Magnetic Confinement Fusion - Plasma Theory: Heating

and Current Drive

Rémi Dumont

To cite this version:

Rémi Dumont. Magnetic Confinement Fusion - Plasma Theory: Heating and Current Drive. Ency-
clopedia of Nuclear Energy, Elsevier, In press, 9780128197257. �cea-03134128�

HAL Id: cea-03134128

https://cea.hal.science/cea-03134128
Submitted on 8 Feb 2021

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Magnetic Confinement Fusion - Plasma Theory: Heating and Current Drive

R. J. Dumont
CEA, IRFM, F-13108 Saint-Paul-lez-Durance, France
[email protected]

Abstract
Plasma heating is required to reach and maintain the conditions compatible with power production in
magnetic fusion devices. In addition, current drive is essential to the operation and performance of
tokamaks. This chapter reviews the main physics processes underlying heating and current drive in the
framework of the kinetic theory, in particular radiofrequency wave propagation and absorption, and
the collisional relaxation of energetic particles. The self-heating of burning plasmas is described, as
well as the main auxiliary heating methods relevant to modern fusion devices, including neutral beam
injection and radiofrequency waves in several range of frequencies.

Keywords
Burning plasmas, Coulomb collisions, Current drive, Cyclotron damping, Electron cyclotron (EC) waves, Ion
cyclotron (IC) waves, Kinetic theory, Landau damping, Lower hybrid (LH) waves, Neutral beam injection (NBI),
Plasma heating, Quasilinear theory, Radiofrequency (RF) waves
I. Introduction
The principle of heating and current drive of magnetic confinement fusion plasmas consists of
transferring energy produced by a power source to the plasma species. Whereas heating results in an
increase of their temperature, more sophisticated schemes, among which current drive, are used as a
means to control the plasma and ensure its state remains optimal with respect to fusion power
production. In present day devices, the electrical grid typically supplies this power. On the other hand,
the burning plasmas contained in future reactors will essentially be self-sufficient in terms of power:
one fifth of the energy produced by the D-T reactions comes in the form of kinetic energy of 4He (alpha)
particles, which is much larger than the average energy of any other species present in the plasma.
Upon collisional slowing-down on the background species, the alpha particles will yield their energy to
the plasma. A fraction of the fusion power produced in-situ will also be recirculated to feed external
generators used mainly for plasma control. These considerations apply to the two main classes of
reactors envisaged today, i.e. tokamaks (Zohm, 2020) and stellarators (Yamada, 2020). In modern
fusion installations, there exist two main classes of external power sources: 1) radiofrequency systems,
2) neutral beam injection systems.

Every radiofrequency (RF)-based heating and current drive method relies on an external power source
on the one hand and an adapted antenna located near the plasma edge on the other hand. How this
power travels between the source and the antenna depends on the frequency range. In an optical-like
description of the process, in all cases, an electromagnetic wave eventually leaves the antenna and
propagates to the plasma core where it is absorbed by the plasma species by two main non-collisional
mechanisms: Landau damping and cyclotron damping. These mechanisms result in the direct damping
of the wave power by one or several thermal species that will be heated as a result, and/or the creation
of energetic particle populations (i.e. particles with typical energies much larger than the thermal
energy characterizing the background plasma species) which eventually heat the plasma by collisional
relaxation.

Neutral beam injection (NBI) consists of producing high-energy ions in an external accelerator, and
neutralizing them at the end of the acceleration process. These neutral particles, being immune to the
effects of the magnetic trap in which the thermal species are confined, can penetrate deep in the
plasma. In this process, the vast majority of the energetic particles are re-ionized. Since the re-
ionization occurs when the injected particle energy is much larger than that of the thermal species, the
result is a population of energetic ions that eventually slow down through collisions with the
background particles, to which they transfer their energy.

Finally, fusion products are usually born at very high energies, the most prominent being the alpha
particles from D-T reactions born around 3.5MeV. The relaxation process is essentially similar to the
one taking place for NBI ions, i.e. results in a transfer of the fusion product energy to the background
electrons and/or fuel ions and the subsequent heating of the plasma. A plasma can be qualified as
“burning” when this self-heating process provides a source sufficient for the fusion reactions to be
maintained with little or no input required from auxiliary power sources.

Current drive is essential in tokamaks, which require the presence of a poloidal magnetic field
produced by the toroidal plasma current (Wesson, 2011). This current can be induced in the plasma by
the central solenoid, part of the poloidal circuit (ohmic current) - an inherently non-stationary process
- or self-generated by the plasma pressure gradient (bootstrap current). The ohmic and bootstrap
 currents generally need to be supplemented by additional sources to achieve stationary plasma
discharges. Furthermore, current profile control is known to be beneficial in terms of plasma stability
and performance (Jenko, 2020), but requires a level of flexibility only achievable with external current
drive sources. Some NBI systems are flexible enough for part of the energy to be injected parallel to
the magnetic field, which has the effect of driving toroidal current, albeit with a limited flexibility in
terms of current profile control. RF wave systems have also been applied for non-inductive current
drive for several decades, and are planned to be used for this application in next-step devices. In
stellarators, the magnetic field, including its poloidal components, is entirely generated by the coil
system, resulting in more modest requirements in terms of external current drive. Nevertheless, the
capability to drive non-inductive current is still desirable to compensate for the self-generated
bootstrap current.

A global summary of the process of heating and current drive is shown in Figure 1.

Figure 1: Summary of the main heating and current drive sources in modern magnetically confined fusion plasmas.

II. Kinetic theory elements

Heating and current drive results in modifications of the velocity distribution function, fs, of a given
plasma species, denoted s. From this quantity, it is possible to construct the so-called fluid moments
(Sarazin, 2020), which include the density, ns=d3v fs, the current, js=qsd3v vfs, and the temperature,
Ts=d3v mv2fs/2. To lowest order, plasma heating and current drive from external sources results in
modifications of these three fluid quantities. Clearly, the knowledge of fs is therefore the first and most
important element of the description of the physics processes involved. Furthermore, the RF wave
heating and current drive processes involve wave-particle resonances which are dependent on the
particle velocities on an individual basis (as opposed to the fluid velocity, which is a macroscopically
averaged quantity). As a result, the kinetic theory is the appropriate framework in which these
phenomena must be described.

1. Fokker-Planck equation
The Fokker-Planck equation is the classical tool employed to describe heating and current drive
processes (Brambilla, 1998). It is derived from the more generic kinetic Boltzmann equation assuming
that Coulomb collisions result in multiple small angle deviations for a given particle, a valid assumption
in the context of magnetically confined fusion plasmas. Formally, this equation takes the form of an
evolution equation for the distribution function, fs, in velocity space under the effects of collisions,
particle source and particle losses:

dfs/dt=C(fs)+Ss-Ls. (1)

Ss is the particle source (relevant in the case of NBI and/or fusion-born alphas, in which new particles
are permanently injected in the system) and Ls is a loss term describing the particles leaving the system
by various mechanisms (orbit losses, charge exchange…), while C(fs) represents the effect of collisions,
and can be written in the form:

C(fs)=-v.(<v>/c fs)+1/2 vv.(<vv>/c fs). (2)

c is a collision time. v is a differential operator with respect to components of the velocity, v. In Eq.
2, the first term represents a convection process, whereas the second term represents a random walk.
The Fokker-Planck equation therefore describes a convection-diffusion process taking place in velocity
space. The quantity F≡<v>/c is called the coefficient of dynamic friction, whereas D≡<vv>/c is the
diffusion tensor.

In its generic form - consistent with the assumptions made to obtain the Fokker-Planck equation -, F
and D involve all species present in the plasma, making C a non-linear operator which can be quite
complex to evaluate. It becomes much more tractable when assuming that the background species,
i.e. all species but the heated one(s) have Mawxellian velocity distributions, an assumption often
legitimate in the context of heating and current drive in magnetic fusion plasmas.

2. Collisional processes
As described previously, the Fokker-Planck equation (1) describes the collisional processes governing
the relaxation of any energetic particle population, and hence the energy transfers taking place
between this population and the background plasma species (Goldston and Rutherford, 1995). This
equation can be solved efficiently on any modern computer using various numerical methods, typically
in a Monte Carlo form, or using finite differences/elements. However, much insight into the relaxation
process itself can be gained by examining the relaxation of a single test ion on the background plasma
species, which are assumed fixed. From the coefficients <v> and <vv> derived to construct the
collisional term (Eq. 2), trajectories in velocity space of an ion with charge number, Zf, and mass
number, Af, injected (or born) at a given velocity, Vf, and undergoing the influence of Coulomb
collisions can be computed.

In the range of energies relevant to energetic ions (NBI or fusion-born), Vf always satisfies
vth,i<<Vf<<vth,e, with vth,i2≡2kBTi/mi and vth,e2≡2kBTi/me the (squared) thermal velocities of background
ions (of mass mi) and electrons (of mass me). Te and Ti are the corresponding temperatures and kB is
the Boltzmann constant. Using this assumption, the dominant collisional processes are found to be 1)
a decrease of the test ion velocity resulting from the collisional friction against the background species;
2) a scattering of the test ion pitch-angle (relative to the magnetic field), i.e. a modification of the
direction of its velocity with respect to its initial velocity.

A closer examination of the magnitude of these processes allows three important quantities to be
identified: 1) the Spitzer slowing-down time (in seconds), se≡6.32x1014(Af/Zf2)(Te3/2/ne)ln(), with ln()
the Coulomb logarithm and Te (resp. ne) the background electron temperature in eV (resp. density in
m-3); 2) the critical energy, Ec≡14.8(Zi2/Ai)AfTe, with Ai and Zi the background ion mass and charge
number; 3) the thermalization time, th≡seln(1+(Ef/Ec)3/2)/3. The Spitzer time describes the rate of
change of the test particle velocity caused by the friction on background electrons, i.e. dvf/dt=vf/se
(i.e. assuming only electrons are involved in the slowing-down process). The critical velocity
corresponds to the limit below which pitch-angle scattering and drag by background ions exceeds the
scattering and drag caused by electrons. When the test ion reaches Ec, both the background electrons
and ions contribute equally to its slowing down. However, the fact that the masses of the test and
background ions are not very different also induces significant pitch-angle scattering by the latter. The
corresponding decrease in the test particle energy does not result in background plasma heating, but
rather in an increased spread in the energetic population velocities. Finally, the thermalization time
corresponds to the delay for the test ion to reach the thermal velocities. It should be noted that th
andse can be quite different, as illustrated by the data shown in Table 1.

3. Quasilinear description of RF heating

The first step towards describing plasma heating by RF waves is to solve the wave equation for the
wave electric field, E. The wave magnetic field, B, can always be deduced from E using Maxwell’s
equations. Assuming a time-harmonic field varying at an angular frequency,  (imposed by the
generator), i.e. E exp(-it), with t the time, the wave equation is a straightforward consequence of
Maxwell’s equations, and is written here in the symbolic form:

xxE-2/c2(E+iJp/0)=i0Jant. (3)

In this expression,  is a differential operator (in real space), 0 is the vacuum permeability, 0 the
vacuum permittivity, and c is the speed of light. Jant represents the boundary condition imposed by the
antenna. Despite its apparent simplicity, this equation is quite complicated because the response
plasma current, Jp=Jp(E), has a complex, temporally and spatially non-local, dependence on the wave
electromagnetic field. A first simplification occurs assuming that this dependence is linear, a hypothesis
generally valid in the context of plasma heating and current drive, where the electromagnetic field is
a perturbation for the magnetically confined plasma.

Another source of difficulty related to Eq. 3 is the fact that Jp depends itself on the distribution function
of the various plasma species, including the species which directly absorb the injected power. In other
words, it is often necessary to solve the wave equation and an equation to describe the modification
of the distribution function in a self-consistent fashion. This two-step approach relies on the fact that
the timescales related to the wave propagation and damping are small compared to the timescales
related to the secular modifications of the distribution function and Coulomb collisions, which
correspond to the macroscopic modifications of the plasma caused by the wave power.
 Including RF wave effects on the distribution function requires additional assumptions with respect to
those made to derive the Fokker-Planck equation (Eq. 1). The quasilinear equation is obtained from
the Vlasov equation assuming that the non-linear effect of the wave on the distribution function is the
product of interactions taking place between the wave field, as deduced from the linearized wave
equation (Eq. 3) on the one hand, and of the linear plasma response on the other hand, hence the
name quasilinear. This procedure requires an averaging over space (resp. time) scales larger than the
wavelength,  (resp. wave period), to eliminate all terms varying on the small space scale, , and fast
time scale, . This quasilinear equation - customarily also referred to as Fokker-Planck equation -
therefore applies to Fs, the time and space-averaged version of fs. Following this procedure, it can be
written in the form:

dFs/dt=C(Fs)+Q(Fs)+Ss-Ls. (4)

Q is the quasilinear operator, and represents the effect of the waves interacting with the particles of
species s on the corresponding distribution function, Fs. In the framework of the quasilinear theory,
this term also takes the form of diffusion/friction term involving a quasilinear diffusion tensor, Dql:

Q(Fs)=v.(Dql.vFs). (5)

Depending on the level of sophistication of the description used, Q can have a rather complicated form.
Important features of the quasilinear operator can nevertheless be isolated by writing any of its
components in the symbolic form:

Dql=A∑p|d(p)(E)|2(-pcs-k//v//). (6)

In this expression, A is a constant. d(p)(E) is a differential operator acting on the electromagnetic field
and k// is the projection of the wave vector in the direction parallel to the magnetic field lines. These
two elements are essentially representative of the wave field. cs=qsB0/ms is the (algebraic) cyclotron
frequency, i.e. the frequency of the cyclotron motion of the considered particles (with charge, qs, and
mass, ms) around the magnetic field line. cs and v//, the particle velocity along the field line, are
characteristics of the interacting particles trajectories. p, the harmonic number, is an integer which can
have either sign, or be 0.(-pcs-k//v//) is non-zero only when the particles are resonant with the
wave, i.e. when the wave-particle resonance condition, =pcs+k//v//, is satisfied.

III. RF wave propagation and damping

In RF heating processes, generators located at some distance from the plasma confinement device
generate oscillating currents. Depending on the frequency range, the corresponding power is
transported to an antenna located inside the vacuum vessel, usually close to the plasma edge (i.e.
within several cm). The wave coupling is the process by which the power emanating from the radiating
structure in the form of a vacuum electromagnetic wave is transferred to the plasma species. This
electromagnetic wave can excite plasma waves of various types: 1) Electromagnetic waves are waves
for which the propagation does not rely on the plasma species. 2) Electrostatic waves are characterized
by an oscillating electric field in the direction of propagation. This electric field induces an oscillating
plasma current and associated space charge, which itself is responsible for a response electric field
further away in the plasma. 3) Magnetohydrodynamic (MHD) waves are characterized by an oscillating
magnetic field which induces a plasma current, itself inducing a magnetic field further away in the
plasma. It should be emphasized that such a separation is an idealized view: actual plasma waves
generally involve several of these processes at the same time (Stix, 1992). As a result, the separation
between the reactive part of the wave power, reversibly exchanged between the electromagnetic field
and the plasma species and therefore linked to wave propagation, and the resistive part of the wave
power, irreversibly transferred to the plasma species and therefore responsible for plasma heating, is
in practice a delicate task (Swanson, 2003).

1. Wave propagation
There exist several methods to solve the wave equation, Eq. 3. When the wavelength  remains much
smaller than the typical space scales characterizing the plasma equilibrium quantities, LB, the WKB
approximation is applicable, allowing Eq. 3 to be solved by ray-tracing or beam-tracing (quasi-optical
methods). On the other hand, regimes in which  becomes comparable to, or larger than, LB, or in
which the wave undergoes reflection or resonance phenomena during its propagation make it
necessary to solve Eq. 3 directly, by resorting to a full-wave approach. Although more intensive
numerically, modern computers allow this type of calculation to be performed in a routine fashion,
including for higher frequency waves.

However, in all cases, considerable insight can be gained by analyzing the dispersion relation
corresponding to the wave equation. This relation is obtained by assuming that the plasma is stationary
and homogeneous. Whereas the former assumption is relatively safe in the context of plasma heating
by RF waves, the homogeneity hypothesis is often questionable. Despite this limitation, the dispersion
relation is always very useful to analyse wave propagation in plasmas. Writing the local electric field
as a monochromatic plane wave characterized by wavevector, k, i.e. E(r,t) exp(i(k.r-t)), the
dispersion relation takes the generic form, D(,k)=0, and therefore expresses an implicit relation
between the wave frequency, and wavevector, k, or equivalently the refractive index, N≡kc/.
Usually, multiple solutions to D(,N)=0 at a given point in the medium are possible, indicating the
potential presence of several modes, each one characterized by a different dependence of N on the
plasma parameters (density, magnetic field…). Once a given solution (or mode) is chosen, the
dispersion relation can be used in the wave equation to deduce the wave polarization, i.e. the ratio
between the various components of the wave electromagnetic field. Among these components, the
parallel, E//, left-handed, E+, and right-handed, E-, turn out to be of particular importance because they
physically correspond to a wavefield oscillating in a direction parallel to the magnetic field (E//), and
rotating in the same direction as ions (E+) or electrons (E-).

In magnetically confined plasmas, the projection of N along the magnetic field, N//, is essentially
determined by the antenna geometry, and usually does not vary much as the wave propagates. The
perpendicular component, N, on the other hand, largely depends on the plasma parameters. It is
instructive to investigate wave propagation in terms of N in the context of the cold plasma theory, in
which all species are assumed to have zero temperature. In such plasmas, the dispersion relation can
be written as D(,N)=A(x)N4+ B(x)N2+C(x), where the coefficients A, B and C are function of the local
plasma parameters (Stix, 1992), parametrized here by the variable x. The consequence is that only two
modes can exist at a given point, x, in a cold plasma. Furthermore, there may exist points verifying
N2(x)=0, called cut-offs. In this situation, the wave can only propagate up to the cut-off where it is
reflected. Points such that N2(x)∞ are called resonances.

The inclusion of finite temperature effects in the model regularizes the singular behavior observed at
cold resonances. Wave absorption, a phenomenon absent in the cold plasma theory, takes place and
results in an irreversible power transfer through the resonance mechanism involving the wave field
and the particle motion. Retaining thermal effects also has the consequence of allowing more than
two modes to propagate in the plasma. In particular, it is possible to encounter situations for which
two modes characterized by N1 and N2 propagating independently reach a point such as N1=N2. In
this case, part of the wave power in a given mode can be partially transferred to another mode, in a
process called mode-conversion.

In magnetic fusion plasmas, these three processes i) wave cut-off, ii) resonant wave absorption and iii)
mode conversion, are quite prevalent and therefore need to be included in any description of wave
propagation and absorption (Stix, 1992).

2. Wave absorption
As discussed previously (see section “Quasilinear description of RF heating”), wave-particle resonance
occurs when parameters characterizing a wave (, k//) and parameters characterizing a given plasma
species are such that a resonance condition is fulfilled. k, on the other hand, is determined by the
dispersion relation. Typically, such a resonance occurs when the wave frequency matches the local
gyration motion of electrons or ions, =cs+k//v// (cyclotron resonance), one of its harmonics,
=pcs+k//v// (harmonic cyclotron resonance, with p>1) or the parallel particle motion, =k//v//
(Cerenkov resonance). A synchronization process takes place between the wave-field and the particle
motion. However, this process is reversible by nature. The transformation of reactive power to
resistive power through the wave-particle resonance requires one or several mechanisms responsible
for decorrelating the field oscillation from the plasma particle response. These processes are often
referred to under the generic term of phase-mixing mechanisms. One such decorrelation process is
the Coulomb collisions themselves. By their stochastic nature, collisions are clearly able to destroy any
wave/particle correlations. However, experimental evidence pointed to the possibility of wave
damping even in collisionless plasmas, triggering an intense effort to interpret these results and
identify non-collisional damping mechanisms.

a) Landau damping
Non-collisional damping has been the center of vigorous debates. Several mathematical derivations
have been put forward to describe this phenomenon. However, L. Landau is to be credited for the
unambiguous explanation of this effect in 1946 (Landau, 1946)(Ryutov, 1999). The mathematical
demonstration is rather subtle, and involves causality as a key element. The underlying physics idea
can be understood by considering a particle with initial parallel velocity v//0 in the presence of an
electromagnetic field characterized by phase velocity, v=/k//. A resonance occurs when the Cerenkov
resonance condition, =k//v//0, is fulfilled. More precisely, during the process, particles with velocities
near the wave phase velocity feel the wave effect, and see their motion modified in the
synchronization process with the field. It can then be shown that in this case, the particle is either
accelerated or decelerated with equal probability. As such, no net acceleration is imparted by the wave
power. However, when considering an assembly of particles, the slower particles, i.e. those with
v//0</k//, which are accelerated, tend to feel the wave resonance with more intensity. On the other
hand, those decelerated tend to resonate even less with the wave. Statistically, therefore, wave power
will be employed to accelerate slower particles. The symmetrical situation involves faster particles, i.e.
those with v//0>/k//. In contrast with the previous case, the decelerated particles will become more
resonant than the accelerated ones. As a result, kinetic energy from the particles will tend to amplify
the wave. However, fusion plasmas are typically characterized by Maxwellian distributions, which have
a decreasing dependence with parallel velocity, v//. The consequence is that slower particles are more
numerous than faster particles, so that the net result of the two processes described previously is an
absorption of the wave by the particles. This explains that , the wave damping rate, which
characterizes the field attenuation with time, i.e. E exp(-t), depends linearly on the slope of the
distribution function considered at the wave-particle resonance, i.e.  dfs/dv|v. The result of this
interaction on the distribution function of the interacting species is described by Eq. 4, and takes the
form of a local flattening in velocity space around the resonance.

b) Cyclotron damping
Cyclotron damping is related to the process of Landau damping. The resonance condition, however,
involves the particle gyromotion, and can be written as =pcs+k//v//, with p a non-zero integer. The
cyclotron resonance occurs when the wave frequency matches the cyclotron frequency cs (p=1), i.e.
the frequency characterizing the particle gyro-motion or one of its harmonics (p>1). The last term,
k//v//, is the Doppler shift correction caused by the parallel motion of the particles along the field lines.
Although it bears some similarities with Landau damping, the dependence of the cyclotron frequency
with the confinement magnetic field magnitude, B0, induces essential differences in the resonance
phenomenon. By construction, B0 has a spatial dependence, proportional to 1/R (the distance to the
torus revolution axis) in tokamaks – but more complex in stellarators. The result is that the resonance
only occurs in a spatially localized region. The particles having periodical motion, the resonance occurs
at multiple times as the particle travels through the plasma, which introduces another possible
decorrelation mechanism in the system. The other fundamental difference is that the cyclotron motion
is, by nature, perpendicular to the magnetic field. The resulting velocity-space diffusion occurs mainly
in the perpendicular direction, which tends to increase the perpendicular particle velocities. The
heated distribution functions in velocity space can therefore exhibit pronounced anisotropies, with a
perpendicular energy content often significantly exceeding the parallel energy content.

3. Wave-induced current Drive

The process of RF-induced current drive (CD) is inherently linked to electron Landau damping. A wave
characterized by angular frequency, , and parallel wavevector, k//, can resonate with electrons having
a parallel velocity fulfilling the Cerenkov resonance condition, v//=k//. The consequent absorption of
an energy quantum, E=ħ, by an electron results in an increase of its parallel momentum,
p//=ħk//=k//E/. Therefore, an injected wave characterized by an asymmetric spectrum in terms of
k// will impart momentum to the electrons, thereby inducing a net parallel current.

As illustrative as this physics picture may be, it must be stressed that the mechanism of momentum
transfer from the wave to the plasma described previously is not the dominant one in RF current drive
applications. Indeed, it turns out that even waves imparting only perpendicular momentum to the
electrons can drive current in a relatively efficient fashion. This results from the fact that wave-induced
current drive relies for the most part on the variation of the plasma collisionality with the energy of
resonant particles. Therefore, creating a population of electrons characterized by an excess of
perpendicular energy can be exploited for current drive purpose, as long as the distribution function
is asymmetric for the parallel velocity. This process is known as the Fisch-Boozer mechanism (Fisch,
1987).

A figure of merit of any current drive method is the amount of current driven, j, for a given amount
of power expended, p. If we consider an electron receiving parallel momentum, p//, from the wave,
 the corresponding current increase is j=-ep///me, with e the absolute value of the electron charge
and me the electron mass. Collisions, on the other hand, tend to oppose this momentum increase, a
process which can be characterized by the counteracting force, F=–ep//, with e an effective collision
frequency. By identifying the wave power spent to the work variation required to maintain the current
drive process, we obtain the elemental current drive efficiency:

j/p=e/meev//. (7)

In order to maximize the efficiency, it would therefore seem natural to target thermal electrons,
characterized by relatively low values of v//. Early current drive schemes, using Alfvén waves, were
based on this idea. However, a large fraction of thermal electrons is trapped in the magnetic mirrors
constitutive of the magnetic confinement structure, and therefore do not complete full revolutions in
the toroidal direction. Consequently, they are unable to carry any parallel current and the resulting
efficiency is too low to be of any interest. By contrast, in the range of velocities characterizing
superthermal (or energetic) electrons, the effective collision frequency, e, varies as(1/v//)3, so that by
virtue of Eq. 7, the efficiency increases with v//2. Finally, superthermal electrons also suffer less from
trapping effects than thermal electrons. As a result, modern RF-based current drive schemes, i.e. those
exploiting lower hybrid and electron cyclotron waves, are based on energetic electrons.

Finally, it should be noted that whereas Eq. 7 is interesting from a physics standpoint, it is hardly
adequate to characterize the global efficiency of a given CD system, i.e. to obtain an estimate of the
level of driven current corresponding to the power available on a given fusion device. For this purpose,
it is worthwhile realizing that the absorbed power (in W) scales as Pabs~2Ra02pabs, whereas the total
current (in A) scales as ICD~a02j, with R the major radius, a0 the minor radius, pabs the power density
(in W.m-3) and j the current density (in A.m-2). Accordingly, a useful quantity is the normalized current
drive efficiency, a global figure of merit valid for any CD scheme and defined as CD≡neRICD/Pabs, with
ne the appropriately averaged electron density.

IV. Plasma heating & current drive by energetic ions

Regardless of the heating method employed, collisions often play an essential role since they are
responsible for the transfer of power between the energetic particles and the thermal plasma species.
Indeed, one of the most prevalent processes in modern heating and current drive applications involves
the collisional relaxation of energetic ions, injected by the NBI system, born from fusion reactions or
generated by RF acceleration of initially thermal species. The transfer of energy occurring between
these energetic ions and the background plasma results from the various collisional processes at play.
Their respective influence depends on the initial energies of the injected particles, Ef, on the Spitzer
slowing-down time, se, and on the critical velocity, Ec. It must be pointed out that energetic ions (NBI-
driven, RF-driven or fusion products) can trigger instabilities, potentially inducing a radial redistribution
of these ions. The impact of these processes on heating, current drive efficiency and fusion
performance, as well as control schemes aimed at minimizing them (including using RF waves), is
currently an active topic of experimental and numerical studies to prepare for the operation of next-
step devices (Gorelenkov and Sharapov, 2020).

1. Neutral beam heating and current drive

Neutral beam injection is an auxiliary heating method which involves the production of high-energy
ions in an accelerator, and their neutralization as they are ejected towards the confined plasma. These
neutral particles are immune to the effects of the magnetic field in which the plasma is confined, and
can thus penetrate to the plasma core. They are progressively re-ionized as they travel through the hot
plasma. Since they are injected at an energy much larger (typically in the range 50-150keV in ongoing
devices, and 1MeV in ITER) than the typical confined plasma particle energies, they slow down through
collisions with these particles. Adapted numerical codes (Schneider et al., 2011) taking into account
the beam and plasma parameters are required to compute the neutral beam and corresponding power
deposition. The NBI system installed in JET, as well as an illustration of a typical power deposition
profile in the plasma is shown in Figure 2.

Consistently with the collisional relaxation process discussed in section “Kinetic theory elements”, the
result of plasma heating by NBI ions depends on the background plasma and injected ion parameters.
As an illustration, NBI ions with typical injection energies of Ef~100keV are used in JET and result in
dominant thermal ion heating (see Table 1). In ITER, on the other hand, the plasma size and density
make it necessary to resort to much higher beam energies to deposit the beam power in the plasma
core. Negative NBI ions with Ef~1MeV have been adopted to fulfil this requirement, and tend to create
a balanced heating on thermal electrons and ions, with little dependence on whether the background
plasma consists of D ions only, or of mixed D-T fuel.

Figure 2: NBI heating in the JET tokamak. This device is equipped with two separate NBI injectors located in octants 4 and 8,
each consisting of eight beams arranged in a tangential bank and a normal bank (Source: EUROfusion). In the inset is shown
an illustration of the power deposition profile in the plasma corresponding to the tangential bank of the injector located in
octant 8.
 Background plasma Initial Critical Spitzer Thermalization Fraction of power to
/ test particle energy energy time [s] time [s] ions / to electrons
[keV] [keV]
JET D / D NBI 100 187 1.45 0.16 0.87 / 0.13
JET DT / DT NBI 100 206 1.81 0.18 0.89 / 0.11
ITER D / D NBI 1000 466 2.64 1.25 0.53 / 0.47
ITER DT / D NBI 1000 413 2.64 1.34 0.50 / 0.50
Table 1 : Main parameters relevant to the collisional relaxation of NBI ions in JET D, JET DT, ITER D, and ITER DT plasmas.

Depending on the NBI system geometry, in some situations, it is possible to inject energetic ions with
a significant component of the velocity direction parallel to the confinement magnetic field, i.e. able
to impart a significant amount of parallel momentum to the plasma. As a result, a finite level of toroidal
current is driven, a scheme referred to as Neutral Beam Current Drive (NBCD). Assuming beam ions
with mass, mf, and density, nf, the parallel momentum injected in the system is mfnf<v//>, the average
being performed on the (injected) energetic NBI ion distribution. Because of collisional relaxation, the
beam ions will transfer this momentum to the background species through friction. Assuming, for
simplicity a situation where Ef>>Ec, this relaxation essentially takes place on electrons. The subsequent
force on electrons in the parallel direction tends to increase their parallel momentum. In a stationary
regime, the friction with thermal ions compensates this variation of parallel electron momentum.
Taking into account the respective relaxation times of these processes and plasma quasi-neutrality, it
can be shown that the resulting driven current takes the form JNBCD=Jf(1-Zf/Zi), with Jf=Zfnfe<v//> the
current carried by the beam ions, and Zf (resp. Zi) the charge number of beam (resp. thermal) ions. In
many situations, Zf and Zi are comparable, which induces a large reduction in the NBI driven current
because of the friction exerted by the background species on the energetic ions. A more accurate
description of the process shows that the current drive efficiency predicted is generally larger than this
analytical estimate. A first correction consists of taking into account trapped electron effects, which,
in the case of NBCD, reduce the amount of electron current resulting from the cancellation of the beam
current, and therefore entail an increase of the net efficiency (Fisch, 1987). The largest efficiency
obtained to date is CD=0.155x1020A.W-1.m-2 using 360keV negative ion NBI in the JT-60U tokamak
(Inoue et al., 2020). Predictions indicate that efficiencies of the order CD~0.2-0.4x1020 A.W-1.m-2 should
be achievable in ITER (Gormezano et al., 2007).

Finally, it must be pointed out that NBI also plays an essential role in driving toroidal rotation of the
plasma, which is known to have a favorable effect on confinement (Jenko, 2020) and MHD stability
(Lao et al., 2020) in ongoing devices. Predictions indicate that this effect will be less significant in
reactor-scale devices, however. A larger plasma volume and density translates into a larger moment
of inertia, whereas the higher beam ion energy results in a reduction of the injected torque.

2. Burning plasmas
Burning plasmas consist of a mix of D-T fuel, which produces alphas born at 3.5MeV able to heat the
background species in a self-sufficient fashion (or nearly so). So far, only the tokamaks TFTR (USA) and
JET (European Union) have operated with mixed D-T fuels. In these devices, the alpha power
production remained small compared to auxiliary heating, making the effect of self-heating by alphas
difficult to isolate. ITER is the first magnetic fusion device expected to contain a genuine burning
plasma. In Table 2 are compared the situations in JET and in ITER with respect to plasma heating by
the alphas. In the two cases, alphas predominantly heat the background electrons. However, the larger
 density (and thus collisionality) in ITER high-performance scenarios results in a more balanced heating
partition between the electrons and the fuel ions, which is favorable for plasma fusion performance.

Background plasma Initial Critical Spitzer Thermalization Fraction of power to

/ test particle energy energy time [s] time [s] ions / to electrons
[keV] [keV]
JET DT / alpha 3500 330 0.73 0.86 0.17 / 0.83
ITER DT / alpha 3500 826 1.32 1.00 0.35 / 0.65
Table 2 : Collisional plasma heating resulting from D-T alphas in JET and in ITER.

V. Plasma heating & current drive by radiofrequency waves

1. Ion Cyclotron (IC) Range of Frequency
In the Ion Cyclotron Range of Frequencies (ICRF, 20-80MHz), the antenna consists of a set of metallic
straps enclosed in a box located at the edge of the magnetized medium and is capable of exciting a
wave propagating in the plasma. In this frequency range, the system plasma-antenna must be
considered as a global electrical circuit. This has the practical consequence that the antenna needs to
be adapted to the plasma, by ensuring that their respective impedances match. In general, coupling
ICRF power to the plasma is a delicate problem and a considerable effort is being devoted to this topic.

Examination of the dispersion relation in this range of frequencies reveals the existence of two waves,
the so-called slow wave and fast magnetosonic wave. The slow wave, characterized by N2~-pe2/ci2
is evanescent with a small wavelength (1mm), which means that it is useless for plasma heating
since it cannot propagate away from the antenna vicinity. Here, ps2=nsqs2/ms0 is the squared plasma
frequency and ci=ZieB0/mi is the cyclotron frequency of the main ions. It should be noted that the
slow wave plays an important role in the problem of wave coupling, and in the development of
performance limiting electrical sheaths, which can occur in some situations. The fast wave dispersion
relation for a cold plasma is given by N2~(N//2-R)(N//2-L)/(S-N//2) with R~-(c/va)2ci/(-ci),
L~(c/va)2ci/(+ci) and S~-pi2/(2-ci2), va being the Alfvén velocity . It is clear from these expressions
that the fast wave exhibits two cut-offs: the right cut-off, N//2=R, the left cut-off, N//2=L; and a
resonance called the Alfvén (or hybrid) resonance, N//2=S. The presence of the right cut-off implies that
the fast wave does not propagate below a finite density, ne~1x1018m-3, which corresponds to a location
significantly inside the plasma scrape-off layer. In other words, like the slow wave, the fast wave is
evanescent near the plasma edge. However, with a space decay typically of the order of 10cm, a
significant fraction of the power can tunnel through the evanescence layer to reach the right cut-off,
and propagate onward to the plasma core. The left cut-off, on the other hand, is located well inside
the plasma core. Once coupled to the plasma, the fast wave is able to propagate between these two
cut-offs, where it remains trapped until dissipation mechanisms damp the power.

The dominant damping mechanism of the fast magnetosonic wave is cyclotron damping by ions
fulfilling the resonance condition, =pci+k//v//. The corresponding generic form of the quasilinear
diffusion coefficient (see Eq. 6) is

Dql=A∑p|E+Jp-1(ki)+E-Jp+1(ki)|2(-pci-k//v//), (8)

with E+ (resp. E-) the left-handed (resp. right-handed) component of the electric field. Jp is the Bessel
function of order p and i=v/ci the (heated) ion Larmor radius. This expression shows that to lowest
order, damping is determined by the left-handed component of the electric field, rotating in the ion
direction. Furthermore, for small arguments, Jp(x)~xp. In other words, fundamental heating (p=1) is in
principle very efficient since Dql is at lowest order independent of ki. From a wave propagation
standpoint, the situation described previously would seem to be ideal, since the cyclotron layer =ci,
around which the fundamental damping occurs, is always located between the two cut-offs. However,
it turns out that the dispersion relation also dictates that the ratio of the left to right polarized
components of the wave field be such as E+/E-=(N//2-R)/(N//2-L), which goes to zero as ci. The result
is that fundamental damping of the fast wave is a weak phenomenon, forbidding its applicability in a
“pure” plasma, i.e. a plasma including a single ion species.

Figure 3: Depiction of IC wave heating in a tokamak plasma consisting of 95% deuterium+5% hydrogen. (a) General principle
illustrated via a schematic wave propagating from the antenna across the plasma magnetic field surfaces (gray lines). The red
color corresponds to locations where significant wave damping occurs; (b) Full-wave calculation of the left-handed electric
field (shown in colour contours). The vertical dashed lines show the approximate location of the fundamental ion cyclotron
resonance layer. Efficient wave absorption by the minority ions results in a low amplitude residual electric on the left of the
resonance.

This problem can been overcome by using a harmonic of the cyclotron resonance, i.e. p=2 or more.
This scheme, named harmonic heating, does not suffer from the same limitation as fundamental
damping and can be used even in pure plasmas. However, Eq. 8 shows that the corresponding term is
now proportional to ki, which depends itself on the pressure of the heated ion. To be efficient,
second harmonic heating thus requires an already significant temperature of the heated ions.
Nevertheless, second harmonic IC resonance heating (ICRH) has been successfully employed in various
tokamaks as a heating method, and second harmonic tritium heating is a scenario which is planned to
be used in ITER (Gormezano et al., 2007).

A possibility to exploit fundamental heating despite the limitation exposed previously is to introduce
a small fraction (typically 2-8%) of a different ion in the plasma in addition to the majority ion and to
perform fundamental cyclotron heating of this minority ion, hence the denomination: minority heating
regime. In this case, it is possible to show that the left cut-off, N//2=L, and the hybrid resonance, N//2=S,
are located very close to each other. The advantage is that the field polarization is now mainly imposed
by the majority ion, and therefore E+ does not cancel at the fundamental minority resonance. The
second advantage is that the fundamental cyclotron layer, =ci, is very close to the cut-off/hybrid
resonance couple. This results in a locally large electric field magnitude and associated stored reactive
power. The outcome is efficient heating of the minority species, and this scheme has been routinely
used in magnetic fusion devices for decades. The most widespread scenario is hydrogen heating in
deuterium plasmas, with a typical hydrogen concentration of 3-5%. Since the power per absorbing ion
is quite large, minority heating schemes tend to create very energetic ion populations and one relies
on collisional relaxation on thermal electrons and/or ions for plasma heating. In ITER, fundamental
heating of helium-3 (3He) ions is currently one of the envisaged scheme to heat DT plasmas with ICRF
waves. It is even envisaged to initiate the heating phase by injecting a small quantity of 3He ions in the
plasma. Minority damping, which is to lowest order independent of the heated ion temperature, will
dominate and increase the ion temperature. Once the tritium temperature is sufficient and as the 3He
concentration naturally decreases, second harmonic tritium damping (located at the same position,
since =c3He=2cT) is expected to take over, thus allowing smaller quantities of the expensive 3He gas
to be used. Figure 3 illustrates the principle of ICRF heating in a tokamak such as JET, as well as an
actual calculation of the left-handed electric field in a typical hydrogen minority heating scenario
performed with a full wave code adapted to the description of waves in the Alfvén and IC range of
frequencies (Dumont, 2009).

More recently, the so-called 3-ion scheme has been proposed and tested in several tokamaks (Kazakov
et al., 2017). This involves the targeted heating of a third ion species characterized by charge number,
Z, and mass number, A, in a plasma with two ion species (Z1, A1, Z2, A2). It can be shown that when the
condition Z1/A1<Z/A<Z2/A2 is satisfied, very efficient damping by the third species can be obtained, as
long as it is present in very small concentrations (typically 0.1-1%). This scheme has been proposed to
generate energetic ions in stellarators in order to experimentally document their confinement, and to
provide efficient ICRF heating during phases of ITER operation at reduced magnetic field. It is also
envisaged to use this scheme for ITER D-T plasmas, exploiting the natural presence of the beryllium
(Be) impurity in the plasma. In a D-T mixture, Be satisfies ZT/AT=1/3<ZBe/ABe=4/9< ZD/AD=1/2. The
advantage is that the heavy Be ions tend to heat the fuel ions by collisions, rather than the electrons,
which is beneficial in terms of fusion power production. It must be pointed out that, owing to its
relatively novel character, this scheme is not as well established as the more traditional minority or
harmonic heating schemes. Nevertheless, its flexibility and the fact that it proposes to exploit
impurities naturally present in the plasma make it an appealing tool for future magnetic fusion devices.

A comparison of thermal plasma heating by typical RF-accelerated ions is shown in Table 3.

Background plasma Initial Critical Spitzer Thermalization Fraction of power to

/ test particle energy energy time [s] time [s] ions / to electrons
[keV] [keV]
JET D / H 2000 93 0.73 1.11 0.09 / 0.91
JET D / He3 700 280 0.54 0.29 0.49 / 0.51
ITER DT / He3 700 619 0.99 0.26 0.72 / 0.28
ITER DT / Be 700 1858 0.74 0.05 0.92 / 0.08
Table 3 : Collisional plasma heating resulting from RF-accelerated ions in JET and in ITER.

By increasing the minority concentration to ~10-15%, the fundamental cyclotron layer, =ci, on the
one hand and the cut-off/resonance structure on the other hand move further apart. In this case, it is
possible to have a substantial part of the fast wave power tunnel through the evanescence layer and
undergo a mode conversion towards the pressure driven Ion Bernstein Wave (IBW). In this mode
conversion regime, one relies on the damping of the small wavelength IBW by the thermal electrons
to obtain a localized electron heating source. It should be pointed out that despite good results
obtained in several tokamaks, this scheme remains difficult to implement in practice, and requires
delicate tuning of the magnetic field, minority concentration and antenna toroidal phase spectrum.

The reader should be aware that the physics of IC waves is very rich and many topics related to their
use in fusion plasmas have been omitted. Among these, Fast Wave Electron Heating (FWEH) has been
successful in some past experiments, and relies on the damping of the fast wave by the thermal
electrons. FWEH is achieved by excluding all relevant cyclotron resonance layers by a careful choice of
frequency/magnetic field combination. Using an asymmetric toroidal phase spectrum, furthermore,
allows Fast Wave Current Drive (FWCD) to be achieved. FWEH is not envisaged in ITER, because the
exclusion of all relevant cyclotron layers does not appear to be achievable in such a large device. FWCD,
on the other hand, can be interesting since a moderate level of central non-inductive current can be
obtained as a by-product of the IC heating wave damping on the electrons, simply by modifying the
antenna current distribution and thus the toroidal phase spectrum. Other possibilities include MHCD
(Minority Heating Current Drive), which relies on finite orbit width effects to drive a current related to
the energetic ions; MCCD (Mode Conversion Current Drive) obtained by exciting a mode-converted
wave with an asymmetric spectrum; HHFW (High Harmonic Fast Wave heating) relying on high
frequency fast waves for localized electron heating/current drive, using traveling wave antennas to
optimize wave coupling and toroidal directivity. To this day, these schemes are not considered as
workhorse ICRH scenarios for ITER.

2. Lower Hybrid (LH) Range of Frequency

Between 1-10GHz, Lower Hybrid (LH) waves can be used for non-inductive current drive in tokamaks.
In this frequency range, ci<<~LH<ce, with LH the LH resonance frequency LH~pi/(1+pe2/ce2)1/2.
The cold dispersion relation reveals two waves: the fast wave already discussed for IC waves, is useless
for plasma heating because the evanescence region between the antenna and the R-cutoff is too wide
compared to the wavelength in this range of frequencies; the slow wave, on the other hand, is
characterized by the approximate dispersion relation N2~N//2(mp/me)LH2/(LH2). This expression
shows that for given a magnetic field and density, the LH wave propagates at a finite angle, =(),
with respect to the magnetic field direction. This kind of configuration is known as a resonance cone
wave (Swanson, 2003). A further examination of the dispersion relation shows that the electric field
polarization is essentially along the wave vector. The LH wave is therefore an electrostatic wave, and
does not have any magnetic field component associated to it.

Originally, the idea was to target ion heating by damping of the wave near the LH resonance, hence
the name of this scheme. Indeed, as discussed in section “Wave propagation”, this cold resonance is
characterized by k∞, so that efficient heating of ions at relatively low velocities, v~/k, (i.e.
thermal) by perpendicular Landau damping was anticipated. However, experimental evidence pointed
to several phenomena preventing this damping mechanism from taking place. For instance, the large
values of N attained when approaching this resonance result in the parametric decay of the wave, a
non-linear effect responsible for a degraded wave penetration in the plasma. Fortunately, it turns out
that efficient wave damping by electrons satisfying the Cerenkov resonance, v//~/k//, is also
achievable in this range of frequencies, provided a slow wave with N//2>1 can be excited from an
external antenna. Like the fast wave, the slow wave is subject to a low-density cut-off at ~pe, but
the corresponding value is of the order 1017m-3 in typical conditions. As a result, the evanescence
region is quite thin compared to the wavelength in vacuum, and it is possible to design efficient slow
wave antennas with the required toroidal periodicity.

Once the slow wave has entered the plasma, it propagates towards the plasma center, with a
corresponding increase of N. In this range of frequencies, however, if N// is smaller than a given N//*
which depends on density and magnetic field, a confluence between the fast and slow waves takes
place in the plasma. Unless damping takes place before the slow wave reaches this confluence point,
the converted fast wave propagates back towards the antenna. For all practical purposes, therefore,
N//> N//* places a restriction on useful parallel refractive indices, densities and magnetic fields required
for the slow wave to reach the plasma core (Brambilla, 1998), and is known as the Stix-Golant
accessibility condition.

Typical values of the parallel refractive indices achieved with standard LH antennas are N//,ant~2. For a
plasma characterized by a core electron temperature, Te=10keV, the range of velocities of the resonant
electrons is therefore found to be v///vth,e=c/(N//vth,e)~2.5. For a Maxwellian distribution function, the
corresponding number of electrons is very small. LH wave damping therefore relies on the changes of
N// as it propagates. Indeed, it can be shown that various effects, among which are toroidicity and
spectral broadening by density fluctuations, cause N// to vary during the propagation (Peysson et al.,
2017). At some point, N// becomes sufficiently large for Landau damping by electrons with parallel
velocities, v//~vth,e, to take place. The regime in which the rays representative of the LH wave have to
travel a long distance around the torus before damping is known as the multi-pass regime, and
characterizes most past and current experiments using LH waves. If, on the other hand, Te and/or N//
are sufficiently large from the onset, the wave can be absorbed on its first pass onward the plasma
core, a situation referred to as the single-pass regime, which is what would happen in a next-step
tokamak containing a very hot plasma. Figure 4 illustrates the difference between these regimes in
two discharges conducted in the Tore Supra tokamak, as computed with a Fokker-Planck code
associated to a ray-tracing code (Decker, 2014).

Figure 4: LH wave current drive in the Tore Supra tokamak. Ray trajectories are shown projected in the poloidal cross-
section of the tokamak, but also propagate toroidally around the plasma. (a) In a relatively dense and cold plasma (volume
averaged density <ne>~3.8x1019m-3, central temperature Te0~2.4keV), the LH wave is in the multi-pass regime and the
rays propagate on large distances before damping takes place. (b) In a less dense and hotter plasma, on the other hand
(<ne>~1.3x1019m-3, Te0~5.7keV), damping occurs relatively early during the propagation and the single-pass regime is
approached.
In all cases, absorption initially occurs at large values of N//, and involves electrons close to the thermal
velocity. An energetic electron tail is progressively created starting from these thermal electrons,
eventually extending up to superthermal electrons at energies up to a few hundred keV. The
quasilinear diffusion induced by the wave is in the parallel velocity direction. As a result, the electron
distribution function features a flat region in terms of v//, known as the quasilinear plateau. Using an
asymmetric N// spectrum therefore results in an asymmetric electron distribution function with respect
to v//, with a substantial superthermal population. Electrons diffusing in the parallel direction do not
experience trapping effects as much as electrons diffusing in the perpendicular direction. Furthermore,
in addition to creating an asymmetric resistivity, the LH directly transfers parallel momentum to the
electrons. These advantages explain the large LH current drive (LHCD) efficiency achieved in several
devices, up to LHCD=0.34x1020A.W-1.m-2 in the large tokamaks JET and JT-60U. Projections for ITER
predict that LHCD~0.24x1020A.W-1.m-2 could be achieved, with a deposition profile peaked significantly
off-axis (normalized radius ~0.7) because of the high density and temperature (Gormezano et al.,
2007). Despite these results, at this point, uncertainties related to the LH wave penetration and current
drive efficiency in dense plasmas, as well as parasitic wave damping by the fusion alphas in a D-T
plasma, make their future applicability in reactors uncertain.

3. Electron Cyclotron (EC) Range of Frequency

Electron Cyclotron (EC) waves are typically injected in the frequency range 50-200GHz, which coincides
with the fundamental cyclotron resonance of electrons in most modern magnetic fusion devices.
Lower field operation opens the possibility to use harmonic schemes. In this frequency range,
ci<<~ce~pe. The wave is exclusively damped by resonant electrons. For these
frequencies/wavelengths, the propagation is quasi-optical and is quite adequately described by ray-
tracing or beam-tracing codes. This has the considerable practical advantage of allowing a coupling of
the wave to the plasma using an antenna located at some distance, generally consisting of a set of
steerable mirrors to vary the deposition location, and possibly to use these waves for current drive by
injecting them obliquely with respect to the magnetic field.

The dispersion relation reveals two distinct waves, called the ordinary and extraordinary modes. The
ordinary (O) mode has the simple dispersion relation, N2=1-pe2/2, showing that it can propagate to
the plasma core, at least up to the plasma cut-off, pe, which corresponds in practice to central
densities larger than 1020m-3. Fundamental absorption of the ordinary mode is therefore a viable
scheme for most tokamaks, including ITER. The extraordinary (X) mode has the somewhat more
complex dispersion relation, N2=2-R2)2-L2)/(2(2-UH2)), with R and L the right and left cut-
off frequencies, and UH the upper hybrid frequency. The right cut-off prevents the X-mode from
reaching the fundamental cyclotron resonance, ce, from the low field side of the plasma. Although
conceptually appealing, the solution of installing an antenna on the high field side does not appear to
be realistic in the context of a reactor. Alternatively, one can resort to the second harmonic resonance
2ce. The drawback of this method is that the right cut-off, while not preventing the resonance at
2ce from being reached, imposes a cut-off density which is only 50% of the cut-off density of the
O-mode for a given frequency and therefore requires operating a relatively higher frequency. The two
schemes envisaged for EC resonance heating (ECRH) and current drive (ECCD) in modern magnetic
fusion devices are therefore fundamental ordinary mode heating (O-1) or second harmonic
extraordinary mode heating (X-2), and sometimes even third harmonic X-mode heating (X-3).
The fundamental resonance condition for EC waves is given by =ce+k//v//, where ce=eB0/me is the
electron cyclotron frequency and me is the electron mass corrected for relativistic effects, which turn
out to play a major role in the resonance condition. It can be shown that the resonance curve is an
ellipse in velocity or momentum space, (p, p//), with p=mev the electron momentum. Writing
=ce(Rres), with Rres the resonant position in terms of major radius R, shows that the energy of
resonant electrons is given by:

E=mec2(Rres/R-1+N//p///mec). (9)

For a given position, R, the resonance parameters /ce and N// thus locate the interaction in the (p,
p//) plane, which results in a coupling between the wave damping properties in physical space and the
location of distribution function modifications induced by the wave momentum space. The quasilinear
diffusion that is induced occurs mainly in the perpendicular direction near the resonance ellipse.

Studying wave damping more carefully requires the calculation of the full hot relativistic dielectric
tensor (Brambilla, 1998). Some insight into the damping process can nevertheless be gained by
realizing that for EC waves, the electric field-dependent coefficient in the quasilinear diffusion
coefficient (Eq. 6) is, to lowest order, proportional to |d(p)(E)|~|vE-Jp-1(ke)|2, with E- the right-
handed component of the electric field, Jp-1 the Bessel function of order p-1 and e the electron Larmor
radius. For small arguments, Jp-1(ke)~(ke)p-1, showing that fundamental damping does not depend
directly on the electron temperature, whereas second harmonic damping is a finite temperature
effect. This would seem to indicate that O-1 damping is more efficient than X-2 damping, but, as for IC
waves, the electric field polarization near =ce is such that E- is relatively small (0 in the cold-plasma
approximation), thereby decreasing the absorption strength. This is confirmed by comparing the
respective optical depths for the O-1 and X-2 modes, i.e. the ratio between the power crossing the
resonance layer and the incident power, which shows that X-2 damping is typically 5-10 times more
efficient than O-1 damping. It should be pointed out that despite this difference, this damping is very
efficient in both cases, so that the power deposition profiles are essentially always narrow.

Figure 5: Schematic of EC wave heating and current drive in a tokamak such as ITER. (a) Poloidal cut; (b) Top view. The beam
is shown in blue, with red colouring at locations where significant absorption occurs. The (fictitious) EC steering mirror is
shown in magenta, and is used in this representation to inject the wave with a toroidal angle with respect to the magnetic
field direction, a situation typical of ECCD. The red dashed curve shows the cold EC resonance.
Furthermore, as explained previously, EC antennas are composed of steerable mirrors (Farina, 2014).
As a result, it is straightforward to inject the wave with a certain angle with respect to the magnetic
field to 1) locate the damping at a precise position in radius, and 2) drive a finite toroidal current. This
is illustrated in Figure 5. As discussed in section “Wave propagation”, even though no parallel
momentum is imparted by an EC wave to the plasma since the quasilinear diffusion is in the
perpendicular direction, the Fisch-Boozer mechanism still makes EC waves suitable for current drive
applications. Successful ECCD in a tokamak requires that the damping take place before the
fundamental resonance, so that, by virtue of Eq. 9, it occurs on significantly energetic electrons. In
theory, the achievable CD efficiency is only a factor of ¾ lower than that attained with LH waves (the
difference being due to the absence of any parallel momentum input). However, owing to trapped
electron effects and to the relatively modest energies of resonant electrons, ECCD is significantly less
efficient than LHCD, except at very high temperatures. The record ECCD efficiency to date,
ECCD=0.042x1020A.W-1.m-2, was obtained in JT-60U. In ITER, the EC launcher geometry and plasma
parameters should allow ECCD~0.2x1020A.W-1.m-2 to be achieved (Gormezano et al., 2007). The good
localization of the deposition and the excellent flexibility offered by the injector arguably compensate
for this lower global efficiency – the driven current density being locally quite large at a position chosen
by an operator or by an automated control system. For this reason, EC waves have a wide range of
applications, including current profile control, MHD control, and plasma start-up assist (required for
stellarators, useful in large tokamaks to enlarge the domain of parameters for plasma breakdown), and
their exploitation should increase further as the technological development of efficient and powerful
EC generators advances.

Acknowledgements
Careful manuscript reading and insightful comments by M. Schneider, G. Giruzzi, D. Keeling and Y.
Peysson are gratefully acknowledged.

References
 Brambilla, M., 1998. Kinetic Theory of Plasma Waves. Oxford: Clarendon Press.
 Decker, J., Peysson, Y., Artaud, J.-F., Nilsson, E., Ekedahl, A., Goniche, M., Hillairet, J., and Mazon,
D., 2014. Damping of lower hybrid waves in large spectral gap configurations. Physics of Plasmas,
21(9), p.092504.
 Dumont, R., 2009. Variational approach to radiofrequency waves in magnetic fusion
devices. Nuclear Fusion, 49(7), p.075033.
 Farina, D., Henderson, M., Figini, L. and Saibene, G., 2014. Optimization of the ITER electron
cyclotron equatorial launcher for improved heating and current drive functional
capabilities. Physics of Plasmas, 21(6), p.061504.
 Fisch, N., 1987. Theory of current drive in plasmas. Reviews of Modern Physics, 59(1), pp.175-234.
 Goldston, R., and Rutherford, P., 1995. Introduction to Plasma Physics. Bristol: Institute of Physics
Publishing.
 Gorelenkov, N. and Sharapov, S., 2020. Magnetic Confinement Fusion - Plasma Theory: Energetic
Particle Physics. Chapter 01208, this encyclopedia.
 Gormezano, C., Sips, A., Luce, T., Ide, S., Becoulet, A., Litaudon, X., Isayama, A., Hobirk, J., Wade,
M., Oikawa, T., Prater, R., Zvonkov, A., Lloyd, B., Suzuki, T., Barbato, E., Bonoli, P., Phillips, C.,
Vdovin, V., Joffrin, E., Casper, T., Ferron, J., Mazon, D., Moreau, D., Bundy, R., Kessel, C., Fukuyama,
A., Hayashi, N., Imbeaux, F., Murakami, M., Polevoi, A. and John, H., 2007. Chapter 6: Steady state
operation. Nuclear Fusion, 47(6), pp.S285-S336.
 Helander, P., and Sigmar, D., 2005. Collisional Transport in Magnetized Plasmas. Cambridge:
Cambridge University Press.
 Inoue, T. et al., 2020. Magnetic Confinement Fusion - Technology: 'Plasma Engineering'. Chapter
01214, this encyclopedia
 Jenko, F., 2020. Magnetic Confinement Fusion - Plasma Theory: Transport and Confinement.
Chapter 01205, this encyclopedia.
 Kazakov, Y., Ongena, J., Wright, J., Wukitch, S., Lerche, E., Mantsinen, M., Van Eester, D.,
Craciunescu, T., Kiptily, V., Lin, Y., Nocente, M., Nabais, F., Nave, M., Baranov, Y., Bielecki, J., Bilato,
R., Bobkov, V., Crombé, K., Czarnecka, A., Faustin, J., Felton, R., Fitzgerald, M., Gallart, D.,
Giacomelli, L., Golfinopoulos, T., Hubbard, A., Jacquet, P., Johnson, T., Lennholm, M., Loarer, T.,
Porkolab, M., Sharapov, S., Valcarcel, D., Van Schoor, M. and Weisen, H., 2017. Efficient generation
of energetic ions in multi-ion plasmas by radio-frequency heating. Nature Physics, 13(10), pp.973-
978.
 Landau, L., 1946. On the vibrations of the electronic plasma. J. Phys. USSR, 10, p.25.
 Lao, L. et al., 2020. Magnetic Confinement Fusion - Plasma Theory: Magnetohydrodynamic
Stability. Chapter 01206, this encyclopedia.
 Peysson, Y., Bonoli, P., Chen, J., Garofalo, A., Hillairet, J., Li, M., Qian, J., Shiraiwa, S., Decker, J.,
Ding, B., Ekedahl, A., Goniche, M. and Zhai, X., 2017. Current Challenges in the First Principle
Quantitative Modelling of the Lower Hybrid Current Drive in Tokamaks. EPJ Web of Conferences,
157, p.02007.
 Ryutov, D., 1999. Landau damping: half a century with the great discovery. Plasma Physics and
Controlled Fusion, 41(3A), pp.A1-A12.
 Sarazin, Y., 2020. Magnetic Confinement Fusion − Plasma Theory: Kinetic and Fluid Analysis.
Chapter 01204, this encyclopedia.
 Schneider, M., Eriksson, L., Jenkins, I., Artaud, J., Basiuk, V., Imbeaux, F. and Oikawa, T., 2011.
Simulation of the neutral beam deposition within integrated tokamak modelling
frameworks. Nuclear Fusion, 51(6), p.063019.
 Stix, T., 1992. Waves in plasmas. New York: Springer-Verlag.
 Swanson, D., 2003. Plasma Waves. Second edition. Bristol: Institute of Physics Publishing.
 Wesson, J., 2011. Tokamaks. Fourth edition. Oxford: Oxford University Press.
 Yamada, H., 2020. Magnetic Confinement Fusion – Experimental physics: Stellarators. Chapter
01211, this encyclopedia.
 Zohm, H., 2020. Magnetic Confinement Fusion – Experimental physics: Tokamaks. Chapter 01210,
this encyclopedia.
Glossary
 Alfvén velocity: The group and phase velocities associated to Alfvén waves are of the order
of the Alfvén velocity, va≡B0/(0nimi)1/2, with B0 the confining magnetic field, 0 the vacuum
permeability and ni (resp. mi) the ion density (resp. mass). This ratio reflects the antagonistic
effects of the magnetic field line tension on the one hand, and of the ion inertia on the other
hand.
 Alfvén waves: A kind of magnetohydrodynamic wave characterized by a periodic oscillation
of the plasma ion drift motion associated to a perturbation of the magnetic field. It is
characterized by a frequency much lower than the ion cyclotron frequency. Alfvén waves play
an important role in magnetic fusion plasmas because they can interact efficiently with energetic
ions and, in some situations, induce a radial displacement of these particles (Gorelenkov and
Sharapov, 2020).
 Boltzmann equation: A kinetic equation describing statistical properties of a
thermodynamic system out-of-equilibrium. In practice, the generic form of the collision term
appearing in this equation prevents it from being directly applied to fully ionized plasmas in
which long-range Coulomb collisions dominate.
 Bootstrap current: A contribution to the toroidal plasma current resulting from the fact
that the plasma pressure is maximal near the plasma center and decreases towards the edge.
Because of this non-uniformity, the friction between trapped electrons and passing electrons
results in a net transfer of parallel momentum from the former to the latter, inducing a finite
parallel current (Helander and Sigmar, 2005). In tokamaks, the (beneficial) bootstrap
contribution to the total current can be quite significant (typically ~50% in ITER, depending on
the regime of operation).
 Coulomb logarithm: This quantity appears in the theory of Coulomb collisions taking place
between the charged particles comprising a plasma. Denoting  the approximate number of
particles contained in the Debye sphere, its logarithm, ln(), is much larger than unity in weakly
coupled plasmas, in which collective effects dominate (Goldston and Rutherford, 1995). Fusion
plasmas are examples of weakly coupled plasmas and it is found that ln() ranges from ~17
(current fusion experiments) to ~18 (fusion reactors).
 Fokker-Planck equation: An equation deduced from the Boltzmann equation assuming that
long-range collisions dominate, as is appropriate in hot magnetic fusion plasmas. The ensuing
accumulation of multiple small angle deviations for a given particle results in a diffusion
equation for the distribution function in velocity space.
 keV: In magnetic fusion plasmas, it is customary to express energies in terms of electronvolt
(eV), kilo-electronvolt (keV) or mega-electronvolt (MeV). 1eV is approximately equivalent to
1.602176x10-19 Joule. By extension, this unit is also employed for plasma temperatures: 1eV is
approximately equivalent to 1.160450x104 Kelvin.
 Quasilinear equation: A kinetic equation describing the secular (i.e. long duration) effect of
the interactions taking place between an electromagnetic wave and the plasma particles. It is
obtained by averaging the Vlasov equation over a time much longer than the wave period, and
over a space region with a large extension compared to the wavelength.
 Trapped /passing particles: In a magnetic confinement device, the confining magnetic field
B0 is non-uniform. The velocity vector, v, of a given particle of mass, m, can be written in terms
of parallel (v//) and perpendicular (v) components with respect to the magnetic field direction.
Magnetic confinement imposes that the particle kinetic energy, E=m(v//2+ v2)/2, and its
magnetic moment, =mv2/(2B0), be conserved to lowest order. By manipulation of these two
equations, it can be demonstrated that particles can be classified into two main categories: some
are characterized by trajectories with v//≠0 at any point, whereas others periodically see their
parallel velocity cancel. The former are passing particles, and perform complete toroidal
revolutions. The latter, on the other hand, travel back and forth in the toroidal direction and are
called trapped particles. In particular, trapped electrons do not carry any toroidal current.
 Vlasov equation: A kinetic equation obtained by neglecting inter-particle correlations in the
kinetic description of the plasma, i.e. assuming that all particle interactions are mediated by a
self-consistent average electromagnetic field satisfying Maxwell’s equations. Since collisions are
neglected in this process, this equation is applicable to timescales much smaller than the typical
collision time. It is sometimes referred to as the collisionless Boltzmann equation.
 WKB (Wentzel-Kramers-Brillouin) approximation: A special case of multiple scale analysis.
In the context of waves propagating in plasmas, it is relevant when the plasma varies on a space
scale much larger than the wavelength. In this case, at a given position x, the electromagnetic
field can be written in the form of E(x)=E0(x)exp(iS(x)), where E0(x) is assumed to vary slowly
compared to the phase S(x). It is applicable outside of wave cut-offs or resonances (Swanson,
2003).