Plasma Electronics Applications in Microelectronic Device Fabrication

Plasma Electronics:
Applications in Microelectronic
Device Fabrication
© 2006 by Taylor & Francis Group, LLC

Series in Plasma Physics
Series Editor:
Steve Cowley, Imperial College, UK and UCLA, USA
Other books in the series:

An Introduction to Inertial Confinement Fusion
S Pfalzner
Aspects of Anomalous Transport in Plasmas

R Balescu
Non-Equilibrium Air Plasmas at Atmospheric Pressure

K H Becker, R J Barker and K H Schoenbach (Eds)
Magnetohydrodynamic Waves in Geospace: The Theory of ULF Waves and their

Interaction with Energetic Particles in the Solar-Terrestrial Environment
A D M Walker
Plasma Physics via Computer Simulation (paperback edition)

C K Birdsall, A B Langdon
Plasma Waves, Second Edition

D G Swanson
Microscopic Dynamics of Plasmas and Chaos

Y Elskens and D Escande
Plasma and Fluid Turbulence: Theory and Modelling

A Yoshizawa, S-I Itoh and K Itoh
The Interaction of High-Power Lasers with Plasmas

S Eliezer
Introduction to Dusty Plasma Physics

P K Shukla and A A Mamun
The Theory of Photon Acceleration

J T Mendonça
Laser Aided Diagnostics of Plasmas and Gases

K Muraoka and M Maeda
Reaction-Diffusion Problems in the Physics of Hot Plasmas

H Wilhelmsson and E Lazzaro
The Plasma Boundary of Magnetic Fusion Devices

P C Strangeby
Non-Linear Instabilities in Plasmas and Hydrodynamics

S S Moiseev, V N Oraevsky and V G Pungin

Series in Plasma Physics
Plasma Electronics:
Applications in Microelectronic
Device Fabrication
T Makabe
Keio University, Japan
Z Petrovic´
Institute of Physics Belgrade, Serbia
New York London

Published in 2006 by
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

CRC Press is an imprint of Taylor & Francis Group
No claim to original U.S. Government works
Printed in the United States of America on acid-free paper
10 9 8 7 6 5 4 3 2 1
International Standard Book Number-10: 0-7503-0976-8 (Hardcover)
International Standard Book Number-13: 978-0-7503-0976-9 (Hardcover)
Library of Congress Card Number 2005056888
This book contains information obtained from authentic and highly regarded sources. Reprinted material is
quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts
have been made to publish reliable data and information, but the author and the publisher cannot assume
responsibility for the validity of all materials or for the consequences of their use.
No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic,
mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and
recording, or in any information storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com
(http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive,
Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration
for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate
system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only
for identification and explanation without intent to infringe.
Library of Congress Cataloging-in-Publication Data
Makabe, T. (Toshiaki).
Plasma electronics : applications in microelectronic device fabrication / by T. Makabe and I. Petrovic
p. cm. -- (Series in plasma physics)
Includes bibliographical references and index.
ISBN 0-7503-0976-8
1. Plasma engineering. I. Title. II. Series.
TA2020.M35 2005
621.044--dc22 2005056888
Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com
Taylor & Francis Group and the CRC Press Web site at
is the Academic Division of Informa plc. http://www.crcpress.com

Preface
Over the past three decades low-temperature plasma applications have been
extended from primarily lighting to the fabrication of microelectronic devices
and new materials, far exceeding our expectations. Radio frequency plasmas
ranging from 105 Hz to 109 Hz, in particular, are now used to process metal-
lic, semiconductor, and dielectric materials for the fabrication of ultra large-
scale integrated (ULSI) circuits and to deposit various kinds of functional
thin films and to modify the surface properties. Without plasma-produced
ions and dissociated neutral radicals for etching and deposition on wafers,
microelectronics manufacturing for ULSI circuit would simply be unfeasible.
The advent of ULSI fabrication has greatly changed how the field of plasma
science is approached and understood. Low-temperature non-equilibrium
plasmas are sustained mainly by electron impact ionization of a feed gas
driven by an external radio frequency power source. These low-temperature
plasmas acquire characteristics intrinsic to the feed gas molecules as deter-
mined by their unique electron-collision cross-section sets. This uniqueness
means that plasmas must be understood using quantum, atomic, and molecu-
lar physics. The disparate time and spatial scales involved in low-temperature
plasma processing (submeter to nanometer and seconds to nanoseconds)
makes plasma processing an inherently stiff problem. The characteristics of
low-temperature plasmas contrast markedly with highly ionized equilibrium
plasmas that are ensembles of charged particles whose behavior can be un-
derstood through their long-range Coulomb interactions and collective effects
and characterized by plasma frequency, Debye length, and electron tempera-
ture. The fundamental collision and reaction processes occurring both in gas
phase and on surface in low-temperature plasmas are the bases for under-
standing their behavior and exploiting them for practical applications.
In the emerging nanotechnology era, device design, reliability, and the
design of integrated plasma processes for their fabrication are tightly cou-
pled. Being able to predict feature profile evolution under the influence of
spatio-temporally varying plasmas is indispensable for the progress of nano-
technology. Prediction of plasma damage and its mitigation are also crucial
prediction & mitigation are 2 process and will be performed through a series
of vertically integrated numerical modeling and simulations ranging from the
reactor scale to those cognizant of device elements subtending the plasma.
Motivated by the important role of plasmas in technology and the need for
simulations to understand the associated complex processes, we emphasize

in this book academic fusion among atomic and molecular physics, surface
physics, the Boltzmann transport theory, electromagnetic theory, and com-
putational science as plasma electronics. We do this to describe and predict
the space and time characteristics of low-temperature plasmas and associated
processing intrinsic to specific feed gases. An underlying theme through this
work is computer-aided plasma analysis and synthesis, with emphasis on
computational algorithms and techniques.
This book is based on a series of lectures presented at Keio University as
part of its graduate program. The university’s interest in the subject matter
and feedback were essential parts of developing this text. We believe that the
book is well suited as an instrument for self-instruction through its topical
exercises and problems arranged in each chapter.
It is a pleasure to acknowledge our debt to David Graves and Robert
Robson, who have helped in a variety of ways during the long period of
our research life composing plasma electronics. Finally, we are indebted to
T. Yagisawa for figure preparation and his attention to detail.
Toshiaki Makabe and Zoran Lj. Petrovic

Keio University

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Plasma and Its Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Application of Low-Temperature Plasma . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Academic Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Phenomenological Description of the Charged

Particle Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Transport in Real (Configuration) Space . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Momentum Balance of Electrons . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Energy Balance of Electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Transport in Velocity Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Electron Velocity Distribution and
Swarm Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Ion Velocity Distribution and Mean Energy . . . . . . . . . . . . . . . 18
2.3 Thermal Equilibrium and Its Governing Relations . . . . . . . . . . . . . . . 20
2.3.1 Boltzmann Distribution in Real Space . . . . . . . . . . . . . . . . . . . . 21
2.3.2 Maxwell Distribution in Velocity Space . . . . . . . . . . . . . . . . . . . 22
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3 Macroscopic Plasma Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Quasi-Neutrality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
3.3 Charge-Separation in Plasmas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
3.3.1 Spatial Scale of Charge-Separation . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.2 Time Scale for Charge-Separation . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Plasma Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4.1 Debye Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4.2 Metal Probe in a Plasma. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .31
3.5 Particle Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.5.1 Ambipolar Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.5.2 Spatial and Time Scale of Diffusion . . . . . . . . . . . . . . . . . . . . . . . 35
3.6 Bohm Sheath Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.6.1 Bohm Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.6.2 Floating Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4 Elementary Processes in Gas Phase and on Surfaces. . . . . . . . . . . . . . . . .41
4.1 Particles and Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.1.1 Particle Representation in Classical and
Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.1.2 Locally Isolated Particle Group and Wave Packets . . . . . . . . 44
4.2 Collisions and Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.2.1 Conservation Laws in Collisions . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2.2 Definition of Collision Cross Sections. . . . . . . . . . . . . . . . . . . . .49
4.2.3 The Distribution of Free Paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2.4 Representation of Collisions in Laboratory and
CM Reference Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.3 Classical Collision Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .57
4.3.1 Scattering in Classical Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3.2 Conditions for the Applicability of the Classical
Scattering Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4 Quantum Theory of Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4.1 Differential Scattering Cross Section σ (θ) . . . . . . . . . . . . . . . . . 66
4.4.2 Modified Effective Range Theory in
Electron Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 Collisions between Electrons and Neutral
Atoms/Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.5.1 Resonant Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.6 Electron–Atom Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.6.1 Energy Levels of Atoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.6.2 Electron–Atom Scattering Cross Sections . . . . . . . . . . . . . . . . . 77
4.7 Electron–Molecule Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.7.1 Rotational, Vibrational, and Electronic Energy
Levels of Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.7.2 Rotational Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.7.2.1 Rotational Energy Levels . . . . . . . . . . . . . . . . . . . . . . . . 83
4.7.2.2 Rotational Excitation Cross Sections . . . . . . . . . . . . . 84
4.7.3 Vibrational Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.7.3.1 Vibrational Energy Levels . . . . . . . . . . . . . . . . . . . . . . . 87
4.7.3.2 Vibrational Cross Sections . . . . . . . . . . . . . . . . . . . . . . . 88
4.7.4 Electronic Excitation and Dissociation . . . . . . . . . . . . . . . . . . . . 89
4.7.4.1 Electronic States of Molecules. . . . . . . . . . . . . . . . . . . .89
4.7.4.2 Cross Sections for Electronic Excitation
of Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.7.5 Electron Collisions with Excited Atoms
and Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.8 Nonconservative Collisions of Electrons with Atoms
and Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.8.1 Electron-Induced Ionization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.8.2 Electron Attachment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.8.2.1 Dissociative Electron Attachment . . . . . . . . . . . . . . . . 99
4.8.2.2 Nondissociative Electron Attachment . . . . . . . . . . . 100

4.8.2.3 Ion Pair Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.8.2.4 Electron Attachment to Excited Molecules . . . . . 101
4.8.2.5 Rate Coefficients for Attachment . . . . . . . . . . . . . . 102
4.8.3 Electron–Ion and Ion–Ion Recombination . . . . . . . . . . . . . . 103
4.8.4 Electron–Ion and Electron–Electron Collisions. . . . . . . . . .105
4.9 Heavy Particle Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.9.1 Ion–Molecule Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.9.1.1 Charge Transfer, Elastic, and Inelastic
Scattering of Ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.9.1.2 Ion–Molecule Reactions . . . . . . . . . . . . . . . . . . . . . . . 109
4.9.2 Collisions of Fast Neutrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.9.3 Collisions of Excited Particles . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.9.3.1 Chemi-Ionization and Penning Ionization . . . . . 112
4.9.4 Collisions of Slow Neutrals and Rate Coefficients . . . . . . 116
4.9.4.1 Quenching and Transport of Excited States . . . . 116
4.9.4.2 Kinetics of Rotational and
Vibrational Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.10 Photons in Ionized Gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.10.1 Emission and Absorption of Line Radiation . . . . . . . . . . . 119
4.10.2 Resonant Radiation Trapping . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.11 Elementary Processes at Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.11.1 Energy Levels of Electrons in Solids . . . . . . . . . . . . . . . . . . . 124
4.11.2 Emission of Electrons from Surfaces. . . . . . . . . . . . . . . . . . .127
4.11.2.1 Photo-Emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.11.2.2 Thermionic Emission . . . . . . . . . . . . . . . . . . . . . . . 128
4.11.2.3 Field-Induced Emission . . . . . . . . . . . . . . . . . . . . . 129
4.11.2.4 Potential Ejection of Electrons from Surfaces
by Ions and Excited Atoms . . . . . . . . . . . . . . . . . . 132
4.11.3 Emission of Ions and Neutrals from Surfaces . . . . . . . . . . 135
4.11.3.1 Surface Neutralization . . . . . . . . . . . . . . . . . . . . . . 136
4.11.3.2 Surface Ionization . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.11.4 Adsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5 The Boltzmann Equation and Transport Equations of

Charged Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.2 The Boltzmann Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.2.1 Transport in Phase Space and Derivation of the
Boltzmann Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
5.3 Transport Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.4 The Transport Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
5.4.1 Conservation of Number Density . . . . . . . . . . . . . . . . . . . . . . 151
5.4.2 Conservation of Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
5.4.3 Conservation of Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.5 Collision Term in the Boltzmann Equation . . . . . . . . . . . . . . . . . . . . 153

5.5.1 Collision Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.5.2 Collision Integral between an Electron and a
Gas Molecule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.5.2.1 Elastic Collision Term J elas . . . . . . . . . . . . . . . . . . . . . . 155
5.5.2.2 Excitation Collision Term J ex . . . . . . . . . . . . . . . . . . . 158
5.5.2.3 Ionization Collision Term J ion . . . . . . . . . . . . . . . . . . . 158
5.5.2.4 Electron Attachment Collision Term J att . . . . . . . . 159
5.6 Boltzmann Equation for Electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
5.6.1 Spherical Harmonics and Their Properties . . . . . . . . . . . . . . 160
5.6.2 Velocity Distribution of Electrons . . . . . . . . . . . . . . . . . . . . . . . 163
5.6.2.1 Velocity Distribution under Uniform
Number Density: g 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.6.2.2 Velocity Distribution Proportional to
∇r n(r, t): g 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .165
5.6.3 Electron Transport Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 170
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6 General Properties of Charged Particle Transport in Gases . . . . . . . . . 175

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
6.2 Electron Transport in DC-Electric Fields . . . . . . . . . . . . . . . . . . . . . . . . 175
6.2.1 Electron Drift Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
6.2.2 Diffusion Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
6.2.3 Mean Energy of Electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
6.2.4 Excitation, Ionization, and Electron
Attachment Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
6.3 Electron Transport in Radio-Frequency Electric Fields . . . . . . . . . . 184
6.3.1 Relaxation Time Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
6.3.2 Effective Field Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . 189
6.3.3 Expansion Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
6.3.4 Direct Numerical Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
6.3.5 Time-Varying Swarm Parameters . . . . . . . . . . . . . . . . . . . . . . . 199
6.4 Ion Transport in DC-Electric Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
7 Modeling of Nonequilibrium (Low-Temperature) Plasmas. . . . . . . . .205

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
7.2 Continuum Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
7.2.1 Governing Equations of a Continuum Model . . . . . . . . . . . . 207
7.2.2 Local Field Approximation (LFA) . . . . . . . . . . . . . . . . . . . . . . . 210
7.2.3 Quasi-Thermal Equilibrium (QTE) Model . . . . . . . . . . . . . . . 211
7.2.4 Relaxation Continuum (RCT) Model . . . . . . . . . . . . . . . . . . . . 212
7.2.5 Phase Space Kinetic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
7.3 Particle Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
7.3.1 Monte Carlo Simulations (MCSs) . . . . . . . . . . . . . . . . . . . . . . . 216
7.3.2 Particle-in-Cell (PIC) and Particle-in-Cell/Monte Carlo
Simulation (PIC/MCS) Models . . . . . . . . . . . . . . . . . . . . . . . . . 219

7.4 Hybrid Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
7.5 Circuit Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .221
7.5.1 Equivalent Circuit Model in CCP . . . . . . . . . . . . . . . . . . . . . . . 221
7.5.2 Equivalent Circuit Model in ICP . . . . . . . . . . . . . . . . . . . . . . . . 223
7.5.3 Transmission-Line Model (TLM) . . . . . . . . . . . . . . . . . . . . . . . . 224
7.6 Electromagnetic Fields and Maxwell’s Equations . . . . . . . . . . . . . . . 225
7.6.1 Coulomb’s Law, Gauss’s Law, and Poisson’s Equation . . . 225
7.6.2 Faraday’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
7.6.3 Ampere’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
7.6.4 Maxwell’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
8 Numerical Procedure of Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

8.1 Time Constant of the System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
8.1.1 Collision-Oriented Relaxation Time . . . . . . . . . . . . . . . . . . . . . 231
8.1.2 Plasma Species-Oriented Time Constant. . . . . . . . . . . . . . . . .232
8.1.3 Plasma-Oriented Time Constant/Dielectric
Relaxation Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
8.2 Numerical Techniques to Solve the Time-Dependent
Drift-Diffusion Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
8.2.1 Time-Evolution Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
8.2.1.1 Finite Difference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
8.2.1.2 Digitalization and Stabilization . . . . . . . . . . . . . . . . . 237
8.2.1.3 Time Discretization and Accuracy . . . . . . . . . . . . . . 239
8.2.2 Scharfetter–Gummel Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
8.2.3 Cubic Interpolated Pseudoparticle Method . . . . . . . . . . . . . . 243
8.2.4 Semi-Implicit Method for Solving
Poisson’s Equation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .245
8.3 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
8.3.1 Ideal Boundary — Without Surface Interactions . . . . . . . . . 246
8.3.1.1 Dirichlet Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . .246
8.3.1.2 Neumann Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
8.3.1.3 Periodicity Condition . . . . . . . . . . . . . . . . . . . . . . . . . . 247
8.3.2 Electrode Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
8.3.2.1 Metallic Electrode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
8.3.2.2 Dielectric Electrode. . . . . . . . . . . . . . . . . . . . . . . . . . . . .249
8.3.3 Boundary Conditions with Charge Exchange . . . . . . . . . . . . 250
8.3.4 Boundary Conditions with Mass Transport . . . . . . . . . . . . . . 250
8.3.4.1 Plasma Etching. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .251
8.3.4.2 Plasma Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
8.3.4.3 Plasma Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
8.3.5 Moving Boundary under Processing . . . . . . . . . . . . . . . . . . . . 252
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
9 Capacitively Coupled Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255

9.1 Radio-Frequency Capacitive Coupling . . . . . . . . . . . . . . . . . . . . . . . . 255

9.2 Mechanism of Plasma Maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
9.2.1 Low-Frequency Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
9.2.2 High-Frequency Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
9.2.3 Electronegative Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
9.2.4 Very High-Frequency Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . 266
9.2.5 Two-Frequency Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
9.2.6 Pulsed Two-Frequency Plasma . . . . . . . . . . . . . . . . . . . . . . . . . 271
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
10 Inductively Coupled Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

10.1 Radio-Frequency Inductive Coupling . . . . . . . . . . . . . . . . . . . . . . . . 275
10.2 Mechanism of Plasma Maintenance. . . . . . . . . . . . . . . . . . . . . . . . . .275
10.2.1 E-mode and H-mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
10.2.2 Mechanism of Plasma Maintenance . . . . . . . . . . . . . . . . . . 277
10.2.3 Effect of Metastables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
10.2.4 Function of ICP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
10.3 Wave Propagation in Plasmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
10.3.1 Plasma and Skin Depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
10.3.2 ICP and the Skin Depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
11 Magnetically Enhanced Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

11.1 Direct-Current Magnetron Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
11.2 Unbalanced Magnetron Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
11.3 Radio-Frequency Magnetron Plasma. . . . . . . . . . . . . . . . . . . . . . . . .294
11.4 Magnetic Confinements of Plasmas . . . . . . . . . . . . . . . . . . . . . . . . . . 295
11.5 Magnetically Resonant Plasmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
12 Plasma Processing and Related Topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
12.2 Physical Sputtering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .301
12.2.1 Target Erosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
12.2.2 Sputtered Particle Transport . . . . . . . . . . . . . . . . . . . . . . . . . 306
12.3 Plasma Chemical Vapor Deposition . . . . . . . . . . . . . . . . . . . . . . . . . .308
12.3.1 Plasma CVD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
12.3.2 Large-Area Deposition with High Rate . . . . . . . . . . . . . . 311
12.4 Plasma Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
12.4.1 Wafer Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
12.4.1.1 On Electrically Isolated Wafers (without
Radio-Frequency Bias) . . . . . . . . . . . . . . . . . . . . . 313
12.4.1.2 On Wafers with Radio-Frequency Bias . . . . . 313
12.4.2 Selection of Feed Gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
12.4.3 Si or Poly-Si Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
12.4.4 Al Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317

12.4.5 SiO2 Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
12.4.6 Feature Profile Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
12.4.7 Plasma Bosch Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
12.4.8 Charging Damage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
12.4.8.1 Surface Continuity and Conductivity . . . . . . 326
12.4.8.2 Charging Damage to Lower Thin
Elements in ULSI. . . . . . . . . . . . . . . . . . . . . . . . . .328
12.4.9 Thermal Damage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330

1
Introduction
1.1 Plasma and Its Classification

The plasma state is defined as the fourth state of matter as distinct from
the solid, liquid, and gas phases. It consists of free positive and negative
charges with electrical quasi-neutrality in addition to feed gas components.
The word “plasma” was first introduced by Langmuir in 1928 and derives
from the Greek πλασµα meaning “something formed” [1]. The strength of the
plasma state is described in terms of the ionization degree HDI = n p /N, where
ne (= n p ) and N are the number density of the charged particles and neutral
molecules in a plasma. Strongly ionized plasma is realized in HDI ≥ 10−3 .
In particular, electron transport in a strongly ionized plasma is subject to
the long-range Coulomb interaction with surrounding electrons and positive
ions, and the smooth trajectory makes it impossible to identify changes in
momentum by collision during flight (Figure 1.1a). Therefore, strongly ionized
plasma is termed collisionless plasma. On the other hand, the character-
istics of a weakly ionized plasma are properly represented by the short-
range interaction between the electrons and the neutral molecules in the feed
gas under external electric or magnetic fields. In such a collision-dominated
plasma, the electron trajectory is characterized by both flight in the field and
collision with surrounding molecules (Figure 1.1b). In comparison, a neu-
tral molecule has a straight trajectory under a short-range collision in gases
(Figure 1.1c).
In the gas phase, a plasma state can be produced electrically, thermally,
or optically through the ionization of neutral molecules in the feed gas. In
particular, nonequilibrium plasma or low-temperature plasma, which is pro-
duced by the collisional ionization of free electrons under an external electrical
power, has a property that the electron energy is much higher than that of
the neutral gas. We herein distinguish low-temperature plasma in the gener-
ation mechanism from thermal plasma produced by the thermal ionization
of neutral molecules.
Low-temperature plasmas from micrometer to meter size are artificially
maintained even in an electrodeless reactor as well as between metallic or
dielectric electrodes by using electrical power sources ranging from direct
current to radiofrequency levels (tens of kHz to several hundred MHz
and GHz).

2 Plasma Electronics: Applications in Microelectronic Device Fabrication
(a) (b) (c)

E
FIGURE 1.1
Electron trajectory in three different media. In strongly ionized plasma (a), in weakly ionized
plasma (b), and in neutral gas (c).
1.2 Application of Low-Temperature Plasma

Low-temperature plasma technology differs from that of collisionless plasma
in that a low-temperature plasma is produced and maintained in a collision-
dominated region and exhibits proper characteristics and functions intrinsic
to the quantum state of the feed gas molecules. This is one of the primary
advantages of low-temperature plasmas for material processing and device
fabrication requiring a variety of surface processes and very different reac-
tions among materials adjacent to each other. The technology assisted by
low-temperature plasma is generally referred to as plasma processing and is
classified into plasma-enhanced chemical vapor deposition, plasma etching,
sputtering, ashing, surface modification, and so on, on a scale of size ranging
from nanometers to meters (Figure 1.2). Typical geometrical arrangements of
Pressure (Pa)
1.0 101 102 103 104 105
1013
µ -plasma
PDP
1012 ICP ICP
ECR MEMS
Plasma density (cm-3)
1011 etching ashing

doping trimming
CCP ICP
1010
surface
9
Magnetron film deposition modification
10 sputter- CCP
deposition
108 Barrier-
discharge
107
10-3 10-2 10-1 1.0 10-3 102 103
Pressure (Torr)
FIGURE 1.2
Low-temperature plasma and material processes.

Introduction 3
Passivation
M6
M5 Global
interconnect
(up to five layers)
M4 Via
M3
Cu
Low-k
Intermediate
interconnect
M2 (up to four layers)
M1 Local
interconnect
SiO2
n+ n+ STI p+ p+
p-well n-well
nMOS pMOS
substrate
FIGURE 1.3
Typical large scale integrated circuits.
an ultra-large-scale integrated circuit demonstrated partly in Figure 1.3 are

manufactured through more than 100 plasma processes.
The principal plasma processes for material fabrication are etching and
deposition, which are highly competitive. In an era of nanotechnology, plasma
technology has developed into a combination of top-down and bottom-up
processes. The bottom-up (chemical) approach is effectively achieved under
plasma-enhanced conditions.
1.3 Academic Fusion

In the past two decades, great progress has been made in the application of
low-temperature plasma [2]. Most currently used models of low-temperature
radiofrequency plasma were proposed between the mid-1980s and the early

Low-Temperature
Plasma and
Its Technologies
FIGURE 1.4
Academic fusion for low-temperature plasmas.
1990s [3]. Further achievements in plasma technology at the design level will
lead to greater miniaturization of ultra-large-scale integrated (ULSI) circuits
in microelectronics and nanoelectronics and to the further functionalization of
new materials. These developments will be synchronized with the progress of
computers utilizing high-speed and high-performance ULSI chips. We stand
at the advent of the design era of sophisticated plasmas based on atomic and
molecular physics both in the gas phase and in the surface phase.
The low-temperature plasma and related technologies will join the excit-
ing and practical fields of academic fusion reconstructed by computational
science with quantum atomic and molecular physics, surface physics and
chemistry, Boltzmann equation of particles, and Maxwell’s electromagnetic
theory as is shown in Figure 1.4.
To address the design issue, it is essential to prepare a two- or three-
dimensional image of the plasma structure in a reactor and feature profile on
a substrate surface by experimental observations or computational modeling.
References
1. Goldston R.J. and Rutherford P.H. 1995. Introduction to Plasma Physics. Bristol:
IOP.
2. American Institute of Physics. 2003. 50 years of science, technology, and the AVS
(1953–2003), J. Vac. Sci. Technol. A (Special Issue) 21:5.
3. Makabe, T., Ed. 2002. Advances in Low Temperature RF Plasmas, Basis for Process
Design. Amsterdam: Elsevier.

2
Phenomenological Description of the Charged
Particle Transport
In this chapter we attempt to lay out the foundation of the phenomenol-

ogy of charged particle transport in gases. We are primarily concerned with
electrons, although ions are also covered briefly. We leave the more detailed
description of transport theory for one of the later chapters; here we describe
aspects of transport that may be understood from a phenomenological con-
sideration of the charged particles either in the real or in velocity space. We
derive the corresponding velocity distributions and transport (swarm) prop-
erties and some of their basic relationships. We also derive general equilib-
rium distributions of charged particles in real and velocity space in thermal
equilibrium.
2.1 Transport in Real (Configuration) Space

Typically, without the external electric field, a charged particle (we discuss
mainly the case of electrons here) swarm will acquire a Gaussian spatial profile
in density. This profile is the result of the particle’s random thermal motion
and collisions with gas molecules, provided that initially all particles start
from the same position. Long-range Coulomb interactions between charged
particles have no influence on the development of the swarm in a weakly
ionized gas. As a result, electrons will have a net movement in the negative
direction of the field superimposed on their random motion by collisions in all
directions under an external field. When a field is added, the Gaussian spatial
profile will be slightly shifted and even skewed in the direction opposite to
that of the field.
A swarm of electrons and its development are considered in a direct cur-
rent (DC) electric field E(= −Ek), where k is a unit vector along the z-axis.
For the purpose of this chapter, we assume that the transport properties are
uniformly distributed throughout the electron swarm, and we mainly follow
the semiquantitative analysis of Parker and Lowke [1].

n(r) I E
II
III z
0
r
FIGURE 2.1
Spatial profile of electron density for a swarm released from a single point in space and allowed
to develop under the influence of electric field and collisions with background molecules.
2.1.1 Momentum Balance of Electrons

The effect of field will also result in the net flux of electrons (Γ) in the direction
opposite to the field. In other words, a net effective velocity (v0 ) may be used
to describe the average motion of a swarm of electrons with number density
distribution n(r). If a swarm introduced into an electrical field acquires a uni-
form distribution, it will attain an average, drift velocity (vd ). When we release
a swarm of particles from a single point as described above, then there will be
a spatial density distribution of charged particles n(r), as shown in Figure 2.1.
Thus the density gradient will result in flux due to diffusion (−D∂n/∂r) in
the direction opposite to that of the gradient. The total flux of particles in
one-dimensional space along a DC electric field E = −Ek is then given by
∂n
Γ = nv0 = nvd − D k, (2.1)
∂z
where D is the diffusion coefficient.

The momentum balance of electrons will be applied to derive the trans-
port parameters. Electrons gain velocity and consequently average momen-
tum only along the axis of the electrical field. Then, the total change of the
momentum of the swarm per unit time is equal to the impulse eE. The mo-
mentum change is caused by the collision with background gas, mainly the
elastic collision with the rate of Rm in a low-energy electron. The momentum
balance equation is therefore written as

d D ∂n
(mnvz k) = e(−E)n − mn vd − k Rm , (2.2)
dt n ∂z
where m and e(> 0) are the mass and charge of the electron, respectively. In
the stationary state, dvz /dt = 0, the total differential on the left-hand side
of Equation 2.2 is separated into two parts, one of which is zero. Applying

Phenomenological Description of the Charged Particle Transport 7
transformation vz = ∂z/∂t, we separate the Equation 2.2 is into two with

respect to the electron number density n(r) and its spatial derivative
0 = (−eE − mvd Rm ) n,
∂n
0 = mvz2 − mDRm . (2.3)
∂z
It is justified (as shown later) to assume that the z component of the velocity
is not substantially perturbed by the electric field-induced drift, and there-
fore 13 v 2 = vx2 = v 2y ∼ vz2 . In that case we may simplify the second part of
Equation 2.3 to give the approximate value of the diffusion coefficient as
eE
vd = − . (2.4)
mRm
v2
D= . (2.5)
3Rm
PROBLEM 2.1.1
Derive Equations 2.4 and 2.5.
2.1.2 Energy Balance of Electrons

The former derivation of the properties of the electron swarm relied only on
momentum balance as given by Equation 2.2. Now, we consider the energy
balance starting from the concept of the mean energy εm . The change of the
total energy of the system of electrons with n(r) is given by

d D ∂n 2m
(nεm ) = e(−E)n · vd − k −n εm Rm , (2.6)
dt n ∂z M
where M is the mass of the gas molecules. The first term on the right-hand
side represents the energy gain by the movement along the field axis in unit
time, and the second term represents the collisional energy losses, given by
the sum of the elastic loss, 2mεm /M. When we substitute Equations 2.4 and
2.5 in the energy balance of Equation 2.6, we obtain for the stationary state
e2 E2 2e Eεm 1 ∂n 2m
− − εm Rm = 0. (2.7)
mRm 3mRm n ∂z M
We consider solving this equation under certain conditions, particularly in
regard to the spatial profile of the electron number density n(z).
1. ∂n/∂z = 0: This corresponds to the point I in Figure 2.1, that is,
to the maximum of the density distribution. From Equation 2.7 we
then obtain

M eE 2
εm0 = εm | ∂n
∂z =0
= . (2.8)
2m2 Rm

2. ∂n/∂z =
0: This corresponds to all points except for I in Figure 2.1.
For the region with a negative density gradient (region II in
Figure 2.1) the mean energy is higher than that at the peak (point I):
εm = εm0 +
ε.
So far we have assumed that the collisional rate was independent of the
electron energy. However, this is normally not the case. Thus we use a Taylor
expansion around the mean energy at the peak I for the collision rate:
∂ Rm
Rm (εm ) = Rm0 + |0
ε,
∂ε
which in combination with Equations 2.7 gives
 
2ε 1 ∂n
εm = εm0 1 − .
m0
(2.9)
Rm εm0 n ∂z
3eE 1 + 2 ∂∂ε |0 Rm0
Exercise 2.1.1
Discuss the value of the mean electron energy in two cases, (a) ∂n/∂z < 0
and (b) ∂n/∂z > 0, as compared with the mean energy at the peak of the
swarm εm0 . For case (a), the gradient of the electron number density shown
in Figure 2.1 is negative as in region II, and therefore εm > εm0 . However,
for case (b), the gradient of the electron number density is positive (see region
III in Figure 2.1) and therefore εm < εm0 . Thus, unless the collision rate Rm
changes very rapidly, we may expect the mean energy of electrons to increase
toward the front of the swarm where flux due to diffusion adds to the drift-
induced component. The spatial profile of energy may have consequences on
the transport coefficients, provided that some energy-dependent processes
also exist, which is generally the case.
PROBLEM 2.1.2
Derive expression 2.9 for the mean electron energy εm .
In the presence of density gradients in the swarm, random motion under

collisions with gas molecules will result in diffusion. So far we have con-
sidered diffusion only in one dimension along the direction of the electric
field. When, however, we consider three-dimensional space defined by the
Cartesian coordinates r(x, y, z) and the corresponding unit vectors (i, j, k),
while keeping the same direction of the electric field E = −Ek, the electron
number density develops as n(x, y, z). The conditions of the zero gradient in
one direction now become lines along the surface of the density profile. The
total flux of electrons in Equation 2.1 are revised for the three-dimensional
case as

∂n ∂n ξ ∂n
Γ = nvd k − D0 i+ j − D0 1 − k, (2.10)
∂x ∂y 1 + 2ξ ∂z

where

εm0 ∂Rm
ξ= 0. (2.11)
Rm0 ∂ε
As a result of the influence of the electric field, the diffusion along the
z-axis (direction of the field) has special properties and is in general
different from the diffusion in perpendicular directions. The diffusion of
electrons (and charged particles in general) becomes anisotropic with two
components of the diffusion tensor D: DL (longitudinal diffusion coefficient)
and DT (transverse diffusion coefficient). These two coefficients may be
given as

ξ
DL = D0 1 − , (2.12)
1 + 2ξ
and
v2
DT = D0 = , (2.13)
3Rm
which are valid only for approximations that were involved in the derivation
of the formulae. In the case where the collision rate Rm is independent of the
mean energy, the diffusion will be isotropic. The same is true for the zero
electric field.
In this section we have seen that the transport of electrons (the same is
true for ions) may be described with the aid of transport coefficients. We
have given approximate relations for the drift velocity and components of
the diffusion tensor. In general these properties will depend on the mean
electron energy. However, it is not practical to use mean energy for tabulating
the data for transport coefficients. It is better to use an external parameter
associated with the value of the external electric field E. The reduced electric
field is defined as the ratio between the magnitude of the electric field E
and the neutral gas number density N, that is, E/N. It is shown that the
reduced field is proportional to the energy gained by electrons between two
collisions. E/N is expressed in units known as Townsend (in honor of the
founder of gaseous electronics), which is defined as 1 Td = 10−21 Vm2 =
10−17 Vcm2 .
PROBLEM 2.1.3
By substituting Equation 2.9 into Equation 2.6, derive the longitudinal DL
(Equation 2.12) and transverse DT (Equation 2.13) coefficients. Assuming analytic
dependence of the collision rate Rm on energy find the ratio DL /DT in analytic form.
Discuss the conditions when the ratio is equal to, less than, or greater than 1. (A hint:
choosing Rm ∼ (εm /εm0 )( +1)/2 is particularly convenient.)

PROBLEM 2.1.4
Pressure p may be used to represent the gas number density N; find the relationship
between E/N and E/ p given that
E/N[Vcm2 ] = 1.036 × 10−19 Tg [K]E/ p[Vcm−1 Torr−1 ],
where Tg is gas temperature. Here, at Tg = 273 K, E/P of 1 Vcm−1 Torr is equal to

2.823 Td (= 10−17 Vcm2 ). Prove that the mean energy gain between two collisions
is proportional to E/N.
2.2 Transport in Velocity Space

An ensemble of electrons may be studied in velocity space. Whereas in the real
space the spatial density distribution n(r) was used to describe the swarm,
in velocity space we may consider the probability distribution of velocities.
The velocity distribution g(v) is defined so that the quantity g(v)dv is the
probability of finding particles with velocity v within the small element dv.
Due to the symmetry with respect to the vz -axis, it is advantageous to use a
polar coordinate system (v, θ, ϕ). Thus dv = v 2 sin θ dθ dϕ. The magnitude of
the velocity v = |v| is associated with the energy of electrons ε = mv 2 /2, and
in the case of thermal equilibrium the distribution is given by the so-called
Maxwellian distribution:
m 3/2
mv 2
g M (v) = exp − . (2.14)
2π kT 2kT
At the same time, we may define the velocity distribution function g(v, r, t),
which describes swarm development with respect to both velocity and real
space and with respect to time.
2.2.1 Electron Velocity Distribution and Swarm Parameters

Thermal equilibrium, however, does not occur for low-density swarms of
charged particles in an electric field. Normally, an electric field affects elec-
trons, which gain energy but are not able to dissipate it to give translational
motion to gas molecules or ions, because elastic collisions are not particularly
effective in energy transfer to the target due to the very different masses;
that is, m/M 1 . Thus electrons in swarms and in low-density collisional
plasmas are normally not in thermal equilibrium. Two distinct conditions
may occur for the electron swarm in an electric field. In the case that the
properties of the swarm are uniform in the real space, we will have a quasi-
equilibrium. Then, the energy gained by the field is dissipated in collisions,
both elastic and inelastic. As a result, properties of electrons given by electron
energy distribution do not change within the swarm. In such a case (i.e., under

g( )
vr
0
vz
FIGURE 2.2
Typical electron velocity distribution g(v) as a function of axial vz and radial vr velocities under
a DC electric field.
hydrodynamic conditions), it will be possible to separate spatial and velocity

distributions. However, close to the initial stage or boundary wall or soon
after the release the properties of the electrons may not be uniform in space
or time or both. These nonequilibrium conditions are not expressed in terms
of thermodynamic equilibrium; however, in relation to the balance between
energy gain and loss, the swarm will develop until the balance is achieved
or the swarm is lost. Such cases are described as a nonlocal (in space) or not
relaxed electron swarm transport.
In this section we focus on relaxed spatially uniform swarms that are in
the hydrodynamic regime. Thus the velocity distribution provides a complete
description of the swarm. Figure 2.2 shows a velocity distribution in velocity
space. Here, we take advantage of the azimuthal symmetry of the swarm in
velocity space and present the distribution as a function of the axial (axis of
electric field) velocity vz and radial (perpendicular) velocity vr . The electric
field produces asymmetry of the distribution with a peak shifted in one direc-
tion. The small difference between the components in the direction and op-
posite to the direction of the field produces a net drift of particles. At the same
time the velocities directed in opposite directions occur at all points through-
out the swarm profile. In the bulk of the spatial swarm profile, loss of particles
moving outside a small section is compensated by gain from the nearby sec-
tions that have a similar density; this, however, is not the case at the edges.
The velocity distribution thus contains information about both drift and
diffusion.
PROBLEM 2.2.1
The drift velocity and mean energy of electrons in argon under the uniform field,
E = 100 Vcm−1 , and at pressure of 1 Torr (at temperature of 300 K) are vd = 2.3×107
cms−1 and 8 eV, respectively. Show, then, that the mean random velocity vr and drift
velocity vd satisfy the relation (vd /vr ) 1.

vr
E
v v'
θ
vz
0 0'
∆V
FIGURE 2.3
Contour plot of electron velocity distribution g(v) as a function of axial vz and radial vr velocities.
When the electric field is turned on, a small component of velocity
V is added to all electrons
in the direction opposite to the field E.
In Figure 2.3 we show the effect of the electric field on the velocity distribu-
tion g(v) at each velocity v with and without the field. When the electric field
E is turned on, a small component of velocity
V is added to all electrons in
the direction opposite to the field. In the simplified model the circular contour
representing random (thermal) isotropic motion of electrons is slightly shifted
as if its center were moved to the position O from O. The resulting velocity
is v , which is given by: v − v ∼
=
V cos θ (where θ is the angle between v
and −E). We can now estimate the effect of the electric field on the electron
velocity distribution as
dg0 (v)
g(v ) = g(v +
V cos θ) ∼
= g0 (v) +
V cos θ , (2.15)
dv
where we applied the condition | v | |
Vcos θ |. The distribution g(v ) is
composed of an isotropic part in all directions and having, in general, a non-
Maxwellian term and a small anisotropic term. Both terms are influenced
by the electric field. This expansion is analogous to a two-term expansion in
spherical harmonics of the velocity distribution, where the second term is the
anisotropic term (see Chapter 5).
Now we consider the momentum conservation of electrons with velocity
distribution g(v) in gases under an electric field. First, let us define Qm (v)
as the momentum transfer cross section of the electron by elastic scattering
with the background gas molecule with number density N. Then, for one
group of electrons with velocity v, the collision frequency of the momentum
exchange is NQm (v)v, and the total momentum dissipated in collisions in the
direction opposite to the field E is mv cos θ NQm (v)v. The total momentum
exchange, therefore, is obtained by averaging over all velocities by using the

distribution, and in the steady state we have

∂(nm < v >)
= e(−E)n − nmvNQm (v)vg(v)dv = 0. (2.16)
∂t

Here, g(v) is normalized as g(v)dv = 1, having dv = 2π sin θdθv 2 dv. When
we substitute Equation 2.15 into Equation 2.16, we obtain the following for
the second term in Equation 2.16:

dg0 (v)
nmv NQm (v) g0 (v) +
V cos θ
2
cos θ 2π sin θ dθv 2 dv
dv

4π dg0 (v)
= nmv 4 NQm (v)
V dv.
3 dv
After we apply partial integration in order to replace the derivative of the

velocity distribution and take into account the symmetry of the distribution
g0 (v), we obtain

4π d 4
− m v NQm (v)
V g0 (v)dv = eE. (2.17)
3 dv
This is an integral equation that is satisfied for an arbitrary field E and cross
sections Qm (v) only if
V = − [eE/mNQm (v)v]. In this case we write the
expansion for g(v) as
eE cos θ dg0 (v)

g(v) = g0 (v) − . (2.18)
mNQm (v)v dv
The second term in Equation 2.18 shows the effect of an external field on
the velocity distribution. The cosine term is the lowest-order development of
the actual distribution in the velocity space in addition to the symmetric g0 (v)
term. Usually the cross section for the momentum transfer of elastic scattering
is much greater than the sum of cross sections for other inelastic collisions.
So far we have taken into account only elastic collisions. The energy balance
of electrons will consist of energy loss in such collisions, as we mentioned
before, 2mε/M. Therefore, the total energy loss in elastic collisions in unit
time is expressed as
2m 1 2
mv NQm (v)vg(v)dv.
M2
The energy gain, on the other hand, is determined by the motion of elec-
trons along the electric field axis:
eEv cos θg(v)dv.

If we make the balance and apply the expansion 2.18 we obtain

m eE cos θ dg0 (v)
mv NQm (v)v g0 (v) −
2
dv
M mNQm (v)v dv

eE cos θ dg0 (v)
= eEv cos θ g0 (v) − dv. (2.19)
mNQm (v)v dv
After integration over the polar angles θ and ϕ and by equating the terms
inside the integrals we obtain the following first-order differential equation:
m 4π eE dg0 (v)
4π mv 2 NQm (v)vg0 (v) = − eEv .
M 3 mNQm (v)v dv
The solution will be obtained by direct integration without making any
assumptions on the velocity dependence of the cross section Qm (v):

3m v NQm (v)v 2
g0 (v) = A exp − vdv , (2.20)
M 0 eE/m

where A should be determined from the normalization g0 (v)dv = 1. Now
we perform the conversion to electron energy distribution f (ε):
1 2
g0 (v)4πv 2 dv = f (ε)dε; mv = ε. (2.21)
2
The Maxwellian distribution in Equation 2.14 in thermal equilibrium is rewrit-
ten as a function of energy :

√ 3ε
f M (ε) = A ε exp − , (2.22)
2ε
and the mean energy is given by ε = 3kTe /2. In a more general case when
there is an electric field but for a constant cross section Qm , we obtain from
Equation 2.20 the so-called Druyvestyn distribution:
2
√ ε
f D (ε) = A ε exp −0.548 . (2.23)
ε
Exercise 2.2.1
Plot the distributions f (), f M (), and f D () for electrons in Ar at 100 Td. Here
the mean energy < > is 6.47 eV.
Figure 2.4 shows the Maxwellian, Druyvestyn, and nonequilibrium ( f 0 )
energy distributions for electrons at the same mean energy. Note the dif-
ferences among the three distributions.
Now we proceed to determine the transport coefficients based on the

energy or velocity distributions in velocity space. The drift velocity is defined

f0
0.10 100 Td
fD 6.47 eV
f (ε)
0.05
fM
0.00
0 5 10 15 20 25 30
Electron energy (eV)
FIGURE 2.4
Comparison among f (), f M () , and f D () with the same mean energy.
in this chapter as the average velocity of a swarm; that is,

vd = vg(v)dv . (2.24)
v
Combination of expression 2.18 and Equation 2.24 gives

eE cos θ dg0 (v)
vd = v cos θ g0 (v) − 2πv 2 sin θ dθ dv
v,θ mNQm (v)v dv

4π eE v2 dg0 (v)
=− dv. (2.25)
3 m NQm (v) dv
When we make the transition to the energy distribution in Equation 2.21 we
obtain

1 2 ε d f (ε)
vd = − eE √ dε, (2.26)
3 m NQm (ε) dε ε
where the value of (2/m)1/2 e E of an electron with 1 eV is 5.93 × 107 cm−1 .

The transverse diffusion coefficient DT may be derived on the basis of the
standard formula λv/3 without field (i.e., for neutral molecules), where λ
is the mean free path between two collisions given by (NQm )−1 . Then, by
using g(v), DT is expressed in an integral form:

1 v 4π v3
DT = g(v)dv = g0 (v)dv. (2.27)
3 NQm (v) 3 NQm (v)

The expression by the energy distribution is

√
1 2 ε
DT = f (ε)dε. (2.28)
3 m NQm (ε)
In addition to elastic collisions, there are a number of inelastic processes
that may occur. The most significant is the ionization whereby electron col-
lision with a molecule of the background gas produces the pair of a new
electron and positive ion. The inelastic processes have a threshold energy
as the molecular quantum structure and they are usually characterized by
the rate (number of events per second per electron). The ionization rate is
calculated from the ionization cross section Qi () as

Ri = NQi (v)vg(v)dv
v

eE cos θ dg0 (v)
= NQi (v)v g0 (v) − 2π v 2 sin θ dθ dv
v,θ mNQm (v)v dv

= NQi v(v)g0 (v)4π v 2 dv
v

2 √
= NQi (ε) ε f (ε)dε. (2.29)
m ε
Another traditional way to describe ionization events is to define the num-
ber of ionization events that one electron makes while crossing a unit distance
α (also known as Townsend’s ionization coefficient), which can be easily de-
fined as
Ri
α = . (2.30)
vd
Normalization of the energy distribution f (ε) used in all calculations of
transport coefficients is given by f (ε)dε = 1 (see Equation 2.21). The process
that leads to production of a negative ion and loss of the original electron is the
electron attachment described by the cross section Qa (). These two processes,
ionization and electron attachment, change the number of electrons and thus
are labeled as nonconservative. All these coefficients or rates are directly used
in deriving electron transport (fluid) equations.
There are also a number of inelastic but conservative (number-
conserving) processes (Qk ) that are associated with a wide range of possible
energy losses (thresholds). Typically these energy losses εk are quite large,
and these processes are much more efficient in controlling the mean energy
of electrons than elastic processes, which have an energy loss of approximately
2mε/M. On the other hand, the momentum loss (exchange) is most influ-
enced by the elastic scattering, whose cross section is usually much greater
than the sum of the inelastic processes, Qm (ε) Qi (ε) + Qa (ε) + Qk (ε).
Therefore, we measure the contribution of the elastic scattering to the elec-
tron transport in terms of the momentum transfer. Rates and spatial excitation

coefficients are defined in the same way as for ionization. However, these co-
efficients do not directly enter the transport equations, although they may be
used in some corrections and energy balance equations.
Thus momentum transfer is dominated by elastic collisions, whereas
energy balance is primarily controlled by inelastic processes (with the excep-
tion of rare gases and some metallic vapors that have no inelastic processes at
low energies). However, it is not practical to use mean energy as the swarm
parameter because it cannot be measured directly, so we traditionally define
a property of electron swarms known as characteristic energy εk = e DT /µ,
where µ = vd /E is the mobility. Characteristic energy has the dimension of
energy and is usually within 30% of the mean energy (although sometimes
the differences may be up to a factor of two or more). This quantity is usually
expressed in units eV or, even more often, by the analogous quantity D/µ,
which is expressed in volts. The definition of characteristic energy is given by
combining Equations 2.26 and 2.28:
√
ε
DT DT −e E 1
3
2
m NQm (ε)
f (ε)dε
εk = e = e E = ,
µ vd 1 2
e E NQεm (ε) dε
d f√(ε)
dε
3 m ε
which for the thermal Maxwellian distribution in Equation 2.22, in particular,

becomes

ε ε
NQm (ε)
− 32 ε
exp dε 2
εk = = ε [eV]. (2.31)
ε ε 3
3
2ε NQm (ε)
exp − 32 ε dε
In the case of the thermal equilibrium where ε = 3kT/2, we obtain the
following relation:
DT kT
= . (2.32)
µ e
This is also known as Einstein’s relation, because Einstein derived an iden-

tical equation for the Brownian motion of particles. Moreover, Nernst derived
the equation for the electrolytic transport of ions, and Townsend derived it for
the electrons in gas discharges even before Einstein, so this relation is often
referred to as the Nernst–Townsend relation. This relation is valid only for
thermal equilibrium, although it is often used, especially in plasma model-
ing, for nonequilibrium conditions. In that case, however, it is possible to use
corrected forms that are more appropriate.
PROBLEM 2.2.2
Discuss the momentum transfer loss of inelastic collision, and compare the influence
of the collisional effect between the energy loss and the momentum loss.

2.2.2 Ion Velocity Distribution and Mean Energy

Ions, in principle, follow the same laws that determine the electron transport
in gases under the influence of electric or magnetic fields. The basic differences
are that the mass of ions Mp is of the same order of magnitude as that of
gas molecules M, and that the charge q may not be equal to e and is either
positive or negative (here we use the subscript p to denote positive ions, but
the conclusions are generally valid for the transport of negative ions as well,
only the collisional processes are of a different nature). The energy exchange
in elastic collisions is generally significant in an ion swarm.
Let us consider an ion swarm with uniform density in real space in a steady
state in gases under an external field E = Ek. At low kinetic energy of ions, the
collision between the ion and neutral molecule is subject to an induced-dipole
interaction, that is, constant mean free time τ , which is given by
1/2
1 0 1/2 1 Mp M
τ= α −1/2 ,
1.105 π qN Mp + M
where α is the polarizability of the ion in the dipole interaction with the
neutral molecule [2]. The governing equation of the velocity distribution G(v)
of ions is then given in the form of the relaxation equation using the BGK
approximation [3, 4],
dG(v) qE dG(v) G(v) − G 0 (v)

= =− . (2.33)
dt Mp dvz τ
Here G 0 (v) is the distribution just after a collision and is approximated by a

two-temperature displaced Maxwellian velocity distribution,
1/2
β⊥ β
G 0 (v) = exp − β⊥ vr2 − β (vz − < vz > + q Eτ/Mp )2 , (2.34)
π π
where β⊥ = 1/(2 < vx2 >), and β = 1/2[< vz2 > − < vz >2 −(q Eτ/Mp )2 ].
Equation 2.33 is then solved as

β⊥ vz − < vz > + qEτ/Mp 1
G(v) = exp −β⊥ vr2 − +
2π(qEτ/Mp) qEτ/Mp 4β (qEτ/Mp )2

1/2 qEτ 1 vz − <vz > + qEτ/Mp
× erfc 2β − , (2.35)
Mp 4β (qEτ/Mp )2 2(qEτ/Mp )
where erfc is a complementary error function.

The momentum transfer of an ion in the elastic collision is M(v p − V)/
(Mp + M) where V is the velocity of the gas molecule. When we introduce the

relative velocity vγ = v p − V, we can write the momentum balance as

M
Mp vγ z k(1 − cos ω)Nσm (vγ , ω)vγ G(vγ )dvγ = qE , (2.36)
Mp + M
where σm is the differential cross section, ω is the angle of scattering in the

center of mass system, and G(vγ ) is the relative velocity distribution of ions.
As the collision between the ion and the molecule at low energy is subject
to the induced-dipole interaction, (1 − cos ω)Nσm (vγ , ω)vγ d is a constant
(Rm ), and Equation 2.36 reduces to

Mp M
Rm vγ z kG(vγ )dvγ = qE .
Mp + M
The integral represents the mean velocity, that is, drift velocity vdp in velocity
space, which is then equal to
Mp + M qE
vd p = . (2.37)
Mp M Rm
When an ion elastically affects a molecule and is isotropically scattered in the

center-of-mass system, the averaged energy loss per collision is given by
Mp M 2

εl = Mp v p − MV 2 . (2.38)
(Mp + M) 2
E / N = 20 Td T g = 300 K
1.0 3
Energy distribution f log (log10ε)
Collision frequency (a.u.)
200 Td
120 Td
Qm( γ) γ
0.5 2
3
kT
2 g
0 1
0.01 0.1 1 10
Relative kinetic energy (eV)
FIGURE 2.5
Energy distribution at three different E/N and elastic momentum transfer cross section of
O− in O2 .

PROBLEM 2.2.3
Derive the relation 2.38.
In a steady state, the energy balance of the ion swarm in unit time is given
by

Mp M 2 Mp + M 1
qE · vdp −
εl Rm =qEvdp 1− Mp v p −MV
2
=0,
(Mp + M)2 Mp M vd2 p
(2.39)
which is satisfied when

(Mp + M )vd2 p + MV 2 − Mp v 2p = 0 .
By using this relation we obtain the expression of the mean ion energy, which
is known as Wannier’s formula [2]:
1 1 1
ε p = Mp v 2p = (Mp + M)vd2 p + MV 2 . (2.40)
2 2 2
The physical meaning of Equation 2.40 is that the mean energy of ions is
separated into three parts: one due to the effect of the electric field given by
the drift velocity (first term), one due to the apparent kinetic energy of the
encounter molecule (second term), and the thermal energy equal to 3kTg /2
(third term).
PROBLEM 2.2.4
When an ion swarm has a shifted Maxwellian velocity distribution in gases with a
Maxwell distribution of Tg , derive that the relative velocity distribution between the
ion and the molecule G(vr ) also has the Maxwellian velocity distribution.
2.3 Thermal Equilibrium and Its Governing Relations

In the case of thermal equilibrium, all processes are balanced by opposite
processes, so it is possible to make some quite general distributions both in
real and velocity space. In this section, we deal primarily with electrons inas-
much as extensions to other charged particles are trivial. One should bear in
mind that in thermal equilibrium all particles should in principle have the
same temperature and that the temperature remain constant throughout the
whole region, which is in equilibrium. This rule of homogeneous temperature
is often difficult to satisfy when it comes to plasmas. We therefore introduce
the term “local thermodynamic equilibrium” (LTE), which means that the
conditions for thermal equilibrium are satisfied at each point in the plasma
but that the value of the effective temperature varies with position. In the

case of LTE, the temperature is merely a parameter used to apply the laws of
thermodynamic equilibrium to a plasma, but this approximation is neverthe-
less quite effective.
2.3.1 Boltzmann Distribution in Real Space

When there is an external electric field, the flux of charged particles consists
of drift and diffusion components. Let us consider the z-axis along the field
direction. In a weak field, the drift velocity vd is proportional to E and is
written as vd = µE using the mobility µ. Therefore, the flux in the z direction
is expressed as
Γ = nv
∂n
= nvd − D∇n = nµE − D k. (2.41)
∂z
Given that, in thermal equilibrium under randomization, the mean velocity

must be zero v = 0, Equation 2.41 leads to a differential equation that
connects the spatial density distribution of charged particles n(z) and the
spatial potential V(z):
µ −µ ∂V(z) 1 ∂n
E= = . (2.42)
D D ∂z n ∂z
This equation is solved by direct integration if we define the boundary condi-

tion such that for z = 0, the potential is V(0) = 0 and the density is n(0) = n0 .
In this case the density distribution is
µ
n(z) = n0 exp − V(z) . (2.43)
D
We substitute the Einstein relation 2.32 into Equation 2.43 and obtain

eV(z)
n(z) = n0 exp − , (2.44)
kT
which is known as Boltzmann’s distribution law. This equation describes

how the population of charged particles in thermal equilibrium with effective
energy kT will be distributed under the space potential V(z). Thus, even for a
potential that corresponds to an energy much larger than the thermal energy,
there will be a certain number of electrons, although this number is very small.
The Boltzmann distribution is often applied to determine the population of
excited molecules at different discrete energy levels, in which case it is actually
applied to bound electrons and their population at discrete bound levels of
atoms or molecules.

gM
0
Vz
Vx
FIGURE 2.6
Equilibrium velocity distribution in vx –v y space.
2.3.2 Maxwell Distribution in Velocity Space

The probability that a charged particle will be found in a small volume of
velocity space defined by (vx , vx + dvx ), (v y , v y + dv y ), and (vz , vz + dvz ) is
given by
g(vx , v y , vz )dvx dv y dvz , (2.45)
where g(vx , v y , vz ) is the velocity distribution. Under conditions of thermal

equilibrium, the distribution g(vx , v y , vz ) must satisfy the following
conditions:
(a) Each of the components of the velocity distribution must be fully

independent of each other, and
(b) The velocity distribution must have the same magnitude in all
directions, and the mean values of velocities in all directions must
be zero.
In Figure 2.6 we show a typical distribution function in two dimensions.

When we apply conditions (a) and (b), we expand the distribution into three
independent functions
g(vx , v y , vz )dvx dv y dvz = G(vx )dvx G(v y )dv y G(vz )dvz

= g(v 2 )dvx dv y dvz , (2.46)
where the magnitude of the velocity v is defined by v 2 = vx2 + v 2y + vz2 , and the
distribution in terms of v 2 is then defined as
g(v 2 ) = G(vx )G(v y )G(vz ). (2.47)
We must choose the function g to satisfy the functional Equation 2.47.

Exercise 2.3.1
Show that the solution to Equation 2.47 is

g vx2 = a 2exp −αvx2 , (α > 0). (2.48)
Let us observe a special case of one-dimensional motion where v y = vz = 0.

In this case, Equation 2.48 is written as

g vx2 = G(vx )G(0)G(0)
= a 2 G(vx ),
where a = G(0) and, by the same argument, g(v 2y ) = a 2 G(v y ) and g(vz2 ) =
a 2 G(vz ). Therefore, Equation 2.47 is rewritten as
1
g(v 2 ) = g vx2 + v 2y + vz2 = 6 g vx2 g v 2y g vz2 . (2.49)
a
Note that the functional Equation 2.49 satisfies Equation 2.48 as the solution
of each component.
Therefore, we obtain

g(v 2 ) = g vx2 + v 2y + vz2 = A exp −α vx2 + v 2y + vz2 , (2.50)
where the value of the constants A and α must be determined. When we

normalize the velocity distribution g(vx , v y , vz ) to 1,

1 = g vx2 + v 2y + vz2 dvx dv y dvz
∞ ∞ ∞

= A exp − αvx2 dvx exp − αv 2y dv y exp − αvz2 dvz
−∞ −∞ −∞
3
π
= A ,
α
and therefore
3/2
α
A= .
π
Now we apply the distribution g(vx , v y , vz ) to determine the mean kinetic

energy of charged particles with mass m at temperature T according to the
conditions of thermodynamic equilibrium
3 1 α 3/2 m
vx2 + v 2y + vz2 e −α(vx +v y +vz ) dvx dv y dvz
2 2 2
kT = mv 2 =
2 2 π 2
α 3/2 3m 1 π 3/2
= ,
π 2 2α α

which leads directly to

m m 3/2
α= and A= . (2.51)
2kT 2π kT
We use the constants 2.51 to obtain the final form of the Maxwellian distribu-
tion from Equation 2.50,

m 3/2 m vx2 + v 2y + vz2
g M = g(vx , v y , vz ) = exp − . (2.52)
2π kT 2kT
In the previous derivation of Maxwell’s velocity distribution, we used the

∞ √ ∞ √
following integrals: 0 e −αx d x = 1/2 π/α, and 0 x 2 e −αx d x = 1/2α π/α.
2 2
In place of Maxwell’s velocity distribution, the Maxwellian speed distribution

F (v) is defined as

F (v)dv = g M (v 2 )v 2 sin θ dθdϕdv
θ ϕ
= 4πv g M (v 2 )dv .
2
(2.53)
The following averaged velocities are defined for the Maxwellian speed
distribution: (i) most probable speed ṽ, (ii) root mean square velocity v 2 ,
and (iii) mean speed v. These are obtained as follows.
i. The most probable speed ṽ is obtained from the condition

dF /dv = 0, and its value is

2kT 1/2
ṽ = . (2.54)
m
ii. The root mean square velocity (v 2 )1/2 is the square root of the mean
value of v 2 :

3kT
v 2 = v 2 F (v)dv = . (2.55)
m
iii. The mean speed is obtained as

8kT
v = vF (v)dv = . (2.56)
πm
We can easily compare the magnitudes of these three mean values as

ṽ < v < v 2 . (2.57)
Figure 2.7 shows the plot of the normalized Maxwell speed distribution
F (v) and gives the values of the three possible choices for averaging the ve-
locity. Different physical quantities may be averaged with different powers of

0.8
0.6
F(v)
0.4
0.2
0 v
0 0.5 1.0 1.5 2.0 ~
v
< v> 兹<v2>

~
v ~
v
Normalized velocity
FIGURE 2.7
Normalized Maxwell speed distribution and the relationships among these averaged velocities.
speed and will be associated with different effective velocities. It is clear that
the mean speed may be used to establish the momentum transfer, that the
mean square speed is relevant for consideration of the mean energy, and that
the most probable speed is the peak of the probability distribution. Although
one may directly produce averages of physical phenomena, associating them
with mean values of speed provides a better physical insight into the phe-
nomenon under consideration.
References
1. Parker, J.H., Jr. and Lowke, J.J. 1969. Phys. Rev. 181:290–301.
2. Wannier, G.H. 1953. Bell Syst. Tech. J. 32:170.
3. Bhatnagar, P.L., Gross, E.P., and Krook, M. 1954. Phys. Rev. 94:511.
4. Whealton, J.H. and Woo, S.B. 1971. Phys. Rev. A 6:2319.

3
Macroscopic Plasma Characteristics
3.1 Introduction
Plasma is a compound phase with electric quasi-neutrality, generally consist-
ing of electrons, positive ions, and neutral molecules. In particular, strongly
ionized plasma consists of electrons and positive ions. The phase is called
collisionless plasma, as the electron (ion) has few short-range binary colli-
sions with neutral molecules, and the plasma system is subject to long-range
Coulomb interactions. When the short-range two-body collision is major, the
system is collisional, that is, collision dominated. Collisionless plasma shows
unique characteristics. We briefly describe the characteristics [1–3].
3.2 Quasi-Neutrality
The presence of electrons and ions with density ne and np will produce an
electric field subject to Poisson’s equation,
n p − ne
divE = e , (3.1)
0
where e(>0) is the elemental charge and 0 is the permittivity (dielectric con-
stant) of vacuum. A plasma where the density of positive ions is exactly
equal to that of electrons has no electric field. Naturally, based on the great
difference in mass between the electron and positive ion, a quasi-neutral
state with np ∼ ne is maintained in plasmas on a macroscopic scale. This
causes a finite positive plasma potential with respect to the wall earthed to the
ground.
Notice that a low-temperature plasma externally excited by direct current
(DC) or rf power source has a plasma production region and holds a structure
bifurcating into a positive ion sheath with a high field to produce an electron–
ion pair and a bulk plasma with quasi-neutrality.
27

plasma
(ne~np)
p-ion
d electron
FIGURE 3.1
Plasma fluctuation and the charge separation in a collisionless plasma with macroscopic quasi-
neutrality.
3.3 Charge-Separation in Plasmas

3.3.1 Spatial Scale of Charge-Separation
In a bulk plasma, quasi-neutrality will be macroscopically kept in space and
time. The quasi-neutral condition, however, will be locally disturbed by the in-
trinsic random motion of the charged particles in a plasma. We first assume the
presence of a localized space–charge layer as shown in Figure 3.1 as a result of
the random fluctuation of charged particles in a bulk plasma with macroscopic
quasi-neutrality. In one-dimensional space, the local field and the potential
difference caused by the two layers is given by Poisson’s Equation 3.1 as
en end
E= and V= ,
ε0 ε0
where n(∼ ne ∼ n p ), is the thickness of each of the layers, and d is the
distance between layers.
In a collisionless plasma, when the potential energy of the charged particle
in the layers is less than the random thermal energy, kTe or kTp ,
end
e < kTe , (3.2)
ε0
the local disturbance of the charge-separation will be kept. Considering the
scale of ∼ d,

ε0 kTe 1/2
d< ≡ λ D. (3.3)
ne 2
Here, λ D defined in Equation 3.3 gives a microscopic maximum spatial
scale for the charge-separation and is known as the electron Debye (shielding)
length or Debye radius.

Macroscopic Plasma Characteristics 29
Exercise 3.3.1
Calculate the Debye length when plasma density and electron temperature
in a collisionless plasma are 1016 m−3 and 3 eV, respectively.

ε0 kTe 1/2 kT[eV] 1/2
λD = ∼ 7.43 × 10 3
× [m]. (3.4)
ne 2 n[m−3 ]
That is,
1/2
3
λ D = 7.43 × 10 × 3
= 1.29 × 10−4 [m].
1016
3.3.2 Time Scale for Charge-Separation

As there exists a huge difference in mass between the electron and the posi-
tive ion, the electron layer in Figure 3.1 will go toward the positive ion layer
under the Coulomb force. In a collisionless plasma, the local drift motion of
the electron layer passes through the massive positive layer by inertia, and
the pair of charged layers is oppositely reformed. The electron layer will con-
tinue to vibrate around the massive positive ions in the absence of two-body
collision between the neutral molecules. The temporal motion is described by
the harmonic oscillation,
d2 e 2n
m (t) = − (t). (3.5)
dt 2 ε0
The oscillation is known as the electron plasma oscillation or Langmuir
oscillation. The frequency of the electron plasma oscillation, electron plasma
frequency, is numerically given as
1/2
e 2n 1/2 −1
ωe = ∼ 56.4 n[m−3 ] [s ], (3.6)
mε0
The time scale of the charge-separation in a collisionless plasma τes , given by
ωe−1 , has the relation with λ D ,

λD kTe 1/2
= ≈ Ve , (3.7)
τes me
where Ve is the electron thermal speed.
PROBLEM 3.3.1
Discuss that the plasma frequency ω p of the ion with charge Ze is defined in the same
way as the electron,
1/2 1/2
Z2 e 2 n p m
ωp = =Z ωe . (3.8)
Mp ε0 Mp

Also discuss that practical plasma oscillation is given by (ωe2 + ω2p )1/2 =
ωe (1 + z2 me /Mp )1/2 .
PROBLEM 3.3.2
Estimate the values of ωe and ω p in collisionless Ar plasma with density 1016 m−3 .
PROBLEM 3.3.3
A plasma is externally irradiated by an electromagnetic wave with frequency of ω.
Derive the condition that the external disturbance does not immerse deeply in the
plasma.
3.4 Plasma Shielding

3.4.1 Debye Shielding
When a material is inserted into a plasma, the surface of the material is imme-
diately charged regardless of metal or insulator, and it is electrically shielded
from the surrounding plasma (see Figure 3.2). We consider the spatial scale
of the shielding caused by a spherical metal inserted into a plasma with zero
space potential. The potential distribution close to the metal is obtained by
Poisson’s equation as a function of radial distance r ,

1 d dV (r ) e
∇ 2 V (r ) = r2 =− n p − ne . (3.9)
r 2 dr dr ε0
There are number of electrons and ions in the shielded region in front of the
sphere. The number density of electron and ions, ne (r) and n p (r), follows the
V n
np(r)
n0
q
ne(r)
4πε0r
-V(r)
0 r
λD
FIGURE 3.2
Plasma shielding of a charged sphere.

Boltzmann distribution in space, discussed in Section 2.3,

eV (r )
n p = n0 exp − , (3.10)
kTp

eV (r )
ne = n0 exp , (3.11)
kTe
where n0 is the quasi-neutral plasma density far from the influence of the
impurity metal. At a long distance from the metallic surface satisfying eV(r )
kTe , kTp , the exponential terms in Equations 3.10 and 3.11 will be expanded
by the Taylor series, and we obtain

1 d dV (r ) en0 eV (r ) eV (r )
r2 =− exp − − exp
r 2 dr dr ε0 kTp kTe
e 2 n0 Tp + Te
≈ V (r ) . (3.12)
ε0 kTp Te
The general converging solution of Equation 3.12 is

A r
V (r ) = exp − ,
r λD
where A (> 0) is constant and λ D is
ε0 kTp Te
λ2D = . (3.13)
e 2 n0 Tp + Te
The potential close to the metal with a surface charge q 0 is expressed by the
Coulomb potential, q 0 /4π0r . Then, the final form is

q0 r
V (r ) = exp − . (3.14)
4π ε0r λD
Noted that V(r ) is negative due to q 0 (< 0) under me
Mp . Equation 3.14
shows that the scale length of the shielding is on the order of the Debye
length λ D .
3.4.2 Metal Probe in a Plasma

A small metallic probe with variable potential is inserted into a plasma in a
steady state. A static shielding sheath is formed there without plasma pro-
duction or loss. We consider the voltage-current characteristics I (V) when
the potential at the probe is externally varied with respect to the wall of the
plasma. They are known as probe characteristics in a plasma. Figure 3.3 shows
typical probe characteristics.

I
insulator metal a
b
(I)
plasma (II)
λD
VS
Vfl VS
d c V
Isi
(III)
FIGURE 3.3
Ideal Langmuir probe characteristics in a plasma.
When the probe has a sufficiently large surface, a saturated electron cur-
rent will be collected at V > Vs where Vs is the space potential of the plasma.
Practically due to the finite surface area, I (V) characteristics at V > Vs will
increase gradually as shown in Figure 3.3. The probe potential at c where the
current is zero is named the floating potential. The name comes from the zero
net flux at the surface of an insulator inserted into a plasma.
The probe will be given an absorbing boundary that absorbs electrons and
ions incident on the surface. At the region bc that satisfies V < Vs , the electron
sheath will be formed in front of the probe (Vs − V < 0). The electron incident
right on the probe surface will be retarded by the field, and only the electron
with kinetic energy mv 2 /2 > |e(Vs − V)| will be absorbed as the electron cur-
rent. We consider that the electron has the Maxwellian velocity distribution
with temperature Te in the plasma and has no collision in the passive sheath
in front of the probe. Then the electron probe current will be given by
∞ ∞ ∞ 3/2
m
I (V) = ene S 2e|V | vz
vx =−∞ v y =−∞ vz = m
p 2π kTe

m(vx2 + v 2y + vz2 )
× exp − dvx dv y dvz
2kTe
3/2 ∞ ∞
m mvx2 mv 2y
= ene S exp − dvx exp −
2πkTe −∞ 2kTe −∞ 2kTe
∞
mvz2
× dv y exp − dvz
vz 2kTe
3/2 √ √
m 2 π 2 π 1 eVp
= ene S m exp − , (3.15)
2πkTe 2 2kT m
2 m
kTe
kTe
e 2kTe

where S is the effective area of the probe. By considering the mean speed
ve = (8kTe /πm)1/2 in Equation 3.15, the probe I (V) characteristics are
described by

ene ve V − Vs
I (V) = S exp . (3.16)
4 kTe
Equation 3.16 shows that the electron temperature Te in the plasma is

obtained by the gradient of ln I (V ). At ab satisfying V > Vs (electron saturated
region), all electrons approaching the probe will be collected, the lower limit
of the integral in Equation 3.15 will be zero, and
ene ve
I = S. (3.17)
4
Equation 3.17 means that electron current consists of ne /2 in one-

dimensional position space and ve /2 in velocity space. In principle, by us-
ing the electron temperature Te , the plasma density (electron density) will be
given by the saturated curve, Equation 3.17, in the region of bc. The probe
having one small metal is called the single probe or the Langmuir probe.
Other types of probes are also used as a simple tool for plasma diagnostics.
Exercise 3.4.1
Discuss the pressure condition that the electron temperature Te is estimated
from the curve in region (II) in bc in Figure 3.3.
As a typical plasma we assume, ne = 1015 m−3 , kTe = 3.0 eV. Then, the Debye
length is
1/2 √
kTe [eV] 7.43 × 103 × 3
λe = 7.43 × 10 × 3
= √ = 4.07 × 10−4 [m].
ne [m−3 ] 10 × 107
The mean speed of electrons is

ve = (8kTe /πm)1/2 = 6.71 × 107 × kTe [eV] ≈ 1.16 × 106 [ms−1 ].
The collision rate R is roughly approximated at 107 p[Pa s−1 ], and the flight
time of the electron in the static sheath in front of the probe is

λe 4.07 × 10−4 1 1
≈ ≈ 3.51 × 10−10
≈ .
ve 1.16 × 106 R 107 p
Therefore, the collisionless condition is obtained as
1 1
p
× ≈ 2.85 × 102 [Pa].
107 3.51 × 10−10

plasma wall
(ne ~ np) A+
e
Γe ~ Γp
Ea
FIGURE 3.4
Fluxes of charged particles in front of a wall.
3.5 Particle Diffusion

3.5.1 Ambipolar Diffusion
We consider a plasma without production in a vessel far from a plasma source.
Electrons and ions in the plasma have a random motion with a kinetic energy.
Even in the same kinetic energy (temperature) between electrons and ions,
the electron flux incident on the wall as a result of the random motion will
be much higher than the massive positive ion, and the relation e p is
immediately attained (see Figure 3.4). Then, an electric field that accelerates
positive ions and decelerates electrons toward the wall surface (i.e., ambi-
polar field) will be formed, and in a macroscopically steady state, the charged
particles’ flow to the wall with the relation
e ≈ p (3.18)
is realized.
We deal semiquantitatively with the ambipolar diffusion phenomena
caused between the static plasma and the surrounding reactor wall. Each
of the fluxes of electrons and ions with density, ne and n p , is given,
dne
Γe = ne ve = ne vde − De , (3.19)
dr
dn p
Γ p = n p v p = n p vd p − Dp , (3.20)
dr
where v j , vd j , and D j are the mean velocity, drift velocity, and diffusion co-
efficient of electrons and positive ions, respectively. In a steady state as men-
tioned above, the electron and ion have the same velocity toward the wall,
ambipolar diffusion velocity va ,
1 dn
va = −Da , (3.21)
n dr

where Da is the ambipolar diffusion coefficient given by
µ p De + µe Dp
Da = , (3.22)
µ p + µe
where µe and µ p are the mobility of electrons and ions, respectively. In

particular, at low reduced field the mobility may be constant. In thermal
equilibrium of electrons and ions with Maxwellian velocity distribution, the
ambipolar diffusion coefficient is expressed in terms of the electron and ion
temperatures, Te and Tp ,

Dp 1 + TTep
Da = . (3.23)
µ
1 + µep
In the case of Te Tp ,
Te
Da = Dp . (3.24)
Tp
In particular, in Te = Tp ,
Da = 2Dp . (3.25)
PROBLEM 3.5.1
Derive the ambipolar diffusion coefficient, Equation 3.22, and show that the ambipolar
diffusion field is written as
Dp − De 1 dn
Ea = . (3.26)
µe + µ p n dr
3.5.2 Spatial and Time Scale of Diffusion

We consider the time and spatial scale of particles in a collisionless or col-
lisional plasma. In the case where the production and loss of the particle is
negligible, the continuity equation of the particles is
∂
n(r, t) = D∇ 2 n(r, t). (3.27)
∂t
The diffusion Equation 3.27 implies that the spatial distribution and time
scale of the diffusion are determined by the geometry of the reactor. Here, we
divide n(r, t) into two independent functions n(r) and T(t). Then we have
1 dT (t) 1 1
= D∇ 2 n(r) ≡ − . (3.28)
T (t) dt n(r) τD
Here, τ D (> 0) is the separation constant.

PROBLEM 3.5.2
Discuss the physical meaning that the r.h.s. of Equation 3.28 has negative value.
Equation 3.28 is written as
T (t) = T0 exp(−t/τ D ) , (3.29)
and
1 n(r)
∇ 2 n(r) = − n(r) = − 2 , where 2 = Dτ D . (3.30)
Dτ D
Equation 3.30 shows that the spatial characteristics of particle diffusion 2

are determined by the geometrical boundary condition. is called the char-
acteristic diffusion length and τ D the diffusion decay time.
PROBLEM 3.5.3
Calculate the characteristic diffusion length in the case where the reactor geometry
is (a) an infinite parallel plate (separation of d0 ), (b) an infinite rectangle (two sides,
d0 and l0 ), (c) a cylinder (radius r0 ), and (d) a sphere (radius r0 ), respectively.
1
= (π/d0 )2 ; infinite plates,
2
= (π/d0 )2 + (π/l0 )2 ; infinite rectangle,
= (2.405/r0 )2 ; cylindrical,
= (π/r0 )2 ; spherical. (3.31)
PROBLEM 3.5.4
Neutral molecules diffuse in three-dimensional space without boundary. Calculate
the number density of molecules n(r, t) in Equation 3.27. In particular, when the
initial condition of the density is given by n(r, t = 0) = δ(r), derive the density
distribution
3/2
1 r2
n (r, t) = exp − . (3.32)
4π Dt 4Dt
The above stochastic process is known as the Wiener process and the
density distribution is the normal distribution N(0, 2Dt) with mean value of
0 and variance of 2Dt.
The characteristic diffusion length D is much influenced by the loss
mechanism of the particle. For example, we consider the electron diffusion in
electronegative gases with density N and with the electron attachment rate
coefficient ka . Then the electron continuity equation is
∂
n(r, t) = −ka n(r, t)N + D∇ 2 n(r, t). (3.33)
∂t

Equation 3.33 is reduced to

1 dT (t) 1 1
= −ka N + D∇ 2 n(r) ≡ − . (3.34)
T (t) dt n(r) τ Dr
As a result, the characteristic diffusion length changes from Equation (3.30)
to

1 ka N −1
=
2
− . (3.35)
Dτ D D
Exercise 3.5.1
Molecules excited to an optically forbidden level are called metastables and
have a long lifetime. The metastable state is de-excited without photo-
emission when it reacts with the wall with absorbed molecules and exhausts
its excess inner energy to the wall. Estimate the effective lifetime of metasta-
bles in a reactor of volume Vol with reflecting wall. Reflection coefficient γr at
the wall is given by γr = 1 − St , where St is the surface sticking coefficient,
2 2
2 2Vol (2 − β)
τ= + . (3.36)
π D St vβ
3.6 Bohm Sheath Criterion

3.6.1 Bohm Velocity
A fluid description of a region between a plasma and material boundary (i.e.,
plasma sheath) will give a simple physical image from the ambipolar diffusion
transport of the charged particle. As described in the previous section, the net
current to the surface disappears as a result of the ambipolar diffusion in the
steady state, and the surface is kept at the floating potential. The electron and
ion dynamics are valid in the sheath under the following assumptions.
1. The frequency of the external power source is much higher than the
collision rate ω R (i.e., the collisionless sheath).
2. In the momentum continuity equation (see details in Chapter 5),
the pressure gradient dP(z)/dz = kTdne (z)/dz is dominant for elec-
trons, whereas other terms except the pressue are predominant for
ions.
3. The ion flux is constant, that is, n p (z)vdp (z) = const, without ion-
ization and recombination in the collisionless sheath. Under these
conditions, the momentum equations of electrons and ions in the
steady state are, respectively,
dP(z) dV(z)
me = −ene (z) , (3.37)
dz dz

dvdp dV(z)
Mp n p (z)vd p = en p (z) . (3.38)
dz dz
From the above ion flux continuity and Equation 3.38, we obtain
n p (z0 )vd p (z0 )

n p (z) =
vdp (z)
1
= n p (z0 ) 1/2 .
2e[V(z)−V(z0 )]
1− Mp vd p (z0 )2
In a local thermal equilibrium (LTE) under Equation 3.37, the electron

number density has the Boltzmann distribution (see Section 2.3). The region
from a bulk plasma/sheath boundary at z0 to the sheath terminal, consisting
of electrons and ions, is expressed by Poisson’s equation
dV(z)2 e
= − [n p (z) − ne (z)]
dz2 ε0

ene (z0 ) 2e[V(z) − V(z0 )] −1/2 e[V(z0 ) − V(z)]

=− 1− − exp − .
ε Mp vd p (z0 )2 kTe
In the sheath, positive ions are dominant; that is, n p (z) − ne (z) > 0. Accor-
dingly,
−1/2
2e[V(z) − V(z0 )] e[V(z) − V(z0 )]
1− − exp > 0. (3.39)
Mp vdp (z0 )2 kTe
In particular, at the position z close to the boundary between the bulk plasma
and the sheath, e[V(z) − V(z0 )]
kTe will be satisfied. By using a Taylor
expansion of the exp term in Equation 3.39,

2e[V(z) − V(z0 )] 2e[V(z) − V(z0 )]
1− < 1− .
kTe Mp vd p (z0 )2
Finally, we have the initial directional velocity of ions just entering the
sheath from the bulk plasma (i.e., Bohm velocity):
1/2
kTe
vd p (z0 ) > . (3.40)
Mp
3.6.2 Floating Potential

Here, we define the sheath edge at the position z0 where vd p is equal to
(kTe /Mp )1/2 .

Mp vd p (z0 )2 kTe
Vplasma − V(z0 ) = = . (3.41)
2e 2e


Therefore, the electron density at z0 is

e[Vplasma − V(z0 )]
ne (z0 ) = ne0 exp − = 0.61ne0 . (3.42)
kTe
When a small metal plate isolated electrically from the reactor is immersed in
a bulk plasma with a plasma potential Vplasma and electron temperature Te , the
surface potential Vf l , called the floating potential, is given at the condition of
zero net current ee = e p as

kTe Mp
Vf l = Vplasma − ln . (3.43)
2e 2.3m
Exercise 3.6.1
Derive the above expression.
Although ee = e p is satisfied, positive ions have a strong directional flux,
and electrons show an isotropic flux with random speed < ve >= (8kTe /π m)1/2
(see Chapter 2). Therefore,

1 e[Vplasma − Vf l ] kTe 1/2
ne0 exp − < ve >(= (n p (surf )vd p (surf )) = n p (z0 ) .
4 kTe Mp

By using Equation 3.42 under ne (z0 ) ∼ n p (z0 ),

e[Vplasma − Vf l ] π m 1/2
− = ln 2.4 .
kTe 8Mp
Finally we will obtain Equation 3.43.
References
1. Chen, F.F. 1984. Introduction to Plasma Physics and Controlled Fusion, Vol. 1.
New York: Plenum.
2. Golant, V.E., Zhilinsky, A.P., and Sakharov, I.E. 1980. Fundamentals of Plasma
Physics. New York: John Wiley & Sons.
3. Cherrington, B.E. 1979. Gaseous Electronics and Gas Lasers. Oxford: Pergamon.

4
Elementary Processes in Gas Phase
and on Surfaces
A large number of individual elementary processes are important in

determining the kinetics of low-temperature plasma. The basic processes
in the gas phase include collisions of electrons, ions, and fast neutrals with
molecules of the feed gas and with surfaces. Collisions between charged par-
ticles, collisions between excited particles, and collisions of radicals with gas
molecules are also important. In addition, one needs to consider ion–molecule
reactions and different types of photon interactions with molecules, excited
species, and surfaces. It is essential to understand the nature of elementary
processes in order to investigate the collective phenomena and kinetics of
plasmas as a whole. Electron collisions are particularly important, because
they can produce new electrons and ions under most conditions and also
produce the chemically active species required for numerous plasma appli-
cations.
Gas phase collisions are dominant in low-temperature plasmas. If we
change the background feed gas in a low-temperature plasma reactor, the
nature of the plasma will change considerably. In highly ionized plasmas,
the nature of the gas is relatively unimportant (unless thresholds for ioniza-
tion change dramatically), because they are dominated by charged particle
collisions. In an earlier chapter we have seen that the motion of charged
particles between two collisions is classical unless the gas density is very
high. However, quantum effects do occur for a very brief period of time dur-
ing the collision. Thus we must briefly review the quantum representation of
particles and particle beams and quantum scattering theory [1.2]. Classical
collision theory is also reviewed in this chapter, because it proves useful in
some cases. We then identify the basic characteristics of different collision
processes [3.4.5]. We analyze electron collisions over a very wide range of
energies while briefly discussing ion and fast neutral scattering over ranges
relevant for practical plasmas. Finally, we discuss collisions and reactions of
neutrals and ions at energies close to room temperature, which is appropriate
for low-temperature plasmas.
Numerous elementary processes take place at surfaces. In this chapter we
consider surfaces as a source of charged particles both through collisions and
41

arising from surface heating and surface reactions. A special group of surface
processes associated with plasma etching are also described briefly here and
in greater detail in Chapters 8 and 12.
4.1 Particles and Waves

4.1.1 Particle Representation in Classical and Quantum Mechanics
An individual particle in classical physics is represented by its mass m,
its energy ε, and its momentum p. Assuming that the particle is moving
along √ the z-axis, the components of the momentum are p( px = 0, p y = 0,
pz = 2mε). In Figure 4.1 we show a classical representation of a moving
particle. In quantum mechanics, the motion of a particle is described by a
matter wave (de Broglie wave). Individual atomic and elementary particles
are subject to quantum mechanical laws; therefore, their properties are sub-
ject to the Heisenberg uncertainty principle, which relates the uncertainty of
the spatial coordinate r with the uncertainty of the momentum p. The
uncertainty relation can be written as
h
r · p ≥ (= h̄), (4.1)
2π
where h is Planck’s constant.
In quantum mechanics, the energy and the momentum of the particle (ε, p)
are related to the angular frequency and the wave number of the plane wave
(ω, k) through
ε = h̄ω (4.2)
p = h̄k. (4.3)
A particle moving along the z-axis is represented as a plane wave
(r, t) = A0 exp[ j (k · r − ωt)], (4.4)
as shown in Figure 4.1. Here, A0 is the amplitude, and k · r = const expresses

the wave front. This plane wave is fully defined by two pairs of quantities,
(ω, k).
PROBLEM 4.1.1
Show that the plane wave satisfies the uncertainty principle.
Exercise 4.1.1
Calculate the de Broglie wavelength λe for the electron that has crossed a
potential drop V0 (see Figure 4.2).

Elementary Processes in Gas Phase and on Surfaces 43
wave
front
k . r = const.
at t = t0
r
p
z
m 0 k λ λ
(particle) (plane wave)
FIGURE 4.1
Classical and quantum (plane wave) description of a particle with the mass m and momentum
p(0, 0, pz ).
The energy of a particle is equal to the potential drop V0 times the elementary
charge e, so we can obtain the momentum of the particle as a function of V0 :
ε = p 2 /2m = eV0 .
Bearing in mind the de Broglie relation in Equation 4.3, we obtain

h 2π h
p = h̄k = · = .
2π λe λe
Finally, we have
h
λe =
(2me V0 )1/2
1/2
6.63 × 10−34 1.50
= = [nm]. (4.5)
(2 × 9.11 × 10−31 × 1.6 × 10−19 × V0 )1/2 V0 [V]
C A G
λe
electron beam
emitter
V0
FIGURE 4.2
A stationary electron beam with mono-energetic energy eV0 produced in a vacuum.

Exercise 4.1.2
Obtain the dispersion relation between the frequency and the wave number
for a plane matter wave (r, t) in free space (see Equation 4.4).
If we start from the time- and space-dependent Schrödinger equation
∂
H (r, t) = jh̄ (r, t), (4.6)
∂t
with the Hamiltonian for a free particle given by
h̄ 2 2
H=− ∇ + V(r, t), (4.7)
2m r
then the solution in free space (V = 0) to this system with the plane wave in
Equation 4.4 gives us
h̄ 2 k 2
(r, t) = h̄ω (r, t).
2m
Finally, we obtain the dispersion relation (between the wave number k and
the angular frequency ω)
h̄k 2
ω(k) = . (4.8)
2m
PROBLEM 4.1.2
The wave-front of a plane wave is defined as k · r = const at time t, and in that case
the wave front is perpendicular to the wave vector k. Show that a spherical wave with
a circular wave front is expressed as
A0
(r, t) = exp[ j (k · r − ωt)]. (4.9)
r
4.1.2 Locally Isolated Particle Group and Wave Packets

An isolated group of particles moving in the z-direction is expressed as a wave
packet. To obtain a wave packet, we can sum a group of plane waves in respect
to k within the boundaries (k0 ± k). It is assumed that (k − k0 ) = ξ k0 .
The summation may be converted to an integral as follows:
0 +k
k
(z, t) = (z, t; k)
k0 −k
k0 +k
= A(k) exp[ j (kz − ωt)]dk. (4.10)
k0 −k

λ0
FIGURE 4.3
Wave packet of a group of particles moving in the z-direction.
Furthermore, we apply the Taylor expansion to the dispersion relation around

k = k0 :
ω = ω(k − k0 )

dω 1 d 2 ω
= ω0 + (k − k ) + (k − k0 )2 + · · ·,
dk 0 2 dk2 0
0
and, as a result, we obtain

k
dω
(z, t) = A(k) exp j (k0 + ξ )z − j ω0 + ξ t dξ
−k dk 0
k
dω
∼ A(k0 ) exp[ j (k0 z − ω0 t)] exp j ξ z − ξ t dξ.
dk −k 0
By combining the above equation with

k
e ja k − e − ja k 2 sin(a k)
exp( ja θ )dθ = =
−k ja a
we obtain

dω
z− t k
dk 0
sin
(z, t) ∼ 2A(k0 ) exp[ j (k0 z − ω0 t)]. (4.11)
z − dω t
dk 0
In Figure 4.3, we show a plot of Equation 4.11. The locally isolated particle
group is represented by a strongly modulated wave with the fundamental fre-
quency ω0 (k0 ). There is an envelope with a maximum in the position where a
particle should be in a classical model (i.e., the center-of-mass) and the undu-
lations extend approximately over the one wavelength λ0 . The exponential
term in Equation 4.11 determines the phase of the wave.

The position of the maximum of the wave packet (z = dω/dk|0 t) moves

with a velocity

dz dω
Vg = = , (4.12)
dt dk 0
also known as the group velocity. For the particular example given in
Equation 4.8, the result is

d h̄k 2 h̄k0
Vg = = . (4.13)
dk 2m 0 m
If we remember that p(= mv) = h̄k0 , then Vg corresponds to the classical

velocity v.
Exercise 4.1.3
Derive the phase velocity of the wave packet in Equation 4.11.
The phase in Equation 4.11 is
= k0 z − ω0 t = const,
and the phase velocity Vp is equal to

dz d ω0 t ω0
Vp = = = .
dt dt k0 k0
Using Equation 4.8 allows the phase velocity to be related to the group velocity
1 h̄k02 h̄k0 Vg
Vp = = = .
k0 2m 2m 2
PROBLEM 4.1.3
A pulsed electron beam is formed in a vacuum by using the electrical shutter between
grids A and G in Figure 4.2. The half width of the group z is 1 mm, and V0 is 150 V.
Discuss the uncertainty principle using vz /vz .
4.2 Collisions and Cross Sections

An atomic collision is a stochastic process that takes place when two particles
approach one another at a sufficiently short distance so that their interaction
through different forces becomes appreciable. In weakly ionized plasmas, we
deal mainly with two-body, short-range collisions, which are then regarded
as occurring very quickly and over a very short distance as compared with the
mean (free) time and mean (free) path between successive collisions. Because

(a) scattered (b) scattered

beam spherical wave
dΩ dΩ
incident
plane wave
v
z z
m M
incident target
particle molecule
FIGURE 4.4
Classical (a) and quantum (b) representation of a collision.
the Coulomb force has infinite range, collisions between charged particles are
described as many-body, long-range collisions.
There is also a whole group of three-body processes where the third
particle is required to satisfy the conservation laws. A three-body process
occurs, for example, when one of the species is left in an excited state after
the two-body collision and the excited species needs a subsequent collision
with a third body to transfer its extra energy for stabilization. Three-body
processes will occur only if the gas density is sufficiently high.
Here we deal primarily with two-body, short-range collisions. In clas-
sical mechanics, particles have definite identities, positions, and velocities
(momenta). The classical collision of a light projectile of mass m and veloc-
ity v on a heavy target of mass M is depicted in Figure 4.4a. After the
collision the light particle will be scattered at an angle θ with respect to its
original direction z and into the solid angle d(= sin θ dθ dφ). It will have
a definitive velocity and path, and the differential cross section σ (θ, φ, ε)d
will give the probability of scattering into the solid angle d. In the quantum
case, an incoming free particle will be represented as a plane wave described
by Figure 4.4b and Equation 4.4. Then, as a result of spherical scattering,
the outgoing wave will be given by the sum of the plane and the spherical
waves.
4.2.1 Conservation Laws in Collisions

First we assume that the collision event occurs in a force-free space, in which
the momentum and energy of the two particles are conserved. This is true, for
example, for an electron collision with a neutral molecule, which is the most
frequent type of collision occurring in collisional plasmas, where the interac-
tion time between the two particles is on the order of 10−16 s, and the effect
of the external force due to an imposed electric field is negligible compared
with the huge internal force between the two particles. The conservation of

momentum is then simply given by
mv + MV = mv + MV, (4.14)
where (m, M) are, respectively, the masses of the charged and neutral particle,
and (v , V ) are the velocities before and (v, V) the velocities after the colli-
sion. A similar conservation law is written for the energy. However, in this
case internal excitation can occur and the inelastic process should also be
accounted for. In the energy conservation law, it is convenient to change the
frame of reference to a center-of-mass (CM) frame, and the kinetic energy is
represented as
1 2 1 1 1
εkin = mv + MV 2 = (m + M)vg2 + µr vr2 , (4.15)
2 2 2 2
where the CM velocity vg and the relative velocity vr are given as
mv + MV
vg = (4.16)
m+M
and
vr = v − V. (4.17)
The CM velocity remains constant before and after the collision under the
conservation of momentum expressed in Equation 4.14. As a result, the trans-
formation from the standard laboratory (LAB) system to the CM system effec-
tively allows us to describe a two-particle process as a single-particle process.
The reduced mass of the effective particle is therefore
mM
µr = . (4.18)
m+M
Using these transformations to the CM frame, we can write the energy
conservation law as
1 1
µr vr = µr vr2 + ε R ,
2
(4.19)
2 2
where ε R is the energy loss (reaction energy) due to the excitation of internal
energy levels of one or both particles. We define three kinds of collisions, each
defined by their value and sign of the reaction energy.
i. ε R > 0: Collisions of the first kind, in which we have loss of the total
kinetic energy to inelastic processes;
ii. ε R < 0: Collisions of the second kind, in which we have gain of the
total kinetic energy from the internal excitation energy; and
iii. ε R = 0: Elastic collisions, in which the total kinetic energy is
conserved.

PROBLEM 4.2.1
Explain that the CM energy is conserved in the case of binary collision, and that in
the case of inelastic scattering the energy to excite the molecule is provided from the
relative kinetic energy.
4.2.2 Definition of Collision Cross Sections

The concept of a collision cross section can be best understood when we con-
sider a collision between two particles represented as hard spheres with radii
of a 1 and a 2 , approaching with a relative velocity vr along the z-axis. A hard
sphere is an object for which the interaction potential is zero for distance r > a 0
and infinitely repulsive for r ≤ a 0 . We now introduce the impact parameter b
to the binary collision with b representing the perpendicular distance from the
z-axis along the initial relative velocity to the center of the target. We then ob-
serve (Figure 4.5) that only those particles with an impact parameter b less than
a 1 + a 2 will collide, whereas others will continue their motion without chang-
ing their direction. Thus the circle with radius a 0 = a 1 + a 2 is the effective size
of the target for projectiles represented as points. The area of this circle πa 02 =
π(a 1 + a 2 )2 is the effective collision cross section for the two hard spheres.
PROBLEM 4.2.2
Derive the energy transfer as a function of scattering angle in an elastic collision of
two hard spheres (r, m, v ) and (R, M, 0), both defined by their radius, mass, and
velocity before collision.
The concept of an “effective area” associated with target particles for any
pair of colliding particles (two-body collision) is useful even if the hard sphere
approximation is not valid. There are, however, some conditions that must
be met for it to be applicable.
i. The collisions between incident particles should be negligible; that
is, the density of the incident particles should be sufficiently low;
a2
v a1 a0
z
V
FIGURE 4.5
Binary encounter of solid spheres with radii of a 1 and a 2 , and the concept of the collision
cross section.

ii. The density of the target particles should be low in order to avoid
collisions of the incident particles with more than one target at the
same time. That is, the de Broglie wavelength of the incident particle
should be shorter than the mean distance between target molecules.
In defining the collision cross section we return to the differential cross section
that was mentioned earlier. When a certain number of projectiles n with
energy ε are directed toward a target consisting of N particles, the number of
particles dn scattered into the surface element dS on the spherical surface at
a distance r from the target center will be proportional to the density of the
projectiles and the targets and also proportional to the area of the surface dS
and r −2 , that is, the solid angle in the collision experiment:
dn ∝ nN ds/r 2 ,
where ds/r 2 = d(= sin θ dθ dφ).

It is known that dn depends on the incident energy of the projectile and
the quantum property of the target molecule. If we define the constant of
proportionality as σ (θ, φ; ε), then the relation above becomes
dn = σ (θ, φ; ε)nNd. (4.20)
In Equation 4.20, σ (θ, φ; ε)d is the probability that one projectile is scat-
tered into a small solid angle d at scattering angle (θ, φ). σ (θ, φ; ε) has the
dimensions of the area and so is termed the “differential cross section.”
The differential cross section may be expanded into the Legendre polyno-
mials Pn (cos θ ) under the azimuthal symmetry
∞
1
σ (θ ; ε) = (2n + 1)Qn (ε)Pn (cos θ), (4.21)
4π 0
where Qn (ε) is the nth integral cross section. Multiplying the equation by
Pn (cos θ ) and using the property of orthogonality of the Legendre polynomi-
als (see Section 5.6.1), we obtain the following definition of the integral cross
section Qn (ε):
π
Qn (ε) = 2π σ (θ; ε)Pn (cos θ) sin θ dθ. (4.22)
0
The coefficients Qn (ε) and the differential cross section σ (θ ; ε) are defined
for each of the many scattering processes that can occur. They are interpreted
as the scattering probability averaged over all angles and weighted by some
Legendre polynomial. For example, Q0 (ε) is the total cross section for the
process (i.e., the probability of scattering at any angle) and corresponds to
the area of the target as seen by a point projectile. The most commonly used
angular averages of the differential cross section are the following:

• Total cross section

π
Q0 (ε) = σ (θ, φ; ε)d = 2π σ (θ; ε) sin θ dθ; (4.23)
0
• Momentum transfer cross section

Qm (ε)(= Q0 − Q1 ) = (1 − cos θ)σ (θ; ε)d
π
= 2π (1 − cos θ)σ (θ ; ε) sin θ dθ ; and (4.24)
0
• Viscosity cross section

2
Qv (ε) = (Q0 − Q2 ) = (1 − cos2 θ)σ (θ; ε)d
3
π
= 2π (1 − cos2 θ)σ (θ; ε) sin θ dθ. (4.25)
0
These integral forms of differential cross section play a special role in transport
theory. For inelastic collisions with threshold energy ε j , the definitions are the
same except that the momentum transfer cross section is defined as
π
εj
Qm (ε) = 2π 1− 1− cos θ σ (θ; ε) sin θdθ, (4.26)
0 ε
> ε j when the scattering is nearly isotropic, Qm (ε) will be very
Thus, at ε ∼
small.
Exercise 4.2.1
Explain the physical meaning of the momentum transfer cross section for
elastic scattering.
Assuming that the initial momentum is p and that particles are scattered at
an angle θ with no loss of the magnitude of the velocity, the difference of
momentum along the axis of the initial velocity is
p = p − p cos θ = p(1 − cos θ).
In other words, the fractional change of the momentum is
p
= (1 − cos θ). (4.27)
p
Therefore, the cross section Qm in Equation 4.24 is in some way a representa-
tion of the momentum transfer to the gas molecules. Qm for elastic collision
has a finite magnitude, and the elastic energy loss (2m/M)ε of the electron
with energy ε is negligibly small.
PROBLEM 4.2.3
Discuss the physical meaning of the viscosity cross section.

Exercise 4.2.2
Gas molecules with density N and temperature Tg are present in a reactor.
Estimate the limitations of the two-body approximation for the collisions
between molecules.
This criterion may be related through the relationship between the mean dis-
tance between molecules d and the de Broglie wavelength λde of the molecule,
d λde .
The mean distance between the gas molecules may be obtained from the gas
number density N as d = N−1/3 . On the other hand, the wavelength, which is
used to approximate the range over which quantum effects are appreciable,
will be calculated for the most probable speed of molecules, 2kTg /M. Hence
the de Broglie wavelength is given by
h h h
λde = ≈ = .
p Mṽ 2MkTg
The condition therefore becomes
λde h N1/3
≈ 1. (4.28)
d 2MkTg
PROBLEM 4.2.4
Calculate Q0 , Qm , and Qv for the following models of differential scattering cross
sections:
• Isotropic (σ (θ ) is independent of angle);

• Forward (σ (θ ) is a delta function at zero angle);
• Backward (σ (θ ) is a delta function at π); and
• Right angles (σ (θ ) is a delta function at π/2).
Discuss the relative magnitudes Qm /Q0 and Qv /Q0 in these cases.
The unit usually used in presenting a collision cross section between an

electron and a molecule is 10−16 cm2 . This unit originates from the Angstrom
(Å) an old unit of length equal to 10−8 cm (0.1 nm). Other units are often
applied as well. For example, it is convenient to represent the cross section by
using the Bohr radius (i.e., the first Bohr orbit of H), a 0 = 0.529 × 10−8 cm as
[a 02 ] = 0.280 × 10−16 cm2 4.29a )
The effective cross section of a Bohr radius is then given as
[πa 02 ] = 0.880 × 10−16 cm2 4.29b)
In the case of photon scattering, it is common to use the “barn”
[barn] = 10−24 cm2 4.29c)

f (ε)
1 exp
λ ( λz )
0 z
ε collision
electron
beam
FIGURE 4.6
Distribution of the mean free path f (z; ε) for a beam of particles with energy ε incident on a gas
target with random distribution.
Finally, we note that the total collision probability Pc is often used to represent
the number of collisions in a gas at some pressure p and at temperature Tg [K].
The total collision probability is equal to Pc = ( p/kTg )Q. Here, the standard
value that is used is p = 1Torr at room temperature, where we have
Pc = 9.66 × 1018 Q[cm2 ]/Tg [K]cm−1 Torr−1 (4.29d)
for the standard collision probability at 1 Torr = 133Pa .
4.2.3 The Distribution of Free Paths

In order to obtain the distribution of the free paths of a particle between colli-
sions, we consider a beam of particles with density n0 and energy ε entering
gas at z = 0 (see Figure 4.6). It can be shown that the distribution function of
the free path is given by

1 z
f (z; ε) = exp − , (4.30)
< λ(ε) > < λ(ε) >
where < λ(ε) > is the mean free path of a particle with energy ε.
Exercise 4.2.3
Derive the distribution function of the free path (Equation 4.30).
When we define F (z) as the integral distribution of collisions on the segment
(0, z), the number of collisions in the small segment (z, z + z) is proportional

to the number of particles remaining in the beam (1 − F (z)) and to dz, so we

have

F (z + dz) − F (z) = c{1 − F (z)}dz,
where c is a constant. Next we obtain the equation
−1 F (z + dz) − F (z)
= −c,
1 − F (z) dz
which has a solution
ln{1 − F (z)} = −cz or F (z) = 1 − exp(−cz).
From the integral distribution of collisions F(z) we determine the distribution

of the free path between collisions:

dF (z)
f (z) = = c exp(−cz).
dz
Considering the definition of the mean free path, we have

∞
zf (z)dz 1
< λ > = 0∞ = ,
0 f (z)dz c
and thus the formula (Equation 4.30) for the distribution function of the
free path is obtained. The distribution f (z) can be used to determine the
probability of collision (or of the length of the free path) of a single particle as
well. One should bear in mind that the mean free path depends on the energy
of the incident particles, that is, < λ(ε) >. Thus if we have a distribution of
incident energies we perform additional averaging and obtain the mean free
path for the whole ensemble.
4.2.4 Representation of Collisions in Laboratory and

CM Reference Frames
In Section 4.2.1, we have shown that the motion of two particles of masses
m and M and velocities v and V can be treated as a single-particle motion
with relative mass µr and relative velocity vr by considering the total energy
conservation
1 2 1 1 1
εkin = mv + MV 2 = (m + M)vg2 + µr vr2 , (4.31)
2 2 2 2
where vg is the CM velocity. A convenient way to describe a binary collision
is to make a transfer from the laboratory frame of reference to the CM frame
of reference of the relative mass µr and the velocity vr (see Figure 4.7).

v'
v
π−ω c
2
ωc
vg
ωL
V
V'
FIGURE 4.7
Velocities of two particles before and after collision and the corresponding angles in the CM
frame.
Exercise 4.2.4
Obtain the relationship between differential cross sections for the laboratory
and the CM reference frames in the elastic scattering.
Consider two particles, (m) and (M), moving toward one another with
velocities (v , V ) before and (v, V) after the collision in the LAB frame. The
magnitude of the relative velocity vr is conserved before and after the elas-
tic scattering in the CM frame, (| v − V |=| v − V |=| vr |). We may thus
represent the motion of particles before and after the scattering by two straight
lines of equal lengths (|vr |) intersecting at the point vg (CM velocity) (see
Figure 4.7). In the CM frame, the velocities of the two particles before and
after the collision are located on spheres with radii of v = vr M/(m + M) and
V = vr m/(m + M), respectively.
We can identify one equilateral triangle with the CM frame scattering
angle ωc , that is, the triangle [vg , v , v]. Thus we write
v v Mvr /(m + M)
= ,
sin ωc sin[(π − ωc )/2]
where the length between points [v, v ] is denoted v v. If, however, we take
the triangle [V , v , v], the following relation is obtained by using the LAB
frame scattering angle of particle m, ω L :
v v vr
= .
sin ω L sin[π − ω L − (π − ωc )/2]

Combining the results from the two triangles we obtain
sin ωc
tan ω L = (4.32a )
(m/M) + cos ωc
and
(m/M) + cos ωc
cos ω L =
1/2 . (4.32b)
1 + 2(m/M) cos ωc + (m/M)2
Because the same number of particles is scattered through the solid angle
in both systems, we have
σ L (ω L ; ε)d L ≡ σc (ωc ; ε)dc ,
or
d cos ωc
σ L (ω L ; ε) = σc (ωc ; ε)
d cos ω L

3/2
1 + 2(m/M) cos ωc + (m/M)2
= σc (ωc ; ε) . (4.33)
1 + (m/M) cos ωc
It is possible to show that the energy transfer from the projectile to the target
(i.e., the energy loss to the first particle) is then equal to
2
εM MV 2 /2 M 2mv

= = cos ω L
εm mv 2 /2 mv 2 m + M
4mM
= cos2 ω L . (4.34)
(m + M)2
PROBLEM 4.2.5
Determine the scattering angles ω L and ωC M when (a) m = M and (b) m M.
PROBLEM 4.2.6
Show that the energy loss in elastic collisions is given by Equation 4.34, and that for
the electron-molecule scattering it is simplified after averaging to (2m/M)ε.
PROBLEM 4.2.7
Prove that for a head-on (ω L = 0) collision of a heavy particle with a light target
m M, the maximum velocity of the target will be V = 2v .
It is also of interest to determine the velocities after the collision in the case
of inelastic processes characterized by the inelastic energy transfer ε. After
inelastic scattering, the velocity of the target in the LAB frame is equal to

1 m
v= mv cos θ L ± [mMv 2 cos 2 θ L − 2(m + M)ε] (4.35)
m+M M

and the velocity of the projectile is

1 M
V= mv cos ω L ± v (M2 − m2 sin2 ω L ) − 2(m + M) ε ,
2
m+M m
(4.36)
where θ L is the scattering angle of the target (M) in the LAB frame. From
Equations 4.35 and 4.36 it is possible to obtain the minimum velocity of the
incident particle necessary to achieve inelastic excitation with a loss ε:
1 2 (m + M)M
mv = ε, (4.37)
2 M2 − m2 sin2 ω L
which for head-on collisions reduces to
m+M
εm = ε.
M

For particles of the same mass (m = M) one needs at least εm = 2ε
to achieve an inelastic process, whereas for light projectiles (m M) the

minimum energy is εm = ε.
4.3 Classical Collision Theory

In the classical theory of collisions it is possible to determine the differential
cross section based on the interaction potential and conservation laws in a
binary collision. It is always useful to reduce the problem of scattering of two
particles to the scattering of one particle with a reduced mass. It is assumed
that the target molecule has a centrifugal potential that permits spherical
symmetry in the scattering event. With the exception of the head-on collision,
the incoming particle will pass the target at some distance. The point of the
closest approach is known as the impact parameter (b). Even in the case
when there is an external force field to deflect the particle, we may define the
instantaneous impact parameter for any point as a function of the spherical
coordinates that describe the motion.
It is evident that impact parameter b will define the degree of inter-
action and final trajectory for a given interaction potential. It is also clear
that the outcome of scattering is described by the angle of scatering θ . In
Figure 4.8 we show several classical trajectories for a projectile scattered
from a target molecule. There, we define the scattering angle θ and the corre-
sponding annular solid angle. It is also obvious that, in general, the smaller
the impact parameter, the greater the scattering angle, because the particle
will approach close to the target and thus be subjected to a stronger interac-
tion. There is a unique relation between the scattering angle and the impact
parameter b = b(θ ) (note this may not be true in some special circumstances,
e.g., when orbiting occurs).

v0
v0 dθ
db θ
b
z
0
2πbdb
dΩ = 2πsinθdθ
FIGURE 4.8
Definition of the scattering angles and the impact parameter in classical scattering
phenomenology.
The number of particles scattered into the angle θ is proportional to the

solid angle d and the differential cross section σ (θ). The scattered particles
with probability σ (θ )d correspond to the particles incident on the area of
the annular surface 2πbdb. Therefore,
σ (θ ) = 2π bdb/d

db(θ ) dθ

= b(θ )
dθ sin θ dθ

1 db(θ )
=
b(θ ) . (4.38)
sin θ dθ
Our goal is thus to determine the dependence b(θ ) for any given potential
interaction.
4.3.1 Scattering in Classical Mechanics

We now proceed to calculate the function b(θ ) for scattering on a target placed
at the scattering center r = 0. The projectile has velocity v and reduced
mass µ. The interaction potential is V(r ), which is assumed to be spheri-
cally symmetric (i.e., a centrifugal potential). Additionally, the scattering is
assumed to be purely elastic. Because we must satisfy the basic laws of con-
servation, the conservation of angular momentum L yields
L = µv0 b = µ(r dξ/dt)r = µv0 b . (4.39)

v0
1 (π − θ)
2 b
1 (π − θ)
2
v0′ rmin θ
r
b′ ξ
FIGURE 4.9
The geometry of classical scattering on a target with interaction potential V(r ).
The conservation of the total energy E yields
E = µv02 /2 = µ{(dr/dt)2 + r 2 (dξ/dt)2 }/2 + V(r )

= µv0 /2,
2
(4.40)
where ξ is defined as in Figure 4.9. At positions sufficiently far from the

scattering center, Equations 4.39 and 4.40 give the relations of the velocities
and the impact parameters before and after collision:
v0 = v0 , b = b. (4.41)
In Figure 4.9 we define the coordinates for the scattering and basically
the incoming particle with reduced mass µ follows the angle ξ . From
Equation 4.39 it is shown that
dξ/dt = L/µr 2 , dr/dt = (dξ/dt)(dr/dξ ) = (L/µr 2 )(dr/dξ ),
and from Equation 4.40 it follows that
E = µ{(L/µr 2 )2 (dr/dξ ) + r 2 (L/µr 2 )2 }/2 + V(r )

= {L 2 /2µr 4 }{(dr/dξ )2 + r 2 } + V(r ).
The equation describing the particle trajectory is therefore
(dr/dξ )2 = {2µr 4 /L 2 }{E − V(r )} − r 2 . (4.42)
It is also worth noting that
L2 µv 2 b 2
+ V(r ) = + V(r ) = Veff (r ), (4.43)
2µr 2 2r 2
and Veff (r ) is the effective potential in the radial direction. The interaction
potential usually has a minimum; as a result, the projectile is for a while
accelerated to the target, but it will not be bound as it has a positive energy.
The addition of the centrifugal term to the interaction potential leads to an
increase in the potential in the region of the minimum as the centrifugal term

increases. The potential may even become repulsive. It may also be possible
that a maximum is formed at some distance rc with a shallow minimum for
smaller distances. Near the maximum the radial motion will be very small,
but it will take several revolutions before the particle can leave after scatter-
ing as the angular momentum is large. This effect is known as “orbiting.”
Orbiting allows longer interaction times and therefore a greater probability
for processes that involve transitions with finite lifetimes.
To calculate the trajectory from Equation 4.43, we must obtain an equation
that will describe the trajectory through the angle ξ ,
dξ 1
= ± 1/2 . (4.44)
dr 2µr 4
L2
{E − V(r )} − r 2
When dξ/dr < 0, the incoming particle approaches the target and the
interaction force increases. When dξ/dr > 0, the projectile leaves the target.
Because of the symmetry of the scattering (see Figure 4.9), we may perform
the integration in two parts. First we perform it up to the point of the closest
approach rmin , from ξ = 0 (for r → ∞) up to ξ = (π − θ)/2. We obtain
(π −θ)/2 rmin
dξ = −[{2µr 4 /L 2 }{E − V(r )} − r 2 ]−1/2 dr
0
∞∞
(π − θ )/2 = [{2µr 4 /L 2 }{E − V(r )} − r 2 ]1/2 dr. (4.45)
rmin
The limit rmin (the point of the closest approach) may be obtained from
dr/dξ = 0, which in combination with Equation 4.45 leads to {2µrmin 4
/L 2 }
{E − V(rmin )} − rmin = 0. Introducing u = 1/r , we derive the following result
2
from Equation 4.45:

umax
(π − θ )/2 = [{2µ/L 2 }{E − V(1/u) − u2 }]−1/2 du;
0
(2µ/L 2 ){E − V(1/umax )} − u2max = 0. (4.46)
An analogous equation may be obtained using the impact parameter b:

umax
(π − θ )/2 = [(1/b 2 ){1 − V(1/u)/E} − u2 ]−1/2 du;
0

1 − V(1/umax )/E − b 2 u2max = 0. (4.47)
Exercise 4.3.1
Determine the dependence of the impact parameter b(θ ) for classical
hard
sphere elastic scattering, that is, for the potential V(r ) = ∞(r ≤ r0 )
0(r > r0 )
. Also deter-
mine the effective cross section for the scattering.
In Figure 4.10 we show that the hard sphere potential is a very good
approximation for many classical potentials. The point of minimum approach

∞
V(r)
hard sphere
realistic potential
r
0 r0
π−θ
2 π−θ
2
π−θ
2
b (θ)
0 r0
FIGURE 4.10
Comparison between the hard sphere interaction potential and a realistic potential (dashed line)
and geometry of the hard sphere scattering for a given impact parameter.
is rmin = r0 = 1/umax and may be used in Equation 4.47 to obtain

1/r0
(π − θ)/2 = [(1/b 2 ) − u2 ]−1/2 du
0
= sin−1 [u/(1/b)]0 = sin−1 [b/r0 ]; (b ≤ r0 )

1/r0
= no collision; (b > r0 ).
We obtain the solution, sin[(π − θ)/2] = b/r0 ; that is,
θ (b) = 2 cos−1 (b/r0 ) (b < r0 )

= 0 (b > r0 ).
Substituting the above relation into Equation 4.38 we obtain the expression
for the differential cross section:
σ (θ ) = (1/ sin θ)b(θ )|db(θ )/dθ |

= (1/ sin θ )r0 cos(θ/2)| − (r0 /2) sin(θ/2)|
= r02 sin θ/(4 sin θ) = r02 /4, (4.48)

and from the definition of the total cross section Equation 4.23 we obtain

Q0 = σ (θ)d = (r02 /4) sin θ dθ dφ
= (r02 /4)4π = πr02 . (4.49)
PROBLEM 4.3.1
Derive the momentum transfer cross section Qm for hard sphere elastic scattering,
and discuss the relationship between momentum transfer and the total cross section.
Exercise 4.3.2
Derive the impact parameter dependence of the scattering angle θ(b) and of
the differential cross section σ (θ) for a light H+ (with charge e and mass m)
and massive ion (with charge Ze and mass M) in a fully ionized plasma (i.e.,
collisionless plasma) by considering the classical Rutherford scattering in a
Coulomb field.
Classical Rutherford scattering and its physical quantities are shown in
Figure 4.11. We may regard Coulomb scattering as resulting in small angle
scattering or small changes of the momentum under the long-range interac-
tion. Thus it is reasonable to assume that
sin θ ∼ θ = pT / p.
Here, the Coulomb force will induce a velocity perpendicular to the original
velocity and the resulting momentum is denoted by pT . We may calculate the
transverse momentum by integrating the effect of the force FT in the region
of interaction
∞ ∞
pT = FT dt = (Ze 2 /4π ε0r 2 ) cos βdt
−∞ −∞
∞
= {Ze 2 /4π ε0 (b 2 + v 2 t 2 )}{b/(b 2 + v 2 t 2 )1/2 }dt
−∞
= 2Ze 2 /4πε0 bv.
A
e vt
p = mv e
θ
b pT
β
r
Ze
FIGURE 4.11
Classical Rutherford scattering between a light H+ (with charge e and mass m) and a massive
ion (with charge Ze and mass M) in a Coulomb field.

Therefore,
θ (b) = pT / p
2Ze 2 1 2Ze 2
= · = , (4.50)
4π ε0 bv mv 4π ε0 mv 2 b
and the differential cross section can be obtained from Equation 4.38 as
σ (θ ) = (1/ sin θ)b(θ )|db(θ )/dθ |

1 2Ze 2 2Ze 2 −1
∼
θ 4π ε0 mv 2 θ 4π ε0 mv 2 θ 2
2
Ze 2 4
∼ . (4.51)
4π ε0 mv 2 θ4
A more detailed analysis gives θ(b) and σ (θ) for the Rutherford scattering:

−1 Ze 2
θ (b) = 2 tan ;
4π ε0 mv 2 b
2
1 Ze 2 1
σ (θ ) = . (4.52)
4 4π ε0 mv 2 sin4 (θ/2)
It appears that it is not possible to integrate the differential cross section
of the Rutherford scattering into total or momentum transfer cross sections
as there is a singularity for θ ∼ 0. However, we should remember that the
long-range effect of the Coulomb force will cancel out in ionized gas due to
the effect of other charged particles. The effect of the Coulomb interaction
is considered only within the so-called Debye radius λ D ; that is, we need to
consider the interaction only at b < λ D with the minimum scattering angle
θmin at b = λ D . In other words, there is a limit on impact parameter b ≤ b max .
The momentum transfer cross section is therefore
π 2
1 Ze 2 1
Qm = (1 − cos θ)2π sin θdθ
θmin 4 4πε0 mv 2
sin (θ/2)
4
2
Ze 2 2
= 2π ln .
4πε0 mv 2 1 − cos θmin
Here, 1 − cos θ = 2/{1 + [b/(Ze 2 /4π ε0 mv 2 )]2 } in Figure 4.11, and we may
determine the cross section Qm by using
1 − cos θmin = 2[(Ze 2 /4π ε0 mv 2 )/b max ]2 , where b max = λ D .
We thus obtain
Qm = 4π(Ze 2 /4π ε0 mv 2 )2 ln[4π ε0 mv 2 λ D /Ze 2 ]

Z2 e 4
= ln . (4.53)
4πε02 (mv 2 )2

Here ln is defined as

4π ε0 mv 2 λ D
ln = ln (4.54)
Ze 2
and is known as the Coulomb logarithm.
PROBLEM 4.3.2
Show that the Coulomb logarithm corresponds to the ratio b max /b min .
4.3.2 Conditions for the Applicability of the Classical Scattering Theory

Two conditions must be met to allow us to use the classical scattering theory:
i. The trajectory of the particle must be clearly defined. In other words,

the de Broglie wavelength λde must be much shorter than the effec-
tive range of the interaction potential d0 , or
λde = (h/µvr ) d0 . (4.55)
ii. The change of the momentum p must be much greater than the
uncertainty pd h. For an effective range of the potential V0 , the
conservation of energy gives vr p ∼ eV0 . The condition is thus
eV0 d/vr h. (4.56)
4.4 Quantum Theory of Scattering

Collisions are considered in quantum theory under three major conditions:
(a) The incident particle is described by a plane wave.

(b) The total energy of the incident particle is arbitrary positive.
(c) The scattered particle is analyzed at a point far from the scattering
center, where the influence of the interaction potential is negligible.
We discuss the elastic scattering of a particle with energy E. A schematic

diagram of the scattering is shown in Figure 4.12, where the region is divided
into three points (a), (b), and (c).
(a) Incident wave. In this subsection, the incoming particle is represented

as the wave function (r, t), which can be written as
(r, t) = exp[i(k · r − ωt)], (4.57)
where p = h̄k and E = h̄ω. The description in Equation 4.57 is valid

only at a long distance where the interaction potential is negligible.

f (θ, φ) ikz
r e
(spherical wave)
π−θ
V(r)
Aeikz
(plane wave)
FIGURE 4.12
The planar wave of the incoming particle and the spherical wave of the scattered particle in
quantum theory of elastic scattering.
When the velocity of the incoming particle is along the z-direction

in the steady state, then
(z) = exp(ikz). (4.58)
(b) Under scattering potential. Here, the collision is characterized by the

Schrödinger equation having a continuous positive energy E of
the incoming particle and the interaction potential V(r ). Hence the
wave equation reads as
−(h̄ 2 /2µ)∇ 2 (r) + V(r)(r) = E(r). (4.59)
(c) After scattering. From the condition that scattered particles eventu-
ally escape to far distances unaffected by V(r ) and distribute them-
selves spherically around the scattering center (target), we may
conclude that the scattered particle can be described by a spheri-
cal wave. Thus, the total wave function is the sum of the scattered
spherical wave and the unaffected plane wave:
f (k; θ, φ)
(r) = exp(ikz) + exp(ikr ). (4.60)
r
The spherical term is normalized by f (k; θ, φ), which is known as the scatter-
ing amplitude. For example, if there is no interaction, the scattering amplitude
is zero and the planar wave is unaffected. The number of particles n scattered
into a small area dS normalized to the total number of the incident particles
n represents the probability of the scattering in a given direction (θ, φ) and is
therefore equal to
2
f (k; θ, φ) 1
n/n = exp(ikr ) dS ·
r exp(ikz)2
dS
= | f (k; θ, φ)|2 .
r2

Here, the solid angle is d = dS/r 2 , and by definition of the differential

cross section σ , we have
n/n = σ (k; θ, φ)d.
We can then obtain the relation between the differential cross section and the
scattering amplitude in quantum theory as
σ (k; θ, φ) = | f (k; θ, φ)|2 . (4.61)
4.4.1 Differential Scattering Cross Section σ(θ)

We can solve Equation 4.59 by using Equation 4.60, making a solution in the
CM system for a stationary target and projectile with a reduced mass µ. The
wave function can be separated into two terms, the radial (r ) and angular
(θ, φ) functions; that is,
(r ) = R(r )Y(θ, φ).
In the case of azimuthal symmetry, it is useful to use Legendre

polynomials (and take advantage of their orthogonality):
(r) = R(r )Pl (cos θ). (4.62)
This makes it possible to derive the differential equation for the radial part of
the wave function R(r ):

1 d 2 dR(r ) 2µE 2µ h̄ 2 l(l + 1)
r + − 2 V(r ) + R(r ) = 0. (4.63)
r 2 dr dr h̄ 2 h̄ 2µ r 2
Here, we have used quantization of the angular momentum. We define the

effective radial potential as
h̄ 2 l(l + 1)
Veffl (r ) = V(r ) + . (4.64)
2µ r 2
As in the classical case, the effective radial potential is the result of
the combined effect of the interaction potential and the centrifugal term. When
we make transformation R(r ) = u(r )/r , the differential equation can be writ-
ten as
h̄ 2 d 2 u(r )
− + Veffl (r )u(r ) = Eu(r ). (4.65)
2µ dr2
We now consider the asymptotic behavior of the solutions. From Veffl → 0
when r → ∞ it follows that
u(r ) → Asin(kr ) + B cos(kr )
→ Asin(kr + ηl − lπ/2); at r → ∞.

The radial component of the wave function has the effect of scattering
through a phase-shift (ηl − lπ/2) at a long distance from the scattering cen-
ter (target). It should be noted that without scattering u(r ) is exactly equal to
the incoming plane wave, expressed below as Asin(kr −lπ/2). Consequently,
the solution for R(r ) is
R(r ) = (1/kr ) sin(kr − lπ/2 + ηl ) at r → ∞, (4.66)
and by combining Equations 4.59, 4.62, and 4.66, we obtain the general
solution for the wave function as

(r) = Al R(r )Pl (cos θ )
l

∼ Al (1/kr ) sin(kr − lπ/2 + ηl )Pl (cos θ) (at r → ∞)
l

∼ Al (1/2ikr ){exp[i(kr − lπ/2 + ηl )]
l
− exp[−i(kr − lπ/2 + ηl )]}Pl (cos θ) (at r → ∞).
We must now match the asymptotic forms of the above general solution
and Equation 4.60. For that purpose we must expand the incoming plane
wave, exp(ikz) in Equation 4.60 for r → ∞ as
exp(ikz) = exp(ikr cos θ )

(2l + 1) 1
= Pl (cos θ ) exp(ikrt)Pl (t)dt t = cos θ
l
2 −1
(Rayleigh formula)
2l + 1 l
= Pl (cos θ ) 2i jl (kr ).
l
2
Recall some of the mathematical relations described above, including the

orthogonal functions

(2l + 1) 1
f (x) = a l Pl (x) and a l = f (t)Pl (t)dt, (4.67)
l
2 −1
the integral expression of the spherical Bessel function

1
jl (x) = (1/2i l ) exp(i xt)Pl (t)dt, (4.68)
−1
and the asymptotic form of the spherical Bessel function in the limit x → ∞:
jl (x) = (π/2xl )1/2 J l+1/2 (x)

∼ (1/x) cos (x − (2l + 1)π/2) (at x → ∞).

Thus, we may write the asymptotic expansion of the planar wave in the limit
r → ∞ as
exp(ikz)r →∞

∼ (2l + 1)i l (1/kr ) sin(kr − lπ/2)Pl (cos θ)
l
(2l + 1)i l
lπ

lπ

∼ exp i kr − − exp −i kr − Pl (cos θ)
l
2ikr 2 2
(4.69)
As a result, the scattering wave is obtained from
(r) − exp(ikz)
2l + 1 l
= Al R(r )Pl (cos θ ) − Pl (cos θ) 2i jl (kr )
l=0 l
2

1 lπ lπ
∼ Al exp i kr − + ηl − exp −i kr − + ηl Pl (cos θ)
l=0
2ikr 2 2
(2l + 1)i l
lπ

lπ

− exp i kr − − exp −i kr − Pl (cos θ)
l
2ikr 2 2

1 lπ
∼ Pl (cos θ ) exp i kr − Al exp(iηl ) − (2l + 1)i l
l=0
2ikr 2

1 lπ
− exp −i kr − [Al exp(−iηl ) − (2l + 1)i l ] (at r → ∞). (4.70)
2ikr 2
Equation 4.70 should be matched to the outgoing spherical wave

exp(ikr )/r . Therefore, the second term in Equation 4.70 must satisfy
[Al exp(−iηl ) − (2l + 1)i l ] = 0, i.e., Al = (2l + 1)i l exp(iηl ).
The asymptotic form of the scattered spherical wave is
f (k; θ )
exp(ikr )
r
1 lπ
∼ exp i kr − [(2l + 1)i l exp(2iηl ) − (2l + 1)i l ]Pl (cos θ)
l=0
2ikr 2
1
ilπ l exp(ikr )
∼ exp − i (2l + 1){exp(2iηl ) − 1}Pl (cos θ) , (4.71)
l=0
2ik 2 r
and from the expression for f (k; θ), it is possible to obtain the differential
cross section σ (θ ):
1
f (k; θ ) = (2l + 1){exp(2iηl ) − 1}Pl (cos θ) (4.72)
2ik l=0

u(r)
V(r) = 0
V(r) > 0
r
kr0
ηl
FIGURE 4.13
The effect of phase-shift on the wave function.
σ (θ ) = | f (k; θ )|2
2
1 2l + 1

= 2 {exp(2iηl ) − 1}Pl (cos θ) . (4.73)
k l=0
2
Hence the only quantities that describe the effect of the scattering are the
phase-shifts ηl . The use of angular momentum quantum numbers l is analo-
gous to the use of an impact parameter in classical scattering. The scattering
cross section is obtained by summing up the different partial waves in respect
to l = 0, 1, 2, . . . , and the procedure shown here is known as the partial wave
method. This technique was developed for light scattering by Rayleigh and
applied to particle scattering by Faxen and Holtsmark.
PROBLEM 4.4.1
Show that the total cross section Q0 of the elastic scattering is given in partial wave
expansion as

Q0 (ε) = | f (k; θ)|d (4.74)
4π
= (2l + 1) sin2 ηl .
k 2 l=0
Exercise 4.4.1
Explain the phase difference ηl in Equation 4.66.
In Figure 4.13, we show two functions u(r ), one without any interaction with
a target (V(r ) = 0; solid line) and one with an interaction (V(r ) > 0; dashed
line). When the potential is repulsive, the outgoing wave function is pushed
to the outside as compared with the wave at V(r ) = 0, and then ηl > 0. On
the other hand, if it is attractive, then ηl < 0.
Exercise 4.4.2
In Figure 4.14 we show the spherical well potential:
V(r ) = −V0 r ≤ r0
= 0 r > r0 .

V(r)
0 r0
r
-V0
FIGURE 4.14
The spherical well potential (spherical attractive potential).
Calculate the total cross section and the phase-shifts for elastic scattering of
a low-energy particle.
From Equation 4.65, we have
d 2 u(r )
2µ l(l + 1)u(r )
22
+
{E − V(r )}u(r ) − = 0,
dr h̄ r2
and for the potential it follows that
<
{d 2 /dr2 + K 2 − l(l + 1)/r 2 }ul,k (r ) = 0; K 2 = (2µ/h̄ 2 )(E + V0 ); (r ≤ r0 ),
2 >
{d /dr + k − l(l + 1)/r
2 2 2
}ul,k (r ) = 0; k = (2µ/h̄ )E;
2 2
(r > r0 ).
The solution for R(r ) = u(r )/r must be finite at r = 0, and the u(r ) must be
zero at r = 0. For low-energy scattering that satisfies K 2 − l(l + 1)/r 2 > 0,
the component in the partial wave expansion will mainly be l = 0 (s-wave).
Therefore,
u0,k (r ) = u<
0,k (r ) = Asin K r
= u>
0,k (r ) = B sin(kr + η0 ).
The wave function u(r ) and its derivative satisfying the Schrödinger equation
must be continuous at r = r0 . Then,
Asin K r0 = B sin(kr0 + η0 ),
K Acos K r0 = k B cos(kr0 + η0 ),
and we obtain
(1/K ) tan(K r0 ) = (1/k) tan(kr0 + η0 ),

k
(or) η0 = tan−1 tan(K r0 ) − kr0 .
K
As a result, the scattering cross section for l = 0 (s-wave) is

4π 4π k
Q0 (k) = 2 sin2 η0 = 2 sin2 tan−1 tan K r0 − kr0 . (4.75)
k k K

V(r)
V0
r
0 r0
FIGURE 4.15
The profile of the spherical repulsive potential.
PROBLEM 4.4.2
In Figure 4.15 we show the spherical repulsive potential:
V(r ) = + V0 r ≤ r0
= 0 r > r0 .
Calculate the total cross section and the phase-shifts for elastic scattering of a low-
energy particle. Show that the total cross section is equal to Q0 = 4πr02 in the
limit V0 (r ) → ∞. Considering that the classical hard sphere cross section is equal
to Q0C = πr02 (see Exercise 4.3.1), explain why the quantum result is so different
(Q0q = 4πr02 ).
4.4.2 Modified Effective Range Theory in Electron Scattering

We have already established that the differential cross section in the elas-
tic collision in basic quantum theory is expressed as σ (ε) = | f (θ ; ε)|2 , where
the scattering amplitude f (θ ) in the partial wave method is given by
Equation 4.72
1
f (θ ) = (2l + 1){(exp(2iηl ) − 1}Pl (cos θ)
2ik l=0
in the partial wave method. Here k is the wave number of the incoming
electron with energy ε. It follows that
4π
Qm (ε) = 2 (2l + 1) sin2 (ηl − ηl+1 ) (4.76)
k l
and also that
4π
Q0 (ε) = (2l + 1) sin2 ηl . (4.77)
k2 l
These formulae are sometimes used to analyze the differential cross section
of the electron and perform analytic extrapolations or integrations of the
experimental differential cross section.

A method for making an analytic representation of cross sections for

targets without permanent dipole moments was proposed by O’Malley and
coworkers under the name of the modified effective range theory (MERT).
This method yields analytic expansion of phase-shifts:

4α 2 πα 2
tan η0 = −Ak 1 + k ln(a 0 ) − k + Dk 3 + F k 4 , (4.78)
3a 0 3a 0
πα 2
tan η1 = k − A1 k 3 , (4.79)
15a 0
where α and a 0 are the polarizability and Bohr radius, respectively. For
all higher-order phase-shifts (l > 1), the Born approximation is sufficiently
accurate:
π αk 2
tan ηl = .
(2l + 3)(2l + 1)(2l − 1)
The parameters A (the scattering length), A1 , D, and F should be
regarded as fitting parameters that may be obtained from transport coeffi-
cients of electrons and available cross section data. The value of parameter
A may be determined using the dimension of the molecule at zero energy
as Q0 = Qm = 4π A2 . For example, the best fit for argon is obtained for
A = −1.459a 0 , A1 = 8.69a 03 , D = 68.93a 03 , and F = −97a 04 . MERT was found
to work well for scattering of electrons on atoms at low energies where the
elastic collisions are the only scattering process. In other words, this approx-
imation is valid for electron–atom scattering below 1 eV.
PROBLEM 4.4.3
Calculate Q0 , Qm for argon from 0 eV to 1 eV. Show that the Ramsauer–Townsend
minimum (RTM) occurs when the contribution of both η0 and η1 are close to zero for
the same energy. Show that the RTM occurs at different energies for Q0 and Qm and
explain why.
4.5 Collisions between Electrons and Neutral Atoms/Molecules

Collisions of electrons with gas phase molecules are very frequent and are
the most important in a majority of low-temperature nonequilibrium plas-
mas. Accordingly, these collisions are described in greater detail than other
processes. A specific property of the collision between an electron and a
molecule is the large difference in mass, which is more than four orders of
magnitude. Thus the recoil of the molecule in collisions is small and the en-
ergy loss of electrons in elastic collisions is very small, whereas the energy
gain for molecules is negligible. This is the basic property that allows forma-
tion of nonequilibrium plasmas. On the other hand, the transfer of energy to

inelastic processes is close to 100% efficient. We can classify electron–molecule

collisions by several criteria. The most important is division into elastic and
inelastic processes. Another important distinction is between conservative and
nonconservative processes of the electron, where, for example, ionization and
attachment fall into the latter category.
A theoretical description of the electron–molecule collision is quite com-
plex and an analytic representation of the cross sections is generally im-
possible. Classical and semiclassical treatments are only of very limited use
in specific processes. Quantum theory is therefore almost always necessary
and the Hamiltonian consists of the part describing the isolated molecule,
the kinetic energy of the electron, and the electron–nuclei and the electron–
molecule interactions. Inasmuch as the electrons in the outer orbit in a mole-
cule and the projectile electron are fermions, the Pauli principle dictates that
the corresponding potential is repulsive and nonlocal. There is no classical
analogue for this interaction. This effect is of importance when the electron
is close to the target, that is, in the near region. There it will compete with
numerous electrostatic interactions.
If we move farther from the target the polarization effect becomes the
dominant one. Polarization is induced by the projectile electron, and although
the semiclassical picture is useful for understanding it, one needs quantum
theory for practical calculations. At long distances r , the polarization potential
is approximated by
1 α(r )e 2
Vpol (r ) → − , (4.80)
8π ε0 r 4
where α is the polarizability of the molecule. If there is a permanent dipole
moment or if the target is charged, then the interaction may be dominated
in the far region by electrostatic terms. Molecules with a permanent dipole
moment, although neutral, have a very long-range potential, and thus the
cross sections may be large, especially for low-incident energies. As a result
of these interactions and of the internal degrees of freedom of molecules, there
are numerous electron–molecule collisions that may occur; these are listed in
Table 4.1.
4.5.1 Resonant Scattering

When an electron comes close to a molecule, it may be bound to the target
molecule for a brief time (τ between 10−10 s and 10−15 s) depending on its
initial energy and the structure of the molecule. When the time significantly
exceeds the time required to cross the interaction region, we may talk about
the formation of a compound state or temporary negative ion. The third term
that may be used to describe this system is a resonance. Atoms in excited
states may have positive electron affinity εa so that the energy of the bound
electron is εex − εa . However, this system is unstable as compared with the
ground state, so it may decay into a free electron with an energy close to
εex − εa and the target molecule. In principle, if the state were long-lived it

TABLE 4.1
List of Electron–Molecule Collisions
Process Reaction Scheme Energy Loss ∆ε(Typical)
ε
Elastic e + M −→ e + M ∼ 2(m/M)ε ∼0
ε>εex
Electronic excitation e + M −→ e + M j εex 10 eV
ε>εr
Rotational excitation e + M(r ) −→ e + M(r ) εr kT
ε>εv
Vibrational excitation e + M(v) −→ e + M(v ) εv 0.1 eV
ε>εd
Dissociation e + AX −→ e + A + X εd 10 eV
ε>εi
Ionization e + M −→ e + e + M+ εi 15 eV
ε>εa
Attachment e + M −→ M− εa ∼ kT
ε>εa
Dissociative attachment e + AX −→ A + X− εa ∼ eV
ε>εi p
Ion pair formation e + AX −→ e + A+ + X− εi p 20 eV
ε
Superelastic collision e + M j −→ e + M 0.01∼ –20eV
could survive long enough to suffer an additional collision with another gas
molecule and it would be stabilized, but usually the system is unstable and
it decays by autodetachment. Resonances are characterized by their energy
width ε, which may be associated with the compound’s lifetime through
the uncertainty principle.
PROBLEM 4.5.1
Calculate the energy width of the resonance that has a lifetime of 10−14 s. Describe
the elastic and the inelastic resonant scattering on a molecule.
There are two types of resonances:
i. Feschbach resonances. Here a temporary potential well is induced

when the target is excited by an electron and the electron is trapped.
These lie from 0 eV to 0.5 eV below the parent excited state.
ii. Shape resonances. The resonance state is above the parent state and
the scattering process strongly depends on the shape of the potential
well.
It is evident that electrons must satisfy a very narrow energy condition

in order to be able to excite the resonances. At the same time, enhanced in-
teraction, due to a long-lasting bound state, will result in a much higher
cross section, sometimes even by several orders of magnitude. It is difficult to
distinguish whether scattering is resonant if the lifetime is very short (or the
resonance is very broad). Yet resonant scattering is very useful in describing
the shape of the cross sections.
The term resonant usually refers to a specific situation where two processes
have to be matched in a very narrow range of conditions (in our case energy)
in order to interact. We use the term in this broader meaning elsewhere in this
book. A specific term resonant scattering, as described in this section, refers to
electron molecule collisions.

4.6 Electron–Atom Collisions

4.6.1 Energy Levels of Atoms
In the elastic collision between a light electron (mass m) and a heavier gas
molecule (mass M), the loss of kinetic energy is on average equal to 2(m/M)ε,
where ε is the kinetic energy of the electron. Thus for low energies, the effi-
ciency of energy transfer in one elastic collision is 10−4 as compared with the
value for typical inelastic processes. In order to understand inelastic collisions
between an electron and atom, one should first consider the energy levels of
an atom. The energies between different levels determine the energy of the
transition
ε = εk+1 − εk . (4.81)
The threshold energy εex required for the upper excited state is equal to the
excitation energy
ε = εex , (4.82)
and below that energy the cross section for the electron-induced excitation is
equal to zero. In Figure 4.16 we show the energy level for helium. Only the
allowed radiative transitions for bound electrons are possible (both absorp-
tion and emission). The lifetimes of excited states that have allowed radiative
transitions to lower levels are on the order of 10−9 s.
Bound electrons can also be excited to triplet states, and these will eventu-
ally decay to the lowest excited triplet states that have no radiative transition
to the ground state by dipole emission. Transitions by quadrupole and higher-
order moments are weak, and the lifetimes of these levels are typically in the
range of 102 s ∼ 100 s or even longer. These are the so-called metastable states.
Of the lowest excitation levels of He (see Figure 4.16), two, 23 S and 21 S, are
metastable states having lifetimes 7.9 × 103 s and 1.95 × 10−2 s, respectively.
The levels such as 21 P and 23 P that have allowed transitions to the ground
state are usually known as resonant levels, and their lifetimes are on the order
of nanoseconds. If the electron collides with the excited metastable, then step-
wise excitation or ionization is possible. This mode is more efficient, because
the threshold for excitation of the higher level or the ionization is lower than
that for the excitation from the ground state. On the other hand, the efficiency
depends on the density of metastable atoms (or excited atoms in general).
As the energy level increases, the energy gaps between excited states become
smaller and smaller. The energy levels eventually converge to the ionization
continuum where the excited electron is no longer bound to the atom forming
an ion and free electron.
Exercise 4.6.1
Using the selection rules, explain the optically allowed and forbidden transi-
tions in an atom.

25
Ionization +
24.6 He (2S)
24 51D 511P 51S 53D 53P 53S

41D 4P 41S 43D 43P 43S
4922Å 3965Å 5048Å 4471Å 3188Å 4713Å
23 31D 31P 31S 33D 33P
??
33S
Energy (eV)
6678Å 5015Å 7281Å

5875Å 3889Å 7065Å
22
21 21P
2.06µ 23P
21S 1.08µ
20 He singlet
584Å 23S
19 He triplet
He
0 11S
FIGURE 4.16
Energy levels of helium. Radiative transitions between singlet and triplet states are not allowed.
A typical term of an atom in a Russell–Saunders state is, for example, 43 P0 .

The first number is the principal quantum number n (in this case n = 4).
Values of orbital angular momentum L are denoted by capital letters:
S(L = 0), P(L = 1), D(L = 2), . . ., (in this case L = 1). The spin angular
momentum S is shown through multiplicity (2S + 1) and written as a super-
script in front of the letter annotating the value of L (S = 1). The subscript to
the right is equal to the total angular momentum J = L + S (J = 0), which
has integer values for odd multiplicity and takes values 1/2, 3/2, . . . for even
multiplicity. Parity is determined from the azimuthal quantum number for
individual electrons li as (−1) li . The superscript to the right, if present,
denotes parity (written for odd terms and not written for even terms). The
magnetic quantum number is M. The selection rules are:
i. J = 0, ±1, except for 0 → 0;

ii. Parity must change;
iii. L = 0, ±1, except for 0 → 0;
iv. S = 0;
v. l = ±1; and
vi. M = 0, ±1.

Rules (iii) to (v) are only approximate, but even strict rules may be broken if
there is, for example, strong spin-orbit coupling in strong magnetic fields, due
to nuclear perturbations or for autoionizing states. Rule (iv) does not allow
transitions among different multiplicities (e.g., triplet to singlet transitions).
Transitions with L= 0 are very rare. For our purposes, it suffices to note that
transitions to the ground state of helium 11 S from 23 S and 21 S are forbidden
because of the S = 0 rule in the former case and because of the L = 0 → 0
rule in the latter case.
4.6.2 Electron–Atom Scattering Cross Sections

In Figure 4.17 we show a set of cross sections for electrons in helium. The
elastic momentum transfer cross section at low energies is almost constant
with a value close to 6 × 10−16 cm2 . At those energies helium is an excellent
(and very rare) example of a constant cross section. Above 10 eV the elas-
tic cross section falls. There are no inelastic losses until the incident energy
exceeds 19.8 eV needed for the first electronic excitation. One should notice
that the excitation to triplet states peaks close to the threshold and drops
down rapidly due to the requirement of the spin exchange of the two bound
electrons from s( 12 , − 12 ) to s( 12 , 12 ) by the aid of the incident electron. On the
other hand, the cross section for excitation of singlet states is broad, peak-
ing at around 100 eV, and resembles the ionization cross section in energy
dependence.
-15
10
Qm
-16
10
Cross section (cm2)
Qi
3 1
-17 Q2 S Q2 P
10
1
3
Q3 P
Q2 P
-18 Q others
10 1
Q2 P
1
Q2 S
3
Q3 P 1
Q3 S
-19
10 -2 -1 2 3
10 10 1 10 10 10
FIGURE 4.17
Set of collision cross sections of electrons in helium as a function of electron energy. The subscript
others denotes the summed effect of all other higher levels, and the subscript i indicates ionization.
All other levels are denoted by their terms.

35 4p4 4p 4P7/2 35.25 eV

34 434.8 nm
33
32 4S4 4s 4P5/2 32.40 3V
+
Ar 2
P3/2 15.76 3V
15.76 3s2 3p5 2
P1/2
15 3p5
3p6
Energy (eV)
5
3p 5p 3p7
3p8
3p9
14 3p10
2p1 13.48 eV
2p2 13.33 eV
419.8 nm 3p5 4p 2p7 13.15 eV
13 2p9 13.08 eV
420.1 nm 2p10
772.376 nm
750.4 nm 772.421 nm
811.5 nm 1
12 1s2 P1 11.83 eV
3
5 1s3 P0 11.72 eV
3p 4s 1s4 3
P1 11.62 eV
1s5 3
P2 11.55 eV
104.8 nm
11
106.7 nm
Ar 1
0 3s2 3p6 S0
FIGURE 4.18
Energy levels of argon. The Paschen notation is also given.
Quite often Paschen notation is used to label the levels, especially for
heavier rare gases (see Figure 4.18). This is really just an assignment of levels
in their order rather than a notation associated with quantum numbers. Thus,
in this notation the ground state would be 1s1 , and the lowest metastable level
would be 1s5 . Two resonant lines from the four low-lying levels are in the
ultraviolet region, but most lines from the 4p and 5p states are in the visible
and near-infrared regions and are often used for diagnostics of argon plasmas.
Properties of the ground state, of the low-lying levels, and of the ionization
limit for the rare gases are given in Table 4.2.
Exercise 4.6.2
In the case of argon (see Figure 4.18), the ground state is 1 S0 . Explain the
allowed and forbidden transitions by using selection rules (and their possible
breakdown). Discuss the selection rules in the case of other low-lying levels
of the rare gases as given in Table 4.2.

TABLE 4.2
Energies; Configurations; and Terms of Ground States, Low-Lying Levels, and
Ionization Continuum for Rare Gases
Excited Levels
Gas Configuration Term Resonant (eV) Metastable (eV) Ionization (eV)
He 1s2 11 S0 2 3 P1 20.96 2 3 S1 19.82 2S

3/2 24.59
2 1 P1 21.12 2 1 S0 20.61
Ne [He]2s 2 2 p 6 2 1 S0 3 3 P1 16.67 3 3 P2 16.62 2P
3/2 21.56
3 1 P1 16.85 3 3 P0 16.72
Ar [Ne]3s 2 3 p 6 3 1 S0 4 3 P1 11.62 4 3 P2 11.55 2P
3/2 15.76
4 1 P1 11.83 4 3 P0 11.75
Kr [Ar]4s 2 3d 10 4 p 6 4 1 S0 5 3 P1 10.03 5 3 P2 9.92 2P
3/2 14.00
5 1 P1 10.64 5 3 P0 10.56
Xe [Kr]5s 2 4d 10 5 p 6 5 1 S0 6 3 P1 8.48 6 3 P2 8.32 2P
1/2 12.13
6 1 P1 9.54 6 3 P0 9.45
In the case of argon we have four low-lying levels: two of these are metastable
(3 P2 and 3 P0 ), and the other two (3 P1 and 1 P1 ) are resonant (with allowed
transitions to the ground state 1 S0 ). Transitions from the two metastables are
strongly forbidden by the rule that only J = 0, ±1 transitions, except for 0 →
0, are allowed. In this case, the 3 P1 level has a resonant transition to the ground
state in spite of the weaker rule that only S = 0 transitions are allowed.
All rare gases with the exception of helium have a similar configuration of
metastable and resonant states.
Cross sections for electron–argon collisions are shown in Figure 4.19. One
should observe that for energies below the threshold for the lowest metastable
(11.55 eV in this case) there are no inelastic losses, as in helium. However,
the elastic momentum transfer cross section has a broad minimum at around
0.23eV. This is the RTM, which is the first quantum effect observed for particles
with nonzero mass. When the RTM is present, the minimum value of the cross
section is two orders of magnitude lower than the maximum (and the value
that one would expect without the RTM). RTM is present in heavier rare gases
(Ar, Kr, Xe) and in many molecules.
One should observe that above the four lowest levels there are a large
number of excited levels, having transitions to those four levels including the
two metastables. If one wants to calculate the excitation rate to a particular
level, one needs to consider the excitation to all higher states and include the
radiative cascading by properly taking into account the branching ratios from
higher levels. In addition to excitation by electrons and radiativ ede-excitation,
there are several oher processes that provide nonradiative transitions to lower
or nearby resonant levels. These represent collisional quenching with other
gas molecules, electrons, or walls of the chamber and are discussed later. We
note, however, that some of the measurements of cross sections for excitation
suffer from the effect of cascading. All higher levels are also excited if the

-14
10
Qm
Qi
-16
10
Cross section (cm2)
Qex
Qexm
-18
10 Qex,(2p1)
Qex,(3p5)
-20
10 Qi,4p( D)
4
-2 -1 2 3
10 10 1 10 10 10
FIGURE 4.19
Cross sections for electron–argon scattering as a function of electron energy. Excitation is repre-
sented by effective summed cross sections and is separated into excitation of metastables and
excitation of higher levels. Three cross sections for specific transitions are added, two for ex-
citations into a neutral and the other for ionization into an excited ion from the ground state
neutral. These are employed for comparison between modeling and the optical diagnostics of a
low-temperature plasma in Ar.
electron energy is sufficient. Short-lived radiative states decay and some of

the transitions populate the excited state that is being studied. Thus the pop-
ulation of the excited state without collisional quenching will be
Ak j
Nj = Ndirect
j + Nk ,
k> j i<k Aki
where Ak j isthe probability of the transition to level j from the higher-level

k and Ak = i<k Aki is the inverse of the lifetime of the higher level. Because
the direct population is proportional to the cross section (for a monoenergetic
beam of electrons), the effective cross section consists of two parts: one due
to direct excitation to the desired state j and the second due to cascading
effects (where each of the higher levels is also subject to cascading). One
should pay attention to the source of the data, the technique that was used,
and other information in order to establish whether a cross section is subject
to the contribution due to cascading.

4.7 Electron–Molecule Collisions

Elastic collisions between an electron and molecule have cross sections and
energy dependencies similar to those for electron–atom scattering. There
are examples both with and without the RTM. However, there are also two
additional features important in electron–molecule scattering. For molecules
with a permanent dipole moment, there is an enhanced elastic (momentum
transfer) cross section peaking at zero energy. The second feature is the ex-
istence of either a sharp or broad structure due to resonant scattering for
energies below the threshold for electronic excitation. Both are discussed
later when complete sets for several molecules are presented. Let us begin
by examining the most salient features of molecules from the viewpoint of
electron–molecule collisions. Molecules having internal degrees of freedom
bring something new to the kinetics of electron transport: the excitation to
rotational and vibrational energy levels leads to inelastic energy losses at
much lower energies. Hence with a molecular gas it may be possible to control
the energy of electrons more precisely in a plasma.
4.7.1 Rotational, Vibrational, and Electronic Energy Levels of Molecules

When two atoms combine into a molecule, their dynamics will be determined
by the forces that act between them. A combination of repulsive and attractive
forces results in a potential energy for each internuclear position. A potential
curve is plotted for each combination of atoms as a function of internuclear
distance. The molecule will be stabilized at an equilibrium internuclear dis-
tance that presumably corresponds to the minimum potential. In Figure 4.20
we show the potential energy curves for a nitrogen molecule. The ground
state is X1 g+ . Superimposed on the potential diagram are the vibrational
(εv ) and rotational (εr ) energy levels. The energies of rotation and vibration
of molecules are quantized; that is, the energy levels are discrete. The total
energy of the bound electron in a molecule is thus
εT = εr + εv + εex . (4.83)
In Figure 4.20 we show only vibrational levels, inasmuch as rotational levels

would need a much higher resolution to be observed. The excited state A3 u+
also has vibrational and rotational levels. At room temperature molecules
populate a large number of rotational levels and, consequently, the emission
in transitions between two electronically excited states is broad and we talk
about molecular bands rather than lines. The third potential curve, depicted in
Figure 4.20, has no minimum. It is a repulsive curve, because the two atoms
composing the molecule would separate with some kinetic energy if they
found themselves on it.

16
12
N(3S) + N(2D)
A3 Σu+
v' = 20
X1 Σg+
Potential (eV)
=15 N(4S) + N(4S)

8 =10
=5
=0
v = 25
v = 20
4 v = 15
v = 10
v=5
0
0.4 0.8 1.2 1.6 2.0 2.4 2.8
Internuclear distance (Å)
FIGURE 4.20
Potential energy curves for a nitrogen molecule.
The total energy of a diatomic molecule with internal potential energy

V(r ) can be written using the relative velocity vr and reduced mass µ in the
CM frame in classical mechanics Equation 4.15 as
1 2
T= µv + V(r ),
2 r
which can be transformed to the spherical coordinate system
2 2
1 2 dθ 1 dr
T= µr + µ + V(r ), (4.84)
2 dt 2 dt
which corresponds to the Hamiltonian in quantum mechanics. In completing

the Hamiltonian, we should remember that the angular momentum L is con-
served and equal to L = r × P = r µ(r dθ/dt). This introduces the Hamiltonian
along with the relation P(= −ih̄d/dr) for quantization of momentum. The
Hamiltonian as the operator of the total energy is given as
L2 h̄ 2 d 2
H= − + V(r ). (4.85)
2µr 2 2µ dr2

The Hamiltonian may be separated into two terms, one corresponding to the
rotational θ -motion
L2
Hrot = (4.86)
2µr 2
and the other to the vibrational r -motion
h̄ 2 d 2
Hvib = − + V(r ). (4.87)
2µ dr 2
4.7.2 Rotational Excitation

4.7.2.1 Rotational Energy Levels
Rotational energy levels are obtained as eigenvalues εr of the appropriately
chosen eigenfunctions φr from the eigenvalue equation
Hrot φr = εr φr , (4.88)
where
L2 L2
εr = = .
2µr 2 2I
Here, I is the moment of inertia of the molecule, I = µr 2 , where r is the equi-

librium distance of the two nuclei. At the same time the angular momentum
is quantized as L 2 = l(l + 1)h̄ 2 with the quantum number l = 0, 1, 2, · · ·. As a
result, we have the rotational energy levels
h̄ 2l(l + 1)
εr = = Be l(l + 1), l = 0, 1, 2, · · ·. (4.89)
2I
Here, Be = h̄ 2 /2I is a rotational constant, and for most molecules it is on the
order of 10−4 eV∼ 10−3 eV, that is, meV. The transition energy between two
successive levels εr with l = 1 is
h̄ 2 h̄ 2
εrot = {l(l + 1) − (l − 1)l} = 2l = Be 2l. (4.90)
2I 2I
For molecules with a quadrupole moment, the selection rule is l =
±2 (l, l + 2), so in that case the energies of rotational transitions are εrot =
Be (4l + 6). (In both cases the transitions from the ground state are obtained by
choosing l = 0.) So far it has been assumed that the molecule is rigid; that is,
the distance between two atoms does not change as the molecule rotates. In
reality, however, the distance between molecules changes and so the moment
of inertia changes. This effect leads to the energy levels being described by

h̄ 2l(l + 1) 1
εr = = Be + α v + l(l + 1), l = 0, 1, 2, · · ·,
2I 2

where v is the quantum number of the vibrational level (discussed in the next
section) and α is a constant tabulated in books on spectroscopy.
Exercise 4.7.1
Estimate the energy of rotational transitions of an HCl molecule, and deter-
mine the ratio of the population of l = 1 to that of l = 0 at 300 K based on this
approximation. Prepare a table of exact values of Be for different molecules
and calculate exact energy losses for the first four levels of H2 , D2 , N2 , O2 ,
and CO2 .
For HCl, estimated values for the relevant quantities are
MH MCl
r ≈ 1.27 × 10−10 m, µ= ≈ 1.63 × 10−27 kg,
MH + MCl
which give the following estimate for the order of magnitude of the first
rotational level:
h̄ 2 h̄ 2 (6.63 × 10−34 /2π )2
εr (l = 1) = = ≈
I µr0 2 1.63 × 10−27 × (1.27 × 10−10 )2
≈ 0.423 × 10−21 J ≈ 2.64 × 10−3 eV.
The energy of rotational transition is on the order of meV. This should be

compared with the thermal energy of molecules at room temperature, 3/2kTg
(where k is the Boltzmann constant). That is, the rotational state populations
Nr (l) obey the Boltzmann distribution with statistical weight G l . The statistical
weight (factor of degeneracy) is (2l + 1) and therefore from the estimate given
above we get

Nr (l = 1) G1 εr (l = 1) − εr (l = 0)
= exp −
Nr (l = 0) G0 kTg

2.64 × 10−3
= 3 × exp − ≈ 2.71.
1.38 × 10−23 × 300/1.60 × 10−19
If we pursue this task more accurately we may assemble a table of values of Be
from the textbooks on spectroscopy or we may even calculate the values based
on the data for stable configurations of molecules. Then one may determine
the energy losses in the lowest rotational transitions. The results are shown in
Table 4.3. It is noted that energies (and Be ) in spectroscopy are usually tabu-
lated in units of cm−1 and that 1eV = 8.0655 × 105 m−1 . In Table 4.4, we show
typical wavelengths of radiation required to make rotational, vibrational, and
electronic transitions and the corresponding energies.
4.7.2.2 Rotational Excitation Cross Sections

Rotational inelastic losses are typically small, but because they occur at small
energies and there are no other losses in that range, it is important to in-
clude them properly. Rotational excitation is negligible for highly spherically
symmetric molecules such as CH4 . As mentioned above, rotational excitation

TABLE 4.3
Rotational Transition Energy of Several Molecules with a
Quadrupole Moment
∆ε (eV)
Molecule q4 (ea20 ) Be (eV) 0–2 1–3 2–4 3–5
H2 0.586 7.56E-3 0.0441 0.0735 0.1029 0.1323

D2 0.2487 3.79E-3 0.0221 0.0368 0.0515 0.0662
N2 1.04 2.48E-4 0.00146 0.00243 0.00340 0.00437
O2 1.80 1.78E-4 0.00107 0.00178 0.00249 0.00320
CO2 3.00 4.85E-5 0.00029 0.00048 0.00068 0.00087
TABLE 4.4
Typical Wavelengths and Transition Energies for Rotational,
Vibrational, and Electronic Excitations
Transition Wave Wavelength (m) Energy(eV)
Rotational Microwave 1 ∼ 10−3 1.24 × 10−6 ∼ 1.24 × 10−4

Vibrational Infrared 10−3 ∼ 0.8 × 10−6 1.24 × 10−4 ∼ 1.55
Electronic Visible 0.8 × 10−6 ∼ 0.4 × 10−6 1.55 ∼ 3.10
Electronic UV 0.4 × 10−6 ∼ 10−8 3.10 ∼ 30
will have the selection rule l = ±1 for molecules with a permanent dipole
and l = ±2 for molecules with quadrupole moment. Gases such as hydro-
gen have very large rotational energies that are comparable with and even
larger than their mean thermal energies at room temperature. Thus, we may
expect that a only few rotational levels will have any significant population.
For this reason, it should be possible to develop an accurate theory to repre-
sent rotational excitation and perhaps to use a normalizing factor for all the
levels. For parahydrogen at 77 K, in which only the ground rotational level
was populated, it was actually possible to determine the rotational excitation
cross section (see Figure 4.21) for the l = 0 → 2 transition from the transport
data. It is then possible to extend the analysis to the l = 1 → 3 transition from
the data for normal hydrogen. The good agreement with available theories
provided justification for the use of theory to determine cross sections for a
few higher transitions that exist at room temperature.
Exercise 4.7.2
What are para, the ortho, and the normal hydrogen? Calculate the population
of rotationally excited states under thermal equilibrium at room temperature
and at 77 K for hydrogen and deuterium. The nuclear spin for hydrogen
is I = 1/2 and that for deuterium is I = 1.
The para and ortho states of diatomic molecules are those in which the
nuclear spins of the two atoms are in the same direction and in the opposite

1.0
Qr (1-3)
Cross section (10 cm )
2
Qv (0-1)
-16
0.5
Qr (0-2)
Qv (0-2)
0.0
-2 -1
10 10 1 10
FIGURE 4.21
Rotational and vibrational excitation cross sections for hydrogen obtained from the swarm data.
directions, respectively. In normal hydrogen there is a 1:3 mixture of para to

ortho states, and in deuterium the ratio is 2:1. The abundance of a rotational
level of molecules in thermal equilibrium at temperature T is
Nl pl exp(−εr /kT)
= ∞ . (4.91)
N l=0 pl exp(−εr /kT)
Here, the statistical weight is equal to pl = (2I + 1)(I + a )(2l + 1). Here a = 0
for even l and a = 1 for odd l for molecules like H2 . The values of a are exactly
the opposite for molecules like D2 and N2 , which depend on the value of the
nuclear moment. The results of the calculations are given in Table 4.5.
TABLE 4.5
Rotational-Level Populations of H2 and D2
l=0 l=1 l=2 l=3 l=4 l=5
p-H2 77 K 0.9946 0 0.0054 0 0 0

n-H2 77 K 0.2487 0.75 0.0013 0 0 0
n-H2 293 K 0.135 0.67 0.112 0.079 0.003 0
n-D2 77 K 0.5721 0.3307 0.0945 0.0026 0 0
n-D2 293 K 0.1832 0.2060 0.3858 0.1137 0.0924 0.0134

4.7.3 Vibrational Excitation

4.7.3.1 Vibrational Energy Levels
The energies of molecular vibrational states may be obtained as eigenvalues
of the eigenvalue equation
h̄ 2 d 2
Hvib φv = εv ϕv ; H vib = − + V(r ). (4.92)
2µ dr2
Furthermore, we may describe the interaction potential V(r ) as a function of
departure from the equilibrium position r0 . The potential is approximated by
a parabolic curve with minimum V(r0 ) (see Figure 4.20):
V(r ) = A2 (r − r0 )2 + V(r0 ). (4.93)
If we replace x = r − r0 in Equation 4.93, we obtain
d 2 φv (x) 2µ

2
+ 2 εv − V(r0 ) − A2 x 2 φv = 0. (4.94)
dx h̄
The solution to this equation is obtained by using Hermite polynomials Hn as
√ 1/2
2µν/h̄ µν µνx 2
φv (x) = Hv x exp − , ν 2 = 2A2 /µ, (4.95)
2v v! h̄ 2h̄
where v is the vibrational quantum number. The energy of the vibrational
levels is then given by

1
εv = v + h̄ν + V(r0 ), v = 0, 1, 2 · ··, (4.96)
2
and the transition energy between successive vibrational levels is
εv = εv − εv−1 = h̄ν. (4.97)
PROBLEM 4.7.1
A good approximation of the realistic potential of diatomic molecules is the Morse
potential
2
V(r ) = D 1 − e −a (r −r0 ) .
Show that for small departures from the equilibrium position r0 , this potential may
be represented by the harmonic potential 4.93. Perform the calculations and plot
both potentials for the example of KCl where D = Equation 4.42 eV and a 2 =
0.602 × 10−20 m−2 .
PROBLEM 4.7.2
mass µ
Derive the solution of the classical vibrational motion of a particle of reduced
for force −2A r . Compare the angular frequency of the oscillation ν = 2A2 /µ to
2
that given in Equation 4.95. Relate the classical energy to that in Equation 4.96 and
explain.

Qm
Q eC
Q e1s2s
Q re02
-16
Q di
10 Q re13
Cross section (cm ) Q eB
2
Qi
-18 Q v1
10
Q v2 Q eE
Q re24
Q e1sn3
10
-20 Q re35 Q elsls
Q e2p2s
-2 -1 2 3 4
10 10 1 10 10 10 10
FIGURE 4.22
A set of cross sections for electron scattering of H2
PROBLEM 4.7.3
Estimate the magnitude of εv for H2 .
Exercise 4.7.3
Assuming that the potential energy of a diatomic molecule at equilibrium
distance is zero (V(r0 ) = 0), the energy of the lowest vibrational state is
εv=0 = h̄ν/2. Discuss this zero point energy from the standpoint of quantum
mechanics.
Quantum phenomena have a finite energy even at 0 K as fluctuation. It is well
known that the nonzero energy arises from the uncertainty principle.
4.7.3.2 Vibrational Cross Sections

Vibrational excitation cross sections for H2 are shown in Figure 4.21. For
hydrogen the threshold for the v = 0 − 1 vibrational excitation is 0.5 eV,
which is quite large. Other molecules have smaller vibrational energies on
the order of 0.1 eV to 0.2 eV. In the case of hydrogen one also needs to con-
sider the difference of losses for rotationally elastic (0 − 0) and inelastic (0 − 2)
vibrational excitations. It is important to note that there is a systematic differ-
ence between the data for the vibrational excitation cross section obtained by
fitting the transport coefficients and those from binary collision experiments
or theory. The difference exceeds the combined uncertainties. On the other
hand, the agreement for elastic scattering and rotational excitation between
both methods is excellent. The “hydrogen problem” for vibrational excitation
remains an open issue.

10
Q elast. Q tot.
Cross section (10-16 cm2)
1 Q vib.
Q att.?~× 50 Q ion.
0.1 Q exc.
Q rot.
-2
10
Q vib.
-3
10 -2
0.1 1 10 10
FIGURE 4.23
A set of cross sections for electron scattering on NO. Vibrational excitation shows a strong reso-
nance that is also present in the total scattering cross section.
The smooth dependence of vibrational excitation cross sections for

hydrogen as shown in Figure 4.22 is, however, not typical. In most molecules
there exist resonances that induce a very pronounced structure in the cross
section. Very strong resonances exist for N2 and O2 . In Figure 4.23 we show
a set of cross sections for NO, which includes resonant vibrational excita-
tion peaking at around 1 eV. However, above 10 eV there is another peak,
which may be obscured by competing with electronic excitation and dissoci-
ation, both of which are more efficient due to their larger energy losses. There
are no useful analytic formulae for vibrational excitation. In most cases the
v = 0 − 1 transition is the most important, but one may need to add data for
higher-level excitations for completeness. In such a case, relative magnitudes
or complete cross sections derived from the theory or beam experiments are
useful.
4.7.4 Electronic Excitation and Dissociation

4.7.4.1 Electronic States of Molecules
As shown in Figure 4.20, there are both bound and repulsive states of
molecules above the ground state potential curve. Molecules are usually
formed when electrons in the outer orbit are shared by both atoms. These
electrons are in different states. Angular momentum in this case is denoted
by the Greek letters , , , · · ·, and the values ( = 0, 1, · · ·) correspond to
the equivalent Latin letters S, P, D,· · ·.

V(r)
5
4
Ec c> 0
3
Potential energy
2
1
v=0
excited state
grounded state
VD
v=0
Pr (v)
r
Transition r0
probability 0 Internuclear distance
FIGURE 4.24
Dissociative electronic excitation and Franck–Condon principle of a molecule.
Exercise 4.7.4
List the most important selection rules for transitions between excited states
of molecules.
Some of the most important selection rules are:
i. = 0, ±1, where is the orbital angular momentum;
ii. = 0, where is the spin quantum number and = S,
S − 1, · · ·, −S;
iii. = 0, ±1, where = | + |, and for 0 → 0 it follows that
J = 0; and
iv. When the vibrational quantum number v changes, the v = ±1
transitions are the strongest but other transitions are not forbidden.
PROBLEM 4.7.4
Interpret the molecular terms X1 g+ and A3 u+ of N2 and discuss whether the upper
state is metastable.
A very important rule in electronic excitation of molecules is that transi-

tions take place over very short periods of time (on the order of 10−16 s), and
therefore the intermolecular distance cannot change during these transitions.
Thus, the so-called Franck–Condon principle states that the region of the
upper potential curve that is directly below the lower vibrational state will
be excited. In Figure 4.24 we show one such example. The ground state of
the molecule at the v = 0 level may be excited to the v = 0 − 5 levels of the
excited state. The most probable transition (around r0 ) will be to the v = 2
level. One may also observe that the levels v = 0 − 2 will be bound, although
there is a possibility of dissociation with the levels v = 3 − 5. In the case

TABLE 4.6
Dissociation Energy (VD ) of Ground-State
Molecules and Their Ions
Molecule VD (eV)
H2 4.48
H+ 2 2.63
B2 2.93
C2 6.24
N2 9.76
O2 5.11
F2 1.59
Cl2 2.48
Br2 1.97
I2 1.54
of dissociation, the potential curve may be reached above the dissociation

limit and the additional energy converted to kinetic energy of the molecular
fragments. The maximum kinetic energy will be E c , as shown in the figure.
If the upper state is repulsive then all excited molecules will be dissociated
with considerable kinetic energies of the fragments; these energies can be as
large as 10 eV to 20 eV. Dissociative excitation is an efficient way of producing
heavy particles of high kinetic energy.
Exercise 4.7.5
Describe a typical example of dissociation of molecules in thermal and
nonequilibrium plasmas.
Dissociation will occur if energy sufficient to cross the dissociation barrier VD
in Figure 4.24 is available. Data for dissociation energies for several molecules
are given in Table 4.6. In thermal plasmas, molecules will be dissociated
through a number of rotational and vibrational transitions by electron–
molecule, and even more so by molecule–molecule or ion–molecule colli-
sions. However, in nonequilibrium low-temperature plasmas, only electron–
molecule collisions to upper electronic excitation states will result in the dis-
sociation due to the low kinetic energy of heavy particles.
4.7.4.2 Cross Sections for Electronic Excitation of Molecules

In addition to direct excitation of bound molecular states, it is also possible
to have dissociative excitation of molecules. For example,
CH4 + e → CH2 + H(n = 3, 4, . . .) + H

→ CH(A2 ) + H∗ + H2 , · · ·,
and the cross sections for excitation of Balmer α and β emission from a CH4
molecule are shown in Figure 4.25.

-1
10
beam experiment
swarm analysis

-2
10
Hα
-3
10
Hβ
-4
10 2 3
10 10 10
FIGURE 4.25
Cross section for electron-induced dissociative excitation of CH4 into H(n = 3, 4) levels.
4.7.5 Electron Collisions with Excited Atoms and Molecules

When an electron collides with an excited molecule or an excited atom, then
there is a chance that the molecule will be de-excited with the electron gaining
energy equal to the excitation energy of the molecule. These are “superelastic
collisions” in which the electron may gain a lot of energy and jump from small
thermal energies to the high-energy tail of the energy distribution. The cross
section for a superelastic collision Qs may be calculated from the cross section
Q j for the excitation of state j (with the threshold energy of excitation ε j and
statistical weight g j ) by using the principle of detailed balance in a thermal
equilibrium situation,
g0 ε + ε j
Qs (ε) = Q j (ε + ε j ), (4.98)
gj ε
where g0 is the statistical weight of the ground state. As an example we show
data for the vibrational excitation of methane and the corresponding super-
elastic cross section (see Figure 4.26).
Exercise 4.7.6
Prove Equation 4.98 for a cross section of the superelastic collision based on
the principle of detailed balance.
The principle of detailed balance states that, in thermal equilibrium (at a
temperature Tg ), the number of transitions of electrons from higher energies
to lower energies due to inelastic collisions is equal to the opposite process
of superelastic transition from low to high energies. In other words, for a

σs24 superelastic
1
-1
10 inelastic
σv24
-2
10 -1 2
10 1 10 10
FIGURE 4.26
Cross section for electron-induced excitation and superelastic de-excitation of 2–4 mode of
vibration of CH4 .
combination of populations for ground state N0 and excited state Nj separated

by excitation energy ε j , we have a balance
√ sup
N0 ε + ε j Qexc0 j (ε + ε j ) f (ε + ε j ) = Nj ε Q j0 (ε) f (ε),

where the electron energy distribution is normalized to√ unity, f (ε)dε = 1.
Bearing in mind that for thermal equilibrium f (ε) = ε exp(−ε/kTg ) and
Nj /N0 = g j /g0 exp(−ε j /kTg ), where g j is the statistical weight of the excited
state j, we obtain

ε + εj gj εj
N0 ε + ε j Q0 j (ε + ε j ) ε + ε j exp −
exc
= N0 exp −
kTg g0 kTg

√ sup √ ε
× ε Q j0 (ε) ε exp − ,
kTg
which eventually yields Equation 4.98.
In Figure 4.27 we show data for a process of excitation from vibrationally

excited nitrogen (N2 (X1 g ; v = 1) to even higher levels. It is especially
important to provide data for metastable levels because these are highly pop-
ulated in nonequilibrium plasmas (depending on the electron density and the
presence of quenching impurities and other losses). In Figure 4.28 we show
the cross sections for collisional coupling between the metastable Ar(3 P2 ) and
the resonant Ar(3 P1 ). This coupling results in a significant loss of metastables,
that is, electron-induced quenching. Excitation to a higher state Ar(2 p) is also

8 6
6
σ (10-16 cm2)
1-1 4 1-2
4
2
2
0 0
2 3 4 5 2 3
4 4
1-3 1-4
σ (10-16 cm2)
2 2
0 0
2 3 2 3
Energy (eV) Energy (eV)
FIGURE 4.27
Cross section for the electron-induced vibrational excitation of a N2 (X1 g ; v = 1). 1–1 denotes
elastic scattering, and the excitations to higher vibrational state are indicated by 1–v.
10 2 3
P2 - 2P
3 3P1
P2 -
10
Cross section (10 -16 cm 2)
1 3P2 - ionization
10 -1 1P1
10 -2
3
P2
10 -3
10 -1 1 10 10 2 10 3
FIGURE 4.28
Cross section for the electron excitation of the metastable Ar(3 P2 ).

shown together with ionization. The thresholds for these processes are 1.55
eV and 4.16 eV, respectively. Hence these processes may become very impor-
tant once the metastable density is sufficiently high. For comparison we have
also shown the cross sections to the lowest metastable state Ar(3 P2 ) and the
resonant state Ar(1 P1 ) from the ground state Ar(1 S0 ).
4.8 Nonconservative Collisions of Electrons

with Atoms and Molecules
Collisions in which the number of electrons changes either being produced or
removed are regarded as nonconservative. Ionization, attachment, electron-
induced detachment from negative ions, and electron–ion recombination all
fall into this category. Although these processes may not differ significantly
from the viewpoint of atomic and molecular physics, they may have a marked
influence on the electron transport properties of any plasma (see Chapter 5).
4.8.1 Electron-Induced Ionization

If we excite a molecule (or atom) to the ground state of a singly charged ion,
then ionization of the molecule (or atom) is said to have occurred and an
additional electron is free to leave the target particle. One typical example of
this process is written as:
e + A → e + e + A+ . (4.99)
The energy loss of the incoming (primary) electron with ε is εi (i.e., the ioniza-
tion energy). The remaining energy (ε − εi ) is split between the two electrons,
and usually the slower one is regarded as secondary. The double differential
cross section, as a function of energy ε of the primary electron and εs of the
secondary [7] is approximated by an analytic formula provided from beam
measurements:
Qi (ε) 1
σ (ε, εs ) = i εs 2.1 .
ε tan−1 ε−ε2ε 1+ ε
In Table 4.7 we show the data required to determine the double differential
cross section σ (ε, εs ) for atomic ionization.
The ionization cross section is usually determined very accurately in
electron beam experiments and it is often used directly in cross section sets
for plasma modeling. In Figure 4.29 we show data for ionization of an argon
atom to several different charge states, Ar+ , Ar2+ , and so on. We should also
note that there is a possibility that excited states of an ion or even multiply
charged ions may be excited/ionized directly from the ground state of the
neutral. For most nonequilibrium plasmas for which applications currently
exist, only the ionization leading to a singly charged ion needs to be consid-
ered. The collisions of ions and electrons leading to further ionization may be

TABLE 4.7
Double Differential Cross Section Parameters for
Ionization of Several Atoms and Molecules
Molecule εi (eV) ε (eV)
He 24.6 15.8
Ne 21.6 24.2
Ar 15.8 10.0
Kr 14.0 9.6
Xe 12.1 8.7
H2 15.4 8.3
O2 12.2 17.4
N2 15.6 13.0
CH4 13.0 7.3
neglected. The ionization cross section peaks at a few hundreds of eV, and at
these energies it becomes the dominant collision process, even exceeding the
elastic scattering. For molecules, ionization is often accompanied by disso-
ciation. Simultaneous ionization and excitation often occur from the ground
state of the neutral molecule, and emission from the excited ions may be used
as a probe for high-energy electrons. Close to the ionization threshold, the
cross section increases as
Qi ∼ (ε − εi )1.127 ,
i=0
-16
10
Cross section (cm )
2
i=1
-17 i=2
10
i=3
i=4
i=5
-18
e + Ar i+ Ar (i+1)++ 2e
10
2 3
10 10 10
FIGURE 4.29
Ionization cross section for e + Ai+ → A(i+1)+ + 2e (i = 0–5). This cross section is different from
one for multiple ionizations from the ground state of a neutral atom.

according to Wannier’s classical theory. For very high energies the cross
section decays according to the Bethe–Born approximation as
1
Qi ∼ (Aln ε + B),
ε
where A and B are the terms for optically allowed and forbidden transitions,
respectively.
4.8.2 Electron Attachment

For some molecules it is possible that an incoming electron may be attached
and a stable negative ion formed. Electron attachment is also a nonconserva-
tive process. In this case, we need to consider a possibility that the electron
will be subsequently released in some detachment processes. Cross sections
for electron attachment are typically smaller than ionization cross sections.
On the other hand, they may occur at lower energies between zero and the
ionization energy. There are different processes that may lead to production
of negative ions, including dissociative electron attachment (DEA), nondisso-
ciative electron attachment (NDEA), and ion pair formation. These processes
may be standard in both two-body and three-body processes. They are shown
through the following equations:
(4.100)
Q0 (ε) Pd
e + AX − −→ AX − ∗ −→A + X−
(+M) → AX− + (energy) (4.101)
and
e + AX → (AX∗ + e) → A+ + B− + e. (4.102)
In the first step, the electron is attached to the excited state [AB]∗ . The state
may decay, through dissociation with probability pd , to form two fragments,
one being a negative ion (Equation 4.100). The other possible channel is for the
excited state to be stabilized in collisions with the third particle M, preferably
a heavy buffer gas molecule. It will then proceed to form a stable ground-state
negative ion of the parent molecule (Equation 4.101). Excess energy will be
taken away by the third particle. Another way to describe these processes
is by following the dynamics on potential energy curves V(r ), as shown in
Figure 4.30.
i. (AX− )∗ is formed through excitation to the repulsive upper state.

As the molecule breaks apart, while r ≤ rc there is a chance for au-
todetachment to occur and the electron to be released again, leaving
a neutral molecule. When r > rc the molecule will definitely break
up into one negative ion and one neutral fragment.
ii. The excited state has a minimum. In this case, a group of molecules,
excited to the potential curve at an energy higher than the dissociation
limit, will undergo dissociative attachment. Other molecules that do

(I) (II) (III)

(AX-)* (AX-)*
Potential energy
A+X A+X A+X
AX AX AX
A+X-
A+X-
AX-
r0 rc
Internuclear distance
FIGURE 4.30
Potential energy curves for dissociative and nondissociative electron attachment.
not have the energy to dissociate will oscillate in the potential well
until autodetachment occurs. Even if the minimum of the excited
state is lower than the ground state of the neutral molecule there is
a threshold for this process to occur.
iii. If the potential bottom of the excited state (AX− )∗ is below that of the
lowest level of the neutral molecule, then there is a chance that the
curves will cross close to the minimum and that even electrons with
zero energy may be attached. The excited negative ion will spend
time oscillating until it is stabilized by collision with the third body
or until autodetachment releases the bound electron.
The overall rate of attachment will strongly depend on the ratio of the
mean free time between collisions with gas molecules and the lifetime of the
excited state. A good example of the potential curves of ground-state neutrals
and negative ions is shown in Figure 4.30. An example of the potential curves
is given in Figure 4.31 for oxygen. There, one may expect nondissociative
attachment to occur due to crossing of the ground state of the molecule by
the potential curve for the negative ion O− 2 . The electron affinity E a is the
energy released if a zero-energy electron is detached from a negative ion. The
values of affinities for electrons for several atoms and molecules are shown
in Table 4.8.
TABLE 4.8
Electron Affinity
Atom Ea (eV) Molecule Ea (eV)
F 3.40 F2 3.08
Cl 3.61 Cl2 2.38
Br 3.36 Br2 2.51
I 3.07 I2 2.58
O 1.46 O2 0.44
S 2.08 SF6 0.60
C 1.26 CF2 2.10
CF3 1.92

15 (a)
Potential energy (eV)
10 • 2
-
O2 ( Π u )
(b)
• 2
- +
O2 ( Σ g ) 6
-
O ( ∆g
1 2
Πu )
QDA (10-18 cm2)

-
O ( 2 P) +O( 1 D)
5 O( 3 P) +O( 3 P) 4 -
O ( Σg
3 - 2
Πu )
-
O ( 2 P) +O( 3 P) - +
O ( ∆g Σg )
1 2
2
O2 ( ∆ g )
1
0 - 0
O2 ( Σ g )
3
2 4 6 8 10
-2
1.0 2.0 3.0 Electron energy (eV)
FIGURE 4.31
Potential energy curves for oxygen showing the ground state O2 (3 g− ), metastable O2 (1 g ), and
higher excited states (a), cross sections for dissociative attachment of the O2 (3 g− ) and O2 (1 g )
states (b).
4.8.2.1 Dissociative Electron Attachment

Let us take molecular oxygen as an example of molecules having dissociative
electron attachment. The efficiency of attachment or the effective cross section
is given as Q0 (ε) × pd (where Q0 is the cross section to produce an excited
unstable negative ion). The set of electron cross sections of O2 is given as Qa
in Figure 4.32. One may observe resonant vibrational excitation, a number
of electronic excitation cross sections, and the ionization cross section, which
begins to dominate just below 100 eV. Finally, we can see that dissociative
attachment is indicated as a small narrow cross section peaking at around
10−18 cm2 . The process that leads to production of O− ions may be described
as
ε>3.7eV 2 ∗
e + O2 (3 g− )−−−−→ O− 2 u → O− (2 P) + O(3 P). (4.103)
The O− ions produced may then undergo any of the following reactions:
ε>1.02eV
O− + O2 −−−−−→ O + O−
2 (charge transfer), (4.104)
ε>0.39eV
O− + O2 −−−−−→ O3 + e (associative detachment), (4.105)
or
ε>1.46eV
O− + O2 −−−−−→ O2 + O + e (electron detachment). (4.106)

-15
10
Qm Qi
B3Σu
-16
10
c1Σu+C3∆u
Cross section (cm )
vib.
2
10
-17 a1∆g others
+
A3Σu
-18
10
+
b1Σg
Qa Qp
-19
10 -2 -1 2 3
10 10 1 10 10 10
FIGURE 4.32
A set of cross sections for electrons in O2 .
4.8.2.2 Nondissociative Electron Attachment

The first example of nondissociative attachment is the electron-SF6 attachment
at low energies. The process may be described as
+M
e + SF6 −→ [SF−6]
∗
−→ SF−6,
−
τad ∼ 25 µs SF5 + F. (4.107)
Here, the lifetime of the excited state is quite long and it is almost certain
that the molecule will be stabilized in collisions with the third particle M.
Effectively, this process will behave like a two-body process rather than a
three-body process. It is also possible to produce SF−
5 in a dissociative attach-
ment process. Low-energy attachment to SF6 is close to the upper theoretical
limit for attachment for s-wave scattering dominant at low energies. A set of
cross sections for electrons in SF6 is shown in Figure 4.33.
A second example for nondissociative attachment is that of O2 , in which a
real three-body process will be observed at normal pressures due to the short
lifetime of the excited state. The process may be described as
ν1
−→ O2 + e
Q0 (ε) ∗
e + O2 −→ O− 2 (4.108)
ν
τad ∼ 0.1 ns (+M) −→ O−
2 + M,
where ν is the collision rate with a third-body M with pressure dependence,

and ν1 is the frequency to the autodetachment. Thus the effective cross section

-14 Q a1
10
Qm Qi
Cross section (cm )
2
-16
10 Qe
Q a3
Q a2 Q a4
Q a5 + +
10
-18 Q [N2 (B3 Σu )]
Qv
10
-20 Q [N2 (C3 Πu)]
-2 -1 2 3
10 10 1 10 10 10
FIGURE 4.33
A set of cross sections for electrons in SF6 . There is one nondissociative (a 1 ) and several disso-
ciative channels for attachment. Cross sections for excitation of two levels of N2 are also shown.
of the electron attachment will depend on pressure by way of ν:

ν
Qa 3 (ε) = Q0 (ε). (4.109)
ν1 + ν
4.8.2.3 Ion Pair Formation

The cross section for ion pair formation in electron–oxygen collisions
ε>17.28 eV
e + O2 −−−−−→ O+ + O− + e (4.110)
is shown as Q p in Figure 4.32. The threshold is at 17.28 eV and the maximum

of 0.48×10−18 cm2 is at 34 eV.
4.8.2.4 Electron Attachment to Excited Molecules

At elevated temperatures, populations of higher rotational and even vibra-
tional levels increase, and one may expect that effective thresholds for
inelastic processes and attachment decrease. However, in nonequilibrium
plasma, populations of excited states usually greatly exceed the thermal pop-
ulations for the gas temperature. Thus, effective rotational and vibrational
temperatures on the order of 500 K to 3000 K are not uncommon. A classic
example of the large enhancement of the cross section for dissociative attach-
ment occurs in hydrogen (and even more so in deuterium). In Table 4.9 we
show the dissociative attachment cross section to hydrogen as a function of
rotational and vibrational excitation. v and r are the vibrational and rotational
quantum numbers, respectively.

TABLE 4.9
Cross Sections for Dissociative
Attachment to H2 (1 g+ )
v r εa (eV) Qa (cm2 )
0 0 3.73 1.6 10−21

0 10 3.13 2.2 10−20
1 0 3.23 5.5 10−20
2 0 2.73 8.0 10−19
3 0 2.28 6.3 10−18
4 0 1.85 3.2 10−17
5 0 1.45 1.1 10−16
9 0 0.13 4.8 10−16
In nonequilibrium low-temperature plasmas, metastable molecules may

also be populated with considerable concentrations and may provide the
basis from which attachment and ionization can proceed. Here, we give
as one example the metastable state, O2 (a 1 g ), which is 0.98 eV above the
ground state (see Figure 4.31) with a lifetime of 2700 s. Electron impact exci-
tation for this state may be described as
ε>0.98 eV
e + O2 (3 g− ) −−−−−→ e + O2 (a 1 g ). (4.111)
The excitation of the ground state O2 (3 g− ) to the higher O∗−

2 (
2
u ) state dis-
sociates to two atoms. On the other hand, when the electron collides with the
metastable O2 (a 1 g ), electron attachment proceeds with threshold energy
2.76 eV, which is 0.98 eV less than that of the ground state as
ε>2.76 eV ∗
e + O2 (a 1 g ) −−−−−→ O− 2(
2
u) → O− (2 P) + O(3 P), (4.112)
and it will result in dissociative attachment; however, the difference of 0.98

eV in threshold energy will result in a large increase in cross section (by a
factor of almost four).
4.8.2.5 Rate Coefficients for Attachment

Many experiments on electron attachment to molecules have been conducted
in such a way that the rates for such processes are defined by

2 √
Ra = NQa (ε) ε f (ε)dε,
m ε
where the energy distribution of electrons, f (ε), is normalized to unity. Ra
describes directly the loss of electrons and the increase of negative ion density
in plasma kinetics. The data for nonequilibrium conditions as a function of
E/N are provided by swarm experiments with drift tubes. There is a sep-
arate group of experiments including the Flowing afterglow, the Cavalleri
diffusion experiment, and others that provide attachment rates for thermal

TABLE 4.10
Thermal Attachment Rate Coefficients (300 K)
Molecule ka = Ra /N (cm3 s−1 )
Cl2 2.8 × 10−10

F2 3.1 × 10−9
SF6 2.27 × 10−7
CCl4 2.8 × 10−7
CH3 I 7 × 10−8
CH3 Br 6.7 × 10−12
CF2 Cl2 1.2 × 10−9
equilibrium. In Table 4.10, we show the thermal attachment rate coefficients

for several molecules. Measurements were performed for thermal electrons
in equilibrium with background gas at room temperature (300 K).
4.8.3 Electron–Ion and Ion–Ion Recombination

Although attachment leads to loss of more mobile electrons, it still maintains
the same number of free charges in plasma. On the other hand, the process
of recombination leads to decay of the density of charges. For example, if
electrons with density ne and positive ions with density n p recombine and
the rate coefficient of the loss is krm , then the charged particle kinetics are
described by
d
ne (t) = −krm ne (t)n p (t), (4.113)
dt
where the rate coefficient is associated with the cross section through
krm = Qrm (εr )vr .
Recombination processes are numerous, and some of the most important

ones for nonequilibrium plasmas are listed in Table 4.11. Dissociative recom-
bination is very fast but nondissociative recombination is very weak, and a
three-body process is more efficient in high-pressure plasma conditions.
If there is a stable ion AB + then its potential curve must have a mini-
mum that is considerably higher than that of the AB molecule. The AB + ion
will require some finite dissociation energy VD+ to dissociate into A+ and B.
However, there will be a chance that this potential curve will be crossed in the
region of the minimum by some of the repulsive potential curves of neutral
molecule AB, which would allow dissociation when the electron approaches
the ion. There will also be a finite probability that autoionization will occur
before fragments separate very far. The fragments A and B will be produced
with a large kinetic energy (εk ). In Figure 4.34 we show a diagram of dissocia-
tive recombination. One of the most important processes in the ionosphere is

TABLE 4.11
Recombination Processes and Estimated Rate Coefficients (300 K)
Recombination Thermal Rate
Process Reaction Scheme Coefficient
Radiative A+ + e → A j + hν ∼ 10−12 cm3 s −1

Dissociative AX+ + e → (AX)∗∗ → A j + X + (εkin ) ∼ 10−8 cm3 s −1
Dielectronic A+ + e → (A)∗∗ → A j + hν —
Ion pair A+ + X − → A j + B ∼ 10−7 cm3 s −1
Three-body A+ + e + X → A + X + hν ∼ 10−26 cm6 s −1
A+ + X− + M → A + X + M + (εkin ) ∼ 10−25 cm6 s −1
A+ + e + e → A j + e ∼ 10−19 cm6 s −1
the dissociative recombination of the NO+ ion
e + NO+ → N + O + εk .
Exercise 4.8.1
Explain why radiative attachment and recombination are less effective than
other nonradiative processes.
Radiative attachment is described by
A + e → A− + hν.
The lifetime of the excited state is typically on the order of 10−8 s, and the
time that the electron spends in the neighborhood of the atom is on the order
of 10−15 s (dimension 10−7 cm and velocity 108 cm s−1 ). Thus, the probability
V(r)
AB
A+ + B
AB+
Potential energy
VD+
εk
AB
A+B
FIGURE 4.34
Potential diagram for dissociative recombination of the AB+ molecular ion.

that radiative stabilization of the transient negative ion will occur is 10−7 .
Therefore, the effective cross section will be on the order of 10−21 cm2 . For
slower electrons the cross section will be larger. For pairs of electrons and
ions, the interaction will also occur at a greater distance, which will lead to
a somewhat longer interaction time and a larger cross section. In the case
of dissociative processes we do not need photon emission to reduce the en-
ergy of the electron from a free to a bound state. Thus, the efficiency will be
defined by the collision cross section only. Radiative attachment and recom-
bination will be roughly seven to five orders of magnitude less effective than
the corresponding dissociative processes.
Exercise 4.8.2
Explain the inverse processes of radiative recombination and photo-
ionization. Define and explain the dissociative and dielectronic recombination
and their inverse processes.
It is obvious that the photo-ionization, that is,
hν + A ↔ A+ + e,
is the opposite of radiative recombination. Dielectronic recombination

follows a dielectronic capture, where the incoming electron excites the ionic
target but, as a result, loses energy and is bound to some high (Rydberg)
excited state. Doubly excited targets may relax without losing the incoming
electron. If the excited electron returns to the ground state, by emission of a
photon, the electron in the Rydberg state is left bound to the target (dielec-
tronic recombination). The opposite process is autoionization, in which one
electron is relaxed by photon emission, which is providing energy for the
other (Rydberg) electron to leave the target. Dielectronic recombination is of
great importance for astrophysical plasmas, but it is not significant in most
nonequilibrium plasma applications.
4.8.4 Electron–Ion and Electron–Electron Collisions

In addition to electron–ion recombination, a number of different collisions
may occur between two charged particles. Because inelastic electron–ion
processes are similar to those for neutrals, we are primarily interested in
momentum transfer collisions between two charged particles. We have al-
ready discussed the Rutherford scattering of two charged particles. In general,
collisions of charged particles thermalize electrons and increase the temper-
ature of ions and neutrals.
4.9 Heavy Particle Collisions

The term “heavy particle collisions” covers a wide range of very different
processes involving ions, fast neutrals, slow neutrals, excited species, and
chemical reactions.

TABLE 4.12
Ion–Molecule Collisions
Process Reaction Scheme
Elastic Scattering X± + AB → X± + AB
Positive Ions
Nonresonant charge transfer X+ + AB → AB+ + X
Resonant charge transfer AB+ + AB → AB + AB+
Dissociative charge transfer X+ + AB → A+ + B + X
Projectile excitation X+ + AB → X+∗ + AB
Target excitation X+ + AB → X+ + AB∗
Target ionization X+ + AB → X+ + AB+ + e
Atom interchange X+ + AB → XA+ + B
Ion interchange XB+ + A → X + AB +
Addition X+ + AB → XAB+
Three-body association X+ + A + B → XA+ + B
Negative Ions
Heavy atom transfer X− + AB → XA + B−
Associative detachment X− + AB → XAB + e
Detachment X− + AB → X + AB + e
4.9.1 Ion–Molecule Collisions

Some types of ion–molecule collisions are listed in Table 4.12. The list could
be extended, especially by a large number of additional possibilities for ion–
molecule chemical reactions.
In gas discharges the mean energy of ions is quite low and normally ions
cannot achieve the energies required to perform excitation or ionization. How-
ever, this is not always true, and recently some low-pressure discharges have
been studied with the result that, at very high E/N and low pressures, con-
ditions may be met for significant heavy particle excitation and sometimes
even ionization. Such conditions may be found in a low-pressure magnetron
sputtering, sheaths of DC and rf discharge plasmas, and plasma thrusters. A
set of cross sections for the elastic, charge transfer, and inelastic processes of
the ion is shown for two ions in their parent gases. The first is an example of
Ar+ ions in Ar (see Figure 4.35) and the second an example of N+ ions in N2
(see Figure 4.36).
4.9.1.1 Charge Transfer, Elastic, and Inelastic Scattering of Ions
In principle, in the case of ions and molecules of the same mass, charge ex-
change (or charge transfer [CT]) could be regarded as elastic scattering in
the backward direction; that is, it is highly anisotropic. The process of charge
exchange is defined as
A + + A → A+ + A
(fast) (slow) (slow) (fast)
whereby the ion that has gained energy from the electric field collides with
a slow particle, and in the process it appears that an electron jumps from the

-14 Qe.m.t.
10
Qc.t.
Cross section (cm )
2
-15
10
-16
10
-17
10 Qiso.
(Ar+ + Ar)
-18
10 -1 2 3 4
10 1 10 10 10 10
Relative energy εr (eV)
FIGURE 4.35
A set of cross sections for the Ar+ ions in Ar.
neutral to the ion. Therefore, the fast neutral continues with almost the same
kinetic energy as the incoming ion. The ion is left with a very small energy,
equal to that of the target atom.
If we have ions in their parent gas, the charge transfer is said to be res-
onant (because of the matching energy levels) or symmetric. This process is
-14
10
Qm
-15
10
Cross section (cm )
2
-16
10
Q chtr
-17
10
Q j (B3 )
-18
10
Q j (391nm)
-19
10 -1 2 3 4
10 1 10 10 10 10
Labratory energy (eV)
FIGURE 4.36
A set of cross sections for N+ ions in N2 .

extremely efficient and it appears to dominate the transport of ions in their

parent gas at all energies, though it has a maximum cross section at zero
energy. A good example of this process is the rare gas ions in their parent
gas (see Figure 4.35). The symmetric (resonant) CT cross section may be as
large or even larger than the gas kinetic cross section. The cross section de-
creases slowly with the energy and thus controls the energy of ions at all
energies.
It is, however, important to be aware of one possible source of confusion
if using data for the cross section of the CT QC T from the literature. Most
of these data originate from the analysis of ion mobilities. The result of that
analysis is Qm . These data were converted to the CT cross section under the
assumption that the momentum transfer cross section corresponds only to
the backward scattering, so that QC T = Q0 = Qm /2. It was thus implicitly as-
sumed that all processes were negligible except for QC T . However, these data
could also be used assuming isotropic elastic scattering so that Q0 = Qm . The
momentum transfer data were applied in both approaches in the literature.
Regardless of the choice, one would correctly predict the mobility of ions, and
thus it was not possible to test the data. However, other transport coefficients
could be incorrect. Most importantly, the transverse diffusion NDT is calcu-
lated incorrectly. In the first approach NDT = 0, and thus effectively such
an application of mobility data would effectively be equivalent to the one-
dimensional model. A more complicated and more difficult approach would
be to calculate the differential cross sections and use them to analyze the ex-
perimental data and for plasma models. However, at low energies we can
split the cross section into two parts, isotropic Qiso and backward scattering
Qb , so that Qm = Qiso + Qb [8].
Obviously, additional information is required, and the best way is either
to use a measured differential cross section and fit it or to use available NDT
data. In the low-energy limit the isotropic part of the total cross section will
reach the polarization limit of the cross section. In Figure 4.35 we can see
a typical example of the ion–atom cross sections with resonant CT. Elastic
isotropic scattering is very high at zero energy, but it diminishes and is neg-
ligible as compared with CT scattering beyond 100 eV. However, it is con-
siderable for the largest range of nonequilibrium plasma conditions (i.e., for
energies up to 10 eV). The resonant CT cross section is almost constant at all
energies and it effectively controls the energy of ions. Still, at very high E/N
ion energies may become high enough that inelastic processes may become
observable.
When an ion collides with a neutral atom (or molecule) other than those of
the parent gas, the CT is not resonant; such a transfer is called an asymmetric
CT. This situation is typical for ions such as H+ and H+ 3 in H2 . Isotropic
elastic scattering, however, behaves in the same way. The asymmetric CT will
be small, and there is the chance of an ion runaway (creation of a beamlike
distribution where ions keep gaining energy with less and less chance to suffer
collisions and dissipate the energy). One example of data with an asymmetric
CT is given in Figure 4.36. Unfortunately, detailed data that are of interest for

plasma processing exist only for a limited number of ion–background gas

combinations. Some data exist for rare gases; basic diatomic molecules such
as H2 , N2 , and O2 ; and alkaline ions in numerous gases. Recently, data have
begun to be made available for asymmetric CT of rare gas ions in other rare
gases.
4.9.1.2 Ion–Molecule Reactions

Apart from elastic scattering and CT that occur at thermal energies, there
are numerous ion–molecule processes that are of possible interest. A brief
list is given in Table 4.12. It would require a separate book to give a de-
tailed presentation of all possible low-energy ion molecule reactions. We
just give several examples that have been shown to play a role in some
discharges. First, we must separate these processes into those involving pos-
itive and those involving negative ions.
In addition to the processes given quite generally in Table 4.12, we add
some more specific processes. A very important class of chemical processes
involves the transfer of a hydrogen atom or its ions as follows:
proton transfer: H2 S+ + H2 O → H S + H3 O+ ;
(hydrogen) atom transfer: C2 H4+ + C2 H4 → C2 H5+ + C2 H3 ;
hydride ion transfer: Ra+ + Rb H → Ra H + Rb+ .
A particularly important reaction that determines the dominant ion in

hydrogen discharges at low energies is
H2+ + H2 → H3+ + H.
Examples involving more general rearrangement processes with positive ions

are
N2+ + O → NO+ + N and

O+ + N2 → NO+ + N,
which play a role in atmospheric chemistry. Cross sections (rates) for ion–
molecule reactions are quite high and often are almost as large as the
orbiting limit. Orbiting occurs for special combinations of the potential and
centrifugal forces, when the radial component of velocity vanishes and the
projectile makes many revolutions around the target before being released
(see Section 4.5). The extended interaction time makes it possible to complete
long-lived transitions.
As an example of reactions with negative ions, typical reactions involving
oxygen ions are
O− + H2 O → OH − + OH,
O− + 2O2 → O3− + O2 and
O− + O2 → O2− + O,

but there are also reactions involving thermal detachment (including associa-
tive detachment):
O− + O → O + O + e,
O− + O2 → O + O2 + e,
O2− + O2 → O2 + O2 + e,
O2− + O → O3 + e.
Detachment is an extremely important process, and in addition to heavy

particles it may be induced by electrons or photons. Finally, we should include
processes forming clusters.
Exercise 4.9.1
Determine the rate for orbiting collisions between a slow ion and neutral
with an attractive interaction potential given by V(r ) = −αe 2 /8π ε0r 4 for
r ≥ a and V(r ) = ∞ for r < a , where the interaction is due to the polarization
determined by polarizability α.
The effective radial interaction energy is equal to
b2 E
Veff (r ) = V(r ) + ,
r2
where the second term is due to the centrifugal force, E is the energy of the
incoming particle at r → ∞, and b is the impact parameter (see Equation 4.43).
The radial velocity is equal to

vradial = 2(E − Veff (r ))/M,
and for orbiting motion, the radial velocity must be equal to zero. Combining
these two conditions one can derive the value of the critical impact parameter
b c , which corresponds to the orbiting motion as
1/4
αe 2
bc = .
2π ε0 E
The orbiting cross section is thus equal to
1/2
αe 2
Qor = π b c2 =π .
2π ε0 E
For a finite probability of transition, the cross section should be multiplied by
probability ζ for the transition during orbiting.
4.9.2 Collisions of Fast Neutrals

Fast neutrals produced in a CT collision continue with very high energy.
These fast neutrals will have quite different cross sections from those for ions.

-15
10
Qtotal
Cross section (cm )
2
-16 Qm
10
Qi
Q(811nm)
-17 Q(795nm)
10
-1 2 3 4
10 1 10 10 10 10
Labratory energy (eV)
FIGURE 4.37
A set of cross sections for fast Ar atoms in argon. Ionization and ultraviolet excitation cross
sections are identical.
Studies of Townsend discharges at very high E/N revealed that fast neutrals
are actually more effective in producing excitation (and possibly ionization)
than are ions. The energy distribution of fast neutrals upon formation is the
same as that of ions, but they do not gain energy. In Figure 4.37 we show a
set of data for cross sections for fast neutral Ar(fast)-Ar collisions [9].
It was discovered that the production of fast neutrals may also take place
on a surface. Collisions of fast ions with a surface often result in reflection and
neutralization, sometimes with small or negligible energy loss. Furthermore,
for molecular ions, the breakup of molecules occurs and molecular neutral
fragments are reflected.
4.9.3 Collisions of Excited Particles

Excited atoms or molecules play a considerable role in the kinetics of plas-
mas. Metastables constitute a large store of energy that may critically affect
processes that sustain plasmas. On the other hand, they may induce chemical
processes and participate in processes at surfaces. Thus, we are particularly
interested in ionizing collisions in the gas phase and at surfaces. In this sec-
tion, we concentrate on electronically excited atoms (and molecules). Some of
the interesting processes are listed in Table 4.13. A very special process is the
so-called energy-pooling process: A∗ + B ∗ = A+ + B+ e. For argon metastable–
argon metastable collisions, the rate coefficient is 6 × 10−10 cm3 s−1 . It may be
one of the dominant sources of ionization in high-density plasmas containing
high metastable densities.

TABLE 4.13
Collisions of Excited Species
Process Reaction Scheme Energy Condition
Leading to Ionization
Penning ionization X∗ + AB → X + AB+ + e ε j (X) ≥ εi (AB)

Dissociative ionization X∗ + AB → X + A + B+ + e ε j (X) ≥
Associative ionization X∗ + AB → (XAB)+ + e εd (AB)+εi (B)
Heavy atom ionization transfer X∗ + AB → XA+ + B + e ε j (X) ≥ εi (XA)
+εd (AB) − εd (XA)
Energy pooling X∗ + X∗ → X+
2 +e
Ion pair formation X∗ + AB → X + A− + B+
Leading to De-excitation or Excitation Transfer
Excimer formation X∗ + X + M → X∗2 + M

Transition to resonance levels X M + X→ Xres + X → 2 X + hν
Collisional-radiative decay X j + X→ Xk + X → 2 X + hν
Collision-induced emission X∗ + X →2 X + hν
Transfer of electronic excitation X∗ + AB → X + AB∗
Dissociative excitation transfer X∗ + AB → X + A∗ + B + (εkin )
Exciplex formation X∗ + AB → XA∗ + B
4.9.3.1 Chemi-Ionization and Penning Ionization

The processes shown in the first half of Table 4.13 lead to some form of ioniza-
tion. The necessary condition for these processes to occur is that the energy
of the excited state is greater than the ionization potential. Metastables (and
all higher excited states) of Ne and He have enough energy to ionize most
molecules, so we use Ne as an example (εm (Ne) > εi (AB)). If excited neon
atoms collide with molecule AB, there are several different processes that may
occur:
Ne∗ + AB → Ne + AB+ + e,
→ Ne + A+ + B + e,
→ NeAB+ + e,
→ NeA+ + B + e.
The first channel is known as Penning ionization, and the other three
are examples of the general term “chemi-ionization.” The second process is
dissociative ionization, the third is associative ionization, and the fourth is
one of the possible rearrangement channels. The metastables Ne(3 P2,0 ) have
energy of 16.62 eV and 16.72 eV, respectively, which is above the ionization
limit of the Ar(1 S0 ) ground state, 15.76 eV. Thus the combination of Ne and Ar
is an excellent example of a mixture that demonstrates Penning and chemi-
ionization, and the data for this combination are shown in Figure 4.38. In
addition to Penning and chemi-ionization, we show elastic scattering
Ne(3 P2 ) + Ne(1 S0 ) → Ne(3 P2 ) + Ne(1 S0 )

Ne (3P0.2) + Ar (1S0)
-15 Qi (Penning)
10
Cross section (cm )
2
Qi (associative)
-16
10
Qelas × 0.01
-17
10
-3 -2 -1 2 3
10 10 10 1 10 10 10
Relative energy (eV)
FIGURE 4.38
Cross sections for the Penning ionization, chemi-ionization, and elastic scattering.
for comparison. One can see that Penning ionization is 10 times more efficient
at relatively low energies (which is relevant for applications, because standard
collisions occur at room temperature, which is equivalent to 24 meV). At
higher energies the cross section for the associative process becomes equal
to and even larger than that for the Penning ionization. The cross section
for the elastic process is another 10 times larger. Thus a small mixture of Ne
in Ar will have a very efficient ionization, and as such it is used in many
applications.
PROBLEM 4.9.1
By using the cross sections from Figure 4.38 and thermal Maxwellian distributions
for Ne(3 P0,2 ) and Ar(1 S0 ) for room temperature (300 K), calculate the rate coefficient
k(cm3 s−1 ) for Penning and chemi-ionization.
Exercise 4.9.2
Explain how the ionization rate in the mixture of Ar and Ne exceeds that of
either Ar or Ne as shown in Figure 4.39.
For the purpose of this evaluation, it is better to represent ionization through
the number of ionizations per unit time, ki = vd (α/E) · (E/N), where vd is the
drift velocity. The production of electrons is given by
dne ∗
= ki−Ar ne NAr + ki−Ne ne NNe + kPenn NAr NNe .
dt
Here, ki denotes the ionization rate coefficients for two gases and kPenn the
rate coefficient for Penning ionization. At the same time, the population of

-3 -2 -1
10 10 10 (= NAr / NNe+Ar)
-4
10
-1
3×10
-2
10
-5
10
α / E (V )
-1
-6
3×10
1
-3 (Ar 100 %)
10 -6
0
10
(Ne 100 %)
10 100 1000
E / N (Td)
FIGURE 4.39
Ionization coefficients normalized by the field (α/E[V −1 ] number of ionizations per unit of
potential difference) for neon–argon mixtures.
neon metastables is
∗
dNNe ∗ ∗
= k Ne ∗ ne NNe − kPenn NAr NNe − k Q−Ne NNe NNe .
dt
Here, k Ne ∗ is the electron excitation coefficient, k Q is the two-body quenching
coefficient of the neon metastable by the ground-state neon (quenching is
discussed in the next section), and we have safely assumed that Penning
ionization dominates the quenching of neon metastables by argon. Under
stationary conditions this leads to
∗ k Ne ∗ ne NNe
NNe = .
kPenn NAr + k Q−Ne NNe
The effective growth of ionization is therefore
dne
(= Ri ne ) = ki−Ar ne NAr + ki−Ne ne NNe (4.114)
dt
k Ne ∗ ne NNe
+ kPenn NAr .
kPenn NAr + k Q−Ne NNe
We can see that the effective ionization will be modified by a term that depends
strongly on the abundance of the constituents in the mixture. As Ar density in-
creases, the third term will increase substantially. Subsequently, Ar will affect
the energy distribution through its inelastic processes with lower thresholds

+
He
25
* 1
He 2S
20 + 2 -7
3 Cd D3/2 (τ ∼ 10 s)
2S 2
Penning D5/2
Energy (eV)
15 ionization λ = 325 nm 441.6 nm2

P3/2
-9
(τ ∼ 10 s) 2
P1/2
+
Cd 1
10 S1/2
1 1
1S S
He Cd
FIGURE 4.40
Energy diagram of He and Cd.
than those for Ne. However, when the argon density becomes large enough
that the Penning term exceeds the quenching by Ne, the result will simply
be
dne
(= Ri ne ) = ne (ki−Ar NAr + ki−Ne NNe + k Ne ∗ NNe ). (4.115)
dt
For small E/N the ionization rate Ri will be equal to the excitation coefficient
for metastable Ne. Thus, a way to measure the excitation rate could be to
determine the ionization rate for the Penning mixture.
Exercise 4.9.3
Explain the principle of operation of the He–Cd laser (its energy diagram is
shown in Figure 4.40).
In collisions of metastable He(3 S1 , 1 S0 ) with the ground state of Cd(1 S0 ), it is
possible to ionize and excite Cd to high-energy levels Cd+ (2 D3/2,5/2 ), which
are then more strongly populated than the lower-energy levels Cd+ (2 P3/2,1/2 ).
Because inverse distribution of the population is achieved, it is possible to
achieve lasing on the two lines that originate from the transitions (λ = 325
nm and 441.6 nm).

TABLE 4.14
Quenching of Excited States Often Used in Plasma Diagnostics
Molecule Feed Excitation Rad-Lifetime Emission Quenching Coefficient
(Excited State) Gas Energy (eV) τrad (ns) hν (nm) kq (cm3 s−1 )
Ar(2 p1 ) Ar 13.57 21 750.4 0.16 10−10

Ar(3 p9 ) Ar 14.57 90 419.8 2.7 10−10
N2 (C 3 u ; v = 0) N2 11.03 36 337.1 1.5 10−11
N+
2 (B u ; v = 0)
2 N2 18.75 63 391.4 4.5 10−10
H(n = 3) H2 12.1 15 656.3 3.8 10−9
H(n = 3) Ar 12.1 15 656.3 4.6 10−10
H(n = 3) CH4 12.1 15 656.3 3.5 10−9
H(n = 2) H2 10.2 15 656.3 2.5 10−9
H2 (d 3 u ; v = 0) H2 13.8 39 601.8 3.3 10−9
CH(A2 ) CH4 12.08 565 431.1 2.0 10−11
CH(B 2 ) CH4 12.43 357 387.1 6.8 10−11
O2 (3 p 3 P) O2 10.98 34 844.6 6.3 10−10
O2 (3 p 5 P) O2 10.73 34 777.4 10.8 10−10
4.9.4 Collisions of Slow Neutrals and Rate Coefficients

4.9.4.1 Quenching and Transport of Excited States
There are a large number of processes in addition to radiative transitions that
may depopulate excited states. We have considered several processes leading
to ionization. However, there are numerous processes that lead to different
excited states of the same or other molecules, some of which are listed in
Table 4.13. Therefore, if one wishes to study particular channels and their
products, data for each channel are required. However, often we just want to
follow the kinetics of the excited state and thus need to evaluate total losses
of population. In such cases we can define the total quenching coefficient kqm
(of the excited state m) of Nm with a radiative lifetime τm = 1/Am . The kinetics
of one excited state m for steady-state conditions is then described by
dNm
= 0 = kexc
m
ne N − Am Nm − kqm Nm N,
dt
which can be solved as
m
kexc ne N
Nm = ,
Am + kqm N
and therefore
m
kexc ne N
Nm Am = Im = kqm N
.
1+ Am
If one observes the intensity of emission (Im = Am Nm ) for a fixed source

of excitation, the intensity will decrease with pressure. The inverse of the
intensity (Im ) is plotted as a function of N (the so-called Stern–Volmer dia-
gram) and the intersection of the extrapolated line with zero pressure gives

the value of intensity without the effect of quenching (which may be used to
exc
determine the excitation rate km for the state m from the intensity of radia-
tions).
Several quenching processes are listed in Table 4.14. Quenching of excited
molecules may be induced by collisions (collision–radiative decay) that transfer
molecules to a different excited state that subsequently decays radiatively.
This can happen both for short- and long-lived states. In the latter case the
most important transitions are made to nearby resonant states. Thermal col-
lisions with molecules may provide enough energy to make the transition to
nearby excited levels (which are within several times of kT from the orig-
inal level). This may happen within one multiplet. Emission of a photon,
even from a metastable state, may be induced by collisions (collision-induced
emission). During collisions a temporary dipole moment is induced (or a tran-
sient molecule is formed), which provides a mechanism for radiative decay.
Special attention should be paid to the quenching of metastable levels.
Two-body quenching consists mainly of collision-radiative and collision-
induced emission processes, as outlined above. Three-body quenching mainly
leads to the formation of eximers [A∗2 ]:
A∗ + A + A = A∗2 + A(→ A + A + hν).
Metastable atoms and molecules, regardless of whether they have a short

effective lifetime due to quenching, may still diffuse to the walls of the dis-
charge vessel, where they will have plenty of opportunity to be de-excited.
Thus we can calculate the losses to the walls by using diffusion coefficients.
However, at moderate pressures (0.5–5 Torr and higher), gas phase collisions
will dominate metastable loss. Typically, diffusion coefficients for rare gas
metastables are approximately a factor of two smaller than those of the
ground-state atoms. In Table 4.15 we show data for the quenching and diffu-
sion of rare gas metastables within their parent gases.
PROBLEM 4.9.2
Determine the effective lifetimes of the metastable levels of rare gases at 1 Torr and at
10 Torr. Compare them with natural radiative lifetimes.
TABLE 4.15
Radiative Lifetimes, Two- and Three-Body Quenching, and Diffusion
Coefficients of Rare Gas Metastable States in Their Parent Gases
Metastable Natural Lifetime Two-Body Three-Body Diffusion
State τrad (s) k2b 3 −1 k3b 6 −1 ND(1018 cm−1 s−1 )
q (cm s ) q (cm s )
He(21 S0 ) 0.02 6 10−15 — 15

He(23 S1 ) 9000 — 2 10−34 16
Ne(33 P2 ) 24 1.6 10−15 5 10−34 5
Ar(43 P2 ) 56 1.2 10−15 9 10−33 1.8
Kr(53 P2 ) 85 2.3 10−15 4.2 10−32 1.03
Xe(63 P2 ) 56 5.8 10−15 4.9 10−32 0.55

TABLE 4.16
Rate Coefficients for Quenching of Rare Gas
Metastables by Molecular Gases (10−10 cm3 s−1 )
Metastable N2 H2 O2 NO CH4 CF4 SF6 Cl2
He(21 S0 ) 1.7 0.49 5.8 4.3 3.7 — 6.7 —

He(23 S1 ) 0.7 0.32 2.1 2.4 1.4 — 2.6 —
Ne(33 P2 ) 0.84 0.5 — — — — 4.1 —
Ar(43 P2 ) 0.36 0.66 2.1 2.2 3.3 0.4 1.6 7.1
Ar(43 P0 ) 0.16 0.78 2.4 2.5 5.5 0.4 1.7 7.2
Kr(53 P2 ) 0.04 0.3 1.6 1.9 3.7 — 1.8 7.3
Xe(63 P2 ) 0.19 0.16 2.2 — 3.3 — 2.8 6.5
In addition to parent gas quenching, even the smallest amount of im-

purities leads to quenching of the metastables. Two types of processes are
responsible, depending on the relative positions of the energy levels: Pen-
ning ionization and all possible channels of chemi-ionization and excitation
transfer (sensitized fluorescence), including chemi-luminescence. These pro-
cesses are very effective, and even small amounts of impurities may change
the effective lifetimes significantly. In Table 4.16 we show a survey of total
quenching coefficients for combinations of rare gas metastables with impuri-
ties. It should be remembered that excitation transfer from metastables (with
or without additional ionization) may be used as a way to detect the rare
gas metastables. For example, the N2 (C 3 u ) state almost coincides with the
Ar metastable and has been used as a probe of Ar metastable population
densities. Similarly, N+2 (B u ) may be used to probe the concentration of He
2
metastables [10].
PROBLEM 4.9.3
Calculate the values of the cross sections for quenching of rare gas metastables by
their parent molecules from the rate coefficients given in Table 4.15.
4.9.4.2 Kinetics of Rotational and Vibrational Levels

When molecules that are excited to some rotational or vibrational levels col-
lide, several processes may occur. First, it is possible that rotational and vibra-
tional excitation may be relaxed to translational energy (R-T, V-T). It is also
possible to have a transfer of rotational or vibrational quantum to the other
molecule (R–R, V–V). For example,
AB(v1 ) + AB(v2 ) = AB(v1 + 1) + AB(v2 − 1).
Thus, if we have a relatively large population of vibrationally excited

molecules, then higher-order vibrational levels may be populated step by step
to achieve nonthermal populations of very high levels. Due to a harmonicity,
transfer of energy will be effective between nearby levels v1 and v2 . If the
V–V processes are effective, the distribution of populations of vibrationally
excited levels will be different from the Boltzmann distribution.

If the effective temperature of the first vibrational level is Tv , which may be

quite different from the gas temperature Tg , in the absence of V–T processes
one obtains Treanor’s distribution:

Nv ε01 2ε01 δv
= exp − .
Nv+1 kTv kTg
Here, δ is the factor of unharmonicity and ε01 is the energy of the first vibra-
tional level. This distribution will have a plateau and even a maximum at
moderate vibrational quantum numbers. The actual vibrational distributions
should be calculated from kinetic equations, and they will be affected by the
electron excitation from the ground and from vibrationally excited levels and
by the V–T and V–V processes. However, excitation to higher levels will be
mainly due to V–V transfers. On the other hand, highly excited molecules
may dissociate to neutral fragments or have a very large cross section for
dissociative attachment.
4.10 Photons in Ionized Gases

Several processes involving photons may be of interest for nonequilibrium
plasmas. These include emission, absorption, photo-ionization, and photo-
detachment in the gas phase and photo-emission from surfaces. Photo-
detachment may be important in the kinetics of negative ions in the majority
of low-pressure nonequilibrium plasmas in electronegative gases. It can also
be used as a diagnostic technique to establish the identity of negative ions,
because the threshold for detachment is a unique characteristic of each ion.
In Figure 4.41 we show the cross section for photo-detachment of negative
ions in oxygen (O− and O− 2 ). Photo-detachment may lead to nonlinear phe-
nomena in plasmas. Photo-ionization is not important under most conditions;
photo-emission from surfaces are discussed in the next section. Absorption,
followed by re-emission, are now discussed. Radiation originating from plas-
mas consists of lines from well-defined transitions between discrete levels or
molecular bands, and there are also broad continua that may originate either
from transitions to repulsive lower states of molecules or from free–free tran-
sitions. The relevant spectral range starts in the ultraviolet and extends to the
far infrared. In any case, line radiation and molecular bands are the dominant
features in low-temperature plasmas.
4.10.1 Emission and Absorption of Line Radiation

The kinetics of emission and absorption of line radiation can be represented
by Einstein’s probability coefficients. If we take two discrete levels with an
energy difference ε21 = ε2 − ε1 , the lower with density N1 and the upper
with density N2 , then the probability of spontaneous emission per unit time

–
O
Cross section (10 -18 cm 2) 6
–
2 O2
-20
3.5×10 cm 2
0
0.5 1.0 1.5 2.0 2.5 3.0
Photon energy (eV)
FIGURE 4.41
Cross sections for photon-induced detachment for negative ions in oxygen.
is defined by coefficient A21 , and thus the total number of emitted photons is
equal to A21 N2 . If the transition 2 → 1 is the only radiative decay of the level
2, then the lifetime of the excited level is equal to τ2 = A−1 21 . In the presence
of radiation with a density (per unit frequency) of line radiation of transition
2 → 1 equal to ρ(ν21 ), it is possible that either the radiation will be absorbed
by the atoms in the lower level with probability B12 ρ(ν21 ) or a stimulated
emission by the photons of exactly the same frequency will occur from the
atoms in the upper levels, with probability B21 ρ(ν21 ).
Exercise 4.10.1
Assuming thermal equilibrium for a system with two excited levels 2 and 1,
find the relationships between Einstein’s coefficients.
In thermal equilibrium, the number of emissions (spontaneous plus stimu-
lated) must equal the number of absorptions, so we have
A21 N2 + B21 ρ(ν21 )N2 = B12 ρ(ν21 )N1 .
At the same time, the populations of two excited states can be determined
from the Boltzmann distribution
N2 g2
= e hν21 /kT .
N1 g1
Therefore, we can solve the equation for radiation density as
AN2 A21 A21

ρ(ν21 ) = = = g2 hν21 /kT .
B12 N1 − B21 N2 N1
B12 N − B21 B12 g1 e − B21
2

Using Planck’s law
8π hν21 3
/c 3
ρ(ν21 ) = , (4.116)
e hν21 /kT − 1
we obtain the following relations:
g1 B12 = g2 B21 , (4.117)
3
8π hν21
A21 = 3
B21 . (4.118)
c
Stimulated emission reduces absorption, and if we have population inver-
sion, that is, if the density of the excited state exceeds the density of the
lower state, we will have overall amplification of emission. This is a possi-
bility in nonequilibrium discharges and is the fundamental principle behind
gas lasers. As discussed above, the kinetics of excited states will be more
complicated due to possible nonradiative transitions such as quenching.
The probability of spontaneous emission can be related to the oscillator

strength f 21 through the formula
g1 2π e 2 ν 2
A21 = f 21 . (4.119)
g2 ε0 mc 3
The absorption of radiation can be represented by an absorption coefficient,
and it will be associated with the attenuation of a beam of radiation of intensity
Iν = cρν as it passes through a gas of absorbers
1 dIν hν
− dν = [B=12 dN1 − B21 dN2 ] kν dν, (4.120)
Iν dx c
where dN refers to a subgroup of atoms that are absorbed within the fre-
quency interval dν, and kν is the absorption coefficient. If we integrate over
all frequencies, we have

hν hν c 3 A21 g2 g1 N2
kν dν = [B12 N1 − B21 N2 ] = 3 g
N1 1 − . (4.121)
c c 8π hν21 1 g2 N1
In addition to the integration over all frequencies covered by the line profile,
one can independently perform integration along the length of the absorbing
medium (where one can expect variation in the density of absorbers). Thus
the solution to Equation 4.121 for a uniform absorbing medium of length l is
I (l) = I (0)e −kν l , (4.122)
and therefore one can determine the absorption coefficient kν from the exper-
imental measurement of absorption profiles. One can then define the optical
l
depth as τν = 0 kν dx = kν l. The frequency profiles of lines can be determined

by Doppler broadening with the profile given by

mc 2 ν − ν0 2
kν = k0 exp − , (4.123)
2kT ν0
with the half width equal to

2ν0 2RT ln 2 −7 T
δν D = = 7.18 × 10 ν0 . (4.124)
c M M
The line profile may also have a Lorentzian shape
e2 γ /4π
kν = N1 f 12 , (4.125)
(4πε0 )mc (ν21 − ν)2 + (γ /4π )2
where γ is the width of the line and (ν21 γ ). If we integrate the absorption
coefficient over all frequencies (i.e., over the line profile) we obtain [11]
∞
e2
kν dν = N1 f 21 .
0 4ε0 mc
PROBLEM 4.10.1
Evaluate the profile due to Doppler broadening, and determine k0 and the width of
the line by assuming a thermal distribution of absorbers at temperature Tg .
PROBLEM 4.10.2
Plot the Lorentzian and Doppler profiles (normalized to the same integral value) with
the same half width and compare the center of the line and the width.
4.10.2 Resonant Radiation Trapping

Absorption of photons is most probable between the same levels that partic-
ipated in the emission (i.e., if there is a resonance). Therefore, if the gas den-
sity is sufficient, resonant levels (those with strong transitions to the ground
level) will have their emission easily reabsorbed. As a result the resonant
radiation may be trapped. In other words, the resonant radiation does not
leave the discharge vessel immediately but rather diffuses out through many
absorptions and re-emissions, hence the effective lifetime, and consequently
the populations of the excited levels, may increase. Typically for few Torr,
the populations of rare gas resonant levels are the same order of magnitude
as the populations of metastables. The effective transition probability in the
case of radiation trapping is denoted g Ajk , where g is the trapping factor
(g ≤ 1). There are approximate theories and Monte Carlo simulations that es-
timate these factors, and these may even be important for pressure ranges of
interest in many nonequilibrium plasmas. However, these effects are unim-
portant for the transitions between two excited states that are frequently used
in diagnostics.

Calculations of the resonant radiation trapping are performed, either by

Monte Carlo simulations or by using some form of the propagator method
where an equation similar to the Boltzmann equation is solved. In these
studies a typical approximation is that of the complete redistribution of fre-
quency, whereby the frequencies of the initial and re-emitted photons are not
correlated. Approximate analytic solutions were found for g by Holstein. For
Doppler broadening we have
• g= √1.875
k0 l πln(k0 l)
, for parallel plane geometry, and
• g= √ 1.6
k0 R πln(k0 R)
, for cylindrical geometry,
and for collisional broadening we have
• g = √1.15 , for parallel plane geometry, and

π kρ l
• g = √1.115 , for spherical geometry.

π kρ R
These formulae have been updated in numerous studies. However, it is

important to state that g may be quite small even at pressures below 0.1 Torr.
This may reduce the transition probability so that the lifetime and, conse-
quently, the densities of the resonant levels are almost equal (less than an
order of magnitude smaller) to those of the metastable levels. Under these
circumstances the resonant states may behave in a similar fashion as metasta-
bles and may even suffer similar quenching and reaction processes.
4.11 Elementary Processes at Surfaces

With the exception of most astrophysical and atmospheric conditions, gas
discharge plasmas are always maintained and enclosed inside a vessel. There
are a large number of processes that may occur between active particles of the
plasma and the reactor wall and these may strongly affect the plasma itself
[12]. For example, the main sustaining mechanism of low-pressure gas dis-
charges in DC and low-frequency power sources is the surface production of
electrons at the cathode by ions and other active particles from the discharge.
In Figure 4.42 we give a schematic representation of a few of the many
possible interaction processes between the plasma and the wall of a reactor
or other surfaces such as electrodes. As can be seen, it is possible to have
photo-emission of electrons from the surface; to have a secondary electron
production by an ion or a metastable molecule impact; to have sputtering of
surface material by an ion or fast atom bombardment; to have ion neutraliza-
tion or charging at the surface; to have chemical reactions at the surface; to
have adsorption or desorption; and finally, if the surface is heated or subjected
to very high fields, to have emission of charged particles from the surface.

Source
Insulator Reactor wall
B
+ M e A
+
hv
B
A
+ e e
plasma +
e A
W
H
– * e
A
H W
e
Metal
FIGURE 4.42
A schematic representation of plasma surface interactions in a typical plasma chamber.
Because production and loss of charged particles are of key importance for
the maintenance of plasmas, we primarily consider the production and the
loss process of charged particles on a surface. On the other hand, for plasma
technologies we require additional processes such as chemical reactions at the
surface, physical sputtering, and implantation. A special form of interaction
between a plasma and a surface is the formation of the sheath that protects
plasma from losing the charged particles to the conductive surfaces. Thus,
the properties of charged particles arriving at the surface and reaching the
plasma from the surface will strongly depend on the sheath potential. We
should note that inasmuch as we mainly have surface modification in mind
when we discuss plasma processing, the understanding of plasma surface
interaction is critical.
4.11.1 Energy Levels of Electrons in Solids

Interaction of electrons that are bound to a nucleus forming an atom is repre-
sented by a bound potential with a negative value and discrete levels. Due to
the laws of quantum physics, each of the electrons must possess a different
state, and each must fill a different energy level. For example, for the sodium
atom Na, the electron configuration is
Na : (1s)2 (2s)2 (2 p)6 (3s).
The potential energy of these bound electrons is negative, as shown in

Figure 4.43a and approximated by the Coulomb potential
11e
V(r ) = − . (4.126)
4π ε0r

Elementary Processes in Gas Phase and on Surfaces
(a) (b)
ε=0
ε=0
bound electron
3p
3p band
3s
Potential energy of
2p 3s
2s 2p
11e
V(r) = 2s
4πε0r
1s
Na level
r 1s
Na Na Na Na
a a a
FIGURE 4.43
Energy levels of a single atom of Na (a) and of atoms in a solid (a one-dimensional array of Na atoms) (b).
125
Consider a sodium lattice consisting of N atoms in one-dimensional space.

The bound potential between two adjacent atoms with a distance a is
11e 11e
V(r ) = − − (4.127)
4π ε0r 4π ε0 (a − r )
11ea
=− . (4.128)
4π ε0r (a − r )
When the electrons of the outermost shell (3s) with energy level εk keep the
following relation at the maximum value of V(r ),
|V(a /2)| > (εi − εk ), (4.129)
then each of (3s) electrons is not potentially restricted in one atom but free
between N atoms (see Figure 4.43b). Here, εi is the ionization energy of Na.
When the external field is applied from both sides of the one-dimensional
lattice, (3s) electrons can move to the neighboring atoms as free electrons.
The electrons that may freely leave atoms will combine their levels into a
band because electrons are fermions. The band is known as the conduction
band, and the electron as the conduction electron (or valence electron). If no
levels populated by electrons satisfy the criterion 4.129 then the material is
nonconducting.
Now we should consider a lattice of atoms and the potential at the edge
of the solid, that is, at the surface (see Figure 4.44). The last atom facing
the surface will have a potential, which increases toward the vacuum level
unperturbed by the neighbours, and conduction electrons cannot leave the
surface potential barrier. To determine this potential more accurately we ob-
serve the force acting on one electron that is leaving the surface. The attraction
of one charged particle at a distance z from the surface may be described by
using the image charge whereby the whole surface is replaced by the opposite
charge at a distance −z.
Within the conduction band in the metal of temperature T, the electrons
will fill the lowest states according to the Fermi–Dirac distribution,
1
f (ε) = ε−ε , (4.130)
exp kT
F
+1
where ε is the energy level of the conduction band, and ε F is the Fermi level.
At zero temperature (T = 0), the maximum energy level of electrons is equal
to the Fermi energy ε F . Electrons at the top of the conduction band require
energy equal to φ to leave the surface, and this energy is known as the work-
function. In Table 4.17 we show the work-functions and the Fermi energies of
several metals. Alkaline metals have the lowest work-function, typically on
the order of 2 eV.

ε V(z)
distance from surface
W 0 z
φ
W
εF
V(z)
N(ε)
0
state density
a a a
crystal lattice vacuum
surface
FIGURE 4.44
Potential energy and state density of bound electrons in a metal.
4.11.2 Emission of Electrons from Surfaces

4.11.2.1 Photo-Emission
Photons may provide sufficient energy to bound electrons to allow them to
leave the surface, as proposed by Einstein in his explanation of the photo-
electric effect 100 years ago (1905). In photo-electric emission, the energy of
the photon hν is spent on electrons that overcome the potential barrier φ to
leave the metal, and the remaining energy is converted to the kinetic energy
εk of the free electron:
1 2
εkinmax = mv = hν − φ. (4.131)
2 max
TABLE 4.17
Work-Function of Typical Metals
Metal εF (eV) φ(eV) Metal εF (eV) φ(eV) Metal εF (eV) φ(eV)
Li 4.7 2.38 Cu 4.3 2.80 In — 3.8

Na 3.1 2.35 Sr — 2.35 Ga — 3.96
K 2.1 2.22 Ba — 2.49 Tl — 3.7
Rb 1.8 2.16 Nb — 3.99 Sn — 4.38
Cs 1.5 1.81 Fe — 4.31 Pb — 4.0
Cu 7.0 4.4 Mn — 3.83 Bi — 4.4
Ag 5.5 4.3 Zn — 4.24 Pt — 5.3
Au 5.5 4.3 Cd — 4.1 W — 4.5
Be 14.3 3.92 Hg — 4.52 Ta — 4.2
Mg — 3.64 Al 5.6 4.25 Ni — 4.6

Al
-1
10
Sus
Au
Photo-electric yield (e / h )
Au covered with C (500A)

-3
10
1.0
-5
10
Reflectivity
-1
10
-2
10
-7 10
5 15 25
Photon energy (eV)
FIGURE 4.45
Photo-emission yield for some metals as a function of photon energy.
As a result, there is a cutoff frequency (wavelength) of a photon that can

produce an electron regardless of the intensity (i.e., number of photons) of
the light (hν − φ = 0). That is, the kinetic energy of electrons does not de-
pend on the photon flux but on the frequency. The photo-electron yield is
the probability that one photon will induce emission of one electron, and it
depends on the material and the surface condition. Data for several metals
are shown in Figure 4.45. Photo-emission was recently recognized as one of
the major mechanisms of secondary electron production in the maintenance
of a high-pressure DC discharge.
PROBLEM 4.11.1
Calculate the cutoff wavelengths for the materials in Table 4.17. Compare them with
the well-known wavelengths of light sources, such as a mercury lamp, nitrogen lamp,
and rare gas lamps.
4.11.2.2 Thermionic Emission

When a metal is heated, there is a chance that some of the bound electrons will
gain sufficient energy to leave the metal surface. Integrating the Fermi–Dirac

distribution, one obtains the Richardson–Dushman equation

φ
J = AT exp −
2
, (4.132)
kT
where J is the current density, A = 4π emk 2 / h 3 = 1.2 × 106 Am−2 K −2 , and k

is the Boltzmann constant. In practice, either tungsten (coated by thorium)
or oxide (e.g., BaO) cathodes are used as thermionic sources of electrons. A
similar law may be defined for the thermal emission of positive ions that may
also be released from the surface. Heating may also lead to the evaporation
of neutral atoms.
4.11.2.3 Field-Induced Emission

If we apply an electric field to the surface of the metal, the effective potential
of the bound electrons is a combination of the attractive potential due to the
image charge (at distance 2z) and the potential due to external field E:
e2
V(z) = − − e E z. (4.133)
16π ε0 z
√
The bound potential has a negative minimum at zm = e/16π ε0 E that is
equal to

e eE
Vmax = − (4.134)
2 π ε0
and is thus considerably smaller than the work-function φ. This potential is

shown in Figure 4.46. Effectively, a barrier with the width wb is formed. There
is therefore a chance that electrons on the left-hand side with energy ε will
pass through the barrier by tunneling, even though their energy is smaller
than the height of the barrier. The probability of tunneling through a square
barrier of the width wb and height V0 for an electron of mass m is equal to

ε ε 2 {2m(V0 − ε)}1/2
T ∼ 16 1− exp − wb . (4.135)
V0 V0 h̄
The current density of the field-induced emission is given by the Fowler–

Nordheim equation
J = αe E 2 exp(−βe /E), (4.136)
where αe and βe are constants that are best determined by fitting the ex-
periments and both depend on the work-function and Fermi energy of the
metal. Fields of the order of E ∼ 106 Vcm−1 should be required for significant
emission. However, even smaller fields are sufficient to produce the emission.
For example, the microroughness of the surface produces local fields that are

(a) V(z)
z
eEz
εF
Wb
Tt
bound electron emitted electron

(b) ϕ(z)
(stational wave) (exponential

decay ) (transmitted wave)
FIGURE 4.46
Field-induced emission is a result of the changes in potential due to external field: (a) formation
of the potential barrier, (b) wave function of electrons.
very high. In addition, very thin layers of surface oxide may have deposited
charges on them, which produce very high fields and lead to enhanced emis-
sion.
PROBLEM 4.11.2
Prove that the effective reduction of the work-function under the influence of the
external field E is given by Equation 4.134.
Exercise 4.11.1
Explain the principle of operation of the scanning tunneling microscope
(STM) as shown in Figure 4.47.
The STM consists of a very sharp probe made of W or Pt that is brought close to
the surface at a distance of 0.1 to 0.5 nm. We label the quantities related to the
sample substrate (metal) with subscript s and those related to the probe with
the subscript p. A potential is applied between the probe and the sample, and
the current is measured; therefore, STM may be applied only to conductive
materials. The direction of current depends on the relative work-functions of
the probe tip and of the substrate. The principle of the STM is described in
Figure 4.47.
There are two possible modes of operation of an STM. It is possible to
operate in a constant height mode, where the tip is held at a fixed position

Ef
φp – eV
tip metal
φp
z=0
eV
Vb(z)
Tt
z = zs
φs
Ef
(b)
surface (z = zs)
tip (z = 0)
tunneling
V
current
z=0
(a)
FIGURE 4.47
Schematic diagram of the scanning tunneling microscope (STM) and the relevant potentials.
while variations in the current are recorded. Consider what happens as the
probe tip gradually approaches the sample. At a distance of ∼1 nm (1) un-
der the condition, φ p < φs , the conduction electrons tunnel from the tip to
the sample substrate until both Fermi levels become equal, and the electric
dipole with potential Vb = φs − φ p is formed between the two surfaces. When
the external potential V is applied at the probe with respect to the sample, the
tunneling current is detected at the probe. By scanning the substrate surface
in the constant height mode, the quantum state close to the Fermi level is
observed in the two-dimensional image.

(a) (b)
ε=0
φ
Excited wb
state W
εi
εex
Metal
Ground
state
Atom (or Ion) Atom (or Ion)
FIGURE 4.48
Schematics of a potential energy diagram for interaction of a particle with the surface (a) at a
distance and (b) in closest contact.
The other mode in which STMs operate is the fixed current mode, where
the position of the tip is modified to keep the current constant. As current due
to tunneling decays exponentially within the barrier, the measured current
is extremely sensitive to the distance between the tip and the substrate. That
also means that most of the current flows through the atom at the top of
the tip, and thus resolution may be very high (subatomic) provided that the
movements of the tip are controlled to within one hundredth of an atomic
diameter.
4.11.2.4 Potential Ejection of Electrons from Surfaces by Ions

and Excited Atoms
When an ion or long-lived excited molecule approaches a surface, there is a
chance that an electron will be ejected from the surface (potential ejection),
depending on the energy levels of the incident particle and of the surface.
This phenomenon is different from the emission of electrons by the kinetic
energy of an incoming particle that deposits sufficient energy onto the surface.
We consider more closely the potential ejection mechanism. The schematic of
such a particle–surface encounter (ion arriving at the surface) is given in
Figure 4.48. There are four possible processes:
(1.) Resonance neutralization
(2.) Resonance ionization
(3.) Auger de-excitation
(4.) Auger neutralization
These four processes are discussed in detail in this section. We first explain
the Auger processes. As a particle with excitation and ionization energy
εex and εi approaches the surface with the work-function φ, the potential
barrier between the particle and the surface is reduced in height and espe-
cially in the width wb , becoming so small that it assumes atomic dimensions.

e
emission
ε=0 ε=0 ε=0
φ
εi
εex
+ *
A A A
(initial) (transient) (final)
FIGURE 4.49
Electron ejection in the case of (εi − εex ) ≥ φ.
There are two interesting situations for the relative magnitude of potentials as
follows:
i (εi − εex ) > φ. In this case, the interaction during approach is as

shown in Figure 4.49. As an ion A+ approaches the surface, the first
step (1) will be a transfer of one electron from the metal to the ion,
and a long-lived excited state (metastable) A∗ will be formed as the
intermediate state:
(1) A+ + Nem → A∗ + (N − 1)e m ,
where N is the number of conduction electrons e m in the metal. De-

excitation of the metastable will lead to the emission of excitation
energy εex, which may be absorbed by one electron in the conduction
band, which will then be free to leave the surface e(↑) as
(3) A∗ + Nem → A + e(↑) + (N − 1)e m .
The maximum kinetic energy of the released electron e(↑) will occur
if it originates from the top of the conduction band, that is,
εmax = εex − φ,
whereas the minimum kinetic energy of the secondary electron

occurs when electrons from the bottom of the conduction band are
released:
εmin = εex − W.
The production of electrons by incoming ions on the electrodes is
one of the major mechanisms to sustain low-temperature plasmas.
A similar process of electron ejection (only step (3)) will occur if
a metastable atom approaches the surface. Metastable atoms are

e
emission
ε=0 ε=0 ε=0
φ
εi
εex
* +
A A A
(initial) (transient) (final)
FIGURE 4.50
Electron ejection in the case of (εi − εex ) < φ.
perturbed close to the surface and may also experience radiative

transitions to the ground state.
ii (εi − εex ) < φ. In this case, the most energetic electrons in the con-
duction band cannot tunnel to the excited state of the incoming ion
at the surface. However, if an excited atom approaches the surface,
then it is possible that the electron will tunnel through the atom
to the metal and into an unoccupied state in the conduction band
(resonance ionization) as shown in Figure 4.50. The ion is formed
as the intermediate state:
(2) A∗ + Nem → A+ + (N + 1)e m .
This process is followed by a transition of the conduction electron

to the ground state of the ion A+ (Auger neutralization):
(4) A+ + Nem → A + e(↑) + (N − 2)e m ,
with a release of energy equal to εi − ϕ and the subsequent release

of an electron from the conduction band through absorption of the
irradiated energy. In this case, the maximum and the minimum
energies of secondary electrons are:
εmax = εi − 2φ and, εmin = εi − 2W.
Ejected electrons typically have broad distributions with a mean

energy of several eV.
PROBLEM 4.11.3
When He*(3 S1 ) is approaching the W-surface, estimate the maximum kinetic energy
of electrons ejected from W by Auger potential ejection.

0.32
0.28 +
Ne
0.24
Yield (molecule / ion) He
+
0.20
0.16
0.12 +
Ar
0.08 +
Kr
0.04 +
Xe
0
0 200 400 600 800 1000
Ion energy (eV)
FIGURE 4.51
Secondary electron yields for rare gas ions colliding with a clean tungsten surface.
The efficiency of absorbing the released energy may be limited and there
is always a chance that the electron may dissipate its energy on the way
to the surface. Ejected electrons may be trapped at the surface by adsorbed
molecules or may be reflected from gas molecules back to the metal surface.
Therefore, there is a limited efficiency of secondary electron release per in-
cident ion, which is described by the secondary electron yield. Examples of
secondary electron yields for rare gas ions colliding with a tungsten surface
are given in Figure 4.51.
4.11.3 Emission of Ions and Neutrals from Surfaces

From the viewpoint of the incident particles, the two processes in Figures 4.49
and 4.50 are important. In the latter case there is a chance that the excited
atom will be released on a surface after the transfer of the electron and before
the subsequent Auger neutralization. In that case the process would be an
ionization process. In addition, there is an efficient resonant neutralization
and ionization.
One should pay attention to the fact that processes shown in Figure 4.49,
where the bound electron makes a transition to the excited state of an ion and
then de-excites and the process where the transition is made directly to the
ground state of an ion (see Figure 4.50), are not the same and may differ in the
energy distribution of ejected electrons. These processes may be adiabatic or
nonadiabatic. In the former case, there is no energy exchange between the nu-
clear motion and the electrons. Adiabatic processes dominate at low energies.

4.11.3.1 Surface Neutralization

For those modeling a plasma–surface interaction, it may be interesting to
calculate the probability that a beam of ions will be neutralized upon surface
collisions. Neutralization may proceed through three processes: the radiative,
the resonant, and the Auger neutralization. The first is highly unlikely as the
transitions require 10−8 s, which is many orders of magnitude longer than
the time of interaction. Both the resonant and Auger processes are probable.
Because the transitions of electrons required to neutralize the ion take place
through barriers, the tunneling effect will determine the probability, and the
basic parameters will be associated with the time spent at the place closest
to the surface. This, in turn, will lead to three principal parameters: the per-
pendicular velocity V⊥ and two quantities, A(s−1 ) and a (cm−1 ), which are
associated with the barrier height and width. Thus the probability of neutral-
ization of a beam is given in a form typical for tunneling processes:

A
Pneut (V⊥ ) = 1 − exp − . (4.137)
a V⊥
One should pay attention to the fact that ions may be neutralized while ap-
proaching the surface and also while leaving it. Thus the probability will
depend on both the initial and final velocities (Vi and Vf ), which may be
quite different:

A 1 1
Pneut (V⊥ ) = 1 − exp − + . (4.138)
a Vi⊥ Vf ⊥
If there is a penetration of the solid, the formula becomes more complex and
will include probabilities for CT in binary encounters inside the solid.
4.11.3.2 Surface Ionization

Neutral atoms on a surface may be ionized depending on their molecular
properties and the work-function of the material. Regarding the relative mag-
nitude of the ionization potential εi , the electron affinity εa of the molecule,
and the work-function φ, there are three distinct situations:
i. εi < φ: The production of positive ions.

This is the situation in which ions with small ionization potential,
such as alkaline metal atoms (Li, Na, K, Rb, Cs), interact with sur-
faces with relatively high work-function, and therefore a large per-
centage of the neutral atom density (na ) is converted into positive
ions (n p ). The degree of ionization at the surface ηi is given by
−1
np e(εi − φ)
ηi = = 1 + w0 exp , (4.139)
n p + na kT
where T is the surface temperature, and w0 is a parameter depend-

ing on the nature of the surface.

(a) (b)
0 0
r r
φ εi φ
εa( εi)
A
A
Metal Metal
FIGURE 4.52
Potentials for surface ion production: (a) εi < φ: production of positive ions, (b) εa > φ: produc-
tion of negative ions.
ii. εa > φ: The production of negative ions.

If, however, we combine atoms with very high electron affinity (F2 ,
Cl2 , Br2 , I2 ) with surfaces that have low work-function, negative
ions of the molecule will emerge from the surface.
iii. εi > φ > εa .
If the work-function of the surface is between the electron affinity
and ionization energy of the molecule, then the surface will leave the
molecule as neutrals without exchange of electrons. For example,
hydrogen molecules H2 (εi = 13.6 eV, εa = 0.7 eV) will scatter as
neutrals off the tungsten surface (φ = 4.5 eV).
PROBLEM 4.11.4
Explain the processes of (i) surface ionization and (ii) negative ion formation. The
potential energy curves and the corresponding transitions are shown in Figure 4.52.
Even in case (iii) it is possible to have surface ionization if a strong electric

field is applied externally. Typically, fields on the order of 108 V cm−1 are
required to achieve significant field ionization at the surface. One should
bear in mind that this process is different from field emission from the solid
and is also different from possible ionization of a free atom or molecule. In
this case the field allows the process of surface ionization by affecting the
potential of particles close to the surface.
Exercise 4.11.2
A hydrogen atom resides at the surface of tungsten (W; φ = 4.5 eV) in an
external field 2 × 108 V cm−1 . Determine the critical distance zc required to
achieve field ionization.
Because the work-function of tungsten is φ = 4.5 eV and the ionization energy
of the ground-state hydrogen atom H(2 S1/2 ;1s) has εi = 13.6 eV, we have
εi−φ>0. Hence the surface ionization without an external field is not possible
(see the dashed line in Figure 4.53). If we add the electric field, the potential

15 Ea = 2 × 108 V cm-1
10
VHE
5
Potential energy (eV)
zc
0 4 8 12
0 z (Å)
φ
-5
εF
VHO
W
-10
1s
-15
surface H
FIGURE 4.53
Potentials for a hydrogen atom on a W surface: (a) without external field (dashed line), (b) with
the external field (solid line).
of the bound electrons will be modified and the condition for ionization will
be given by
(eE zc + φ − εi ) > 0. (4.140)
At that point the electron from the neutral H may tunnel through the barrier
and a positive ion H+ may emerge from the surface. Therefore the condition
is achieved for zc > (εi − φ)/eE. The probability of tunneling will strongly
depend on zc and thus ionization will occur for a narrow range of distances
as the atom approaches and leaves the surface.
4.11.4 Adsorption
Atoms and molecules may be bound to surfaces in a process known as ad-
sorption. Molecules may be physisorbed (weakly bound) or chemisorbed
(strongly bound). In the former case there is a Van der Waals interaction be-
tween the molecule and the surface, whereas in the latter case the interaction is
chemical, with an electron exchange, and therefore the value of the interaction
potential is much larger. The schematic potential curves for both processes

Potential energy
2O + M
B C A εdiss
O2 + M
0
εphys z
εchem
metal surface
FIGURE 4.54
Potential energy diagrams for adsorption of an oxygen molecule on a metal surface.
are shown in Figure 4.54. Physisorption allows formation of several layers of

gas molecules and is nonspecific. In this case, the heat of adsorption is less
than 2 ∼ 3 times the heat of evaporation. In the case of chemisorption the
reaction is very specific. It may involve dissociation of molecules, and only
a monolayer is usually formed. The heat of adsorption is more than 2 ∼ 3
greater than the heat of evaporation.
We describe the processes at the surface by using the surface coverage
number of adsorbed molecules

θ=
maximum uptake of molecules
and the sticking probability:
rate of adsorption
s= .
rate of collisions with the surface
The rate of adsorption should be proportional to the rate of collisions with
the surface ka , the pressure, and the fraction of vacant sites, ka p(1 − θ), and
the rate of desorption should be proportional to the surface coverage and the
specific rate of desorption kd :
kd θ.
Under thermal equilibrium these two processes are balanced:

θ ka
= p = bp, (4.141)
1−θ kd

where b is the equilibrium adsorption constant. As a result we obtain a cor-

relation between the pressure of the external gas and the surface coverage as
bp
θ= . (4.142)
1 + bp
In the limit of low pressures, we obtain θ∼ p, and at high pressures the correct
limit θ ∼ 1 is obtained. At intermediate pressures one obtains θ ∼ p 1/n , which
is known as Freundlich’s isotherm. We typically have n = 2.
The lifetime of particles at the surface (i.e., the residence time) is another
way to describe the kinetics of desorption:
τs = τ0 exp(ε phys /kT), (4.143)
where τ0 ∼ 10−12 s. Thus, one may estimate the residence times for small and
large values of activation energy.
A number of processes proceed on the surface. Molecules may be dissoci-
ated upon adsorption, or atoms may recombine at the surface. In the case of
hydrogen it was found that recombination of atoms into molecules leads to a
release of excited molecules in high vibrational levels and that the properties
of these molecules are dependent on the surface temperature. The recombi-
nation of two atoms of nitrogen to form a molecule at the surface leads to
formation of a molecule in the metastable A3 state. However, at the sur-
face, the metastable molecule is able to release an electron through the Auger
process. Thus, surface recombination of atoms may provide a source of sec-
ondary electrons in gas discharge. In nitrogen discharges, it was found that
the atoms may remain in afterglow for very long periods of time, on the or-
der of hours. This is shown by the Lewis–Rayleigh afterglow, which is due
to emission of excited states formed in the gas phase. The recombination of
these atoms at the surface releases the electrons and makes it much easier to
start the discharge even hours after the previous discharge.
There are two possible mechanisms for surface reactions. In the Langmuir–
Hinshelwood mechanism, the reaction proceeds through adsorption of two
atoms on the neighboring sites. In other words, two atoms that have been
adsorbed come in proximity at the surface and react. In the Eley–Rideal mech-
anism, a single gas phase atom reacts with the atoms already adsorbed on the
surface. As a result, the two processes have a different pressure dependence.
References
1. French, A.P. and Taylor, E.F. 1978. An Introduction to Quantum Physics. Boston:
MIT Press.
2. Eisberg, R. and Resnick, R. 1974. Quantum Physics of Atoms, Molecules, Solids,
Nuclei and Particles. New York: John Wiley & Sons.

3. Massey, H.S.W., Burhop, E.H.S., and Gilbody, H.B. 1969. Electronic and Ionic
Impact Phenomena, Vol 1–4. Oxford: Oxford Clarendon Press.
4. Massey, H. 1979. Atomic and Molecular Collisions. London: Taylor and Francis.
5. McDaniel, E.W. 1989. Atomic Collisions: Electron and Photon Projectiles. New York:
John Wiley & Sons.
6. O’Malley, T.F., Rosenberg, L., and Spruch, L. 1962. Phys. Rev. 125:1300.
7. Opal, C.B., Peterson, W.K., and Beaty, E.C. 1971. J.Chem. Phys. 8:4100–4106.
8. Phelps, A.V. 1994. J. Appl. Phys. 76:747–753, J. Phys. Chem. Ref. Data 20:557.
9. Phelps, A.V. 1990, 1991. J. Phys. Chem. Ref. Data 19:653; J. Phys. Chem. Ref. Data
20:1339.
10. Golde, M.F., 1976. Gas Kinetics and Energy Transfer. Chem. Soc. London 2:123–
174.
11. Thorne, A.P., 1988. Spectrophysics. London: Chapman and Hall.
12. Thomas, E.W. 1984. In Data Compendium for Plasma-Surface Interactions, Special
Issue, Eds. R.A. Langley, J. Bohdansky, E. Eckstein, P. Mioduszevski, J. Roth, E.
Taglauer, E. W. Thomas, H. Verbeek, and K.L. Wilson, Nuclear Fusion, IAEA,
Vienna.

5
The Boltzmann Equation and Transport
Equations of Charged Particles
5.1 Introduction
The Boltzmann equation is the basis for the kinetics of charged particles and
neutral molecules in nonequilibrium and low-temperature plasmas that are
maintained and controlled under a short-range two-body collision with feed
gas molecules [1]. Of course, gas molecules have their own quantum states
through their electronic configuration in a low-temperature plasma. Each of
the atoms and molecules with a different quantum state exhibits unique char-
acteristics in collision with an electron or ion. This provides a very interest-
ing world of low-temperature plasmas with different functions reflected by
the quantum structure even in a fixed external plasma condition. Therefore,
it is essential to understand the terms of the collisions—including various
kinds of two-body collisions—between electron and neutral molecules in the
Boltzmann equation.
The Boltzmann equation is the transport equation in phase space, and
usually we transfer the Boltzmann equation to the kinetic transport equation
in real space in order to be readily able to calculate the plasma structure [2].
Accordingly, we derive the transport theory in detail. The purpose of this
chapter is to provide a foundation, that is, the kinetic transport theory, and par-
ticularly electron transport theory, for the modeling and design of nonequi-
librium and low-temperature plasmas.
5.2 The Boltzmann Equation

The Boltzmann equation is associated with the famous Austrian physicist
Ludwig Boltzmann (1844–1906), who published his kinetic treatment of gases
in 1872. The form of the Boltzmann equation has not changed since its
inception, but the equation for electrons requires a more detailed form of
the collision integral between the electron and molecule.
143

5.2.1 Transport in Phase Space and Derivation of the Boltzmann Equation

The subject of this chapter is the theoretical representation of the transport of
ensembles of charged particles in external electric (E) and magnetic (B) fields,
which can be expressed as
F = mα(r, t) = eE(r, t) + ev × B(r, t), (5.1)
where α is the acceleration due to the force, and m and e(> 0) are the mass
and charge of the charged particle. Here, the number density of the swarm of
the charged particle n is a function of position r and time t:
n = n(r, t). (5.2)
In addition, each of the particles has a different velocity v and the velocity
changes over time. The transport of the charged particle is therefore desig-
nated in phase space, that is, both in the configuration (real) space and in
the velocity space. The number of charged particles within a small volume
defined by dr around a position r and within dv around a velocity v is desig-
nated dn, which can be defined as
dn = g(v, r, t) dv dr. (5.3)

In Equation 5.3, in particular, when dn = 1 is satisfied, g(v, r, t) is the
probability density function and is called the velocity distribution function
in physics (see Figure 5.1). The change of coordinates in the phase space after
a short interval dt as a result of the effects of both the self-motion and the
external forces is r (= r + dr) = r + vdt, and v (= v + dv) = v + αdt. That
is, the number of charged particles changes in a small time increment dt to
dn = g(v , r , t + dt)dv dr . (5.4)
v-space
g(v, r, t)
r-space
Vx
n(r, t)
x E(r, t) Vz
dr
E(r, t) z
FIGURE 5.1
Example of the density distribution of charged particles n(r, t) in real space and the velocity
distribution g(r, v, t) in phase space.

The Boltzmann Equation and Transport Equations of Charged Particles 145
Here, the small element of phase space dv dr after dt is connected to the initial
element dv dr by a Jacobian
∂(r , v )
dv dr = dv dr, (5.5)
∂(r, v)
and from Equations 5.3 and 5.4 we obtain
∂ x ∂ x ∂ x ∂ x
∂x ∂y . .
∂v y ∂vz
∂y
∂x . . . . .

∂(r , v ) . . . .

= = 1. (5.6)
∂(r, v) . . . .

∂v y . . .
∂x
∂v
z . . ∂vz
∂x ∂vz
The net change of the number of the particles during dt is given as
dn − dn = [g(v + αdt, r + vdt, t + dt) − g(v, r, t)] dv dr

∂ ∂ ∂
= +v· +α· g(v, r, t) dv dr dt, (5.7)
∂t ∂r ∂v
where we have used the Taylor expansion in v, r, and dt in order to obtain
the second line of this equation. Equation 5.7 is equal to zero without the
inflow or outflow from the element dv dr due to a binary collision between
the particle and the gas molecule. However, in general, we have to consider
collisions that result in a change of the number of particles in any cell of the
phase space, and in that case Equation 5.7 is not equal to zero. The effect
of collisions is proportional to dv dr dt and is described by using a function
J (g, F ) as dn − dn = J (g, F )dv dr dt. Here, J (g, F ) is named the collision
operator (or collision integral), and F = F (V, r, t) is the velocity distribu-
tion function of gas molecules. Usually gas molecules have a Maxwellian
velocity distribution function at a gas temperature of Tg ; in other words, gas
molecules are in thermal equilibrium. In the realm of the so-called low-
temperature or nonequilibrium plasmas, however, the charged particle is not
in thermal equilibrium, and the velocity distribution satisfies
∂ ∂ ∂
g(v, r, t) + v · g(v, r, t) + α · g(v, r, t) = J (g, F ). (5.8)
∂t ∂r ∂v
Equation 5.8 demonstrates that the temporal change of the velocity distri-
bution function of charged particles is, in principle, caused by three factors:
diffusion in configuration space (second term on the left-hand side), accel-
eration by external fields (third term on the left-hand side), and collisions
(right-hand side).
PROBLEM 5.2.1
Confirm Equation 5.6 (or dv dr = dv dr ) by using the condition ∇v · F = 0
for external forces in Equation 5.1.

Equation 5.8 is known as the Boltzmann equation. The physical require-

ment of the independence of events among diffusion, external acceleration,
and collision limits the interaction of a charged particle with another charged
particle or gas molecule to a short-range interaction, that is, a two-body
encounter.
On the other hand, there exists a system in which a long-range interac-
tion of charged particles (i.e., a many-body collision) is dominant in plasmas.
This occurs in a highly ionized plasma in which the Coulomb collision, repre-
sented by a large amount of small-angle scattering in collisions, is dominant.
Then, Equation 5.8 is the Fokker–Planck equation. Under specific conditions
the effect of collisions may be neglected, and then we have the so-called Vlasov
or collisionless Boltzmann equation:
∂ ∂ ∂
g(v, r, t) + v · g(v, r, t) + α · g(v, r, t) = 0. (5.9)
∂t ∂r ∂v
5.3 Transport Coefficients

When we integrate the velocity distribution function of charge particles
g(v, r, t) in velocity space we obtain macroscopic transport parameters in
configuration space. By definition, the number density n(r, t) is determined
from

n(r, t) = g(v, r, t)dv. (5.10)
In general, the ensemble average of a quantity A(v, r, t) in velocity space is

given by

A(v, r, t)g(v, r, t)dv
A(r, t) = . (5.11)
g(v, r, t)dv
Under the influence of an external field there appears a directional velocity
v(r, t) in charged particles, such that the mean velocity is not equal to zero
(v =

0). It is convenient to divide the mean velocity into two components:
the directional component vd = v(r, t) and the random component vr , as
v(r, t) = vd (r, t) + vr (v, r, t), (5.12)
where vr = 0. The ensemble average of the velocity, often labeled as fluid
velocity, is given by

1
v(r, t) = vg(v, t)dv, (5.13)
n
and the mean (kinetic) energy is

1 1 2 1
ε(r, t) = mv g(v, r, t)dv = m vd2 + vr2 . (5.14)
n 2 2

In the case where the number density has a spatial nonuniformity, n(r, t),
the velocity distribution function g(v, r, t) may be represented in powers of
the density gradient ∇r n as
g(v, r, t) = g 0 (v, t)n(r, t) + g1 (v, t) · ∇r n(r, t)

+ g2 (v, t) ∇r2 n(r, t) + · · · . (5.15)
Here the coefficients in the density expansion, gk , are tensors of rank k and
indicates a kfold scalar product. The functions independent of position are
normalized as

1; k = 0
gk (v, t)dv = (5.16)
0; k =

0 .
The expansion 5.15 is known as hydrodynamic expansion, and its conver-

gence is usually assumed in the Boltzmann equation of electrons (hydrody-
namic approximation) [3]. In situations in which the convergence of 5.15 is not
assured, the velocity distribution function is labeled as nonhydrodynamic. In
hydrodynamic approximation the fluid velocity 5.13 is written as

1 1
v(r, t) = vg 0 (v, t)dv + vg1 (v, t)dv · ∇r n(r, t). (5.17)
n n
On the right-hand side the first term is the drift term due to the influence of
external fields, and the second term is the diffusion term (tensor) under the
density gradient, ∇r n(r, t). Thus we have the following definition of the drift
velocity and diffusion tensor, respectively:

1
vd (t) = vg 0 (v, t)dv (5.18)
n

1
D(t) = vg1 (v, t)dv. (5.19)
n
When the external electric field is applied in the z-direction, for exam-
ple, the diffusion tensor 5.19 has two components, the longitudinal (DL ) and
transverse (DT ), and the tensor is represented by a diagonal matrix

DT 0 0

D(t) = 0 DT 0 . (5.20)

0 0 DL
Each of the components of the diffusion tensor is represented by the ensemble

average of the first-order velocity distribution g1 (v, t) as

1 1
DT (t) = vx gx1 (v, t)dv = v y g 1y (v, t)dv, (5.21)
n n

1
DL (t) = vz gz1 (v, t)dv. (5.22)
n

PROBLEM 5.3.1
In the case where there is a magnetic field as well as an electric field, the drift tensor
has six components. Derive each component of the drift velocity.
The total pressure tensor is defined as

PT (r, t) = m vvg(v, r, t)dv,

= m vr vr g(v, r, t)dv + mnvd vd ,
= mn(r, t)vr vr + mn(r, t)vd vd . (5.23)
The first term on the right-hand side (pressure tensor) shows the contribu-
tion of the thermal or random velocity, and the second term is due to drift
(directional) velocity. As random velocity is usually much larger than the
directional velocity, the total pressure tensor PT is almost isotropic with
equal components in the x-, y-, and z-directions, and each component (scalar
pressure p) is written according to

Px 0 0

1
P = 0 Py 0 , p = px = p y = pz = mnvr2 . (5.24)
3
0 0 Pz
The energy flux vector is defined as the energy carried by charged particles
that cross unit area per unit time

1 2
Q(r, t) = mv vg(v, r, t)dv. (5.25)
2
Q represents the third-order velocity moment and
v 2 v = [(vd + vr ) · (vd + vr )] (vd + vr )

= vd2 vd + vd2 vr + 2vd vd · vr + 2vd · vr vr + vr2 vd + vr2 vr

= vd2 vd + 2vd · vr vr + vr2 vd + vr2 vr .
Here we combine Equations 5.14 and 5.24 to obtain

1
Q(r, t) = mnv 2 v
2
1
1
= mn vd2 + vr2 vd + mnvr vr · vd + mn vr2 vr

2 2
1 2
= n(r, t)ε(r, t) vd + P · vd + mn(r, t) vr vr . (5.26)

2
The first term in Equation 5.26 is the energy transported by drift, the second
is the work performed, and the third is the energy transported by the random
motion. This third term is called the thermal flux q and is defined as
1
q(r, t) = mn(r, t) vr2 vr . (5.27)

2

PROBLEM 5.3.2
The diffusion coefficients, DL and DT , are obtained by the trace of the trajectory of
particles in the configuration space in accordance with
1d
DL (t) = (z(t) − z(t) )2

2 dt
= z(t)vz (t) − z(t) vz (t) . (5.28)
1d
DT (t) =
x(t)2 + y(t)2
4 dt
1
= (x(t)vx (t) + y(t)v y (t) ). (5.29)
2
In particular, in addition to E = Ek the magnetic field B = Bi is present, electrons
have a complicated behavior and each of the diagonal components of the diffusion
tensor in E × B fields is expressed as
1d
DE =
(< z2 > − < z >2 ),
2 dt
1d
DE×B = (< y2 > − < y >2 ),
2 dt
1d
DB = (< x 2 > − < x >2 ).
2 dt
In addition to the above components, the Hall diffusion coefficient, representing the
sum of the off-diagonal components of the diffusion tensor, will appear and is given by
1d
(< y >< z > − < yz >).
DHall =
2 dt
Derive the above relationship from the statistical physics.
Exercise 5.3.1
An isolated swarm of charged particles with delta function in space is injected
into gases under a uniform dc-field. After a finite collisional relaxation time,
the swarm attains a quasi-equilibrium with time independent transport coef-
ficients, vd and D. Discuss the transport and the coefficient in an initial stage
before the relaxation.
Higher-order spatial moments about the center of mass of the swarm
< r >,
Mk (t) = < (r(t) − < r(t) >)k >. (k = 3, 4..)
have a finite magnitude in the initial phase of the steep density distribution.
In particular, M3 and M4 represent skewness and kurtosis of the density dis-
tribution, and the 3rd- and 4th-order transport coefficients, D3 and D4 are
related to the skewness and kurtosis, respectively,
1 d
D3 = M3 (t),
3! dt
1 d
D4 = M4 (t) − 3(M2 (t))2 .
4! dt

Then, in an initial stage of the transport, the fluid velocity expressed by the
drift velocity and diffusion tensor in Equation 5.17 will be extended to a form
including the 3rd- and 4th-order transport coefficients.
5.4 The Transport Equation

When we start from the general form of the Boltzmann Equation 5.8, multiply
it by an arbitrary function A(v, r, t), and integrate it in velocity space, we
obtain the transport equaton of particles in configuration space:

∂g(v, r, t) ∂g(v, r, t)
A(v, r, t) dv + A(v, r, t)v · dv
∂t ∂r

∂g(v, r, t)
+ A(v, r, t)α · dv = A(v, r, t)J (g, F )dv. (5.30)
∂v
The first term may be expanded as

∂g ∂ ∂A
A dv = Agdv − gdv
∂t ∂t ∂t

∂ ∂A
= (n(r, t)A ) − n(r, t) , (5.31)
∂t ∂t
and the second term is equal to

∂g ∂ ∂
Av · dv = · (vAg)dv − · vA gdv,
∂r ∂r ∂r

∂ ∂
= · (n(r, t)vA ) − n(r, t) · vA . (5.32)
∂r ∂r
The third term can be expanded as

∂g ∂ ∂
Aα · dv = · αAgdv − · αA gdv
∂v ∂v ∂v

∂ ∂
= · (n(r, t)αA ) − n(r, t) · αA
∂v ∂v

∂
= −n(r, t) α · A . (5.33)
∂v

Now we substitute Equations 5.31 to 5.33 into Equation 5.30 and obtain
the general continuity equation as

∂ ∂A(v, r, t) ∂
(n(r, t)A(v, r, t) ) − n(r, t) + · (n(r, t)vA(v, r, t) )
∂t ∂t ∂r

∂ ∂A(v, r, t)
− n(r, t) · vA(v, r, t) − n(r, t) α ·
∂r ∂v

= A(v, r, t)J (g, F )dv. (5.34)
5.4.1 Conservation of Number Density

If we use A = 1 and replace it in the general form 5.34, the second, fourth,
and fifth terms are equal to zero. The integral on the right-hand side is written
as n R 0 by using the production rate R 0 , and the general moment equation
reduces to
∂ ∂
n(r, t) + · (n(r, t)v ) = ne (r, t)R 0 (r, t). (5.35)
∂t ∂r
This equation represents the conservation of particle number density. The first
term describes the temporal changes and the second describes the changes
due to flow of the number density of charged particles. The term on the right-
hand side gives the production and loss due to the collision between the
particle and the neutral molecule.
PROBLEM 5.4.1
Show that for the number density continuity of electrons, the term R 0 on the right-
hand side of Equation 5.35 is expressed through the ionization rate Ri and electron
attachment rate Ra as
R 0 (r, t) = Ri (r, t) − Ra (r, t).
5.4.2 Conservation of Momentum

If we substitute A = mv in Equation 5.34, the second and the fourth terms
are equal to zero, and we obtain

∂ ∂ ∂
(mn(r, t)v ) + · (mn(r, t)vv ) − mn(r, t) α · v = m vJ dv.
∂t ∂r ∂v
(5.36)
According to Equation 5.12 the velocity v is divided into two parts, the direc-
tional and thermal velocities, vd and vr . The first term on the left-hand side
of Equation 5.36 is replaced by using v = vd and Equation 5.35 as
∂ ∂ ∂ ∂
mn(r, t) vd + mvd n(r, t) = mn(r, t) vd − mvd · (n(r, t)vd )
∂t ∂t ∂t ∂r
+ mne (r, t)vd R 0 ,

the second term of Equation 5.36 is expanded as
∂ ∂ ∂
· (mn(r, t)vdvd ) + · (mn(r, t)vr vr ) = vd · (mn(r, t)vd )
∂r ∂r ∂r
∂ ∂
+ mn(r, t) vd · vd + · (mn(r, t)vr vr ),
∂r ∂r
and finally, the third term is written as

∂
−mn(r, t) α · v = −n(r, t)e(E + v × B).
∂v
As a result, the conservation of momentum is expressed as

∂ ∂ ∂
mn(r, t) vd + mn(r, t)vd · vd + m · (n(r, t)vr vr )
∂t ∂r ∂r
= n(r, t)e(E + v × B) − mne (r, t)vd R 0 + mvJ . (5.37)
Here the third term on the left-hand side is expressed by using the pressure
tensor P (which is defined in Equation 5.24) as ∇r ·P.
PROBLEM 5.4.2
Show that the last term in the momentum conservation Equation 5.37 is approximated
by using the rate of the momentum transfer collision Rm as
mvJ = −mn(r, t)vd Rm . (5.38)
Exercise 5.4.1
Consider the case that the directional velocity of charged particles is inde-
pendent of space and time under a constant E without B. Then, derive the
current envd .
From Equation 5.37 we obtain the relation under Rm R 0
∂
mnvd Rm = enE − m · (nvr vr ).
∂r
The current is expressed by using the equation
∂
J = envd = enµE − e (Dn),
∂r
where µ and D are equivalent as µ = e/mRm and D = vr vr z /Rm = vr2 /3Rm ,
where the electric field is directed along the z-axis. Here µ is the mobility
and D is the diffusion coefficient and these equations are consistent with
simple expressions 2.4 and 2.5 in Chapter 2.

5.4.3 Conservation of Energy

If we replace A with mv 2 /2 in Equation 5.34, the second and the fourth terms
are equal to zero and the fifth term on the left-hand side is simplified as

1 ∂ mv 2
−n(r, t) [eE(r, t) + ev × B(r, t)] · = −n(r, t)eE · v ,
m ∂v 2
and as a result we obtain

∂ mn(r, t) 2 ∂ mn(r, t) 2
v + · v v − n(r, t)eE · vd
∂t 2 ∂r 2

mv 2
= J dv. (5.39)
2
The first term is the change in the mean energy ε(r, t), and the second is the
change in the energy flow. When we substitute Equation 5.26 in the second
term we obtain
∂ ∂ ∂
(n(r, t)ε(r, t) ) + · (n(r, t)ε(r, t) vd ) + · (P · vd )
∂t ∂r ∂r

∂ mv 2
+ · q − n(r, t)eE · vd = J dv. (5.40)
∂r 2
PROBLEM 5.4.3
Derive the collisional integral term in the energy conservation Equation 5.40 for
electrons that is given by

mv 2 2m
J dv = ε(r, t) Rm ne (r, t)
2 M
1/2
2ε
+ ε j R j + εi Ri − εNQa (ε) ne (r, t), (5.41)
m
where Rm , R j , Ri , and Ra are the rates of elastic momentum transfer, electronic

excitation, ionization, and electron attachment collisions, respectively. Also ε j , εi ,
and εa are the corresponding threshold energies.
5.5 Collision Term in the Boltzmann Equation

5.5.1 Collision Integral
So far we have not specified the exact structure of the collision term on
the right-hand side of the Boltzmann Equation 5.8. In Chapter 4 we have
discussed specific types of collisions and how to represent their quantitative

properties by the differential cross section σ (θ, φ; ε) and the integral cross
sections Q(ε). The velocity distribution of neutral gas molecules is represented
by F (V , r,t), where gas molecules have a mass M and velocity V before col-
lision. The charged particles of mass m and velocity v before collision are de-
scribed by the corresponding velocity distribution function g(v , r, t). When
we consider a small element of the phase space dv dr, then the number of the
charged particles that enter this element during a short time dt is equal to

J in = F (V , r, t)dV g(v , r, t)dv vγ σ (θ , φ ; vγ )d dr dt. (5.42)
v
During the same time dt, the number of charged particles leaving the same
element dv dr is

J out = F (V, r, t)dV g(v, r, t)dvvγ σ (θ, φ; vγ )ddr dt, (5.43)
v
where vγ and vγ are the relative speed between the gas molecule and the
charged particle just before and after the collision. The resulting change in
the number of charged particles in the element dv dr in time dt is
J dv dr dt = {J in − J out }. (5.44)
5.5.2 Collision Integral between an Electron and a Gas Molecule

In Figure 5.2 we show a schematic representation of a collision between
a charged particle (lowercase symbols) and a gas molecule (uppercase
symbols). The properties before collisions are denoted by (’). In the labo-
ratory frame of reference, a complete set of momentum and energy balance
equations should be solved to describe the collisions. However, the calcula-
tion is simplified in the center-of-mass frame where we subtract the motion
of the center of mass.
We have learned in Chapter 2 that even under a substantial electric field
in gases, the directional velocity of charged particles vd is much smaller than
the random component vr . This is particularly valid for electrons in a low-
temperature plasma, and we mainly concentrate on the electron transport
under these physical circumstances. In practice, this fact means that the elec-
tron velocity distribution is nearly spherical with a small perturbation in the
field direction. Thus it is convenient to expand the velocity distribution in
spherical harmonics in velocity space

g(v) = gmn (v)Ymn
e
(θ, ϕ), (5.45)
mn
where Ymne
(θ, ϕ) = Pnm (cos θ ) cos mϕ are the spherical harmonics by Morse–
Feshbach notation [4]; this expansion is discussed further in Section 5.6.1. The
spherical harmonic expansion helps us take care of the dependence on three
components of the vector v by taking advantage of the cylindrical symme-
try of the problem and turning the solution of the velocity dependence to a

v
m
m M
v' V'
M
V
(before collision) (after collision)
(a) position space
ω
V' v'
vg
V
M v M v
m+M r m+M r
(b) velocity space
FIGURE 5.2
The definition of a collision in (a) a laboratory frame of reference (real space) and (b) a center-of-
mass frame of reference (velocity space).
series of equations and corresponding distribution functions that depend on

a single variable, the magnitude of the velocity v. Furthermore, it is possible
to use the energy of particles as a variable. Solving the velocity dependence is
the critical step in solving the Boltzmann equation, and quite often the equa-
tions for the velocity (energy) dependence are referred to as the Boltzmann
equation.
5.5.2.1 Elastic Collision Term J elas

The elastic collision term is established by considering the momentum and
energy conservation equations. In this process we are able to take advantage
of the following simplifications:
i. The relative velocity between the electron and the gas molecule is
not changed after the collision;

ii. The velocity of gas molecules is much less than the velocity of elec-
trons, |V| |v|, and we represent the velocity distribution of gas
molecules with the number density N by F (V) = Nδ(V), where δ is
Dirac’s delta function; and
iii. From the momentum and energy conservation equations before and
after collisions, v and v, we have the relation
v 2 − v 2 2m
= (1 − cos ω), (5.46)
v 2 M+m
where ω is the scattering angle as defined in Figure 5.2b.
From Equation 5.46 we obtain dv /dv = (v /v)3 , and thus we replace the small
velocity element dv by dv. As a result, we obtain the elastic term of electrons
as follows:

J elas dv = Nδ(V )g(v )vγ σ (vγ , ω)d dv dV

V

− Nδ(V)g(v)vγ σ (vγ , ω)ddv dV
V

v 3
=N g(v )v σ (v , ω)d − g(v)vσ (v, ω)d dv. (5.47)
v
3

The velocity distribution g(v) is expanded in spherical harmonics

g(v) = gmn (v)Ymn
e
(θ, ϕ), (5.48)
mn
and in order to divide the differential cross section into spherical (σ0 ) and
forward-directed (σ1 ) parts, and the like, we expand the differential cross
section σ (θ, ϕ; ε) in terms of the Legendre function Pn (cos θ) according to

σ (v, ω) = σn (v)Pn (cos ω). (5.49)
n
By using the addition theorem for the spherical harmonics (see

Equation 5.73) we exchange θ to θ in Equation 5.47.

v 4
N 3 gmn (v )σn (v ) Ymne
(θ , ϕ )Pn (cos θ )Pn (cos θ)d
v mn n
n
(n − m)! e e
+2 Ymne
(θ , ϕ ) Ymn (θ , ϕ )Ymn (θ, ϕ)
m=1
(n + m)!

4
v 4π
+ Ymn 0
(θ , ϕ )Ymn
0
(θ, ϕ) d = N 3 g0n (v )σn (v ) Pn (cos θ)
v n
2n +1

n
4π

+ gmn (v )σ (v ) Y (θ, ϕ) .
e
m=1 n
2n + 1 mn

Here we consider
the orthogonality property of the Legendre function, 4π/
(2n + 1) = Pn (cos ω)Pn (cos ω)d (see Equation 5.74). Then the elastic colli-
sion term is given as
J elas
1 4
=N e
Ymn (θ, ϕ) v gmn (v )σ (v , ω)Pn (cos ω) − v 4 gmn (v)σ (v, ω) d
mn v 3
1 4
=N e
Ymn (θ, ϕ) v gmn (v )σ (v , ω) − v 4 gmn (v)σ (v, ω) Pn (cos ω)d
mn v 3

−N e
Ymn (θ, ϕ)gmn (v)vσ (v, ω) {1 − Pn (cos ω)} d. (5.50)
mn
When we compare the magnitudes of velocities before and after the elastic
collision by using Equation 5.46, we conclude that both of the velocities have
nearly the same magnitude, and v 2 = v 2 − v 2 . Therefore we are able to
expand the first term in the right-hand side of Equation 5.50 by the Taylor
series with the variable v 2 = v 2 − v 2 as
1 2m ∂[v 4 gmn (v)σ (v, ω)]
∼
J elas = N e
Ymn (θ, ϕ) (1 − cos ω) Pn (cos ω)d
mn
v M+m ∂(v 2 )

−N e
Ymn (θ, ϕ)gmn (v)vσ (v, ω) {1 − Pn (cos ω)} d. (5.51)
mn
In the particular case of the isotropic part of the distribution (m = 0, n = 0),

by considering P0 (cos ω) = 1, Equation 5.51 reduces to
2m 1 ∂ 4
00
J elas =N v g00 (v)Qm (v) . (5.52)
M + m 2v 2 ∂v
Here, Qm (v) is the momentum transfer cross section defined as

Qm (v) = (1 − cos ω)σ (v, ω)d.

For thermal equilibrium, electrons have the Maxwellian distribution. For zero
field, the temperature of electrons Te are equal to Tg when the gas temperature
is not zero; that is, Tg =

0. However, under all circumstances collisions with
thermal gas molecules lead to an additional velocity of electrons that is on
the order of O(V). Thus the total velocity has two components, one for zero
temperature and an additional component due to thermal motion (Tg =
0):
v = v0 + O(V).
For thermal equilibrium the distribution function has to be a Maxwellian,
00
and therefore J elas = 0. As a result, we obtain the order of V 2 according to the
principle of equipartitioning of energy:

2
2 ∂g00 v02 2kTg
O(V ) = −g00 v0 = . (5.53)
∂(v )
2 m

Finally, we obtain the elastic collision integral for a general case of nonzero
gas temperature (Tg =
0):

m 1 ∂ kTg ∂
00
J elas =N v 4
Qm (v) g00 (v) + g00 (v) . (5.54)
M + m v 2 ∂v mv ∂v
5.5.2.2 Excitation Collision Term J ex

Inelastic collisions may be represented in the Boltzmann theory under the
following assumptions:
i. The change of kinetic energy in jth inelastic collisions ε j usually
satisfies the relation ε j kTg with gas molecule F (V) = Nδ(V);
and
ii. From the energy conservation we have
1 2 1
mv = mv 2 + ε j . (5.55)
2 2
From Equation 5.55, that is, v dv = vdv, it is possible to transform the small
velocity element dv to dv by dv=(v/v)dv, and finally, we obtain the collision
operator for excitation

J exj dv = N g(v )v σ j (v , ω)δ(V )d dv dV
V

−N δ(V)g(v)vσ j (v, ω)ddv dV

1
=N g(v )v 2 σ j (v , ω)d − g(v)v 2 σ j (v, ω)d dv, (5.56)
v
which is further simplified by using the spherical harmonic expansion for
g(v),
J exj

1 2
=N Ymn (θ, ϕ) gmn (v )v σ j (v , ω)Pn (cos ω)d − gmn (v)v Q j (v) ,
2
v
(5.57)
where the integrated cross section is defined on the basis of the corresponding
differential cross section as

Q j (v) = σ j (v, ω)d.

5.5.2.3 Ionization Collision Term J ion

Ionization is a specific process by which a new electron is created in the
inelastic transition to the ionization continuum. Ionization by electrons is a
nonconservative process as the number density of electrons is not conserved.
With regard to the ionization collision operator, the following simplification
is made:

i. The change of kinetic energy in ionization εi usually satisfies the

relation εi kTg with gas molecule F (V) = Nδ(V).
ii. In the principle of indistinguishability, the incoming and newly pro-
duced electrons at ionization cannot be distinguished after collision,
but experimentally we know that there exist a pair of electrons with
high and low energy. We assume that the rest of the kinetic energy
is shared by two electrons according to the ratio (1 − ):. Then,
the velocity element can be expressed as
1 v 1 v
dv = dv2 or dv = dv1 . (5.58)
v (1 − ) v
iii. From the energy conservation, we have
1 2 1 1
mv = mv12 + mv22 + εi . (5.59)
2 2 2
On the basis of these assumptions and the spherical harmonic expansion, we
derive the collision integral for ionization as

1 2 1
J ion = N g(v )v σi (v , ω)d + N g(v )v 2 σi (v , ω)d
(1 − )v v

1
−N g(v)v 2
σi (v, ω)d
v

1 1
=N Ymn (θ, ϕ) gmn (v )v 2 σi (v , ω)Pn (cos ω)d
mn
v (1 − )

1 2
+ gmn (v )v σi (v , ω)Pn (cos ω)d − Ngmn v Qi (v) , 2
(5.60)

where the integrated ionization cross section is defined as

Qi (v) = σi (v, ω)d.
5.5.2.4 Electron Attachment Collision Term J att

Electron attachment is a very specific and nonconservative process wherein
an electron with energy ε is lost in collision with threshold energy εa . The
collision operator is simplified under J in = 0 as

J a tta = −Ng(v)v σa (v, ω)d

= −N e
Ymn (θ, ϕ)gmn (v)v Qa (v), (5.61)
mn
where the integrated attachment cross section is given by

Qa (v) = σa (v, ω)d.


5.6 Boltzmann Equation for Electrons

Charged particles in gases under the influence of an electric field are not in
thermal equilibrium but may be in quasi-equilibrium in the case where energy
gained from the field is dissipated in inelastic (and even elastic) collisions with
gas molecules. Under these conditions, transport coefficients of charged parti-
cles are well defined and the density gradient is relatively small. It is therefore
possible to separate the spatial and velocity parts of the distribution function.
This approximation is termed a “hydrodynamic description” or “density gra-
dient expansion” [3]. We have already mentioned that the directed velocity
of charged particles, and of electrons in particular, is much smaller than the
random velocity. This is the rationale for applying the spherical harmonic
expansion to simplify the theory in the velocity space. Both components of
the distribution function may be dependent on time and may have a complex
time dependence or relaxation to quasi-stationary conditions. In this section
we describe the detailed equations arising from the Boltzmann equation. The
first step is to apply both density gradient and spherical harmonics expan-
sions to the velocity distribution function g(v, r, t):

g(v, r, t) = g k (v, t) ⊗ (∇r )k n(r, t) (5.62)
k

= k
gmn (v, t)Ymn
e
(θ, ϕ) ⊗ (∇r )k n(r, t), (5.63)
k mn
where tensor gk (v, t) has components that are velocity distributions of differ-
e
ent order in density gradient expansion, and Ymn (θ, ϕ) are spherical harmonics
in Morse–Feshbach representation [4].
5.6.1 Spherical Harmonics and Their Properties

By definition, spherical harmonics have the following form:
1/2
2n + 1 (n − m)!
Ynm (θ, ϕ) = (−1)m Pnm (cos θ)e imϕ . (5.64)
4π (n + m)!
Their orthogonality is given by

Ynm1 1 (θ, ϕ)Ynm2 2 (θ, ϕ)d = δn1 ,n2 δm1 ,m2 . (5.65)
ϕ θ
The dependence on two angles (θ, φ) may be separated by using Ynm (θ, ϕ) =
(θ )m (ϕ), where
1
m (ϕ) = √ e imϕ (5.66)
2π

e
Y00 (θ, φ) P0 = 1
e
Y01(θ, φ) P1 = cosθ
0
Y11(θ, φ) sinθ sinφ
e
Y11(θ, φ) sinθ cosφ
FIGURE 5.3
Four spherical harmonics of the lowest order in Morse–Feshbach notation.
and
1/2
2n + 1 (n − m)!
(θ ) = (−1)m Pnm (cos θ), −n ≤ m ≤ n. (5.67)
2 (n + m)!
Here, Pnm (cos θ ) is the associated Legendre polynomial. The spherical har-
monics is written in Morse–Feshbach representation as
e
Ymn (θ, ϕ) = Pnm (cos θ) cos mϕ (5.68)
and
0
Ymn (θ, ϕ) = Pnm (cos θ) sin mϕ. (5.69)
e 0
A few terms of Ymn and Ymn are shown in Figure 5.3 and thus the veloc-
ity distribution function is expanded as g(v, t) = gmn (v, t)Ymn
e
(θ, ϕ) (see
Equation 5.63). Normalization of the distribution function requires the fol-
lowing integrals of spherical harmonics:

2π π 2
e
Ymn (θ, ϕ) or 0
Ymn (θ, ϕ) d
0
0
 4π (n + m)!
, n = 1, 2, 3, . . .
= 2(2n + 1) (n − m)! (5.70)
 4π , n = 0.
Recurrent relations for these functions are
(2n + 1) cos θ Ymn

e
(θ, ϕ) = (n + m)Ym(n−1)
e
(θ, ϕ) + (n − m + 1)Ym(n+1)
e
(θ, ϕ)
(5.71)

and
∂
(2n + 1) cos2 θ Ye (θ, ϕ)
∂ cos θ mn
= (n + 1)(n + m)Ym(n−1)
e
(θ, ϕ) − (n − m + 1)Ym(n+1)
e
(θ, ϕ), (5.72)
and these are also useful. Because the spherical harmonics in Morse–Feshbach
notation is connected to the associated Legendre polynomials by Ymn e
=
Pn cos mϕ and Y0n = Pn , Pn and Pn also satisfy the relations 5.71 and 5.72.
m e m
The addition theorem for spherical harmonics is given by
Pn (cos ω) = Pn (cos θ1 )Pn (cos θ2 )

n
(n − m)! m
+2 Pn (cos θ1 )Pnm (cos θ2 ) cos m(ϕ1 − ϕ2 )
m=1
(n + m)!
= Pn (cos θ1 )Pn (cos θ2 )
n
(n − m)! e
+2 Ymn (θ1 , ϕ1 )Ymn
e
(θ2 , ϕ2 ) + Ymn
0
(θ1 , ϕ1 )Ymn
0
(θ2 , ϕ2 ) .
m=1
(n + m)!
(5.73)
Some of the low-order Legendre and associated Legendre polynomials are

given in Tables 5.1 and 5.2, respectively. The orthogonality property of the
TABLE 5.1
Legendre Polynomials
Pn
P0 (cos θ ) 1
P1 (cos θ ) cos θ
P2 (cos θ ) (3 cos2 θ − 1)/2
P3 (cos θ ) (5 cos3 θ − 3 cos θ )/2
P4 (cos θ ) (35 cos4 θ − 30 cos2 θ + 3)/8
TABLE 5.2
Associated Legendre Polynomials
Pnm
P11 (cos θ ) sin θ

P21 (cos θ ) 3 cos θ sin θ
P22 (cos θ ) 3 sin2 θ
P31 (cos θ ) 3(5 cos2 θ − 1) sin θ/2
P32 (cos θ ) 15 cos θ sin2 θ
P33 (cos θ ) 15 sin3 θ

Legrendre function is given by


π  2
, n = n
Pn (cos θ )Pn (cos θ) sin θ dθ =
 2n 0+ 1 ,
(5.74)
0
n .
n=
If we expand a function f (cos θ) in Legendre polynomials as
∞

f (cos θ) = a n Pn (cos θ), (5.75)
n=0
the expansion coefficients a n (n = 0, 1, 2, . . .) are given by

2n + 1 π
an = f (cos θ)Pn (cos θ) sin θ dθ, (5.76)
2 0
and therefore we have

∞ π
2n + 1
f (cos θ ) = f (cos θ )Pn (cos θ ) sin θ dθ Pn (cos θ). (5.77)
n=0
2 0
If we have a function of both azimuthal and polar angles f (θ, ϕ), then the
expansion is in terms of Ymne
(θ, ϕ):

f (θ, ϕ) = e
a mn Ymn (θ, ϕ), (5.78)
where
π
2(2n + 1) (n − m)! 2π
a mn = f (θ, ϕ)Ymn
e
(θ, ϕ)d. (5.79)
4π (n + m)! 0 0
5.6.2 Velocity Distribution of Electrons

When we expand the velocity distribution function of electrons g(v, r, t) in
the Boltzmann Equation 5.8 by using the spatial density-gradient expansion
5.63, we obtain the chain of difference differential equations. The three lowest
terms, g 0 (v, t), g1 (v, t), and g2 (v, t), are defined by the equations
∂ 0 ∂ 0
g (v, t) + α(t) · g (v, t) + R 0 (t)g 0 (v, t) = J (g 0 , F ), (5.80)
∂t ∂v
∂ 1 ∂ 1
g (v, t) + α(t) · g (v, t) + R 0 (t)g1 (v, t)
∂t ∂v
= J(g 1 , F ) + vg 0 (v, t) − vd (t)g 0 (v, t), (5.81)
and
∂ 2 ∂ 2
g (v, t) + α(t) · g (v, t) + R 0 (t)g2 (v, t)
∂t ∂v
= J(g 2 , F ) + vg1 (v, t) − vd (t)g1 (v, t) + D(t)g 0 (v, t), (5.82)

where R 0 is the effective production rate of electrons usually equal to

Ri − Ra . vd and D are the drift velocity and the diffusion tensor, respectively.
Equation 5.80 describing g 0 (v, t) is independent of and uncoupled with other
distributions, g1 and g2 , and so on, and thus it enables us to solve the set of
equations successfully as g 0 (v, t) → g1 (v, t) → g2 (v, t).
5.6.2.1 Velocity Distribution under Uniform Number Density: g 0

The lowest-order term in the density-gradient expansion of the velocity dis-
tribution function g 0 (v, t) is calculated from Equation 5.80. g 0 is axisymmetric
with respect to the external field E or α = eE/m, and usually it is further ex-
panded in spherical harmonics for m = 0, that is, for Legendre polynominals
Pn (θ ), according to the equation

g 0 (v, t) = 0
g0n (v, t)Y0n
e
(θ, ϕ) = gn0 (v, t)Pn (θ ). (5.83)
n=0 n
Then, the second term in the right-hand side of Equation 5.80 is rearranged
by the aid of the addition theorem of the Legendre function as
∂ 0
α(t) · g (v, t)
∂v
∂g 0 sin2 θ ∂g 0
= αz (t) cos θ +
∂v v ∂ cos θ
∂gn0 (v) sin2 θ 0 ∂ Pn (θ )
= αz (t) cos θ Pn (θ ) + αz (t) gn (v) .
n
∂v n
v ∂ cos θ
In the above equation, cos θ Pn and sin2 θ∂ Pn /∂ cos θ are replaced by using
Equations 5.71 and 5.72, and the equation is simplified to
∂ 0
α(t) · g (v, t)
∂v
n+ 1 n
0
∂gn (v)
= αz (t) Pn+1 (θ ) + Pn−1 (θ )
n
2n + 1 2n + 1 ∂v
g 0 (v)
n(n + 1) n(n + 1)
+ αz (t) n
− Pn+1 (θ ) + Pn−1 (θ ) .
n
v 2n + 1 2n + 1
The above equation is further arranged after binding each term associated
with Pn (θ ) as
∂ 0
α(t) · g (v, t)
∂v

n ∂gn−1
0
(v) n + 1 ∂gn+1
0
(v)
= αz (t) +
n
2n − 1 ∂v 2n + 3 ∂v

(n − 1)n gn−1
0
(v) (n + 1)(n + 2) gn+1
0
(v)
− + Pn (θ )
2n − 1 v 2n + 3 v

n

∂ n−1

= αz (t) − 0
gn−1 (v, t)
n
2n − 1 ∂v v

n+1 ∂ n+2
+ αz (t) + 0
gn+1 (v) Pn (θ ). (5.84)
2n + 3 ∂v v
Finally, we derive a hierarchy of equations in terms of spherical harmonics

gn0 (v, t)(n = 0, 1, 2, 3, . . .) for the zeroth order in the density-gradient expan-
sion of the velocity distribution function of electrons g 0 (v, t):

∂ 0 n ∂ n−1
gn (v, t) + αz (t) − 0
gn−1 (v, t)
∂t 2n − 1 ∂v v

n+1 ∂ n+2
+ αz (t) + 0
gn+1 (v, t)
2n + 3 ∂v v
+ R 0 (t)gn0 (v, t) − J (g 0 , F ) = 0. (5.85)
In Figure 5.4 we show one example of the solution for g 0 (v) in Ar at

100 Td. Cylindrical symmetry is assumed and the exact solution is plotted
in velocity space (vz , vr ) where vr = vx2 + v2y . We also show the relative mag-
nitude of various components in the Legendre polynomial expansion, gn0 (v)
(n = 0, 1, 2, . . .).
5.6.2.2 Velocity Distribution Proportional to ∇r n(r, t): g 1

First, we divide the vector function g1 (v, t) into spatial components, gx1 , g 1y ,
and gz1 :
g1 (v, t) = gx1 (v, t)i + g 1y (v, t)j + gz1 (v, t)k. (5.86)
When the external force is along the z-axis, we have α(t) = −αz (t)k. From
Equation 5.81 we obtain
∂ 1 ∂ 1 J(gx1 , F )
g (v, t) αz (t) g (v, t) gx1 vx g 0
∂t x ∂vz x
∂ 1 ∂ 1
g y (v, t) + αz (t) g (v, t) + R 0 g 1y + vy g0 = J(g 1y , F ) . (5.87)
∂t ∂vz y
∂ 1 ∂ 1
g (v, t) αz (t) g (v, t) gz1 (vz − vd )g 0 J(gz1 , F )
∂t z ∂vz z
Here, the component along the external field (i.e., the z-axis), gz1 (v, t), has a
directional component of velocity vd , and the other two components vx and v y

g0 (υ )
3.3
Vr
(1
3.3
0
8
0.0 0.0
-3.3
cm
-1
Vz (1 08
cm s )
s
-1
)
g00(ε)
Distribution (arb.)
0.5
g20(ε) g0(ε) g0(ε)

3 1
0
0 5 10 15 20 25
FIGURE 5.4
Lowest-order solution for the velocity distribution function for electrons at 100 Td in Ar: (a) g 0 (v)
and (b) gn0 (v) (n = 0, 1, 2, . . .).
are symmetric to each other. The longitudinal component gz1 (v, t) is symmetric
with respect to the external field direction (i.e., the z-axis) and is expanded by
spherical harmonics for m = 0:

gz1 (v, t) = gz10n (v, t)Y0n
e
(θ, ϕ) = gz1n (v, t)Pn (θ ). (5.88)
n n
The term (vz − vd )g 0 (v, t) in Equation 5.87 is expanded by using Equa-

tion 5.71:
(vz − vd )g 0 (v, t)

= (v cos θ − vd )gn0 Pn (θ )
n

n n+1

= v Pn−1 (θ )gn + v
0
Pn+1 (θ )gn − vd gn Pn (θ )
0 0
n
2n + 1 2n + 1
n+ 1 n

= vgn+1
0
(v, t) + vgn−1
0
(v, t) − vd gn0 (v, t) Pn (θ ), (5.89)
n
2n + 3 2n − 1
where we use recurrent relations for Pn (θ ).

The governing equations for gz1n (v, t) are obtained by considering
Equation 5.89 and the master Equation 5.85 for g 0 :

∂ 1 n ∂ n−1
g (v, t) + αz (t) − gz1n−1 (v, t)
∂t zn 2n − 1 ∂v v

n+1 ∂ n+2 1
+ αz (t) + gzn+1 (v, t) + R 0 (t)gz1n (v, t) − J gz1 , F
2n + 3 ∂v v
n+1 0 n
=− vgn+1 (v, t) − vg 0 (v, t) + vd gn0 (v, t). (5.90)
2n + 3 2n − 1 n−1
The hierarchy of the first-order functions gz1n (v, t)(n = 0, 1, 2, . . .) is derived

numerically by using the known functions gn0 (v, t)(n = 0, 1, 2, . . .) under uni-
form density from Equation 5.85. One example of gz1 is shown in Figure 5.5.
The perpendicular components gx1 (v, t) and g 1y (v, t) are determined by
equations in the first and second row in Equation 5.87. The appropriate ex-
pansion for gx1 (v, t) is the spherical harmonics for m = 1 in Equation 5.68,

gx1 (v, t) = gx11n (v, t)Y1n
e
(θ, ϕ) = gx1n (v, t)Pn1 (θ ) cos ϕ. (5.91)
n n
The term αz (t)∂gx1 (v, t)/∂vz in Equation 5.87 is expanded as

∂ 1 ∂gx1 sin2 θ ∂gx1 ∂ϕ ∂gx1
αz (t) g (v, t) = αz (t) cos θ + +
∂vz x ∂v v ∂ cos θ ∂vz ∂ϕ
∂gx1n gx1 ∂ P 1 (θ )
= αz (t) cos θ Pn1 (θ ) cos ϕ + αz (t) n cos ϕ sin2 θ n .
n
∂v n
v ∂ cos θ
When we use Equations 5.71 and 5.72 to replace cos θ Pn1 (θ ) and sin2 θ∂ Pn1 (θ )/
∂ cos θ, we obtain
n n+1 1

∂gx1n (v, t)
= αz (t) cos ϕ P 1 (θ ) + P (θ )
n=1
2n + 1 n+1 2n + 1 n−1 ∂v
n2 (n + 1)2 1
1
gxn (v, t)
+ αz (t) cos ϕ − 1
Pn+1 (θ ) + Pn−1 (θ ) ,
n=1
2n + 1 2n + 1 v

gz1 (υ )
3.3
Vr
(1
3.3
0
0.0
8
0.0
-3.3
cm
-1
8 )
Vz (10 cm s
s
-1
)
g1z 2(ε)
0
g1z 3(ε)
Distribution (arb.)
g1z 1(ε)
-0.5
g1z 0(ε)
-1.0
0 5 10 15 20 25
FIGURE 5.5
The longitudinal component (z) of the first-order solution for the velocity distribution function
for electrons at 100 Td in Ar: (a) gz1 (v) and (b) gzn
1 (ε) (n = 0, 1, 2, . . .).
and after we order the right-hand side of the above equation in terms of Pn1 (θ ),
we have

n − 1 ∂gx1n−1 (v, t) n + 2 ∂gx1n+1 (v, t)
= αz (t) +
n=1
2n − 1 ∂v 2n + 3 ∂v

(n − 1)2 gx1n−1 (v, t) (n + 2)2 gx1n+1 (v, t)
− + Pn1 (θ ) cos ϕ
2n − 1 v 2n + 3 v
n−1

∂ n−1

= αz (t) − gx1n−1 (v, t)
n=1
2n − 1 ∂v v

n+2 ∂ n+2
+ αz (t) + gx1n+1 (v, t) Pn1 (θ ) cos ϕ. (5.92)
2n + 3 ∂v v

The term vx g 0 (v, t) is expanded by the aid of the recurrence formula in

Legendre polynomials:

vx g 0 (v, t) = v sin θ cos ϕ Pn (θ )gn0 (v, t)
n=0

1 1
= vgn0 (v, t) cos ϕ 1
Pn+1 (θ ) − 1
Pn−1 (θ )
n=0
2n + 1 2n + 1

1 1
= v 0
gn−1 (v, t) − v 0
gn+1 (v, t) P(θ ) cos ϕ. (5.93)
2n − 1 2n + 3
Finally, when we substitute all the terms in Equations 5.92 and 5.93 into the
first row of Equation 5.87, we have the governing equation of gx1n (v, t)(n =
1, 2, . . .):

∂ 1 n−1 ∂ n−1
g (v, t) + αz (t) − gx1n−1 (v, t)
∂t xn 2n − 1 ∂v v

n+2 ∂ n+2 1
+ αz (t) + gxn+1 (v, t) + R 0 (t)gx1n (v, t)−J gx1 , F
2n + 3 ∂v v
1 1
=− vgn−1
0
(v, t) + vg 0 (v, t). (5.94)
2n − 1 2n + 3 n+1
We numerically obtain gx1n (v, t)(n = 1, 2, . . .) by using the known function
gn0 (v, t)(n = 0, 1, 2, . . .) derived from Equation 5.85. One example of gx1 is
shown in Figure 5.6.
Because of the symmetry of the problem when no magnetic field is present,
we need solve for only one transverse term, gx1 . Thus, by using Equation
5.94 we have three series of equations that must be solved consecutively.
First we obtain the solution for g 0 from the governing Equation 5.85, then
the longitudinal component gz1 from Equation 5.90, and finally the transverse
component gx1 from Equation 5.94. Each of these governing equations actually
represents a hierarchy of equations in the expansion of spherical harmonics.
In Table 5.3 we show expanded collision integrals for different terms in
the gradient and spherical harmonic expansions. The expansion in terms of
number density gradients may be extended to higher terms. In addition, the
expansion in terms of spherical harmonics may be extended to an order suf-
ficient to capture the angular distribution in velocity space accurately. The
lowest-order solution would be to use only the lowest-order term in gradient
expansion and the first two terms (n = 0, 1) in spherical harmonic expansion.
This is known as a two-term expansion and is discussed later. The solution
of equations of different order may proceed by expansion in terms of a par-
ticular velocity distribution, which results in additional requirements in the
expansion and convergence of the solution, or by applying finite difference or
similar numerical techniques. Such techniques may be applied at any level of
expansion even before the density gradient or spherical harmonic expansions.
However, if no expansions are made the numerical solution will need to be
made in seven dimensions in (v, r, t), which under cylindrical symmetry may

gx1 (υ )
3.3
Vr
3.3
(1
0.0 0.0
0
8
-3.3
cm
-1
8 )
Vz (10 cm s
s
-1
)
0.5 g1x 1(ε)

Distribution (arb.)
g1x 3(ε)
0
g1x 4(ε)
g1x 2(ε)
0 5 10 15 20 25
FIGURE 5.6
The transverse component (x) of the first-order solution for the velocity distribution function for
electrons at 100 Td in Ar: (a) gx1 (v) and (b) gx1n (v) (n = 1, 2, . . .).
reduce to five. Even at a very poor resolution of 100 points per dimension, the
number of mesh points is formidable, though it may be within the range of
modern computers. Nevertheless, expansions make the problem more com-
plex but numerically much easier to handle.
5.6.3 Electron Transport Parameters

In Section 5.3, we defined the general form of transport parameters and their
definitions. In this section, we evaluate the expressions for transport coeffi-
cients in terms of the velocity distribution function g(v, r, t). First we write
down the expanded form of the distribution function in configuration and

TABLE 5.3
Each of the Collision Terms Appearing in J(g, F )
Collision Type Collision Integral Expanded Collision Term

m 1 ∂ kTg ∂g00
Elastic J m g00 NQm (v)v 4 g00 +
M v 2 ∂v mv ∂v
m 2m
! 1 ∂
" 3
!
kTg
" #
= ε − kTg NQm (ε) + − NQm f 00
4π M 2 ∂ε 2 4ε
m 2m ∂
# ∂
+ (ε + kTg )NQm (ε) + kTg ε NQm (ε) f0
4π M ∂ε ∂ε 0
m 2m ∂ 0
+ kTg εNQm (ε) f
4π M ∂ε 0
m 1 ∂
J m g10 −NQm (v)vg10 + (NQv (v) − NQm (v)) v 4 g10
M v 2 ∂v
m
= NQm (ε) f 10
4π
m 2m 3
+ (NQv (ε) − NQm (ε))
4π M 2
∂
#
+ε [NQv (ε) − NQm (ε)] f 10
∂ε
2m ∂ 0
+ ε (NQv (ε) − NQm (ε)) f1
M ! 3" ∂ε
3 m
J m g20 − NQv (v)vg20 = − NQv (ε)
2 4π 2
m
J m gx011 −NQm (v)vgx111 = − NQm (ε) f x111
4π ! "
3 m 3
J m gx012 − NQv (v)vgx112 = − NQv (ε) f x112
2 4π 2
1 2
Excitation J j g00 v NQ j (v )g00 (v ) − v 2 NQ j (v)g00 (v)
v
ε+ε j
m 1 ∂ √
= √ εNQ j (ε) f 00 (ε)dε
4π ε ∂ε ε
m
J j g10 org20 −NQ j (v)vg10 org20 = − NQ j (ε) f 10 or f 20
4π
m
J j gx011 orgx012 −NQ j (v)vgx111 orgx112 = − NQ j (ε) f x111 or f x112
$ 4π %
1 1 + k 2
Ionization J i g00 v NQi (v )g00 +(1 + k)v 2 NQi (v )g00 −v 2 NQi (v)g00
v k
(1+k)ε+εi
m 1 ∂ √
= √ εNQi (ε) f 00 (ε)dε
4π ε ∂ε
1+k
ε
ε+εi
∂ k √
+ εNQi (ε) f 00 (ε)dε
∂ε 0
m
J i g10 org2 0
−NQi (v)vg10 org20 = − NQi (ε) f 10 or f 20
4π m
J i gx011 orgx012 −NQi (v)vgx111 orgx112 = − NQi (ε) f x111 or f x112
4π
m
Attachment J a g00 −NQa (v)vg00 = − NQa (ε) f 00
4π
m
J a g10 org20 −NQa (v)vg10 org20 = − NQa (ε) f 10 or f 20
4π
m
J a gx011 orgx012 −NQa (v)vgx111 orgx112 = − NQa (ε) f x111 or f x112
4π

velocity space:

g(v, r, t) = gk (v, t) ⊗ (∇r )k n(r, t)
k

= k
gmn (v, t)Ymn
e
(θ, ϕ) ⊗ (∇r )k n(r, t)
k mn

= gn0 (v, t)Pn (θ ) n(r, t)
n=0

+ gx11n (v, t)Y1n
e
(θ, ϕ)i + g 1y1n (v, t)Y1n
e
(θ, ϕ)j
n=1 n=1

∂
+ gz1n (v, t)Pn (θ )k · n(r, t) + O ∇r2 n
n=1
∂r

3 cos2 θ − 1
= g00 (v,
t) + g10 (v,
t) cos θ + t) g20 (v, + · · · n(r, t)
2
∂
+ gx11 (v, t) sin θ cos ϕ + gx12 (v, t)3 cos θ sin θ cos ϕ + · · · i · n(r, t)
∂r
∂
+ g 1y1 (v, t) sin θ cos ϕ + g 1y2 (v, t)3 cos θ sin θ cos ϕ + · · · j · n(r, t)
∂r

3 cos2 θ − 1 ∂
+ gz10 (v, t) + gz11 (v, t) cos θ + gz12 (v, t) + · · · k · n(r, t)
2 ∂r

+ O ∇r2 n . (5.95)
Exercise 5.6.1
Discuss and develop equations for two-term approximation in the case of
electron transport in gases. Discuss the order of errors and applicability.
Two-term approximation consists of taking the first two terms in the spherical
harmonic expansion in the lowest order of gradient expansion (k = 0) in a
steady state. From Equation 5.95 we directly obtain

g(v) = g00 (v) + g10 (v) cos θ + O g20 ne ,
where the order of error in spherical expansion is given by O(g20 ), and higher-
order terms in the gradient expansion are neglected. In order for this solution
to be accurate (i.e., in order that the spherical harmonic expansion converges),
we need to satisfy the condition g00 (v, t) g10 (v, t). The corresponding equa-
tions of the two-term expansion are given by

m 1 ∂ 1 eE ∂ 2
− NQm (v)v 4 g00 + R 0 g00 (v) + + g10 (v) = 0, (5.96)
M v ∂v
2 3 m ∂v v

where g10 (v) is given as
1 eE ∂g00 (v)
g10 (v) = − & . (5.97)
NQk (v)v m ∂v
Finally, it is noticed that the velocity distribution of electrons expressed by

the two-term theory, Equations 5.96 and 5.97, is nearly isotropic. On the other
hand, an anisotropic velocity distribution of electrons is realized under a
higher E/N or high degree of inelastic collsions that strengthen the directional
component of electrons to the external field E. Thus we conclude that the two-
term theory is of acceptable accuracy for situations in which the cross section
for inelastic processes is much smaller than that for elastic processes. It has
often been assumed that the theory is exact for atomic gases at low E /N when
only elastic collisions occur.

First we must normalize the distribution function g(v, r, t)dv = 1.
Bearing this normalized function in mind, and on the basis of Equation the
definition of the directed or drift velocity in Equation 5.18, we derive the
equation for the drift velocity vd (t) as

vd (t) = vg 0 (v, t)dv
π
= v cos θgn0 (v, t)Pn (θ )v 2 dvd
n 0

4π
= v 3 g10 (v, t)dv. (5.98)
3
We can derive the expression for the transverse diffusion coefficient DT (t)
and for the longitudinal diffusion coefficient DL (t) from Equations 5.21 and
5.22 as

DT (t) = vx gx1 (v, t)dv = v y g 1y (v, t)dv

= v sin θ cos ϕgx11n (v, t)Y1n
e
(θ, ϕ)v 2 dvd
n

4π
= v 3 gx11 (v, t)dv, (5.99)
3
and

DL (t) = vz gz1 (v, t)dv

= v cos θgz1n (v, t)Pn (θ )v 2 dvd
n

4π
= v 3 gz11 (v, t)dv. (5.100)
3

The effective production rate of electrons is described by combining the

ionization and electron attachment rates, Ri and R a :
R 0 (t) = Ri (t)
− Ra (t)
= N [Qi (v) − Qa (v)] vg(v, t)dv

= 4π N [Qi (v) − Qa (v)] v 3 g00 (v, t)dv. (5.101)
The rates of excitation of electronically excited levels are also defined in a

similar way as the ionization or attachment rates: R j (t) = 4π N Q j (v)v 3 g00
(v, t)dv. The ensemble average of the electron energy ε(t) as defined by
Equation 5.14 can be described as

1 2
ε(r, t) = mv g(v, r, t)dv
2
1
= mv 2 gn0 (v, t)Pn (θ )v 2 dvd
n
2
1
−∇z nk mv 2 gz1n (v, t)Pn (θ )v 2 dvd
n
2
1
−∇x ni mv 2 gx11n (v, t)Y1n
e
(θ, ϕ)v 2 dvd
2
n
= 2πm v g0 (v, t)dv − 2π m v gz0 (v, t)dv ∇z nk. (5.102)
4 0 4 1
PROBLEM 5.6.1
Derive the following expression of the diffusion tensor of electrons in gases without
inelastic collisions in a direct-current field:

v(v − vd )g 0 (v) v emE · ddv g1 (v)
D= dv + dv, (5.103)
NQm (v)v(1 + (v)) NQm (v)v(1 + (v))
where (v) is a function of M and Tg of gas molecules,
kTg d
(v) = ln[NQm (v)v].
Mv dv
References
1. Chapman, S. and Cowling, T.G. 1970. The Mathematical Theory of Non-Uniform
Gases, 3rd ed., Cambridge: Cambridge University Press.
2. Shkarofsky, I.P., Johnston, T.W., and Bachynski, M.P. 1966. The Particle Kinetics of
Plasmas. Reading, MA: Addison-Wesley.
3. Kumar, K., Skullerud, H.R., and Robson, R.E. 1980. Kinetic theory of charged
particles swarms in neutral gases. Aust. J. Phys. 33:343–448.
4. Morse, P.M. and Feshbach, H. 1953. Methods of Theoretical Physics. New York:
McGraw-Hill.

6
General Properties of Charged
Particle Transport in Gases
6.1 Introduction
Swarm parameters are generally applied to plasma modeling and simula-
tions. At the same time, the nonequilibrium regime in discharges is well repre-
sented under a broad range of conditions by using the Boltzmann equation, as
discussed in the previous chapter. In this chapter we discuss the basic features
of transport coefficients (swarm parameters) of electrons and also, to a lesser
degree, of ions. Together with transport data we compile cross sections of the
electron for interesting gases. First we discuss the basic characteristics of the
E/N dependence of the coefficient for swarms in the hydrodynamic regime.
We consider swarms in direct current (DC) fields and also in time-varying ra-
dio frequency (rf) fields. Collective phenomena that cannot be easily described
on the basis of single electron dynamics are labeled kinetic phenomena, and
some special examples are described. Differences between the transport in
conservative and nonconservative situations (i.e., when there are number-
changing collisions such as ionization and attachment) are briefly considered.
6.2 Electron Transport in DC-Electric Fields

Swarm experiments were initially developed to study the motion of charged
particles in gases. These experiments were made at relatively high pressures
to ensure that nonhydrodynamic effects were negligible. The discussion of
experimental techniques developed to measure transport coefficients is be-
yond the scope of this book [1]. At present we have an opportunity of highly
accurate transport data available for a low-temperature plasma modeling.
Data on the swarm parameters for electrons up to the early 1980s have been
compiled by Dutton et al. [2].
6.2.1 Electron Drift Velocity

Drift velocities are usually measured in two types of experiments [1, 3].
Experiments with shutter grids are of high accuracy but are limited to low
175

103
electronic
excitation
III IV
polyatomic gas
2
vde (arb.) 10
II atomic gas
10 I ∝ (E / N)1/2
∝E/N
1
1 10 102 103
DC- E / N (arb.)
FIGURE 6.1
Typical characteristics of electron drift velocity as a function of DC-E/N.
E/N. The pulsed Townsend experiments are made between two parallel
electrodes with a swarm of electrons produced by a photo-emission. Their
accuracy is somewhat smaller, but it may still be possible to achieve an un-
certainty of less than ±3%. The measurements may be extended to higher
E/N. There is a class of the drift velocity experiment measuring the optical
emission from short-lived excited species in the electron swarm that can be
performed only at higher E/N.
In Figure 6.1 we show the typical characteristics of the drift velocity of
electrons in gases as a function of DC-E/N. At sufficiently low E/N, where
an electron loses all the kinetic energy equal to the gain from the field E at
one elastic collision, the drift velocity is proportional to E/N (region I). In
region I the mean energy of electrons is close to the thermal energy 3kTg /2
and the velocity distribution is approximated by the Maxwellian. A finite
kinetic energy of electrons is steadily maintained when the energy gain from
the field is balanced with the energy loss by elastic collisions during a unit
time. Then, the region II appears in Figure 6.1. Even at low E/N, polyatomic
molecules have a probability of vibrational (and rotational) excitations with
the electron as well as the elastic scattering. This means that there is also a
region III that mixes II with I as shown in Figure 6.1. The magnitude of the
drift velocity in the mixed region strongly depends on the magnitude of the
vibrational cross section. A rapid increase in the drift velocity at high E/N
(region IV) is caused by electronic excitations (including ionization) with a
high energy loss from several eV to 20 eV [4].
The E/N dependence of the drift velocity of electrons in He, Ar, H2 ,
HCl, CH4 , CF4 , and SiH4 is shown in Figure 6.2. As qualitatively shown in
Figure 6.1, CF4 , SiH4 and CH4 have more distinct characteristics than He and
Ar in the region of moderate E/N. That is, the drift velocity has a peak in re-
gion III in CF4 , SiH4 , and CH4 . The first significant increase is often known as

General Properties of Charged Particle Transport in Gases 177
8 Ar
10 H2
He
CH4
Drift velocity (cm s )
-1
CF4 HCl
10
7 CH4
SiH4
SiH4
H2
6
10
Ar
HCl
5
10 2 3
1 10 10 10
DC- E / N (Td)
FIGURE 6.2
Electron drift velocities in various gases as a function of DC-E/N.
enhanced mobility, and the decrease of the drift velocity is known as negative
differential conductivity (NDC). The name of the conductivity originated
from the fact that in a weakly ionized plasma the conductivity σ is defined as
je = ene vd = σ E.
Observe, however, that the conductivity is proportional to both the drift ve-
locity vd and the number density ne , and thus the NDC will also occur if the
electron number density changes with E/N, which will be possible in the
case of the nonconservative collisions, that is, under the electron attachment
or ionization. The most interesting phenomena associated with the NDC arise
from the negative slope of vd with respect to E/N, and we use the term NDC
just to describe the negative slope of the drift velocity. Typical gases that have
NDC are CF4 , SiH4 , CH4 , and mixtures of Ar with molecular gases such as N2 .
Exercise 6.2.1
Derive the conditions for NDC based on the momentum and energy balance
of electrons. If we start from the energy balance between the energy gain from
the field E and dissipation in inelastic collisions with a threshold energy ε j
and rate R j in a steady state with uniform ne (here, we neglect the energy loss
in elastic collisions)
d
< εm >= e Evd (E/N) − ε j R j (E/N) = 0.
dt

The momentum balance with the momentum transfer collision rate Rm is

written in the form
d
(mvd (E/N)) = e E − mvd R j (E/N) = 0.
dt
After eliminating E from both of the balanced equations, we obtain the
following condition for NDC:

∂ eε j R j 1 ∂ Rj 1 ∂ Rm
vd (E/N) = − .
∂ E/N 2mvd Rm R j ∂ E/N Rm ∂ E/N
Here, we observe directly that if we want the slope of the drift velocity to
be negative, then the collision rate for inelastic scattering R j should decrease
as a function of E/N and the collision rate for the elastic momentum trans-
fer should increase. Although the above relation is derived for very simple
assumptions, the conclusions are generally applied for the NDC.
The magnitude of the drift velocity is a strong indicator of the anisotropy

of the velocity distribution g 0 (v) in a uniform density of electrons. The effect
of enhanced mobility due to inelastic processes, for example, in SiH4 , can be
observed from the plots of the velocity distribution g 0 (v) in Figure 6.3 at DC-
E/N of 20 Td, 100 Td, and 500 Td corresponding to the maximum, minimum,
and much higher value of the drift velocity. The asymmetric behavior of g 0 at
θ = 0 axis is caused such that the part of the distribution in the direction of
acceleration is continuously supplied by the effect of the electric field, but the
relative shrinking of the distribution in other directions is directly the result
of inelastic collisions.
6.2.2 Diffusion Coefficients

The diffusion coefficients NDT and NDL are directly used in plasma mod-
eling, though the anisotropy of the diffusion tensor is often neglected. In
Chapter 2 we discussed the reasons for the anisotropy of the diffusion tensor.
In Equation 2.12 of the diffusion coefficient, ξ is the logarithmic derivative
of the momentum transfer collision rate Rm on mean energy. When Rm is in-
dependent of the energy the diffusion tensor is isotropic. If ξ 1, it follows
that DL = DT /2. This is the case in many molecular gases for the region of
moderate E/N, which overlaps the typical applied conditions in a nonequi-
librium and low-temperature collision-dominated plasma. At higher mean
electron energies where the collision rate starts to decay, the dependence of
the collision rate becomes close to a constant, and thus at very high E/N the
components of the diffusion tensor again become similar in magnitude. In
gases that have a chance for a rapid reduction of the collision rate (e.g., due to
the Ramsauer–Townsend minimum), there is a chance that DL will be greater

(a)
g0 (υ )
8.7
8.7
0.0
0.0 -8.7
Vr
(10 7
cm -1
7 cm s )
s -1 Vz (10
)
(b)
g0 (υ )
2.2
2.2
0.0
0.0 -2.2
Vr
(10 8
cm -1
8 cm s )
s -1 Vz (10
)
(c)
g0 (υ )
4.0
4.0
0.0
0.0 -4.0
Vr
(10 8
cm -1
8 cm s )
s -1
) Vz (10
FIGURE 6.3
Velocity distribution g 0 (v) as a function of DC-E/N in pure SiH4 : (a) 20 Td, (b) 100 Td, and
(c) 500 Td.

NDLL
23 NDTT
10 Ar
Diffusion coefficients (cm s )

-1
-1
Ar
CF4
22
10 SiH4
SiH4
21
10 2 3
1 10 10 10
DC- E / N (Td)
FIGURE 6.4
Diffusion coefficients of electrons in various gases as a function of dc-E/N. Transverse diffusion
coefficient (· · ·), longitudinal diffusion coefficient (—).
than DT at very low E/N. In Figure 6.4 we show data for the components of
the diffusion tensor in Ar, CF4 , and SiH4 .
The variations of the DT /DL ratio attain a magnitude as large as 10 in gases
such as Ar, CH4 , CF4 , or SiH4 . In particular, the most striking feature is the
large difference between the two components of the diffusion tensor between
10 Td and 100 Td, which is almost one order of magnitude. This feature is a
direct result of the very strong Ramsauer–Townsend minimum in the elastic
scattering cross section.
PROBLEM 6.2.1
Discuss the numerical algorithm to calculate g 0 (v) as a boundary condition problem.
6.2.3 Mean Energy of Electrons

Figure 6.5 shows the characteristics of the electron mean energy in He, Ar,
H2 , N2 , CF4 , HCl, SiH4 , SF6 , and CH4 in a pure DC electric field. The en-
ergy is determined from the balance between the gain from the field and the
loss by collisions with the molecule. The electron mean energy is the dom-
inant one when we consider the electron swarm in gases, though it does
not appear in the continuity equation of the number density, unlike other
transport coefficients such as the drift velocity, diffusion coefficients, and col-
lision rates. Usually we have no way to measure the mean energy with high
accuracy.

2
10
H2
Mean energy (eV) 10 SF6
Ar
CH 4
He
1
N2
CF4
10
-1 HCl
SiH4
2 3
1 10 10 10
DC- E / N (Td)
FIGURE 6.5
Electron mean energy in gases as a function of DC-E/N.
6.2.4 Excitation, Ionization, and Electron Attachment Rates

The rates of inelastic processes (and of elastic processes) are represented as
i. Collision frequency (number of events per unit time, per electron with
enegy ε, and per gas molecule : ν j (ε)[cm3 s−1 ]);
ii. Rate constant (number of events per unit time per electron and per
molecule: k j [cm3 s−1 ]);
iii. Rate (number of events per unit time per electron R j [s−1 ]);
iv. Net rate (number of events per unit time: j [cm−3 s−1 ]); and
v. Townsend’s spatial coefficients (number of events per unit distance
traveled by an electron: α j [cm−1 ]).
Excitation processes that include dissociation are a special group of trans-
port properties of electron swarms. Excitation rates are used to determine the
absolute intensity of emission and dissociation and the density of short-lived
species and long-lived radicals and to normalize the collision cross sections.
There is a particular need for excitation rates for optimization of plasma
displays, light sources, and particle detectors and for diagnostics of plasma
structure and characteristics.
Ionization and electron attachment rates enter the transport equations
as zeroth-order transport coefficients describing the change in the number

(a)
10
10
Ri
8
10
Rex
Rexm
Rex(2p1)
6
10
Rate (s )
-1
Rex(3p5)
4
10
2
10
0
10 2 3
1 10 10 10
DC- E / N (Td)
(b)
10
10
Rv3
8
10 Rv1
Rv4
6
10
Rate (s )
-1
4
10
Ri1
Rex
2 Rd1
10
Ra Rd2
Rd3
0
10 2 3
1 10 10 10
DC- E / N (Td)
FIGURE 6.6
Excitation rates for electrons in Ar (a) and CF4 (b) as a function of DC-E/N.
density. All these collision processes have common characteristics having a

threshold energy as a minimum energy required for the process to occur. Thus
only a part of the energy distribution is convoluted to give the rates or net
rates for those processes, normally the higher energy tail of the distribution.

10
10
ionization
attachment
8
10 SF6-
SF5-
SF6
SF4-
F-
6
10
Rate (s )
-1
CF4 SiH4
CF4
4
10 Ar SiH4
H2 N2
2
10
SF6
0
10 2 3
1 10 10 10
DC- E / N (Td)
FIGURE 6.7
Ionization and attachment rates of electrons in Ar, H2 , N2 , CF4 , SF6 , SiH4 , and CH4 as a function
of DC-E/N.
In Figure 6.6 we show some of the excitation rates with ionization and electron
attachment rates in Ar and CF4 . One can see that for a range of E/N there is a
negligible rate, and then at some point the rate starts to increase very rapidly.
The point where this occurs is related to the magnitude of the energy loss
for that process, and the rapid increase marks the overlap of the high-energy
tail of the energy distribution with the cross section for that process. At some
point the increase slows down until a maximum is reached, and eventually
the rate starts to decay. The rate (constant) is a key to calculate the kinetics of
various plasma chemical processes and densities of excited states and is thus
used directly in plasma modeling.
The ionization rates are significant for relatively high mean energies (i.e.,
high E/N) and are essential for establishing the accuracy of the energy distri-
bution of electrons and checking the cross section data at higher energies (see
Figure 6.7). Dissociative electron attachment also has a threshold energy at
moderate energy. The electron attachment to a parent gas at very low thermal
energy may occur. Thus sometimes the electron attachment rate will have a
maximum at thermal energy. Attachment has a threshold and peak at lower
E/N than ionization, and thus there is an E/N where ionization catches up
and becomes equal to attachment, as shown in Figure 6.7. Slightly above that
point is the usual electrical breakdown point in electronegative gases. Thus
in electronegative gases, the E/N that satisfies Ri = Ra will give an ideal bulk
electric field in low-temperature plasmas.

6.3 Electron Transport in Radio-Frequency Electric Fields

For time-varying electric fields we consider the electron transport at all fre-
quencies. However, for low frequencies, quasi-DC behavior is achieved at
each of the phases of a time-varying rf field, because the electron energy distri-
bution has enough time to relax to the profile at the instantaneous field by col-
lision with neutral molecules. The most interesting phenomena under general
circumstances will occur for radio frequencies at which electrons do not have
enough time to relax their energy or momentum on the time scale of the
changes of the field. For extremely high frequencies, electrons are not able
to react to the electric field on the time scale of the period of oscillations
and behave as a DC-like swarm. Ions will fail to relax their properties even
at much lower frequencies due to the large mass in comparison with elec-
trons. Here, we use the term “transport in rf fields” for all frequencies used
in low-temperature plasma applications.
We consider the rf electron transport in an electric field
E(t) = E0 cos ωt = Re[E0 exp(iωt)], (6.1)
where ω and E 0 are the angular frequency and the amplitude, respectively,
and E0 is usually taken to be −E 0 k. The rf transport is characterized as a
function of E 0 /N and ω/N in a gas with number density N, whereas the
DC-transport is only a function of E 0 /N.
6.3.1 Relaxation Time Constants

We start from the velocity distribution g 0 (v, t) of electrons with a uniform
density in an rf field 6.1. Usually we have time constants for momentum and
energy of electrons migrating in the field under collisions with surrounding
molecules in a gas or in a collision-dominated plasma. These two time con-
stants of the electron are related to the first two-term expansion of the
Boltzmann equation (see Section 5.6.2). We have the relaxation equation for
energy,
∂ 0 1 0
g0 (v, t) + ζ E 0 , ω, g10 (v, t) = − g (v, t), (6.2)
∂t τe (v) 0
and for momentum relaxation
∂ 0 1
g1 (v, t) + ξ E 0 , ω, g00 (v, t) = − g 0 (v, t), (6.3)
∂t τm (v) 1
where g00 (v, t) and g10 (v, t) are the isotropic and directional parts of the velocity
distribution g 0 (v, t) expressed by the two-term approximation,
g 0 (v, t) = g00 (v, t) + g10 (v, t)cos θ,

where θ is the angle between v and −E; that is, θ = −(v · E)/vE. The two time
constants τe (v) and τm (v) are defined, respectively, as

2m εv d
−1
τe (v) ∼Nv Qm (v)+ (Qv (v)v) + Q j (v)+Qi (v)+Qa (v) ,
M ε dv
(6.4)

τm (v)−1 = Nv Qm (v) + Qv (v) + Q j (v) + Qi (v) + Qa (v) . (6.5)
Typically, at 1 Torr in a low-temperature plasma the mean energy of electrons

is on the order of 2 eV ∼ 6 eV and then τm ∼ = 10−10 s, whereas τe ∼= 10−5 s in
atomic gases or τe ∼ = 10 −6
s - τ ∼
e = 10 −8
s in molecular gases. In Figure 6.8 we
show the collisional relaxation times in Ar at 1 Torr together with the cross
section Q j and collision frequency ν j . We see that at low energies (ε < 11.55 eV)
the energy relaxation is slow as it is conducted purely by elastic collisions,
whereas at higher energies first electronic excitation and then ionization set
in and shorten τe by several orders of magnitude. We should recognize that
the relaxation of the high-energy tail of the electron energy distribution will
be much faster than that of the bulk of the distribution. The relaxation of the
electron transport in rf fields can be determined directly by the solution to
the Boltzmann equation or by Monte Carlo simulation. In Figure 6.9 we show
a relaxation of the mean energy < ε(t) > of electrons in rf fields for 1 MHz,
10 MHz, and 100 MHz. The transition region may be easily recognized, and
then quasi-stationary conditions develop where the same periodic function is
repeated even though the electrons may not be fully relaxed to the properties
they would have for the instantaneous electric field. Nevertheless, a periodic
steady state or quasi-stationary distribution develops.
The characteristics of the electron transport can be divided into three dis-
tinct regions depending on the relative magnitudes of the relaxation time
constants and period of oscillations of the rf field.
i. τm (v)
τe (v)
ω−1 : Low frequencies.
Under these circumstances, at all times the electron swarm and the ve-
locity distribution are relaxed to the DC field conditions corresponding
to the instantaneous electric field E(t). Thus this is usually labeled the
regime of “instantaneous field approximation.”
ii. τm (v)
ω−1 τe (v)
: High frequencies.
Under these conditions the momentum will mostly be relaxed at all
times, but the energy of the electron swarm will not be relaxed and this
will lead to numerous phenomena that may not be explained easily
on the basis of DC-field transport. In all cases there will be a delay
introduced between the field and the transport coefficients.
iii. ω−1 τm (v)
τe (v): Very (ultra) high frequencies.
For this set of circumstances both the energy and momentum will be
delayed in response to the electric field, and therefore all transport
coefficients will have a considerable phase shift to the electric field.

Ar CF4
(a)
-15
10 10
-15 Qv3
Qi Qm
Cross section (cm )
Cross section (cm )

Qm
2
2
Q1 Q
-16 -16 Qv1 d1
10 10
Qex
Qexm Qv4 Qex Q Qd3
-17 -17 d2 Qi3
10 10 Qi4
Qex(2p1) Qi2 Qi5
-18 -18
10 10 Qi6
Qex(3p5) Qi7
Qa
-19 -19
10 -2 -1 2
10 -2 -1 2
10 10 1 10 10 10 10 1 10 10
Electron energy (eV) Electron energy (eV)
(b) 10-6 10
-6
Collision frequency (cm s )
Collision frequency (cm s )
i
3 -1
3 -1
-7 -7 d1
10 10 i1
ex
v3 d3
-8 -8
10 exm 10 i3
m d2 i4
m v1 ex
-9 -9
10 ex(2p1) 10 i5
i2 i6
-10 -10 v4
10 10 i7
ex(3p5)
a
-11 -11
10 -2 -1 2
10 -2 -1 2
10 10 1 10 10 10 10 1 10 10
(c) 10-2 10
-2
τe τe
-4 -4
10 10
Relaxation time (s)
Relaxation time (s)
-6 -6
10 10
τm
-8
10 10
-8
τm
-10 -10
10 10
-2 -1 2 -2 -1 2
10 10 1 10 10 10 10 1 10 10
FIGURE 6.8
A set of cross sections of electrons in pure Ar and CF4 : (a) cross section, (b) collision frequency,
and (c) collisional relaxation time at 1 Torr.

10 MHz
6.0
Mean energy (eV)
100 kHz 6000 π

4.0
1 GHz
2.0
0 2π 4π 6π 8π 10π
ωt
FIGURE 6.9
Examples
√ of the relaxation of the mean energy of electrons initiated at g M (v) at 2 eV at E 0 /N =
50 2 Td and 1 Torr in Ar: (a) 100 kHz, (b) 10 MHz, and (c) 1 GHz.
The ratio of the alternating-current component to the DC component

of the coefficient will decrease with the increase of ω of E (t ), and the DC
component is caused by the superimposition of the property during
the one period.
The properties in the second and third region are observed in the ioniza-
tion rate Ri in Ar in Figure 6.10. Periodic steady-state characteristics with
twice the frequency of the external field are realized as a function of ωt.
The time-averaged ionization rate is kept at an almost constant value up to
100 MHz. At frequencies greater than 100 MHz, the DC component rapidly
decreases, because electrons do not have enough time to get energy from the
external field before its direction changes. That is, electrons fall into a spa-
tial trapping as is shown in the term of the phase delay with respect to the
external E(t).
Exercise 6.3.1
By considering the time evolution of the electron number density ne (t) due to
ionization and electron attachment, estimate the form of ne (t) under a periodic
steady-state condition.
The time-dependent distribution function G 0 (v, t) may be separated into
the product of a time-varying number density ne (t) and a velocity distribution
function g 0 (v, t) normalized to unity under a periodic steady state

G 0 (v, t) = ne (t)g 0 (v, t) where g 0 (v, t)dv = 1. (6.6)
v

(a)
8.0
Ionization rate Ri(t) (107 s-1) 1 MHz
6.0 200 MHz
4.0
2.0
1 GHz
5 GHz
0.0
0 π 2π
ωt (rad.)
(b)
90
4.0
Ri(t)DC (107 s-1)
3.0 60 ϕi (deg.)
2.0
30
1.0
0.0 0
106 107 108 109 1010
Frequency (Hz)
FIGURE 6.10
√
Ionization rate Ri (t) at E 0 /N = 300 2 Td at 1 Torr in Ar as a function of external frequency:
(a) time-modulation of Ri (t), (b) DC-component and phase delay.
ne (t) and g 0 (v, t) are separated using a separation function RT (t):
∂ 0 eE(t) ∂ 0
g (v, t) + · g (v, t) = J (G, F ) − RT (t)g 0 (v, t), (6.7)
∂t m ∂v
where RT (t) represents the difference between the ionization and attachment
rates, Ri (t) − Ra (t), and is related to the electron number density by
t
ne (t) = no exp RT (t)dt , (6.8)
0
where no is the density at t = 0.

6.3.2 Effective Field Approximation

In atomic gases such as He and Ar, most of the energy loss of electrons is
due to elastic collisions with neutral atoms up to moderate E(t)/N. Under
these conditions, the energy relaxation time τe will be much longer than the
period of the field, τe ω−1 , whereas τm is much shorter (see for example
Figures 6.11 and 6.12). Then, the time dependence of the velocity distribution
g 0 (v, t) is expressed as
g 0 (v, t) = g00 (v) + g10 (v)cos θ cos ωt
in terms of an effective DC-field E e f f (v),
E0 1 E0 1
E e f f (v) = √ 1/2 = √ 2 1/2 . (6.9)
2 1 + (ωτm (v)) 2 2 ω
1+ R
m(v)
The method of using the effective DC-field instead of a time-varying

field 6.1 is very simple and has been widely used to study swarm trans-
port. With this procedure, the transport coefficients defined in terms of g00 (v),
such as < ε > and Ri , are constant in time, though the drift velocity given by
g10 (v, t) has a periodic time response. For molecular gases with large vibra-
tional excitations from low energy on the order of 0.1 eV to several eV, the
situation is not so simple even at high frequency. In that case, the isotropic
part of the distribution g00 (v, t) as well as g10 (v, t) is modulated in time owing
to τe < ω−1 even at high frequency.
Exercise 6.3.2
Derive the effective field approximation 6.9 by using the momentum balance
equation of electrons.
The momentum balance of electrons with a uniform density in gases is
given by
d
m vd (t) = −eE(t) − Rm mvd (t),
dt
where vd is the drift velocity and Rm is the momentum transfer rate. The
solution is

eE0 1 −1 ω
vd (t)=− cos(ωt+φ) where φ= tan . (6.10)
mRm [1+( Rωm )2 ]1/2 Rm
In the limit of low frequency ((ω/Rm )2 1), the drift velocity converges to an
instantaneous DC-like solution, vd (t) = −e E 0 /mRm cos ωt, and for the high
frequency limit (ω/Rm )2 1) the solution is
eE0
vd (t) = − cos(ωt + φ). (6.11)
mω

H2 N2
(a)
10-14 10-14
Qm
Qm
Cross section (cm2)
Cross section (cm2)

QeC 3 + Qi
Qi B Σg
10-16 Qre02 10-16 Qd
Qre13 Qvib. others
1
a Πg
QeE 3 +
B' Σu
Qv1 Qe1s2s Qe2p2s
10-18 10-18 3
QeB Qe1sn3 C Πu
Qre24 Qv2 Qdi 3 +
A Σu
Qre35 Qe1s1s
10-20 -2 10-20 -2
10 10-1 1 10 102 10 10-1 1 10 102
(b)
10-6 10-6
νeC
Collision frequency (cm3 s-1)
Collision frequency (cm3 s-1)
νi νi
νm νeB 10 -7
3 + νd
10-8 B Σg
νeE νe2p2s
νre02 νe1sn3 10-8
νm ν vib. others
νdi
νre13 a Πg
1
10-10 νv1 νe1s1s

3 +
B' Σu
νre24 νe1s2s 10-9
νre35 νv2
3 +
A Σu C3 Πu
10-12 -2 10-10 -2
10 10-1 1 10 102 10 10-1 1 10 102
(c)
10-4 10-2
10-4
Relaxation time (s)
Relaxation time (s)
τe
10-6
τe
10-6
10-8
τm 10-8 τm
10-10 10-10
10-2 10-1 1 10 102 10-2 10-1 1 10 102
FIGURE 6.11
A set of cross sections of electrons in H2 and N2 : (a) cross section, (b) collision frequency, and
(c) collisional relaxation time at 1 Torr.

SiH4 SF6
(a) -14
10
Qm Qa1
10 -14
Qd Qm
Qi
Cross section (cm 2 )
Cross section (cm 2 )

Qi
10 -16
10 -16 Qex
Qa2 Qa3
Qa4
-18 100 Qa 10 -18

10 Qa5
Qv
Qv24 Qv13
10 -20
10 -20 -2
10 10 -1 1 10 10 2 10 -2 10 -1 1 10 10 2
(b)
10 -6 10 -6
ν a1
νi
Collision frequency (cm 3 s -1 )
Collision frequency (cm 3 s -1 )
νd ν ex
νi νm
10 -7 -8
νm 10
ν a3
10 -8 ν a2 ν a4
10 -10
ν a5
100 ν a
10 -9 νv
ν v24 ν v13 10 -12
-10
10
10 -2 10 -1 1 10 10 2 10 -2 10 -1 1 10 10 2
(c) 10 -4
10 -4 τe
Relaxation time (s)
Relaxation time (s)
10 -6
10 -6
τe
10 -8
10 -8
τm
τm
10 -10 10 -10
10 -2 10 -1 1 10 10 2 10 -2 10 -1 1 10 10 2
FIGURE 6.12
A set of cross sections of electrons in SiH4 and SF6 : (a) cross section, (b) collision frequency,
and (c) collisional relaxation time at 1 Torr.

We can see that in the limit of (Rm /ω)2 → 0 the electron motion is always in
the field direction and thus the electron cannot gain any energy from the field.
If there are collisions (represented by Rm ), electrons are able to attain some
mean energy. Thus collisions with gas or any other dissipation of momentum
will lead to a net gain of energy for electrons in rf fields. Next, the power
absorption of electrons is achieved by the drift in an rf field:
π π
PW = −eE(t) · vd (t)dωt/ dωt
0 0
2
1 e E0 e 2 E e2f f
= √ = . (6.12)
mRm 2(1 + (ω/Rm )2 )1/2 mRm
Here, E e f f is equal to the effective field in Equation 6.9.
Although the effective field approximation may give averaged properties

of electrons in rf fields, it will be not so accurate in the most interesting fre-
quency region and it certainly cannot provide time-dependent results. We
proceed by discussing in greater detail the time-dependence of the transport
data for quasi-stationary (relaxed) conditions.
6.3.3 Expansion Procedure

A finite phase delay with respect to an external field 6.1 will produce higher-
order harmonics of the fundamental wave with frequency ω in the veloc-
ity distribution of electrons. We consider the temporal harmonic behavior
of electron transport under a purely sinusoidal field 6.1. Inasmuch as the
velocity distribution g 0 (v, t) is symmetric with respect to the field direction,
the velocity distribution can be expanded in terms of spherical harmonics,
and the temporal behavior is determined by the sum of the higher-order har-
monics of the fundamental wave with frequency ω. Expanding g 0 (v, t) in
Legendre polynomials in velocity space and in a Fourier series in time, we
obtain

g (v, t) = Re
0
gs0k (v)Ps (cos θ)exp(ikωt) , (6.13)
k s
where θ is the angle between −E(t) and v is expressed as θ = cos−1 [−(v·E)/vE.

gs0k (v) is real for k = 0 and complex for k > 0, expressing the phase lag
of electron transport with respect to the applied rf field 6.1. It should be
noted that the isotropic part of the distribution go0k (v) possesses only even
harmonics in time, whereas the directional-drift part, g10k (v), possesses only
odd harmonics. Physically, this is because under a pure sinusoidal field 6.1
the energy gain described by the isotropic velocity component is proportional

to E(t)2 , that is, even time harmonics. The directional-drift component is a

function of E(t), that is, odd harmonics.
It is convenient to define an energy distribution f s0k (ε) according to the
relation
f s0k (ε) = (4π/m)vgs0k (v); ε = mv 2 /2, (6.14)

∞
where f s0k (ε) is normalized to 0 f 00 (ε)dε = 1. In a pure sinusoidal field,
symmetry considerations show that only components f s0k (ε) where s + k is an
even number can exist; that is, s + k = 2β, β = 1, 2, . . . . In that case, insertion
of expansion 6.13 into the Boltzmann equation without a density gradient
5.85 gives a set of coupled differential/difference equations of the form [5]:
ikω(m/2)1/2 f sk ()

e ER s d s
+ √ [λ1 f s−1 () + f s−1 ()] − [λ1 f s−1 () + f s−1 ()]
k−1 k+1 k−1 k+1
2 2s − 1 d 2

e ER s + 1 d s +1
+ √ [λ1 f s+1 () + f s+1 ()] +
k−1 k+1
[λ1 f s+1 () + f s+1 ()]
k−1 k+1
2 2s + 3 d 2
m 1/2
+J [ f k (), Tg ] s=0
= Isk () 0
2 −N[Q m () + Q j () + Q i () + Q a ()] f s
k
() s= 0,
(6.15)
where λ1 = 1 for all k = 1, and λ1 = 2 for k = 1. Tg is the gas temperature,

and Isk () satisfies the relation

RT (t)Re f s ()exp(ikωt) = Re
k
Is ()exp(ikωt)
k
k k
and can be expressed as

∞
λo k k−k 1 k+k k ∗
k

Isk () = RT f s () + [RT f s () + RTk ∗ f sk+k ()], (6.16)
2 k =0 2 k =0
where λo = 1 for all k = 0, and λo = 2 for k = 0. RkT is the kth component in

the Fourier series of RT (t), and the symbol ∗ denotes the complex conjugate.
The collision term in Equation 6.15 is expressed as follows:
J [ f 0k (), Tg ]

√ 2m d 2 d −1/2 k
= NQm () f o () + k B Tg NQm ()
3/2 k
[ f o ()]
M d d
d + j
+ −1/2 1/2 NQ j () f ok ()d
j
d

o
−1/2 d
+ 1/2 NQa () f ok ()d
d a
/δ+i /(1−δ)+i
d
+ −1/2 + 1/2
NQi () f ok ()d ,
d o
(6.17)
where k B is the Boltzmann constant. ε j , εi , and εa are the threshold energies

for excitation, ionization, and electron attachment, respectively. δ :(1 − δ) is
the energy partition ratio between two ejected electrons after ionization.
In particular, at ultra-high frequency, most of the component f 000 (ε) lies in
an energy range less than the electronic excitation threshold ε j , and collisions
are almost completely limited to elastic, rotational, and vibrational scatterings.
Under these circumstances, f 000 (ε) can be expressed in an analytical form from
the Boltzmann Equation 6.15 as
 

√  1+ [Qr ()r +Qv ()v ]/(2m/M) Qm () 
f 000 ()=A exp− (e E R /N)2 d .
0 k B Tg + √ 2
/3 M Qm () Q()
2m
1+[ω/ 2/mN Q()]
(6.18)
This is identical to the expression of Margenau and Hartman [6], except for
the presence of rotational and vibrational collisions. In atomic gases subject
only to elastic scattering, the distribution f 000 (ε) in the limit, ω−1 τm (ε), has
a Maxwellian form with an effective gas temperature

e ER 2 3k B Tg
Te f f = Tg 1 + / . (6.19)
mω M
The presence of rotational and vibrational collisions with low threshold en-
ergies in molecular gases changes the shape of the distribution substantially
in comparison with that obtained in atomic gases. That is, f 000 (ε) can be ex-
pressed in the asymptotic form of Equation 6.19 as

√ 1 [Qr ()r + Qv ()v ]
f 0 () = A exp −
00
1+ d . (6.20)
k B Te f f 0 (2m/M) Qm ()
The second term of the integral in Equation 6.20 shows the ratio of the colli-
sional energy loss between inelastic and elastic scattering. Recalling the fact
that the higher the threshold energy, the faster the energy relaxation time, one
would expect that ε/ε j should be weighted by the energy relaxation time for
inelastic scattering and that the effective relaxation time is expressed as

τee f f (ε)−1 = N[(2m/M)Qm (ε) + Q j (ε)ε j /ε]v (6.21)
when referring to inelastic collision processes. The electrons with energy ε

follow the time-varying field closely when ω−1 τee f f (ε).

The macroscopic behavior of the swarm is described in terms of swarm

parameters (transport parameters), which are ensemble averages of quantities
with respect to the time-dependent velocity (energy) distribution.
The mean energy, < ε(t) >, is expressed as

1
< ε(t) > = Re mv g (v, t)dv
2 0
2
v ∞

= Re 2π e2ikωt mv 4 go2k (v)dv
k 0

∞
= Re e 2ikωt
f 02k ()d ; (6.22)
k 0
the ionization rate, Ri (t), as

Ri (t) = Re NQi (v)vg 0 (v, t)dv
v ∞

= Re 4πe2ikωt NQi (v)v 3 go2k (v)dv
k 0

∞
= Re (2/m) 1/2 2ikωt
e NQi () 1/2
f o2k ()d ; (6.23)
k i
and the drift velocity as

Vd (t) = Re vg 0 (v, t)dv
v ∞
4π
= Re ei(2k+1)ωt
v g1 (v)dv
3 2k+1
k
3 0
∞
1
= Re (2/m)1/2 ei(2k+1)ωt 1/2 f 12k+1 ()d . (6.24)
k
3 0
6.3.4 Direct Numerical Procedure

It is sometimes difficult to obtain the time-dependent energy distributions
from the set of Equations 6.15 by using the expansion procedure both in
velocity space and in time as described in the preceding section. Develop-
ments in large memory computing hardware, however, permit the application
of direct numerical methods. We describe a direct numerical procedure (DNP)
for solution of the Boltzmann equation in an rf field [7], and the results are
shown to illustrate the usefulness of the procedure. Because the spatially ho-
mogeneous Boltzmann equation represents the time evolution of the velocity
distribution only in velocity space, the Boltzmann Equation 5.8 is completely

(a)
ω t =0
g0 (υ )
5.0
Vr
(1
5.0
0
8
0.0 0.0
cm
-5.0 -1
Vz (10 8
cm s )
s
-1
)
(b)
0.5 ω t =0
g00 (ε)
Distribution gn0 (ε) (arb.)
g02 (ε)
0
g03 (ε) g10 (ε)
0 10 20
FIGURE 6.13
Electron velocity distribution g 0 (v, ωt √ = 0) (a) and each component of Legendre polynomials
gn0 ()(n = 0, 1, 2, 3) (b) at E 0 /N = 300 2 Td and 1 MHz at 1 Torr in Ar.
equivalent to the expression
g(v + v, t + t) = g(v, t) + J (g, F )t, (6.25)
when t approaches 0. The velocity increment v is related to t as follows,
v = [eE(t)/m]t. (6.26)
Equation 6.25 suggests that it should be possible to calculate the time evolu-
tion from an arbitrary initial distribution to a final periodic steady state under
the rf field by a differential method. It is essential to consider the change both
of the incident angle and of the energy at collision in terms of a deterministic
finite probability for each type of collision. For this purpose, it is convenient to

evaluate the collisional scattering in spherical coordinates. The electron velo-

city also changes, however, due to acceleration or deceleration in the electric
field, and Cartesian coordinates are more suitable for this case. In fact, the
transformation of quantities between the two coordinate systems is a serious
practical issue, and, therefore, a simulation in velocity space instead of energy
space may be essential.
The collision term is expressed in the case of a cold gas, Tg = 0, as
3
v1
J (g, F ) = N g(v1 , t)v1 σel (v1 , θ)d
v
3
v
+N 2
g(v2 , t)v2 σ j (v2 , θ)d
j v

1 v3
+N g(v3 , t)v3 σi (v3 , θ)d
(1 − δ) v

1 v4
+N g(v4 , t)v4 σi (v4 , θ)d
δ v

−N g(v, t)v[σel (v, ϑ) + σ j (v, θ) + σa (v, ϑ) + σi (v, θ)]d.

(6.27)
Here, σ (v, θ ) is a differential cross section, defined as the number of electrons

scattered into a solid angle d = v 2 sin θ dθ dφdv at scattering angle θ . The
subscripts of σ (v, θ ), el, j, a , and i denote elastic, excitation, attachment, and
ionization collisions, respectively. The velocity v1 of the incoming flux prior
to elastic scattering is related to the velocity of the outgoing flux v as
v
v1 = .
1 − (m/M)(1 − cos ϑ)
For excitation, v2 is expressed as v2 = (v 2 + 2ε j /m)1/2 . Similarly, v3 = (v 2 /δ +
2εi /m)1/2 and v4 = (v 2 /(1 − δ) + 2εi /m)1/2 are the velocity relations for two
electrons after ionization. When the scattering is not strongly dependent on
angle θ , we may approximate it by isotropic scattering, which means that
the differential cross section σ is independent of θ. For isotropic scattering,
σ (v, θ ) is replaced by Q(v)/4π, and Equation 6.25 can be rewritten in a form
for numerical calculation of the time evolution:
g(vx , v y , vz +(e E(t)/m)t, t+t) = g(vx , v y , vz , t)+ J (vx , v y , vz , t)t, (6.28)
where

NQm (v1 )v14
J (vx , v y , vz , t) = g(v1 , t)d
4πv 3

Q j (v2 )v22
+N g(v2 , t)d
4π v

NQi (v3 )v32
+ g(v3 , t)d
4πv(1 − δ)


NQi (v4 )v42
+ g(v4 , t)d
4πvδ

−N[Qm (v)+ Q j (v)+Qa (v) + Qi (v)]vg(v, t). (6.29)
Because there is axial symmetry around the z-axis, g is stored only as

a function of vz and vx (or v y ) in such a way that the velocity increment
vz satisfies the relation vz = (e E z (t)/m)t. By so doing, evaluation of
g in Equation 6.28 merely involves a shifting of the two-dimensional array
g(vx , vz ) along vz at each time interval t and addition of the corresponding
collision term J (vx , v y , vz , t)t to each array element. The collision integrals
for each vx and vz at time t are calculated by integrating the distribution of
v and θ over both the velocity v and the polar angle θ . The time step t
in an explicit method is limited to a very short time in order to satisfy the
Courant–Friedrichs–Lewy condition relating to the velocity cell size, v,
[eE(t)/m]t ≤ v, (6.30)
and to satisfy the differential expression
Max[τm (v)−1 ] × t << 1. (6.31)
A method not restricted by the relation in Equation 6.30 has been proposed [8].
In Figure 6.14 we show the √ time modulation of the velocity distributions
g 0 (vz , t) in Ar at E 0 /N = 50 2 Td and 13.56 MHz in a periodic steady state.
74 25.6
6.4
Tim 34 0
e ( 6.4 eV)
ns) gy (
0 25.6 Ener
FIGURE 6.14
√
Time modulation of the electron velocity distribution g 0 (vz , t) at E 0 /N = 50 2 Td and 13.56 MHz
at 1 Torr in Ar.

The distribution at each phase is not symmetric, leading to a directional

velocity component in a direction opposite that of the field and the asym-
metry changes with the field, E 0 cos ωt. At low energy of less than 2 eV,
especially in the Ramsauer–Townsend minimum, the collisional momentum
relaxation is not complete. The energy relaxation time in Ar is much longer
than the one period, 74 ns, except energies greater than the excitation thresh-
old 11.56 eV. As a result, g 0 (vz , t) exhibits temporal characteristics peculiar
to the rf field. Most important, the effects of the limited times of relaxation
become observable for some parts of the periods and also for some parts of
the energy distribution, and the time behavior of the transport coefficients
become quite complex. This leads to general kinetic phenomena, and here
we describe two such processes: the negative differential conductivity and
anomalous anisotropic diffusion intrinsic to the rf field.
6.3.5 Time-Varying Swarm Parameters

The time dependence of most transport coefficients of electron swarms in
rf fields is explained by starting from the instantaneous field approxima-
tion and adding temporal relaxation for momentum and energy. Figure 6.15
exhibits the √ time modulation of the electron swarm parameters in SiH4 at
E 0 /N = 160 2 Td as a function of applied frequency f (= ω/2π ). The ensem-
ble average of the energy < ε(t) > has several interesting characteristics. One
is its time modulation with a frequency of 2ω. Another is a gradual increase
of the minimum value due to failure to achieve full relaxation of the mean en-
ergy. A third is a decrease of the maximum due to the lack of energy gain from
the field during a half period. As a result, the fourth is the deduction of the
amplitude of the modulation and of the mean value. In addition, we observe
a phase delay with respect to the waveform of the external field E 0 cos ωt by
lack of collisions for momentum transfer during a half period with increasing
frequency f (ω/2π).
The drift velocity vd (t) in Figure 6.15 changes its shape and the amplitude
as a function of frequency. At low frequency, we observe an NDC similar to
that in a DC field, for example, at the first and third maximum at 100 kHz.
From the DC E/N dependence of the drift velocity and assuming instan-
taneous relaxation, one will expect symmetric nonsinusoidal profiles of the
drift velocity with two or even three maxima (and minima) depending on the
maximum value of E/N, which is confirmed at a low frequency of 100 kHz
in Figure 6.15. With increasing frequency, the third peak disappears. The first
peak is skewed and then disappears at a much higher frequency. As a result,
a sinusoidal shape with one peak will appear at microwave frequency. The
temporal drift velocity profiles at low and high frequency suggest that the
power deposition to electrons, given by the convolution between the electric
field and the drift velocity, will be quite different from the expected sinu-
soidal profile. As the frequency (in detail ω/N) increases, a phase lag appears
in the drift velocity. This coincides with the beginning of electron trapping.

100 kHz 13.56 MHz 100 MHz 1 GHz

Drift velocity (10 7 cm s -1 )
vde
1.0
0.0
-1.0
0 π 2π 0 π 2π 0 π 2π 0 π 2π
3.0
>ε >
Mean energy (eV)
2.0
1.0
0.0
0 π 2π 0 π 2π 0 π 2π 0 π 2π
(10 22 cm -1 s -1 )
Diffusion coefficients
1.5
NDT
1.0
NDL
0.5
0.0
0 π 2π 0 π 2π 0 π 2π 0 π 2π
1.0
0.8 Ri
Rate (10 6 s -1 )
0.6
0.4
0.2 Ra
0.0
0 π 2π 0 π 2π 0 π 2π 0 π 2π
ωt ωt ωt ωt
FIGURE 6.15
√
Time modulation of the swarm parameters in SiH4 at E 0 /N = 160 2 Td frequencies of 100 kHz,
13.56 MHz, 100 MHz, and 1 GHz.
Finally, collisionless phenomena occur between electrons and molecules

during one cycle of the driving field. The electron swarm behaves as a group
vibrating in a vacuum under the effect of the field E0 cos ωt. The instantaneous
group velocity has the form
vd (t) = −Re[(eEo /mω)cos(ωt + π/2), (6.32)
and its magnitude decreases with ω.
PROBLEM 6.3.1
Discuss how to calculate the velocity distribution g 0 (v, t) of electrons in a periodic
steady state by using a time evolution algorithm of the DNP.

The second kinetic phenomenon observed only in rf fields is the anoma-

lous behavior of the longitudinal component of the diffusion tensor [7, 9], and
the example is shown in SiH4 in Figure 6.15. The effect is observed in the time
dependence of the transverse and longitudinal diffusion coefficients. The in-
stantaneous field approximation gives minima equal to the thermal values
for both components of the diffusion tensor. The transverse component fol-
lows the expected behavior with a somewhat delayed minimum that is not
as deep as predicted due to a failure to relax to the instantaneous field at
this very high frequency. However, the longitudinal component shows two
unexpected aspects. When the electric field changes its sign, a peak, rather
than a minimum, occurs in the longitudinal diffusion coefficient. Sometimes
that peak even exceeds the value of the transverse diffusion coefficient, which
is also unexpected, because we expect the transverse component to be larger
for the cross sections that were used for all mean energies except the thermal
energy (see DL at 13.56 MHz in Figure 6.15). The importance of the effect
of anomalous diffusion is that it is a kinetic effect occurring when the field
changes sign, which cannot be predicted on the basis of any effect for DC
fields. This effect may be relevant for the calculation of power input into
electrons in capacitively and inductively coupled plasmas.
6.4 Ion Transport in DC-Electric Fields

Extending the discussion to all aspects of ion transport in neutral gases would
require another volume [10–13]. Here, we briefly mention some basic prop-
erties of ion transport, avoiding the details of the ion molecule reactions and
interaction potentials as well as other details. The data on the swarm param-
eters in gases up to the early 1980s have been compiled by Mason et al. [14].
A typical example of the cross section of ions in gas is shown in
Figure 6.16. Here, the total momentum transfer cross section is shown to-
gether with the cross section for isotropic-elastic scattering and backward
scattering equivalent to the symmetric charge exchange. In principle, ions
exchange a lot of energy in elastic collisions and therefore require a much
higher E/N to reach nonthermal conditions of energy as compared with elec-
trons. The dependence of the energy distribution of ions on E/N is shown in
O− ions in O2 in Figure 2.5 in Chapter 2. The energy distributions give general
characteristics as a function of relative energy between an ion and molecule.
That is, the relative energy distribution is Maxwellian at a low E/N of 20 Td
and has a broad profile at a moderate E/N where the attractive and repulsive
forces are balanced between the ion and molecule. At higher E/N where a
repulsive force is predominant, the energy distribution will shrink in width.
Examples of the drift velocity of ions are shown in Figure 6.17. The drift
velocity is almost proportional to E/N at E/N < 100 Td. In addition to the
characteristics dominated by the charge transfer, there are many other features

-14 Qe.m.t.
10
Qd.c.t.3 Qc.t.
-15
Cross section (cm )
10
2
-16
10
-17 Qd.c.t.2 Qiso.

10 (Ar+ + Ar)
Qd.c.t.1
-18
10
-1 2 3 4
10 1 10 10 10 10
Relative ion energy εr (eV)
FIGURE 6.16
Cross sections for Ar+ in Ar (—) and CF4 (· · ·).
O-/ O2
+
CH3 /Ar
5
Drift velocity (cm s )
10 Cl-/ Ar
-1
+
O2 / O2
+
Ar / Ar
4
10
2 3
10 10 10
DC-E / N (Td)
FIGURE 6.17
Ion drift velocities in gases as a function of DC-E/N.

NDLL
NDTT
Diffusion coefficients (cm s )

-1
-1
O-/ O2
10
19 Cl-/ Ar
+
O2 / O2
18
10 2 3
10 10 10
DC-E / N (Td)
FIGURE 6.18
Diffusion coefficients of ions in gases as a function of DC-E/N. Transverse diffusion coefficient
(· · ·) and longitudinal diffusion coefficient (—).
that are possible for different ions. For example, ions that are not in their par-
ent gas (such as N+ in N2 or Cl− in Ar) may not have such a large charge
transfer cross section at high energies. This leads to a situation in which the
high energy ions may have a small cross section and may increase their energy
continuously, never coming to quasi-equilibrium with the external field (i.e.,
runaway). Fast neutrals produced in charge transfer collisions should be fol-
lowed separately, as they may participate in a number of processes, including
gas phase excitation (or even ionization), and secondary electron production
and sputtering and etching on a surface. Diffusion coefficients are classified
into longitudinal and transverse coefficients with respect to the direction of
the external electric field. Some diffusion coefficients are shown in Figure 6.18
as a function of E/N.
References
1. Huxley, L.G.H. and Crompton, R.W. 1973. The Drift and Diffusion of Electrons in
Gases. New York: Wiley Interscience.
2. Dutton, J. 1983. J. Phys. Chem. Ref. Data 4: 577; Gallagher, J.W., Beaty, E.C., Dutton,
J., and Pitchford, L.C. J. Phys. Chem. Ref. Data 12:109.
3. Tagashira, H., Sakai, Y., and Sakamoto, S. 1977. J. Phys. D 10:1051; Taniguchi, T.,
Tagashira, H., and Sakai, Y. J. Phys. D 10:2301.

4. von Engel, A. 1983. Electric Plasmas: Their Nature and Uses. London: Taylor &
Francis.
5. Goto, N. and Makabe, T. 1990. J. Phys. D 23:686.
6. Margenau, H. and Hartman, L.M. 1948. Phys. Rev. 73:309.
7. Maeda, K., Makabe, T., Nakano, N., Bzenić, S., and Petrović, Z.Lj. 1997, 1994.
Phys. Rev. E 55:5901; Maeda, K. and Makabe, T. Physica Scripta T53:61; Maeda, K.
and Makabe, T. Europhys. Conf. Abstr. 18E:151.
8. Matsui, J., Shibata, M., Nakano, N., and Makabe, T. 1998. J. Vac. Sci. Technol.
A 16:294.
9. White, R.D., Robson, R.E., and Ness, K.F. 1995. Aust. J. Phys. 48:925.
10. McDaniel, E.W. and Mason, E.A. 1973. The Mobility and Diffusion of Ions in Gases.
New York: John Wiley & Sons.
11. Mason, E.A. and McDaniel, E.W. 1988. Transport Properties of Ions in Gases.
New York: John Wiley & Sons.
12. Farrar, J.M. and Saunders, W.H. Jr. Ed. 1988. Techniques for the Study of Ion-
Molecule Reactions. New York: John Wiley & Sons.
13. Lindinger, W., Mark, T.D., and Howorka, F. Ed. 1984. Swarms of Ions and Electrons
in Gases. Wien: Springer Verlag.
14. Mason, E.H. and Viehland, L.A. 1976, 1978, 1984. Atomic Data and Nuclear Data
Tables, 17: 177; Ellis, H.W., McDaniel, E.W., Albritton, D.A., Viehland, L.A., Lin,
S.L., and Mason, E.A., ibid. 22:179; Ellis, H.W., Thackston, G., McDaniel, E.W.,
and Mason, E.A. ibid. 31:113.

7
Modeling of Nonequilibrium
(Low-Temperature) Plasmas
7.1 Introduction
Early attempts to model plasmas were based on zero-dimensional and
phenomenological models. An example of this approach is the circuit model,
which represents plasma behavior using passive circuit elements, resistance,
capacitance, and inductance. Although such an approach is useful for repre-
senting specific plasmas and studying how they behave once connected to
the external circuit, it is beyond the scope of this chapter to specify plasma
characteristics intrinsic in a feed gas molecule. There is a rate equation model
describing zero-dimensional plasma, which has proven quite useful for the
kinetic optimization of a gas laser, mainly by utilizing a bulk plasma.
During the 1980s and early 1990s, it was recognized that plasma-enhanced
chemical vapor deposition of amorphous silicon and plasma etching for
integrated circuits could not be improved further by a trial-and-error
approach, and that a complete understanding of these processes was required
for a new generation of technologies. This was partially motivated by the
increasing cost of the development of the equipment and complexity of reac-
tion chambers.
The need for radio frequency (rf) plasmas arose from the fact that, in these
technologies, one needs also to treat metals, semiconductors, and dielectrics
at the same time. There is an important difference between modeling of direct
current (DC) and alternating current (AC) discharge plasmas, especially those
operating in an rf range. Rfs have periods on the order of nanoseconds, much
shorter than the characteristic lifetimes of many of the gas-phase processes,
and yet the discharge clearly develops and exhibits complex behavior on such
short time scales, so the characteristic time step of the numerical simulation
must be significantly shorter than the period. At the same time, one must take
into account slow processes such as the kinetics of chemically active neutrals
and the flow of gas through the system that will occur on time scales from
milliseconds to seconds. As a result of this vast difference in time scales, we
need to cover as many as 10 orders of magnitude in time development. Thus
the rf plasma for material processings belongs to the class of the stiff system,
and special attention should be paid to the numerical techniques or physical
205

approximations. Under these circumstances, the basic numerical difficulty

that sometimes limits the stability and accuracy of computer simulation will
arise from the orders-of-magnitude difference among the time constants (re-
laxation times) associated with the governing differential equations. The gov-
erning equation system is classified as a set of stiff differential equations [1].
Another issue associated with the use of rf plasma is that, due to the
change of the amplitude and the direction of the local field, the properties of
electrons and ions will not be relaxed to the instantaneous electric field, or,
at the very least, temporal relaxation will take up a considerable part of the
period. Both DC and rf discharge plasmas, however, engender the problems
with the spatial relaxation of the properties of charged particles, particularly
electrons.
As a result of an extensive scientific effort, a number of low-temperature
plasma models were developed in the 1980s and early 1990s. The general
characteristics of these models are that they are predominantly numeric and
that they (to differing degrees) try to represent the entire plasma, often in-
cluding external circuits. In this chapter, we present several techniques for
modeling low-temperature (nonequilibrium) plasmas [2, 3]. See Figure 7.1.
Governing Equation System in VicAddress

Data Base Governing Equations Source
Collision/Reaction
Processes Boltzmann Eq. Maxwell’s Eq.
Conservation Poisson’s Eq. Voltage
of Particles n-E field V (t)
electrons neutrals
Gas-Phase p,n-ions (radicals)
e+A Q(ε) PIC/MCS Faraday’s Law Current
A++B- Momentum induced B,E fields I (t)
A+AB Kj Relaxation
electrons p,n-ions
ejected-particles
Ampere’s Law Antenna
Energy Relaxation J-induced B fields P (t)
electrons
Interface
(surface)
Navier-Stokes Eq.
Y (e, θ) Feed Gas
St(e, θ) Φw Kr
Gas flow
Solid-Phase Boundary Condition (Reactor, wall, wafer, ...)

ΦB ε Feature profile
Surface evolution Eq.
Etching, Deposition
Current damage Schrödinger Eq.
Heat transfer Eq. Thermal damage
FIGURE 7.1
Governing equation system of a low-temperature plasma for material processing.

Modeling of Nonequilibrium (Low-Temperature) Plasmas 207
7.2 Continuum Models

Continuum models are based on the moments of the Boltzmann equation for
both charged particles and neutral molecules. As outlined in Chapter 5, these
moments are obtained by integrating the Boltzmann equation with various
powers of velocity (see Section 5.4). The corresponding conservation laws
are

particle number density ⇐ (Boltzmann Eq.) dv (7.1)

momentum balance ⇐ (Boltzmann Eq.) vdv (7.2)

energy balance ⇐ (Boltzmann Eq.) v2 dv. (7.3)
These laws should be written for all particles involved and should be
completed by a set of Maxwell’s equations. This model treats particles as
fluids. Continuum models have a key and practical advantage of having a
short computation time, and the validity is kept under conditions that the
mean free path of charged particles, especially electrons, is shorter than or
equivalent to the characteristic length of a plasma reactor. In principle, these
Equations 7.1 to 7.3 are not closed, so additional equations should be added.
For example, in the case of thermal equilibrium, the equation of state is used to
complete the system. In nonequilibrium plasmas, the electron energy distri-
bution, which is dominated by the short-range electron–molecule collisions,
is non-Maxwellian, and we need a procedure to find solutions to the corre-
sponding transport (Boltzmann) equation.
Typically, nonequilibrium rf plasma consists both of a bulk plasma
with quasi-neutrality between electrons and positive ions and of ion sheaths.
Both regions are periodically modulated in time even when quasi-stationary
conditions are achieved. In order to deal with such a complex system, we
need to establish a modeling hierarchy that involves several different
models.
7.2.1 Governing Equations of a Continuum Model

The minimum set of equations required to describe nonequilibrium plasmas
in time and configuration space, as mentioned above, consists of continuity
equations (i.e., conservation of number density) for electrons, of positive and
negative ions, and of Poisson’s equation.

In the one-dimensional space along the axial field direction (z-axis), the
continuity equation for electrons has the form
∂ne (z, t) ∂ne (z, t) ∂ 2 ne (z, t)

= −Vde (z, t) + DLe (z, t)
∂t ∂z ∂z2
+ {Ri (z, t) − R a (z, t)}ne (z, t) − Rr e (z, t)ne (z, t)n p (z, t). (7.4)
For positive ions it is given by
∂n p (z, t) ∂n p (z, t) ∂ 2 n p (z, t)

= −Vd p (z, t) + DL p (z, t) + Ri (z, t)ne (z, t)
∂t ∂z ∂z2
− Rr e (z, t)ne (z, t)n p (z, t) − Rri (z, t)nn (z, t)n p (z, t), (7.5)
and for negative ions it is given by
∂nn (z, t) ∂nn (z, t) ∂ 2 nn (z, t)

= −Vdn (z, t) + DLn (z, t)
∂t ∂z ∂z2
+ R a (z, t)ne (z, t) − Rri (z, t)nn (z, t)n p (z, t). (7.6)
Poisson’s equation for the electric field E(z, t) and the space potential V(z, t) is
∂ E(z, t) ∂ 2 V(z, t) e
=− = {n p (z, t) − ne (z, t) − nn (z, t)}. (7.7)
∂z ∂z2 ε0
Here, the number densities of electrons, positive ions, and negative ions are
ne (z, t), n p (z, t), and nn (z, t), respectively; Vdj and DL j are the drift velocity
and the longitudinal component of the diffusion tensor of the jth particles
(e, p, n); Ri and R a are the ionization and electron attachment rates; and Rr e
and Rri are rates of recombination in electron–ion and ion–ion collisions. The
data for transport coefficients Vdj , DL j , Ri , and so on, which are taken from a
database of experimental or theoretical data, are provided for the local and in-
stantaneous electric field E(z, t). One example of a set of data for the collision
rate for O2 is shown in Figure 7.2. In Equation 7.7, e(> 0) is the elementary
charge and ε0 is the permittivity (dielectric constant) of the vacuum.
The set of Equations 7.4 to 7.7 is numerically calculated for a time-varying
one-dimensional capacitively coupled plasma (CCP) with infinite parallel
plates. In a finite plasma source or axisymmetric plasma, these equations are
developed in the two- or three-dimensional configuration space.
PROBLEM 7.2.1
Derive the continuity of the current density in the one-dimensional position space
∂ JT ∂
= J e (z, t) + J p (z, t) + Jn (z, t) + JDis (z, t) = 0 (7.8)
∂z ∂z
from the Equations 7.4 to 7.7. Here Je , J p , and Jn are the current density of elec-
trons and of positive and negative ions, respectively. J Dis is the displacement current
density.

FIGURE 7.2
Collision rates of the electron in pure oxygen as a function of DC-E/N.
We need information on the induced magnetic and electric fields, B(t)

and E(t), in addition to the static Poisson field in the case of inductively cou-
pled plasmas (ICPs), and for that purpose we must solve Maxwell’s equa-
tions. In the system it is necessary to extend the calculation to the two-
or three-dimensional position space. For cylindrical coordinate (r, θ, z), the
Equation 7.7 is replaced by the following system of Maxwell’s equations,
which includes Poisson’s equation:
1 ∂ 1 ∂E θ (r, θ, z, t) ∂E z (r, θ, z, t)
(r Er (r, θ, z, t)) + +
r ∂r r ∂θ ∂z

1 ∂ ∂V 1 ∂2V ∂2V
=− r − 2 −
r ∂r ∂r r ∂θ 2 ∂z2
e
= (n p − ne − nn );
ε0
(7.9)
Faraday’s law:
1 ∂E z ∂E θ Br
−
r ∂θ ∂z
∂Er ∂E z ∂
− =− Bθ ; (7.10)
∂z ∂r ∂t
1 ∂ 1 ∂Er
(r E θ ) − Bz
r ∂r r ∂θ

Ampere’s law:
1 ∂Bz ∂Bθ Er Jer J pr
−
r ∂θ ∂z
1 ∂Br ∂Bz ∂
− = −ε0 E θ + Jeθ + J pθ ; (7.11)
µ0 ∂z ∂r ∂t
1 ∂ 1 ∂Br
(r Bθ ) − Ez Jez J pz
r ∂r r ∂θ
and Coulomb’s law in magnetics:
1 ∂ 1 ∂Bθ ∂Bz
(r Br ) + + = 0; (7.12)
r ∂t r ∂θ ∂z
where µ0 is the magnetic permeability of vacuum and where µ0 ε0 = c −2
(c is the speed of light).
PROBLEM 7.2.2
Derive the following continuity equation for a short-lived excited molecule with a
number density Nj (k)
∂ Nk
Nk (t) = j k jk ne Nj + l m klk Nl Nm − j l kk j Nk Nl − k , (7.13)
∂t τrad
where k jk is the collisional rate constant from the state j to k. τrad

k
is the radiative life-
time of the molecule in the k state. The system describing a linkage among collisional
excited molecules Nj ( j = 1, 2, . . .) with short lifetimes is given by the simultaneous
equations similar to Equation 7.13. The model is called a collisional radiative (CR)
model.
7.2.2 Local Field Approximation (LFA)

When the transport coefficients in the continuity Equations 7.4 to 7.6 are
given at each time as a function of reduced field strength E(z, t)/N, given
by the number density of the feed gas N and the local instantaneous field
E(z, t), we can solve the governing Equations 7.4 to 7.7 and obtain the number
densities of charged particles and the space potential. This approximation,
known as the local field approximation (LFA), assumes a quasi-equilibrium of
charged particles at all points in space and for all times. As mentioned earlier,
this assumption allows the direct and fastest possible application of swarm
data through an appropriate database. However, when the external driving
frequency of the plasma increases, the temporal variation of the electric field
occurs over a very short period of time, that is, comparable with or even
shorter than the relaxation times for charged particles in an rf plasma, and the
spatial variation of the field will be less than the relaxation length. Application
of LFA becomes more difficult as the gas number density N is reduced, because
the collisional relaxation time of the charged particle is proportional to N−1 .

Even for higher neutral density (pressure), the spatial variation of the field
in the sheath close to the electrode is very large and the application of LFA
for these regions will be inaccurate. In practical terms, the LFA algorithm
provides very fast numerical calculation and thus should be the first algorithm
to be attempted whenever there exists a database of the transport (swarm)
parameters.
7.2.3 Quasi-Thermal Equilibrium (QTE) Model

When electrons are in thermal equilibrium in a plasma, their energy dis-
tribution is Maxwellian. Maxwellian distribution is determined by a single
parameter, electron temperature Te , which significantly simplifies the calcu-
lation of collision rates. On the other hand, the temperature will be spatially
dependent, and it should certainly depend on the time in an rf plasma, that
is, Te (z, t). Nevertheless, this simplifies the calculation tremendously. If we
assume that the cross sections for the ionization and electron attachment are
constant with threshold energies εi and εa , respectively, we obtain rates in a
very simple analytic form. For the ionization rate,

εi
Ri (z, t) = Nki exp − , (7.14)
kTe (z, t)
where k is the Boltzmann constant and ki is the constant intrinsic to the feed
gas. The electron temperature is obtained from the energy 3 kTe (z, t)/2 con-
servation equation given as

∂ 3 ∂q e (z, t)
ne (z, t)kTe (z, t) +
∂t 2 ∂z

∂ne (z, t)
= e E(z, t) −ne (z, t)Vde (z, t) + DLe
∂z

εi
− εi ne (z, t)Nki exp −
kTe (z, t)

εa
− εa ne (z, t)Nka exp −
kTe (z, t)

εj
− ε j ne (z, t)Nk j exp − , (7.15)
kTe (z, t)
where the term q e (z, t) is the enthalpy defined for electrons as

5 ∂ne (z, t) 3
q e (z, t) = ne (z, t)Vde (z, t) − DLe kTe . (7.16)
3 ∂z 2
This model is based on the second-order velocity moment, that is, energy
balance. It is applicable to a plasma modeling, if one can represent data

TABLE 7.1
Low-Temperature Plasma Model and the Variable
Model Variable References
LFA E(r, t)/N Boeuf [4]

QTE < εe (r, t) > Graves [5]
RCT E emf f (r, t)/N for momentum Makabe [6]
E ee f f (r, t)/N for energy
Phase space εe (r, t)/N Sommerer [7]
Particle εe (r, t), (ε p (r, t)) Birdsall [8]
Hybrid — Kushner [9]
Circuit V(t) or I (t) —
for the drift velocities and diffusion coefficients Vd j and DL j as a function

of mean electron energy or, consequently, electron temperature Te (z, t). The
QTE model assumes no phase delay between the mean electron energy εm (z, t)
given by the bulk of the energy distribution and the ionization rate Ri pro-
duced by high-energy electrons to be small, whereas in reality it is large and
thus the ionization in a plasma will be overestimated.
Conditions that favor applicability of QTE are those in which there is a
high degree of ionization so that Coulomb forces couple charged particles.
In these circumstances, electrons lose their energy through the long-range
interactions in e–e scattering and in e–p scattering, which can be represented
as a large number of momentum transfers with small scattering angles (see
Chapter 4). In such collisions, electrons are thermalized and reach a velocity
distribution closer to the Maxwellian. The conditions in a low-temperature
plasma used in integrated circuit production or thin film preparation and
coating, and so on, range mainly from a low degree of ionization in which
the velocity distribution of electrons is in nonthermal equilibrium to a high
degree in which it is Maxwellian. Thus one must pay attention to which
circumstances apply.
The shape of the Maxwellian energy distribution does not depend directly
on the cross sections of the feed gas (see Section 2.3.2), where the Maxwellian
is a logarithmically straight line depending only on the electron temperature
Te . By contrast, nonequilibrium energy distribution is strongly affected by the
intrinsic short-range collisions and the cross sections, and the energy tail of
the distribution will drop down very rapidly and to a much greater degree
than in the Maxwellian distribution in the range greater than the threshold
energy of the electronic excitation collisions. See Table 7.1.
7.2.4 Relaxation Continuum (RCT) Model

As mentioned earlier, an rf plasma maintained externally by an rf power
source has two distinct regions: a positive ion sheath and a quasi-neutral
bulk plasma. In the sheath region, the electric field is very strong and changes
rapidly both in space and time. It is usually impossible for charged particles

to come to equilibrium under such conditions. In electropositive gases, the

electron relaxation process is the most important in sustaining the rf plasma,
whereas in electronegative gas both the massive positive ions and the negative
ions should be considered in addition to the light electrons.
The relaxation continuum (RCT) model considers the momentum and
energy relaxations in a simple and physically natural way. First we consider
the momentum relaxation. From Equation 7.2, we write the momentum con-
servation in the form of a relaxation equation:

∂
ne (z, t)mVde (z, t)
∂t
ne (z, t)mVde (z, t) ∂
= eE(z, t)ne (z, t) − − ne (z, t)mVde (z, t) Vde (z, t), (7.17)
τm ∂z
where τm is the effective time constant for momentum transfer (the inverse
of the momentum transfer collision rate and approximately equal to mVde0 /eE
by use of the DC-drift velocity Vde0 ). The first term on the right-hand side is
the momentum gain due to electric field acceleration, the second term is the
loss due to momentum transfer collisions, and the third term describes the
effect of the spatial gradient of drift velocity on relaxation.
The relaxation of the mean energy of electrons, εm (z, t), is described by the
conservation of energy,

∂ ∂ne (z, t)
ne (z, t)εm (z, t) = eE(z, t) − ne (z, t)Vde (z, t) + DLe
∂t ∂z

2m
− Rm εm (z, t) + R j ε j + Ri εi ne (z, t)
M

∂ ∂ne (z, t)
− ne (z, t)Vde (z, t) − DLe εm (z, t) . (7.18)
∂z ∂z
The r.h.s in Equation 7.18 is described by the terms of the energy gain due
to flux of charged particles in the electric field, followed by energy losses
in elastic momentum transfer, inelastic and ionization collisions, and finally
the spatial gradient of the mean energy. We introduce the effective field E eff ,
where E(z, t)2eff = εm /e µτe , which gives a field modified from the actual field
E(z, t) to take into account the energy relaxation from the proper transport
data under quasi-equilibrium. The relaxation equation for the effective field
may be written as

∂ E eff (z, t)2 − E(z, t)2 ne (z, t) ∂

E eff (z, t) ne (z, t) =
2
− q e (z, t), (7.19)
∂t τe ∂z
where the enthalpy is defined as
∂
q e (z, t) = ne (z, t)Vde (z, t)E eff (z, t)2 − DLe (z, t) ne (z, t)E eff (z, t)2 . (7.20)
∂z

-15 Qv3
10
Qm
Qil
-16
Qv1 Qdl
Cross section (cm )
10
2
Qd3
Qv4 Qex
Qd2 Qi3
-17
10
Qi4
Qi5
Qi2
-18
10 Qi6
Qi7
Qa
-19
10 -2 -1 2
10 10 1 10 10
FIGURE 7.3
Set of collision cross sections in CF4 as a function of electron energy.
The temporal relaxation time for electron diffusion is of the same mag-
nitude as that for the mean energy. However, the energy relaxation time of
electrons with energy greater than the excitation threshold j is more diffi-
cult to estimate than those for drift velocity and mean energy, which are the
result of the effect of the entire energy distribution of electrons. In particular,
it is important to establish the relaxation time constant of the ionization rate
τ Ri . The ionization is the result of only the electrons at the highest end of the
energy distribution, and these have kinetics quite different from the bulk of
the distribution. Therefore, for τ Ri we must make an estimate from the relax-
ation times for the mono-energetic groups of electrons (see Figure 7.3). The
RCT model is a theory that maintains all the advantages of continuum models
with respect to speed, simplicity, and ease of interpretation.
PROBLEM 7.2.3
Discuss the time behavior of the drift velocity of electrons as a function of external
field E(t) = E 0 cos ωt from the momentum relaxation Equation 7.17 in the case of
τm
ω−1 (a) and τm ω−1 (b). For simplicity, we assume that the electron number
density and the drift velocity are distributed uniformly in space.
7.2.5 Phase Space Kinetic Model

In order to determine the transport and rate coefficients of electrons precisely
in space and time, we must solve the equation for the electron transport
in phase space. We have already shown methods to solve the Boltzmann

equation (see Section 5.6), but the most difficult issues are the need for huge
CPU time and the possibility of numerical diffusion if the resolution is
reduced in order to speed up the computation. Here we discuss the two-
term approximation (TTA) as a means of solving the Boltzmann equation for
the time- and space-modulated electric field E(z, t). The equation for the en-
ergy distribution, f (ε, z, t), based on the TTA of the velocity distribution of
electrons is
∂
f (ε, z, t)
∂t

2 2m ∂ Me E(z, t)ε ∂ ∂
= NQm (ε)ε 3/2 f + eE (ε −1/2 f )+ (ε−1/2 f )
m M ∂ε 6mNQT (ε) ∂ε ∂z

2 ε ∂2 ∂2
+ eE(z, t) (ε −1/2 f ) + 2 (ε −1/2 f )
m 3NQT (ε) ∂ε∂z ∂z
ε + εj εa
2 ∂ 2 ∂
+ ε 1/2 NQ j (ε) f dε − ε 1/2 NQa (ε) f dε
m j ∂ε ε m ∂ε 0
1+k

(1 + k)ε + εi k ε + εi
2 ∂
+ + ε 1/2 NQi (ε) f dε . (7.21)
m ∂ε ε 0
Here, the energy distribution function f (actually the probability density) is

normalized as f (ε, z, t)dε = 1. The sum of the cross sections Qm , Q j , Qa ,
and Qi is the total cross section QT = Qm + Q j + Qa + Qi . j , i , and a are
the threshold energies of the excitation, ionization, and electron attachment,
respectively. (1:k) is the ratio of the energy partition between two electrons
after ionization.
Equation 7.21 should be included within Equations 7.4 to 7.7 to make
a complete set through the transport coefficients given from the integrated
forms of f (ε, z, t). In TTA, the longitudinal diffusion coefficient DL e is not
well defined and should be estimated, as it is the most important com-
ponent of the diffusion tensor in plasma modeling. However, the problem
is also to properly represent the boundary conditions. In addition, in the
region between the sheath and bulk plasma, where the net ionization rate is
at a maximum, the velocity distribution of electrons is far from spherical and
the accuracy of the TTA will go down.
We have already shown the (DNP) numerical procedure to solve the
Boltzmann equation (see Sections 5.6 and 6.3). One of the procedures to over-
come the appearance of the anisotropic velocity distribution in a high field
and low-pressure region is the use of the DNP of the Boltzmann equation
(see Section 6.3.4). In these circumstances, the most difficult issues are the
need for huge CPU time and the possibility of numerical diffusion at low
resolution.

7.3 Particle Models

At low pressures the mean free path of electrons λe becomes comparable
with or even larger than the characteristic dimension of the reactor d,
λe ≥ d, (7.22)
and then the characteristics of plasma depend greatly on particle–surface in-

teraction. It is difficult for electrons to relax their energy and momentum by
binary collisions with the feed gas molecules. Thus, in the gas phase, electron
transport strongly depends on previous history (non-Markovian process, or
processes with memory), that is, on when and where the electron was pro-
duced. Under such conditions it is difficult to treat plasmas as fluids by using
a continuum model.
An rf field and magnetic field are able to spatiotemporally trap electrons
and ions. Effective pressure of the feed gas under an rf or a magnetic field (or
magnetic fields) will become higher as the path of the electron is curved in
such a way that the mean free path may be reduced by an increasing possibil-
ity of collision with the feed gas. In a magnetic field strong enough to confine
charged particles, it is possible to describe transport by the Boltzmann equa-
tion, but the fluid description becomes increasingly difficult at lower pressure
due to the increase of the transport coefficient and the complex behavior of
the charged particles.
Under such conditions the so-called particle model is better. The model
may be applied at higher pressure as well. In principle, this model treats indi-
vidual particles, and it is based on some form of Monte Carlo simulation [10]
and treatment of individual collisions in the gas phase and at surfaces. The
main advantage of a particle model as opposed to a fluid model is that no as-
sumption for the velocity or energy distribution of charged particles is made
implicitly.
7.3.1 Monte Carlo Simulations (MCSs)

The binary collision between particles in a low-temperature plasma is a
stochastic process described by the set of collision cross sections (i.e., collision
probabilities). The collision events are simulated by Monte Carlo Simulation
(MCS) by using the pseudo-random numbers generated in a computer under
the database of the set of collision cross sections in the feed gas (see Figure 7.4).
MCS traces the trajectory of a number of particles under the external electric E
or magnetic field B by considering the collision event as a function of kinetic
energy. In the case of an electron at (v, r, t), for example, the trajectory of the
electron at t + t is tracked by Newton’s equation
d2 e
2
r(t) = [E(r, t) + v × B(r, t)] , (7.23)
dt m

and during the short period t, the collision occurs when the total collisional
probability NQT (ε)vt satisfies
NQT (ε)vt ≥ ξr . (7.24)
Here, {ξr } is the random number distributed uniformly in [0, 1]. Of course,
the time step t must satisfy the condition
1
t
. (7.25)
NQT (ε)v
PROBLEM 7.3.1
The probability that the electron will have a collision in the time interval (0, t) is
given by
t
p(t) = NνT (ε(t))exp − NνT (ε(t))dt ,
0
where νT (ε(t)) is the time-dependent total collisional frequency and given by

QT (ε(t))v(t). Derive the above relation.
At the same time, the scattering angle (θ, φ) after the collision must be
determined by the random number. In the case of a collision, the energy and
(a)
Qm Qex1 Qex2 … Qi
collision probability
NQT v∆t
uniform random number

0 ξr 1.0
(b) ϕ
v'
v (θ, ϕ) = ( cos-1 (1-2γr), 2πγr)

collision point
FIGURE 7.4
Determination of the collision and the kind by the random number {ξr } (a) and the scattering
angle at collision {γr } (b).

the velocity of the particle change, subject to the momentum and the energy
conservation rule between two colliding particles (see Chapter 4). The trace
of the electron with (v, r, t + t) after the collision restarts during the next
t, and the same procedure is carried out.
Exercise 7.3.1
An electron has a binary collision with a neutral molecule and is scattered
isotropically. Derive the distribution of the scattering angle (θ, φ) with respect
to the incident direction given by
1
p(θ ) = sin θ, (7.26)
2
1
p(φ) = . (7.27)
2π
The function p(θ, φ) (the probability density function) distributed uniformly
in a unit sphere (1, θ, φ) is expressed by
sin θ dθ dφ
p(θ, φ) = .
4π
The θ and φ in the spherical coordinates are independent of each other, and
we can write
p(θ, φ)dθ dφ = pθ (θ )dθ pφ (φ)dφ,
where pθ (θ ) and pφ (φ) are normalized to unity:

π 2π
pθ (θ )dθ = 1 and pφ (φ)dφ = 1. (7.28)
0 0
Therefore,
2π
2π sin θ dθ 1
pθ (θ )dθ pφ (φ)dφ = = sin θ dθ, (7.29)
0 4π 2
π
2dφ 1
pφ (φ)dφ pθ (θ )dθ = = dφ . (7.30)
0 4π 2π
Exercise 7.3.2
Obtain the scattering angle (θ, φ) distributed isotropically by using a series
of random numbers {γr } uniformly distributed between [0, 1].
From Equation 7.29,
θ
1
sin θ dθ = γr → cos θ = 1 − 2γr . (7.31)
0 2

In the same way,

φ
1
dφ = γr → φ = 2π γr . (7.32)
0 2π
An expression of the stochastic judgment is possible by using additional ran-

dom numbers to more physically precise or complicated collision processes,
for example, anisotropic scattering or energy share to two electrons after
ionization.
The formulae used to sample transport coefficients are taken from the
stochastic theory. We can also determine the energy distribution and velocity
distribution of electrons. The transport coefficient of electrons is determined
from the position r in the mean position of the electron swarm. That is, the
effective ionization rate is given by
1 d
R0 = ne (r, t), (7.33)
ne (r, t) dt
the drift velocity is obtained as
d
vd = r(t) , (7.34)
dt
and the diffusion tensor is given by
1d
D= (r(t) − r(t) )2 . (7.35)
2 dt
7.3.2 Particle-in-Cell (PIC) and Particle-in-Cell/Monte Carlo Simulation

(PIC/MCS) Models
A high-density plasma source in a low-temperature plasma is operated at
low pressure, sometimes under the support of an external magnetic field. A
strongly ionized plasma consists of electrons and ions interacting through
the electromagnetic fields. The charged particles are subject to the long-range
Coulomb interaction among particles. These plasmas are classified as colli-
sionless plasmas. It is quite difficult to trace the motion of a huge number of
charged particles in the collisionless plasma. One of the practical methods to
overcome this difficulty is the use of the particle-in-cell (PIC) model [8]. The
PIC model considers:
i. A finite number of superparticles, consisting of h electrons (ions)

with constant (e/m) rather than all the electrons and ions in the
reactor; and
ii. A particle cloud with a finite spread in place of all the independent
particles.
Then, the superparticle with charge eh is traced by using Newton’s equation

under the external field at time step t. When a superparticle is found at z j ,

a eh ∆z - a
z
zi zj z i+1
∆z
FIGURE 7.5
Superparticle and the charge sharing between the nearest two grids.
at distance a from the lattice z j in Figure 7.5, the charge is divided into aeh
and (z − a )eh at z j and z j+1 , respectively. The electric field at the position of
the superparticle is given by linear interpolation, after the field is calculated
by Poisson’s equation at each of the lattices
E(z j ) = a E(zi ) + (z − a )E(zi+1 ) . (7.36)
In order to avoid numerical instability in the explicit particle simulation, the

time step t and mesh space z are respectively taken as
t < ω−1
pe , (7.37)
z < λ D , (7.38)
where ω pe and λ D are the plasma frequency and Debye length in a collisionless
plasma (see Chapter 3).
When a plasma is maintained by electron impact ionization of a feed gas
molecule, we must consider the short-range interaction even in the case where
a strongly ionized plasma is mainly subject to the long-range interaction.
The binary collision is considered by MCS. Thus the model is classified as a
PIC/MCS model. PIC/MCS is widely used in low-pressure plasmas. When
the operating pressure of a plasma increases, the number of collision events
between the electron and feed gas molecule is huge, and thus the continuum
fluid model is much more appropriate for describing the system. The main
advantage of the particle model as opposed to the fluid model is that, in
the former, no assumption for the velocity or energy distribution of charged
particles is made implicitly. In compensation, the cost is high because the PIC
model is very time consuming.
7.4 Hybrid Models

As described in the section on particle models, one of the difficulties in apply-
ing the particle model to a practical system is the cost of the computational
time. One of the proposed means of overcoming this difficulty is to combine
the continuum model with the particle model.
An example of such combination is the LFA model coupled with the
MCS of electrons to identify the nonhydrodynamic behavior of the transport

coefficient. Another is the system combining the PIC/MCS model of electrons

and the RCT model of massive ions in a low-pressure plasma, especially in
a magnetron plasma, where the drift and diffusion of electrons exhibit local
and complex characteristics (see Chapter 11).
7.5 Circuit Model

Electrical properties, such as the sustaining voltage Vsus (t), discharge cur-
rent I (t), dissipated power W(t), and space potential V(z) are the basis for
the understanding of discharge plasmas. These electrical properties have
been modeled by using the electrical impedance of the sheath and bulk
plasma [11, 12]. A knowledge of the impedance is related to the plasma pa-
rameters and is used for the impedance matching of the rf power supply (see
Chapter 9). The estimation of the plasma characteristics through the electri-
cal impedance is achieved with an equivalent circuit model of the rf plasma
(see Figure 7.6).
7.5.1 Equivalent Circuit Model in CCP

We first consider a typical CCP sustained between two electrodes. There exist
a quasi-neutral bulk plasma and two ion sheaths in front of both electrodes.
In a uniform bulk plasma with electrons and positive ions, the current is
composed of the conduction current of electrons and ions, Je (z, t) and J p (z, t),
and the displacement current, J Dis (z, t). Due to the great difference of the
mass between the electron and the ion, the ion current will be neglected as
compared with the electron in electropositive plasmas. Then,
J T (z, t) ∼ Je (z, t) + JDis (z, t).
power matching sheath bulk sheath

source network plasma
R sh1 Rb R sh2
L1 Cb
IT
50 Ω Vsus
C1 C2 C sh1 Lb C sh2
FIGURE 7.6
Typical equivalent circuit of a capacitively coupled rf plasma with an external impedance match-
ing network.

In the uniform bulk plasma, a drift current is dominant and by considering

the phase delay φ of the drift velocity of electrons vde (t) with respect to the
field with sinusoidal waveform Eb exp( jωt), the momentum conservation of
electrons is
d
{mvde exp j (ωt − φ)} = eEb exp( jωt) − mvde exp[ j (ωt − φ)]Rm ,
dt
where Rm is the total collision rate of electrons. Then, the electron conduction
current density is given as
ne e 2
Je (t) = ene vde (t) = Eb exp( jωt)
m(Rm + jω)

ne e 2 Rm 1 1
= −j Eb exp( jωt) .
mω ω 1 + ( Rωm )2 1 + ( Rωm )2
Then, the total bulk plasma current density is

ne e 2 Rm 1 1 ε0 m 2
J T (t) = −j +j ω Eb exp( jωt). (7.39)
mω ω 1 + ( Rωm )2 1 + ( Rωm )2 ne e 2
As a result, the first in-phase oscillation term gives a power dissipation at the
resistive impedance (plasma resistance Rb ), the second and third out-of-phase
oscillations correspond to an inductive and capacitive impedance ( jωL b and
1/jωCb ) with no power dissipation (reactance).

mω ω Rm 2 db ω 2 1 Rm 2 ε0 Sb −1
Rb = 1+ = 1+ ,
ne e 2 Rm ω Sb ω pe Rm ω db
(7.40)
2 2 −1
mω0 Rm db 1 Rm ε0 Sb
Lb = 1+ = 2 1+ , (7.41)
ne e 2 ω Sb ω pe ω db
0 Sb
Cb = , (7.42)
db
where ω2pe ( = e 2 ne /mε0 ) is the electron plasma frequency. Sb and db are the
cross sectional area and thickness of the bulk plasma. In fact, the third in
Equation 7.39, displacement component will be negligible in electropositive
plasmas. In a collision-dominated plasma with Rm ω, the impedance of the
bulk plasma is represented by a resistive component R b .
In the sheath of a typical rf discharge plasma, the circumstances change
the relation completely among the current components, Je , J p , and JDis . The
sheath region having n p ne is formed in front of electrodes or a wafer in a
discharge plasma, and the displacement current occupies a large part of J T ,
although finite currents of positive ions and electrons flow under a DC self-
bias voltage in a time-averaged fashion with (Je (t) + J p (t)) = 0. As a result,
the equivalent circuit of the sheath having a voltage drop Vsh is composed

power matching current plasma

source network coil
L1 Cb
I sus
M
Rc
50 Ω Vsus Lc Lp Rp
C1 C2
FIGURE 7.7
Typical equivalent circuit of an inductively coupled rf plasma with an external impedance match-
ing network.
of the capacitance Csh (= ε0 Ssh /dsh ) and resistance R sh , where Ssh and dsh are
the cross sectional area and thickness of the sheath. The power is dissipated
at the resistance Rsh by a collisional effect between the charged particle and
feed gas molecule.

1
J T (t) ∼ JDis + J p ∼ jωCsh + Vsh . (7.43)
Rsh
We show a typical equivalent circuit of the CCP with metallic electrodes con-
nected to an external impedance matching network in Figure 7.6.
When there is a dielectric substrate on a metallic electrode, we have to
add a series capacitor between the sheath circuit and the matching network
(or grounded-metal electrode). In front of the reactor wall, a wall sheath is
formed and is modeled by the capacitance similar to the above model. The
equivalent circuit model is usually created by using the passive elements (i.e.,
resistance, capacitance, and inductance). Sometimes, active elements such as
the diode and current source may be included in the circuit to realize the
detailed system.
7.5.2 Equivalent Circuit Model in ICP

A plasma externally driven by an rf coil current is classified as an ICP or
transformer coupled plasma (see Chapter 10). The equivalent primary and
secondary circuits of the transformer are used to model the electrical proper-
ties of the total system of an ICP as shown in Figure 7.7. That is, the equiva-
lent circuit is described as a nonideal transformer with a coupling coefficient
k = (M/L c L p )1/2 between the inductance of the current coil (L c ) and of the
plasma (L p ) with resistance (R p ) caused by collisions of electrons. The plasma
current flows through the inductor with L p and resistor of R p .

(a) Z/ 2 Z/ 2
Vi-1, j Vi, j
(b) Z/ 2 Z/ 2
Vi-1, j Vi, j
(c) Z/ 2 Z/ 2
Vi-1, j Vi, j
Ii-1, j ZE Y
FIGURE 7.8
Transmission line, modeled by T-type equivalent circuit (a) and the other elemental circuits (b)
and (c).
7.5.3 Transmission-Line Model (TLM)

A very large area processing with a wafer of (sub) meter in size is needed
in industrial applications of plasma deposition or etching. With increasing
the frequency of the external rf source, the dimension of the plasma reactor
becomes comparable with a quarter of the free-space wavelength of the rf
power source. Under these circumstances, it is of first importance to estimate
the local distribution of the surface potential on the powered electrode and
to perform the plasma uniformity, because the surface potential is influenced
by the size of and rf feed position on the electrode. Transmission-line model
(TLM) is appropriate to the estimation of the local surface potential as a func-
tion of feed position of an rf power on the electrode [13]. An equivalent circuit
of a rod to plane electrode system is shown in Figure 7.8 as a transmission line
in a small length x along the rod. The circuit is named the T-type equiva-
lent circuit, and the series impedance and parallel admittance per unit length
are written as Z and Y. In Figure 7.8 a, we have two relations between V(x)
and I(x),
dV(x) dI(x)
= −ZI(x) and = −YV(x).
dx dx

These are arranged as the Helmholtz equation in the one-dimensional rod to

plane electrode system,
d 2 V(x)
= ZYV(x). (7.44)
dx 2
Here note that V(x), I(x), Z, and Y are complex numbers, expressing the am-
plitude and phase shift. In the TLM, three types of T-type circuits are usually
used (see Figure 7.8). We deal with a plasma sustained between the rod and
plane with distance l. The series impedance Z and parallel admittance Y have
to be investigated. The widths of the sheath and bulk plasma are dsh and db ,
perpendicular to the electrodes, respectively; that is, l = 2dsh + db . Then, the
impedance Z is written as
Z = R + jωL ,
   
1+ r02 + 1
µ0 
= ρ/(2πr t) + jω log   − r02 + 1 + r0 , (7.45)
2π r0
where ρ is the resistivity of the rod electrode with radius r0 . The rf current
flows in a thin region r0 ≥ r ≥ (r0 − dskin ) along the rod. dskin is the skin
depth. µ0 is the permeability in vacuum. The plasma admittance including
the sheath Y per unit length is given in the form similar to Section 7.5.1,
1 Rb 1 Rb 2
= 2 + j 2 −j , (7.46)
Y ωL b ωCsh
1+ Rb
ωL b
1+ Rb
ωL b
where R b and L b are given by Equations 7.40 and 7.41, respectively.
7.6 Electromagnetic Fields and Maxwell’s Equations

In this section we give a brief summary of some of the laws in electromagnetic
theory that are relevant for understanding generation and maintenance of
plasmas.
7.6.1 Coulomb’s Law, Gauss’s Law, and Poisson’s Equation

Electric field E(r) surrounding a charge e is determined from the electrostatic
force F(r) on a test particle (of charge q ) and it is given by Coulomb’s law
F(r) e r
= E(r) = , (7.47)
q 4π ε0r r
2

where ε0 is the permittivity of vacuum. Equation 7.47 allows us to determine

the integral of the magnitude of the field over the surface of a sphere at a
distance r
e
|E(r )|4πr 2 = ,
ε0
which can be written in an integral form

e
E(r) · ndS = , (7.48)
ε0
that is often known as Gauss’s law. If we have more than one charge inside
the volume and based on the linear additivity of electric field, we write Gauss’s
law in a more general form, which can be further expanded by using Gauss’s
theorem

1
E(r) · ndS = div E(r)dV = ρ(r)dV.
S V ε0 V
From this equation we may directly obtain the differential form that is often
known as Poisson’s equation:
1
div E(r) = ρ(r) . (7.49)
ε0
In principle, Gauss’s law is valid for all fields that have sources, and it states
that the flux of the field over a closed surface is equal to the charges that are
inside the volume defined by that surface. Because a magnetic field has no
(known) particles that may be its source, we have
div B(r) = 0. (7.50)
Poisson’s equation is often used in plasma physics to determine the field in a

system with space charge.
7.6.2 Faraday’s Law

In the year 1831 Faraday established that a time-varying magnetic field in-
duces voltage u(V) or current I (A) in a closed conductor with resistance
R(). See Figure 7.9. The induced current is actually proportional to the time
derivative of flux of magnetic field (t),
d(t)
u = IR = − . (7.51)
dt
The total induced potential is equal to the line integral of the local induced
electric field E(r, t). The flux of magnetic field (t) is defined in a similar way
as the flux of an electric field. Thus we have

u = E(r, t) · dl

E(t)
E(t)
I(t)
B(t)
I(t)
I 0 sinωt
FIGURE 7.9
Schematic representation of the effect of a time-varying current I0 sin ωt that induces external field
E(t). However, the external time-varying magnetic field B(t) induces a field in the conductor
(dashed lines) and affects the resulting current.

= B(r, t) · ndS. (7.52)
S
If we combine Equations 7.51 and 7.52 we obtain

d
u = E(r, t) · dl = − B(r, t) · ndS .
dt S
When we replace the line integral by a surface integral by using Stokes’s
theorem, we obtain

∂
rotE(r, t) + B(r, t) · ndS = 0,
S ∂t
which gives directly
∂
rotE(r, t) + B(r, t) = 0, (7.53)
∂t
known as Faraday’s law.
7.6.3 Ampere’s Law

If, on the other hand, we wish to calculate the magnetic field induced by a
conductor that has a current I , we obtain the line integral of magnetic field
according to

B(r) · dl = µ0 I, (7.54)
where µ0 is the magnetic permeability in vacuum. The current is separated

into two terms, convection and displacement currents, and the total current

that passes through a closed loop defined in an area S is obtained by integrat-

ing the current density j(r, t) and the displacement current defined through
the time-varying electric field E(t) as

∂E(r, t)
I (r) = j(r, t) + ε0 · ndS, (7.55)
∂t
which when combined with Equation 7.54 leads to

∂E(r, t)
B(r, t) · dl = µ0 j(r, t) + ε0 · ndS.
S ∂t
If we replace the term on the left-hand side by using Stokes’s theorem we
obtain

B(r, t) · dl = rotB(r, t) · ndS .
S
The same equation may be written in a differential form

1 ∂E(r, t)
rotB(r, t) = j(r, t) + ε0 , (7.56)
µ0 ∂t
which is also known as Ampere’s (or the Ampere–Maxwell) law.
7.6.4 Maxwell’s Equations

A complete set of differential equations giving the relationship between elec-
tric and magnetic fields and their external sources (j and ρ) is given as
∂B(r, t)
rotE(r, t) = −
∂t
1 ∂E(r, t)
rotB(r, t) = j(r, t) + ε0 (7.57)
µ0 ∂t
divE(r, t) = ρ(r)/ε0
divB(r, t) = 0 .
These are Maxwell’s equations, which provide a basis for understanding the
development of electric and magnetic fields.
References
1. Chua, L.O. and Lin, P.M. 1975. Computer-Aided Analysis of Electronic Circuit.
Englewood Cliffs, NJ: Prentice-Hall.
2. Makabe, T. Ed. 2002. Advances in Low Temperature RF Plasmas: Basis for Process
Design. Amsterdam: Elsevier.

3. Kim, H.C. Iza, F., Yang, S.S., Radjenovic, M.R., and Lee, J.K. 2005. J. Phys. D.
Topical Review 38: R283.
4. Boeuf, J.-P. 1987. Phys. Rev. A 36: 2782.
5. Graves, D.B. 1987. J. Appl. Phys. 62: 88.
6. Dkazaki, K., Makabe, T., and Yamaguchi, Y. 1989. Appl. Phys. Lett. 54: 1742.
Makabe, T., Nakano, N. and Yamaguchi, Y. 1992. Phys. Rev. A 45: 2520.
7. Sommerer T.J., Hitchon, W.N.G., and Lawler, J.E. 1989. Phys. Rev. Lett. 63: 2361.
8. Birdsall, C.K. and Langdon, A.B. 1985. Plasma Physics via Computer Simulation.
New York: McGraw-Hill.
9. Kushner, M.J. 1986. IEEE Trans. Plasma Sci. PS-14: 188.
10. Binder, K. and Heermann, B.W. 1988. Monte Carlo Simulation in Statistical
Physics. Berlin: Springer-Verlag.
11. Koenig, H.R. and Maissel, L.I. 1970. IBM J. Res. Develop. 14: 168.
12. Piejak, R.B., Godyak, V.A., and Alexandrovich, B.M. 1992. Plasma Sources
Sci. Technol. 1: 179.
13. Satake, K., Yamakoshi, H., and Noda, M. 2004. Plasma Sources Sci. Technol.
13:436.

8
Numerical Procedure of Modeling
8.1 Time Constant of the System

The low-temperature radio frequency (rf) plasma used for a microelectronic
device fabrication includes a number of physical and chemical collision/
reaction/transport processes with very different time constants or relaxation
times, and the system is expressed by stiff differential equations. This stiffness
constitutes a fundamental numerical problem, and we consider methods for
overcoming it below (see Figure 8.1).
8.1.1 Collision-Oriented Relaxation Time

In a collision-dominated plasma, the electron changes the direction of mo-
tion in a field by a collision with a neutral molecule. Rf plasma, sustained
by an external rf power source, changes the direction of the bulk field once
per period. Each of the bulk electrons can follow the local field with a finite
time delay through a two-body collision, and the time constant of the mo-
mentum change is the collisional momentum relaxation time τm , which is
derived as
1
= Rm + R j , (8.1)
τm
where Rm and R j are the rates of the elastic momentum transfer and the in-
elastic scattering with threshold energy j , respectively. The energy relaxation
time of electrons by two-body collisions in a time-varying field are obatined
from the conservation of energy in Chapter 7:
1 2m j R j
= Rm + , (8.2)
τ M m
where m and M are the mass of the electron and the molecule, respectively,
and m is the averaged energy of electrons.
231

102
etching
Time Constant and Relaxation Time (s)

gas residence
10-2 diffusion (radicals)
10-4
energy transfer
-6 (1 Torr)
10
ion transit
ωpe
-1
10 -8 τrad
momentum transfer
(1 Torr)
10-10 ωpp
-1
dielectric
~1 ~10
εe (eV)
FIGURE 8.1
Time constant and relaxation time of each of the processes in a low-temperature plasma.
8.1.2 Plasma Species-Oriented Time Constant

Characteristics of the species in a plasma in a periodic steady state depend
on the transport. For neutral species, the characteristic diffusion time,
2
τD = , (8.3)
D
is necessary to obtain a steady-state profile of neutral molecules. Here, D and
are the diffusion coefficient and characteristic diffusion length, respectively.
In the case of a system including a reaction to another molecule, the effective
diffusion time τ Dr is shortened from τ D to
2 1
τ Dr = , (8.4)
D 1 + r NDr 2
k
where kr is the reaction rate constant, and Nr the reactant number density. For
charged particles, the characteristic drift time is given by
L
τd = , (8.5)
vd
where vd is the drift velocity and L is the characteristic drift length.

Numerical Procedure of Modeling 233
The degree of activation (i.e., ionization, dissociation to radical) of feed

gases introduced to a plasma reactor depends on the residence time. The
residence time is defined as
p Vvol
τres = , (8.6)
Qflow
where p, Vvol , and Qflow are the pressure, effective plasma volume, and total
mass flow rate, respectively.
Exercise 8.1.1
A plasma reactor with volume 103 cm3 is operated at pressure 1 Torr and flow
rate 50 sccm (standard cubic cm per minutes). Estimate the residence time of
the feed gas.
1 Torr
× 103 cm3
τres = 760 Torr
50 cm3
∼ 1.6 s.
60 s
PROBLEM 8.1.1
Derive Equation 8.4 from the continuity equation of the system,
∂
Nj (r, t) = −kr Nk Nj + D∇ 2 Nj . (8.7)
∂t
In Equation 8.4, kr Nr 2 /D is called the Damkohler number D̃. Show that the sys-
tem is controlled by diffusion in the case of D̃
1 and by reaction in the case of
D̃ 1.
8.1.3 Plasma-Oriented Time Constant/Dielectric Relaxation Time

In general, when we simulate a plasma structure with a number density n
by using a time-development method, the time step t is proportional to
1/n. The dielectric relaxation, however, gives the relationship between t
and 1/n. Here we consider a system without production or loss of charged
particles and with an almost uniform number density of electrons and ions.
The continuity equation for total charge is expressed by
∂ρT (t)
+ div(eΓT ) = 0 , (8.8)
∂t
where ρT (t) = e{n p (t) − ne (t)}. Under these circumstances, ΓT is a drift flux
only because the uniformity of the charge distribution and the drift flux may
be related to the electric field by apparent mobility µ:
ΓT = n p vd p − ne vde = (µ p n p + µe ne )E. (8.9)

-9
10
∆tnew
-10
10
t (s)
-11
10
τd
-12
10
-13
10 9 10 11 12
10 10 10 10
Plasma density (cm-3)
FIGURE 8.2
Dielectric relaxation time constant τd and the time steps used in the semi-empirical method as a
function of plasma density.
When we consider Poisson’s equation, divE = ρT (t)/ε0 , Equation 8.8 may be

converted to
∂ρT (t) e(µ p n p + µe ne )

∼ −e(µ p n p + µe ne )divE = − ρT (t) , (8.10)
∂t ε0
which forms the relaxation equation for total charge, leading to the solution

t
ρT (t) = ρT0 exp − ,
τd
and therefore we may define the dielectric relaxation time from Equation 8.10
as
ε0
τd = . (8.11)
e(µ p n p + µe ne )
τd is also known as the Maxwell relaxation time. Solutions to the system are
obtained only if the time step t is shorter than the relaxation time (during
relaxation time τd the charge may decay by a factor of 1/e):

ε0
t < τd = , (8.12)
e(µ p n p + µe ne )
which gives an additional limitation on the time steps in a numerical calcu-

lation. The dependence of the dielectric relaxation time limit as a function of
charged particle density is given in Figure 8.2.

Exercise 8.1.2
Estimate the dielectric relaxation time for plasmas with density
(a) 109 cm−3 and (b) 1012 cm−3 .
ε0 ε
τd = ∼ 0 µp np
e(µ p n p + µe ne ) eµe ne 1 + µe ne
8.854 × 10−12 1
∼ × (s),
1.6 × 10−19 µe [m2 s −1 V −1 ]ne [m−3 ]
where for µe ∼ 70 m2 /sV, the calculated values are (a )τd ∼ 5.5 × 107 × 1.4 ×
10−2 × 10−15 ∼ 10−10 s and (b)τd ∼ 10−13 s.
The dielectric relaxation time becomes very small with increasing plasma
density. For example, for n ∼ 1012 cm−3 , the relaxation time is τd ∼ 10−13 s,
which creates a difficult situation in the time-development modeling method
due to the excessive CPU time required.
8.2 Numerical Techniques to Solve the Time-Dependent

Drift-Diffusion Equation
The numerical techniques involved in solving continuity equations and un-
derstanding their accuracy and convergence are discussed below.
8.2.1 Time-Evolution Method

The governing equation systems of low-temperature plasmas are both time
and space dependent, as described in Chapter 7. Using a time-development
solution to solve the governing equations, we can simulate a plasma with a
periodic steady-state structure and periodic steady-state characteristics, in
addition to simulating the initiation of the discharge. The time-evolution
method consists of the following solutions from time tm to time tm+1 . The
present-day computer has become a powerful tool that allows modeling of rf
plasmas by following both time and space profiles in a reasonable amount of
CPU time. However, it is beyond the capability of present-day computers to
obtain a time solution and follow the time modulation in microwave plasmas
at 2.45 GHz. Under these circumstances, the computer does have sufficient
memory and speed to allow development of two- and three-dimensional
codes in space for plasmas excited by external sources from direct current
(DC) to very high frequency (VHF).
The low-temperature rf plasma used for a microelectronic device fabrica-
tion includes a number of physical and chemical collision/reaction/transport
processes with very different time constants or relaxation times, and the sys-
tem is expressed by stiff differential equations. This stiffness constitutes a
fundamental numerical problem, and we consider methods for overcoming
it below.

Exact
Backward
Central
Forward
∆z ∆z
i-2 i-1 i i+1 i+2 z
FIGURE 8.3
Example of a finite difference scheme.
8.2.1.1 Finite Difference

The modeling of a low-temperature plasma basically requires a numerical so-
lution to a time-dependent conservation equation, specifically, the partial dif-
ferential drift-diffusion equation. We briefly review the discretization method
for obtaining the finite difference between the first and second derivatives of
the conservation equation. The forward (FD), central (CD), and backward
(BD) difference schemes in Figure 8.3 are expressed in the case of uniform
grid size (z = zi+1 − zi = const.) as

∂ ni+1 − ni z ∂ 2 n
n(z, t) ∼ − (FD); (8.13)
∂z z 2 ∂z2 i

∂ ni+1 − ni−1 (z)2 ∂ 3n
n(z, t) ∼ − (CD); (8.14)
∂z 2z 6 ∂z3 i

∂ ni − ni−1 z ∂ 2n
n(z, t) ∼ + (BD); (8.15)
∂z z 2 ∂z2 i

∂2 ni+1 − 2ni + ni−1 (z)2 ∂ 4n
n(z, t) ∼ − (CD). (8.16)
∂z2 (z) 2 12 ∂z4 i
Exercise 8.2.1
A low-temperature rf plasma is described by a system of stiff differential
equations. It is known that Gear’s algorithms of orders 1 to 6 are stiffly stable
[2]. Discuss the algorithms.

For example, for a differential equation ∂n(z, t)/∂t = d(n(z, t), t), Gear’s
algorithms for the first three orders are
ni+1 = ni + t[d(ni+1 , ti+1 )] (first order)

4 1 2
= ni − ni−1 + t d(ni+1 , ti+1 ) (second order)
3 3 3

18 9 2 6
= ni − ni−1 + ni−2 + t d(ni+1 , ti+1 ) (third order).
11 11 11 11
Note that the kth-order Gear’s algorithm requires k starting values, ni ,
ni−1 , . . . , ni−(k−1) .
8.2.1.2 Digitalization and Stabilization

The drift-diffusion equation without production of new charged particles is
given in the one-dimensional space,
∂n ∂n ∂ 2n
= −vd + D 2. (8.17)
∂t ∂z ∂z
Here, we use the central difference for uniform spatial mesh (index i) and
calculate the spatial density nim+1 = n(tm+1 , zi ) at time tm+1 = tm + t as

nm − ni−1 m
nm + ni−1 m
− 2nim
nim+1 = nim + −vd i+1 + D i+1 t, (8.18)
2z (z)2
where z is the spatial step. The solution is rewritten as [1],
c m c m
nim+1 = (1 − 2d)nim + d − ni+1 + d + n , (8.19)
2 2 i−1
where
t
d=D (> 0). (8.20)
(z)2
d is the ratio between the time step t and the characteristic diffusion time
(z)2 /D (i.e., the time it takes to move by z due to diffusion).
t
c = vd (> 0) (8.21)
z
is the ratio between the time step and the characteristic drift time (z/vd ),
and c is also known as the Courant number. Finally, we rewrite Equation 8.19
as
nim+1 = A nim , (8.22)
where A is a tridiagonal matrix. The distribution at tm+1 , nm+1 may be ob-

tained by applying matrix A (m + 1) times starting from the initial assumed
distribution n0 and bearing in mind that the values of the matrix elements

change both with time and with changing conditions in the spatial profile
of the plasma. The difference between the densities obtained in subsequent
iterations is:

2
ε = nim − nim−1 = nim − nim−1 , (8.23)
i
and the difference ε may be associated with the error of calculations. If we

consider eigenvalue λ of the matrix A, the solution for the number density
may be written as
nim = λm exp( jαi), (8.24)
where j 2 = −1 and α is an arbitrary constant in wavenumber units. If λ is less

than 1 then ε will disappear. Combining Equations 8.19 and 8.24 we obtain
the solution for the eigenvalue λ:
λ = 1 + 2d(cos α − 1) − jc sin α, (8.25)
where the number density nim is stable under |λ|2 < 1. There are three possible
situations (see Figure 8.4).
i. Without diffusion (d = 0), λ is always greater than 1 for arbitrary

values of c, and the solution is stable;
ii. Without drift (c = 0), λ has a maximum for cos α = −1, and the
solution is stable for d < 1/2;
iii. With diffusion and drift, the system is stable for d< 1/2 and
c < 2d, is:
d < 1/2, c < 2d (i.e., c < 1). (8.26)
PROBLEM 8.2.1
Derive the relations in Equation 8.26 by reference to the trajectory of Equation 8.25
in the complex plane in Figure 8.4.
By combining Equations 8.20 and 8.26 we obtain
(z)2
t < . (8.27)
2D
This equation indicates that if we reduce the size of the spatial mash z (or
increase the spatial resolution) by a factor of 2, the time step should be reduced
by a factor of 1/4, and therefore the overall CPU time should increase by a
factor of 8. Also, by combining Equations 8.21 and 8.26 we have
z
t < . (8.28)
vd

Im
1
c
Re
0 1-2d 1
-c
FIGURE 8.4
Trajectory of the eigenvalue Equation 8.25.
To determine the proper temporal resolution for solving the drift-diffusion

equations, we set the following condition:

z (z)2
t < min , . (8.29)
vd 2D
Sometimes we may redefine the condition 8.26, c < 2d, by using the Peclet
number Pe , which is defined as
c vd z
Pe = = <2. (8.30)
d D
It can be shown that no oscillation of the solution (i.e., stabilization) occurs
when the local Peclet number is Pc < 2.
8.2.1.3 Time Discretization and Accuracy

For the analysis of time development, the drift-diffusion equation may be
rewritten as
du(t)
= βu(t), (8.31)
dt
and Equation 8.31 is solved explicitly by Euler’s method. We can then write
uim+1 − uim
= βuim .
t
This means that at time tm+1 the solution um+1 may be represented by the
previous time tm so that
uim+1 = (1 + βt)uim , (8.32)

and if we start at t = 0 from the initial distribution u(0) after m steps or after
time t = mt(= T), we obtain
uim = (1 + βt)m ui (0). (8.33)
Convergence of the procedure (stability) is achieved when
|1 + βt| ≤ 1, (8.34)
and therefore
−βt ≤ 2. (8.35)
From Equation 8.33 and in the limit t → 0 one obtains a continuous solution
βT
(1 + βt)m = (1 + βt)T/t = (1 + βt)1/βt → e βT ,
in which we use the definition of exponential function. As a result, we

obtain

(1 + βt)T/t β2 β2
∼ exp − Tt ∼ 1 − Tt,
e βT 2 2
which may be used to define the relative error |δ|,
1
|δ| = (βT)(βt). (8.36)
2
When we choose very small t for large T = mt, the relative error will
approach 1. The relation 8.36 may be rewritten as a condition of accuracy:
2δ0
t < . (8.37)
β2T
8.2.2 Scharfetter–Gummel Method

When we simulate a discharge plasma we have to consider both the bulk
plasma and the ion sheath region close to the electrodes, wafer, or walls. In
the region close to the electrode, the variations of fluxes of charged particles
may be very pronounced. In the bulk plasma the flow of particles is dominated
by diffusion, whereas in the sheaths it is dominated by drift.
When gas pressure is relatively high, the flux due to diffusion becomes
small. In the case of a low-pressure plasma sustained between two par-
allel plates, the drift flux will be dominant in the positive ion sheath. In
that case, a region satisfying D∂n/∂z
vd n appears in front of the elec-
trodes, wafer, or side wall. Then, the Peclet number Pe rapidly increases and
stability of the solution of the drift-diffusion equation is not achieved (see

Γi + 1
ni 2 ni + 1
i i+ 1 i+1
2
FIGURE 8.5
Staggered mesh method. Scalar quantities are defined on integer i mesh points and vector quan-
tities on mid i + 1/2 points.
Equation 8.30). Under these circumstances the drift-diffusion equation be-

comes inappropriate, so one has to use the drift equation
∂n ∂n
= −vd . (8.38)
∂t ∂z
The drift equation should be solved from the upwind to the downwind
direction. There are a few methods that may be used to achieve stability of the
drift-diffusion Equation 8.17 when condition 8.30 is not met. One is known as
the Patankar difference method, but the practical application that is discussed
here is the so-called Scharfetter–Gummel (SG) method [3]. In this procedure
we use the staggered mesh method for spatial distribution; in this method,
all scalar quantities are defined on integer mesh points and vector quantities
are defined on midpoints, as described in Figure 8.5.
The flux at time tm is a combination of drift and diffusion fluxes:
∂nm
m = nm vdm − Dm , (8.39)
∂z
where the values of m and vdm are represented at the midpoint (i + 1/2)
between the mesh point i and i + 1. The first-order differential Equation 8.39
is analytically solved in the range of z(0 < z < z) as

vdm i+1/2 i+1/2
m
nm (z) = C exp z + , (8.40)
m
Di+1/2 vd m i+1/2
and we take into account the boundary condition nm (0) = nim at i. We derive
the value of the constant as
i+1/2
m
C = nim − . (8.41)
vdm i+1/2

As a result, nm (z) = ni+1
m
is expressed as

i+1/2
m vdm i+1/2 i+1/2
m
m
ni+1 = nim − exp z + , (8.42)
vdm i+1/2 m
Di+1/2 vdm i+1/2
and this equation may be used to solve for i+1/2

m
at the midpoint i + 1/2:
vm
nim exp d i+1/2
m
Di+1/2
z − ni+1
m
i+1/2
m
= vd i+1/2
m vd i+1/2
m
. (8.43)
exp m
Di+1/2
z − 1
Equation 8.43 is the so-called SG expression for drift-diffusion flux. By ap-

plying the SG expression to the r.h.s. of Equation 8.17, the spatial derivative
of the flux expressed at the midpoints i − 1/2 and i + 1/2 is estimated by the
quantities for i − 1, i, and i + 1:
vm vm m
i+1/2
m
− i−1/2
m m
ni+1 − exp Ddim z + 1 nim + exp Ddim z ni−1
− =
i
vm
i
. (8.44)
z exp Ddim z − 1 z
vm i di
PROBLEM 8.2.2
Prove that Equation 8.44 reduces to: (i) for Pe → ∞, then the equation is expressed
by:
m
vdi
− nim − ni−1
m
(8.45)
z
and we have the backward difference equation for the drift flux, and (ii) for Pe → 0,
we obtain the diffusion flux
Dim m
ni+1 − 2nim + ni−1
m
, (8.46)
(z) 2
where vd and D have no change in the small range between i − 1 and i + 1.
Equation 8.44 is thus able to represent the whole range of conditions occur-
ring in the sheath and in the bulk where both drift-dominated and diffusion-
dominated fluxes occur.
Exercise 8.2.2
Show that the backward difference Equation 8.45 is equivalent to a drift-
diffusion equation with an artificial coefficient D = vd z/2.

If we rearrange Equation 8.45 as
m
vdi m ni+1 − ni−1 v m z ni+1 − 2nim + ni−1
m m m m
− nim − ni−1
m
= −vdi + di , (8.47)
z 2z 2 (z)2
we obtain two terms, respectively describing the derivatives of the drift flux
and the effective diffusion flux with the artificial diffusion coefficient, vd z/2,
in the central difference scheme. The truncation error with order O(z)2 in
Equation 8.47 has the effect of an artificial diffusion.
PROBLEM 8.2.3
When the drift is along the positive z-axis, show that Equation 8.45 should be ex-
pressed as
|vd | m
− ni − ni−1
m
. (8.48)
z
On the other hand, when the drift is along the negative z-axis, show that it is given
as
|vd | m
− ni − ni+1
m
. (8.49)
z
This method, in which the artificial diffusion in the central diffusion scheme is reduced
by checking determining the flow direction, is called the upwind scheme.
8.2.3 Cubic Interpolated Pseudoparticle Method

The cubic interpolated pseudoparticle (CIP) method is applied to the drift
problem typical of the propagation of a soliton, which is a single wave with a
constant spatial profile in a plasma [4]. In Exercise 8.2.2, we have shown that
the numerical solution of the drift equation will bring artificial diffusion in
the procedure. Let us return to a consideration of the drift equation with drift
velocity vd :
∂n(z, t) ∂n(z, t)
= −vd . (8.50)
∂t ∂z
The Equation 8.50 has an analytical solution n(z, t) = n(z − vd t, t), which de-
notes the density propagation to the negative direction with the same spatial
form (see Figure 8.6). The CIP method allows us to consider the analytical
solution. That is, the density at (m + 1)t under a constant vd is expressed by
nim+1 (zi , t + t) nim (zi − vd t, t). (8.51)
n(z) at zi ≤ z ≤ zi+1 is given by the interpolation as

dnim (z)
ñ(z) = nim (zi ) + (z − zi ) + h i (z − zi )3 + ki (z − zi )2 , (8.52)
dz

t t+∆t
z
zi-1 zi
zi - vd∆t
FIGURE 8.6
Temporal development of the number density subject only to the drift equation.
where the constants h i and ki are obtained by n(zi+1 ) and dn(zi+1 )/dt in
Equation 8.52. Then,

dnm
nim+1 = nim + − h im vdi
m
t + kim vdi
m
t − i vdi
m
t (8.53)
dz
dnim+1 dnim m m m
= + 3h i vdi t − 2kim vdi t, (8.54)
dz dz
where by using the sign, that is, isgn = 1 at vd < 0 and isgn = −1 at vd > 0,
we have
m
d
dz
nim + ni+isgn
m
2 ni+isgn − nim
h im = × isgn −
(z)2
(z)3
m m
d
dz
2ni + ni+isgn
m
3 ni+isgn − nim
ki = −
m
× isgn + .
z z)2
The diffusion part of the drift-diffusion equation is numerically calculated in

an implicit form by using Equation 8.53
nim+1 − ni∗ m+1

ni+1 − nim+1 nim+1 − ni−1
m+1
= Di+1/2
m
− Dm
i−1/2 + Sim . (8.55)
t (z)2 (z)2
Then, by collecting the terms in Equation 8.55 with respect to the position
grids, i − 1, i, and i + 1, we obtain

m
Di−1/2 t m
Di+1/2 t Di−1/2
m
t m+1 Di+1/2
m
t m+1 ∗ m
− m+1
ni−1 + 1+ + ni − ni+1 =ni +Si t.
z2 z 2 z 2 z 2
(8.56)

After we write the Equation 8.56 for all the spatial meshes, the simultaneous
equations are expressed in the matrix form
 m+1   
  n1 d1
b1 c1 0  m+1  
a  n2   d2  
 2 b2 c2   
  . 
  .
..   .. 
 ... ...  
 
 
= . (8.57)
  ..   .. 
 ... ...    . 
  . 
 a I −2 b I −2 c I −2  
 m+1  
 
 n I −2   d I −2 
0 b I −1 c I −1
nm+1
I −1
d I −1
The above matrix is solved by the appropriate numerical method, and we

obtain the solution nim+1 of the drift-diffusion equation in Equation 8.17.
∂nim+1 /∂z is given by
m+1 ∗

∂nim+1 ∂n∗ n − ni−1
m+1
− ni+1 + ni−1
m+1
= i + i+1 . (8.58)
∂z ∂z 2z
The CIP calculation in the next time step proceeds by using the set (nim+1 ,
∂nim+1 /∂z).
8.2.4 Semi-Implicit Method for Solving Poisson’s Equation

In the explicit method, the selected value of the time step t must be much
smaller than the dielectric relaxation time, and it is difficult to apply the
time-development method of the governing equation with increasing plasma
density. We here present a method for optimizing this calculation [5]. By
performing a Taylor expansion of the r.h.s. of Poisson’s equation, the total
charge density ρT is defined as
∂ρT (z, t)
−ε0 ∇ 2 V(z, t + t) = ρT (z, t + t) ∼ ρT (z, t) + t. (8.59)
∂t
Here, we use ρT (z, t) = −ε0 ∇ 2 V(z, t) and ∂ρT /∂t = −div(eΓT (t)), and thus
we obtain
−ε0 ∇ 2 V(z, t + t) = −ε0 ∇ 2 V(z, t) − et divΓT (z, t) (8.60)
and, consequently,
∇ 2 {V(z, t + t) − V(z, t)} e

= divΓT (z, t). (8.61)
t ε0
This means that the potential V(z, t) at time t + t is expressed by the value
at the time t (semi-implicit method) and from the potential we determine the
electric field. The technique described here allows more efficient calculation
of the potentials and fields, as can be seen in Figure 8.2.

8.3 Boundary Conditions

Nonequilibrium plasmas are maintained in reactors that have electrodes and
walls. The properties of the plasma may be strongly affected by the nature
of the interactions with these surfaces. It is thus very important to properly
specify the boundary conditions. The key physical and chemical processes
in plasma etching, deposition, and many other plasma technologies occur at
surfaces and may even involve material transport. Therefore, in this section
we discuss methods for defining the boundary conditions for plasma models.
8.3.1 Ideal Boundary — Without Surface Interactions

8.3.1.1 Dirichlet Condition
In some cases, the value of the function in the governing equation is given in
the Dirichlet-type boundary. For example, the boundary condition at r = r0
is given for the electron number density as
ne (r, t)|r0 = const. (8.62)
Note that the reflection (as well the absorption) at the surface affects the
velocity distribution function close to the surface, g(v, r, t). At the perfect
absorbing boundary, the velocity distribution function satisfies
g(v, r, t; v · nT > 0)|r0 = 0. (8.63)
In the case of the ideal perfect reflection surface, the velocity after reflection
v only changes direction perpendicular to the surface:
g(v , r, t)|r0 = g(v, r, t)|r0 ,

where v = v + 2(v · nT )nT . (8.64)
The general surface including random reflection may be expressed in terms

of the reflection coefficient αref (v) as
g(v , r, t)|r0 = αref (v) g(v, r, t)|r0 , (8.65)
where αref = 0 for the perfect absorbing boundary, and αref = 1 for specular
reflection in Equation 8.64.
In Figure 8.7 we schematically show the effect of reflection on the flux of
particles and on the velocity. In addition, this schematic shows that, due to
surface irregularities, the direction of the particles after scattering may not
have the same angles as the incoming particles.

plasma surface
nT
v'
θ
θ
v
v'
θ
v
r = r0
FIGURE 8.7
Reflection of particles from the surface: reflection of a flux of particles, velocities before v and
after reflection v against an ideally flat and against a realistic irregular surface.
8.3.1.2 Neumann Condition

The boundary condition for the first derivative is known as the Neumann
condition. For example, for the electron density, this condition is

∂ne (r, t)
= const. (8.66)
∂r r0
Chantry introduced the effect of the reflection of the species incident on the
boundary,

∂n(r, t) n(r, t) 1 + αref
= β = λ cos θ , (8.67)
∂r r0 β 1 − αref
where λ is the mean free path, and αref is the reflection coefficient of the species.
β is defined as a distance between r0 and the point where the tangential line
of n intercepts the r-axis (see Figure 8.8). The sticking coefficient st on the
boundary surface is frequently used instead of the reflecton αref in the field
of thin film deposition. Then,
st = 1 − αref (8.68)
√
A value 2/3 or 1/ 3 is taken for cos θ [6, 7].
8.3.1.3 Periodicity Condition

In an rf plasma externally sustained at a voltage source V0 sin ωt, the plasma
characteristics (v, r, t) in a periodic steady state satisfy periodic boundary

plasma surface
n
∼λ
r = r0
FIGURE 8.8
Boundary condition of Chantry and Phelps and the particle density close to the surface
conditions in time:

iπ
(v, r, t) = v, r, t + i = 1 or 2. (8.69)
ω
The series of structures on a wafer will come into contact with a plasma having
a spatial periodicity defined as
(v, r, t)|r0 = (v, r + ik, t)|r=0 i = 1, 2, . . . , (8.70)
where k is the spatial wavenumber of the surface structure.
8.3.2 Electrode Surface

There practically exists some type of electrode, wafer, or substrate to be
processed in plasma processing for material fabrications.
8.3.2.1 Metallic Electrode

On a metallic electrode the incident-charged particles are immediately recom-
bined. In most cases, the metallic electrode is biased through the blocking
capacitor Cb to the external source having Vrf (t) = V0 sin ωt (see Figure 8.9).
Then, the metallic surface potential is different from the external source,
Vsus (t) = V0 sin ωt − Vself (t), (8.71)
where the self-bias voltage Vself (t) is given by the incident fluxes of electrons,
and positive and negative ions, e , p , and n , from the plasma to the metallic
surface S, and the displacement current density J dis ,
Vself (t) = Q(t)|sur /Cb , (8.72)
where
t
1
Q(t)|sur = e{ p (r, t) − e (r, t) − n (r, t) + J dis (r, t)} d Sdt. (8.73)
−∞ S e

V(t)
Vrf (t)
Vp (t)
Vrf
Vsus Vp
0 t
Cb
Vdc
electrode
Vsus (t)
FIGURE 8.9
Potentials at points around an electrode with voltage source Vrf .
Usually, Vself (t) is temporally modulated and DC self-bias voltage Vdc is ex-
perimentally defined as
T T
1 1
Vdc = Vsus (t)dt = −Vself (t)dt. (8.74)
T 0 T 0
8.3.2.2 Dielectric Electrode

On the other hand, when a dielectric electrode or wafer is exposed to a plasma,
the incident-charged particles are locally accumulated on the surface,
t
σ (r, t)|sur = e{ p (r, t) − e (r, t) − n (r, t)}dt. (8.75)
0
When an insulated wafer set on a metallic electrode is biased externally, then

the surface potential is considered as a series of capacitors, Cb and Cwafer , in
circuit theory. The insulator thickness of the wafer is usually very thin, and
the voltage drop-off of the wafer will be very low as compared with that in
the blocking capacitor Cb .
Exercise 8.3.1
There is a trench structure with a finite surface conductivity σ on an SiO2
wafer. The trench is exposed by fluxes of electrons and positive ions from a
plasma. When the surface etching and deposition are negligible, derive the
equation of the local number density of the surface charge.
By using the local surface fluxes incident from the plasma, e (r, t) and p (r, t),
a simple continuity equation of electrons and positive ions at the surface can

be given as
∂ne
+ dive (r, t) = −div[je surf (r, t)/e] − Rr ne n p , (8.76)
∂t
∂n p
+ div p (r, t) = −Rr ne n p , (8.77)
∂t
where je surf (r, t) = σ (r, t)Esuf (r, t) is the surface electron current density, and
Rr is the surface recombination coefficient. The Equations 8.76 and 8.77 are
simultaneously solved with Poisson’s equation. A self-consistent solution is
obtained with respect to the potential of the metallic electrode behind the
wafer. The Laplace equation must, of course, be used inside the dielectric
SiO2 .
8.3.3 Boundary Conditions with Charge Exchange

Normally, charged particles are lost or reflected at the surface. Sometimes
they may also be generated at surfaces. Secondary production of electrons
by the impact of positive ions ( p) with the surface proceeds through two
processes (see Chapter 4): Auger potential ejection and kinetic mechanisms.
In a similar fashion, fluxes of fast neutrals (n ), photons ( ph ), and meta-
stables (m ) induce electrons exclusively through either a kinetic (n) or a
potential mechanism ( ph, m). Under normal conditions, we have fluxes of all
of these particles arriving at the electrode in a plasma, and then the produced
electrons are ejected and contribute in part to further kinetics in a plasma.
The resulting flux of electrons is
e (r, t)|r0 = {γ p p (r, t) + γn n (r, t) + γm m (r, t) + γ ph ph (r, t)}|r0 , (8.78)
where γ p , γn , γ ph , and γm are the secondary electron emission yields for

positive ions, fast neutrals, photons, and metastables, respectively. These
processes are shown in the upper panel of Figure 8.10.
8.3.4 Boundary Conditions with Mass Transport

The most important goal in the application of plasmas to integrated cir-
cuit production is to perform plasma etching. In the case of plasma etching,
sputtering, or deposition, material is transported between the plasma and
surface. This material usually originates from a surface but may also be gen-
erated in the gas phase of the plasma. The transport of material is induced
or facilitated by fluxes of physically or chemically active ions and neutrals
that interact with the surface, which leads to the modification of the surface
profile as well as superficial changes in the substrate material.

plasma surface
Γp, Γn, Γph Γeo

γ p, γ n, γ ph
Γp
Yetch
Γn
Ydepo
FIGURE 8.10
Boundary conditions with charge exchange and mass transfer.
8.3.4.1 Plasma Etching

When there exist high-energy ions with a mass of Mp , velocity distribution
g p (v, r, t), number density n p (r, t), and chemically active molecules (i.e., rad-
icals) with Mr , gr (v, r, t) and nr (r, t) incident on a wafer, we may calculate the
flux of ejected materials from the surface by dry etching as
Ns Vs |r0

p 2ε p 1/2
= n p (r0 ) Yetch (ε p ) g p (v, r0 , t)dv
Mp

2εr 1/2
+ nr (r0 ) Yetch (εr )
c
gr (v, r0 , t)dv, (8.79)
Mr
p
where Yetch ( p ) and Yetch
c
(r ) are the etching yield of ions ( p) and radicals (r ),
and Ns and Vs the number density and velocity of the molecule ejected from
the wafer. The r.h.s of Equation 8.79 represents the sum of the contributions
from the physical etching and chemical etching. The velocity distribution
functions are normalized as g p (v, r, t)dv = 1 and gr (v, r, t)dv = 1.
The rate of change of the material surface with atomic density ρ by etching,
that is, the etch rate, is

n p (r0 ) p 2ε p 1/2
Retch = Yetch (ε p ) g p (v, r0 , t)dv
ρ Mp

nr (r0 ) 2εr 1/2
+ Yetch (εr )
c
gr (v, r0 , t)dv. (8.80)
ρ Mr
In general, g p (v, r, t) is a beamlike, high anisotropic distribution, whereas
gr (v, r, t) is an isotropic thermal distribution (Maxwellian) in front of the
wafer or electrode.

8.3.4.2 Plasma Deposition

Dissociated neutral radical species generated in a plasma may be deposited
on the surface. This is known as plasma deposition. Often the precursors
of deposition are first deposited on a surface that may be activated by ion
bombardment. When the incident flux of neutral radicals is denoted by r (z),
and the sticking coefficient as St , the rate of deposition to the surface is

St
R depo = r (r) , (8.81)
ρ r0
where ρ is the atomic density of the deposition film.
8.3.4.3 Plasma Sputtering

Plasma sputtering is a purely physical process of surface etching, and the
sputtering rate is defined by an equation similar to Equation 8.80:

n p (r0 ) p 2ε p 1/2
Rsputter = Ysputt (ε p ) g p (v, r0 , t)dv. (8.82)
ρ Mp
8.3.5 Moving Boundary under Processing

A surface affected by etching, sputtering, or deposition in a plasma will
undergo a change in geometrical and physical profiles, and then the surface
boundary is classified as a moving boundary. A spatiotemporally varying
surface boundary as shown in Figure 8.11 can be generally described by the
Φ(r, t) = d
d Φ(r, t) > 0
Φ(r, t) = 0
t
Φ(r, t) < 0
Φ(r, t+∆t) = 0
t+∆t
FIGURE 8.11
The moving boundary in material processing.

surface evolution equation known as the Hamilton–Jacobi equation [8]:

∂ ∂(r, t)

(r, t) = Rreact (r, t) , (8.83)
∂t ∂r
where (r, t) is the surface function defined as
(r, t) > 0 in gas phase;

= 0 on a boundary surface; and
< 0 inside a material. (8.84)
Note that the local surface reaction rate as the local surface velocity Rreact (r, t)
is a negative value in etching and sputtering (see Equations 8.80 and 8.82)
and a positive value in deposition (see Equation 8.81).
An approach to the surface evolution based on the numerical solution of
the Hamilton–Jacobi Equation 8.83 is named the Level Set method. The Level
Set method was developed in [8] and is robust in two- and three-dimensional
evolution problems.
References
1. Ferziger, J.H. and Penc, M. 1996. Computational Methods for Fluid Dynamics. Berlin:
Springer Verlag.
2. Chua, O.L. and Lin, P.M. 1975. Computer-Aided Analysis of Electronic Circuit.
Englewood Cliffs, NJ: Prentice-Hall.
3. Sharfetter, D.L. and Gummel, H.K. 1969. IEEE Trans. on Electron Devices ED-16:64.
4. Takewaki, H., Nishiguchi, A., and Yabe, T. 1985. J. Comput. Phys. 61:261–268.
Nakamura, T. and Yabe, T. 1999. Comput. Phys. Commun. 120:125–154.
5. Ventzek, P.L.G., Hoekstra, R.T., and Kushner, M.J. 1994. J. Vac. Sci. Technol.
B12:461.
6. Chantry, P.J. 1987. J. Appl. Phys. 62:1141.
7. Phelps, A.V. 1990. J. Res. Natl. Inst. Stand. Technol. 95:407.
8. Osher, S.J. and Sethian, J.A. 1988. J. Comput. Phys. 79:129. Sethian, J.A. and
Strain, J. 1992. J. Comput. Phys. 98:231.

9
Capacitively Coupled Plasma
9.1 Radio-Frequency Capacitive Coupling

Plasma is generally sustained capacitively or inductively by a radio-
frequency (rf) power supply. Part of the rf power input to the plasma is re-
flected back to the power supply as a reactive power. The maximum power
dissipation in a plasma is supplied by an external rf source when the dis-
charge plasma impedance Z is equal to the impedance at the external power
source z (the maximum-power transfer theorem). Plasma impedance is de-
termined self-consistently both by external plasma parameters and by the
internal fundamental property of feed gas molecules. An impedance match-
ing network is, therefore, required between the rf electrode and the power
supply. A matching network with an equivalent capacitor, Cb (blocking ca-
pacitor), is typical. The discharge plasma connected by way of a capacitor to
the rf power source is named capacitively coupled plasma (CCP). The value
of Cb should be much greater than the sheath capacitance of the reactor.
9.2 Mechanism of Plasma Maintenance

In this chapter, we consider a parallel plates reactor made by metallic elec-
trodes as shown in Figure 9.1. A powered electrode is connected to an rf
voltage source, Vrf (t), defined as
Vrf (t) = V0 sin ωt, (9.1)
through a blocking capacitor Cb. Here, V0 and ω are the amplitude and angular
frequency, ω(= 2π f ), of the rf power supply, respectively. The other electrode
as well as the side metallic wall is grounded to the earth.
When an rf voltage Vrf (t) is applied to the parallel plate reactor in a
vacuum, the current IT (t) between electrodes leads the applied rf voltage
waveform Vrf (t) by π/2. The current is known as the displacement current or
charging/discharging current between the two electrodes, and the external
source has no power dissipation to the reactor; that is, V rf (t) · I T (t) = 0.
255

sheath
Cb
bulk
plasma
V0 sinωt
FIGURE 9.1
A typical CCP reactor.
When an rf voltage appropriate to sustain a discharge plasma is supplied

between the two electrodes filled with feed gas molecules at pressure p, a
finite discharge current flows through the closed circuit shown in Figure 9.1.
Of course, the total current IT (t) is continuous as a function of axial position,
whereas the magnitude of each of the elements of IT (t) is different in each po-
sition (see Chapter 7). When a low-temperature plasma is sustained between
electrodes, the sustaining voltage at the powered electrode Vsus (t) is given by
t
1
Vsus (t) = Vrf (t) − IT (t)dt, (9.2)
Cb −∞
where the second term on the right-hand side is the voltage drop between
Cb to ensure a net direct current (DC) of zero in a periodic steady state. As a
result, the sustaining voltage has a finite negative bias voltage expressed by
the second term on the right-hand side of Equation 9.2.
Capacitively coupled rf plasma is generally classified into two regions
by the phase-shift between the total current IT (t) and the sustaining voltage
Vsus (t). Low-frequency CCP is defined simply when the ions in the plasma
satisfy the relation
eff 1
Vd > d, (9.3)
2f
eff
where Vd is the effective drift velocity of positive ions between parallel plates
with distance d. That is, at low-frequency plasma, the ions can follow the local
field change in time, and the current-sustaining voltage characteristics are
resistive without phase difference. Note that the sheath behaves resistively at
low-frequency CCP.

Capacitively Coupled Plasma 257
1000
SF6/N2(10%)
Minimum sustaining voltage 800
CH4(10%)/H2
600 H2
(Vpp) [V]
400 Ar
He
O2
200
0
10k 100k 1M 10M 100M
Frequency [Hz]
FIGURE 9.2
Minimum sustaining voltage (experimental) in a quasi-symmetric parallel plate CCP as a function
of applied frequency at 1 Torr and d = 2 cm.
On the other hand, at high-frequency plasma, it is difficult for massive

ions to follow the instantaneous local field,
eff 1
Vd d. (9.4)
2f
At high-frequency CCP, the total current IT (t) has a finite phase lead with
respect to the sustaining voltage waveform Vsus (t), and the sheath behaves
capacitively. Usually there exists a boundary between the low- and high-
frequency plasmas at several MHz of the external frequency (see Figure 9.2),
and the sustaining mechanism of the discharge plasma is completely different
between the two regions.
Exercise 9.2.1
Discuss the maximum-power transfer theorem in the CCP system sustained
by a sinusoidal voltage in Equation 9.1 in Figure 9.1.
The circuit current I T (t) in the series circuit with plasma impedance Z and
external variable impedance z is given by V rf (t)/(z + Z), and the average
power dissipated in the discharge plasma, Pa v , is expressed by
T
1 1 2 Re(Z)
Pav = V sus (t) · I T (t)dt = V , (9.5)
T 0 2 0 |z + Z|2

where T is the period of the external rf voltage source. The maximum power
dissipation to the discharge plasma is obtained at the condition of Z = z
(impedance matching). That is, the impedance of the discharge plasma Z
must be matched to the conjugate of the power supply, z (maximum-power
transfer theorem).
9.2.1 Low-Frequency Plasma

As the amplitude of the applied voltage waveform is increased, the space
between parallel plates filled by neutral gas will change from the Laplace
field to Poisson’s field by way of the transition region from the Townsend
discharge to glow discharge having a plasma phase (see [1, 2]). Low-frequency
plasma is sustained by the electron multiplication of the secondary electrons
ejected mainly by the impact of positive ions on the powered electrode by
Auger potential ejection (see Section 4.11.2.4). The maximum flux and kinetic
energy of ions incident on the powered electrode from the plasma occur at the
phase of the lowest surface potential, that is, at 3π/2 in Equation 9.2. In order
to keep a maximum current condition under a given rf voltage at the powered
electrode, the space between the two electrodes is bifurcated into two regions,
an ion sheaths region and a bulk plasma region. In front of the powered
electrode, an ion sheath is constructed so as to allow a sufficient amount
of electron ejection from the electrode and a sufficient acceleration of the
secondary electrons, which generate ionization multiplications to maintain
the low-frequency plasma. Therefore, a high field with a strong potential
gradient is essential to the ion sheath (active sheath). A small ion sheath
appears in front of the grounded electrode in the conventional parallel plates
reactor. Note that, in an ideal symmetric electrode system, a pair of active
ion sheaths is constructed in front of both electrodes during one rf period (see
Figure 9.3). A bulk plasma region with quasi-neutrality (ne ∼ n p ) having a low
(a) (b)
Λj (1011cm-3s-1)
5.6 240 W 10 4
V
Current [mA]
2.8 120 5 2
Power [W]
Voltage [V]
0.0 0 0 0
20 0 5 10
I
D Time [µs]
is -120 -5 -2
ta 10
nc 10
e 5 -240 -10 -4
(m
m 0 (µs)
) 0 T e
im
FIGURE 9.3
Typical example of the net excitation rate (a) and voltage-current characteristics (b) during one
period in a low-frequency CCP at 100 kHz at 1 Torr in H2 .

3 (a)
(b)
Ion flux
2
0
0 200 400 600 800 1000
Energy (eV)
FIGURE 9.4
Time-averaged ion energy distribution incident on the powered electrode at a low-pressure rf
plasma. The (a) low- and (b) high-frequency cases are shown.
field is formed between the two ion sheaths. The whole system is completely
controlled by Poisson’s field.
The total current in the sheath, which consists mainly of a conduction com-
ponent of positive ions and electrons, coincides with the sustaining voltage
waveform Vsus (t) without phase difference. Even in the case where the sus-
taining voltage is a sinusoidal waveform, the total current is nonsinusoidal in
time, and the plasma potential behaves nonsinusoidally. Under the condition
of a low-frequency CCP in which ions are influenced by the instantaneous
sheath field, the energy of ions incident on the electrode through the sheath
is characterized by a saddle-shaped profile, that is, bimodal distribution (see
Figure 9.4). The energy dispersion between the maximum and minimum en-
ergy peaks decreases as the frequency of the rf source increases to the high-
frequency region.
The high degree of loss of ions from the bulk plasma to both electrodes or
the side wall during a half-period of the low-frequency plasma exhibits some
peculiar characteristics:
i. Formation of low-density plasma with a strong and thick sheath;
ii. Production of high-energy ions incident on the powered electrode;
and
iii. A considerably high value of the sustaining voltage.
PROBLEM 9.2.1
A parallel plates reactor is connected to a low-frequency voltage source by a matching
network consisting of inductance L m . Discuss the DC self-bias voltage at the powered
electrode.
The effective surface area of the powered electrode connected to an exter-

nal rf source with Cb is usually different from that of the opposite electrode
in a CCP. It is typical when the opposite metallic electrode is grounded to

the earth as well as the grounded chamber wall. Under these circumstances,
the asymmetry of the discharge plasma is enhanced as there exists a large
difference in the effective area between the two electrodes.
Exercise 9.2.2
Consider the pair of asymmetric parallel plate electrodes with the surface
areas AP and AG shown in Figure 9.1. Estimate the ratio of each of the drops
in sheath potential in front of both electrodes, Vsh P and VshG , in the case of a
space-charge-limited regime at low frequency.
In a space-charge-limited regime, the ion current density incident on the
sheath with width of d is given by the Child–Langmuir expression,
3/2
K o Vsh
Jp = 1/2
, (9.6)
Mp d 2
where K o is constant, and Mp is the mass of the ion. Provided that the ion
current is dominant in the sheath, the ion currents incident on both electrodes
during one period are equal to each other; that is, J p P AP = J pG AG . Assuming
that the time-averaged potential drops in the sheaths with capacitances C P
and C G are related to the total charge Q flowing to each of the sheaths as
Q = C P Vsh P = C G VshG , (9.7)
where the two sheath capacitances are given by C P =

0 AP /d P and C G =

0 AG /dG , respectively. The ratio of the sheath voltage is derived from the
above two equations as a function of the electrode area:
2
Vsh P AG
= . (9.8)
VshG AP
This relation is valid for low-frequency plasma or DC glow discharge under

the condition that the main sheath current is supplied by the ion conduction.
9.2.2 High-Frequency Plasma

As the frequency of the voltage supply increases, the ions will have a finite
phase delay with respect to a time-varying instantaneous field. This means
that the ion flux incident on the powered electrode and the number of the sec-
ondary electrons will decrease as the rf frequency satisfies the relation (9.4).
At the same time, due to the difference of the phase between IT (t) and Vsur (t),
the number of electrons and ions released to the electrodes will decrease. In
turn, electrons released from the bulk plasma toward the powered electrode
will be effectively reflected in the positive ion sheath as wave-riding electrons.
The reflected electrons will be accelerated toward the bulk plasma through
the sheath, and ionization multiplication will occur at the boundary between
the sheath and bulk plasma (see Figure 9.5). This is a sustaining mechanism
in a high-frequency plasma. As the external frequency increases, electrons

(a) (b)
4.1 60 V W 24 1.2
Λj (1011cm-3s-1)
Current [mA]
Power [W]
Voltage [V]
2.1 30 12 0.6
0.0 0 37 74 -0.6
0 0
20 Time (ns)
D 0
is -30 -12
ta 10
nc 74
e I -1.2
(m 37 -60 -24
m 0 (ns)
) 0 T e
im
FIGURE 9.5
A typical example of the net excitation rate (a) and voltage-current characteristics (b) during one
period of a high-frequency CCP at 13.56 MHz in H2 .
in addition to ions will gradually enter into their trap in space. Due to the
presence of electron trapping inside the reactor, an apparent field of ambipo-
lar diffusion will drop in front of the powered electrode in a time-averaged
fashion. This will introduce a reduction of the minimum sustaining voltages
as shown in Figure 9.2. The boundary between the low- and high-frequency
plasma under conventional external conditions will appear at approximately
several MHz.
When a sinusoidal high-frequency voltage is supplied at the powered
electrode, the total current in the sheath, consisting of the predominant dis-
placement current and the conduction of electrons and ions, has a sinu-
soidal waveform. The capacitive sheath results in a sinusoidal variation of
the plasma potential. The high-frequency plasma has several proper charac-
teristics:
i. Formation of high-density plasma with a weak thin sheath;
ii. A relatively low sustaining voltage;
iii. Production of low-energy ions incident on an electrode; and
iv. A sinusoidal current waveform IT (t) leading the sustaining voltage
waveform Vsus (t) is formed, and the discharge is defined as capaci-
tive.
The electron charge flowing into the blocking capacitance Cb through a
small powered electrode during the positive potential should be equal to
the positive ions during the rest of one period in a periodic steady state.
The great difference between the mass of the electron and that of the ion will
cause an excess negative charge in the capacitor during one period. Therefore,
a negative bias voltage Vdc to the small electrode is needed in order to keep
the zero net DC current in a periodic steady state through the Cb . That is, the
surface of the powered electrode is negatively biased as
Vsus (t) = Vrf (t) − Vdc (t). (9.9)

TABLE 9.1
Classification of Sheaths Formed in Discharge Plasmas
Sheath Circumstances Position Formed Physical Law
Positive Ion In Front Of
Passive Without ionization Wafer without bias Ambipolar diffusion

(floating) Electric probe
Reactor wall
Active Production of e/ion rf powered electrode Plasma maintenance

Dynamic sheath DC cathode Plasma maintenance
Biased electrode Energetic ion injection
Negative Ion
Active Negative ion plasma Instantaneous anode e & n-ion acceleration
Here, −Vdc (t) is the negative self-bias voltage with time variation. The time-
averaged DC self-bias voltage,
2π
1
Vdc = Vdc (t) = Vsus (t) d(ωt), (9.10)
2π 0
is practically observed in the experiment.

At high-frequency CCP, under which the ion transit time across the sheath
is sufficiently longer than the half-period of the external source, the ion in-
cident on the electrode reflects the time-averaged sheath characteristics (see
Table 9.1). As a result, a single-peaked energy distribution is formed with
a maximum at about the time-averaged sheath voltage (see Figure 9.4). The
sheath formed in discharge plasmas is classified in Table 9.1.
PROBLEM 9.2.2
At low pressure a collisionless sheath with thickness dsh , bulk electrons with average
velocity < v > diffusing to the plasma sheath boundary will interact electrically
with the moving boundary having an velocity V sh (t). This is not a binary collision
between the electron and the neutral molecule but a wavelike interaction resulting
from the long-range Coulomb interaction. Provided that the interaction is perfectly
elastic under dsh /v > 2π/ω, the reflected electrons will change their velocity from
− < v > → < v > + V sh , (9.11)
close at instantaneous cathode phase. Therefore, the phenomenon is a collisionless

heating and is referred to as a stochastic heating of electrons, Fermi heating, or wave
riding. Derive the relation 9.11 from the momentum balance of the electrons.

9.2.3 Electronegative Plasma

A conventional low-temperature rf plasma consists of electrons, positive ions,
and feed gas molecules as the components. In a reactive plasma for etching,
electronegative gases are widely used due to a strong chemical reactivity of
the feed gas or the dissociated molecules (radicals) on a surface. There are
hidden and interesting characteristics of an rf plasma for dry etching. That is,
these reactive gases have the ability to produce negative ions by dissociative or
nondissociative electron attachment with a threshold of several eV or thermal
energy (see Section 4.8.2). With increasing pressure, the percentage of nega-
tive ions in the plasma increases due to the increase of the degree of spatial
trapping of massive negative ions and of the collisional chance of electron at-
tachment (see Table 9.2). The quasi-neutrality in the bulk plasma is realized as
n p ∼ (ne + nn ). (9.12)
The electronegativity Hen is defined by the terms of the number density of

negative and positive ions, nn and n p , as
nn
Hen = . (9.13)
np
Fully negative ion plasma is performed under Hen = 1, though the plasma is
not maintained without electrons. In an electronegative plasma with densities
ne , n p , nn , and N, the macroscopic measure of the plasma is characterized by
both the electronegativity Hen and the degree of ionization Hdi given by
np
Hdi = . (9.14)
N
Let us consider again the CCP system consisting of light electrons and
massive positive ions as the components of the charged particle. The system
TABLE 9.2
Negative Ions in Processing Plasmas
Feed Gas Negative Ion Material Processing
SF6 SF−j (6 ≥ j ≥ 1), F− Si, W etching

Reactor cleaning
Cl2 Cl− Si, Al etching
BCl3 Cl− Si, Al etching
CF4 F− , CF−j SiO2 etching
C4 F8 F− , CF−j SiO2 etching
CF3 I I− , CF−j SiO2 etching
O2 O− , O −
2 Oxidization, photoresist ashing
Surface treaming
SiH4 SiH− a-Si:H deposition
H2 /N2 H− Reduction, low-k(organic) etching
HCl Cl− Si etching

has typical plasma characteristics with a bulk plasma under quasi-neutrality

(ne ∼ n p ) and with positive ion sheaths (i.e., electropositive plasma). That is,
one is the presence of a positive space potential (plasma potential) Vs in the
bulk plasma. The other is the strong sheath potential Vsh , which is essential
to maintaining the plasma by the electron impact ionization. Both are caused
by the great difference in speed between light electrons and massive positive
ions. The general characteristics in the electropositive plasma will gradually
change to a different phase in the presence of negative ions. Generally, in a
high-frequency electronegative plasma consisting primarily of positive and
negative ions, with fewer electrons, a thin, higher sheath field is realized as
well as the presence of a strengthened bulk field. This is mainly caused by
massive negative ions having a drift velocity of two orders of magnitude less
than electrons. As a result, there exist three different regions of the plasma
production during the half-period in electronegative plasmas at higher pres-
sure, as shown in Figure 9.6a. One is the production by reflected electrons
in front of the instantaneous cathode (I), another is by the bulk electrons in
a high field (II), and the other is by electrons accelerated through a double
layer close to the instantaneous anode (III). Also the phase difference between
the total current IT (t) and the sustaining voltage Vsur (t) is shortened with in-
creasing electronegativity Hen , as compared with electropositive plasma at all
frequencies. In other words, the characteristics of an rf plasma change from
capacitive to resistive in a strong electronegative plasma (see Figure 9.6).
A typical electronegative plasma is formed in O2 , SF6 , Cl2 , CF4 , and so on
with a finite collision cross section of electron attachment at high pressure.
Electron capture by molecules excited to electronic or vibrational states is
known to have a larger cross section as compared with that by the ground-state
molecule. Negative ion production by way of electronically or vibrationally
excited states will make a considerable contribution to the electronegativity
in plasma, particularly in a high-density plasma. O2 (a 1 ) in oxygen plasma
is an example of the excited states that strongly contribute to the electron
(a) (b)
Λj[SF5+](1016cm-3s-1)
Current density (mAcm-2)
2.8
Power density (Wcm-2)
III 200 W 1.4 0.28

1.4 I III
Voltage (V)
100 0.7 0.14

0.0 II 0 0 0
20 II 0 37 74
I Time (ns)
-100 -0.7 -0.14
Di
10 74 V
st
an
37 -200 I -1.4 -0.28

ce
(ns)
0 0 Time
(m
m
)
FIGURE 9.6
Typical example of net excitation rate in electronegative high-frequency CCP in SF6 /N2 (10%) at
1 Torr.

electrode
total charge density
np >> (ne + nn)
p-ion sheath
np ~ (ne + nn)
0
bulk plasma
np << (ne + nn)

n-ion layer
FIGURE 9.7
Double layer formation in front of the instantaneous anode in an rf electronegative plasma.
attachment, because the peak of the attachment cross section is several times
higher than the the ground state O2 (3 g− ), and the threshold energy stands at
2.76 eV (see Section 4.8.2).
Exercise 9.2.3
In an rf electronegative plasma at a range of Torr, a negative charge-layer is
formed, facing the electrode in the instantaneous anode phase (see Figure 9.7).
Discuss the negative layer formation (it is usually called double layer forma-
tion).
Minority electrons in an rf plasma with Hen ∼ 1 vibrate between the two
electrodes. At the instantaneous anode phase, the total charge in front of the
anode, ρ = n p + nn + ne , is negative due to the flow of electrons from the
bulk plasma to the sheath, though the sheath is positive except for the anodic
phase.
PROBLEM 9.2.3
In an electronegative plasma with electronegativity Hen ∼ 1, negative ion density at
low power is the rate-limiting step between the production by an electron attachment
to the feed gas molecule and the loss by a recombination with positive ion. Then,
derive the simple relation between nn and ne as
1/2
Ra ne
nn N , (9.15)
Rr
where Ra and Rr are the electron attachment rate and recombination rate, respectively.

Frequency Wave Length

[Hz] [m]
4
10 4
LF 10
6
10 MF 2
HF 10
8
10 VHF
UHF 0
10
10 SHF -2
10
EHF
12
10 SEHF -4
10
14
IR
10 VR -6
10
UVR
16
10 VUVR -8
FIGURE 9.8 10
Frequency and its wavelength of electromagnetic 10
18
wave.
XR 10
-10
9.2.4 Very High-Frequency Plasma

A low-temperature plasma source has developed the range of driving fre-
quency from low frequency (LF) ( f <∼ MHz) to ultra-high frequency UHF
(3 × 108 Hz ≤ f ≤ 3 × 109 Hz), through high frequency (HF) (3 × 106 Hz ≤ f ≤
3 × 107 Hz) and very high frequency VHF (3 × 107 Hz ≤ f ≤ 3 × 108 Hz) (see
Figure 9.8). The most pronounced characteristic expected for a parallel plate
CCP is radial uniformity of the plasma structure derived from the uniform
surface potential of the electrode connected to an rf power source.
The VHF source has the advantage of maintaining a high-density plasma
under a low self-bias voltage at the powered electrode and is used primarily
to obtain a good quality film at high deposition rate and a high-efficiency
etching. One of the issues in a large area and high-density plasma sustained
using the VHF technique is the radial nonuniformity of the surface voltage
due to a standing wave effect, when the dimension of the powered electrode
exceeds or is comparable with λ/4 or λ/8 of the free space wavelength λ of the
VHF. The Si wafer in ultra-large scale integration (ULSI) has grown to a large
size from 100 mm in diameter in 1977 to 300 mm in 2000. The glass substrate
of 1 m2 is standard in a-Si:H deposition for thin film solar cells and in plasma
display panels and active matrix liquid crystal displays. In the UHF source,
the plasma is sustained by the aid of an adequate antenna or waveguide due
to a wavelength that is comparable with or shorter than the reactor geometry.
Figure 9.9 shows the experimental amplitude of the surface potential on
a metallic rod electrode coupled at the edge to a VHF (100 MHz) [3]. The rod
electrode is faced to a large area substrate, and VHF plasma is maintained be-
tween them. Here, it is shown that a surface standing wave propagates with
a reduced short wavelength. As shown in the example, radial uniformity of
the plasma structure in a large-size reactor is disturbed by the presence of

(a) 100 MHz
Plasma
0 100 cm 0 6 cm
(b)
Electrode voltage (V)
80
Numerical
40
Experimental
0 20 40 60 80 100
Position (cm)
FIGURE 9.9
Surface potential of a powered electrode exposed to a VHF plasma at 100 MHz at 13 Pa in Ar.
Power is supplied at z = 0, and z = 100 cm is the open end.
harmonics of the external source frequency in the range of VHF. Adequate

multicoupling of the VHF source on the electrode is one of the practical meth-
ods to maintain the uniformity.
The external electrical power supplied by rf sources is first deposited to
the electron in the plasma in the form of transport/conduction. The exter-
nal power is transferred to emission of various kinds of lights by way of
molecules excited by the electron impact. The heavy particle collisions be-
tween the ion and feed gas molecule and between the excited molecule and
the feed gas transfer their kinetic or potential energy to the feed gas transla-
tional energy. These lead to the heating of the feed gas molecule, gas heating.
The total dissipated power is divided into the element, electron, positive ion,
and negative ion. In the bulk plasma, almost all power is deposited to elec-
trons in electropositive gases under a low bulk field. However, in the case
of electronegative gases, the circumstances change. With increasing the elec-
tronegativity Hen = nn /n p , both the negative ion density and the bulk field
increase, because a high field is needed to carry the majority of negative ions
in the bulk plasma. Then, the negative ions in the bulk plasma share the
power among the electron and positive ions. In the sheath, the power is de-
posited into positive ions in front of both electrodes. The asymmetry of the
power deposition in each of the sheaths arises from the presence of the DC
self-bias voltage at the powered electrode as compared with the grounded
electrode. Various applications in a low-temperature plasma technology uti-
lize the plasma element in a form of maximizing or optimizing the function

TABLE 9.3
Time-Averaged Power Deposition
Feed Gas Active Sheath Bulk Plasma Passive Sheath
Electro positive P-ions Electrons P-ions

Electro negative
(Weakly) P-ions Electrons P-ions
(Strongly) P-ions Electrons P-ions
N-ions, P-ions
of the element. Then, the control of the power deposition in space and in time
is of first importance to the technology. See Table 9.3.
PROBLEM 9.2.4
Scaling lows of internal plasma parameters (plasma density, sheath width, etc.) are
prepared with respect to the external parameters (frequency, amplitude, pressure,
etc.) in a simplified condition. Discuss the relationship between plasma density and
driving frequency in a high-density plasma at a sufficiently low pressure.
Exercise 9.2.4
There exists an external electromagnetic wave with frequency ω, which is
lower than the plasma (electron) frequency ω < ω pe ; then, it is difficult for
the external electromagnetic wave to penetrate the plasma. Discuss that this
does not mean none of plasma production/maintenance.
By applying the rf wave with frequency ω and amplitude V0 through an
antenna to a neutral gas molecule, the interaction with the gas molecule will
cause the electron multiplication when we increase the V0 . A periodic steady
state plasma will be formed through an active sheath in front of the antenna.
Then, inside of the formed bulk plasma a cut-off density nc (= ω2pe m
e −2 ) will
exist.
9.2.5 Two-Frequency Plasma

A two-frequency CCP reactor has two types of arrangement of rf power
sources. One is the configuration of a different rf source at each of two elec-
trodes (see Figure 9.10). The other is the system of an electrode with two rf
sources and a grounded electrode. For the choice of the rf frequency, it is of
the first importance to clarify the function in the plasma. For example, in a
low-temperature plasma reactor for plasma etching of SiO2 , a high-density
plasma production and high-energy ions incident on a wafer electrode are
two functions that are required simultaneously. External power sources some-
times interact with each other in a similar frequency range. Both functions are
competitive in a two-frequency (2f) CCP reactor for sustaining a high-density
plasma at VHF with low-voltage amplitude and for biasing the wafer at LF
with high amplitude in each of the electrodes at low pressure. One example
of the functional separation is shown in Figure 9.11 in terms of net ionization
rate as a function of bias frequency (i.e. 1 MHz and 10 MHz).

Drive (VHF)
symmetric
axis
metallic wall
electrode
z wafer insulator
Bias (LF)
FIGURE 9.10
Schematics of a 2f-CCP reactor for SiO2 etching.
(a)
VHF (100 MHz)
6.1E+16 6.7E+16
0.0 0.0
LF (1 MHz)
300 V 700 V
(b)
VHF (100 MHz)
6.1E+16 7.3E+16
0.0 0.0
RF (10 MHz)
300 V 700 V
FIGURE 9.11
Example of the functional separation in a 2f-CCP in CF4 (5%)/Ar at 50 mTorr in terms of net
ionization rate.

The values of the negative self-bias voltage Vdc at each of the electrodes
are considerably different, with the value being very high at LF and very low
at VHF. That is,
i. A high Vdc at the bias electrode is capable of generating the beamlike
ions with an anisotropic velocity distribution at the surface;
ii. High-energy secondary electrons produced by the high-energy ion
bombardment are not confined by a very low Vdc at the opposite
VHF electrode; and
iii. The contribution of the fast secondary electrons to an excess de-
gree of dissociation is significantly reduced in the 2f-CCP at low
pressure.
In particular, high Vdc is realized at the biased electrode without distur-
bance of the plasma profile in the VHF/LF-CCP system. A narrow gap 2f-CCP
is operated by three potentials, Vsur at the powered and biased electrodes, and
a wall potential usually grounded to the earth. The plasma potential in the
system is kept considerably high. One example of the sheath potential drop in
front of the wafer and the surrounding wall is shown in Figure 9.12 as a func-
tion of radial position. The sheath potential drop is defined as the difference
between the plasma potential and the surface potential at the surface. The po-
tential drop is strongly distorted between the wafer edge and the grounded
metallic wall. The phase of the maximum potential drop depends on the ra-
dial position. On the SiO2 wafer the maximum value appears at ωt = 3π/2
when the wafer surface has the lowest potential, and on the grounded metallic
side wall the maximum value appears at ωt= π/2 when the plasma potential
is highest. The magnitude and the energy of charged particles incident on the
surface are influenced by the temporal change of the radial field caused by
the edge characteristics of the potential drop.
The velocity distribution of ions incident on the wafer gives critical infor-
mation for the reactive ion etching of chemically active plasma. There are a
number of investigations of the influence of the bias frequency on the incident
ion energy distribution. It is known that at high-frequency regions the ions
can not relax their energy to the temporal change of the wafer potential; the
energy distribution exhibits a stational profile. However, at low-frequency,
ions follow the instantaneous potential change at the wafer, and a strong time
modulation arises on the energy distribution with a maximum and minimum
at low pressure (see Figures 9.4 and 9.12). Note that the radial characteristics of
the velocity distribution of the incident ions are the key to realizing a uniform
etch process on a wafer. The vicinity of the wafer edge is strongly influenced
by the potential difference between the wafer and the outer grounded wall,
whereas the central part of the wafer keeps the intrinsic characteristics of the
ion velocity distribution as a function of sheath potential drop.
In the other system of 2f-CCP, both of the power sources are supplied to
the same electrode and the second electrode is grounded. This system is less
functional for the material etching device.

6.0 nAr+ (z = zsh) 2.5
Density (1010 cm-3)

(a)
Flux (1016 cm-2 s-1)

Γe 2.0
2.0
1.0
4.0
ΓAr+
0.0 0.0
0.0 2.0 4.0 6.0 8.0 10.0
400 (b)
V plasma
200 Ez Ez
Potential (V)
V surface
0
Vdc ~ -312 (V)

-200
Er
-400
SiO2 Wafer Insulator Metallic wall
0.0 2.0 4.0 6.0 8.0 10.0

Radial position (cm)
FIGURE 9.12
Time-averaged plasma density and flux (a) and potential (b) in a 2f-CCP maintained under
conditions of Figure 9.11a.
9.2.6 Pulsed Two-Frequency Plasma

A modern plasma reactor has to be equipped with a multifunction for process-
ing. One is a selective injection of particles accumulated to a sufficient density
in a bulk plasma. In a CCP, even in a 2f-CCP at low pressure, it is difficult
to accelerate negative charges, electrons, or negative ions to a wafer surface,
because the sheath field in front of the wafer is always formed to decelerate
the particles in a cw operation. Pulsed operation of plasma production by a
VHF source makes it possible to inject negative charged particles in an accel-
eration mode to the biased wafer in a 2f-CCP (see Figure 9.13). It is found that
in Figure 9.13b secondary electrons produced by a high-energy ion impact
on the wafer have little influence on the plasma production in a high-density
2f-CCP. An injection of negative charged particles with high energy are much
enhanced in a system where both pulsed operations of VHF/LF sources are
synchronized in a 2f-CCP [4].

(a) 100 MHz
Λj (1015 cm s )
6.8
-3 -1
3.4
0.0
20
10.4
10
z
(m
m) 0
0 )
t (µs
500 kHz
(b) 100 MHz
5.6
Λj (1014 cm s )
-3 -1
2.8
0.0
20
20
10
z
(m
m) 0 10.4 )
t (µs
500 kHz
(c) 200 1.5

Drive
Z = 0 Λj (1015 cm cm s )
-3 -1
0
Electrode potential (V)
0
-3
-200
-400 Bias
-600
0 10 20
Time (µs)
FIGURE 9.13
Pulsed-plasma source operation in 2f-CCP with VHF (100 MHz) and LF (1 MHz) at 25 mTorr in
CF4 /Ar: (a) net excitation rate during on phase and (b) during off phase and (c) driving voltage
on both electrodes.

References
1. von Engel, A. 1983. Electric Plasmas Their Nature & Uses, London: Taylor &
Francis.
2. Hirshi, M.N. and Oskam, H.J. Eds., 1978. Gaseous Electronics, Vol. 1. New York:
Academic Press.
3. Satake, K., Yamakoshi, H., and Noda, M. 2004. Plasma Sources Sci. Technol. 13:436–
445.
4. Ohmori, T., Goto, T., and Makabe, T. 2004, 2003. J. Phys. D. 37:2223, and Appl.
Phys. Lett. 83:4637.

10
Inductively Coupled Plasma
10.1 Radio-Frequency Inductive Coupling

Radio-Frequency (rf) plasma is sustained by the rf electromagnetic field ra-
diated from an rf current coil. This is named the inductively coupled plasma
(ICP) or transformer coupled plasma (TCP) based on the difference of the
current coil arrangement, as shown in Figure 10.1 [1]. Usually ICP is initi-
ated locally by the capacitive coupling between both terminals of the cur-
rent coil with increasing the voltage amplitude between terminals of the coil
(E-mode). A locally sustained low-density plasma diffuses into gas, and an
abrupt change occurs from a capacitive coupling between the terminals (E-
mode) to an inductive coupling between the plasma and coil (H-mode) when
the power (or coil current) is further increased. The phase transition is known
as an E–H transition in ICP with a hysteresis loop between a rise and fall
in the power. ICP and TCP in Figure 10.1 are examples of the electrodeless
plasma. The pressure range of ICP in plasma etching is between a few mTorr
to some tens of mTorr, whereas for thin film deposition the feed gas pressure
is usually distributed over 100 mTorr and 10 Torr.
10.2 Mechanism of Plasma Maintenance

10.2.1 E-mode and H-mode
The transition between the E-mode and the H-mode is one of the most marked
features of the ICP. One example of the E–H transition is shown in Figure 10.2
in terms of the net excitation rate of the short-lived Ar(2p1 ) in the coil plane of a
cylindrical glass reactor, sustained by a single-turn current coil at 13.56 MHz
in Ar, in Figure 10.1a as a function of amplitude of the coil current [2]. A
sudden transition of the emission is observed at a point (I) with an increase
in intensity of two orders of magnitude, and from the point (II) a gradual
increase continues with increasing coil current. A hysteresis is observed at
high pressures of 100 mTorr and 300 mTorr upon the return to smaller currents.
The ICP in the H-mode has a threshold for maintenance.
275

(a) z (b) z
I sinωt
I sinωt
0 r 0 r
insulator insulator
wafer wafer
FIGURE 10.1
A typical example of ICP (a) and TCP (b) reactors.
In the E-mode the power deposition in the discharge is low, and electron
density is low whereas the mean electron energy is high with a nonequilib-
rium distribution. In the H-mode high power is dissipated with a low-power
reflection. High density of plasma with a lower mean energy of electrons is
sustained in the H mode. The differences of the electron energy diminish at
low pressures due to the nonlocal transport of electrons.
Figure 10.3 shows the computerized tomography (CT) images of the three-
dimensional number density of the excited Ar(2p1 ) in pure Ar at 300 mTorr
and 13.56 MHz in E-mode (a) and H-mode (b) [2]. Both images are scanned
close to the coil plane. These figures demonstrate that in the E-mode the
coupling of energy into the plasma is capacitive and a strong ion sheath is
105
(II)
300 mTorr
Net ionization rate (arb.)
104
15 mTorr
103
102 50 mTorr
(I)
100 mTorr
10
5.0 10.0 15.0 20.0
Amplitude of coil current (A)
FIGURE 10.2
Net excitation rate of Ar(2p1 ) as a function of coil-current amplitude for pressures of 15 mTorr
(), 50 mTorr (), 100 mTorr (×), and 300 mTorr (◦) in ICP at 13.56 MHz in Ar in the reactor
shown in Figure 10.1a.

Inductively Coupled Plasma 277
(a) E-mode (b) H-mode
1.6 5.7
Nj (106 cm-3)
Nj (107 cm-3)
0.8 2.8
0.0 0.0
-60 60 -60 60
x (m 0 0 )
m x (m 0 0 )
m
m) 60 -60 y (m m) 60 -60 y (m
FIGURE 10.3
Emission CT images of short-lived excited atom Ar(2p1 ) in Ar. External conditions are 300 mTorr,
50 sccm, and amplitude 10 A of coil current at 13.56 MHz. (a) E mode (7 W) and (b) H mode
(30 W) correspond to (I) and (II) in Figure 10.2.
formed. The voltage on the coil is divided between the capacitances of the
sheath and the glass wall, so the transfer of power is azimuthally nonuni-
form. As a result, a local peak appears near the powered terminal. At lower
pressure, the nonlocal behavior of electrons distributes the excitation over the
entire radius and the azimuthal anisotropy is lost.
10.2.2 Mechanism of Plasma Maintenance

In this section we discuss the sustaining mechanism of an ICP in the H-mode.
We consider a cylindrical glass tube wound tightly by a ribbon-shaped coil
driven at 13.56 MHz. We then simplify the system to a one-dimensional system
as a function of radial position r and time t. The external coil current, I0 sin ωt,
generates an rf magnetic field Bz (r, t) under Ampere’s law,
Bz (r, t) = a (r )µ0 I0 sin ωt. (10.1)
Here, a (r ) is the radial component, µ0 is the permeability, and ω is the angular

frequency of the rf current source. At the same time, the azimuthal electric
field
Sra(r )dr
E θ (r, t) = − ωµθ Iφ Tθ cos ωt (10.2)
r
is induced by Faraday’s law. Also the ambipolar diffusion field Er (r, t) is
formed close to the wall and results in a passive wall sheath in a stational ICP.
ICP in the H-mode is mainly sustained by the collisional ionization under an
electron heating in the θ direction having power absorption
Pθ (t) = Jeθ (t) · E ∼ E θ2 ∼ ne ω2 I02 (1 + cos 2ωt) (10.3)

in a very low electromagnetic field, E θ ∼10 Vcm−1 and Bz ∼ 10 G. Under

a high-pressure condition the plasma production is limited in a space close
to the glass wall, and at low pressure of several mTorr, nonlocal electrons
affect the maintenance of ICP through the ionization collision allover reactor.
Supplementary ionization due to the E × B drift motion of electrons is expected
except in the case of low pressure.
PROBLEM 10.2.1
Prove that the ionization event under E × B fields leads by π/4 with respect to that
in the E θ -drift motion of electrons.
Exercise 10.2.1
Discuss the measurement method of the power deposition in an ICP sustained
by an external coil current.
When the coil is supplied from an rf source with variable power and an
appropriate matching box, the power deposited into the ICP is determined
by monitoring the coil current I (t) and subtracting Reff I (t)2 from the power
W(t) measured by the power meter, that is, W(t)− Reff I (t)2 . Here, the effective
impedance Reff is determined from the dissipated power W (t) in a vacuum
with no plasma in the same reactor at the same external coil current I (t). It is
assumed that the phase difference between W (t) and I (t) is negligible.
10.2.3 Effect of Metastables

ICP produces a diffusive high-density plasma. When the electron density
increases, we may expect the metastable or dissociated radical species to in-
crease as the electron density does. This increase can be estimated by using
a numerical modeling based on the fundamental collision and reaction pro-
cesses of the related species. Then, the stepwise ionization and quenching of
the metastable by electrons are the two basic and competitive mechanisms of
the maintenance of ICP.
We consider an ICP in Ar driven by a one-turn rf-coil current. The fun-
damental processes of the metastable Ar(3 P2 ,3 P0 ) in Ar are summarized in
Table 10.1. In a periodic steady state, the continuity equation of the metastable
density N∗ is
D∗ ∇ 2 N∗ (r, z) + kdi ne (r, z)Ng + ksi ne N∗ + kmp N∗ N∗

− keq ne N∗ − k2bq N∗ Ng − k3bq N∗ Ng Ng = 0, (10.4)
where each of the reaction rate coefficients corresponds to the value in

Table 10.1. We compare the rates of ionization caused by some of the pro-
cesses in Table 10.1 to estimate the effect of metastables on the maintenance.
The influence of the direct and stepwise ionization and the metastable pool-
ing on the plasma structure are simply estimated by comparison of each of
the rates (see Figure 10.4). The ratio of the net rate between the stepwise and

TABLE 10.1
Collision and Reaction Processes of Metastable Ar*(3 P2 ,3 P0 ) in Ar
Process Collision/Reaction Rate Coefficient(cm3 s−1 )
Excitation by electron e + Ar(1 S0 )→ e + Ar* B. Eq.

Electron quenching e + Ar* → e + Ar(upper state) B. Eq. (2 × 10−7 )
Superelastic e-quenching e + Ar* → e + Ar(1 S0 ) B. Eq.
Stepwise ionization e + Ar* → e + e + Ar+ B. eq.
Elastic collision Ar* + Ar(1 S0 )→ Ar* + Ar(1 S0 )
Metastable pooling Ar* + Ar* → e + Ar+ + Ar 6.2 × 10−10
Two-body quenching Ar* + Ar(1 S0 )→ 2Ar(1 S0 ) 2.1 × 10−15
Impurity quenching Ar* + Cl2 → Ar(1 S0 ) + Cl2 7.1 × 10−10 (at 300 K)
Ar* + CF4 → Ar(1 S0 ) + CF4 4 × 10−11 (at 300 K)
Ar* + SF6 → Ar(1 S0 ) + SF6 1.6 × 10−10 (at 300 K)
Ar* + N2 → Ar(1 S0 ) + N2 6.0 × 10−11 (at 300 K)
Three-body quenching Ar* + 2Ar(1 S0 )→ 3Ar(1 S0 ) 1.1 × 10−31 (cm6 s−1 )
Surface de-excitation Ar* + Wall → Ar
Diffusion coefficient NDAr ∗ 1.5 × 1018 (cm−1 s−1 )
direct ionization in Ar is
si ksi ne N∗ ksi N∗
= = , (10.5)
di kdi ne Ng kdi Ng
and for the rate of the metastable pooling to the direct ionization,
2
mp kmp N∗ N∗ kmp N∗ Ng
= = . (10.6)
di kdi ne Ng kdi Ng ne
A notable contribution of the metastable to the plasma maintenance be-

comes important at low plasma density at high pressure. The influence of the
(a) (b)
4.1 4.6
Λdi (1016 cm-3 s-1)
Λsi (1016 cm-3 s-1)
2.0 2.3
0.0 50 0.0 50
0 0
25 25
25 25
50 0 50 0
m ) m)
r (m z (m r (m z (m
m) m )
coil coil
FIGURE 10.4
Two-dimensional net ionization rate in ICP at 13.56 MHz and 50 A at 50 mTorr in Ar.

1.0
15 mTorr
0.8
0.6 50 mTorr
Λsi + Λdi
Λdi
0.4
0.2 100 mTorr
0.0
0.0 0.5 1.0 1.5 2.0 2.5
Power (kW)
FIGURE 10.5
Contribution of direct ionization to the maintenance of ICP at 13.56 MHz in Ar as a function of
dissipated power.
atomic metastable on the plasma is rapidly reduced through the molecular

quenching by impurity molecules (see, for example, Table 10.1). Figure 10.5
shows a comparison of the net ionization rate i among the production
terms in the ICP at 13.56 MHz in Ar as a function of power. Notice that
the direct ionization is dominant at low pressure (15 mTorr), and the step-
wise ionization through Ar* is predominant at high pressure (100 mTorr). At
50 mTorr, with increasing power the dominant process of ionization changes
from stepwise to direct ionization. The contribution of the metastable pool-
ing to the ionization is negligible as compared with the conditions mentioned
above.
The two-dimensional–density distribution of Ar* in the ICP in Figure 10.1a,
predicted by RCT-model, is shown in Figure 10.6 for comparison with the ex-
perimental profile by laser absorption spectroscopy. The spatial profile of the
metastable Ar* in the ICP reflects the production and the destruction mecha-
nisms. That is, the shallow minimum at the center is mostly caused by electron
quenching, and the broad peak in front of the wall is caused by production by
the collision between the electron and the feed gas Ar(1 S0 ). The database of the
fundamental process and the absolute value give the first accuracy. The pre-
dicted number density of Ar* coincides with the experimental value within a
factor of two. This difference in detail may be attributed to the difference of
the dissipated power and the possibility of impurity molecules in the reactor
during the experiment. Metastable molecules are produced both by the direct
excitation to the metastable state and by the cascade transition from the upper

(a) (b)
4.4 1.8
N* (1011 cm-3)
N* (1011 cm-3)
2.2 0.9
0.0 34 0.0 35
0 17 0 20
25 26
50 0 52 5
r (m m) r (m m)
m) z (m m) z (m
coil coil
FIGURE 10.6
Time-averaged number density of Ar*(r, z). (a) numerical value at 386 W (0.49 Wcm−3 ) and (b)
experimetal value at 400 W (0.25 Wcm−3 ). The other external conditions are 15 mTorr in Ar at
13.56 MHz.
excited states. The number density of the metastables is typically 1011 cm−3
under the balance between the production and loss by electron quenching
and diffusion to the wall. In a high-density plasma with ne ∼ 1011 cm−3 , the
effect of ionization of the metastables by a low-energy electron impact be-
comes less important under a strong electron quenching. On the other hand,
under a low-power and low-density plasma condition, metastable species
make a great contribution to the plasma maintenance through a low-energy
electron impact ionization (stepwise ionization) or metastable-metastable col-
lision (metastable pooling).
Exercise 10.2.2
An Ar atom has two metastable states, Ar(3 P2 ,3 P0 ). Investigate the density
ratio of the two metastables in an ICP in pure Ar.
The steady-state value after production depends on the electron quenching.
The value is 4 × 10−7 cm3 s−1 for Ar(3 P2 )–Ar(3 P0 ) and 1.5 × 10−7 cm3 s−1
for Ar(3 P0 )–Ar(1 P1 ), respectively. As a result, the density ratio N[Ar (3 P2 )]/
N[Ar (3 P0 )] is measured at 100 W as 5.7 at 50 mTorr and 7.3 at 5 mTorr, respec-
tively.
10.2.4 Function of ICP

ICPs and TCPs in H-mode are able to keep a high-density plasma in dif-
fusive mode, though the plasma is nonuniform in the production area (see
Figure 10.7). This means that a rather long distance is necessary between
the production region and wafer in order to realize a radial uniformity of the

6.0 15 mTorr
10 × N*
100 mTorr
ne
Number density (1012 cm-3)
4.0
5 × ne
30
Potential (V)
2.0
20
Vp 10
0 0
0 10 20 30 40 50
FIGURE 10.7
Typical plasma density and potential distribution at coil plane in ICP sustained at 13.56 MHz in
Ar: (a) 15 mTorr and 50 A, (b) 100 mTorr and 30 A.
plasma in front of the substrate. As a result, a high degree of dissociation of the

feed gas consisting of polyatomic molecules is induced in the large volume be-
tween the plasma production and the wafer. These characteristics in ICPs lead
to the marked contrast to the narrow-gap capacitively coupled plasma (CCP).
PROBLEM 10.2.2
The plasma potential in an ICP is usually 10 V to 20 V and the ambipolar diffusion
of electrons to the reactor wall controls the plasma density in the case where there is
no extinction of electrons in the bulk plasma. Estimate the time constant of electron
loss at 50 mTorr in Ar.
Exercise 10.2.3
Estimate the typical Larmor frequency ω L in an ICP and compare the value
with the rf frequency ω and the total collision rate Rt .
The typical ICP in Figure 10.1 is driven at 100 mTorr at 13.56 MHz at a current
amplitude of 30 A. Then, the Lamor frequency (cyclotron frequency) ω L of
the electron at 1 cm inside the glass wall is
eB 1.60 · 10−19 × 1.26 · 10−6 × 30/(2π 10−2 )

ωL = = = 1.06 × 108 s−1 .
m 9.11 · 10−31
On the other hand, the external frequency ω is 8.5 × 107 s−1 , and RT is usually
108 s−1 . The relation among these three frequencies is ω L ∼ ω ∼ RT .

10.3 Wave Propagation in Plasmas

10.3.1 Plasma and Skin Depth
We have learned that there exist two types of bulk plasma in gases: collisional
and colllisionless plasma. Here, the designation of collisional or collisionless
is derived from the presence or absence of two-body collisions between the
electron and the neutral molecule in the plasma. In the collision-dominated
plasma, the collisional event of electrons is represented by the rate of mo-
mentum transfer for the elastic scattering Rm , because the magnitude of Rm
is usually much greater than the rate for the inelastic collisions.
We investigate the influence of two-body collisions Rm , plasma density ne ,
and frequency ω pe on the propagation of an external electromagnetic wave
E0 e jωt incident on a uniform plasma (see Figure 10.8). When we consider the
phase delay φ of the electron drift velocity vde exp( jωt − φ) in a plasma, the
momentum conservation of electrons is given by

d
mvde e j (ωt−φ) = eE0 e jωt − mvde e j (ωt−φ) Rm . (10.7)
dt
By using the drift velocity derived from Equation 10.7, the electron current
density in the plasma is given as
ne e 2
Je (t) = ene vde (t) = E0 exp( jωt)
m(Rm + jω)

ne e 2 Rm 1 1 E0 exp( jωt)
= Rm 2 − j Rm 2 . (10.8)
m ω 1+ 1+ ω ω
ω
Here, the conduction of massive ions and the displacement current are ne-
glected in an rf wave. The wave propagation is described by the Maxewll
electromagnetic
wave (ω)
ω < ωpe
δ
plasma (ωpe)
electromagnetic
ω > ωpe wave (ω)
FIGURE 10.8
Propagation, cut-off, and decay of an electromagnetic wave in a uniform plasma.

equations (see Section 7.6)
1 ∂
rotB(t) = µJe (t) + E(t), (10.9)
c 2 ∂t
∂
rotE(t) = − B(t). (10.10)
∂t
If we take the rotation of Equation 10.10 and substitute it into Equation 10.9,
we derive the wave equation
ω2
∇ 2 E(t) + E(t) = jωµJe (t). (10.11)
c2
When the wave propagates in the z-direction in the plasma, the wave equation
is divided into the z- and r -directions:
d 2 E z (t) ω2
+ 2 E z (t) = jωµJ ez (t), (10.12)
dz2 c
ω2
Er (t) = jωµJ er (t). (10.13)
c2
We eliminate the conduction current density J e (t) in Equation 10.12 by using
Equation 10.8, and then, the axial component of the wave E z (t) satisfies

d2 ω2 ω2pe Rm ω2pe
E z (t) = − 1− 2 −j 2 E z (t), (10.14)
dz2 c2 ω + Rm
2 ω + Rm ω
2
where ω pe is the plasma electron frequency, (e 2 ne /m0 )1/2 . The solution has
the form
E z (t) = E 0 exp[ j (ωt − kz)]. (10.15)
Here, the complex wavenumber k is expressed as

1/2
ω ω2pe Rm ω2pe
k = (kr − jki ) = 1− 2 −j 2 . (10.16)
c ω + Rm
2 ω + Rm ω
2
As a result, we rewrite the propagating wave as
E z (z, t) = E 0 e −ki z e j (ωt−kr z) ,
where ki corresponds to the attenuation constant, and the wave is extin-

guished during the propagation of a length L in a plasma in the case of
ki L 1. The penetration length of the wave into the homogeneous plasma
is estimated practically by the distance that the wave decays to e −1 of the am-
plitude of the incident point, referred to as the skin depth (see Figure 10.8).

102 ωpe = 1010 s-1

-3
10
10 Rm / ω = 103
Skin depth (cm)
10-2
1.0
10-1 102
10
10-1
1.0
10-1 1.0 10 102

ωpe / ω
FIGURE 10.9
Skin depth of electromagnetic waves incident on a plasma.
That is, the skin depth δ at Rm =

0 is given as
 
√ 2 2 1/2 −1/2
1 2c  ω pe
2
ω pe Rm
2
ω pe
2

δ= =  1− 2 +  − 1−
2 
,
ki ω ω + Rm 2 ω + Rm ω
2 2 ω + Rm
2
(10.17)
where δ is shown in Figure 10.9 as a function of ω pe /ω and Rm /ω. The skin
depth is a concept describing the propagation of the electromagnetic wave
in a plasma and not for activation of the plasma production of neutral gases.
See Table 10.2.
TABLE 10.2
Typical Skin Depth as a Function of ω, ω pe , and Rm
Plasma Region ω and ωpe Skin Depth
c
Collisionless (Rm ω) ω pe ω ∼ ω pe
Rm ∼ ω ω pe ω ∼ √ 2 c
( 2+1)1/2 ω pe
Rm ∼ ω ω pe ∼ ω ∼ √ 2 c
( 2−1)1/2 ω pe
√
2c
Rm 1/2
Collisional (Rm ω) ∼ ω pe ω

10.3.2 ICP and the Skin Depth

Electrons in an ICP are heated by the azimuthal field induced by an external
current coil (see Figure 10.1a). The azimuthal electric field E θ satisfies the
wave equation

∂ 1 ∂ ∂2 ω2 ω2pe ω(ω + j Rm )
(r E θ ) + 2 E θ + 2 E θ = 2 , (10.18)
∂r r ∂r ∂z c c ω2 + Rm 2
and E θ is given by

3.83r
E θ (r, z, t) = E 0 J 1 exp[ j (ωt − kz)], (10.19)
r0
where J 1 is the first term of the Bessel function of the first kind and r0 is the
radius of the cylindrical reactor. k is the complex wavenumber and is obtained
from

3.83 2 ω2 ω2pe ω(ω + j Rm )
− − k2 + 2 = 2 .
r0 c c ω2 + Rm2
The skin depth is obtained by the reciprocal of the imaginary part of k,

2 2 1/2
c 2 1 + Rm /ω
δ= , (10.20)
ω pe b 1 + 1 + R2 /ω2 b 2 1/2
m
where b is
2 2
c 3.83 ω 2 2 2
b= 1+ − 1 + Rm /ω + 1.
ω pe r0 c
Exercise 10.3.1
There is a collisionless plasma that has a linear density distribution n(z) = n0 z.
Discuss the profile of the wave propagation as a function of z.
In a collisionless plasma (Rm ω) the dispersion relation is given by

c2k2 ω2pe
ξ2 ≡ 2 =1− 2 . (10.21)
ω ω
As shown in Figure 10.10, the propagation properties of electromagnetic
waves are classified into three regions with different modes as a function
of ne (z). An incident wave propagates from left to right. As the plasma fre-
quency ω pe is proportional to ne (z)1/2 , the wavelength of the electromagnetic
wave 2π/k becomes longer, and at ω pe = ω the wavelength is infinity. That
is, the wave is almost stationary in space and oscillates with ω in time. In the
region of ω pe ω, k is complex, and the wave is attenuated with the atten-
uation constant Im{k}. The frequency at ω = ω pe gives the cut-off frequency.
The position at ω pe = ω acts as the reflective wall, and the wave is reflected
to the left. Here, ξ is known as the refractive index.

n (z) = n0 z
ωpe (z) ∝ √z
reflection wave
attenuated wave
incident wave ξ (z)

0 z
ωpe < ω ωpe = ω ωpe > ω
FIGURE 10.10
Propagation of electromagnetic waves into a plasma with a spatial profile n0 z.
Exercise 10.3.2
When the intensity of electromagnetic waves is low enough not to perturb
the bulk plasma, the plasma density is measured by using the principle that
a wave propagating through a plasma has a phase shift relative to the wave
propagating in a vacuum. Practically, a microwave is used in order to obtain a
spatial resolution of ∼cm [3]. Derive a simple relation between the phase-shift
φ and and the plasma density ne .
The phase shift between two path φ(rad) is given as a function of the electron
plasma frequency, ω pe (z) = (e 2 ne (z)/m0 )1/2 :
1/2
ω pe (z)2
φ = (k0 − kplasma )dl = k0 1− 1− dz. (10.22)
l l ω2
Then,

φ = 2.82 × 10−17 λ0 ne (z)dz. (10.23)
l
In particular, when the plasma is uniform in the radial direction of the ICP
reactor, we obtain the plasma density ne (cm−3 ):
φ
ne = 1.18 f , (10.24)
L
where f (Hz) is the microwave frequency and L(cm) is the effective width of
the plasma investigated.

PROBLEM 10.3.1
The earth is surrounded with the ionospheric layer of plasma density, ∼106 cm−3 .
A shortwave broadcasting (frequency: ∼MHz-30 MHz) is operated on the ground
station. A satellite-based broadcasting is serviced at frequency greater than GHz.
Discuss the difference of the available broadcasting-frequency.
References
1. Ventzek, P.L.G., Hoekstra, R.J., and Kushner, M.J. 1994. J. Vac. Sci Technol. B
12:461.
2. Miyoshi, Y., Petrovic, Z.Lj., and Makabe, T. 2002, 2005. IEEE Trans. on Plasma
Sci. 30:130. Miyoshi, Y., Miyauchi, M., Oguni, A., and Makabe, T. IEEE Trans. on
Plasma Sci. 33:362.
3. Auciello, O. and Flamm, D.L. 1989. Plasma Diagnostics, Vol. 1. San Diego:
Academic Press.

11
Magnetically Enhanced Plasma
Magnetron plasmas and electron cyclotron resonance (ECR) plasma are used
in dry plasma processing at low pressure in which an external permanent
magnetic field makes a significant contribution to the plasma maintenance
by reducing electron losses to walls and electrodes. Helicon wave plasma
and magnetic neutral loop discharge (NLD) are also magnetized plasmas.
Collisionless electron heating through an electron cyclotron or a helicon wave
is the main mechanism to sustain these plasmas at very low pressure. On the
other hand, an induced magnetic field operates on an inductively coupled
plasma (ICP) and surface wave plasma (SWP).
11.1 Direct-Current Magnetron Plasma

Direct-current (DC) magnetron plasma has been widely used to deposit metal-
lic film on a large area substrate in electronic and photonic device fabrication.
Functional glass is prepared by material film coating. Sputter deposition of
Cu atoms ejected from a copper target in a DC magnetron is practically used
for interconnect in a trench or hole on SiO2 film. See Table 11.1.
Typical DC magnetron plasma (planar magnetron) is maintained between
two parallel plate electrodes, named the target and the substrate. Permanent
magnets are arranged radially from the center on the back of the target (cath-
ode) in order to supply a doughnutlike magnetic field in front of the target
(see Figure 11.1). DC magnetron discharge is usually sustained at about 200
V in nonreactive Ar at several mTorr under a permanent magnetic field par-
alllel to the target, several hundred Gauss. At the low-pressure condition, an
external magnetic field is essential for the maintenance of a plasma in a DC
source, and the effective lifetime of electrons in the magnetron is prolonged
(see Exercise 11.1.1). In particular, the net ionization rate is greatly enhanced
at a point where the magnetic field component parallel to the electric field is
zero, as compared with the plasma under no magnetic field.
The Lamor radius of ions is on the order of centimeters in contrast to the
sheath thickness with several mm in front of the target. Thus, ions in mag-
netron plasma are considered to be uninfluenced by the external magnetic
field. Inert Ar is used as the feed gas in a magnetron sputtering plasma for
several reasons. Inert Ar is chemically nonreactive on the target surface, the
massiveness of the Ar+ ion enhances sputtering on the target, and Ar is highly
289

TABLE 11.1
Magnetron Plasma for Sputtering
Type DC Magnetron rf Magnetron
Power source DC rf (13.56 MHz)

Magnet Permanent Permanent
Feed gas Ar Ar
Admixture O2 , N2 O2 , N2
Target material Metal Dielectric, metal
Film depo. Metallic Dielectric
Oxide metal
Nitride metal
Wire depo. Cu interconnect
abundant in the earth. Oxide or nitride metal film is processed in a DC mag-

netron with a metal target in an admixture of a small amount of O2 or N2 with
Ar. The process is known as a reactive sputter deposition.
Exercise 11.1.1
Both the electric E(=Ek) and the magnetic field B(=Bj) are applied in a gas at
a number density of N. Derive the neutral density N estimated by apparent
electron changes to
2 1/2
eB 1
N 1+ 2
. (11.1)
m Rm
The momentum conservation of a single electron with mass m and velocity

v is
d
(mv) + mvRm = e(E + v × B), (11.2)
dt
where E = Ek, B = Bj, v = (vx i + v y j + vz k), and Rm is the total collision rate
of the electron. We simply consider the case of Rm independent of v. Then,
v is

eB eE 1 eE Rm
v= − , 0, − . (11.3)
m m R2 + eB 2 m R2 + eB 2
m m m m
The electron mean free path λe under B is as

v eE 1
λe ∼ = eB 2 1/2 . (11.4)
Rm mRm
Rm 1 + m 1
Rm2
Equation 11.4 means that the gas number density changes from N to the above
expression 11.1. A decisive effect of a magnetic field on plasma production
is expressed under the condition of the electron cyclotron frequency (Larmor
frequency) ωce (= eB/m) Rm .

Magnetically Enhanced Plasma 291
(a) symmetric axis

metallic wall
feed gas substrate
z
target
r
N S N
DC
yoke
(b)
40 129
Br (Gauss)
Bz (Gauss)
-150 -241
-340 -611
30 30
z( z(
m m
m N 88 m N
88
) 0 ) 0
0S 0S
r (mm) r (mm)
FIGURE 11.1
A typical DC magnetron plasma reactor describing a magnetic field line (a); radial and axial
components of B(z, r ) (b).
Figure 11.2a shows a typical structure of a DC magnetron plasma in terms

of the number density distribution of electrons and ions obtained by a hybrid
model (see Section 7.4). The density distribution has a strong doughnutlike
peak at position r (34 mm, 6 mm) with Bz = 0 in front of the target. This is
caused by the presence of the peak net ionization rate at the position. It means
that the radial nonuniformity of the plasma density (i.e., of the ion flux to the
target) is the property intrinsic in a magnetron plasma, though it is capable of
sustaining a plasma at lower pressure, several mTorr, as compared with a DC
glow discharge or radio-frequency (rf) capacitively coupled plasma (CCP).
Figure 11.2b shows the potential distribution in the reactor configuration in

(a)
3.5 ne
ne, np (1010 cm-3)

np
1.7
0.0
30
z
(m
m 88
) N
0
0S
r (mm)
ne
np
(b)
5
Potential (V)
-98
-200
30
z
(m
m N 88
) 0
0S
r (mm)
FIGURE 11.2
Plasma density (a) and potential (b) distributions in a DC magnetron in Ar at 5 mTorr driven at
200 V in the reactor in Figure 11.1.
Figure 11.1a. A thin sheath region with a strong potential difference appears
in front of the target, and the potential in the bulk plasma (plasma poten-
tial) is very low when the plasma is surrounded by the metallic reactor wall
grounded to the earth. In these potential distributions, electrons are mainly
produced by the collisional ionization at the radially localized region with
Bz ∼ 0 in the sheath edge, and these electrons diffuse to the bulk plasma in
the very low electric and magnetic fields. On the other hand, ions produced
locally as the pair of the ionization are strongly accelerated to the target with
a beamlike energy, and these ions sputter the target material. It is easy to
estimate a local erosion profile by ion sputtering from the local ion flux inci-
dent on a target. Magnetron plasma is usually operated in a region of current
source. Removal of the external magnetic field, after the magnetron plasma
is formed, introduces the discharge plasma with extinction.

PROBLEM 11.1.1
Discuss the reasons why maximum ionization efficiency is realized at the point where
the magnetic field crosses at a right angle with the electric field in front of the target
(cathode) in a magnetron discharge.
PROBLEM 11.1.2
An azimuthal drift current in the plasma ring in front of the target is on the order of
few amperes. Estimate the magnetic field generated by the loop current of electrons
and compare the value with the external permanent magnetic field.
11.2 Unbalanced Magnetron Plasma

Almost all magnetic field lines from the north pole of the magnet arranged
behind the target terminate at the south pole through the gas phase in the
reactor. The magnetron plasma with these magnetic field arrangements is
called a balanced magnetron (BM). An unbalanced magnetron (UBM) has a
proper magnetic field configuration in which a finite degree of the field lines
from the outer magnetic pole diverge to the substrate, though the rest of the
lines finish on the inner pole behind the target (see Figure 11.3b). Sufficient
FIGURE 11.3
Comparison between a balanced and unbalanced magnetron (a) and (b).

plasma density and a positive ion current on a metallic substrate even at a

large distance from the target can be achieved in the UBM as compared with
the BM.
11.3 Radio-Frequency Magnetron Plasma

Magnetron plasma with a dielectric target is sustained only by an external
rf source, which is distinguished from a DC magnetron plasma with a metal
target. An rf power source at 13.56 MHz having a voltage waveform,
Vrf (t) = V0 sin ωt, (11.5)
is usually used through a blocking capacitor Cb . Therefore, the target surface

is negatively biased and the surface potential takes the form
t
1
Vsus (t) = Vrf (t) − Vdc (t) where Vdc (t) = IT (t)dt. (11.6)
Cb −∞
An rf magnetron sustained at 13.56 MHz has a weak temporal change of the

electron density close to the target except for the region trapped deeply by the
magnetic field. The other difference from the DC magnetron is the presence of
the phase to release the spatiotemporal electron trapping by electromagnetic
fields, based on E(r, t) = 0 twice during one rf period. This allows a radially
spreading distribution of the net ionization rate in front of the target (see
Figure 11.4 ) as compared with that in a DC magnetron. The phenomena
increase the efficiency both of the target utilization and of a much radially
uniform deposition on the substrate. Note that the value of Cb is carefully
arranged in the case of the dielectric target, especially for a ferrodielectric
target with a high dielectric constant.
Figure 11.4 shows the spatial profile of the electron density at each of the
external voltage phases of ωt = 0, π/2, π , and 3π/2 in Ar at 5 mTorr. A local
area close to the target (P B ) with Bz ∼ 0 is weakly modulated in time, because
the electrons under a large component of Br are spatially trapped by the
E(r, t)× B(r) fields in front of the target. As a result, the sustaining mechanism
of the rf plasma is classified into two regions depending on the radial position.
One is that of the conventional rf CCP, that is, the ionization multiplication
in front of the target by the reflected electrons. The other is the ionization
multiplication by the secondary electrons caused by the positive ion impact on
the target. That is, rf magnetron plasma is maintained by mixed mechanisms
between an rf CCP and a DC glow discharge. At low pressure such that rf CCP
cannot be sustained, the first mechanism disappears. Figure 11.5 shows the
ion flux incident on the Cu target (a) and the erosion profile (b) as a function
of radial position [1].

ωt = 0 ωt = π/2
1.2 1.2
ne [1010cm-3]
ne [1010cm-3]
0.6 0.6
0.0 0.0
S 8.0 S 8.0
4.0 4.0 4.0 4.0
N N
r[cm] 8.0 0.0 z[cm] r[cm] 8.0 0.0 z[cm]
ωt = 3π/2 ωt = π
1.2 1.2
ne [1010cm-3]
ne [1010cm-3]
0.6 0.6
0.0 0.0
S 8.0 S 8.0
4.0 4.0 4.0 4.0
N N
r[cm] 8.0 0.0 z[cm] r[cm] 8.0 0.0 z[cm]
FIGURE 11.4
Typical electron density distribution in an rf magnetron sustained in Ar at 5 mTorr at 13.56 MHz
and V0 = 400 V.
11.4 Magnetic Confinements of Plasmas

When the operating pressure of a plasma drops to the order of mTorr, electrons
and ions produced in a plasma rapidly diffuse and are lost to the reactor wall
by the lack of binary collisions between the electron (or ion) and neutral feed
gas molecule. The multipolar magnetic field arrangement on the reactor side
wall is effective for the confinement of electrons inside the reactor, when the
mean free path of fast electrons is equal to or larger than the chamber size. A
magnetically confined reactor is arranged by an array of parmanent magnets
with N-poles and S-poles alternately, positioned outside the reactor as shown
in Figure 11.6a. The reactor with surface cusp magnetic fields is widely used
in the application of a low-pressure plasma to material processings.

(a)
0
-1
cm s )
-3 -1
-2
15
Dielectric target
Ions flux(10
-3
Metallic target
-4 PB
Target
-5
0 20 40 60 80
r (mm)
(b)
PB
Depth (arb.)
-1
Target
0 20 40 60 80
r (mm)
FIGURE 11.5
Ion flux incident on Cu target (a) and erosion profile (b) in an rf magnetron plasma.
S
Er
B
N
θp
FIGURE 11.6
Magnetically confined plasma reactor. Arrangement of multipolar magnets and net ionization
rate.

The magnetic mirror effect near the cusp of two magnetic poles suppresses
the electron diffusion in the direction normal to the magnetic field line in ad-
dition to the electron reflection in the wall sheath on the reactor. A set of
straight magnets is arranged in the axial direction on the outer surface of the
cylindrical reactor. The resulting line cusp magnetic field near the chamber
wall confines electrons and ions. Such a field configuration has the proper-
ties of a magnetic mirror and has a very small effect on the confinement of
slow electrons. That is, with increasing the number, the electron confinement
becomes stronger, and the escape of electrons from the loss cone becomes
significant [2]. The electron mean energy rises near the wall due to the pres-
ence of the high-energy electrons trapped by the multipolar magnetic field.
The net ionization rate is expected to have a local peak. The local electron
energy increases when the magnitude of the magnetic field increases due to
the increase of the trapped high-energy electron. There is an optimum for the
number of the magnetic poles. The magnetic field has a very small effect on
the confinement of slow electrons. At pressure range where the electron mean
free path is less than the reactor dimension, the magnetic field has little effect
on the electron confinement by the trapping.
11.5 Magnetically Resonant Plasmas

We consider an elctromagnetic wave, propagating parallel to the applied mag-
netic field B in a cold and uniform plasma. The wave has a frequency in that
ion motion is negligible , ω ω pi . It is known that there are two modes, R-
and L-modes, for electromagnetic waves traveling along the magnetic field.
Each of the refractive indexes, ξ+ and ξ− , are given by [3–5]

k2c2 ω2pe
ξ±2 ≡ 2 =1− , (11.7)
ω (ω ± ce )(ω ± cp )
where ce = eB/m (< 0) and cp = eB/M (> 0) are the cyclotron frequencies
of electrons and ions, respectively. Equation 11.7 exhibits that the R- and
L-modes have cut-off frequencies, ω+ and ω− (see Chapter 10),
1/2
( p − e )2 cp + ce
ω± = ω2pe + ∓ . (11.8)
4 2
We see that the rotational direction of the R-wave corresponds to the direction
of the cyclotron motion of electrons in the magnetic field. Note that the de-
nominator of Equation 11.7 becomes zero when the frequency ω of the R-wave
approaches |ce |. Then the electrons are continuously accelerated and result
in the absorption of the energy of the electromagnetic wave. It is known as
the ECR between the electron and the wave. It is noted that the L-wave has
no resonances with electrons but with ions.

(a) (b)
microwave quartz
electron-cyclotron wave guide tube
wave
quartz window
antenna rf power
coil
B
coil
parmanent
magnet
ECR layer
B
plasma
wafer plasma
wafer
rf bias rf bias
FIGURE 11.7
Representative for magnetically resonant plasma. ECR plasma source (a) and helicon source (b).
ECR plasma is excited by the resonance mechanism of the R-mode de-

scribed above. A typical ECR plasma source is driven at microwave power
with 2.45 GHz in low pressure of mTorr. It is the principle to synchronize
the electron cyclotron frequency |ce | with the UHF (microwave) frequency
2.45 GHz at B of 875 G by adjusting the external coil current. In the resonance
point, electrons efficiently get the external microwave energy and a diffusive
high-density plasma is produced at very low pressure in a reactor as shown in
Figure 11.7a. Microwave power through a wave guide is fed into the reactor
through a quartz window.
Wave heated plasmas are described in detail by Lieberman and
Lichtenberg [6].
PROBLEM 11.5.1
Calculate the magnitude of the magnetic field at the ECR condition of electrons in a
collisionless plasma supplied by the microwave power at 2.45 GHz.
References
1. Kuroiwa, S., Mine, T., Yagisawa, T., and Makabe, T. 2005. J. Vac. Sci. Technol. B,
23:2218.
2. Takekida, H. and Nanbu, K. 2004. J. Phys. D (37:1800).

3. Boyd, T.J.M. and Sanderson, J.J. 2003. The Physics of Plasmas. Cambridge:
Cambridge University Press.
4. Sturrock, P.A. 1994. Plasma Physics. Cambridge: Cambridge University Press.
5. Nicholson, D.R. 1983. Introduction to Plasma Theory. New York: John Wiley &
Sons.
6. Lieberman, M.A., and Lichtenberg, A.J. 2005. Principles of Plasma Discharges
and Materials Processing (2nd Edition). Hoboken: John Wiley & Sons.

12
Plasma Processing and Related Topics
12.1 Introduction
Plasma processes (i.e., sputtering, deposition, etching, surface treatment, etc.)
require information about the surface reactions of active species and their
probability, as well as information on gas-phase collision/reaction processes
and their cross sections. Physical or chemical quantities describing the plasma
surface interaction are given by an effective surface reaction probability con-
sisting of the sticking coefficient and the yield for etching or deposition.
Historically, the collision cross section of the electron or ion in the gas
phase has been continuously accumulated theoretically and experimentally
as a function of the impact energy, in addition to the study of the quan-
tum characteristics of mono- and polyatomic molecules in the field of atomic
and molecular physics. In the same way, we expect that the surface reaction
process and the probability of neutral molecules and ions with a material
surface will be rapidly elucidated and accumulated by using first-principle
molecular dynamics and measurements.
12.2 Physical Sputtering

Metallic film deposition is usually performed by using a direct-current (DC)
magnetron plasma with a metal target in pure Ar (see Figure 12.1). A rel-
atively large number of data are available on ion sputtering in ion–surface
interactions. Most of these data were collected for systems between metal
targets and rare-gas ions with energy ranging from several hundred eV to
several hundred keV under a clean surface [1, 2]. Very little is known about
systems with low-energy ions with p < 100 eV. Reactive sputtering is a tech-
nique widely used for deposition of compound materials, for example, oxides
(Al2 O3 , Ta2 O5 , etc.) and nitrides (Si3 N4 ). A sputtering system of a pure metal
target in admixture with reactive gas (O2 or N2 ) and Ar holds the advan-
tage over a system in a compound target in Ar from the viewpoints of the
stoichiometry of the film and the power density and thermal conductivity
of the target. Chemical sputtering is generally considered to be a multistep
301

Substrate
St
deposition
+
M , εp
sputtered
impact
Y(ε)
Target erosion
FIGURE 12.1
Physical sputtering system.
process finally leading to the formation of a volatile molecule that escapes

into the gaseous phase. The sputtering efficiency of the system between an
ion with incident energy ε and angle θ and a solid surface is described by the
sputtering yield Y(ε, θ ) as shown in Figure 12.2 as examples [3].
Exercise 12.2.1
A zero-dimensional model of a reactive sputtering process is successfully
used for compound film deposition in a DC magnetron sputtering with a
metallic target in admixture with reactive gas and Ar. Derive the governing
equations of the system in which a tantalum (Ta) target is sputtered in an
Ar/O2 mixture to deposit Ta2 O5 compound on a substrate [4].
The reaction on the surface in this system is divided into three regions: the tar-
get; substrate; and side wall with areas of AT , A S , and AW (see Figure 12.3a).
A fraction (1 − X) of the target material sputtered by the Ar+ ions with flux
p deposits on the substrate, and X arrives at the side wall. T , S , and W
give the degree of the compound (Ta2 O5 ) formation on the target, substrate,
and reactor wall, respectively. The values of T and S are a practical indi-
cator of the reactive sputtering from pure metallic target ( T = 0) resulting
in deposition of a stoichiometric compound film at the substrate ( S = 1).
st , ss , and sw are the sticking coefficients of O2 to the metallic part of the
target, substrate, and wall, respectively. The sticking of the sputtered parti-
cles on the target is assumed to be unity. YTm and YTc are the sputtering yields
of the metal and compound material at the target. The number ratio of the
reactive atom (O) between the reactive gas (O2 ) and the compound molecule
(Ta2 O5 ) is denoted by h r .
At the target surface, the balance equation of the compound is
∂
T (t)AT = h r O2 st (1 − T )AT − p YTc T AT . (12.1)
∂t

Plasma Processing and Related Topics 303
(a)
Incident ion energy (eV)
100 300 500
101
1
Sputtering yield Y(ε)
+
Ar / Cu
10-1 3.0
Y(250 eV; θ)
Y(ε; θ)
2.0
10-2
Y(100 eV; θ) 1.0
10-3 0
0 30 60 90
Incident angle (deg.)
(b)
0 50 100 150 200 250
1.2
10 eV
1.0
15 eV
Sticking probability
0.8
Sputtering yield
25 eV
0.6
+
Cu / Cu
0.4
35 eV
0.2 Y(ε)
50 eV 75 eV
0
100 eV
0 20 40 60 80 100
Incident angle (deg.)
FIGURE 12.2
Sputtering yield Y(ε) and Y(ε, θ ) of Ar+ -Cu target (a), and Y(ε) of Cu+ –Cu target and sticking
coefficient St(θ ) of Cu on Cu substrate (b).
At the substrate surface, the balance equation of the compound is described

as
∂
S (t)AS = b m h r O2 ss (1 − S )AS + (1 − S )b m (1 − X) p YTc T AT
∂t
− S (1 − X) p YTm (1 − T )AT , (12.2)

(a) Substrate : AS
(1-ΘS) ΘS
SS
ΘW
Wall : AW
reactive gas O2
(1-ΘW)
positive ion SW
ΓO2
Γp
ST
YTm YTc
(1-ΘT) ΘT
Target : AT
(b) 1
Deposition rate (arb.)
0.5
0
0 1 2
Reactive gas flow (arb.)
FIGURE 12.3
The 0th-order reactive sputtering model (a), and the predicted deposition rate of a compound
film as a function of partial pressure (b).
where b m is the number of metal atoms in the compound molecule. In the

same way, for the side wall, the surface metal balance is described as
∂
W (t)AW = b m h r O2 sw (1 − W )AW + b m (1 − W )X p YTc T AT
∂t
− W X p YTm (1 − T )AT . (12.3)
The total supply q of the reactive gas into the reactor is equal to the consump-
tion on the surface and the quantity pumped out to the outside. That is,
q = st O2 (1− T )AT + sw O2 (1− W )AW + ss O2 (1− S )AS + pr Vpump , (12.4)
where pr is the partial pressure of the reactive gas and related to O2 =

pr /(2π MkT)1/2 , and Vpump is the pumping speed of the gas. Here, the plasma
parameters, incident ion flux p , partial gas pressure pr , and pumping speed

ppump are given. Unknown parameters, X, T , S , and W can be solved in

the steady state from Equations 12.1 to 12.4.
The average deposition rate per unit area on a substrate R depo is given as
AT
R depo = p {YTc T + YTm (1 − T )}(1 − X). (12.5)
AS
The deposition rate R depo in the reactive sputtering has a range with multival-
ues as a function of reactive gas flow, as shown in Figure 12.3b.
12.2.1 Target Erosion

Magnetron plasma is widely applied to thin film deposition. The target in
a magnetron discharge, that is, cathode, is eroded by physical sputtering by
high-energy ion impact under a low-pressure and highly localized plasma
condition. The target erosion is described by the sputtering yield,
atoms removed
Y( p ; target) = , (12.6)
incident ion
for the system between an incident ion with energy p and target material.
The databases are widely available in the literature [1, 2]. The number of
target atoms ejected by the sputtering of ions with velocity v incident on the
target surface dS from the magnetron plasma is estimated by the velocity
distribution of the incident ion, g p ( p , θ), as

ρ dSdl = n p Y( p ; target)vg p ( p , θ, r)dθ d p dt dS, (12.7)
θ p
where ρ is the atomic number density of the target material, dl is the erosion
depth during a small time dt, and g p ( p , θ, r) is normalized to unity as

g p ( p , θ, r)dθd p = 1. (12.8)
p
The sputter rate Rsp (r) is obtained by

dl n p (r)
Rsp (r) = = Y( p ; target)vg p ( p , θ, r)dθ d p . (12.9)
dt ρ θ p
The erosion depth profile of the target is estimated numerically by the time
development of the ion sputtering in a magnetron plasma.
PROBLEM 12.2.1
Even in a low-pressure magnetron plasma, ions affect a target with a finite angle of
incidence. Then, we must consider the angular dependence of the sputtering yield
Y(ε, θ ). Revise Equation 12.9 of the sputter rate by considering Y(ε, θ).

12.2.2 Sputtered Particle Transport

Next we consider the flux and energy of the atoms sputtered from the target.
The ejected flux of the target atom is given at the surface rt by

Γs (rt ) = n p Y( p ; target)vg p ( p , θ, r)dθ d. (12.10)
θ p
On the other hand, the energy spectrum of the sputtered atom is estimated
by the Thompson formula [5],
1 − {(Ut + )/γ p }1/2

f s (; p , rt ) = A , (12.11)
2 (1 + Ut /)
where A is constant. p and are the kinetic energy of the incident ion and
the sputtered atom, respectively. Ut is the binding energy of a target material.
γ is the energy transfer factor in the elastic collision between the incident ion
and the target atom with masses of Mp and Ms , respectively, and is given by
4Mp Ms
γ = . (12.12)
(Mp + Ms )
PROBLEM 12.2.2
Discuss the reason that the angular distribution of the ejected neutral is usually
approximated by the cos θ -law in the physical sputtering process.
The Thompson formula is derived under the condition that the sputtered
atoms come from a well-developed collision cascade in a material. The colli-
sion cascade is realized in a system between an incident heavy ion and light
atom in the material. In a system between a light incident ion with low energy
and a target consisting of a massive atom, however, the ion from the sheath
will be easily backscattered by a massive target atom and will knock off an
atom in the top layer of the target. Then, the sputtering is caused by a single
knock-on mechanism. A modified formula has been proposed for the system
between a light ion and massive target atom [6]:

γ (1 − γ ) p 2
f s (; p , rt ) = A ln , (12.13)
( + Ut )α+1 + Ut
where α = 3/5 for the H+ –Fe system. The angular distribution of the ejected
atom is described by cosine law.
We consider the arrival flux of the sputtered atom at the substrate rs . First,
the flux without collision in gas phase is given by

Γs (rs ; ) =
0
Γs (rt )exp(−NQl)dS, (12.14)
St
where l is the distance between rt and rs . Q is the collision cross section

between the sputtered atom and feed gas molecule. Here we estimate the

transport of the relaxed component Nsrel by collision with feed gas molecules.
The neutral transport is described by the diffusion equation
∂ rel
N (r, t) = Ds ∇ 2 Nsrel (r, t) + rel
s (r, t), (12.15)
∂t s
where Ds is the diffusion coefficient of the sputtered atom in the feed gas
molecule. The production rate of the relaxed component rel s is given under
the assumption that the sputtered atom is randomized in energy and direction
after one collision with feed gas molecules. Then, from the difference of the
ejected flux distribution at small distance (l, l + dl),
rel
s (r) = −∇ · s (r).
0
(12.16)
In a steady-state (∂/∂t = 0), the spatial transport of the sputtered atom under
relaxation is obtained by
Ds ∇ 2 Nsrel (r) − ∇ · s0 (r) = 0. (12.17)
The arrival flux of the sputtered atom, collisionally relaxed in the gas phase,
is given by
srel (rs ) = −Ds ∇ · Nsrel (rs ). (12.18)
As a result, the total flux arriving at the substrate at rs is
stotal (rs ) = s0 (rs ) + srel (rs ). (12.19)
Figure 12.4 shows the spatial density distribution of Ar+ ions and sput-
tered Cu neutrals in a DC magnetron sputtering system operated between
1.0 × 109(cm-3) 3.0 × 108(cm-3)

1.5 × 109
1.0 × 1010 np
nCu 1.5 × 1010 1.5 × 109

3.0 × 109
r
N S N
FIGURE 12.4
Relations of the distribution between the ion density n p (r) and sputtered atom Ns (r) in a DC
magnetron plasma.

TABLE 12.1
Chemical Vapor Deposition of SiH4 (+H2 )
Deposited Film Process Radicals Conditions Ref.
a-Si:H, µc-Si:H Plasma process H, Si, SiH2 , SiH3 400 K, 1–103 Pa [7]
Poly-Si Thermal process SiH2 , H2 , 900 K, 10–103 Pa —
a-Si:H, µc-Si:H Catalytic process H, Si 2000 K, 0.5–10 Pa [8]
a-Si:H Photo process — Room temperature —
axisymmetric parallel plates at 5 mTorr in Ar sustained at Vsus of −200 V and

maximum B
of 300 G on a Cu target surface.
12.3 Plasma Chemical Vapor Deposition

Chemical vapor deposition (CVD) is a process for fabricating thin films or par-
ticles using chemical reactions of gas molecules. CVD is classified according
to the various forms of the external energy source into plasma CVD, thermal
CVD, photo CVD, and catalytic CVD. In these different processes, chemi-
cally active species (precursors) produced by dissociation of gas molecules
have a strong influence on the deposition. To distinguish it from CVD, a de-
position through a physical vacuum evaporation or sputtering is termed a
physical vapor deposition (PVD). The quality of the film deposited is influ-
enced by the properties of the precursor molecules, which depend on the
process and the external condition. There are several different methods for
depositing Si films (see Table 12.1): low-temperature plasma processing, ther-
mal decomposition, photo decomposition, and catalytic decomposition on a
hot tungsten filament.
12.3.1 Plasma CVD

Capacitively coupled parallel plates plasma is widely used for hydrogenated
amorphous silicon (a-Si:H) deposition. A good-quality a-Si:H film with a de-
position area of up to a few m2 can be manufactured at a very low-power
condition, 50 mWcm−2 , in SiH4 /H2 in parallel plate capacitively coupled
plasma (CCP). Plasma damage to the film quality caused by an ion impact
during deposition is prevented by a wafer arrangement on the grounded
electrode and by using a high-frequency source at 13.56 MHz and more ef-
ficiently at very high frequency, 100 MHz. Recently, microcrystalline silicon
(µc-Si:H) thin film has been developed into the basic materials of large-area
solar cell and thin film transistor (TFT) for a large-area flat pannel display.
Hydrogenated amorphous carbon (a -C:H) or diamondlike carbon (DLC) is
the other application of the low-temperature plasma CVD. a -C:H film is hard

and wear resistant. The head and disk surfaces in magnetic disk drives are
usually coated with a thin layer of DLC to minimize wear and corrosion.
Thin films deposited by plasma polymerization (surface grafting) have
a large variety of applications. The plasma has a high ability to modify the
surface. These materials are widely used as biocompatible films with hy-
drophilicity and are suitable for various biomedical and pharmaceutical mate-
rials, for example, optical lenses, implants, and drug delivery devices. Plasma
polymerization is usually performed in a glass reactor excited by helical coils
driven at 13.56 MHz (i.e., inductively) in carrier gas Ar at several hundred
mTorr with monomer vapor. Excitation of the carrier gas takes place in radio-
frequency (rf) Ar plasma, generating active species (ions and radicals) in
the gas phase. These active species interact with the substrate and generate
reactive sites (mainly free radicals) on the wafer surface. The monomer va-
pors in the plasma chamber readily react with these radicals, yielding grafted
surfaces.
Exercise 12.3.1
Diamond thin film grows on a diamond substrate in a thermal plasma in
CH4 /H2 . The surface under deposition is divided into two sites: the diamond
lattice growth with probability D and nth layers of amorphous carbon with
probability Cn . Atomic H and C, dissociated from the feed gases CH4 /H2 , in
thermal plasma diffuse to the substrate surface. The thin film growth is highly
selective on the site and highly competitive between deposition and etching
and is modeled by the birth and death process in the stochastic process [9–11].
That is, on the substrate surface, the evolution equation is given in the form
of the 0th-order simultaneous rate equations
dD
= kc H C1 − ks C D + ketch H C1 (12.20)
dt
dC1
= −kc H C1 + ks C D − ketch H C1 − ks C C1 + ketch H C2 (12.21)
dt
dCn
= ks C Cn−1 − ketch H Cn − ks C Cn + ketch H Cn+1 (n > 1), (12.22)
dt
where ks is the sticking coefficient of the C atom. ketch is the etching probability
of a -C:H by an H atom. kc is the conversion probability of a -C:H to diamond
lattice by an H atom, and
∞

D+ Cn ≡ 1. (12.23)
n=1
1. Derive the surface structure (C1 , Cn , and D) in the steady-state con-

dition.
2. Derive the growth rate of the diamond film R depo .

By summing up Equations 12.20 to 12.22 in the steady state, we obtain the

relation
∞
dD dCn
+ = −ks C Cn + ketch H Cn+1 = 0.
dt n=1
dt
Therefore,
ks C
Cn = χCn−1 where χ= .
ketch H
We obtain Cn as
(1 − χ n )
Cn = C1 . (12.24)
(1 − χ )
By substituting Equation 12.24 into Equation 12.23, we have
∞
(1 − χ n ) C1
D + lim C1 = 1 and D = 1 − . (12.25)
n→
n=1
(1 − χ ) (1 − χ )
Also, from Equation 12.20 we have

χD
C1 = .
1 + kketch
c
Equations 12.24 and 12.25 give D , C1 , and Cn as

(1 − χ ) 1 + kketch
c
D= ,
1 + (1 − χ ) kketch
c
(1 − χ )χ
C1 = ,
1 + (1 − χ ) kketch
D
(1 − χ n )χ
Cn = . (12.26)
1 + (1 − χ ) kketch
C
The deposition rate of diamond R depo is

R depo = kc H C1 − ketch
D
H D,
D
where the etching rate coefficient ketch is less than the other rate constant, and
ks (1 − χ ) kketch
c
c
R depo ∼ kc H C1 = . (12.27)
1 + (1 − χ ) kketch
c
At kc ketch ,
ks (1 − χ )C
R depo (χ ) = ∼ kc C . (12.28)
(1 − χ ) + kketch
c
At kc ketch ,
kc
R depo (χ ) = ks (1 − χ ) C . (12.29)
ketch

12.3.2 Large-Area Deposition with High Rate

High-quality thin film of hydrogenated amorphous silicon (a-Si:H) is fabri-
cated under a low-energy ion impact in a very high frequency (VHF) plasma.
A VHF source provides a higher density plasma with low-energy ions based
on a high degree of spatial trap of electrons as compared with that at
13.56 MHz. However, inhomogeneity of plasma arises from the nonuniform
surface potential on the powered electrode caused by a standing wave of the
VHF voltage. Figure 9.9 shows the potential nonuniformity of the rod elec-
trode as a function of position in a plasma of 2×109 cm−3 sustained at 100 MHz
at 13 Pa in Ar. The numerical value, simulated by the transmission-line model
(TLM) (see Section 7.5.3) reproduces the experimental observation. Numeri-
cal design of the surface potential uniformity is achieved by time-averaging
the surface potential through the control of the modulation rate of the phase
of the VHF source at the feed position in a one- or two-dimensional electrode
system.
12.4 Plasma Etching

Dry etching for micro- and nanoelectronic device fabrication should be highly
selective of one material over others and highly competitive with the
deposition process. A competitive surface between deposition and etching
is described by two physical quantities: the etching yield Yetch () and the
sticking coefficient St (θ ) of incident active particles. Plasma etching has two
extremely different phases, that is, isotropic chemical etching by neutral rad-
icals and anisotropic etching assisted by energetic ion impacts (i.e., reactive
ion etching). It is much more complicated in a practical etching that has a
surface process between a purely physical sputtering and a spontaneous
chemical etching as well as deposition as a function of radical-to-ion ratio inci-
dent on the wafer. The practical etching in ultra-large scale integrated (ULSI)
manufacturing is mainly devoted to two processes: Si-gate etching and SiO2 -
contact hole (trench) etching (see Table 12.2 and Figure 12.5). In general, the
atomic scale mechanisms that enhance the surface process by ion impact are
i. Sticking coefficient of radicals;
ii. Diffusion of etchant (i.e., surface coverage);
iii. Impact damage; and
iv. Volatile- and nonvolatile-molecule productions, and the like.
Together, these effects mean that ion-assisted etching in a plasma process
produces an increase in the local surface coverage of the etchant and enhances
the effective etching yield compared with chemical etching under the same
number of radicals incident on the surface.
The etching yield Yetch (ε) is defined as the ratio between the number of
removed substrate molecules and the number of the incident species on the

TABLE 12.2
Plasma Etching of Materials
Materials Feed Gas Property
Aluminum (Al) Cl2 Direct reaction between Al and Cl2

sidewall passivation of C compounds
Nat. oxided-Al BCl3 Deoxidization + ion-assisted etching
Silicon (Si) Cl2 , HBr, etc. Low-energy ion-assisted
sidewall passivation of SiO2
SiO2 C4 F8 , CHF3 , etc. High-energy ion-assisted, high-selectivity
sidewall passivation of polymer (Ci F j )
SiOCH (low-k) C4 F8 /O2 High-energy ion assisted
Organic low-k N2 /H2 , NH3 High-energy ion assisted
sidewall passivation of Ci N j Ok
Photoresist O2 Plasma trimming
Comp. semicon. Comp. of halogen or H
Metal (high b.p.) Compounds of F or Cl
Transition group Difficult for reactive etching
substrate. More specifically, the etching yield of ions is a function of the impact
energy ε p and the incident angle θ , Y(ε p , θ). The etching rate R etch (r) of ions
with velocity distribution g p (v, r) incident on the substrate r is related to the
etching yield Y(ε p , θ ) when we define R etch (r) as

n p (r)
R etch (r) = Y(ε p , θ)vg p (v, r)dv/ g p (v, r)dv, (12.30)
ρn
where n p (r) is the number density of ions close to the wafer. ρn is the atomic
number density of the substrate material.
Reactive plasma
Cl2 - plasma Cx Fy - plasma
resist resist
Passivation layer
(SiOClx) Passivation layer
(CFx-Polymer)
Mixed layer
poly-Si SiO2
(a) (b)
FIGURE 12.5
Typical etching profile in Si-ULSI: Si-gate etching (a) and trench or hole etching of SiO2 (b).

PROBLEM 12.4.1
Reactive ion etching (RIE) is based on a first adsorption of radical species on a
surface and successive ion impact with high energy. Discuss the necessary condition
of radicals and ions incident on the surface in the RIE.
12.4.1 Wafer Bias

The wafer to be etched is set on the substrate holder, which is usually elec-
trically isolated from the reactor base potential. The simple relation between
the wafer potential and the impact ion energy is briefly described in advance.
12.4.1.1 On Electrically Isolated Wafers (without Radio-Frequency Bias)

When the wafer is set on an electrically insulated holder from the system,
the wafer surface is kept at floating potential Vf l , which provides the am-
bipolar diffusion with ee = e p on the surface in a time-averaged fashion.
In a collisionless sheath, Bohm’s sheath criterion predicts the relation (see
Section 3.6),

kTe Mp
Vplasma − Vf l = ln , (12.31)
2e 2.3m
where Vplasma and Te are the time-averaged potential and electron temperature
in the bulk plasma. Mp and m are the mass of the positive ion and the electron,
respectively.
12.4.1.2 On Wafers with Radio-Frequency Bias

In the case of a high-energy ion-assisted etching, we must apply an rf bias
voltage Vbias (t) on the holder. For sophisticated modern plasma etching, an
ion-assisted etching with very high energy, 500 eV to 1 keV, is practical. Then,
it is essential to perform the functional separation between plasma produc-
tion and ion acceleration, because some degree of ionization is unavoidable in
the active sheath in front of the biased wafer. As already described in Chapter
8, a negative DC self-bias voltage Vdc appears on the wafer under the am-
bipolar diffusion in the steady state. Accordingly, in a time-averaged fashion,
the wafer surface is irradiated by positive ions with energy ε p , which is
defined as
ε p = e(Vplasma − Vdc ). (12.32)
As a result, the impact energy of ions on the wafer is roughly estimated by

the relation with/without bias voltage.

kTe Mp
ε p = ln : without bias (floating);
2 2.3m
= e(Vplasma − Vdc ) : with rf bias;
= e Vplasma : on the ground.

Another important particle on the surface exposed to the plasma etching is

the chemically active radicals. The neutral radical transport is usually treated
by the random motion in the gas phase. That is, the radical flux incident on
a wafer with/without rf bias voltage is estimated by using the Maxwellian
velocity distribution in the quasithermal condition
1/2
Nr 8kTg
radical = , (12.33)
4 π Mr
where Tg is the gas temperature. Nr and Mr are the number density and mass
of the neutral radical, respectively.
12.4.2 Selection of Feed Gas

Dry etching in a low-temperature plasma utilizes a physical and chemical
reactivity on material surfaces based on ions and dissociated neutrals (radi-
cals) in a plasma. Feed gases for plasma etching are carefully prepared under
conditions (see Table 12.2):
i. The reactive product caused by the surface reaction must be in

gas phase (volatile) in order to remove the surface material, and
thus the vapor pressure will be high for the reactive product (see
Table 12.3);
ii. The binding energy of reactive products must be lower than that of
the material to be etched; and
TABLE 12.3
Reactive Product in Plasma Etching and the Boiling Point
and Vapor Pressure
Materials Reactive Product Boiling Point (C) Vapor Pressure
Al AlF3 1291 —
AlCl3 183* 0.6
AlBr3 255* —
Si SiF4 –90 ≤ 760
SiCl4 57 —
SiBr4 5.4 —
Fe FeF2 ≤ 1000 —
FeCl3 319 —
FeBr2 684 —
Ni NiCl2 1000 —
Ni(CO)4 -25 — —
W WF6 17* —
WCl6 346* —
Co CoF2 1200 —
CoCl2 1050 —
∗ Sublimation temperature.

iii. It is preferable that the feed gas, the dissociated molecules, and the
reactive products are all nontoxic and all have low global warming
potentials.
Admixture gases are usually used for plasma etching for the following
purposes:
i. Density control of the dissociated chemically active species (i.e.,

radicals);
ii. Control of the absorption site on the material surface to be etched;
iii. Control of the surface reactivity;
iv. Improvement of heat transfer in gases and of surface cooling; and
v. Control of the plasma density (impedance) by way of the plasma
structure.
The influence of each of the gases on surface etching is very complicated. Most
of the feed gases used for plasma etching are greenhouse gases. That is, green-
house gases absorb infrared radiation and trap energy in the atmosphere. The
atmospheric lifetime and global warming potential (GWP) characterize the
effect of the greenhouse gas. The GWP of a greenhouse gas is the ratio of
global warming from one unit mass of a greenhouse gas to that of one unit
mass of CO2 over a period of time. Hence this is a measure of the potential
for global warming per unit mass relative to CO2 (see Table 12.4).
12.4.3 Si or Poly-Si Etching

Si and poly-Si material are etched by halogen and halogen compounds. A
fluorine atom can be used to etch the material without the assistance of ion
impact, and isotropic etching is naturally realized. F atoms incident on the
clean Si surface saturate the dangling bond and insert into the Si-Si bond,
resulting in an ejection of volatile particles, SiF4 , as the major etching product
TABLE 12.4
Global Warming Potential (GWP) and
Lifetime of Gases for Plasma Etching
Feed Gas GWP (Units of 500 yr) Lifetime (yr)
CO2 1.0 1.8 × 102

CF4 8.9 × 103 5 × 104
CHF3 1 × 104 2.6 × 102
C3 F8 1.2 × 104 2.5 × 103
SF6 3.2 × 104 3.2 × 103
c-C4 F8 9100 3200
l-C4 F8 100 1.0
C3 F6 5 × 10 1.0
l-C4 F6 5 × 10 1.0

at room temperature:
4F + Si → SiF4 . (12.34)
The chemical etching probability (yield) of F atoms and Si crystals is sensitive

to the surface condition, contamination, roughness, and so on, and the value
distributes between 0.025 and 0.064. With increasing substrate temperature
(T S > 500 K), the fraction of SiF4 gradually decreases and that of SiF2 increases
rapidly. On the other hand, the etching probability (yield) of crystalline Si is
0.005 for a Cl atom at room temperature, and the dominant etching product
is SiCl4 . With increasing surface temperature, the dominant etching product
becomes SiCl2 :
4Cl + Si → SiCl4 at TS < 500 K,

2Cl + Si → SiCl2 at TS > 750 K. (12.35)
Ion energy-assisted etching of Si-crystal or poly-Si enhances anisotropic

etching in an rf plasma in chlorine or bromine compounds. Therefore, the
anisotropic etching of Si or poly-Si is performed on the wafer biased at less
than 100 V in a high-density chlorine plasma in inductively coupled plasma
(ICP) or rf-magnetron, and so on. It should be noted that the etching proba-
bility of crystalline Si by molecular F2 and Cl2 will be very small.
PROBLEM 12.4.2
The volume density of neutral etching products is simply estimated without redepo-
sition in a plasma as
1 R etch ρ Seff
Netch = , (12.36)
V R j ne + kpump
where Seff is the effective area of a substrate exposed to etching, V is the plasma
volume, R j is the rate of the destruction processes of the species in gas phase, and
kpump is the pumping speed. Derive the expression 12.36.
Plasma etching is used for fine-pattern transfer in ULSI processing. The

target of the present technology is the fabrication of sub-45 nm rule or less.
The validity of the model depends directly on knowledge of both the gas-
phase collision processes and the physics/chemistry of the surface exposed
to the plasma. Figure 12.6 shows the etching yield Y(ε) of Si as a function of
incident ion energy [12]. Note here that in a typical plasma etching, the Cl
flux is larger than the ion flux by a factor of 100. Attention should be paid to
the difference of the yield between the nonreactive Ar+ ion and reactive Cl+ 2
ion. As described in Equation 12.12, the magnitude of the physical etching is
the direct result of the mass ratio, Mp /Ms . Judging from the fact that the mass
ratio of Cl+ +
2 in Cl2 plasma is similar to that of Ar in Ar plasma a predominant
element comes from the chemical reactivity, that is, chemical etching, of the
neutral radical species, Cl. In fact, this is confirmed from the values of the yield
of Cl+ +
2 /Cl2 as compared with that in Ar /Ar in Figure 12.6. Active fluxes of

(a)
6 +
4 Cl2 /Cl2
Yield (Si atoms/ion)

2
1 Ar+/Cl2
8
6
4
Ar+/ Ar
2
0.1
8
6
4
0 500 1000
(b)
2.0 +
CF3
Etching Yield (SiO2 /ion)
1.5 +
CF2
1.0 +
CF
Ar+
0.5
F+
0.0
0 500 1000 1500 2000
Ion energy (eV)
FIGURE 12.6
Etching yield of Si (a) and SiO2 (b) as a function of ions from plasmas.
ions and neutral radicals incident on the surface and passive ejected flux of
neutral etch products can be, in principle, extracted from a self-consistent
modeling of the total system of the dry etching. However, it may be difficult
to perform the modeling due to the lack of an available database.
12.4.4 Al Etching
Al etching is generally performed by chlorine and the compounds. However,
fluorine compound is not valid for this purpose, because the reactive etching
product has a high boiling point and is usually nonvolatile (see Table 12.3).
The surface etching is described by the direct reaction between Cl2 and Al:
3Cl2 + 2Al → 2AlCl3 . (12.37)

As a result, the etching speed depends on the gas number density of Cl2 . On
the other hand, in BCl3 the plasma density controls the etching speed because
the surface reaction proceeds by way of the dissociated Cl2 in BCl3 plasma:
e(plasma) + BCl3 → (Cl, Cl2 , BCl, BCl2 ) + e

3Cl2 + 2Al → 2AlCl3 . (12.38)
That is, the etching speed in rf plasma in Cl2 is independent of the dissipated
power, and the speed is proportional to the power in BCl3 plasma. This means
that ion bombardment from the plasma has no effect on the etching rate, and
isotropic etching is performed. The native surface oxide is removed by the
ion impact in BCl3 plasma.
12.4.5 SiO2 Etching

SiO2 etching for electrical contact holes or trenches in ULSI circuits is usually
performed by a CCP. This process requires a high-energy ion-assisted etching
with several hundred eV or 1 keV in order to maintain a high etching rate and
high selectivity of SiO2 to the Si surface. For this purpose, a two frequency
CCP with a different frequency source, functionally separated for sustain-
ing a high-density plasma and for biasing the wafer, at each of two parallel
electrodes has been adopted by a combination of VHF and low-frequency
sources.
SiO2 thin film is generally etched by an rf plasma in a fluorocarbon C j Fk
gas system. In SiO2 the characteristics of a fluorocarbon polymer deposition
as well as the high-energy ion-assisted etching determine the feature pro-
file of the hole or trench. The control of the profile with a high aspect ratio
is performed by the sidewall passivation film of the polymer. The selective
etching of SiO2 over Si and photoresist is due to the selective formation of
protective fluorocarbon polymer film over the Si surface. In a steady plasma
etching, a mixed amorphous interfacial layer, Sil Cm Fn , is formed on SiO2 un-
der impact of energetic CF+ 3 ions. In actual practice, the apparent etching
yield of SiO2 in the continuous plasma irradiation is given for the value of a
mixed interfacial layer rather than for the pure SiO2 surface. Carbon in the
form of CFi radicals and CF+j ions from fluorocarbon plasma reacts with the
oxygen on the SiO2 surface under energetic ion impact to form volatile prod-
ucts, whereas fluorocarbon is deposited on the sidewall to form a protective
layer of Ci F j polymers that inhibit lateral etching. This leads to “anisotropic
etching.” Low-k materials with a relative permittivity r smaller than 3.8 of
SiO2 are used as the dielectric in a multilayer interconnect system in ULSI.
CCP maintained in admixture of H2 and N2 is used to etch organic low-k
materials. The etching yield is shown in Figure 12.7.
PROBLEM 12.4.3
Discuss the effects of additive gases (Ar, O2 , CO, H2 ) on SiO2 etching in fluorocarbon
plasma (see Table 12.5).

2.0
N2 /H2 plasma
1.5
Sputtering yield (C / ion)
N2 plasma
1.0
Ar plasma
0.5
H2 plasma
0.0
0 400 800 1200
Ion energy (eV)
FIGURE 12.7
Etching yield of organic low-k as a function of ions from plasmas.
12.4.6 Feature Profile Evolution

The space and time evolution of a patterned wafer surface exposed to par-
ticles of positive ions and neutral radicals from a plasma as shown in terms
of the velocity distribution of ions in Figure 12.8 is estimated by the Level
Set method based on the Hamilton–Jacobi-type equation under a moving
boundary [13] (see Chapter 8):
∂
(x, z, t) = R etch (x, z, t)|∇|. (12.39)
∂t
Here, R etch is the etching rate (speed function) of a material surface. The sur-
face as a function of Cartesian coordinates (x, z) and time t in Figure 12.9 is
TABLE 12.5
Function of Additive Gas in SiO2 Etching in Fluorocarbon
Plasma
Gas Property
O2 Decelerates polymerization through the production

of volatile CO, CO2 , COF2
H2 Scavenges [F] to form volatile HF, increasing the [C]/[F] ratio,
and shifting the chemistry from an etching to a polymerization
CO Tunes gas to increase the selectivity in SiO2 /Si
Increase of relative CF
Ar Serves as a buffer gas for control of the dissociation of C j Fk

1.0
0.0
0
p
+20
0
1500 -20
En
erg
y(
eV (deg)
) Angle
FIGURE 12.8
Example of the energy and angular distributions of ions incident on a wafer as a function of
radial position.
z
ion trajectory
Φ=0 Mp
(i+1, j) (1-r) r (i, j)
Φ<0 Φ>0
l rp-1 s
r0
n (1-s)
rp m Gas
(i+1, j+1) (i, j+1)
Solid
FIGURE 12.9
Schematic diagram of the Level Set method and the surface evolution.

defined as
(x, z, t) = 0. (12.40)
That is, the gas and solid phase are given at > 0 and < 0 at t, respec-
tively. The dielectric surface (r, t) = 0 is usually locally charged by ions and
electrons incident on the surface during etching, and the metallic surface is
kept at equipotential.
A brief description of the Level Set method is given below. First, we in-
vestigate the relation between the time constant τ0 for one monolayer etching
and the time step to trace an ion trajectory ttr in order to estimate a surface
evolution t under the following physical requirement:
ttr τ0 < t.
Also t is adjusted to satisfy the condition that a renewal of the surface

during t does not exceed the spatial mesh size, x and z; that is, R etch t
x(and z).
Step 1: At a position rintl unaffected by a topographically and locally
charged wafer surface, the velocity distribution g p(e) (v, rintl ) and the density
n p(e) (rintl ) of positive ions and electrons are given from the plasma structure in
the reactor simulated, in advance, as the initial conditions of ions and electrons
incident on the wafer surface (see Figure 12.9). Then, the ion (and electron)
trajectory incident in a trench from rintl is traced by Monte Carlo simulation
with time step ttr by using Poisson’s equation and Newton’s equation under
the consideration of collisions with the gas molecule and on the surface. At
each position r p (ttr ) of the ion trajectory, (r p ) is interpolated from i, j at
four grid nodes surrounding r p as (see Figure 12.9)
p (r p ) = (1 − r )(1 − s)i, j + (1 − r )si, j+1

+ r (1 − s)i+1, j + r si+1, j+1 , (12.41)
where r = (x p − xi )/(xi+1 − xi ) and s = (z p − zi )/(zi+1 − zi ). When p (r p ) > 0

is satisfied, the trace of the ion flight is continued at ttr + ttr . Otherwise, if
p (r p ) < 0 is satisfied, go to Step 2.
Step 2: We estimate the point r0 that intersects with the surface, (r0 ) = 0,
from (r p ) and the value one step before (r p−1 ) at ttr − ttr as

p x p−1 − p−1 x p p yp−1 − p−1 yp
r0 (x0 , y0 ) = , . (12.42)
p − p−1 p − p−1
When the ion with energy ε p is reflected at the material surface having a
reflection coefficient αref (ε p ), that is, when the condition
ξ ≤ αref (ε p )
is satisfied, the trace of the ion trajectory in gas phase is continued at Step 1.
Here, ξ is a uniform random number distributed over [0, 1].

In the case of ξ > αref (ε p ), the information of the flux velocity Γ(v0 , r0 ) at
the surface is transcribed into the adjacent two-grid nodes (see Figure 12.9):
1/2
−1
s 2 + (1 − r )2
Γ(v)i+1, j = 1 + Γ(v0 , r0 ),
r 2 + (1 − s)2
1/2
−1
r 2 + (1 − s)2
Γ(v)i, j+1 = 1 + Γ(v0 , r0 ). (12.43)
s 2 + (1 − r )2
Then, the incident angle of the transcribed flux vector θ with respect to the
normal vector n = ∇/|∇| at both nodes is obtained by (r) and the deriva-
tive of the nodes:
∂
−1 ∂x
· vx + ∂
∂y
· vy
θ = cos . (12.44)

∇φ
·
v
Then we have the component of the etching yield at the node as a function
of the ion energy ε p and incident angle θ as Y(ε p , θ). The above procedure
is carried out for the number of ions at initial position rintl by considering
g(v, rinitl ) and n p (rintl ). Then, the flux velocity Γ(v, r) at each node adjacent to
the surface = 0 is accumulated during t.
Step 3: After tracing the ion trajectories incident on a trench structure from
a plasma during t = m ttr (m ∼ 100), the accumulated flux velocity (ε p , r),
energy ε p , and angle θ at each of the grid nodes adjacent to the surface = 0
enables us to estimate the new surface = 0. That is, at each of the nodes,
the etching rate (speed function) is given as

R etch (r) = n pintl vg p (v, r)Y(ε p , θ)dε p dθ/ vg p (v, r)dε p

= Γ(ε p , r)Y(ε p , θ)d ε p d θ. (12.45)
As a result, the new value of (r, t) is obtained at each node from

Equation 12.39
t+t
(r, t + t) = R etch (r, t)|∇(r, t)|dt. (12.46)
t
Step 4: By using the information of the renewed surface ((r, t) = 0),

at each grid surrounding the surface (r, t) is redefined. At the same time,
the local charge distribution on the new surface is also renewed by using the
previous ones. Successive simulation of the etching is performed by repeating
the loop from Step 1 to Step 4.
The same procedure is adopted for chemical etching by neutral radicals
incident on the surface. In general, the neutral radicals have a Maxwellian
velocity distribution with a temperature of Tg . Plasma etching is a more-or-
less competitive process between deposition and etching. As described in
Section 12.3, neutral radicals are the predominant species for deposition, and

FIGURE 12.10
A typical example of feature profile evolution by plasma etching: (a) without charging and
deposition, (b) with charging and without deposition, and (c) with charging and deposition.
etching is achieved by positive ion impact in addition to the chemical etching

of neutral radicals. In particular, the surface reaction of neutral species is a
function of surface temperature as predicted by Arrhenius’ equation. It is pos-
sible to carry out the simulation of the competitive process by considering a
surface evolution by neutral radicals and ions using the Level Set method
under the database of Y(ε p , θ) of ions and St (θ ) of radicals. Figure 12.10
exhibits self-consistent results of a feature profile evolution based on plasma
reactor-scale and feature-scale simulation.
PROBLEM 12.4.4
Discuss the string model of the feature profile evolution in etching in comparison
with the Level Set method.
12.4.7 Plasma Bosch Process

Plasma etching of a large hole/trench with high aspect ratio (depth-to-width)
in Si is a basic process to fabricate integrated components for micro-
electromechanical systems (MEMS) [14]. Holes or trenches in MEMS range
from 1 µm to 100 µm in width and from 10 µm to several 100 µm in aspect

ratio. The etching needs a high etch rate with anisotropy and selectivity to the
mask material. Fluorocarbon plasma, maintained in CF4 , C4 F8 , SF6 , and so
on, in admixture with O2 are used for the processes under the basic reactions
that F isotropically etches Si wafer with rapid chemical etching, and a fluo-
rinated silicon oxide passivating layer is formed on the sidewall to keep the
anisotropic etching.
A series of alternating processes of etching and depositing is effective in
forming MEMS construction. The time-multiplexed deep etching is known
as the Bosch process. One of the effective and rapid plasma processes consists
of three steps: isotropic etching of Si/polymer in SF6 , polymer passivation in
C4 F8 , and polymer depassivation in O2 . The resulting feature profile keeps a
high anisotropy, though the etching is chemically isotropic. The Plasma Bosch
process is widely carried out by using a high-density ICP reactor at pressure
between 1.33 Pa (10 mTorr) and 13.3 Pa (100 mTorr). The dimension of the
MEMS structure is usually comparable with the sheath thickness or larger.
It means that the bulk plasma and sheath structure exposed to the substrate
will change in time (plasma molding).
Exercise 12.4.1
In a plasma for deposition or etching, a large size polyatomic molecule may
grow up in circumstances of high degree of dissociation of the feed gas. The
generated species is a named particle, dust, or powder [15]. Discuss the tool
for detection of these large polyatomic molecules in a plasma.
The plasma potential is always a positive value during the cw operation.
Therefore, positive ions and neutrals are detected through an orifice on the
electrode or reactor wall. However, it is not easy to detect negative ions from
the electrode. Technically, a pulsed operation of the plasma is introduced to
measure negative ions in the afterglow phase. Mie scattering is an in situ,
active procedure (see Table 12.6).
12.4.8 Charging Damage

We have basically two types of materials to be etched in ULSI. One is metal
or poly-Si with adequate electric conductivity. The other is dielectric SiO2
or low-k materials. These are known as Si-gate etching and the contact hole
(via or trench) etching in SiO2 , which is affected by an anomalous etching
caused by a local charging of the surface exposed to a plasma as shown in
Figure 12.11. The surface resistivity changes with the flux composition from
the plasma, that is, ions, electrons, and neutral radical species.
TABLE 12.6
Detection of Polyatomic Molecule or Cluster in Plasmas
Method Molecular Size Mass Resolution Ref.
Quadrupole mass spectrometer M/e < 103 — [16, 17]

Time-of-flight mass spectrometer M/e < 105 5×103 [18]
Laser Mie scattering d0 > µm — [19]

Plasma Processing and Related Topics
(a) (b) f e (ε,θ) f p (ε,θ)
Sheath
E
f e (ε,θ) f p (ε,θ)
Sheath
E
M+
Wafer e
e
surface Wafer
Open area surface Resist
Resist
M+
SiO2
Poly-Si Tunneling
current
SiO 2
Metal
Damage
Si current SiO2
Damage
Si current
Cb
Cb
FIGURE 12.11
325
Schematic diagram of the local charging damage during plasma etching: gate-Si etching (a) and SiO2 trench etching (b).

plasma
quasi-conductive
layer
Γe
Γe
jesurf
Γp
SiO2 jesurf
charge
Si
FIGURE 12.12
Local surface charges and conduction in the trench (hole) of SiO2 during etching.
12.4.8.1 Surface Continuity and Conductivity

A number of experiments have emphasized that the etching of a high aspect
ratio structure of SiO2 is a highly competitive process between etching of SiO2
and deposition of a thin CFx polymer, and that a thin CFx polymer is deposited
on the active surface exposed to photons, neutral radicals, electrons, and ions.
These facts imply that electrons may be conducted on the surface of the thin
polymer layer (passivation film) under a field distribution due to a local accu-
mulation of electrons and ions and photo-irradiation from the plasma. Then,
the polymer film would prevent the wall from charging by a recombination
process of the accumulated positive ions through a fast electron transport.
The surface conduction in conjunction with the incident local fluxes of pos-
itive ions and electrons is described by a simple surface continuity equation
(see Figure 12.12):

∂ne (r, t)
+ divΓe (r, t) = −div[je sur f (r, t)/e] − R r ne n p , (12.47)
∂t sur f

∂n p
+ divΓ p (r, t) = −R r ne n p , (12.48)
∂t sur f

where Γe and Γ p are the instantaneous local fluxes of electrons and positive
ions incident on the surface, respectively. je sur f (r, t) = σ (r, t)Esur f (r, t) is a sur-
face current density of the electron conduction, σ is the electrical conductivity
of the CFx thin polymer, and R r is the surface recombination coefficient. The
time evolution of the local surface potential Vsur f (r, t) is solved at t in the sys-
tem including the continuity Equations 12.47 and 12.48 and Poisson’s equa-
tion under consideration of the incident fluxes of electrons and ions from the
plasma through the sheath as described in the previous section.
The resulting potential distribution across the gas, surface, and solid phase
is calculated by solving
ρs (r, t)
∇ 2 (r, t) = − : surface of SiO2
ε0 εr
= 0 : inside of SiO2 (12.49)
n p (r, t) − ne (r, t)
= −e : in gases,
ε0
where ρs is the surface charge density on SiO2 , and ε0 and ε0 εr are the
permittivities of a vacuum and SiO2 , respectively. That is, the overall po-
tential from the gas phase to the bulk SiO2 is simultaneously solved by
changing the mesh size in gas and solid. Then, the distribution of the sur-
face charge at the boundary between SiO2 and lower poly-Si (or metal) in
Figure 12.12 is iteratively calculated under the principle that the equipo-
tential of the surface of lower poly-Si (or metal) must be maintained. The
origin of the local charging of electrons and positive ions arises from the
significantly different velocity distribution. In a steady state, the velocity dis-
tribution of ions incident on a wafer with a beamlike component is quite
different from an isotropic distribution of electrons in the positive ion sheath,
and both of the charged fluxes incident on a flat surface exposed to the
plasma have the same magnitude in a time-averaged fashion. The great dif-
ference of the velocity distribution leads to a local accumulation of electrons
at the upper part of the trench (hole) and ions at the lower part and bottom,
that is, charging on the inside wall of a hole or trench with a high aspect
ratio.
Figure 12.11 demonstrates the schematic diagram of the charging damage
in the gate-Si etching by Cl2 plasma (a) and SiO2 trench etching by CF4 /Ar
plasma (b). Both of the chargings seriously damage the profile and lower-level
device elements (i.e., the thin gate etc.).
Exercise 12.4.2
Discuss the relationship among the time constants of charging, radical depo-
sition, and ion etching in a typical SiO2 etching (see Figure 12.13.)
Typical fluxes of radicals and ions incident on a wafer are 1018 cm−2 s−1 and
1016 cm−2 s−1 , respectively. Also, the time constant for local charging is usually
ms and the time for effective monolayer etching needs 100 ms.

1.6
AR = 2.0
Incident flux (1016 cm-2 s-1)

1.2
5.0
0.8
10.0
0.4
ion flux
electron flux
0.0
0.0 1.0 2.0 3.0 4.0 5.0
Time (ms)
FIGURE 12.13
Time constant for local charging as a function of aspect ratio in typical trench etching of SiO2 .
12.4.8.2 Charging Damage to Lower Thin Elements in ULSI

The current damage through thin gate oxide is typical of a high-density
plasma processing (see Figure 12.11). The current densities of the direct tunnel
and Fowler–Nordheim tunnel, J DT and J FNT , are expressed, respectively [20],
as

e2 Vox Vox 1
J DT = 2
φB − exp −4π dox 2em∗DT φ B − , (12.50)
2π hdox 2 2 h
 
4π 2m∗FNT φ B
3/2
e 3 Vox2
J FNT = 2 φ
exp− , (12.51)
8π hdox B 3π eh
where dox is the thickness of the oxide film, and Vox is the applied voltage
between both sides of the thin oxide. h is the Planck constant, φ B is the barrier
height, and m∗ is the effective mass of the electron. Plasma current damage
is typical in a metal gate etching as shown in Figure 12.11a, and besides, in
a multilayer interconnect system we take care of the current damage during
trench or via etching of SiO2 (see Figure 12.14).
12.4.9 Thermal Damage

High-energy ions are indispensable for the etching of SiO2 . There exists a high-
energy deposition at a very short time within narrow spots near the material
suface. The ion irradiation induces a rapid local heating and gradual cooling

Γe Γp
resist
V
SiO2
metal
I
poly-Si
gate oxide LOCOS
0 5 ms 10 ms
Si
FIGURE 12.14
Current damage at the gate oxide during plasma etching.
with high-temperature gradients, leading to thermal damage such as thermal

stress. The thermal damage appears on the lower-level device elements, for
example, gate oxide in addition to the etching surface. In the case of the elastic
binary collision at the surface, the fractional energy of the incident ion with ε p ,
4Mp Ms
ε = cos θ 2 ε p , (12.52)
(Mp + Ms )2
is transferred into an atom in the surface layer of the wafer. Here, θ is the in-
cident angle of the ion to the wafer. Except for the case of Mp = Ms and
θ = 0, some of the energy of the incident ion is dissipated in the form
of excitation of the substrate atoms. After a short relaxation time of the
highly nonequilibrium local state, a high-temperature local spot with T(r0 , t)
diffuses thermally to the circumference. The thermal diffusion except for the
highly nonequilibrium state during a very short time is described by

∂ T(r, t) ∂T
ρc(T) = ∇r k(T) + S(r0 , t0 ) (12.53)
∂t ∂r
where ρ, c(T), and k(T) are the mass density, the heat capacity, and the heat
conductivity of the substrate material, respectively. S is the volume power
density of the heat source, that is, the fractional ion energy dissipated to the
surface.
In a typical feature-scale etching by using the 2f-CCP reactor, the positive
ion flux to the wafer has the magnitude on the order of 1016 cm−2 s−1 . It
corresponds to the one ion impact every 1 µs. Under these conditions, the
time constant of thermal diffusion is on the order of 0.1 µs to 1 µs. That is, in
the micro- or nanometer scale etching, the etching by the aid of ion impact

is a discontinuous process in time subject to the stochastic process. Also note

that when the scale of the material to be etched approaches nanometer scale,
we should pay attention to the change of the physical quantity, c(T) and k(T)
and so on.
References
1. Behrisch, R., Ed. 1981. Sputtering by Particle Bombardment I. Topics in Appl. Phys.
47. Berlin: Springer Verlag.
2. Behrisch, R., Ed. 1983. Sputtering by Particle Bombardment II. Topics in Appl. Phys.
52. Berlin: Springer Verlag.
3. Kress, J.D., Hansen, D.E., Voter, A.F., Liu, C.L., Liu, X.-Y., and Coronell, D.G.
1999. J. Vac. Sci. Technol. A 17:2819.
4. Engelmark, F., Westlinder, J., Nyberg, T., and Berg, S. 2003. J. Vac. Sci. Technol. A
21:1981.
5. Thompson, M.W. 1968. Philos. Mag. 18:377.
6. Kenmotsu, T., Yamamura, Y., and Ono, T. 2004. Jpn. Soc. Plasma. Sci. Nucl. Fusion
Res.
7. Matsuda, A. 2004. Thin-film silicon (Inv. Review paper), Jpn. J. Appl. Phys.
43:7909.
8. Matsumura, H., Umemoto, H., and Masuda, A. 2004. J. Non-Cryst. Solids
19:338–340.
9. van Kampen, N.G. 1981. Stochastic Processes in Physics and Chemistry. Amsterdam:
North-Holland.
10. Gardiner, C.W. 1983. Handbook of Stochastic Methods for Physics, Chemistry and the
Natural Sciences. Berlin: Springer Verlag.
11. Ford, I.J. 1995. J. Appl. Phys. 78:510.
12. Balooch, M., Moalem, M., Wang, W-E., and Hamza, A.V. 1996. J. Vac. Sci.
Technol. A 14:229.
13. Sethian, J.A. and Strain, J. 1992, 1993. J. Comp. Phys. 98: 231; Sethian, J.A.
and Chopp, D.L. 1993, J. Comp. Phys. 106:77.
14. Esashi, M., and Ono, T. 2005. J. Phys. D, Topical Review 38:R223.
15. Bouchoule, A., Ed. 1999. Dusty Plasmas. Chichester: John Wiley & Sons.
16. Auciello, O. and Flamm, D.L. 1989. Plasma Diagnostics, Vol. 1. San Diego:
Academic Press.
17. Bruno, G., Capezzuto, P., and Madan, A. 1995. Plasma Diagnostics of Amorphous
Silicon-Based Materials. San Diego: Academic Press.
18. Saito, N., Koyama, K., and Tanimoto, M. 2003. Jpn. J. Appl. Phys. 42:Part 1, 5306.
19. Bohren, C.F., and Huffman, D.R. 1983. Absorption and Scattering of Light by Small
Particles. New York: John Wiley & Sons.
20. Hirose, M. 1996. Mater. Sci. Eng. 41:35.

Plasma Electronics Applications in Microelectronic Device Fabrication

Uploaded by

Copyright:

Available Formats

Plasma Electronics Applications in Microelectronic Device Fabrication

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Plasma Electronics Applications in Microelectronic Device Fabrication

Uploaded by

Copyright:

Available Formats

Plasma Electronics:

© 2006 by Taylor & Francis Group, LLC

Other books in the series:

Aspects of Anomalous Transport in Plasmas

Non-Equilibrium Air Plasmas at Atmospheric Pressure

Magnetohydrodynamic Waves in Geospace: The Theory of ULF Waves and their

Plasma Physics via Computer Simulation (paperback edition)

Plasma Waves, Second Edition

Microscopic Dynamics of Plasmas and Chaos

Plasma and Fluid Turbulence: Theory and Modelling

The Interaction of High-Power Lasers with Plasmas

Introduction to Dusty Plasma Physics

The Theory of Photon Acceleration

Laser Aided Diagnostics of Plasmas and Gases

Reaction-Diffusion Problems in the Physics of Hot Plasmas

The Plasma Boundary of Magnetic Fusion Devices

Non-Linear Instabilities in Plasmas and Hydrodynamics

© 2006 by Taylor & Francis Group, LLC

New York London

© 2006 by Taylor & Francis Group, LLC

© 2006 by Taylor & Francis Group, LLC

Library of Congress Cataloging-in-Publication Data

Visit the Taylor & Francis Web site at

© 2006 by Taylor & Francis Group, LLC

© 2006 by Taylor & Francis Group, LLC

Toshiaki Makabe and Zoran Lj. Petrovic

© 2006 by Taylor & Francis Group, LLC

2 Phenomenological Description of the Charged

3 Macroscopic Plasma Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

© 2006 by Taylor & Francis Group, LLC

© 2006 by Taylor & Francis Group, LLC

5 The Boltzmann Equation and Transport Equations of

© 2006 by Taylor & Francis Group, LLC

6 General Properties of Charged Particle Transport in Gases . . . . . . . . . 175

7 Modeling of Nonequilibrium (Low-Temperature) Plasmas. . . . . . . . .205

© 2006 by Taylor & Francis Group, LLC

8 Numerical Procedure of Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

9 Capacitively Coupled Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255

© 2006 by Taylor & Francis Group, LLC

10 Inductively Coupled Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

11 Magnetically Enhanced Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

12 Plasma Processing and Related Topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301

© 2006 by Taylor & Francis Group, LLC

© 2006 by Taylor & Francis Group, LLC

1.1 Plasma and Its Classification

© 2006 by Taylor & Francis Group, LLC

(a) (b) (c)

1.2 Application of Low-Temperature Plasma

1011 etching ashing

© 2006 by Taylor & Francis Group, LLC

an ultra-large-scale integrated circuit demonstrated partly in Figure 1.3 are

1.3 Academic Fusion

© 2006 by Taylor & Francis Group, LLC

© 2006 by Taylor & Francis Group, LLC

In this chapter we attempt to lay out the foundation of the phenomenol-

2.1 Transport in Real (Configuration) Space

© 2006 by Taylor & Francis Group, LLC

2.1.1 Momentum Balance of Electrons

where D is the diffusion coefficient.

© 2006 by Taylor & Francis Group, LLC