4.19 Beryllium as a Plasma-Facing Material for Near-Term
Fusion Devices
G. Federici
Fusion for Energy, Garching, Germany

R. Doerner
University of California at San Diego, San Diego, CA, USA

P. Lorenzetto
Fusion for Energy, Barcelona, Spain

V. Barabash
ITER Organization, St Paul Lez Durance, France

ß 2012 Fusion for Energy (F4E). Published by Elsevier Ltd. All rights reserved.

4.19.1
4.19.2
4.19.2.1
4.19.2.2
4.19.2.3
4.19.3
4.19.3.1
4.19.3.1.1
4.19.3.1.2
4.19.3.1.3
4.19.3.1.4
4.19.3.2
4.19.3.2.1
4.19.3.2.2
4.19.3.3

4.19.3.3.1
4.19.3.3.2
4.19.4
4.19.4.1
4.19.4.1.1
4.19.4.1.2
4.19.4.2
4.19.4.3
4.19.4.4
4.19.4.4.1
4.19.4.4.2
4.19.4.4.3
4.19.4.4.4
4.19.4.4.5
4.19.5
4.19.5.1
4.19.5.1.1
4.19.5.1.2
4.19.5.2
4.19.6
4.19.6.1
4.19.6.1.1
4.19.6.1.2
4.19.6.2

Introduction
Background
Synopsis of PWIs in Tokamaks
Brief History of Plasma-Facing Materials in Fusion Devices
Experience with Beryllium in Tokamaks

Beryllium PWI Relevant Properties
Beryllium Erosion Properties
Physical sputtering of beryllium
Mixed-material erosion
Chemically assisted sputtering of beryllium
Enhanced erosion at elevated temperatures
Hydrogen Retention and Release Characteristics
Implantation
Beryllium codeposition
Mixed-Material Effects
Be–C phenomena
Be–W alloying
Main Physical and Mechanical Properties
General Considerations
Physical properties
Mechanical properties
Selection of Beryllium Grades for ITER Applications
Considerations on Plasma-Sprayed Beryllium
Neutron-Irradiation Effects
Thermal conductivity
Swelling
Mechanical properties
Thermal shock effects
Bulk tritium retention
Fabrication Issues
Joining Technologies and High Heat Flux Durability of the Be/Cu Joints
Be/Cu alloy joining technology
High heat flux durability of unirradiated Be/Cu joints
Thermal Tests on Neutron-Irradiated Joints
Tokamak PFC Design Issues and Predictions of Effects in ITER During Operation

PFC Design Considerations
Design of the beryllium ITER-like wall at JET
Design of the beryllium ITER wall
Predictions of Effects on the ITER Beryllium Wall During Operation

623
624
624
626
628
629
629
629
631
632
632
633
633
635
637
637
637
638
638
639
639
640
642
643
643

643
644
644
644
644
644
645
646
648
650
650
650
651
653

621


622

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

4.19.6.2.1
4.19.6.2.2
4.19.6.3
4.19.7
References

Safety issues in ITER
Erosion/damage of the ITER Be wall

Prospect of Using Beryllium in Beyond-ITER Fusion Reactors
Concluding Remarks

Abbreviations
Alcator
C-Mod

ANFIBE
ASDEXUpgrade

ATC
CFC
CIP
CP
DIII-D

DIMES

DS-Cu
EAST

The name Alcator was given to a
class of tokamaks designed and
built at the Massachusetts Institute
of Technology; these machines are
distinguished by high magnetic
fields with relatively small
diameters. The high magnetic field
helps create plasmas with
relatively high current and particle

densities. The present incarnation
is Alcator C-Mod
Computer code for ANalysis of
Fusion Irradiated BEryllium
Axially Symmetric Divertor
Experiment. The original ASDEX,
located in Garching, Germany, and
decommissioned in about 1990,
would qualify today as a medium
sized tokamak. It was designed for
the study of impurities and their
control by a magnetic divertor.
Its successor, ASDEX-Upgrade
(a completely new machine, not
really an ‘upgrade’), is larger and
more flexible.
Adiabatic Toroidal Compressor
Carbon-fiber composite
Cold isostatic pressing
Cold pressing
A medium-sized tokamak, but the
largest tokamak still operational in
the United States. Operated by
General Atomics in San Diego
Divertor Material Evaluation
Studies, a retractable probe that
allows the insertion and retraction
of test material samples to the DIIID divertor floor, for example, for
erosion/deposition studies.
Dispersion-strengthened copper

Experimental advanced
superconducting tokamak – an
experimental superconducting

ELMs
FISPACT
FZJ
HIP
INEEL

ISX

ITER

JET

653
654
659
659
662

tokamak magnetic fusion energy
reactor in Hefei, the capital city of
Anhui Province, in eastern China
Edge localized modes
Inventory code included in the
European Activation System
Forschungszentrum Juelich,
Germany

Hot isostatic pressing
Idaho National Engineering and
Environmental Laboratory. Now
Idaho National Laboratory (INL)
Impurity study experiment (ISX-A
and ISX-B where two tokamaks
operated at Oak Ridge National
Laboratory)
ITER, the world’s largest tokamak
experimental facility being
constructed in the South of France
to demonstrate the scientific and
technical feasibility of fusion
power. The project is being built on
the basis of an international
collaboration between the
European Union, China, India,
Japan, Russia, South Korea, and
the United States. The
international treaty was signed in
November 2006 and the central
organization established in
Cadarache. Most of the
components will be provided in
kind by agencies set up for this
purpose in the seven partners
Joint European Torus – a large
tokamak located at the Culham
Laboratory in Oxfordshire,
England, jointly owned by the

European Community. First
device to achieve >1 W of fusion
power, in 1991, and the machine
that has most closely approached
Q ¼ 1 for DT operation (Q ¼ 0:95
in 1997)


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

JUDITH

Juelich Divertor Test Facility in Hot
Cells
KSTAR
Korea Superconducting Tokamak
Advanced Reactor – a long-pulse,
superconducting tokamak built in
South Korea to explore advanced
tokamak regimes under steady
state conditions
LANL
Los Alamos National Laboratory
LCFS
Last closed flux surface
MAR
ITER Materials Assessment Report
MIT
Massachusetts Institute of
Technology

MPH
ITER Materials Properties
Handbook
NBI
Neutral beam injection
NRA
Nuclear reaction analysis
NRI
Nuclear Research Institute in the
Czech Republic
PDX
Poloidal divertor experiment
PFCs
Plasma-facing components
PISCES
Plasma Interaction with Surface
Components Experimental
Station. It is a plasma simulator
located at the University of
California San Diego in the United
States (originally at University of
California, Los Angeles) that is
used to test materials and
measure sputtering, retention, etc.
expected in tokamaks
PLT
Princeton Large Torus
PWIs
Plasma–wall interactions
RES

Radiation enhanced sublimation
RMP
Resonance magnetic perturbation
SNL
Sandia National Laboratory
SOL
Scrape-off-layer
ST
Symmetric tokamak (in this
chapter)
STEMET 1108 Brazing alloy: Cu–Sn–In–Ni
TFR
Torus Fontenay-aux-Roses
TPE
Tritium plasma experiment
TRIM
Transport of ion in matter code
UCSD
University of California, San Diego
UNITOR
One of the first small tokamaks
where beryllium was used
UTIAS
University of Toronto Institute for
Aerospace Studies
VDE
Vertical displacement event
VHP
Vacuum hot pressing


623

4.19.1 Introduction
Beryllium, once called ‘the wonder metal of the future,’ 1
is a low-density metal that gained early prominence
as a neutron reflector in weapons and fission research
reactors. It then found a wide range of applications
in the automotive, aerospace, defense, medical, and
electronic industries. Also, because of its unique
physical properties, and especially favorable plasma
compatibility, it was considered and used in the past
for protection of internal components in various
magnetic fusion devices (e.g., UNITOR, ISX-B, JET).
Most important future (near-term) applications in this
field include (1) the installation of a completely new
beryllium wall in the JET tokamak, which has been
completed by mid of 2011 and consists of $1700 solid
Be tiles machined from 4 t of beryllium; and (2) ITER,
the world’s largest experimental facility to demonstrate the scientific and technical feasibility of fusion
power, which is being built in Cadarache in the South
of France. ITER will use beryllium to clad the first
wall ($700 m2 for a total weight of about 12 t of Be).
Although beryllium has been considered for other
applications in fusion (e.g., as neutron multiplier in
the design of some types of thermonuclear breeding
blankets of future fusion reactors and for hohlraums
in inertial confinement fusion), this chapter will
be limited to discussing the use of beryllium as a
plasma-facing material in magnetic confinement
devices, and in particular in the design, research,

and development work currently underway for
the JET and the ITER tokamaks. Considerations
related to health and safety procedures for the use
of beryllium relevant for construction and operation
in tokamaks are not discussed here.
Designing the interface between a thermonuclear
plasma and the surrounding solid material environment has been arguably one of the greatest technical
challenges of ITER and will continue to be a challenge for the development of future fusion power
reactors. The interaction between the edge plasma
and the surrounding surfaces profoundly influences
conditions in the core plasma and can damage the
surrounding material structures and lead to long
machine downtimes for repair. Robust solutions for
issues of plasma power handling and plasma–wall
interactions (PWIs) are required for the realization
of a commercially attractive fusion reactor. A mix
of several plasma-facing materials is currently proposed in ITER to optimize the requirements of areas
with different power and particle flux characteristics


624

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

(i.e., Be for the first wall, carbon-fiber composite
(CFC) for the divertor strike point tiles, and W
elsewhere in the divertor). Inevitably, this is expected
to lead to cross-material contamination and the formation of material mixtures, whose behavior is still
uncertain and requires further investigation.
The use of beryllium for plasma-facingcomponent (PFC) applications has been the subject

of many reviews during the last two decades (see,
e.g., Wilson et al.2 and Raffray et al.3 and references
therein). Much of this fusion-related work has been
summarized in a series of topical workshops on beryllium that were held in the past, bringing together
leading researchers in the field of beryllium technology and disseminating information on recent
progress in the field.4 Comprehensive reviews have
also appeared recently in specialized journals5,6 containing state-of-the-art information on a number of
topics such as manufacturing and development of coating techniques, component design, erosion/deposition,
tritium retention, material mixing and compatibility
problems, safety of beryllium handling, etc.
This chapter reviews the properties of beryllium
that are of primary relevance for plasma protection
applications in near-term magnetic fusion devices
(i.e., PWIs, thermal and mechanical properties, fabricability and ease of joining, chemical reactivity, etc.)
together with the available knowledge on performance and operation in existing fusion machines.
Special attention is given to beryllium’s erosion and
deposition, the formation of mixed materials, and the
hydrogen retention and release characteristics that
play an important role in plasma performance, component lifetime, and operational safety. The status of
the available techniques presently considered for
joining the beryllium armor to the heat sink material
of Cu alloys for the fabrication of beryllium-clad
actively cooled components for the ITER first wall
is briefly discussed together with the results of the
performance and durability heat flux tests conducted
in the framework of the ITER first-wall qualification
programme. The effects of neutron irradiation on the
degradation of the properties of beryllium itself and
of the joints are also briefly analyzed.
This chapter is organized as follows. Section

4.19.2 provides some background information for
the reader and briefly reviews (1) the problem of
PWIs in tokamaks; (2) the history of plasma-facing
materials in fusion devices and the rationale for
choosing beryllium as the material for the first wall
of JET and ITER; and (3) the experience with the use
of beryllium in tokamaks to date. Section 4.19.3

describes in detail the beryllium PWI-relevant properties such as erosion/deposition, hydrogen retention
and release, and chemical effects such as material
mixing, all of which influence the selection of beryllium as armor material for PFCs. Section 4.19.4
briefly reviews a limited number of selected physical and mechanical properties of relevance for
the fabrication of heat exhaust components and the
effects of neutron irradiation on material properties.
Section 4.19.5 describes the fabrication issues and
the progress of joining technology and high heat flux
durability of beryllium-clad PFCs. Section 4.19.6
describes the main issues associated with the JET
and ITER first-wall designs and discusses some constraints foreseen during operation. The prospects of
beryllium for applications in fusion reactors beyond
ITER are briefly discussed. Finally, a summary is
provided in Section 4.19.7.

4.19.2 Background
4.19.2.1

Synopsis of PWIs in Tokamaks

A detailed discussion on this subject is beyond the
scope of this review. The relevant PWIs are comprehensively reviewed by Federici et al.7,8 More recent

interpretations of the underlying phenomena and
impact on the ITER device can be found in Roth
et al.9 Here we summarize some of the main points.
PWIs critically affect tokamak operation in many
ways. Erosion by the plasma determines the lifetime
of PFCs, and creates a source of impurities, which cool
and dilute the plasma. Deposition of material onto
PFCs alters their surface composition and, depending
on the material used, can lead to long-term accumulation of large in-vessel tritium inventories. This latter
phenomenon is especially exacerbated for carbonbased materials but there are still some concerns with
beryllium. Retention and recycling of hydrogen from
PFCs affects fuelling efficiency, plasma density control,
and the density of neutral hydrogen in the plasma
boundary, which impacts particle and energy transport.
The primary driver for the interactions between
the core plasma, edge plasma, and wall is the power
generated in the plasma core that must be handled by
the surrounding structures. Fusion power is obtained
by the reaction of two hydrogen isotopes, deuterium
(D) and tritium (T), producing an a-particle and a
fast neutron. Although the kinetic energy carried by
the 14.1 MeV neutron escapes the plasma and could
be converted in future reactors beyond ITER to
thermal energy in a surrounding blanket system, the


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

kinetic energy of the a-particle is deposited in the
plasma. The fraction of this power that is not radiated

from the plasma core as bremsstrahlung or line radiation (and that on average is distributed uniformly on
the surrounding structures) is transported across field
lines to the edge plasma and intersects the material
surfaces in specific areas leading to intense power
loads. The edge plasma has a strong influence on
the core plasma transport processes and thereby on
the energy confinement time. A schematic representation of the regions of the plasma and boundary
walls in a divertor tokamak is portrayed in Figure 1
taken from Federici et al.7
The outermost closed magnetic field surface forms
an X-point of zero poloidal magnetic field within the
vessel. This boundary is called the ‘last closed flux

Magnetic flux surfaces
Separatrix (LCFS)
Edge region
Scrape-off
layer

First wall

Plasma core

Separatrix (LCFS)
X-point

Divertor region

Baffle


Vertical divertor
target plate

Private flux
region
Separatrix strike point
Pump

Figure 1 Poloidal cross-section of a tokamak plasma with
a single magnetic null divertor configuration, illustrating the
regions of the plasma and the boundary walls where
important PWIs and atomic physics processes take
place. The characteristic regions are (1) the plasma core,
(2) the edge region just inside the separatrix, (3) the
scrape-off-layer (SOL) plasma outside the separatrix, and
(4) the divertor plasma region, which is an extension of the
SOL plasma along field lines into the divertor chamber.
The baffle structure is designed to prevent neutrals from
leaving the divertor. In the private flux region below the
X-point, the magnetic field surfaces are isolated from the rest
of the plasma. Reproduced with permission from Federici, G.;
Skinner, C. H.; Brooks, J. N.; et al. Plasma-material
interactions in current tokamaks and their implications
for next-step fusion reactors. Nucl. Fusion 2001, 41,
1967–2137 (review special issue), with permission from IAEA.

625

surface’ (LCFS) or ‘separatrix.’ Magnetic field surfaces
inside the LCFS are closed, confining the plasma ions.

The edge region, just inside the LCFS, contains significant levels of impurities not fully ionized, and perhaps
neutral particles. Impurity line radiation and neutral
particles transport some power from here to the wall.
The remaining power enters the region outside the
LCFS either by conduction or, depending on the
degree to which neutrals penetrate the plasma, by
convection. This region is known as the scrape-offlayer (SOL) as here power is rapidly ‘scraped off ’ by
electron heat conduction along open field lines, which
are diverted to intersect with target regions that are
known as ‘divertors.’ Poloidal divertors have been very
successful at localizing the interactions of plasma ions
with the target plate material in a part of the machine
geometrically distant from the main plasma where any
impurities released are well screened from the main
plasma and return to the target plate.
The plasma density and temperature determine
the flux density and energy of plasma ions striking
the plasma-wetted surfaces. These, in turn, determine the rate of physical sputtering, chemical sputtering, ion implantation, and impurity generation.
The interaction of the edge plasma with the surrounding solid material surfaces is most intense in
the vicinity of the ‘strike point’ where the separatrix
intersects the divertor target plate (see inset in
Figure 1). In addition, the plasma conditions determine where eroded material is redeposited, and
to what degree codeposition of tritium occurs at the
wall. The plasma power flow also determines the level
of active structural cooling required.
Typical plasma loads and the effects expected
during normal operation and off-normal operation
in ITER are summarized in Table 1.
Because of the very demanding power handling
requirements (predicted peak value of the heat flux in

the divertor near the strike-points is >10 MW mÀ2)
and the predicted short lifetime due to sputtering
erosion arising from very intense particle fluxes
($1023–1024 particles mÀ2 sÀ1) and damage during
transient events, beryllium has been excluded from
use in the ITER divertor and is instead the material
selected for the main chamber wall of ITER.
Recent observations in present divertor tokamaks
have shown that plasma fluxes to the main wall
are dominated by intermittent events leading to fast
plasma particle transport that reaches the PFCs along
the magnetic field (see Loarte et al.10 and references
therein). The quasistationary heat fluxes to the main
wall are thought to be dominated by convective


626
Table 1

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

Major issues associated with operation of ITER PFCs

PFCs

Divertor –
strike-point
regions

Plasma loads


Candidate
armor

Effects

 Radiation and particle heat
 Large particle fluxes
 Disruptions
 ELM’s
 Slow-high power

CFCa

Chemical erosion evaporation
brittle destruction and
tritium codeposition

 Radiation heat
 Disruptions
 Radiation heat (MARFE’sb)
 $100 eV ions and CX

 Erosion lifetime and
component replacement

 High tritium inventory
and safety

transients


Divertor –
baffle region
Dome

Issue

W
W

High sputtering evaporation/
melting
High sputtering evaporation/
melting

 Plasma contamination
 Erosion lifetime

neutrals

First wall

Start-up
limiters

 Moderate power transients
 Plasma contact during
VDEsc
 Disruptions and runaway
electrons

 ELMs
 High start-up heat loads
 Plasma contact during
VDEs
 Disruptions

Be

Evaporation/melting

 T retention in beryllium
codeposited layers

 Chemical reactivity
especially with Be dust
Be

High sputtering evaporation/
melting

 Erosion lifetime

a

W is also considered as an alternative.
Multifaceted asymmetric radiation from the edge (MARFE).
Vertical displacement event (VDE).

b
c


transport,11 but still remain to be clarified. Although
the steady-state parallel power fluxes associated with
these particle fluxes will only be of the order of
several MW mÀ2 in the ITER QDT ¼ 10 reference
scenario, local overheating of exposed edges of main
wall PFCs can occur because of limitations in the
achievable alignment tolerances. Similarly, transient
events are expected to cause significant power fluxes
to reach first-wall panels in ITER along the field lines.
Edge localized modes (ELMs) deposit large amounts
of energy in a short time, and in some cases in a
toroidally localized fashion, which can lead to strong
excursions in PFC surface temperatures. While the
majority of ELM energy is deposited on divertor
surfaces, a significant fraction is carried to surfaces
outside the divertor. There are obvious concerns that
ELMs will lead to damage of the divertor and the first
wall.12 An additional concern is that even without
erosion, thermal shock can lead to degradation of
material thermomechanical properties, for example,
loss of ductility leading to an enhanced probability of
mechanical failure or spalling (erosion). Research
efforts to characterize the ELMs in the SOL are
described elsewhere.13–15 There are still large uncertainties in predicting the thermal loads of ELMs on
the ITER beryllium first wall and the range of parallel

energy fluxes varies from 1.0 MJ mÀ2 (controlled
ELMs) to 20 MJ mÀ2 (uncontrolled ELMs).16,17 Even
for controlled ELMs, such energy fluxes are likely to

cause melting of up to several tens of micrometers of
beryllium at the exposed edges,18 which could cause
undesirable impurity influxes at every ELM.10,11
4.19.2.2 Brief History of Plasma-Facing
Materials in Fusion Devices
PWIs have been recognized to be a key issue in
the realization of practical fusion power since the
beginning of magnetic fusion research. By the time
of the first tokamaks in the 1960s in the USSR and
subsequently elsewhere, means of reducing the level of
carbon and oxygen were being employed.19,20 These
included the use of stainless steel vacuum vessels and
all-metal seals, vessel baking, and discharge cleaning.
Ultimately, these improvements, along with improved
plasma confinement, led to the first production of
relatively hot and dense plasmas in the T3 tokamak
($1 keVand $3 Â 1019 mÀ3).21,22 These plasmas, while
being cleaner and with low-Z elements fully stripped in
the core, still had unacceptable levels of carbon, oxygen, and metallic impurities. The metallic contamination inevitably consisted of wall and limiter materials.


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

Early in magnetic fusion research, it was recognized that localizing intense PWIs at some type of
‘sacrificial’ structure was desirable, if only to ensure
that more fragile vacuum walls were not penetrated.
This led to the birth of the ‘limiter,’ usually made to
be very robust, from refractory material and positioned to ensure at least several centimeters gap
between the plasma edge and more delicate structures like bellows, electrical breaks, vacuum walls,
etc. Typical materials used for limiters in these

early days included stainless steel in Adiabatic
Toroidal Compressor (ATC)23 and ISX-A24 and
many others, molybdenum in Alcator A25 and Torus
Fontenay-aux-Roses (TFR),26 tungsten in symmetric
tokamak (ST)27 and Princeton Large Torus (PLT),28
and titanium in poloidal divertor experiment (PDX).29
Poloidal divertors have been very successful at localizing the interactions of plasma ions with the target
plate material in a part of the machine geometrically
distant from the main plasma where any impurities
released are well screened from the main plasma and
return to the target plate.30 By the early 1980s, it was
also recognized that in addition to these functions, the
divertor should make it easier to reduce the plasma
temperature immediately adjacent to the ‘limiting’ surface, thus reducing the energies of incident ions and the
physical sputtering rate. Complementary to this, high
divertor plasma and neutral densities were found. The
high plasma density has several beneficial effects in
dispersing the incident power, while the high neutral
density makes for efficient pumping. Pumping helps
with plasma density control, divertor retention of
impurities and, ultimately, in a reactor, helium exhaust.
By the late 1970s, various tokamaks were starting to
employ auxiliary heating systems, primarily neutral
beam injection (NBI). Experiments with NBI on PLT
resulted in the first thermonuclear class temperatures
to be achieved.28,31,32 PLT at the time used tungsten
limiters, and at high powers and relatively low plasma
densities, very high edge plasma temperatures and
power fluxes were achieved. This resulted in tungsten
sputtering and subsequent core radiation from partially

stripped tungsten ions. For this reason, PLT switched
limiter material to nuclear grade graphite. Graphite
has the advantage that eroded carbon atoms are fully
stripped in the plasma core, thus reducing core radiation. In addition, the surface does not melt if overheated – it simply sublimes. This move to carbon by
PLT turned out to be very successful, alleviating the
central radiation problem. For these reasons, carbon
has tended to be the favored limiter/divertor material
in magnetic fusion research ever since.

627

By the mid-1980s, many tokamaks were operating
with graphite limiters and/or divertor plates. In
addition, extensive laboratory tests and simulations
on graphite had begun, primarily aimed at understanding the chemical reactivity of graphite with
hydrogenic plasmas, that is, chemical erosion. Early
laboratory results suggested that carbon would be
eroded by hydrogenic ions with a chemical erosion
yield of Y $ 0.1 C/Dþ, a yield several times higher
than the maximum physical sputtering yield. Another
process, radiation-enhanced sublimation (RES), was
discovered at elevated temperatures, which further
suggested high erosion rates for carbon. Carbon’s ability to trap hydrogenic species in codeposited layers
was recognized. These problems, along with graphite’s
poor mechanical properties in a neutron environment
(which had previously been known for many years
from fission research33), led to the consideration of
beryllium as a plasma-facing material. This was primarily promoted at JET.34 A description of the operation experience to date with Be in tokamak devices is
provided in Section 4.19.2.3.
At present, among divertor tokamaks, carbon is the

dominant material only in DIII-D. Alcator C-Mod
at Massachusetts Institute of Technology (MIT),
USA35 uses molybdenum. ASDEX-Upgrade (Axially
Symmetric Divertor Experiment) is fully clad with
tungsten,36 and JET has completed in 2011 a large
enhancement programme37 that includes the installation of a beryllium wall and a tungsten divertor. New
superconducting tokamaks, such as Korea Superconducting Tokamak Advanced Reactor (KSTAR) in
Korea38 and experimental advanced superconducting
tokamak (EAST) in China,39 employ carbon as material
for the in-vessel components, but with provisions to
exchange the material later on in operation.
The current selection of plasma-facing materials
in ITER has been made by compromising among
a series of physics and operational requirements,
(1) minimum effect of impurity contamination on
plasma performance and operation, (2) maximum
operational flexibility at the start of operation, and
(3) minimum fuel retention for operation in the DT
phase. This compromise is met by a choice of three
plasma-facing materials at the beginning of operations (Be, C, and W). It is planned to reduce the
choices to two (Be and W) before DT operations
in order to avoid long-term tritium retention in carbon codeposits during the burning plasma phase.
Beryllium has been chosen for the first-wall PFCs
to minimize fuel dilution caused by impurities
released from these surfaces, which are expected to


628

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices


have the largest contamination efficiency.40–44 Moreover, the consequences of beryllium contamination
on fusion performance and plasma operations are
relatively mild. This has been demonstrated by
experiments in tokamaks (see Section 4.19.2.3).
The main issues related to the use of beryllium in
ITER are (1) the possible damage (melting) during
transients such as ELMs, disruptions, and runaway
electron impact, and its implications for operations
and (2) the codeposition of tritium with beryllium
which is eroded from the first wall and deposited at
the divertor targets (and possibly also locally redeposited into shadowed areas of the shaped ITER first
wall). Both issues are part of ongoing research, the
initial results of which are being taken into account in
the ITER design so that the influence of these two
factors on ITER operation and mission is minimized.
This includes ELM control systems based on pellets
and resonance magnetic perturbation (RMP) coils,
disruption mitigation systems, and increased temperature baking of the divertor to release tritium from
the beryllium codeposited layers. Carbon is selected
for the high power flux area of the divertor strike
points because of its compatibility with operation
over a large range of plasma conditions and the
absence of melting under transient loads. Both of
these characteristics are considered to be essential
during the initial phase of ITER exploitation in
which plasma operational scenarios will require
development and transient load control and mitigation systems will need to be demonstrated.
4.19.2.3 Experience with Beryllium in
Tokamaks

Only three tokamaks have operated with beryllium as
the limiter or first-wall material. The first experiments were performed by UNITOR,45 which were
then followed by ISX-B.46 Both tokamaks investigated the effects of small beryllium limiters on
plasma behavior (UNITOR had side limiters at two
toroidal locations and ISX-B had one top limiter) in
support of the more ambitious beryllium experiment
in JET (see below). The motivation to use beryllium
came from the problem of controlling the plasma
density and impurities when graphite was used.
Both UNITOR and ISX-B showed that once
beryllium is evaporated from the limiter and coated
over a large segment of the first wall, oxygen gettering
leads to significant reduction of impurities. When the
heat load on the beryllium limiter was increased to
the point of evaporating beryllium, the oxygen

concentration was decreased dramatically. Although
the concentration of beryllium in the plasma was
increased, its contribution to Zeff (the ion effective
charge of the plasma Zeff provides a measure for
impurity concentration) was more than compensated
by the reduction of oxygen, carbon, and metal impurities.45 The plasma Zeff was observed to be reduced
from 2.4 to near unity with beryllium. It must be noted
that there was a negative aspect associated with beryllium operation during the ISX-B campaign. The
reduction in plasma impurities was not observed
until the limiter surface was partially melted causing
beryllium to be evaporated and coated on the first
wall. Once melting did occur, the droplets made
subsequent evaporation more likely but hard to control. The consequent strong reduction in plasma
impurities associated with gettering then made discharge reproducibility hard to obtain. However, if a

much larger plasma contact area is already covered
with Be, one does not need to rely on limiter melting
to obtain the beneficial effect of beryllium. This effect
could be achieved by using large area beryllium limiter, or coating the inside wall with beryllium which
was the approach taken by JET when it introduced
beryllium in 1989.
Large tokamak devices such as JET had found
it very difficult to control the plasma density
with graphite walls as the discharge pulse length got
longer. Motivated by the frequent occurrence of
a phenomenon that plagued the earlier campaigns –
the so-called carbon blooms due to the overheating of
poorly designed divertor tiles and the subsequent
influx of carbon impurities in the plasma due
to evaporation – JET decided to use beryllium as a
plasma-facing material.
Thin evaporated beryllium layers on the graphite
walls were used initially ($100 A˚ average thickness
per deposition) on the plasma-facing surface of the
device. Subsequently, beryllium tiles were installed
on the toroidal belt limiter.
The main experimental results with beryllium can
be summarized as follows:
1. The concentration of carbon and oxygen in the
plasma were 4–7% and 0.5–2%, respectively,
when graphite was used as belt limiter. With a
beryllium belt limiter, the carbon content was
reduced to 0.5% and oxygen became negligible,
because of oxygen gettering by beryllium. During
ohmically heated discharges, the concentration

of beryllium remained negligible even though
beryllium was the dominant impurity.


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

2. While the value of Zeff was $3 using the graphite
limiter and auxiliary heating power of 10 MW, Zeff
was $1.5 even with additional heating powers of
up to 30 MW with a beryllium limiter.
3. The fuel density control was greatly improved
with the beryllium limiter and beryllium evaporated wall. Gas puffing to maintain a given plasma
density was typically 10 times larger when using
beryllium than graphite.
Following the beryllium limiter experience, divertor beryllium targets were installed in JET for two
configurations. An extensive set of experiments with
toroidally continuous X-point divertor plates was carried out in JET in the period 1990–1996 to characterize
beryllium from the point of view of its thermomechanical performance, as well as its compatibility with
various plasma operation regimes.47–50
In the JET Mk I experiments, melting of the
beryllium tiles was reached by increasing (in a progressive way) the power flux to a restricted area of the
divertor target in fuelled, medium density ELMy
H-mode discharges (Pinp $ 12 MW). Large beryllium
influxes were observed when the divertor target temperature reached $1300  C. In these conditions, it
became difficult to run low-density ELMy H-mode
discharges (Pinp $ 12 MW) without fast strike point
movement (to achieve lower average power load) and
the discharges either had very poor performance
or were disrupted. However, no substantial plasma
performance degradation was observed for medium

density H-modes with fixed strike point position,
or if fast strike point movement was applied in lowdensity H-modes, despite the large scale distortion of
the target surface caused by the melt layer displacement and splashing due to the previous $25 high
power discharges48,51 (see Figure 252). This demonstrated that it was possible to use the damaged Be
divertor target as the main power handling PFC if the

Figure 2 Melting of the Joint European Torus Mk
I beryllium target plate tiles after plasma operation. Image
courtesy of EFDA-JET.

629

average power load was decreased, either by increasing plasma density and radiative losses, or by strike
point sweeping. The damage did not prohibit
subsequent plasma operation in JET, but would seriously limit the lifetime of Be PFCs in long-pulse
ITER-like devices.
The latest results of the operation of JET
with beryllium have been reviewed recently by
Loarte et al.10

4.19.3 Beryllium PWI Relevant
Properties
This section describes the present understanding
of PWIs for beryllium-containing surfaces. First,
it focuses on the erosion properties of ‘clean’ beryllium surfaces at different temperatures. Retention of
plasma fuel species in both bulk and codeposited
layers of beryllium is then described. As beryllium
will not be used as the exclusive plasma-facing material in future confinement devices, issues associated
with mixed, beryllium-containing surfaces are also
addressed.

4.19.3.1

Beryllium Erosion Properties

The term erosion is used to describe a group of
processes that remove material from a surface subjected to energetic particle bombardment. Included
under the general classification of erosion are processes such as physical sputtering, chemically assisted
physical sputtering, chemical sputtering, and thermally
activated release from surfaces. Of these processes,
only chemical sputtering, where volatile molecular
species are formed on the surface, appears to be
inactive in beryllium.
4.19.3.1.1 Physical sputtering of beryllium

Physical sputtering results from the elastic transfer
of energy from incoming projectiles to atoms on the
surface of the target material. Target atoms can be
sputtered when the energy they receive from the
collisional cascade of the projectile exceeds the binding energy of the atom to the surface. The physical
sputtering rate is usually referred to as the sputtering
yield, Y, which is defined as the ratio of the number of
atoms lost from a surface to the number of incident
energetic particles striking the surface. The lower
the binding energy of surface atoms, the larger the
physical sputtering yield. As physical sputtering can
be approximated using a series of binary collisions
within the surface, it is relatively easy to estimate


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices


the physical sputtering yield of given projectile-target
scenarios. Monte-Carlo based simulation codes (such
as transport of ions in matter (TRIM))53 have been
used to generate extensive databases of sputtering
yields based on incident particle angle, energy, and
mass, for a variety of targets54 including beryllium.
Measurement of the physical sputtering yield from
a beryllium surface is complicated by the natural
affinity of beryllium for oxygen. A beryllium surface
will quickly form a thin, stable, passivating oxide
surface layer when exposed to atmosphere. In ion
beam devices, it is possible to clean any oxides from
the beryllium surface before a measurement and with
careful control of the residual gas pressure, make
the measurements before the oxide reforms on the
surface and alters the measurement.55 It has also been
shown that it is possible to deplete the beryllium
surface of oxide by heating the sample to temperatures exceeding 500  C, where the beryllium below
the oxide can diffuse through the oxide to the
surface,56 thereby allowing measurements on a clean
beryllium surface. The comparison between the calculated sputtering yield and measurements made using
mass-selected, monoenergetic ion-beams devices
impinging on clean beryllium surfaces is fairly good.57
Measurements of sputtering yields in plasma
devices, however, are complicated by several factors.
In plasma devices, the incident ions usually have a
temperature distribution and may contain different
charge state ions. Each different charge state ion
will be accelerated to a different energy by the electrostatic sheath in the vicinity of the surface. When

hydrogenic plasma interacts with a surface, one must
also account for a distribution of molecular ions
striking the surface. In the case of a deuterium
plasma, for example, the distribution of molecular
þ
ions (Dþ, Dþ
2 , D3 ) must be taken into account
as the incident molecule disassociates on impact
with the surface and a Dþ
2 ion becomes equivalent
to the bombardment of two deuterium particles with
one-half the incident energy of the original Dþ
2 ion.
Figure 3 shows the change to the calculated sputtering yield when one includes the effects of molecular
ions in a plasma, compared to the calculated sputtering yield from pure Dþ ion bombardment.
The trajectory of the incoming ions can also be
altered by the presence of electrostatic and magnetic
sheaths. Plasmas also contain varying amounts of
impurity ions, originating either from PWIs in other
locations of the device, or ionization of residual background gas present in the device and these impurity
ions, or simply neutral gas atoms, may interact with

the surface. Finally, the incident flux from the plasma
is usually so large that the surface being investigated,
and its morphology, becomes altered by the incident
flux and a new surface exhibiting unique characteristics may result.
Nevertheless, the physical sputtering yield from
beryllium surfaces exposed to plasma ion bombardment has been measured in several devices. Unfortunately, there is little consensus on the correct value of
the physical sputtering yield. In JET, the largest
confinement device to ever employ beryllium as a

PFC sputtering yield measurements range from
values far exceeding47 to values less58 than one would
expect from the predictions of TRIM. In the Plasma
Interaction with Surface Components Experimental
Station B (PISCES-B) device, systematic experiments
to measure the physical sputtering yield routinely
show values less59–61 than those expected from
TRIM. This difference is shown in Figure 3, where
the energy dependence of the calculated yield is compared to experimental measurements.
Another primary difference between the conditions in an ion beam device and those encountered
in a plasma device has to do with the neutral density
near the surface being investigated. In an ion beam
experiment, the background pressure is kept very low
0.1

Physical sputtering yield

630

0.01

0.001

0.0001

10-5
0

20


40

60 80 100 120 140 160
Ion energy (eV)

Measured yield in D plasma
Calculated yield (D+ ions only)
Calculated yield (D+, D+2 , D+3 ions)
Figure 3 Calculated sputtering yields from pure Dþ
bombardment at normal incidence compared to that
þ
calculated for a (0.25, 0.47, 0.28) mix of Dþ, Dþ
2 , and D3 ; also
shown is the measured yield from such a plasma.


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

so that the surface being probed maintains its clean
properties. On the other hand, the incident flux in a
plasma device is usually several orders of magnitude
larger than in an ion beam device, ensuring that the
surface remains clean because of the large incident
flux. However, this plasma-facing surface undergoes
not only energetic ion bombardment, but also bombardment by neutral atoms and molecules.
The neutral density in plasma generators is typically on the order of 1020 mÀ3 (a few millitorr) which
is necessary for breakdown of the plasma. The estimated neutral atom flux is approximately equal to the
incident ion flux to the surface61 and it is often not
possible to alter significantly this flux ratio. In the
case of a beryllium surface which can form a hydride

(see Section 4.19.3.1.3), the presence of adsorbed
deuterium on the surface could affect the measured
sputtering yield by decreasing the beryllium concentration at the surface and altering the binding energy
of surface beryllium atoms.
Some evidence of this effect may be discerned
in data from JET measurements of the beryllium
sputtering yield. Two sets of sputtering yield measurements have been reported from JET; one from
beryllium divertor plate measurements and the other
from beryllium limiter measurements. In the divertor
region, one expects a neutral density similar to that
encountered in plasma generators (1020 mÀ3 or more)
and the measured sputtering yield is lower than that
predicted by TRIM calculations.58 When sputtering
measurements are made on the limiter, where the
neutral density is typically lower, the sputtering
yield agrees with, or exceeds, the calculated value.47
Of course, other issues such as impurity layers on the
divertor plate and angle of incidence questions tend to
confuse the results. However, the data sets from JET
are consistent with the impact of neutral absorption
on the beryllium plasma-facing surface.
Effects associated with plasma operation will need
to be taken into account when predicting sputtering
yields from different areas of confinement devices.
In addition to the low-energy neutral atom flux and
higher-energy charge exchange neutral flux, the
impact of small impurity concentrations in the incident plasma flux will also have a large impact on the
expected sputtering yield. Some of the implications
of the formation of a mixed-material surface are
discussed in the next section and in Section 4.19.3.3.

4.19.3.1.2 Mixed-material erosion

As was pointed out in the previous section, it is important to have accurate knowledge of a target’s surface

631

composition to predict its erosion rate. A small impurity concentration contained within the incident
plasma can drastically alter the surface composition
of a target subjected to bombardment by the impure
plasma. Oxygen impurities in the plasma, either from
ionization of the residual gas, or due to erosion from
some other surface, will readily lead to the formation of
beryllium oxide on the surface of a beryllium target.
Depending on the arrival rate of oxygen to the surface
compared to the erosion rate of oxygen off the surface,
one can end up measuring the sputtering rate of a clean
beryllium surface or a beryllium oxide surface. Careful
control of the residual gas pressure in ion beam sputtering experiments55 has documented this effect.
Unfortunately, it is not always so easy to control the
impurity content of an incident plasma.
In the case of a magnetic confinement device
composed of groups of different plasma-facing material surfaces, erosion from a surface in one location of
the device can result in the transport of impurities
to other surfaces throughout the device. Mixedmaterial surfaces are the result. To first order, a
mixed-material surface will affect the sputtering of
the original surface material in two ways. The first is
rather straightforward, and is true even for materials
which do not form chemical bonds, in that the surface
concentration of the original material is reduced
thereby reducing its sputtering rate. The second effect

changing the sputtering from the surface results from
changes in the chemical bonding on the surface, which
can either increase, or decrease the binding energy
of the original material. If the chemical bonds increase
the binding energy, the sputtering rate will decrease.
If the bonding acts to reduce the surface binding
energy, the sputtering rate will increase (assuming the
change in surface concentration does not dominate this
effect). A recent review of mixed materials62 provides
some background information on the fundamental
aspects of general mixed-material behavior.
If a plasma incident on a beryllium target contains
sufficient condensable, nonrecycling impurities (such
as carbon), it will affect the sputtering rate of the
beryllium. This effect was first referred to as ‘carbon
poisoning.’ 5,9,63 A simple particle balance model
has been used to adequately explain the results for
formation of mixed carbon-containing layers on
beryllium at low surface temperature.64 However, as
the target temperature increases, additional chemical
effects, such as carbide formation, have to be
included in the model.
An interesting change occurs when the bombarding species is a mixture of carbon and oxygen.


632

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

Measurements of the chemical composition of a

beryllium surface bombarded with a COþ ion beam
showed almost exclusive bonding of the oxygen
to the beryllium in the implantation zone.65,66
The formation of BeO on the surface left the carbon
atoms easily chemically eroded. The amount of oxygen present in the incident particle flux plays a strong
role in the final chemical state of the surface atoms
and their erosion behavior.
The inverse experiment, beryllium-containing
plasma incident on a carbon surface, has also been
investigated.67–69 In the case of beryllium impurities
in the plasma, a much more accurate measurement of
the impurity concentration was possible. Contrary
to the carbon in beryllium experiments, a simple
particle balance model could not account for the
amount of beryllium remaining on the surface after
the plasma exposure. Clearly, the inclusion of chemical effects on the surface needs to be taken into
account to interpret the results.
Beryllium carbide (Be2C) was observed to form on
the surface of carbon samples exposed to berylliumcontaining deuterium plasma even during bombardment at low surface temperature. Carbide formation
will also act to increase the binding of beryllium
atoms to the surface and decrease the binding of
carbon atoms. This effect will result in an increase
in the concentration of beryllium on the surface
compared to a simple particle balance equation and
must be included to understand the evolution of the
surface. In addition, the formation of the carbide was
correlated with the decrease of carbon chemical erosion70 (see Section 4.19.3.3.1 for more discussion of
the chemical erosion of the beryllium–carbon system).
4.19.3.1.3 Chemically assisted sputtering
of beryllium


The term chemically assisted physical sputtering
refers to the transfer of energy from an incident
particle to a molecule on the surface. The energy
gained is sufficient to break any remaining bonds of
the molecule to other atoms on the surface resulting
in the release of the molecule, or a fragment of the
molecule, from the surface. In the case of beryllium
bombarded by deuterium plasma, the sputtering of
beryllium deuteride was first recorded in JET71 during operation with a beryllium divertor plate. Since
that time, a series of systematic investigations were
performed in PISCES-B to quantify the magnitude of
this erosion term.72,73
The results from PISCES-B show a surface temperature dependence of the sputtering rate72 of BeD

molecules. The maximum in the BeD sputtering rate
(at $175  C) corresponds with the onset of thermal
decomposition of BeD2 molecules73 from a standardized sample of BeD2 powder. Even at the maximum
loss rate, the chemical sputtering remains smaller
than the physical sputtering rate of beryllium atoms
from the surface over the incident energy range
examined (50–100 eV). Molecular dynamics simulations have predicted,74 and subsequent measurements have validated the prediction, that chemical
sputtering can dominate physical sputtering of beryllium as the incident deuterium ion energy decreases
below 50 eV.
A distinction should be made between chemical
sputtering and chemically assisted physical sputtering.
Chemical sputtering involves the formation and loss of
volatile molecules from a surface. In the case of beryllium deuteride, the molecule decomposes into a deuterium molecule and a beryllium atom before it
becomes volatile, so at least to date there is no evidence
for chemical sputtering of beryllium during deuterium

particle bombardment. Documentation of the chemically assisted physical sputtering of beryllium may be
important for determining material migration patterns
in confinement devices and the identification of beryllium deuteride molecular formation in plasmaexposed surfaces may also help explain the hydrogenic
retention properties of beryllium.
4.19.3.1.4 Enhanced erosion at elevated
temperatures

In addition to the temperature-dependent chemical
sputtering of beryllium when exposed to deuterium
plasma, another temperature-dependent loss term is
present in beryllium exposed to plasma bombardment at elevated temperature. The classical picture
of the temperature dependence of erosion from
chemically inert surfaces exposed to energetic particle bombardment is composed of the superposition
of a constant physical sputtering yield with an exponentially varying thermal sublimation curve. The
classical picture is contradicted, however, by experiments that show an exponential increase in erosion
at lower temperature that cannot be explained by
classical thermodynamic sublimation. First observed
by Nelson75 for a variety of metal surfaces, similar
results have been measured for Be,76,77 W,78 and
C79,80 surfaces. In the case of carbon, this mechanism
has been called RES.
In the case of beryllium, two explanations have
been proposed and both rely on the large flux of ions
incident during plasma bombardment to modify the


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

4.19.3.2 Hydrogen Retention and
Release Characteristics

4.19.3.2.1 Implantation

The use of beryllium as a plasma-facing material in
tokamaks has prompted many experimental studies
of retention and emission of hydrogen implanted into
beryllium-like metals from ion-beams or plasmas.
References and discussions of these studies can be
found in reviews.82–85 Here, we review those studies
which are relevant to H retention in Be in a fusion
plasma environment. This section is mainly excerpted
from Federici et al.7 Two basic parameters for understanding H retention are the hydrogen diffusivity and

Solubility (H/metal atom Pa-1/2)

solubility. Studies of solubility and diffusivity are
reviewed in Causey and Venhaus85 and Serra et al.86
Figures 487–90 and 587,91,92 show experimental values
for hydrogen solubility and diffusivity in W and Be.
For Be there are significant differences between
results from various studies. These differences may
be due to effects of traps and surface oxide layers.
The presence of bulk traps tends to increase the
measured values of solubility and to decrease the measured values of diffusivity (see Federici et al.7),
especially under conditions where the concentration

10-7
Be

1


2
10-8
3

W

10-9

0.5

1.5

1.0
1000/T(K)

Figure 4 Measured solubility of hydrogen in tungsten
(dashed line87) and beryllium (solid lines 1,88 2,89 and 390).
Reproduced with permission from Federici, G.; Skinner,
C. H.; Brooks, J. N.; et al. Plasma-material interactions in
current tokamaks and their implications for next-step
fusion reactors. Nucl. Fusion 2001, 41, 1967–2137
(review special issue), with permission from IAEA.

1´10-7
1´10-8

Diffusivity (m2 s−1)

plasma-facing material surface. In the first, the incident plasma ions, in addition to creating sputtered
atoms from the surface, also create a population of

surface adatoms. An adatom is an atom from a lattice
site on the surface that has gained sufficient energy to
leave its lattice location, yet does not have sufficient
energy to escape from the surface as a sputtered
atom. The atom then occupies a site on top of the
regular lattice sites. Because an adatom does not have
the same number of adjacent atoms as those in the
lattice, it is less strongly bound to the surface and can
therefore sublime at a lower temperature than one
associates with equilibrium thermodynamic sublimation. In the second explanation, incident plasma ions
that have thermalized somewhere below the surface
of the lattice exert a stress on the surface atoms of the
target again resulting in a lower binding energy of
the surface atoms to the bulk of the material.
Measurements show atoms are being lost from the
surface at thermal energies,77 rather than the energy
associated with sputtered particles (i.e., on the order
of electron volts). This seems to verify the loss mechanism that occurs because of the thermal release of an
ensemble of particles with a lower surface binding
energy than that of bulk atoms of that element. Additional measurements at elevated temperature have
documented the variation in Be atom surface loss
rate with changes of the incident flux of energetic
particles.81 The larger the incident flux, the lower
the onset temperature for the enhanced erosion.
The implication of this enhanced loss rate at elevated temperature is a reduction of the permissible
operating temperature of any plasma-facing material,
or alternatively that the lifetime of a component
operating at extreme temperature may be less than
that expected based on the predictions from classical
surface loss terms.


633

W

1´10-9
1´10-10

1a

10-11

2

Be

1b

10-12
3
10-13

0.5

1.0
1000/T(K)

1.5

Figure 5 Measured diffusivity of hydrogen in tungsten

(dashed line87) and beryllium (solid lines 1a&b,91 2,89
and 392). Reproduced with permission from Federici, G.;
Skinner, C. H.; Brooks, J. N.; et al. Plasma-material
interactions in current tokamaks and their implications for
next-step fusion reactors. Nucl. Fusion 2001, 41, 1967–2137
(review special issue), with permission from IAEA.


634

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

of hydrogen in solution is smaller than the concentration of traps. For this reason, studies done on
materials of higher purity and crystalline perfection,
and at higher temperatures and with higher concentrations of hydrogen in solution, tend to give more
reliable results. The porosity and oxide inclusions
present in beryllium produced by powder metallurgy
are also likely to lead to inconsistent results in measurements of hydrogen solubility and diffusivity.
In the Be experiments, the effects of traps were not
characterized and may be dominant. One firm conclusion is that the solubility of hydrogen is very low
in both Be and W.
Many studies have been done on the retention and
emission of H implanted into materials to provide
data needed to predict H retention in fusion reactor
environments. Figure 6 shows the retention of 1 keV
deuterium implanted into Be at 300 K versus incident fluence, measured by thermal desorption.93
D retention in Be was close to 100% at low fluences
but saturated at high fluences. Earlier nuclear reaction analysis (NRA) measurements of D retained in
Be within $1 mm of the surface gave very similar
results.94 This saturation behavior indicates that

D implanted into Be at 300 K does not diffuse, but
accumulates until it reaches a limiting concentration
of $0.3–0.4 D/Be within the implantation zone.
At high fluences, the implanted zone becomes porous
allowing additional implanted D to escape. This

saturation mechanism is confirmed by electron microscopy, which shows bubbles and porosity in the implantation zone after high fluence H implantation.95
Saturation of retention by the same mechanism is
observed for D implanted into stainless steel at
150 K where the D is not mobile.96 H retention in Be
increases with increasing ion energy and decreases
with increasing sample temperature.84,97 The retention
of 1 keV deuterium implanted into W and Mo at
300 K98 is also shown for comparison in Figure 6.
Figure 784 shows retention of deuterium and tritium as a function of incident particle fluence from a set
of high fluence experiments in which Be specimens
were exposed to laboratory ion-beams (Idaho National
Engineering and Environmental Laboratory (INEEL),
University of Toronto Institute for Aerospace Studies
(UTIAS)), linear plasma devices (Sandia National
Laboratory (SNL)/Los Alamos National LaboratoryTritium plasma experiment (LANL-TPE), University
of California, San Diego-Plasma Interaction with
Surface Components Experimental Station B (UCSDPISCES-B)), a tokamak divertor plasma (DIIID-DIMES), and a neutral beam (NB-JET). In some
of these studies carbon deposition or formation of
carbide or oxide surface layers occurred, which is
likely to affect D retention. The figure shows the
D retention in Be observed under a wide range of
exposure conditions. The high fluence saturated concentration tends to be lower at higher temperatures.

1022


100% retention

Retained fluence (D m−2)

3
2

Be

1021

W

3
2

1020

Mo, slope = 0.345

3
2
19

10 20
10

3 keV D+3 (1020 D m–2 s)
2


21

10

2

Be, Mo, or W (300 K)
10

22

2

Incident fluence

1023 2

1024 2

1025

(D m−2)

Figure 6 Retention of 1 keV deuterium implanted into Be and W, at 300 K versus incident fluence, measured by thermal
desorption. Adapted and reproduced with permission from Haasz, A. A.; Davis, J. W. J. Nucl. Mater. 1997, 241–243,
1076–1081, Federici, G.; Skinner, C. H.; Brooks, J. N.; et al. Plasma-material interactions in current tokamaks and their
implications for next-step fusion reactors. Nucl. Fusion 2001, 41, 1967–2137 (review special issue), with permission from IAEA.



Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

635

1023
SNL/LANL-TPE

Retained quantity (D (T) m–2)

PISCES-B-1

1022

PISCES-B-2
PISCES-B-3
INEEL(200–400)
UTIAS(27)

1021

100

DIII-D(DiMES)(130)

250

100
NB-JET(120)

200

250
250

1020

200

200

200
500

500
500

40
200
300

300
500

500

500
540
500
700

1019 21

10

1022

1023
1024
1025
Particle fluence (D (T ) m–2)

1026

1027

Figure 7 Retention of deuterium and tritium in Be as a function of incident particle fluence. For purposes of comparing
results from different experiments using different ion energies, the data have been scaled to correspond to an equivalent
100 eV deuterium ion energy. Numerical values next to the symbols and in the legend are specimen exposure temperatures,
in degrees Celsius. Reproduced with permission from Anderl, R. A.; et al. J. Nucl. Mater. 1999, 273, 1–26.

It must be noted that this phenomenon is very
important because it implies that tritium inventories
and permeation due to implantation in beryllium
for ITER PFC applications should be significantly
lower than was previously estimated using classical
recombination-limited release at the plasma surface.
A first attempt to model this saturation by allowing
the recombination coefficient to become exponentially large as the mobile atom concentration near
the plasma-facing surface approaches a critical
value was made by Longhurst et al.99 For Be, calculations suggest that the critical concentration is related
to the yield strength using Sieverts’ law of solubility.
On the basis of the results of these calculations, it can

be concluded that the inventory of tritium in the
beryllium first wall of a device such as ITER, because
of implantation, diffusion, trapping, and neutroninduced transmutation, will be of the order of 100 g
rather than the kilogram quantities estimated previously,100,101 and most of that will result from neutroninduced transmutations in the Be itself and from
trapping in neutron-induced traps. Current predictions of tritium inventory in ITER are briefly discussed in Section 4.19.6.2.1.
Fusion neutrons will create vacancies and interstitials in plasma-facing materials. For metals at reactor
wall temperatures, these defects will be mobile and
will annihilate at sinks (e.g., surfaces or grain boundaries), recombine, or agglomerate into defect clusters.
Vacancy agglomeration may also lead to the formation

of voids. In beryllium, neutron-induced nuclear reactions produce helium and tritium, which may be
trapped at defects or precipitate as gas bubbles.
These defects, resulting from neutron irradiation,
will increase the retention of hydrogen, by increasing
the concentration of sites where diffusing hydrogen
can precipitate as gas or become trapped as atoms.
The effect of neutron irradiation on hydrogen retention in metals is complex, but, in principle, this can be
modeled, provided the material parameters are
known, such as hydrogen diffusivity, solubility, trap
binding energy, and defect microstructure produced
by the neutron irradiation. For many metals, most of
these parameters are known well enough to attempt
such modeling. For beryllium, however, uncertainties
in solubility, diffusivity, and trapping of hydrogen
make such modeling of hydrogen retention difficult.
The problem of production of helium and tritium
by nuclear transmutation in beryllium itself is discussed in Section 4.19.4.4.5.
4.19.3.2.2 Beryllium codeposition

As deuterium retention in plasma-exposed beryllium

targets saturates after a given ion fluence (see Section
4.19.3.2.1), it is apparent that retention in codeposits
will eventually be the dominant accumulation mechanism with respect to beryllium PFCs. This is primarily due to the fact that the thickness of a
codeposit will continue to grow linearly with time.
It is, therefore, critical to understand both the


636

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

1

1
Mayer et al.
Mayer et al.

0.1

0.1
Causey
et al.
TPE

0.01

Present
data
PISCES
400


O/Be

D/Be

Present
data
PISCES

Causey and
Walsh
TPE

Causey and
Walsh
TPE
0.01

600
800
Temperature (K)

400

800

Figure 8 Comparison of D/Be levels in beryllium codeposits with the O/Be levels in the same codeposits.
Reproduced with permission from Baldwin, M. J.; Schmid, K.; Doerner, R. P.; Wiltner, A.; Seraydarian, R.; Linsmeier, Ch.
J. Nucl. Mater. 2005, 337–339, 590–594.


retention amounts and the release behavior of hydrogen isotopes from beryllium codeposits. In this section, a ‘codeposit’ includes both the codeposition
(where a BeD or BeD2 molecule is deposited on a
surface) and co-implantation (where deposited layers
of beryllium are bombarded with energetic hydrogen
isotopes) processes.
Initial interpretation of studies of beryllium codeposits were made difficult by relatively high oxygen
impurity content within the codepositing surface.102,103
Subsequent measurements104 with lower oxygen content seemed to indicate that the oxygen level within
the codeposit was correlated to the level of hydrogen
isotope retention in the codeposit. The other variable
that was identified to impact the retention level in
these studies was the temperature of the codepositing
surface.
Measurements seriously questioning the importance of oxygen on the retention level in beryllium
codeposits were made by Baldwin et al.105 In this data
set, the oxygen content throughout the codeposit was
measured by depth profiled X-ray photoelectron
spectroscopy and the oxygen content did not correlate with the deuterium retention level (Figure 8),
although the temperature of the codepositing surface
was still a dominating term in determining the deuterium retention level. Later, more detailed measurements confirmed that the presence of a beryllium

oxide surface layer was not correlated with an
increase in retention in beryllium.106
A systematic study of beryllium codeposition followed,107 identifying three experimental parameters
that seemed to impact the retention level in a codeposit. Along with the surface temperature, the incident deuterium energy and the beryllium deposition
rate were determined to be influential scaling parameters. The previously reported data in the literature
was also evaluated using the derived scaling and
found to agree with the predictions of the retention
levels measured under the various experimental
conditions present in the different machines. Later

the derived scaling was revised108 to use the ratio of
the fluxes of the codepositing species, rather than
the deposition rate to permit more accurate extrapolation to conditions expected in the edge of confinement devices.
The ability to predict the level of tritium retention
in beryllium codeposits is an important aspect of
a safety program; however, developing techniques
to remove the trapped tritium from codeposits is
a more important issue. The deuterium release
behavior during thermal heating of beryllium codeposits has been investigated.109 The results show that
the maximum temperature achieved during a bakeout is the figure of merit for determining the amount
of deuterium release from beryllium. Increasing


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

the time spent at lower baking temperatures did
not increase the amount of deuterium released from
the beryllium codeposits. These results, along with
the retention level predictions, should make it possible to design baking systems for different areas of a
confinement device to control the accumulation rate
of tritium to a desired level.
4.19.3.3

Mixed-Material Effects

A recent review of mixed-material effects in ITER62
provides background information on mixed-material
formation mechanisms and plasma–surface interaction effects. Here, the focus is on berylliumcontaining mixed-material surfaces (i.e., Be/C and
Be/W) and the conditions when one might expect
these surfaces to dominate the observed plasma–

surface interactions. In addition to plasma interactions with mixed-material surfaces, which will be
discussed here, other aspects such as changes to thermal conductivity, material strength, and ductility, the
impact of impurities on material joints, etc., must also
be carefully evaluated.
4.19.3.3.1 Be–C phenomena

Beryllium and carbon have been observed to begin
thermally interdiffusing at a temperature of around
500  C,56 resulting in the formation of a beryllium
carbide layer. However, beryllium carbide has also
been observed to form during energetic carbon ion
bombardment of beryllium surfaces at room temperature.110 As mentioned in Section 4.19.3.1.2, the
change in the binding energy of the carbide molecule
affects the sputtering yield of both the beryllium and
carbon atoms. In addition, the formation of beryllium
carbide also has a dramatic effect on the chemical
erosion properties of a carbon surface bombarded
with energetic beryllium ions.67,68,111
The presence of beryllium carbide on the surface of
a carbon sample exposed to deuterium plasma has been
shown to correlate with the reduction of chemical
erosion of the carbon surface.70 The speculation for
the cause of this effect is that the carbide enhances the
recombination of deuterium in the surface, thereby
lessening the amount of deuterium available to interact
with carbon atoms on the surface. This is similar to the
impact of small amounts of boron carbide in a graphite
surface affecting chemical erosion.112 However, the
difference here is that instead of obtaining the carbide
through an expensive production technique, the carbide forms naturally as beryllium ions in the plasma

interact with the carbon surface.

637

A systematic study of the time necessary to suppress chemical erosion of a graphite surface due to
the interaction with beryllium-containing plasma has
been carried out.69 Increasing the surface temperature of the graphite was seen to have the biggest
impact on reducing the suppression time. Increasing
the beryllium content of the plasma also reduced the
suppression time in a nonlinear fashion. An increase
of the incident particle energy was observed to
increase the time necessary to suppress the chemical
erosion of the surface, presumably due to an increase
in the removal of the carbide-containing surface
layer. A subsequent study showed that applying
heat pulses to a graphite surface interacting with
beryllium-containing plasma, to simulate surface
heating due to intermittent events, acted to reduce
the time necessary for the carbide surface to form and
suppress the chemical erosion of the surface.113
4.19.3.3.2 Be–W alloying

The existence of tungsten beryllide alloys (i.e., Be2W,
Be12W, and Be22W) is an excellent example of the
importance of mixed-material surface formation in
plasma-facing components.114 Figure 9 shows the
tungsten–beryllium phase diagram. Each of the beryllides shown in the figure exhibits a lower melting
temperature than one would expect from a tungsten
plasma-facing surface. If plasma containing beryllium
impurities interacts with a tungsten surface, there is

a possibility of these lower melting temperature
beryllide alloys being formed.
In thermodynamic equilibrium, various beryllide
alloys of tungsten have been observed to form,115 and
their reaction rates have been measured,116 at temperatures in excess of 800  C. However, as was seen with
beryllium carbide forming during plasma bombardment at lower temperature than expected thermodynamically, the concern exists that tungsten beryllide
could form at temperatures below 800  C as well.
Well controlled laboratory measurements in vacuum117 and in plasma simulators118 have shown that
although thin, nanometer scale, Be2W layers form
at the interface between beryllium and tungsten
surfaces, their growth below 800  C is negligible. In
addition, above 800  C, rapid beryllium sublimation
from surfaces can act to limit the amount of beryllium available for reacting with tungsten and thereby
also limit the growth rate of the alloys. In the present
low wall temperature confinement devices, modeling
shows that the divertor strike point locations are the
only areas where significant beryllide growth might
be expected and in these regions there does not


638

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

Weight percent tungsten
90
95

0 4060 70 80


100

3500
3422 ºC

3000

L

<2250 ºC
1289 ºC

?

2000

2100 ± 50 ºC
-60

-95

W

<1750 ºC
?

?
1500

Be2W


Temperature ( ºC)

2500

?
(bBe)
(aBe)
Be12W

Be22W

1000

500
0
Be

10

20

50
60
70
30 40
Atomic percent tungsten

80


90

100
W

Figure 9 Phase diagram for the Be–W system. Reproduced with permission from Doerner, R. P.; Baldwin, M. J.;
Causey, R. A. J. Nucl. Mater. 2005, 342, 63–67.

appear to be enough beryllium deposition to raise
significant concerns.119 One caveat to these predictions would be the existence of intermittent events
that raise the temperature of surfaces where significant beryllium deposits are located, thereby possibly
allowing the optimized beryllide growth conditions.
Another concern with regard to thin Be2W surface
layers on plasma-exposed tungsten is the impact of
these layers on tritium retention. While a thin
Be2W surface layer is not likely to retain much tritium itself, the thin beryllide surface layer could alter
the recombination characteristics of the bulk material
and change the accumulation rate of tritium within
the device. To date, there is little or no data available
to address this issue.
While it appears likely that the most serious issues
of tungsten beryllide formation may be avoided in
present confinement devices, the issues associated
with these alloys highlight the uncertainties and importance of understanding and predicting mixed-material

formation in plasma environments. Mixed materials
often interact with plasma in much different ways
than their elemental components. In the case of the
beryllium–carbon system (Section 4.19.3.3.1), the
mixed material appears to offer the potential for beneficial effects, whereas in the case of the beryllium–

tungsten system, the mixed material appears likely to
be detrimental to the operation of the device. Each
mixed-material system must, therefore, be individually
evaluated to determine its potential impact on all
aspects of operating surfaces in contact with plasma.

4.19.4 Main Physical and Mechanical
Properties
4.19.4.1

General Considerations

A comprehensive, although not recent, review of
the science and technology of beryllium can be
found in Beryllium Science and Technology.120


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

Several reviews have been published recently related
to use of beryllium in tokamaks and the status of the
investigations of the Be properties for the fusion
application.3,121–126 Various production and processing methods of beryllium metal fabrication have
been reviewed in Dombrowski.127 The majority of
methods are based on powder metallurgy and include
powder preparation from cast product by grinding
(i.e., attrition milling, impact grinding, ball mill
grinding); further powder consolidation (i.e., by cold
pressing (CP), cold isostatic pressing (CIP), vacuum
hot pressing (VHP), hot isostatic pressing (HIP)); and

possible additional mechanical treatment (e.g., extrusion, rolling, forging). Beryllium protective armor
can also be produced by plasma spray (see Section
4.19.4.3) and vapor deposition.
Several proposals were made at the beginning of
the ITER Research Programme during the ITER
Engineering Design Phase to develop a fusion grade
beryllium with high ductility, high resistance to heat
flux, and high radiation resistance. However, it was
recognized that this development would require significant efforts and could not be supported only by
requests from the fusion community.
There are various beryllium grades, which have
been developed for different applications. These grades
differ by chemical composition (BeO content, impurities), by method of powder preparation, by method of
consolidation, etc. The nonexhaustive list of various
beryllium grades from the US and the Russian Federation is presented in ITER Materials Properties
Handbook (MPH).128 Grades with similar composition are under production in Kazakhstan and in China.
We briefly discuss below some of the most relevant
physical and mechanical properties of beryllium,
in relation to its application as armor for PFCs.
4.19.4.1.1 Physical properties

The physical properties of beryllium are summarized
in Table 2, which is taken from ITER MPH.128
These properties have been used for design and
performance assessments. In addition to its low
atomic number, beryllium has several excellent thermal properties that make it well-suited for heat
removal components. The thermal conductivity is
comparable with that of graphite or CFC at low and
high temperatures but, in contrast to C-based materials, is not significantly degraded as a result of
neutron-irradiation. The specific heat of beryllium

exceeds that of C-based materials typically by a
factor of 2 over the temperature range of interest
for operation. However, Be has poor refractory

Table 2

Physical properties of beryllium

Atomic number
Atomic weight
Crystal structure
Density (kg mÀ3)
Melting temperature ( C)
Thermal conductivity (W mÀ1  CÀ1)
Specific heat (J kgÀ1  CÀ1)
Latent heat of fusion (kJ kgÀ1)
Latent heat of vaporization (kJ kgÀ1)
Electrical resistivity (mO cm)
Thermal expansion coefficient
10À6  CÀ1
Emissivity

639

4
9.013
Hexagonal closepacked
1830–1850
1283–1287
$200 (RT)

$82 (800  C)
$1900
$1300
$3.66 104
$4.4 (RT)
$11.6 (RT)
$14.96 (400  C)
0.1–0.5* (at
300  C)

Source: ITER MPH, ITER Final Design Report 2001 (internal
project document distributed to the ITER participants).
RT, room temperature.
* Depending on quality of surface

properties, such as low melting temperature and
high vapor pressure. The high heat capacity and
good thermal conductivity of Be can be used to
maintain low surface temperatures in PFCs during
normal operation, but its low melting temperature
and high vapor pressure cause great design difficulties from the standpoint of survivability from offnormal events such as vertical displacement event
(VDE), ELMs, disruptions, and runaway electron
impact (see Section 4.19.6.2).
For the beryllium hexagonal close packed crystal
structure, the main physical properties, such as the
coefficient of thermal expansion, elastic modulus etc.
have some anisotropy. However, for the polycrystalline grades these properties could be, in the first
approximation, considered as isotropic. Some anisotropy is also typical for the highly deformed grades.
The physical properties (thermal conductivity, specific heat, elastic modulus, etc.) in first approximation
are the same for beryllium grades with similar BeO

and other impurity content and they are produced by
the same fabrication method.
4.19.4.1.2 Mechanical properties

Beryllium is known to be a brittle material, with a
typical elongation to failure in room temperature
tensile tests of roughly $0.8–6%. For material with
strong anisotropy (e.g., rolled plate or sheet), elongation in the rolling direction could be higher, but
in the transverse direction the elongation is


640

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

typically significantly lower than 1%. Recently, the
mechanical properties of beryllium have been
summarized in ITER MPH128 and ITER Materials
Assessment Report (MAR).129
The mechanical properties of beryllium depend
on the production method used and they are sensitive
to a variety of factors including BeO and impurity
content (which varies from less than 1% to 2–3%
for various grades), method of powder preparation
(impact grinding, attrition grinding), method of consolidation, and further treatments. The main problem
in using beryllium is its low ductility related to the
hexagonal-close-packed structure. There is limited
slip in directions not parallel to the basal planes,
resulting in very small ductility perpendicular to
the basal direction. Depending on the production

method, ductility of beryllium can be severely
anisotropic. The grain size is an important factor in
determining the ductility of various beryllium components. Much of the fine grain size present in the
starting powder is retained during hot pressing at
1060  C. Without an oxide network, grain growth
occurs at a much lower temperature, about 800  C.
Among various beryllium grades, it was found that
grade S-65C VHP (production of Brush Wellman,
US) has the highest guaranteed fracture elongation
at room temperature (minimum 3%; typical is more
than 4–5%). This grade is produced using impact
grinding powder and has a guaranteed BeO content
<1%. The level of impurities is also controlled.
The high ductility of the grade is one of the advantages of this material. Because of the VHP production
method, there is some anisotropy of properties in
relation to hot pressing direction, but the differences
are not significant.
As typical for all metals, the tensile properties of
beryllium depend on the testing temperature. As the
testing temperature increases, a decrease of the ultimate tensile and yield strength are observed. However, rupture elongation increases with increasing
test temperature and could reach a value higher by
40–50% for temperatures around $300–350  C (see
as example data for grade S-65C VHP in the ITER
MPH128). A further increase in the test temperature
leads to a decrease of the elongation. At temperatures
above 600  C, the ductility depends on the impurity
content, mainly aluminum, which tends to segregate
at grain boundaries, impairing the mechanical properties. By heat treatment in the temperature range
650–800  C, aluminum can be combined with other
elements, mainly iron and beryllium itself, to form a

stable beryllide as AlFeBe4. However, the stable

beryllide dissolves progressively when heated at temperatures >850  C. This last feature is important for
the selection of the joining technology for
manufacturing of the PFCs.
Further details on mechanical properties, such
as creep and fracture toughness, can be found elsewhere (see, e.g., ITER MAR129).
4.19.4.2 Selection of Beryllium Grades for
ITER Applications
For ITER PFC applications, various commercially
available beryllium grades from the United States
(Brush Wellman Inc.) and from the Russian Federation, listed in Table 3, were evaluated more than a
decade ago as potential candidates during the ITER
Engineering Design Activity (EDA).
The selection of the optimum grade for ITER
PFC applications is driven mainly by the requirements of ITER operation for structural integrity and
stability against various thermal loads, and in particular, the absence or minimization of macrodamage.
It is believed that ion-induced and thermal erosion
at elevated temperatures is very similar for various
grades of Be. However, performance under high heat
fluxes, especially under transient thermal loads such
as disruptions, VDE, and ELMs resulted in different
behavior and damage mechanisms. It is considered
that the ease of joining beryllium to copper alloys
(see Section 4.19.5) is not so sensitive to BeO content, impurity levels, and method of consolidation,
which are the parameters defining the grade of beryllium material.
It should be noted that for tokamak applications
(see Section 4.19.6) beryllium is used in the form of
tiles. Some surface cracking of the tiles could be
acceptable, if there is no macrodamage or delamination along the surface of tiles, which leads to the loss

of macroscopic particles.
The resistance to thermal fatigue is the most
important factor that affects the material selection
Table 3

Candidate grades of beryllium

Producer

Grades

Brush Wellman, US

S-65C VHP, S-65C HIP, S-65 CIP,
SR-200 VHP
S-200F HIP, S-200F VHP, I-400 VHP
DShG-200, TShG-56, TR-30,
TGP-56
TShGT, DIP-30, TShG-200

Russian Federation

VHP, vacuum hot pressing; HIP, hot isostatic pressing;
CIP, cold isostatic pressing.


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

because cracking could lead not only to enhanced
armor erosion, delamination, and loss of particles, but

also potentially to crack propagation to the heat sink
structure. Neutron irradiation resistance is another
factor to be taken into account because it may affect
the thermal performance and structural integrity.
Because of some of the uncertainties in the ITER
thermal loads, especially during transient events,
preference is given to beryllium grade(s) with potentially higher resistance to transient thermal loads.
The selection of the reference grades was made on
the basis of comprehensive assessment of the results
of various tests carried out during the ITER EDA.
The detailed analysis is presented in ITER MAR.129
Among the various studies, the following shall be
mentioned:

641

the molten zone, whereas for some grades the cracks
propagated to the bulk of the sample.
 Results of VDE simulation tests have been
reported in Linke et al.132,133 Severe melting of Be
was observed for energy densities of 60 MJ mÀ2
($1 s pulse duration); however, no cracks were
observed between molten and unmolten material
and in the bulk of unmolten parts for S-65C
VHP grade.
On the basis of the available data, Be S-65C VHP
(Brush Wellman, US) was selected as the reference
material on the basis of excellent thermal fatigue and
thermal shock behavior, and for the good available
database on materials properties, including neutron

irradiation effects. DShG-200 (produced in the Russian Federation) was proposed as a backup, but this
grade is no longer commercially available.
Recently, China and the Russian Federation, that
are two of the seven International Parties engaged in
the construction of ITER, have proposed the fabrication of additional first-wall grades as part of their
ITER contribution. The Russian Federation proposes
to use beryllium grade TGP-56-FW. This grade is
produced by VHP in almost the final form of the tiles
foreseen for the first wall. The recent results on
development of this grade have been reported in
Kupriyanov et al.134 China proposes instead to use a
grade called CN-G01135 that is produced from

 Screening low cycle fatigue test of 21 different
beryllium grades was performed in the past.130
It was shown that the grades with the best thermal
fatigue resistance are S-65C VHP, DShG-200,
TShG-56, and TShGT. Figure 10 shows the results
of the comparative low cyclic thermal fatigue study
of different grades of beryllium.
 Various grades of beryllium were also tested in
conditions simulating the disruption heat loads.131
The tests show that crack formation and behavior
after surface layer melting in different grades are
quite different. For Be S-65C, all cracks stopped in

2500
S-65C (L)

Grades with best fatigue

performance

DShG-200 (T)
Cycles to crack initiation

2000
S-65C (T)
S-65-H
TShGT(T) TShG-56 (T)

1500

S-200F (T)

98% S-65
1000
Extruded (T)
S-200F-H
SR-200

500

Be/60% BeO
0

0

94% S-65

S-200F (L)

TGP-56
Extruded (L)
I-400 (T)

Be/30% BeO
0.5
1
1.5
Side crack propagation depth (mm)

2

Figure 10 Results of low cycle thermal fatigue tests of different Be grades: number of cycles to crack initiation versus
crack propagation depth. Reproduced with permission from Watson, R.; et al. Low cyclic fatigue of beryllium. In Proceedings
of the 2nd IEA International Workshop on Beryllium Technology for Fusion, Wyoming, Sept 6–8, 1995; p 7.


642

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

impact ground powder (similar to powder used for
S-65C grade) by VHP. The grade is produced by
Ningxia Orient Non-Ferrous Metals Co. Ltd.
In order to accept these newly proposed beryllium
grades a specific qualification program is underway.
4.19.4.3 Considerations on
Plasma-Sprayed Beryllium
In the past, plasma spraying was considered as a
high deposition rate coating method, which could

offer the potential for in situ repair of eroded or
damaged Be surfaces. Development work was
launched during the early phase of the ITER
R&D Program in the mid-1990s.136 In the plasma
spray process, a powder of the material to be deposited is fed into a small arc-driven plasma jet, and the
resulting molten droplets are sprayed onto the target
surface. Upon impact, the droplets flow out and
quickly solidify to form the coating. With recent
process improvements, high quality beryllium coatings ranging up to more than 1 cm in thickness have
been successfully produced. Beryllium deposition
rates up to 450 g hÀ1 have been demonstrated with
98% of the theoretical density in the as-deposited
material. Several papers on the subject have been
published.136–138 A summary of the main achievements can be found in Table 4.
However, based on the results available, the initial
idea of using plasma-sprayed beryllium for in situ
(in tokamak) repair was abandoned for several reasons. First was the complexity of the process and
requirements to control a large number of parameters, which affect the quality of the plasma sprayed

Table 4
together)

coatings. Some of the most important parameters
include plasma spray parameters such as (1) power,
gas composition, gas flow-rate, nozzle geometry, feed,
and spray distance; (2) characteristics of the feedstock
materials, namely, particle size distribution, morphology, and flow characteristics; (3) deposit formation
dynamics, that is, wetting and spreading behavior,
cooling and solidification rates, heat transfer coefficient, and degree of undercooling; (4) substrate
conditions, where parameters such as roughness,

temperature and thermal conductivity, and cleanliness play a strong role; (5) microstructure and
properties of the deposit, namely, splat characteristics, grain morphology and texture, porosity, phase
distribution, adhesion/cohesion, and physical and
mechanical properties; and (6) process control, that
is, particle velocity, gas velocity, particle and gas
temperatures, and particle trajectories. Second,
plasma-sprayed beryllium needs (1) inert gas pressure, (2) reclamation of the oversprayed powder
(more than 10%), and (3) strict control of the substrate temperature. The higher the temperature the
higher the quality of the plasma-sprayed coating, but
unfortunately, an easy and reliable method to heat
the first wall to allow in situ deposition was not found.
Finally, tools to reliably measure the quality of the
coating and its thickness are not available today and
a strict control of the coating parameters is difficult
to achieve.
Thus, it was concluded that plasma-sprayed beryllium for in situ repair is too speculative for ITER
without further significant developments. Nevertheless, this method still remains attractive and could be
used for refurbishment of damaged components in

Main achievements of ITER-relevant plasma-sprayed technology (summary of best results, not always achieved

Parameter

Value/results

Comments

Residual porosity (%)
Thermal conductivity (W mKÀ1)


$2
Up to 160 at RT

Bond strength (MPa)
Substrate temperature ( C)

100–200
>450

Substrate preparation

Negative
transfer arc
4.5
>10
>90
5/680; 1/3000

Could be more than 5%
Depends on temperature of substrate, maximum achieved at
T $ 600–800  C with addition of H
Reasonable
Very important for good strength, adhesion, and thermal
conductivity. Keep in mind that CuCrZr temperature should not be
higher than 500  C for several hours due to overageing of CuCrZr
Needed, but very difficult to do in situ

Deposition rate (kg hÀ1)
Thickness (mm)
Deposition efficiency (%)

Thermal fatigue (MW mÀ2/
number of cycles)

Reasonable
Reasonable
It means that more than 10% of powder will be lost in chamber
For first-wall conditions tested


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

hot cell, albeit it may be cheaper to replace a damaged component with a new one.
4.19.4.4

Neutron-Irradiation Effects

Several authors have reviewed the properties of
neutron-irradiated beryllium for fusion applications
in the past.139–141 Neutron irradiation leads to complex changes in the microstructure, such as the
radiation-induced change of volume in beryllium,
which is dominated by the nucleation and growth of
He bubbles.
There are two important pathways for gas production. One is the (n, 2n) reaction in which the 9Be is
reduced to 8Be, which then splits into two 4He atoms.
The second is the (n,a) reaction where the 9Be
absorbs a neutron and then splits to form a 4He and
a 6He. The 6He rapidly undergoes a bÀ decay to
become 6Li. The 6Li then reacts with a thermal
neutron to produce 4He and 3H. These processes
have been incorporated into the inventory code

FISPACT,142 which is used (see, e.g., Forty et al.143)
to estimate the generation rates of gas and other
reaction products in a tokamak.
Helium generation has significant effects on the
properties of materials, especially at elevated temperatures. Helium is initially trapped within the beryllium
lattice in submicroscopic clusters. At higher neutron
fluence massive helium-bubble-induced swelling
occurs, especially at elevated irradiation or postanneal
temperatures. Because of the atomistic nature of the
helium bubble nucleation and growth, porous beryllium microstructures, such as from powder metallurgy
or plasma spray technology, were not found to be
effective in releasing significant amounts of helium
under fusion reactor conditions.2
The maximum neutron-induced damage and
helium production expected in Be for ITER firstwall applications (fluence of 0.5 MWamÀ2) are
$1.4–1.7 dpa and $1500 appm, respectively and the
expected irradiation temperatures are in the range
of 200–600  C. The maximum temperature is on
the surface of beryllium tile and depends on thickness and heat flux. Tritium production in beryllium
is expected to be about 16 appm. Recently, Barabash
et al.144 have analyzed the specific effects of neutroninduced material property changes on ITER PFCs
foreseen during ITER operation.
Typically, property changes induced by neutron
irradiation are investigated by exposing samples/
mock-ups in fission reactors. However, the differences between the fission and fusion neutron spectra

643

are important to interpret and predict the effects.
The key difference is transmutation production,

which needs to be considered for the correct prediction of the material performance.145 During irradiation in fission reactors, for example, the typical value
of the ratio (appm He per dpa) is 100–250, whereas
for a fusion neutron spectrum this value is $1000.
Depending on operational temperature, the dpa or
He transmutation must be used as a reference neutron damage parameter. For beryllium, during lowtemperature irradiation (<$300  C) the dpa value
must be considered. For high-temperature irradiation
(more than $500  C), the He generation must be
taken as the reference parameter.
A detailed discussion on this subject is beyond the
scope of this review. We summarize only some of the
main findings with emphasis on results for ITER
relevant grades. Considerations of the effects of neutron irradiation of duplex Be/Cu alloy mock-ups are
provided in Section 4.19.5.
4.19.4.4.1 Thermal conductivity

For S-65C Be grade irradiated up to 1025 n mÀ2
($0.74 dpa) at $300  C, the thermal conductivity
was found to be similar, within experimental error,
to that of the unirradiated material.146 Similarly,
no effect was seen for Be S-65C after irradiation
at 350 and 700  C to a damage dose $0.35 dpa.147
Significant changes in the thermal conductivity
were observed only for conditions that lead to significant changes of the beryllium structure, such as
the formation of a high density of radiation defects
(especially at low irradiation temperature and high
dose) or high (more than tens of percent) swelling.144
Other physical properties (elastic modulus, coefficient of thermal expansion, etc.) are not influenced
by neutron irradiation (at least at the fluence
and temperature ranges relevant for the beryllium
armor for the ITER PFCs).

4.19.4.4.2 Swelling

It is well known that beryllium swells when irradiated
by neutrons, especially during high temperature
irradiation. Reviews of the available swelling data
for different Be grades can be found elsewhere (see,
e.g., ITER MAR,129 Billone,139 and Barabash et al.141).
The computer code ANFIBE (ANalysis of Fusion
Irradiated BEryllium), has been developed and
applied in the past as an interpretative and predictive
tool148 for the prediction of beryllium swelling. The
driving force for the swelling is the presence of
He, which forms He bubbles, especially during


644

Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

high-temperature irradiation (more than $400  C) or
after high-temperature annealing. The maximum
values of swelling could reach approximately tens of
percent at temperatures more than 600  C and
helium content more than several thousand atomic
parts per million. Swelling depends on the structure
of the beryllium: beryllium grades with small grain
size ($8–10 mm) and high BeO content ($3–4 wt%)
have a higher resistance to swelling than conventional
Be grades.141 As concluded in ITER MAR,129 for an
irradiation temperature <$400  C, swelling of beryllium containing 1500 appm He is <$1%. At higher

temperature, swelling could reach the value of a few
percent at the end of life.
4.19.4.4.3 Mechanical properties

Beryllium undergoes hardening from dislocation
pinning and grain-boundary decohesion from the
helium bubbles at the interfaces. The general qualitative trends in the strength and ductility behavior of
neutron-irradiated Be were summarized in Gelles
et al.140: (1) at low and moderate irradiation temperatures (20–500  C) the strength typically increases,
while ductility decreases, in some cases to zero
value; (2) at high temperature (more than $600  C)
the ductility decreases without an increase in the
strength; and (3) an increase of the dose leads to
saturation of the changes of the strength and ductility.
The main effect is the embrittlement of Be at low
irradiation temperature (<$200  C) due to accumulation of the radiation defects in the form of dislocation loops, and at high irradiation temperature (more
than $400  C) due to He bubble formation at grain
boundaries. For the range of conditions expected
at the ITER first wall (e.g., Be temperature range
200–600  C and damage level at the end of life of
$1 dpa), it is expected that the ductility of Be will be
at the level <$1%. Recent irradiated Be mechanical
property studies (at ITER relevant low temperatures)
demonstrated that no catastrophic embrittlement or
thermal conductivity degradation occurred.149
4.19.4.4.4 Thermal shock effects

First-wall beryllium protection tiles in ITER will be
subjected to thermal shocks (disruptions, VDEs). The
damage during these events is a complex function of

the heat loads and material properties, which, as
described above, are neutron irradiation dependent.
Only limited studies to investigate the combination of
these effects have been performed,150,151 but more
work in this area is necessary. Several Be grades
(S-65C, plasma sprayed Be, TShG-56, etc.) have

been irradiated at 350 and 700  C to a fluence
of $0.35 dpa and corresponding He content of
$55 appm. An increase of erosion after neutron irradiation (up to 100%) was observed for different
grades of beryllium for a thermal shock disruption
load of $15 MJ mÀ2. It was concluded that in this
case thermal erosion is not caused by simple evaporation, but by the loss of some particles due to the
brittle destruction of the surface.
4.19.4.4.5 Bulk tritium retention

Several studies have been conducted in the
past to investigate tritium trapping in irradiated
beryllium152,153 and it has been found that the tritium
retention can increase with neutron fluence by a
factor 3–10. Tritium loading in these experiments
was performed by exposure to tritium in the gaseous
phase. Tritium atoms are indeed trapped by neutroninduced defects such as dislocation loops and helium
bubbles. Bulk tritium retention due to implantation,
and, as a result, the effect of the damaged microstructure on this retention, is expected to be less serious
than previously anticipated.154

4.19.5 Fabrication Issues
4.19.5.1 Joining Technologies and High
Heat Flux Durability of the Be/Cu Joints

One of the most critical aspects of the design of the
ITER first wall is the attachment of the Be tiles to the
actively cooled copper alloy substrate. The primary
threat to this attachment comes from the heat flux.
However, secondary effects also arise from the
fact that there will be thermal gradients as well as
mechanical loads from disruptions, which will cause
distortion. These effects cannot easily be simulated
experimentally. It should, therefore, be the responsibility of the design to minimize these effects.
At the time of writing this chapter, the first wall of
ITER was undergoing a major redesign. Because of
the lack of more detailed information on the new
design, most of the considerations presented here,
albeit general, are on the basis of the design of
actively cooled ITER first-wall panels, which consist
of stainless steel tubes in a copper heat sink with
10 mm thick beryllium tiles bonded to the plasmafacing surface. This technology was deemed to be
adequate for handling the previously assumed surface
heat flux of 0.5 MW mÀ2, and could potentially withstand 3 MW mÀ2(155) for a limited number of cycles.
However, because of fatigue it would be incompatible


Beryllium as a Plasma-Facing Material for Near-Term Fusion Devices

with areas subjected to higher power densities during
each pulse. In such cases, other options for the heat
exhaust technology are being considered,156 using
thinner Be tiles.
4.19.5.1.1 Be/Cu alloy joining technology
4.19.5.1.1.1 Background information


The main problem of bonding Be to Cu alloys is that
Be reacts with almost all possible metals (except Al,
Si, Ag, and Ge) and forms brittle intermetallic
phases.157–159 Such bond joints have poor mechanical
integrity. More robust joints use metal interlayers to
act as either diffusion barriers and/or strain accommodating compliant layers to avoid the formation of
deleterious phases and to assist in the accommodation of thermal cycling-induced strains.160
R&D has been performed over a number of years
to develop the design and manufacturing techniques
required to meet the demanding design requirements. Significant experience has been gained with
these manufacturing techniques and the associated
inspection techniques. It must be noted that in the
1990s the best joining technology developed for
manufacturing the Be/Cu actively cooled components was brazing with Ag-base alloys (e.g., InCuSil
with $41.75% Ag) which was successfully used in
JET. However, the use of Ag base brazing alloy was
not allowed in ITER mainly because of the transmutation by neutron irradiation to Cd ($5 wt% Cd will
be produced in Ag–Cu eutectic alloy at 1 MWamÀ2)
whose presence would (1) reduce the melting temperature of the braze; (2) lead to the formation of
highly radioactive isotopes; and (3) affect the pumping system in case of Cd release to the vacuum
chamber and codeposition in the cryopumps panels.
During the early stage of the ITER first-wall
design development, dispersion-strengthened copper
(DS-Cu) alloys (e.g., Glidcop Al25) were considered
as the first option because (1) the stresses were
within the design allowable, and (2) they had better
thermal stability under the manufacturing route,
which required a first wall to be integrated with a
4 t shield. The main developments for fabrication

of joints between Be and DS-Cu alloys are reported
in ITER MAR129 and Lorenzetto et al.161 However,
as a result of a design change that took place
from an integrated first-wall panel to a separated
first-wall panel design, a precipitation-hardened
copper–chromium–zirconium alloy (CuCrZr), was
subsequently chosen. This was because the fracture
toughness of DS-Cu is very low above 200  C even
for unirradiated material. Fracture toughness of the

645

unirradiated and irradiated CuCrZr alloy decreases
with increasing temperature, but it remains at a relatively high level in the ITER working temperature
range and it is significantly higher than fracture toughness of DS-Cu. The use of separable first-wall panels
makes it possible to perform heat treatments with fast
cooling rates, which are mandatory to adequately
retain the mechanical properties of precipitationhardened materials.
Thus, extensive studies were then performed
during the last 10 years to develop reliable silver
free Be/CuCrZr alloy joining techniques and to
modify the joining conditions to minimize the
mechanical strength loss of the CuCrZr alloy. Different methods have been considered and investigated. Some of these were eliminated because of bad
results (e.g., explosive bonding, inertial welding,
joint rolling, and some types of brazing). Two methods gave good results and were kept for further
investigations: HIPping and brazing. Good results
were achieved with the HIP joining technique by
lowering the HIP temperature as close as possible
to the CuCrZr alloy ageing temperature (about
480  C) and with the brazing technique in the

development of a fast brazing method to minimize
the holding time at high temperature. The latter was
achieved by induction brazing in Europe and by fast
heating and cooling using an e-beam test facility in
the Russian Federation.
4.19.5.1.1.2

HIP joining technique

An extensive development programme performed
especially in Europe has enabled the production of
very good Be/CuCrZr alloy joints by HIPping.
The progress on the fabrication of Be/CuCrZr joints
in Europe is described in Lorenzetto et al.155,161,162
and Sherock et al.163 The HIP joining temperatures
ranged from 540 to 580  C. Different interlayers such
as Cr, Ti, and Cu were tested and reported. The
selection of the joining conditions used for the fabrication of representative first-wall mock-ups was done
on the basis of mechanical test results performed with
guillotine shear test specimens. The best results were
obtained with Ti and Cu interlayers at 580  C, in
which the shear strength exceeded the yield strength
of the parent materials as creep of the CuCrZr part
or rupture of the Beryllium part was observed. Performance achieved with representative first-wall
mock-ups exceed the present ITER design requirements, namely, 30 000 cycles at 0.5 MW mÀ2 peak
heat flux plus transient events up to 1.4 MW mÀ2
for about 1000 cycles (see Section 4.19.5.1.2).