 2002 Fusion Sci. Technol. 42 327
experiments should also address the effect on T removal rates [5] Kukushkin A.S. et al 2003 Fusion Eng. Des. 65 355
introduced by different levels and types of mixed materials, [6] Reichle R. et al 1991 Proc. 18th EPS Conf. on Control
Fusion and Plasma Physics (Berlin, Germany, 1991)
as well as the efficacy of T removal from difficult-to-reach
vol 15C (ECA) D27 III-105
locations (e.g. tile gaps). [7] Asakura N. et al 1999 Nucl. Fusion 39 1983
He and H exhaust was demonstrated in the IPB to be [8] Reichle R. et al 2004 Phys. Scr. t111 107
of a level needed to keep ITER He concentrations under an [9] Andrew P. et al 2005 J. Nucl. Mater. 337–339 99
acceptable value. New experiments with ITB plasmas have [10] DelaChambre E. et al 2005 J. Nucl. Mater. 337–339 1069
[11] Hildebrandt D. et al 2005 J. Nucl. Mater. 337–339 1064
shown that, although the situation is slightly worsened due to [12] Stangeby P.C. 2000 The Plasma Boundary of Magnetic
the improved impurity confinement, it is still within acceptable Fusion Devices (Plasma Physics Series) (Bristol: IOP
limits for current tokamaks. This bodes well for a BPX-like Publishing)
device. [13] Riccardo V. et al 2001 Plasma Phys. Control. Fusion 43 881
The key neutral issues which remain to be addressed [14] Matthews G.F. et al 2001 J. Nucl. Mater. 290–293 668
[15] Herrmann A. et al 2003 J. Nucl. Mater. 313–316 759
experimentally in conjunction with modelling are those related [16] Leonard A.W. et al 1999 J. Nucl. Mater. 266–269 109
to BPX specific conditions: the effect of n–n and ion–neutral [17] Eich T. et al 2001 28th EPS Conf. on Controlled Fusion and
reactions on the divertor solution and the trapping of Lyman Plasma Physics (Funchal, Portugal, 2001) vol 25A (ECA)
series radiation lines. These effects can already be seen to a 1809
lesser extent in present experiments than in a BPX and can [18] Fundamenski W. and Sipilä S. 2004 Nucl. Fusion 44 20
[19] Fundamenski W. et al 2005 J. Nucl. Mater. 337–339 305
have a large influence on the expected divertor conditions [20] Kukushkin A.S. et al 2003 Nucl. Fusion 43 716
(detachment, ionization balance and neutral pressure) and [21] Fundamenski W. et al 2005 Nucl. Fusion 45 950
He removal. Such studies should be the subject of future [22] Kallenbach A. et al 2004 Plasma Phys. Control. Fusion
experimental/code validation studies for which the models are 46 431
being presently developed. [23] Kempenaars M. et al 2003 30th EPS Conf. on Controlled
Fusion and Plasma Physics (St. Petersburg, Russia, 2003)
Fuelling of current experiments and how it scales with vol 27A (ECA) P-1.56
plasma conditions is fairly well understood. Pellet injection [24] Fundamenski W. et al 2003 J. Nucl. Mater. 313–316 787
from the inside edge of the torus seems the most appropriate [25] Chankin A. 1997 J. Nucl. Mater. 241–243 199
tool to provide the required fuelling in ITER. Further [26] Sipila S. et al 2003 30th EPS Conf. on Controlled Fusion and
information is needed on alternative techniques, such as Plasma Physics (St. Petersburg, Russia, 2003) vol 27A
(ECA) P-3.144
supersonic gas and compact toroid injection. In addition, [27] Pitts R.A. et al 2005 J. Nucl. Mater. 337–339 146
detailed modelling of well diagnosed fuelling experiments with [28] Loarte A. et al 1999 J. Nucl. Mater. 266–269 587
edge parameters (plasma pressure, density and temperature) [29] Kukushkin A.S. et al 2002 Nucl. Fusion 42 187
close to those expected at the edge of ITER would be necessary [30] Lasnier C.J. et al 1998 Nucl. Fusion 38 1225
to make more accurate predictions of the fuelling efficiency. [31] Eich T. et al 2005 J. Nucl. Mater. 337–339 669
[32] Schaffer M.J. et al 1997 Nucl. Fusion 37 83
In summary, ITER will itself be an experiment in edge [33] Rognlien T.D. et al 1999 J. Nucl. Mater. 266–269 654
physics and plasma–wall interactions whose detailed outcome [34] Borrass K. 2001 J. Nucl. Mater. 290–293 551
has many uncertainties. However, we believe that, by [35] Schneider R. et al 2000 Contrib. Plasma Phys. 40 328–33
focusing current experimental capabilities on preparing for [36] Kim J.W. 2002 An analysis of the anomalous transport of the
ITER operation, we will be able to ensure that we have the plasma edge in ASDEX Upgrade PhD Thesis University
of Muenchen
operational tools and understanding of the underlying physics [37] Simonini R. et al 1994 Contrib. Plasma Phys. 34 368
to avoid or deal with problems as they arise. [38] Itoh S.I. and Itoh K. 1994 Plasma Phys. Control. Fusion
36 1845
[39] Connor J.W. et al 1999 Nucl. Fusion 39 169
Acknowledgments [40] Harrison M.F.A. et al 1983 Nucl. Technol. Fusion 3 432
[41] Counsell G.F. et al 1999 J. Nucl. Mater. 266–269 91
This article has been made possible thanks to the help and [42] Erents S.K. and Stangeby P.C. 1998 Nucl. Fusion 38 1637
contributions of those working in the area of plasma edge [43] Neuhauser J. et al 1989 Plasma Phys. Control. Fusion
and plasma–wall interaction research. The first author is 31 1551
particularly thankful to his wife Marta for help in solving [44] McCormick G.K. and Pietrzyk Z.A. 1989 J. Nucl. Mater.
162–164 264
many editorial problems during the elaboration of this chapter. [45] McCormick G.K. et al 1993 Proc. 20th Eur. Conf. on
Figures 4, 5, 8, 10–12, 15, 19, 21–25, 28, 30 and 32 are Controlled Fusion and Plasma Physics (Lisbon, Portugal,
reproduced in this chapter by kind permission of Elsevier 1993) (Geneva: European Physical Society) vol 2 587
S257
A. Loarte et al
[46] Bosch H.S. et al 1995 J. Nucl. Mater. 220–223 558 [108] Loarte A. et al 1993 Proc. 20th Eur. Conf. on Controlled
[47] Rowan W.L. et al 1995 J. Nucl. Mater. 220–222 668 Fusion and Plasma Physics (Lisbon, Portugal, 1993)
[48] Fenstermacher M.E. et al 1995 J. Nucl. Mater. 220–222 330 vol 17C Part II (Geneva: European Physical Society) 555
[49] Boedo J.A. et al 2001 Phys. Plasmas 8 4826 [109] Pitcher C.S. et al 1993 Proc. 20th Eur. Conf. on Controlled
[50] LaBombard B. et al 1995 Phys. Plasmas 2 2242 Fusion and Plasma Physics (Lisbon, Portugal, 1993)
[51] LaBombard B. et al 2001 J. Nucl. Mat. 290–293 995 vol 17C Part I (Geneva: European Physical Society) 291
[52] Asakura N. et al 1997 J. Nucl. Mater. 241–243 559 [110] Asakura N. et al 2000 Phys. Rev. Lett. 84 3093
[53] Erents S.K. et al 1997 J. Nucl. Mater. 241–243 433 [111] Erents S.K. et al 2000 Plasma Phys. Control. Fusion 42 905
[54] Boedo J.A. et al 1998 Rev. Sci. Instrum. 69 2663 [112] Asakura N. et al 2002 Plasma Phys. Control. Fusion 44 2101
[55] Pitts R.A. et al 2001 J. Nucl. Mater. 290–293 940 [113] Boedo J.A. et al 1999 J. Nucl. Mater. 266–269 783
[56] Umansky M.V. et al 1998 Phys. Plasmas 5 3373 [114] LaBombard B. et al 2004 Nucl. Fusion 44 1047
[57] Umansky M.V. et al 1999 Phys. Plasmas 6 2791 [115] Erents S.K. et al 2004 Plasma Phys. Control. Fusion 46 349
[58] LaBombard B. et al 2000 Nucl. Fusion 40 2041 [116] Hutchinson I.H. 1988 Phys. Rev. A37 4358
[59] LaBombard B. et al 1997 J. Nucl. Mater. 241–243 149 [117] Jablonski D. et al 1997 J. Nucl. Mater. 241–243 782
[60] Shimizu K. et al 1992 J. Nucl. Mater. 196–198 476 [118] LaBombard B. et al 1999 J. Nucl. Mater. 266–269 571
[61] Tsui H.Y.W. et al 1995 J. Nucl. Mater. 220–222 672 [119] Pitcher C.S. et al 1999 J. Nucl. Mater. 266–269 1009
[62] Pitcher C.S. et al 2000 Phys. Plasmas 7 1894 [120] Isler R.C. et al 1999 J. Nucl. Mater. 266–269 376
[63] Pitcher C.S. et al 2001 J. Nucl. Mater. 290–293 812–19 [121] Stangeby P.C. and Chankin A.V. 1996 Nucl. Fusion 36 839
[64] Lipschultz B. et al 2002 Plasma Phys. Control. Fusion [122] Chankin A.V. and Stangeby P.C. 1996 Plasma Phys. Control.
44 733 Fusion 38 1879
[65] Schneider R. et al 1999 J. Nuc. Mater. 266–269 175 [123] Porter G.D. et al 2003 J. Nucl. Mater. 313–316 1085
[66] Bosch H.S. et al 1999 Plasma Phys. Control. Fusion 41 A401 [124] Rozhansky V.A. et al 2001 Nucl. Fusion 41 387
[67] Kallenbach A. et al 2003 Nucl. Fusion 43 573 [125] Bonnin X. et al 2003 J. Nucl. Mater. 313–316 909
[68] Whyte D.G. et al 2005 Plasma Phys. Control. Fusion 47 1579 [126] Chankin A.V. et al 2001 J. Nucl. Mater. 290–293 518
[69] Loarte A. 2001 Plasma Phys. Control. Fusion 43 R183 [127] Bonnin X. et al 2005 J. Nucl. Mater. 337–339 301
[70] Allen S. et al 1999 J. Nucl. Mater. 266–269 168 [128] LaBombard B. et al 2001 Phys. Plasmas 8 2107
[71] Vlases G.C. et al 1999 J. Nucl. Mater. 266–269 160 [129] Boswell C.J. et al 2001 J. Nucl. Mater. 290–293 556
[72] Lipschultz B. et al 2000 Proc. 18th Int. Conf. on Fusion [130] Boedo J.A. et al 2000 Phys. Plasmas 7 1075
Energy 2000 (Sorrento, Italy, 2000) (Vienna: IAEA) [131] Asakura N. et al 2003 J. Nucl. Mater. 313–316 820
CD-ROM file EX5/6 and http://www.iwww.iaea.org/ [132] Leonard A.W. et al 1997 Phys. Rev. Lett. 78 4769
programmes/ripc/physics/fec2000/html/node1.htm [133] Hatayama A. et al 2000 Nucl. Fusion 40 2009
[73] McCormick G.K. et al 1992 J. Nucl. Mater. 196–8 264 [134] ITER Technical Basis ITER EDA Documentation Series No
[74] LaBombard B. et al 1995 J. Nucl. Mater. 220–222 976 24 (Vienna: IAEA) 2002
[75] McCormick G.K. et al 1999 J. Nucl. Mater. 266–269 99 [135] Dumortier P. et al 2002 Plasma Phys. Control. Fusion
[76] Suttrop W. et al 1999 J. Nucl. Mater. 266–269 118 44 1845
[77] Neuhauser J. et al 2002 Plasma Phys. Control. Fusion 44 855 [136] Sakurai S. et al 2001 J. Nucl. Mater. 290–293 1002
[78] LaBombard B. et al 2002 Proc. 19th Int. Conf. on Fusion [137] Kubo H. et al 2001 Nucl. Fusion 41 227
Energy 2002 (Lyon, France, 2002) (Vienna: IAEA) [138] Konoshima S. et al 2001 Plasma Phys. Control. Fusion
CD-ROM file EX/D2-1 and http://www.iaea.org/ 43 959
programmes/ripc/physics/fec2002/html/fec2002.htm [139] Konoshima S. et al 2003 J. Nucl. Mater. 313–316 888
[79] Kim J.W. et al 2001 J. Nucl. Mater. 290–293 644 [140] Takenaga H. et al 2005 Nucl. Fusion 45 1618
[80] Greenwald M. 2002 Plasma Phys. Control. Fusion 44 R27 [141] Asakura N. et al 2004 Nucl. Fusion 44 503
[81] Whyte D.G. 2004 Physica Scripta t111 34 [142] Matthews G.F. 2005 J. Nucl. Mater. 337–339 1
[82] Rudakov L. et al 2002 Plasma Phys. Control. Fusion 44 717 [143] Kallenbach A. et al 1996 Plasma Phys. Control. Fusion
[83] Boedo J.A. et al 2003 J. Nucl. Mater. 313–316 813 38 2097
[84] Zweben S.J. et al 2002 Phys. Plasmas 9 1981 [144] Rapp J. et al 2004 Nucl. Fusion 44 312
[85] Terry J.L. et al 2003 Phys. Plasmas 10 1739 [145] Matthews G.F. et al 1999 Nucl. Fusion 39 19
[86] Boedo J.A. et al 2003 Phys. Plasmas 10 1670 [146] Kubo H. et al 2003 J. Nucl. Mater. 313–316 1197
[87] Hidalgo C. et al 2003 J. Nucl. Mater. 313–316 863 [147] Mertens V. et al 2003 Fusion Eng. Des. 66–68 119
[88] Zweben S.J. and Gould R.W. 1985 Nucl. Fusion 25 171 [148] Lang P.T. et al 2003 Nucl. Fusion 43 1110
[89] Endler M. et al 1995 Nucl. Fusion 35 1307 [149] Kallenbach A. et al 2001 J. Nucl. Mater. 290–293 639
[90] Filippas A.V. et al 1995 Phys. Plasmas 2 839 [150] Puiatti M. et al 2002 Plasma Phys. Control. Fusion 44 1863
[91] Nielsen A.H. et al 1996 Phys. Plasmas 3 1530 [151] Rice J.E. et al 2002 Nucl. Fusion 42 510
[92] Joseph B.K. et al 1997 Phys. Plasmas 4 4292 [152] Neu R. et al 2003 J. Nucl. Mater. 313–316 116
[93] Heller M.V.A.P. et al 1999 Phys. Plasmas 6 846 [153] Dux R. et al 2003 Plasma Phys. Control. Fusion 45 1815
[94] Carreras B.A. et al 1999 Phys. Plasmas 6 1885 [154] Takenaga H. et al 2003 Nucl. Fusion 43 1235
[95] Jha R. et al 2003 Phys. Plasmas 10 699 [155] Federici G. et al 2003 Plasma Phys. Control. Fusion 45 1523
[96] Antar G.Y. et al 2003 Phys. Plasmas 10 419 [156] Wuerz H. et al 2001 J. Nucl. Mater. 290–293 1138
[97] Zweben S.J. and Medley S.S. 1989 Phys. Fluids 1 2058 [157] Pestchanyi S.E. et al 2002 Plasma Phys. Control. Fusion 44
[98] LaBombard B. and Lipschultz B. 1987 Nucl. Fusion 27 81 845
[99] Tynan G.R. et al 1992 Phys. Rev. Lett. 68 3032 [158] Bazylev B. et al 2005 J. Nucl. Mater. 337–339 766
[100] Lipschultz B. et al 2003 Proc. 30th EPS Conf. on Controlled [159] Landman I.S. et al 2005 J. Nucl. Mater. 337–339 761
Fusion and Plasma Physics (St. Petersburg, Russia, 2003) [160] Loarte A. 2004 Phys. Scr. t111 13
vol 27A (ECA) P-3.197 [161] Federici G. et al 2005 J. Nucl. Mater. 337–339 684
[101] Lipschultz B. et al 2005 Plasma Phys. Control. Fusion [162] Eich T. et al 2003 J. Nucl. Mater. 313–316 919
47 1559 [163] Fenstermacher M. et al 2003 Plasma Phys. Control. Fusion
[102] Stangeby P.C. 1984 J. Nucl. Mater. 121 55 45 1597
[103] Stangeby P.C. 2002 Phys. Plasmas 9 3489 [164] Loarte A. et al 2004 Phys. Plasmas 11 2668
[104] Janeschitz G. et al 1992 J. Nucl. Mater. 196–198 380 [165] Loarte A. et al 2003 J. Nucl. Mater. 313–316 962
[105] Matthews G.F. et al 1992 J. Nucl. Mater. 196–198 374 [166] Loarte A. et al 2003 Plasma Phys. Control. Fusion 45 1549
[106] Lipschultz B. et al 2001 J. Nucl. Mater. 290–293 286 [167] Kirk A. et al 2003 J. Nucl. Mater. 313–316 1081
[107] Whyte D.G. et al 2001 Nucl. Fusion 41 1243 [168] Bergmann A. 2002 Nucl. Fusion 42 1162
S258
[169] Tskhakaya D. et al 2004 Proc. Workshop on Theory of [219] Loarte A. et al 1997 Proc. 24th EPS Conf. on Controlled
Fusion Plasmas (Varenna, Italy) Fusion and Plasma Physics (Berchtesgaden, Germany,
[170] Jachmich S. et al 2002 Proc. 28th EPS Conf. on Controlled 1997) vol 21A part III p 1049
Fusion and Plasma Physics (Funchal, Portugal, 2001) [220] Petrie T.W. et al 2005 J. Nucl. Mater. 337–339 216
vol 25A (ECA) P-4.079 [221] Lisgo S. et al 2005 J. Nucl. Mater. 337–339 139
[171] Pitts R.A. et al 2003 Nucl. Fusion 43 1145 [222] Stotler D. et al 2001 J. Nucl. Mater. 290–293 967
[172] Loarte A. et al 2002 Plasma Phys. Control. Fusion 44 1815 [223] Kukushkin A.S. et al 2004 Proc. 20th Int. Conf. on Fusion
[173] Leonard A.W. et al 2002 Plasma Phys. Control. Fusion 44 Energy 2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA)
945 CD-ROM file IT/P3-24 and http://www-naweb.iaea.org/
[174] Urano H. et al 2003 Plasma Phys. Control. Fusion napc/physics/fec/fec2004/datasets/index.html
45 1571 [224] Reiter D. et al 2002 Plasma. Phys. Control. Fusion 44 1723
[175] Eich T. et al 2003 Phys. Rev. Lett. 91 195003 [225] Maggi C.F. et al 1999 Proc. 26th EPS Conf. on Controlled
[176] Evans T. et al 1995 J. Nucl. Mater. 220–222 235 Fusion and Plasma Physics (Maastricht, The Netherlands,
[177] Counsell G.F. et al 2003 Proc. 30th EPS Conf. on Controlled 1999) vol 23J (ECA) p 201
Fusion and Plasma Physics (St Petersburg, Russia, 2003) [226] Tsitrone E. et al 2002 Plasma Phys. Control. Fusion 44 701
vol 27A (ECA) P-3.202 [227] Rapp J. et al 2003 Proc. 30th EPS Conf. on Controlled
[178] Leonard A.W. et al 1997 J. Nucl. Mater. 241–243 628 Fusion and Plasma Physics (St Petersburg, Russia, 2003)
[179] Asakura N. et al 1999 J. Nucl. Mater. 266–269 182 vol 27A (ECA) P-1.160
[180] Counsell G.F. et al 2002 Plasma Phys. Control. Fusion [228] Maingi R. et al 2004 Nucl. Fusion 44 909
44 827 [229] Horton L.D. et al 1999 Nucl. Fusion 39 1
[181] Counsell G.F. et al 2002 Plasma Phys. Control. Fusion [230] Reiter D. et al 1990 Nucl. Fusion 30 2141
44 B23 [231] Sakasai A. et al 1999 J. Nucl. Mater. 266–269 312
[182] Petrie T.W. et al 2003 J. Nucl. Mater. 313–316 804 [232] Stork D. et al 1999 Proc. 26th EPS Conf. on Controlled
[183] Connor J.W. et al 1998 Phys. Plasmas 5 2687 Fusion and Plasma Physics (Maastricht, The Netherlands,
[184] Snyder P.B. et al 2002 Phys. Plasmas 6 2037 1999) vol 23J (ECA) 205
[185] Groth M. et al 2003 J. Nucl. Mater. 313–316 1071 [233] Bosch H.S. et al 2001 J. Nucl. Mater. 290–293 836
[186] Leonard A. et al 2003 Phys. Plasmas 10 1765 [234] Sakasai A. et al 2001 J. Nucl. Mater. 290–293 957
[187] Zeng L. et al 2004 Plasma Phys. Control. Fusion 46 A121 [235] Zastrow K.-D. et al 2002 Proc. 29th EPS Conf. on Controlled
[188] Herrmann A. 2004 Plasma Phys. Control. Fusion 46 971 Fusion and Plasma Physics (Montreux, Switzerland, 2002)
[189] Nunes I. et al 2004 Nucl. Fusion 44 883 vol 26B (ECA) O-5.02
[190] Ghendrih Ph. et al 2003 J. Nucl. Mater. 313–316 914 [236] Horton L.D. et al 1993 Proc. 14th Int. Conf. on Plasma
[191] Gonçalves B. et al 2003 Plasma Phys. Control. Fusion Physics and Controlled Nuclear Fusion Research 1992
45 1627 (Wuerzburg, 1992) (Vienna: IAEA) vol 1 p 423
[192] Gill R.D. et al 1998 Nucl. Fusion 38 1461 [237] Zastrow K.-D. et al 2005 Nucl. Fusion 45 163
[193] Chankin A.V. et al 2003 J. Nucl. Mater. 313–316 828 [238] Finken K.-H. et al 2001 Proc. 28th EPS Conf. on Controlled
[194] Fundamenski W. et al 2004 Plasma Phys. Control. Fusion Fusion and Plasma Physics (Funchal, Portugal, 2001)
46 233 vol 25A (ECA) p 1793
[195] Krasheninnikov S.I. 2001 Phys. Lett. A 283 368 [239] Groth M. et al 2001 J. Nucl. Mater. 290–293 867
[196] D’Ippolito D.A. et al 2002 Phys. Plasmas 9 222 [240] Klepper C.C. et al 1993 Nucl. Fusion 33 533
[197] Silva C. et al 2005 J. Nucl. Mater. 337–339 722 [241] Groth M. et al 2002 Nucl. Fusion 42 591
[198] Leonard A.W. et al 2003 J. Nucl. Mater. 313–316 768 [242] Goetz J.A. et al 2001 Nucl. Fusion 41 1751
[199] Janeschitz G. et al 2001 J. Nucl. Mater. 290–293 1 [243] Wade M.R. et al 1998 Nucl. Fusion 38 1839
[200] Oyama N. et al 2004 Nucl. Fusion 44 582 [244] Itami K. et al 1999 J. Nucl. Mater. 266–269 1097
[201] Wilson H.R. et al 2004 Phys. Rev. Lett. 92 175006 [245] Goetz J.A. et al J. Nucl. Mater. 266–269 354
[202] Loarte A. et al 1998 Nucl. Fusion 38 331 [246] Fichtmueller M. et al 1999 J. Nucl. Mater. 266–269 330
[203] Doyle et al 2007 Progress in the ITER Physics Basis Nucl. [247] Coster D.P. et al 2001 J. Nucl. Mater. 290–293 845
Fusion 47 S18–S127 [248] Reiser D. et al 2001 J. Nucl. Mater. 290–293 953
[204] Schneider R. et al 1997 Proc. 16th Int. IAEA Conf. on Fusion [249] Hogan J.T. et al 2003 J. Nucl. Mater. 313–316
Energy 1996 (Montreal, Canada, 1996) (Vienna: IAEA) [250] Chung T. et al 2005 J. Nucl. Mater. 337–339 109
vol 2 p 465 [251] Abou-Gabal H.H. and Emmert G.A. 1991 Nucl. Fusion 31
[205] Monier-Garbet P. et al 2003 Proc. 30th EPS Conf. on 407
Controlled Fusion and Plasma Physics (St Petersburg, [252] Bachmann P. et al 1992 J. Nucl. Mater. 196–198 865
Russia, 2003) vol 27A (ECA) P-1.095 [253] Kubo H. et al 1999 Plasma Phys. Control. Fusion 41 747
[206] Loarte A. et al 2000 Proc. 18th Int. Conf. on Fusion Energy [254] Kukushkin A.S. et al 2002 Plasma Phys. Control. Fusion 44
2000 (Sorrentto, Italy, 2000) (Vienna: IAEA) CD-ROM 931
file ITERP/11R and http://www.iaea.org/programmes/ [255] Lumma D. et al 1997 Phys. Plasmas 4 2555
ripc/physics/fec2000/html/node1.htm [256] Isler R.C. et al 1997 Phys. Plasmas 4 2989
[207] Wenzel U. et al 1999 Nucl. Fusion 39 873 [257] McCracken G.M. et al 1998 Nucl. Fusion 38 619
[208] Uesugi Y. et al 2001 J. Nucl. Mater. 290–293 1134 [258] Terry J.L. et al 1998 Phys. Plasmas 5 1759
[209] Ohno N. et al 2001 Nucl. Fusion 41 1055 [259] Krasheninnikov S.I. et al 1997 J. Nucl. Mater. 241–243 283
[210] Takamura S. et al 2002 Plasma Sources Sci. Technol. [260] Reiter D. et al 1997 J. Nucl. Mater. 241–243 342
11 A42 [261] Lipschultz B. et al 1999 Phys. Plasmas 6 1907
[211] Uesugi Y. et al 2002 Proc. 29th EPS Conf. on Plasma [262] Lipschultz B. et al 1998 Phys. Rev. Lett. 81 1007
Physics and Controlled Fusion (Montreux, Switzerland, [263] Maggi C.F. et al 1999 J. Nucl. Mater. 266–269 867
2002) vol 26B (ECA) P-1.087 [264] Hollmann E.M. et al 2002 Phys. Plasmas 9 1226
[212] Monk R. (JET Team) 1999 Nucl. Fusion 39 1751 [265] Fantz U. and Heger B. 1998 Plasma Phys. Control. Fusion
[213] Bosch H.S. et al 1999 J. Nucl. Mater. 266–269 462 40 2023
[214] Kawashima H. et al 1999 Nucl. Fusion 39 1679 [266] Fantz U. et al 2001 Plasma Phys. Control. Fusion 43 907
[215] Niemczewski A. et al 1997 Nucl. Fusion 37 151 [267] Fantz U. et al 2003 J. Nucl. Mater. 313–316 743
[216] Takenaga H. et al 2001 Nucl. Fusion 41 1777 [268] Fantz U. et al 2001 J. Nucl. Mater. 290–293 367
[217] Maingi R. et al 1999 Nucl. Fusion 39 1187 [269] Ohno N. et al 1998 Phys. Rev. Lett. 81 818
[218] Guo H.Y. et al 2000 Nucl. Fusion 40 379 [270] Nishijima D. et al 2002 Plasma Phys. Control. Fusion 44 597
S259
A. Loarte et al
[271] Tonegawa A. et al 2003 J. Nucl. Mater. 313–316 1046 [328] Kallenbach A. et al 1999 Nucl. Fusion 39 901
[272] Coad J.P. et al 2001 J. Nucl. Mater. 290–293 224 [329] Polevoi A.R. et al 2004 Proc. 20th Int. Conf. on Fusion
[273] Brooks J.N. et al 2003 J. Nucl. Mater. 313–316 424 Energy 2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA)
[274] Federici G. et al 2003 J. Nucl. Mater. 313–316 11 CD-ROM file IT/P3-28 and http://www-naweb.iaea.org/
[275] Federici G. et al 2001 Nucl. Fusion 41 1967 napc/physics/fec/fec2004/datasets/index.html
[276] Philipps V. et al 2002 Vacuum 67 399 [330] Janev R.K. and Reiter D. 2004 Phys. Plasmas 11 780
[277] Roth J. et al 2004 Nucl. Fusion 44 L21 [331] Pugno R. et al 2003 30th EPS Conf. on Controlled Fusion
[278] Thoma A. et al 1997 Plasma Phys. Control. Fusion and Plasma Physics (St Petersburg, Russia, 2003) vol 27A
39 1487 (ECA) P-1.153
[332] Stamp M. et al 2001 J. Nucl. Mater. 290–293 321
[279] Wampler W.R. et al 1999 J. Nucl. Mater. 266–269 217 [333] Vietzke E. et al 1987 Amorphous Hydrogen Films (Paris: Les
[280] Naujoks D. et al 1996 Nucl. Fusion 36 671 Editions de Physique) 1987 p 351
[281] Gafert J. et al 1999 J. Nucl. Mater. 266–269 365 [334] Roth J. et al 2005 J. Nucl. Mater. 337–339 970
[282] Fenstermacher M.E. et al 1999 Plasma Phys. Control. Fusion [335] Brooks J.N. et al 1999 J. Nucl. Mater. 266–269 58
41 A345 [336] Wuerz H. et al 2002 J. Nucl. Mater. 307–311 60
[283] Krieger K. et al 2003 J. Nucl. Mater. 313–316 327 [337] Landman I.S. et al 2004 Phys. Scr. t111 206
[284] Kirschner A. et al 2001 J. Nucl. Mater. 290–293 238 [338] Naujoks D. et al 2004 Phys. Scr. t111 80
[285] Krieger K. et al 2002 J. Nucl. Mater. 307–311 139 [339] Rohde V. et al 2004 Phys. Scr. t111 49
[286] Pütterich T. et al 2003 Plasma Phys. Control. Fusion [340] von Seggern J. et al 2003 J. Nucl. Mater. 313–316 439
45 1873 [341] Masaki K. et al 2005 J. Nucl. Mater. 337–339 553
[287] Federici G. et al 2001 J. Nucl. Mater. 290–293 260 [342] von Seggern J. et al 1999 Phys. Scr. t81 31
[288] Neu R. et al 2003 Fusion Sci. Technol. 44 692 [343] Pegourie B. et al 2004 Phys. Srip. t111 23
[289] Higashijima S. et al 1999 J. Nucl. Mater. 266–269 1078 [344] Hsu W.S. et al 1989 J. Vac. Sci. Technol. A7 1065
[290] Isler R.C. et al 2003 J. Nucl. Mater. 313–316 873 [345] Coad J.P. et al 1997 J. Nucl. Mater. 241–243 408
[291] Lipschultz B. et al 2001 Nucl. Fusion 41 585 [346] Roth J. et al 1997 Fusion Eng. Des. 37 465
[292] Geier A. et al 2003 J. Nucl. Mater. 313–316 1216 [347] Thomas P.R. et al 1990 J. Nucl. Mater. 176–177 3
[293] Krieger K. et al 2005 J. Nucl. Mater. 337–339 10 [348] Campbell D.J. et al 1997 J. Nucl. Mater. 241–243 379
[294] Krieger K. et al 1997 J. Nucl. Mater. 241–243 684 [349] Anderl R.A. et al 1999 J. Nucl. Mater. 273 1
[295] Brooks J.N. and Whyte D.G. 1999 Nucl. Fusion 39 525 [350] Causey R.A. and Walsh D.S. 1998 J. Nucl. Mater. 254 84
[296] Whyte D.G. 1999 J. Nucl. Mater. 266–269 67 [351] Mayer M. 1997 J. Nucl. Mater. 240 164
[297] Coad J.P. et al 2003 J. Nucl. Mater. 313–316 419 [352] Baldwin M.J. et al 2005 J. Nucl. Mater. 337–339 590
[298] Whyte D.G. et al 2001 J. Nucl. Mater. 290–293 356 [353] Doerner R.P. et al 2001 J. Nucl. Mater. 290–293 166
[299] McCracken G.M. et al 1999 Nucl. Fusion 39 41 [354] Doerner R.P. et al 2004 J. Appl. Phys. 95 s4471
[300] Neu R. et al 1997 J. Nucl. Mater. 241–243 678 [355] Whyte D.G. and Davis J.W. 2005 J. Nucl. Mater.
[301] Groth M. et al 2005 J. Nucl. Mater. 337–339 425 337–339 560
[302] Strachan J.D. et al 2003 Nucl. Fusion 43 922 [356] Loarte A. et al 2005 J. Nucl. Mater. 337–339 816
[303] West W.P. et al 2003 J. Nucl. Mater. 313–316 1211 [357] Lingertat J. et al 1997 J. Nucl. Mater. 241–243 402
[304] Neu R. et al 2001 J. Nucl. Mater. 290–293 206 [358] Philipps V. et al 2000 Plasma Phys. Control. Fusion 42 B293
[305] Krieger K. et al 1999 J. Nucl. Mater. 266–269 207 [359] Yoshida N. 1999 J. Nucl. Mater. 266–269 206
[306] West W.P. et al 2001 J. Nucl. Mater. 290–293 783 [360] Ciotti M. et al 1992 J. Nucl. Mater. 196–198 725
[307] Tsitrone E. et al 2001 J. Nucl. Mater. 290–293 331 [361] Hutchinson I.H. et al 1994 Phys. Plasmas 1 1511
[308] Mayer M. et al 2004 Phys. Scrip. t111 55 [362] Pospieszczyk A. et al 2001 J. Nucl. Mater. 290–293 947
[309] Rubel M. et al 2003 J. Nucl. Mater. 313–316 321 [363] Neu R. et al 1997 Plasma Phys. Control. Fusion 38 A165
[310] Likonen J. et al 2003 Fusion Eng. Des. 66–68 219 [364] Neu R. et al 2005 Nucl. Fusion 45 209
[311] Vainonen-Ahlgren E. et al 2005 J. Nucl. Mater. 337–339 55 [365] May M. et al 1999 Plasma Phys. Control. Fusion 41 45
[312] Gotoh Y. et al 2003 J. Nucl. Mater. 313–316 370 [366] Apicella M. et al 2003 J. Nucl. Mater. 313–316 269
[313] Field A.R. et al 1996 Nucl. Fusion 36 119 [367] Dux R. et al 2003 J. Nucl. Mater. 313–316 1150
[314] Asakura N. et al 2002 Proc. 19th Int. Conf. on Fusion Energy [368] Rapp J. et al 1997 Plasma Phys. Control. Fusion 39 1615
2002 (Lyon, France, 2002) (Vienna: IAEA) CD-ROM file [369] Apicella M. et al 1997 Nucl. Fusion 37 381
EX/D1-3 and http://www.iaea.org/programmes/ripc/ [370] Neu R. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
physics/fec2002/html/fec2002.htm 2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA)
[315] Ohya K. et al 2003 Japan. J. Appl. Phys. Part 1 42 5769 CD-ROM file EX/10-5 and http://www-naweb.iaea.org/
[316] Matthews G.F. et al 2003 Proc. 30th EPS Conf. on Controlled napc/physics/fec/fec2004/datasets/index.html
Fusion and Plasma Physics (St Petersburg, Russia, 2003) [371] Koch R. et al 1995 Fusion Eng. Des. 26 103
vol 27A (ECA) P-3.198 [372] Neu R. et al 2002 Plasma Phys. Control. Fusion 44 811
[317] Likonen J. et al 2005 J. Nucl. Mater. 337–339 60 [373] Stober J. et al 2003 Nucl. Fusion 43 1265
[318] Esser H.G. et al 2004 Phys. Scr. t111 129 [374] Dux R. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
[319] Wade M.R. et al 2005 J. Nucl. Mater. 337–339 737 2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA)
[320] Asakura N. et al 2005 J. Nucl. Mater. 337–339 712 CD-ROM file EXP6-14 and http://www-naweb.iaea.org/
[321] Wienhold P. et al 2003 J. Nucl. Mater. 313–316 311 napc/physics/fec/fec2004/datasets/index.html
[322] Kawakami R. and Ohya K. 2003 Japan. J. Appl. Phys. Part 1 [375] Maddaluno G. et al 1997 J. Nucl. Mater. 241–243 908
42 3623 [376] Loarte A. et al 2004 Proc. 20th Int. Conf. on Fusion Energy
[323] Merola M. et al 2002 J. Nucl. Mater. 307–311 1524 2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA)
[324] 2005 ITER Project Integration Document CD-ROM file IT/P3-34 and http://www-naweb.iaea.org/
G-A0-GDRD-6-04-09-09-R0.2-version 1.0 ITER napc/physics/fec/fec2004/datasets/index.html
International Team Report, Naka Joint Work Site, Japan [377] Landman I.S. and Würz H. 2003 J. Nucl. Mater. 313–316 77
[325] Kakudate S. and Shibanuma K. 2003 Fusion Eng. Des. [378] Pappas D. et al 1999 J. Nucl. Mater. 266–269 635
65 133 [379] Verbeek H. et al 1997 Proc. 24th EPS Conf. on Controlled
[326] Tesini A. et al 2005 ITER IT, Technical Memorandum on Fusion and Plasma Physics (Berchtesgaden, Germany,
requirements for first wall shutdown in ITER rev. 3 ITER 1997) vol 21A part IV p 1457
International Team Report, Naka Joint Work Site, Japan [380] Rohde V. et al 2002 Proc. 19th Int. Conf. on Fusion Energy
[327] Matthews G.F. et al 1997 J. Nucl. Mater. 241–243 450 2002 (Lyon, France, 2002) (Vienna: IAEA) CD-ROM file
S260
EX/D1-4 and http://www.iaea.org/programmes/ripc/ CD-ROM file EX/P5-22 and http://www-naweb.iaea.org/

physics/fec2002/html/fec2002.htm napc/physics/fec/fec2004/datasets/index.html
[381] Causey R.A. 2002 J. Nucl. Mater. 300 91 [431] von Seggern J. et al 2004 Phys. Scr. t111 118
[382] Ogorodnikova O. et al 2003 J. Nucl. Mater. 313–316 469 [432] Federici G. et al 2005 J. Nucl. Mater. 337–339 40
[383] Rubel M. et al 2002 J. Nucl. Mater. 307–311 111 [433] Wienhold P. et al 2001 J. Nucl. Mater. 290–293 362
[384] Ueda Y. et al 2004 Nucl. Fusion 44 62 [434] Kirschner A. et al 2004 J. Nucl. Mater. 328 62
[385] Venhaus T. et al 2001 J. Nucl. Mater. 290–293 505 [435] Esser H.G. et al 2005 J. Nucl. Mater. 337–339 84
[386] Wang W. et al 2001 J. Nucl. Mater. 299 124 [436] Ehrhardt A.B. and Langer W.D. 1987 Collisional Processes
[387] Nishijima D. et al 2005 Nucl. Fusion 45 669 of Hydrocarbons in Hydrogen Plasmas Princeton Plasma
[388] Nishijima D. et al 2004 J. Nucl. Mater. 329–333 1029 Physics Laboratory Report PPPL-2477, Princeton
[389] Jotaki E. et al 1997 Fusion Eng. Des. 36 447 University; Princeton, USA
[390] Nachtrieb R.T. et al 1999 J. Nucl. Mater. 266–269 896 [437] Janev R.K. and Reiter D. 2002 Collision Processes of
[391] Gauthier E. et al 1997 J. Nucl. Mater. 241–243 553 Hydrocarbon Species in Hydrogen Plasmas Jül-3966
[392] Esser H.G. et al 1997 J. Nucl. Mater. 241–243 861 [438] Alman D.A. and Ruzic D.N. 2003 J. Nucl. Mater.
[393] Xie J.K. et al 2001 J. Nucl. Mater. 290–293 1155 313–316 182
[394] Brakel D. et al 2001 J. Nucl. Mater. 290–293 1160 [439] Eckstein W. 1991 Computer Simulation of Ion-Solid
[395] de la Cal E. and Gauthier E. 2005 Plasma Phys. Control. Interactions (Berlin: Springer)
Fusion 47 197 [440] Eckstein W. 2002 Calculated sputtering, reflection and range
[396] Hu J.S. et al 2005 Plasma Phys. Control. Fusion 47 1271 values MPI-Garching Report IPP 9/132
[397] Linsmeier Ch. et al 2001 J. Nucl. Mater. 290–293 25 [441] Kirschner A. et al 2003 Plasma Phys. Control. Fusion 45 309
[398] Goldstrass P. and Linsmeier Ch. 2001 J. Nucl. Mater. [442] Tabarés F.L. et al 2004 Plasma Phys. Control. Fusion
290–293 71 46 B381
[399] Okamoto H. and Tanner L.E. 1991 Phase Diagrams of Binary [443] Rubel M. et al 2004 Phys. Scr. t111 112
Tungsten Alloys ed S V Naidu and P Rao (Calcutta : [444] Litnovsky A. et al 2005 J. Nucl. Mater. 337–339 917
Indian Institue of Metals) [445] Doerner R.P. et al 2004 Phys. Srip. t111 75
[400] Doerner R.P. et al 2004 Proc. 20th Int. Conf. on Fusion [446] Schleußner D. et al 1999 J. Nucl. Mater. 266–269 1296
Energy 2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA) [447] McCarthy K.A. et al 1998 Fusion Eng. Des. 42 45
CD-ROM file IT/P3-18 and http://www-naweb.iaea.org/ [448] Piet S.J. et al 1998 Proc. 17th IEEE/NPSS Symp. on Fusion
napc/physics/fec/fec2004/datasets/index.html Engineering (San Diego, CA, USA, 1997) (Piscataway, NJ:
[401] Doerner R.P. et al 2005 J. Nucl. Mater. 342 63 IEEE) IEEE 97CH36131 vol 1 167
[402] Wiltner A. and Linsmeier Ch. 2005 J. Nucl. Mater. [449] Rubel M. et al 2001 Nucl. Fusion 41 1087
337–339 951 [450] Bader A. et al 2004 Rev. Sci. Instrum. 75 370
[403] Schmid K. and Roth J. 2003 J. Nucl. Mater. 313–316 302 [451] Counsell G.F. and Wu C.H. 2001 Phys. Scr. t91 70
[404] Roth J. et al 1997 J. Nucl. Mater. 250 23 [452] Skinner C.H. et al 2004 Phys. Scr. t111 92
[405] Schmid K. and Roth J. 2001 J. Nucl. Mater. 290–293 96 [453] Skinner C.H. et al 2003 J. Nucl. Mater. 313–316 496
[406] Schmid K. et al 2005 J. Nucl. Mater. 337–339 862 [454] Haasz A.A. et al 1996 J. Vac. Sci. Technol. A 14 184
[407] Goldstrass P. et al 1999 J. Nucl. Mater. 266–269 581 [455] Alberici S. et al 1999 J. Nucl. Mater. 266–269 754
[408] Balden M. et al 2005 J. Nucl. Mater. 337–339 980 [456] Haasz A.A. and Davis J.W. 1998 J. Nucl. Mater. 256 65
[409] Davis J.W. and Haasz A.A. 1998 J. Nucl. Mater. 255 214 [457] Davis J.W. and Haasz A.A. 2001 Phys. Scr. t91 33
[410] Poon M. et al 2000 J. Nucl. Mater. 283–287 1062 [458] Philipps V. et al 1999 J. Nucl. Mater. 266–269 386
[411] Davis J.W. and Haasz A.A. 1999 J. Nucl. Mater. 266–269 478 [459] Moormann R. et al 2000 Fusion Eng. Des. 49–50 295
[412] Davis J.W. et al 2002 J. Nucl. Mater. 290–293 66 [460] Lyssoivan et al 1998 Proc. 2nd Europhys. Top. Conf. on RF
[413] Barabash V. et al 2003 J. Nucl. Mater. 313–316 42 Heating and Current Drive of Fusion Devices (Brussels,
[414] Skinner C.H. et al 1996 J. Vac. Sci. Technol. A 14 3267 Belgium) vol 22A (Europhysics Conf. Abstracts) p 85
[415] Mueller D. et al 1996 Fusion Technol. 30 840 [461] Xiao B. et al 2003 J. Nucl. Mater. 313–316 163
[416] Andrew P. et al 1999 Fusion Eng. Des. 47 233 [462] Borrass K. et al 1999 Nucl. Fusion 39 843
[417] Peacock A.T. et al 2000 Fusion Eng. Des. 49–50 745 [463] Kallenbach A. et al 2002 Nucl. Fusion 42 1184
[418] Haasz A.A. and Davis J.W. 2005 Hydrogen Retention in and [464] Mahdavi M.A. et al 2002 Nucl. Fusion 42 52
Removal from Carbon Materials (Springer Series on [465] Kallenbach A. et al 2005 J. Nucl. Mater. 337–339 381
Chem. Physics vol 78) (Berlin : Springer) [466] Porter G.D. et al 1998 Phys. Plasmas 5 1410
[419] Bekris N. et al 2002 J. Nucl. Mater. 307–311 1649 [467] Mahdavi M.A. et al 2003 Phys. Plasmas 10 3984
[420] Skinner C.H. et al 2001 J. Nucl. Mater. 290–293 486 [468] Borrass K. et al 1997 Nucl. Fusion 37 523
[421] Mueller D. et al 1998 Proc. 17th IEEE/NPSS Symp. on [469] Krasheninnikov S.I. et al 1999 J. Nucl. Mater. 266–269 251
Fusion Engineering (San Diego, CA, USA, 1997) vol 1 [470] Stober J. et al 2001 Nucl. Fusion 41 1123
(Piscataway New Jersey: IEEE) IEEE 97CH36131 p 279 [471] Saibene G. et al 2002 Plasma Phys. Control. Fusion 44 1769
[422] Bekris N. et al 2003 J. Nucl. Mater. 313–316 501 [472] Yao L. et al 1998 Nucl. Fusion 38 631
[423] Philipps V. et al 2004 Proc. 20th Int. Conf. on Fusion Energy [473] Yao L. et al 2001 Nucl. Fusion 41 817
2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA) [474] Gao X. et al 2000 Nucl. Fusion 40 1875
CD-ROM file EX/10-1 and http://www-naweb.iaea.org/ [475] Baldzuhn J. et al 2003 Proc. 30th EPS Conf. on Controlled
napc/physics/fec/fec2004/datasets/index.html Fusion and Plasma Physics (St Petersburg, Russia, 2003)
[424] Mayer M. et al 2005 J. Nucl. Mater. 337–339 119 vol 27A (ECA) P-4.166
[425] Rohde V. et al 2005 J. Nucl. Mater. 337–339 847 [476] Jiafu D. et al 2002 Plasma Phys. Control. Fusion 44 371
[426] Tanabe T. 2004 Kyushu University private communication [477] Bucalossi J. et al 2002 Proc. 19th Int. Conf. on Fusion Energy
[427] Hirohata Y. et al 2005 J. Nucl. Mater. 337–339 609 2002 (Lyon, France, 2002) (Vienna: IAEA) CD-ROM file
[428] Bucalossi J. et al 2001 28th EPS Conf. on Controlled Fusion EX/P4-04 and http://www.iaea.org/programmes/ripc/
and Plasma Physics (Funchal, Portugal, 2001) vol 25A physics/fec2002/html/fec2002.htm
(ECA) 1629 [478] Loarer T. et al 2003 Proc. 30th EPS Conf. on Controlled
[429] Mertens V. et al 2003 30th EPS Conf. on Controlled Fusion Fusion and Plasma Physics (St Petersburg, Russia, 2003)
and Plasma Physics (St Petersburg, Russia, 2003) vol 27A vol 27A (ECA) P-1.161
(ECA) P-1.128 [479] Tsitrone E. et al 2003 Proc. 30th EPS Conf. on Controlled
[430] Loarer T. et al 2004 Proc. 20th Int. Conf. on Fusion Energy Fusion and Plasma Physics (St Petersburg, Russia, 2003)
2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA) vol 27A (ECA) O-2.5A
S261
A. Loarte et al
[480] Lang P.T. et al 2003 Proc. 30th EPS Conf. on Controlled [522] Mukhovatov V. et al 2003 Nucl. Fusion 43 942
Fusion and Plasma Physics (St Petersburg, Russia, 2003) [523] Pacher G.W. et al 2004 Plasma Phys. Control. Fusion
vol 27A (ECA) P-1.129 46 A257
[481] Lang P.T. et al 2004 Proc. 20th Int. Conf. on Fusion Energy [524] Rozhansky V.A. et al 2003 J. Nucl. Mater. 313–316 1141
2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA) [525] Wiesen S. et al 2003 Proc. 30th EPS Conf. on Controlled
CD-ROM file EX2/6 and http://www-naweb.iaea.org/ Fusion and Plasma Physics (St Petersburg, Russia, 2003)
napc/physics/fec/fec2004/datasets/index.html vol 27A (ECA) P-3.194
[482] Baylor L.R. et al 2003 J. Nucl. Mater. 313–316 530 [526] LaBombard B. et al 2003 J. Nucl. Mater. 313–316 995
[483] Lang P.T. et al 1997 Phys. Rev. Lett. 79 1487 [527] Smick N. et al 2005 J. Nucl. Mater. 337–339 281
[484] Müller H.W. et al 1999 Phys. Rev. Lett. 83 2199 [528] Kirnev G.S. et al 2005 J. Nucl. Mater. 337–339 271
[485] Pégourié B. et al 2003 J. Nucl. Mater. 313–316 539 [529] Coster D.P. et al 2005 J. Nucl. Mater. 337–339 366
[486] Baylor L.R. et al 1997 Nucl. Fusion 37 445 [530] Heikkinen J.A. et al 2001 J. Comput. Phys. 173 527
[487] Lang P.T. et al 2001 Nucl. Fusion 41 1107 [531] Kurki-Suonio T. et al 2002 Nucl. Fusion 42 725
[488] Rozhansky V.A. et al 1995 Plasma Phys. Control. Fusion [532] Fundamenski W. et al 2002 Plasma Phys. Control. Fusion
37 399 44 761
[489] Parks P.B. et al 2000 Phys. Plasmas 7 1968 [533] Hassanein A. and Konkashbaev I. 1999 J. Nucl. Mater.
[490] Baylor L.R. et al 1999 J. Nucl. Mater. 266–269 457 273 326
[491] Jones T.T.C. et al 2000 Proc. 27th EPS Conf. on Controlled [534] Hassanein A. 2002 Fusion Eng. Des. 60 527
Fusion and Plasma Physics (Budapest, Hungary, 2000) [535] Wuerz H. et al 2001 Fusion Sci. Technol. 40 191
vol 24B 2000 paper OR04 [536] Hassanein A. and Konkashbaev I. 2003 J. Nucl. Mater.
[492] Takenaga H. et al 2001 Phys. Plasmas 8 2217 313–316 664
[493] Kubo H. et al 2002 Phys. Plasmas 9 2127 [537] Pestchanyi S.E. et al 2003 Proc. 30th EPS Conf. on
[494] Geraud A. et al 1998 Proc. 20th SOFT Conf. on Fusion Controlled Fusion and Plasma Physics (St Petersburg,
Technology (Marseille, France) vol 2 p 942 Russia, 2003) vol 27A (ECA) P-2.164
[495] Andelfinger C. et al 1993 Rev. Sci. Instrum. 64 983 [538] Adams M.L. et al 2001 J. Quant. Spectrosc. Radiat. Transfer
[496] Geraud A. et al 2003 Fusion Eng. Des. 69 5 71 117
[497] Geraud A. et al 2005 J. Nucl. Mater. 337–339 485 [539] Lovegrove T. et al 1995 Proc. 22nd EPS Conf. on Controlled
[498] Watson M.J. and Milnes J. 2003 EFDA Task Final Report Fusion and Plasma Physics (Bournemouth, UK, 1995)
TW2-TTFD-VD25 vol 19C (ECA) III-301
[499] Polevoi A.R. et al 2003 Nucl. Fusion 43 1072 [540] Reiter D. et al 2003 J. Nucl. Mater. 313–316 845
[500] Polevoi A.R. and Shimada M. 2001 Plasma Phys. Control. [541] Krasheninnikov S.I. et al 1988 Contrib. Plasma Phys. 28 443
Fusion 43 1525 [542] Behringer K. 1998 Escape factors for line emission and
[501] Raman R. et al 1997 Nucl. Fusion 37 967 population calculation MPI-Garching Report IPP 10/11
[502] Ogawa T. et al 2001 J. Nucl. Mater. 290–293 454 [543] Marchand R. et al 1992 Phys. Fluids B 4 924
[503] Loarte A. et al 1999 J. Nucl. Mater. 266–269 1123 [544] Wan A.S. et al 1995 J. Nucl. Mater. 220–222 1102
[504] Igitkhanov Yu. et al 1994 Proc. 21st EPS Conf. on Controlled [545] Adams M.L. et al 2002 Contrib. Plasma Phys. 42 395
Fusion and Plasma Physics (Montpellier, France, 1994) [546] Scott H.A. and Adams L.A. 2004 Contrib. Plasma Phys.
vol 18B (ECA) p 766 44 51
[505] Kukushkin A.S. and Runov A.M. 1994 Contrib. Plasma [547] Reiter D. et al 2004 The EIRENE code http://www.eirene.de
Phys. 34 204 [548] Adams M.L. and Scott H.A. 2004 Contrib. Plasma Phys.
[506] Neuhauser J. et al 1999 Proc. 26th EPS Conf. on Controlled 44 262
Fusion and Plasma Physics (Maastricht, The Netherlands, [549] Reiter D. et al 2005 Fusion Sci. Technol. 47 172
1999) paper P4.040 [550] Fantz U. et al 1999 J. Nucl. Mater. 266–269 490
[507] Kukushkin A.S. et al 2000 Contrib. Plasma Phys. 40 233 [551] Sawada K. and Fujimoto T. 1995 J. Appl. Phys. 78 2913
[508] Pacher H.D. et al 1999 J. Nucl. Mater. 266–269 1172 [552] Pigarov A.Yu. 2002 Phys. Scr. t96 16
[509] Taroni A. et al 1995 Proc. 22th EPS Conf. on Controlled [553] Kubo H. et al 2005 J. Nucl. Mater. 337–339 161
Fusion and Plasma Physics (Bournemouth, UK, 1995) [554] Pitts R.A. et al 2000 Proc. 18th Int. Conf. on Fusion Energy
vol 19C (ECA) Part IV (Geneva: European Physical 2000 (Sorrento, Italy, 2000) (Vienna: IAEA) CD-ROM
Society) p 297 file EXP4/23 and http://www.iaea.org/programmes/ripc/
[510] Wesson J. 1997 Tokamaks (Oxford: Oxford University physics/fec2000/html/node1.htm
Press) [555] Wischmeier M. et al 2004 Contrib. Plasma Phys. 44 268
[511] Cohen R. and Ryutov D.D. 2004 Contrib. Plasma Phys. [556] Janev R.K., Kato T. and Wang J.G. 2000 Phys. Plasmas
44 111 7 4364
[512] Kukushkin A.S. et al 1999 Proc. 26th EPS Conf. on [557] Janev R.K. et al 2003 Collision processes in low-temperature
Controlled Fusion and Plasma Physics (Maastricht, The hydrogen plasmas Report Forschungszentrum Jülich JUEL
Netherlands, 1999) vol 23J (ECA) p 1545 4105
[513] Kukushkin A.S. et al 2005 J. Nucl. Mater. 337–339 50 [558] Lisgo S. et al 2003 J. Nucl. Mater. 313–316 1089
[514] Kukushkin A.S. et al 2003 Proc. 30th EPS Conf. on [559] Cambe A. et al 2003 J. Nucl. Mater. 313–316 364
Controlled Fusion and Plasma Physics (St Petersburg, [560] Kirschner A. et al 2003 Proc. 30th EPS Conf. on Controlled
Russia, 2003) vol 27A (ECA) P-3.195 Fusion and Plasma Physics (St Petersburg, Russia, 2003)
[515] Kirschner A. et al 2003 J. Nucl. Mater. 313–316 444 vol 27A (ECA) P-3.196
[516] Kukushkin A.S. et al 2005 Nucl. Fusion 45 608 [561] Ohya K. et al 2003 J. Nucl. Mater. 313–316 568
[517] Kukushkin A.S. et al 2001 J. Nucl. Mater. 290–293 887 [562] Hyodo I. et al 2003 J. Nucl. Mater. 313–316 1183
[518] Pacher H.D. et al 2003 J. Nucl. Mater. 313–316 657 [563] Stangeby P.C. and Elder J.D. 1999 Simulating net erosion
[519] Coster D.P. et al 2002 Proc. 19th Int. Conf. on Fusion Energy near the inside target of JET using the DIVIMP code
2002 (Lyon, France, 2002) (Vienna: IAEA) CD-ROM file Tritium Retention Meeting, JET, Culham Science Centre,
TH/P2-13 and http://www.iaea.org/programmes/ripc/ UK
physics/fec2002/html/fec2002.htm [564] Coster D.P. et al 2003 J. Nucl. Mater. 313–316 868
[520] Monk R.D. et al 1997 Proc. 24th EPS Conf. on Controlled [565] Alman D.A. et al 2000 Phys. Plasmas 7 1421
Fusion and Plasma Physics (Berchtesgaden, Germany, [566] Janev R.K., Wang J.G. and Kato T. 2001 NIFS-DATA-64
1997) vol 21A (ECA) p 117 [567] Janev R.K. and Reiter D. 2003 J. Nucl. Mater. 313–316 1202
[521] Pacher G.W. et al 2003 Nucl. Fusion 43 188 [568] Traskelin P. et al 2003 J. Nucl. Mater. 313–316 52
S262
[569] Salonen E. et al 2003 J. Nucl. Mater. 313–316 404 [589] Mazurenko A. et al 2002 Phys. Rev. Lett. 89 225004-1
[570] Pappas D. 2000 Study of molybdenum sources and screening [590] Rognlien T.D. et al 2004 Contrib. Plasma Phys. 44 188
in the Alcator C-mod tokamak PhD Thesis MIT published [591] Xu X.Q. et al 2003 Phys. Plasmas 10 1773
as PSFC Report PFC/RR-00-6 [592] Sarazin Y. and Ghendrih Ph. 1998 Phys. Plasmas 5 4214
[571] Kawakami R. and Ohya K. 2003 J. Nucl. Mater. 313–316 107 [593] Garbet X. et al 1999 Nucl. Fusion 39 2063
[572] Strachan J.D. et al 2004 Nucl. Fusion 44 772 [594] Bian N. et al 2003 Phys. Plasmas 43 1382
[573] Stangeby P.C. et al 2003 J. Nucl. Mater. 313–316 883 [595] Ghendrih Ph. et al 2003 Nucl. Fusion 43 1013
[574] Coster D.P. et al 2002 Plasma Phys. Control. Fusion 44 979 [596] Garcia O. et al 2004 Phys. Rev. Lett. 92 165003
[575] Wootton A.J. et al 1990 Phys. Fluids B 2 2879 [597] Sarazin Y. et al 2000 Phys. Plasmas 7 1085
[576] Hidalgo C. et al 1996 Proc. 16th Int. Conf. on Fusion Energy [598] Garbet X. et al 1988 Proc. 12th. Int. Conf. on Plasma Physics
(Montreal, Canada, 1996) (Vienna: IAEA 1997) vol 1 and Controlled Nuclear Fusion Research 1988 (Nice,
p 617 France, 1988) (Vienna: IAEA) vol 2 p 163
[577] Antar G.Y. et al 2001 Phys. Plasmas 8 1612 [599] Sarazin Y. et al 2003 J. Nucl. Mater. 313–316 796
[578] Connor J.W. and Helander P. 1999 Plasma Phys. Control. [600] Falchetto G.L. et al 2004 Proc. 20th Int. Conf. on Fusion
Fusion 41 251 Energy 2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA)
[579] Nedospasov A.V. et al 1989 Sov. J. Plasma Phys. 15 659 CD-ROM file TH/1-3Rd and http://www-naweb.iaea.org/
[580] Garbet X. et al 1991 Nucl. Fusion 31 967 napc/physics/fec/fec2004/datasets/index.html
[581] Beyer P. and Spatschek K.H. 1996 Phys. Plasmas 3 995 [601] Lin Z. et al 1998 Science 281 1835
[582] Xu X.Q. and Cohen R.H. 1998 Contrib. Plasma Phys. 36 [602] Diamond P.H. and Hahm T.S. 1995 Phys. Plasmas 2 3640
158 [603] Ghendrih Ph. et al 2004 Proc. 20th Int. Conf. on Fusion
[583] Berk H.L. et al 1991 Phys. Fluids B 5 1336 Energy 2004 (Vilamoura, Portugal, 2004) (Vienna: IAEA)
[584] Ribeiro T. and Scott B. 2005 Plasma Phys. Control. Fusion CD-ROM file TH/1-3Ra and http://www-naweb.iaea.org/
47 1657 napc/physics/fec/fec2004/datasets/index.html
[585] Krasheninnikov S.I. et al 2002 Proc. 19th Int. Conf. on [604] Ghendrih Ph. et al 2005 J. Nucl. Mater. 337–339 347
Fusion Energy 2002 (Lyon, France, 2002) (Vienna: IAEA) [605] Aydemir A.Y. 2005 Phys. Plasmas 12 062503
CD-ROM file TH/4-1 and http://www.iaea.org/ [606] Peng Y.K.M. et al 1994 Proc. 15th Int. Conf. on Plasma
programmes/ripc/physics/fec2002/html/fec2002.htm Physics and Controlled Nuclear Fusion Research 1994
[586] Xu X.Q. et al 2000 Nucl. Fusion 40 731 (Seville, Spain, 1994) (Vienna: IAEA) vol 2 p 643
[587] Xu X.Q. et al 2004 Contrib. Plasma Phys. 44 105 [607] Hidalgo C. 2003 private communication Euratom-CIEMAT
[588] McKee G.R. et al 2003 Phys. Plasmas 10 1712 association
S263
View publication stats

Progress ITER Physics Basis Chapter 4 2007

Uploaded by

Copyright:

Available Formats

Progress ITER Physics Basis Chapter 4 2007

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Progress ITER Physics Basis Chapter 4 2007

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Chapter 4: Power and particle control

Article in Nuclear Fusion · June 2007

Alberto Loarte Bruce Lipschultz

SEE PROFILE SEE PROFILE

A. S. Kukushkin Nobuyuki Asakura

SEE PROFILE SEE PROFILE

Greek Comparatives View project

The user has requested enhancement of the downloaded file.

Chapter 4: Power and particle control

Received 23 August 2005, accepted for publication 2 May 2006

a Author to whom any correspondence should be addressed.

PACS numbers: 52.40.Hf, 52.55.Fa, 52.55.Rk, 52.50.Kj, 52.65.Pp

2.5.5. Wall conditioning and particle control

1. Introduction The steady-state heat fluxes to divertor plates are now a

the efficiency of the fuelling methods in tokamak devices and 18

Heat Load (MW/m2)

2.1. Steady-state transport in the SOL and its implications 4

20 profiles on both targets, and also predicts a strong modification

12 figure 5) and the ASDEX-Upgrade scaling for DIV II from

are qacc = 10 MW m−2 , Rt = 5.5 m, PSOL ∼ 80 MW and

ne and temperature Te,i (see section 2.1.1) together with 1.0

∆ (net and contributions)

Near/Far SOL and Fluctuations Regression Analysis of Deff,

0.1 0.2 0.3 0.4 0.5

temperature is obtained from real-time measurements of the

τIR , with significant power fluxes arriving at the divertor during

Outer Divertor Target

2MA q95 = 4.6 35

separatrix [175–177]. As a consequence, only a small toroidal

0.6 3186 3188 3190 3192 3194

# PDP-16775 ; time - 2.14356 s ; averaged : 0.0 s

parallel heat flux / MW/m 2

of the dependence of the radial velocity on experimental

ASDEX-Upgrade [188] during Type I ELMs, as shown in

ELM energy loss) [188,194]. This result is tentative at present.

200 ~ 2.2 x (τII -175)

ITER-size tokamak but are generally found at too high pedestal

Mark I Mark IIA 3.5 Mark IIGB

Divertor neutral pressure (10-3 mbar)

2.0 Mark IIA/P

in carbon clad fusion devices. A tokamak demonstration of a

gross erosion (m-2s-1)

retains D at similar concentrations as carbon [350, 351], with

400 K. As described in section 2.5.3, recent data from PISCES-

the D retention in Be with impurities. A C–Be mixed-material

[eV] 100 importance in ITER operation, remains an active field of

Table 3. Compilation of fuel retention data from various fusion devices.

tion in the device. This approach has been followed in Tore

Carbon deposition (C/cm )

a rate of about 2.5 × 1020 D s−1 , which is about 50% of the

injected fuel. The amount of fuel retained in the analysed

and, thus, further research in this area is on-going to resolve

tween post-mortem analysis and fuel retention determination Discharge number

Be eroded from the main wall, which decreases the C surface

detail in section 2.5. Due to concerns of the compatibility

density widths ratio

(inside sep. / full)

The normalized separatrix density, nsep /nGreen , from 1.2