Vladimir Rozhansky

Plasma
Theory
An Advanced Guide for Graduate
Students
Plasma Theory
Vladimir Rozhansky

Plasma Theory
An Advanced Guide for Graduate Students
Vladimir Rozhansky
Peter the Great St. Petersburg Polytechnic University
St. Petersburg, Russia

ISBN 978-3-031-44485-2 ISBN 978-3-031-44486-9 (eBook)

https://doi.org/10.1007/978-3-031-44486-9

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by
similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Paper in this product is recyclable.

Introduction

Lectures on plasma theory are part of a program for students of the Plasma Physics
department of Peter the Great St. Petersburg Polytechnic University. This is an
advanced course, and students are supposed to be familiar with the basics of plasma
physics in the framework of introductory courses such as “Introduction to Plasma
Physics” and “Elementary Processes in Plasma.” Lectures on plasma theory are
planned for two semesters. In the first semester, students study kinetic theory and
transport processes in plasma, while the second semester is devoted to plasma
dynamics, including MHD theory, equilibrium, and stability. More advanced prob-
lems, such as neoclassical theory, stochastization of the magnetic field lines, and
edge plasma physics, are also considered. Waves in plasma are not included in this
course and should be studied separately. Only low-frequency waves and instabilities
that are closely connected with the dynamics and transport of plasma, such as MHD
and drift waves, are considered.
The distinctive feature of this course compared to most courses on plasma physics
is that phenomena in both low- and high-temperature plasmas are considered
simultaneously so that the theories of slightly ionized and fully ionized plasmas
are presented. Therefore, this course might be useful for a wide audience of students
and specialists working in different areas, such as nuclear fusion, gas discharge
physics and low-temperature plasma applications, space, and astrophysics. At the
end of the chapters, one can find examples of using theoretical expressions obtained
in the chapter in various applications, experiments, and numerical simulations. For
further reading, see [1–15].
The author is grateful to Dr. E. Kaveeva and Dr. I. Senichenkov for useful
discussions on different problems considered in this book and to I. Rozhanskaia
for help in manuscript preparation for publishing.

v
Contents

1 Plasma Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Boltzmann Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Collision Operator for Coulomb Collisions . . . . . . . . . . . . . . . . 6
1.2.1 General Expression for a Flow in the Velocity
Space Caused by Collisions . . . . . . . . . . . . . . . . . . . . 6
1.2.2 Deceleration and Diffusion of Test Particles
Cloud in the Velocity Space . . . . . . . . . . . . . . . . . . . . 8
1.2.3 Momentum and Energy Loss of the Test Particles . . . . 12
1.2.4 Landau Collision Operator . . . . . . . . . . . . . . . . . . . . . 14
1.3 Relativistic Collision Operator . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4 Fokker-Planck Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5 Runaway Electrons in Fully Ionized Plasma . . . . . . . . . . . . . . . 20
1.6 Distribution Function of Electrons in Slightly Ionized
Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
→
1.6.1 Approximation f 0 , f 1 . . . . . . . . . . . . . . . . . . . . . . . . 26
1.6.2 Distribution Function in the Electric Field . . . . . . . . . . 30
1.6.3 Impact of Electron-Electron Collisions . . . . . . . . . . . . . 32
→
1.6.4 General Expression for f 1 . . . . . . . . . . . . . . . . . . . . . 32
1.7 Transport Coefficients for Electrons in Slightly Ionized
Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.8 Drift Kinetic Equation in a Stationary Electric
and Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.9 Gyrokinetic Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.10 Pellet Ablation in a Tokamak . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2 Transport Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.1 Momentum Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 Transport Coefficients in Fully Ionized Plasma. Method
of Chapman and Enskog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

vii
viii Contents

2.3 Summary of the Results for the Fully Ionized Plasma . . . . . . . . 56

2.4 Transport Coefficients in Fully Ionized
Plasma. Qualitative Considerations . . . . . . . . . . . . . . . . . . . . . . 59
2.4.1 Friction Caused by the Relative Mean Velocity
and Thermal Force . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.4.2 Spitzer Conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.4.3 Heat Flux: Conductive and Convective Parts . . . . . . . . 62
2.4.4 Collisional Heat Production . . . . . . . . . . . . . . . . . . . . 64
2.4.5 Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.5 Equation for Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.6 Viscosity in the BGK Approximation . . . . . . . . . . . . . . . . . . . . 68
2.7 Thermal Force for Impurities . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.8 First Ionization Potential Effect and Impurity
Retention in a Tokamak Edge . . . . . . . . . . . . . . . . . . . . . . . . . 71
3 Quasineutral Plasma and Sheath Structure . . . . . . . . . . . . . . . . . . . 75
3.1 Quasineutrality Maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.2 Collisionless Sheath at the Material Surfaces . . . . . . . . . . . . . . . 78
3.2.1 Electrons in a Capacitor with a Reflecting
Electric Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.2.2 Particle and Energy Fluxes to the Material
Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.2.3 Current-Voltage Characteristics of the
Sheath. Floating Potential . . . . . . . . . . . . . . . . . . . . . . 83
3.2.4 Sheath Structure. Bohm Criterion . . . . . . . . . . . . . . . . 83
3.3 Impact of Electron Emission. Double Sheath . . . . . . . . . . . . . . . 86
3.4 Sheath in Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.5 Thermoelectric Current Between Two Electrodes . . . . . . . . . . . 91
4 Diffusion in Partially Ionized Unmagnetized Plasma . . . . . . . . . . . . 95
4.1 Ambipolar Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2 Examples of Solutions of the Ambipolar Diffusion
Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.2.1 Decay of Initial Perturbation in Infinite Plasma . . . . . . . 98
4.2.2 Positive Column of Glow Discharge . . . . . . . . . . . . . . 99
4.2.3 Diffusive Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.2.4 Diffusive Probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.3 Diffusion of Slightly Ionized Multispecies Plasma . . . . . . . . . . . 105
4.4 Diffusion in the Ionosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5 Diffusion of Partially Ionized Magnetized Plasma . . . . . . . . . . . . . . 111
5.1 Diffusion and Mobility in a Magnetic Field . . . . . . . . . . . . . . . 111
5.2 One-Dimensional Diffusion in Magnetized Plasma . . . . . . . . . . 115
5.2.1 Diffusion Across a Magnetic Field . . . . . . . . . . . . . . . 115
5.2.2 1D Diffusion at an Arbitrary Angle with a
Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Contents ix

5.3 Diffusion of Perturbation in Unbounded Plasma . . . . . . . . . . . . 120

5.4 Diffusion in Plasma Restricted by Dielectric Walls . . . . . . . . . . 126
5.5 Diffusion in a Cylinder with Conducting Walls . . . . . . . . . . . . . 130
5.6 Diffusive Probe in Magnetic Field . . . . . . . . . . . . . . . . . . . . . . 132
5.7 Experiments in Laboratory Plasma . . . . . . . . . . . . . . . . . . . . . . 134
6 Partially Ionized Plasma with Current . . . . . . . . . . . . . . . . . . . . . . 137
6.1 Plasma with Net Current in the Absence of a Magnetic
Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
6.1.1 Small Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.1.2 Nonlinear Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.2 Magnetized Plasma with Current . . . . . . . . . . . . . . . . . . . . . . . 143
6.2.1 One-Dimensional Evolution . . . . . . . . . . . . . . . . . . . . 143
6.2.2 Multidimensional Evolution of Small
Perturbation in Unbounded Plasma . . . . . . . . . . . . . . . 144
6.2.3 Effect of Conductivity Recovery in a Weak
Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.3 Plasma Clouds in the Ionosphere . . . . . . . . . . . . . . . . . . . . . . . 147
6.3.1 Redistribution of Metal Ions in the Polar
Ionosphere. Sporadic Layers . . . . . . . . . . . . . . . . . . . . 147
6.3.2 Active Experiments with Barium Clouds . . . . . . . . . . . 148
7 Transport in Strongly Ionized Plasma Across
a Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.1 Classical Diffusion of Fully Ionized Plasma
Across a Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.2 Transport of Impurities in Fully Ionized Plasma
Across a Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.3 Partially Ionized Magnetized Plasma with an
Inhomogeneous Neutral Component . . . . . . . . . . . . . . . . . . . . . 157
7.4 Penetration of Neutral Particles into Hot Tokamak Plasma . . . . . 160
8 Drift Waves and Turbulent Transport . . . . . . . . . . . . . . . . . . . . . . 163
8.1 Drift Waves in Inhomogeneous Plasma . . . . . . . . . . . . . . . . . . 163
8.2 Drift-Dissipative Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
8.3 Universal Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
8.3.1 Fluid Ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
8.3.2 Kinetic Ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.4 Instabilities Caused by the Temperature Gradient . . . . . . . . . . . 176
8.5 Turbulent Transport Caused by Random Electric Fields . . . . . . . 178
8.6 Effect of Magnetic Shear on Plasma Instabilities . . . . . . . . . . . . 181
9 Dynamics of Fully Ionized Plasma in the Absence
of a Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
9.1 Ion Acoustic Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
9.2 Nonlinear Dynamics. Self-Similar Solutions . . . . . . . . . . . . . . . 192
x Contents

9.3 Simple Nonlinear Waves. Overturn . . . . . . . . . . . . . . . . . . . . . 195

9.4 Nonlinear Ion Acoustic Waves with Dispersion . . . . . . . . . . . . . 197
9.5 Plasma Expansion During Pellet Injection . . . . . . . . . . . . . . . . . 201
10 Magnetohydrodynamics (MHD) . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
10.1 Magnetohydrodynamic Equations . . . . . . . . . . . . . . . . . . . . . . 205
10.2 Magnetic Field Frozen in and Skin Effect . . . . . . . . . . . . . . . . . 209
10.3 MHD Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
10.4 Nonlinear MHD Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.5 Magnetosonic Waves with Dispersion . . . . . . . . . . . . . . . . . . . 222
10.6 Alfven Masers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
11 Dynamics of Plasma Blobs and Jets in a Magnetic Field . . . . . . . . . 229
11.1 Plasma Motion Across Magnetic Field in Vacuum . . . . . . . . . . 229
11.2 Deceleration of the Plasma Jet by Ambient Plasma . . . . . . . . . . 234
11.3 Edge Localized Modes and Filaments . . . . . . . . . . . . . . . . . . . . 238
12 Plasma Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
12.1 On the Possibility of Equilibrium in the Absence
of a Vacuum Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . 242
12.2 Equilibrium of a Pinch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
12.3 Magnetic Flux Surface Functions . . . . . . . . . . . . . . . . . . . . . . . 245
12.4 Grad-Shafranov Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
12.5 Integral Equilibrium in a Tokamak . . . . . . . . . . . . . . . . . . . . . . 252
12.6 Plasma Equilibrium in a Tokamak with
Circular Cross-Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
12.7 Coordinates for Arbitrary Flux Surfaces . . . . . . . . . . . . . . . . . . 260
12.8 Force-Free Equilibrium and Pinch with
Canonical Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
12.9 2D Modeling of the Tokamak Edge . . . . . . . . . . . . . . . . . . . . . 267
13 Transport Phenomena in Tokamaks . . . . . . . . . . . . . . . . . . . . . . . . 269
13.1 Fluid Regime (Pfirsch-Schlueter Regime) . . . . . . . . . . . . . . . . . 271
13.1.1 Qualitative Estimates . . . . . . . . . . . . . . . . . . . . . . . . . 271
13.1.2 Heat Conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
13.1.3 Plasma Flows on the Flux Surface,
Density, Temperature and Potential Perturbations . . . . . 276
13.1.4 Particle Fluxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
13.2 Radial Electric Field, Poloidal and Toroidal Rotation . . . . . . . . . 281
13.3 Neoclassical Transport in Collisionless Regimes . . . . . . . . . . . . 285
13.3.1 Particle Trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . 285
13.3.2 Ware Drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
13.3.3 Estimation of Transport Coefficients
in the Plateau Regime . . . . . . . . . . . . . . . . . . . . . . . . . 291
13.3.4 Estimation of Transport Coefficients
in the Banana Regime . . . . . . . . . . . . . . . . . . . . . . . . . 293
13.4 Distribution Function in the Collisionless Regimes . . . . . . . . . . 294
Contents xi

13.4.1 Plateau Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294

13.4.2 Banana Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
13.5 Particle and Heat Balance Equations . . . . . . . . . . . . . . . . . . . . . 301
13.6 Transport Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
14 Instabilities in Magnetized Plasma . . . . . . . . . . . . . . . . . . . . . . . . . 307
14.1 Rayleigh-Taylor Instability in Fluids . . . . . . . . . . . . . . . . . . . . 307
14.2 Flute Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
14.3 Dissipative Modifications of Flute Instability . . . . . . . . . . . . . . 315
14.3.1 RT Instability in Partially Ionized Plasma . . . . . . . . . . 315
14.3.2 Flute Instability in Plasma Contacting Metal
Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
14.3.3 Gravitational-Dissipative Flute Instability . . . . . . . . . . . 318
14.4 Energy Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
14.5 Kink Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
14.6 Tearing Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
14.7 Geodesic Acoustic Mode and Zonal Flows . . . . . . . . . . . . . . . . 331
14.8 Equatorial Plasma Bubbles . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
15 Magnetic Islands and Stochastic Magnetic Field . . . . . . . . . . . . . . . 335
15.1 Magnetic Islands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
15.2 Stochastic Instability and Magnetic Field Line Diffusion . . . . . . 339
15.3 Transport in Stochastic Magnetic Field . . . . . . . . . . . . . . . . . . . 344
15.4 Resonant Magnetic Perturbations in Tokamak . . . . . . . . . . . . . . 346
15.5 Simulation of Resonant Magnetic Perturbations Effects
with Codes and Examples of Experimental Results . . . . . . . . . . 350
16 Improved Confinement Regime (H-Mode) . . . . . . . . . . . . . . . . . . . . 353
→ →
16.1 E × B Drift Shear and Transport Barriers . . . . . . . . . . . . . . . . . 353
16.2 Transition from Low to High Confinement Regime
(L-H Transition) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
16.3 L-H Transition Power Threshold . . . . . . . . . . . . . . . . . . . . . . . 358

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
Chapter 1
Plasma Kinetics

1.1 Boltzmann Equation

The plasma state is generally described by a set of distribution functions

→ →
f α r , V , t for all the plasma components and their quantum states. The distribu-
tion function can be denoted as a particle density in a six-dimensional coordinate and
→ → → →
velocity phase space, while the quantity dnα r , V , t = f α d r d V is the number of
particles in an infinitesimal element of a phase volume. Subscript α here represents
different particles, neutral or ionized ones, as well as different quantum states of
atoms, molecules, or ions. Classical ideal nonrelativistic plasma is considered below.
The variation in the number of particles in the six-dimensional phase space in the
absence of collisions is caused by a flow of a “phase liquid” to the neighboring
regions of the phase space and a change in the number of particles over time. It is
controlled by a six-dimensional continuity equation

6
∂f α ∂
þ ðf α x_ i Þ = 0:
∂t i=1
∂x i

In the phase space, six “coordinates” xi in the Cartesian coordinate system are
three space coordinates ri and three components of the velocity Vi, and “velocities” x_ i
correspondingly consist of three components of the velocity Vi and three components
of the acceleration V_ i . Therefore, the continuity equation in the phase space has the
form

3 3
∂f α ∂ ∂
þ ðf α V i Þ þ f α V_ i = 0:
∂t i=1
∂r i i=1
∂V i

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 1

V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_1
2 1 Plasma Kinetics

The first term on the l.h.s. represents temporal variation of the distribution function,
the second term corresponds to the divergence of a flux in the real space, and the
third term corresponds to the flux divergence in the velocity space. The acceleration
V_ i is produced by forces applied to a particle. In the plasma

→_ Zαe → → → →
V = E þ V×B þ g,
mα
→ →
where Zα is the charge number of a particle, mα is the particle mass, E and B are
→
electric and magnetic fields, respectively, and mα g is the gravitational force.
The space coordinates and velocities are the independent variables; hence, ∂Vi/
∂ri = 0. Additionally, since the Lorentz force is perpendicular to the velocity of a
particle, we have ∂V_ i =∂V i = 0. Therefore, the continuity equation is reduced to the
form

∂f α → ∂f α →_ ∂f α
þ V → þ V → = 0, ð1:1Þ
∂t ∂r ∂V

or

∂f α → ∂f α Z α e → → → ∂f α → ∂f α
þV →þ E þ V×B → þ g → = 0: ð1:2Þ
∂t ∂r mα ∂V ∂V

This equation is known as the Vlasov equation.

The Vlasov equation can be rewritten in any generalized coordinates qi and
momentum pi. The Vlasov equation in the form of Eq. (1.1) in the general case is
derived from the continuity equation in the phase space using a relation
∂q_ i =∂qi þ ∂p_ i =∂pi = 0, which follows from the Hamilton equations
q_ i = ∂H=∂pi ; p_ i = - ∂H=∂qi .
The left-hand side of the Vlasov equation is equal to a full derivative dfα/dt.
Therefore, in the stationary case, according to Liouville’s theorem, the distribution
function is constant along the phase-space trajectories of the system. An important
statement that follows from Liouville’s theorem is that the stationary distribution
function in a collisionless case should be a function of integrals of motion. This
important notice provides an opportunity to find the distribution function in various
collisionless problems.
Accounting for collisions changes Eqs. (1.1) and (1.2), since the distribution
function is not constant along the trajectories even in the stationary case. In the
process of collisions, the velocities of the particles change as well as their quantum
states (we shall not consider the change in the particle positions in the process of
collisions assuming that the plasma is an ideal gas). In the presence of collisions, the
kinetic equation is given by
1.1 Boltzmann Equation 3

df α ∂f α → ∂f α Z α e → → → ∂f α → ∂f α
 þV →þ Eþ V×B → þ g → = St α , ð1:3Þ
dt ∂t ∂r mα ∂V ∂V

where a collision operator on the r.h.s. Stα is responsible for the change in the
distribution function during collisions. Equation (1.3) is known as a Boltzmann
equation. The collision operator is a sum over all species

St α = St α β f α , f β : ð1:4Þ
β

Each summand corresponds to collisions of species α with all species in the plasma,
including particles α.
We shall consider elastic collisions when the quantum states of particles remain
the same in the process of collisions and the energy and the momentum of the
particles are conserved. We assume that before the collision, two species α and β
→ →
have velocities V α and V β , while after the collision, they change their velocities to
→0 →0
the values V α and V β correspondingly without changing the spatial coordinate.
→
After each collision, the particle with the velocity V α escapes from the infinitesimal
→
volume in the velocity space d V α . The full number of such escapes caused by the
collisions of the particles of α species with the particles of species β with
→ → →0 →0 →
V α , V β → V α , V β per second in the infinitesimal velocity space volume d V α for
→
a fixed value of V α is given by the expression

→ → → → → → →
dQα-β = d r d V α fα V α fβ V β dσα β =dΩ V α - V β dΩd V β :
Vβ Ω
→

Here, dσαβ/dΩ is a differential cross-section of scattering to the solid angle Ω. In

→
addition to losses in the infinitesimal velocity space volume d V α , there is also a
→0 →0 → →
source caused by the collisions V α , V β → V α , V β that transfer particles with
→0 →0 →
velocities V α , V β to the velocity space volume d V α :

→ →0 →0 →0 →0 →0 →0
dQþ
αβ = d r fα V α fβ V β dσ α β =dΩ V α - V β dΩd V α d V β :
→0 →0 Ω
Vβ Vα

→0 →0
Velocities V α and V β under the integral are not independent but connected by the
conservation laws. Indeed, in the process of collision of the particles with velocities
→0 →0 →
V α , V β , the particle of species α obtains the velocity V α . Let us change variables in
→ →
the integral and integrate over the velocities V α , V β using the conservation of the
4 1 Plasma Kinetics

→ → →0 →0
relative velocity V α - V β = V α - V β . As is known from classical mechanics,
→0 →0 → →
the Jacobian of this transformation is equal to unity: d V α d V β = d V α d V β . Finally,
→ →
since integration over d V α is carried out in the vicinity of a chosen value V α , one
obtains

→ → →0 →0 → → →
dQþ
αβ = d r dV α fα V α fβ V β dσα β =dΩ V α - V β dΩd V β :
Vβ Ω
→

→ →
Combining sources and sinks in the volume d r d V α , we find

→ →
dQþ -
α β - dQα β = St αβ d r d V α ,

where the collision operator is

→ → →
St αβ f α, f β = f 0α f 0β - f α f β dσαβ =dΩ V α - V β dΩd V β : ð1:5Þ
→
Vβ Ω

→0 →0
Here, f 0α  f 0α V α , f 0β  f 0β V β . Equation (1.5) is known as the Boltzmann
collision operator.
Boltzmann kinetic equation (1.3) is an integro-differential equation that contains
all distribution functions of the particles in the system. Therefore, a system of
coupled equations for all distribution functions is to be solved.
In the process of deriving the collision operator, we assumed that during the
collision, the particle coordinate remains unchanged. This is justified provided that
the potential energy is significantly smaller than their average kinetic energy, which
is of the order of their temperature T. For the charged particles, the average potential
energy of the Coulomb interaction is ZαZβe2/4πε0hrαβi, and the average distance
between the charged particles hrαβi is of the order of n-1/3, where n is the plasma
density. Therefore, the criterion of ideal plasma used in the derivation has the form

Z α Z β e2 n1=3 =ð4πε0 TÞ << 1:

In other words, the plasma should be sufficiently hot and not very dense.
An inelastic collision operator can be constructed in a similar way. However, it is
seldom used in the general case; in practice, special processes – ionization, recom-
bination, excitation, etc. – are analyzed, so integrals over the velocities are used,
which are easier to obtain directly.
Moments of the distribution function describe macroscopic plasma parameters.
They are defined according to
1.1 Boltzmann Equation 5

→ → *
M αj,k...n = V j V k . . . V n f α r , V , t dV : ð1:6Þ

The most important of them are the following. Particle density:

*
nα = f α dV : ð1:7Þ

The flux density of the particles is defined as

→ → → *
Γ α = nα u α = V f α dV , ð1:8Þ

→
where u α is the fluid or average velocity. The average energy of chaotic motion (for
single atoms) is:

→ → 2
3 mα V - u α →
n T = f αd V : ð1:9Þ
2 α α 2

A quantity Tα is called a temperature, which is the temperature definition. A heat flux

density is defined according to

→ mα → → 2 → → →
qα= V - uα V - u α f αd V : ð1:10Þ
2

Frequent collisions in plasma tend to establish a local Maxwellian distribution

function, which is defined as

→ → 2
nα mα V - u α
fM
α = exp - : ð1:11Þ
ð2πT α =mα Þ 3=2 2T α

In the absence of all forces in the stationary case in the homogeneous plasma,
Boltzmann equation (1.3) is reduced to equation Stα = 0. It is easy to see that the
Maxwellian distribution functions with the common temperature and common fluid
velocity turn to zero the collision operator. Indeed, substituting Maxwellian func-
tions, Eq. (1.11), into the collision operator, Eq. (1.5), we find that the bracket
f 0α f 0β - f α f β is zero, which is the consequence of the energy conservation during
an act of collision. Therefore, the Maxwellian distribution functions of all particles
correspond to thermodynamic equilibrium. A stronger statement is known in statis-
tical physics as the Boltzmann H-theorem – in the absence of all forces, a system
tends to reach thermodynamic equilibrium and hence to establish Boltzmann distri-
bution functions.
6 1 Plasma Kinetics

1.2 Collision Operator for Coulomb Collisions

1.2.1 General Expression for a Flow in the Velocity Space

Caused by Collisions

The Boltzmann form of the collision operator is inconvenient in the case of Coulomb
collisions. This is connected with the specific character of the Coulomb collisions,
where small angle scattering for large impact parameters dominates and determines
the Coulomb cross-section. The fact that the velocity variation in a single act of
collision is small with respect to the particle velocity can be used to simplify the
Boltzmann collision operator. Indeed, during the collision process, “phase fluid”
arrives at a given phase volume from the neighboring regions; therefore, almost
continuous flow of “phase fluid” takes place. In other words, the collision operator
can be rewritten as a divergence of a flow in the velocity space
→
→V
St α = - ∇ →V  Γα : ð1:12Þ

With such a form of the collision operator, the kinetic equation takes the form of a
continuity equation in the phase space with the account of collisions. The general
expression for the flow induced by the collisions has the form (summation over
subscript k is assumed)

→ F St ∂f
f - Dαjk α :
jα
ΓjαV = ð1:13Þ
mα α ∂V k

The higher derivatives over the velocities are neglected here, and the corresponding
series are truncated since the neighboring regions give the major contributions to the
→ St
flow. The quantity F α is known as a dynamical force, while quantity Dαjk is a
diffusion tensor in the velocity space. For collisions with different species, the flow
in the velocity space is defined as
→ →
ΓjαV = Γjαβ
V
:
β

→ St
Now, we shall demonstrate that quantities F α and Dαik are connected with
the problem of deceleration and scattering of the test particles α due to collisions
with the background particles of the species β. Let us consider a cloud of test
→ →
particles that had the same velocity V = V 0 at the initial moment t = 0. The mean
→ →
velocity u , which at t = 0 coincides with V 0 , decreases with time due to collisions.
Simultaneously, due to the process of scattering, diffusion of the cloud takes place.
The evolution of the cloud of test particles in the velocity space is shown
1.2 Collision Operator for Coulomb Collisions 7

Fig. 1.1 Scheme of the

evolution of the test particles
cloud

schematically in Fig. 1.1. In homogeneous plasma in the absence of forces, the time
derivative of the cloud mean velocity can be rewritten using the kinetic equation:
→
∂uj ∂ 1 → 1 ∂f → 1 →V →
= V jf αd V = Vj α d V = - V j ∇ →V  Γα dV :
∂t ∂t nα nα ∂t nα

Integrating by parts, taking into account that the distribution function vanishes at
→
→ →V
infinity, fα → 0 at V → 1, and, therefore, according to Eq. (1.13), Γ α → 0, and one
obtains

∂uj 1 → → 1 F St
jα f α ∂f → 1 F St ∂Dαjk →
- Dαjk α d V =
jα
= ΓjαV d V = fα þ dV :
∂t nα nα mα ∂V k nα mα ∂V k
ð1:14Þ

Substitution of the test particle distribution function at t = 0 in the form of

→ →
f α jt = 0 = nα δ V - V 0 into Eq. (1.14) yields

∂uj F St
jα ∂Dαjk
= þ : ð1:15Þ
∂t t=0
mα ∂V k

The diffusion of the test particle cloud is described by the dispersion tensor
→ → → →
V-u V-u . Here, averaging is defined as
j k

→
hgi = gfd V =n:
8 1 Plasma Kinetics

Evaluating the expression for the dispersion in the same manner as for the time
derivative of the mean velocity, it is easy to find

→ → → →
∂ V-u V-u
j k = 2Dαjk : ð1:16Þ
∂t
t=0

→ St
Hence, we demonstrated that the dynamic force F α , the diffusion tensor Dαik , and
consequently the collisional flow in the velocity space and the collision operator are
connected with the parameters of the test particles cloud in the velocity space.

1.2.2 Deceleration and Diffusion of Test Particles Cloud

in the Velocity Space

Let us calculate the quantities on the l.h.s. of Eqs. (1.15) and (1.16). The collisions
between the particles of species α and β should be considered in the center of mass
reference frame as a scattering of a particle with the mass mαβ = mαmβ/(mα + mβ) and
→r → →
the relative velocity u = V α - V β on a central potential. The change in the velocity
in the laboratory frame is proportional to the change in the relative velocity:

→ mβ →r
ΔV α = Δu :
mα þ mβ

The relative number of scattering events with a given impact parameter ρ and
azimuth angle φ for a single test particle is

→0 →0 →0 →0 →0
dnβ V ur dS = f β V d V ur dS = f β V d V ur ρdρdφ:

→
Multiplying this expression by a change in the velocity Δ V α and integrating over the
ambient particle velocities, we obtain

∂uj →0 →0 mβ
= fβ V wj d V , wj = Δurj ur dS: ð1:17Þ
∂t t=0
mα þ mβ

Similarly,
1.2 Collision Operator for Coulomb Collisions 9

→ → → →
∂ V-u V-u →0 →0
j k = fβ V wjk d V ,
∂t ð1:18Þ
t=0
2
mβ
wjk = Δurj Δurk ur dS:
mα þ mβ

Here, vector wj and tensor wjk are functions of the relative velocity; therefore, due to
the tensor invariance, they can be expressed according to

urj urj urk

wj = A, w jk = δ jk B þ C, ð1:19Þ
ur ð ur Þ 2

where А, В, and С are scalars. Let us calculate these tensors in the reference frame
with the z-axis parallel to the vector of the relative velocity. In the process of
scattering at a deflection angle θ in the central mass system, the change in the
relative velocity components is

Δurx = ur sin θ cos φ,

Δury = ur sin θ sin φ,
Δurz = - ur ð1 - cos θÞ:

For Coulomb collisions, the impact parameter is expressed through the deflection
angle according to

θ rs
tg = ,
2 ρ

where a strong force radius rs is defined as

Z α Z β e2
rs = :
4πε0 mαβ ður Þ2

Hence, the change in the relative velocity components as a function of the impact
parameter is given by

ρr s
Δurx = 2ur cos φ,
ρ2 þ r 2s
ρr
Δury = 2ur 2 s 2 sin φ,
ρ þ rs
r 2s
Δurz = - 2ur :
ρ2 þ r 2s
10 1 Plasma Kinetics

Let us substitute these expressions into Eqs. (1.17) and (1.18). The integrals over ρ
tend to infinity and should be truncated at some value ρ = ρmax since in the plasma,
the Coulomb potential is screened at distances of the order of the Debye radius. We
shall choose ρmax = rd, where the Debye radius is defined as

ε0 T
rd = : ð1:20Þ
ne2

The resulting integrals have logarithmic accuracy and are proportional to a

Coulomb logarithm Λ, where Λ is a large quantity of the order of 10–15:

Λ = lnðr d =r s Þ: ð1:21Þ

For more details see, for example, [6]. In particular,

ρmax
ρdρ
= Λ,
ρ2 þ r 2s
0
ρmax
ρ3 dρ
2
= Λ - 1=2 ≈ Λ:
ρ2 þ r 2s
0

After integrating in Eqs. (1.17) and (1.18) over ρ and φ, we find

urj
wj = - 1 þ mα =mβ Lαβ ,
4πður Þ3
urj urk 1
wjk = Lαβ δjk - ,
ð ur Þ 2 4πur
2
Z α Z β e2
Lαβ = Λ :
ε0 m α

The time derivatives of the mean velocity and dispersion tensor are given by

∂uj mα Lαβ urj →0 →0

= - 1þ fβ V dV ,
∂t t=0
mβ 4π ð ur Þ 3
→ → → →
∂ V-u V-u ð1:22Þ
j k Lαβ δjk urj urk →0 →0
= - fβ V dV :
∂t 4π u r
ð ur Þ 3

t=0
1.2 Collision Operator for Coulomb Collisions 11

Note that the time derivative of the higher moment

→ → → → → →
V-u V-u V-u does not contain parameter Λ, and, hence, its
i k j
contribution to the flux in the velocity space can be neglected. With an account of
expressions

2
∂ ur δjk urj urk
= r - ,
∂V j ∂V k u ð ur Þ 3
∂ 1 urj
= - ,
∂V j ur ð ur Þ 3

Equation (1.22) can be rewritten in the form

→0
fβ V
∂uj mα ∂ 1 →0
= 1þ L dV ,
∂t mβ αβ ∂V j 4π → →0
t=0 V -V

→ → → →
∂ V-u V-u 2
j k ∂ 1 → →0 →0 →0
= Lαβ V - V fβ V dV :
∂t ∂V j ∂V k 4π
t=0
ð1:23Þ

To rewrite Eq. (1.23) in one more form, it is useful to introduce “Rosebluth

potentials”

→0
fβ V
1 →0
φβ = - → →0
dV ,
4π
V -V ð1:24Þ
1 → →0 →0 →0
Ψβ = - V - V fβ V dV :
8π

The “potentials” φβ and Ψβ are linked by the relations

Δ →V Ψβ = φβ ,
ð1:25Þ
Δ →V φβ = Ψβ ,

where Δ →V is the Laplace operator in the velocity space. One can derive Eq. (1.25)
using the following identities:
12 1 Plasma Kinetics

1 → →0
Δ →V → →0
= - 4πδ V - V ,
V -V
→ →0 2
Δ →V V - V = → →0
:
V -V

Considering Eq. (1.24), one obtains Eq. (1.23) in the form

∂uj mα ∂φβ
= - 1þ Lαβ : ð1:26Þ
∂t t=0
mβ ∂V j

→ → → →
∂ V-u V-u ∂ Ψβ
2
j k = - 2Lαβ : ð1:27Þ
∂V j ∂V k
∂t
t=0

1.2.3 Momentum and Energy Loss of the Test Particles

The derived equations are of special interest since they can be implemented to find
→
the momentum and energy losses of the test particles with velocity V α in the
background medium of β species. The particle momentum is decreasing according
to Eq. (1.26). Let us derive the momentum loss of electrons due to collisions with
ions. We assume that the electron velocity is much larger than that of background
ions. Then, neglecting the ion velocity in the denominator of the integrand in
Eq. (1.24), one obtains φi = - ni/4πVe, and neglecting the mass ratio, we find

∂pe Λe4 Z 2 ni
= - ν e pe , νe = : ð1:28Þ
∂t t=0 4πε20 m2e V 3e

Here, the collision frequency νe corresponds to the inverse characteristic time scale
for electron deceleration at the initial moment when all electrons have the same
velocity. The collision frequency is inversely proportional to the cube of the electron
velocity: decreases with the velocity, which is typical for Coulomb collisions. Note
that one cannot use this equation at later stages since electrons not only decelerate
but are also deflected during electron-ion collisions. The collision frequency given
by Eq. (1.28) can be called a slowing-down collision frequency. A collision fre-
quency that corresponds to the characteristic time of their deflection can be obtained
using Eq. (1.27). Choosing the z-axis in the direction of the electron velocity, we
obtain
1.2 Collision Operator for Coulomb Collisions 13

→ → 2
→ →
∂ V-u ∂ V-u
→
V-u
→
⊥
j ¼ x x
j
∂t t¼0 ∂t t¼0
→ → → → ð1:29Þ
∂ V-u V-u
y y
þ j
∂t t¼0
¼ 2νe V : 2

Hence, the deflection collision frequency is twice as large as the slowing-down

collision frequency.
The characteristic frequency for the decrease in the kinetic energy of electrons can
be calculated using the following derivation:

→ → → →
∂ V-u V-u →
∂ε ∂m 1 j j *∂ u
= VV =m þu :
∂t t=0 ∂t 2 j j t=0 2 ∂t ∂t t=0
t=0

Here, we use identity

→ → → → →2
V-u V-u = V jV j - u :
j j

After substitution of Eqs. (1.26) and (1.27), one finds

∂εα →
= - mα Lαβ φβ þ 1 þ mα =mβ V α ∇ →V φβ : ð1:30Þ
∂t t=0

Let us calculate, for example, the change in the kinetic energy for electrons in
electron-ion collisions. By analogy with electrostatics, where the potential and
electric field outside of the cloud of distributed charge density are known to be
→
φi = - ni/4πV, ∇ →V φi = V ni
4πV 3
, we obtain

∂εe 2me
= - νεe εe, νεe = ν: ð1:31Þ
∂t t=0
mi e

We see that the collision frequency for energy exchange in the process of
electron-ion collisions νεe is 2me/mi times smaller than the slowing-down collision
frequency νe. This is an expected result since the kinetic energy of electrons changes
rather slowly in collisions with ions because of the small mass ratio. If the velocities
14 1 Plasma Kinetics

of particles are comparable, the corresponding expressions can be derived from

Eq. (1.30).

1.2.4 Landau Collision Operator

Using the expressions for the dynamic force and diffusion tensor in the velocity
space, Eqs. (1.15) and (1.16) and Eqs. (1.26)–(1.27), we find the flow in the velocity
space, Eq. (1.13):

mα ∂φβ
2
→
∂ Ψβ ∂f α
ΓjαV = - Lαβ fα - : ð1:32Þ
mβ ∂V j ∂V k ∂V j ∂V k

Let us use the following relations:

∂φβ 1 ∂f 0β → 0
=- U jk dV ,
∂V j 8π ∂V 0k
2
ð1:33Þ
∂ Ψβ 1 →0
=- U jk f 0β d V :
∂V j ∂V k 8π

Here, we introduced a tensor

δjk urj urk

U jk = r - : ð1:34Þ
u ð ur Þ 3

The first relation in Eq. (1.33) can be obtained according to

∂ f 0β → 0 urj →0 1 ∂U jk 0 → 0 1 ∂f 0β →0
dV = - f0 dV = -
3 β
f dV = 0 U jk d V ,
∂V j ur ð ur Þ 2 ∂V 0k β 2 ∂V k

with account of

urj 1 ∂U jk 1 ∂U jk
=- = :
ð ur Þ 3 2 ∂V k 2 ∂V 0k

The second relation is obtained according to

1.3 Relativistic Collision Operator 15

∂Ψβ 1 ∂ ∂ur 0 → 0 1 ∂ urk 0 → 0

=- f dV = - f dV
∂V j ∂V k 8π ∂V j ∂V k β 8π ∂V j ur β
1 →0
=- U jk f 0β d V :
8π

Finally, substituting Eq. (1.33) into Eq. (1.32), we obtain the flow in the velocity
space in the form

0
→ Z 2α Z 2β e4 Λ f α ∂f β f 0β ∂f α →0
ΓjαV = U jk 0 - dV : ð1:35Þ
8πε20 mα mβ ∂V k mα ∂V k

The Landau collision operator is equal to the divergence of this flow with an
opposite sign:

0
Z 2α Z 2β e4 Λ ∂ f α ∂f β f 0β ∂f α →0
St αβ = - U jk 0 - dV : ð1:36Þ
8πε20 mα ∂V j mβ ∂V k mα ∂V k

1.3 Relativistic Collision Operator

In the previous section, relativistic effects were neglected. For electrons, this is
justified if their temperature is significantly smaller than their rest mass
(~0.5 MeV). In the opposite case, one shall use more general expression for the
collision operator. In this section, we shall restrict consideration to the practically
frequent situation of relativistic particles moving through the background of nonrel-
ativistic particles, which could be considered practically stationary. Since the deflec-
tion angle is assumed to be small, as in the nonrelativistic case, one can neglect
radiation with respect to the collisional energy losses.
The momentum of relativistic particle α moving in the z-direction is

p = pz = γmα V, ð1:37Þ

where mα is the rest mass and γ = (1 - V2/c2)-1/2 is the relativistic mass factor. The
particle energy is

Eα = γmα c2 = m2α c4 þ p2 c2 : ð1:38Þ

The initial energy of nonrelativistic particle β is Eβ = mβc2. After the collision, the
energies of particles α and β are given by
16 1 Plasma Kinetics

E0α =
2
m2α c4 þ pz þ δpz c2 þ ðδp⊥ Þ2 c2 : ð1:39Þ

E0β = m2β c4 þ δp2z c2 þ ðδp⊥ Þ2 c2 :

Here, δp⊥ is the momentum change in the direction perpendicular to the z-axis. For
the small change in momentum during the collision

2pz δpz þ ðδp⊥ Þ2 2

E 0α = Eα 1 þ c, ð1:40Þ
2E 2α

2
δpz þ ðδp⊥ Þ2
E 0β = mβ c2 1 þ :
2m2β c2

We neglect here (δpz)2 since from the energy conservation one obtains

2
E α ðδp⊥ Þ
δpz = - 1 þ << pz :
m α c2 2pz

The change in the momentum in the perpendicular direction depends on the impact
parameter as (see previous section)

2mα Vρr s
δp⊥ = :
r 2s þ ρ2

The total effect of collisions with all impact parameters is described as

rd
∂ ðδp⊥ Þ2 2πρ3 Vnβ 2m2α c3
2
= ð2mα r s V Þ dρ = ναβ : ð1:41Þ
∂t r 2s þ ρ2
2 V
0

∂ δpz E α m2α c3
= - 1þ ν :
∂t mβ c2 pz V αβ

Here, it is taken into account that the number of encounters in unit time with impact
parameters in the interval (ρ, ρ + dρ) is nβV2πρdρ. The collision frequency at the
speed of light is
1.4 Fokker-Planck Equation 17

Z 2α Z 2β e4 Λ
ναβ = : ð1:42Þ
4πε20 m2α c3

Analogously

∂hδpk δpl i
= μðmα cÞ3 ναβ Pkl : ð1:43Þ
∂t

where

1=2:
p2 δkl - pk pl p2
Pkl = , μ = 1 þ
p3 m2α c2

The relativistic collision operator is given by

mα ∂ 2 ∂ Pkl ∂ðμf α Þ
St αβ = m3α c3 ναβ μ fα þ : ð1:44Þ
mβ p2 ∂p ∂pk 2 ∂pl

1.4 Fokker-Planck Equation

The Landau collision operator is still too complicated. In some cases, it could be
further simplified. In particular, this can be done for the test impurity of heavy ions in
the ambient plasma with light ions. Let fα be a distribution function of test heavy
ions, so that nα < < nβ. We assume that the impact of heavy ions on the distribution
function of light ions fβ is negligible, and function fβ is assumed to be Maxwellian.
Taking into account that for f β = f Mβ , the identity

∂f 0β mβ 0 0
=- V f ,
∂V 0k Tβ k β

is valid, we can simplify the Landau collision integral Eq. (1.36). The first term can
be rewritten as

0
∂ f α ∂f β → 0 ∂ f α 0 0 →0
U jk dV = - U jk V f dV
∂V j mβ ∂V 0k ∂V j Tβ k β
∂ fα →0
=- U jk V k - urk f 0β d V :
∂V j Tβ

Here, the second term under the integral is equal to zero,

18 1 Plasma Kinetics

U jk urk = 0: ð1:45Þ

→0 → →r →0
Since V >> V , we have u ≈ - V , and, therefore,

∂ fα →0 ∂ δjk V 0j V 0k f α →0
- U jk V k f 0β d V = - 0 - V k f 0β d V :
∂V j Tβ ∂V j V 0
ðV Þ T β3

In the second term, the contribution from components with j ≠ k turns to zero (in the
→r
reference frame where the mean velocity u is absent); therefore, the remaining
integral is given by

2
0 2 0
f α f 0β ðV Þ - V j
1=2
∂ →0 4 mβ ∂ →
- V dV = - p n
3=2 β →
fαV :
∂V j Tβ j ðV 0 Þ3 3 2π T β ∂V

The second term in the Landau collision integral Eq. (1.36) is simplified analo-
gously. Finally, the kinetic equation with the simplified collision operator has the
form

df α ∂ → T β ∂f α
= ναβ → V fα þ , ð1:46Þ
dt ∂V mα ∂ →
V

where a collision frequency ναβ is defined as

p 1=2
2mβ ΛZ 2α Z 2β e4 nβ
ναβ = 3=2
: ð1:47Þ
12π3=2 ε20 mα T β

Equation (1.46) is known as the Fokker-Planck equation. In homogeneous plasmas

in the absence of external forces, this equation describes the process of relaxation of
a distribution function fα to the Maxwellian distribution function with the tempera-
ture Tβ

3=2
mα mα V 2
α = nα
fM
2πT β
exp -
2T β

-1
with the characteristic time scale τ = ναβ . Indeed, this Maxwellian distribution
M
function f α turns to zero the r.h.s. of Eq. (1.46).
The Fokker-Planck equation is a linear equation and can be solved analytically.
Let us demonstrate how the Fokker-Planck equation can be used to calculate the
mobility of impurities – the coefficient that connects the applied electric field and
1.4 Fokker-Planck Equation 19

mean (fluid) velocity of impurities. In the stationary homogeneous plasma in the

electric field, the Fokker-Planck equation has the form

→
Z α e E ∂f α ∂ → T β ∂f α
= ναβ → V fα þ → : ð1:48Þ
mα ∂ → V ∂V mα ∂ V

→
Let us multiply Eq. (1.48) by V and integrate over velocities. After integrating by
part, one obtains that the first term on the r.h.s. is proportional to the particle flux
→
nα u α , while the second term turns to zero. Integral on the l.h.s. is equal
→
to - Z α nα e E =mα . Finally,

→ → Zαe
u α = bα E , b α = : ð1:49Þ
mα ναβ

Thus, the mobility is given by Eq. (1.49) with the numerical coefficient equal to
unity. The distribution function can also be obtained from Eq. (1.48). In relatively
small electric fields when the fluid velocity of impurities is much smaller than the
thermal velocity, the solution can be sought in the form of a sum of the Maxwellian
distribution plus small correction: f = f M + f 1. In the linear approximation, the
distribution function on the l.h.s. of Eq. (1.48) can be taken as the Maxwellian one
fα =fM α since the electric field is small, while on the r.h.s. f α = f α can be kept since
1

the Maxwellian distribution function turns to zero the r.h.s. Then, it is easy to obtain

mα → →
f 1α = u αV fM
α , ð1:50Þ
Tβ

where the mean velocity is given by Eq. (1.49). In other words, the distribution
function is given by expansion of a shifted Maxwellian distribution

→ → 2
mα
3=2 mα V - u α
f α = nα exp -
2πT β 2T β
ð1:51Þ
3=2
mα mα V 2 mα → →
≈ nα exp - 1þ u V :
2πT β 2T β Tβ α

It is, however, worthwhile to note that in the general case, the distribution function in
the electric field does have such a simple form.
20 1 Plasma Kinetics

1.5 Runaway Electrons in Fully Ionized Plasma

Let us analyze distribution function of electrons in the homogeneous plasma without

a magnetic field in the weak electric field.

e3 Λn
E << ED = : ð1:52Þ
ε0 T e

Here, the field ED is known as the Dreicer field. This condition is equivalent to an
assumption that the electric force is smaller than the friction force acting on the main
body of electrons with thermal velocities V T e = 2T e =me :

ne4 Λ
enE << nme νe ðV Te ÞV Te  nme V Te :
ε0 m2e V 3Te

From this inequality, one obtains Eq. (1.52). In the opposite case, the main part of the
electrons is infinitely accelerated since the friction force is unable to balance the
electric force due to the inverse dependence of the friction force on the velocity.
In the weak electric field when the condition, Eq. (1.52), is satisfied, only a small
fraction of electrons is accelerated by the electric field, so the distribution function
→
has a tail in the direction - E (Fig. 1.2). The main bulk of electrons is approximately
described by the Maxwellian distribution function, while the flux to the tail in the
velocity space is caused by Coulomb collisions. Since the number of runaway
electrons is small (exponentially small, as shown below), the problem of runaway
electrons can be treated as a stationary problem. Indeed, the density in the main body
of the distribution function is very slowly decreasing or is compensated by a small
source of electrons.
Let us restrict ourselves by calculating the number of electrons per second that are
accelerated and become runaway electrons, i.e., we shall find stationary flux in the
velocity space towards larger energies. The kinetic equation in the electric field is
given by

Fig. 1.2 Surfaces of the

constant distribution
function in the electric field
1.5 Runaway Electrons in Fully Ionized Plasma 21

→ →
∂f e E ∂f →V
- → þ ∇ →V  Γ = 0: ð1:53Þ
∂t me ∂ V

Let us introduce spherical coordinates in the velocity space with the z axis in the
→
- E direction. The second term on the l.h.s. of Eq. (1.45) can be rewritten in the
form (dependence on azimuth angle is absent due the symmetry of the problem)

→
e E ∂f eE ∂f sin θ ∂f
- → = m cos θ -
me ∂ V e ∂V V ∂θ
eE cos θ ∂ 1 ∂
= V 2f - f sin 2 θ :
me V 2 ∂V V sin θ ∂θ

Divergence of the collisional flux in the velocity space (the third term on the l.h.s. of
Eq. 1.53)) in spherical coordinates has the form
→
→V 1 ∂ →
1 ∂ →
∇V  Γ
→ = V Γ
2 V
þ sin θΓ V
θ :
V 2 ∂V V V sin θ ∂θ

We are interested in the collisional flux averaged over the angles in the velocity
space

2π π
→ →
V 1
ΓV = dφ dθ sin θΓVV :
4π
0 0

Now, one can also average the kinetic equation for fast electrons over the angles
assuming cosθ ≈ 1 since the distribution function of fast electrons is strongly
→
elongated in the - E direction. From Eq. (1.53), one obtains

→
∂f eE ∂ 1 ∂ V
þ V 2f þ 2 V 2 ΓV = 0: ð1:54Þ
∂t me V ∂V
2
V ∂V

The average distribution function f is sufficient to calculate the net flux in the
velocity space. In the collision operator, it is sufficient to take into account collisions
of the fast electrons with the slow Maxwellian background. Collisions of fast
electrons with ions contribute little to the average flux in energies due to the small
mass ratio, and collisions between fast electrons are negligible.
According to Eq. (1.35), a component of fast electron flux in the velocity space is
22 1 Plasma Kinetics

→0 →0
∂f M V fM V
→
Λe 4
f ∂f →0
ΓjV = U jk - dV
8πε20 me me ∂V 0k me ∂V k
ð1:55Þ

Λe4 →0 T e M → 0 ∂f →0
=- U jk V 0k ff M V þ f V dV :
8πε20 me T e me ∂V k

Here, f M is the unshifted Maxwellian distribution function. Let us evaluate the first
term in the integrand. In the zero approximation, the relative velocity coincides with
→r →
the fast electron velocity u = V , and tensor Ujk is independent of V′ so that the
integral is zero. In the first approximation using the expansion ur =
V 1 - V j V j0 =V 2 and expanding the tensor Ujk, one obtains

0 0
→0 →0 δjk V i V 0i V 0k V j V k V k V j V k V 0k V j V k 3V i V 0i 0
U jk V 0k ff M V dV = 2
þ 3
þ 3
- 3 Vk
V V V V V V2
:
→0 →0
× ff M V dV

ð1:56Þ

The last and the previous terms cancel during integration, since

V k V 0k V i V 0i M → 0 → 0 V2 →0 →0
V 0k f M V d V
2
3 2
f V d V = 3 k2
V V
V2 →0 →0
= k2 ðV 0 Þ f M V d V :
2
V

The first two terms in Eq. (1.56) are identical. Hence,

→0 →0 2T e
U jk V 0k ff M V dV = V jf :
me V 3

The second term in Eq. (1.53) can be evaluated analogously. One has to take into
account the inequality V⊥ < < V, and expansion of the tensor Ujk up to the second-
*
order terms proportional to (V′)2 is needed. Finally, the flux in the V direction is

→
T e ∂f
ΓVV = - νe ðV Þ Vf þ ,
me ∂V

where
1.5 Runaway Electrons in Fully Ionized Plasma 23

ne4 Λ
νe ðV Þ =
4πε20 m2e V 3

is the collision frequency of fast electrons. After averaging over the angles in the
velocity space

→
V T e ∂f
ΓV = - νe ðV Þ Vf þ : ð1:57Þ
me ∂V

The net flux in the velocity space is as follows from Eq. (1.54) is given by
→
V eE
SV = ΓV þ f: ð1:58Þ
me

After neglecting the time derivative in Eq. (1.54), one obtains the conservation of the
net flux in the velocity space

4πV 2 SV = const = n_ r : ð1:59Þ

Here, a constant n_ r is the number of fast electrons per second escaping in the velocity
space, which is to be found. This equation should be considered as the equation for
the averaged distribution function of fast electrons. In the dimensionless form,

e α
Vm1=2 1=2
E
α= , u= :
ED T 1=2
e

Equation (1.59) can be rewritten as

df u - u3 n_ r m2e
þ f = - Cu, C= : ð1:60Þ
du α 16π2 ne2 Λα1=2

The solution of Eq. (1.60) is given by

u
u0 1 u4
f = F - CF du0 , F = A exp - u2 : ð1:61Þ
F ð u0 Þ 2α 2
0

→
When u → 0, the distribution function f → f ðM Þ . This condition determines the
constant А:
24 1 Plasma Kinetics

3=2
me
A=n :
2πT e

For u → 1, the function F → 1, while the function f remains finite, so the ratio
f =F → 0. This condition determines the constant С

1 1
3=2
-1 u 1 2πT e 1 u4
C = du = exp - - u2 udu:
F n me 2α 2
0 0

This integral can be calculated using the Laplace method. Finally, the number of
runaway electrons per time unit is proportional to

n_ r = βðαÞnνe ðV Te Þ expð- E D =4E Þ: ð1:62Þ

A function β(α) is some dimensionless power function of the parameter α, which

cannot be found in the framework of approximation considered. The reason is
connected with the fact that the Dreicer electric field, Eq. (1.52), is proportional to
the Coulomb logarithm Λ, and the latter is under the exponent in Eq. (1.62). Since
the Coulomb logarithm Λ is specified with logarithmic accuracy, the quantity β can
be correctly calculated using only the next approximation. Therefore, Eq. (1.62)
determines the number of runaway electrons with logarithmic accuracy.
For an electric field larger than the Dreicer electric field, the main mechanism of
runaway electron generation is associated with the so-called avalanche effect mul-
tiplication of fast electrons due to low impact parameter collisions of fast electrons
with thermal electrons. During such collisions, electrons with velocities of the order
of the speed of light c produce new fast electrons, and the generation speed is
proportional to the existing number of fast electrons. The collision frequency for
the electrons moving with the speed of light in accordance with Eq. (1.28) is

Λe4 n
νe ðcÞ = : ð1:63Þ
4πε20 m2e c3

If the electric field is larger than the critical electric field (Ec = ED(c))

Λe3 n
Ec = , ð1:64Þ
ε0 m e c2

the fast electrons can transfer part of their energy to the slow electrons. For
E > > Ec, electrons with velocities V > V0 = c(Ec/E)1/2 are accelerated since
their friction force is smaller than the electric field. Since the Coulomb cross-section
is inversely proportional to the cube of the scattering angle and, therefore, inversely
proportional to the cube of the transferred velocity, such electrons are born as a result
of scattering of fast electrons with frequency νe = 4πεe2 mn2 V 3 . The number of such
4

0 e
1.6 Distribution Function of Electrons in Slightly Ionized Plasma 25

electrons born per second by an order of magnitude can be estimated as ( fr is a 1D

distribution function of fast electrons that is independent of the velocity)

c
fr c2 e4 n
νe f r dV = - 1 :
8πε20 c2 V 20 m2e
V0

Assuming that the distribution function of fast electrons is constant, so that their
density is nr = cfr, one obtains an estimate for the number of fast electrons born per
time unit

n_ r = nr νe ðcÞðE=Ec - 1Þ=ð2ΛÞ: ð1:65Þ

Here, the factor (E/Ec - 1) demonstrates that the avalanche effect has a threshold
character.

1.6 Distribution Function of Electrons in Slightly Ionized

Plasma

As in the previous section, we shall consider the electrons to be the test particles.
Then, the collision operator could be linearized, and the kinetic equation could be
significantly simplified. Slightly ionized plasma is defined as the plasma where the
following inequality is fulfilled:

νee << νeN , ð1:66Þ

where νee is the electron-electron collision frequency (for clarity, let us consider νee
to be identical to νe(V ) in the previous section) and νeN is the electron-neutral
collision frequency. However, the condition given by Eq. (1.66) is not sufficient to
treat electrons as test particles. There is another frequency responsible for the change
in electron energy in the electron-neutral collisions:

2me
νεeN = δeN νeN = ν : ð1:67Þ
mN eN

This frequency contains the electron to neutral mass ratio, while in electron-electron
collisions, the energy of the electron is changed practically during one collision at a
time scale  νee- 1 . Hence, the Coulomb collisions can be neglected only if a stronger
condition is satisfied:
26 1 Plasma Kinetics

νee << νεeN : ð1:68Þ

The intermediate case

νεeN << νee << νeN ð1:69Þ

also corresponds to slightly ionized plasma, and due to frequent electron-electron

collisions, the distribution function of electrons is close to Maxwellian.

→
1.6.1 Approximation f 0 , f 1

Let us assume the condition Eq. (1.66) to be fulfilled. Due to the small energy loss of
electrons in the electron-neutral collisions, the mean velocity of electrons is small
with respect to the chaotic velocity. As a result, the anisotropic part of the distribu-
tion function should also be small with respect to the isotropic part. Hence, it is
reasonable to seek the distribution function as an expansion over spherical har-
monics decaying with number:

1 l
→ → →
f r , V,t = fm
l l ðθ, φÞ:
r , V, t Y m ð1:70Þ
l=0 m= -l

The coefficients f m
l depend on the velocity absolute value. Spherical functions can be
expressed through the associated Legendre polynomials

l = Pl ðcos θÞ expðimφÞ:
Ym m

The first spherical functions are

Y 00 = 1, Y 01 = cos θ, Y 11 = sin θðcos φ þ i sin φÞ:

Let us truncate expansion Eq. (1.70) and keep the first two terms. Then, Eq. (1.70)
can be rewritten in the form

→ → (
→ → →
f 1 r , V, t V
f r , V , t = f 0 r , V, t þ : ð1:71Þ
V

This distribution function should be substituted into the kinetic equation for
electrons. Below is the result of the evaluation of different terms of the kinetic
equation. The first term:
1.6 Distribution Function of Electrons in Slightly Ionized Plasma 27

→ →
∂f ∂f V ∂f 1
= 0þ : ð1:72Þ
∂t ∂t V ∂t

The term with the spatial derivative has the form

→ → → → →
V ∇ →r f = V ∇ →r f 0 þ V ∇ →r f 1 V =V :

Using the identity known from vector algebra

→ → → → → → → → → →
∇ F  G = F ∇ G þ G∇ F þ F × ∇ × G þ G × ∇× F ,

one obtains

→
→ → V → →
V ∇ →r f = V ∇ →r f 0 þ V ∇ →r f 1 : ð1:73Þ
V

Other terms are evaluated analogously:

e → → →
- E þ V×B ∇ →V f
me
→ → → → →→ → ð1:74Þ
e E V ∂f 0 e E → f 1 eE f 1 e → → f 1
=- - V ∇ →V - - V×B :
me V ∂V me V me V me V

The collision integral is also separated into two components (for homogeneous
neutral gas at rest):

St eN = St 0eN þ St 1eN ,
→
which depends on f0 and f 1 correspondingly. Now, multiplying the kinetic equation
by (1/4π)dΩ, where dΩ is the solid angle in the velocity space, for simplicity, the
→
axis could be directed along f 1 . After integrating over the solid angle, we obtain the
averaged kinetic equation:

→
→ →
∂f 0 V → e E V ∂ f 1 =V eE →
þ ∇ →r  f 1 - - f = St 0eN : ð1:75Þ
∂t 3 me 3 ∂V me V 1

Multiplying the kinetic equation by quantities Vj/V and (3/4π)dΩ, after integrating
over the solid angle dΩ, one obtains three equations. After summation, we have one
equation in the vector form
28 1 Plasma Kinetics

→ →
∂f 1 e E ∂f 0 e → →
þ V∇ →r f 0 - þ B × f 1 = St 1eN : ð1:76Þ
∂t me ∂V me

The collision integral for the electron-neutral collisions can be obtained from the
general expression, Eq. (1.5). It is also possible to derive such an expression directly
assuming the neutral particles to be cold and staying at rest. Indeed, since the relative
velocity coincides with the velocity of electrons, the number of particles coming and
leaving the differential volume in the velocity space is (for brevity, the differential
cross-section is denoted as σ)

→ →0 →
St eN d V = nN σðV 0 , θÞV 0 f 0 dΩd V - nN σðV, θÞVfdΩd V :
Ω Ω

Electron velocities before and after collision are connected by the relation

1=2
2me
V0 = V 1 þ ð1 - cos θÞ :
mN

It follows from the above expression that dV′/V′ = dV/V. Hence,

! V 03 V 03 →
dV 0 = V 0 dV 0 dΩ =
2
dVdΩ = 3 d V :
V V

Substituting this expression into the collision integral, one obtains

nN
V 0 f 0 σðV 0 , θÞ - V 4 f σðV, θÞ dΩ:
4
St eN = ð1:77Þ
V3
Ω

Let us first calculate the collision integral Eq. (1.69) in the first approximation
with respect to ratio me/mN, i.e., the mass ratio is set to zero. In this approximation,
V′ = V and, therefore, f 0 = f 00 , so according to Eq. (1.77) St 0eN = 0. Substituting the
→ →
correction f 1 V =V into Eq. (1.77), we find

St 1eN = nN V jf 1 j σ V, θ cos θ0 - σ V, θ cos θ dΩ: ð1:78Þ

Ω

→ →0
In this equation, the scattering angle between vectors V and V is denoted as θ,
→
while θ, φ and θ′, φ′ are the angles related to the spherical coordinates for vectors V
→0
and V , respectively. Let us now turn from the integration over the scattering angle
to the equivalent integration over the angles θ′, φ′. Let us evaluate the first term
1.6 Distribution Function of Electrons in Slightly Ionized Plasma 29

under the integral. Expansion of the cross-section over Legendre polynomials is

given by

σ V, θ = σk Pk cos θ :
k

The cosine of the angle between two vectors is given by

cos θ = cos θ cos θ0 - sin θ sin θ0 cosðφ - φ0 Þ:

According to the addition theorem for spherical functions,

Pk ðcos θ cos θ0 - sin θ sin θ0 cosðφ - φ0 ÞÞ = Pk ðcos θÞPk ðcos θ0 Þ

k
ðk - mÞ! m 0 0
þ2 P ðcos θÞPm
k ðcos θ Þ cosðmðφ - φ ÞÞ:
m=1
ð k þ mÞ! k

Integrating the first term in Eq. (1.70) with account of orthogonal properties of
Legendre polynomials, one obtains

σ V, θ cos θ0 dΩ = σ1 cos 2 θ0 dΩ = ð4π=3Þσ1 :

Ω Ω

It is convenient to rewrite this expression again with account of orthogonality in the

form

σ V, θ cos θ0 dΩ = σ V, θ cos θdΩ:

Ω Ω

Combining two terms in Eq. (1.70), we find

→ →
St 1eN = - νeN ðV Þ f 1 V =V, ð1:79Þ

where

νeN ðV Þ = nN V ð1 - cos θÞσðV, θÞdΩ ð1:80Þ

Ω

is known as the transport collision frequency for electron-neutral collisions. After

averaging with the weight 3Vj/(4πV ), we have
30 1 Plasma Kinetics

→
St 1eN = - νeN ðV Þ f 1 : ð1:81Þ

To obtain a nonzero value of St 0eN , it is necessary to keep the terms of the order of
me/mN in the collision integral Eq. (1.77). Substituting the value f0 into Eq. (1.77)
and expanding the integrand in series, one obtains

nN ∂
V0 - V2
2
St 0eN = V 4 f 0 σ dΩ: ð1:82Þ
V3 ∂V 2
Ω

Finally, substituting the value V′ into Eq. (1.74), we have

2me 1 ∂
St 0eN = ν ðV ÞV 3 f 0 : ð1:83Þ
mN V ∂V 2 eN

Collison integrals given by Eqs. (1.81) and (1.83) represent the right-hand sides of
→
equations for f0 and f 1 and, therefore, close the system for isotropic and anisotropic
parts of the distribution function.

1.6.2 Distribution Function in the Electric Field

Let us find a distribution function of electrons in the homogeneous plasma in the

presence of a stationary homogeneous electric field in the absence of a magnetic
field. In this case, Eqs. (1.81) and (1.83) have the form

→
→ →
∂f 0 e E V ∂ f 1 =V eE → 2me 1 ∂
- - f = ν ðV ÞV 3 f 0 ,
∂t me 3 ∂V me V 1 mN V ∂V 2 eN ð1:84Þ
→ →
∂ f 1 e E ∂f 0 →
- = - νeN ðV Þ f 1 :
∂t me ∂V

Since there is only one featured direction along the electric field, the distribution
→ →
function f 1 k E . As follows from the second equation, the characteristic time scale
-1
for establishing stationary distribution function is given by νeN , i.e., occurs practi-
cally at a single collision and corresponds to the relaxation time scale for the
momentum. In the stationary case

→
→ eE ∂f 0
f 1= : ð1:85Þ
me νeN ðV Þ ∂V

The equation for f0 can be rewritten in the form

1.6 Distribution Function of Electrons in Slightly Ionized Plasma 31

→
→
1 e E ∂ f 1V
2
∂f 0 m 1 ∂
- = e 2 ν ðV ÞV 3 f 0 : ð1:86Þ
∂t 3V 2 me ∂V mN V ∂V eN
→
After substituting f 1 into Eq. (1.86), we have

∂f 0 1 e2 E 2 ∂ V 2 ∂f 0 m 1 ∂
- = e 2 ν ðV ÞV 3 f 0 : ð1:87Þ
∂t 3V me ∂V νeN ðV Þ ∂V
2 2 mN V ∂V eN

The process of f0 relaxation has a diffusive character, as seen from Eq. (1.87), the
-1
stationary distribution function is established at a νeN mN =me time scale. The
stationary distribution function can be easily found by integrating Eq. (1.87) for
∂f0/∂t = 0:

V
3m3e V 0 ν2eN ðV 0 Þ 0
f 0 = A exp - dV : ð1:88Þ
e2 E 2 m N
0

This distribution function of electrons is determined by the velocity dependence of

the electron-neutral collision frequency.
If the collision frequency is velocity independent (such a situation is typical for
Не in a wide range of velocities), then according to Eq. (1.88)

me V 2
f 0 = A exp - , ð1:89Þ
2T eff

where

e2 E2 mN
T eff = : ð1:90Þ
3m2e ν2eN

For the constant collision frequency, the distribution function is a Maxwellian

function with an effective temperature given by Eq. (1.90). Since in accordance
with Eq. (1.87), the process of gaining energy has a diffusive character with the
-1
diffusion coefficient DV  e2 E 2 = m2e νeN ; then, during time scale νeN mN =me , the
-1
average square of the velocity reaches V  DV νeN mN =me , and the average energy
2

is of the order of that given by Eq. (1.90).

As a second case, let us consider a constant mean free path λeN. The collision
frequency is then proportional to the velocity: νeN(V ) = V/λeN~V. According to
Eq. (1.88)

3m3e V 4
f 0 = A exp - : ð1:91Þ
4e2 E 2 mN λ2eN
32 1 Plasma Kinetics

This distribution function is known as the Druyvesteyn distribution function. Here,

the velocity dependence is much stronger than for the case of the Maxwellian
distribution function.

1.6.3 Impact of Electron-Electron Collisions

When the condition Eq. (1.69) is satisfied, electron-electron collisions have a strong
impact on the distribution function of electrons. On the r.h.s. of equations for f0 and
→
f 1 (1.75) and (1.76), it is necessary to take into account the electron-electron
collision integral Stee. This additional collision integral in the equation for f0 is of
the order of νeef0, while the electron-neutral collision term St0eN  ðme =mN ÞνeN f 0 .
Hence, when condition Eq. (1.69) is fulfilled, the main term on the r.h.s. of the
equation for f0 is Stee( f0, f0). Weak electric fields and gradual density and tempera-
ture gradient terms on the l.h.s. of Eq. (1.75) are small with respect to νeef0, and the
solution of the equation for f0 is a function that turns to zero the r.h.s. Stee( f0, f0) = 0,
i.e. it coincides with the Maxwellian distribution function f M. On the other hand, on
→
the r.h.s. of the equation for f 1 Eq. (1.76), the additional term from electron-
*
electron collisions St ee  νee f 1 remains small with respect to St 1eN . Therefore,
→
the equation for f 1 remains the same as that without electron-electron collisions,
→
and f 1 is expressed through f0 according to Eq. (1.76), where f0 is equal to the
Maxwellian distribution function.
In the inhomogeneous plasma, this approach is valid if the characteristic spatial
scale of the inhomogeneity L exceeds the relaxation length of the distribution
function λf. Relaxation of f0 to f M takes place at a time scale νee- 1 ; hence, the length
λf corresponds to the random walk shift during νee- 1 . Along the magnetic field, the
step of the random walk is the mean free path λeN, and across the magnetic field, the
step is the electron gyroradius ρce. Hence,

λf k = λeN ðνeN =νee Þ1=2 , λf ⊥ = ρce ðνeN =νee Þ1=2 :

→
1.6.4 General Expression for f 1

→
The general solution of Eq. (1.76) for f 1 in the case of slow processes when the time
derivative can be neglected has the form

→
→ ∂f 0 e E ∂f 0
f 1= -M V → - m , ð1:92Þ
∂r e ∂V

where
1.7 Transport Coefficients for Electrons in Slightly Ionized Plasma 33

νeN ωce
- 0
ω2ce þ ν2eN ω2ce þ ν2eN
M= ωce νeN ð1:93Þ
0
ω2ce þ ν2eN ω2ce þ ν2eN
0 0 1=νeN

Here, ωce = eB/me. The function f0 is close to the Maxwellian function provided the
condition Eq. (1.69) is satisfied.

1.7 Transport Coefficients for Electrons in Slightly Ionized

Plasma

Particle and heat fluxes in slightly ionized plasma with a distribution function close
to the Maxwellian one could be obtained by direct integration of the distribution
function. The particle flux of electrons is

→ → 1
→ → → f 1V → 4π →
Γe=nue= V f0 þ dV = f 1 V 3 dV: ð1:94Þ
V 3
0

→
Since the function f 1 is linear with respect to density, temperature and potential
gradients, the particle flux can be written in the form

→ → T
Γ e = - De ∇n - be E n - De n∇T e =T e , ð1:95Þ
T
where De is a diffusion tensor, De is a tensor of thermal diffusion, and be is a
→
mobility tensor. Substituting f 1 as a function of f0, Eq. (1.92) into Eq. (1.94), for f0
equal to the Maxwellian distribution function, one obtains

1
4πe
be = V 4 M ðV Þf 0 dV,
3nT e
0
1
4π
De = V 4 M ðV Þf 0 dV, ð1:96Þ
3n
0
1
T 4π me V 2 3 dD
De = V 4 M ðV Þ - f dV = T e e :
3n 2T e 2 0 dT e
0

These tensors can be calculated when the velocity dependence of the electron-
neutral collision frequency νeN(V ) is known. According to Eq. (1.96), the mobility
and diffusion tensors are linked by the Einstein relation
34 1 Plasma Kinetics

Te
De = b: ð1:97Þ
e e

The heat flux is defined as

→ →
→ me → → 2 → → f 1V →
qe= V - ue V - ue f0 þ dV : ð1:98Þ
2 V

Assuming the mean velocity to be small with respect to the chaotic velocity, we have

1
→ 2πme → 5 →
qe= M ðV ÞV 5 f 1 dV - nT u : ð1:99Þ
3 2 e e
0

The electron heat flux can be written in the form

→ →
q e = - κe ∇T e þ cT nT e u e , ð1:100Þ

where κe is the thermal conductivity tensor and cT is a dimensionless tensor. Here,

the first term corresponds to the heat flux caused by the temperature gradient, while
the second term represents the heat flux caused by the mean velocity. Additionally, a
heat conductivity tensor χe is often used, which is connected to κe by the relation

3
κe = nχ : ð1:101Þ
2 e

Using Eq. (1.92), tensors κe ðχe Þ and cT can be calculated for known dependence
νeN(V ). They are linked to tensors of diffusion and thermal diffusion as

T
3 3 T dD T -1 T
κe = nχe = n De þ De - T e e þ De De De ,
2 2 dT e ð1:102Þ
T T -1
c = De De - I,

where I is the unity tensor. The diffusion and thermal diffusion coefficients are the
same order quantities.
When the collision frequency νeN is the velocity independent quantity, expres-
sions for diffusion, thermal diffusion and mobility coefficients have a simple form.
From (1.96) and (1.102), it follows that
1.8 Drift Kinetic Equation in a Stationary Electric and Magnetic Field 35

T Te 3
De = De = b = χ
e e 5 e
1 ωce =νeN
- 0
ω2ce =ν2eN þ1 ωce =ν2eN þ
2 1
ð1:103Þ
Te
= ωce =νeN 1 :
me νeN 0
ω2ce =ν2eN þ 1 ω2ce =ν2eN þ 1
0 0 1

For weak dependence νeN(V ) or when a detailed description of the transport

coefficients is not needed, a so-called elementary theory is used (elementary theory
corresponds to the quasihydrodynamic approximation, see the next chapter). In this
approximation, Eq. (1.95) is used with the average transport collision frequency
→
νeN ðT Þ = hνeN ðV Þi = νeN ðV Þf 0 d V =n:For the strong magnetic field ωce >> νeN(T ),
Eq. (1.95) remains exact for arbitrary dependence νeN(V ), which can be easily seen
from Eq. (1.96) and definition of νeN(T ). More accurate approximations for arbitrary
dependence νeN(V ) without magnetic field can be found in [7].
For ions, simplified expressions of the elementary theory are normally used:

T Te 3
Di = Di = b = χ
e i 5 i
1 ωci =νiN
0
ω2ci =ν2iN þ 1 ω2ci =ν2iN þ 1
ð1:104Þ
Ti
= ω =ν 1 :
mi νiN - 2 ci 2 iN 0
ωci =νiN þ 1 ω2ci =ν2iN þ 1
0 0 1

For the important case of ions in their own gas, the charge exchange collision
frequency is

1=2
32σex T i
νiN = nN : ð1:105Þ
3π1=2 mi

1.8 Drift Kinetic Equation in a Stationary Electric

and Magnetic Field

For fully ionized plasma immersed in a strong magnetic field, it is often useful to
switch to the description of the plasma by following the guiding centers of the
Larmor orbits. It can be done when the ion Larmor radius is small with respect to the
typical scale of the problem. In the guiding center approximation of the particle
36 1 Plasma Kinetics

motion, averaging on the cyclotron motion is performed, and only the guiding center
motion is considered. In this approximation, the guiding center moves with the
parallel velocity Vk along the magnetic field and across the magnetic field with the
drift velocity. The expression for the drift velocity is obtained by averaging the
particle motion equation over the small Larmor circle with the velocity V⊥. The
result of such averaging [8] is

•
→ → 1 → → mV 2⊥ → ∇B
R = Vk h þ 2
E×B þ h×
B 2ZeB B
ð1:106Þ
mV 2k → → →
þ h × h∇h ,
ZeB
→ → → →
where h = B =B. The second term on the r.h.s. is the E × B drift, the third term is
known as gradB drift, and the last term is the curvature drift. The expression for the
→ →
E × B drift is valid only for the weak electric field so that the drift velocity is smaller
than the perpendicular velocity: E/B << V⊥.
The variation in the parallel and perpendicular velocity is [8]:

→→
• e E h V 2⊥ → Vk → → → →
V k¼ þ div h þ E h × h∇ h
m 2 B
mV 2⊥ V k ∇B → → → mV 2⊥ V k → → →
- h × h∇ h - h rot h ∇ h ,
2ZeB B 2ZeB
• V kV ⊥ → V → → V → → ∇B
V ⊥¼ - div h - ⊥ h rot E þ ⊥ E h × ð1:107Þ
2 2B 2B B

V⊥ → → → → mV 2k V ⊥ ∇B → → →
þ E h h rot h þ h × h∇ h
2B 2ZeB B

mV 2k V ⊥ → → →
þ h rot h ∇ h :
2ZeB

The terms with the electric field represent the work of the electric field during particle
motion along the field and the work of the inductive electric field for the time-
dependent magnetic field at a particle rotating over its Larmor radius. The terms
connected with magnetic field inhomogeneity represent the change in the parallel
velocity while the magnetic moment and the full energy remain constant.
→
It is convenient to introduce five variables: R , μ, V k , where

mV 2⊥
μ= ð1:108Þ
2B

is the magnetic moment of a rotating particle and is adiabatic invariant of motion.

Then, the equation for the parallel velocity can be reduced to
1.8 Drift Kinetic Equation in a Stationary Electric and Magnetic Field 37

• μ Ze → Vk →
Vk = - ∇B þ ∇φ b þ ∇× b : ð1:109Þ
m m ωc

This expression can be derived by using μB þ mV k2 =2 þ Zeφ = ε = const

• d 2 μ dB dφ
Vk = ε - μB - Zeφ = - þ Ze ðε - - μB - eφÞ - 1=2 :
dt m m dt dt
ð1:110Þ

→_ →_
dt = R ∇B and dt = R ∇φ. With account of
dφ
Here, dB

1 → → mV 2⊥ → ∇B mV 2k → → →
2
E × B þ h × þ h × h∇h ðμ∇B - e∇φÞ = 0,
B 2ZeB B ZeB

→_
after substituting Eq. (1.106) for R into Eq. (1.110), we obtain Eq. (1.109).
The drift kinetic equation could be derived by averaging the Boltzmann kinetic
equation over gyromotion (see below). Alternatively, it is possible to introduce a
distribution function of the guiding centers in the corresponding 5-dimensional
→
space. This distribution function F depends on R , V k and μ. The phase space volume
element is defined to give the number of particles guiding centers in the space unit

2π →
dV = Bd R dV k dμ, dN = FdV : ð1:111Þ
m

In the absence of collisions, the continuity equation for the guiding center fluid in the
phase space leads to the Vlasov kinetic equation for the guiding center distribution
function

•
∂F → ∂F • ∂F
þ R → þ Vk = 0: ð1:112Þ
∂t ∂R ∂V k

This is the collisionless drift kinetic equation. It could also be extended to the
collisional case by using the collision integral when it is possible to neglect the
difference between the particle position and guiding center coordinate. Note also that
the calculation of a self-consistent magnetic field requires knowledge of real plasma
currents, including e.g. diamagnetic currents, which differ from guiding center
currents, so the guiding center kinetic equation cannot be used for this purpose.
38 1 Plasma Kinetics

1.9 Gyrokinetic Equation

The gyrokinetic equation is more general and can treat time-dependent cases in a
strong magnetic field for low frequencies. It can be used even for k⊥ρci~1, where k⊥
is a wave vector across the magnetic field, so waves and turbulence with scales of the
order of ion Larmor radius can be considered. It is also assumed that

ω ρ eφ δB
 ci  kk ρci    OðεÞ,
ωci L Te B ð1:113Þ
k⊥ ρci  Oð1Þ:

Let us start with Vlasov kinetic Eq. (1.2),

∂f α → ∂f α Z α e → → → ∂f α
þV →þ E þ V×B → = 0: ð1:114Þ
∂t ∂r m α ∂V

Now, we introduce gyrokinetic variables

→ → →
f α r , V , t → f α R , V k , μ, θ, t , ð1:115Þ

where the links between the initial and gyrokinetic variables are

→ → → → → → → → →
r = R þ ρ , ρ = h × V ⊥ =ωcα , V ⊥ = V ⊥ cos θ e 1 þ sin θ e 2 ,
→
ð1:116Þ
→ →
V = V k h þ V ⊥:

Here, θ is the gyrophase angle. The electric field is assumed to be electrostatic

→
E = - ∇φ. Let us express derivatives ∂∂→r and ∂→ through the new gyrokinetic
∂V
variables:

→
∂ ∂R ∂ ∂μ ∂ ∂V k ∂ ∂θ ∂
→ = → → þ → þ → þ → ,
∂r ∂ r ∂R ∂r ∂μ ∂r ∂V k ∂ r ∂θ
→ ð1:117Þ
∂ ∂R ∂ ∂μ ∂ ∂V k ∂ ∂θ ∂
→ = → → þ → þ → þ → :
∂V ∂V ∂ R ∂V ∂μ ∂V ∂V k ∂ V ∂θ
∂
For simplicity, let us consider a uniform magnetic field. Since derivatives →
∂r
→
are taken at V = const, only the first term on the r.h.s. is nonzero, and since
→ → → →
r = R þ ρ and ∂ →ρ = 0, we have ∂∂→r = ∂→ . For derivatives over velocity
∂R ∂R
1.9 Gyrokinetic Equation 39

→→ →2
∂ V h ∂ V - V 2k mα =2B →
∂V k → ∂μ m V
→ = → = h, → = → = α ⊥,
∂V ∂V ∂V ∂V B
→ → → → →
→ ∂ r - h × V =ωcα ∂ h × V =ωcα
∂R
→ = → =- → :
∂V ∂V ∂V

Hence,

→ →
→ ∂ h × V =ωcα ∂
∂ ∂→ mα V ⊥ ∂ 1 → → ∂
→ = h þ - h × e⊥ - → *
∂V ∂V k B ∂μ V ⊥ ∂θ ∂V ∂R

Different terms in the kinetic equation are

→ ∂ → ∂ → ∂
V → = Vk h → þ V⊥ → ,
∂r ∂R ∂R
→ → →
→→ E h × V⊥ ∂ → →
Zαe → ∂ Z e ∂ E V ⊥ mα ∂ E×B ∂
E → = α Ek þ - þ ,
mα ∂ V mα ∂V k B ∂μ V 2⊥ ∂θ B2 ∂ →
R
→ → → →
Zαe → → ∂ V ×B B×V ∂ → → → ∂
V×B → = - ωcα þ V×B ×B →
mα ∂V B2 V 2⊥ ∂θ ∂R
∂ → ∂
= ωcα - V⊥ → :
∂θ ∂R

The electrostatic potential and electric field should also be expressed in the new
→ →
variables. Since φ = φ R þ ρ , we have

→ → → →→
∂φ ∂φ ∂ r ∂φ ∂ ρ V ∂φ E V⊥
= → = → = ⊥ → =- ,
∂θ ∂r ∂θ ∂r ∂θ ω cα ∂r ωcα

and

→→ 2
E V⊥ ∂ Zαe ∂φ ∂
Zαe =- mα :
B ∂μ mα ∂θ ∂μ

Now, we can rewrite the Vlasov equation in the new variables

40 1 Plasma Kinetics

→ →
∂f α → E×B ∂f α Z α e ∂f α ∂f Z eω ∂φ ∂f α
þ Vk h þ þ E þ ωcα α - α cα
∂t B2 ∂R
→ mα k ∂V k ∂θ B ∂θ ∂μ
→ → →
E × B V ⊥ ∂f α
- ωcα = 0:
B2 V 2⊥ ∂θ
ð1:118Þ

The distribution function can be sought in the form f α = f 0α þ f 1α with expansion

parameter ε, Eq. (1.113). Keeping the largest term in the zero-order approximation,
∂f 0
one obtains ωcα ∂θα = 0, i.e., f 0α is independent of the gyrophase angle. Therefore,
2π
f α = hf α i þ f α1 , where hf α i = 1
2π f α dθ.
0
In the first-order approximation,

→ →
∂f 0α → E×B ∂f 0α Z α e ∂f 0α ∂f 1
þ Vk h þ → þ Ek þ ωcα α
∂t B2 ∂R mα ∂V k ∂θ

Z α eωcα ∂φ ∂f 1α
- = 0: ð1:119Þ
B ∂θ ∂μ

The last small term in Eq. (1.118) is neglected. After averaging Eq. (1.119) over the
gyrophase angle, we have

→ →
∂ hf α i → E ×B ∂ hf α i Zαe ∂ hf α i
þ Vk h þ þ E = 0: ð1:120Þ
∂t B2
∂R
→ mα k ∂V k

This is the gyrokinetic equation in a uniform magnetic field in the electrostatic

approximation. In this equation, the electric field is averaged over the gyrophase
* 2π *
→
angle E = 1
2π E r dθ. In the general case of a nonuniform magnetic field
0
drift, the kinetic equation can be derived in a similar way and is given by

•
∂ hf α i → ∂ hf α i • ∂ hf i
α
þ R → þ Vk = 0, ð1:121Þ
∂t ∂R ∂V k

where

•
→ →1 → mα μ → ∇B
→ mα V 2k → → →
R = Vk h þ 2 E ×B þ h× þ h × h∇h ,
B Zαe B Z α eB
ð1:122Þ
1.9 Gyrokinetic Equation 41

• Zαe → μ → Vk →
Vk = E - ∇B b þ ∇× b : ð1:123Þ
mα mα ωcα

Now, we have to calculate the gyrophase average of the potential and electric field:

→ → → →→ →→
φ r =φ R þ ρ = φ →k exp i k R þ i k ρ
→
k
→→
ð1:124Þ
= φ →k exp i k R þ ik ⊥ ρ sin θ :
→
k

We use the Fourier-Bessel expansion

expðik ⊥ ρ sin θÞ = J n ðk⊥ ρÞ expðinθÞ: ð1:125Þ

n

The averaged value is

2π
1
h expðik ⊥ ρ sin θÞi = J n ðk⊥ ρÞ expðinθÞdθ = J 0 ðk⊥ ρÞ: ð1:126Þ
2π n
0

Finally,

→→
h φi = φ →k J 0 ðk⊥ ρÞ exp i k R : ð1:127Þ
→
k

With this expression, one can solve the gyrokinetic Eq. (1.121) with account of finite
Larmour radius effects.
An additional problem in the gyrokinetic approach arises from the fact that the
particle density and current calculated using the gyro-averaged distribution function
are different from the real particle density and current due to the guiding center
approach. For example, to calculate the real density, which one can use in Poisson’s
equation, one has to calculate the correction f 1α to the averaged distribution function.
This can be done using Eq. (1.119). The sum of the first four terms on the l.h.s. of
Eq. (1.119) is small, with an average field, it is zero, according to Eq. (1.120), so we
can neglect it. Integrating the remaining terms, we obtain

Z αe ∂ hf α i
f 1α = ð φ - h φi Þ : ð1:128Þ
B ∂μ

The real density is given by

42 1 Plasma Kinetics

→ B → → →
nα = 2π d R dV k dμ hf α iδ R þ ρ - r
mα
ð1:129Þ
→ B → → →
þ d R dV k dμdθ f 1α R þ ρ - r :
mα

1.10 Pellet Ablation in a Tokamak

Below we present an example of using the expression for electron energy loss
derived in this chapter to calculate the ablation rate of a hydrogen pellet in hot
tokamak plasma. Pellet injection into tokamak and stellarator plasmas is one of the
most promising methods for fueling reactors. Pellets are also used to diagnose the
tokamak plasma and to control its parameters. We shall consider here frozen
hydrogen (deuterium or tritium) pellets. The typical radius of such a pellet is a few
mm, and it is injected into hot plasma with a velocity of a few km/s.
The pellet injected into the tokamak plasma evaporates due to heat fluxes carried
by electrons and ions of the background plasma, producing a neutral gas cloud
around it, which then ionizes (see Figs. 1.3 and 1.4). Both neutral cloud and cold
plasma produced by ionization of neutrals shield the pellet from the hot plasma
particle flux, thereby governing the plasma particle energy loss and ablation rate.
Neutrals after ablation expand radially, while secondary plasma expands mainly
along the magnetic field, producing a cigar-like structure. Below, we shall consider
only neutral gas shielding, the so-called neutral gas shielding (NGS) model,
neglecting shielding of the cold secondary plasma. We shall consider hot plasma
electrons decelerating in an expanding spherical neutral gas cloud. This model
makes it possible to obtain simple formulas for the dependence of the ablation rate
_ cloud temperature TN, and cloud density nN on the pellet radius rp, the temper-
N,
ature T0 and density n0 of the background plasma, and the atomic mass MI of the
pellet material.
The main idea of the NGS model consists of the fact that almost all energy of
ambient electrons is spent in the neutral cloud surrounding a pellet, and only a very
small fraction of the energy flux reaches the pellet surface and evaporates pellet
material. This assumption is justified since frozen hydrogen pellets evaporate very
easily. Let r be the characteristic radius of the neutral gas cloud. Thus, the neutral
gas density nN and the expansion velocity V are related by

N_
nN = : ð1:130Þ
4πr2 V 
1.10 Pellet Ablation in a Tokamak 43

Fig. 1.3 Schematic representation of a cloud surrounding a pellet

The energy flux of the background electrons is of the order of qinq  n0 T 0 =m1e =2
3=2

→ 3=2
(ion flux can be neglected), and its divergence is ∇  q inq  n0 T 0 = r  m1=2e . The
latter estimate represents the heat that is released within the cloud and goes primarily
into the expansion of the cloud. Insignificant energy losses by ionization and
radiation emission are also ignored. A steady-state expansion is considered. There-
fore, from the estimate that the heat of ambient electrons is spent on neutral cloud
expansion, we have

→ 3=2
∇  q inq  n0 T 0 = r  m1=2
e  nN T N V  =r  :

According to Eq. (1.28), the deceleration of a monoenergetic electron beam in the

cloud by Coulomb collisions is described by the equation
44 1 Plasma Kinetics

Fig. 1.4 Image of a D2 pellet in ASDEX-Upgrade. The pellet is injected from the right to the left.
Short-time exposures are taken with a CCD camera from above of a pellet cloud. The locations of
the plasmoids correspond to the momentary position of the pellet, their radii roughly are the local
ionization radii, and their extent along the magnetic field lines represent the momentary length of
the ionized plasmoid fraction

∂V Λe4 Z 2 þ Z nN
= - νe V νe = : ð1:131Þ
∂t t=0 4πε20 m2e V 3

Here, the ion density is replaced by the neutral density because electrons with high
energy collisions with neutral particles are Coulomb collisions. We also took into
account collisions with Zn electrons (either free or bound). In our elementary
analysis, we ignore inelastic collisions and the electron distribution over pitch
angles. In Eq. (1.131), we replace ∂V
∂t t = 0
by VdV/ds assuming that electrons move
→
along the magnetic field, with s being a coordinate along B , and integrate over s to
obtain

1
V1 Λe4 Z 2 þ Z
1 4
V = nN ds, ð1:132Þ
4 V0 4πε20 m2e
s0
1.10 Pellet Ablation in a Tokamak 45

where s0 is the coordinate corresponding to the pellet surface, V0 is the electron

velocity at this surface, and V1 is the initial electron velocity in the background
plasma. Since the sublimation energy in hydrogen is small, the shielding should be
efficient enough for the energy of the electrons that reach the pellet surface to be very
low. Consequently, we can set V0 = 0. The electrons that reach the pellet surface
have initial velocities higher than the electron thermal velocity VT = (2T0/me)1/2;
these are the electrons from the tail of the distribution function. To take this effect
into account, we use the estimate V1 = αVT and retain the factor α > 1 in further
manipulations. Note that the case α = 1 corresponds to the approximation of the
Maxwellian electron distribution by a monoenergetic electron beam with an energy
equal to the electron temperature. This approach was used in the earliest versions of
the NGS model. The integral in Eq. (1.132) can be estimated as nNr, so Eq. (1.132)
can be rewritten as

Λe4 Z 2 þ Z
α4 T 20 = nN r  : ð1:133Þ
4πε20

This relationship implies that the mean free path of the background electrons should
be on the order of the cloud radius r. The set of Eqs. (1.130) and (1.133) should be
closed by the additional assumption that the neutral cloud expands with the sound
speed

V = T N =M I = T N = μI m p , ð1:134Þ

with mp being the proton mass. Possible numerical coefficients are neglected here.
With the same accuracy, we can assume that r = rp because the pellet radius is the
only characteristic dimension in the problem under consideration (the mean free path
of the background electrons is of the order of rp).
After some algebra from Eqs. (1.130), (1.133), and (1.134), we obtain

- 2=3 4=3 1=3 11=6 - 1=3

N_ = 7:2 × 1023 α8=3 Z 2 þ Z r p n0 T 0 μ I , ð1:135Þ

2=3 2=3 2=3 - 1=3 1=3

T N = 2:33α - 8=3 Z 2 þ Z r p n0 T 0 μI , ð1:136Þ

α4 T 20
nN = 3:84 : ð1:137Þ
Z 2 þ Z rp

The temperature T0 of the background plasma is expressed in kiloelectronvolts, the

density n0 of the background plasma in units of 1013cm-3, the pellet radius in
millimeters, the ablation rate in, the temperature TN of the neutral gas cloud in
46 1 Plasma Kinetics

electron volts, and the density nN of the neutral gas cloud in units of 1018cm-3. More
accurate treatment of neutral gas expansion, account of Maxwellian distribution of
fast electrons, account of contribution of shielding by ionized cigar-like structure,
etc., give very similar scaling for ablation rate.
The ablation rate strongly depends on the ambient plasma temperature and has a
rather weak dependence on its density. For a given pellet radius and velocity,
knowing the radial density and temperature profiles of the ambient plasma, from
Eq. (1.135), one can calculate pellet penetration depth in a tokamak or stellarator.
For present day tokamaks, frozen pellets with radii of the order of a few mm should
have initial velocities of the order of km/s to be able to penetrate deep into the hot
plasma. Note that in Fig. 1.4, the pellet ablation rate decreases at the end of its
trajectory (left image) because of the smaller pellet radius since almost all of the
ablation material has evaporated by this time.
Chapter 2
Transport Equations

2.1 Momentum Equations

Momentum balance equations are derived by multiplying the kinetic equation by the
factor VjVk. . .Vn and integrating over velocities. As a result, an infinite coupled chain
of momentum equations is obtained. The first 5 momentum equations are obtained
by multiplying the kinetic equation by 1, mαVj and mαV2/2.
Multiplying by unity and integrating over velocities leads to the particle conti-
nuity (or particle balance) equation

∂nα →
þ ∇  Γ α = 0: ð2:1Þ
∂t

We consider here only elastic collisions; therefore, the integral of the collision
operator over velocities is zero – no particles are produced or disappear. If inelastic
collisions are present, the corresponding terms (ionization, recombination, etc.) can
be added to the r.h.s. of Eq. (2.1). Equation (2.1) describes particle conservation.
After integrating the kinetic equation with the weight mαVj, one obtains three
equations

∂Γαj ∂M αjk → →
mα þ = Z α enα Ej þ u α × B þ Rαj : ð2:2Þ
∂t ∂xk j

→ → *
The moments of the distribution function M αj,k...n = V j V k . . . V n f α r , V , t dV
are defined according to Eq. (1.6).

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 47

V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_2
48 2 Transport Equations

The quantity

→
Rαj = mα V j St α d V ð2:3Þ

is known as a friction force. Similar to the collision operator, the friction force is a
→ →
sum of particles of species β : R α = R αβ . Combining Eq. (2.2) with the particle
β
continuity equation (2.1), separating velocities in the tensor Mαjk into the chaotic and
mean velocities, one obtains

∂uαj ∂uαj ∂p ∂παjk → →

mα nα þ uαk =- α- þ Z α enα E j þ u α × B þ Rαj : ð2:4Þ
∂t ∂xk ∂xj ∂xk j

Here, partial pressure is defined as

m α nα → → 2
pα = V - uα = nα T α , ð2:5Þ
3

and the quantity

→ → 2
παjk = παkj = mα nα V j - uαj ðV k - uαk Þ - δjk V - u α =3 ð2:6Þ

is known as the viscosity tensor.

Equation (2.4) is known as a momentum balance equation, and physically, it is
the equation of motion. Its l.h.s. contains a hydrodynamic derivative that corre-
sponds to the acceleration of a given element volume. The r.h.s. is the sum of all
forces applied to this volume element: partial pressure gradient, viscous force,
electric force, Lorentz force, and friction. The vector form of Eq. (2.4) is given by

→
∂uα → → $ → → → →
mα nα þ u α ∇ u α = - ∇pα - ∇  π α þ Z α enα E þ u α × B þ R α : ð2:7Þ
∂t

Integration of the kinetic equation with the weight mαV2/2 results in an energy
balance equation

∂ m α nα 2 3 ∂ m α nα 2 5
uα þ n α T α þ u þ n T u þ παjk uαk þ qαj
∂t 2 2 ∂xj 2 α 2 α α αj ð2:8Þ
= Z α eE j uαj nα þ Rαj uαj þ Qα :
2.1 Momentum Equations 49

Here

→ mα nα → → 2 → →
qα= V - uα V - uα ð2:9Þ
2

is the heat flux, and

→ → 2
mα V - u α →
Qα = St α d V ð2:10Þ
2

is the collisional heat production. It is also the sum of the species: Qα = Qαβ . On
β
the l.h.s. the first term represents the temporal change in the unit volume energy,
which consists of thermal and kinetic energies. The second term is the divergence of
→
the conductive heat flux q α and the convective flux of kinetic and thermal energies

m α nα 2 3 → →
∇ u þ n T uαþ qα ,
2 α 2 α α

plus work of pressure and viscosity

∂
pα δjk þ παjk uαk ,
∂xj

taken with the opposite sign. As a result, the coefficient in front of nαTα is equal to
5/2. The r.h.s. of Eq. (2.8) in addition to heat production Qα contains the work of the
electric field and that of the friction force, while the Lorentz force does not appear in
Eq. (2.8) since its work is zero because the Lorentz force is perpendicular to the
particle velocity.
The energy balance equation (2.8) can be combined with the continuity equation
and momentum balance equation (2.7) and rewritten in two equivalent forms:

3 ∂ 3 → → ∂uαj →
ðn T Þ þ ∇  nα T α u α þ nα T α ∇  u α þ παjk þ ∇  q α = Qα : ð2:11Þ
2 ∂t α α 2 ∂xk

3 ∂ → → ∂uαj →
n þ u α∇ T α þ nα T α ∇  u α þ παjk þ ∇  q α = Qα : ð2:12Þ
2 α ∂t ∂xk

These two equations describe the evolution of the thermal energy or temperature and
can be considered thermal balance equations.
The process of integrating the kinetic equation could be continued to obtain an
infinite system of moment equations. Since this system is derived from the kinetic
equation, it is rigorously correct. However, it cannot be used practically in the
general case. Indeed, to obtain the density profile from the particle continuity
→ →
equation, one needs to know the average velocity u α . To obtain u α from the
50 2 Transport Equations

momentum balance equation, one needs to have partial pressure pα, viscosity tensor
→ →
παjk and friction force R α . There are new quantities q α and Qα in the energy balance
equation, etc.
To close the system of particle, momentum, and energy balance equations, it is
→ →
necessary to express the quantities παjk, q α , R α , and Qα through other quantities nα,
→
u α , Tα, and their spatial derivatives. Such a procedure is possible in hydrodynamic
or fluid approximation when the typical spatial and temporary scales are large and
collisions are sufficiently frequent. The criterion is the small value of the mean-free
pass λα along the magnetic field or ion gyroradius ρcα across the magnetic field with
respect to plasma spatial scales Ln = |d ln n/dx|-1(or LT = |d ln T/dx|-1, etc.). The
typical time scale should be larger than the inverse collision frequency να- 1 and
-1
inverse cyclotron frequency ωcα . The electric field should also be small enough, for
fully ionized plasma, for example, this means E << ED, where ED is the Dreicer
field, Eq. (1.52). In an inhomogeneous magnetic field, the criterion for the applica-
bility of the fluid approximation could be more severe due to the guiding center drift,
and the corresponding condition is Δr << Ln, LT, where Δr is the guiding center
shift between the collisions. Special treatment requires regions at the plasma bound-
aries restricted by material surfaces, where plasma density tends to zero and the scale
Ln becomes small. In these regions, kinetic description is required.
→ →
In the thermodynamics of irreversible processes, the quantities R αβ , q α , παjk, and
Qαβ are called thermodynamic “fluxes” jm. The factors that are responsible for
→ →
deviation from thermodynamic equilibrium ∇Tα, Tα - Tβ, u α - u β and

∂uαj ∂uαk 2 →
W αjk = þ - δjk ∇  u α ,
∂xk ∂xj 3

are known as thermodynamic “forces” xn. The quantity Wαjk is known as the rate-of-
strain tensor. In the fluid approximation, “fluxes” jm are linear functions of “forces”
xn. The latter can be expressed as a linear function of the correction to the Maxwel-
lian distribution function f 1. The general linear form is given by

jm = Lmn xn , ð2:13Þ
n

and а matrix Lmn is known as a matrix of transport coefficients. Therefore, to

calculate “fluxes” in the momentum equations, one has to find the correction to the
Maxwellian distribution function f 1 and then use it to calculate transport coefficients.
Momentum equations with known transport coefficients are called transport
equations.
2.2 Transport Coefficients in Fully Ionized Plasma. Method of Chapman and Enskog 51

2.2 Transport Coefficients in Fully Ionized Plasma. Method

of Chapman and Enskog

We shall demonstrate how to calculate transport coefficients in the fully ionized

plasma. For simplicity, let us consider Z = 1. Let us seek the distribution function as
a series

f α = f 0α þ f 1α þ :: . . . ð2:14Þ

Here, f 0α is the Maxwellian distribution function (nonshifted due to mean velocity),

and f 1α is a small correction. Let us consider the electron distribution function.
In the kinetic equation

∂f e → ∂f e e → → → ∂f e
þV → - E þ V×B → = St e
∂t ∂r m e ∂V

the zero-order terms are the collision integral and the term with the magnetic field.
Other terms contain a small electric field, small density gradient, and small temporal
derivative since the processes are assumed to be slow enough.
The Maxwellian distribution function turns to zero electron-electron collision
operator

St ee f 0e , f 0e = 0, ð2:15Þ

while the electron-ion collision integral is small due to the small mass ratio me/mi.
Since the Maxwellian distribution function is a function of the velocity absolute
value, it also turns to zero the term with magnetic field:

→ → ∂f 0e
V×B → = 0: ð2:16Þ
∂V

Hence, one has to keep the correction f 1e in the electron-electron collision integral
and in the term with the magnetic field. In other “small” terms, it is sufficient to keep
only the zero-order distribution function f 0e . Thus, we have

∂f 0e → ∂f 0e e → ∂f 0e e → → ∂f 1e
þV → - E → - V×B → = St ee f e , f e
0 1
∂t ∂r me ∂ V me ∂V ð2:17Þ
þSt ee f 1e , f 0e þ St ei ðf e , f i Þ:

Here, the electron-electron collision operator is linearized.

Now, since the density and the temperature in the Maxwellian distribution
function are time-dependent quantities, one can express them through the spatial
52 2 Transport Equations

derivatives using the momentum equations. The resulting equation for correction f 1e
is an integro-differential equation. After solving this equation, it is possible to
express f 1e as a function of thermodynamic forces and then use f 1e to calculate
thermodynamic fluxes and finally obtain transport coefficients.
Let us demonstrate how this approach can be used for homogeneous isothermal
plasma in the absence of the magnetic field placed in a constant homogeneous
electric field. We start with the evaluation of the electron-ion collision operator

e4 Λ ∂ f e ∂f i f ∂f e →0
St ei = - U jk 0 - i dV : ð2:18Þ
8πε20 me ∂V j mi ∂V k me ∂V k

Due to the small mass ratio me/mi ion distribution function in Eq. (2.18) could be
replaced by the Maxwellian one. Since the relative velocity in the tensor Ujk is close
to the electron velocity, in the first approximation, we have Ujk = Vjk, where

δjk V jV k
V jk = - 3 : ð2:19Þ
V V

Neglecting small terms of the order of me/mi, one can keep only the second term on
→0
the r.h.s. of Eq. (2.18). After integration over V , one obtains

e4 nΛ ∂ ∂f
St ei = V jk e : ð2:20Þ
8πε20 m2e ∂V j ∂V k

First, let us calculate the friction force in the so-called quasihydrodynamic

approximation when the electron distribution function is assumed to be the Max-
→ → →
wellian one shifted by u e - u i = u (in the reference frame where the ion mean
velocity is set to zero). For the shifted Maxwellian function

me → →
f e = f 0e 1 þ uV ,
Te

using identity VjkVk = 0, one finds

→0 → → ne4 Λ → ∂ ∂ →→ →
R ei = me V St ei d V = V V jk f0 u V dV
8πε20 T e ∂V j ∂V k e
ne4 Λ → ∂ →
= V V jk f 0e uk d V :
8πε20 T e ∂V j

Using the relation

∂V jk 2V
= - 3k
∂V j V
2.2 Transport Coefficients in Fully Ionized Plasma. Method of Chapman and Enskog 53

one obtains

→0 →
R ei = - me nνei u : ð2:21Þ

Here, the electron-ion collision frequency is defined as

p
2 ne4 Λ
νei =  νe : ð2:22Þ
12π3=2 ε20 1=2
me T 3=2
e

The numerical coefficient in Eq. (2.21) is equal to unity. This is the result of the
assumption that the electron distribution function is Maxwellian. In reality, however,
the electron distribution function differs from the shifted Maxwellian function since
fast electrons have fewer collisions than cold electrons and the distribution function
is deformed and stretched in the direction opposite to the electric field. As a result of
this deformation, the correct numerical coefficient in Eq. (2.21) becomes almost
twice as smaller than unity.
To correctly calculate the friction force, it is convenient to shift to the reference
→
frame moving with the mean electron velocity u e . The electron-ion collision
operator can be divided into two parts:

→
e E ∂f 0e
St ee f 0e , f 1e þ St ee f 1e , f 0e þ St ei f e , f 0i = - 0
→ - St ei f e , f i - f i : ð2:23Þ
me ∂ V

The term St ei f e , f 0i is added to the electron-ion collision integral, and the same term
is subtracted from it. Here, f 0i is the ion distribution function shifted to make the
ion mean velocity coincide with the electron mean velocity. The summand
St ei f e , f i - f 0i , which arises on the r.h.s of Eq. (2.23), contains a small quantity
f i - f 0i , and, therefore, it is possible to put f e = f 0e in it. This term is small since it
contains a small mass ratio me/mi. The correction f 1e is sought in the form

→
f 1e = f 0e Φ V : ð2:24Þ

Now let us substitute Eq. (2.24) into Eq. (2.23). The electric field in Eq. (2.23)
should be expressed through the friction from the momentum balance equation for
→ →
electrons en E = R ei . The term St ei f 0e , f 0i can be evaluated as follows. According to
Eq. (2.20)

e4 nΛ ∂
St ei f 0e , f 0i = - V jk V k f 0e ,
8πε20 me T e ∂V j
54 2 Transport Equations

where the main part of the tensor V jk is calculated for ions with the mean velocity
→
- u:

δjk V j þ uj ð V k þ u k Þ
V jk = → - →
:
→ → 3
V þ u V þ u

∂V jk
With an account of identities V jk ðV k þ uk Þ = 0 and ∂V j
=- 2V k
V3
, one finds

e4 nΛ → → 0
St ei f 0e , f 0i = - V u f e:
8πε20 me T e

As a result, Eq. (2.23) is reduced to the form

→0 →1
→→ R ei þ R ei
ne Λ V u 0
4 → 0
I ee ðΦÞ þ I ei ðΦÞ = f þ V f e: ð2:25Þ
8πε20 me T e V 3 e nT e

The friction force here is divided into two parts:

* →0 →1
R ei = R ei þ R ei :

→0
The first contribution R ei Eq. (2.21) is determined by the Maxwellian distribution
→
function f 0e shifted by u with respect to the ions. The second contribution is caused
by the correction equation (2.24) to the electron distribution function:

→1 → ←
R ei = me V I ei ðΦÞd V : ð2:26Þ

Here

I ei ðΦÞ = St ei f 0e Φ, f 0i ,
ð2:27Þ
I ee ðΦÞ = St ee f 0e , f 0e Φ þ St ee f 0e Φ, f 0e :

Therefore, Eq. (2.25) can be considered an integro-differential equation for the

correction to the electron distribution function.
It is assumed that the density, temperature, and mean velocity (the latter is zero in
the chosen reference frame) are determined by f 0e . Therefore, the correction should
satisfy the following conditions:
2.2 Transport Coefficients in Fully Ionized Plasma. Method of Chapman and Enskog 55

→ * → →
f 0e Φd V = 0, V f 0e Φd V = 0, V 2 f 0e Φd V = 0: ð2:28Þ

The solution of Eq. (2.25) should be vector-invariant and hence could be sought in
the form

→→
Φ = A V2 u V :

It is convenient to seek function A(V2) using expansion over associated Laguerre

polynomials

1
me 3=2 me V 2
A V2 = ak Lk ðxÞ, x= : ð2:29Þ
Te k=1
2T e

The associated Laguerre polynomials are orthogonal so that

1
ðp þ mÞ!
xm expð- xÞLm
p ðxÞLq ðxÞdx =
m
δpq : ð2:30Þ
p!
0

The first two polynomials are Lm0 = 1, L1 = m þ 1 - x. Expansion in Eq. (2.29) starts
m

from k = 1 to satisfy Eq. (2.28). Multiplying Eq. (2.25) by

1=2
4 me → 3=2 me V 2 →
- V Lk dV
15n T e 2T e

and integrating over velocities, one ends up with the infinite system of algebraic
equations

akl þ a0kl al = δ1k , k = 1, 2:: . . . , ð2:31Þ

where akl, a0kl are dimensionless matrices:

4 me 3=2 3=2 →
akl = - Lk ðxÞV β I ee Ll ðxÞV β d V
15nνei 2T e
ð2:32Þ
4 me →
a0kl = -
3=2 3=2
Lk ðxÞV β I ei Ll ðxÞV β d V :
15nνei 2T e

The matrices akl and a0kl were calculated by Braginsky [9]. Truncating the expansion
equation (2.29) and using the few first polynomials, it is possible to solve the finite
algebraic system, Eq. (2.31), and find coefficients al and therefore obtain correction
to the distribution function.
The friction force calculated with account of this correction has the form
56 2 Transport Equations

→ →0 →1 → →
R ei = R ei þ R ei = - 0:51nme νei u e - u i : ð2:33Þ

Therefore, it is almost twice as small as that calculated under the assumption of the
shifted Maxwellian distribution function given by Eq. (2.21).
In the general case, the fluxes and forces are calculated in a similar way.

2.3 Summary of the Results for the Fully Ionized Plasma

Below is a summary of the results obtained for fully ionized magnetized plasma by
Braginsky [9]. Here, the ion charge number is set to unity Z = 1, and a strong
magnetic field ωce >> νei, ωci >> νii is assumed.
The friction force consists of two parts
→
→ → →u →T
R ei  R = R þR : ð2:34Þ

The first part of the friction force is connected with the relative velocity of electrons
and ions:
→
→u → → → → →
R = - nme νei 0:51 u k þ u ⊥ , u = u e - u i: ð2:35Þ

The electron-ion collision frequency in accordance with Eq. (2.22)

p
2 ne4 Λ
νei = , ð2:36Þ
12π3=2 ε20 m1=2
e Te
3=2

where the Coulomb logarithm is given by

Λ = 23:4 - 1:15lgn þ 3:45lgT e , T e < 50eV;

:
Λ = 25:3 - 1:15lgn þ 2:3lgT e , T e > 50eV:

The density in the expression for the Coulomb logarithm should be inserted in CGS
units, while the temperature is in eV. The second part is known as the thermal force
and depends on the electron temperature gradient:

→
→T 3 nνei B
R = - 0:71n∇k T e - × ∇T e : ð2:37Þ
2 ωce B
2.3 Summary of the Results for the Fully Ionized Plasma 57

Ion heat flux

→
→ 5 nT i B
q i = - κik ∇k T i - κi⊥ ∇⊥ T i þ × ∇T i , ð2:38Þ
2 eB B

where the heat conductivities along and across the magnetic field are:

3:9nT i 2nT i νii

κik = , κi⊥ = , ð2:39Þ
mi νii mi ω2ci

with the ion-ion collision frequency defined as

1 ne4 Λ
νii = : ð2:40Þ
12π ε0 mi T 3=2
3=2 2 1=2
i

The electron heat flux is a sum of two contributions:

→
→ → T →u
qe= qe þ qe : ð2:41Þ

The flux caused by heat conductivity has the form

→
→T 5 nT e B
qe = - κek ∇k T e - κe⊥ ∇⊥ T e - × ∇T e , ð2:42Þ
2 eB B

where

3:16nT e 2nT e νei

κek = , κe⊥ = : ð2:43Þ
me νei me ω2ce

The second part of the electron heat flux is caused by the relative velocities of
electrons and ions:

→
→
→u → 3 nT e B →
qe = 0:71nT e u k - ν ×u : ð2:44Þ
2 ωce ei B

Sources and sinks in the heat balance equations correspond to heat transfer
between electrons and ions and Joule heating. The latter exists only in the electron
equation:

3me
Qi = Q Δ = nν ðT - T i Þ,
mi ei e ð2:45Þ
→→
Qe = - R u - QΔ :
58 2 Transport Equations

The viscosity tensor consists of two parts:

→ →
$ $u $q
π=π þπ : ð2:46Þ

The first part depends on the mean velocities [9], and the second part depends on the
heat fluxes [10]. In the absence of a magnetic field, the first part is:

$u
→
∂uj ∂uk 2 →
π jk = - η0 W jk = - η0 þ - δjk ∇  u , ð2:47Þ
∂xk ∂xj 3

→ →
In the strong magnetic field ( z k B )
→
πzzu = - η0 W zz ,
→ η η
πxxu = - 0 W xx þ W yy - 1 W xx - W yy - η3 W xy ,
2 2
→ η0 η1
πyy = -
u
W xx þ W yy - W yy - W xx þ η3 W xy ,
2 2 ð2:48Þ
→ → η
πxyu = πyxu = - η1 W xy þ 3 W xx - W yy ,
2
→ →
πxzu = πzxu = - η2 W xz - η4 W yz ,
→ →
πyzu = πxyu = - η2 W yz þ η4 W xz :

Ion viscosity coefficients are

ηi0 = 0:96nT i =νii ,

nT ν
ηi1 = 0:3 i2 ii , ηi2 = 4ηi1 ,
ωci ð2:49Þ
nT i
ηi3 = , ηi4 = 2ηi3 :
2ωci

For electrons, the viscosity is usually small with respect to the ion viscosity. The
→
second part becomes important when the conductive heat flux q i is of the order of
→
the convective heat flux nT i u i , and the corresponding expressions can be found in
[10]. Viscosity heat production is mostly determined by the main ions, and its main
part is

2
∂V j η0 ∂V x ∂V y ∂V z
Qvis = - πjk = þ -2 : ð2:50Þ
∂xk 3 ∂x ∂y ∂z

Numerical coefficients for Z > 1 and intermediate values of the magnetic field can
be found in [9, 10].
2.4 Transport Coefficients in Fully Ionized Plasma. Qualitative Considerations 59

2.4 Transport Coefficients in Fully Ionized Plasma.

Qualitative Considerations

Let us consider qualitative mechanisms responsible for transport coefficients.

2.4.1 Friction Caused by the Relative Mean Velocity

and Thermal Force
→
→u
An estimate for the part of the friction force associated with the relative velocity R
can be easily obtained as a change of the momentum caused by relative mean
velocity of the electrons with respect to the ions multiplied by electron mass
→ →
me u e - u i and by particle density n during the time between collisions νei- 1 .
For the motion across the magnetic field, the electron distribution function is close to
→
the Maxwellian one shifted by u since the rotation over the Larmor radius reduces
the anisotropy of the distribution function. Corrections contain a small factor νei/ωce.
Hence, the perpendicular part of the friction force coincides with that obtained in the
quasihydrodynamic approximation given by Eq. (2.21). In the parallel direction, due
to distortion of the distribution function, the numerical coefficient and friction force
are almost twice as small as those in the perpendicular direction.
The second part of the friction force, which is proportional to the gradient of the
electron temperature, is known as the thermal force. This force is connected with the
velocity dependence of the electron-ion collision frequency for Coulomb collisions.
Let us consider electrons in the absence of mean velocity that collide with ions at rest
in the plasma without a magnetic field. Electron chaotic fluxes from the mean free
path distance through the unit area situated at z = z0, Fig. 2.1, have different
directions and are given by an estimate Γ+~Γ-~nVT with VT being the thermal
velocity. Due to the momentum loss of electrons during collisions, two forces
proportional to the corresponding fluxes arise R+~R-~menVTνei, which are applied
to the ions. In the plasma with a homogeneous electron temperature, these forces
balance each other so that the net force is zero. However, in the presence of a
temperature gradient, electrons that cross the area come from regions with different
electron temperatures. The temperature difference can be estimated as δTe = (dTe/
dz)λe, where λe is the mean free path. Hence, R+ ≠ R- and unbalanced force arises. If,
for example, the electron temperature increases from left to right, then since the
Coulomb collision frequency decreases with increasing velocity, the force R+ is
smaller than R-. As a result, we have an unbalanced force extracting ions into the hot
region. This force can be estimated as
60 2 Transport Equations

Fig. 2.1

Fig. 2.2

dνei dT
RT  me δT nV  n e : ð2:51Þ
dT e e T dz

According to Newton’s law, the same force with the opposite sign is applied to
electrons.
Across the magnetic field, electrons rotate over the Larmor circle, and, therefore,
an unbalanced force arises in the direction perpendicular to the temperature gradient
(Fig. 2.2). Since electrons bring their momentum from a distance of the order of ρce,
the temperature difference should be estimated as δTe = (dTe/dz)ρce. As in the
absence of the magnetic field, Eq. (2.51), the force applied to ions is

→
ν B
R  n ei
T
× ∇T e , ð2:52Þ
ωce B

while the same force with the opposite sign is applied to electrons.
2.4 Transport Coefficients in Fully Ionized Plasma. Qualitative Considerations 61

2.4.2 Spitzer Conductivity

In homogeneous plasma in the absence of a magnetic field, the electric force is

balanced by electron-ion friction. As a result, the current is flowing, and the density
is proportional to the electric field:

→ →
j = σk E : ð2:53Þ

Here, the Spitzer conductivity is defined as

ne2
σk = : ð2:54Þ
0:51me νei

Since the Coulomb collision frequency is proportional to the electron density, the
Spitzer conductivity is independent of the density and is determined by the electron
temperature, σk  T 3=2e . It is also worthwhile to note that in the fully ionized plasma,
mobility is not defined (in contrast to the slightly ionized plasma); it is possible to
obtain only the relative velocity of electrons with respect to the ions from Eq. (2.53)
but not the velocities itself. This fact is connected with the absence of the special
reference frame, which in slightly ionized plasma is bound to neutral gas.
In the literature, a quantity σ⊥ = ne2/(meνei) known as perpendicular conductivity
is often used, which, however, does not determine the real physical conductivity
across the magnetic field. Indeed, a homogeneous electric field applied across the
magnetic field could be turned to zero by the Lorentz transformation to the reference
→ →
frame moving with E × B =B2 velocity. Hence, the homogeneous electric field in
the fully ionized plasma does not cause current in the perpendicular direction. In this
sense, the perpendicular conductivity in fully ionized plasma is zero, as is the
mobility in the direction of the electric field.
In the plasma with Z > 1, a numerical coefficient in the expression for friction
force in Eq. (2.35) differs from 0.51. Therefore, the expression for Spitzer conduc-
tivity has the following form:

ne e 2
σk = ð2:55Þ
с me νei
u

with
p
2 ni Z 2 e 4 Λ
νei = : ð2:56Þ
12π3=2 ε20 m1=2
e Te
3=2
62 2 Transport Equations

The numerical coefficient is [9]:

сu = 0:51 for Z = 1, сu = 0:44 for Z = 2, сu = 0:4 for Z = 3, сu = 0:38 for Z = 4,

сu = 0:3 for Z = 1:

If plasma is a mixture of different species, it is convenient to introduce Zeff according

to

nj Z 2j nj Z 2j
j j
Z eff = = : ð2:57Þ
nj Z j ne
j

Since the friction force is a sum of friction forces due to collisions of electrons with
different species, assuming that all ions have velocities smaller than the electron
velocity so that the current is determined by the electron velocity, the Spitzer
conductivity can be rewritten as

e 2 ne
σk = f Z eff , ð2:58Þ
me νei Z eff

where the collision frequency of electrons is given by Eq. (2.36), and a function f is
approximated according to [10]

4:95Z eff
f Z eff = : ð2:59Þ
1 þ 1:54Z eff

For Zeff = 1, expression (2.54) is recovered. Note that in the literature, a simpler
approximation for Spritzer conductivity with f = 1 is often used, which differs
significantly from the correct approximation given by Eq. (2.59).

2.4.3 Heat Flux: Conductive and Convective Parts

Similar to the friction force, the heat flux is a sum of two contributions. The first
conductive part is a linear function of the temperature gradient and is caused by heat
conductivity. It has the form

→T 3
qα = - κα ∇T α = - nχ ∇T α : ð2:60Þ
2 α

Here, κα is the heat conductivity tensor, and χα is the heat diffusivity tensor. An
estimate for the tensor components along and across the magnetic field is obtained
similar to the case of slightly ionized plasma. The diagonal components of the heat
2.4 Transport Coefficients in Fully Ionized Plasma. Qualitative Considerations 63

diffusivity tensor are estimated as the square of the random walk step multiplied by
the collision frequency. Along the magnetic field (in the absence of it)

χek  λ2e νei , χik  λ2i νii : ð2:61Þ

Since the electron and ion mean-free paths for comparable temperatures are of the
same order, we have χek =χik  mi =me , i.e., the parallel electron heat diffusivity is
much larger than the ion heat diffusivity. Across the strong magnetic field, particles
are shifted at a distance of the order of Larmor radius, and, therefore,

χe⊥  ρ2ce νei , χi⊥  ρ2ci νii : ð2:62Þ

The perpendicular heat diffusivity coefficients for the comparable temperatures have
an inverse relation in contrast to the parallel ones -χi⊥ =χe⊥  mi =me . The estimate
for the heat fluxes along the temperature gradients can be easily obtained as the
difference in the heat fluxes brought from the mean free path or Larmor radius
distance. Along the magnetic field

dT α
qαk  qαkþ - qαk -  nV T δT α  - nV T λα ,
dz

while across the field

dT α
qα⊥  qα⊥þ - qα⊥ -  nρcα ναα δT α  - nρ2cα ναα :
dx

In the magnetized plasma, there are also off-diagonal Hall heat fluxes that
correspond to off-diagonal components of the heat conductivity tensor. The
corresponding flux is shown in Fig. 2.2 for electrons. Along the у-axis arises an
unbalanced heat flux through the unit area. A one-sided flux is given by an estimate
nVTT. Its unbalanced part is of the order of nVTρcdT/dx. As a result, for electrons

→T nT e →
q eΛ  - B × ∇T e : ð2:63Þ
eB2

A similar Hall heat flux for ions has the opposite sign. The flux equation (2.63) is
directly perpendicular to the temperature gradient along isotherms, and, therefore,
for a constant density, its divergence is equal to zero, i.e., this flux does not lead to a
change in the temperature. In the inhomogeneous magnetic field, the divergence of
the Hall flux equation (2.63) is not zero. Physically, its divergence coincides with the
divergence of the convective flux associated with the drift of the particles in the
inhomogeneous magnetic field.
64 2 Transport Equations

Fig. 2.3

For electrons, in addition to the conductive flux, there is also a convective flux
→
→u
associated with the relative velocity, q e in Eq. (2.41). Similar to the thermal force,
its origin is connected with the velocity dependence of the electron collision
frequency. Distortion of the electron distribution function with respect to the shifted
Maxwellian function in the absence of a magnetic field leads to a larger contribution
→
→ →u →
of fast electrons to the mean electron velocity u . As a result, heat flux q e  nT e u
arises.
The mechanism for the similar convective heat flux across the magnetic field is
illustrated in Fig. 2.3. The friction force associated with the electron velocity
decelerates electrons rotating over the Larmor radius and crossing the area y = 0
in the positive у-direction and accelerates those moving in the negative у-direction at
y = 0. The work acting on electrons meνeiuxρce results in the fact that the energy of
electrons crossing the area in the downwards direction is smaller than the energy of
electrons moving upwards. Multiplying the difference in energies meνeiuxρce by the
particle flux nVT, one obtains an estimate for the heat flux in the у-direction

→
→
→u nT B →
q eΛ  e νei ×u : ð2:64Þ
ωce B

2.4.4 Collisional Heat Production

During the collision of the light particle with the heavy particle, the fraction ~me/mi
of the energy is transported from one particle to another one. Estimating the energy
transported from electrons to ions as (3/2)(me/mi)n(Te - Ti) during the time scale
νei- 1 , one obtains the heat exchange rate between species
2.4 Transport Coefficients in Fully Ionized Plasma. Qualitative Considerations 65

me
QΔ  nν ðT - T i Þ:
mi ei e

This term enters the heat balance equations for electrons and ions with different
signs.
→→
The term - R u enters only the heat balance equation for electrons and consists
→
→u →
of two parts. The summand - R u represents the Joule heating of electrons and
can be rewritten in the form

→
→u → j2k j2
-R u= þ ⊥: ð2:65Þ
σk σ⊥

→T →
The second contribution - R u could be positive or negative, and the
corresponding heating is reversible.

2.4.5 Viscosity

Viscosity is associated with the transport of momentum. Let us consider plasma

without a magnetic field in the absence of temperature gradients. Consider, for
example, the case when the mean velocity in the y-direction uy is changing in the
x-direction. Similar to the heat flux, the difference in one-sided momentum fluxes
mnVTuy from the mean free path distance is estimated as

duy
πxy  πþ -
xy - πxy  mnV T δuy  - mnV T λ :
dx

After introducing the viscosity coefficient, we have

duy nT
πxy = - η0 , η0  : ð2:66Þ
dx ν

For ions, the ion-ion collision frequency should be substituted in Eq. (2.66);
therefore,

ηi0  nmi χik : ð2:67Þ

Substituting Eq. (2.66) into the momentum balance equation for ions, Eq. (2.4)
yields

2
∂uy ∂ uy
=D 2 :
∂t ∂x
66 2 Transport Equations

This is the diffusion equation with the diffusion coefficient D = Ti/(miνii). Thus, the
viscosity represents the diffusion of momentum in the direction of the velocity
gradient. Analogous transport of other velocity components takes place. In the
general case, the viscosity tensor is a linear function of the rate-off-strain tensor
given by Eq. (2.46): πjk = - η0Wjk. Tensor Wjk is a symmetrical tensor with a zero
→ → →
trace Wjj = 0. This tensor vanishes when plasma rotates as a whole u = ω × r or
→ →
stretches as a whole u  r . Viscosity components that are proportional to the heat
→ →
flux have the same order of magnitude if q α  nT α u α .
In magnetized plasma, the viscosity tensor has a more complicated character.
Components of the viscosity tensor that contain coefficients η1 and η2 correspond to
the reduction of the random walk step across the magnetic field to the Larmor radius
value. Components with coefficients η3 and η4 are known as gyroviscosity terms;
they are independent of the collision frequency and are inversely proportional to the
gyrofrequency. Gyroviscosity terms correspond to momentum flow in the Hall
direction analogously to the terms, Eq. (2.63), in the heat flow.
$
Note also that viscous forces ∇  π are small with respect to the pressure gradient
∇p by factors λ/Lk and ρc/L⊥ along and across the magnetic field, respectively.

2.5 Equation for Entropy

Entropy per single particle, for example, for electrons, is given by

3
se = ln T e - ln n þ const: ð2:68Þ
2

The equation for the entropy of electrons can be evaluated by combining the particle
balance equation with the heat balance equation. It has the form

→
∂nse → q QΔ
þ ∇  se n u e þ e þ = θe : ð2:69Þ
∂t Te Te

Here, entropy production per unit volume is

→ →→ 1
T e θe = - q e ∇ ln T e - R u - π W : ð2:70Þ
2 ejk ejk

The l.h.s. of this equation describes entropy variation in time, entropy transport in
space, and entropy transfer between electrons and ions. Substituting the expressions
for friction, heat flux, and viscosity from Sect. 2.3, one obtains the entropy produc-
tion in the form
2.5 Equation for Entropy 67

κek 2 κe⊥ j2k j2 1

T e θe = ∇k T e þ ð∇⊥ T e Þ2 þ þ ⊥ þ ηe0 W ejk W ejk : ð2:71Þ
Te Te σk σ⊥ 2

Therefore, in this form, it is clear that the entropy production is essentially positive.
For ions, analogously,

→
∂nsi → q Q
þ ∇  s i n u i þ i - Δ = θi , ð2:72Þ
∂t Ti Ti

κik 2 κi⊥ 1
T i θi = ∇k T i þ ð∇⊥ T i Þ2 þ ηi0 W ijk W ijk : ð2:73Þ
Ti Ti 2

The equation for the total entropy sn = sen + sin is of the form

→ →
∂ns → → q q
þ ∇  se n u e þ si n u i þ e þ i = θe þ θi þ θei ,
∂t Te Ti
2
ð2:74Þ
1 1 3me ðT e - T i Þ
θei = QΔ - = nν :
Ti Te mi ei T e T i

If the time scales for heat transport, heat exchange, and heat production are large in
comparison with the density and temperature variation time scales, so dissipative
processes can be neglected, the process can be treated as adiabatic. In this situation,
the equation for electron or ion entropy reads

∂nsα →
þ ∇  nsα u α = 0: ð2:75Þ
∂t

This equation has the same form as the particle continuity equation, and, therefore,
quantities nsα and n change identically so that their ratio (entropy per particle) is
conserved. Combining Eq. (2.75) with the particle balance equation, one obtains

∂sα →
þ u α ∇ sα = 0 : ð2:76Þ
∂t

Therefore, entropy per particle is conserved, and, hence, conserved is the quantity

α =n = const :
T 3=2 ð2:77Þ
68 2 Transport Equations

The thermodynamic “fluxes” and “forces” discussed in Sect. 2.1 and, connected
by Eq. (2.13), are called conjugate if

θ= j m xm :
m
→ → $
According to Eqs. (2.71), (2.73), and (2.74), “fluxes” q α , R , π αjk , QΔ, and “forces”
→
∇Tα/Tα, u , 1/2Wαjk, (Te - Ti)/(TeTi) are the conjugate ones. For transport coeffi-
cients Lmn, which connect “fluxes” and “forces” according to Eq. (2.13), the follow-
ing relation is fulfilled:

→ →
Lmn B = Lnm - B :

This relation represents the general principle of Onsager symmetry of transport

coefficients.

2.6 Viscosity in the BGK Approximation

The viscosity tensor can be derived in a simplified form using the so-called BGK
(Bhatnagar, Gross, and Krook) form of the collision operator

St ðf α Þ = - να f α - f 0α , ð2:78Þ

where f 0α is the Maxwellian distribution function and ν is some collision frequency

representing relaxation to the Maxwellian distribution function. Let us take the
→ → → →
V - uα V - u α moment of the kinetic equation to obtain
j k

∂M 0jk ∂ ∂uαj 0 ∂u
þ V l M 0jk þ M 0ljk þ M þ αk M 0lj
∂t ∂xl ∂xl lk ∂xl ð2:79Þ
eBl 0 0
- ε M þ εkml M mj = - να παjk
mα jml mk

Here, moments are defined as

→ → *
M 0αj,k...n = V j - uαj ðV k - uαk Þ . . . ðV n - uαn Þf α r , V , t dV ,

and εjml is the Levi–Civita completely antisymmetric unit tensor, εjml = 1 for an even
permutation of the numbers 123, εjml = - 1 for odd permutation, and εjml = 0 if any
two indices coincide.
2.6 Viscosity in the BGK Approximation 69

As discussed in Sect. 2.2 in the fluid approximation, there are two “large” terms
where the deviation from the Maxwellian distribution function is important-term
with the magnetic field and the collision operator. In the other “small” terms, the
distribution function can be considered as the Maxwellian one. After substituting the
Maxwellian distribution function, we find

∂pα ∂ ∂uαj ∂uαk

þ ðuαl pα Þ δjk þ p þ þ να παjk
∂t ∂xl ∂xk ∂xj
ð2:80Þ
eBl
= ε M 0 þ εkml M 0mj :
mα jml mk

The l.h.s. of this equation can be simplified using Eq. (2.11) in the zero approxima-
tion, where dissipative terms are neglected:

∂pα ∂ 2 ∂
þ ðuαl pα Þ = - pα u
∂t ∂xl 3 ∂xl αl

The r.h.s. of Eq. (2.80) in the matrix form with the z-axis along the magnetic field is

2παxy παyy - παxx παyz

eBl eB
ε M 0 þ εkml M 0mj = παyy - παxx - 2παxy - παxz
mα jml mk mα
παyz - παxz 0

Finally, Eq. (2.80) can be simplified to

→ $
pα W α þ να π α
2παxy παyy - παxx παyz
eB : ð2:81Þ
= παyy - παxx - 2παxy - παxz
mα
παyz - παxz 0

Resolving this algebraic system leads to the viscosity tensor equation (2.48) with the
simplified form of viscosity coefficients (for a strong magnetic field)

ηα0 = nT α =να ,
nT ν
ηi1 = 0:25 α2 α , ηα2 = 4ηα1 ,
ωcα
nT α
ηi3 = , ηα4 = 2ηα3 :
2ωcα

Viscosity coefficients ηi3 and ηα4 coincide with those obtained by Braginsky,
Eq. (2.49), other coefficients are very similar.
70 2 Transport Equations

2.7 Thermal Force for Impurities

In the plasma with impurities of charge Z and species I, there are thermal forces due
to collisions with the electrons and the main ions. In the absence of a magnetic field,
the thermal force on impurity ions from electrons can be estimated in the
following way.
Let us consider the case of test impurities with

nI Z 2 << 1: ð2:82Þ

Electron thermal fluxes acting on impurity ions have different directions and are
given by an estimate Γ+~Γ-~neVTe, with VTe being the thermal velocity of electrons.
Due to the momentum loss of electrons during collisions with impurity ions, two
forces proportional to the corresponding fluxes arise: R+~R-~meneVTeνeI, where
p
2 nI Z 2 e 4 Λ
νeI = : ð2:83Þ
12π3=2 ε20 1=2
me T 3=2
e

In the presence of an electron temperature gradient, electrons that cross the area
come from regions with different electron temperatures. However, since the
electron-electron collision frequency is much larger than the collision frequency
between electrons and impurity ions in Eq. (2.84), the temperature difference can be
estimated as δTe = (dTe/dz)λe, where λe = VTe/νe is the mean free path due to
electron-electron collisions, νe is given by Eq. (2.22). Hence, the thermal force from
electrons is estimated as

∂ ∂T
RTeI  ðne me νeI V Te Þλee  nI Z 2 e : ð2:84Þ
∂z ∂z

This thermal force applied to impurity ions is directed towards the region of higher
electron temperature.
Another part of the thermal force applied to the impurity ions is caused by
collisions with the main ions. Heavy impurities mI >> mi can be treated analogously
(for the main ions, we assume Z = 1)

∂ ∂T
RTiI  ðni mi νiI V Ti Þλii  nI Z 2 i : ð2:85Þ
∂z ∂z

This part of the thermal force is directed towards the regions of higher temperature of
the main ions.
2.8 First Ionization Potential Effect and Impurity Retention in a Tokamak Edge 71

The resulting thermal force on impurities is

∂T e ∂T
RTI = αnI Z 2 þ βnI Z 2 i , ð2:86Þ
∂z ∂z

where α and β are numerical coefficients calculated from kinetic theory.

2.8 First Ionization Potential Effect and Impurity Retention

in a Tokamak Edge

In recent years, understanding the mechanism of impurity retention/leakage has

become a more pressing issue, as impurity seeding into the edge plasma has been
found to be a reliable way to obtain divertor detachment. This allows for reducing the
target heat loads down to an acceptable level through impurity radiation, and
attaining such a regime with impurity seeding and radiative divertor operation is
considered to be essential for the operation of future thermonuclear devices. To
obtain this highly radiated regime, gas (Ne, Ar, N, and others) is seeded into the edge
plasma, and after ionization, it should stay mainly in the divertor region. On the other
hand, the thermal force, Eqs. (2.85) and (2.86), extracts impurities upwards towards
the separatrix and further into the main plasma, where the radiation of impurities
might lead to unacceptable energy loss in the main plasma.
To understand in detail the process of impurity retention/leakage in the divertor,
one should analyze in detail the parallel momentum balance for ionized impurities
and the behavior of both neutral and ionized components at the edge plasma. Edge
tokamak plasma is described by 2D modeling, where equations for ionized compo-
nents are solved on the basis of transport equations discussed in this chapter and
neutral particles are followed by the Monte-Carlo kinetic code. The typical simula-
tion region and computational mesh are shown in Fig. 2.4. Here, x is the poloidal
coordinate, the full magnetic field has a projection on it, and y is directed perpen-
dicular to the flux surfaces.
The flows of the main deuterium ions in the divertor region are governed by the
recycling process. Neutral particles are reflected from divertor plates and move into
the divertor plasma. After ionization, most of the ions return back to the divertor
plates along magnetic field lines, and, hence, along x, driven by the pressure gradient
(Fig. 2.5). Their poloidal velocity is of the order of a poloidal sound speed
1=2
bx cs = ðBx =BÞðT e þ T i Þ1=2 =mi ; for further details, see Chap. 9. Part of the deute-
rium ions that ionize at larger distance from the plates move in the opposite direction
towards upstream also driven by the pressure gradient. Ions that move towards the
plate recombine at the plate and return back as neutrals.
72 2 Transport Equations

Fig. 2.4 Computational meshes for plasma (violet quadrangular cells) and neutrals (dark-yellow
triangular cells) used in the simulations. Shown are the in-vessel structures and locations of gas
puff/seeding and pumping

To understand impurity behavior, let us consider the stationary impurity force

balance in the direction of the magnetic field

→ →
∇ mI nI u I ukI = - ∇k ðnI T I Þ - ZenI ∇k φ þ RIku þ RTIk : ð2:87Þ

→
Here, a part of the friction force RIu is caused mainly by ion-impurity collisions due
→
to different velocities, RIu = mi ni νiI uki - ukI , and thermal force RTI is given by
Eq. (2.86). Parallel derivatives here can be expressed through poloidal derivatives as
|∇k| = bx∂/∂x. Since collisions between impurities are sufficiently strong (the
corresponding criterion is νiI >> bxcs/L with L being the poloidal spatial scale),
the main contribution to the parallel momentum balance equation (2.86) is produced
by the last two terms, so that
→
RIku þ RTIk ≈ 0: ð2:88Þ

In other words, the friction force for impurity ions is close to zero. This equation
determines the parallel (poloidal) velocity of impurities: (uIk = bxuIx)
2.8 First Ionization Potential Effect and Impurity Retention in a Tokamak Edge 73

Fig. 2.5 Illustration of leakage/retention mechanism. Zones of maximal ionization are shown
schematically by ellipses for deuterium (green ellipses), nitrogen (blue ellipses), and neon (red
ellipses). Dotted lines schematically represent the stagnation points of poloidal flow (the difference
in their location between species exists in reality, but is not shown here). Solid circles show points
of ionization and arrows of corresponding color represent the poloidal flow directions

∂T e ∂T i
uIx = uix þ αnI Z 2 bx þ βnI Z 2 bx =ðbx mi ni νiI Þ: ð2:89Þ
∂x ∂x

One can see that impurity ions are strongly coupled to the main ions. The thermal
force changes their velocity with respect to the main ions and is important in the
regions with a strong temperature gradient, i.e., in the region where ionization takes
place. Since temperatures rise from the plate towards upstream, thermal force
extracts impurities away from the divertor.
Now, one can understand the first ionization potential (FIP) effect. Impurities as
neutrals are reflected at the plates and move away from the plates. Impurity with a
first ionization potential smaller than that of deuterium (N, e.g.) is mainly ionized
below the stagnation point of the main ion poloidal flow (closer to the plates than the
deuterium ions). Therefore, after ionization, they are dragged to the plates by the
main ions. Only a small fraction of them is ionized above the stagnation point and is
dragged away upstream (Fig. 2.5). Impurities with ionization potentials higher than
that of deuterium (Ne, for example) are mainly ionized above the stagnation point
74 2 Transport Equations

Fig. 2.6 Nitrogen and neon net density distribution in a tokamak. Modeling by the SOLPS-ITER
transport code

and are dragged upstream by the main ion parallel flow. As a result, elements with
low first ionization potential are better retained in the divertor region than those with
high ionization potential. Thermal force can only shift the stagnation boundary for
impurity poloidal flow with respect to that for deuterium. Examples of simulations of
the impurity density distribution are shown in Fig. 2.6. N with a lower first ionization
potential is indeed better retained in the divertor region than Ne.
Chapter 3
Quasineutral Plasma and Sheath Structure

3.1 Quasineutrality Maintenance

Plasma is a quasineutral medium where the densities of the charged particles

multiplied by their charge for positive and negative charges coincide with high
accuracy. The charge separation in plasma can be estimated from the Poisson
equation, which for Z = 1 reads

→ e
∇
E = ðn - ne Þ: ð3:1Þ
ε0 i

Indeed, estimating the potential in plasma as T/e, from the Poisson equation, one
obtains

ε 0 j φj ε 0 T
jni - ne j   2,
L2 eL

where L = |∇ ln n|-1 is a typical spatial density scale. Dividing |ni - ne| by ion or
electron density, we obtain a condition to be satisfied in the quasineutral plasma

jni - ne j r 2d
 2 ≪ 1, ð3:2Þ
n L

where the Debye radius is defined as r D = ε0 T=ne2 . Hence, the plasma can be
considered a quasineutral medium provided that the Debye radius is much smaller
than the characteristic scale of the problem. This condition is violated at the plasma
boundaries where scale L decreases since the plasma density is reduced. Everywhere
else in plasma, the paradigm of quasineutrality consists of the following. The
densities of the charge species are considered to be equal, ni = ne, and the self-
consistent electric field, providing that quasineutrality takes place, is to be found

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 75

V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_3
76 3 Quasineutral Plasma and Sheath Structure

from the plasma equations with ni = ne. The space charge can be calculated
afterwards from the Poisson equation if necessary.
In some situations, for example, when a strong current is flowing through the
plasma, the potential in the plasma can significantly exceed the value T/e. In this
case, the value T/e in the quasineutrality condition, Eq. (3.2), should be replaced by a
corresponding value of the plasma potential so that the Debye radius should be
replaced by the corresponding effective Debye radius.
Now let us consider a process of maintaining quasineutrality when the
quasineutrality does not hold from the very beginning, i.e., nonquasineutral pertur-
bations of electrons and ions exist at t = 0. Let us consider the evolution of small
perturbations δne and δni on a uniform quasineutral background ne = ni = n0 in a
slightly ionized plasma without a magnetic field and in the absence of a plasma
current as an example. Let us start with a particle balance for electrons and ions with
the electron flux given by Eq. (1.95) and similar for ions:

∂ne
þ ∇
ð- De ∇ne þ ne be ∇φÞ = I - R, ð3:3Þ
∂t
∂ni
þ ∇
ð- Di ∇ni - ni bi ∇φÞ = I - R: ð3:4Þ
∂t

A constant temperature is assumed, so thermal diffusion is absent. Terms on the

r.h.s. represent sources and sinks of the charged particles. In the absence of a
magnetic field, mobility and diffusion tensors are reduced to scalars, and according
to the Einstein relation Dα/bα = Tα/e. Systems (3.3) and (3.4) should be
supplemented by Poisson equation (3.1).
The linearized system of Eqs. (3.1), (3.3), and (3.4) has the form:

∂δne
= ∇
ðDe ∇δne - n0 be ∇φÞ,
∂t
∂δni ð3:5Þ
= ∇
ðDi ∇δni þ n0 bi ∇φÞ,
∂t
e
Δφ = ðδne - δni Þ:
ε0

The solution is sought as the Fourier integral

→ 1 →→ →
δnα r , t = δnα →k exp i k r d k ,
ð2πÞ3
→ 1 →→ →
φ r ,t = φ →k exp i k r d k :
ð2πÞ3
3.1 Quasineutrality Maintenance 77

After substituting into Eq. (3.5), we have for the Fourier components

∂δn →
ek
= - De k2 δn → þ be n0 k 2 φ → ,
∂t ek k
∂δn → ð3:6Þ
ik
= - Di k δn → - bi n0 k φ → :
2 2
∂t ik k
- k2 φ → = e δn → - δn → =ε0 :
k ek ik

Eliminating the potential from the third equation, one can seek the solution of the
remaining equation system for δne →k and δni →k assuming that the time dependence is
given by exp(-iωt). Then, we obtain the algebraic system

- iωδn → = - De k2 δn → þ be n0 e δn → - δn → =ε0 ,
ek ek ik ek
- iωδn → = - Di k δn → - bi n0 e δn → - δn
2 → =ε0 :
ik ik ik ek

The algebraic system has a nontrivial solution provided its determinant is equal to
zero. Solving the corresponding quadratic equation for ω, one obtains two roots:

ω1 = - iDe =r 2d ; ω2 = - ik 2 Di ð1 þ T e =T i Þ, ð3:7Þ

where the Debye radius is defined as rd = (ε0Te/ne2)1/2. A general solution for the
Fourier components is

δn → = C 1 expð - iω1 t Þ þ C 2 expð - iω2 t Þ,

ek ð3:8Þ
δn → = - ðbi C1 =be Þexpð - iω1 t Þ þ C 2 expð - iω2 t Þ:
ik

During the derivation of Eq. (3.8), the inequalities De ≫ Di and krd ≪ 1 were used.
Coefficients C1, C2 are determined from the initial conditions

C 1 = δn0 → - δn0→ ,
ek ik
C 2 = δn0→ - bi C 1 =be δn0 → ≈ δn0→ ,
ik ek ik

where δn0 → = δne,i →k ðt = 0Þ.

e,i k
Since ω1- 1 ≪ ω2- 1 , evolution is characterized by two different time scales. The
first stage is determined by the time scale ω1- 1 , which corresponds to the diffusion of
electrons at a distance of the order of the Debye radius. At this fast stage, the ion
density perturbation remains practically constant, while the electron density pertur-
bation is strongly redistributed. At the end of this stage, the quasineutrality is
established so that δne = δni, as follows from Eq. (3.8). Therefore, quasineutrality
maintenance occurs at the so-called Maxwellian time scale
78 3 Quasineutral Plasma and Sheath Structure

τM = r 2d =De = ε0 =σe , ð3:9Þ

where σe = enbe is the electron conductivity. The Maxwellian time does not depend
on the spatial scale of the perturbation. The second root ω2 corresponds to the joint
quasineutral evolution of electrons and ions at a longer time scale. This process of
ambipolar diffusion will be considered in the next chapter.
It is worth noting that the process of maintaining quasineutrality takes place due
to the mobility of electrons in a self-consistent electric field and not by their
diffusion. Indeed, neglecting diffusive terms in Eq. (3.6), subtracting the second
equation from the first one and substituting potential from the Poisson equation, one
finds for the Fourier components

∂ δne →k - δni →k
= - De r d- 2 δne →k - δni →k : ð3:10Þ
∂t

This equation describes relaxation to the quasineutral state at the Maxwellian time
scale. The quasineutrality maintenance occurs at the Maxwellian time scale in the
slightly ionized plasma with high collision frequency when τMνeN ≫ 1. In the
opposite case, oscillation at plasma frequency takes place, which decays with the
-1
time scale νeN , which in this situation is the time scale for establishing
quasineutrality.

3.2 Collisionless Sheath at the Material Surfaces

Near the material surfaces, the quasineutrality condition is violated, and the fluid
approximation simultaneously becomes invalid since the typical spatial scale of the
density variation L = |∇ ln n|-1 tends to zero. Here, a sheath of space charge is
formed with a width of the order of Debye radius. Let us analyze a sheath structure in
the absence of a magnetic field for the case when both the electron and ion mean-free
paths are larger than the sheath width. Such a sheath is known as a collisionless
sheath.
Let for definiteness consider a surface that is biased negatively with respect to the
plasma. Near the surface at a distance of the order of the Debye radius, a sheath of
positive space charge is located (Fig. 3.1). Further away from the surface at the
plasma side of the sheath, a so-called presheath is situated. Its width is of the order of
the mean-free path of ions, and the preasheath separates the quasineutral plasma and
the sheath. In the preasheath, ions are freely accelerated towards the surface by the
electric field, so the fluid description is not applicable here; however, this region is
still a quasineutral one, while further in the plasma, the fluid description becomes
applicable.
At the sheath edge, ions and electrons move in completely different ways. Ions
are accelerated in the sheath; therefore, all ions that move towards the surface reach
3.2 Collisionless Sheath at the Material Surfaces 79

Fig. 3.1

Fig. 3.2

it. In contrast, a significant fraction of electrons is reflected back in the sheath, and
only electrons with large energy can reach the surface. To understand how the
electron distribution function is formed, we consider below an electron distribution
function in a capacitor with an electric field that decelerates electrons.

3.2.1 Electrons in a Capacitor with a Reflecting Electric Field

Consider a capacitor with an electric field that decelerates electrons and a potential
difference Δφ (Δφ is a positive quantity). The flux of electrons is directed from right
to left (Fig. 3.2), and their distribution function is assumed to be Maxwellian for
positive velocities Vx. Part of the electrons is reflected back by the decelerating
electric field, while the other part, which has sufficient energy to overcome the
potential difference Δφ, appears at the left from the capacitor. In the stationary case
80 3 Quasineutral Plasma and Sheath Structure

and in the absence of collisions, the distribution function is conserved along the
particle trajectories in accordance with Liouville’s theorem. Therefore, in the sta-
tionary case, the distribution function should be a function of the integral of motion.
Since the full energy E = meV2/2 - eφ is conserved, the distribution function is a
function of the full energy.
To the right of the capacitor, we assume the distribution function to be Maxwel-
lian for electrons moving from right to left with positive velocities (the potential here
is set to zero φ = 0)

m1=2 mV 2x
f þ ð V x ≥ 0Þ  f M = A e
exp - : ð3:11Þ
ð2πT e Þ1=2 2T e

Hence, f = f M(E). To overcome the reflecting potential Δφ, electrons should have a
velocity larger than V 0 = 2eΔφ=me , while electrons with a velocity V0 after
passing through the capacitor are decelerated to zero velocity. Therefore, to the
left of the capacitor for electrons with positive velocities

eΔφ m1=2 mV x 2
f - ðV x ≥ 0Þ = A exp - e
exp - : ð3:12Þ
T e ð2πT e Þ 1=2 2T e

There are no electrons with negative velocities in this region:

f - ðV x < 0Þ = 0: ð3:13Þ

In contrast, to the right of the capacitor, there are reflected electrons with kinetic
energies smaller than eΔφ, while electrons with larger energies flowing to the right
are absent; they are passed through the capacitor. Therefore,

m1=2 mV x 2
A e
exp - , jV x j ≤ V 0
f þ ð V x < 0Þ = ð2πT e Þ 1=2 2T e : ð3:14Þ
0, jV x j > V 0

Finally, inside the capacitor

m1=2 me V x 2 - eφðxÞ
f in ðV x ≥ 0Þ = A e
exp - ,
ð2πT e Þ1=2 2T e
m1=2 me V x 2 - eφðxÞ
f in ðV x < 0Þ = A e
exp - , jV x j ≤ V 0 ,
ð2πT e Þ1=2 2T e ð3:15Þ
f in ðV x < 0Þ = 0, jV x j > V 0 ,
2e
V 0 = ðΔφ þ φÞ:
me
3.2 Collisionless Sheath at the Material Surfaces 81

Note that even to the right of the capacitor, the electron density does not coincide
with А; in particular, it is smaller than А. Indeed, integration of Eqs. (3.12) and (3.14)
over velocities yields

A
nþ = 1 þ erf ðeΔφ=T e Þ , ð3:16Þ
2

where the error function is defined as

z
2
erf ðzÞ = p exp - t 2 dt:
π
0

The electron density inside the capacitor is

A eφ
nin = 1 þ erf eðφ þ ΔφÞ=T e exp : ð3:17Þ
2 Te

If the potential drop Δφ > Te/e is large enough, almost all electrons are reflected
back, and the error function in Eq. (3.16) is close to unity, so that n+ ≈ A. In this
situation, the electron density decreases inside the capacitor starting from the right
plate according to the Boltzmann law

eφ
nin = nþ exp : ð3:18Þ
Te

At the left plate, the density, Eq. (3.17), does not follow the Boltzmann law due to
the absence of a sufficient number of reflecting electrons. In particular, to the left of
the capacitor in accordance with Eq. (3.12)

A eΔφ
n- = exp - :
2 Te

3.2.2 Particle and Energy Fluxes to the Material Surfaces

One can easily calculate particle and energy fluxes to the surface coming from the
sheath edge. Since the electron particle flux is conserved in the sheath, it can be
calculated using the Maxwellian distribution function, Eq. (3.11), for the positive
velocities at the sheath edge (at the right plate of the capacitor). Only electrons with
energies larger than eΔφ contribute to the flux:
82 3 Quasineutral Plasma and Sheath Structure

1
A Te eΔφ
Γe = V x f þ dV x = p exp - : ð3:19Þ
2π me Te
V0

The same expression can be obtained by integrating with the distribution function at
the surface, Eq. (3.15), but then one should integrate over all positive velocities. If
the potential drop in the sheath is significantly larger than Te/e, so that the error
function in Eq. (3.16) is close to unity, then the flux of electrons is (ns  n+)

n Te eΔφ
Γe = p s exp - : ð3:20Þ
2π me Te

The ion flux to the wall can be easily obtained assuming that at the sheath edge,
all ions have the same velocity u0 directed to the surface. This is justified, as shown
below, if Te ≫ Ti and ions gain energy equal to Te/2 due to the acceleration in the
presheath, so that their velocity at the sheath edge should be u0 = (Te/mi)1/2. The ion
flux at the sheath edge coincides with their flux at the surface since all ions are
accelerated in the sheath and reach the surface. Hence,

Γ i = ns u0 : ð3:21Þ

While calculating an energy flux, one has to distinguish the flux to the surface and
the flux leaving the plasma, these fluxes do not coincide. The easiest way to calculate
the electron energy flux to the surface is to use the distribution function, Eq. (3.15),
at the surface. One has to use a 3D distribution function. For Δφ ≫ Te/e

1
me V 2 ns m3=2 me V 2 þ 2eΔφ
qe = Vx e
exp - dV x dV y dV z
2 ð2πT e Þ3=2 2T e ð3:22Þ
0
= 2T e Γe :

While calculating the energy flux of ions to the surface, it is necessary to take into
account the ion energy gain in the sheath eΔφ. With an account of energy gain in the
presheath Te/2, one obtains

qi = ðeΔφ þ T e =2ÞΓi : ð3:23Þ

The energy fluxes given by Eqs. (3.22) and (3.23) cause heating of the surface.
To obtain the energy flux of electrons at the sheath edge leaving the plasma, qes,
one has to integrate the energy flux from V0 to infinity over velocities to obtain

qes = ðeΔφ þ 2T e ÞΓe : ð3:24Þ

3.2 Collisionless Sheath at the Material Surfaces 83

The ion energy flux leaving the plasma, qis, is (assuming the small ion temperature)

1
qis = T Γ: ð3:25Þ
2 e i

The energy fluxes are not conserved in the sheath, the energy is redistributed
between the charged particles due to the work performed by the electric field. It is
also worth noting that with the account of finite ion temperature, their energy flux
increases considerably.

3.2.3 Current-Voltage Characteristics of the Sheath. Floating

Potential

For many plasma physics problems, it is necessary to know the connection between
the potential drop in the sheath and the current to the surface. This relation, known as
the current-voltage characteristic of the sheath, is used as a boundary condition for
the potential distribution inside the plasma volume. The collisionless sheath current-
voltage characteristic can be obtained from Eqs. (3.20) and (3.21). For eΔφ ≫ Te,
the current density to the surface is

1=2 1=2
Te 1 Te eΔφ
j = ens -p exp - : ð3:26Þ
mi 2π m e Te

In the absence of current, the surface is biased negatively with respect to the
plasma, and the potential drop is called the floating potential. The isolated surface in
plasma is, therefore, biased to the floating potential (by absolute value). In accor-
dance with Eq. (3.26), the floating potential (plasma potential with respect to the
surface in the absence of current) is

1=2
Te mi
φfl = ln : ð3:27Þ
e 2πme

The floating potential is of the order of few Te/e (for hydrogen plasma, e.g., of the
order of 3Te/e).

3.2.4 Sheath Structure. Bohm Criterion

Let us now find the potential distribution inside the sheath for the negatively biased
surface. The potential distribution is described by the Poisson equation
84 3 Quasineutral Plasma and Sheath Structure

d2 φ
= eðne - ni Þ=ε0 : ð3:28Þ
dx2

For simplicity, we assume a sharp boundary between the plasma and the sheath
located at x = 0, and, at this boundary, we assume quasineutrality ne = ni = ns and
zero electric field. The electron density in the sheath decreases in accordance with
the Boltzmann distribution, Eq. (3.18). Since the ion flux is conserved in the sheath,
the ion density in the sheath can be obtained from the flux conservation condition
(φ(x = 0) = 0 is assumed):

1=2
2eφ
Γi = ns u0 = ni ðxÞ u20 - : ð3:29Þ
mi

The ion density decreases deeper in the sheath since the ions are accelerated by the
electric field. The electron density, however, should decrease even faster to provide a
positive space charge in the sheath (Fig. 3.1). The Poisson equation (3.28) has the
form

d2 φ eφ u0
= ε0 ens exp - : ð3:30Þ
dx2 Te u0 - 2eφ=mi
2 1=2

In the dimensionless variables

eφ x ε0 T e
Φ= - , ξ= , rd =
Te rd ns e 2

one has

d2 Φ 1
= - expð- ΦÞ þ , ð3:31Þ
dξ2 ð1 þ Φ=Φ0 Þ1=2

where Φ0 = mi u20 =2T e .

Let us study the character of the solution for small values of Φ. Expanding the
r.h.s. of Eq. (3.31), we find that it is equal to Φ 1 - T e =mi u20 . The space charge
value, which is proportional to the r.h.s. of Eq. (3.31), tends to zero at the sheath
boundary, while its sign depends on the u0 value. If u0 < (Te/mi)1/2, which corre-
sponds to a negative space charge, the solution of Poisson equation (3.31) has an
oscillating character, so it is impossible to construct a solution that corresponds to the
negative potential drop in the sheath and, hence, a positive space charge. In contrast,
if u0 ≥ (Te/mi)1/2, the monotonic potential profile exists corresponding to the
potential profile as shown in Fig. 3.1. A condition
3.2 Collisionless Sheath at the Material Surfaces 85

u0 ≥ ðT e =mi Þ1=2 ð3:32Þ

is known as the Bohm criterion.

The separation into plasma and sheath is, of course, conventional. In reality, at a
considerable distance from the surface, the charge separation, and the electric field
strength smoothly increase towards the surface; this region is often referred to as the
presheath. The characteristic presheath scales equal the mean-free path of attracting
particles (for negatively biased surface to ion mean-free path). This means that this
problem demands rigorous kinetic treatment. The value u0 depends on the problem
and type of collision. Rigorous analysis gives the quantity of u0 close to

u0 = ðT e =mi Þ1=2 : ð3:33Þ

We shall demonstrate how this value can be obtained from the simplified fluid
approach.
Let us consider one-dimensional stationary plasma flow towards the surface.
Summing up momentum balance equations of electrons and ions neglecting electron
inertia and electron-neutral collisions, we have in plasma

dui d ð pe þ p i Þ
mi nui =- - nmi νiN ui : ð3:34Þ
dx dx

Combining this equation with the particle balance equation in the absence of sources
and sinks yields

c2s dui
1- = - νiN : ð3:35Þ
u2i dx

Here, cs = (dp/dρ)1/2 is the sound speed in the plasma, p = pe + pi, ρ = nmi. From
Eq. (3.35), one can see that ions are accelerated up to the sound speed at the sheath
edge, where acceleration turns to infinity (the density gradient here also turns to
infinity as follows from momentum balance equation (3.34). In reality, of course,
acceleration remains finite and takes place at the last ion mean free path. In the case
considered Te ≫ Ti, the sound speed value cs coincides with u0 given by Eq. (3.33).
Equation (3.31) is analogous to the Newton equation of motion. After multiplying
both sides of the equation by dΦ/dξ and integration, one obtains

2
1 dΦ
þ W ðΦÞ = 0: ð3:36Þ
2 dξ

Here, the “potential energy” known as the Sagdeev potential is defined as

W ðΦÞ = 1 - expð- ΦÞ - 2Φ0 ðΦ=Φ0 þ 1Þ1=2 þ 2Φ0 : ð3:37Þ

The solution of Eq. (3.36) for Φ0 = 1/2 is

86 3 Quasineutral Plasma and Sheath Structure

Φ
1 dΦ0
ξ= p : ð3:38Þ
2 0 1=2
ð1 þ 2Φ Þ - 2 þ expð- Φ Þ 0
0

For a large potential drop in the sheath, assuming Φ ≫ 1, the potential in the sheath
is determined by the Child-Langmuir law

25=4 3=4
ξ= Φ : ð3:39Þ
3

Inserting the full potential drop and sheath width into Eq. (3.39), in the initial
coordinates, we find the relation between them:

3=4
25=4 eΔφ
Lsh = r : ð3:40Þ
3 d Te

Hence, the sheath width is of the order of the Debye radius and increases with the
size of the sheath potential drop. Analysis for the case of a positively biased surface
can be done analogously.

3.3 Impact of Electron Emission. Double Sheath

The structure of the collisionless sheath changes in the presence of additional

electron flux from the surface, caused by thermal electron emission or secondary
electron emission. We assume that the surface emits electrons with a Maxwellian
distribution with temperature Tec, which is much smaller than the temperature of
electrons in plasma Teh. Let us restrict ourselves by considering a surface at floating
potential.
The emitted electrons partially neutralize the positive space charge in the sheath,
thus reducing the potential drop in the sheath and increasing the flux of hot electrons
from the plasma to the surface. For a small flux of cold electrons Γec, the zero current
condition reads

Γeh - Γec = Γi , ð3:41Þ

where the fluxes Γeh and Γi are defined according to Eqs. (3.20) and (3.21). Then,
Eq. (3.41) determines the potential drop as a function of emitting flux Γec.
A further increase in the cold electron flux leads to the formation of a negative
space charge near the surface, which leads to partial reflection of the cold electron
flux. The so-called virtual cathode arises, and a double sheath is formed with a
potential minimum close to the electrode (Fig. 3.3). The negative space charge and
the negative potential drop restrict the flux of cold electrons so that only part of their
3.3 Impact of Electron Emission. Double Sheath 87

Fig. 3.3

flux Γec = Γ0ec is injected through the double layer into the plasma, while the rest are
reflected back to the surface. The potential difference between the surface and
potential minimum Δφmin is connected with the flux of cold electrons reaching the
plasma by the relation

Γðec0Þ = Γec exp - eΔφðmin Þ =T ec ,

Therefore, the value Δφmin should be of the order of few Tec/e. Since Tec ≪ Teh, it is
possible to neglect Δφmin with respect to the potential drop Δφ between the plasma
and the surface.
Inside the double layer, the potential profile is determined by the Poisson equation

d 2 φ=d x2 = eðneh þ nec - ni Þ=ε0 : ð3:42Þ

Charged particle densities on the r.h.s. of Eq. (3.42) are the functions of the potential.
The ion density is still given by Eq. (3.29), where ns is the ion density at the plasma
side of the double sheath. In contrast, the density of hot electrons should be described
in a more accurate way than the Boltzmann distribution since the potential drop in
the double layer is significantly smaller than the floating potential. We shall take the
distribution function given by Eqs. (3.15) and (3.16). According to Eq. (3.17)

1
neh ðΦÞ = A expð- ΦÞ 1 þ erf Φðmin Þ - Φ : ð3:43Þ
2

where Φ = - eφ/Teh.
The cold electrons are accelerated inside the double sheath, so neglecting their
initial energy, one obtains
88 3 Quasineutral Plasma and Sheath Structure

T eh Φðmin Þ - Φ T eh Φðmin Þ - Φ
nec ðΦÞ = nec Φðmin Þ exp 1 - erf
T ec T ec

1 T ec
≈ p nec Φðmin Þ ,
π T eh Φðmin Þ - Φ
ð3:44Þ

where nec(Φmin) is the cold electron density at the potential minimum. It is connected
with the flux of cold electrons:
p
2 T ec
Γðec0Þ = p nec Φðmin Þ : ð3:45Þ
π me

Now, we can integrate the Poisson equation assuming a zero electric field at both
boundaries of the double sheath – at the plasma side of the sheath and at the potential
minimum:

Φðmin Þ

ðneh þ nec - ni Þ dΦ = 0: ð3:46Þ

0

Combining this equation with Eqs. (3.43) and (3.44) and the expression for the ion
density, Eq. (3.29), with an account of the zero current condition, Eq. (3.41) and the
quasineutrality constraint nehs + necs = ns at the plasma side of the sheath, one
obtains an equation for the plasma drop in the double sheath for a floating surface:

1 þ erf ΔΦ fl ð1 - aÞ
exp - ΔΦ fl = a ;
1þp
πΔΦfl
ð3:47Þ
mi u20 2T eh
a= 1þ ΔΦ fl - 1 ,
T eh mi u20

Here, ΔΦfl = eΔφfl/Teh. We neglected the terms containing a small factor (me/mi)1/2.
For Teh ≫ Ti accounting for the Bohm criterion mi u20 = T eh , we find

eΔφfl =T eh = ΔΦfl = 0:95: ð3:48Þ

We see that the potential drop in the double sheath that is formed for the case of
strong electron emission is significantly reduced with respect to the floating potential
Eq. (3.47).
3.4 Sheath in Magnetic Field 89

In general, a double sheath structure can be formed not only near the emitting
surface but also in many other situations, particularly at the boundary of plasmas
with different parameters, so their role in plasma physics is quite fundamental.

3.4 Sheath in Magnetic Field

The sheath structure for the magnetic field normal or slightly inclined to the surface
does not change significantly with respect to the unmagnetized plasma. However, for
a magnetic field strongly inclined or parallel to the surface, the sheath structure
changes significantly. The reason consists of the following. If the Debye radius is
smaller than the charged particle Larmor radii, a so-called magnetic presheath is
formed in front of the surface, as shown in Fig. 3.4, with a width of the order of
Larmor radius. For negatively biased surface, it is ion Larmor radius.
Let us consider the general situation of the inclined magnetic field forming an
angle Θ with the surface. At distances of the order of the ion Larmor radius, ions
reach the negatively biased surface at a time scale of the order of the inverse
cyclotron frequency ωci- 1 . Their distribution functions strongly differ depending on
whether their orbits intersect or do not intersect the surface. At the orbits intersecting
the surface, the ion density is small since the ions escape to the surface with the
thermal velocity, while these orbits are filled due to slower ion flow from the bulk
plasma along the magnetic field or by rare collisions with other ions. Closer to the
surface, the number of intersecting orbits is larger, and the density at these orbits is
correspondingly smaller. At the surface, practically all orbits intersect, and the
density ns here can be estimated by equalizing the thermal ion flux ns(T/mi)1/2 and
normal component of the ion flux from the bulk plasma n(T/mi)1/2 sin Θ. As a result,
the density at the sheath edge ns~n sin Θ should be significantly smaller than that at

Fig. 3.4
90 3 Quasineutral Plasma and Sheath Structure

the outer side of the magnetic presheath. The density drop in the magnetic presheath
is especially large for small angles Θ when the magnetic field line is almost parallel
to the surface.
Electrons in the magnetic presheath are trapped by the electric field to prevent
their fast escape along the magnetic field lines, so the potential profile here corre-
sponds to the Boltzmann distribution for electrons. The potential drop in the
presheath is, therefore, quite large, it is of the order of Δφps = (Te/e) ln
n/ns = (Te/e) ln (sin-1Θ). The detailed simulation of the density profile inside the
magnetic presheath, however, is rather complicated since the kinetic approach is
required here. On the other hand, one can easily obtain the current-voltage charac-
teristics of the sheath and the magnetic presheath by taking the sum of the potential
drops in the sheath and presheath so that the potential drop is calculated from the
outer presheath edge to the surface. The normal flux of electrons to the surface is a
projection of their parallel flux along the magnetic field:

n sin Θ Te eΔφ
Γe = p exp - : ð3:49Þ
2π me Te

The ion flux is conserved and coincides with the normal component of ion flux at the
outer edge of the magnetic presheath:

Te
Γi = n sin Θ : ð3:50Þ
mi

The ion velocity in Eq. (3.50) corresponds to the Bohm criterion, which can be
derived as without the magnetic field. The current-voltage characteristic, which
generalizes Eq. (3.26), is thus given by

1=2 1=2
Te 1 Te eΔφ
j = en sin Θ -p exp - : ð3:51Þ
mi 2π me Te

For very small angles when the magnetic field is almost parallel to the surface,
however, this expression is not applicable. At small angles, the orbits that intersect
the surface are filled not by the parallel flux from the bulk plasma but by the
collisions of ions from nonintersecting orbits with neutrals or with ions. The orbits
intersecting the surface remain almost empty since the time scale of escape to the
surface is of the order of the inverse cyclotron frequency ωci- 1 , while the time scale of
filling these orbits is νi- 1 . In the stationary case, the ion flux to the surface is equal to
the ion flux in the velocity space from nonintersecting to intersecting orbits. The
latter can be estimated as
3.5 Thermoelectric Current Between Two Electrodes 91

Γi ≈ nρci νi : ð3:52Þ

The ion collision frequency here is either the ion-neutral collision frequency or
ion-ion collision frequency. The exact expression for the ion flux can be found in
[7] and requires kinetic derivation. Therefore, the current-voltage characteristic
equation (3.51) can be used for angles Θ > νi/ωci. In the general case, one should
use the ion flux to the surface

Te
Γi = n sin Θ þ nρci νi : ð3:53Þ
mi

3.5 Thermoelectric Current Between Two Electrodes

The current voltage characteristic of the sheath, equation (3.26), can be used to
calculate the thermoelectric current between two electrodes adjacent to the plasma
with different electron temperatures. Suppose we have two connected grounded
electrodes with the same (zero) potential with different electron temperatures
Tþ -
es , T es at the sheath entrances, as shown in Figs. 3.5–3.6. Let us assume that the
plasma between electrodes has zero potential drop and that the magnetic field is
normal to the surfaces or absent. The potentials at the sheath entrances in the general
case are not equal to the floating potentials given by Eq. (3.27). Indeed, if the
electron temperature at the right electrode marked by + is larger than the temperature
at the left electrode marked by -, then φþ -
fl > φfl and the potential difference is
applied to plasma to produce infinite current for infinite plasma conductivity.
Therefore, for uniform plasma, sheath potential drops in both sheaths must coincide,
so that

Fig. 3.5 Net current through the divertor plates for various seeding rates
92 3 Quasineutral Plasma and Sheath Structure

jth

– +
Tes < Tes
B, z

Fig. 3.6 Scheme of thermoelectric current between two electrodes

φs- = φþ
s = φ: ð3:54Þ

This potential should be lower than the floating potential at the right electrode
φ < φþ -
fl and larger than the floating potential at the left electrode φfl < φ. As a
result, the current should flow through both sheaths from right to left, i.e., from hotter
to colder plasma. In a more general situation of nonuniform plasma with the
potential drop Δφ over it, we have

φs- þ Δφ = φþ
s : ð3:55Þ

Since the current through each sheath must be the same according to the current
continuity, one finds

T es- 1=2
1 T es- 1=2
eφs-
jth = ens- -p exp - , ð3:56Þ
mi 2π me T es-

Tþ Tþ eφþ
1=2 1=2
1
jth = enþ - es
þp es
exp - s
: ð3:57Þ
s mi 2π me Tþes

This current is called the thermoelectric current. To calculate the potential drop over
the plasma, one has to add parallel momentum balance equation (2.7) for electrons.
Neglecting electron inertia, we have

∂pe ∂φ ∂T
0= - þ en þ 0:51me νei jth - 0:71n e : ð3:58Þ
∂z ∂z ∂z

One can integrate Eq. (3.58) over z to obtain

3.5 Thermoelectric Current Between Two Electrodes 93

zþ zþ
1 ∂pe 0:51me νei
Δφ = φþ - φs- = dz þ jth dz þ 0:71 T þ -
es - T es =e: ð3:59Þ
s en ∂z en
z- z-

Equations (3.59), (3.56), (3.57), and (3.55) determine the thermoelectric current
between the electrodes and potentials in front of both electrodes. Note that the
current through each sheath flows because at one electrode, the electron flux is larger
than the ion flux, while at the other electrode, the situation is reversed (Fig. 3.6).
Figure 3.5 shows an example of the thermoelectric current flowing from the outer
divertor target to the inner target in the ASDEX-Upgrade tokamak as a function of
the nitrogen seeding rate. At a small seeding rate, the outer divertor plasma is
significantly hotter than the inner divertor plasma, and the thermoelectric current is
large. With an increase in the seeding rate, the outer divertor plasma becomes colder
so that T þ -
es becomes closer to T es , and the thermoelectric current decreases by
absolute value.
Chapter 4
Diffusion in Partially Ionized Unmagnetized
Plasma

4.1 Ambipolar Diffusion

Consider pure plasma with single species ions and constant and homogeneous
particle temperatures. Neutral gas is also supposed to be homogeneous. We shall
choose a reference frame where neutral gas is at rest: uN = 0. Considering the
quasineutrality condition from Eqs. (3.3) and (3.4), one obtains initial equations for
plasma density and potential:

∂n
þ ∇ . ð - De ∇n þ nbe ∇φÞ = I - R,
∂t ð4:1Þ
∂n
þ ∇ . ð - Di ∇n - nbi ∇φÞ = I - R:
∂t

The potential could be excluded from this equation system. Multiplying the first
equation by bi and the second equation by be, after summation, we obtain an
equation for ambipolar diffusion:

∂n
- Da Δn = I - R: ð4:2Þ
∂t

Here, the coefficient Da is known as the ambipolar diffusion coefficient:

De bi þ Di be
Da = : ð4:3Þ
be þ bi

With an account of the Einstein relation

Da = Di ð1 þ T e =T i Þ: ð4:4Þ

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 95

V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_4
96 4 Diffusion in Partially Ionized Unmagnetized Plasma

The ambipolar diffusion equation gives a solution for plasma density indepen-
dently of a potential distribution; however, the density profile boundary conditions at
the material surfaces that restrict plasma should be imposed. This can be done in the
following way. As discussed in the previous section, the ion flux at the edge of the
sheath is Γ = ns T e =mi . On the other hand, this flux from plasma is controlled by
the ambipolar diffusion and, according to Eq. (4.2), is of the order of Γ~nDa/L,
where L is the typical plasma scale and n is the bulk plasma density. By equalizing
two fluxes, we obtain an estimate for plasma density at the sheath edge (here, λiN is
the ion mean free path)

ns =n - λiN =L ≪ 1: ð4:5Þ

Therefore, while solving the equation for ambipolar diffusion, it is possible to

impose a zero boundary condition at the surface

nj s = 0 : ð4:6Þ

Potential distribution can be found, if necessary, by subtracting one of the

equations in (4.1) from the other. After subtraction, we obtain an equation where
the divergence of the ion flux is equal to the divergence of the electron flux:

→ →
∇ . Γ e = ∇ . Γ i: ð4:7Þ

Another form of this equation is the divergence-free condition for the plasma
current:

→
∇ . j = 0: ð4:8Þ

After the substitution of fluxes into Eq. (4.7), we have the equation for the
potential

∇ . ½nðbe þ bi Þ∇φ] = ∇ . ½ðDe - Di Þ∇n]: ð4:9Þ

The solution for the potential could be sought as a sum of two contributions

φ = φ d þ φc : ð4:10Þ

The diffusive part of the potential φd is determined by the density gradient and
→ →
corresponds to equal electron and ion fluxes Γ e = Γ i . From the equal fluxes
condition

- De ∇n þ be n∇φd = - Di ∇n - bi n∇φd , ð4:11Þ

one obtains
4.1 Ambipolar Diffusion 97

Fig. 4.1 Plasma polarization in the process of ambipolar diffusion: (а) j = 0, φ = φd, (b) additional
polarization φc, produced by the current

De - Di T
φd = ln n þ const ≈ e ln n þ const: ð4:12Þ
be þ bi e

The second part of the potential φc is proportional to the current that flows
through the system and satisfies the equation

∇ . ½nðbe þ bi Þ∇φc ] = 0: ð4:13Þ

These phenomena will be considered in Chap. 6, while here, we shall assume the
absence of current flowing through the plasma so that φc = 0. In this case, the
potential equation (4.12) corresponds to the Boltzmann distribution for electrons.
Polarization of the positive density perturbation is shown in Fig. 4.1a. The
physical mechanism responsible for this polarization consists of the following.
Electrons with a large diffusion coefficient start to diffuse with a large velocity.
Due to the quasineutrality constraint, an electric field arises that decelerates electrons
and reduces their flux to be equal to the ion flux. Since the De ≫ Di electric field
strongly reduces the flux of electrons, the resulting potential corresponds to the
Boltzmann potential for electrons. The ion flux becomes (1 + Te/Ti) larger than the
unipolar diffusive flux (flux in the absence of an electric field) and reaches the value
→
Γ = - Da ∇n. As a result, density evolution occurs with equal fluxes of the charged
particles.
If there is a current flowing through the plasma, then additional polarization
arises, as shown in Fig. 4.1b. Indeed, in the 1D case integrating Eq. (4.13), one
obtains

→
- ∇φc = j =neðbe þ bi Þ, ð4:14Þ
→
where j is a constant current density. Hence, a part of the electric field connected
with the current decreases inside the positive density perturbation. However, in the
case of pure plasma with a single ion species, the current flowing through the plasma
has no impact on the profile evolution. The latter is determined by the ambipolar
diffusion equation (4.2). In plasma with several ion species, evolution is more
complicated and is to a large extent determined by convection (see Chap. 6).
98 4 Diffusion in Partially Ionized Unmagnetized Plasma

4.2 Examples of Solutions of the Ambipolar Diffusion

Equation

4.2.1 Decay of Initial Perturbation in Infinite Plasma

Consider the decay of arbitrary initial perturbation. We shall seek a solution to the
equation

∂n
- Da Δn = 0 ð4:15Þ
∂t

with the boundary conditions

n ð r → 1 Þ = n0 , φðr → 1Þ = 0:

The initial perturbation is taken in the Gaussian form:

→ N
δn r , t = 0 = exp - r 2 =a2 , ð4:16Þ
π3=2 a3

where N is the total number of ionized particles injected into the plasma. The
solution is sought as the Fourier integral

→ 1 →→ →
δn r , t = δn →k exp i k r d k : ð4:17Þ
ð2πÞ3

After substituting Eq. (4.17) into Eq. (4.15), one obtains the equation

∂δn →k
þ Da k2 δn →k = 0: ð4:18Þ
∂t

Its solution is

δn →k = δn →k ð0Þ exp - Da k 2 t :

The density perturbation in real space according to Eq. (4.17) is given by the
integral

→ 1 →→ →
δn r , t = δn →k ð0Þ exp - Da k 2 t exp i k r d k : ð4:19Þ
ð2πÞ3

The spectral density at t = 0 can be obtained using inverse Fourier transformation

4.2 Examples of Solutions of the Ambipolar Diffusion Equation 99

→ →→ →
δn →k = δn r , t exp - i k r d r ð4:20Þ

of the initial Gaussian density perturbation equation (4.16):

δn →k ð0Þ = N exp - a2 k 2 =4 : ð4:21Þ

After substitution of Eq. (4.21) into Eq. (4.19) and performing integration, we
find

→ N r2
δn r , t = exp - : ð4:22Þ
π3=2 ða2 þ 4Da t Þ3=2 a2 þ 4Da t

This solution describes the temporal diffusive evolution of the initial Gaussian
density perturbation.p At time scales exceeding a2/4Da, the spatial scale of the density
profile increases as 4Da t , while the density decreases as t-3/2, keeping the total
number of particles constant.
Plasma potential with respect to infinity is easily calculated using Boltzmann
distribution for electrons, Eq. (4.12):

Te δn þ n0
φ= ln : ð4:23Þ
e n0

4.2.2 Positive Column of Glow Discharge

Let us obtain a stationary density profile in an infinite cylinder with radius a and an
ionized particle source due to the ionization of neutral gas by electrons. Plasma
escape is controlled by radial ambipolar diffusion to the walls. Recombination is
neglected with respect to diffusion. The balance of the charged particles is then given
by the equation

1 d dn
- rDa = nZ: ð4:24Þ
r dr dr

Here, Z is the number of ionizations per second produced by one electron.

Equation (4.24) should be solved with the zero boundary condition at the cylinder
wall n(r = a) = 0 and the condition of finite n at r = 0. This is a classical Sturm-
Liouville problem for eigenvalues and eigenfunctions. In other words, Eq. (4.24)
with the boundary conditions could be satisfied only for the special values of the
ionization source; otherwise, a stationary solution for the density would not exist.
100 4 Diffusion in Partially Ionized Unmagnetized Plasma

Since Eq. (4.24) is the Bessel equation, its solution, which is finite at r = 0, is the
Bessel function

nðr Þ = nð0ÞJ 0 r Z=Da : ð4:25Þ

Zero boundary condition n(r = a) = 0 determines the plasma source intensity Z

ð2:4Þ2 Da
Z= , ð4:26Þ
a2

Here, the quantity 2.4 corresponds to the first root of the Bessel function. Other
roots imply negative density and, therefore, do not represent physically justified
solutions. The radial density profile is, therefore, given by

nðr Þ = nð0ÞJ 0 ð2:4r=aÞ: ð4:27Þ

Plasma potential corresponds to the Boltzmann distribution for electrons

Te
φ = Δφ - ln½nð0Þ=n]: ð4:28Þ
e

The potential difference Δφ between the center of the plasma and the cylinder
wall is the sum of two contributions: the potential difference between the center and
the sheath edge and the potential jump in the sheath. The potential difference
between the center and sheath edge is

Te
Δφpl = ln½nð0Þ=ns ]:
e

The potential jump in the sheath can be found by equalizing the electron and ion
fluxes to the wall in Eq. (3.26)

Te ns T 1=2
Δφsh = ln e
:
e ð2πme Þ1=2 Γi ðaÞ

From Eq. (4.27), flux to the wall Γi(a) = - Dadn/dr(r = a) can be expressed
through the central density:

2:4Da nð0Þ
Γi ðaÞ = J 1 ðr=aÞ: ð4:29Þ
a

Finally,
4.2 Examples of Solutions of the Ambipolar Diffusion Equation 101

Te T 1=2
e a
Δφ = Δφpl þ Δφsh = ln , ð4:30Þ
e 1=2
αme Da

where the numerical coefficient α = 2.4J1(2.4)(2π)1/2 = 3.1.

4.2.3 Diffusive Decay

When the plasma source is switched off and plasma recombination is still neglected,
plasma decay is governed by ambipolar diffusion to the walls. We shall consider a
restricted cylinder of length L and radius a. Plasma decay is described by Eq. (4.15)
with the zero boundary conditions at the walls. Its solution is obtained by the method
of separation of variables and has the form

1
nðr, z, t Þ = Ajk exp - t=τjk J 0 ðξk r=aÞ sinðjπz=LÞ: ð4:31Þ
j, k = 1

Here, the decay time is defined as

1 Da ξ2 D j2 π2
= 2k þ a 2 , ð4:32Þ
τjk a L

where ξk are the Bessel function’s roots and coefficients Ajk are determined by the
initial conditions. Since the sequence of the Bessel function’s roots rises fast, at large
time scales, all terms in the sum in Eq. (4.31) except the first one disappears, and the
main diffusive mode is established in plasma:

nðr, z, t Þ = nð0Þ expð- t=τ11 ÞJ 0 ð2:4r=aÞ sinðπz=LÞ: ð4:33Þ

4.2.4 Diffusive Probe

The probe is a small electrode inserted into the plasma for diagnostic purposes. It is
biased with respect to the wall or some large electrode and thus is biased with respect
to the surrounding plasma. Varying the potential of the probe, the current-voltage
characteristic is measured, and it contains information on the plasma parameters.
Two types of probes are known. If the mean free path of electrons and ions is larger
than the probe size, the probe is called a Langmuir probe. In the opposite case, when
the mean free path of the charged particles and the sheath width are smaller than the
probe size, electron and ion fluxes to the probe are governed by collisions, i.e., are
102 4 Diffusion in Partially Ionized Unmagnetized Plasma

determined by diffusion and mobility. We shall consider this collisional case, and
such a probe is known as a diffusive probe.
The density profile in the vicinity of the diffusion probe is described by the
ambipolar diffusion equation. In the absence of ionization and recombination, it is
reduced to the Laplace equation

Δn = 0: ð4:34Þ

The boundary conditions at the probe surface and at infinity are

→
nj s = 0; n r → 1 = n0 :

For a spherical probe of radius а, the solution of Eq. (4.34), which satisfies the
boundary conditions, is

n = n0 ð1 - a=r Þ: ð4:35Þ

We see that there is a depletion region near the probe at a scale of the order of the
probe size, and the density profile is independent of the probe potential or current to
the probe. Let us consider the situation when a large negative potential is applied to
the probe. For a large negative potential of the probe, the electrons are reflected back,
and the electron current to the probe should be zero. Hence, electrons are also
trapped in plasma, and the plasma electric field corresponds to the zero flux of
electrons: φ = (Te/e) ln n/n0. This field increases the ion radial flux up to
→
Γ i = - Da ∇n. Using the profile Eq. (4.35), one obtains the ion flux at the probe
surface Γi(a) = Dan0/a. Multiplying this flux by the probe surface and electron
charge, we obtain the ion saturation current

i = 4πen0 Da a:
I sat ð4:36Þ

In the more general case, the probe can be approximated by an ellipsoid of

rotation with semiaxes (a, a, b). Then, the Laplace equation can be solved in elliptic
coordinates

x2 y2 z2
þ þ = 1:
a2 þ ξ a2 þ ξ b2 þ ξ

Here, ξ is the elliptic coordinate, and ξ = 0 corresponds to the probe surface. The
solution of the Laplace equation in elliptic coordinates is (the equivalent electrostatic
problem is solved in [11])
4.2 Examples of Solutions of the Ambipolar Diffusion Equation 103

Arth b 2 - a2 = b 2 þ ξ
1- ; b > a;
Arth 1 - a2 =b2
n=n0 = ð4:37Þ
arctg b2 - a 2 = b 2 þ ξ
1- ; b < a:
arctg 1 - a2 =b2

The expression for the saturation current could be obtained using electrostatic
analogy. The potential profile created by a solitary conductive body in an electro-
static field is also described by the Laplace equation with a constant potential at the
surface. The ion saturation current is proportional to the normal component of the
density gradient integrated over the probe surface. A similar integral of the potential
gradient in the electrostatic problem over the surface of the conductor is proportional
to the conductor charge, i.e., to its capacity С. Similarly, the ion saturation current to
the probe is proportional to its capacity

i = 4πen0 Da C:
I sat ð4:38Þ

The expression for the capacity of the conductor having the form of an ellipsoid
of rotation with semiaxes a, a, b can be found, for example, In [11]:
p
b2 - a2
; b>a
Archðb=aÞ
p
C= a 2 - b2 : ð4:39Þ
; a>b
arccosðb=aÞ
a; a=b

For the strong positively biased probes, ions are trapped in the plasma and do not
flow to the probe, while the electron saturation current is calculated analogously

e = 4πen0 De ð1 þ T i =T e ÞC:
I sat ð4:40Þ

The electron saturation current is significantly larger than the ion saturation
current since De ≫ Di. The typical current-voltage (I–V) characteristic of the probe
is presented in Fig. 4.2.
In the intermediate part of the I–V characteristic, there are two typical potentials.
One is the floating potential φf when the current to the probe is zero. The other is the
plasma potential (which we assume to be zero) when there is no electric field both in
the plasma and in the sheath. Since I sat e ≫ I i , a considerable part of the I–V
sat

characteristic corresponds to the negative probe potentials. The ion current to the
probe for a large negative φ is equal to the ion saturation current, and the ion flux to
the probe is Γi(a) = Dan0/a. The electron current to the probe is I þ I sat
i . For I ≪ I e ,
sat
104 4 Diffusion in Partially Ionized Unmagnetized Plasma

Fig. 4.2 Typical current-voltage characteristic of a diffusive probe

the electric field in the plasma and the main part of the sheath corresponds to the
Boltzmann distribution for electrons. Employing the electron flux to the surface for
trapped electrons Eq. (3.20) and adding the potential drop in plasma Te/e ln (n0/ns),
we obtain for the spherical probe

ðsat Þ
eφ I þ Ii 3a2
= ln - ln : ð4:41Þ
Te I ðesatÞ r 2d ð1 þ T i =T e Þ2

where rd is the Debye radius with electron temperature. This expression determines
the I–V characteristic for large negative probe potentials |φ| ≫ Te/e.
The floating potential φfl is negative and can be calculated by substituting I = 0 in
Eq. (4.41):

eφfl Da 3a2
= ln - ln 2 : ð4:42Þ
Te De ð1 þ T i =T e Þ r d ð1 þ T i =T e Þ2

For the probe at plasma potential φ = 0, there is no electric field in the plasma,
and the sheath vanishes. The density profile still obeys Eq. (4.35), while the current
to the spherical probe is given by the difference between two unipolar diffusive
fluxes:

I 0 = 4πeðDe - Di Þn0 a: ð4:43Þ

For positive probe potentials, the I–V characteristics can be obtained in a way
similar to Eq. (4.41).
4.3 Diffusion of Slightly Ionized Multispecies Plasma 105

4.3 Diffusion of Slightly Ionized Multispecies Plasma

The ambipolar diffusion equation (4.2) for the pure plasma, which was investigated
in the previous section, has several striking peculiarities. First, it is formulated solely
in terms of the plasma density. The self-consistent electric field that is always present
in the inhomogeneous plasma remains “hidden” in the ambipolar diffusion coeffi-
cient Eq. (4.3). The self-consistent problem of calculating the plasma density and
potential profiles is reduced to two separate problems. The plasma density profile can
be found from Eq. (4.2) with the zero boundary condition independent of the
potential profile. The latter can be found, if necessary, from Eq. (4.9). In other
words, the evolution of the density profile does not depend on the presence of dc
current through the plasma inhomogeneity. The rate of the diffusion process is
determined by less mobile particles, i.e., by ions. The electrons are hindered by
the electric field, which drags the ions. This drag force enhances ion diffusion and is
responsible for the factor (1 + Te/Ti) in the expression for the ambipolar diffusion
coefficient, Eq. (4.3). The second remarkable property of the ambipolar diffusion
process consists of the fact that despite the nonlinearity of the initial problem, the
equation for density is linear (in the absence of sources).
These properties, and the striking simplicity of the ambipolar diffusion equation,
stimulated numerous attempts to apply such an approach to more complicated
situations – to multispecies plasmas, to plasmas in magnetic field, etc. These
attempts have not attained serious success over several decades. The reason is that
such an approach is principally wrong. It follows from the fact that the ambipolar
diffusion problem corresponds to a degenerate case. This means that such a reduc-
tion of the nonlinear problem to a system of two linear equations for plasma density
and potential is possible only for the case of pure plasma with scalar constant
diffusion and mobility coefficients. In this situation, several essentially nonlinear
effects strictly compensate for each other. However, in the general case, such
compensation is impossible, and the problem cannot be reduced to the ambipolar
diffusion equation and remains essentially nonlinear and current dependent. Even in
the case of diffusion in currentless plasma, the electric field is determined by the
density gradients of all charged particles. Hence, the problem becomes essentially
nonlinear since every partial field-driven flux is proportional to the product of the
density of a given species and the field strength, which is determined by a linear
combination of the density gradients of all species.
To understand these nonlinear mechanisms, we consider the simplest 1D exam-
ple, when N test ions are injected at t = 0 into a small region of infinite and uniform
ambient weakly ionized plasma with density n0 and Te = Ti. We neglect the
ionization–recombination processes and assume that the mobility and diffusion
coefficients of the injected and ambient ions coincide. The only distinction with
the ambipolar problem equation (4.2) consists of the fact that we shall now distin-
guish between the injected and the ambient particles. The evolution process for the
injected n1 and for the ambient n2 ion densities is described by the equations
106 4 Diffusion in Partially Ionized Unmagnetized Plasma

∂n1
þ ∇ . ð - D1 ∇n1 - n1 b1 ∇φÞ = 0,
∂t : ð4:44Þ
∂n2
þ ∇ . ð - D2 ∇n2 - n2 b2 ∇φÞ = 0:
∂t

The quasineutrality condition is satisfied, so n = n1 + n2. Since D1 = D2 (b1 = b2)

for the electron profile n(x, T ), Eq. (4.2) is valid, with the ambipolar diffusion
coefficient Da = 2D1. Its solution for the case of the point initial condition
δn(x, t = 0) = n1(x, t - 0) = Nδ(x) is

N x2
δnðx, t Þ = p exp - : ð4:45Þ
4πDa t 4Da t

The potential profile φ = (Te/e) ln n/n0 corresponds to the Boltzmann distribution

for electrons. The total Gaussian profile δn with an effective width of the order of
(4Dat)1/2 is, nevertheless, formed in a rather complex way by the diffusive and field-
driven fluxes of the injected and ambient particles. In the linear case, n1 ≪ n0, the
electric field perturbation is small, and its influence on the motion of the injected ions
is negligible. Their profile is thus determined by the unipolar ion diffusion:

N x2
n1 ðx, t Þ = p exp - : ð4:46Þ
4πD1 t 4D1 t

Perturbation of the ambient ion density is caused by the electric field and
diffusion. It can be found as the difference between Eqs. (4.45) and (4.46). The
ambient ions are extracted by the electric field from the neighboring zone
x ≤ (4D1t)1/2, where the injected ions dominate and are piled up outside of it
(Fig. 4.3a).
If n1 > n0, the problem becomes nonlinear. The net profile equation (4.45)
consists mainly of the injected ions. Their diffusion is practically ambipolar: the
field here simply doubles their diffusive flux. The ambient ions are almost absent in
the neighboring zone, as they are dragged out by the electric field. In Fig. 4.3b, an
example of the numerical solution for such a situation is displayed.
As the second example of plasma separation into regions with different ion
compositions, we shall consider the problem of diffusive decay of a cylindrical
positive column of weakly ionized plasma that consists of electrons and two ion
species – positive ions with density n+(x, t) and negative ions with density n-(x, t).
The quasineutrality condition demands n = n+ - n-. Expressing the radial electric
field from the zero-current condition at the tube wall Γe þ Γn - = Γnþ , we obtain the
radial flux of the negative ions
4.3 Diffusion of Slightly Ionized Multispecies Plasma 107

Fig. 4.3 (a) Small perturbation. Density profiles of injected ions, Eq. (4.36), (1) and perturbation of
ambient ions (2) for equal mobilities and point initial perturbation. The curve 3 is the perturbation of
electron density described by ambipolar diffusion, Eq. (4.45). The spatial coordinate ξ is measured
in units of (4D1t)1/2. (b) Numerical simulation for strong disturbance: 1 – injected ions, 2 – ambient
ions, and 3 – electron density. The densities are in units of n0
108 4 Diffusion in Partially Ionized Unmagnetized Plasma

n - ∇n De bn - - bnþ Dnþ - n∇n - Dn - be þ Dn - bnþ - 2n - ∇n - Dn - bnþ

Γn - =
bn þ nþ þ bn - n - þ be n
De bn - n - ∇n - Dn - be n∇n - - 2n - Dn - bnþ ∇n -
≈ :
bn þ nþ þ bn - n - þ be n
ð4:47Þ

Imposing the zero boundary condition on all three partial densities and neglecting
all the plasmachemical processes, we find that the problem of plasma decay has a
solution that corresponds to the fundamental diffusive mode, Eq. (4.33).
For the case of cylindrical geometry, it means that all the densities are propor-
tional to the Bessel function J0(ζ1r/a) and to time-dependent factors N-(t), N+(t),
N(t):

2
dN - N D dN 2N - N þ Dn - bnþ ζ1
= - n- = : ð4:48Þ
dt NDe dt N - bn - þ N þ bnþ þ Nbe a

One can see that the decay time for electrons is De =Dn - smaller than that for
negative ions. The radial field is

De N þ Dn - N - - Dnþ N þ d
Er = - ðln J 0 ðζ1 r=aÞÞ: ð4:49Þ
Nbe þ N þ bnþ þ N - bn - dr

At Te = Ti, the electric field, Eq. (4.49), corresponds to the Boltzmann law both
for electrons and for negative ions until the Bessel partial profiles satisfy the
condition

N ≫ N þ bnþ =be : ð4:50Þ

Since the relaxation of the partial temperatures usually occurs considerably faster
than that of the particle densities, the plasma is isothermal during the main part of the
decay, Te = Ti = TN. In this case, the flux Γn - , according to Eq. (4.47), is small, and
the decay process at this stage is determined by electron-ion ambipolar diffusion
when the radial positive ion flux is practically equal to the electron flux. During this
stage, plasma is continuously enriched by the negative ions, and when the relation
(4.50) is violated, the field, Eq. (4.49), cannot hinder electrons. The rate of electron
loss increases drastically up to a value that corresponds to the electron-free unipolar
diffusion, and the whole tube volume is filled with the ion-ion plasma. The second
stage of the decay process is determined by ion-ion ambipolar diffusion.
4.4 Diffusion in the Ionosphere 109

Fig. 4.4 Temporal dependence of the square of the cloud half-width: × – longitudinal scale of the
ionized barium cloud; o – longitudinal scale of the neutral barium cloud; Δ – transverse scale of the
ion barium cloud. The faster longitudinal expansion of the ionized cloud is connected with the
factor (1 + Te/Ti) in the diffusion coefficient Da. The decrease of Λk at the later stage is apparently
connected with the influence of ambient brightness

4.4 Diffusion in the Ionosphere

In the ionosphere at heights of 200–300 km, we have an example of strongly

magnetized partially ionized plasma. The evolution of plasma inhomogeneities is
controlled by complicated 3D processes of diffusion and mobility. Understanding
complicated plasma evolution is very helpful for analyzing so-called active exper-
iments in space. In some of these experiments with barium clouds, neutral barium
vapor is injected into the ionosphere from the spacecraft. The evolution of a neutral
barium cloud has a diffusive character in the framework moving with the velocity of
the ionospheric wind of the neutral component. Simultaneously, photoionization of
the cloud takes place, and a second ionized Ba+ cloud is formed. The evolution of an
ionized cloud is different from that of the neutral cloud since ionospheric plasma is
magnetized. Experiments with barium clouds are performed at dawn or sunset so that
sunlight on the one hand can produce photoionization and, on the other hand, makes
observation of the clouds from the Earth possible, which can be done even with
pure eye.
Since the density in the clouds is very large, the evolution of Ba+ clouds along the
Earth’s magnetic field is controlled by 1D ambipolar diffusion (see Sect. 4.1). The
global evolution of the plasma cloud is discussed in Sect. 6.3. Figure 4.4 plots the
parallel dimensions of two clouds (halfwidth Λk) versus time for the first Soviet
Union experiment “Spolokh.” One can see that the square dimensions of the clouds
increase linearly with time, which is typical for diffusive processes. One would
expect, in accordance with the results of Sect. 4.2.1,
110 4 Diffusion in Partially Ionized Unmagnetized Plasma

Λ2k = 4Da t ð4:51Þ

for plasma clouds, and

Λ2k = 4Di t ð4:52Þ

for neutral clouds. Ion and neutral diffusion coefficients are identical. Indeed,
in Fig. 4.4, the slope of the curve for ions is larger than that for neutrals since
Da/Di = (1 + Te/Ti). Using Eqs. (4.51) and (4.52), one can obtain information
on ionospheric parameters at the height of the experiment, neutral density, and
temperature ratio.
Chapter 5
Diffusion of Partially Ionized Magnetized
Plasma

5.1 Diffusion and Mobility in a Magnetic Field

The diffusion of pure plasma with constant temperatures of electrons and ions is
governed by particle balance equations similar to Eqs. (3.3) and (3.4):

∂n
- ∇
De ∇n - be n∇φ = I - R;
∂t ð5:1Þ
∂n
- ∇
Di ∇n þ bi n∇φ = I - R:
∂t

Mobility and diffusion tensors are taken in the approximation of elementary theory
(1.103) and (1.104). The dependence of the diagonal (perpendicular) components on
the magnetic field is shown in Fig. 5.1. The following notations are used:

ωce ωci
xe = ; xi = :
veN viN

Since xe ≫ xi, when the magnetic field is rising from B = 0 at first, the value xe = 1 is
reached. For larger magnetic fields xe ≫ 1, the electron diffusion coefficient
De⊥ = Dek = 1 þ x2e starts decreasing with the magnetic field as B-2. For such a
magnetic field, electrons are considered magnetized. On the other hand, since xi ≪ xe,
at xe = 1, the perpendicular ion diffusion coefficient remains practically equal to the
parallel ion diffusion coefficient, and ions are not magnetized. When the magnetic
field increases further at xexi = 1 (for equal temperatures of electrons and ions), the
perpendicular diffusion coefficients of electrons and ions become equal. However,
we still have xi ≪ 1, so ions remain unmagnetized. For larger magnetic fields, when

xe xi ≫ 1 ð5:2Þ

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 111
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_5
112 5 Diffusion of Partially Ionized Magnetized Plasma

Fig. 5.1 Dependence of the perpendicular diffusion coefficients on magnetic field. (1) Electron and
(2) ion diffusion coefficient

ion perpendicular diffusion coefficient exceeds that of electrons (the same is true for
the mobility coefficients due to the Einstein relation). Plasma with inequality (5.2)
fulfilled is called magnetized plasma. Finally, for larger magnetic fields with xi ≫ 1,
the ions become magnetized, and their perpendicular diffusion coefficient decreases
with the magnetic field as B-2.
Let us discuss the physical meaning of perpendicular transport diffusion and
mobility coefficients for a strong magnetic field when both electrons and ions are
magnetized. In the diffusion tensor (α = i, e)

Dα⊥ ± DαΛ 0
Dα = ∓ DαΛ Dα⊥ 0 ð5:3Þ
0 0 Dαk

the diagonal perpendicular diffusion coefficients in the strong magnetic field xi ≫ 1

are

T α vαN
Dα⊥ =  ρ2cα vαN : ð5:4Þ
mα ω2cα

Hence, the perpendicular diffusion coefficients of the charged particles can be

estimated as the square of the random walk step during collisions with neutrals
(Larmor radius) multiplied by the collision frequency.
Nondiagonal components of the tensor, Eq. (5.3), for magnetized ions do not
depend on the collision frequency:

Tα
DαΛ = ð5:5Þ
eB
5.1 Diffusion and Mobility in a Magnetic Field 113

Fig. 5.2 Hall (diamagnetic) fluxes of electrons and ions, and diamagnetic current

They are connected with the flux known as the Hall or diamagnetic flux, which is
directed perpendicular to the density gradient (Fig. 5.2). The flux value can be
estimated as the thermal velocity of the charged particles multiplied by the density
difference at the Larmor radius scale. For example, if the density gradient is parallel
to the x-axis, the flux in the y direction is given by

dn dn
Γαy = ∓ DαΛ  V Tα ρcα : ð5:6Þ
dx dx

In the homogeneous magnetic field, these fluxes do not cause density redistribution
since they are directed perpendicular to the density gradient along equidensities. It is
easy to demonstrate that

0
∇
Dα ∇n = ∇
Dα ∇n , ð5:7Þ

0
where diagonal tensor Dα is defined as

Dα⊥ 0 0
0
Dα = 0 Dα⊥ 0 :
0 0 Dαk

Diffusive Hall fluxes in a uniform magnetic field, therefore, do not contribute to the
equation system, Eq. (5.1). The difference in electron and ion Hall fluxes leads to the
formation of a diamagnetic current. For magnetized electrons and ions, one obtains
114 5 Diffusion of Partially Ionized Magnetized Plasma

→ → → ðT e þ T i Þ →
j Λ = e Γ iΛ - Γ eΛ = B × ∇n ð5:8Þ
B2

The diamagnetic current, Eq. (5.8), reduces the vacuum magnetic field. Using the
→ →
Maxwellian equation ∇ × B = μ0 j , we find that the magnetic field perturbation is
of the order of

δB 2μ nðT e þ T i Þ
 β= 0 : ð5:9Þ
B B2

Below, we shall assume that β ≪ 1 and thus shall neglect magnetic field perturba-
tions produced by plasma currents.
Nondiagonal components of the mobility tensor in the strong magnetic field
xi ≫ 1 coincide for electrons and ions:

1
bαΛ = : ð5:10Þ
B
→ → →
The corresponding Hall flux caused by the electric field Γ αΛ = n E × B =B2 is just
→ →
E × B drift of electrons and ions with the same velocity. Note that this is correct
only for magnetized ions; in contrast, for xi ≪ 1, the ion Hall flux is significantly
smaller than that for electrons.
Diagonal perpendicular components of the mobility tensor, similar to the diffu-
sion tensor, in a strong magnetic field are proportional to the collision frequency

evαN
bα⊥ = : ð5:11Þ
mα ω2cα

The corresponding fluxes in the direction of the electric field can be interpreted as
drift produced by friction with neutrals (Fig. 5.3). Indeed, an electric field parallel to
the y-axis causes joint drift of electrons and ions along the x-axis with equal
velocities uex = uix = E/B. Due to collisions with neutrals, a friction force in the x-
direction arises, acting to decelerate charged particles: RαN = - mαnvαNuαx. In turn,
these friction forces produce the drift of the particles along the electric field with
velocity uαy = ± RαN/(neB). As a result, we have a flux along the electric field
Γαy = ± nbα⊥E with mobility, Eq. (5.11). Since the ion-neutral friction force is
much larger than the electron-neutral force, the ion flux along the electric field is
much larger than that of electrons provided that the plasma is magnetized, Eq. (5.2).
For the mobilities, therefore, we have bi⊥ ≫ be⊥.
Everywhere below, with one exception, we shall consider the case of magnetized
plasma.
5.2 One-Dimensional Diffusion in Magnetized Plasma 115

Fig. 5.3 Flux along the electric field as drift caused by friction with neutrals

5.2 One-Dimensional Diffusion in Magnetized Plasma

5.2.1 Diffusion Across a Magnetic Field

One-dimensional (1D) diffusion across a strong magnetic field is very similar to

diffusion in the absence of a magnetic field; only the role of charged particles is
inversed since in magnetized plasma, ion mobility and diffusion coefficients are
much larger than electron mobility and diffusion coefficients. Similar to the case
without a magnetic field, multiplying the first equation in Eq. (5.1) by bi⊥, and the
second one by be⊥ after summation, one obtains the equation of ambipolar diffusion:

∂n
- Da⊥ Δn = I - R: ð5:12Þ
∂t

Here, the coefficient of ambipolar diffusion is

De⊥ bi⊥ þ Di⊥ be⊥ T

Da⊥ = ≈ De⊥ 1 þ i : ð5:13Þ
be⊥ þ bi⊥ Te

In the absence of current through the plasma, the potential corresponds to the
Boltzmann distribution for ions:

Di⊥ - De⊥ T
φd = - ln n þ const ≈ - i ln n þ const: ð5:14Þ
bi⊥ þ be⊥ e

Hence, in contrast to the situation without a magnetic field, positive plasma

perturbation is biased negatively, so that the electric field confines ions and accel-
erates electrons to keep their fluxes equal.
116 5 Diffusion of Partially Ionized Magnetized Plasma

5.2.2 1D Diffusion at an Arbitrary Angle with a Magnetic

Field

Here, we shall study the situation when the plasma density depends only on the
coordinate ζ when the density gradient forms an angle β with a magnetic field, as
shown in Fig. 5.4. It is convenient to introduce a new coordinate system (x, η, ζ)
instead of the initial coordinate system (x, y, z), where the z-axis is parallel to the
magnetic field. A new coordinate system is obtained from the initial coordinate
system by rotation over the x-axis; therefore, the diffusion and mobility tensors in the
new system have the standard form

0 0 0 0
D = ADA ; b = AbA , ð5:15Þ

where

1 0 0
^=
A 0 cos β - sin β
0 sin β cos β
0
is the operator of rotation over the x-axis, and A is the inversed operator. In the new
coordinate system, the diffusive tensor has the form

jη
z, B
η

β
0
y

→ →
Fig. 5.4 1D diffusion ∇n k ζ for μe ≪ μ ≪ μi, particle fluxes are shown by arrows
5.2 One-Dimensional Diffusion in Magnetized Plasma 117

D⊥ Dxy cos β - Dxy sin β

0
D = - Dxy cos β D⊥ cos 2 β þ Dk sin 2 β Dk - D⊥ cos β sin β ð5:16Þ
Dxy sin β Dk - D⊥ cos β sin β Dk cos β þ D⊥ sin β
2 2

Here, the quantitates Dk, D⊥, and Dxy correspond to the components of the tensor
Eq. (5.3). The analogous form has the mobility tensor. In the new coordinates initial
equation system, Eq. (5.1) has the form

∂n ∂ ∂n ∂φ
- Dαζζ ± bαζζ n = I - R, ð5:17Þ
∂t ∂ζ ∂ζ ∂ζ

where α = i, e. Similar to previous cases, after eliminating the electric field, one
obtains

∂n ∂ ∂n
- D μ2 = I - R, ð5:18Þ
∂t ∂ζ ∂ζ

where μ = cos β,

Deζζ biζζ þ Diζζ beζζ

Dðμ2 Þ =
biζζ þ beζζ
= Dek μ2 þ De⊥ ð1 - μ2 Þ bik μ2 þ bi⊥ ð1 - μ2 Þ ð5:19Þ
þ Dik μ2 þ Di⊥ ð1 - μ2 Þ bek μ2 þ be⊥ ð1 - μ2 Þ
-1
× bek þ bik μ2 þ ðbe⊥ þ bi⊥ Þð1 - μ2 Þ :

The potential profile can be found by equalizing the electron and ion fluxes in the ζ
direction:

Dek - Dik μ2 þ ðDe⊥ - Di⊥ Þð1 - μ2 Þ

φ= ln n þ const: ð5:20Þ
bek þ bik μ2 þ ðbe⊥ þ bi⊥ Þð1 - μ2 Þ

Equation (5.18) has the form of the ambipolar diffusion equation. However, this
equation cannot be obtained by equalizing the full fluxes of the charged particles
since Γek ≠ Γik, Γe⊥ ≠ Γi⊥ coincides only the ζ projections of the fluxes. In other
words, even in the absence of current in the ζ direction, there is still current flowing
along the η-axis.
Let us discuss the physical mechanism that controls the formation of particle
fluxes. If the density gradient is almost perpendicular to the magnetic field, then it
satisfies inequality μ ≪ μ0, with

1=2
μ0 = bi⊥ =bek , ð5:21Þ

Equation (5.19) can be simplified. In this case, we have

118 5 Diffusion of Partially Ionized Magnetized Plasma

D μ2 = De μ2 = ð1 þ T i =T e Þ Dek μ2 þ De⊥ 1 - μ2 : ð5:22Þ

The diffusion coefficient, Eq. (5.22), is (1 + Ti/Te) times larger than the unipolar
coefficient Deζζ. In particular, for μ = 0, when the density gradient is perpendicular
to the magnetic field, the coefficient, Eq. (5.22), coincides with the coefficient of
ambipolar diffusion, Eq. (5.13). For μ ≪ μ0, the potential given by Eq. (5.20) is
negative since electrons are less mobile than ions. The electric field accelerates
electrons, while the potential corresponds to the Boltzmann distribution of ions.
As seen from Eq. (5.22), there is a critical angle

1=2
μe = be⊥ =bek : ð5:23Þ

For μ < μe, electrons diffuse mainly across the magnetic field, and the diffusion
coefficient, Eq. (5.22), practically coincides with the coefficient of ambipolar diffu-
sion across the magnetic field, Eq. (5.13). On the other hand, at angles μe ≪ μ ≪ μ0,
→
electrons move mainly along the magnetic field B , and the diffusion coefficient
D(μ2) exceeds the perpendicular coefficient of ambipolar diffusion. Ions in both
cases diffuse mainly across the magnetic field since their unipolar coefficient
Diζζ ≈ Di⊥. As a result, the fluxes of electrons and ions along and across the
magnetic field are essentially different.
For large angles μ > μ0 between the density gradient and direction normal to the
magnetic field diffusion coefficient, Eq. (5.19) is reduced to

D μ2 = Di μ2 = ð1 þ T e =T i Þ Dik μ2 þ Di⊥ 1 - μ2 : ð5:24Þ

We see that for these angles, diffusion is controlled by ions. The potential for
Te~Ti corresponds to the Boltzmann distribution for electrons. Now, electrons move
→
mainly along the magnetic field B , while the direction of the flux depends on the
value of μ with respect to a parameter

1=2
μi = bi⊥ =bik : ð5:25Þ

For μ < μi, the ion flux is mainly caused by the perpendicular motion of ions, while
in the opposite case, ions move mainly along the magnetic field.
At a special angle μ0 = (Di⊥/Dek)1/2 = μ0(Ti/Te)1/2, the unipolar diffusion
coefficients of electrons and ions become equal Deζζ = Diζζ = Di⊥, diffusive fluxes
in the direction of the density gradient are also equal to each other, and, therefore, an
electric field is absent. At this angle, although the density gradient is almost
→
perpendicular to B , coefficient D(μ2) is equal to Di⊥. It is much larger than the
ambipolar perpendicular diffusion coefficient given by Eq. (5.13). The dependence
D(μ2) is shown in Fig. 5.5.
Since in the two limiting cases μ = 0, 1, the initial equation system is reduced to
the single equation of ambipolar diffusion, one would suppose that diffusion at an
5.2 One-Dimensional Diffusion in Magnetized Plasma 119

Fig. 5.5 Function D(μ2) (solid line) for Te = Ti, xe = 103, xi = 10. Dotted line – different
approximations: 1 – Eq. (5.22), 2 – Eq. (5.24), 3 – Eq. (5.27)

arbitrary angle with a magnetic field could be described by a simple equation of

anisotropic diffusion

∂n
- ∇k
Dak ∇k n - ∇⊥
ðDa⊥ ∇⊥ nÞ = I - R, ð5:26Þ
∂t

with the coefficients defined by Eqs. (4.3) and (5.13). However, Eq. (5.26) can be
obtained by simultaneously assuming Γek = Γik and Γe⊥ = Γi⊥, which requires a
nonpotential electric field. Hence, Eq. (5.26) is incorrect. For the 1D problem from
Eq. (5.26), it follows that

Da μ2 = Dak μ2 þ Da⊥ 1 - μ2 : ð5:27Þ

This expression strongly differs from the correct Eq. (5.19), especially for μ~μ0,
Fig. 5.5.
The current along the equidensities can be calculated using

∂n ∂φ
jη = en - Diηζ - Deηζ - n biηζ þ beηζ : ð5:28Þ
∂ζ ∂ζ

Substituting potential equation (5.20), we find

enDek ð1 þ T i =T e Þbi⊥ sin β cos β ∂n

jη = : ð5:29Þ
bek cos 2 β þ bi⊥ sin 2 β ∂ζ

The current is zero at β = 0, π/2.

120 5 Diffusion of Partially Ionized Magnetized Plasma

5.3 Diffusion of Perturbation in Unbounded Plasma

Consider diffusion of small perturbation δn on the constant background n0 in a

magnetized unbounded plasma. Let us assume for simplicity Te = Ti. The basic
equations in the plasma with constant temperatures of the charged particles are given
by Eq. (5.1) in the absence of sources and sinks

∂n
- ∇
De ∇n - be n∇φ = 0,
∂t ð5:30Þ
∂n
- ∇
Di ∇n þ bi n∇φ = 0:
∂t

The boundary conditions for density and potential are:

→ →
n r → 1 = n0 , φ r → 1 = 0:

Let us choose the initial perturbation in the form of the delta-function

→ →
δn r , 0 = Nδ r :

The linearized Eq. (5.30) has the form:

∂δn
- ∇
De ∇δn - be n0 ∇φ = 0,
∂t ð5:31Þ
∂δn
- ∇
Di ∇δn þ bi n0 ∇φ = 0:
∂t

The solution is sought as the Fourier integral

→ 1 →→ →
δn r ; t = 3
δn → exp i k r d k ,
ð2πÞ k
1 →→ → :
→
φ r ;t = φ → exp i k r d k :
ð2πÞ3 k

After substitution into Eq. (5.31), one obtains the equation system for the Fourier
harmonics:
5.3 Diffusion of Perturbation in Unbounded Plasma 121

∂n →k → → → →
= - k De k δn →k þ n0 k be k φ →k ,
∂t : ð5:32Þ
∂n →k → → → →
= - k Di k δn →k - n0 k bi k φ →k :
∂t

Its solution is

→ → → →
k D k - k D k δn →k
φ →k = → e → → i→ ,
k b k þ k b k n0
e i : ð5:33Þ
→
δn →k ðt Þ = δn →k ð0Þ exp - D k t ,

where

→ →→ →
→ 2 k De k k Di k
D k = → → → → = k 2 D μ2 : ð5:34Þ
k De k þ k Di k
→
Here, μ = cos β is the cosine of the angle between the vector k and the magnetic
→→
field: μ = k B =ðkBÞ. The diffusion coefficient D(μ2) here coincides with that given
by Eq. (5.19):

2 Dek μ2 þ De⊥ ð1 - μ2 Þ Dik μ2 þ Di⊥ ð1 - μ2 Þ

D μ2 = : ð5:35Þ
Dek þ Dik μ2 þ ðDi⊥ þ De⊥ Þð1 - μ2 Þ

For the density perturbation, we have

→ N →→ →
δn r , t = exp i k r - D μ2 k2 t d k : ð5:36Þ
ð2πÞ3
→
To calculate this integral, let us divide k -space into three regions. In the first
→
region, the vector k is almost perpendicular to the magnetic field: μ ≪ μ0. In this
region, as has already been discussed in the previous section, the diffusion coeffi-
cient Eq. (5.35) has a simple form of Eq. (5.22):

D μ2 = De μ2 = 2 Dek μ2 þ De⊥ 1 - μ2 = 2 Dek k2z þ De⊥ k 2⊥ =k 2 :

In the integral Eq. (5.36) let us switch to Cartesian coordinates. Neglecting small
contributions, one can extend integration to infinity. Let us then introduce new
coordinates:
122 5 Diffusion of Partially Ionized Magnetized Plasma

1=2 1=2
z = z Dek =De⊥ , kz = kz Dek =De⊥ :

In the new variables, the integral equation (5.36) is reduced to integral equation
(4.19), which can be easily calculated. Hence, the contribution from region μ ≪ μ0 in
→
the k -space is (in the initial coordinates)

N exp - z2 =8Dek t - ðx2 þ y2 Þ=8De⊥ t

ne = p 1=2
: ð5:37Þ
16 2π3=2 t 3=2 Dek De⊥

→
The second contribution in the k -space gives region 1 > μ ≫ μ0. In this region,
the diffusion coefficient, Eq. (5.35), can be approximated as

D μ2 = Di μ2 = 2 Dik μ2 þ Di⊥ 1 - μ2 = 2 Dik k2z þ Di⊥ k2⊥ =k2 :

→
In the integral over k , the integration region can be extended to zero (neglecting
small corrections). Introducing, analogously, new coordinates

1=2 1=2
z = z Dik =Di⊥ , kz = kz Dik =Di⊥ ,

and calculating the integral, we find that the contribution from region 1 > μ ≫ μ0 is

N exp - z2 =8Dik t - ðx2 þ y2 Þ=8Di⊥ t

ni = p 1=2
: ð5:38Þ
16 2π3=2 t 3=2 Dik Di⊥

Finally, the contribution from region μ~μ0 can be estimated, and it is of the order of
n0h, where

N
h= p 1=2
: ð5:39Þ
16 2π3=2 t 3=2 Dek Di⊥

In the center of inhomogeneity, this contribution is a small correction.

Therefore, the density perturbation may be expressed as a sum

→
δn r , t = ne þ ni þ Oðn0 hÞ: ð5:40Þ

The potential is calculated analogously and is given by an integral

5.3 Diffusion of Perturbation in Unbounded Plasma 123

→ → →
Fig. 5.6 Equidensities δn r , t =δnð0, t Þ. Parameters xe = 30, xi = 0.3; z B ; distances in units
of (8Di||t)1/2. Depletion regions are dashed

→ N Dek μ2 - Di⊥ ð1 - μ2 Þ 1 →→ →
φ r ,t = exp i k r - D μ2 k2 t d k : ð5:41Þ
ð2πÞ3 bek μ2 þ bi⊥ ð1 - μ2 Þ n0

In region μ ≪ μ0, the multiplier in front of the exponent equals -T/e, and in region
1 > μ ≫ μ0, it equals T/e. Integration in Eq. (5.41) yields

→ T ne T n i T
φ r ,t = - þ þ O n0 h : ð5:42Þ
e n0 e n0 e

According to Eq. (5.40), the density perturbation is approximately the sum of two
Gaussian profiles – “electron” ne and “ion” ni, and lines of constant density represent
the superposition of two ellipsoids (Fig. 5.6). The potential is shown in Fig. 5.7.
The physical mechanisms responsible for the unusual behavior of the density
profile are as follows. Electrons, whose mobility along the magnetic field is the
→
largest, start diffusing along B . Correspondingly, the potential along the magnetic
field becomes negative. Across the magnetic field, the situation is inversed-ions
→
mobility across B in the magnetized plasma is larger than that of electrons, and,
therefore, the potential here is positive. As a result, the potential of a quadrupole
configuration arises, as shown in Fig. 5.7, with the saddle point at the origin. Since
φ(r → 1) = 0, a potential minimum arises at z = ± z0 along the magnetic field and a
→ →
potential maximum across B at ρ = ρ0 (ρ = ðx2 þ y2 ). The φð r → 1Þ = 0
→
electric field is not able to confine electrons along B everywhere, as well as ions
across the magnetic field. As a result, electrons and ions diffuse practically
124 5 Diffusion of Partially Ionized Magnetized Plasma

Fig. 5.7 Equipotentials. Potential in units of Tδn(0,t)/en0, other notations as in Fig. 5.6

→
independently along and across B , correspondingly forming two separate ellipsoids.
Particles with other charges cannot diffuse to these ellipsoids from the origin due to
their low mobility. Hence, electrons and ions from the ambient plasma should come
to “ion” and “electron” ellipsoids correspondingly to compensate space charge to
maintain quasineutrality. Their fluxes are driven by an electric field. Therefore, in the
ambient plasma depletion regions with reduced ambient plasma density should be
formed (dashed regions in Fig. 5.6). The dimensions of the depletion regions are
→ →
(8Dekt)1/2 along B and (8Di⊥t)1/2 across B . Ions are extracted from this region
→
across B by the electric field towards the electron “ellipsoid,” and electrons – along
→
B to the ion “ellipsoid.” These fluxes form eddy “short circuiting” currents in the
ambient plasma, which are responsible for fast, practically unipolar diffusion of the
perturbation.
Angle μ0 with respect to the magnetic field, at which depletion regions are
formed, can be estimated by equating the characteristic time scales for electrons to
→ →
move along B driven by the electric field and for ions across B . From the condition
z2/(bekφ)~ρ2/(bi⊥φ), one obtains μ0 = ρ/z = (Di⊥/Dek)1/2.
The potential has a quadrupole character, and, hence, at large distances, it should
decrease as r-3. Since far from ellipsoids particle motion is mainly determined by
fluxes driven by the electric field, divergence of the fluxes there decreases with
distance as r-5, while density perturbations decrease as δn~t/r5.
5.3 Diffusion of Perturbation in Unbounded Plasma 125

Let us calculate the density perturbation and potential far from the ellipsoids. In
→ → → →
the condition of equal divergences of the fluxes ∇
Γ e = ∇
Γ i , which follows
from Eq. (5.30), one can neglect small terms of the order of De⊥/Di⊥ and Dik/Dek.
Let us introduce a new variable ζ and a new function Ψ according to

1=2
ζ = z bi⊥ =bek , Ψ = φ - ðT=eÞ lnðn=n0 Þ: ð5:43Þ

Then, the condition of equal divergences of the fluxes is reduced to the form

2T
∇
ðn∇ΨÞ = - Δ n: ð5:44Þ
e ⊥

This Eq. (5.44) is the Poisson equation. The net “charge” is

ε0 T
Q= Δ ndζdxdy = 0: ð5:45Þ
en0 ⊥

The “dipole moment” is also zero due to the symmetry of the problem. “Quadrupole
moment” is given by the integral

ε0 T 8Tμ0 N
M= 2ζ2 - x2 - y2 Δ ndζdxdy = - , ð5:46Þ
en0 ⊥ en0

and the quadrupole potential is

2ζ2 - x2 - y2
Ψ=M 5
: ð5:47Þ
16πε0 ζ2 þ x2 þ y2

At large distances φ = Ψ, in the initial coordinates outside ellipsoids, we have

NT μ0 2μ0 cos α - sin α

2 2 2
→
φ r = , ð5:48Þ
2πen0 r 3 μ2 cos 2 α þ sin 2 α 5=2
0

where α is the angle with the z-axis.

Density perturbation could be obtained, for example, from the ion equation,
where only perpendicular mobility should be taken into account,

∂n
= ∇⊥
ðbi⊥ n∇⊥ ΨÞ = bi⊥ n0 Δ⊥ Ψ: ð5:49Þ
∂t

Substituting potential Eq. (5.48), one finds

126 5 Diffusion of Partially Ionized Magnetized Plasma

→ NTDi⊥ t
δn r , t =
πμ40 r 5
12 cos 2 α - 36 cos 2 α sin 2 α=μ20 þ ð9=2Þ sin 4 α=μ40 ð5:50Þ
× 9=2
:
cos 2 α þ sin 2 α=μ20

This expression corresponds to the density perturbation outside ellipsoids. One can
see that at angles α~μ0, the density perturbation is negative, so depletion regions are
formed in the ambient plasma.
When the density perturbation becomes stronger, the depletion regions become
deeper, and, at some critical density, the short-circuiting mechanism by the ambient
plasma becomes ineffective. In the limiting case of a very large initial density
perturbation, when the depth of the depletion region becomes of the order of n0,
the main part of electrons and ions of the initial perturbation diffuse together, and the
evolution of the main part is governed by the equation of anisotropic ambipolar
diffusion Eq. (5.26). Only a small fraction of the ionized particles diffuse due to the
unipolar mechanism, forming “electron” and “ion” ellipsoids. For further details of
nonlinear evolution, see [7].

5.4 Diffusion in Plasma Restricted by Dielectric Walls

Let us analyze plasma decay in a cylinder with dielectric walls of length L and radius
a with the z-axis parallel to the magnetic field. The initial perturbation is supposed to
have azimuthal symmetry, with arbitrary dependence on r and z. The boundary
conditions at the dielectric surfaces are:

nðz = 0, LÞ = 0, nðr = aÞ = 0,
ð5:51Þ
Γek ðz = 0, LÞ = Γik ðz = 0, LÞ, Γe⊥ ðr = aÞ = Γi⊥ ðr = aÞ:

It is instructive to introduce four characteristic time scales for the diffusion of

electrons and ions along and across the magnetic field

L2 a2
ταk = , τα⊥ = : ð5:52Þ
π2 Dαk ð2:4Þ2 Dα⊥

In the magnetized plasma, τe⊥ ≫ τi⊥ and τek ≪ τik.

Let us separate diffusion processes into fast and slow. We shall consider pro-
cesses to be fast if they occur at time scales τf = max (τek, τi⊥). At such time scales,
→ →
it is possible to neglect ion diffusion along B and electron diffusion across B , so the
initial equations are reduced to
5.4 Diffusion in Plasma Restricted by Dielectric Walls 127

∂n ∂ ∂n ∂φ
= Dek - nbek ,
∂t ∂z ∂z ∂z
ð5:53Þ
∂n 1∂ ∂n ∂φ
= r Di⊥ þ bi⊥ n :
∂t r ∂r ∂r ∂r

At small time scales of the order of τf, the particles with small diffusion coefficients
→ →
in the corresponding direction, i.e., ions along B and electrons across B , to not have
enough time to escape from plasma to the walls. Since the current to dielectric walls
should be zero, and Γik(z = 0, L ) = 0, Γe⊥(r = a) = 0 at τ~τf, we have for τ~τf:

Γek ðz = 0, LÞ = Γik ðz = 0, LÞ = 0, Γe⊥ ðr = aÞ = Γi⊥ ðr = aÞ = 0: ð5:54Þ

Let us seek the solution of Eq. (5.53) for the case when (the opposite case could
be considered analogously)

τek ≪ τi⊥ : ð5:55Þ

When this inequality is satisfied, each term on the r.h.s. of the electron Eq. (5.53) is
larger than the terms in the ion equation. Since the r.h.s. of these two equations are
equal, the terms on the r.h.s. of the electron equation must practically compensate for
each other. The corresponding potential is given by the Boltzmann distribution for
electrons:

Te
φ= ln n þ Ψðr Þ, ð5:56Þ
e

where Ψ(r) is an arbitrary function of the radial coordinate. Let us now integrate
Eq. (5.53) over z from 0 to L along the cylinder length. Since the electron flux at the
cylinder ends is zero, Eq. (5.54), we find

∂N 1 ∂N 1 1 ∂ ∂N 1 ∂Ψ
= 0, = r ð1 þ T e =T i ÞDi⊥ þ bi⊥ N 1 , ð5:57Þ
∂t ∂t r ∂r ∂r ∂r

L
where N 1 = ndz. For time scales τ~τf, the particles do not leave the plasma volume
0
to the cylinder ends, and, therefore, according to Eq. (5.57), the quantity N1 = const(t)
is conserved. Hence, it can be calculated using the initial density profile

N 1 ðr Þ = nðr, z, t = 0Þdz:
0
128 5 Diffusion of Partially Ionized Magnetized Plasma

The r.h.s. of ion equation of Eq. (5.57) is zero, the ion flux to the sidewalls is also
zero. Hence,

Te þ Ti
Ψ= - ln N 1 þ const: ð5:58Þ
e

Substituting potential Eqs. (5.56) and (5.58) into the ion equation of Eq. (5.53), we
have

∂n 1 ∂ ∂n ðT þ T i Þ ∂ ln N 1
= r ð1 þ T e =T i ÞDi⊥ - bi⊥ e n : ð5:59Þ
∂t r ∂r ∂r e ∂r

This is the equation for the density evolution at a time scale of the order of τ~τi⊥.
During this evolution, the density profile changes over time due to particle redistri-
bution in the plasma volume without escaping to the walls. At the end of this fast
stage, the profile, which corresponds to Γi⊥ = 0, is established. According to
Eq. (5.59) at the end of this fast stage

n = N 1 ðr Þf ðzÞ:

The arbitrary function f(z) can be found by integrating this expression over the radius
and introducing a new integral N2(z), which is also conserved, according to

a a

N 2 ðzÞ = 2πnrdr = 2πnðr, z, t = 0Þrdr,

0 0
L a L a

N= 2πnrdzdr = 2πnðr, z, t = 0Þrdzdr:

0 0 0 0

As a result, at the end of the fast stage, the following density profile is established:

N 1 ðr ÞN 2 ðzÞ
n= : ð5:60Þ
N

The potential, which corresponds to Eq. (5.60) can be obtained from Eqs. (5.56),
(5.58):

Te T
φ= ln N 2 - i ln N 1 þ const: ð5:61Þ
e e

Therefore, at time scales τ~τf, the profiles that are given by Eqs. (5.60) and (5.61)
are formed due to the redistribution of ions and electrons inside the plasma volume,
while integrals N1, N2, and N remain constant. An example of such redistribution is
5.4 Diffusion in Plasma Restricted by Dielectric Walls 129

Fig. 5.8 Scheme of short-circuiting of the electron and ion fluxes in a dielectric cylinder at time
scales τ~τf

shown in Fig. 5.8. The short-circuiting currents that flow during plasma redistribu-
tion are similar to those analyzed in the previous section.
The case τek ≫ τi⊥ could be analyzed in a similar way, and after plasma
redistribution at time scale τ~τf, the profiles Eqs. (5.60) and (5.61) are established.
For larger time scales t > τf, particle escape to the walls should be taken into
account, and the full equation system (5.50) should be solved. However, as the initial
condition, the density distribution Eq. (5.60) could be taken. This is an important
simplification since this profile is a product of two functions depending on r and
z correspondingly. Hence, a solution for density could also be sought in the form of
the product of two functions: n = n1(r)n2(z), while the potential could be sought as a
sum φ = φ1(r) + φ2(z). The latter at t = 0 coincides with Eq. (5.61). Using the
separation of variables method, one obtains

n= Aj Bk J 0 ðζk r=aÞ sinðπjz=LÞ exp - t=τjk , ð5:62Þ

j, k = 1

where

1
= ð1 þ T e =T i ÞDik ðπj=LÞ2 þ ð1 þ T i =T e ÞDe⊥ ðζk =aÞ2 :
τjk
130 5 Diffusion of Partially Ionized Magnetized Plasma

The density profile Eq. (5.62) coincides with the solution of the equation of
anisotropic ambipolar diffusion

∂n
= ∇k
Dak ∇k n þ ∇⊥
ðDa⊥ ∇⊥ nÞ:
∂t

At large times τ > τf, ionized particles escape from the volume with equal fluxes, and
the decay time is determined by the smallest of two large time scales
τs = min (τik, τi⊥).
Finally, we see that plasma decay in the dielectric tube occurs in two stages.
During the first one, at a fast time scale of the order of τf, the profile Eq. (5.60) is
formed due to electron diffusion along the magnetic field and ion diffusion across the
field without escape to the dielectric walls. During this stage, short-circuiting
currents flow in the plasma. At the second slow stage with the time scale τs joint
ambipolar diffusion of electrons and ions takes place, and the density profile is
described by Eq. (5.61).

5.5 Diffusion in a Cylinder with Conducting Walls

In a cylinder with conducting walls, short-circuiting currents can flow inside the
walls of the conducting vessel. In this case, ionized particles can escape to the walls
at a time scale τf = max (τek, τi⊥), i.e., much faster than in the tube with dielectric
walls. This phenomenon is known as the Simon effect.
Let us analyze plasma decay in such a vessel, and for simplicity, we shall assume
up-down and azimuthal symmetry and consider a “short” cylinder with τek ≪ τi⊥. In
contrast to the previous case, boundary conditions should be imposed on the
potential – φ = 0 at the surfaces and not on the fluxes. Due to condition τek ≪ τi⊥,
the potential of the end wall is strongly negative with respect to the plasma, and if the
sheath adjacent to this wall can be treated as collisionless, the potential difference
between the wall and a point in the plasma (close enough to the end that the problem
can be assumed 1D along the magnetic field), according to Eq. (3.20), is
p
T 2πΓek ðr, z = 0, L, t Þ
φðr, z, t Þ = - e ln : ð5:63Þ
e n T e =me

Here, we used the fact that the potential at the sheath edge and the potential in plasma
at the same magnetic field line are connected by the Boltzmann relation for electrons:
φ/φs = (Te/e) ln n/ns. In a “short” device (τe|| < τi⊥), we can again, as in the case of a
dielectric device, seek the potential in the form of Eq. (5.56). Hence, Eq. (5.63) holds
nearly throughout the whole plasma volume, except the immediate vicinity of the
sidewall. We also assume, as in the previous section, that both slow times τi|| and τe⊥
exceed the two fast times τe|| and τi⊥, so that the small coefficients De⊥ and Di|| can
be neglected in the transport equations. In this approximation, the electron flux to the
end walls is related to the change in the number of particles in column N1(r,t) by
5.5 Diffusion in a Cylinder with Conducting Walls 131

∂N 1
= - 2Γej j ðr, z = 0, t Þ: ð5:64Þ
∂t

Substituting Eq. (5.64) into Eq. (5.63), we find

π=2j∂N 1 =∂t j
eφðr, z, t Þ = - T e ln : ð5:65Þ
n T e =me

Substitution of the potential into the second Eq. (5.53) leads to

2
∂n T 1 ∂ ∂n T 1 ∂ n∂ N 1 =∂r∂t
= 1þ e rDi⊥ - e rDi⊥ : ð5:66Þ
∂t T i r ∂r ∂r T i r ∂r ∂N 1 =∂t

There is a small region δ adjacent to the sidewall, where electron motion cannot be
considered 1D, and Eq. (5.66) is not correct. Near the sidewall, the electrons are
trapped both along and across the magnetic field. The potential with respect to the
end wall is given by Eq. (5.63), while the potential with respect to the sidewall is
described by an analogous equation provided that the electron gyroradius exceeds
the sheath width (see Sect. 3.4): near the sidewall, the perpendicular electron flux can
be expressed through the potential as

Γe⊥  nρce νeN expð- eφ=T e Þ ð5:67Þ

Combining Eq. (5.67) with Eq. (5.63), we obtain

ρce νeN Γe⊥ ðr = a, z, t Þ


T e =me Γek ðr = a - δ, z = 0, t Þ

The ratio Γe⊥/Γe|| is thus very small, and only in the thin layer of width δ near the
sidewall, where the divergences of parallel and perpendicular fluxes are comparable,
does the electron motion become two-dimensional. Therefore, the width of this layer
is

δ  LΓe⊥ =Γek  LνeN =ωce : ð5:68Þ

Equation (5.66) has the particular solution

nðr, z, t Þ = AJ 0 ðζ1 r=aÞf ðzÞ expð- t=τi⊥ Þ, ð5:69Þ

which corresponds to the decay with the “short-circuiting” τi⊥, where τi⊥ is defined
by Eq. (5.52). Due to the proper account for the potential drop in the sheath adjacent
to the end walls, this decay time differs by the factor (1 + Te/Ti) from the original
value derived by Simon from qualitative consideration. The difference becomes
132 5 Diffusion of Partially Ionized Magnetized Plasma

particularly important in the case Te ≫ Ti, which is typical for devices with gas
discharge plasmas.
The potential drop between a point in the plasma and the end wall is determined
by Eq. (5.65). For the profile Eq. (5.69), the radial dependencies of the density and
particle escape rate in the column, ∂N1/∂t, coincide, and, as a result, the potential is
independent of the radius. In other words, the radial electric field is absent almost in
the whole volume of the tube, with the exception of a thin layer of the order of δ,
Eq. (5.68), near the sidewalls.
By applying voltage between the end and sidewalls, it is possible to control the
decay time in the range from τf to τs; for further details, see [7].

5.6 Diffusive Probe in Magnetic Field

In the magnetized plasma, the ionized particles are collected by the probe in a way
different from that analyzed in Sect. 4.2. First, in a strong magnetic field, the motion
of charged particles in the vicinity of a small electrode (probe) can be described by
the transport equations even when their mean free paths exceed the probe dimen-
sions. We shall show below that the criterion of applicability of the electron
description in terms of diffusion and mobility is

ρce ≪ a, ð5:70Þ

where ρce is the electron gyroradius and a is the probe size perpendicular to the
magnetic field. The reason is that in the strongly magnetized plasma for a positively
biased probe, the plasma density is perturbed in a region with the scale lk = aωce/
veN = axe along the magnetic field. Hence, the condition λeN ≪ lk is equivalent to
inequality (5.70). For magnetized ions, the fluid description is valid if their
gyroradius is smaller than the probe size. However, in the opposite situation, part
of the results obtained below remains valid for the positively biased probe. There-
fore, the fluid approach in magnetized plasma can be used even in plasma with
relatively rare collisions.
Density and potential profiles in the fluid approximation are described by the
equation system

→ → →
∇
De ∇n - be n∇φ = 0;
→ → → ð5:71Þ
∇
Di ∇n þ bi n∇φ = 0:

→ →
with the boundary conditions n r → 1 = n0 ; φ r → 1 = 0. For electron
saturation current, potential distribution coincides with the Boltzmann distribution
5.6 Diffusive Probe in Magnetic Field 133

for ions φ = - (Ti/e) ln n/n0, which corresponds to the absence of the ion current to
the probe. The electron equation in Eq. (5.71) then has the form

2
De⊥ ∂ ∂n ∂ n
r þ Dej j 2 = 0: ð5:72Þ
r ∂r ∂r ∂z

By coordinate transformation z′ = z(De⊥/Dek)1/2, this equation is reduced to the

Laplace equation

Δn = 0: ð5:73Þ

The probe, which has the form of an ellipsoid of rotation with semiaxes a, a, b, in the
new coordinates is transformed into a disk with semiaxes a, a, b/xe. The problem is,
therefore, reduced to one considered in Chap. 4 for an unmagnetized probe. The
solution is sought in ellipsoidal coordinates, in which the initial variables have the
form

x2 þ y2 z2
þ 2 = 1: ð5:74Þ
a þ ξ b þ ξ x2e
2

The plasma density is given by

1 - γ2e
→
arctg γ2e þξ=a2
n r  n0 ð 1 - n Þ = n0 1 -
e
, ð5:75Þ
1 - γ2e
arctg γ2e

where γe = b/axe. From Eq. (5.75), one can see that the longitudinal scale of the
depletion region is lk = aωce/veN = axe. For b ≪ axe, this longitudinal scale is much
larger than the longitudinal probe size.
The electron saturation current in the new coordinates coincides with Eq. (4.40),
while the capacity of a disk with small longitudinal semiaxis b/xe is C = 2a/π. When
returning to the initial coordinates, one must take into account that the saturation
current is also transformed. Finally,

e = 8en0 a xe ð1 þ T i =T e ÞDe⊥ :
I sat ð5:76Þ

The large longitudinal scale of the perturbed region in plasma, from which the
electron saturation current is collected, leads to a strong restriction on the probe’s
perpendicular size. Diagnostics should not change the global plasma distribution,
and this is true if lk ≪ L, where L is the typical plasma scale along the magnetic field.
This condition is not always satisfied in magnetized plasma.
134 5 Diffusion of Partially Ionized Magnetized Plasma

Fig. 5.9 Dependence of the transverse ambipolar diffusion coefficient in a helium plasma on the
magnetic field; sharp increase at B = Bc is due to transition to the turbulent state. Dashed line
corresponds to the classical theory, a = 2 cm, L = 75–80 cm

5.7 Experiments in Laboratory Plasma

Diffusion in magnetized plasma has been studied in many laboratory experiments.

The first example is related to diffusion perpendicular to the magnetic field in
dielectric tubes. Classical ambipolar diffusion has been observed, e.g., in decaying
plasma where instabilities have been stabilized by a metal electrode perpendicular to
the magnetic field. In Fig. 5.9, the perpendicular ambipolar diffusion coefficient,
which was calculated from the measured plasma lifetimes, is displayed versus
magnetic field strength. The result is in reasonable agreement with the classical
values given by Eq. (5.13), the ambipolar diffusion coefficient is inversely propor-
tional to the squared magnetic field strength.
The second example is connected with the density gradient inclined to the
magnetic field considered in Sect. 5.2.2. If the magnetic field is parallel to the side
surface of the dielectric tube, the particle lifetime in decaying plasma is governed by
the perpendicular ambipolar diffusion coefficient. For small inclination angles, the
decay time dramatically decreases (Fig. 5.10), in accordance with Eq. (5.19).
The plasma decay in a cylinder with conducting surfaces is shown in Fig. 5.11.
When the end and sidewalls are disconnected, the time of plasma decay is the same
as in the dielectric tube, Eq. (5.62). The acceleration of the decay process can be
initiated at any arbitrary moment by short-circuiting of the end and the sidewalls.
Before short-circuiting, the slow ambipolar decay process, as described in the
preceding section, took place, and after the short-circuiting of the walls, the regime
of fast diffusion was observed. The decay rate after short-circuiting increased
drastically by more than two orders of magnitude.
5.7 Experiments in Laboratory Plasma 135

Fig. 5.10 Effect of inclination of dielectric tube on the particle lifetime. Experiments in He in a
tube of length 90 cm, B = 0.1T. Angle between the tube axis and the magnetic field is α = Δx/L.
Different types of dots correspond to the inclination in two perpendicular directions. Solid line –
calculations by Eq. (5.19) for L → 1 with d replaced by πa/ζ

Fig. 5.11 Decay of helium plasma in a metallic chamber; a = 2 cm, p = 13.3 Pa, B = 0.2T. The
end electrode was simulated by a cylinder in a weak field. Arrow indicates moment when the end
and side electrodes are connected; x – ambipolar decay, o – short circuited walls
Chapter 6
Partially Ionized Plasma with Current

6.1 Plasma with Net Current in the Absence of a Magnetic

Field

As shown in Chap. 4, in pure plasma, which consists of electrons and one species of
ions, the net current through the plasma does not change the density evolution in the
absence of a magnetic field. The equation system, Eq. (4.1), for pure plasma is
reduced to the equation for ambipolar diffusion equation (4.2) as in the absence of
the current, while the current reveals itself only in the additional electric field,
Eq. (4.14), which is proportional to the current.
The situation becomes essentially different when two or more ion species are
present in plasma. Then, as we shall see, the convection of different quasineutral
perturbations takes place. Let us consider the evolution of multispecies plasma with
р ionized particles: p - 1 ion species and electrons. The basic equations are

∂nα ∂ ∂n
- Dα α - Z α nα bα E = 0;
∂t ∂x ∂x
∂ne ∂ ∂n
- De e þ be ne E = 0; ð6:1Þ
∂t ∂x ∂x
p-1
ne = Z α nα ,
α=1

where subscript α corresponds to ion species. The boundary conditions for

Eq. (6.1) are:

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 137
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_6
138 6 Partially Ionized Plasma with Current

nα ð x → ± 1 Þ = n0α ;
ne ðx → ± 1 Þ = n0 ; ð6:2Þ
E ð x → ± 1Þ = E 0 ≠ 0:

The last condition implies that a finite electric field E0 exists at infinity, which is
responsible for the flow of current j.
Subtracting the sum of ion particle balance equations from the electron balance
→
equation in Eq. (6.1), one obtains the current continuity equation ∇  j = 0. In the
1D case, this condition corresponds to the current conservation

p-1 p-1
j = eE0 be n0 þ Z 2α bα nαð0Þ = eE be ne þ Z 2α bα nα
α=1 α=1
ð6:3Þ
p-1
∂n ∂n
þe De e - Dα Z α α :
∂x α=1
∂x

Evaluating the electric field from this equation and substituting it into ion balance
equation (6.1), we obtain the equation system for ion densities

p-1
∂nα ∂ Z α bα nα
þ E0 n0 Z β be þ Z β bβ
∂t ∂x β = 1 β p-1
be þ Z β bβ Z β nβ
β=1
p-1 ∂nβ ð6:4Þ
De - Dβ Z β
∂ ∂n β=1 ∂x
þ - D α α - Z α bα nα p - 1 = 0:
∂x ∂x
be þ Z β bβ Z β nβ
β=1

In the pure plasma with ions of one species p = 2 with Z = 1 second term on the
l.h.s. turns to zero since the sum in the denominator is equal to (be + bi)n1 and cancels
with the nominator, so Eq. (6.4) is reduced to ambipolar diffusion equation (4.2). In
contrast, in multispecies plasma with several species of ions, this term is quite
important, and the ambient electric field E0 to a large extent determines the evolution
of the inhomogeneity. Hence, the situation in pure plasma should be considered a
degenerated case.
Let us analyze the solution of Eq. (6.3) in a so-called drift approximation, when
the electric field associated with the current is strong enough, the following inequal-
ity is satisfied:

E0 ≫ T=ðeLÞ, ð6:5Þ

where L is the spatial scale of the inhomogeneity. In this situation, diffusive

processes play a modest role, and the corresponding term (third one on the l.h.s. of
6.1 Plasma with Net Current in the Absence of a Magnetic Field 139

Eq. (6.4)) can be neglected. For simplicity, let us restrict ourselves to two species of
ions, 1 and 2, with Z1, 2 = 1. Then, in the drift approximation, the equation system
(6.4) with account of b1, 2 ≪ be is reduced to

∂n1 ∂ n1
þ b1 E 0 n0 = 0,
∂t ∂x n1 þ n2
ð6:6Þ
∂n2 ∂ n2
þ b2 E 0 n0 = 0:
∂t ∂x n1 þ n2

6.1.1 Small Perturbations

Let us find the solution of Eq. (6.6) in the drift approximation for small perturbations
of ion densities

nα = δnα þ n0α , δnα ≪ n0α :

The solution of linearized Eq. (6.6) is sought as the Fourier integral

þ1
1
δn1,2 ðx, t Þ = δn1,2k expðikx - iωt Þdk:
2π
-1

After substitution into linearized Eq. (6.6), one obtains an algebraic system of
equations for δn1k and δn2k. Since the determinant of this algebraic system should
be zero, two roots for the frequency are obtained

ω2 b1 n02 þ b2 n01
ω1 = 0, = E0 = V: ð6:7Þ
k n01 þ n02

The corresponding Fourier transforms are

b2 n01
δn1k ðt Þ = ½δn2k ð0Þ þ δn1k ð0Þb2 =b1 
þ b2 n01
b1 n02
þ - δn2k ð0Þ þ δn1k ð0Þn02 =n01 expð - ikVt Þ ;
ð6:8Þ
b1 n02
δn2k ðt Þ = ½δn1k ð0Þ þ δn2k ð0Þb1 =b2 
b1 n02 þ b2 n01
þ - δn1k ð0Þ þ δn2k ð0Þn01 =n02 expð - ikVt Þ :
140 6 Partially Ionized Plasma with Current

Fig. 6.1 Small perturbations in current carrying plasma. Injected ions of species 1 with Gaussian
profile of width а are injected into plasma with ambient ions of species 2. (а) Initial moment, (b)
moment t = 5a/(b1E0). Solid line – injected ions, dash-dotted line – ambient ions, dotted line –
electron density. Mobility ratio: b1 = 2b2 ≪ be

Density perturbations contain moving contribution and zero-velocity parts. The

stable contribution corresponds to zero frequency ω1, while the moving part contains
the density of two species moving with velocity V.
This phenomenon of density perturbation propagation in the electric field is
known as ambipolar mobility, while velocity V is known as ambipolar mobility
velocity.
Let us consider a special case when initially background plasma consists of the
ions of species 2, so that n01 = 0, and ions of species 1 are injected into the
background plasma at t = 0, so that δn2(x, t = 0) = 0. Then, in the limit n01 → 0 in
Eq. (6.8), one finds

b2
δn1k ðt Þ = δn1k ð0Þ expð- ikb1 E0 t Þ, δn2k ðt Þ = δn1k ð0Þ ½1 - expð- ikb1 E 0 t Þ: ð6:9Þ
b1

In the initial coordinates

δn1 ðx:t Þ = δn1 ðx - b1 E 0 t Þ,

b ð6:10Þ
δn2 ðx, t Þ = 2 ðδn1 ðx, t = 0Þ - δn1 ðx, t ÞÞ:
b1

The corresponding profiles are shown in Fig. 6.1.

We see that at t = 0, injected ions of species 1 are localized in the place of
injection, while the perturbation of background ions of species 2 is absent. The
perturbation of injected ions moves with the constant velocity b1E0. In other words,
injected ions move in the unperturbed electric field. A small perturbation of the
electric field has an impact only on the background ions. According to Eq. (6.1), at
the place where ions of species 1 were injected, the peak of the background ions is
formed with an amplitude larger or smaller than that of injected ions, depending on
the ratio of species mobilities b1/b2. The second negative perturbation of the
6.1 Plasma with Net Current in the Absence of a Magnetic Field 141

background ions of the same amplitude accompanies the peak of injected ions. It is
interesting to note that in the special case b1 = b2, the moving perturbation of
electron density disappears since positive and negative perturbations of injected and
background ions compensate for each other. The initial perturbation of the electron
density simply stays at rest, which corresponds to the solution for the case of pure
plasma, which is described by the ambipolar diffusion equation. In the drift approx-
imation, this implies the absence of any evolution. Hence, even if the mobilities of
ions coincide, a moving signal exists that can be observed, provided the injected and
background ions differ by some other physical feature.
In the general case in multispecies plasma, the arbitrary initial perturbation splits
into р - 1 signals, with one of them having zero velocity. From the solution obtained
above, we can conclude that the fact that in the pure current-carrying plasma, the
density evolution is described by the same ambipolar diffusion equation as in the
currentless plasma results from the exact compensation of several rather complicated
disturbances of ambient and injected particle densities. In plasmas of more complex
composition, in addition to the fixed density disturbance, several modes of propa-
gating weakly damping signals exist. In each of these modes, the densities of all the
plasma components are perturbed.

6.1.2 Nonlinear Evolution

We shall start the nonlinear analysis with the simplest case b2 = 0 when the mobility
of one species of positive ions, in particular ambient ions, is negligible. The well-
known situation of this kind corresponds to an n-type semiconductor, where posi-
ð0Þ
tively charged fixed donor ions with spatially uniform density n2 are present. Some
situations in gaseous and dusty plasmas can also be described by this approximation.
Assuming n2 = n02 = const and n01 = 0, in the drift approximation from Eq. (6.6), we
obtain the equation for injected ions

∂n1 ∂n
þ V ðn1 Þ 1 = 0, ð6:11Þ
∂t ∂x

where V(n1) is a nonlinear ambipolar drift velocity:

2
b1 E0 n02
V ðn1 Þ = 2
: ð6:12Þ
n1 þ n02

Equation (6.11) describes propagation in a purely nonlinear wave, and its solution is

n1 = n1 ðx - V ðn1 Þt Þ: ð6:13Þ
142 6 Partially Ionized Plasma with Current

Fig. 6.2 Evolution of nonlinear initial Gaussian profile of injected ions in plasma with ambient ions
with zero mobility. Dimensionless density is n1 = n1 =n02 , time is t = tb1 E 0 =a, initial perturbation
1 ðt = 0Þ = 2. (а) Initial profile and (b) t = 15 after shock formation
amplitude nmax

In other words, each point of the initial profile of injected ions with density n1
propagates with its own velocity V(n1). According to Eq. (6.12), parts of the profile
with larger densities move slower than those with smaller densities; hence, the front
of the inhomogeneity becomes shallower with time, while the rear steepens
(Fig. 6.2). In the evolution process, a multivalued density profile is formed, which
is the consequence of the drift approximation used. This phenomenon is similar to
the wave overturn in the standard hydrodynamics.
To obtain a physically meaningful solution, it is necessary to keep diffusion terms
in the initial equation system. Indeed, in the process of evolution, steepening of the
profile takes place, and a sharp density gradient – diffusion shock – is formed, where
Eq. (6.5) becomes invalid. Typical spatial scale of this diffusion shock

T
lT = ð6:14Þ
eE

corresponds to diffusive and convective terms in Eq. (6.4) being of the same order.
Shock position, as in hydrodynamics, could be obtained from the “area rule”. At the
multivalued profile, the vertical line is chosen so that the area below the multivalued
solution and the area below the profile with shock should coincide (Fig. 6.2). The
total number of particles is conserved and remains the same for both profiles, while
the gradual part of the new profile with shock satisfies the initial equations since
inequality (6.5) is fulfilled. Shock velocity propagation is also determined by particle
conservation:

Γ1 nþ -
1 - Γ1 n1
W= , ð6:15Þ
nþ -
1 - n1

where Γ1 = n1b1E is the particle flux, and quantities nþ -

1 and n1 correspond to the
density on the right and left sides of the shock, respectively. The shock structure, as
in standard hydrodynamics, could be obtained by transfer to the reference frame
moving with the velocity W and reduction of the partial derivative equations to the
ordinary derivative equations.
6.2 Magnetized Plasma with Current 143

6.2 Magnetized Plasma with Current

6.2.1 One-Dimensional Evolution

In the magnetized plasma, the phenomenon of ambipolar mobility reveals itself even
in pure plasma with one species of ions. Let us consider plasma evolution in the same
geometry as in Sect. 5.2.2 when plasma density depends only on coordinate ζ,
forming an angle β with the magnetic field (Fig. 5.4). In contrast to Sect. 5.2, we
→
now assume the existence of an arbitrary ambient electric field E 0 . In the coordi-
nates (x, η, ζ), the initial equations, analogous to those considered in Sect. 5.2, have
the form

∂n ∂ ∂n ∂φ ∂φ ∂φ
- Dαζζ ± bαζζ n ± bαζx n ± bαζη n = I - R, ð6:16Þ
∂t ∂ζ ∂ζ ∂ζ ∂x ∂η

where α = i, e, and the components of the diffusion and mobility tensors are given by
Eq. (5.16). The boundary condition for the potential is more general:

→ →
∇φ r → 1 = - E 0 : ð6:17Þ

Note that the components of the electric field Eη and Ex are independent of coordi-
nate ζ and coincide with the corresponding ambient values: Eη = E0η and Ex = E0x.
This follows from the electrostatic character of the electric field, so that ∂Eη/
∂ζ = ∂Eζ/∂η = 0 and ∂Ex/∂ζ = ∂Eζ/∂x = 0. The electric field in the ζ direction
can be eliminated similar to Sect. 5.2. Multiplying equations for ions and electrons
by beζζ and biζζ correspondingly, after summing them up, we obtain

∂n ∂ ∂n ∂n
- D μ2 þ V ζ μ2 = I - R, ð6:18Þ
∂t ∂ζ ∂ζ ∂ζ

where Vζ(μ2) is the ζ component of a vector

→ →
→ bek μ2 þ be⊥ ð1 - μ2 Þ bi E 0 - bik μ2 þ bi⊥ ð1 - μ2 Þ be E 0
V μ 2
= : ð6:19Þ
bek þ bik μ2 þ ðbi⊥ þ be⊥ Þð1 - μ2 Þ

According to Eq. (6.18), in addition to the diffusion plasma, perturbation has a

convective velocity – ambipolar drift velocity. According to Eq. (6.19), if the electric
field belongs to the plane (η, ζ), the ambipolar drift velocity is zero for the ζ-axis
parallel or perpendicular to the magnetic field but remains finite for an arbitrary angle
between the ζ-axis and magnetic field direction. The ambipolar velocity caused by
→ →
the Ex component for magnetized ions corresponds to E × B drift.
144 6 Partially Ionized Plasma with Current

The electric field in the ζ-direction is obtained by subtraction of the ion equation
from the electron equation and integration over ζ. The quantity Eζ contains a
→
diffusive part and a part linear to the ambient field E 0 .

6.2.2 Multidimensional Evolution of Small Perturbation

in Unbounded Plasma

The evolution of small perturbation δn at a constant ambient background with

→
density n0 in the presence of the ambient electric field E 0 qualitatively differs
from the diffusive evolution considered in Sect. 5.3. Initial equations for perturbed
→
density and potential in the presence of the ambient electric field E 0 are given by

∂n →
- ∇  De ∇n - be n∇φ þ be n E 0 = 0,
∂t ð6:20Þ
∂n →
- ∇  Di ∇n þ bi n∇φ - bi n E 0 = 0
∂t
→
with the same boundary conditions as without E 0 :

→ →
n r → 1 = n0 , φ r → 1 = 0:

Solution of the linearized equation system, Eq. (6.20), with the initial perturbation in
→ →
the form of delta-function δn r , 0 = Nδ r with the Fourier method leads to the
expression

→ N → → → →
δn r , t = exp i k r - V μ2 t - D μ2 k 2 t d k , ð6:21Þ
ð2πÞ3
→
where velocity V ðμ2 Þ is defined according to Eq. (6.19). This integral was analyzed
in detail in [7]. In particular, it was demonstrated that the initial perturbation is
separated with time into several moving perturbations. An example of such an
evolution when the electric field is perpendicular to the magnetic field is presented
in Fig. 6.3. The initial perturbation here is separated into two blobs moving with
velocities close to the velocities of electrons and ions in the ambient electric field
(Fig. 6.3a). Polarization of the blobs also leads to their deformation, especially for
electron blob along the magnetic field. As in the process of diffusion, quasineutrality
is maintained by the short-circuiting currents in the ambient plasma and the forma-
tion of depletion regions that move together with the blobs. Strong perturbation,
when the density of the ambient plasma is not sufficient to provide quasineutrality,
6.2 Magnetized Plasma with Current 145

→ →
Fig. 6.3 Evolution of density perturbation for E 0 ⊥ B . (a) Small perturbation and (b) strong
perturbation. Here OO1 and OO2 are the direction of blobs motion, the depletion regions are dashed

mainly remains at rest while its peripheral parts drift (see Fig. 6.3b). As a result,
nonlinear deformation of the blob occurs.

6.2.3 Effect of Conductivity Recovery in a Weak Magnetic

Field

Here, we shall consider conductivity in a weak magnetic field xexi ≪ 1 when plasma
remains unmagnetized. On the other hand, the magnetic field is assumed to be strong
enough so that electrons become magnetized, xe ≫ 1. In such plasma, the perpen-
dicular electron mobility is significantly smaller than the parallel electron mobility
be⊥ ≪ bek, but in contrast to the situation considered previously, the perpendicular
ion mobility is small with respect to the electron mobility, bi⊥ ≪ be⊥. This case is
typical, for example, for the plasma of MHD generators or low ionosphere plasma.
Let us consider plasma flow in a channel restricted by two infinite dielectric
→
surfaces. An electric field E 0 is applied in the x-direction parallel to the surfaces, as
shown in Fig. 6.4. The boundary condition at the dielectric surfaces corresponds to
the absence of the current in the y-direction: jy = 0. Note that the initial electric field
→
E 0 causes not only current in the х-direction but also the Hall current of electrons in
the у-direction. Therefore, the plasma is polarized, as shown in Fig. 6.4, and an
additional electric field in the y-direction, Ey, arises. The ion fluxes in both the х- and
у-directions are negligibly small and do not contribute to the current. For simplicity,
we shall also neglect the diffusive electric field of the order of T/eL with respect to
146 6 Partially Ionized Plasma with Current

Fig. 6.4 Unmagnetized

plasma polarization in a
channel with dielectric
surfaces

the field E0. Then, the polarization field in the у-direction, Ey, can be found from the
condition of zero current normal to the surface

jy = - e - be^ nE0 - be⊥ nE y = 0: ð6:22Þ

The field Ey obtained from Eq. (6.22) is negative:

E y = - E 0 beΛ =be⊥ : ð6:23Þ

Since electrons are magnetized, the ratio beΛ/be⊥ ≫ 1, and the polarization field in
the y-direction is much larger than the initial field E0. This polarization field Ey
→ →
causes E y × B drift of electrons in the х-direction. The latter together with the small
current caused by the field E0 determines the net current along x:

jx = - enbeΛ E y þ enbe⊥ E 0 = enE 0 b2eΛ =be⊥ þ be⊥ = enbek E0 : ð6:24Þ

The effective conductivity is the coefficient, which connects the current density with
the applied electric field and coincides with the conductivity in the absence of a
magnetic field σk = ebekn,

jx = σeff E 0 = σk E0 : ð6:25Þ

In other words, conductivity across a magnetic field is “recovered” and coincides

with the conductivity in the absence of a magnetic field. The main contribution to
this effect of conductivity recovery is due to the polarization electric field, which by
factor xe exceeds the initial electric field and produces the Hall current of electrons.
This result is of course a direct consequence of the chosen boundary condition at
the material surfaces, jy = 0. In the opposite limiting case of the two short-circuiting
highly conductive surfaces, the current in the у-direction can flow freely, and the
effect of conductivity recovery is absent. The polarization electric field in this case
does not exist, and the current along the х-axis is simply jx = enbe⊥E0. In addition to
6.3 Plasma Clouds in the Ionosphere 147

these two limiting cases of short-circuited, jy = 0, and disconnected, Ey = 0,

surfaces, there are also possible intermediate cases with finite resistance between
the plates. The dependence of the effective conductivity on the resistance value is
easily calculated analogously to the limiting cases.

6.3 Plasma Clouds in the Ionosphere

Ionospheric plasma provides many examples of density evolution in the presence of

currents and global electric fields.

6.3.1 Redistribution of Metal Ions in the Polar Ionosphere.

Sporadic Layers

The first example is the formation of plasma layers in the Polar ionosphere. The
traditional explanation of the formation of sporadic layers in the inhomogeneous
ionosphere plasma consists of the so-called wind shear model. In the reference frame
that moves with the neutral gas across the magnetic field, due to the Lorenz
transformation, an electric field arises. The inhomogeneous Hall fluxes in these
crossed fields result in the piling up of the plasma density. In the lower part of the
polar ionosphere, narrow layers of high plasma density with sharp gradients are often
observed. To explain these layers in terms of the wind shear theory, it is necessary to
assume that the neutral wind changes its direction on a very short spatial scale. On
the other hand, the lifetime of these layers is very large, reaching several hours. This
contradicts the fact that the characteristic ionization and recombination times of the
main ionospheric ions (NO+, O2+) are of the order of 10–20 s. Therefore, these
sporadic layers probably consist of metal ions, n1 which are characterized by very
long recombination times. Furthermore, in the Polar ionosphere, a considerable
current parallel to the Earth’s magnetic field flows. This current in the multispecies
ionospheric plasma can cause the formation of shocks analyzed in this chapter in
Sect. 6.1. Indeed, the ambient ion density n2 = n02 can be considered constant due to
fast recombination. The injected ions of species 1 can be associated with metal ions
of meteor origin with zero recombination. Their density may significantly exceed the
background density. Then, the situation is similar to that presented in Fig. 6.2 and is
described by Eq. (6.11).
The shock arises at the rear side of a density hump (if the current outflows from
the ionosphere, the shock is formed at the lower side of an inhomogeneity). For
typical ionospheric parameters, the values of lT in Eq. (6.14) are of the order of
several hundred meters and L is of the order of several kilometers. These numbers
are within the experimentally observed range.
148 6 Partially Ionized Plasma with Current

6.3.2 Active Experiments with Barium Clouds

Drift and deformation of strong plasma inhomogeneities were studied in many

experiments where barium (and some other) clouds were injected into the iono-
sphere. In these experiments, the density of injected Ba+ ions was rather high with
respect to the ambient plasma density. Therefore, the main fraction of the injected
ions for a long time remained at rest in the reference frame of the neutral wind (the
main cloud). Simultaneously, part of the injected ions outflowed in the perpendicular
→ →
E × B direction, creating a tail. The motion of this tail is possible due to the short-
circuiting effect (see Sect. 6.2.2). Only one of the two separate plasma clouds, where
barium ions are located, can be seen by spectroscopy measurements, and the second
blob consists of ambient ions and remains invisible. A photo of the Ba+ cloud is
presented in Fig. 6.5. One can see the shock on the rear side of the cloud and the
→ →
outflow of Ba+ in the E × B direction. The cloud is separated into striations due to
→ →
specific instability. Small-density Ba+ clouds move in the E × B direction without
significant deformation, and they can be used to measure the electric field in the
ionosphere.

Fig. 6.5 Photo of the barium cloud evolution in the “Spolokh” experiment made from the different
points of observation (a) and (b). Plasma is separated into striations stretched along the Earth
magnetic field. In (b), the line of sight is perpendicular to the outflow direction
Chapter 7
Transport in Strongly Ionized Plasma
Across a Magnetic Field

7.1 Classical Diffusion of Fully Ionized Plasma Across

a Magnetic Field

First, let us demonstrate that evolution along the magnetic field (or in the absence of
a magnetic field) in fully ionized plasma (in contrast to slightly ionized plasma) is not
of diffusive character. To show this, let us sum two momentum balance equations for
electrons and single-charged ions:

→
due → →
me n = - ∇pe - en E þ R ei ;
dt ð7:1Þ
→
dui → →
mi n = - ∇pi þ en E þ R ie :
dt

Both electric forces and friction are canceled during summation, and neglecting
electron inertia, one has

→
dui
mi n = - ∇p, ð7:2Þ
dt

where p = pe + pi is the net pressure.

This Eq. (7.2) describes density evolution with the velocity of the order of a sound
speed. Indeed, estimating the inertia term as mi nu2i =L, where L is the characteristic
spatial scale, assuming |∇p|~p/L, we obtain ui~cs = [(Te + Ti/)mi]1/2 with cs being the
sound speed. The dynamics of such plasma evolution are considered in Chap. 9.
In contrast, the evolution of fully ionized plasma across a magnetic field has a
diffusive character, similar to the case of partially ionized plasma. The pressure
gradient across the magnetic field is balanced by the Lorentz force, and the radial
velocity is significantly smaller than the sound speed so that the inertia term in the
momentum balance equation can be neglected.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 149
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_7
150 7 Transport in Strongly Ionized Plasma Across a Magnetic Field

We consider diffusion in an infinite cylinder placed in a strong homogeneous

magnetic field parallel to the z-axis. The plasma density and temperatures depend
only on the radial coordinate. Neglecting viscosity and inertia terms in the momen-
tum balance equations, we have (see Sect. 2.3)

→
→ → → → 3 nνei B
- ∇pe þ en∇φ - en u e × B - nme νei u e - u i - × ∇T e = 0,
2 ωce B
→
→ → → → 3 nνei B
- ∇pi - en∇φ þ en u i × B þ nme νei u e - u i þ × ∇T e = 0:
2 ωce B
ð7:3Þ

The sum of these equations gives us the equilibrium equation

→ →
∇p = j × B : ð7:4Þ

According to Eq. (7.4), the pressure gradient is balanced by the Lorentz force, which
is caused by the diamagnetic current flowing in the azimuthal direction. The
mechanism of diamagnetic current formation is discussed in Chap. 5. Substituting
→ →
the current from the Maxwellian equation ∇ × B = μ0 j , after integrating over
radius, one obtains

B2 B2
pþ = const = 0 , ð7:5Þ
2μ0 2μ0

where B0 is the vacuum magnetic field. As follows from Eq. (7.5), such equilibrium
is possible if

β = 2μ0 p=B2 < 1: ð7:6Þ

Note that for β < < 1, the vacuum magnetic field is only slightly perturbed by the
magnetic field of diamagnetic currents.
Radial projections of the momentum balance equations (7.3) read

dpe dφ
- þ en - enueϑ B = 0,
dr dr ð7:7Þ
dp dφ
- i - en þ enuiϑ B = 0:
dr dr

Note that the radial component of the friction force is absent; it is zero since the
radial velocities of electrons and ions must coincide; otherwise, a radial nonzero
current should flow, which is restricted by the quasineutrality constraint. The thermal
7.1 Classical Diffusion of Fully Ionized Plasma Across a Magnetic Field 151

force also does not contribute to the radial momentum balance; only the poloidal
component of the thermal force exists.
The azimuthal velocities according to Eq. (7.7) are:

1 dpe 1 dφ
ueϑ = - þ = upe þ V 0 ,
eBn dr B dr ð7:8Þ
1 dpi 1 dφ
uiϑ = þ = upi þ V 0 :
eBn dr B dr

These velocities consist of two contributions. The first part upe, upi represents the
diamagnetic flux of electrons and ions associated with rotation over the Larmor
radius in inhomogeneous plasma. The origin of diamagnetic fluxes is discussed in
→ →
Sect. 5.2. The second contribution corresponds to E × B drifts in the radial electric
field with velocity V0.
The azimuthal projections of the momentum balance equations have the form

3 nνei dT e
enuer B - nme νei ðueϑ - uiϑ Þ - = 0,
2 ωce dr
ð7:9Þ
3 nνei dT e
- enuir B þ nme νei ðueϑ - uiϑ Þ þ = 0:
2 ωce dr

Substituting the difference in the azimuthal velocities (ueϑ - uiϑ) obtained from
Eq. (7.8) into Eq. (7.9), one finds

me νei dp 3 νei 1 dT e
uer = uir = ur = - þ : ð7:10Þ
e2 B2 n dr 2 ωce eB dr

In this approximation, electrons and ions have equal radial velocities that are
independent of the radial electric field value. This situation is quite different from
the case of partially ionized plasma, where ion and electron radial fluxes are
essentially different. Sometimes this fact is known as a statement that in fully ionized
plasma transport is “automatically ambipolar.”
The radial velocity can be interpreted as a particle drift caused by the azimuthal
friction force (Fig. 7.1). Indeed, the azimuthal friction force associated with the
particle’s diamagnetic velocities with opposite directions and the azimuthal thermal
force is applied both to electrons and ions. These forces cause radial drift of ions with
velocity uri = Rieϑ/(eBn) and drift for electrons with velocity ure = - Reiϑ/(eBn).
Since according to the third Newton law Reiϑ = - Rieϑ and the signs of particle
charges are also different, we have uer = uir.
Note that for plasma consisting of ions with charge number Z and density ni
(n = Zni), the radial electron velocity remains the same, given by Eq. (7.10), while
the radial velocity of ions is smaller uri = Rieϑ/(ZeBni). Their radial flux is Z times
smaller than that of electrons:
152 7 Transport in Strongly Ionized Plasma Across a Magnetic Field

Fig. 7.1 Fluid velocities in a cylinder of fully ionized plasma oriented along magnetic field

ni uir = nuer =Z:

The radial current, however, is still zero since each ion carries charge Ze.
In the approximation described above, the radial electric field remains arbitrary.
To find the radial electric field, one has to go to the next approximation for the
momentum balance equations and take into account corrections connected with the
ion viscosity. Indeed, since the azimuthal velocity of ions given by Eq. (7.8) depends
on the radius, there is an azimuthal viscous force applied to ions. This force is
proportional to the radial gradient of the azimuthal velocity ∂uiϑ/∂r. In the absence
of an ion temperature gradient, this force, in accordance with Eq. (2.48), is given by

$ d 1 d
- ∇ π = η ðru Þ : ð7:11Þ
ϑ dr 1 r dr iϑ

A similar electron viscosity is very small and can be neglected. The additional
viscous force applied to ions causes additional ion radial flux, which is absent for
electrons, and hence produces radial current

1 d 1 d dpi
jr = - 2 dr
η1 r Er - : ð7:12Þ
B r dr endr

Since the radial current must be zero in nonuniform plasma, the following radial
electric field is established:

T i d ln n
Er = : ð7:13Þ
e dr

This field corresponds to the Boltzmann distribution for ions. In the plasma with an
inhomogeneous ion temperature, the additional term in Eq. (7.13) arises, which is
caused by the viscosity associated with the azimuthal heat flux.
The radial velocity in Eq. (7.10) can be rewritten as
7.1 Classical Diffusion of Fully Ionized Plasma Across a Magnetic Field 153

dn D 1 dT e dT i
ur = - D þ - , ð7:14Þ
ndr T e þ T i 2 dr dr

where diffusion coefficient D is defined as

me νei ðT e þ T i Þ
D= : ð7:15Þ
e2 B2

The first term on the r.h.s. of Eq. (7.14) corresponds to diffusive flux, while the
second one represents thermodiffusion. The diffusion coefficient D corresponds to a
simple estimate

D  ρ2ce νei ð7:16Þ

is of the order of square of the electron Larmor radius multiplied by the electron–ion
collision frequency. Note that the diffusion coefficient is of the order of the electron
heat conductivity coefficient χe and is (mi/me)1/2 times smaller than the ion heat
conductivity coefficient, which can be estimated as χi  ρ2ci νii . The ion–ion colli-
sions do not lead to particle diffusion because during the collision process, the
leading centers of the two ion orbits shift in the opposite direction without a center
mass shift.
A special feature of fully ionized plasma is the absence of conductivity perpen-
dicular to the magnetic field, in contrast to the partially ionized plasma case. Indeed,
→ →
the electric field causes only E × B drift for both electrons and ions with the same
velocity, and, therefore, there is no friction between them, in contrast to the case of
partially ionized plasma. In partially ionized plasma, friction with neutral particles
causes additional drift of particles, mainly ions, in the direction of the electric field,
which results in current, while in fully ionized plasma, such drift is absent. The
perpendicular current in the direction of the electric field can flow only in a
nonhomogeneous electric field; in this case, it is produced by viscosity as the current
given by Eq. (7.12).
Classical diffusion in magnetized fully ionized plasma has a specific feature: the
diffusion coefficient is proportional to plasma density because the electron-ion
collision frequency is proportional to plasma density. As a result, the corresponding
diffusion equation becomes nonlinear. In plasma with a constant temperature, the
diffusion equation has the form
p
∂n e ðT e þ T i ÞΛ
2π e2 m1=2
- εΔn = 0, D = 2εn; ε =
2
: ð7:17Þ
∂t 3ð4πε0 Þ 2
B2 T 3=2
e

We shall demonstrate the specific character of diffusion in fully ionized plasma by

constructing a self-similar solution of Eq. (7.17). As an example, let us consider a
one-dimensional (1D) case. A self-similar solution can be sought in the form
154 7 Transport in Strongly Ionized Plasma Across a Magnetic Field

nðx, t Þ = t γ nðξÞ; ξ = x=t δ : ð7:18Þ

Transformation to the new variables ξ, t yields

∂n dn
= γt γ - 1 n - δt γ - 1 ξ ,
∂t dξ
2 2 2 2
d n d n
2
= t 2γ - 2δ 2 :
dx dξ

Equation (7.17) is reduced to an ordinary differential equation if the powers of t in all

terms of Eq. (7.17) coincide, which is satisfied for γ = 2δ - 1. One more relation
should be obtained from the net particle conservation condition

ndx = const = t δþγ ndξ,

which is satisfied if δ = - γ. From these two equalities, one finds

1
δ = - γ, nðx, t Þ = n x=t 1=3 : ð7:19Þ
t 1=3

Function nðξÞ satisfies the ordinary differential equation

d 2 n2 ξ dn n
ε þ þ = 0: ð7:20Þ
dξ2 3 dξ 3

Its solution is

1
nð ξÞ = C - ξ2 :
12ε

In the initial variables, it is convenient to rewrite the solution, replacing the integra-
tion constant as a ratio of two new constants:

2=3
a2 x2 t 0
nðx, t Þ = 1- : ð7:21aÞ
2=3
12εt 1=3 t 0 a2 t 2=3

According to the obtained solution, the plasma density becomes zero at a finite
distance, and, at the moment t = t0, the density becomes zero at x = a. This fact is a
consequence of the dependence of the diffusion coefficient on density – it turns to
zero with density. The typical spatial scale of the profile rises with time as t1/3, and
the density is reduced correspondingly as t-1/3. The solution in the cylindrical
geometry can be obtained in a similar way and has the form
7.2 Transport of Impurities in Fully Ionized Plasma Across a Magnetic Field 155

1=2
a2 r2 t0
nðx, t Þ = 1- : ð7:21bÞ
1=2
16εt 1=2 t 0 a2 t 1=2

7.2 Transport of Impurities in Fully Ionized Plasma Across

a Magnetic Field

Consider the transport of heavy impurities with mass mI, density nI, and charge
number Z in the cylindrical geometry, the same as in the previous section. The
density of the main ions with Z = 1 is ni, nI < < ni, and subscript i also corresponds
to other parameters of the main ions. The main friction force applied to impurities is
the friction force with the main ions, while the friction with electrons is (me/mi)1/2
times smaller and, therefore, can be neglected. Since we assume mi < < mI, the
friction associated with the difference in velocities and thermal force has the same
form as in the case of electron-ion collisions (see Sect. 2.3):
→
→u → →
R iI = - ni mi νiI u i - u I ,
→
→T 3 νiI B
R iI =- n × ∇T i , ð7:22Þ
2 i ωcI B
p 4 2
2e Z ΛnI
νiI = p 3=2
:
12π3=2 ε20 mi T i

Azimuthal fluxes of impurity ions are obtained in the same way as for main ions:

c dpI c dφ
uIϑ = þ = upI þ V 0 : ð7:23Þ
eZBnI dr B dr

Diamagnetic fluxes of the main ions and impurity ions are shown in Fig. 7.2. For not
very steep profile of impurities when

d ln ni 1 d ln nI
>> ð7:24Þ
dr Z dr

diamagnetic velocity of impurity ions is significantly smaller than that of main ions.
Therefore, in plasma with constant ion temperatures, the direction of the azimuthal
friction force applied to impurities coincides with the main ion diamagnetic velocity.
This force causes the drift of impurity ions to the core of a plasma cylinder. In the
general case of an arbitrary ion temperature profile, the azimuthal momentum
balance with account of Eqs. (7.23) and (7.8) has the form
156 7 Transport in Strongly Ionized Plasma Across a Magnetic Field

Fig. 7.2 Fluid velocities of the main ions and impurities

eZnI 3 ni νiI dT i
u B þ ni mi νiI ðuiϑ - uIϑ Þ þ =0 ð7:25Þ
c Ir 2 ωci dr

Substituting azimuthal velocities into this equation, one obtains the radial flux of
impurity ions

ni mi νiI c2 T i 1 dni 1 dnI 1 dT i

ΓIr  nI uIr = 2 2
- - : ð7:26Þ
Ze B n i dr Zn I dr 2 dr

Since the collision frequency is proportional to the impurity density, the flux
Eq. (7.26) contains the diffusive part (second term on the r.h.s.) and convective
part (first and third terms on the r.h.s.). For the condition given by Eq. (7.24),
practically all radial flux is convective and directed towards the core of the plasma
pinch. In the absence of sources and sinks, the core will be enriched by impurity ions
until the outwards diffusive flux compensates for the inwards convective flux so that
the net flux turns to zero and a stationary profile is established. It corresponds to
equality

1 dni 1 dnI 1 dT i
- - = 0,
ni dr ZnI dr 2 dr

and is rather peaked:

- Z=2
nI = AnZi T i : ð7:27Þ

In other words, impurities are gathered in the center of the plasma.

Note that the radial flux of impurity ions does not produce radial current and thus
does not lead to an additional radial electric field. Indeed, since azimuthal friction
forces RiIϑ and RIiϑ differ only by the sign, we have
7.3 Partially Ionized Magnetized Plasma with an Inhomogeneous Neutral Component 157

ZΓIr = - Γir : ð7:28Þ

The counter flux of the main ions caused by impurities is Z times larger than that of
impurities, so the resulting radial current equals zero, and the presence of impurities
does not change the radial electric field.

7.3 Partially Ionized Magnetized Plasma

with an Inhomogeneous Neutral Component

Partially ionized plasma with inhomogeneous neutral density in a magnetic field is a

more complicated object. In contrast to Chap. 5, we shall consider neutral gas to be
strongly inhomogeneous. Neutral gas driven by the pressure gradient tries to expand
freely with a velocity of the order of the sound speed. However, in the presence of a
rather small fraction of ionized particles (small ionization degree), such expansion in
the collisional plasma could be stopped due to the interaction of the perpendicular
current with the magnetic field.
Let us analyze the diffusion of a slightly ionized plasma pinch oriented along the
magnetic field. Two conditions are assumed to be satisfied. First,

β = 2μ0 p=B2 < < 1: ð7:29Þ

Here, in contrast to Sect. 7.1, p = pN + pe + pi is the net pressure, which is defined as

the sum of the pressures of electrons, ions, and neutral particles. Second, the mean
free pass of neutral particles with respect to ion-neutral collisions λNi is supposed to
be smaller than the plasma density scale L:

λNi < < L: ð7:30Þ

We shall demonstrate that neutral gas is confined due to the interaction of magnetic
field with a small fraction of ionized particles provided that these two conditions are
satisfied, and the evolution of neutral particles has a diffusive character.
The momentum balance for neutral and ionized particles has the following form
provided that electron inertial and viscous forces are neglected:

→ → → → →
0 = - ∇pe - en E þRei þReN - en ue × B ;
→
du → → → → →
mi n i = - ∇pi þ en E þRie þ RiN þ en ui × B ; ð7:31Þ
dt
→
du → →
mN nN N = - ∇pN þ RNi þ RNe :
dt
158 7 Transport in Strongly Ionized Plasma Across a Magnetic Field

The sum of these three equations reads

→ →
duN dui → →
mN n N þ mi n = - ∇p þ j ×B : ð7:32Þ
dt dt

If ion and neutral inertia terms are neglected, then

→ →
∇p = j × B : ð7:33Þ

Substituting the current from the Maxwellian equation, we obtain an equation

similar to Eq. (7.5)

B2 B2
pþ = const = 0 , ð7:34Þ
2μ0 2μ0

but with full pressure p = pN + pe + pi, which includes the pressure of the neutral
component. This equation can be satisfied provided that inequality (7.29) is fulfilled.
In plasma with a small ionization degree restriction given by Eq. (7.29) is much more
severe than that in fully ionized plasma, Eq. (7.6).
The mechanism of establishing equilibrium could be analyzed using momentum
balance Eq. (7.31). We shall take into account only the largest friction force between
ions and neutral particles. Homogeneous temperatures are assumed, so thermal force
is neglected. The radial force balance for neutral particles in the absence of inertia
reads

dpN
= RNir = - mN nN νNi uNr , ð7:35Þ
dr

where the ion-neutral collision frequency νNi = nhVσiNi is proportional to the plasma
density. Here, the radial ion velocity is set to zero since the radial current is absent in
the cylindrical plasma. From Eq. (7.35), the radial velocity of neutral particles is
derived:

1 dpN
uNr = - : ð7:36Þ
nnN mN hVσiNi dr

If condition Eq. (7.30) is satisfied, this velocity is smaller than the sound speed, and
the transport of neutral particles has a diffusive character.
According to the azimuthal momentum balance or ions in the absence of their
radial velocity, the azimuthal friction force should be zero RiNϑ = 0, and, therefore,
the azimuthal velocities of ions and neutrals must coincide:

uNϑ = uiϑ : ð7:37Þ

7.3 Partially Ionized Magnetized Plasma with an Inhomogeneous Neutral Component 159

Radial momentum balance for ions yields

dpi
- þ enE r þ enuiϑ B þ RiNr = 0: ð7:38Þ
dr

Since the radial friction between ions and neutrals is balanced by the pressure
gradient of neutral particles, Eq. (7.35), from Eq. (7.38), we have

dp
Er þ uiϑ B = : ð7:39Þ
endr

This equation remains correct for any azimuthal velocity of ions and the same for
neutrals. In the absence of special forces that could produce rotation of neutral gas
and plasma in the azimuthal direction, these velocities should turn to zero due to
azimuthal viscosity forces. As a result, a radial electric field is established

dp
Er = , ð7:40Þ
endr

which is directed from the periphery to the core. This field is much larger than the
field Eq. (7.13) for slightly ionized plasma if pN > > pi. The large radial electric field
→ →
Eq. (7.40) causes E × B drift of electrons in the azimuthal direction, which creates
an azimuthal current while ions are at rest. The account of electron–neutral friction
leads to ambipolar classical diffusion of electrons and ions, as described above in
Sect. 7.1.
In the general two-dimensional (2D) case, the velocities of neutrals are obtained
from the momentum balance for neutral particles

→ →
∇pN = - mN nN νNi u i - u N : ð7:41Þ

Ion and electron velocities obeyed equations, which are generalizations of the
equations of Chap. 1 for moving neutrals

→ → → → →
n u i - u N = - Di ∇n þ bi n E þ u N × B ,
→ → : ð7:42Þ
→ → →
n u e - u N = - De ∇n - be n E þ u N × B

The latter could be derived by shifting to the reference frame moving with the
velocity of neutral particles and using the Lorentz transformation for the electric
field.
160 7 Transport in Strongly Ionized Plasma Across a Magnetic Field

7.4 Penetration of Neutral Particles into Hot Tokamak

Plasma

Penetration of neutral particles into the hot tokamak plasma may have a diffusive
character if the mean free path of neutral particles is smaller than the typical spatial
scale of plasma density variation. In the edge plasma, neutral particles flow from the
divertor plates due to the recycling process and then move into the plasma. In the
collisional case, their motion is controlled by charge-exchange collisions with ions.
They are also ionized by plasma electrons. Therefore, in the steady state, their
penetration is described by the equation:
→
∇  nN u N = - nN νion , ð7:43Þ

where νion = nhVσiion is the ionization frequency, and the velocity of neutral
particles is given by the 2D analog of Eq. (7.36),

→ 1
uN= - ∇pN : ð7:44Þ
nmN hVσiNi

Substituting Eq. (7.44) into Eq. (7.43), one obtains an estimate for the penetration
depth of neutral particles into the plasma:

1 m1=2
Lpen = - p e
: ð7:45Þ
n σNi σion m1=2
i

Here is taken into account the fact that a velocity in ionization frequency is the
electron thermal velocity, while the charge-exchange frequency is determined by the
ion thermal velocity.

Fig. 7.3 Radial particle flux as a function of tokamak major radius in ASDEX-Upgrade tokamak
7.4 Penetration of Neutral Particles into Hot Tokamak Plasma 161

In large tokamak reactors, the penetration depth is smaller than the distance from
divertor plates to the separatrix, so neutral particles practically do not penetrate to the
core. In contrast, for smaller tokamaks, the penetration depth can be comparable to
the distance to the separatrix, and neutral particles can cross the separatrix, providing
an ionization source nNνion inside it. Due to this ionization source, the flux of ionized
particles from the tokamak core rises towards the separatrix, leading to a strong
impact on the radial density profile. To obtain quantitative results of this impact, one
must run 2D transport code outside the separatrix, such as SOLPS-ITER, and 1D
code inside the separatrix. An example of the corresponding particle flux in the core
plasma obtained with such a combined simulation is presented in Fig. 7.3 for the
ASDEX-Upgrade tokamak.
Chapter 8
Drift Waves and Turbulent Transport

8.1 Drift Waves in Inhomogeneous Plasma

In inhomogeneous magnetized plasma, there are special waves generated by the

gradient of plasma parameters that can propagate across the magnetic field. Consider
fully ionized inhomogeneous plasma with a density gradient parallel to the х-axis,
n0(x), Fig. 8.1. The homogeneous magnetic field is perpendicular to the density
gradient and is parallel to the z-axis. The temperatures of electrons and ions are
assumed to be constant Te, i = const(x). We consider density perturbation, which is
periodic in y and z:

n1 = AðxÞ exp - iωt þ ik y y þ ik z z : ð8:1Þ

The problem will be considered in the local approximation, and we shall assume that
the density scale of unperturbed plasma in the x-direction satisfies the condition
kyLx ≫ 1. The spatial scale for perturbation in the x-direction is also large, so if one
performs Fourier transformation for x, then for typical wave vectors, a condition
should be satisfied: kx ≪ ky. On the other hand, the spatial scale of perturbation
should not exceed that of the unperturbed plasma: kx- 1 < Lx . Along the magnetic
field, the wavelength is supposed to be large, so that kz ≪ ky. Wavelengths in the y-
and z-directions should be larger than the ion Larmor radius and mean-free path,
respectively, so that the fluid description should be applicable.
Parallel momentum balance for electrons in the first approximation, when elec-
tron inertia and friction with ions are neglected, is reduced to balance between the
electron pressure gradient and the electric force. For a constant electron temperature,
this corresponds to the Boltzmann distribution for electrons

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 163
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_8
164 8 Drift Waves and Turbulent Transport

Fig. 8.1 Isodensities in the drift wave

Te n þ n1 T n1
φ1 = ln 0 = e ð8:2Þ
e n0 e n0

A positive potential perturbation corresponds to a positive density perturbation,

while a negative density perturbation is biased negatively (Fig. 8.1). The ion pressure
gradient and the electric field cause ion motion along the magnetic field. However,
let us assume that the longitudinal shift of ions during time scale ω-1 is significantly
smaller than k z- 1 ; the corresponding criterion is presented below. Therefore, ion
motion along the magnetic field is neglected.
Perpendicular velocities of ions and electrons are the sum of contributions from
→ →
E × B and diamagnetic drifts provided inertia terms are neglected

→1 →
E ×B
→1 T e ½∇n1 × B
u e⊥ = 2
þ ,
B en0 B2
ð8:3Þ
→1 →
E ×B
→1 T i ½∇n1 × B
u i⊥ = 2
- :
B en0 B2

Since diamagnetic fluxes in the homogeneous magnetic field are divergence free
→ →
(their divergence is zero), only E × B drifts contribute to particle balance equations.
Their velocities are

→ →
→ → →
E×B
u ed = u id = u d = : ð8:4Þ
B2

Drifts caused by the electrostatic electric field in the homogeneous magnetic field are
incompressible
8.1 Drift Waves in Inhomogeneous Plasma 165

→ →
E×B
∇⊥  = 0: ð8:5Þ
B2

This statement can be easily checked directly for the electrostatic electric field.
Considering this fact, the particle balance equation is given by

∂n →
þ u d ∇⊥ n = 0: ð8:6Þ
∂t

After linearization, one finds

dn0
- iωn1 þ u1dx = 0: ð8:7Þ
dx

According to Eq. (8.4), u1dx = E 1y =B = - ik y φ1 =B. Substituting potential Eq. (8.2)

into Eq. (8.7), we obtain the dispersion relation

T e d ln n0
ω = ωd = - k y : ð8:8Þ
eB dx

The frequency ωd is known as the drift frequency, and the wave is called the drift
wave.
Since the obtained frequency is a real number, the density perturbation oscillates
with drift frequency and propagates along the y-axis with a phase velocity

ωd T d ln n0
V py = =- e : ð8:9Þ
ky eB dx

The latter coincides with the diamagnetic velocity of electrons. The phase velocity in
the z-direction is Vpz = ωd/kz. Particles in the drift wave oscillate in the x-direction
due to the drifts caused by Ey, which on the inhomogeneous background n0(x) leads
to the propagation of the drift way along the y-axis. The frequency of the drift wave
is real because the electric field E 1y and particle drift u1dx = E 1y =B are shifted in phase
by π/2 with respect to the density perturbation. Hence, density changes in the knots,
which reveals density propagation along the y-axis.
The above derivation is valid provided the following inequality is satisfied:

Te þ Ti ω Te
<< d << : ð8:10Þ
mi kz me

The left part is required to neglect the parallel flux of ions in the particle balance
equation, so the parallel phase velocity should exceed the ion velocity. The upper
limit for the parallel phase velocity is the electron velocity. This condition is required
166 8 Drift Waves and Turbulent Transport

to neglect parallel electron inertia and is necessary for establishing the Boltzmann
distribution, which is assumed. Substituting the drift frequency into the left part of
inequality (8.10), we obtain

kz < < ky ρci =Lx : ð8:11Þ

Since the ion Larmor radius ρci ≪ Lx, this inequality could be satisfied only for rather
→ →
small parallel wave vectors kz ≪ ky. For kz > kyρci/Lx, perpendicular E × B drift
becomes unimportant with respect to parallel ion motion, and the drift wave trans-
forms into another type of wave-ion sonic wave, which will be analyzed later. One
should also keep in mind that if the so-called Alfven velocity cA = (B2/μ0n0mi)1/2 is
smaller than the electron thermal velocity, the right part of inequality (8.10) should
be replaced by a condition ωd/kz ≪ cA. In the opposite case, the drift wave transforms
into an Alfven wave, which will also be analyzed later.
During derivation, it was supposed that an unperturbed electric field in the x-
direction is absent. In reality, such an electric field does exist, and in plasma with
constant temperatures, it is given by Eq. (7.13) (in slab geometry Ex = (Ti/e)d ln n0/
dx). This field causes drifts of both electrons and ions along the y-axis with velocity
V0 = - Ex/B. The frequency Eq. (8.8) is the drift frequency in the reference frame of
the plasma, and, hence, with an account of plasma drift in the laboratory reference
frame

ω = ωd þ k y V 0 , ð8:12Þ

i.e., Doppler shift arises. The phase velocity Eq. (8.9) is also calculated with respect
to the moving plasma, in the laboratory frame phase velocity is

ωd ðT þ T i Þ d ln n0
V py = þ V0 = - e : ð8:13Þ
ky eB dx

The phase velocity of the drift wave in the moving plasma is still directed towards
the diamagnetic drift of electrons.

8.2 Drift-Dissipative Instability

Since the wave frequency obtained in the previous section is a real number, it is
necessary to take into account small terms, which were neglected above, to under-
stand where the drift ways are unstable or decay. In addition to the previous
consideration, we shall take into account the friction between electrons and ions
and the inertia of ions. We shall again consider a reference frame moving with the
velocity V0 where the ambient electric field is absent.
8.2 Drift-Dissipative Instability 167

The electron parallel velocity is determined by the parallel momentum balance for
electrons (parallel ion velocity could still be neglected if condition Eq. (8.10) is
satisfied)

∂n
- Te - enE z - 0:51nme νei uez = 0: ð8:14Þ
∂z

After linearization

- ik z T e n1 þ ik z en0 φ1 - 0:51n0 me νei u1ez = 0: ð8:15Þ

Thus, the parallel electron velocity is

ik z T e n1 eφ1
u1ez = - - : ð8:16Þ
0:51me νei n0 Te
→ →
The component of the electron E × B drift velocity in the x-direction is still
the same:

ik y φ1
u1edx = - : ð8:17Þ
B

The linearized particle balance equation for electrons with an account of parallel
velocity has the form

dn0
- iωn1 þ ik z n0 u1ez þ u1edx = 0: ð8:18Þ
dx

After substituting expressions for the velocities Eqs. (8.16) and (8.17), one obtains

k2z T e n1 k2z T e eφ1

ωþi þ - ωd - i = 0: ð8:19Þ
0:51me νei n0 0:51me νei T e

For ions, the parallel velocity is neglected, uik = 0. The perpendicular ion velocity is
obtained from the linearized perpendicular momentum balance equation

→1
∂ui →1 →1 →
n0 m i = - T i ∇n1 þ en0 E þ en0 u i × B : ð8:20Þ
∂t

If ion inertia is neglected, which corresponds to zero r.h.s. of Eq. (8.20), the ion
perpendicular velocity according to Eq. (8.3) is the sum of the diamagnetic velocity
→ →
and E × B velocity. After substitution of Eq. (8.3) to the l.h.s. of Eq. (8.20) and
168 8 Drift Waves and Turbulent Transport

→
multiplying the resulting equation by B , one finds the ion velocity in the next
approximation:

→1 →
E ×B →1
→1 mi ∂ E T ½∇n1 × B m ∂ T i ∇n1
u i⊥ = þ 2 ∂t
- i - 2 i2 : ð8:21Þ
B2 eB en0 B 2
e B ∂t n0

Divergence of the fluxes connected with the last two terms on the r.h.s. is zero, so
only the drift velocity of ions produces a nonzero contribution,

mi ∂E y
1
ik y φ1 iωmi 1 m
u1idx = - ; u1idy = =- E = - i2 ωky φ1 : ð8:22Þ
B eB2 ∂t eB2 y eB

Inserting Eq. (8.22) into the linearized ion balance equation

dn0
- iωn1 þ u1idx þ ik y n0 uidy = 0, ð8:23Þ
dx

one finds

n1 k2y mi T e eφ1
ω þ - ωd þ ω 2 2 = 0: ð8:24Þ
n0 e B Te

Equation systems (8.19) and (8.24) have nontrivial solutions when the determinant
of the system equals zero. From this condition, we obtain the dispersion equation

k2z ωce ωci k2z T e k2z ωce ωci

ω2 þ iω þ - iωd = 0: ð8:25Þ
k2y 0:51νei me 0:51νei k2y 0:51νei

The second term in the brackets can be neglected in the fluid approximation kyρci ≪ 1.
After introducing the definition

k 2z ωce ωci
ωs = , ð8:26Þ
k 2y 0:51νei

dispersion equation can be rewritten:

ω2 þ iωωs - iωd ωs = 0: ð8:27Þ

Let us analyze two limiting cases ωs ≫ ωd and ωs ≪ ωd. In the first case, the roots of
the dispersion equation are
8.3 Universal Instability 169

ω1 = ωd þ iω2d =ωs ; ω2 = - iωs : ð8:28Þ

In the opposite case of small ωs

ω1,2 = ± ð1 þ iÞðωd ωs =2Þ1=2 : ð8:29Þ

In both cases, the first root corresponds to instability with the increment

γ = ω2d =ωs ; ωs > > ω d ,

1=2
ð8:30Þ
γ = ðωd ωs =2Þ ; ωs < < ωd :

In the most interesting low collisionality case when ωs ≫ ωd, the increment of the
instability is significantly smaller than the real part of frequency ωd. When the
collision frequency is reduced to the low limit so that the electron mean-free pass
becomes comparable with the wavelength along the magnetic field kz(Te/me)1/2/
νei~1, the ratio of the imaginary to the real part is very small:

ωd =ωs  ωd = kz ðT e =me Þ1=2 k2y ρ2ci < < 1:

Therefore, we demonstrated that with the account of collisions, the drift wave
considered in the previous section becomes unstable with the increment given by
Eq. (8.30). The physical mechanism of wave growth consists of the following. In the
absence of a friction force, the potential perturbation has the same phase as the
density perturbation, while the electric field and particle drift along x are shifted by π/
2 with respect to the density perturbation. Hence, the density rises in the knots of the
wave, and the frequency remains real. If friction is taken into account, a shift
between density and potential perturbation arises, and the phase shift between the
perturbed electric field and density perturbations differs from π/2. As a result, in the
regions where perturbation is maximal, density rise takes place due to perpendicular
particle drift. The drift for the inhomogeneous density profile reveals itself as
instability. It is also worth mentioning that in the laboratory reference frame, the
Doppler shift kyV0 associated with plasma motion caused by the ambient electric
field is added to the real part of the frequency.

8.3 Universal Instability

8.3.1 Fluid Ions

In low collisional plasma with the mean-free path exceeding k z- 1 , the fluid approach
considered in the previous section is not applicable. Nevertheless, in such low
collisional plasma drift, waves are also unstable. Here, Landau damping plays a
170 8 Drift Waves and Turbulent Transport

role of friction force, and the corresponding instability is known as universal

instability. Ions are still assumed to be fluid kyρci < 1. We shall use for analysis
the drift kinetic equation for electrons derived in Chap. 1. In the homogeneous
→ →
magnetic field, guiding centers move across the magnetic field due to E × B drift
and along the magnetic field with parallel velocity. Therefore, the drift kinetic
Eq. (1.112) for electrons is reduced to

→ →
∂f ∂f E×B eE ∂f
þ Vz þ ∇⊥ f - z = St : ð8:31Þ
∂t ∂z B 2 me ∂V z

One has to keep the collisional integral on the r.h.s. this equation despite collisions
being rare.
We seek a solution in the form f = f0 + f 1, where f0 is an unperturbed distribution
function that corresponds to the unperturbed density profile n0(x), and f 1~ exp (-
iωt + ikyy + ikzz) is a small correction. After linearizing the electron drift kinetic
equation, one obtains

E 1y ∂f 0 eE1z ∂f 0
- iωf 1 þ ik z V z f 1 þ - = St 1 : ð8:32Þ
B ∂x me ∂V z

→1 →
Here, E = - i k φ1 is the wave electric field. For the collision integral, the simplest
model representation may be chosen: St1 = - νf 1. The real collisional integral is
given by the Landau form in Eq. (1.36), but as demonstrated below, the result is
insensitive to the collisional frequency value since ν → 0, so one can use a simplified
form of the collision integral. From Eq. (8.32), correction to the distribution function
can be found:

φ1 ky ∂f 0 ek z ∂f 0
f1 = - þ : ð8:33Þ
ω - k z V z þ iν B ∂x me ∂V z

The distribution function has real and imaginary parts. The multiplayer in Eq. (8.33)
can be rewritten after multiplying both the nominator and the denominator by
complex conjugate values:

1 Ω - kz V z iðν þ γÞ
= - :
ω - kz V z þ iν ðΩ - k z V z Þ2 þ ðγ þ νÞ2 ðΩ - k z V z Þ2 þ ðν þ γÞ2

Here, ω = Ω + iγ. Because the universal instability increment γ is much smaller than
the real part Ω, the resonance corresponds to Ω = kzVz. Two summands demonstrate
different behaviors in the vicinity of this resonance point Ω = kzVz. The first real term
turns to zero in the resonance point, and for ν → 0 and γ → 0 is antisymmetric with
respect to Vz = Ω/kz. The second term for ν → 0 and γ → 0 is small everywhere
except the resonance point, where it goes to infinity. These features can be used for
8.3 Universal Instability 171

the calculation of macroscopic plasma parameters. The real part of the density
perturbation for ν → 0 and γ → 0 equals the principal-value integral

φ1 k y ∂f 0 ek z ∂f 0 →
Re n1 = V:P: - þ dV z d V ⊥ : ð8:34Þ
Ω - kz V z B ∂x me ∂V z

Since the frequency real part is of the order of the drift frequency, we can, in
accordance with Eq. (8.10), consider Ω ≪ |kzVz| and Ω - kzVz~kzVz. Hence,

φ1 k y ∂f 0 e ∂f 0 →
Re n1 = V:P: - dV z d V ⊥ : ð8:35Þ
V z k z B ∂x me ∂V z

Choosing the Maxwellian distribution function for f0, we find that the first term on
the r.h.s. of Eq. (8.35) becomes zero since it is an even function of Vz. In the second
term ∂f0/∂Vz = - meVzf0/Te, singularity Vz disappears, and the integral is easily
calculated:

eφ1
Re n1 = n : ð8:36Þ
Te 0

Therefore, the real part of the density perturbation corresponds to the Boltzmann
distribution for electrons.
While calculating the imaginary part, we shall take into account that the integrand
has a sharp maximum in the vicinity of the resonance point. Therefore, all functions
except the resonant multiplier can be taken at Vz = Ω/kz:

νþγ k y ∂f 0 ekz ∂f 0 →
Im n1 = - φ1 - þ dV z d V ⊥ :
2
ðΩ - k z V z Þ þ ðν þ γÞ 2 B ∂x me ∂V z V z = Ω=k z

ð8:37Þ

Since

þ1
νþγ π
dV z = ,
2
ð Ω - k z V z Þ þ ðγ þ ν Þ 2 j k zj
-1

the imaginary part is

p
eφ1 π me
Im n1 = p ð ω - ωd Þ n 0 : ð8:38Þ
Te 2 jk z j T e

The resonant multiplier in Eq. (8.33) for ν → 0 and γ → 0 is convenient to write in

the form
172 8 Drift Waves and Turbulent Transport

1 1
=P - iπδðω - kz V z Þ: ð8:39Þ
ω - kz V z ω - kz V z

This implies that integrals with such multipliers should be calculated in the
following way: with the first term as the principal value and with the second term
as the integral with δ-function.
Finally, combining Eqs. (8.36) and (8.38), we have
p
eφ1 π me
n1 = 1þi p ð ω - ωd Þ n0 : ð8:40Þ
Te 2 jk z j T e

This result is independent of the collision frequency value, which justifies the choice
of the model collision integral. Collisional imaginary correction to the wave fre-
quency in Eq. (8.40) is caused by friction of the resonant particles with the wave
known as Landau damping.
To derive the dispersion equation, one must also analyze the continuity equation
for ions, which connects the ion density and potential perturbations. Since parallel
ion motion can be neglected according to inequality (8.10), which is assumed to be
satisfied, the equation for ions coincides with Eq. (8.24) obtained in the previous
section in the fluid approximation. Hence, the equation system is a pair of Eqs. (8.24)
and (8.40):

n1 k2y mi T e eφ1
ω þ - ωd þ ω = 0:
n0 e2 B2 Te
p
eφ1 π me
n1 - 1þi p ð ω - ωd Þ n 0 = 0 :
Te 2 jk z j T e

The determinant of the system should be zero, which gives us the dispersion relation
(for increments much smaller than the real frequency)
p
ωd γ πωd k2y mi T e
Ω= ; = < < 1: ð8:41Þ
k2y mi T e ωd jk z j
2
2T e =me e2 B2 1 þ ky m2 i T2 e
1þ e2 B2 e B

The increment of the universal instability reaches its maximum at kyρci~1 and is
equal to γ=ωd  ωd = kz 2T e =me < < 1. The increment strongly decreases for
smaller ky.
In summary, in plasma with rare collisions, the drift wave is unstable with a small
increment compared to the drift frequency. At the edge of applicability for kzλei~1,
increment Eq. (8.41) turns into the first Eq. (8.30).
8.3 Universal Instability 173

8.3.2 Kinetic Ions

This instability also exists in a more general situation for kyρci ≥ 1. When the
perpendicular wavelength is comparable to or smaller than the ion Larmor radius,
one must use a kinetic equation to calculate the ion density perturbation. The
linearized kinetic equation for ions has the form

∂f 1 ∂f 1 e → → ∂f 1 e → 1 ∂f 0
þV → þ V×B → = - m E → : ð8:42Þ
∂t ∂r m ∂V ∂V

Here, f 1 is a correction to the main ion distribution function f0. Note that the small
→ → →
second-order term me E ∂f→ is omitted and me V × B ∂f→0 = 0. This equation can be
1

∂V ∂V
integrated along unperturbed particle trajectories. Let at the moment t0 particles have
→ →
coordinate r 0 and velocity V 0 . Then,

τ
→ → → → → →
r r 0, V 0, τ = r 0 þ V 0 r 0 , V 0, τ0 dτ0 ,
t0
τ ð8:43Þ
→ → → → e → → → →
V r 0, V 0τ = V 0 þ V r 0 , V 0 τ0 × B dτ0 :
m
t0

Since

→ →
∂f ∂f ∂f ∂r ∂f ∂V
→ → = → → þ → → → → þ → jt,r → → ð8:44Þ
∂τ r 0, V 0 ∂t r ,V
∂r t, V ∂τ r 0, V 0
∂V ∂τ r 0, V 0

and

→ →
∂r → ∂V e → →
→ → = V, → → = V×B , ð8:45Þ
∂τ r 0, V 0 ∂τ r 0, V 0 m

the kinetic Eq. (8.42) is reduced to

∂f 1 e → 1 ∂f 0
→ → =- E → : ð8:46Þ
∂τ r 0, V 0 m ∂V

Correction to the distribution function is connected with the perturbed electric field;
→1
in the absence of E , the distribution function is conserved along trajectories. Let us
assume that perturbation is switched on at t = - 1 from the zero level and then rises
over time due to instability. Then, according to Eq. (8.46),
174 8 Drift Waves and Turbulent Transport

t
e →1 ∂f 0
f =-
1
E → dt 0 : ð8:47Þ
m ∂V
-1

The unperturbed distribution function should be a function of motion integrals;

→
hence, f 0 V = F ε⊥ = mV ⊥ 2 =2, V z , so

∂f 0 → ∂F → ∂F
→ = mV ⊥ þ ez : ð8:48Þ
∂V ∂ε⊥ ∂V z

→→
Potential perturbation φ1 is proportional to exp - iωt þ i k r . Then, from
Eq. (8.47), one obtains

eφ1 ∂F ∂F ∂F
f1 = m - ðω - kz V z Þm þ kz Þ I, ð8:49Þ
m ∂ε⊥ ∂ε⊥ ∂V z

where

1 τ
→ → → →
0
I = -i exp iωt - i k V r 0 , V 0 , τ0 dτ0 dt 0 : ð8:50Þ
0 τ - t0

Here, integration by parts is used with account of the relation

dφ1 →
= V ∇φ1 - iðω - kz V z Þφ1 : ð8:51Þ
dt

In a uniform magnetic field, the ion velocity is

V x ðt Þ = V ⊥ cos½α0 - ωc ðt - t 0 Þ,
V y ðt Þ = V ⊥ sin½α0 - ωc ðt - t 0 Þ, ð8:52Þ
V z = const, α0 = arctg V y ðt 0 Þ=V x ðt 0 Þ :

Integral in the exponent in Eq. (8.50) is easily calculated:

τ
→ → → →
k V r 0 , V 0 , τ0 dτ0 = k z V z t 0-ðζ sinðα0 - ωc τ - ΨÞ-sin½α0 -ωc ðτ - t 0 Þ-ΨÞÞ,
τ - t0
ð8:53Þ
8.3 Universal Instability 175

where

ζ = k ⊥ V ⊥ =ωc , Ψ = arctg k y =k x :

Part of the exponent in Eq. (8.50) can be expanded using the generating function for
the Bessel functions

exp½ - iζ sinðα0 - ωc ðτ - t 0 Þ - ΨÞ =

n=1
ð8:54Þ
ζn J n exp½ - in½α0 - ωc ðτ - t 0 Þ - Ψ:
n= -1

After integration over t′, one finds

n=1
I = exp½iðk⊥ V ⊥ =ωc Þ sinðα - ΨÞ ζn J n exp½ - in½α - Ψ: ð8:55Þ
n= -1

Here, α = arctgVy/Vx, ζn = (ω - nωc - kzVz)-1.

Now, integrating correction to distribution functions Eqs. (8.49) and (8.55) over
velocities both for electrons and ions for Maxwellian unperturbed distribution
functions, from the quasineutrality condition n1e = n1i = n1 , one obtains the dispersion
relation

p ω
2þi π W ðjkz jV T e ÞI 0 ðk⊥ ρce Þ2 exp - ðk⊥ ρce Þ2
jkz jV T e
p ω ð8:56Þ
þi π W ðjk z jV T i ÞI 0 ðk⊥ ρci Þ2 exp - ðk⊥ ρci Þ2 = 0
jkz jV T i

Here, ω << ωci is assumed, V T α = ð2T α =mα Þ1=2 , ρcα = T α =mα =ωcα , and I0 is a
Bessel function of the imaginary argument,

x
2i
W ðxÞ = exp - x 2
1þp exp t 2 dt:
π
0

The general dispersion relation is valid for arbitrary k⊥ρci (k⊥ ≈ ky). In the fluid limit
k⊥ρci << 1, general Eq. (8.56) is reduced to Eq. (8.41). In the kinetic limit k⊥ρci >> 1
from Eq. (8.41), one obtains

ω p ωd
2þ p d -i π =0 ð8:57Þ
ω 2πk⊥ ρci jkz jV T e
176 8 Drift Waves and Turbulent Transport

Maximal increment is reached at

p
π ωd
kz = , ð8:58Þ
2 V Te

and

V
γmax = Ω = pti , ð8:59Þ
8 πL

where L = |d ln n/dx|-1 is the density spatial scale length. Therefore, in the kinetic
limit, the frequency and the increment become independent of the wave vector.

8.4 Instabilities Caused by the Temperature Gradient

The instabilities considered in the previous section have increments significantly

smaller than the drift frequency. More dangerous are instabilities caused by the
temperature gradients. Therefore, let us consider the instability caused by the ion
temperature gradient, known as ηi instability, which is to a large extent responsible
for turbulent transport in fusion devices.
For simplicity, we shall consider a situation with only the ion temperature
gradient, while the density gradient and the electron temperature gradient are absent:
dT 0i =dx ≠ 0, dn0 =dx = dT 0e =dx = 0. Let us demonstrate that in this situation, ηi
instability exists with a frequency in the range

< < jωj < < ky e - 1 B - 1 jdT i =dxj:

1=2
kz ðT i þ T e Þ1=2 =mi

We shall analyze it in the fluid approximation kyρci ≪ 1 in the collisionless plasma.

In this type of wave, perpendicular particle drift caused by a perturbed electric
field at an inhomogeneous temperature background leads to ion temperature pertur-
bations. Parallel pressure gradient caused by ion temperature perturbations drives
ions along the magnetic field. Their parallel motion leads to the density perturbation
and, due to the Boltzmann distribution of electrons, causes potential perturbation.
The equation for ion temperature we shall take in the form of Eq. (2.12), where
dissipative terms with heat conductivity, viscosity, and heat exchange are neglected
(the corresponding time scales are supposed to be large with respect to ω-1):

3 ∂ → →
n þ u i∇ T i þ nT i ∇  u i = 0: ð8:60Þ
2 ∂t

The linearized particle, momentum, and ion heat balance equations are
8.4 Instabilities Caused by the Temperature Gradient 177

- iωn1 þ ik z n0 u1iz = 0,
- iωn0 mi u1iz = - ik z n0 T 1i þ n1 T 0i - ik z en0 φ1 , ð8:61Þ
cφ dT 0i
1
- iωn0 T 1i - ik y n0 þ ik z n0 T 0i u1iz = 0:
B dx

With an account of Boltzmann distribution for electrons

n1 eφ1
= 0 ð8:62Þ
n0 Te

1=2
one obtains a closed equation system. Assuming k z ðT i þ T e Þ1=2 =mi < < jωj, from
the condition of its zero determinant, we find the dispersion relation

T e 1 d ln T i
ω3 = - k2z k : ð8:63Þ
mi y eB dx

This equation has one real root and two imaginary roots, and one of the imaginary
roots corresponds to the increment
p 1=3
3 2 T e 1 d ln T i
γ= kz ky : ð8:64Þ
2 mi eB dx

More detailed analysis for plasma with a density gradient gives the following
criterion for ηi instability (for the same sign of the density gradient as the ion
temperature one):

d ln T i
ηi = > 2: ð8:65Þ
d ln n

In the kinetic limit k⊥ρci ≫ 1, a similar ηi instability also exists with the increment of
the order of that given by Eq. (8.59).
A similar instability caused by an electron temperature gradient is known as ηe
instability. In contrast to ηi instability, it is short-scaled, kyρci ≫ 1, and high
frequency, jωj > > kz ðT i þ T e Þ1=2 =m1=2
e . Ions in this wave have a Boltzmann distri-
bution since their Larmor radius exceeds the perpendicular wavelength:

n1 eφ1
=- 0 ð8:66Þ
n0 Ti

For electrons, however, kyρce ≪ 1, and their perpendicular motion is just drift in the
perturbed electric field. The linearized equation system for electrons coincides with
that for ions Eq. (8.43) with the replacement of subscripts «i» with «е». The
increment of ηe instability is analogously
178 8 Drift Waves and Turbulent Transport

p 1=3
3 2 Ti 1 d ln T e
γ= kz k : ð8:67Þ
2 me y eB dx

Similar to ions, the criterion of instability is

ηe > 2: ð8:68Þ

8.5 Turbulent Transport Caused by Random Electric

Fields

As a result of the nonlinear evolution of drift-type instability with various wave

vectors, plasma transforms into a turbulent state with turbulent transport of particles,
energy, and momentum across the magnetic field. We shall consider this process by
taking turbulent diffusion produced by universal instability as an example. The
increment of this instability in the fluid limit is much smaller than the real part of
the frequency, which is close to the drift velocity. In such situations, it is possible to
use the so-called quasilinear approximation. In plasma, many waves with different
wave vectors are excited, but due to a small increment, the main impact is produced
on the average distribution function. The evolution of the latter leads to saturation of
the instabilities at a small level before the interaction of different waves becomes
important.
As in the linear case, we consider the kinetic equation for electrons in the drift
approximation

→ →
∂f ∂f E×B eE ∂f
þ Vz þ ∇⊥ f - z = St: ð8:69Þ
∂t ∂z B2 me ∂V z

The distribution function is sought in the form

→→
f = f 0 þ f 1; f 1 = f *k 0 exp - iω →k t þ i k r
→
k
→→
= f *k exp - iΩ →k t þ i k r , ð8:70Þ
→
k

where amplitudes f →k are slow functions of time, varying with a time scale smaller
than the time scale ω →- 1 due to a small increment. The function f0 is averaged over a
k
8.5 Turbulent Transport Caused by Random Electric Fields 179

time scale larger than ω →- 1 and over a spatial scale larger than the wavelengths. For
k
potential, analogously

→→ →→
φ1 = φ*k 0 exp - iω →k t þ i k r = φ*k exp - iΩ →k t þ i k r : ð8:71Þ
→ →
k k

Summation in Eqs. (8.70) and (8.71) includes positive and negative values of wave
→ →
→ E×B
vectors k . Since the potential is real, φ - →k = φ→ . Nonlinear terms B2
∇⊥ f and
k
eE z ∂f
me ∂V zare neglected in the linearized equation; in other words, nonlinear interaction
between different modes is neglected due to their small amplitudes.
In the linear approximation, similar to Eqs. (8.32) and (8.33),

φ →k ky ∂f 0 ek z ∂f 0
f →k = - þ : ð8:72Þ
Ω →k - k z V z þ iγ →k þ iν B ∂x me ∂V z

The equation for the slowly changing distribution function f0 can be found from the
averaged Eq. (8.69):

→1 →
E ×B
∂f 0 eE 1z ∂f 1
þ ∇⊥ f 1 - = hSt i: ð8:73Þ
∂t B2 me ∂V z

While averaging in Eq. (8.73) the sums of the type

→ →0
A →k B → 0 exp i k r - iΩ →k t þ γ →k t exp i k r - iΩ → 0 t þ γ → 0 t one must take
→ →0 k k k
k k
→0 →
into account that only terms with k = - k remain finite, while other terms vanish.
Additionally, according to Eq. (8.41), Ω →k = - Ω - →k and γ →k = γ - →k . Hence,

k ω ∂ ky φ →k
2 2
∂f 0 ∂ ∂ kω ∂
- - z ce πδ Ω →k - k z V z - z ce f
∂t → ∂x ky ∂V z B2 ∂x k y ∂V z 0
k
= hSt i:
ð8:74Þ

This equation describes the evolution of the averaged distribution function.

Potential perturbations grow with nonlinear increment γ →k , which depends on the
distribution function f0. The nonlinear increment differs from the linear increment in
Eq. (8.41), calculated for the Maxwellian distribution function. Keeping the arbitrary
averaged distribution function in the equations of Sect. 8.3, one obtains
180 8 Drift Waves and Turbulent Transport

Ω →k πT e ky ∂f 0 ek z ∂f 0 →
γ →k ðt Þ = - þ δ Ω →k - k z V z dV z d V ⊥ : ð8:75Þ
en0 B ∂x me ∂V z

The average distribution function f0 is controlled by the balance of the diffusion in

the velocity space for resonant Particles and the collisional integral. Diffusion in the
velocity space makes the distribution function more gradual in an attempt to form a
plateau, while collisions lead to its Maxwellization. Solving Eq. (8.75), it is possible
to find f0 and, hence, nonlinear increment γ →k . Integration of Eq. (8.75) over
velocities with account of Eq. (8.74) leads to the diffusion equation

∂n0 ∂ ∂n0
- D =0 ð8:76Þ
∂t ∂x ∂x

With the diffusion coefficient

k2y φ2→ γ →k
D= k
: ð8:77Þ
→ B 2
Ω2→
k k

Ion contributions to diffusion produce nonresonant particles. The simplest way to

→ →
calculate ion turbulent flux is to average E × B drifts in the direction of the density
gradient:

cE y ∂n
Γx = n = -D 0: ð8:78Þ
B ∂x

During averaging, density perturbations shifted in phase with respect to potential

perturbations contributing to the flux. Density perturbations can be found using the
drift kinetic equation for ions, similar to Eq. (8.72). From Eq. (8.78), it follows that
the ion diffusion coefficient in Eq. (8.78) coincides with that for electrons Eq. (8.77).
For other types of instabilities with larger increments, the quasilinear approach is
not applicable since nonlinear interaction between different modes becomes impor-
tant. The latter determines the wave spectrum. Nevertheless, the quasilinear diffu-
sion coefficient can be used for rough upper estimates of transport coefficients. Let
us assume that modes with the largest wavelengths give the main contribution to the
transport and keep only one main term in the sum Eq. (8.77). The amplitude of
density perturbations can be estimated as k y n →k  dn0 =dx. At such a level, the
density gradient of the perturbation becomes comparable to the average density
gradient, and large-scale perturbations are the source of smaller-scale instability.
Energy is transported from the larger scales to the smaller scales, while the main
large-scale mode saturates. Substituting this estimate into (8.77), keeping only one
term in the sum, with account of n →k =n0 ≈ eφ →k =T e and Ω →k  ωd , one finds an
estimate for the diffusion coefficient
8.6 Effect of Magnetic Shear on Plasma Instabilities 181

D  k y- 2 γ →k : ð8:79Þ

The same estimate can be obtained assuming that the increment for the main large-
scale mode γ →k is balanced by damping produced by turbulent diffusion Dk 2y , which
is caused by the whole spectrum of oscillation.
The largest possible diffusion coefficient for drift waves is reached for the
wavelength of the order of density scale: kyL~1, where L = |d ln n0/dx|-1 and
increment of the order of drift frequency. Hence, for γ →k  Ω →k  ωd , we have

Te
D  DB = : ð8:80Þ
eB

This diffusion coefficient is known as the Bohm diffusion coefficient and is an upper
estimate for drift turbulence. In general, to obtain the turbulent diffusive flux,
knowledge of the turbulent spectrum is required to account for the nonlinear
interaction between different modes.

8.6 Effect of Magnetic Shear on Plasma Instabilities

Magnetic shear (variation of magnetic field in space) generally suppresses plasma

instabilities. Let us analyze its impact using drift instability as an example. First, we
define shear in the slab geometry. Consider a magnetic field

→ → →
B = Bz ðxÞ e z þ By ðxÞ e y : ð8:81Þ
→ → →
Such a magnetic field is produced by the current j = jy ðxÞ e y þ jz ðxÞ e x , which is
connected with the magnetic field by Maxwell equations -∂Bz/∂x = μ0jy and ∂By/
∂x = μ0jz. All perturbations in wave Ψ still have the form

Ψ = AðxÞ exp - iωt þ ik y y þ ik z z : ð8:82Þ

Since the magnetic field turns in space, from the scalar product
→*
k B = kk B = k z Bz þ k y By , one obtains the parallel wave vector

k z Bz þ ky By
k k ð xÞ = : ð8:83Þ
B

Let the parallel wavelength tends to infinity at some value x0, where kk(x0) = 0. This
is possible when wave vectors kz and ky have different signs. Expanding the value kk
in the vicinity of x0 assuming for simplicity that only By is a function of x with
By ≪ Bz and Bz/B ≈ 1, we have
182 8 Drift Waves and Turbulent Transport

d By =B
k k ð xÞ = k y ðx - x0 Þ: ð8:84Þ
dx

Let us introduce typical plasma scale length L along the x-axis, for example, L = |
d ln n0/dx|-1) and define magnetic shear as

d By =B
Θ=L : ð8:85Þ
dx

Then,

x - x0
k k ðxÞ = ky Θðx0 Þ : ð8:86Þ
L

The magnetic shear is the geometric characteristic of the magnetic field; on the other
hand, the parallel wavelength depends on magnetic shear.
In the cylindrical geometry, magnetic shear can be introduced analogously. Let
perturbation have the form

Ψ = Aðr Þ expð- iωt þ imθ þ ik z zÞ, ð8:87Þ

→ → →
B = Bz e z þ Bθ ðr Þ e θ , and only the azimuthal component of the magnetic field Bθ
depends on the radius. The azimuthal wavelength is connected with the azimuthal
wavenumber according to kθ = m/r. Correspondingly,

kz Bz þ ðm=r ÞBθ
kk ðr Þ = : ð8:88Þ
B

In the vicinity of the resonant flux surface with radius r0, where kk(r0) = 0, the
parallel wave vector is

r - r0
kk ðr Þ = kθ Θðr 0 Þ , ð8:89Þ
r0

where magnetic shear is defined as

dðBθ =rBÞ
Θ = r2 : ð8:90Þ
dr

In the general case, when both components depend on the spatial coordinate,
8.6 Effect of Magnetic Shear on Plasma Instabilities 183

x - x0
k k ðxÞ = k⊥ ðx0 ÞΘðx0 Þ ;
L
k ⊥ = k y Bz =B - kz By =B; ð8:91Þ
B2 d By =Bz
Θ = L z2
B dx

for the slab geometry, and

r - r0
kk ðxÞ = k⊥ ðr 0 ÞΘðr 0 Þ ;
r0
m
k⊥ = B =B - k z Bθ =B; ð8:92Þ
r z
B2 dðBθ =rBz Þ
Θ = r 2 z2
B dr

for the cylindrical geometry. In the toroidal geometry, the definition of magnetic
shear will be introduced later.
Magnetic shear in general leads to stabilization of the oscillations. Indeed, the
parallel wave vector strongly increases with distance from the resonant flux surface,
→
and the wavelength along B decreases. At some distance δ from the resonant flux
surface, the parallel wave vector reaches the maximal value kmax k when the wave
increment turns to zero. For a universal or ηi instability maximal value of k max
k ,
according to inequality (8.10),

Te þ Ti
kmax
k = ωd = : ð8:93Þ
mi

For larger values of the parallel wave vector, perpendicular ion drift in the drift wave
is replaced by their parallel motion, and a drift wave cannot exist. According to
Eq. (8.86)

kmax
k L
δ= : ð8:94Þ
ky Θ

The quantity δ is a region of drift wave localization in the vicinity of a resonant flux
surface. Outside of this region, the drift wave attenuates. Substituting the value of
kmax
k Eq. (8.93) and the expression for drift frequency Eq. (8.8), assuming Te~Ti, we
have

ρci
δ : ð8:95Þ
Θ

This estimate is justified for sufficiently large magnetic shear when δ ≪ L, in other
words, for Θ > ρci =L.
184 8 Drift Waves and Turbulent Transport

Therefore, for large magnetic shear, a new characteristic scale δ arises in the
direction of the initial inhomogeneity x instead of L. As a result, the local approx-
imation, discussed in the previous sections when the wave amplitude is supposed to
be a gradual function of x, is valid only for short wavelengths kyδ > 1. The maximal
wave vector in the y direction is thus determined by the condition

y δ = 1:
k min ð8:96Þ

Drift waves with larger wavelengths are not generated in plasma with large shear.
Restriction of the maximal perpendicular wavelength leads to a reduction in the
turbulent transport coefficients. Let us use Eq. (8.79) to estimate the ion heat
conductivity coefficient caused by ηi instability. Substituting k min
y Eq. (8.96) and δ
Eq. (8.95), assuming the increment to be of the order of the drift velocity, one finds

T e ρci
χi  : ð8:97Þ
eB ΘL

The ion heat conductivity coefficient is thus δ/L times smaller than the Bohm
coefficient. In contrast to the Bohm coefficient, it is inversely proportional to B2
and depends on the plasma characteristic scale L. This dependence is known as
Gyro-Bohm scaling. It is important to mention that for this type of instability, the
diffusion coefficient should be much smaller than the ion heat conductivity coeffi-
cient due to the Boltzmann distribution of electrons in the wave so that the averaged
→ →
E × B drift vanishes.
For ηe instability, the maximal value of the parallel wave vector is

Te
k max
k = ωd = : ð8:98Þ
me

Hence, the localization region for this instability is much smaller:

ρce
δ : ð8:99Þ
Θ

The corresponding estimation for electron heat conductivity is, therefore, smaller
than that for ions:

T e ρce
χe  : ð8:100Þ
eB ΘL

The strong magnetic shear local approach considered previously in this chapter is no
longer valid. Indeed, the typical scale of the wave amplitude variation in the x-
direction becomes comparable with the wavelength in the y-direction, and, hence,
the electric fields and polarization current along x and y are of the same order. Even
8.6 Effect of Magnetic Shear on Plasma Instabilities 185

for smaller perpendicular wavelengths, the divergence of the parallel ion flux
changes from the small one for kk → 0 up to values when the wave starts to dampen
at x~δ. The rigorous approach implies that the solution is sought in the form of
Eq. (8.1), but the equations are reduced to ordinary differential equations with
respect to x.
Let us demonstrate this approach for collisionless drift waves in the absence of
temperature gradients, which corresponds to universal instability in the absence of
magnetic shear. In the particle balance for ions in addition to Eq. (8.23), we shall take
into account the polarization drift in the x-direction and their parallel velocity:

ik y φ1 dn0 m m d 2 φ1
- iωn1 þ in0 k k ðxÞu1ik - - iωk 2y n0 i2 φ1 þ iωn0 i2 = 0: ð8:101Þ
B dx eB eB dx2

It is taken into account that the wave amplitude spatial scale in the x-direction is
much smaller than L. Here, kk(x) changes linearly in the vicinity of the resonance
surface (8.86). Parallel ion velocity could be found from the parallel momentum
balance equation for ions (Te ≫ Ti)

- iωn0 mi u1ik = - ik k ðxÞT e n1 : ð8:102Þ

One more relation between density and potential perturbations is obtained from
electron Eq. (8.40). Neglecting the imaginary part of Eq. (8.40), which is associated
with Landau damping, we have n1 = (eφ1/Te)n0 – Boltzmann distribution for
electrons. Combining Eqs. (8.101) and (8.102) with an account of this relation,
one obtains

2
d 2 φ1 k2y T e Θ ðx - x0 Þ2 ω
ρ2ci þ - 1 þ k2y ρ2ci - d φ1 = 0, ð8:103Þ
dx2 mi ω2 L2 ω

where the Larmor radius is defined as ρci = (Te/mi)1/2(mi/eB). In the new variables

1=2
x - x 0 ωd Θ
ξ= expðiπ=4Þ ð8:104Þ
ρci ω

Eq. (8.103) is reduced to the equation of the quantum oscillator

d 2 φ1
þ 2E - ξ2 φ1 = 0, ð8:105Þ
dx2

with energy
186 8 Drift Waves and Turbulent Transport

i ω ω
E= 1 þ k 2y ρ2ci - d : ð8:106Þ
2 ωd Θ ω

The eigenvalues for the oscillator are

1
E = n þ , n = 0, 1, 2 . . . : ð8:107Þ
2

Therefore, from Eqs. (8.106)–(8.107), one obtains the dispersion relation. When n is
not too large, the frequency is

Re ω = Ω = ωd = 1 þ k2y ρ2ci ;
ð8:108Þ
Imω = γ = - ωd Θð2n þ 1Þ:

Eigenfunctions are given by Hermite polynomials Hn:

φ1 = H n ðξÞ exp - ξ2 =2 : ð8:109Þ

In accordance with Eq. (8.90), mode slightly damps. At x → 1, eigenfunctions

oscillate

ωd Θ x 2
φ1 ðx → 1Þ = exp - i ð8:110Þ
ω 2ρ2ci

1=2
with the typical scale ρci =Θ . Due to the small imaginary correction, the frequency
eigenfunctions diverge at infinity. To make them converge, it is necessary to take
into account an additional term in the electron equation, which is responsible for
universal drift instability. The corresponding analysis leads to an ordinary differen-
tial equation that can be solved only numerically. The imaginary part of the
eigenvalues remains negative, so drift waves in the slab geometry with magnetic
shear are damped. Drift waves caused by the temperature gradient remain unstable in
the presence of magnetic shear for kyδ > 1. Note, however, that other factors, such as
an inhomogeneous magnetic field, can make drift waves unstable even in the
presence of magnetic shear.
Chapter 9
Dynamics of Fully Ionized Plasma
in the Absence of a Magnetic Field

9.1 Ion Acoustic Waves

In plasma, ion acoustic or sound waves are a type of oscillation of electrons and ions
similar to acoustic waves traveling in neutral gas. However, in contrast to neutral
gas, electrons and ions can interact with the electrostatic field that arises in plasma. In
the fluid approximation, fully ionized plasma without impurities for Z = 1 is
described by the equation system considered in Chap. 2:

∂n →
þ ∇  n u α = 0,
∂t
→
duα $ → →
mα n = - ∇pα - ∇  π α ± en E þ R α , ð9:1Þ
dt
3 dT α → ∂uαj →
n þ nα T α ∇  u α þ παjk þ ∇  q α = Qα :
2 dt ∂xk

Summing up two momentum balance equations for electrons and ions, neglecting
the electron inertia term and ion viscosity, one obtains the net momentum balance
equation, where the electric field and friction force are absent:

→
dui
nmi = - ∇p: ð9:2Þ
dt

The latter forces cancel each other due to different signs in the partial momentum
balance. Here, p = pe + pi is the total pressure. The viscosity term can be neglected
with respect to the pressure gradient provided that the mean-free path is smaller than
the typical scale L of plasma, which is the criterion of applicability of the fluid
approach.
As mentioned in Chap. 7, it is possible to estimate the typical velocity in
inhomogeneous fully ionized plasma from Eq. (9.2). This typical velocity is of the

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 187
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_9
188 9 Dynamics of Fully Ionized Plasma in the Absence of a Magnetic Field

order of ui~[(Te + T )i/mi]1/2, and its exact value depends on the problem type. We
shall first consider the evolution of small perturbations in unbounded homogeneous
plasma with density n0 in the isothermal case. Let all variables be functions on
z only. The density and ion fluid velocity are sought in the form:

n = n 0 þ n1 , ui = u1i , n1  exp ð- iωt þ ikzÞ, u1i  exp ð- iωt þ ikzÞ: ð9:3Þ

Linearizing the ion particle balance equation and net momentum balance Eq. (9.2),
one obtains

- iωn1 þ iku1i n0 = 0, - iωmi n0 u1i þ ik ðT e þ T i Þn1 = 0: ð9:4Þ

This equation system has a nontrivial solution when its determinant equals zero,
which leads to the dispersion relation

ω = ± kcs , ð9:5Þ

where for plasma with constant temperatures

Te þ Ti
cs = : ð9:6Þ
mi

Here, the quantity cs is known as the ion sound speed. A more general definition of
the ion sound speed is

dp
cs = , ð9:7Þ
dρ

where ρ = nmi. For Te, i = const, this definition coincides with Eq. (9.6). We see that
in accordance with Eq. (9.5), density perturbation propagates with constant phase
velocity cs, and sign ± corresponds to the propagation in one or opposite direction.
Dispersion, i.e., dependence of phase velocity on wavelength, is absent for the ion
acoustic wave.
According to continuity equations for electrons and ions, in the absence of plasma
current, ue = ui = u, i.e., electrons and ions move together. Hence, the electron-ion
friction force is zero, and from the electron momentum balance, we obtain the
Boltzmann distribution for electrons

Te T n1
φ1 = lnðn=n0 Þ ≈ e : ð9:8Þ
e e n0

The positive density perturbation has positive potential, while negative perturbation
is biased negatively. The electric field that corresponds to this Boltzmann potential
keeps electrons coupled to ions so that their velocities in the acoustic wave coincide.
9.1 Ion Acoustic Waves 189

If the time scale typical for ion acoustic waves ω-1 = (kcs)-1 is smaller than all
dissipative time scales – a heat conductivity time scale (k2χe, i)-1 and a heat
exchange time scale (2meνei/mi)-1 – then ion acoustic waves follow an adiabatic
law. In this situation, entropy per volume s and particle temperatures change with
density as Te, i~n2/3 and total pressure p~n5/3. In the general case system, Eqs. (9.4) is
given by

- iωn1 þ iku1i n0 = 0,
dp ð9:9Þ
- iωi mi n0 þ ik n1 = 0:
dn

The dispersion law is still given by Eq. (9.5), while the adiabatic sound speed is

dp 5ð T e þ T i Þ
cs = j = : ð9:10Þ
dρ s = const 3mi

Finally, a situation is possible when the time scale ω-1 is small with respect to the
ion heat conductivity time scale and the heat exchange time scale but is large with
respect to the electron heat conductivity time scale. Such a situation may exist due to
the high electron heat conductivity with respect to that of ions. In this case, the ion
sound speed is isothermal for electrons and adiabatic for ions, and the sound speed is

T e þ ð5=3ÞT i
cs = : ð9:11Þ
mi

Since dispersion for ion acoustic waves is absent, i.e., waves with different wave-
lengths propagate with the same velocity, and any perturbation of arbitrary shape,
which can be decomposed into Fourier harmonics, propagates with the same veloc-
ity. Hence, the general solution for linearized Eqs. (9.1) and (9.2) has the form

n1 = f 1 ðz - cs t Þ þ f 2 ðz þ cs t Þ: ð9:12Þ

This becomes quite clear if linearized equations are reduced to the form

∂n1 ∂u1
þ n0 i = 0,
∂t ∂z
ð9:13Þ
∂u1i ∂n1
n0 = - c2s :
∂t ∂z

Taking the time derivative in the first equation and the spatial derivative in the
second equation and subtracting one from another, one comes to the wave equation
190 9 Dynamics of Fully Ionized Plasma in the Absence of a Magnetic Field

Fig. 9.1 Splitting of rectangular perturbation into two moving ones of half amplitude

2 2
∂ n1 1 ∂ n1
- 2 = 0, ð9:14Þ
∂z 2 cs ∂t 2

in which the solution coincides with Eq. (9.12).

As an example, let us consider the evolution of rectangular density perturbation,
n1 = A for -a ≤ z ≤ a and n1 = 0 outside of this interval, as shown in Fig. 9.1. The
initial fluid velocity is set to zero. Such perturbation can be considered as a sum of
two perturbations of half amplitudes, which in accordance with Eq. (9.12), propagate
with the velocities ±cs. In the course of time, the initial perturbation splits into two
perturbations moving in different directions (Fig. 9.1).
Ion acoustic waves can also propagate in the collisionless plasma if the electron
temperature exceeds the ion temperature, Te > > Ti. This case can be considered in
the following way. According to the momentum balance for electrons, the electric
force is balanced by the electron pressure gradient. Therefore, the electric force
exceeds the ion pressure gradient and viscosity in the momentum balance equation
for ions, so both the ion pressure gradient and the ion viscosity can be neglected. As
9.1 Ion Acoustic Waves 191

a result, replacing electric force with electron pressure gradient in the ion momentum
balance equation, one ends up with Eq. (9.2) with only the electron pressure gradient

→
dui
nmi = - ∇pe : ð9:15Þ
dt

The sound velocity coincides with Eq. (9.6). For comparable temperatures, the
situation becomes more complicated, ion acoustic waves strongly dampen due to
collisionless Landau damping since the sound speed is of the order of ion thermal
velocity, and kinetic analysis is needed.
When the perturbation wavelength becomes comparable with the Debye radius,
special consideration is required since the quasineutrality is violated and the Poisson
equation should be solved. Let us analyze this situation for collisionless plasma with
Te > > Ti. The initial equation system in the general case of an arbitrary wavelength is

∂ni →
þ ∇  ni u i = 0,
∂t
∂ne →
þ ∇  ne u e = 0,
∂t
→
dui * ð9:16Þ
mi ni = eni E ,
dt
→
0 = - T e ∇ne - ene E ,
→
∇  E = e ðni - ne Þ=ε0 :

From the electron momentum balance equation, one obtains the Boltzmann distri-
bution for electrons φ = (Te/e) ln (ne/n0), where n0 is an unperturbed density for both
→
electrons and ions. From the ion momentum balance, we have mi ni ddtu i = - T e ∇ne .
The linearized system of particle balance for ions, momentum balance for electrons
and ions and Poisson equation in the 1D case is, therefore,

- iωn1i þ ikn0 u1i = 0,

eφ1 =T e = n1e =n0 ,
ð9:17Þ
- iωmi n0 u1i = - ikT e n1e ,
k2 φ1 = e n1i - n1e =ε0 :

After eliminating the potential and ion velocity, one obtains an equation system for
two unknowns n1e , n1i . From the condition of zero determinant, one obtains a disper-
sion relation
192 9 Dynamics of Fully Ionized Plasma in the Absence of a Magnetic Field

k2 c2s
ω2 = , ð9:18Þ
1 þ k2 r 2d

where the Debye radius is defined as r 2d = T e ε0 =ðn0 e2 Þ. One can see that in the
general case, ion acoustic waves have dispersion since the phase velocity depends on
the wavelength. For a large wavelength, Eq. (9.18) turns into Eq. (9.5), while for a
small wave length krd > > 1, the wave frequency becomes independent of the
wavelength and coincides with the ion plasma frequency:

n 0 e2
ω = ωpi = : ð9:19Þ
m i ε0

In multicomponent plasma with p different ion species, there are p roots for phase
velocities not counting the direction of propagation. For the Te > > Ti case, p-1
modes propagate with small velocities of the of the order of ion thermal velocities,
and one fast mode propagates with a velocity proportional to Te1/2. The general
expression for this fast mode can be easily obtained similarly to the above derivation,
and the sound speed for the fast mode for a large wavelength is

T e Z I mp =mI
2
cs = , ð9:20Þ
mp hZ I i

where mp is the proton mass, mI the ion mass, ZI the charge number, and

hX I i = X I nI = nk . In the practically important case for the fusion application

I k
of a deuterium–tritium mixture (half deuterium, half tritium) for Te = Ti, isothermal
electrons and adiabatic ions, two sound speeds are

1:1T e 1:8T e
c1s = , c2s = , ð9:21Þ
mp mp

9.2 Nonlinear Dynamics. Self-Similar Solutions

In this section, we shall consider nonlinear evolution in the fully ionized plasma in
the situation when the initial sizes of an inhomogeneity are unimportant. Then, the
solutions of different problems could be obtained using a self-similar approach when
plasma density and ion fluid velocity are sought as a function of one variable, which
is a combination of spatial coordinates and time. As an example, let us consider 1D
plasma expansion into a vacuum (ambient plasma of small density). At z = 0, a
9.2 Nonlinear Dynamics. Self-Similar Solutions 193

_ ðzÞ. Plasma from the source spreads in

source of ionized particles is located: I = Nδ
both the positive and negative directions of z. Near the source, a stationary density
profile is established, and the plasma flux can be obtained by integrating the particle
balance equation over the region in the source neighborhood:
2Γðz = 0Þ = Idz = N. _ Let s first consider isothermal case Te = const, Te > > Ti.
Such a situation can be realized due to the large electron heat conductivity, so that a
large parallel heat flux compensates for the work spent on plasma expansion. In the
equations (subscript for fluid velocity is dropped)

∂n ∂ðnuÞ
þ = 0;
∂t ∂z
ð9:22Þ
∂u ∂u ∂p
nmi þu =-
∂t ∂z ∂z

we introduce a new variable ζ = z/t. Both plasma density and ion fluid velocity are
supposed to be functions of ζ only: n = n(ζ), u = u(ζ). Derivatives are transformed in
the following way:

∂ ∂ζ d 1 d ∂ ∂ζ d 1 d
= =- ζ ; = = : ð9:23Þ
∂t ∂t dζ t dζ ∂z ∂z dζ t dζ

In the new variables,the equation system has the form

dn d ðnuÞ
-ζ þ = 0;
dζ dζ
ð9:24Þ
du du dn
n -ζ þ u = - c2s :
dζ dζ dζ

Here, the sound speed is defined according to Eq. (9.7). From the first equation, the
derivative du/dζ can be obtained as a function of the density derivative. After
substitution of the latter into the second equation, one obtains (the sign corresponds
to expansion into the positive direction of the z-axis)

u = cs þ ζ: ð9:25Þ

In the isothermal case, the sound speed is constant; hence, du/dζ = 1, and from the
second Eq. (9.24), we have the density

dn n
þ = 0: ð9:26Þ
dζ cs

Its solution is
194 9 Dynamics of Fully Ionized Plasma in the Absence of a Magnetic Field

ζ
n = A exp - : ð9:27Þ
cs

_
In the initial variables with an account of the condition nuðz = 0Þ = Γðz = 0Þ = N=2,
we have for z > 0

Γðz = 0Þ z z
n= exp - , u = cs þ : ð9:28Þ
cs cs t t

Hence, plasma density decreases exponentially with a spatial scale cst, while ion
velocity increases linearly with distance. The same is true for z < 0.
In the adiabatic limit, sound speed Eq. (9.7) depends on the density as n1/3 ( p~n5/3).
The fluid velocity is still given by Eq. (9.25) From the second Eq. (9.24), one finds a
general relation between u and n:

dn
u= - cs : ð9:29Þ
n

Integrating Eq. (9.29) with an account of Eq. (9.25) yields

3
z
n = n0 1 -
4cs ðn0 Þt ð9:30Þ
3z
u = cs ð n0 Þ þ :
4t

Here, n0cs(n0) = Γ(z = 0). In the adiabatic limit, the plasma density turns to zero at a
finite distance from the source, and the fluid velocity remains finite.
Another problem that can be analyzed using the obtained solution is the problem
of initial discontinuity decay. Let us consider the initial plasma density discontinuity
at t = 0

nðz ≤ 0Þ = n0 , nðz > 0Þ = 0: ð9:31Þ

Let us restrict ourselves to the isothermal case. Then, the solution can be constructed
using Eqs. (9.25)–(9.27). The solution that satisfies initial condition Eq. (9.31) is

z z
n = n0 exp - -1 , u = cs þ , z > - cs t;
cs t t ð9:32Þ
n = n0 , u = 0, z ≤ - cs t:

The density profile is shown in Fig. 9.2. The plasma from the region -cst < z < 0
moves to the right, while a wave of rarefaction propagates to the left with the sound
speed. It is interesting that the density at z = 0 remains constant and equals
9.3 Simple Nonlinear Waves. Overturn 195

Fig. 9.2 Evolution of initial density shock (dotted line). In time plasma (solid line) flows to the
right, while wave of rarefaction moves to the left

n(z = 0) = n0 exp (-1). The solution in the adiabatic case can be constructed
analogously.

9.3 Simple Nonlinear Waves. Overturn

The evolution of one-dimensional nonlinear propagating waves, which are nonlinear

analog of ion acoustic waves, is more complicated. Let us seek a partial solution of
Eq. (9.22) in the form of a propagating wave. All physical quantities in the wave
(plasma density, pressure, and fluid velocity) are supposed to be functions of z - Vt,
where a propagation velocity V, in contrast to the linear case, depends on the wave
amplitude. Since density, pressure, and fluid velocity propagate together with the
same velocity, one physical quantity can be considered a function of the other. Using
this fact, the particle balance equation can be transformed to (ρ = min):

∂ρ dðρuÞ ∂ρ
þ = 0: ð9:33Þ
∂t dρ ∂z

Here, the product ρu is considered to be a function of ρ. In the momentum balance

equation

∂u ∂u 1 ∂p
þu þ =0 ð9:34Þ
∂t ∂z ρ ∂z

pressure can be considered as a function of u : ∂p/∂z = (dp/du)∂u/∂z. Then, the

equation for velocity is reduced to
196 9 Dynamics of Fully Ionized Plasma in the Absence of a Magnetic Field

∂u 1 dp ∂u
þ uþ = 0: ð9:35Þ
∂t ρ du ∂z

dðρuÞ
The quantities V ðρÞ = dρ in Eq. (9.33) and V ðuÞ = u þ 1ρ dp
du in Eq. (9.35)
represent propagation velocities for density and fluid velocity correspondingly.
Since these quantities propagate together, two velocities must coincide
V(ρ) = V(u), so

dðρuÞ 1 dp
= uþ : ð9:36Þ
dρ ρ du

Taking into account dp=du = c2s dρ=du with the sound speed defined according to
Eq. (9.7), from Eq. (9.36), we find

du c
= ± s, ð9:37Þ
dρ ρ

and after integration

cs
u= ± dρ: ð9:38Þ
ρ

This expression determines u(ρ) since sound speed is a known function of density;
see Sect. 9.1. Hence, the propagation velocity V ðρÞ = dðdρuÞ
ρ is

V ð ρ Þ = V ð uÞ = u ± c s : ð9:39Þ

One can see that the propagation velocity is different for different points of a density
profile and depends on the density. According to Eq. (9.33), one obtains for the
density:

ρ = ρðz - V ðρÞÞt: ð9:40Þ

In other words, each point of the initial profile moves with a constant velocity,
which, however, depends on the corresponding density. For example, in the case of
isothermal plasma, when cs = const, Eqs. (9.38) and (9.39) yield

ρ
V ðρÞ = cs ln ± cs , ð9:41Þ
ρ0

where ρ0 is the constant, e.g., constant density at infinity. Points of the profile with a
larger density propagate faster than those with a smaller density (Fig. 9.3). The
density profile becomes multivalued with time, and nonlinear steepening and over-
turn take place. A nonphysical multivalued density profile arises because dissipative
9.4 Nonlinear Ion Acoustic Waves with Dispersion 197

Fig. 9.3 Evolution of simple nonlinear wave

terms, in particular ion viscosity, are neglected in the initial equations. However, the
real density profile can be approximately obtained from the equations without
viscosity using the “area rule.” Let us add a vertical line, as shown in Fig. 9.3, so
that the dashed areas should be equal to each other. Then, the density profile, shown
by the solid line, is a solution of the initial equations since it consists of a gradual
profile and discontinuity. The total number of particles for such a profile is conserved
and equals the initial number of particles. At the profile front, a shock is formed,
which is similar to the standard shock in gases.
As in hydrodynamics, to investigate a shock structure, it is necessary to shift to a
reference frame moving with the shock velocity and to take into account the ion
viscosity. The density profile in the shock is obtained from ordinary differential
equations, and its typical scale is of the order of the ion mean free path. However, for
a large mean-free path situation, plasma differs from gaseous case. When the mean-
free path exceeds the Debye radius, the shock structure is determined by the
quasineutrality violation and not by dissipation. In such situations, the shock is
known to be collisionless, and to analyze its structure, kinetic equations should be
solved together with the Poisson equation. The density inside such collisionless
shock has an oscillating character.

9.4 Nonlinear Ion Acoustic Waves with Dispersion

The propagation of nonlinear density perturbations with the typical scale of the order
of the Debye radius has some special features compared to those considered in the
previous section. Dispersion, which exists for linear waves, in the nonlinear situation
may be some special case that compensates for nonlinear steepening and overturning
of the wave. The corresponding solution is known as a soliton. Let us consider an
example of collisionless plasma in the limit of cold ions. The initial equations are
198 9 Dynamics of Fully Ionized Plasma in the Absence of a Magnetic Field

∂ni ∂ðni ui Þ
þ = 0,
∂t ∂z
∂ui ∂u ∂ϕ
mi ni þ ui i = - ni e ,
∂t ∂z ∂z
ð9:42Þ
∂n ∂ϕ
0 = - T e e þ ne e ,
∂z ∂z
2
∂ ϕ
= - eðni - ne Þ=ε0 :
∂z2

We shall seek a partial solution in the form of localized nonlinear perturbation

propagating with a constant velocity, so all quantities are functions of a variable
z - Vt : Ψ = Ψ(ζ) = Ψ(z - Vt). In contrast to the case considered in the previous
section, the propagation velocity is assumed to be constant for all profile points.
The particle balance equation for ions is reduced to

dni dðni ui Þ
-V þ = 0: ð9:43Þ
dζ dζ

Integration yields

ni ðui - V Þ = const = - n0 V: ð9:44Þ

The following boundary conditions are supposed to be satisfied: ne(1) = ni(1) = n0

and ui(1) = 0. The momentum balance for ions in the new variables is

dui du dφ
mi - V þ ui i þe = 0: ð9:45Þ
dζ dζ dζ

After integration:

u2i
mi - ui V þ eφ = 0: ð9:46Þ
2

The integration constant is chosen to satisfy the boundary condition φ(1) = 0. From
the quadratic Eq. (9.43), one finds

ui = V ± V 2 - 2eφ=mi : ð9:47Þ

Choosing minus in the solution, combining Eq. (9.47) with Eq. (9.44), we obtain the
ion density as a function of the potential:

n0 V
ni = : ð9:48Þ
V - 2eφ=mi
2
9.4 Nonlinear Ion Acoustic Waves with Dispersion 199

The momentum balance for electrons gives electron density as a function of potential

eφ
ne = n0 exp : ð9:49Þ
Te

Finally, after the substitution of Eqs. (9.48) and (9.49) into the Poisson equation, one
obtains

d2 φ eφ V
ε0 = en0 exp - : ð9:50Þ
dζ2 Te V - 2eφ=mi
2

This equation coincides with Eq. (3.30), but here, we seek solutions that satisfy
condition φ(0) = 0. Let us introduce dimensionless variables

eφ ζ ε0 T e
Φ= , ζ = , rd = :
Te rd n0 e 2

In the new variables, the Poisson equation has the form

d2 Φ 1
2
= expðΦÞ - 1=2
: ð9:51Þ
dζ 1 - 2Φ=M 2

Here, M = V/cs is the Mach number. Introducing potential energy so that the
r.h.s. equals -dU/dΦ, we have

1=2
U ðΦÞ = - expðΦÞ - M 2 1 - 2Φ=M 2 þ 1 þ M2: ð9:52Þ

Constant choice corresponds to U(1) = 0. This potential energy is known as the

Sagdeev potential. For small Φ, the potential energy is given by a parabola

Φ2 Φ2
U ðΦÞ = - þ : ð9:53Þ
2 2M 2

For M > 1, the potential energy at small Φ is negative and decreases with Φ, as
shown in Fig. 9.4. For larger values of Φ, the potential energy rises, passes through
zero andpbecomes positive. Indeed, when a square root in (9.52) becomes zero at
Φ = M= 2, we have

U M 2 =2 = 1 þ M 2 - exp M 2 =2 :

If M < 1.6, then this value is positive. Hence, for M < 1.6, the potential energy has
the form shown in Fig. 9.4. Here, Φmax corresponds to the root of equation
U(Φmax) = 0.
200 9 Dynamics of Fully Ionized Plasma in the Absence of a Magnetic Field

Fig. 9.4 Potential energy profile for 1 < M < 1.6

Fig. 9.5 Ion acoustic soliton

Let us seek a solution of Eq. (9.51), corresponding to zero energy level E = 0 in

the range
1 < M < 1:6: ð9:54Þ

Using mechanical analogy (ζ represents time and Φ – coordinate), one can notice
that a solution has oscillatory character, and, at Φ → 0, a particle spends infinite
time. The potential profile is given by the equation corresponding to the energy
conservation
2
1 dΦ
= - U, ð9:55Þ
2 dζ

and is shown schematically in Fig. 9.5. The potential profile is given by the integral

dΦ
ζ= ± : ð9:55Þ
1=2
- 2 - expðΦÞ - M 2 1 - 2Φ=M 2 þ 1 þ M2

The solution is the localized potential perturbation moving with a constant velocity
V. Such a solution is known as a soliton. The corresponding density perturbation is
determined by Boltzmann distribution Eq. (9.49). Each part of the profile moves
9.5 Plasma Expansion During Pellet Injection 201

with the same constant velocity, and the profile keeps its shape Eq. (9.55). The
propagation velocity of a soliton depends on its amplitude. This relation can be
derived from the equality U(Φmax) = 0, so we have

½expðΦmax Þ - 12
M ðΦmax Þ = : ð9:56Þ
2½expðΦmax Þ - 1 - Φmax 

The typical spatial scale of the soliton is of the order of the Debye radius, since the
dimensionless variable ζ is the distance measured in units of the Debye radius, the
potential amplitude is of the order of Te/e, and the propagation velocity is supersonic.
Therefore, in contrast to quasineutral nonlinear waves, a special partial solution
exists in the form of a soliton when nonlinearity is balanced by dispersion. Solitons
exist in the range of Mach numbers given by Eq. (9.54). The maximal velocity is
M = 1.6, and the corresponding soliton amplitude according to Eq. (9.56) is also
maximal and equals Φlim max = 1:3. For larger Mach numbers, soliton solutions do not
exist. At M = 1.6 and Φlim
max = 1:3, the square root in Eq. (9.52) becomes zero, and for
larger Mach numbers, it becomes imaginary. Physically, this means that at large
amplitudes, dispersion cannot balance nonlinearity, so nonlinear perturbations
should overturn.

9.5 Plasma Expansion During Pellet Injection

During pellet injection into a tokamak, evaporated neutral particles first expand
spherically, and then, after ionization by electrons move mainly along the magnetic
field, forming a cigar-like structure (see Figs. 1.3 and 1.4). When the pellet stays at a
given flux tube, it may be considered a given constant source of particles, and its
parallel expansion is similar to that analyzed in Sect.9.2. The only difference with the
self-similar solution Eq. (9.28) with constant cloud temperature consists of the fact
that in real tokamak plasma, expanding cold plasma is heated by ambient electrons
with high energy. To obtain density and velocity parallel profiles in this situation,
Eqs. (9.22) should be supplemented by the energy balance equation:

∂T ∂T ∂u ∂qek
3n þu þ nT =- : ð9:57Þ
∂t ∂z ∂z ∂z

We assume Te = Ti due to strong coupling between electrons and ions in cold

plasma. To obtain r.h.s. of this equation,principle kinetic treatment is required, since
hot electrons have a large mean-free path and the fluid treatment does not correctly
describe the heating process. However, in the modeling, one can use an expression
for nonlocal heat conductivity, which, in the situation when the electron mean-free
path is comparable to the cloud dimension, is equivalent to the kinetic description.
According to this approach,
202 9 Dynamics of Fully Ionized Plasma in the Absence of a Magnetic Field

1
∂T
qek = - κek W ðz, z0 Þdz0 , ð9:58Þ
∂z0
0

where

z
3:16nT e 0 1 dz0 0 n z0 0
κek = , W ðz, z Þ = exp - ,
me νei 2λðz0 Þ nðz0 Þλðz0 Þ
z0
p 2 2
ð4πε0 Þ T
λ = 32 2λ0 , λ0 = :
8πne4 Λ

An example of simulations of Eq. (9.22) and (9.58) is presented in Fig. 9.6. The
dotted line corresponds to the self-similar solution Eq. (9.28) for the isothermal

Fig. 9.6 (a) Density; (b) electron temperature; (c) hydrodynamic velocity profiles for t = 25μs. The
plasma source is switched on at t = 0 and is situated at z = 0. Dotted line: calculation according to
Eq. (9.28). The plasma and pellet parameters correspond to: n0 = 3 × 19m-3, Te0 = 500eV,
N_ = 1023 s - 1 . Initially cold plasma had temperature Te0 = 5eV
9.5 Plasma Expansion During Pellet Injection 203

plasma expanding into a vacuum produced by a constant source. We see that the
bulk of the ablatant remains fairly cold. A discrepancy between the self-similar
solution for the density and the result of the calculation appears at densities that are
an order of magnitude smaller than those at z = 0. The elongation of the tail is due to
ablatant heating by the ambient electrons. The velocity profile, which should be
linear for isothermal plasma expanding into the vacuum, drops off to zero at
distances where the ablatant density becomes comparable with the ambient density
and the self-similar solution loses its validity.
Chapter 10
Magnetohydrodynamics (MHD)

10.1 Magnetohydrodynamic Equations

The equation system of the transport equations for density and fluid velocities in a
magnetic field for pure plasma is as follows:

∂n →
þ ∇  n u e = 0,
∂t
∂n →
þ ∇  n u i = 0,
∂t ð10:1Þ
→ → → →
0 = - ∇pe - en E - ne u e × B þ R ,
→
dui → → → →
nmi = - ∇pi þ en E þ ne u i × B - R :
dt

Electron inertia here is neglected. It is supposed that the pressure is a known function
of density and is determined, for example, by the adiabatic law, which replaces
energy balance equations. These basic equations are known as two-fluid magneto-
hydrodynamic (MHD) equations.
Let us transform the initial two-fluid MHD equations into a new equation system
where electric fields and plasma currents are excluded. After summing up the two
momentum balance equations for electrons and ions, the electric field and friction
force are canceled, and with an account of the current density definition
→ → →
j = en u i - u e , one obtains

→
dui → →
nmi = - ∇p þ j ×B : ð10:2Þ
dt

This equation can be interpreted in the following way. Two forces are applied to the
plasma as a whole: the total pressure gradient and the Lorentz force. They cause

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 205
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_10
206 10 Magnetohydrodynamics (MHD)

plasma acceleration, which practically coincides with the acceleration of ions since
the total momentum is practically the ion momentum due to their large mass.
According to the Maxwellian equation

→
→ → 1 ∂E
∇ × B = μ0 j þ :
c2 ∂t

For slow processes, which are considered in this chapter, displacement current
density (the last term on the r.h.s.) can be neglected. Indeed, since drift velocity is
connected with electric field, u = E/B, the ratio

→
∂E 1 → EL u2
= ∇ × B   < <1
∂t c2 c2 Bt c2

is small. Hence, the Maxwellian equation has the form

→ →
∇ × B = μ0 j : ð10:3Þ

Substituting Eq. (10.3) into Eq. (10.2) yields

→
dui 1 → →
nmi = - ∇p þ ∇× B × B : ð10:4Þ
dt μ0

With an account of identity

→ → → → 1
∇× B × B = B∇ B - ∇B2
2

the total momentum balance Eq. (10.4) is reduced to

→
dui B2 1 → →
nmi = - ∇p - ∇ þ B∇ B: ð10:5Þ
dt 2μ0 μ0

The second term on the r.h.s. is known as the magnetic pressure gradient. Magnetic
pressure B2/(2μ0) is added to the gas kinetic pressure and accelerates plasma in a
nonuniform magnetic field. The last term is called “magnetic tension” and is
associated with magnetic field line curvature. Both terms physically represent the
Lorentz force caused by plasma currents in a magnetic field.
To obtain one more equation, we express the electric field from the electron
momentum balance:

→
→ ∇pe → → R
E=- - ue× B þ : ð10:6Þ
en en
10.1 Magnetohydrodynamic Equations 207

Let us take the curl of both parts of Eq. (10.6). Considering the Maxwellian equation,
the l.h.s. is transformed to

→
→ ∂B
∇× E = - :
∂t

Now, for different contributions to the r.h.s, we have the following. First

∇pe 1
-∇× = - ∇ × ðT e ∇ ln n þ ∇T e Þ:
en e

With an account of identity

→ → →
∇ × a b = a∇ × b þ ∇a × b ,

we find

∇pe 1 ∇n
-∇× = × ∇T e : ð10:7Þ
en e n

For the second term:

→ → → → 1 * →
- ∇ × u e × B = - ∇ × u i × B þ ∇ × j =n × B
e
→ → →
ð10:8Þ
→ 1
= -∇× ui× B þ ∇ × ∇ × B =n × B :
eμ0

In the third term, we neglect the thermal force, which contributes to slow diffusive
evolution. To study such processes, one has to use the initial system of two-fluid
MHD Eqs. (10.1). The remaining part of the friction force caused by fluid velocity
→ →
can be written in the form R = σ - 1 ne j , where σ - 1 is a tensor reversed to
conductivity tensor σ with components σk = ne2/(0.51meνei) and σ⊥ = ne2/(meνei).
Hence,

→ → 1 →
∇ × R =ne = ∇ × σ - 1 j = ∇ × σ - 1∇ × B : ð10:9Þ
μ0

Finally, Eq. (10.6) is reduced to

→
∂B 1 → → 1 → →
= - ½∇n × ∇T e  þ ∇ × u i × B - ∇× ∇ × B =n × B
∂t en μ0 e
1 →
- ∇ × σ - 1∇ × B :
μ0
ð10:10Þ
208 10 Magnetohydrodynamics (MHD)

Now, we divide two parts of momentum balance Eq. (10.4) by density and take curl.
The term with the magnetic field is transformed to

→
1 1 → → dui ∇p
∇× ∇ × B × B = mi ∇ × þ∇×
μ0 n dt n
ð10:11Þ
∇n ∂ → → →
=- × ∇ðT e þ T i Þ þ mi ∇ × u i þ mi ∇ × ∇× u i × u i :
n ∂t

→ → → →
Here, we used identity ∇ × u i × u i = u i ∇ u i - 12 ∇u2i . Combining
Eqs. (10.10) and (10.11), we find

∂ → mi * → → m * 1
B þ ∇ × ui = ∇ × u i × B þ i ∇ × ui þ ½∇n × ∇T i 
∂t e e en
ð10:12Þ
1 →
- ∇ × σ - 1∇ × B :
μ0
*
In a strong magnetic field, the quantity mei ∇ × u i may be neglected with respect to
B, and the parameter is the ratio ρci(ui)/L = miui/(eBL) < < 1, where L is the typical
spatial plasma scale. In other words, the ion Larmor radius calculated using fluid ion
velocity should be smaller than L. Then, Eq. (10.12) in a strong magnetic field is

→
∂B → → 1 1 →
= ∇ × u i × B þ ½∇n × ∇T i  - ∇ × σ - 1 ∇ × B : ð10:13Þ
∂t en μ0

The second term on the r.h.s. is nonzero when the density and ion temperature
gradient are not parallel to each other. We shall not consider this situation and omit
this term.
In the final form, the initial equation system (10.1) is reduced to

→
∂B → → 1 →
= ∇ × u × B - ∇ × σ - 1∇ × B ,
∂t μ0
→
du B2 1 → →
nmi = - ∇p - ∇ þ B∇ B, ð10:14Þ
dt 2μ0 μ0
∂n →
þ ∇  n u = 0,
∂t
p = pðnÞ:

We dropped here the ion velocity subscript i, added the adiabatic law for pressure
and dropped continuity equations for electrons. This system is called a single-fluid
MHD system. Within such an approach, the plasma is considered a single fluid with
→
mean fluid velocity u associated with fluid ion velocity, mass density ρ = min,
pressure p = n(Te + Ti), and magnetic field, which reveals itself as magnetic pressure
and magnetic tension. Single-fluid MHD equations are convenient for various
10.2 Magnetic Field Frozen in and Skin Effect 209

classes of problems, some of which are analyzed below. On the other hand, the
electric field and currents are absent in these equations, and sometimes physical
interpretation of a single fluid MHD is difficult, so one has to go back to two-fluid
equations.
From the second equation of system (10.14), one can see that in addition to the
sound speed in a strong magnetic field, there is one more typical velocity cA,

B2
cA = : ð10:15Þ
μ0 nmi

This velocity is called the Alfven velocity. Indeed, estimating l.h.s. of the second
Eq. (10.14) as nmiu2/L, where L is a typical spatial scale, and r.h.s. as B2/(2μ0L ), one
obtains u~cA.

10.2 Magnetic Field Frozen in and Skin Effect

Comparing the second term on the r.h.s. of the first Eq. (10.14) with the term on the l.
h.s., we obtain a characteristic time

τs = μ0 σL2 , ð10:16Þ

which is known as the skin time. For times smaller than the skin time, it is possible to
neglect the diffusive contribution in the equation for the magnetic field in
Eq. (10.14):

→
∂B → →
=∇× u × B : ð10:17Þ
∂t

This is the equation for frozen in the magnetic field. Let us prove the following
statement: during perpendicular 2D plasma motion across a straight magnetic field,
the ratio B/n is conserved. Indeed, let the magnetic field be directed along the z-axis,
while plasma motion takes place in the x, y plane. Then, Eq. (10.17) is reduced to

∂B ∂ðBux Þ ∂ Buy
þ þ = 0: ð10:18Þ
∂t ∂x ∂y

Combining it with the continuity equation for density

∂n ∂ðnux Þ ∂ nuy
þ þ = 0, ð10:19Þ
∂t ∂x ∂y
210 10 Magnetohydrodynamics (MHD)

we have

dðB=nÞ ∂ðB=nÞ → B
= þ u∇ = 0: ð10:20Þ
dt ∂t n

Hence, in the regions where plasma is compressed, the magnetic field increases
proportionally. In other words, magnetic field lines act as frozen in plasma and
converge or diverge together with plasma. The physical mechanism of the magnetic
field frozen in is connected with the electric field, which, according to the Maxwel-
lian equation, is induced when the magnetic field is changing over time. The induced
→ →
electric field produces E × B drifts, which lead to plasma compression or decom-
pression. Note that the quantity B/n is conserved only for 2D motion across a straight
magnetic field, while in the general 3D case, the magnetic field evolution is
described by the general Eq. (10.17).
In the opposite situation, when plasma motion across the magnetic field can be
neglected, the first equation in Eq. (10.14) is reduced to

→
∂B 1 →
= - ∇ × σ - 1∇ × B : ð10:21Þ
∂t μ0

In the simple case when conductivity may be considered constant and independent
of coordinates, we have

→
∂B 1 →
= ΔB: ð10:22Þ
∂t μ0 σ

Equation (10.22) is a diffusion equation with a coefficient of magnetic field diffusion

1
DB = : ð10:23Þ
μ0 σ

Equation (10.21) or its simplified form Eq. (10.22) describes the diffusive penetra-
tion of the magnetic field into the plasma.
Let us analyze this mechanism using the redistribution of current and poloidal
magnetic field in a tokamak as an example. In the cylindric geometry, Eq. (10.21) for
a poloidal (azimuthal) magnetic field is given by

∂Bϑ 1 ∂ 1 ∂ðrBϑ Þ
= , ð10:24Þ
∂t μ0 ∂r σk r ∂r

where the parallel conductivity on the r.h.s. σk = ne2/(0.51meνei). According to this

equation, current redistribution takes place at a skin time scale τs = μ0σka2, where
10.2 Magnetic Field Frozen in and Skin Effect 211

a is the cylinder radius. Let us consider the case when the net plasma current
I increases instantly from the value I0 to I1. The current distribution in plasma is
determined by Ohm’s law jz ≈ jk = σkEk ≈ σkEz. In the stationary case, the electric
field Ez is constant in time and is independent of the radius; therefore, the radial
current distribution is determined by the radial profile of the parallel conductivity,
e : Therefore, the main part of the current flow is in the
which is proportional to T 3=2
hot central region. Then, for the given temperature profile, it is easy to obtain the
current density profile from the condition

2πrjz dr = I: ð10:25Þ
0

First, the radial distribution of the poloidal magnetic field B0θ ðr Þ corresponds to zero
r.h.s. of Eq. (10.24), and the poloidal magnetic field at the boundary is determined by
the net current value

μ0 I 0
B0θ ðaÞ = : ð10:26Þ
2πa

When the net current rises, these poloidal fields at the plasma boundary rise
simultaneously. Further evolution of the poloidal magnetic field and parallel current,
which is connected to the poloidal field by the Maxwellian equation, is governed by
Eq. (10.25) with the step-like initial condition

Bθ ðr, t = 0Þ = B0θ ðr Þ, r < a;

μ I ð10:27Þ
B0θ ðaÞ = 0 1 , r = a:
2πa

Since at t = 0, the poloidal magnetic field ramps up at the boundary r = a, the

additional current is concentrated inside the thin layer of zero width at the
boundary (Fig. 10.1). Jump in the current density here is caused by the
corresponding jump of the electric field, generated by the variation of magnetic
field. Then, the current starts penetrating into the plasma within a distance of
the order of skin depth δ  1=μ0 σk t, which rises over time. The poloidal
magnetic field changes with the same scale. The redistribution of the poloidal
magnetic field and current ends at t > τs, and the new current profile repeats
the old one, while the current density amplitude becomes larger than the initial
one.
In the general case, when a magnetic field penetrates into a moving plasma, one
must use general Eq. (10.14).
212 10 Magnetohydrodynamics (MHD)

Fig. 10.1 Skinned current density profile during fast net current ramp-up

10.3 MHD Waves

We shall analyze one-dimensional MHD waves using single-fluid MHD Eq. (10.14)
in the collisionless plasma. The solution is sought in the form when all quantities are
functions of coordinate z and time t, and, therefore, are functions of each other. Note
that the magnetic field direction does not coincide with the z-axis. Equation
corresponding to adiabatic approximation we shall replace by equivalent equation
for entropy conservation ds/dt = 0. In the equation for the magnetic field, the term
containing conductivity is neglected. Then equation system (10.14) in 1D case is:

∂ux ∂u B ∂Bx
þ uz x - z = 0,
∂t ∂z μ0 ρ ∂z
∂uy ∂uy B ∂By
þ uz - z = 0,
∂t ∂z μ 0 ρ ∂z
∂uz ∂u B ∂Bx By ∂By 1 ∂p
þ uz z þ x þ þ = 0,
∂t ∂z μ0 ρ ∂z μ0 ρ ∂z ρ ∂z
∂Bx ∂u ∂u ∂B
- Bz x þ Bx z þ uz x = 0,
∂t ∂z ∂z ∂z ð10:28Þ
∂By ∂uy ∂uz ∂By
- Bz þ By þ uz = 0,
∂t ∂z ∂z ∂z
∂Bz
= 0,
∂t
∂p ∂u ∂p
þ ρc2s z þ uz = 0,
∂t ∂z ∂z
∂s ∂s
þ uz = 0:
∂t ∂z

In the ion particle balance equation, we switched to the pressure derivative, using
10.3 MHD Waves 213

∂p dp ∂ρ ∂ρ ∂p dp ∂ρ ∂ρ
= j = c2s , = js = const = c2s :
∂t dρ s = const ∂t ∂t ∂z dρ ∂z ∂z

From the equation for the z-component of magnetic field ∂Bz/∂t = 0, it follows that
→
Bz is the constant in time. On the other hand, from the condition ∇  B = 0, in which
the 1D case is reduced to ∂Bz/∂z = 0, we find that component Bz is also constant in
space. Hence, the projection of the magnetic field on the z-axis remains constant
during MHD wave propagation: Bz = const. The remaining 7 Eqs. (10.28) for seven
variables ux, uy, uz, Bx, By, p, s should be solved simultaneously.
In the matrix form, Eq. (10.28) can be written as

∂Ψ ∂Ψ
þ Z ð ΨÞ = 0, ð10:29Þ
∂t ∂z

where Ψ is a vector in a 7D space, Ψ = (ux, uy, uz, Bx, By, p, s), and Z(Ψ) is a
matrix

Bz
uz 0 0 - 0 0 0
μ0 ρ
Bz
0 uz 0 0 - 0 0
μ0 ρ
Bx By 1
Z ð ΨÞ = 0 0 uz 0 : ð10:30Þ
μ0 ρ μ0 ρ ρ
- Bz 0 Bx uz 0 0 0
0 - Bz By 0 uz 0 0
0 0 ρc2s 0 0 uz 0
0 0 0 0 0 0 uz

The solution is sought in the form of linear waves, i.e., all quantities are supposed to
be a sum Ψ = Ψ0 + Ψ1, where vector Ψ1 represents small perturbations. The
linearized Eq. (10.29) is:

∂Ψ1 ∂Ψ1
þ Z Ψ0 = 0: ð10:31Þ
∂t ∂z

For simplicity, one can choose a reference frame where the unperturbed fluid
velocity along the z-axis is zero, uz = 0, and the unperturbed y-component of the
magnetic field is also zero, By = 0. The last condition can be reached by rotating the
coordinate system over the z-axis. Then, the matrix Z(Ψ0) can be simplified:
214 10 Magnetohydrodynamics (MHD)

Bz
0 0 0 - 0 0 0
μ0 ρ
Bz
0 0 0 0 - 0 0
μ0 ρ
Bx 1
Z Ψ0 = 0 0 0 0 0 : ð10:32Þ
μ0 ρ ρ
- Bz 0 Bx 0 0 0 0
0 - Bz 0 0 0 0 0
0 0 ρc2s 0 0 0 0
0 0 0 0 0 0 0

Vector Ψ1 is supposed to have the form Ψ1 = A exp (ikz - iωt). Then, from
Eq. (10.31), one obtains

ZA = VA, ð10:33Þ

where V = ω/k is the phase velocity. Eq. (10.33) represents the classical problem of
finding eigenvalues and eigenvectors for matrix Z(Ψ0). The determinant of the
matrix (Z - VI) = 0 should be zero, with I being the unit matrix,

detðZ - VI Þ = 0: ð10:34Þ

From Eq. (10.34), seven eigenfrequencies can be obtained.

The first two roots of this equation correspond to the Alfven wave:

V A = ± cA cos θ: ð10:35Þ

Here, the Alfven wave velocity cA is defined according to Eq. (10.15), and θ is the
angle between the magnetic field and the z-axis. Signs ± correspond to two direc-
tions of propagation, i.e., physically the same type of wave. Two second types of
phase velocities represent fast magnetosonic waves:

1=2
2
c2A þ c2s þ c2A þ c2s - 4c2A c2s cos 2 θ
VF = ± : ð10:36Þ
2

The phase velocities of the slow magnetosonic wave are

1=2
2
c2A þ c2s - c2A þ c2s - 4c2A c2s cos 2 θ
VS = ± : ð10:37Þ
2

The last phase velocity corresponds to the entropy wave:

10.3 MHD Waves 215

Fig. 10.2 Phase diagram for MHD waves

V E = 0: ð10:38Þ

The dependence of the phase velocity on the angle with a magnetic field is often
expressed on a diagram, known as the phase polar (Fig. 10.2). It can be seen that the
Alfven wave propagates mainly along the magnetic field, for θ = 0, π phase velocity
coincides with the Alfven velocity VA = cA, and, in the perpendicular direction, the
phase velocity is equal to zero. Fast magnetosonic waves can propagate both along
and across magnetic fields. In the strong magnetic field cA > > cs, its phase velocity
coincides with the Alfven velocity independently of the angle with the magnetic
field. The third mode is called the slow magnetosonic wave. This wave propagates
mainly in the parallel direction, and, in the strong magnetic field, cA > > cs for
θ = 0, π, the phase velocity coincides with the ion sonic velocity cs. The phase
velocity of the entropy wave is zero.
Eigenvectors for these modes for 0 < θ < π/2 are:

AA = ½0, cA cos θ, 0, 0, - Bz , 0, 0,

V F c2A sin θ cos θ Bx V 2
AF = , 0, V F , 2 F 2 , 0, ρc2s , 0 ,
VA - VF
2 2
VF - VA
ð10:39Þ
V S c2A sin θ cos θ Bx V 2
A =S
, 0, V S , 2 S 2 , 0, ρc2s , 0 ,
VA - VS
2 2
VS - VA
AE = ½0, 0, 0, 0, 0, 0, 1:

Eigenvector components show which quantities are exited in the given mode and
what are the relations between them. For example, in the Alfven wave, both the
velocity and the magnetic field along the y-axis are perturbed while all the other
parameters remain constant. Their ratio is u1y =B1y = - c0A cos θ=B0z .
Let us analyze physical processes in different modes for simple cases of wave
propagation. We start with the Alfven wave propagating in the parallel direction,
θ = 0. Assume that there is an electric field directed along x, periodic in space and
216 10 Magnetohydrodynamics (MHD)

→ →
time, E 1x  expð- iωt þ ikzÞ. This field causes E × B drift along the y-axis with
velocity u1y = - E1x =B0 . Simultaneously, the polarization current flows in the x-
direction:

n0 mi ∂E 1x n m
j1x = = - iω 0 2 i E1x : ð10:40Þ
B02 ∂t B0

This current produces magnetic field perturbation according to the Maxwellian

equation:

1 ∂By
1
ik
j1x = - = - B1y : ð10:41Þ
μ0 ∂z μ0

The time-dependent magnetic field in its turn produces an electric field

∂B1y ∂E1x
- = iωB1y = = ikE 1x : ð10:42Þ
∂t ∂z

After eliminating the current and electric field from Eqs. (10.40) to (10.42), one finds
the dispersion relation for the Alfven wave:

2
ω2 = k2 c0A : ð10:43Þ

Particle drift in the Alfven wave takes place at a homogeneous background; hence,
no density perturbation arises. The only two perturbed quantities are velocity u1y and
magnetic field B1y , and their ratio, according to Eqs. (10.40) and (10.41), is
u1y =B1y = - c0A =B0 , which is consistent with the general expression for the eigenvec-
tor of the Alfven mode Eq. (10.39). The drift velocity is therefore phase shifted at π
with respect to magnetic field perturbation, Fig. 10.3. Plasma displacement ξ =
u1y dt is phase shifted at π/2.
A fast magnetosonic wave propagating perpendicular to the magnetic field in the
z-direction, θ = π/2, can be analyzed similarly. The main magnetic field is parallel to
the x-axis (Fig. 10.4) and is assumed to be sufficiently strong, cA > > cs. Consider,
as in the previous case, an electric field parallel to the y-axis and periodic in time and
→ →
space: E 1y  expð- iωt þ ikzÞ. Plasma E × B drift is directed along the z-axis with
velocity u1z = - E 1y =B0 , while the polarization current

n0 mi ∂E y
1
n m
j1y = = - iω 0 2 i E1y : ð10:44Þ
B20 ∂t B0

The current causes magnetic field perturbation in the x-direction:

10.3 MHD Waves 217

Fig. 10.3 Bend of magnetic field line and plasma shift ξ during Alfven wave propagation

Fig. 10.4 Magnetic field and density perturbations in the fast magnetosonic wave

1 ∂B1x ik
j1y = = B1x , ð10:45Þ
μ0 ∂z μ0

which induces an electric field according to

∂B1x ∂E1y
- = iωB1x = - = - ikE 1y : ð10:46Þ
∂t ∂z
218 10 Magnetohydrodynamics (MHD)

From Eqs. (10.44) to (10.46), one obtains the dispersion relation, which coincides
with Eq. (10.43). Therefore, the fast magnetosonic wave propagates with the Alfven
velocity. The perturbation characteristics are illustrated in Fig. 10.4. The magnetic
field is perturbed in this wave; however, the magnetic field lines remain straight. It
can be seen from Eqs. (10.44) to (10.45), the particle velocity is in the same phase as
the magnetic field perturbation, and u1z =B1x = c0A =B0 in accordance with the general
expression Eq. (10.39). The density in this wave is also perturbed, as follows from
the particle balance - iωn1 þ ikn0 u1z = 0, so that n1 =n0 = B1x =B0 . The last expression
corresponds to the magnetic field frozen in Eq. (10.17) and the conservation of the
ratio B/n for motion across the magnetic field. It also corresponds to eigenvector
Eq. (10.39).
Fast magnetosonic waves propagating across a magnetic field resemble sonic
waves, where magnetic pressure B2/2μ0 plays a role of plasma pressure. Indeed, the
Alfven velocity can be obtained by replacing the pressure in the expression for the
sonic velocity with the magnetic pressure. In the limit cA < < cs, a fast
magnetosonic wave is transformed into a sonic wave according to Eq. (10.39). A
fast magnetosonic wave, which propagates along a magnetic field, is in fact an
Alfven wave with different polarization. At θ → 0, the first and fourth terms in the
expression for eigenvector AF diverges as θ-1, since V 2A - V 2f → c2A sin 2 θ and
Bx = B sin θ. Other terms remain finite and can be neglected. The ratio of the first
and fourth terms is u1x =B1x = c0A =B0 , in other words, it corresponds to an Alfven wave
with a perturbed x-component of the magnetic field, with plasma also moving along
the x-axis.
Slow magnetosonic waves propagating in the parallel direction in a strong
magnetic field correspond to sonic waves.
The entropy wave represents entropy perturbation in plasma with unperturbed
pressure. Let us, for example, consider positive density perturbation n1(z) and
negative temperature perturbations T 1e ðzÞ and T 1i ðzÞ, so that
n1 =n0 = - T e þ T i = T e þ T i . Since pressure perturbation is absent, such den-
1 1 0 0

sity perturbation will stay at rest in the absence of dissipation, according to the
3=2
pressure balance equation. At the same time, entropy proportional to ln T e,i =n
remains perturbed.
MHD waves decay due to dissipative processes. In addition, there exists a specific
collisionless decay mechanism caused by the spatial dependence of the phase
velocity. We shall illustrate it for the case of an Alfven wave propagating along
→ →
magnetic field B 0 k z in the inhomogeneous plasma with density depending on the
x-coordinate. According to the pressure balance in Eq. (7.5), p0 + B2/2μ0 = const(x).
If β < < 1, the magnetic field is only slightly perturbed B ≈ B0, and the dependence
of the Alfven wave on the x-coordinate is mainly determined by the inhomogeneous
density

cA ð x Þ = B20 =μ0 mi n0 ðxÞ: ð10:47Þ

10.4 Nonlinear MHD Waves 219

Consider х-localized wave packet of Alfven waves, where the y-component of the
magnetic field is perturbed

δBy = bðxÞ expð- iωt þ ikzÞ: ð10:48Þ

Phase velocity ω/k = cA depends on х; hence, waves with the same phase at z = 0
will diverge at phases for large values of z. In the experiment, magnetic field
perturbation is always averaged in space; hence, averaged over х magnetic pertur-
bation should turn to zero due to phase divergence. This effect manifests itself as
collisionless damping of Alfven waves, in spite of the absence of real collisional
damping. Indeed, the magnetic field perturbation averaged over interval 2δ can be
defined as

xþδ xþδ
1 1
δBy = δBy dx = bðxÞ expð- iωt þ ikzÞdx: ð10:49Þ
2δ 2δ
x-δ x-δ

Since ω(x) = kcA(x), one can integrate over ω:

ωþ
expðikzÞ
δBy = bðωÞ expð- iωt Þdω, ð10:50Þ
2δ
ω-

where

bð x ð ω Þ Þ
bðωÞ = , ω ± = ωðx ± δÞ:
dω=dx

Integrating by parts yields

ωþ
expðikzÞ bðωÞ expð- iωt Þ ωþ 1 dbðωÞ
δBy = - ω- þ expð- iωtÞdω :
2δ it it dω
ω-
ð10:51Þ

Continuing integration by parts, one obtains a series with terms inversely propor-
tional to t. At t → 1, all terms turn to zero, and hδByi also turns to zero, so the
average magnetic field perturbation is damped.

10.4 Nonlinear MHD Waves

In the general case, nonlinear waves are described by Eq. (10.29). In the plane (z, t),
special curves can be defined,
220 10 Magnetohydrodynamics (MHD)

Fig. 10.5 Characteristic in (t, z) plane

z = ζðt Þ, ð10:52Þ

which are called characteristics, Fig. 10.5. At such curve, vector Ψ is defined by
initial conditions, while its value Ψ outside of characteristics could not be found for
known value Ψ at the characteristics. Let us express derivatives ∂Ψ/∂z and ∂Ψ/∂t as
functions of the normal derivative to characteristic ∂Ψ/∂n and derivative ∂Ψ/∂s
along the characteristic:

∂Ψ 1∂Ψ V ∂Ψ
= - ,
∂t 1 þ V ∂s 2
1 þ V ∂n2
ð10:53Þ
∂Ψ V ∂Ψ 1 ∂Ψ
= þ :
∂z 1þV 2 ∂s
1þV 2 ∂n

Here, V = dζ/dt slope of the curve z = ζ(t). Substituting Eq. (10.53) into Eq. (10.29)
yields

∂Ψ ∂Ψ
ðZ - VI Þ = - ðVZ þ I Þ : ð10:54Þ
∂n ∂s

According to the definition, vector Ψ outside of the characteristics could not be

found for known value Ψ, and, therefore, the derivative ∂Ψ/∂n could not be obtained
from Eq. (10.54). On the other hand, ∂Ψ/∂s = 0 at the characteristics. Hence, an
invertible matrix for matrix Z - VI does not exist, and Z - VI is a degenerate matrix.
Therefore, we have the relation

detðZ - VI Þ = 0, ð10:55Þ

which fully coincides with Eq. (10.34). As in the linear case, 7 types of pure waves
and 7 phase velocities dζ(t)/dt = V are obtained from Eq. (10.55). The difference
consists of the fact that matrix Z(Ψ) Eq. (10.30) depends on Ψ, and, therefore,
V = V(Ψ). Since vector Ψ is conserved at the characteristic, the value V is constant
there so that the characteristic is a straight line. At the same time, the slopes of the
characteristics are different, and the propagation velocity V depends on amplitude.
10.4 Nonlinear MHD Waves 221

Let us consider a nonlinear Alfven wave propagating along a magnetic field as an

example of a nonlinear MHD wave. Perturbed components are supposed to be uy and
By. According to the general analysis of Sect. 10.3, the magnetic field component Bz
remains constant in the 1D wave. The second and fifth Eqs. (10.28) are:

∂uy B ∂By
- z = 0,
∂t μ0 ρ ∂z
ð10:56Þ
∂By ∂uy
- Bz = 0:
∂t ∂z

One can easily see that the solution of this equation system coincides with the
solution of the system

∂By ∂By
-V = 0,
∂t ∂z
ð10:57Þ
∂uy ∂uy
-V = 0,
∂t ∂z

if

B2z
V = cA = : ð10:58Þ
μ0 ρ

From the third equation of system (10.28)

By ∂By 1 dp
þ =0 ð10:59Þ
μ0 ρ ∂z ρ dz

one obtains

B2y
pþ = const: ð10:60Þ
2μ0

We see that in the regions where the magnetic field and magnetic pressure increase
(note that Bz = const), the plasma pressure and, hence, plasma density decrease. The
Alfven velocity Eq. (10.58) is proportional to ρ-1/2, and considering adiabatic law
p~ρ5/3, the propagation velocity V~p-0.3; therefore, the propagation velocity
decreases with pressure, and according to Eq. (10.60) increases with the magnetic
field. As a result, regions with a larger magnetic field propagate faster than those with
a smaller magnetic field, so the nonlinear Alfven wave overturns.
In the second example, we consider a fast magnetosonic wave propagating across
a magnetic field Bx. Since for perpendicular motion, the ratio
222 10 Magnetohydrodynamics (MHD)

Bx
b= = const ð10:61Þ
ρ

is conserved, the momentum balance equation along the z-axis can be rewritten in
the form

∂uz ∂u 1 ∂B2x 1 ∂p 1 ∂p

þ uz z = - - = , ð10:62Þ
∂t ∂z 2μ0 ρ ∂z ρ ∂z ρ ∂z

where

b2 ρ 2
p = p þ :
2μ0

Equation (10.62) and particle balance equation coincide with equations for nonlinear
sonic waves, Sect. 9.3, where p is replaced by p. The wave propagates with Alfven
velocity (for cA > > cs), with respect to plasma velocity

dp b2 ρ bBx
V = uz þ c A ≈ cA = = = : ð10:63Þ
dρ μ0 μ0

Since the propagation velocity is proportional to B1=2

x , the wave overturns.

10.5 Magnetosonic Waves with Dispersion

We shall demonstrate using magnetosonic waves as an example that in magnetized

plasma, soliton-type solutions exist, which propagate perpendicular to the magnetic
field without deformation. However, in contrast to sonic solitons, which need
deviation from quasineutrality, magnetosonic soliton solutions require an account
of electron inertia. Let the magnetic field be directed along the z-axis, while a wave
propagates along the x-axis with velocity V. We shall seek solutions with V = const
that are independent of the wave amplitude, so all quantities are sought in the form
Ψ = Ψ(x - Vt).
As basic equations, we take two-fluid MHD equations with zero temperatures in
the collisionless limit. The particle balance equation for electrons and ions is

∂n ∂ðnuxe,i Þ
þ = 0, ð10:64Þ
∂t ∂x

or, since density is n = n(ξ = x - Vt),

10.5 Magnetosonic Waves with Dispersion 223

dn d ðnuxe,i Þ
-V þ = 0: ð10:65Þ
dξ dξ

Integrating Eq. (10.65), assuming n(ξ → 1) = n0 and uxe, i(ξ → 1) = 0, one obtains

nðuxe,i - V Þ = - n0 V: ð10:66Þ

According to Eq. (10.66), the ion and electron velocities in the x-direction coincide:
uxe = uxi = u. The momentum balance equations in the new variables are

duxe,i
me,i ðuxe,i - V Þ = ∓ eE x ∓ euye,i B,
dξ
ð10:67Þ
duye,i
me,i ðuxe,i - V Þ = ∓ eE y ± euxe,i B:
dξ

Electric and magnetic fields are linked by the Maxwellian equation

→ →
∇ × E = - ∂ B =∂t, which for variable ξ = x - Vt is

dEy dB
=V : ð10:68Þ
dξ dξ

Integration with an account of Ey(ξ → 1) = 0 and B(ξ → 1) = B0 yields

E y = V ðB - B0 Þ : ð10:69Þ
→ →
From the second Maxwellian equation ∇ × B = μ0 j , we have

dB
- = μ0 en uyi - uye : ð10:70Þ
dξ

After summing up the x components in the momentum balance for electrons and ions
in Eq. (10.67), with account of uxe = uxi = u and mi > > me, one obtains

du
m i ðu - V Þ = e uyi - uye B : ð10:71Þ
dξ

Combining this equation with Eq. (10.70) and particle balance Eq. (10.66), we find

du B dB
m i n0 V = , ð10:72Þ
dξ μ0 dξ

and after integration:

B2 - B20
u= : ð10:73Þ
2μ0 mi n0 V
224 10 Magnetohydrodynamics (MHD)

Now, the sum of two y-components of momentum balance Eq. (10.67) yields
meuye + miuyi = 0. Hence,

uye > > uyi : ð10:74Þ

In other words, magnetic field perturbation is created by electron current, while ion
current in Eq. (10.70) can be neglected. Substituting the electron velocity uye as a
function of the magnetic field derivative from Eq. (10.70) into the second
Eq. (10.67), with account of Eq. (10.69), Eq. (10.66) for density and Eq. (10.73),
we obtain the equation for the magnetic field

δ2 d dB B2 - B20 B2 - B20 B2 - B20

- -V -V = - V B þ VB0 :
V dξ dξ 2μ0 mi n0 V 2μ0 mi n0 V 2μ0 mi n0 V
ð10:75Þ

Here, a scale

1=2
me
δ= ð10:76Þ
n0 e 2 μ 0

determines the characteristic perturbation spatial scale. This scale is known as

collisionless skin depth. Multiplying both parts of Eq. (10.75) by dB/dξ and inte-
grating, we have

2 2
dB
2
B2 - B20 B2 - B20
- δ2 -V = - V 2 ðB - B0 Þ2 þ const: ð10:77Þ
dξ 2μ0 mi n0 V 4μ0 n0 mi

The constant is chosen to be zero, so that at infinity B = B0 and dB/dξ = 0. Then,

from Eq. (10.77), we obtain

dB B - B0 ðB þ B0 Þ2
δ =± 2
- B20
1- : ð10:78Þ
dξ B
-1 4μ0 mi n0 V 2
2μ0 mi n0 V 2

Curves in the phase space are shown in Fig. 10.6. A soliton solution corresponds to a
rise of the magnetic field from the value B0 up to Bmax (branch dB/dξ > 0) and a
further drop of the field from Bmax back to B0 (branch dB/dξ < 0). At the maximum
dB/dξ = 0, so according to Eq. (10.78)

ðBmax þ B0 Þ2
V= : ð10:79Þ
4μ0 n0 mi
10.5 Magnetosonic Waves with Dispersion 225

Fig. 10.6 Curves in the phase space (B, dB/dξ)

Fig. 10.7 Magnetosonic soliton

This relation connects the velocity of the magnetosonic soliton with its amplitude.
The soliton shape can be obtained by further integration of Eq. (10.78). An example
is shown in schematic Fig. 10.7. The Alfvenic Mach number is introduced as

V V 2μ0 n0 mi
M= = : ð10:80Þ
cA B0

For small perturbations B → B0, the soliton velocity becomes the Alfven velocity
and M → 1. For larger amplitudes, the Mach number rises in accordance with
Eq. (10.79) until the denominator in Eq. (10.78) becomes zero. The critical value
of the magnetic field according to Eqs. (10.78) and(10.79) is Bmax = 3B0 for Mach
number M = 2. In this case, the spatial derivative in the maximum dB/dξ → 1,
Fig. 10.7. Therefore, magnetic solitons exist in the following range of Alfvenic
Mach numbers:

1 ≤ M ≤ 2: ð10:81Þ

For larger Mach numbers, the effects associated with electron inertia are not able to
prevent overturn, and collisionless shock waves arise.
226 10 Magnetohydrodynamics (MHD)

10.6 Alfven Masers

When plasma is restricted by boundaries in the parallel direction, standing Alfven

waves are formed. Such an open magnetic trap is formed in the Earth’s magneto-
sphere with relatively hot collisionless plasma (Fig. 10.8). Alfven waves are
reflected from the ionosphere with high perpendicular conductivity caused by
ion-neutral collisions. Therefore, the perpendicular inductive electric field and the
perpendicular magnetic field at the level of the ionosphere are small and could be put
to zero under imposed boundary conditions, so a standing Alfven wave is formed.

Fig. 10.8 Scheme of Alfven maser: (a) magnetosphere flux tube, (b) open magnetic trap in the
laboratory
10.6 Alfven Masers 227

Similar magnetic traps exist in the laboratory. Particles in such magnetic traps
(magnetic belts in the Earth and planet magnetospheres) form a strongly
nonequilibrium distribution function. Indeed, at the ends of a magnetic tube, per-
pendicular particle velocity increases due to adiabatic invariant conservation, and
since full energy is also conserved, particles with small parallel velocities are
reflected back to the trap. As a result, loss cones are formed in the velocity space,
which becomes a source of various electromagnetic instabilities. Alfven waves in the
“resonator” interact with sources of particles, and a maser is created. Similar
standing Alfven waves interacting with fast particles are also generated in tokamaks
in the vicinity of resonant flux surfaces, where the magnetic field line after a few
turns over tokamak returns to its initial position.
Chapter 11
Dynamics of Plasma Blobs and Jets
in a Magnetic Field

11.1 Plasma Motion Across Magnetic Field in Vacuum

Consider a plasma blob moving at t = 0 across the magnetic field along the x-axis
→
with velocity u 0 , as shown in Fig. 11.1. From the momentum balance equation for
electrons, it follows that such motion is caused by an induced electric field in the y-
direction, and
→ →
→
E0× B0
u0= : ð11:1Þ
B2
→ →
In other words, the blob motion is E × B drift produced by blob polarization. To
analyze the further motion of the blob, let us consider the sum of the momentum
balance Eq. (10.2), electron inertia is neglected. Multiplying both parts of Eq. (10.2)
→
by B × , we obtain for current density

→ → →

→ B × ∇p B × nmi ddtu i
j = þ : ð11:2Þ
B2 B2

The first term on the r.h.s. is a diamagnetic current, which is divergent free in a
homogeneous magnetic field. The second term corresponds to inertial or polarization
current. If the blob as a whole is decelerated, this current is directed downwards.
However, this contradicts the quasineutrality condition

→
∇  j = 0: ð11:3Þ

Hence, vertical current should be zero, and, correspondingly, inertia should also turn
→
to zero. The blob moves along the x-axis with constant velocity u 0 given by
Eq. (11.1). Considering the forces applied to each particle, the Lorentz force eu0B,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 229
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_11
230 11 Dynamics of Plasma Blobs and Jets in a Magnetic Field

Fig. 11.1 Polarization of a blob moving across the magnetic field

Fig. 11.2 Magnetized plasma blob in the presence of gravitational force

which shifts particles in the vertical direction, is balanced by the opposite vertical
→
electrostatic force e E 0 . Strictly speaking, this statement is correct for a blob
extended in the direction of motion х when the electric field inside the blob is
practically homogeneous. For blobs with comparable sizes in the x and
→
y directions, in addition to motion with velocity u 0 , blob deformation takes place
(see below).
In the presence of the force of gravity, the magnetized plasma blob obtains
additional polarization, as shown in Fig. 11.2. Indeed, ion drift caused by gravity
11.1 Plasma Motion Across Magnetic Field in Vacuum 231

→ →
force in the direction mi g × B generates current in the negative direction of the x-
axis:

nmi g
jgx = - : ð11:4Þ
B

The contribution of electrons can be neglected since their drift is mi/me times smaller
than that of ions. Current jgx leads to blob polarization. The polarization electric field
should produce current jp in the opposite direction, which should balance current jgx:

jpx = - jgx : ð11:5Þ

In a fully ionized plasma and uniform electric field, such a current is a polarization or
inertial current caused by the electric field changing over time:

→
→ nm ∂ E
j p = 2i : ð11:6Þ
B ∂t
→
This expression corresponds to the second term in Eq. (11.2), if one puts u i =
→ → → →
E × B =B2 and d u i =dt = ∂ u i =∂t. From Eqs. (11.4) to (11.6), one finds that the
electric field in the x-direction grows linearly with time

E x = Bgt: ð11:7Þ
→ →
Increasing electric field produces E × B drift downwards in the direction of the
gravitational force with the velocity

Ex
Vy = - = - gt, ð11:8Þ
B
→ →
which corresponds to blob motion with acceleration a = g . As in the previous case,
electric field inhomogeneity leads to its additional deformation.
Blob motion in a nonuniform magnetic field is similar to that caused by gravita-
tion. Let us consider as an example the magnetic field of a torus (tokamak) B = B0R0/
R, Fig. 11.3. In such a magnetic field, ionized particles drift vertically with the
velocity

V 2j⊥
mj 2 þ V 2jk
V jy = , ð11:9Þ
ej BR

where Vk, ⊥ are single particle velocities. The first term here is caused by a force
applied to the Larmor orbit trying to push it out to the weaker magnetic field, and the
second term is connected with the centrifugal force in the curvilinear magnetic field.
232 11 Dynamics of Plasma Blobs and Jets in a Magnetic Field

Fig. 11.3 Blob acceleration in nonuniform magnetic field

Electrons and ions drift in opposite directions due to different charge signs, so a
vertical current arises. To obtain the expression for this current, one should perform
averaging over velocities using the Maxwellian distribution function:

* 2nðT e þ T i Þ
jy = e i - V ey f e dV =
V iy f M M
: ð11:10Þ
BR

It is worth noting that the divergence of this current caused by orbit leading centers
vertical drifts coincides with the divergence of the diamagnetic current, given by the
first term in Eq. (11.2). The current (11.10) should be compensated by the opposite
current Eq. (11.6). As a result, the vertical electric field increases with time. This
→ →
field produces E × B outwards drift, so the blob is accelerated in the direction
opposite to the magnetic field gradient, and the acceleration is

2ðT e þ T i Þ
g= : ð11:11Þ
mi R

In the general case, the blob evolution is governed by equation system (11.2) and
(11.3). In the first approximation, the electron and ion velocities coincide since the
→ → * → → →
drift is E × B : u e = u i = E × B =B2 . Let us introduce dimensionless variables

x = x=a; y = y=a; t = t ðg=aÞ1=2 ; φ = cφ= Ba3=2 g1=2 , ð11:12Þ

where a is the characteristic blob size, φ is the electrostatic potential, and acceler-
ation g is defined according to Eq. (11.1). Then, from Eqs. (11.2) and (11.3), one
obtains
11.1 Plasma Motion Across Magnetic Field in Vacuum 233

y
2,5

-0,5

-5,5
x
-3,5 4,5 12,5

Fig. 11.4 Lines of constant densities in nonuniform magnetic field, t = 4

∂φ ∂∇φ ∂ ln n
þ J ðφ, ΔφÞ þ þ J ðφ, ∇φÞ ∇ ln n þ = 0, ð11:13Þ
∂t ∂t ∂y

where the operator J is defined as

∂ α ∂β ∂β ∂α
J ðα, βÞ = - :
∂x ∂y ∂x ∂y

The particle balance equation in the dimensionless variables is reduced to

∂n
þ J ðφ, nÞ = 0: ð11:14Þ
∂t

Here, it is taken into account that in the homogeneous magnetic field

→ →
∇  E × B =B2 = 0, and corrections caused by derivatives of the magnetic field
are small. Eq. (11.14) describes plasma motion along equipotentials. Numerical
modeling of the system Eqs. (11.13) and (11.14) demonstrates that the blob moves
in the x direction with acceleration given by Eq. (11.11), and its shape resembles a
mushroom since its upper and lower parts move in the opposite direction in the
dipole electric field (Fig. 11.4).
234 11 Dynamics of Plasma Blobs and Jets in a Magnetic Field

→
Fig. 11.5 Plasma jet moves across magnetic field, at t = 0 the velocity is u 0

11.2 Deceleration of the Plasma Jet by Ambient Plasma

Consider the dynamics of a plasma jet injected into the ambient magnetized plasma.
For simplicity, let us assume that the jet is infinite in the x-direction, and, at t = 0, the
jet moves with a velocity u0 across the magnetic field, as shown in Fig. 11.5. Plasma
density in the jet nI(y, z) is supposed to be much larger than the ambient plasma
density ni, and jet ions mI could differ from ambient ions mass mi. As discussed in
the previous section, plasma injected in the x-direction is polarized in the y-direction
→
so that electric field E 0 , which satisfies Eq. (11.1). However, in the ambient plasma,
the electric field propagates along the magnetic field as an Alfven wave. At the front
of the Alfven wave, the electric field increases, and, therefore, a polarization current
in the y-direction is generated. This current balances the polarization current inside
the jet in the -y-direction caused by jet deceleration. The scheme of currents is
shown in Fig. 11.6.
Let us find the jet deceleration law for the case u0 ≪ cA, where cA = B/(2μ0nimi)1/2
is the Alfven velocity of the ambient plasma. If this condition is satisfied, the
perturbed magnetic field in the Alfven wave is small: Bx~uxB/cA ≪ B, and the
magnetic field can be considered constant. Alfven wave propagation outside the
jet is described by the wave equation

2 2
∂ ux ∂ u B2
= c2A 2x ; c2A = : ð11:15Þ
∂t 2 ∂z 2μ0 mi ni

Inside jet analogously

2 2
∂ ux B2 ∂ ux
nI mI = : ð11:16Þ
∂t 2 2μ0 ∂z2

Let us integrate Eq. (11.16) along z from -z0 to z0, where z0 is chosen to satisfy
11.2 Deceleration of the Plasma Jet by Ambient Plasma 235

Fig. 11.6 Scheme of currents in the ambient plasma during jet injection across magnetic field for
z > 0. At negative z,the current pattern is symmetrical

ak < z0 ≪ LA = cA t,

where ak is the jet characteristic size along the magnetic field. At this distance,
plasma velocity ux can be considered independent of z since it changes at a much
larger scale LA. Integration yields

2 z0
∂ ux B2 ∂ux
MI = , ð11:17Þ
∂t 2 2μ0 ∂z - z0

where
z0 1

MI = mI nI dz ≈ mI nI dz ð11:18Þ
- z0 -1

is the integral mass of injected ions per unit length. Using symmetry, we consider
only positive values of z. For z0 → 0, we find

2
∂ ux B2 ∂ux
MI = : ð11:19Þ
∂t 2 z=0
μ0 ∂z z=0
236 11 Dynamics of Plasma Blobs and Jets in a Magnetic Field

Fig. 11.7 Velocity profile of the ambient plasma at fixed moment

This expression should be considered as a boundary condition for Eq. (11.15), which
is the equation for Alfven wave propagation in the ambient plasma. The initial
condition for Eq. (11.15) is:

u0 , z=0
ux = : ð11:20Þ
0, z>0

Since the solution of wave Eq. (11.15) is ux = ux(z - cAt), considering the initial
and boundary conditions, one obtains

B2
ux = u0 exp ð z - cA t Þ z ≤ cA t;
μ0 M I c2A : ð11:21Þ
ux = 0 z ≥ cA t :

Velocity profile given by Eq. (11.21) is shown in Fig. 11.7. One can see that ambient
plasma is dragged by the plasma jet in the x-direction at the scale zτ = cAt. The
wavefront propagates along the magnetic field with the Alfven velocity. The jet
velocity coincides with the ambient plasma velocity z = 0 and exponentially
decreases with time:

μ 0 M I cA
ux ð0, t Þ = u0 expð- t=τA Þ, τA = : ð11:22Þ
B2

Characteristic time τA can be interpreted as follows. The net mass of the ambient
plasma per unit length dragged during time τA is (with account of positive and
negative values of z)
11.2 Deceleration of the Plasma Jet by Ambient Plasma 237

M i ðτA Þ = 2zτ ni mi = M I : ð11:23Þ

In other words, at the moment τA, the integral mass of the ambient plasma dragged
by the jet equals the jet integral mass. At the same time, the jet velocity decreases
considerably with respect to the initial velocity.
The electric field that causes plasma drift in the x-direction up to distances
z = ± cAt, according to Eq. (11.21), decreases exponentially with time. Hence, a
negative polarization current flows in the y-direction, as shown in Fig. 11.6. Integral
current (for positive values of z) is obtained by integration along the magnetic field

cA t
ðnI mI þ ni mi Þ ∂E y E y ð z = 0Þ
I = Ic þ I0 = dz = - , ð11:24Þ
B 2 ∂t μ0 cA
0

where Ey(z = 0) = u0B exp (-t/τA). Net current consists of two equal contributions.
The first one Ic is the integral over the jet, during the integration, the electric field Ey
is taken at z = 0, and the integral over nI(z)mI gives MI/2. The second contribution I0
is obtained from the integral over ambient plasma with account of Eq. (11.21) and
identity Ey = uxB. This current flows in the region 0 < z < cAt. The net negative
current I should be compensated by the positive current in the y-direction. Therefore,
at the Alfven wavefront, where ambient plasma is accelerated from zero velocity up
to value u0, a positive current IF = - I flows. Now let us separate integral currents in
a different way. As a first contribution, we shall now take the integral current of the
jet

E y ð z = 0Þ
Ic = - : ð11:25Þ
2μ 0 cA

The second contribution from the ambient plasma is the sum of two currents,
including the current at the Alfven wavefront:

E y ð z = 0Þ
IW = Ic þ IF = ð11:26Þ
2μ 0 cA

These two net currents gain balance each other. The second current in the ambient
plasma can be rewritten as

1
IW = E y ð z = 0Þ ; = , ð11:27Þ
W W 2μ0 cA

where the quantity ΣW is known as the Alfven conductivity.

In other words, ambient plasma with a propagating Alfven wave is equivalent to a
resistor, according to Eq. (11.27) (one has to keep in mind that the same current
238 11 Dynamics of Plasma Blobs and Jets in a Magnetic Field

flows at negative values of z). The paradigm of Alfven conductivity can be used in
various problems of plasma physics.
All solutions discussed above are applicable in low collisional plasma when the
skin time scale is sufficiently large. The corresponding condition is

τs = 2μ0 σa2⊥ ≫ τA , ð11:28Þ

with a⊥ being the perpendicular jet size. In the opposite case, the electric field is
electrostatic, and spreading of the polarization in the ambient plasma has a diffusive
character with diffusion coefficient D  c2A τs .

11.3 Edge Localized Modes and Filaments

In edge tokamak plasma with a strong density gradient, special edge localized modes
(ELMs) can develop. This phenomenon is typical for high confinement regimes,
Chap.16, when the pressure gradient near the separatrix exceeds a critical value,
determined by the threshold for some of the instabilities, such as peeling-ballooning
and kinetic ballooning. As a result of the nonlinear development of the instability,
plasma is split into several blobs or filaments extended in the magnetic field
direction. Photographs of such filaments for the MAST tokamak are shown in
Fig. 11.8. After formation, filaments move radially across the separatrix and further
away towards the major radius, i.e., in the direction opposite to the magnetic field
gradient. The perpendicular size of a single filament is of the order of a few cm, while
its parallel dimension is controlled by plasma expansion along the magnetic field line
analyzed in Chap.9.

Fig. 11.8 Filament

structure during ELMs
11.3 Edge Localized Modes and Filaments 239

∇B
----
E
jy
-
j||
++++

j||+
jw

Fig. 11.9 Currents short-circuiting scheme through divertor targets

Outside of the separatrix, the plasma blob is polarized due to the ∇B driven
vertical current given by Eq. (11.10). In contrast to the situation in vacuum consid-
ered above in Sect.11.1, this vertical current could be short-circuited by parallel
currents in the ambient plasma, which in turn are closed through divertor targets
(Fig. 11.9). The resulting polarization is smaller than that in vacuum, so the
compensating polarization current can be neglected. Parallel currents required to
close the ∇B driven vertical current could be obtained from the current continuity
→
equation ∇  j = 0 by integrating it over the parallel filament dimension. From
Eq. (11.10), we obtain

2nðT e þ T i Þ lk
jk± = , ð11:29Þ
BR l⊥

where lk and l⊥ are the parallel and perpendicular half-widths of the filament (the
perpendicular pressure gradient is replaced here by p/l⊥). This is the absolute value
of the pair of currents, from the blob to the target and from the target to the blob
(Fig. 11.9). Since this current should flow through the sheath to the target, the
potential at the sheath edge (and at the filament for infinite parallel conductivity of
the ambient plasma) should be larger than the floating potential for current to the
target and smaller than the floating potential for current from the target. According to
the current-voltage characteristic of the sheath in Eq. (3.26), one obtains
240 11 Dynamics of Plasma Blobs and Jets in a Magnetic Field

1=2 1=2
T es þ T is 1 T es eφþ
jþ
k = ens -p exp - :
mi 2π me T es
ð11:30Þ
1=2 1=2
T es þ T is 1 T es eφ
jk- = - ens -p exp - :
mi 2π me T es

Here, subscript “s” corresponds to plasma parameters at the sheath edge. It is

assumed that the parallel current does not exceed the saturation ion current given
by the first term on the r.h.s., otherwise short-circuiting through the target is not
possible. Eqs. (11.29) and (11.30) determine the potentials at the sheath edge, and
the same potentials are assumed at the blob level. The vertical electric field of the
blob is

E y = ðφþ - φ - Þ=l⊥ , ð11:31Þ

which produces drift in the direction of the major radius

ux = E y =B = ðφþ - φ - Þ=Bl⊥ : ð11:32Þ

For edge tokamak plasma parameters, its value could be in the order of 1 km/s,
which is consistent with experimental observations. Very dense filaments that ∇B-
driven vertical current could not be short-circuited should still be accelerated, as
described in Sect.11.1.
Chapter 12
Plasma Equilibrium

We shall define equilibrium as a plasma state with flow velocities smaller than sound
or Alfven velocities. In such a situation, it is possible to neglect inertia terms in
Eq. (10.2) for the total momentum balance with respect to the pressure gradient to
obtain

→ →
∇p = j × B : ð12:1Þ

This is the equation for plasma equilibrium. Accordingly, the pressure gradient is
balanced by the Lorentz force caused by plasma currents and self-consistent mag-
netic fields. Plasma equilibrium Eq. (12.1) can be combined with the Maxwellian
equations

→ 1 → →
j = ∇× B , ∇  B = 0: ð12:2Þ
μ0

Substituting the current into Eq. (12.1) yields

B2 1 → →
∇p = - ∇ þ B∇ B: ð12:3Þ
2μ0 μ0

The first term on the r.h.s. can be interpreted as the magnetic pressure gradient, and
the second can be interpreted as the magnetic tension associated with the curvature
of the magnetic field lines. Equations (12.1) and (12.2) or Eq. (12.3) are the basic
equations for equilibrium analysis. Let us emphasize once again that plasma flows
are still present in plasma.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 241
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_12
242 12 Plasma Equilibrium

12.1 On the Possibility of Equilibrium in the Absence

of a Vacuum Magnetic Field

Let us demonstrate that equilibrium cannot be obtained by internal currents flowing

in plasma in the absence of an external magnetic field produced by currents in the
coils. To prove this, let us assume the opposite, so that Eq. (12.3) is supposed to be
satisfied with internal currents only. We introduce the magnetic stress tensor

B2 1
πik = pδik þ δ - BB : ð12:4Þ
2μ0 ik μ0 i k
→
With an account of ∇  B = 0, Eq. (12.3) can be reduced to

∂πik
= 0: ð12:5Þ
∂xk

Let us introduce an integral over an infinite volume, including the region with
plasma and the region outside plasma, which after integration by parts is:

∂πik ∂xi
xi dV = xi πik dSk - πik dV: ð12:6Þ
∂xk ∂xk

In accordance with Eq. (12.5), l.h.s. of this equation is zero. Hence, since ∂xi/
∂xk = δik, one finds

πii dV = xi πik dSk : ð12:7Þ

After substitution of the magnetic stress tensor πik in Eq. (12.4), one obtains

→* →
B2 B2 →
rB B →
3p þ dV = pþ r - dS: ð12:8Þ
2μ0 2μ0 μ0

The l.h.s. of this equation is finite and a positive quantity. On the r.h.s., the integral is
taken over the surface limiting infinite volume. The plasma pressure at this surface
equals zero. If a magnetic field is created by plasma currents, at infinity, it should
decay as r-3 (dipole field) or faster. Therefore, the surface integral on the r.h.s. of
Eq. (12.8) becomes zero. Therefore, one comes to a contradiction—Eq. (12.8)
cannot be satisfied, and hence the assumption that Eqs. (12.3) and (12.5) are
incorrect. Finally, one can conclude that to keep plasma in the equilibrium state, it
is necessary to create a magnetic field by currents flowing in the coils, where
→ →
mechanical forces compensate for j × B forces.
12.2 Equilibrium of a Pinch 243

12.2 Equilibrium of a Pinch

Consider an infinite plasma cylinder in a magnetic field with the cylinder axis
parallel to the magnetic field. Current along the cylinder axis can produce an
additional azimuthal magnetic field. Such a plasma cylinder is known as a pinch.
All pinch plasma parameters are independent of the longitudinal coordinate z and
azimuthal angle θ and are functions of the radial coordinate r only. Maxwellian
→
equation ∇  B = 0 in cylindrical coordinates in the absence of z and θ dependence
is reduced to equation r-1d(rBr/dr) = 0, from which it follows Brr = const. Since the
magnetic field cannot go to infinity at r = 0, its radial component should be zero:
Br = 0. Therefore, only longitudinal and azimuthal components of magnetic fields
exist in the pinch, which are correspondingly produced by azimuthal and longitudi-
→ →
nal currents. From the projections of the Maxwellian equation ∇ × B = μ0 j , we
have

1 dBz 1 dðrBθ Þ
jθ = - , jz = : ð12:9Þ
μ0 dr μ0 r dr

Here, we use identity

→ eijk ∂ðAk hk Þ
∇× A = :
i hj hk ∂xj

Quantity eijk is the Levi-Civita symbol, and each component changes sign when any
two indices are interchanged. It equals ±1 for different indices and zero for coin-
ciding indices. Lame coefficients for cylindrical coordinates are hr = 1, hθ = r, hz = 1.
Equilibrium equation for pinch given by the radial balance

dp
= jθ Bz - jz Bθ : ð12:10Þ
dr

After substituting the current density from Eq. (12.9), one obtains

2
dp 1 dBz B d ðrBθ Þ
þ þ θ = 0: ð12:11Þ
dr 2μ0 dr μ0 r dr

The azimuthal magnetic field is connected with the current flowing inside a surface
of a given radius:

r
μ I ðr Þ
Bθ = 0 , I ðr Þ = jz 2πrdr:
2πr
0
244 12 Plasma Equilibrium

Therefore, Eq. (12.11) can also be written in the form

2
dp 1 dBz μ dI 2
þ þ 20 2 = 0: ð12:12Þ
dr 2μ0 dr 8π r dr

Equation (12.11) or (12.12) determines pinch equilibrium in the magnetic field.

Let us now consider special cases of equilibrium. Pinch without parallel current is
called θ-pinch. In the θ-pinch jz = 0 and Bθ = 0, the equilibrium equation after
integration over the radius is reduced to

B2z B2
pþ = 0 : ð12:13Þ
2μ0 2μ0

Here, B0 is the magnetic field outside the plasma. This equation can be satisfied only
for small parameter β,

2μ0 p
β= < 1: ð12:14Þ
B20

In the opposite case β > 1, there is no equilibrium since the plasma pressure cannot
be balanced by the Lorentz force, and the plasma expands radially with a velocity of
the order of sound speed as without a magnetic field (see Сhap. 9). In contrast, when
the condition Eq. (12.14) is satisfied, equilibrium is established automatically by the
diamagnetic current, which arises in the nonuniform plasma due to particle rotation
over Larmor orbits. The diamagnetic current reduces the magnetic field inside the
plasma column, so Eq. (12.14) is satisfied.
In contrast to the θ-pinch, in a z-pinch, a longitudinal magnetic field is absent,
Bz = 0, and hence, a diamagnetic current is also absent, jθ = 0. The plasma pressure
is balanced by the Lorentz force caused by the interaction of the longitudinal current
and the azimuthal magnetic field produced by the current. According to Eq. (12.12)

dp μ dI 2
- = 20 2 : ð12:15Þ
dr 8π r dr

Let us multiply both parts of Eq. (12.15) by πr2 and integrate over r from zero to
pinch radius a. On the l.h.s., integration by parts yields

a a
dp
πr
2
dr = - 2πrpdr:
dr
0 0

As a result, one obtains the integral equilibrium equation

12.3 Magnetic Flux Surface Functions 245

a
μ 0 I 2 ð aÞ
= 2πrpdr = hpiπa2 , ð12:16Þ
8π
0

which connects pinch net current and average plasma pressure. Therefore, for
equilibrium, the net current in the pinch should be organized in accordance with
Eq. (12.15) or (12.16).

12.3 Magnetic Flux Surface Functions

The magnetic flux surface is a key concept for fusion devices. We shall consider here
a system of nested toroidal flux surfaces. The magnetic field line, which belongs to
the flux surface, either comes to a starting point and closes itself after several turns in
the poloidal and toroidal directions or fills ergodically all flux surfaces. In the last
case, the magnetic field line can be found in the infinite neighborhood of any given
flux surface point. Flux surfaces where the magnetic field line closes itself after
several turns are known as resonant flux surfaces. In a tokamak that is toroidally
symmetric, all quantities are independent of the toroidal angle. Examples of various
flux surfaces are shown in Fig. 12.1.
At the edge, magnetic field lines become open and go to material surfaces. The
magnetic flux surface that separates regions of different topologies is called the
separatrix. Inside the separatrix, nested flux surfaces embrace the field line, which is
called the magnetic axis. In addition to toroidally symmetric magnetic configura-
tions, there are more complicated magnetic traps, such as stellarators with helical
symmetry. Magnetic islands in tokamaks, which can emerge in the vicinity of
rational flux surfaces, also have helical symmetry, and they will be considered
later. Meanwhile, we shall restrict ourselves by toroidally symmetric nested flux
surfaces.
To describe equilibrium, it is convenient to introduce flux surface function
quantities depending only on the flux surface and changes from one flux surface to
another. The first flux surface function is a toroidal magnetic flux ΨT through the
torus cross-section,

→ →
ΨT = Bd S , ð12:17Þ
ST

where the integral is taken over the toroidal cross-section, as shown in Fig. 12.2.
→
Since ∇  B = 0, according to the Gaussian theorem, the magnetic flux through the
closed surface is zero. Hence, the magnetic flux through one toroidal cross-section is
equal to the flux through another toroidal cross-section because together with the
flux surface, they form a closed surface, and the magnetic flux through the flux
246 12 Plasma Equilibrium

Fig. 12.1 Cross-sections of magnetic flux surfaces corresponding to fixed toroidal angle

SP

ST

Fig. 12.2 Toroidal and poloidal cross-sections of a torus

surface is zero since the normal to the flux surface component of the magnetic field is
absent according to the flux surface definition.
Poloidal flux is a magnetic field flux through a surface inside the torus,
Fig. 12.2—through the “hole of the bagel”:

→ →
Ψp = Bd S, ð12:18Þ
Sp

Figure 12.3 shows different variants of poloidal surface choice. The magnetic flux
through each of the poloidal cross-sections is the same since two cross-sections with
corresponding parts of the flux surfaces form a closed surface with zero magnetic
field flux through it. Hence, the poloidal flux is also the flux surface function of the
toroidal flux.
12.3 Magnetic Flux Surface Functions 247

Fig. 12.3 Different variants of poloidal surface choice

The total plasma pressure p is the third flux surface function. Indeed, since the
→
equilibrium inertia terms are neglected, condition B ∇p = 0 is satisfied, which
means that pressure should be constant along the magnetic field line and, hence,
should be constant over the whole flux surface. Note, however, that partial ion and
electron pressures are not flux surface functions.
→ →
From Eq. (12.1) ∇p = j × B , it follows that since ∇p is directed normally to
the flux surface, there is no current normal to the flux surface, and current flows only
along the flux surface (this statement is valid only provided viscosity is neglected).
However, if dissipation and inertia are neglected, current lines belong to the flux
surface as magnetic field lines. Since the current density current continuity equation
→
∇  j = 0 is valid, the net current through the closed flux surface is zero, similar to
the magnetic field flux. Hence, accounting for the current through parts of the
magnetic flux surface being zero, one can introduce another two flux surface
functions:

→ →
IT = j dS ð12:19Þ
ST

net toroidal current inside a given flux surface, and

→ →
Ip = j dS ð12:20Þ
Sp

net poloidal current through the poloidal cross-section.

Safety factor q is another quantity that is defined at a given flux surface. For a
closed magnetic field line, q is defined as the ratio of the number of turns in the
toroidal direction n to the number of turns in the poloidal direction m until it comes to
the starting point: q = n/m. For the ergodic magnetic field line, the limit for infinite
turns should be taken:
248 12 Plasma Equilibrium

d P

Fig. 12.4 Toroidal and poloidal fluxes between two neighboring flux surfaces. Here φ is a toroidal
angle, θ is a poloidal angle, and the third coordinate is the normal to flux surface

n
q = lim : ð12:21Þ
n→1 m

By definition, the safety factor q is the flux surface function. Now let us demonstrate
that the safety factor equals the derivative

dΨT
q= - : ð12:22Þ
dΨp

To prove this, we start with special case q = 1. Consider the volume between two
neighboring flux surfaces, which we unfold into a region between two neighboring
planes (Fig. 12.4). Toroidal angle φ and poloidal angle θ change in the range (0, 2π),
so that the left and right faces of the parallelepiped are glued with each other. The
third “radial” coordinate is perpendicular to the planes. Since after one poloidal and
toroidal turn the magnetic field line comes to a starting point, it should belong to a
dashed surface (not necessary plane), Fig. 12.4. Magnetic flux through this surface is
absent, as well as magnetic flux through near and far parts of flux surfaces. There-
fore, considering a volume restricted by a dashed diaphragm, near and far parts of
flux surfaces, lower and right faces, one can conclude that magnetic flux through the
low face equals flux through the right face: dΨT = - dΨp. Minus represents the fact
that while moving from inner to outer flux surfaces, toroidal flux rises and poloidal
flux decreases. Hence, for q = 1, Eq. (12.22) is satisfied. For q = n/m, the toroidal
flux through the end face increases by factor n corresponding to the number of
magnetic field line intersections before returning to the starting point, while the
poloidal flux increases by factor m in accordance with the number of intersections
through the low face. Therefore, Eq. (12.22) remains valid in this situation. Finally,
for the ergodic field line in the limit of an infinite number of turns, one also obtains
Eq. (12.22).
12.4 Grad-Shafranov Equation 249

12.4 Grad-Shafranov Equation

The shape of flux surfaces for the case of toroidal symmetry can be described by an
equilibrium equation where all quantities are flux surface functions. For analysis, we
shall consider cylindrical geometry with coordinates (R, φ, z) shown in Fig. 12.5.
Coordinate R represents the major radius, and φ is the toroidal angle – all quantities
are independent of this angle. The point belonging to the flux surface is, therefore,
determined by a pair of coordinates (R, z). The magnetic field has two components:
→ → → → →
the poloidal field B p and the toroidal field B T , B = B p þ B T . The poloidal
→ → →
magnetic field in its turn consists of two components: B p = B z þ B R . The poloidal
magnetic flux Ψp by definition is

Ψp = - 2πR0 Bz dR0 : ð12:23Þ

0

After taking the derivative, we have the relation

1 ∂Ψp
Bz = - : ð12:24Þ
2πR ∂R
→
From equation ∇  B = 0, which in cylindrical coordinates is given by

∂Bz 1 ∂
þ ðRBR Þ = 0, ð12:25Þ
∂z R ∂R

Fig. 12.5 Cylindrical coordinate system. R– major radius, φ– toroidal angle

250 12 Plasma Equilibrium

one can obtain the radial component of the magnetic field as a function of the
poloidal flux:

R
1 ∂Bz 0 0 1 ∂Ψp
BR = - R dR = : ð12:26Þ
R ∂z 2πR ∂z
0

Analogously, for poloidal current Ip:

Ip = - 2πR0 jz dR0 : ð12:27Þ

0

Taking the derivative, similar to Eq. (12.24), one finds

1 ∂I p
jz = - : ð12:28Þ
2πR ∂R
→
Since ∇  j = 0, analogously to Eq. (12.26):

R
1 ∂jz 0 0 1 ∂I p
jR = - R dR = : ð12:29Þ
R ∂z 2πR ∂z
0

The toroidal magnetic field can be derived from the integral Maxwellian equation as
a function of the poloidal current:

μ0 I p
Bφ = - : ð12:30Þ
2πR

Toroidal current density

1 → 1 ∂BR ∂Bz
jφ = ∇× B = - =
μ0 φ μ0 ∂z ∂R
2 ð12:31Þ
1 ∂ Ψp ∂ 1 ∂Ψp 1
= þR = Δ  Ψp :
2πμ0 R ∂z2 ∂R R ∂R 2πμ0 R

Here, operator Δ resembles the Laplace operator:

2
∂ ∂ 1 ∂
Δ = þR :
∂z2 ∂R R ∂R

The radial component of the equilibrium equation reads

12.4 Grad-Shafranov Equation 251

∂p
= jφ Bz - jz Bφ : ð12:32Þ
∂R

Let us substitute the components of the magnetic field and current densities
Eqs. (12.24), (12.28), (12.30), and (12.31) to the r.h.s. and use relation ∂p/
∂R = (dp/dΨp)(∂Ψp/∂R) on the l.h.s. After dividing by ∂Ψp/∂R, one obtains

2
dp μ2 dI p
Δ  Ψp = - μ 0 R 2 - 02 : ð12:33Þ
dΨp 8π dΨp

This equation is known as the Grad-Shafranov equation. Its solution is a function

Ψp(R, z), and equality Ψp(R, z) = const determines the equation for magnetic flux
coordinates. The Grad-Shafranov equation can be solved if functions p(Ψp) and
Ip(Ψp) are specified. These functions determine the flux surface shape. The outside
plasma Grad-Shafranov equation is

Δ Ψp = 0: ð12:34Þ

At the plasma, boundary solutions of Eq. (12.33) Ψpi and Eq. (12.34) Ψpe should
coincide as well as their normal derivatives:

∂Ψpi ∂Ψpe
Ψpi s = Ψpe s , = : ð12:35Þ
∂n s ∂n s

It is convenient to calculate the poloidal flux from the magnetic axis,

Ψ = Ψ0p - Ψp , ð12:36Þ

where Ψ0p is the poloidal flux corresponding to the magnetic axis. Poloidal flux Ψ
turns to zero at the magnetic axis and rises towards the separatrix. It also satisfies the
Grad-Shafranov equation.
Let us consider a simple example of equilibrium. Let us put dp/dΨ = const and
dI2/dΨ = const, with constants chosen in the following way:

dp aΨ μ20 dI 2 bΨ
μ0 = - 40 , = - 20 :
dΨ R0 8π2 dΨ R0

Then simple solution of this equation has the form

Ψ 1 a-1 2 2
R40 = bR20 þ R2 z2 þ R - R20 :
Ψ0 2 8

At R = R0 and z = 0 poloidal flux Ψ = 0, this corresponds to the magnetic axis. In the

vicinity of the magnetic axis at R ≈ R0
252 12 Plasma Equilibrium

2
Ψ 1 z2 ða - 1Þ ðR - R0 Þ
= ð b þ 1Þ 2 þ : ð12:37Þ
Ψ0 2 R0 2 R20

Therefore, the magnetic flux surfaces in this example have the form of ellipsoids
with semiaxes a and b.

12.5 Integral Equilibrium in a Tokamak

For better understanding, let us consider a tokamak with a circular cross-section. Let
the last closed flux surface have radius a. We shall also restrict ourselves to the case
of a large aspect ratio.

a
≪ 1: ð12:38Þ
R

Here, R is the major radius (e.g., of the center of the last closed flux surface). If
condition Eq. (12.38) is satisfied, the torus can be in zero approximation considered
as a cylinder, and toroidal effects can be considered as small corrections.
Let us analyze forces applied to plasma as a whole. In the zero (cylindrical)
approximation, forces are directed towards the cylinder center along the minor
radius, the so-called minor radius equilibrium. In this approximation, Eq. (12.11)
or (12.12). is satisfied. After integrating Eq. (12.12) with weight r2 over a cylinder
with radius a using integrating by parts, one finds

B2eφ B2iφ μ0 I 2T ðaÞ

hpi = - þ : ð12:39Þ
2μ0 2μ0 8π2 a2

Here

a a
1 B2iφ 1 B2iφ
h pi = 2 2πrpdr, = 2 2πr dr: ð12:40Þ
πa 2μ0 πa 2μ0
0 0

Magnetic field Biφ is a toroidal magnetic field in plasma that corresponds to magnetic
field Bz in a cylinder, and field Beφ is a vacuum magnetic field outside plasma
(corresponding to field B0 in Eq. (12.12)), IT is the net toroidal current. Equation
(12.39) is an integral equilibrium condition along the minor radius.
Since the torus surface element at the outer side is larger than that at the inner side,
the forces also arise in the direction of the major radius in contrast to the cylinder
case. These forces contain a small parameter a/R. The integral balance of these forces
is called equilibrium along the major radius. To calculate the forces along the major
radius, we consider the virtual shift of the plasma along the major radius and
12.5 Integral Equilibrium in a Tokamak 253

calculate the work A and potential energy W = - A of the plasma pinch. Then, the
force along the major radius would be F = - ∂W/∂R.
The potential energy and corresponding force are associated with three different
factors:

W = W p þ W I þ W B: ð12:41Þ

The first part of the potential energy Wp is connected with the plasma pressure and
difference in surface elements at the outer and inner parts of the torus. When the
plasma pinch is shifted along the major radius, its volume increases, and work is
performed. The corresponding potential energy is

Wp = - pdV = - 2π2 a2 Rhpi, ð12:42Þ

and the force is

∂W p
Fp = - = 2π2 a2 hpi, ð12:43Þ
∂R

The second force is caused by the currents interaction flowing in the toroidal
direction, or, in other words, by the interaction of the toroidal currents with the
poloidal magnetic field: jφBp. The potential energy is given by

LI 2T ðaÞ
WI = , ð12:44Þ
2

with L being the inductivity of the plasma pinch. For torus

L = μ0 R½lnð8R=aÞ - 2 þ li =2, li = B2p =B2p ðaÞ: ð12:45Þ

Here, Bp(a) is a poloidal magnetic field at the plasma boundary, and averaging is
performed similar to Eq. (12.40). When shifting along the major radius, magnetic
flux Ψ(a) = ITL is conserved while the current is changing. Therefore, it is more
convenient to rewrite the potential in the form

Ψ2 ðaÞ
WI = , ð12:46Þ
2L

The corresponding force

∂W I μ I2 8R l
FI = - jΨ = const = 0 T ln -1 þ i : ð12:47Þ
∂R 2 a 2
254 12 Plasma Equilibrium

The third force is caused by the interaction of the poloidal current with the
toroidal magnetic field jpBφ. The corresponding potential energy is

B2iφ B2eφ
WB = dV þ dV: ð12:48Þ
2μ0 2μ0
Vi Ve

Integrals are taken over plasma volume and outside it correspondingly. After
regrouping different terms, one obtains

B2eφ B2iφ B2eφ

WB = dV þ - dV = const
2μ0 2μ0 2μ0
V e þV i Vi

B2iφ B2eφ
þ 2π2 a2 R - : ð12:49Þ
2μ0 2μ0

Here, the first summand is an integral of the vacuum magnetic field over the whole
volume and hence is constant. The second integral increases linearly with the major
radius. The force in the major radius direction, therefore, equals

∂W B B2eφ B2iφ
FB = - = 2π2 a2 - : ð12:50Þ
∂R 2μ0 2μ0

The r.h.s. of Eq. (12.50) can be modified using equilibrium over minor radius
Eq. (12.39). After the substitution of toroidal magnetic pressure, we have

μ0 I 2T
F B = 2π2 a2 hpi - : ð12:51Þ
4

The sum of three forces Eqs. (12.43), (12.47), and (12.51) yields

μ0 I 2T 8R 3 l 4π2 a2 hpi
FR = Fp þ FI þ FB = ln - þ iþ : ð12:52Þ
2 a 2 2 μ0 I 2T

The force FR accelerates plasma on the major radius direction towards the outer part
of the torus. To keep plasma in the equilibrium state, it is necessary to apply
additional force in the opposite direction F R = - F R . Such a force is produced by
a vertical magnetic field BV, which is created by currents flowing in the special coils.
As a result of the interaction of the vertical magnetic field with the toroidal current,
the force arises

F R = - jφ BV dV = - 2πRI T BV : ð12:53Þ
12.5 Integral Equilibrium in a Tokamak 255

Fig. 12.6 X-point formation at tokamak high field side for large βI. The direction of poloidal
magnetic field in different regions is marked by arrows

Condition F R = - F R determines the vertical magnetic field:

μ0 I T 8R 3 l 4π2 a2 hpi
BV = ln - þ iþ : ð12:54Þ
4πR a 2 2 μ0 I 2T

The last term in brackets is known as beta poloidal

4π2 a2 hpi 2μ0 hpi

βI = = 2 , ð12:55Þ
μ0 I 2T B p ð aÞ

and equals the ratio of the average pressure to the poloidal magnetic field pressure.
Plasma equilibrium is violated at large values of βI. Indeed, since a vertical
magnetic field is added to the poloidal magnetic field produced by the plasma
current, at the torus high field side, their sum turns to zero (Fig. 12.6). The point
of zero poloidal magnetic field is called the X-point, and the line separating areas
with different topologies that pass through the X-point is called the separatrix. For
small values of βI, the vertical magnetic field is small, and the Х-point is located
outside the plasma, where the poloidal field created by the toroidal current decreases
as r-1 and balances the vertical field. When βI rises, the Х-point is shifted towards
plasma and then into plasma, and equilibrium cannot be reached since plasma
expands along magnetic field lines. The maximal possible value of βmax I is reached
when

μ0 I T
BV = : ð12:56Þ
2πa

Hence,
256 12 Plasma Equilibrium

2R 8R 3 l
βmax
I = - ln þ - i: ð12:57Þ
a a 2 2

The βmax
I corresponds to maximal βI for tokamaks with circular cross-sections.
For other shapes, for example, for larger elongation, the values of βmax
I could be
larger than those given by Eq. (12.57).

12.6 Plasma Equilibrium in a Tokamak with Circular

Cross-Sections

Here, we consider equilibrium when the tokamak flux surfaces are close to circles.
Let us seek equilibrium with flux surfaces being a system of nested circles with
centers shifted with respect to the center of the last closed flux surface, Fig. 12.7. We
denote the major radius of the flux surface center of the last closed flux surface as R0,
and the major radius of an arbitrary point of the flux surface as R. Minor radius r is
defined as the radius of a given flux surface calculated from its center for the last
closed flux surfacer = a. The shift of the flux surface center with respect to the center
of the last closed flux surface is denoted as Δ(r), and this quantity decreases with
r and Δ(a) = 0. This quantity Δ(r) is known as the Shafranov shift. Approximation
of small toroidicity is assumed

r
ε= ≪ 1: ð12:58Þ
R0

The major radius of the given flux surface center in this approximation is also
≈R0 for the last closed flux surface. The toroidal magnetic field is inversely propor-
tional to the major radius:

Fig. 12.7 Shafranov shift of circle magnetic flux surfaces

12.6 Plasma Equilibrium in a Tokamak with Circular Cross-Sections 257

Fig. 12.8 Surface area between two neighboring flux surfaces

B0T R0
BT = : ð12:59Þ
R

Here, B0T is a toroidal magnetic field in the center of the last closed flux surface.
Considering the geometrical relation R = R0 + r cos θ, Eq. (12.59) reads

B0T
BT = ≈ B0T ð1 - ε cos θÞ: ð12:60Þ
ð1 þ ε cos θÞ

The poloidal magnetic field is sought in an analogous form

Bp = B0p ð1 - εΛðr Þ cos θÞ: ð12:61Þ

Functions Λ(r) and Δ(r) should be obtained from the equilibrium equation. These
functions are not independent. To find the relation between two functions, consider
two neighboring flux surfaces. Let dξ be the distance between them in the normal
direction, as shown in Fig. 12.8. Distance dξ depends on the poloidal angle and
Shafranov shift

dΔ
dξ = dr þ cos θdΔ = 1 þ cos θ dr: ð12:62Þ
dr

Surface element dSp between neighboring flux surfaces, which is crossed by poloidal
magnetic flux, is
258 12 Plasma Equilibrium

dΔ
dSp = 2πRdξ = 2πR0 ð1 þ ε cos θÞ 1 þ cos θ dr
dr
ð12:63Þ
dΔ
= 2πR0 1 þ ε þ cos θ dr:
dr

Quadratic in ε terms are neglected here. Multiplying dSp by Bp, one obtains the
poloidal flux between two surfaces:

dΔ
dΨp = Bp dSp = B0p ð1 - εΛ cos θÞ2πR0 1 þ ε þ cos θ dr
dr
ð12:64Þ
dΔ
≈ 2πR0 B0p 1 þ ε - εΛ þ cos θ dr:
dr

Since the poloidal flux is a surface function, the quantity dΨp should be poloidally
independent, and the coefficient in front of cosθ equals zero. Hence,

dΔ
= εðΛ - 1Þ: ð12:65Þ
dr

Let us now project equilibriumEq. (12.3) normal to the flux surface:

→ → →
dΔ d B2 h B∇ B
1- cos θ pþ = : ð12:66Þ
dr dr 2μ0 μ0
→
Here, h is the unit vector normal to the flux surface. Identity

∂ ∂ dΔ
= 1- cos θ ,
∂ξ ∂r dr

which follows from Eq. (12.62), is also used. The operator on the r.h.s. can be split
into four components:

→ → → → → → → → → →
B ∇ B = B T ∇ B T þ B p ∇ B T þ B T ∇ B p þ B p ∇ B p : ð12:67Þ

The first term on the r.h.s. according to the definition of the curve radius is a vector
parallel to the major radius direction,

→ → B2T → → → → B2T
B T∇ B T = - R, h B T∇ B T = - cos θ: ð12:68Þ
R2 R
12.6 Plasma Equilibrium in a Tokamak with Circular Cross-Sections 259

→ →
The second term on the r.h.s. of Eq. (12.67) B p ∇ B T is a vector in the toroidal
direction and has no projection on the normal to the flux surface. The third compo-
→ →
nent B T ∇ B p is a poloidal vector and has no projection on the normal to flux
surface. The last vector in Eq. (12.67) is directed along the minor radius towards its
center, so

→ → → B2p
h B p∇ B p = - : ð12:69Þ
r

With an account of Eqs. (12.68) and (12.69), Eq. (12.66) is transformed to

2
dΔ d B2 1 B2T Bp
1- cos θ pþ =- cos θ þ : ð12:70Þ
dr dr 2μ0 μ0 R r

Let us put all small terms of the order of ε containing cosθ into r.h.s.:

d B2 B2p B2T
2
dΔ Bp
pþ þ =- þ cos θ: ð12:71Þ
dr 2μ0 μ0 r μ0 R dr μ0 r

In the last term, quantity d( p + B2/2μ0)/dr is replaced by B2p =ðμ0 r Þ, which corre-
sponds to the cylindrical (zero order in ε) approximation. One must keep in mind that
on the l.h.s. of Eq. (12.71), in addition to zero-order terms in ε, there are also first-
order terms proportional to ε cos θ. The latter are the result of substitution of toroidal
and poloidal magnetic fields in the form of Eqs. (12.60) and (12.61). Therefore,
besides the zero-order equation

2
d B0
2 B0p
pþ þ = 0, ð12:72Þ
dr 8π 4πr

one can obtain a first-order equation from the condition that the coefficient in front of
the cosθ term in Eq. (12.71) should be zero. Considering Eqs. (12.65) and (12.72),
we have

d 2 dΔ 2μ r 2 dp 2r 0 2
r B0p = 0 - B : ð12:73Þ
dr dr R0 dr R0 p

After integrating over the minor radius:

2
dΔ 2μ0 ðp - pÞ - Bp r
= 2
, ð12:74Þ
dr R0
B0p
260 12 Plasma Equilibrium

where

r r
1 1 2
0 0 0
2πr0 B0p dr 0 :
2
p= 2 2πr pðr Þdr , Bp = 2
πr πr
0 0

According to Eq. (12.74), shift of flux surfaces is determined by parameter βI,

Eq. (12.55). Derivative dΔ/dr is negative because for pressure, it is decreasing with
radius p < p. Hence, inner flux surfaces are more shifted outwards than the outer
ones in accordance with Fig. 12.7. The Shafranov shift Δ(r) can be obtained by
integrating Eq. (12.74). For a known Shafranov shift, it is easy to calculate the
poloidal magnetic field from Eqs. (12.61) and (12.65).

12.7 Coordinates for Arbitrary Flux Surfaces

To describe plasma behavior in fusion devices with flux surfaces of arbitrary shape,
special coordinates are introduced. It is convenient to use flux surface coordinates,
which determine the position at the flux surface,

xi = ða, θ, ζÞ: ð12:75Þ

Here, for coordinate a, any flux surface function can be chosen to mark a given flux
surface. For example, or circular cross-sections small radius can be taken as a.
Coordinates θ and ζ correspond to poloidal and toroidal angles, respectively, and
change in the interval 0, 2π. Signs for coordinates θ and ζ are chosen so that for circle
surfaces and small toroidicity, they coincide with the same directions of the cylin-
drical coordinates of a pinch, θ and z, Fig. 12.1. If flux surface coordinates are known
as functions of Cartesian coordinates, then the covariant and contravariant basis
vectors are defined as

→i → →
e = ∇xi , e i = ∂ r =∂xi : ð12:76Þ

In these coordinates, the length element is

dl2 = gik dxi dxk , ð12:77Þ

→ →
where metric tensor gik = e i e k . Volume element

p p → → → -1
dV = gdx1 dx2 dx3 , g= e 1× e 2 e 3= ∇x1 × ∇x2 ∇x3

= detgik : ð12:78Þ
12.7 Coordinates for Arbitrary Flux Surfaces 261

Additional relations:

→ →k → p
eie = δik , e i = g ∇xiþ1 × ∇xiþ2 : ð12:79Þ

Arbitrary vector in arbitrary coordinates:

→ →i → →→ → →i
A = Ai e = Ai e i ; Ai = A e i ; Ai = A e : ð12:80Þ

Let us demonstrate that for a magnetic field line, which lies on the magnetic flux
surface, the magnetic field can be presented in the form

→ 1 1 1
B= ½∇Ψ1 × ∇ζ þ ½∇Ψ2 × ∇θ þ ½∇a × ∇η, ð12:81Þ
2π 2π 2π

where Ψ1(a) and Ψ2(a) are flux surface functions and η(a, θ, ζ) is a periodic function
of θ and ζ. Indeed, a vector with a zero component along ∇a can be written as

→
B = x½∇a × ∇θ þ y½∇a × ∇ζ: ð12:82Þ
→
From equation ∇  B = 0, we have

∂x=∂ζ - ∂y=∂θ = 0: ð12:83Þ

Hence, quantities x and y can be expressed through one periodic function

1 ∂η 1 ∂η
x = x 0 ð aÞ þ , y = y0 ð a Þ þ : ð12:84Þ
2π ∂θ 2π ∂ζ

Substituting Eq. (12.84) into Eq. (12.82), one obtains Eq. (12.81), if one denotes

1 dΨ1 1 dΨ2
x0 ð a Þ = , y0 ð aÞ = : ð12:85Þ
2π da 2π da

Function η(a, θ, ζ) must satisfy additional condition – the absence of current normal
to flux surface, see Section 12.3,

→ 1 → 1 →
j ∇a = ∇ × B ∇a = ∇  B × ∇a = 0: ð12:86Þ
μ0 μ0

It is easy to see that quantities Ψ1 and Ψ2 coincide with poloidal and toroidal
magnetic fluxes Ψp and ΨT through the surfaces shown in Fig. 12.2. To demonstrate
this, let us integrate Eq. (12.81) over toroidal and poloidal surfaces. Introducing
→
vector potential A , we have
262 12 Plasma Equilibrium

→ → → → → → → →
Ψp = Bd S = A d l T , ΨT = Bd S = Ad l p , ð12:87Þ
Sp ST

where lp, lT contours embracing toroidal and poloidal cross-sections correspond-

ingly. Vector potential, which corresponds to Eq. (12.81), can be written in the form

→ 1 1 1
A= Ψ ∇ζ þ Ψ2 ∇θ - η∇a þ ∇d, ð12:88Þ
2π 1 2π 2π

where d is an arbitrary function. After substitution of Eq. (12.88) into Eq. (12.87)
with an account of

→ → → → →
∇θd l p = ∇ζd l T = 2π, ∇ad l p,T = ∇θd l T = ∇ζd l p = 0,
→
∇dd l p,T = 0,

ð12:89Þ

one obtains Ψ1 = Ψp, Ψ2 = ΨT. Hence, the magnetic field depends on the poloidal
and toroidal magnetic fluxes according to

→ 1 1 1
B= ∇Ψp × ∇ζ þ ½∇ΨT × ∇θ þ ½∇a × ∇η: ð12:90Þ
2π 2π 2π

In general, angle coordinates could be chosen so that η = 0. Such coordinates are

called flux coordinates with straightened magnetic field lines. In such coordinates,
(they are not necessarily orthogonal ones)

→ 1 1
B= ∇Ψp × ∇ζ þ ½∇ΨT × ∇θ: ð12:91Þ
2π 2π

The equation for the magnetic field line in these coordinates is

dθ dζ
→ = → : ð12:92Þ
B ∇θ B ∇ζ
→ →
Taking the poloidal B ∇θ and toroidal B ∇ζ magnetic fields from Eq. (12.91), we
find

→
dζ B ∇ζ dΨT =da dΨT
= → =- =- = q, ð12:93Þ
dθ B ∇θ dΨ p =da dΨp
12.8 Force-Free Equilibrium and Pinch with Canonical Profiles 263

where safety factor q is defined according to Eq. (12.22). Since the safety factor is a
surface function, the magnetic field line in the chosen coordinates is a straight line.
Such a choice of orthogonal coordinates is natural for tokamaks. Let us denote
inverse gradients as

1 1 1
ha = , hθ = , hζ = : ð12:94Þ
j∇aj j∇θj j∇ζj

One can see that hζ = R, where R is the major radius for the flux surface point, since
after integrating over loop dlT, one should obtain 2π according to Eq. (12.89). In
accordance with Eq. (12.91), physical components of magnetic field

dΨp =da dΨT =da

Bθ  Bp = - , Bζ = BT = : ð12:95Þ
2πRha 2πha hθ

For circular flux surfaces, hθ is just the minor radius r. In the general case, quantities
ha, hθ can be specified numerically for a given flux surface shape. Differential
operators and integrals for known ha, hθ, and hζ are defined according to general
rules. If the poloidal magnetic flux Ψ is calculated from the magnetic axis according
to Eq. (12.36), then

→ 1 1
B=- ½∇Ψ × ∇ζ þ ½∇ΨT × ∇θ, ð12:96Þ
2π 2π

dΨ=da dΨT =da

Bθ  Bp = , Bζ = BT = : ð12:97Þ
2πRha 2πha hθ

12.8 Force-Free Equilibrium and Pinch with Canonical

Profiles

In some special situations when the plasma pressure is small with respect to the
magnetic field pressure, configurations of force-free equilibrium can exist in plasma.
In the equilibrium equation, one can put ∇p = 0. From Eqs. (12.1) and (12.2), we
have

1 → →
∇ × B × B = 0: ð12:98Þ
μ0

According to this equation,

264 12 Plasma Equilibrium

→ →
∇× B =μB: ð12:99Þ

In other words, the plasma current should flow along the magnetic field. Below is
shown that the quantity μ is independent of the magnetic field. Such a configuration
is force-free because current flows along the magnetic field line and the Lorentz force
is absent. The current density is proportional to the magnetic field strength. This
equation is also known as the Beltrami relation. Force-free magnetic equilibrium is
obtained in a toroidal pinch when the toroidal and poloidal magnetic fields are of the
same order.
Pinch equilibrium in the general case is governed by Eq. (12.11) (or Eq. (12.12),
containing two arbitrary functions Bz(r) and Bθ(r). Let us demonstrate that an
additional constraint of minimal magnetic field energy with conservation of a
specific function known as helicity leads to force-free magnetic equilibrium of
type Eq. (12.99). The obtained radial profiles are known as canonical profiles. Let
us first define helicity as a volume integral taken over the flux tube volume:

→*
K= A B dV, ð12:100Þ

→
where A is the vector potential. In plasma with infinite conductivity, this integral is
conserved.
We shall prove this first for a closed field line. The integral can be rewritten in the
form

→ → * →
K= Ad l Bd S , ð12:101Þ

where dS is the magnetic tube cross-section and dl is the differential length along the
magnetic field line. Since the magnetic flux is conserved in the flux tube, it is
necessary to prove that

d → →
A d l = 0: ð12:102Þ
dt

Integral evolves according to

→
d → → ∂A → → → → → →
Ad l = þ V∇ A d l þ A d l ∇ V: ð12:103Þ
dt ∂t

→ → →
Here, is used d l =dt = l ∇ V . Since the magnetic field is frozen into the plasma
for σ → 1 according to Eq. (10.17), the vector potential satisfies the equation
12.8 Force-Free Equilibrium and Pinch with Canonical Profiles 265

→
∂A → →
- þ V × B = ∇χ, ð12:104Þ
∂t

where χ is an arbitrary scalar function. Substituting Eq. (12.104) into Eq. (12.103),
→ → →
taking into account that V × B d l = 0 and combining the remaining terms, we
find

d → → → →→
Ad l = d l ∇ A V þ χ = 0: ð12:105Þ
dt

To demonstrate that this integral is zero, one has to apply Stokes’ theorem and take
→→
into account that ∇ × ∇ A V þ χ = 0.
Therefore, helicity is conserved for closed magnetic tubes and infinite conduc-
tivity. In the general case with finite collisionality, this is not strictly true since
magnetic field lines can reconnect, so the topology is changing. However, there are
heuristic arguments that the general structure of the magnetic field does not change
significantly during reconnections, and the integral Eq. (12.100) is conserved if
integration is performed over the whole plasma volume. Helicity is then redistributed
between flux tubes, while the whole sum is conserved.
Following this assumption, let us find the magnetic configuration corresponding
to the minimum magnetic field energy with restriction of helicity conservation. To
do this, one has to seek a minimum

B2
L= dV - λK, ð12:106Þ
2μ0

where λ is the Lagrange multiplier. From the condition of zero functional variation
δL = 0, one obtains

→
B → →→
δ ∇ × A - λδ A B dV = 0 ð12:107Þ
μ0

Using equality

→→ → → → → → → → →
δ A B = Aδ ∇× A þ BδA = A ∇×δA þ BδA ð12:108Þ

equation (12.107) can be reduced to

→
B → → → →
-λA ∇ × δ A - λ B δ A dV = 0: ð12:109Þ
μ0
266 12 Plasma Equilibrium

The first term can be further modified using the expression

→ → → → → →
∇ a × b = - a ∇× b þ b ∇× a :

Since the volume integral over divergence turns to zero, we have

→
B → → → →
∇× -λA δ A - λ B δ A dV = 0: ð12:110Þ
μ0

Since variation is arbitrary, the Lagrange equation should be satisfied:

→
B → →
∇× -λA - λ B = 0, ð12:111Þ
μ0

or for magnetic field

→
B →
∇× = 2λ B : ð12:112Þ
μ0

Introducing the constant μ = 2μ0λ, we obtain Eq. (12.99) for the field-free
configuration.
In the projections:

1 drBθ
= μBz ,
r dr ð12:113Þ
dB
- z = μBθ :
dr

Substitution of the azimuthal magnetic field Bθ from the second equation into the
first one leads to the Bessel equation:

1 d dB
r z þ μ2 Bz = 0: ð12:114Þ
r dr dr

Hence, canonical profiles of magnetic field components are given by Bessel

functions

Bz = B0 J 0 ðμr Þ,
ð12:115Þ
Bθ = B0 J 1 ðμr Þ:

The radial profiles of the magnetic field components are shown in Fig. 12.9. The
longitudinal magnetic field strongly decreases with radius, while the azimuthal
12.9 2D Modeling of the Tokamak Edge 267

Fig. 12.9 Canonical radial profiles of parallel and azimuthal magnetic field components in a
pinchcomponents in a pinch

component increases towards the periphery, and its projection onto the magnetic
field line also rises. Profile shapes depend on the parameter μa, where a is the pinch
radius. Parameter μa is connected with the ratio of longitudinal current to toroidal
magnetic flux:

μ0 aI
μa = : ð12:116Þ
ΨT

The radial profiles become more peaked with the rise of μa, and, for μa = 2.4, the
longitudinal magnetic field turns to zero at the plasma boundary, and, for μa > 2.4,
the longitudinal magnetic field changes sign at the plasma periphery. Canonical
profile Eq. (12.115) and magnetic field reversal are in good agreement with the
experimental profiles observed in reversed field pinches.

12.9 2D Modeling of the Tokamak Edge

Tokamak edge plasma is a complicated 2D system and, therefore, requires 2D

modeling. For such modeling, one needs equilibrium analysis. As a first step, the
Grad-Shafranov Eq. (12.33) is solved, and the poloidal flux Ψp(R, z) is calculated.
Using this function, a special 2D mesh is generated; see Fig. 12.10 as an example.
This mesh is generated as a rectangular mesh; one side coincides with one of the flux
surfaces in the x-direction, and the other side is perpendicular to the flux surface in
the y-direction. The sizes of each cell are hx, hy correspondingly. In the z-direction,
the Lame coefficient is hz = R. Plasma parameters such as density, temperature,
velocities, and potential are associated with cell centers, and fluxes are associated
268 12 Plasma Equilibrium

j, V
100

–2.5

Z, m
–3 50

–3.5
0
–4

–4.5
–50
4 5 6
R, m

Fig. 12.10 Plasma potential in the 2D modeling of edge tokamak plasma with the SOLPS-ITER
transport code

with cell areas. Cell surfaces and volumes differ from cell to cell, and they are used to
reduce transport equations in the differential form to their numerical realization. All
simulation areas are restricted by magnetic flux at the core side of the plasma,
magnetic flux outside the separatrix, and divertor plates, where boundary conditions
are imposed. An example of the solution of the transport equations with code
SOLPS-ITER is presented in Fig. 12.10.
Chapter 13
Transport Phenomena in Tokamaks

Plasma equilibrium does not imply a full absence of flows over and across flux
surfaces; the only requirement is that corresponding velocities should be smaller
than the sound or Alfven velocity. In this chapter, such flows are analyzed for a
tokamak, and, in particular, particle, energy, and momentum fluxes across flux
surfaces are obtained. The fluxes caused by Coulomb collisions are called neoclas-
sical fluxes, and the corresponding transport coefficients are known as neoclassical
transport coefficients. Here, the prefix “neo” reflects the fact that fluxes are calcu-
lated in complicated geometry; however, the fluxes considered below are purely
classical in contrast to turbulent transport, which was analyzed in Chap. 8.
To analyze transport for arbitrary flux surfaces, one must use the coordinate
system considered in the previous section. However, for simplicity below the
simplest variant of circular flux surfaces is considered, and, at the end of the chapter,
general expressions valid for arbitrary flux surfaces will be presented.
Let us consider the simplest model for a magnetic field, the case of small
toroidicity ε ≪ 1 with circular flux surfaces. The model magnetic field is simplified
with respect to that considered in Sect. 12.6:

B0T
BT = ≈ B0T ð1 - ε cos θÞ,
ð1 þ ε cos θÞ
Bp  Bθ = B0p ð1 - ε cos θÞ = ΘBT ,
ð13:1Þ
2 2
B0 = B0T þ B0p ,
ε = r=R0 :

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 269
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_13
270 13 Transport Phenomena in Tokamaks

Fig. 13.1 Coordinate system for circle flux surfaces used for transport analysis

For this model electric field, the ratio Θ = Bp/BT is independent of the poloidal angle
θ. Let us assume Θ ≪ 1, so that Θ ≈ B0p =B0 . As a coordinate a, which is the flux
surface characteristic, we choose a small radius r. The coordinates θ, ζ correspond to
the poloidal and toroidal angles, as shown in Fig. 13.1. The length element in these
coordinates is

dl2 = dr 2 þ r 2 dθ2 þ R20 ð1 þ ε cos θÞ2 dζ2 , ð13:2Þ

and the components of the metric tensor are

grr = 1, gθθ = r 2 , gζζ = R20 ð1 þ ε cos θÞ2 : ð13:3Þ

Ion Larmor radius ρci is supposed to be small enough, so that

ρci =ðrΘÞ ≪ 1: ð13:4Þ

This inequality is usually fulfilled in tokamaks. The Larmor radius and thermal
velocity are defined as

1=2
2T j =mj 2T j
ρcj = ; V Tj = : ð13:5Þ
ωcj mj

All macroscopic quantities are sought as series

13.1 Fluid Regime (Pfirsch-Schlueter Regime) 271

nðr, θÞ = n0 ðr Þ þ n1 ðr, θÞ . . . ;
T j ðr, θÞ = T 0j ðr Þ þ T 1j ðr, θÞ . . . ;
ð13:6Þ
φðr, θÞ = φ0 ðr Þ þ φ1 ðr, θÞ . . . ;
ujk ðr, θÞ = ujk0 ðr Þ þ ujk1 ðr, θÞ: . . .

The first term corresponds to the cylindrical approximation and is independent of the
poloidal angle, the second term is linear in ε, and the other terms are higher order in
ε. Note that temperatures, density, and potential are not surface functions, in contrast
to full pressure, and are functions of the poloidal angle θ. Here, φ0 is the average
potential on the flux surface, and uik0 corresponds to the average ion velocity along
magnetic field lines.

13.1 Fluid Regime (Pfirsch-Schlueter Regime)

First, we consider the fluid case when the mean-free path of the particles is small
with respect to the length of the magnetic field line after one turn in the toroidal
direction:

V Tj r
λj = < = qR: ð13:7Þ
νj Θ

Here, νi  νii, νe  νei. This regime is typical for small temperatures in small
tokamaks or in the separatrix vicinity for larger machines. In this regime, plasma
is described by the fluid equations of Chap. 2, and transport studies are relatively
simple.

13.1.1 Qualitative Estimates

Here, we present qualitative estimates of particle and heat fluxes caused by

toroidicity. Let us start with an estimate of temperature perturbations on the flux
surface. Electrons and ions drift vertically in an inhomogeneous magnetic field with
the velocities given by Eq. (11.9). This ∇B-driven drift is directed upwards for
electrons and downwards for ions for the normal direction of the magnetic field.
After averaging over the Maxwellian distribution function, the corresponding
velocities are:

2T j
ugj = : ð13:8Þ
ej BR

Additionally, the vertical current of guiding centers arises, Eq. (11.10),

272 13 Transport Phenomena in Tokamaks

2nðT e þ T i Þ
jg = : ð13:9Þ
BR

There are vertical convective heat fluxes associated with these vertical drifts:
qgj = 5/2Tj(nugj). The divergence of these fluxes in plasma with an inhomogeneous
magnetic field is nonzero, which causes an increase in temperature perturbations at
the flux surface. As a result, parallel heat flux along magnetic field lines arises to
compensate for the divergence of vertical convective heat flux. The corresponding
estimate for heat balance reads

n0 T 20j κjk T 1j Θ2
 : ð13:10Þ
eB0 R0 r r2

Here, as a typical spatial scale in the radial direction, a minor radius r is taken, and
for a scale in the parallel direction, the magnetic field length r/Θ = qR is chosen.
From Eq. (13.10) with account of the expression for heat conductivity (κjk~nTj/
(mjνj)), one can estimate the temperature perturbation on the flux surface

T 1j ε νj qνj
 = : ð13:11Þ
T 0j Θ2 ωcj Θωcj

Ion temperature perturbation is positive at the lower part of the torus and negative at
the upper part. For electrons, the upper part is slightly hotter than the lower part, as
shown in Fig. 13.2. In the first approximation in ε, the vertical heat flux through the
upper and lower parts of the torus is the same but is directed inwards and outwards
correspondingly, so the net flux through the flux surface is zero. In the second
approximation in ε, since the lower part of the torus is hotter for ions than the
upper part, the flux through the lower part is larger than that through the upper part.
As a result, unbalanced net ion heat flow, averaged over the flux surface, arises:

nT 0j n0 T 0j
qj  T  q2 ρ2cj νj : ð13:12Þ
eB0 R0 1j r

This heat flow is q2 larger than the heat flow in a cylinder, (q⊥j = - κj⊥dT0j/dr), see
Chap. 2. The electron temperature perturbation is mi =me smaller than the ion
temperature perturbation. Therefore, the electron net heat flux through the flux
surface is mi =me smaller than that of ions in accordance with Eq. (13.12).
Vertical current Eq. (13.9), similar to the vertical heat flux, is closed by a parallel
current. The scheme of its closing is shown in Fig. 13.3. The parallel current, which
closes the vertical current, is called the Pfirsch-Schlueter current. It has opposite
directions on the outer and inner parts of the torus. To estimate it, one has to equalize
the divergency of the vertical current and the divergency of the poloidal projection of
the parallel current:
13.1 Fluid Regime (Pfirsch-Schlueter Regime) 273

Fig. 13.2 Scheme of ion flux formation through the flux surface

Fig. 13.3 Scheme of current closing: (а) cross section. 1 – Vertical current and 2 – poloidal
projection of parallel current; (b) current closing along magnetic field line

p
k =r,
 ΘjPS ð13:13Þ
BRr

so that

qp
k 
jPS
Br
: ð13:14Þ

The Pfirsch-Schlueter current is q times larger than the diamagnetic current.

274 13 Transport Phenomena in Tokamaks

13.1.2 Heat Conductivity

Now, we calculate the ion temperature perturbation on the flux surface from the ion
heat balance equation. Using the expressions for particle fluxes obtained below, one
can demonstrate that the terms containing fluid velocities are small with respect to
the terms containing heat conductivity, so we neglect convective terms from the
beginning. The following inequality is assumed to be satisfied:

ρci νj qR
< 1, ð13:15Þ
Θr V Ti

which implies that the ion temperature perturbation is small, in accordance with
estimate Eq. (13.11). Heat exchange with electrons is also neglected. Then, the
stationary heat balance equation for ions is reduced to

→
∇  q i = 0, ð13:16Þ
→
where ion heat flux q i is defined by Eqs. (2.38) and (2.39):

→
→ 5 nT i B
q i = - κik ∇k T i - κi⊥ ∇⊥ T i þ × ∇T i , ð13:17Þ
2 eB B

with parallel and perpendicular heat conductivity coefficients

3:9nT i 2nT i νi
κik = , κi⊥ = : ð13:18Þ
m i νi mi ω2ci

Let us neglect the small perpendicular heat conductivity and small perpendicular
heat flux associated with the second term on r.h.s. Then, in the chosen coordinates,
the heat balance equation is
p
1 ∂ gςς
p p qiθ = 0: ð13:19Þ
gςς ∂θ gθθ

According to Eq. (13.17), we have the poloidal component of the ion heat flux

5 nT i ð1 þ ε cos θÞ dT i Θ2 ∂T 1i
qiθ = - κik : ð13:20Þ
2 eB0 dr r ∂θ

The first term on the r.h.s. represents the diamagnetic heat flux connected with
rotation over the Larmor radius, see Chap. 2. In an inhomogeneous magnetic field,
the divergence of it is finite and coincides with the divergence of the vertical guiding
center heat flux. The second term represents the poloidal projection of the parallel
heat flux (qik = - κik Θr ∂T
∂θ
1i
). From Eq. (13.19), one obtains
13.1 Fluid Regime (Pfirsch-Schlueter Regime) 275

p
gςς
p qiθ = constðθÞ: ð13:21Þ
gθθ

After linearization of Eqs. (13.20) and (13.21), equalizing the first order in ε terms,
containing cosθ, we have

Θ2 ∂T 1i n T ε dT 0i
κik = 5 0 0i cos θ: ð13:22Þ
r ∂θ eB0 dr

Integration yields

n0 T 0i εr dT 0i
T 1i ðr, θÞ = 5 sin θ: ð13:23Þ
eB0 κik Θ2 dr

The perturbation of ion temperature has a maximum at the lower part of the torus and
corresponds to estimating Eq. (13.11). If condition Eq. (13.15) is satisfied, the ion
temperature perturbation remains small: T1i/T0i < 1.
Temperature perturbation T1i Eq. (13.23) can be used to calculate the additional
radial heat flux caused by toroidicity averaged over the flux surface. Averaged over
the flux surface for quantity f is defined as

1 p
hhf ii = f gθθ gζζ dθdζ, ð13:24Þ
S

where flux surface area S is given by

p p
S= gθθ gζζ dθdζ: ð13:25Þ

For the chosen simple model of magnetic field S = 4π2rR0, the surface average is
defined according to

1
hhf ii = f ð1 þ ε cos θÞdθ: ð13:26Þ
2π

The surface-averaged radial ion heat flux is thus

2π
1 1 5 n0 T 0i ∂T 1i
hhqi ii = qir ð1 þ ε cos θÞdθ = - ð1 þ ε cos θÞ2 dθ
2π 2π 2 eB0 r∂θ
0
2π
1 ∂T 0i
- κi⊥ ð1 þ ε cos θÞdθ:
2π ∂r
0
ð13:27Þ
276 13 Transport Phenomena in Tokamaks

Here, an additional factor (1 + ε cos θ) appears due to the poloidal dependence of the
magnetic field in the denominator. While taking the integral, all terms containing
products n1 and T1i and n2 and T2i turn to zero since they contain factor sinθ cos θ.
Neglecting small terms in Eq. (13.27), one obtains Shafranov’s heat flux

dT 0i
hhqi ii = - 1:6q2 þ 1 κi⊥ : ð13:28Þ
dr

Here, it is taken into account that

25n2 T 2i
= 1:6:
2e2 B2 κik κi⊥

Hence, the heat flux associated with toroidicity (the first term in Eq. (13.28)) is 1.6q2
times larger than the heat flux in the cylinder (the second term in Eq. (13.28)). The
ion flux caused by toroidicity was calculated as the diamagnetic heat flux,
corresponding to the last term in Eq. (13.17); however, physically, it corresponds
to the vertical guiding center vertical heat flux averaged over the flux surface with the
account of temperature perturbation. Note that the main toroidal part of the heat flux
in Eq. (13.28) is quadratic in ε.

13.1.3 Plasma Flows on the Flux Surface, Density,

Temperature and Potential Perturbations

Plasma flows on the flux surface can be found from the particle balance equations.
p p
1 ∂ gςς 1 ∂ gςς
p p nujr þp p nujθ = 0: ð13:29Þ
gςς ∂r grr gςς ∂θ gθθ

Radial particle velocity is obtained from the poloidal projection of the momentum
balance equations, where viscosity and inertia terms are neglected. It is thus

1 ∂φ1 1 ∂p1j
ujr = - - : ð13:30Þ
BT r ∂θ ej nBT r ∂θ

Here, we neglected the perpendicular friction force, which leads to classical radial
diffusive flux in a cylinder (see Chap. 7). Since, as demonstrated below, perturba-
tions on the flux surface are rather small, the contribution of radial fluxes Eq. (13.30)
to the particle balance equations can be neglected. One can check it by substituting
the perturbations obtained below into Eqs. (13.29) to (13.30). For the same reason,
we neglect density perturbations in Eq. (13.29) and replace it by
13.1 Fluid Regime (Pfirsch-Schlueter Regime) 277

p
1 ∂ gςς
p p n u = 0: ð13:31Þ
gςς ∂θ gθθ 0 jθ

From the radial momentum balance, the poloidal velocity is determined:

E r Bp 1 dpj
ujθ = - þ ujT þ : ð13:32Þ
BT BT ej nBT dr

Here, ujT is the toroidal velocity. One can also choose two alternative directions –
parallel to the magnetic field and perpendicular to the magnetic field – with the
perpendicular direction lying on the flux surface. Then, the radial momentum
balance determines the perpendicular velocity component

Er 1 dpj
uj⊥ = - þ = V 0 þ upj : ð13:33Þ
B ej nB dr
→ →
The first term here corresponds to E × B drift caused by the radial electric field, and
the second term represents diamagnetic drift. The first and third terms in Eq. (13.32)
are projections of the perpendicular velocity on the poloidal direction, and due to the
smallness of the poloidal magnetic field, they practically coincide with V0 and upj.
Parallel velocity also has a poloidal component (parallel and toroidal velocities
almost coincide). Therefore, with good accuracy

ujθ = V 0 þ Θujk þ upj : ð13:34Þ

From Eq. (13.31), we have

p
gςς
p ujθ = constðθÞ: ð13:35Þ
gθθ

Substitution of Eq. (13.34) yields

Er 1 dpj
- þ ð1 þ ε cos θÞ2 þ Θujk ð1 þ ε cos θÞ = const ðθÞ: ð13:36Þ
B0 ej nB0 dr

Keeping the first-order terms in ε, one finds parallel velocity

Er 1 dp0j
ujk ðr, θÞ = - 2ε - þ cos θ þ U k ð1 - ε cos θÞ
Bp ej nB0p dr
0
ð13:37Þ
= uPS
jk þ U k ð1 - ε cos θÞ:
278 13 Transport Phenomena in Tokamaks

Here, velocity U k is an average common parallel velocity of electrons and ions,

which is independent of the poloidal angle θ and with sufficient accuracy U k = U T .
We shall also neglect the difference in the mean velocities of electrons and ions due
to the flow of the average current in a tokamak.
The first term uPSjk is a Pfirsch-Schlueter flux. The component of parallel flux,
which is proportional to the pressure gradient, closes diamagnetic fluxes. Their
divergence in a nonuniform magnetic field is nonzero and coincides with the
divergence of the vertical guiding center drifts. A similar term is connected with
→ →
the poloidal E × B drift, whose divergence is also nonzero in a nonuniform
magnetic field. The second term in Eq. (13.37) is associated with the average toroidal
velocity. Since the surface of the low-field side of the torus is larger than the high-
field side surface, the parallel velocity is poloidally perturbed-smaller at the low-field
→
side to fulfil condition ∇  u j = 0.
By subtracting the parallel electron velocity from the ion velocity, after multiply-
ing by the density and electron charge, the expression for the Pfirsch-Schlueter
current is obtained

1 dp
k = - 2ε
jPS cos θ, ð13:38Þ
B0p dr

This is consistent with estimating Eq. (13.14). Here, p = p0e + p0i is the net plasma
pressure.
Due to the relative velocities of electrons and ions, a parallel friction force arises:
→
u
Reik = 0:51me νe jPS
k =e: In addition, a parallel thermal force exists caused by electron
temperature perturbation on the flux surface. The electron temperature perturbation
T1e can be obtained similar to the ion temperature perturbation from the electron heat
balance

→
∇  q е = 0: ð13:39Þ

Here, the electron parallel heat flux contains an additional term proportional to the
Pfirsch-Schlueter current, see Chap. 2, Eq. (2.44),

→
→ 5 nT e B
q e = - κek ∇k T e - κe⊥ ∇⊥ T e - × ∇T e k =e:
- 0:71T e jPS ð13:40Þ
2 eB B

The electron temperature perturbation is

n0 T 0e εr dT 1 dp0
T 1e ðr, θÞ = - sin θ 5 0e þ 1:42 : ð13:41Þ
eB0 κek Θ 2 dr n 0 dr
13.1 Fluid Regime (Pfirsch-Schlueter Regime) 279

It is (mi/me)1/2 smaller than the ion temperature perturbation. The parallel thermal
force is calculated using Eq. (13.41): RTeik = - 0:71Θ∂T 1e =r∂θ. As a result, full
friction force

→ n20 T 0e ε dT 1 dp0
Rk = Reik
u
þ RTeik = - 3:55 0e þ 4:14 cos θ: ð13:42Þ
eB0 κek Θ dr n0 dr

Neglecting inertia and viscosity terms, parallel momentum balance equations for
electrons and ions can be written in the form

1 ∂n1 1 ∂T 1e 1 ∂φ1
- ΘT 0e - Θn0 þ en0 Θ þ Rk = 0, ð13:43Þ
r ∂θ r ∂θ r ∂θ

1 ∂n1 1 ∂T 1i 1 ∂φ1
- ΘT 0i - Θn0 - en0 Θ - Rk = 0: ð13:44Þ
r ∂θ r ∂θ r ∂θ

In these equations, the friction force and electron pressure gradient associated
with electron temperature perturbations are (mi/me)1/2 smaller than the other terms.
Hence, after neglecting them, one obtains

eφ1 n1 T 1i
= =- : ð13:45Þ
T 0e n0 T 0e þ T 0i

The density perturbation is positive in the upper part of the flux surface, and the
poloidal gradient of the full pressure caused by the density gradient balances the
pressure gradient caused by ion temperature perturbation. As a result, full pressure
p = p0 is not perturbed and is a flux surface function in accordance with the result of
the previous chapter. The potential perturbation in the approximation considered
corresponds to Boltzmann’s distribution, and the potential perturbation as the den-
sity perturbation is positive at the upper part of the torus (Fig. 13.4). The poloidal
electric field is negative at the low-field side part of the flux surface and positive at
the high-field side, and the drift caused by this electric field is directed inwards at the
outer part of the flux surface.

13.1.4 Particle Fluxes

Plasma diffusion in a tokamak is a result of specific convection – radial particle flux

is a result of averaging convective fluxes over the flux surface. From Eq. (13.4), one
can get the impression that the average particle flux is directed inwards since the
outer part of the torus has a larger surface than the inner part. However, this is
incorrect, as sine full flux also contains vertical guiding center particle drift
280 13 Transport Phenomena in Tokamaks

→ →
Fig. 13.4 Poloidal electric field and corresponding E × B drifts

(poloidally inhomogeneous fraction of radial diamagnetic flux). In the first approx-

imation, the full flux averaged over the flux surface turns to zero due to Boltzmann’s
distribution Eq. (13.45). Hence, to obtain finite particle flux, it is necessary to keep
all terms in parallel momentum balance equations. Indeed, according to Eq. (13.30),
the radial electron flux is determined by the poloidal electric field and poloidal
pressure gradient:

n ∂φ1 n ∂p1e
Γe  nuer = - - : ð13:46Þ
BT r ∂θ ej nBT r ∂θ

Using the parallel momentum balance for electrons Eq. (13.43), one finds

Rk
Γe = : ð13:47Þ
eBp

This relation can also be easily obtained from the toroidal projection of the
momentum balance equation of electrons. Here, the Lorentz force eBpΓe is balanced
by the toroidal friction force RT ≈ Rk. Substituting Eq. (13.42) into Eq. (13.47), after
flux surface averaging, one obtains

2π
1 Rk
hhΓe ii = ð1 þ ε cos θÞdθ
2π eBp
0 : ð13:48Þ
n m ν q2 1 dp0 dT
= - 0 2e 2e 1:3 - 1:12 0e
e B0 n 0 dr dr

Particle flux, similar to heat flux, is q2 times larger than the flux in a cylinder,
Eq. (7.10),
13.2 Radial Electric Field, Poloidal and Toroidal Rotation 281

n0 me νe 1 dp0 3 dT 0e
Γ0 = - þ :
e2 B20 n0 dr 2 dr

The effective diffusion coefficient Deff = hhΓeii/(-d ln n0/dr) is (mi/me)1/2 times

smaller than the ion heat diffusivity coefficient χi = hhqiii/(-3/2n0dTi/dr), and the
particle flux is not fully diffusive.
The ion flux in the considered approximation coincides with the flux of electrons.
Indeed, from Eqs. (13.30) and (13.44), we have Γi = cRk/(eBp) and Γe = Γi since
both fluxes are expressed through the same friction force in accordance with
Eq. (13.47). To calculate the ambipolar electric field, as in the cylindrical case, it
is necessary to take into account ion viscosity terms in the momentum balance
equations.

13.2 Radial Electric Field, Poloidal and Toroidal Rotation

To analyze the radial electric field and rotation in a tokamak, one has to take into
account the viscosity and inertia of ions. For parallel velocities smaller than the
sound speed, the viscosity dominates over inertia forces. Considering the viscosity,
the total momentum balance equation is given by (subscript i is omitted for brevity)

$
- ∇k p - ð∇  π k Þk = 0: ð13:49Þ

In contrast to the previous section, full pressure is now not a flux surface function. In
$
the viscosity tensor, it is sufficient to consider only parallel viscosity π k . This part of
the viscosity tensor is connected with the coefficient η0 and is presented in Chap. 2
for a magnetic field parallel to the z-axis. In an arbitrary magnetic field, with an
account of the tensorial character of viscosity, it should have the form

Bα Bβ 1
πkαβ = pk - p⊥ - δαβ : ð13:50Þ
B B 3

Here, the difference between parallel and perpendicular pressures is connected with
the rate-of-strain tensor and derivatives of heat fluxes,
→ →
u q
pk - p⊥ = pk - p⊥ þ pk - p ⊥ : ð13:51Þ

In accordance with Eqs. (2.47) and (2.48), the first term, which determines compo-
→
$u
nent π k , equals
282 13 Transport Phenomena in Tokamaks

→
u 3 Bj B
pk - p⊥ =- η W k, ð13:52Þ
2 0 B jk B

where the rate-of-strain tensor Eq. (2.47) in Cartesian coordinates

∂uj ∂uk 2 →
W jk = þ - δjk ∇  u : ð13:53Þ
∂xk ∂xj 3
→
q
The second term pk - p⊥ is analogously connected with heat fluxes.
Let us multiply parallel momentum balance Eq. (13.49) by magnetic field B and
p p
perform volume averaging over the flux surface with weight g = grr gθθ gζζ :

→ → $
- B ∇p - B ∇  π k = 0: ð13:54Þ

Volume averaged (in contrast to surface averaged Eq. (13.24)) is defined as

p p
hf i = f gdθdζ= gdθdζ: ð13:55Þ

After averaging, the first term n in Eq. (13.54) becomes zero since

→ → → →
B ∇p = ∇  p B - p∇  B = ∇  p B
→ → p
= pBd S = gdθdζ = 0 ≠ : ð13:56Þ

→
Here, equation ∇  B = 0 is used, the volume integral is reduced to surface integrals
for neighboring flux surfaces, and the latter is zero since magnetic flux is absent
through the flux surface. Hence, after volume averaging, the parallel momentum
balance is reduced to

→ $
B ∇  π k = 0: ð13:57Þ

In Cartesian coordinates xk, the average viscosity can be rewritten in the form

→ $ ∂ Bi πkik ∂Bi
B ∇  πk = - πkik :
∂xk ∂xk

Substituting Eq. (13.50) into the first term ∂(Biπkik)/∂xk, we find that it is propor-
*
tional to ∇  B and, hence, equals zero. The second term is evaluated as
13.2 Radial Electric Field, Poloidal and Toroidal Rotation 283

∂Bi Bi Bk 1 ∂Bi B ∂B
πkik = p k - p⊥ - δik = p k - p⊥ k :
∂xk B B 3 ∂xk B ∂xk

Therefore, from Eq. (13.57), one obtains

→
→ $ B
B ∇  πk = p⊥ - pk ∇B = 0: ð13:58Þ
B

Let us calculate part of the viscosity tensor associated with nonhomogeneous

→
velocity. Considering condition ∇  u i = 0, which is satisfied on the flux surface,
→
u 3 Bj B
pk - p⊥ =- η W k =
2 0 B jk B
∂uik
- 3η0 Θ - uiθ ε sin θ þ ΘU k ε sin θ : ð13:59Þ
r∂θ

Here, the last two terms are caused by changes in curvilinear coordinates in space
p
and are proportional to ∂ gζζ =∂θ. Substituting Eqs. (13.34) and (13.37) for poloidal
→
and parallel velocities, using identity B =B ∇B = Bp ε sin θ=r, one finds

→
→ $ u 3Θ2 E T ∂n 1 ∂T 0i
B ∇  πk =- η B - 0r þ 0i0 þ 0 þ Uk : ð13:60Þ
2 0
2R0 Bp enBp ∂r eBp ∂r

Accounting for the viscosity caused by parallel heat flux leads to the general
expression

→ $ 3Θ2 E T ∂n 1 ∂T 0i
B ∇  πk = - η B - 0r þ 0i0 þ kT 0 þ Uk : ð13:61Þ
2R20 0 Bp enBp ∂r eBp ∂r

Here, the numerical coefficient kT depends on collisionality and in the fluid approx-
imation considered in this section kT = 2.7 [13].
Condition Eq. (13.57) determines the neoclassical electric field

T 0i ∂n 1 ∂T 0i
Er = þ kT þ B0p U k : ð13:62Þ
en ∂r e ∂r

Thus, the radial electric field, which turns to zero volume average parallel viscosity,
consists of two parts. The two first contributions are proportional to the density and
ion temperature gradients; for decreasing radius density and temperature, these terms
correspond to a negative (directed from periphery to center) electric field. The last
term is caused by the toroidal rotation of plasma. For co-current toroidal rotation
284 13 Transport Phenomena in Tokamaks

(positive in the chosen coordinates), the corresponding contribution to the radial

electric field is positive. This last term can be interpreted as an additional electric
field B0p U T in the reference frame moving with the velocity U T ≈ U k in accordance
with the Lorentz transformation. It is worth noting that the average poloidal velocity
according to Eq. (13.32) depends only on the ion temperature gradient

1 ∂T 0i
uiθ = ð1 - k T Þ : ð13:63Þ
eB0 ∂r

In contrast, the perpendicular velocity ui⊥ depends on the toroidal rotation velocity
according to Eq. (13.33).
Toroidal rotation velocity is determined from the toroidal component of the
momentum balance equation:

$ dU T
jr Bp - ∇  π k = nmi : ð13:64Þ
ζ dt

Here, the term on the r.h.s. symbolically represents the radial transport of toroidal
momentum. In real situations, radial transport is driven by turbulence since neoclas-
sical flux is rather small. A similar term also enters parallel momentum balance
Eq. (13.54); however, it is smaller than the parallel viscosity. In Eq. (13.64), the
same term is retained since the toroidal viscosity averaged over the flux surface, as
shown below, becomes zero. Note that the account of viscosity, in contrast to that
considered in the previous chapter ideal case, leads to the formation of radial current,
and lines of current density do not belong to the flux surface. On the other hand, the
ambipolarity constraint should be satisfied, which means that the surface-averaged
radial current through the flux surface should be zero,

hhjr ii = 0: ð13:65Þ

Ambipolarity condition, according to Eq. (13.64), leads to equation

$
∇  πk nmi dU T
ζ
- = dt
: ð13:66Þ
Bp Bp

Let us demonstrate that the l.h.s. of this equation becomes zero. In curvilinear
coordinates
p p
$ 1 ∂ g πkθζ ∂ gζζ
∇  πk =p p πkθξ þp : ð13:67Þ
ξ g ∂θ g θθ gθθ gζζ ∂θ

Then, we divide the toroidal viscosity by the poloidal magnetic field and average
over the flux surface:
13.3 Neoclassical Transport in Collisionless Regimes 285

$
∇  πk p
ξ 1 1 ∂ g p
= p p πkθξ gθθ gζζ dθdζ
Bp S Bp g ∂θ gθθ
ð13:68Þ
p
1 πkθξ ∂ gζζ p
þ p gθθ gζζ dθdζ:
S Bp gθθ gζζ ∂θ

p
Product Bp grr gζζ is proportional to the poloidal flux between neighboring flux
surfaces and is independent of the poloidal coordinate and, hence, can be removed
p p
from the integral. Using identity g = grr gθθ gζζ , taking the first integral by parts,
one finds that the first integral compensates the second one, so that

$
∇  πk
ζ
= 0: ð13:69Þ
Bp

Ambipolarity constraint Eq. (13.66) is thus reduced to

nmi dU T
dt
= 0: ð13:70Þ
Bp

This equation determines the toroidal rotation profile. In the presence of an external
force in the toroidal direction, for example, from neutral beam injection, the
corresponding force should be added to the r.h.s.

13.3 Neoclassical Transport in Collisionless Regimes

13.3.1 Particle Trajectories

In collisionless plasma with a large mean free path λj > qR, the main contribution to
radial transport is caused by particles with small parallel velocities. Therefore, let us
consider the qualitative character of their motion. In a magnetic field, there are two
integrals of motion – full energy E = mjV2/2 - ejφ and magnetic moment
μj = mj V 2j⊥ =2B. The particle parallel velocity is a function of energy, and the
magnetic momentum is

2
V jk = ± E - ej φ - μ j B : ð13:71Þ
mj

While analyzing the particle motion, it is possible to neglect weak poloidal depen-
dence of the potential. Equations of motion in collisionless plasma are:
286 13 Transport Phenomena in Tokamaks

dr V 2jk þ μj Bmj
=- sin θ,
dt ωcj R
ð13:72Þ
dθ V 2jk þ μj Bmj
r =- cos θ þ ΘV jk þ V 0 :
dt ωcj R

Two first terms on the r.h.s. of Eqs. (13.72) are projections of the guiding centers’
vertical drift in the inhomogeneous magnetic field Eq. (11.9) in the radial and
poloidal directions, respectively. The contribution ΘVjk is a poloidal projection of
→ →
the parallel velocity, and V0 corresponds to the poloidal E × B drift. By dividing the
first equation into the second equation, we can obtain the equation for the trajectory.
The characteristics of motion strongly depend on the parallel velocity Vjk. If
parallel and perpendicular velocities are of the same order (of the order of thermal
velocity, e.g.), then variation of the parallel velocity Eq. (13.71) is small because the
magnetic field inhomogeneity is small. The particle completes the full turn over
poloidal angle. Such particles are called transit or untrapped particles. The trajecto-
ries of transit particles are shown in Fig. 13.5. An ion with positive velocity at the
outer midplane (at point О) moves in the counterclockwise direction along the
poloidal angle and is shifted radially due to vertical drift. Since the drift Vgi for
ions (for the chosen direction of the magnetic field) is directed downwards, the ion in
the upper part of its trajectory is shifted inside the flux surface and returns outside of
it in the lower part. As a result, the trajectory is closed and lies inside a given flux
surface (Fig. 13.5a). The ion trajectory with a negative parallel velocity is shifted
outwards of the magnetic surface. For electrons, the vertical drift Vge is directed
upwards; hence, the trajectories of an electron with positive velocity at point О are
situated outside the flux surface, and the trajectory of an electron with negative
velocity is located inside the flux surface (Fig. 13.5b).

Fig. 13.5 Poloidal projections of transit trajectories: (а) ions and (b) electrons. Trajectories
1 correspond to the positive (co-current) direction of parallel velocity Vjk > 0, and
trajectories 2 correspond to the negative (counter current) direction of parallel velocity Vjk < 0
13.3 Neoclassical Transport in Collisionless Regimes 287

The average radial deviation from the flux surface can be estimated as

-1
Δrj = V gj νbj : ð13:73Þ

Here, the bounce frequency

ΘV jk þ V 0
νbj = ð13:74Þ
r

is the frequency (inverse time) for a particle to turn in the poloidal direction. For
particles with thermal velocities

-1 Tj r
Δrj = ugj νbj   qρcj : ð13:75Þ
eBR ΘV Tj

Here, ugj is the vertical drift velocity for thermal particles in Eq. (13.8). We neglected
her contribution from V0~T/eB with respect to the poloidal projection of thermal
velocity ΘVTj since condition Eq. (13.4) is satisfied.
Let us now discuss the motion of particles with small parallel velocities at the
outer midplane. First, assume V0 = 0. According to Eq. (13.71) during motion in the
poloidal direction, the parallel velocity of a particle decreases due to the rise of the
magnetic field towards high field and energy conservation. For transit particles, the
change in parallel velocity is rather small-of the order of εVkj, while for particles with
small parallel velocity, their parallel velocity turns to zero, and a stagnation point
arises. The trajectory of such particles resembles a banana shape (Fig. 13.6), and
particles are called trapped or banana particles. Ions with positive velocity at the

Fig. 13.6 Poloidal projection of trapped particle trajectories: (а) ions and (b) electrons. Trajectories
1 correspond to the positive (co-current) direction of parallel velocity Vjk > 0 at outer midplane, and
trajectories 2 correspond to the negative (counter current) direction of parallel velocity Vjk < 0 at
outer midplane
288 13 Transport Phenomena in Tokamaks

outer midplane are shifted inside the flux surface (Fig. 13.6a). After reflection from
the stagnation point, ions move poloidally towards the low-field side and continue
shifting downwards. In the lower part of the trajectory, ions move radially towards
the initial flux surface, and, at the reflection point, one more change in parallel
velocity takes place. Finally, the poloidal projection of the trajectory has a banana
shape. Ions with negative velocity at the equatorial midplane are shifted outside the
initial flux surface. Banana trajectories are shown in Fig. 13.6b. Electrons with
positive velocity at point О are located outside the flux surface, while electrons
with positive velocity are located inside the flux surface. The particle trajectory in
space for V0 = 0 is presented in Fig. 13.7a.
Let us find the maximal value of the parallel velocity at the outer midplane V max
k
when the particle remains trapped. For such velocity, the reflection point is situated
at the inner midplane. From Eq. (13.71) applied for such particles for the outer and
inner midplane, one has

2
jk =
V max
mj
E - ej φ - μj Bmin ,
ð13:76Þ
2
0= E - ej φ - μj Bmax :
mj

Here Bmin, Bmax are the minimal and maximal values at the outer and inner
midplanes, respectively. Eliminating full energy yields

2 p
jk =
V max
mj
μj Bmax - μj Bmin ≈ 2εV j⊥ : ð13:77Þ

For thermal particles, the particle remains trapped if the parallel velocity satisfies the
inequality
p
0 ≤ V jk ≤ 2εV Tj : ð13:78Þ

For trapped particles with perpendicular velocities of the order of thermal velocity
shift with respect to flux surface and banana width are given by an estimate

-1 Tj r qρcj
Δrj = ugj νbj  p  p : ð13:79Þ
eBR εΘV Tj ε

The characteristic of untrapped particle motion does not change significantly with
→ →
an account of poloidal E × B drift with the velocity V0, since the latter is smaller
than ΘVTj. In contrast, for trapped particles, the situation is different. Due to radial
electric field kinetic energy changes with radius, Eq. (13.76) has the form
13.3 Neoclassical Transport in Collisionless Regimes 289

2
jk =
V max
mj
E - ej φout - μj Bmin ,
ð13:80Þ
2
0= E - ej φin - μj Bmax ,
mj

where φout, φin are the values of potential at the low- and high-field sides, respec-
tively. Since

qρcj
φout - φin  V 0 BΔr j  V 0 B p , ð13:81Þ
ε

one can expand the expression under square root in series over V0/ΘVTj. Finally, the
particle is trapped at the outer midplane:
p
0 ≤ V jk þ ΘV 0 ≤ 2εV Tj : ð13:82Þ

Hence, the particle whose parallel velocity at the outer midplane satisfies inequality
(13.82) has a reflection point, and its trajectory at the (r, θ) plane still has a banana
shape. At the same time, the parallel (toroidal) velocity remains finite, of the order of
-V0/Θ. Hence particle trajectory is disconnected in the toroidal direction, and
toroidal particle precession takes place with the velocity -V0/Θ, see Fig. 13.7b.

13.3.2 Ware Drift

For banana particles, there is also a specific effect – the average drift of electrons and
ions in the radial direction towards the plasma center. This drift is caused by the
toroidal inductive electric field ET, which exists in a tokamak. In the presence of
electric field, ET trajectories of trapped particles in the plane (r, θ) become discon-
nected, as shown in Fig. 13.8. Particles (ions) with positive velocity at the outer
midplane are accelerated by the electric field in the upper part of the trajectory and
gain additional energy Δmj V 2jk =2. Hence, the position of the reflection point is
shifted in the poloidal direction, and the particle is simultaneously shifted inwards
in the radial direction due to vertical drift in the nonhomogeneous magnetic field. At
the inner part of the trajectory field, ET decelerates particles that now do not reach the
reflection point at ET = 0. As a result, the particle is shifted radially with respect to
the initial flux surface. The average radial drift velocity can be estimated as follows.
The change in the particle radial shift δ(Δr) is connected with the variation in the
particle mean parallel velocity
290 13 Transport Phenomena in Tokamaks

Fig. 13.7 Trajectory of a trapped particle. (а) In the absence of radial electric field, trajectory is
closed; (b) disconnected trajectory with account of radial electric field

Fig. 13.8 Disconnected trajectory of trapped ion in toroidal electric field

ugj r ΔV jk
δ Δr j = , ð13:83Þ
ΘV jk V jk

where velocity variation ΔVjk is caused by the work of electric field eET:

eE T r
ΔV jk = : ð13:84Þ
mj ΘV jk
p
Flux associated with Ware drift with account of small, of the order of δn=n  ε,
fraction of banana particles is hence estimated as
13.3 Neoclassical Transport in Collisionless Regimes 291

p
hhΓ ii  εnδ Δr j νbj :
p
Combining with Eqs. (13.84) and (13.85), substituting εV Tj as a typical parallel
velocity, we obtain

p E
hhΓii  - εn T : ð13:85Þ
Bp

The minus corresponds here to the inwards direction of the Ware drift. Analysis
based on the kinetic equation solution shows that the Ware drifts for electrons and
ions coincide.

13.3.3 Estimation of Transport Coefficients in the Plateau

Regime

The collisionality parameter in this regime satisfies inequality

νj
ε3=2 < qR < 1: ð13:86Þ
V Tj

In other words, the particle mean free path, in contrast to Sect. 13.1, is larger than qR.
In such cases, particles with small parallel velocities produce the main contribution
to radial transport. If the left part of inequality (13.86) is satisfied, then their parallel
velocities are not small enough, and magnetic field nonuniformity cannot change
trajectory topology, so that the main contribution to transport is produced by
untrapped particles with small poloidal velocities.
Let us estimate heat diffusivity as the mean square particle shift between colli-
sions multiplied by collision frequency

δn 2 eff
χj  Δr j νj : ð13:87Þ
n

Factor δn/n takes into account the relative fraction of particles with small poloidal
velocities. Small poloidal velocities correspond to small parallel velocities since we
assume |V0| < ΘVTi, which is equivalent to Eq. (13.4) for eφ0~Ti. Hence,

δn=n  V jk =V Tj : ð13:88Þ

The effective collision frequency νeff

j in Eq. (13.87) is a special collision frequency
that corresponds to the deviation of the parallel velocity by the value of the order of
the parallel velocity itself. One can estimate νeff j from the fact that Coulomb
collisions are successive scattering at small angles, and particle motion in the
292 13 Transport Phenomena in Tokamaks

velocity space has diffusive character. The perpendicular particle velocity according
to Eq. (1.29) rises with time as

2
ΔV j⊥  νj V 2j t: ð13:89Þ

For a particle with full velocity of the order of thermal velocity Vj~VTj and with small
parallel velocity, the latter changes significantly at ΔVj⊥~Vjk. The corresponding
time t~(Vjk/VTj)2/νj, and the effective frequency is thus

νj
νeff
j  2
: ð13:90Þ
V jk =V Tj

The average deviation from the flux surface for particles with small parallel veloc-
ities is inversely proportional to the parallel velocity:

Tj r
Δr j  : ð13:91Þ
eBR ΘV jk

Substituting Eqs. (13.88), (13.90), and (13.91) into Eq. (13.87), one finds

2
Tj r 2 V Tj νj
χj  3
: ð13:92Þ
eBR Θ2 V jk

This expression diverges at Vjk → 0. However, one should note that when the
parallel velocity Vjk decreases, the bounce frequency νbj = ΘVjk/r also decreases,
and the effective collision frequency Eq. (13.90) rises. Hence, a minimum exists for
eff
jk , where νbj = νj . For smaller parallel velocity particles, trajec-
parallel velocity V min
tories are destroyed by collisions before they perform one turn in the poloidal
direction. From condition νbj = νeff
j , we have

1=3
jk = V Tj νj r=Θ
V min : ð13:93Þ
2

Assuming that the main contribution to the heat conductivity coefficient is brought
by particles with parallel velocities of the order of V min
jk , substituting Eq. (13.93) into
Eq. (13.92), we finally obtain

q2 ρ2ci V Tj Θ
χj  : ð13:94Þ
r

Since the heat conductivity is independent of the collision frequency, the

corresponding regime is called the plateau regime.
13.3 Neoclassical Transport in Collisionless Regimes 293

13.3.4 Estimation of Transport Coefficients in the Banana

Regime

For small collision frequencies

νj
qR < ε3=2 , ð13:95Þ
V Tj
p
velocity determined by Eq. (13.93) becomes smaller than εV Tj . In this situation,
the main contribution to transport
p is produced by banana particles with parallel
velocities of the order of V jk  εV Tj . To estimate the heat conductivity coefficient
p
in the banana regime, we use Eq. (13.87), where we put δn=n  εV Tj , the
p
deviation from the flux surface in accordance with Eq. (13.79) Δr j = qρcj = ε, and
the effective collision frequency is taken as νeff
j = νj =ε. As a result,

p
ε
χj  ρ2cj νj : ð13:96Þ
Θ 2

However, one must keep in mind that in reality, the situation is much more
complicated – the main part to transport is produced by trapped and untrapped
particles from a narrow transition region in the velocity space, which separates
trajectories with different topologies.
In all regimes considered, the ion heat conductivity coefficient is  mi =me
times larger than that for electrons. The dependence on collisionality is illustrated in
Fig. 13.9, and different parts of the curve correspond to Eqs. (13.28), (13.94), and
(13.96). Such heat conductivity is called neoclassical heat conductivity. In the
banana regime, the neoclassical heat conductivity coefficient is ε-3/2 times larger

Fig. 13.9 Neoclassical ion heat conductivity. 1 – Banana regime, 2 – plateau regime, and 3 –
Pfirsch-Schlueter regime
294 13 Transport Phenomena in Tokamaks

than that in the Pfirsch-Schlueter regime, and, for intermediate collision frequencies
(plateau regimes), transport is independent of collision frequency. Neoclassical
particle diffusivity is small – of the order of electron heat conductivity.

13.4 Distribution Function in the Collisionless Regimes

13.4.1 Plateau Regime

The starting equation is the drift kinetic equation considered in Sect. 1.8:

• •
∂f j → ∂f j → ∂f j • ∂f j
þ R →þ R þ V⊥ = St j : ð13:97Þ
∂t ∂R ∂V jk ∂V j⊥

The tokamak guiding center drift is given by:

•
→ 1 →
R = V jg - ∇φ × B þ V jk , ð13:98Þ
B2

where the vertical drift in the inhomogeneous magnetic field according to

Eq. (11.9) is:

V 2j⊥
mj 2 þ V 2jk
V jg = : ð13:99Þ
ej BR

The variations in the parallel and perpendicular particle velocities in accordance with
the general expressions of Sect. 1.8 are:

• ej Θ ∂φ V 2j⊥ Θε sin θ Θε sin θV 0 V jk

V jk = - - þ : ð13:100Þ
mj r∂θ 2 r r

• Θε sin θ V jk V j⊥ Θε sin θV 0 V j⊥
V j⊥ = þ : ð13:101Þ
r 2 r

The first term in Eq. (13.101) and the second one in Eq. (13.100) are connected with
magnetic momentum conservation μj = mj V 2j⊥ =2B in a nonhomogeneous magnetic
field. For example, the first term in Eq. (13.101) can be easily obtained in the
following way (for V0 = 0):
13.4 Distribution Function in the Collisionless Regimes 295

• dV j⊥ dlk Θ ∂ 2Bμj Θ ∂ 2B0 μj 1

V j⊥ = = V = 1 - ε cos θ V jk
dlk dt r ∂θ mj jk r ∂θ mj 2
Θε sin θ V jk V j⊥
= :
r 2

The first term in Eq. (13.100) is caused by the work of the poloidal electric field,
which has a projection in the parallel direction. The last terms in Eqs. (13.100) and
→ →
(13.101) are caused by the change in E × B drift in a nonuniform magnetic field.
The distribution function is sought in the form

f j ðr, θÞ = f 0j ðr Þ þ f 1j ðr, θÞ . . . : ð13:102Þ

Here, the zero-order distribution function is the Maxwellian one shifted by the mean
velocity

2
n0 mj V jk - U k mj V 2j⊥
f 0j ðr Þ = exp - - : ð13:103Þ
2πT 0j =mj
3=2 2T 0j 2T 0j

After linearization, the drift kinetic equation has the form

∂f 1j 1 ∂φ1 ∂f 0j 1 V 2j⊥ ∂f 0j
ΘV jk þ V 0 - - þ V 2jk ε sin θ
r∂θ B r∂θ ∂r ej Br 2 ∂r

ej Θ ∂φ1 ∂f 0j 1 mj V 2j⊥
ð13:104Þ
- - ε sin θ þ V 2jk V 0 f 0j
mj r∂θ ∂V jk r Tj 2
2
mj V j⊥
- ε sin θΘU k f = - νj f 1j :
Tj 2 0j

Here, the collision integral is replaced by its simplified form since the later collision
frequency will be turned to zero. Its solution in the complex form is

1 ∂φ1 ∂f 0 j ∂f 0 j
2
-i 1 V j⊥
f~1j = -i þ V 2jk εexpðiθÞ
ΘV jk þ V 0 - iν j B r∂θ ∂r e j Br 2 ∂r
e j Θ ∂φ1 ∂f 0 j im j V 2j⊥ V 2j⊥
k
þ - εexpðiθÞf 0 j þ V 2jk V 0 þ ΘU :
m j r∂θ ∂V jk r T j 2 2
ð13:105Þ

Multiplier (ΘVjk + V0 - iνj)-1 can be divided into real and imaginary parts:
296 13 Transport Phenomena in Tokamaks

1 ΘV jk þ V 0 iνj
= þ : ð13:106Þ
ΘV jk þ V 0 - iνj 2
ΘV jk þ V 0 þ νj
2 ΘV jk þ V 0
2
þ ν2j

Then, taking into account that νj → 0 and keeping in mind that during integration
over parallel velocities, the main contribution to the integral with the second term
gives resonance velocity Vjk = - V0/Θ, Eq. (13.106) can be replaced by

1 1
=P þ iπδ ΘV jk þ V 0 : ð13:107Þ
ΘV jk þ V 0 ΘV jk þ V 0

This means that while integrating over parallel velocities, two integrals should be
calculated. The first one should be considered the principal value integral, while the
second one contains the delta-function and gives the same result as the integral in
Eq. (13.106) with collision frequency νj → 0.
Finally, the real part of the perturbed distribution function f 1j = Re f 1j is

1 ∂
f 1j = P þ πδ ΘV jk þ V 0 Af 0j , ð13:108Þ
ΘV jk þ V 0 ∂θ

where

2
eφ1 ∂ 1 ∂ mj V j⊥ ∂
A = Θ þ - þ V 2jk cos θ
mi ∂V jk ωcj ∂r ej B 2 ∂r
ð13:109Þ
mj ε V 2j⊥ V 2j⊥
- cos θ þ V 2jk V 0 þ ΘU k :
Tj 2 2

This distribution function can be used to calculate all macroscopic parameters. For
example, the parallel velocity is given by the integral

1 1
2π
ujk = f 1j V jk V j⊥ dV jk dV j⊥ ð13:110Þ
n0
-1 0

and coincides with the fluid expression Eq. (13.37). Potential perturbation and
poloidal electric field are found from the quasineutrality constraint

* *
f 1e dV e = f 1i dV i , ð13:111Þ

so that
13.4 Distribution Function in the Collisionless Regimes 297

1=2 1=2
π T 0e 1 mi T 0i ε sin θ
φ1 =
2 e 1 þ T 0e =T 0i eB0 Θ
ð13:112Þ
dln n0 e dφ0 eB0 d ln T 0i
× þ þ ΘU k þ 0:5 :
dr T 0i dr T 0i dr

The radial electric field can be obtained from condition Eq. (13.58). Parallel and
perpendicular ion pressure is calculated according to

mi V 2ik * mi V 2i⊥ *
pik = 2 f dV ; pi⊥ = f 1i dV i : ð13:113Þ
2 1i i 2

Equation (13.58) is satisfied for

T 0i ∂n 1 ∂T 0i
Er = þ 1:5 þ B0p U k : ð13:114Þ
en ∂r e ∂r

After substitution in Eq. (13.112), we have for potential perturbation:

1=2 1=2
π T 0e 1 mi T 0i ε sin θ d ln T 0i
φ1 = - : ð13:115Þ
2 e 1 þ T 0e =T 0i eB0 Θ dr

Density and ion temperature perturbations are given by Eq. (13.45).

The radial average flux is calculated as

→ cmj ε V 2j⊥ =2 þ V 2jk sin θ c ∂φ1

Γj = dV j - - f 0j þ f 1j : ð13:116Þ
eBr eBr ∂θ

In the zero-order approximation with respect to parameter (me/mi)1/2, the radial flux
turns to zero since terms containing the poloidal electric field are compensated by the
collisionless part of the distribution function, so that

→ mj ε V 2i⊥ =2 þ V 2ik sin θ

h h Γ i ii = d V i πδ V 0 þ ΘV ik Ai f 0i ≈ 0: ð13:117Þ
eBr

This, as in the fluid case, is a consequence of Boltzmann’s distribution for particles

with large parallel velocities in the poloidal electric field. In the next approximation,
the ion flux coincides with that of electrons by an order of magnitude:

ΘV Te 2 ∂n
hhΓi ii = hhΓe ii  - ρ2ce q : ð13:118Þ
r ∂r
298 13 Transport Phenomena in Tokamaks

To obtain the exact expression, it is necessary to take into account the deviation of
the zero-order distribution function f0e from the Maxwellian function caused by the
current in the plasma. This effect is discussed qualitatively below for the banana
regime.
The average ion heat flux through the flux surface is calculated as

→ mi ε V 2i⊥ =2 þ V 2ik sin θ 1 ∂φ1

hhq1i ii ¼ d V i V 2i⊥ þ V 2ik - - ðf 0i þ f 1i Þ :
eBr eBr ∂θ

ð13:119Þ

It appears that there is a contribution from the poloidal electric field. However, one
also has to take into account the work of the poloidal electric field in the presence of
Pfirch-Shlueter ion fluxes, which contribute to the final flux:

eΘ ∂φ1
hhq2i ii = nu : ð13:120Þ
r ∂θ ik

Finally, for particle flux, the poloidal electric field does not contribute to the
resulting heat flux:

hhqi ii = hhqi1 þ qi2 ii

→ mi mj ε V 2i⊥ =2 þ V 2ik sin θ :

= d V i πδ V 0 þ ΘV ik V 2i⊥ þ V 2ik Ai f 0i :
2 eBr

ð13:121Þ

Considering Eq. (13.114)

3=2
π ε2 m2i T 0i dT 0i
hhqi ii = - 3 n : ð13:122Þ
2 Θre2 B20 0 dr

13.4.2 Banana Regime

Let us qualitatively discuss the shape of the distribution function. For νj → 0, the
distribution function should be a function of motion integrals and is conserved along
the trajectory. On the other hand, since trajectories of trapped and untrapped particles
are topologically different, the distribution function has discontinuity for the deriv-
ative over the parallel velocity. Collisions make the transition region more gradual.
Banana ions, which have a positive velocity at the outer midplane, come to their
13.4 Distribution Function in the Collisionless Regimes 299

Fig. 13.10 Ion distribution function in the banana regime

point from trajectories lying inside the flux surface. Hence, their density exceeds the
density of ions with negative velocity at the outer midplane by

dn0i
Δni  - Δr i , ð13:123Þ
dr

where banana width Δri is estimated according to Eq. (13.79).p Therefore, in the
velocity space, where ions are trapped, V ik = - V 0 =Θ ± εV Ti , the distribution
function derivative is positive ∂fi/∂Vik > 0, Fig. 13.10. The derivative in his region
can be obtained from the following estimate:
p
Δni Δf i ∂f i εV Ti
  : ð13:124Þ
n fi ∂V ik V ik = - V 0 =Θ
fi

Momentum exchange is most effective between trapped and untrapped particles.

Hence, the ion distribution function in this region should be similar to the Maxwel-
lian function (at dT0i/dr = 0 coincides with the Maxwellian distribution function).
Using this fact, one can replace the derivative by

∂f i f i mi f i mi
=- V ik - U k = V þ ΘU k : ð13:125Þ
∂V ik V ik = - V 0 =Θ
Ti V ik = - V 0 =Θ
ΘT i 0

→ →
Combining Eq. (13.123) with Eq. (13.124), one obtains the poloidal E × B velocity
V0. The corresponding radial electric field is of the order of given by Eq. (13.62), as
in other regimes.
Therefore, the requirement that the distribution function should be close to the
Maxwellian function (more accurately, it should turn to zero ion flux in the velocity
space caused by ion-ion collisions) and that it should be conserved along trajectories
300 13 Transport Phenomena in Tokamaks

Fig. 13.11 Electron distribution function in the banana regime

for small parallel velocities determines the radial electric field. The vertical ∇B drift
of electrons is directed upwards; hence, in the vicinity of -V0/Θ, more electrons
have a negative velocity, and, here, the derivative of their distribution function is
negative. On the other hand, for thermal velocities, the electron distribution function
should be close to the Maxwellian function bound to the ion distribution function by
electron-ion collisions. As a result, the electron distribution function should look as
shown in Fig. 13.11. A change in the derivative corresponds to the transition from
trapped to untrapped particles. Trapped electrons have a mean velocity of the order
of T0/ΘeBrp with respect to ions (we assume Te~Ti~T ). Hence, with account of their
fraction  ε from the full electron density, this flux corresponds to the current
p
j1k  n0 εT 0 =ΘeBr, ð13:126Þ

which depends on the density gradient as well as on the electron and ion temperature
gradients. In addition, there is a second contribution to the current, j2k, which arises
due to collisions between trapped and untrapped electrons. The latter produce a drag
force and shift the whole body of the electron distribution function with respect to
ions. From force balance

me νei j2k =e - me νee j1k =e n0 = 0 ð13:127Þ

one obtains j1k~j2k. Two contributions are of the same order and together form a
bootstrap current jB, which is a function of density and temperature gradients. By
order of magnitude
p
jB  n0 εT 0 =ΘBr ð13:128Þ

The bootstrap current is directed as the main plasma current.

13.5 Particle and Heat Balance Equations 301

The special character of the distribution function also leads to a reduction in the
conductivity current caused by the induced toroidal electric field. Since trapped
particles are not involved in the conductivity process, neoclassical conductivity is
reduced with respect to the Spitzer conductivity,
p
σNEO
k = σk 1 - α ε , ð13:129Þ

where α is a numerical coefficient.

13.5 Particle and Heat Balance Equations

Below are approximation expressions that can be used practically for all
collisionality regimes. The dimensionless collisionality parameter is defined as

1=2
νj mj qR
νj = 1=2
: ð13:130Þ
T j ε3=2

Particle flux

hhΓii = Γ1 þ Γ2 ;
p
2n εT m ν d ln T 0e
Γ1 = 0 2 2 0e 2 e e K 11 Ae þ K 12 ;
e B0 Θ dr
d lnðnT 0e Þ 5 d ln T 0e T 0i dlnðnT 0e Þ 1 - k T d ln T 0i ð13:131Þ
Ae = - þ - ;
dr 2 dr T 0e dr 1 þ ν2e ε2 dr
p 1
Γ2 = - K 13 n0 ε E ,
B0 Θ T

where

1 ε3 νe c2mn =bmn

K mn = K 0mn þ
1þ
1=2
amn νe þ bmn νe 1 þ cmn νe ε3=2

for m and n from 1 to 2,

1
K m3 = K 0m3 :
1=2
ð1 þ cm3 νe ε3=2 Þ 1 þ am3 νe þ bm3 νe

Coefficients are given in the table.

302 13 Transport Phenomena in Tokamaks

mn K 0mn amn bmn cmn

11 1.04 2.01 1.53 0.89
12 1.2 0.75 0.67 0.56
22 2.55 0.45 0.43 0.43
13 2.3 1.02 1.07 1.07
23 4.19 0.57 0.61 0.61
33 1.83 0.68 0.32 0.66

Electron heat flux:

p
5 2n0 εT 20e me νe d ln T 0e
hhqe ii þ hhΓiiT 0e = - K 12 Ae þ K 22
2 e B0 Θ
2 2 2 dr
ð13:132Þ
p 1
- K 23 n0 T 0e ε E :
B0 Θ T

Parallel current:
p
n0 εT 0e d ln T 0e p
jk = - K 13 Ae þ K 23 - σk E T 1 - K 33 ε : ð13:133Þ
ΘB0 dr

Ion heat flux:

5
hhqi ii þ hhΓiiT 0i
2
p ð13:134Þ
n0 εT 20i mi νi d ln T 0i 5 1 - kT
= - 2K 2 þ T 0i hhΓii - :
e B0 Θ
2 2 2 dr 2 1 þ ν2i ε3

Here

1 1:77ε3 νi
K 2 = 0:66 þ : ð13:135Þ
1þ
1=2
1:03νi þ 0:31νi 1 þ 0:74νi ε3=2

Neoclassical radial electric field:

T 0i ∂n 1 ∂T 0i
Er = þ kT þ B0p U k , ð13:136Þ
en ∂r e ∂r

where

1=2
- 0:17 þ 1:05νi þ 3:1ν2i ε3
kT = : ð13:137Þ
1=2
1 þ 0:7νi ð1 þ ν2i ε3 Þ
13.5 Particle and Heat Balance Equations 303

Particle balance equation:

∂n0 1 ∂
þ ðr hhΓiiÞ = I - R, ð13:138Þ
∂t r ∂r

where I and R are particle sources and sinks, respectively.

Heat balance equation:

3 ∂ðn0 T 0e Þ 1 ∂ 5
þ r hhqe ii þ hhΓiiT 0e = Qe ,
2 ∂t r ∂r 2
ð13:139Þ
3 ∂ðn0 T 0i Þ 1 ∂ 5
þ r hhqi ii þ hhΓiiT 0i = Qi :
2 ∂t r ∂r 2

Sources on the r.h.s. in the absence of additional heating are

Qe = jk E T - QΔ - eE r hhΓii,
ð13:140Þ
Qi = QΔ þ eE r hhΓii:

Here, the term QΔ represents the heat exchange between electrons and ions:

me
QΔ = 3 nν ðT - T i0 Þ, ð13:141Þ
mi e e0

and E r is the electric field in the reference frame moving with the toroidal velocity:

T 0i ∂n 1 ∂T 0i
E r = Er - B0p U k = þ kT : ð13:142Þ
en ∂r e ∂r

Additional terms on the r.h.s. of the heat balance equations represent the work of the
radial electric field, and other work of the poloidal electric field and friction are taken
into account during the calculation of radial heat fluxes hhqjii.
One should keep in mind that in the real tokamak, turbulent particle and heat
fluxes can significantly exceed neoclassical values, especially for particle and
electron heat fluxes. In such a situation, Eqs. (13.138) and (13.139) remain the
same. Only the work of the electric field could look different, since for neoclassical
fluxes, part of the wok of the poloidal electric field and friction was taken into
account while integrating over the flux surface.
For arbitrary flux surfaces, one should use the surface coordinates described in
Sect. 12.7. In these coordinates, the local particle balance has the form
p
∂n0 1 ∂ g
þp Γ = I - R, ð13:143Þ
∂t g ∂a ha a
304 13 Transport Phenomena in Tokamaks

p p
where g = ha hθ hζ . Let us multiply his equation by g and integrate over poloidal
and toroidal coordinates. Assuming constant density at the flux surface, the term on
the l.h.s. is transformed to
p
∂ gn0 dθdζ ∂
= ½V 0 ðaÞn0 ,
∂t ∂t

where V is the volume inside a given flux surface and V′(a) is the volume between
neighboring flux surfaces. From the second term, one obtains the net flux through the
flux surface. Multiplying and dividing it by V′(a) yield
p
g →
Γdθdζ = V 0 ðaÞ Γ ∇a :
ha

Here, brackets hi correspond to volume averaging Eq. (13.55). Finally, the averaged
particle balance equation is given by

∂ 0 ∂ →
½V ðaÞn0  þ V 0 ðaÞ Γ ∇a = V 0 ðaÞðI - RÞ: ð13:144Þ
∂t ∂a

As the flux surface coordinate a, it is possible to choose the equivalent radius of the
flux surface ρ, defined via the toroidal magnetic flux, πρ2 B0T = ΨT . Then, Eq. (3.144)
takes the form

∂ 0 ∂ →
½V ðρÞn0  þ V 0 ðρÞ Γ ∇ρ = V 0 ðρÞðI - RÞ: ð13:145Þ
∂t ∂ρ

For circular flux surfaces ρ = r, V′(r)~r and Eq. (3.145) is reduced to Eq. (13.38).
The heat balance equations in the general case are similarly:

3 ∂½V 0 ðρÞn0 T 0e  ∂ → 5 →
þ V 0 ð ρÞ q e ∇ρ þ Γ ∇ρ T 0e = V 0 ðρÞQe ,
2 ∂t ∂ρ 2
3 ∂½V 0 ðρÞn0 T 0i  ∂ → 5 →
þ V 0 ðρ Þ q i ∇ρ þ Γ ∇ρ T 0i = V 0 ðρÞQi :
2 ∂t ∂ρ 2
ð13:146Þ

The local radial electric field is

T 0i 1 d ln n 1 d ln T i B
Ea = þ kT - θ BV k ð13:147Þ
e ha da ha da B
13.6 Transport Codes 305

13.6 Transport Codes

The flux surface averaged transport equations analyzed in the previous section can
be used in 1D transport codes for the core region of a tokamak. However, in real
tokamaks, in addition to neoclassical transport, strong contributions to particle and
heat fluxes are produced by turbulent (anomalous) fluxes. This is taken into account
by adding turbulent diffusion, particle convection and turbulent heat conductivities
for electrons and ions to the corresponding averaged fluxes. The 1D codes can be
separated into two groups. The first group uses some models for turbulent transport
coefficients, and such codes are known as “theory-based” codes. Another approach
has a phenomenological character; that is, such codes’ turbulent transport coeffi-
cients are chosen to obtain a better fit with the experimentally observed density and
temperature radial profiles. Codes of the second group are widely used since “theory-
based” codes are rather complicated and have restrictions due to limited knowledge
of turbulent processes. Example of simulation results with code ASTRA of the
second type and comparison with experimental data is presented in Fig. 13.12.
The 1D codes cannot be used close to the separatrix and outside of it since the
plasma distribution is essentially 2D. In this region, edge plasma simulation can be

Fig. 13.12 Calculated density and temperature profiles, and comparison with experiment for
tokamak Globus-M. In the graphs (b) and (d), the data are the same as in (a) and (c) respectively,
but the scale is changed to show the details near separatrix
306 13 Transport Phenomena in Tokamaks

performed with codes such as SOLPS-ITER (see Sect. 12.9). However, to perform
2D modelling, particle and heat fluxes from the core should be specified at the core
side of the simulation domain as boundary conditions. They could be taken from 1D
code describing the core. On the other hand, plasma sources obtained in 2D
simulations can be used in 1D code after averaging over the flux surfaces, as well
as other averaged plasma parameters obtained in 2D modeling (density, tempera-
tures, potential, etc.). Iterating between 2 codes, after convergence, the simulation
gives final profiles in the whole region, as shown in Fig. 13.12.
Chapter 14
Instabilities in Magnetized Plasma

In this chapter, plasma stability in the situation when plasma equilibrium in a

magnetic field is achieved is considered. The most dangerous instabilities for plasma
confinement are large-scale magnetohydrodynamic (MHD) instabilities, which can
be described by MHD equations. There are also other types of instabilities of
inhomogeneous plasma with finite conductivity, in particular drift waves, which
were analyzed in Chap. 8.
The first type of MHD instability is rather similar to instability in fluids, so we
start with its consideration.

14.1 Rayleigh-Taylor Instability in Fluids

Consider fluid in the gravitational field, Fig. 14.1. Fluid density ρ0 changes with
height, ρ0 = ρ0(z). In equilibrium, the gravity force is balanced by the pressure
gradient,

dp0
- - ρ0 g = 0: ð14:1Þ
dz

Equilibrium becomes unstable if fluid density rises with height (heavy fluid on light
one). In this situation, Rayleigh-Taylor (RT) instability starts developing. The
opposite case is stable. Physically, the equilibrium is unstable to any perturbations:
if a parcel of heavier fluid is displaced downwards with an equal volume of lighter
fluid displaced upwards, the potential energy of the configuration is lower than the
initial state. Thus, the disturbance will grow and lead to a further release of potential
energy, as the denser material moves down under the gravitational field, and the less
dense material is further displaced upwards. This, for example, is observed when
water flows from inverted glass.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 307
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_14
308 14 Instabilities in Magnetized Plasma

Fig. 14.1 Heavy fluid on top of light one – Rayleigh-Taylor instability

The evolution of small perturbations is given by Euler equations

→
∇  u = 0,
∂ρ →
þ u ∇ρ = 0,
∂t ð14:2Þ
→
∂u → → →
ρ þ u ∇ u = - ∇p þ ρ g :
∂t

Perturbations are sought in the form of waves:

ρ1 = αðzÞ expð - iωt þ ikyÞ,

→1 →
u = β ðzÞ expð - iωt þ ikyÞ, ð14:3Þ
p1 = γðzÞ expð - iωt þ ikyÞ:

Amplitudes are unknown functions of coordinate z. The linearized system (14.2) is

given by

∂u1z
iku1y þ = 0,
∂z
∂ρ
- iωρ1 þ u1z 0 = 0,
∂z ð14:4Þ
- iωρ0 uy = - ikp1 ,
1

∂p1
- iωρ0 u1z = - - ρ1 g:
∂z

The velocity y-component u1y from the first equation can be substituted into the third
equation, and the pressure perturbation p1 from the third equation is substituted into
the last equation. Density perturbation ρ1 from the second equation should also be
substituted into the last one. As a result, we have a single differential equation for u1z :
14.1 Rayleigh-Taylor Instability in Fluids 309

d du1 dρ0 1
ω2 ρ0 z = k 2 ρ 0 ω 2 þ g u: ð14:5Þ
dz dz dz z

Here, the partial derivative is replaced by the full derivative since the velocity
amplitude depends on z only.
For a given wave vector k, Eq. (14.5) corresponds to the Sturm-Liouville problem
to seek eigenvalues and eigenfunctions for prescribed boundary conditions. At the
boundaries z = 0, L corresponds to a fixed or free boundary:

du1z ðz = 0, LÞ
u1z ðz = 0, LÞ = 0; = 0: ð14:6Þ
dz

It is possible to analyze the stability of the equilibrium without solving Eq. (14.5)
in the following way. Let us multiply both parts of Eq. (14.5) by complex conjugate

u1z and perform integration over z for the whole region. L.h.s. can be integrated by
parts, and the outside integral term turns to zero for any type of boundary condition.
For frequency, one obtains

dρ0 1 2
g u dz
dz z
- ω2 = 2
: ð14:7Þ
1
1 duz 2
ρ 0
þ u1z dz
k2 dz

Since all functions except derivative dρ0/dz are positive, the sign of r.h.s. depends on
the sign of derivative dρ0/dz. If density monotonically increases with z and dρ0/
dz > 0 for all z values, then ω2 < 0. In this case, ω = ± iγ, and root ω = iγ
corresponds to the increment of instability. In the opposite case, the frequency is real,
and the equilibrium is stable.
For the short wavelength case, kλ >> 1, where λ is the typical spatial density
scale λ = |d ln ρ0/dz|-1, Eq. (14.7) is reduced to

dρ0 1 2
g u dz
dz z
- ω2 = : ð14:8Þ
2
ρ0 u1z dz

In this situation, the increment for RT instability for dρ0/dz > 0 is of the order of
γ  g=λ. If eigenfunctions are strongly localized with scales smaller than λ, then
ρ0 and dρ0/dz can be taken outside the integral and

γ= g=λ: ð14:9Þ
310 14 Instabilities in Magnetized Plasma

To find the exact value of the increment and eigenfunctions u1z in the general case,
one has to solve Eq. (14.5). We shall demonstrate how the solution is obtained using
the example of exponential profile ρ0 = ρ(0) exp (z/λ) for fixed boundary conditions
u1z ðz = 0, LÞ = 0. Eq. (14.5) for this case is given by

d2 u1z ω2 du1z g
ω2 þ - k 2 ω2 þ u1z = 0: ð14:10Þ
dz2 λ dz λ

The solution is sought in the form u1z = A expðbzÞ, where the complex number
b = b1 + ib2. After substituting into Eq. (14.10), we find

ω2 g
ω2 b21 þ 2ib1 b2 - b22 þ ðb þ ib2 Þ - k2 ω2 þ = 0: ð14:11Þ
λ 1 λ

Since at the boundaries the velocity turns to zero, the imaginary part should be
b2 = πn/L, where is n an integer. As the frequency should be a real coefficient in
front of the imaginary terms, it is zero, and, hence, b1 = - 1/(2λ). The characteristic
equation is then given by

ω2 g
ω2 b21 - b22 þ b - k2 ω2 þ = 0, ð14:12Þ
λ 1 λ

and increment, we find

k2 g 1
γ2 = - ω2 = : ð14:13Þ
λ k2 þ 1 2 þ πn 2
4λ L

The real eigenfunctions are:

πnz
u1z = C sin expð- z=2λÞ cos ky expðγt Þ: ð14:14Þ
L

In particular, for short waves kλ >> 1, γ = g=λ.

14.2 Flute Instability

In fully ionized magnetized plasma, an analog of RT instability for fluids can

develop. It is caused by gravitational force or inhomogeneity or curative of magnetic
field. This instability of the RT family is called flute or interchange instability.
→
Consider plasma in a magnetic field parallel to x-axis, B kx, while gravitational
force is directed along z-axis. Equilibrium is described by equation
14.2 Flute Instability 311

Fig. 14.2 Polarization of a perturbation for flute instability

dp0
- - mi n0 g - j0y B = 0: ð14:15Þ
dz

Let us restrict ourselves to the case of small β when the magnetic field is assumed to
be constant and independent of z. The current in the y-direction consists of the sum of
the diamagnetic current and current j0g caused by the gravitational force. Current j0g is
associated with the drift of ions (electron drift is negligible due to small mass):
j0g = - mi gn0 =B.
The instability mechanism for short waves that are localized in the z direction in a
region much smaller than the plasma spatial scale is illustrated in Fig. 14.2. For
periodic density perturbation in the y-direction, perturbed current j1g = - mi n1 g=B
arises. To balance this current, an electric field arises, which produces a polarization
∂E 1
current in the opposite direction: j1p = nB0 m2 i ∂ty . The perturbed electric field rises with
→
time. For short waves, the condition ∇  j = 0 is equivalent to the condition of zero
perturbed current in the y-direction

mi n1 g n0 mi ∂Ey
1
j1g þ j1p = - þ 2 = 0: ð14:16Þ
B B ∂t

The mechanism of electric field formation is similar to that discussed in Sect. 11.1.
Perturbed electric field causes vertical drift with the velocity

u1z = - E 1y =B: ð14:17Þ

The linearized particle continuity equation considering that drift fluxes are diver-
→ →
gence free (∇  E × B =B2 = 0) has the form
312 14 Instabilities in Magnetized Plasma

∂n1 ∂n
þ u1z 0 = 0: ð14:18Þ
∂t ∂z

For perturbations, Eq. (14.3) from Eq. (14.16), one obtains

iBg n1
E1y = : ð14:19Þ
ω n0

Substituting Eq. (14.17) and (14.19) into Eq. (14.18) with account of ∂n1/∂t = -
iωn1 yields

d ln n0
- ω2 = g : ð14:20Þ
dz

Hence, for d ln n0/dz > 0, plasma is unstable with increment

γ= g=λ, ð14:21Þ

where λ = |d ln n0/dz|-1. In the opposite case, the frequency is real, and the plasma is
stable.
In the general case, flute instability is described by the equation system

∂n →
þ ∇  n u e = 0,
∂t
∂n →
þ ∇  n u i = 0,
∂t ð14:22Þ
→
dui → → → →
nmi = - ∇pi þ en E þ en u i × B þ mi n g ,
dt
→ → →
- ∇pe - en E - en u e × B = 0:

From momentum balance equations, perpendicular velocities can be derived:

→ → →
→
E×B ∇pe × B
ue= þ ,
B2 enB2 ð14:23Þ
→ → → → → → →
→
E×B ∇pi × B mi B × d u i =dt mi g × B
ui= - þ þ :
B2 enB2 eB2 eB2

Since diamagnetic fluxes are divergence free in a homogeneous magnetic field and
drift fluxes are incompressible, the continuity equation is reduced to

∂n →
þ u E ∇n = 0, ð14:24Þ
∂t
14.2 Flute Instability 313

→ → → →
where u E = E × B =B2 is the drift velocity. Choosing ∇  j = 0 as a second
equation, one finds

→ → → →
mi B × d u E =dt mi g × B
∇n þ = 0: ð14:25Þ
B2 B2

The expression for the polarization current velocity can be taken in the first approx-
→ →
imation, i.e., only E × B drift can be substituted. The linearized equation coin-
cides with Eq. (14.18) with account of Eq. (14.17). Linearized Eq. (14.25) is

mi n1 g n0 mi ∂Ey ∂ n0 mi ∂E 1z
1
ik - þ 2 þ = 0: ð14:26Þ
B B ∂t ∂z B2 ∂t

Using Eq. (14.18) for density perturbation, one has one equation for potential

d dφ1 dn
ω2 n0 = k 2 n 0 ω 2 þ g 0 φ1 : ð14:27Þ
dz dz dz

This equation coincides with Eq. (14.5) for fluid R-T instability, and, hence, stability
analyses in plasma are the same as in fluids described in the previous section.
Therefore, magnetized plasma is unstable if the gravitational force is directed
opposite to the density gradient. In the laboratory plasma, much more important is
the “effective gravity” associated with nonuniform and curvilinear magnetic fields
(see Chap. 11). In a tokamak, the effective acceleration is given by Eq. (11.11). The
instability increment remains of the order of that given by Eq. (14.21), where g is
replaced by the effective acceleration

2ðT e þ T i Þ
g= : ð14:28Þ
mi R

While developing instability, the wavelength along the magnetic field remains
infinite, so perturbations have a flute shape, which gives rise to instability. Magnetic
shear is a stabilizing factor for this instability and can suppress it. Indeed, perturba-
tion with kk = 0 at a given z in a sheared magnetic field will have kk ≠ 0 at other
z values. Then, positive and negative potential perturbations are “short-circuited”
since they belong to the same magnetic field line.
In magnetic traps, the situation can be rather complicated – effective acceleration
can change sign at different parts of magnetic field lines. In a tokamak, for example,
at the low-field side, the sign of acceleration corresponds to an instability increment,
while at the high-field side, the signs of effective acceleration and density gradient
coincide, which corresponds to stable equilibrium. To analyze stability in such
314 14 Instabilities in Magnetized Plasma

situations, let us consider a magnetic flux tube in plasma with a small β. Such a tube
tries to increase its volume and, therefore, to shift to the region where the tube
volume can become larger. Magnetic field line curvature should not rise during shift,
otherwise magnetic energy would increase. For closed field line, the magnetic tube
volume is

dl dl
V= Sdl = BS =Ψ , ð14:29Þ
B B

where Ψ is the magnetic flux, which is conserved along the tube and remains
constant over time. We shall define quantity U as

dl
U= : ð14:30Þ
B

For an infinite magnetic field line belonging to the flux surface,

1 dl
U = nlim : ð14:31Þ
→1 n B

where n is the number of toroidal rotations. Since the tube tries to expand and hence
to increase U, which is proportional to the tube volume, quantity -U might be
considered an analog of the potential energy. Since integral U is a flux surface
function similar to full pressure, one may consider p = p(U ). For a small shift of a
magnetic flux tube without a change in the field line curvature in accordance with
Eq. (14.29), δV/V = δU/U. For the adiabatic process, the pressure variation in the
tube is

δp = - 5=3pδV=V = - 5=3pδU=U:

In the tube neighborhood, the ambient plasma density also changes. Since p = p(U ),
we have

dp
pðU þ δU Þ = p þ δU: ð14:32Þ
dU

If the ambient plasma pressure rises faster than the tube pressure, the tube returns to
initial equilibrium. Hence, plasma would remain stable if

dp 5 p
>- : ð14:33Þ
dU 3 U

Usually, it is possible to neglect r.h.s. of Eq. (14.33), especially in the vicinity of

the plasma boundary, and the condition for plasma stability is reduced to
14.3 Dissipative Modifications of Flute Instability 315

dp
> 0: ð14:34Þ
dU

In other words, plasma remains stable if inside plasma where pressure is rising,
quantity U also has a maximum. Since the magnetic field in this region is smaller
than that outside, it is usually said that plasma should stay in the “magnetic pit” for
stability. In a tokamak, a “magnetic pit” exists due to the Shafranov shift of the inner
flux surfaces towards the low-field side. Hence, tokamak plasma is stable with
respect to fluid instabilities.
However, ballooning modes can develop in a tokamak. This is a modification of
flute instability where perturbation is a function of poloidal angle. On the low-field
side, the region of unfavorable curvature instability tends to grow, while on the high-
field side, the region of favorable curvature instability is suppressed. Finally, the
resulting instability occurs for a large plasma pressure gradient when it exceeds some
threshold value.

14.3 Dissipative Modifications of Flute Instability

14.3.1 RT Instability in Partially Ionized Plasma

In partially ionized plasma, a dissipative analog of flute (RT) instability exists. It is

also caused by gravitational force as well as magnetic field curvature. Let us consider
the geometry shown in Fig. 14.2 and perform analysis in local approximation.
Plasma is assumed to be magnetized so that xexi >> 1, Chap. 5. In partially ionized
plasma, as in fully ionized plasma, one gravitational force causes current in the y-
direction. For a periodic perturbation along y, the current perturbation equals

j1g = - biΛ mi n1 g: ð14:35Þ

For magnetized ions xi >> 1, the Hall mobility is biΛ = 1/B, and this current
coincides with that in fully ionized plasma. However, in contrast to the fully ionized
plasma case, this current is balanced not by the polarization current but by the ion
conductivity current provided that the ion-neutral collision frequency is sufficiently
large. Instead of Eq. (14.16), we have

- biΛ mi n1 g þ en0 bi⊥ E 1y = 0: ð14:36Þ

Here, the current generated by diffusion en1Di⊥k2 is neglected, which is justified for
(in the case of magnetized) ions
316 14 Instabilities in Magnetized Plasma

T i k2 νiN
<< 1: ð14:37Þ
mi g ωci

The perturbed electric field is obtained from Eq. (14.36),

mi g biΛ n1
E 1y = : ð14:38Þ
e bi⊥ n0

From the particle balance for electrons, Eqs. (14.18) and (14.17), one obtains the
frequency

g d ln n0
ω=i : ð14:39Þ
νiN dz

For d ln n0/dz > 0, the plasma is unstable with increment

g
γ= : ð14:40Þ
νiN λ

RT in partially ionized plasma is responsible, for example, for bubble formation in

the equatorial ionosphere.

14.3.2 Flute Instability in Plasma Contacting Metal Surfaces

If fully ionized plasma is restricted by highly conductive surfaces perpendicular to

the magnetic field, its polarization is to a large extent short-circuited. However, even
in collisionless plasma due to the finite conductivity of the boundary sheath,
polarization can survive, and plasma could be unstable.
Plasma is supposed to be restricted in the x-direction by conductive surfaces. In
the y-direction, the perturbed current is still described by Eq. (14.35), but now it
could be short-circuited by parallel currents along the x-axis. Integrating equation
*
∇  j = 0 over x between two surfaces, assuming symmetry over x, we have

ikI 1g þ 2j1k = 0: ð14:41Þ

Here

L
g g 1
I 1g =- n1 dx = - N ð14:42Þ
mi B mi B
0
14.3 Dissipative Modifications of Flute Instability 317

is the integral perturbed current, and j1k is the parallel current density at surface x = L.
To provide current to the surface, the potential should be perturbed with respect to
the floating potential. Potential perturbation corresponds to sheath current-voltage
characteristics Eq. (3.26). For floating unperturbed potential, linearized potential
perturbation is proportional to the current density

eφ1
j1k = ens cs , ð14:43Þ
Te

where cs = (Te/mi)1/2 is the sound speed velocity and ns is the density at the sheath
entrance. Due to high conductivity, potential perturbation is transported along the
magnetic field and is independent of x. In addition to potential Eq. (14.43),
Boltzmann’s potential φB = (Te/e) ln n exists in the plasma to provide parallel
→ →
momentum balance for electrons. The E × B drift caused by this potential is
divergence free and does not contribute to particle balance.
Particle balance Eq. (14.18) has the form

ickφ1 dn0
- iωn1 þ = 0: ð14:44Þ
B dz

Let us assume for simplicity that unperturbed density could be factorized n0 = n0x(x)
n0y(z), so that |d ln n0/dz|-1 = λ is independent of x. Then, since φ1 = const(x), the
perturbation density profile in the x-direction repeats the unperturbed profile n0(x).
The combination of Eqs. (14.41)–(14.44), hence, yields

i 2 2 Lx d ln n0
ω= k ρci g , ð14:45Þ
2 cs dz

where

L
ðT e =mi Þ1=2
Lx = n0 dx=ns , ρci = :
ωci
0

For d ln n0/dz > 0, the plasma is unstable with increment

g 2 c
γ= , νeff = 2 2 s : ð14:46Þ
νeff λ k ρci Lx
318 14 Instabilities in Magnetized Plasma

14.3.3 Gravitational-Dissipative Flute Instability

This modification of flute instability is connected with finite conductivity along the
→
magnetic field and finite wave length along B . This instability becomes important
when flute instability with kk = 0 cannot develop but local conditions for plasma
polarization, such as unfavorable magnetic field curvature, exist. The analysis is
performed analogously to previous cases, but in the current balance, the parallel
conductivity current is taken into account. The resulting frequency satisfies the
equation

ω2 þ iωs ωg þ ω2g = 0, ð14:47Þ

where

k 2x ωci ωce g
ωs = , ω2g = - : ð14:48Þ
k 2y 0:51νei λ

For ωs << ωg, we have standard flute instability, while for ωs >> ωg

ω2g
γ= << ωg : ð14:49Þ
ωs

14.4 Energy Principle

It is possible to formulate a principle based on energy release to investigate MHD.

We shall define plasma shift as

t
→ → →
ξ = u r , t 0 dt 0 : ð14:50Þ
0

→ →
Here, u r , t is the velocity of the plasma elementary volume during instability
onset.
Linearized MHD equations can be transformed using plasma shift. The linearized
equation for magnetic field perturbation in the collisionless limit is

→1
∂B →1 →0
=∇× u × B : ð14:51Þ
∂t
14.4 Energy Principle 319

Superscript «0» corresponds to equilibrium values, while superscript «1» corre-

→0
sponds to perturbed values, and u = 0. Integrating Eq. (14.51) over time from
→1 →
0 to t, taking into account that the initial perturbation is absent, B r , t = 0 = 0,
one obtains

→1 → → →0
B r ,t =∇× ξ × B : ð14:52Þ

The heat balance for electrons and ions is considered in the adiabatic approximation:

3 ∂ðnT α Þ 3 → →
þ ∇  nT α u α þ nT α ∇  u α = 0: ð14:53Þ
2 ∂t 2

After summing up, the equation for net heat balance is obtained:

3 ∂p 3 → →
þ ∇  p u þ p∇  u = 0: ð14:54Þ
2 ∂t 2

After linearization, we have

3 ∂p1 3 → 1 0 5 0 →1
þ u ∇p þ p ∇  u = 0: ð14:55Þ
2 ∂t 2 2
→
Integrating with account of p1 r , t = 0 = 0 yields

* → 5 0 →
p1 r , t = - ξ ∇p0 - p ∇ ξ: ð14:56Þ
3

Finally, momentum balance

→
du 1 → →
ρ = - ∇p þ ∇× B × B ð14:57Þ
dt μ0

after linearization is given by

→1 2→
∂u ∂ ξ 1 →1 →0 1 →0 →1
ρ0 = ρ0 2 = - ∇p1 þ ∇× B ×B þ ∇× B ×B : ð14:58Þ
∂t ∂t μ0 μ0

After substitution of Eqs. (14.52) and (14.56) into the r.h.s., we have

2→
∂ ξ →
ρ0 =F ξ , ð14:59Þ
∂t 2
320 14 Instabilities in Magnetized Plasma

→ → 5 → 1 →0 → →0
F ξ = ∇ ξ ∇p0 þ p0 ∇  ξ þ rot B × rot ξ × B
3 μ0
ð14:60Þ
1 → →0 →0
þ rotrot ξ × B × B :
μ0
→
Equation (14.59) can be interpreted as the Newton equation, and operator F ξ
corresponds to the force that causes a shift of the plasma element. Multiplying
•
Eq. (14.59) by ξ and integrating over volume, one finds:

• 2
→
• •• ρ 0
ξ •
→→ ∂ → → → → →
ρ ξ ξ dr =
0
dr = ξ F ξ dr: ð14:61Þ
∂t 2

Quantity
• 2
→
ρ0 ξ
→
K= dr ð14:62Þ
2

is the net kinetic energy associated with instability onset. Using Eq. (14.60), one can
demonstrate that the integral on the r.h.s. of Eq. (14.61) is a self-adjoint one,

→ * → * → →
ηF ξ d r = ξF η d r : ð14:63Þ

Hence, the following relation is satisfied:

*_ * → 1 *_ * → 1 * *_ → 1 ∂ * * →
ξ F ξ dr = ξ F ξ dr þ ξF ξ dr = ξ F ξ d r : ð14:64Þ
2 2 2 ∂t

Considering Eqs. (14.61) and (14.64) is reduced to

∂
ðK þ δW Þ = 0, ð14:65Þ
∂t

with
1 * * →
δW = - ξF ξ d r : ð14:66Þ
2

Quantity δW might be considered a variation in the potential energy of the whole

system. If for a given type of plasma shifts δW < 0, the kinetic energy K starts rising,
and plasma becomes unstable. If δW > 0 for all possible shifts, plasma equilibrium
14.5 Kink Instability 321

remains stable. Therefore, the energy principle provides an opportunity to analyze

the sign of δW to predict plasma stability for MHD perturbations.

14.5 Kink Instability

For flute instability, the source of energy is plasma thermal energy, while for kink
instability, magnetic field energy serves as energy storage. Below, several examples
of kink instability are presented.
Let us consider the plasma cylinder to be infinite in the z direction. The current
flows along z, and a parallel magnetic field is absent, so this is the z-pinch considered
in Chap. 12. Such pinch can be unstable with respect to radial perturbations

ξ = Aðr Þ expð- iωt þ imθ þ ikzÞ: ð14:67Þ

Here, m = kθr integer number due to periodicity over azimuthal coordinates. The
plasma element shift ξ here takes place in the radial direction. For m = 0, the
perturbation has azimuthal symmetry and is of the constriction type (Fig. 14.3). In
the region of constriction, where the plasma radius decreases, the azimuthal mag-
netic field increases since the full current I is conserved and Bθ~r-1. Hence, the
Ampere force IBθ rises, pinch contracts, and perturbation further develops. Plasma
pressure also rises but not as quickly since pressure can equalize along pinch, so it
cannot stop contraction.

Fig. 14.3 Constrictions

onset in z-pinch
322 14 Instabilities in Magnetized Plasma

To study kink instability with m = 0, one can use the energy principle. Let us
assume that pinch is restricted in the radial direction by the metal wall to exclude the
vacuum region from consideration. Stability condition δW > 0 Eq. (14.66) is
simplified for m = 0. After integrating by parts in Eq. (14.66), one finds

5 0 B20 → 2 4B20 → 2 dp0 B20 →

δW = p þ ∇ ξ - ξ∇  ξ þ ξ2 þ d r > 0:
3 μ0 μ0 r r dr 2πr
ð14:68Þ

Here

→ 1 d
∇ ξ = ðrξÞ:
r dr

Expression under the integral can be rewritten in the following way:

4B20 2
5 0 B20 → μ0 r →
δW = p þ ∇ ξ - ξ dr
3 μ0 2 5 0
þ μ0
B2
3p 0

ð14:69Þ
4B20 2 2
μ0 r ξ 2 dp 0 2B20 →
þ - þ ξ2 þ d r > 0:
5 B2 r dr μ0 r
4 p0 þ 0
3 μ0

The first integral is positive, so Eq. (14.69) is satisfied if the second integral is also
positive, and, hence,

r dp0 10=3
- < , ð14:70Þ
p0 dr 1 þ 56 β

where β = 2μ0 p0 =B20 . Therefore, inequality (14.70) should be satisfied to stabilize

the kink mode m = 0.
The longitudinal magnetic field stabilizes mode m = 0 since magnetic field lines
during contraction are compressed together with plasma and magnetic pressure
increases. However, on the other hand, mode m = 1 can become unstable. Let us
consider instability for the m = 1 mode and thin current carrying pinch in the
longitudinal magnetic field. The perturbed pinch with m = 1 is shown in
Fig. 14.4. Since the current flows along the perturbed pinch, the current’s azimuthal
component arises, and the radial Lorentz force jθBz leads to instability onset. The
azimuthal current depends both on the radial displacement and on the helix pitch
distance along z. The radial momentum balance reads
14.5 Kink Instability 323

Fig. 14.4 Instability of thin

current-carrying pinch for
m = 1 mode

d2 ξ
ρ = jθ Bz = jk Bz kξ: ð14:71Þ
dt 2

Hence, increment

1=2
IkBz
γ= : ð14:72Þ
πa2 ρ

Expressing current through the azimuthal magnetic field, one obtains

cA
γ= ð2ΘkaÞ1=2 , ð14:73Þ
a

where a is the pinch radius, Θ = Bθ/Bz, and cA = Bz/(μ0ρ)1/2 is the Alfven velocity
calculated for magnetic field Bz. Therefore, the characteristic time for kink instability
onset is determined by the Alfven time, in contrast to flute instability, where it is
determined by the sound speed time.
To obtain the stability criterion for a cylinder of finite radius with respect to mode
m = 1, we consider a simplified model. Let us assume that a helical perturbation is
added that does not change the cylinder shape. Two cross-sections separated by a
quarter of the wavelength are shown in Fig. 14.5. Assuming that at z = 0 pinch is
shifted to the left (dotted line), then at distance z = λ/4, pinch should be shifted
downwards. Let us consider the situation when the magnetic field line at distance
z = λ/4 turns on an angle larger than π/2. Since the magnetic field is frozen into the
plasma, the magnetic field line turns at the same angle with respect to the center of
the perturbed pinch (dotted line). At the same time, with respect to the unperturbed
pinch, the magnetic field line turns on a larger angle. To provide rotation on a larger
angle, the magnetic field at the upper part of the pinch should rise with respect to the
equilibrium angle. The magnetic pressure in the upper part also increases, and,
therefore, the pinch is further shifted downwards, which leads to instability onset.
324 14 Instabilities in Magnetized Plasma

Fig. 14.5 Displacements of pinch cross-sections separated by λ/4 in the process of kink instability
onset for m = 1

Fig. 14.6 Lines of constant phase for radial displacement and magnetic field lines in the plane
(z, θ). Situation corresponds to instability onset

In the opposite case, when the magnetic field line turns on an angle smaller than π/2
for length λ/4, the pinch is stable. The safety factor for the cylinder can be defined as

Bz 2πa Bz ka
qð aÞ = = : ð14:74Þ
Bθ λ Bθ

Therefore, pinch remains stable for

q > 1: ð14:75Þ

This inequality is known as the Kruskal-Shafranov condition.

14.5 Kink Instability 325

For mode m ≥ 2, one can use the following arguments. Let us unveil the pinch
surface as a rectangle, as shown in Fig. 14.6. The vertical axis corresponds to
coordinate z, and the horizontal axis represents azimuth angle θ. Radial perturbations
shift the plasma boundary, and a rippled surface is formed (Fig. 14.6). The longitu-
dinal pinch current also shifts radially together with plasma and current displace-
ments, leading to an increase in perturbations of the magnetic field radial component
B1r . A time-dependent perturbed radial magnetic field induces an electric field. Due
to very high plasma conductivity, this inductive electric field should be practically
perpendicular to the magnetic field with an almost zero parallel component. The
→ →
induced electric field in turn produces radial E × B drift of the plasma element. The
magnetic field perturbation and displacement are connected according to Eq. (14.52)
as B1r  iBz ξkk ≈ iξðkBz - mBθ =aÞ. Here, the sign in the brackets should be negative
to correspond to an outwards current shift. In other words, the induced electric field
should cause outwards plasma drift, further increasing displacement (Fig. 14.6). In
the opposite case, plasma remains stable. Hence, the stability criterion is q > m. For a
cylinder of finite length L, the wave vector k = 2πn/L. In particular, for a torus with
radius R, we have k = n/R. In this situation, the stability criterion has the form

Bz a m
qð aÞ = > : ð14:76Þ
Bθ R n

The condition (14.76) should be satisfied at the plasma boundary. If the current
density is constant as a function of radius, the azimuth magnetic field increases
linearly with radius, and the q value is independent of radius. In this model case,
Eq. (14.76) is satisfied for the whole plasma volume. For any realistic current profile,
the safety factor depends on the radius. For the standard current density distribution,
when the current density decreases with radius, q(r) rises with radius. Hence, inside
the plasma pinch, many flux surfaces exist with radius r = rres, where

m
qðr res Þ = : ð14:77Þ
n

The corresponding flux surfaces are known as resonance flux surfaces. At the
resonance surface, the kink mode is neutral stable, but at different sides, it is either
stable or unstable, since due to dependence q(r) condition q(r) > m/n is satisfied at
one side and is not satisfied on the other side. Stability in such situations depends on
the q derivative, on the rate of q variation with radius, in other words, on magnetic
field shear.
The required shear can be estimated for the following reasons. Let us consider a
mode localized in the vicinity of the resonance flux surface. Pressure perturbation
according to Eq. (14.56) is p1~ - ξdp/dr. Pressure perturbation and magnetic field
curvature RS = rBz/Bθ produce a destabilizing force

dp ξ
F1  - , ð14:78Þ
dr Rs
326 14 Instabilities in Magnetized Plasma

which tries to accelerate plasma outwards. This force is balanced by the stabilizing
force associated with magnetic field deformation, which can be estimated as

→ →
B∇ B iBz kk B1r
F2 =  : ð14:79Þ
μ0 μ0

Using Eq. (14.52), B1r  iBz ξkk , and expansion kk in the resonance flux surface
vicinity Eq. (8.92) k k ðr Þ = k θ Θðr - r res Þ=r res , where Θ is the magnetic shear intro-
duced in Chap. 8, assuming kθ(r - rres)~1, from F1 = F2, we obtain

2
- dp=dr B2z Θ
< : ð14:80Þ
B2z =ð2μ0 Þ B2θ r

This inequality is known as the Suydem stability criterion. Therefore, sufficiently

strong magnetic shear can suppress plasma instabilities localized in the vicinity of
resonance flux surfaces. A similar Mercier criterion was obtained for toroidal
geometry.

14.6 Tearing Instability

As shown in the previous section, MHD instabilities localized in the vicinity of

resonance flux surfaces can be stabilized by magnetic field shear. However, in the
vicinity of resonance flux surfaces, another special instability can develop, which is
connected with a change in magnetic field topology. This is known as the tearing
mode. This instability is connected with finite conductivity and is accompanied by
the formation of magnetic islands – regions with different magnetic field topologies
near resonance magnetic flux surfaces.
Let us analyze the tearing instability for the plasma slab in Fig. 14.7. The
unperturbed magnetic field B0y ðxÞ is directed along the y-axis and is a function of
the x-coordinate (linear, for example). At x = 0, the magnetic field turns to zero,

Fig. 14.7 Unperturbed magnetic field near neutral layer

14.6 Tearing Instability 327

which corresponds to the neutral layer. Such a magnetic field is created by a current
sheet with j0z = ð1=μ0 Þ∂B0y ðxÞ=∂x. Plasma equilibrium is supposed to be satisfied, so
that

∂p0
- þ B0y j0z = 0: ð14:81Þ
∂x

We consider the perturbation periodic in the y-direction and independent of z:

→
B1x  expð- iωt þ ikyÞ. However, according to equation ∇  B = 0, perturbed are
both components of the magnetic field, which are connected by

∂B1x
þ ikB1y = 0: ð14:82Þ
∂x

At some distance from the resonance flux surface, magnetic field lines are
perturbed, as shown in Fig. 14.8, without changing topology. The perturbed time-
dependent magnetic field induces an electric field in the z-direction:

∂B1x ∂E 1z
=- : ð14:83Þ
∂t ∂y

Since tearing instability is aperiodic with only an imaginary part of the frequency,
the perturbed electric field is shifted in phase by π/2 with respect to the B1x
perturbation. Electric field causes plasma drift along the x-axis with the velocity

E1z
u1x = - : ð14:84Þ
B0y

The drift velocity is directed towards the neutral layer x = 0 in the regions where
magnetic field lines are compressed and away from it in the regions of magnetic field
line divergence.
At the neutral layer, the velocity Eq. (14.84) becomes infinity when the
unperturbed magnetic field goes to zero. Hence, in the vicinity of the neutral layer,
it is necessary to take into account finite conductivity, and instead of Eq. (14.84), one
should use the electron momentum balance along the z-axis:

E1z = - u1x B0y þ ηj1z , ð14:85Þ

where resistivity η = σk- 1 = 0:51 me νei =ne2 . Currents in the z-direction flow in a
thin layer in the vicinity of x = 0, where the magnetic field is close to zero. They
create a magnetic configuration with closed field lines – magnetic islands. Indeed, a
positive magnetic field perturbation B1x near x = 0 shifts the magnetic field line to
positive values of x. While moving along x, the unperturbed magnetic field increases,
328 14 Instabilities in Magnetized Plasma

Fig. 14.8 Magnetic island formation

and the magnetic field line starts going in the y direction. The phase of the perturbed
magnetic field B1x starts to change, and, hence, the magnetic field line returns to the
neutral line x = 0 and then shifts further to the region of negative x. Here, the
unperturbed magnetic field changes sign and shifts the magnetic field line in the
opposite -y direction. The phase of the magnetic field perturbation B1x changes
again, and, as a result, the magnetic field line performs a full turn, forming a
magnetic island (Fig. 14.8). Far away from the neutral line, magnetic field pertur-
bation B1x leads only to deformation of the magnetic field line without changing
topology. A line separating closed- and open-field lines is known as the separatrix.
In the vicinity of the neutral line, plasma drift motion in the x-direction caused by
E 1z has different signs for positive and negative x (to the neutral line or away from it)
due to different signs of the unperturbed magnetic field. Hence, here, the plasma
motion along the x-axis should transform into motion along the y-axis.
→
Plasma motion is assumed to be incompressible (∇  u = 0), so that

∂u1x
þ iku1y = 0: ð14:86Þ
∂x

Linearized Eq. (10.14) for magnetic field perturbation is reduced to

2
∂B1x η0 ∂ B1x
= iku1x B0y þ - k 2 B1x : ð14:87Þ
∂t μ0 ∂x2

This equation contains a small coefficient η0 in front of a high derivative. Far from
the neutral layer, the last term on the r.h.s. is small, but while approaching x = 0, the
second derivative becomes large, and the term ∂B1x =∂t is balanced by the second
2
derivative ∂ B1x =∂x2 multiplied by the small coefficient η0. Looking from large
2
distances, the second derivative ∂ B1x =∂x2 has a break, while in reality, the first
14.6 Tearing Instability 329

derivative changes quickly inside a boundary resistive layer with width 2ε. Let us
introduce a quantity

dB1x dB1
Δ0 = ðþεÞ - x ð- εÞ =B1x ð0Þ: ð14:88Þ
dx dx

The value of Δ′ is determined by the solution of MHD equations in the outer region
with corresponding boundary conditions.
Inside the resistive layer, the second derivative can be estimated as

∂ B1x Δ0 B1x
2
 : ð14:89Þ
∂x2 ε

Inside the resistive layer, all terms of Eq. (14.87) are of the same order. Introducing
increment γ, one obtains

2
c2 η0 ∂ B1x
γB1x  : ð14:90Þ
μ0 ∂x2

From (14.89) and (14.90), we have

c2 η 0 Δ 0
γ : ð14:91Þ
μ0 ε

Another relation between the increment and resistive layer width could be obtained
from the momentum balance equation. In the x-direction j1z B0y , the force is balanced
by the pressure gradient ip1/ε:

j1z B0y  ip1 =ε: ð14:92Þ

Pressure perturbation accelerates plasma along the layer in the y-direction, and

ρ0 γu1x
ikp1  ρ0 γu1y  : ð14:93Þ
kε

We used Eq. (14.86). It is supposed that the resistive layer is thin, so that

kε ≪ 1, ð14:94Þ

and since according to Eq. (14.86), plasma is incompressible, u1y ≫ u1x -plasma is
strongly accelerated along the layer. After eliminating the pressure perturbation from
Eqs. (14.92) and (14.93), we find
330 14 Instabilities in Magnetized Plasma

ρ0 u1x γ
j1z  : ð14:95Þ
ðkεÞ2

From the Maxwellian equation, we have

1 ∂ B1x B1x Δ0
2
j1z   , ð14:96Þ
μ0 k ∂x2 μ0 kε

and a combination of Eqs. (14.96) and (14.97) yields

B1x Δ0
γρ0 u1x  kεB0y : ð14:97Þ
μ0

The unperturbed magnetic field can be estimated as

0
B0y  B0y ε: ð14:98Þ

0
Substituting estimate γB1x  ku1x B0y  ku1x B0y ε in accordance with Eq. (14.87) into
Eq. (14.97), after the elimination velocity, one finds

0 2
B0y k 2 ε3 Δ 0
γ2  : ð14:99Þ
μ0 ρ 0

Combining Eqs. (14.99) and (14.91), we find an increment of resistive tearing

instability

0 2=5
k2=5 ðΔ0 Þ
4=5 3=5
B0y ðη0 Þ
γ : ð14:100Þ
ðμ0 Þ4=5 ðρ0 Þ1=5

Let us denote the typical scale of plasma parameter variation along x as L and
introduce two characteristic times

1=2
L Lðμ0 ρ0 Þ L2 μ0
τA = = ; τs = , ð14:101Þ
cA 1=2 η0
B0y

where τA is the Alfven time and τs is the skin time. Then, Eq. (14.100) may be
rewritten in terms of τA and τs, giving
14.7 Geodesic Acoustic Mode and Zonal Flows 331

- 2=5 - 3=5
γ  τA τs : ð14:102Þ

Therefore, resistive tearing instability grows on a time scale that is intermediate

between a short MHD time scale τA and a long skin time scale τs. The resistive layer
width may be obtained from (14.99):

2=5
ε τA
 ≪ 1: ð14:103Þ
L τs

The tearing instability grows exponentially until the magnetic island width
becomes comparable to the resistive layer width ε, after which nonlinear effects
become important. In a cylinder or torus, tearing instability can develop around
resonance flux surfaces, q = m/n; see the next chapter. In a tokamak, the so-called
neoclassical tearing mode can also exist, and the latter is caused by bootstrap current
variation due to the formation of magnetic islands.

14.7 Geodesic Acoustic Mode and Zonal Flows

In a tokamak-specific oscillation of density, poloidal and parallel velocities and a

radial electric field are observed, known as the geodesic acoustic mode (GAM). We
shall consider GAM for a tokamak with a simple circular cross section, Sect. 13.1.
Let us assume that in addition to the stationary radial electric field given by
Eq. (13.62), an oscillating perturbed radial electric field exists. Corresponding
perturbed potential

0 ðr Þ = A ðr Þ expð- iωt Þ
φG ð14:104Þ
G

is independent of the poloidal coordinate. Amplitude AG(r) is an arbitrary function of

radius, so the radial scale of this mode remains unspecified. Perturbed radial electric
→ →
field E G0 = - dφ0 =dr causes poloidal E × B drift in addition to stationary drift with
G

velocity V G0 = - E 0 =B. Corresponding poloidal flows are known as zonal flows

G

analogously to zonal flows in planet atmospheres.

In a nonuniform magnetic field, velocity V G 0 has finite divergence, which pro-
duces close parallel plasma fluxes similar to the formation of Pfirsch-Schlueter fluxes
(Fig. 14.9). These parallel fluxes, in contrast to the Pfirsch-Schlueter fluxes, period-
ically change their direction, and their amplitude is proportional to cosθ. The parallel
velocity, hence, is

ik ðθ Þ = uik cos θ,
uG ð14:105Þ
a
332 14 Instabilities in Magnetized Plasma

Fig. 14.9 Poloidal velocities and density perturbations in geodesic acoustic mode

a
with uik being a periodic function of time. These parallel fluxes are driven by a
parallel pressure gradient, so periodic pressure perturbation proportional to sinθ
should exist, namely,

1 = n1 sin θ:
nG ð14:106Þ
a

Therefore, the perturbed radial electric field has a poloidal wavenumber m = 0, and
the density and parallel velocity perturbations correspond to m = 1.
The linearized particle balance equation has the form

n0 Θuak 2ε
- iωnG
1 - 0 sin θ = 0:
sin θ - n0 V G ð14:107Þ
r r

Here, the second term represents divergence of the poloidal projection of parallel
flux, and the third term corresponds to divergence of the poloidal flux associated
with the poloidal rotation. The parallel momentum balance is given by

Te þ Ti a
- iωmi n0 uaik cos θ = - Θ n1 cos θ: ð14:108Þ
r

We assume that species temperatures remain unperturbed due to large heat conduc-
tivity and heat exchange.
To determine the radial electric field, one needs the current balance equation. In
particular, it is sufficient to put the net current integrated over the flux surface to zero.
Due to density perturbations, the vertical ∇B-driven current integrated over the flux
surface remains finite and is equal to

2π
2ðT e þ T i Þ G 4π2 rðT e þ T i Þna1
hhj∇B ii4π2 rR = - 4π2 rR n1 sin θdθ = - : ð14:109Þ
BR B
0
14.8 Equatorial Plasma Bubbles 333

This current is compensated by the polarization current caused by radial electric field
temporal variation:

mi n0 ∂V G
4π2 rRhhjP ii = - 4π2 rR 0
: ð14:110Þ
B ∂t

From the condition of zero radial current hhj∇Bii + hhjPii = 0, one obtains

c2s ε nG
1
- iωV G
0 = 0, ð14:111Þ
r n0

where cs = ðT e þ T i Þ=mi . From Eqs. (14.107) and (14.108), we have for

amplitudes

na1 Θuk 2ε G
a
iω þ þ V 0 = 0,
n0 r r
ð14:112Þ
Θc n
2 a
- iωuaik þ s 1 = 0:
r n0

By setting the determinant of the system (14.111) and (14.112) to zero, we obtain the
dispersion relation for frequency, whose solution is

c2s 1
ω2 = 2þ 2 : ð14:113Þ
R2 q

In the approximation considered, GAM has a real frequency of the order of ion
bounce frequency. According to nonlinear analysis, GAM has increment or decre-
ment due to nonlinear interaction with drift ways and can stay in a dynamic balance
with them.

14.8 Equatorial Plasma Bubbles

Plasma bubbles that occur in the equatorial ionosphere are plasma depletion regions
with respect to the background ionosphere density. They are observed in the
equatorial Earth F-region of the ionosphere at heights, where ionospheric plasma
density increases with height. Such depletions can be directly measured in situ by
satellite probes. Their formation is a result of the nonlinear development of Ray-
leigh-Taylor instability in the partially ionized plasma considered in Sect. 14.3.1.
Indeed, in the equatorial ionosphere, Earth magnetic field lines are parallel to the
Earth surface and hence are perpendicular to the density gradient. Below the F-
region maximum, the density gradient is directed upwards and is antiparallel to the
gravity force mig. This bottom layer is unstable due to gravitational RT instability
334 14 Instabilities in Magnetized Plasma

Fig. 14.10 Contour plots of density perturbations at 2000 s, 5000 s, and 10,000 s. The plus
and minus signs indicate the enhancements and depletions of the electron density relative to
the background. The dashed curve represents the electron density profile of the background
ionosphere

because it is located below the heavy layer. An increment is given by Eq. (14.40). A
perturbation initiated in the bottom layer rises like a bubble and penetrates the top
layer (layer above the F-region maximum). Electron density irregularities exist along
the trace of the bubble, as well as within and around the rising bubble. Examples of
bubble evolution obtained in the simulations are shown in Fig. 14.10.
Chapter 15
Magnetic Islands and Stochastic Magnetic
Field

15.1 Magnetic Islands

In the previous chapter, it was demonstrated that tearing instability leads to the
formation of magnetic islands. The same configuration can be created by magnetic
field perturbations caused by currents in the external coils. We shall study the
magnetic island structure in more detail independently of the mechanism of their
formation. Let us consider the same magnetic field B0y ðxÞ as in the previous section,
Fig. 14.7, i.e., changing sign at x = 0. Near x = 0, magnetic field is a linear function

0
B0y ðxÞ = B0y x: ð15:1Þ

In the general case, in addition to the magnetic field B0y ðxÞ, there exists a magnetic
field component B0z ; here, for simplicity, we shall assume B0y ðxÞ ≪ B0z . Magnetic field
perturbations are supposed to be stationary and contain the sum of harmonics

B1x = B →k exp ik y y þ ik z z : ð15:2Þ

ky , k z

→
First, we consider the case B0z = 0 and kz = 0. Due to condition ∇  B = 0, it is
possible to introduce a magnetic flux function for the full perturbed magnetic field as

∂Ψ ∂Ψ
Bx = - , By = : ð15:3Þ
∂y ∂x

For a single harmonic with real amplitude B1x = B →k sin k y y

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 335
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_15
336 15 Magnetic Islands and Stochastic Magnetic Field

1 0 0 2 B →k
Ψ = Ψ0 þ B x þ cos ky y: ð15:4Þ
2 y ky

One must keep in mind that both components of the magnetic field B1x and B1y are
perturbed since perturbation B1x is a function of x. Line Ψ = const coincides with the
magnetic field line (Fig. 14.9). Indeed, along this line, dΨ = dx∂Ψ/∂x + dy∂Ψ/
∂y = 0, and, therefore, according to Eq. (15.3), along this line, -Bydx + Bxdy = 0.
The latter equation coincides with the equation for the magnetic field line. According
to Eq. (15.4), at the X-point at the separatrix with coordinates x = 0, y = 0, the value
of the flux function is

B →k
Ψ = Ψ0 þ : ð15:5Þ
ky

At the O-point at the separatrix with coordinates x = W/2, y = π/ky we have

1 0 0 W 2 B →k
Ψ = Ψ0 þ B - : ð15:6Þ
2 y 2 ky

Here, W is the maximal width of the separatrix (island width). From Eqs. (15.5) and
(15.6), the island width is determined:

1=2
B →k
W =4 0 : ð15:7Þ
ky B0y

In the presence of field B0z , the magnetic field line goes mainly along the z-axis.
For a single harmonic with kz = 0, the magnetic field line projection to the xy plane
acts in the same way as for B0z = 0, forming magnetic islands in the vicinity of x = 0.
For harmonics B1x = B →k exp ik z z þ ik y y with finite wave vectors ky and kz, plane
x = x0 is of special character since the magnetic field perturbation phase remains
constant along the unperturbed magnetic field line. The plane x = x0 is called the
resonance plane. Phase kzz + kyy = 0 is constant along the straight line z = - (ky/kz)
y. The equation for the unperturbed magnetic field line is given by the straight line
0
z = yBz0 = B0y x0 . These two lines coincide when

k z B0z
x0 = - 0 : ð15:8Þ
ky B0y

Since wave vectors can have the same signs or different ones, the value x0 could be
positive or negative.
15.1 Magnetic Islands 337

Fig. 15.1 Magnetic islands in the vicinity of resonance flux surfaces

Similar to case B0z = 0, the magnetic field line moves together with the
unperturbed line, making turns around it. For an observer moving along an
unperturbed magnetic field, its end forms an island. Island formation is caused by
a change in the magnetic field y-component when the magnetic field line is shifted
along x and a further change in the perturbation B1x phase. The projection of the
magnetic field line along the unperturbed line on the plane xy forms the structure
shown in Fig. 15.1. The island width in the general case is still given by Eq. (15.7).
In the presence of several harmonics, a system of magnetic islands arises, and each
island is formed in the vicinity of its resonant flux surface in accordance with
Eq. (15.8).
In a tokamak, magnetic islands are formed in a similar way around resonant flux
surfaces. Let, for circular magnetic flux surfaces, us consider radial magnetic field
perturbations, namely,

B1r = Bmn expðimθ - inζÞ: ð15:9Þ

m, n

Resonant magnetic surfaces are defined by Eq. (14.77),

m
qðrres Þ = : ð15:10Þ
n
338 15 Magnetic Islands and Stochastic Magnetic Field

The role of the magnetic field B0y changing with x plays poloidal magnetic field B0θ
changing with radius. One can introduce a poloidal magnetic field additional to that
on the resonant flux surface:

B0
θ = Bθ ðr Þ - Bθ ðr res Þ:
0 0
ð15:11Þ

It is connected to the corresponding flux as


B0
θ = ∇Ψ × ∇ζ: ð15:12Þ

Here, in contrast to Eq. (12.91), 2π is included in the flux function. The resonant
harmonic forms a magnetic island due to the radial poloidal magnetic field depen-
dence in Eq. (15.11), analogous to the slab geometry. The magnetic field line turns in
space around the unperturbed magnetic field line. Analogous to Eq. (15.3), the flux
function for the net magnetic field is introduced for a single harmonic Eq. (15.9):

Ψ = Ψ ðr Þ þ Ψr cosðmθ - nζÞ: ð15:13Þ

Phase ky + kzz in the slab geometry is now replaced by phase mθ - nζ. The perturbed
field B1r is obtained as

∂Ψ
B1r = - : ð15:14Þ
r∂ðθ - ζ=qÞ

As in the slab case, the flux function can be expanded in the vicinity of the resonant
flux surface

1 d 2 Ψ
Ψ = Ψ ðr res Þ þ ðr Þðr - r res Þ2 þ Ψr cosðmθ - nζÞ: ð15:15Þ
2 dr 2 res

Similar to the slab case, one obtains

1=2
Bmn q
W = 4r ð15:16Þ
mBθ rq0

In the general case, radial magnetic field perturbations lead to the formation of
magnetic island chains near resonance flux surfaces given by Eq. (15.8) for the slab
case and by Eq. (15.10) in a toroidal geometry.
15.2 Stochastic Instability and Magnetic Field Line Diffusion 339

15.2 Stochastic Instability and Magnetic Field Line

Diffusion

In the presence of several magnetic field harmonics with sufficient amplitudes, the
magnetic field can become stochastic. This occurs when the distance between
neighboring flux surfaces Δx0 = x0 k y , kz - x00 k 0y , k0z Þ (or Δr res = jr res ðm, nÞ
- r 0res ðm0 , n0 Þj in the cylindrical geometry) becomes smaller than the single island
width. This condition of stochastization is known as Chirikov criterion

W ky , k z þ W 0 k 0y , k0z
Δx0 < : ð15:17Þ
2

As shown below, the equation describing the perpendicular displacement of the

magnetic field line is fully equivalent to the Newton equation, so the mechanical
formalism can be applied. We shall discuss qualitatively corresponding results.
Let us consider a group of magnetic field lines forming a flux tube. One can move
along the magnetic field line marking the passing distance as l and marking the
position of the magnetic field line in the plane perpendicular to the magnetic field.
This process is known as mapping or a Poincare plot. At l = 0, many magnetic field
lines form a circle, as shown in Fig. 15.2a. If the Chirikov criterion is not satisfied,
then the tube cross-section remains similar to the initial cross-section with an
increase in l, with some deformation due to the perpendicular magnetic field. In
contrast, when the Chirikov criterion is fulfilled, the character of the magnetic field
line end motion in the perpendicular plane changes qualitatively. According to

b c

Fig. 15.2 Deformation of magnetic tube cross-section with distance along magnetic field. (a)
Initial cross-section l = 0, (b) start of stochastic instability onset l > Lc, and (c) random walk for
l ≫ Lc
340 15 Magnetic Islands and Stochastic Magnetic Field

general theory, motion becomes unstable according to the Kolmogorov-Arnold-

Moser theorem. The so-called Kolmogorov length Lc exists, so that for l > Lc,
magnetic field lines start to diverge exponentially. This phenomenon is known as
stochastic instability. At larger distances, ergodization starts – a random walk of the
magnetic field line ends in the perpendicular plane. The flux tube cross-section starts
to deform strongly, stretching in one direction and squeezing in the other direction,
since the tube cross-section area should be conserved due to magnetic flux conser-
vation (Fig. 15.2b). With a further increase in distance l, the magnetic tube cross-
section becomes rather complicated, as shown in Fig. 15.2c. The length of the
“sleeves” increases and their width decreases to maintain a constant cross-sectional
area.
The average square of radial displacement, in particular displacement in the x-
direction, at l > Lc is proportional to distance l in accordance with random walk
displacement character:

ðΔxÞ2 = 2Dst l: ð15:18Þ

Here, quantity Dst is the magnetic field line stochastic diffusion coefficient.
To calculate the stochastic diffusion coefficient in the presence of magnetic field
perturbations, one must first analyze the magnetic field line equations for the
perturbed magnetic field line. In the slab geometry

dx Bx dy By x
= ; = = , ð15:19Þ
dl B dl B a
0
where a = B0z = B0y . Integration of the second equation along the magnetic field
line, which at l = 0 corresponds to x = x0 and y = 0, yields

x
xl 1
y= 0 þ x0 dl: ð15:20Þ
a a
x0

For a single harmonic, which satisfies Eq. (15.8), inserting Eq. (15.20) into the
perturbation phase, one obtains the first of Eq. (15.19) in the form

x
dx B →k ky
= 0 exp i x0 dl : ð15:21Þ
dl Bz a
x0

Now, we demonstrate that Eq. (15.21) is equivalent to the Newton equation for
charged particle motion in the field of electrostatic longitudinal waves. For periodic
waves, we have
15.2 Stochastic Instability and Magnetic Field Line Diffusion 341

Fig. 15.3 Phase plane V, x

d2 x dV e
2
= = E 0 cosðωt - kxÞ: ð15:22Þ
dt dt m

In the reference frame moving with wave phase velocity,

ω
x= t þ x, ð15:23Þ
k

The equation of motion can be written as

dV e e
= E 0 cos kx = E 0 cos k Vdt : ð15:24Þ
dt m m

Equation (15.24) is equivalent to Eq. (15.21), and the variables in Eq. (15.21)
correspond to variables for the problem of particle motion in the following way:

B →k e ky
l $ t; x $ V; $ E ; $ k: ð15:25Þ
B0z m 0 a

The particle motion character depends on its energy. If the particle kinetic energy in
the wave framework is small, the particle is trapped between the wave potential hills.
Particles with large kinetic energy remain untrapped with the velocity modulated by
wave potential. Various behaviors are illustrated by the phase plane in Fig. 15.3. The
boundary velocity corresponds to the separatrix and is found from the condition
2
mV c =2 = 2eφmax , so

eφmax 1=2 1=2

eE0
Vc = 2 =2 , ð15:26Þ
m mk

and separatrix width

342 15 Magnetic Islands and Stochastic Magnetic Field

1=2
eE 0
ΔV = 2V c = 4 : ð15:27Þ
mk

Equation (15.27) corresponds to the width of the magnetic island given by Eq. (15.7)
with account of similarity in Eq. (15.25).
If several waves act on charged particles, then in the phase plane, several zones of
finite motion arise for particles trapped in each wave with width Eq. (15.27). In the
case of islands, the overlapping motion of a single particle becomes stochastic. In
this situation, in accordance with ergodic theory, the description of a single particle
is equivalent to the description of an ensemble of particles with their distribution
function. The corresponding kinetic equation is given by

∂f ∂f eE1 ∂f
þV þ = 0, ð15:28Þ
∂t ∂x m ∂V

where the small electric field is a sum of many harmonics

E1 = E k expð- iωk t þ ikxÞ: ð15:29Þ

k

In the quasilinear approximation, a solution is sought in the form

f = f 0 þ f 1; f1 = f k expð- iωk t þ ikxÞ: ð15:30Þ

k

In the quasilinear approximation, the interaction between different harmonics is

neglected, and only their collective impact on distribution function f 0 is taken into
account.
In the linear approximation, the kinetic equation is reduced to equations for a
single harmonic

eE k ∂f 0
ð- iωk þ ikV Þf k = - , ð15:31Þ
m ∂V

so that

eEk ∂f 0 1
fk = -i : ð15:32Þ
m ∂V ωk - kV

Here, in accordance with Eq. (8.39)

1 1
=P - iπδðωk - kV Þ: ð15:33Þ
ωk - kV ωk - kV
15.2 Stochastic Instability and Magnetic Field Line Diffusion 343

The equation for f 0 is obtained with account of quadratic terms in the kinetic
Eq. (15.28) after space averaging:

∂f 0 eE 1 ∂f 1
=- : ð15:34Þ
∂t m ∂V

During space averaging, only terms with k = - k′ and the imaginary part of
Eq. (15.33) survive (see also Sect. 8.5):

∂f 0 ∂ πe2 jE k j2 ∂f 0
= δðωk - kV Þ : ð15:35Þ
∂t ∂V k
m 2 ∂V

This is a diffusion equation in the velocity space with a quasilinear diffusion

coefficient

πe2 jE k j2
D= δðωk - kV Þ: ð15:36Þ
k
m2

Diffusion in the velocity space establishes a plateau in the region of the distribution
function where resonances exist.
Using the analogy between the ergodic motion of the charged particle under the
impact of many wave harmonics and the motion of the magnetic field line end for
overlapping magnetic islands, we introduce the distribution function fB of the field
line ends in the perpendicular plane. Similar to Eq. (15.36), in accordance with
Eq. (15.25), the equation for the magnetic field line end density also has the form of a
diffusion equation:

2
∂f B ∂ B →k x ∂f B
= π δ kz þ ky , ð15:37Þ
∂l ∂x ky , k z B0z
2 a ∂x

where the stochastic diffusivity coefficient is

2
B →k x
Dst = π 2
δ kz þ ky : ð15:38Þ
ky , k z B0z a

Similar to the toroidal geometry with account of identity δ(n/R - mBθ/rBζ) = Rδ

(n - m/q),

jBmn j2
Dst = πR δðn - m=qÞ: ð15:39Þ
m, n B2
344 15 Magnetic Islands and Stochastic Magnetic Field

The coefficient of stochastic diffusion has dimension of length and determines the
mean square of the magnetic field line shift in the perpendicular direction as a
function of length along the magnetic field line, in accordance with Eq. (15.18).

15.3 Transport in Stochastic Magnetic Field

Magnetic field line stochasticity causes stochastic transport across a magnetic field.
It is possible to estimate the diffusion coefficient of a test particle, for example, an
electron, in the following way. Let the electron mean free path λmfp be larger than the
Kolmogorov length Lc. In such a situation, electrons move freely along the magnetic
field line, simultaneously shifting in the perpendicular direction together with
it. Perpendicular displacement becomes irreversible only after collision when the
electron shifts at a distance of the order of its gyroradius ρce. Initially, the electron
was “smeared” over a magnetic tube with radius ρce. At large distances l > Lc, the
perpendicular cross-section of a tube becomes similar to that shown in Fig. 15.2,
while the perpendicular size of the “sleeves” decreases to less than ρce. Therefore,
electrons during rare collisions shift to completely different flux tubes and “forget”
their history; as a result, motion becomes irreversible. Since for time t the electron
passed distance l = Vkt along the magnetic field line, then in accordance with
Eq. (15.18)

ðΔxÞ2 = 2Dst V k t, ð15:40Þ

and test particle diffusion coefficient

Dtest = Dst V k : ð15:41Þ

Replacing the parallel electron velocity by their thermal velocity, one can obtain an
estimate of the electron heat conductivity in a stochastic magnetic field as

χe = Dst V Te : ð15:42Þ

This expression is known as the Rechester-Rosenbluth formula.

Diffusion in a stochastic magnetic field is significantly smaller – it is controlled
by ions with velocities (mi/me)1/2 times smaller than those of electrons. An ambipolar
electric field arises in the plasma, which slows down electrons and accelerates ions,
and the resulting ambipolar diffusion coefficient can be estimated as

D = Dst cs , ð15:43Þ

where cs = ðT e þ T i Þ=mi is the ion sound velocity.

15.3 Transport in Stochastic Magnetic Field 345

If a stochastic magnetic field is produced by plasma turbulence, stochastic

transport coefficients can be estimated in the following way. In the equation for
the magnetic field, Eq. (10.14)

→
∂B → → 1 →
= ∇ × u × B - ∇ × σ - 1∇ × B , ð15:44Þ
∂t μ0

One can estimate collisionless conductivity as

ne2
σ , ð15:45Þ
me ω

where ω is the typical frequency of the turbulence. The typical spatial scale can be
estimated as a scale where the magnetic field line becomes unfrozen, i.e., all three
terms in Eq. (15.44) are of the same order. Comparing the first and third terms

me ωB
Bω = , ð15:46Þ
μ0 ne2 δ2

one obtains the scale

c
δ= , ð15:47Þ
ωpe

where ωpe = ne2 =ε0 me is the electron plasma frequency. Scale δ is known as the
collisionless skin layer. For the tokamak, one can assume that the magnetic field line
displacement coincides with δ at the Kolmogorov length, which could be estimated
as qR. In other words,

c
B →k =B  : ð15:48Þ
ωpe qR

Inserting this estimate into Eqs. (15.42) and (15.39), we obtain an estimate for
electron heat conductivity in a stochastic magnetic field:

c2 V Te
χe  : ð15:49Þ
ω2pe q2 R

This estimate is known as the Okawa formula for electron heat conductivity in the
stochastic magnetic field for developed turbulence. This electron heat conductivity
coefficient is inversely proportional to the plasma density and is independent of the
magnetic field. However, in the real tokamak, such a level of turbulence is probably
not reached.
346 15 Magnetic Islands and Stochastic Magnetic Field

15.4 Resonant Magnetic Perturbations in Tokamak

Radial magnetic perturbations are created in a tokamak by special helical external

coils with periodicities of a few m and n in the poloidal and toroidal directions. The
current in these coils in the toroidal geometry produces resonant magnetic perturba-
tions (RMPs) of the radial magnetic field with many harmonics. In the region near
the separatrix (on the core side), where the value of the safety factor is large and,
hence, the distances between resonant flux surfaces are small, magnetic islands in the
vicinity of resonant flux surfaces overlap, and a stochastic region is established. In
this region, electron transport increases, as demonstrated in the previous section. As
a result, it is possible to control transport in the edge plasma.
Radial particle and heat fluxes associated with RMPs could be calculated in the
quasilinear approximation for the collisionless case. The drift kinetic equation for
electron guiding centers, Eq. (1.112), reads

∂f ∂f E x ∂f eE ∂f
þ V k bx þ - x bx = 0: ð15:50Þ
∂t ∂x B ∂y m ∂V k

Here, the slab geometry is considered bx = Bx/B, where x represents the minor radius
and y corresponds to the poloidal direction.

Bx = B →k exp ik y y þ ik z z = B →k exp ik k l : ð15:51Þ

→ →
k k

The radial electric field Ex = - ∂φ0/∂x. Distribution function is sought in the form

f = f 0 þ f 1; f1 = f →k exp ik x x þ ik y y : ð15:52Þ
→
k

In the linear approximation for a single harmonic, we have

b →k ∂f 0 eE0 ∂f 0
f →k = i Vk - : ð15:53Þ
ky V 0 þ kk V k ∂x m ∂V k

Here, V0 = B-1∂φ0/∂x. As before,

1 1
=P þ iπδ kk V k : ð15:54Þ
ky V 0 þ kk V k kk V k

It is supposed that kyV0 < < kkVTe. Here, we neglect the potential perturbation φ →k :,
assuming that it is smeared out by perpendicular viscosity and inertia due to the
small distance between ergodic field lines in the perpendicular direction.
Radial fluxes can be calculated using definition
15.4 Resonant Magnetic Perturbations in Tokamak 347

→
Γe = V k bx f 0 þ f 1 d V ,
: ð15:55Þ
mV 2 →
qe = V k bx f 0 þ f 1 d V :
2

The radial particle flux Γe is quadratic with respect to the magnetic field perturbation.
Substituting Eqs. (15.52)–(15.54) into Eq. (15.55), one obtains

8T e ∂ ln n e ∂φ0 ∂ ln T e
Γe = - n D - þ 0:5 : ð15:56Þ
πme st ∂x T e ∂x ∂x

The electron heat flux is given by

8T e ∂ ln n e ∂φ0 ∂ ln T e
qe = - 2nT e D - þ 1:5 : ð15:57Þ
πme st ∂x T e ∂x ∂x

Here, the stochastic diffusion coefficient of the magnetic field lines is defined
according to Eq. (15.38) or (15.39).
Since the ion flux in the stochastic magnetic field is much smaller than the
electron flux, Γi  me =mi Γe , the radial electron current flows in the stochastic
magnetic field:

je = σSt E x - E St
x , ð15:58Þ

where

8
σSt = e2 n D , ð15:59Þ
πme T e st

and

T e ∂ ln n ∂ ln T e
x = -
ESt þ 1:5 : ð15:60Þ
e ∂x ∂x

In the absence of an ion current, the radial current should turn to zero due to the
ambipolarity constraint, and, hence, the radial electric field should coincide with ESt
x .
This electric field is positive, i.e., it is directed from the core towards the
tokamak edge.
Ion current in the radial direction can be generated by the ion neoclassical effects
described in Chap. 13. The radial ion current becomes nonzero when the radial
turbulent transport of toroidal (parallel) momentum is taken into account. In the
presence of this turbulent transport neoclassical condition, Eq. (13.57)
→ $
B ∇  π k = 0 should be replaced by a more general one:
348 15 Magnetic Islands and Stochastic Magnetic Field

→ $ dU k
- B ∇  π k = nmi : ð15:61Þ
dt

dU k
The term nmi dt here represents the radial transport of parallel momentum. The
toroidal projection of the momentum balance equation reads

dU T
ji By = nmi : ð15:62Þ
dt

Here, the surface-averaged ion neoclassical toroidal viscosity is zero, see Chap. 13.
The radial current in Eq. (15.62) is an ion current that compensates for the radial
current of electrons in the stochastic magnetic field in Eq. (15.55) in accordance with
the ambipolarity constraint

ji þ je = 0: ð15:63Þ

Ion radial current accelerates plasma in the toroidal direction. Note that there is no
contribution from the electron current since the latter flows along magnetic field lines
→ →
and does not produce a j × B force.
Neglecting the difference between Uk and UT and between volume and surface
averaging, from Eqs. (15.61) and (15.62), one can obtain the radial ion current as a
function of parallel viscosity:

→ $
B ∇  πk
ji = - : ð15:64Þ
BBy

After substituting Eq. (13.61), the ion radial current can be expressed as a function of
the difference between the radial electric field and the neoclassical value:

ji = σNeo E x - ENeo
x , ð15:65Þ

where the neoclassical electric field is defined according to Eq. (13.62)

T i ∂n 1 ∂T i
x =
E Neo
en ∂x
þ kT
e ∂x
þ By U k : ð15:66Þ

Neoclassical conductivity in the Pfirsch-Schluter regime according to Eqs. (15.64),

(15.65), and (13.61)

3
σNeo = η : ð15:67Þ
2B2 R2 0
15.4 Resonant Magnetic Perturbations in Tokamak 349

Ex

x
x=0

2
1

Fig. 15.4 Radial electric field in the outer equatorial midplane at the core side of separatrix. 1 –
Without RMPs and 2 – with RMPs, x = 0 corresponds to separatrix position

Here, η0 is the Braginsky viscosity coefficient. In collisionless regimes, it should be

replaced by neoclassical viscosity coefficients [13].
Ambipolarity constraint Eq. (15.63) determines the radial electric field

x þσ
σst E St Neo Neo
Ex
Ex = : ð15:68Þ
σ þσ
st Neo

In the absence of RMPs, the radial electric field coincides with the neoclassical field
and is negative (directed from the edge towards the core), while in the presence of
RMPs, the electric field becomes less negative (Fig. 15.4), and, in the extreme case,
σSt ≫ σNeo is positive: Ex = E St x . The radial current can be calculated from
Eq. (15.58) or (15.65). Toroidal rotation generated by this current is determined by
Eq. (15.62). If the turbulent transport of toroidal momentum is strong enough, the
resulting toroidal rotation does not contribute to the neoclassical electric field and
can be neglected in Eq. (15.66). In the opposite situation, one has to solve the
differential equation for toroidal rotation, which follows from Eqs. (15.62), (15.65),
and (15.66), and then obtain the current and electric field. For further details,
see [14].
Radial current is associated with radial particle flux Γi = ji/e and, hence, leads to
density reduction in the core, see schematic Fig. 15.5. This is known as the “pump
out” effect in the presence of RMPs.
One more important issue concerns the so-called “plasma response.” In the
stochastic magnetic field, current flows along magnetic field lines and creates
magnetic field perturbations so that the vacuum magnetic field produced by RMPs
coils could be significantly reduced by currents in the plasma. Hence, magnetic field
perturbations should be calculated self-consistently so that all treatment becomes
rather difficult and requires simulations with complicated codes.
350 15 Magnetic Islands and Stochastic Magnetic Field

x=0 x

Fig. 15.5 Plasma density in the outer equatorial midplane at the core side of separatrix. 1 – Without
RMPs and 2 – with RMPs, x = 0 corresponds to separatrix position

15.5 Simulation of Resonant Magnetic Perturbations

Effects with Codes and Examples of Experimental
Results

The impact of RMPs could be simulated by 2D transport codes of the SOLPS-ITER

type. The effects of stochastic magnetic fields could be taken into account by
introducing an additional ion current perpendicular to the flux surfaces (equal to
the electron current) and additional particle and electron heat fluxes associated with
stochasticity, which are considered in the previous section. The stochastic diffusion
coefficient of magnetic field lines is calculated by a special code for given currents in
the helical magnetic coils. Examples of such modeling for MAST tokamak are
shown in Figs. 15.6, 15.7, and 15.8. The modeling results are compared with
experimental measurements. One can see that the radial electric field at the core
side of the separatrix with RMPs switched on becomes less negative than in the shot
without RMPs, in accordance with Eq. (15.68). The pump-out effect is also clearly
seen both in the experiment and in the modeling. The electron temperature profile
does not change significantly. This is explained by two factors acting in opposite
directions. Switching on RMPs leads to an increase in stochastic electron heat
conductivity, which should result in electron temperature reduction. On the other
hand, the density drop due to the pump-out effect causes a temperature rise to
maintain the same power flux from the core. The two factors compensate for each
other while keeping the electron temperature almost unchanged.
15.5 Simulation of Resonant Magnetic Perturbations Effects with Codes. . . 351

2000

Er (V/m) 0

–2000
σST = 0.0 code
σST = 5.0e - 3 code
Icoils = 0 (#21712)
Icoils = 1kA (#21714)
–4000
Icoils = 1.4kA (#21713)

–0.04 –0.02 0.00 0.02 0.04

r - rLCF.S (m)

Fig. 15.6 Comparison of experimental and simulated radial electric field profiles at the outer
midplane for shots with and without RMPs

Fig. 15.7 Electron density at the outer midplane for shots with and without RMPs
352 15 Magnetic Islands and Stochastic Magnetic Field

Fig. 15.8 Electron temperature at the outer midplane for shots with and without RMPs
Chapter 16
Improved Confinement Regime (H-Mode)

→ →
16.1 E × B Drift Shear and Transport Barriers
→ →
Spatial inhomogeneity (shear) of the E × B drift velocity may lead to turbulence
suppression and to a reduction in turbulent transport. If it occurs in the localized
region, a transport barrier is formed with reduced turbulent transport and strong
density and temperature gradients. To understand this effect, we consider a simple
slab geometry, as shown in Fig. 16.1. Here, Fig. 16.1a corresponds to the laboratory
→ →
frame, while Fig. 16.1b corresponds to the moving reference frame where the E × B
drift velocity V0( y) changes sign at y = 0.
Let us first look at the evolution of the paint spot in a river with water flowing in
the x-direction with V0( y). In the p
absence of V0( y), paint evolution has a diffusion
character, so its size increases as Dt (Fig. 16.2a):

t t t 1
′ ″ ′ ″ ′
x 2
= V x t V x t dt dt = dt V x t ′ V x t ′ - τ dτ = 2Dt = y2 :
0 0 0 0

ð16:1Þ

Here, velocities represent the random walk process. In the presence of shear flow,
there is an additional displacement along x, which depends on the y-coordinate,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 353
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9_16
354 16 Improved Confinement Regime (H-Mode)

y,E

V0 ( y)

y,E

V0 ( y)

Fig. 16.1 Shear flow. (a) Laboratory reference frame and (b) moving reference frame

√
√
x

x
3/2
t

Fig. 16.2 Evolution of paint in the river. (a) Diffusion in the absence of shear flow and (b)
stretching of the paint in the moving reference frame
 → →
16.1 E × B Drift Shear and Transport Barriers 355

a
y

x
t 3/2

Fig. 16.3 Vortices in turbulent plasma. (a) Without shear flow and (b) with shear flow

t t t t
0 00 0 00 dV 0 00 dV 0 0 00
x 2
= V x ðt ÞV x ðt Þdt dt = 2Dt þ yð t 0 Þ yðt Þ dt dt
dy dy
0 0 0 0
2 2
dV 0 dV 0
2Dt þ 2Dt t 2 = 2Dt þ 2D t3 :
dy dy
ð16:2Þ

Here, we assume dV dy to be independent of y. One can see that the spot is stretched in
0

the x-direction, and, at large times, its dimension in the x-direction increases as t3/2,
as shown in Fig. 16.2b. Since the number of particles in the spot is conserved, the
dimension in the y-direction decreases accordingly.
A similar effect has shear flow on the vortex in turbulent plasma, as shown in
Fig. 16.3. In the presence of shear flow, vortexes become extended in the direction of
the flow. One can introduce decorrelation length Lc. For the pure diffusion case, the
decorrelation time for typical vortices can be estimated as the time of diffusive
damping:

τ1 = L2c =D: ð16:3Þ

For strong shear jdV 0 =dyj > D=L2c , according to Eq. (16.2), a shorter decorrelation
time can be introduced:
356 16 Improved Confinement Regime (H-Mode)

τ2 = L2=3
c D
- 1=3
ðdV 0 =dyÞ - 2=3 : ð16:4Þ

In the situation without shear, the turbulent diffusion coefficient can be roughly
estimated from the balance D=L2c  Dk 2⊥  γk , where k⊥ is a typical wave vector, so
that D  γk k⊥- 2 . The strong shear turbulence level is determined by estimate
γk  τ2- 1  k⊥ D1=3 ðdV 0 =dyÞ2=3 . Since increment γk increases with k⊥, the balance
2=3

is reached at larger wave vectors, i.e., large-scale vortexes are smeared out. As a
result, the diffusion coefficient in the region of strong shear decreases, and a
transport barrier is formed.

16.2 Transition from Low to High Confinement Regime

(L-H Transition)

The L-H transition is observed in a tokamak when the heating power exceeds some
threshold. After that, a fast transition from gradual density and temperature profiles
to profiles with transport barriers at the edge takes place (Fig. 16.4). The temperature
profiles look similar. A region with a steep density gradient (transport barrier) is
formed in the separatrix vicinity at the closed flux surfaces, and the barrier width is of
the order of few cm at the equatorial midplane.
One can understand L-H transition on the basis of radial electric field evolution
and poloidal rotation shear. Indeed, the radial electric field is close to the neoclassical
field, Eq. (15.66). If we neglect the contribution from toroidal rotation and the ion
temperature gradient for simplicity, then the radial electric field would be close to

n` pedestal
H

x
x=0

Transport barrier

Fig. 16.4 Density profiles in the L and H modes

16.2 Transition from Low to High Confinement Regime (L-H Transition) 357

T i ∂n
x ≈
ENeo
en ∂x
, ð16:5Þ

and the shear of the poloidal rotation (in the y-direction) would be

x =B
∂ ENeo ∂ T i ∂n Ti
ωs = ≈  : ð16:6Þ
∂x ∂x en ∂x eBLn

Here, we neglect the weak dependence on the major radius of the tokamak, and Ln is
the typical density spatial scale in the separatrix vicinity. For the gradual density
profile in the L-mode, Ln is of the order of the tokamak minor radius. When shear
Eq. (16.6) remains smaller than the typical increment of drift instability γ, plasma
turbulence is not suppressed by poloidal rotation shear, and the corresponding
turbulent transport remains at the level typical for the low-confinement regime (L-
mode).
With increasing heating power, the ion temperature at the separatrix also rises,
and the poloidal rotation shear in Eq. (16.6) rises correspondingly. At some thresh-
old value, ωs becomes equal to the typical increment of drift instability γ, and the
turbulent level decreases compared to that in the L-mode, while the turbulent
transport coefficients are slightly reduced. The reduction in the turbulent diffusion
coefficient in turn leads to an increase in the density gradient. Indeed, the density
profile near the separatrix is determined by ionization source I and diffusion:

0
∂n
Γ ð xÞ = - D = I ðx0 Þdx0 : ð16:7Þ
∂x
xðcoreÞ

Particle flux Γ is supposed to be constant in the core. Therefore, while the diffusion
coefficient is reduced, the density gradient rises to maintain the same particle flux Γ.
The radial electric field, which according to Eq. (16.5), is proportional to the density
gradient, becomes larger, the shear of the poloidal rotation further increases, turbu-
lence is further suppressed, the diffusion coefficient is further reduced and the
avalanche-type process of density profile steepening is launched. As a result, the
transition to a profile with a transport barrier (H-regime) takes place (Fig. 16.4).
A question about the factors that determine the final steady state profile in the H-
mode requires special treatment. According to the present-day understanding of very
steep density profiles, some new modes, known as edge localized modes or ELMs,
start to develop, which prevent profiles from further steepening. Since the profile in
the core becomes more gradual, general confinement in the H-mode is better than
that in the L-mode. Transition into the H-mode can be caused not only by an increase
in heating power but also by other factors, such as pellet injection, which leads to
density gradient steepening, geodesic acoustic modes, etc.
358 16 Improved Confinement Regime (H-Mode)

16.3 L-H Transition Power Threshold

The critical shear value for the L-H transition in Eq. (16.6) is proportional to the ion
temperature in the separatrix vicinity and hence to the power coming from the core to
the edge. It is inversely proportional to the toroidal magnetic field. One can also
assume that with increasing density, the edge ion temperature decreases and
increases with decreasing density. Therefore, if the L-H transition occurs when the
shear of the poloidal rotation exceeds a given value of increment of drift instability γ,
one would expect that the critical power for the L-H threshold should have the
following scaling:

P  nB: ð16:8Þ

This can be checked in the modeling with 2D transport codes. An example of such
modeling is shown in Fig. 16.5. The critical poloidal rotation shear was chosen to be
ωs = 3  105s-1 at a reference point 2 cm from the separatrix (in the modeling
interval, around this value was taken). One can see that indeed a linear dependence
of the power threshold on the density and magnetic field Eq. (16.8) is observed.
Similar scaling is observed on many tokamaks. Moreover, the absolute value of the
power threshold is in good agreement with the experimental value, which justifies
the choice ωs = 3  105s-1. Note that for low densities, the power threshold deviates
from the simple scaling in Eq. (16.8). This is explained by the fact that power can be
spent to heat electrons so that the electron temperature at the edge rises with power,
while the ion temperature rises only due to heat transfer from electrons to ions, so at
low densities, the ion temperature can increase slower than for the large density case.

Fig. 16.5 Heating power that is necessary to achieve a given value of poloidal rotation shear
(ωs = 3  105s-1 at a reference point) for ASDEX Upgrade for different plasma densities, I = 1MA,
B = 2T. Simulations with SOLPS-ITER
Bibiliography

1. Landau, Lev D.; Lifshitz, Evgeny M. (1976). Mechanics. Vol. 1 (3rd ed.). Butterworth-
Heinemann. ISBN 978-0-7506-2896-9
2. Landau, Lev D.; Lifshitz, Evgeny M. (1975). The Classical Theory of Fields. Vol. 2 (4th ed.).
Butterworth-Heinemann. ISBN 978-0-7506-2768-9
3. Landau, Lev D.; Lifshitz, Evgeny M. (1980). Statistical Physics. Vol. 5 (3rd ed.). Butterworth-
Heinemann. ISBN 978-0-7506-3372-7
4. Golant V. E., Zhilinsky A. P., Sakharov I. E. Fundamentals of Plasma Physics. New York,
Chichester: Wiley, 1980.
5. Chen F. F. Introduction to Plasma Physics and Controlled Fusion. Third Edition. 2016 Springer.
6. Goldston R. J., Rutherford P. H. Introduction to Plasma Physics. 1995 Taylor & Francis, NY.
7. Rozhansky V. A., Tsendin L. D. Transport Phenomena in Partially Ionized Plasma. 2001 Taylor
& Francis NY.
8. Sivukhin D. V. in Reviews of Plasma Physics Vol 1, ed by M.A. Leontovich, Consultants
Bureau 1965.
9. Braginskii S.I. in Reviews of Plasma Physics Vol 1, ed by M.A. Leontovich, Consultants
Bureau 1965.
10. Zhdanov V. M. Transport Phenomena in Multispecies Plasma. 2002 Taylor & Francis NY.
11. Landau, Lev D.; Lifshitz, Evgeny M.; Pitaevskii, Lev P. (1984). Electrodynamics of Continuous
Media. Vol. 8 (2nd ed.). Butterworth-Heinemann. ISBN 978-0-7506-2634-7
12. Fitzpatrik R. Plasma Physics. An Introduction. 2015 Taylor & Francis NY.
13. Helander P., Sigmar D.J. Collisional Transport in Magnetized Plasma. 2005 Cambridge Uni-
versity Press.
14. Rozhansky V in Reviews of Plasma Physics V.24, ed. by V.D. Shafranov, Springer
-Verlag 2008.
15. Wesson J., Campbell D J. Tokamaks. 2004 Clarendon Press.

© The Editor(s) (if applicable) and The Author(s), under exclusive license to 359
Springer Nature Switzerland AG 2023
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9
Index

A D
Acceleration in nonuniform magnetic field, Deceleration by ambient plasma, 234–238
206, 313 Depletion regions, 102, 123, 124, 126, 133,
Alfven masers, 226–227 144, 145, 333
Alfven wave propagation, 234, 236 Diffusion in the ionosphere, 109–110
Alfven waves, 166, 214–216, 218, 219, 221, Diffusive decay, 101, 106
226, 227, 234, 237 Diffusive shocks, 142
Ambipolar diffusion, 78, 95–109, 115, Dispersion, 7, 8, 10, 165, 168, 172, 175, 177,
117, 118, 126, 130, 134, 137, 138, 186, 188, 189, 191, 192, 197–201, 216,
141, 344 218, 222–225, 333
Ambipolarity, 284 Dissipative flute instabilities, 318
Ambipolarity constraint, 284, 285, Distribution function in low collisionality
347–349 regimes, 294
Ambipolar mobility, 140, 143 Distribution functions, 1–5, 7, 17–23, 25–33,
37, 40, 41, 45, 47, 50–56, 59, 64, 68, 69,
79–82, 87, 89, 170, 171, 173–175,
B 178–180, 227, 232, 271, 294–301, 342,
Bohm criterion, 83–86, 88, 90 343, 346
Bootstrap current, 300, 331 Double sheath, 86–89
Braginsky equations, 69 Drift waves, 163–166, 169, 172, 181,
183–186, 307
Drift-dissipative instability, 166–169
C
Chapman-Enskog method, 51
Chirikov criterion, 339 E
Classical diffusion, 149–155, 159 Electron temperature gradient mode, 56, 59, 70,
Collision integral, 17, 18, 27, 28, 30, 32, 37, 51, 176, 177, 279
53, 170, 172, 295 Electron-ion collisions, 12, 13, 155, 300
Collisionless dissipation, 197 Energy principle, 318–322
Collisionless sheath, 78–86 Entropy, 66–68, 189, 212, 214, 215, 218
Conductivity recovery, 145–147 Equatorial bubbles, 316, 333, 334
Convective flux of impurities, 156 Equilibrium, 5, 50, 150, 158, 241–245, 247,
Current, 37, 61, 76, 96, 113, 137, 150, 184, 188, 249–252, 254–260, 263, 264, 267,
205, 229, 241, 271, 311, 335 269, 307, 309, 310, 313, 314, 319, 320,
323, 327

© The Editor(s) (if applicable) and The Author(s), under exclusive license to 361
Springer Nature Switzerland AG 2023
V. Rozhansky, Plasma Theory, https://doi.org/10.1007/978-3-031-44486-9
362 Index

F M
Fast magnetosonic waves, 214–216, 218, 221 Magnetic islands, 245, 326–328, 331,
Filaments in edge tokamak plasma, 238 335–350
Floating potential, 83, 86–88, 91, 92, 103, 104, Magnetic presheath, 89, 90
239, 317 Magnetic shear, 181–186, 313, 326
Flute instability, 310–318, 321, 323 Magnetized plasma, 56, 63, 66, 112, 114–119,
Flux surface functions, 245–249, 260, 261, 279, 123, 126, 132–134, 143–147, 157–159,
281, 314 163, 222, 230, 234, 307–327
Flux surfaces, 71, 182, 183, 227, 245–249, Magnetohydrodynamic (MHD) equations, 205
251, 252, 256–263, 267, 269–280, Magnetosonic solitons, 222, 225
282–289, 292, 293, 298, 299, 303–306, Mapping, 339
314, 315, 325–327, 331, 332, 337–339, MHD equations, 205, 208, 222, 307,
346, 350 318, 329
Force-free equilibrium, 263–267 MHD waves, 212–222
Friction, 20, 24, 48–50, 52–56, 59–62, 64, 66, MHD waves with dispersion, 218
72, 114, 115, 149–151, 153, 155, 156, Momentum equations, 47–50, 52
158, 159, 163, 166, 169, 170, 172, 187, Multispecies plasma, 105–108, 137, 138, 141
188, 205, 207, 276, 278–281, 303 Mushroom-type shape, 233
Frozen in magnetic field, 209–212, 218,
264, 323
N
Neoclassical fluxes, 269, 284, 303
G Neoclassical transport coefficients, 269
Geodesic acoustic mode (GAM), 331–333
Grad-Shafranov equation, 249–252
Gyrokinetic equation, 38–42 O
1D transport codes, 305
Overturn, 142, 195–197, 201, 221, 222, 225
H
H-mode, 353–358
P
Parallel viscosity, 281, 283, 284, 348
I Partially ionized plasma, 109, 137, 149, 151,
Integral equilibrium, 244, 252–256 153, 157, 315–316, 333
Ion acoustic waves, 187–192, 195, 197–201 Pellet injection, 42, 201–203
Ion temperature gradient mode, 152, 176, 177, Perpendicular ambipolar diffusion, 134
208, 283, 284, 300 Perpendicular viscosity, 346
Pfirsch-Schlueter currents, 272, 273, 278
Pfirsch-Schlueter regime, 271–281, 293, 294
K Pinch, 156, 157, 243–245, 253, 260, 263–267,
Kinetic equation, 2, 4, 6, 7, 18, 20, 21, 321–325
25–27, 35–37, 39, 40, 47–49, 51, 68, Plasma blob, 229, 239
170, 173, 178, 180, 197, 291, 294, 295, Plasma clouds in the ionosphere, 147–148
342, 346 Plasma decay in laboratory plasma, 134
Kink mode, 322, 325 Plasma expansion in a magnetic field, 238
Plasma jet, 234–238
Plateau and banana regimes, 291–301
L Polarization electric field, 146, 231
L-H transition threshold, 356–358 Poloidal rotation shear, 356–358
L-mode, 357 Pump-out effect, 350
Localization region, 184
Index 363

Q T
Quasineutrality, 75–78, 84, 88, 95, 97, 106, Tearing mode, 326, 331
124, 144, 150, 175, 191, 197, 222, Thermal force, 56, 59–60, 64, 70–74, 150, 151,
229, 296 155, 158, 207, 278, 279
Thermoelectric current, 91–93
Tokamaks, 269–294
R Toroidal and poloidal rotation, 281–285
Radial electric field, 106, 132, 151, 152, 156, Transport barrier, 353–357
157, 159, 277, 281–285, 288, 290, 297, Transport coefficients, 33–35, 50–56, 59–66,
299, 300, 302–304, 331–333, 346–351 68, 180, 184, 269, 291–294, 305, 345
Radial profile steepening, 357 Transport in stochastic magnetic field, 344–345
Rayleigh-Taylor instability, 307–310, 333 Trapped and untrapped particles, 293, 298, 299
Rechester-Rosenbluth heat conductivity, 344 Turbulence suppression, 353
Resonant magnetic perturbations (RMPs), Turbulent transport, 176, 178–181, 184, 269,
346–352 305, 347, 349
2D transport codes, 350

S
Self-similar solutions, 192–195 U
Shafranov shift, 256, 257, 260, 315 Unipolar diffusion, 108, 118, 124
Shocks, 142, 147, 148, 197, 225 Universal instability, 169–176, 178, 185
Short-circuiting through ambient plasma, 124,
126, 144
Skin effect, 209–211 V
Slow magnetosonic wave, 214, 215, 218 Viscosity, 48–50, 58, 65–66, 68–69, 150, 152,
Solitons, 197, 200, 201, 222, 224, 225 153, 159, 176, 187, 190, 197, 247, 276,
Sound speed, 45, 71, 85, 149, 157, 158, 188, 279, 281–284, 348, 349
189, 191–194, 196, 209, 244, 281,
317, 323
Spitzer conductivity, 61–62, 301 Z
Splitting of small perturbation, 190 Zonal flows, 331–333
Stochastic magnetic field, 335–350