PLASMA HEATING IN PRESENT-DAY AND FUTURE FUSION MACHINES

Yevgen Kazakov, Dirk Van Eester, and Jef Ongena

Laboratory for Plasma Physics, LPP-ERM/KMS,

EUROfusion Consortium Member, Brussels, Belgium

ABSTRACT plasma, they transfer their energy to bulk ions and

electrons via Coulomb collisions. The second method
A fusion reactor requires plasma pre-heating before relies on launching electromagnetic waves into the
the rate of deuterium-tritium fusion reactions plasma and providing proper conditions for strong
becomes significant. Ohmic heating due to the localized wave damping by ions and/or electrons.
toroidal plasma current that flows in tokamak plasmas Due to the inhomogeneity of the confining magnetic
allows to reach temperatures of several keV only. field in toroidal plasmas, the position of the power
In this lecture we provide a short overview of the deposition for radiofrequency (RF) heating can be
two main auxiliary heating systems relevant for controlled externally, e.g., by a choice of the wave
fusion machines: an injection of high-energy neutral frequency. The supported by plasmas resonant wave
particles and radiofrequency plasma heating. frequencies will be identified below, using a simplified
– yet, good as a first approximation – cold-plasma
approach.
I. INTRODUCTION There are numerous applications of NBI and
RF systems beyond heating itself. Tokamaks
The energy balance for a fusion reactor implies are inherently pulsed machines, and driving
that the energy losses are compensated by the plasma non-inductive current with auxiliary heating systems
heating due to fusion-born alpha particles. The is essential for maximizing the plasma pulse duration,
fusion rate is a strong function of plasma temperature a prerequisite condition for economic viability of
and at low temperatures is negligible. Hence, tokamaks as fusion reactors. Heating systems are also
plasma pre-heating is required before a positive power routinely used for diagnostics purposes, controlling
balance can be reached in a fusion machine. impurities and plasma instabilities, optimizing
A fusion triple product, which characterizes the particle and energy transport, for plasma start-up
performance of a fusion device, is minimized at and landing, etc.
plasma temperature T ≈ 15 keV. The initial heating
in tokamaks comes from the ohmic heating: likewise
an electric current flowing through a metal wire II. NEUTRAL BEAM INJECTION HEATING
heats it up, the toroidal plasma current, which is
used in tokamaks for plasma confinement, dissipates Neutral beam injection has been used as
its energy and heats the plasma. However, ohmic a working horse for plasma heating in most of
heating allows to reach plasma temperature of a few fusion machines. For instance, the world highest
keV only [1, 2]. As the plasma temperature increases, fusion power 16 MW was obtained on JET tokamak
the collisional frequency and the plasma resistivity, by injecting 22 MW of NBI and 3 MW of ICRF
being proportional to ∝ 1/T 3/2 , decrease. The heating. A strong magnetic field is used for plasma
limitations of ohmic heating are discussed in more confinement in magnetic fusion devices, and this
detail in a separate lecture by R. Koch and D. Van prohibits a direct injection of energetic ions into
Eester [3]. There, it is also shown that raising the plasma. Whereas injected energetic ions will be
the plasma current to increase the ohmic power deflected by the magnetic field and can not penetrate
2
PΩ = R Ipl is limited by the necessity to keep the deep into the plasma, neutral particles do not feel
edge safety factor qedge > 2 for MHD stability of the Lorentz force and will travel along straight-line
the plasma. Note that for stellarators the confining trajectories until their ionization by collisions with
magnetic field is solely produced by the external the background plasma. The penetration length
magnetic coils and the plasma current is so low that for a beam in the plasma – before most of its
ohmic heating is negligible. power is deposited due to ionization – depends
There are two main methods for heating fusion essentially on the injection beam energy ENBI . As
plasmas. The first technique is based on injecting the will be discussed in this section, most of present-day
beam of high-energy neutral particles to the plasma tokamaks and stellarators operate with beam energies
(NBI). As the beam particles are ionized in the ENBI ≃ 100 keV, and the application of NBI to

290
ITER and next-step devices requires a fundamental energetic neutrals occurs. The neutralizer consists of
change and an increase of the beam energy by an a simple gas cell, filled with the molecular gas and
order of magnitude up to ENBI ≃ 1 MeV. ENBI open at each end, through which the beam passes.
also determines whether a collisional ion or electron Normally the gas species for the neutralizer is chosen
heating dominates due to the slowing down of the the same as for the ion source. The density of the
NBI-generated fast ions. Present-day NBI systems neutralizing gas is varied to achieve the maximum
usually provide a high fraction of ion heating, while efficiency of ion-to-neutral conversion. Figure 2
(max)
for the future machines most of the injected beam depicts the maximum neutralization efficiency, ηneutr.
power will be transferred to electrons. computed for atomic D and H ion beams as a function
A schematic diagram of NBI system is shown in of beam energy. For the same NBI acceleration
Fig. 1. There is no direct way of accelerating neutral energy, the neutralization efficiency of H beams
particles to high energies, and one still has to start is lower than for D beams because of the larger
with accelerating electrically charged ions, and then energy per nucleon. As follows from this figure,
set a dedicated system to convert energetic ions into the neutralization efficiency for D+ ions at ENBI ≈
energetic neutral particles. Low-energy ion source is 100 keV is reasonably good, ≈ 50%, but at higher
a first step of the NBI system. Beam energies of about beam energies decreases to unacceptably low levels.
100 keV are sufficient for the machine’s and plasma In contrast, for negative beams (both for D− and H− )
parameters of the existing fusion devices, and this (max)
the neutralization efficiency is much higher, ηneutr. ≈
allows using well-established techniques for producing 60%. This is the reason why using negative ions is the
positive ions. However, ion sources based on a plasma inevitable choice for neutral beam systems of ITER
discharge produce not only the required atomic ions and future fusion machines, which will operate at
(e.g., D+ ), but also a significant fraction of molecular MeV level.
ions (D+ +
2 and D3 ). The molecular ions will ultimately
dissociate into atoms with one-half and one-third NBI
Maximum neutralization efficiency (%)

100
energies. Such ENBI /2 and ENBI /3 atoms will have +
Deuterium, D (P-NBI)
−
Deuterium, D (N-NBI)
a lower penetration to the plasma and deposit more +
Hydrogen, H (P-NBI)
heating at the plasma edge, which is not desirable. In 80 −
Hydrogen, H (N-NBI)

future fusion machines, including ITER, a different D

−

technology for the ion source has to be adopted, and 60

H
−

NBI systems will be based on using negative ion

beams (N-NBI). 40

20
+ +
H D

0
10 100 1000
NBI energy, ENBI (keV)

Figure 2: Maximum neutralization efficiency for D

and H atomic ion beams as a function of beam energy.

It is quite easy to compute quantitatively the

maximum value of ηneutr. for N-NBI systems. At high
Figure 1: Layout of neutral beam injection system beam energies 200 − 1000 keV/amu, the two most
(Courtesy: Ursel Fantz, IPP-Garching). important reactions occurring in the neutralizer are:
1) a stripping reaction D− + D2 → D0 + D2 + e
At the second stage, the ions are accelerated (cross-section σ−10 ), which converts a negative ion
to the required energy by applying a high DC to an energetic neutral; and 2) a competing loss
voltage. Because a positive electrostatic voltage process due to a re-ionization of the formed neutral
will be applied to extract negative ions in N-NBI by a collision with background neutrals, D0 + D2 →
systems, there will be no one-half and one-third D+ + D2 + e (σ01 ). Also negative ions can be
beam components present after the extraction. The directly converted to positive ions via a reaction
accelerated beam current is usually about 50 A, and D− + D2 → D+ + D2 + 2e (cross-section σ−11 ).
one can easily compute the consumed input electrical Then, the maximum neutralization efficiency, when
power for this part of NBI system by multiplying the the gas density of the neutralizer is optimized for
acceleration voltage by the beam current. D0 generation, is connected to the cross-sections as
After this stage a directed beam of high-energy follows
ions is formed and it enters the neutralization (max) σ−10
chamber, where the conversion of energetic ions into ηneutr. = pp/(1−p) ≈ 58%. (1)
σ−10 + σ−11

291
Here, p = σ01 /(σ−10 + σ−11 ) ≈ 0.28 and the first After passing the neutralization stage, the
term σ−10 /(σ−10 + σ−11 ) ≈ 0.95. The tabulated remaining neutral beam either impinges onto the
data for the cross-sections can be found in [4]. For calorimeter or continues into the duct leading to
negative ions the neutralization efficiency at high a plasma. Adopting a movable calorimeter allows
beam energies remains nearly constant. At the an off-line commissioning and optimization of the
exit of the neutralizer, the beam consists of 58% of beam system without plasma operation. Combined
high-energy neutrals, and 25% and 17% of positive with measurements of the losses in the neutral beam
and negative ions, respectively. duct, this also allows the injected NBI power to be
One should note here that for present-day NBI determined independently of the plasma device.
systems based on positive ions, the acceleration
Estimating the required NBI energy
energy, beam current and beam power fractions
depend significantly on the ion species used for In this subsection, we estimate the required
injection. Table 1 illustrates this for the NBI beam energies as a function of machine’s size and
system of the tokamak JET, which consists of two operational plasma densities, and show why NBI
neutral injector boxes equipped each with up to eight system for ITER has to provide beams with the
positive ion neutral injectors (PINIs) [5]. The system energy of about 1 MeV.
was recently upgraded to increase the total injected When the neutral beam particles enter the
deuterium neutral beam power to at least 34 MW and plasma, there are several processes causing an
to increase the beam pulse length. If operating with ionization of a high-energy neutral in a plasma:
deuterium, the full-energy beam component carries charge exchange, electron and ion impact
about 50% of the total NBI power, and for the ionization in collisions with plasma electrons,
hydrogen mode of operation a half-energy component ions and impurities. As the neutral beam
is the dominant. penetrates and is absorbed in the plasma, its
flux exponentially decays with the propagation
R ∆l
distance, I(∆l) = I0 exp[− 0 dl n(~r)σ(~r)], where
Table 1: Measured (D2 ) and predicted (H2 and ∆l is a distance along the beam propagation path.
T2 ) parameters of the JET NBI system after the Assuming a constant density profile n ≈ const, one
completion of recent upgrade [5]. The total power can write
is computed for two neutral injector boxes equipped I(∆l) = I0 e−∆l/LNBI . (2)
with up to eight PINIs each. A characteristic distance, at which the initial beam
Parameter \ Gas species H2 D2 T2 intensity is attenuated by a factor of e ≈ 2.7 –
Max. beam energy (keV) 90 125 118 further referred to as the beam decay length – is
Max. beam current (A) 50 65 45 then LNBI = 1/(nσ). For simplicity, we neglect an
Max. power per PINI (MW) 1.0 2.16 2.2 ionization due to electron impact. For E/ANBI ≥
Max. total power (MW) 16.0 34.6 35.2 40 keV/amu, the total cross-section for charge
exchange and ion impact is inversely proportional
Predicted beam power
to the beam energy and can be approximated as
fractions, E0 : E0 /2 : E0 /3 28:44:28 52:39:9 63:26:11
σi+CX (m2 ) ≈ 1.8 × 10−18 / [ENBI (keV)/ANBI ]. Then,
the beam decay length is given by
Since one should avoid of injecting high-energy ENBI (keV)/ANBI
positive/negative ions from NBI system to the plasma LNBI (m)|σ=σCX +σi ≃ , (3)
180ne,20
(after entering the plasma they will likely deposit
their energy on the neutral beam entrance port), where ne,20 is the plasma density expressed in the
the next stage of the NBI system is the residual units 1020 m−3 .
ion dump. Here, the charged beam components Let us consider an updated JET NBI system with
are filtered out by using the deflection magnets or 125 keV D0 beams as an example. Using a typical
the electrostatic field. To increase the wall-plug plasma density of ne = 5 × 1019 m−3 , Eq. (3) predicts
efficiency of NBI systems (in present-day machines LNBI ≈ 0.7 m, which is comparable to the minor
it is ≈ 20 − 30% only), for devices following ITER radius of the machine a ≈ 0.9 m. However, such
a possibility of energy recovery (recirculating the NBI energies are too low for heating ITER plasmas
residual negative ions) is explored [6]. In addition, (a = 2 m, ne = 1 × 1020 m−3 ): according to Eq. (3),
utilizing photo-detachment or plasma neutralizers to at ENBI = 125 keV a decay length will be LNBI ≈
improve the neutralization efficiency beyond 58% 0.35 m (taking into account charge-exchange and ion
– maximum for the standard gas neutralizers – is impact ionization mechanisms only). A short decay
being actively studied in the context of DEMO length (i.e., LNBI ≪ a) is undesirable since most of
development. The wall-plug efficiency of NBI systems the energy is deposited at the outer part of the
has to be improved by a factor of two to become plasma volume. As a result, the beam energy has
reactor-relevant [6]. to be significantly increased for ITER and future

292
larger, higher-density fusion machines in order to For ITER operating with deuterium beams,
allow neutrals to penetrate deeper into the plasma. Eq. (4) yields an estimate of ENBI ≈ 600 keV.
Why are then heating NBI systems of ITER
designed for 1 MeV D0 beams? In fact, at high
beam energies (of several hundreds keV or higher)
Tangential injection
an additional mechanism – multistep ionization
– produces a substantial increase of the beam
stopping cross-section and this reduces the beam
penetration. As discussed in [8, 9], the multistep
∆l1
ionization arises from excitation of the beams
2
α 1 R
and the subsequent ionization of already excited
RR00-–aa R0 R0
Rtang R0 + a
neutral atoms. The enhancement factor of the
ionization against the single step-processes considered
above, σeff = σ(1 + δ) is theoretically predicted to
increase with the beam energy and the electron
Normal injection density, δ = δ(ENBI , ne , ...). While an effect of the
multistep ionization is relatively small for present-day
beam systems, an enhancement of the stopping
Figure 3: Plan view illustrating a very different beam cross-section by a factor of two was measured
propagation length for normal and tangential NBI for 350 keV hydrogen beam in JT-60U tokamak
injection. (δ ≈ 0.8 − 1.05), in accordance with theoretical
predictions. A similar enhancement factor is expected
The required NBI energy for ITER can be for high-energy deuterium beams in ITER. Then,
estimated as follows (see Fig. 3). In ITER, Eq. (3) for estimating the beam decay length in
the two heating neutral beams will inject beams a plasma has to be adopted accordingly to include
tangentially with a tangency radius Rtang = 5.28 m the multistep ionization
(Rtang ≈ R0 − a/2), and the third diagnostic beam ENBI (keV)/ANBI
is designed for a normal injection [7]. The full LNBI (m) ≃ . (5)
180 (1 + δ) ne,20
path of the q tangential beam through the plasma
2 The same multiplication factor (1 + δ) has to
is ∆l2 = 2 (R0 + a)2 − Rtang ≈ 12.5 m, and
be included to Eq. (4), when estimating the
because of the very long geometrical distance the required beam energies and this correction yields
shinethrough fraction (amount of NBI power arriving ENBI ≈ 1 MeV for ITER.
at the wall) is very small. However, a significant Initially ITER will use two neutral beams for
fraction of beam power is deposited in the outer plasma heating that are designed to inject each
shell of the plasma during a relatively short path 16.5 MW of power to the plasma. Operation
of the beam since its entrance to the plasma with deuterium (1 MeV/40 A) and hydrogen
(point ‘1’ in Fig. 3) to the point, when it intersects (870 keV/46 A) will be supported. Note that
the flux surface r/a = 0.5 q (point ‘2’). This whereas tritium has been used with positive ion
distance is given by ∆l1 = 2
(R0 + a)2 − Rtang − based injectors during the D-T experiments on JET
q
2
and TFTR tokamaks, a current status of the ITER
(R0 + a/2)2 − Rtang ≈ 1.4 m. The fraction NBI design does not consider using tritium due
of beam power deposited along this path is to the regulations on tritium handling. The third
pabs ≈ 1 − e−∆l1 /LNBI , and this sets a limitation NBI port in ITER is reserved for the diagnostics
on the acceptable NBI decay length and, hence, neutral beam that will be installed to support the
required beam energies. LNBI should be comparable charge exchange recombination spectroscopy, which
to ∆l1 or even less in order to reduce the beam is an essential diagnostic for the measurement of the
power deposition at the edge. In fact, already at helium-ash density in the core of D-T plasmas. The
∆l1 /LNBI = 0.7, > 50% of the incident beam diagnostics neutral beam in ITER will use a normal
power is deposited within r/a ≥ 0.5 region. For injection of the hydrogen beam with a much smaller
a fusion device with the ITER-like
√ aspect ratio, energy (H− , 100 keV/60 A).
∆l1 can be approximated as aR0 /3, and one can
derive an estimate for the required NBI energy Critical energy
√
(LNBI & 0.5 aR0 ) After the ionization of the injected neutral
p particles in the plasma, the resulting fast ions are
ENBI (keV)|σ=σCX +σi & 90 aR0 ne,20 ANBI . (4) slowed down by Coulomb collisions with bulk plasma
ions and electrons. A Fokker-Planck description of
In Eq. (4), the beam energy is given in keV, and the the test-particle slowing down is normally adopted
machine’s minor and major radii are in meters. to compute the resulting power deposition. The

293
 NBI (JET) NBI (ITER) Alphas (ITER) a much higher beam energy has to be adopted,
1
as discussed in the previous subsection, and the
ratio E0 /Ecrit will be significantly larger than for
Power redistribution fractions

0.8 the present-day NBI systems. As a result, N-NBI

systems with ENBI ≈ 1 MeV will provide a dominant
electron heating. A similar test-particle slowing down
0.6 reasoning applies for fusion-born alpha particles.
pion
pelec Since the birth energy of alpha particles is 3.5 MeV,
0.4 this means that their energy loss is mainly due to
collisions with plasma electrons.

0.2
III. RADIOFREQUENCY HEATING
0
0 2 4 6 8 10 12 Radiofrequency heating is another very efficient
E0/Ecrit method for increasing fusion plasma temperatures.
Magnetized plasmas have several ‘natural’ resonant
Figure 4: Fraction of the fast ion energy transferred frequencies and support the variety of wave modes.
to bulk ions and to electrons as a function of E0 /Ecrit . This results in the existence of many different RF
The shaded areas highlight regimes relevant for JET heating and current drive scenarios. Though RF
(D)
NBI system (E0 =125 keV, T = 3 − 7 keV), ITER heating involves a much more complicated physics
(D)
N-NBI system (E0 = 1 MeV, T = 10−15 keV), and of electromagnetic wave propagation in plasmas and
(α)
for fusion-born alpha particles (E0 = 3.5 MeV). wave-particle interaction, this, in turn, makes RF
heating much more flexible in terms of external
control of the heating region than for neutral
corresponding derivation one can find in a lecture [3].
beams. For example, if the RF heating in the ion
For our discussion it is sufficient to note that whether
cyclotron range of frequencies (ICRF) is applied,
the resultant dominant bulk ion or electron heating
changing power deposition from dominant ion to
occurs, depends on the ratio of the fast-particle
dominant electron can be done by varying the chosen
energy to the so-called critical energy, Ecrit defined as
operational wave frequency or selecting a proper
!2/3 plasma composition. Furthermore, ICRF is the only
heating method in future fusion machines capable of
X
Ecrit = 14.8Afast Te Xi Zi2 /Ai , (6)
i providing a significant fraction of bulk ion heating.
On the other hand, heating in the electron cyclotron
where Xi = ni /ne are the concentrations of various range of frequencies (ECRF) is characterized with
ion species in a plasma. A fast ion with the energy a very localized power deposition, and this allows
E = Ecrit , transfers instantaneously the same amount also to use ECRF as a tool to control the plasma
of power to ions and electrons via collisions. pressure and current density profiles for the MHD
If we consider a slowing down of fast particles plasma stability [10]. Lower hybrid (LH) heating
with the initial (birth) energy E0 to thermal has been actively exploited on different present-day
velocities, then the average fraction of the total machines as a very efficient method of non-inductive
energy given up by the fast particles, which goes into current drive.
the thermal bulk ions of the plasma, is For ITER, in addition to 33 MW of NBI power,
E0 /Ecrit
installing two RF heating systems (ECRH and ICRH)
Ecrit dy
Z
pi (E0 ) = . (7) each providing 20 MW of auxiliary power is foreseen.
E0 0 1 + y 3/2 A further 40 MW upgrade of the ITER heating mix,
with a possibility of installing lower hybrid heating,
The fraction of power flowing to electrons is then
is also considered.
pe (E0 ) = 1 − pi (E0 ). (8) The principle of wave heating is similar for all RF
scenarios. A generator (the type of utilized source is
A plot of pi and pe as a function of E0 /Ecrit is shown very different depending on the operating frequency)
in Fig. 4. Note that at E0 = Ecrit about 75% of the sends waves along the transmission line to a launching
fast ion energy is transferred to the plasma ions, and structure located at the plasma edge [11]. Then, in
an equal total energy transfer to ions and electrons is contrast to NBI heating which has no problem of
reached at E0 /Ecrit ≈ 2.4. injecting neutral particles to the plasma, a special
It is instructive to illustrate these results with care has to be taken to optimize coupling of RF
(D)
a few examples. For ENBI = 125 keV beam particles power from the launcher to the plasma [12]. Once the
injected to 5 keV deuterium plasma, E0 /Ecrit ≈ 1.3 required electromagnetic wave starts its propagation
and pi :pe ≈ 2:1. For NBI system relevant for ITER in the plasma, the focus of the discussion shifts

294
towards understanding how to make RF power to follows from Eq. (10), if B0 6= 0 plasma behaves as
be absorbed at a desired part of the plasma volume, a gyrotropic medium: e.g., vx component is not only
often preferably close to the plasma center. proportional to Ex , but depends also on Ey .
It is impossible to cover all aspects of plasma ~ are known, the current
If vi as a function of E
waves physics in a short lecture note. An interested density can be easily computed ~j =
P
s=e,i qs ns~
vs
reader is referred to lectures [13, 14] for the further ~ (σ is defined as the
and formally written as ~j = σ E
reading, and to books [15, 16] for an in-depth
conductivity tensor). The second equation (9) is often
discussion on plasma waves in magnetized plasmas.
re-written in a simpler form, by combining the two
In this lecture note, we will focus only on a single issue
terms at the right-hand side
of RF plasma heating: determining resonant wave
frequencies suitable for efficient heating and current ~
1 ∂E 4π ~
1 ∂(ǫE)
drive. ~ =
rotB + ~j = , (11)
c ∂t c c ∂t
Cold plasma dielectric tensor
and introducing a quantity ǫ, which is called as the
The propagation of electromagnetic waves in an plasma dielectric tensor. The plasma dielectric and
arbitrary medium is described by Maxwell’s equations the conductivity tensors are related via ǫ = I +
(cgs system is adopted here) (4πi/ω)σ.
Then, by introducing the wave refractive index
 ~
~ = − 1 ∂B
rotE
 ~n = c~k/ω, Eqs. (9) and (11) can be merged into
c ∂t (9) ~
a single equation for the RF electric field E
1 ∂ ~
E 4π
 ~ =
rotB + ~j.
c ∂t c ~ + ǫE
~n × (~n × E) ~ = 0. (12)
Plasma consists of charged particles, ions and In the cold plasma limit, the plasma dielectric
electrons, and it influences the wave propagation tensor is a function of the imposed wave frequency
characteristics via the wave induced current density (see Eqs. (10)), but is independent of the wave
~j. Depending on the level of complexity of describing
number. The Stix’s notation for ǫ is commonly
the system, different plasma models can be adopted. adopted [15]
We consider the simplest case, when a cold plasma  
is immersed in a constant equilibrium magnetic S −iD 0
field B~ = B0~ez and the wave amplitude is small ǫ = ǫ(ω) = +iD S 0, (13)
such that wave propagation can be treated as a 0 0 P
linearized perturbation for every individual mode P 2
B ~ ∼ ∝ ei(~k~r−ωt) (ω is the wave frequency and
~ ∼, E where S(ω) = 1 − s ωps /(ω 2 − ωcs
2
),P D(ω) =
2 2 2 2 2
P
~k is the wave number). Under these assumptions, s (ω cs /ω)(ω ps /(ω − ω cs )), P (ω) = 1 − s ωps /ω ,
2 2
the time derivative and curl operator in Eqs. (9) and ωps = 4πns qs /ms is the square of the plasma
frequency.
can be replaced with ∂/∂t → −iω and rot = i~k×,
Equation (12) is a set of three coupled, linear,
respectively.
homogeneous equations for the three components of
Cold plasma description allows to use a simple
the RF electric field. Requiring non-trivial solutions
equation of motion for an ion/electron, d~v /dt =
~ to exist, the determinant of these equations must
q E/m + (qB0 /mc) ~v × ~ez , to characterize its response
vanish, and one can obtain the wave dispersion
to the electromagnetic field. This equation for the
relation, which defines the refractive index n as
simplified case considered here can be easily solved,
a function of wave frequency ω:
and the individual velocity components as a function
of RF electric field are given by A(ω, θ)n4 − B(ω, θ)n2 + C(ω, θ) = 0, (14)
iq ω q ωc


 vx = Ex − Ey , where θ is an angle between the wave vector and the
 2
m ω − ωc 2 m ω − ωc2
2
confining magnetic field (ky = 0), and the functions


iq ω q ωc

vy = E + E , (10) A(ω, θ), B(ω, θ) and C(ω, θ) can be easily derived
m ω 2 − ω2 y m ω 2 − ω2 x
c c (see, e.g., [15]).


vz = iq Ez ,


Equation (14) is a bi-quadratic, and hence for

mω
any ω there are two – generally different – solutions
and the cyclotron frequency ωc = qB0 /mc has been n2 = n21,2 (ω). This means that within the cold
introduced. The cyclotron frequency depends on q/m plasma approximation for any wave frequency plasma
ratio of a particle, and therefore for electrons it is supports two different wave modes (with a different
about 2000 times larger than for ions. In addition, ωc dispersion relation). In hot plasmas, the dielectric
includes the sign of a particle’s charge, which reflects tensor is also a function of the wave number ~k, which
the fact that ions and electrons rotate in the opposite leads to the appearance of extra new modes and
directions around the confining magnetic field. As solutions, and makes wave physics more complicated.

295
 For heating fusion plasmas, the electromagnetic 4
waves are usually excited to propagate predominantly ω = ωc,D
perpendicular to the confining magnetic field. For 3 ω = ωc,3He
a simplified limiting case of purely perpendicular
propagation θ = π/2, the wave resonances are 2
determined by the condition
X 2
ωps 1
A(ω, θ = π/2) = S(ω) = 1− = 0. (15)
ω 2 − ωcs
2

Z (m)
s=e,i 0
This is a desired equation for computing the resonant
-1
wave frequencies (for a perpendicular, cold plasma
resonance). The solutions of Eq. (15) define the
-2
range of electromagnetic frequencies used for efficient ω = ωc,Be
plasma heating and current drive in fusion plasmas.
-3
Resonant wave frequencies
-4
Before we proceed to finding solutions of Eq. (15),
4 5 6 7 8
it is helpful to note that the electron plasma and
R (m)
electron cyclotron frequencies are comparable in the
√ Figure 5: Location of ion cyclotron resonance layers
core of fusion plasmas (ωpe /ωce ≈ 3 ne,20 /BT ;
BT is the confining magnetic field in Tesla), and the in ITER-like plasma for f = 50 MHz (B0 = 5.3 T).
2 2
relation ωpi /ωci ≫ 1 holds for bulk ions. Another
2
P 2
useful identity is ωpe /|ωce | = i ωpi /ωci – this is just Then, the vacuum and electron contributions are
the charge quasi-neutrality re-written in a different negligible in comparison with the contributions due
way. to ion species, which balance each other.
1) ECRF (electron cyclotron range of frequencies), The solution of Eq. (18) is referred to as the
f ≃ fce ≫ fci . ion-ion hybrid (or Buchsbaum) resonance
2 2 2 2
For this frequency range ions are immobile and 2
ωp1 ωc2 + ωp2 ωc1
Eq. (15) can be simplified by neglecting the ion ωIIH = 2 + ω2 . (19)
ωp1 p2
2 2
contribution term (≃ ωpi /ωce ∝ me /mi ≪ 1)
2 ✟
One can show that the ion-ion hybrid resonant
✟✟ −
ωpi 2
X ωpe frequency is bounded between the two ion cyclotron
✟ω✟
S(ω) = 1 − 2 − ω2
= 0. (16)
ω 2 − ωce
2
frequencies, min(ωc1 , ωc2 ) < ωIIH < max(ωc1 , ωc2 ).
✟i ci
The most successful ICRF heating scenario – minority
This defines the resonant frequency known as the ion heating – relies on wave absorption by a small
upper hybrid (UH) resonance fraction of resonant minority ions (e.g, H) in
2
ωUH 2
= ωce 2
+ ωpe . (17) deuterium majority plasmas (minority and majority
ions should have different q/m). If the concentration
The word ‘hybrid’ refers here that the resonant of minority ions is much less than the concentration
frequency involves both ωce and ωpe . The upper of majority ions, then ωIIH ≈ ωc,mino.
hybrid frequency is somewhat above ωce . Since Wave absorption by ions occurs in the vicinity of
fce ≈ 28 × 109 BT , ECRF heating requires sources the ion cyclotron fundamental resonance (ω = ωci )
in the range 100–200 GHz. Note that gyrotrons – and harmonic (ω = N ωci , N ≥ 2) layers. The latter
sources for ECRF heating – can deliver one or a few is a finite-Larmor-radius effect and usually requires
wave frequencies only. For ITER, f = 170 GHz is significant plasma beta and pre-heating to become
selected and the radiofrequency sources for the ECRF efficient. Since the magnetic field in tokamaks is
system will be composed of 24 gyrotrons (with 1 MW inversely proportional to the distance to the torus
unit output). axis, B(R) ≃ B0 R0 /R, the lines of the ion cyclotron
2) ICRF (ion cyclotron range of frequencies), resonance are (almost) vertical lines in (R, Z) plane
f ≃ fci ≪ fce . and can be determined from the following simple
equation:
For this frequency range, a resonant frequency occurs
when there are two or more ion species present in 15.2 B0 (T)
a plasma RIC ≈ R0 (N × Zi /Ai ) (20)
ω=N ωci f (MHz)
2
ωp1 2
ωp2 2 ✚
ωpe ✚
✚ = 0. (N = 1 for the fundamental resonance, and N ≥ 2
S(ω) = 1✁ − − −
2
ω 2 − ωc1 2
ω 2 − ωc2 ω 2✚− ωce
2 for harmonic resonances). Figure 5 shows the location
✚
(18) of the ion cyclotron resonances in ITER-like plasma

296
for the wave frequency f = 50 MHz (B0 = 5.3 T, IV. FURTHER READING
R0 = 6.2 T). Note that the cyclotron resonance of This note is complementary to the more detailed
3
He2+ ions is located at R ≈ 6.8 m. It can be easily lectures on NBI and fast-particle heating [3], and RF
shifted to the plasma center for core ion heating by heating [10, 13] of the CMSS12. A comprehensive
adopting a different wave frequency f = 54 MHz. introduction to the physics of plasma heating can be
Frequencies used for ICRF heating in present-day found in books [1, 2]. Articles [9] and [17, 18] are
fusion devices usually vary from 20 MHz to about focused on a discussion of heating systems of ITER
100 MHz, depending on the central magnetic field and JET tokamaks. For an in-depth discussion on
and ion species used for plasma heating. Note that – plasma waves, books [15, 16] are recommended.
unlike gyrotrons – ICRF generators are much more REFERENCES
flexible in varying wave frequency, and at every
device a number of different wave frequencies are 1. J. Wesson, “Tokamaks” (Oxford: Clarendon)
normally available for plasma heating. For instance, (2004).
A2 ICRF antennas on JET have been designed to 2. J. Freidberg, “Plasma Physics and Fusion
operate within the frequency range f = 23 − 55 MHz. Energy” (Cambridge University Press) (2007).
For ITER, the bandwidth f = 40 − 55 MHz has been 3. R. Koch and D. Van Eester, “Fast particle
chosen. heating”, these proceedings.
4. C.F. Barnett, “Atomic Data for Fusion, vol. 1,
iii) LH (lower hybrid), f = fLH , fci ≪ fLH ≪ fce . Collisions of H, H2 , He and Li Atoms and Ions
with Atoms and Molecules”, Report ORNL-6086
In the intermediate frequency range between the ion (1990).
and electron cyclotron frequencies, ωci ≪ ω ≪ ωce , 5. D. Ćirić et al., Fusion Engineering and Design
there exists another solution of Eq. (15): 86, 509–512 (2011).
6. R. McAdams, “Beyond ITER: Neutral beams
2 2
X ωpi ωpe for a demonstration fusion reactor (DEMO)”,
S(ω) = 1 − − = 0. (21)
i
2
ω 2 − ωci ω✚
✚2 2
− ωce Review of Scientific Instruments 85, 02B319
(2014).
2 7. R. Hemswoth et al., “Status of the ITER
For this frequency range, one can neglect ωci in
2 heating neutral beam system”, Nuclear Fusion
the denominator of the second term and ω in the
49, 045006 (2009).
denominator of the third term. The corresponding
8. R.K. Janev, C.D. Boley and D.E. Post,
solution of Eq. (21) is known as the lower hybrid
“Penetration of Energetic Neutral Beams into
resonance
P 2 Fusion Plasmas”, Nuclear Fusion 29, 2125–2140
2 i ωpi (1989).
ωLH = . (22)
2 /ω 2
1 + ωpe ce
9. ITER Physics Expert Group on Energetic
Particles, Heating and Current Drive, Nuclear
If the second term in the denominator of Eq. (22) Fusion 39, 2495–2539 (1999).
is much larger than one (case often relevant for 10. E. Westerhof, “Electron cyclotron waves”, these
√ proceedings.
astrophysical plasmas), then ωLH ≈ ωci ωce . The
LH heating utilizes the frequency range f ≈ 1−8 GHz 11. P. Dumortier and A.M. Messiaen, “ICRH
and is mainly used for non-inductive current-drive. Antenna Design and Matching”, these
Another important factor, which influences the proceedings.
wave damping strength, is the wave polarization. 12. F. Louche and R. Koch, “The Coupling
For cyclotron resonance heating, a strong wave of Electromagnetic Power to Plasmas”, these
absorption requires a presence of the wave electric proceedings.
13. R. Koch, “The ion cyclotron, lower hybrid
field component, which rotates in the same sense as
and Alfvén wave heating methods”, these
ions for ICRF heating, and as electrons for ECRF
proceedings.
heating. In inhomogeneous plasmas (which is the
14. D. Van Esster and E. Lerche, “Kinetic Theory of
case for toroidal plasmas since B and ne vary with
Plasma Waves”, these proceedings.
the radial coordinate) – in addition to resonant
15. T.H. Stix, “Waves in Plasmas” (New York: AIP)
frequencies – there can be cutoff frequencies (defined
(1992).
as n(ωcutoff ) = 0), where the incoming wave is
16. M. Brambilla, “Kinetic Theory of Plasma Waves
partially reflected. Furthermore, plasmas support at
– Homogeneous Plasmas” (Oxford: Clarendon)
least two different modes at the same wave frequency,
(1998).
and there are special cases, when one wave mode can
17. J.-M. Noterdaeme et al., Fusion Science and
be transformed into another mode with a different
Technology 53, 1103–1151 (2008).
dispersion relation. This phenomenon is called as
18. M.-L. Mayoral et al., Nuclear Fusion 54, 033002
mode conversion and has been also actively exploited
(2014).
for efficient RF heating.

297