Introduction to

Colloidal (Dusty) Plasmas

Lecture Notes

André Melzer

Institut für Physik

Ernst-Moritz-Arndt-Universität Greifswald

Version of: October 26, 2016

(2nd Edition)
ii CONTENTS

Contents
1. Introduction 1

2. Charging of Dust Particles 6

2.1. Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2. OML Charging Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1. Derivation of the OML Currents . . . . . . . . . . . . . . . . . . . . 7
2.2.2. Discussion of the OML Model . . . . . . . . . . . . . . . . . . . . . 11
2.2.3. Streaming Ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.4. Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3. Other Charging Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.1. Photoelectron Emission . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.2. Secondary Electron Emission . . . . . . . . . . . . . . . . . . . . . 14
2.4. The Particle as a Floating Probe . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.1. OML Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.2. The Capacitance Model . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.3. Floating Potential with Streaming Ions . . . . . . . . . . . . . . . . 17
2.4.4. Floating Potential with Collisions . . . . . . . . . . . . . . . . . . . 18
2.4.5. Floating Potential with Secondary Electron Emission . . . . . . . . 19
2.5. Temporal Evolution of the Particle Charge . . . . . . . . . . . . . . . . . . 21
2.5.1. Charging in the rf Sheath . . . . . . . . . . . . . . . . . . . . . . . 22
2.5.2. Stochastic Fluctuations . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6. Influence of Many Particles and Electron Depletion . . . . . . . . . . . . . 24

3. Forces and Trapping of Dust Particles 27

3.1. Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2. Electric Field Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.1. Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.2. Polarization Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3. Ion Drag Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3.1. Qualitative Discussion: Barnes model . . . . . . . . . . . . . . . . . 31
3.3.2. Quantitative Discussion: Hutchinson/Khrapak model . . . . . . . . 33
3.4. Neutral Drag Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5. Thermophoresis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.6. Laser Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.7. Compilation of Forces, Dust Levitation and Trapping . . . . . . . . . . . . 38
3.8. Vertical Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.8.1. Linear Resonance and Charge Measurement . . . . . . . . . . . . . 41
3.8.2. Parametric Resonances . . . . . . . . . . . . . . . . . . . . . . . . . 45
CONTENTS iii

3.8.3. Nonlinear Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4. Coulomb Crystallization 51
4.1. The One-Component Plasma (OCP) . . . . . . . . . . . . . . . . . . . . . 51
4.2. Yukawa Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3. Coulomb Crystallization in Dusty Plasmas . . . . . . . . . . . . . . . . . . 54
4.4. Crystallization in Bounded Systems . . . . . . . . . . . . . . . . . . . . . . 55
4.5. Structural Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.5.1. Wigner-Seitz Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.5.2. Pair Correlation Function . . . . . . . . . . . . . . . . . . . . . . . 57
4.5.3. Structure Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.5.4. Angular Correlation Function . . . . . . . . . . . . . . . . . . . . . 59
4.5.5. Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6. 3D Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

5. Dust Particle Interaction 63

5.1. Interaction in the Horizontal Plane . . . . . . . . . . . . . . . . . . . . . . 63
5.2. Vertical Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2.1. The wakefield potential (wave model) . . . . . . . . . . . . . . . . . 65
5.2.2. The ion focus (particle model) . . . . . . . . . . . . . . . . . . . . . 68
5.2.3. Non-reciprocal attraction . . . . . . . . . . . . . . . . . . . . . . . . 70
5.3. Measurement of the Attractive Force . . . . . . . . . . . . . . . . . . . . . 72
5.4. Oscillatory Instability of the Vertical Alignment . . . . . . . . . . . . . . . 75
5.4.1. Solution of the Coupled Oscillations . . . . . . . . . . . . . . . . . . 76
5.4.2. General Analysis of the Instability . . . . . . . . . . . . . . . . . . 78
5.4.3. Conditions for the Instability . . . . . . . . . . . . . . . . . . . . . 79
5.5. The Phase Transition of the Plasma Crystal: Experiment . . . . . . . . . . 81
5.6. The Phase Transition of the Plasma Crystal: Simulations and Theory . . . 84

6. Waves in Weakly Coupled Dusty Plasmas 87

6.1. Dust-Acoustic Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.2. Ion-Flow Driven Dust Acoustic Waves . . . . . . . . . . . . . . . . . . . . 92
6.3. Dust Ion-Acoustic Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.4. Other Wave Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

7. Waves in Strongly Coupled Dusty Plasmas 100

7.1. Compressional Mode in 1D . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.2. Dust Lattice Waves in 2D . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7.3. Compressional 1D Dust Lattice Waves: Experiment . . . . . . . . . . . . . 107
7.4. Compressional 2D Dust Lattice Waves: Experiment . . . . . . . . . . . . . 110
7.5. Shear 2D Dust Lattice Waves: Experiment . . . . . . . . . . . . . . . . . . 111
7.6. Mach Cones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
iv CONTENTS

7.7. Transverse Dust Lattice Waves . . . . . . . . . . . . . . . . . . . . . . . . 114

7.8. Dispersion Relation from Thermal Particle Motion . . . . . . . . . . . . . . 116
7.9. A Note on Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

8. Finite Dust Clusters 120

8.1. Formation of Finite Dust Clusters . . . . . . . . . . . . . . . . . . . . . . . 120
8.2. Structural Transitions in 1D Dust Clusters . . . . . . . . . . . . . . . . . . 121
8.3. Structure of 2D Finite Dust Clusters . . . . . . . . . . . . . . . . . . . . . 122
8.3.1. Example of 2D Cluster Configurations: Configuration (0,6) . . . . . 123
8.3.2. Example of 2D Cluster Configurations: Configuration (1,5) . . . . . 127
8.4. Structure of 3D Finite Dust Clusters . . . . . . . . . . . . . . . . . . . . . 128
8.5. Normal Modes in 2D Finite Clusters . . . . . . . . . . . . . . . . . . . . . 130
8.6. Modes from Thermal Particle Motion . . . . . . . . . . . . . . . . . . . . . 133
8.7. Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
8.8. Phase Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

9. Technical Applications of Dusty Plasmas 140

9.1. Particle Growth Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . 140
9.1.1. Cluster Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.1.2. Agglomeration Phase . . . . . . . . . . . . . . . . . . . . . . . . . . 144
9.1.3. Accretion Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
9.2. Technological Impacts of Dusty Plasmas . . . . . . . . . . . . . . . . . . . 146
9.2.1. Surface Contamination by Dust . . . . . . . . . . . . . . . . . . . . 147
9.2.2. Flue Gas Cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
9.2.3. Dust in Fusion Devices . . . . . . . . . . . . . . . . . . . . . . . . . 149
9.2.4. Formation of New Materials . . . . . . . . . . . . . . . . . . . . . . 150
9.2.5. Polymorphous Silicon Films . . . . . . . . . . . . . . . . . . . . . . 151

10.Astrophysical Dusty Plasmas 154

10.1. Noctilucent Clouds and PMSE . . . . . . . . . . . . . . . . . . . . . . . . . 154
10.2. Dust Streams From Jupiter . . . . . . . . . . . . . . . . . . . . . . . . . . 156
10.2.1. Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
10.2.2. Model and Experiment . . . . . . . . . . . . . . . . . . . . . . . . . 158
10.3. Dust Orbits at Saturn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
10.4. Spokes in Saturn’s Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
10.5. Mach Cones in Saturn’s Rings . . . . . . . . . . . . . . . . . . . . . . . . . 164

11.Summary 166
CONTENTS v

Preface to the Second Edition

The field of dusty plasmas has continued to grow over the last years. In the light of new
developments I felt that some parts of this Introduction needed to be revised.

While I still tried to stick to the basic concepts of dusty plasmas I tried to mention
some recent progress. Some of the questions declared as open in the first edition have
been settled, some new questions have arisen.

Changes include, to name a few, the role of collisions in the charging, the state of
understanding of the ion drag force, driven dust-acoustic waves, three-dimensional dust
systems and many more. Further many figures have been revised (not only changed to
color).

Suggestions, comments and improvements are still welcome.

Enjoy!
vi CONTENTS

Preface to the First Edition

This manuscript has emerged from lectures held by the author at the Ernst-Moritz-Arndt-
University Greifswald and the Christian-Albrechts-University Kiel in the years 2002 to
2005. It covers the topics in the field of dusty plasmas for a course of 2 to 3 semester hours.

It is intended as an introductory course into the novel and fascinating field of

complex plasmas. The reader should have a background of general plasma physics on
an introductory level. Dusty plasmas shed a new and different light on various aspects
of plasma physics like the problem of shielding or the mechanism of acoustic waves in
plasmas. The study of dusty plasmas thus provides new insight into plasma physics and
substantiates, revises and strengthens its concepts. The author hopes that these lecture
notes will drive the reader’s interest into this growing and interesting field of physics.

At the University Greifswald, the manuscript can serve as a companion to the Course
“Colloidal Plasmas” which is a compulsory one-hour course for students who have
specialized into plasma physics and to the Course “Colloidal Plasmas II”, a voluntary
two-hour course for students interested in dusty plasmas.

I would like to thank Yuriy Ivanov, Sebastian Käding and Matthias Wolter for
proofreading. Additional errors that I left in the manuscript can be reported to me under
[email protected]. Suggestions, comments and improvements are welcome.

Enjoy!
 1

1. Introduction
Colloidal (or dusty or complex) plasmas are a new and fascinating field of plasma physics.
Colloidal plasmas enable to study basic plasma properties on an “atomic” kinetic level and
the allow to visualize collective plasma phenomena, like oscillations and waves. Moreover,
a vast number of novel phenomena are found in these systems. New features in colloidal
plasmas range from Coulomb crystallization to new types of forces and waves. In these
lecture notes, a general introduction to this active and growing field will be given. Areas
of special interest are covered in a few monographs [1, 2, 3, 4].
Dusty plasmas share a number of physical concepts and similarities with non-neutral
plasmas, like pure ion plasmas in Paul or Penning traps, as well as with colloidal suspen-
sions, where charged plastic particles are immersed in an aqueous solution. In analogy to
these systems, the terms “colloidal plasmas” or “complex plasmas” are frequently used
for dusty plasmas more or less as synonyms. Here, we will mainly use “colloidal plas-
mas” for strongly coupled dust systems with micrometer sized particles. When referring
to particles in astrophysical or technological plasmas, the term “dusty plasmas” will be
used.
Colloidal and dusty plasmas consist of macroscopic solid particles immersed in gaseous
plasma environment of electrons, ions and neutrals. The particles are charged by the inflow
of electrons and ions or by other means. Thus, the dust particles act as an additional
plasma species. Hence, the study of the plasma-particle and particle-particle interaction,
their fundamental properties and their collective effects open up new and interesting
visions on plasma physics.
Dusty plasmas are ubiquitous in astrophysical situations, like interstellar clouds, the
rings of the great planets or comet tails. In his book [1], Verheest characterizes dusty
plasmas by: “If the claim is made that more than 99 % of the observable universe is
in the plasma state then it could be jokingly asserted that the remainder is dust”. For
instance, Saturn’s B-ring consists of micron and submicron dust particles (and larger
boulders) in the plasma environment of Saturn. There, peculiar features have been ob-
served by the Voyager spacecraft (see Fig. 1.1a): Radially extended structures (“spokes”)
develop within minutes and last for hours. This behaviour cannot be explained by pure
gravitational interactions. Other examples of dusty astrophysical plasmas include comets
and interstellar clouds. Comets are “dirty (dusty) snowballs” that evaporate under the
influence of the solar wind. Comets usually form a plasma tail and the bright curved
dust tail. Cometary dust generally has a size distribution with high abundance of small
grains (0.1 µm diameter) and a smaller number of larger particles which then can interact
with the plasma tail near the comet head and the solar wind (see Fig. 1.1b). Finally,
in interstellar clouds, dust particles of 0.01 to 10 µm are found. One of the interesting
questions, here, is the influence of charged particles on star and planet formation.
2 1. Introduction

Figure 1.1: (a) Spokes in Saturn’s rings. The spokes are the radially extended dark
features. (b) Plasma and dust tail of comet Hale-Bopp (1994).

In microchip manufacturing about 70 % of the production steps require plasma pro-

cessing techniques for etching or thin film deposition. Plasma processing allows to manu-
facture multi-layer devices and fine vertical structures that help to increase packing density
(see Fig. 1.2). However, it was found in the end of the 1980’s that in reactive gases used
for chip processing also small dust particles can grow due to polymerization. The grown
particles are trapped in the plasma near the wafers (see Fig. 1.3). It is obvious that dust
particles falling on the etched structures can lead to severe damages and malfunctions of
the chips (“killer particles”). Thus, here the removal of dust particles from the discharge
is a crucial issue. Similarly, particles can also grow in fusion devices and its implication
for operational performance is currently discussed. On the other hand, the incorporation
of submicron dust particles in thin films leads to new materials, like self-lubricating films
with small embedded MoS2 particles or the manufacturing of polymorphous solar cells
with increased efficiency.
In typical laboratory experiments on fundamental aspects of colloidal plasmas
monodisperse spherical particles of 1 to 10 micrometer diameter are trapped in gas dis-
charges. The microspheres attain high negative charges of the order of 103 to 105 elemen-
tary charges due to the inflow of plasma electrons and ions. However, due to the extreme
mass the charge-to-mass ratio that determines the dynamic properties is by orders of
magnitude smaller than that of ions, not to mention electrons.
In colloidal plasmas the microspheres can be viewed individually due to their compar-
atively large size. Due to the very low charge-to-mass ratio the spatial and time scales
 3

Figure 1.2: (a) 300 mm wafer (Infineon), (b) deep silicon etching for MEMS Micro-
Electro-Mechanical Systems (Surface Technology Systems Plc)

Figure 1.3: (a) Dust particles trapped above silicon wafers in a plasma processing dis-
charge [5]. (b) Micrograph of a “killer particle”.

for particle motion are ideal for studying the dynamics of complex plasmas by video
4 1. Introduction

microscopy, e.g. interparticle distances are of the order of hundreds of microns, typical
frequencies of the order of a few Hertz.
Due to the high charges the electrostatic potential energy of the dust particles by far
exceeds the thermal energy of the microspheres which are effectively cooled to room
temperature by the ambient neutral gas: the system is said to be strongly coupled.
The strong-coupling regime is hardly reached in ordinary plasmas. In colloidal plasmas,
the dust particles can arrange in ordered crystal-like structures, the plasma crystal (see
Fig. 1.4).

Figure 1.4: The plasma crystal with two layers, top and side view.

Colloidal plasmas therefore enable to study a vast variety of novel phenomena, like fluid
and crystalline plasmas, phase transitions, strong-coupling effects, waves and Mach cones
in condensed matter and many more. They provide a unique system bridging the fields
of plasma physics, condensed matter and material science. Thus, the leading motive for
the enormous research activity in this field is that dynamic processes of strongly coupled
systems can be investigated on the kinetic level of individual particles.
The above mentioned properties set the stage for the physics of colloidal plasmas. The
topics presented in the Lecture notes are guided by the available experiments on colloidal
and dusty plasmas. The theoretical concepts are developed to a depth necessary to
understand the experimental findings. A complete theoretical description is not intended,
here. In these Lecture notes, I will first give an introduction to the fundamental properties
of dusty plasmas which include particle charging, which is treated in Chapter 2, and forces
 5

on the dust particles in Chapter 3. Strongly coupled systems together with the particle-
particle interaction and phase transitions are described in Chapter 4 and 5. Thereafter,
collective effects like waves in weakly and strongly coupled systems will be discussed in
Chapters 6 and 7, respectively. The above concepts will then also be applied to systems
consisting of only a small number of particles, so-called dust clusters, in Chapter 8.
Finally, in Chapter 9 and 10 I will present phenomena in technological and astrophysical
dusty plasmas that have already been briefly mentioned in this Introduction.
6 2. Charging of Dust Particles

2. Charging of Dust Particles

The charge a particle attains in a plasma is certainly the most fundamental parameter.
Unlike for electrons or ions, the particle charge is not fixed, but determined by the ambient
plasma parameters. The charge may thus change on the particle trajectory which results
in a number of new and interesting phenomena.

2.1. Outline
The particle charging is, from the theoretical point of view, one of the most interesting
but also difficult questions when all the peculiarities of the charging process of a dust par-
ticle are taken into account including trapping in the highly non-neutral, non-equilibrium
sheath environment of a plasma with the presence of streaming ions and ion-neutral colli-
sions. Hence, here, the charging model will be developed in steps starting from idealized
cases to more complex analyses.
The problem of particle charging is closely related to the theory of electrostatic probes
(Langmuir probes) in plasmas. There, the task is to determine electron and ion densities,
electron temperature etc. from the current-voltage characteristic of the probe. For particle
charging, we assume that these plasma parameters are known and the question is which
potential on the dust grain is established by the currents onto the particle.
The particle attains a potential, the so-called floating potential φfl which is determined
from the condition that at floating potential the sum of all currents to the particle vanishes,
i.e.
X dQd
I` (φfl ) = =0 (2.1)
`
dt

where I` denotes the different currents to the probe at the floating potential. Since we
are interested in the equilibrium charge (at least at the moment), the first task will be to
determine this floating potential from the particle currents.
Currents to the particle may arise from the inflow of plasma electrons and ions or from
the emission of secondary electrons and photoelectrons. One can think of many other
charging currents, like electron emission due to strong electric fields etc., which however
are considered unimportant in most cases and which will not be discussed here. Electron
and ion collection will dominate in laboratory plasmas whereas secondary electron and
photoelectron emission are considered important under astrophysical conditions.
2.2. OML Charging Currents 7

2.2. OML Charging Currents

We will start the discussion with electron and ion collection in the ideal case of an isotropic
Maxwellian plasma environment at rest. The currents to the particle can be described
by the OML (“orbital motion limit”) model first derived by Mott-Smith and Langmuir
in 1926 [6]. There, it is assumed that electrons and ions move towards the dust particle
from infinity on collisionless orbits subject only to the electrostatic interaction with the
dust particle. The OML problem is exactly equivalent to Kepler dynamics or Rutherford
scattering.
Since the electrons are much more mobile than the ions the particle will generally
charge negatively. In the calculation of the currents we will adopt that view and assume
that the particle potential is negative with respect to the plasma potential, i.e. φp < 0.
Thus, the ions are the attracted and the electrons are the repelled species. The ion and
electron OML currents are given by the following expressions (given here in advance for
reference, they will be derived in the following section)
s !
2 8kTi eφp
Ii = πa ni e 1−
πmi kTi
s !
2 8kTe eφp
Ie = −πa ne e exp , (2.2)
πme kTe
where a is the particle radius, ni ,ne are ion and electron densities, Ti ,Te the respective
temperatures and mi ,me their masses. For the sake of completeness, the OML currents
for positive particle potentials φp > 0 read as
s !
2 8kTi eφp
Ii = πa ni e exp −
πmi kTi
s !
8kTe eφp
Ie = −πa2 ne e 1+ . (2.3)
πme kTe

2.2.1. Derivation of the OML Currents

Ion Current
As mentioned above, we will study here the usual case of φp < 0 to derive the OML
currents of Eq. (2.2). We start with the ion species that is attracted by the dust. Possible
ion trajectories are depicted in Fig. 2.1. Ions with impact parameters smaller than a
critical parameter bc , i.e. b < bc , will hit the particle, those with b > bc will only be
deflected in the electric field of the dust. The critical parameter bc is defined by the
condition that the ion will hit the dust particle with grazing incidence.
At infinity the angular momentum of such an ion with the critical impact parameter
bc is
L = |~r × ~p| = mi vi,0 bc .
8 2. Charging of Dust Particles

Figure 2.1: Ion trajectories with different values of the impact parameter b.

Since this ion hits the dust at grazing incidence its angular momentum at the particle
surface is
L = m i vi a .
When there are no collisions the angular momentum of the ion is conserved.
The energy balance of the ions is the sum of kinetic and potential energy in the
electrostatic potential of the dust particle
1 1
mi v2i,0 = mi v2i + eφp .
2 2
Here, the energy of the ion at infinity and when arriving on the particle are balanced.
Using the conservation of angular momentum the energy balance can be written as
v2
!
b2c
!
1 1 eφp 1 eφp
mi v2i,0 = mi v2i,0 2i + = mi v2i,0 +
2 2 vi,0 (1/2)mi v2i,0 2 a 2 (1/2)mi v2i,0
and thus
!
2eφp
b2c =a 2
1− . (2.4)
mi v2i,0
We can then easily define the cross section for ion collection as
!
2eφp
σc ≡ πb2c = πa2 1− . (2.5)
mi v2i,0
It is seen that the cross section σc for ion collection is larger than the particle’s geometric
cross section σ = πa2 . Of course, this is due to the attraction by the dust (Remember:
φp < 0) as also easily seen from Fig. 2.1.
2.2. OML Charging Currents 9

Now, the charging current to the particle by ions is given by

dIi = σc (vi ) dji = σc (vi )ni evi f(vi )dvi ,
where ji = ni evi is the ion current density. The ion current has to be integrated over the
ion velocity distribution function f(vi ) that is assumed to be an isotropic Maxwellian in
the following, i.e.
3/2 1 2
! Z
∞
m m v

2 i 2 i i
f(vi ) = 4πvi exp − with f(vi )dvi = 1 ,
2πkTi kTi
0

where Ti is the ion temperature and k is Boltzmann’s constant. The charging current is
then obtained by integration over the Maxwellian velocity distribution
Z
3/2 ∞ ! 1 2
!
m 2eφ m v

i p i i
Ii = 4π2 a2 ni e 1− v3i exp − 2 dvi . (2.6)
2πkTi mi v2i kTi
0

By rearranging we have
 

Z Z
∞ ∞
 
1 1
mi v2i mi v2i
3/2  ! ! 
mi 2eφp
 
Ii = 4π2 a2 ni e  v3i exp − 2
dvi − vi exp −2
dvi  .
 
2πkTi  kTi mi kTi 
{z
|0 } |0 {z }
 

F1 F2
These two integrals can be easily evaluated resulting in∗
2(kTi )2
F1 =
m2i
2(kTi )2 eφp
F2 = .
m2i kTi
Finally, collecting all terms the OML ion current is
s !
8kTi eφp
Ii = πa2 ni e 1− . (2.7)
πmi kTi
This relation can be interpreted as a product of the ion current density ji = ni evth,i at
the ion thermal velocity vth,i = (8kTi /πmi )1/2 onto a dust particle and an effective cross
section. The term πa2 (1 − eφp /kTi ) describes the increased cross section of the dust
particle at the ion thermal energy kTi (which is not exactly the “real” ion thermal energy
(3/2)kTi ).
∗
The definite integral is given by
Z
∞
2 k!
xn e−ax dx = for odd n = 2k + 1 .
2ak+1
0
10 2. Charging of Dust Particles

Electron Current
The electron current can be calculated in complete analogy to the ion’s. The cross section
of the dust particles for electrons is
!
2eφp
b2c = a2 1+ .
me v2e,0
The difference to the ionic cross section is the “+”-sign in the bracket. This results in
a reduction of the effective cross section compared to the dust particle area since the
electrons are the repelled species (still φp < 0). The electron current (with Maxwellian
electrons) is then determined by
Z
∞
1
me v2e
! !
me 3/2 2eφp 3
 
2 2 2
Ie = −4π a ne e 1+ v exp − dve . (2.8)
2πkTe me v2e e kTe
vmin
q
Here, the lower bound of the integral is vmin = −2eφp /me instead of “0” in the ion cur-
rent since only electrons that are energetic enough to overcome the electrostatic repulsion
can reach the dust particle, i.e. electrons with me v2e /2 > me v2min /2 = −eφp are able to
hit the dust.
For the electron currents the analogous integrals I1 and I2 result in†
2(kTe )2
! !
eφp eφp
I1 = exp 1 −
m2e kTe kTe
2(kTe )2 eφp
!
eφp
I2 = 2
exp
me kTe kTe
The electron current finally is
s !
8kTe2 eφp
Ie = −πa ne e exp . (2.9)
πme kTe
One could have obtained the same result for the electron current from a statistical descrip-
tion by arguing that the thermal electron flux onto the dust particle will be reduced by a
Boltzmann factor ne → ne exp(eφp /kTe ) for the repelled electrons: Only the electrons of
the high-energy tail of the Maxwellian distribution will reach the particle.
†
The following integrals result in
Z
∞
2 1 −av2
xe−ax dx = e
2a
v

and
Z
∞
2 1 2
x3 e−ax dx = 1 − av2 e−av


2
.
2a
v
2.2. OML Charging Currents 11

2.2.2. Discussion of the OML Model

General: Our derivation was based on the assumption of collisionless ion trajectories.
This is an essential ingredient in view of conservation of angular momentum. Further, the
calculations were done for an isotropic situation where we have used the isotropic form
of the Maxwell velocity distribution function. Another assumption is that the distribution
function for the electrons and ions is Maxwellian in the first place.
However, all of these conditions are often violated in plasma discharges. Typically, the
ion mean-free path is often not much larger than the Debye length, especially in discharges
in noble gases that have very large cross sections for ion-neutral charge exchange collisions.
Moreover, the particles are trapped in the space charge sheath of a plasma discharge (as
we will see below) where the ion motion is directed towards the electrodes. Finally, the
distribution functions of electrons is in plasma discharges often better characterized by
a bi-Maxwellian or Druyvesteyn-like distribution rather than a pure Maxwellian. Hence,
especially the high-energy tail of the electron distribution that determines the particle
charge can strongly deviate from the Maxwellian assumption. For a full treatment of all
these processes, thorough numerical simulations [7] are required. However, the effect of
streaming ions and ion-neutral collisions will be further discussed below to get a feeling
for these effects.

Energy barrier in the OML model: An interesting point is that for the OML motion an
angular momentum barrier exists: Certain ions that from energy balance considerations
should fall onto the particle will actually not reach the particle [8]. The reason is that the
effective potential Ueff = eφ − L2 /2mr2 has a “hump” at distances r > a where the ions
are already reflected. Their distance of closest approach thus is not the particle surface.
They do not come close to the region near the particle where the above discussed energy
and angular momentum balance hold. This hump in the effective potential appears only
for a small range of angular momenta and energies. For particles much smaller than the
Debye length (a  λD ) that fraction of reflected ions is negligible and the OML results
are still valid [9, 10].

Radial motion theory: In Langmuir probe theory a number of other models are available
to describe the current to a (spherical) probe, like e.g. the radial motion theory (ABR) by
Allen, Boyd and Reynolds [11] where the ions start from infinity with no kinetic energy
and are accelerated radially towards the probe (or the dust particle). However, it has been
shown [12] that in the ABR model the particle floating potential tends to zero for a  λD
which is unphysical. On the contrary, OML gives a finite particle potential independent
of particle size (see Sec. 2.4.1.).
12 2. Charging of Dust Particles

2.2.3. Streaming Ions

In many cases, like in the sheath environment of a discharge, the ions have a drift velocity
ui that can be much larger than the thermal ion velocity vth,i . Thus, the ion species is no
longer described by an isotropic Maxwellian distribution. In that case, the ion current is
modified. The ion current to a dust particle can then be written in the form

Ii = πa2 ni evth,i f(ui ) , (2.10)

where f(ui ) is a (rather complicated) function of the ion drift velocity [13]
" ! #
π 1 eφp 1
r
2
f(x) = x 1+ 2 − 2
erf(x) + √ e−x
4 2x kTi x πx
q
where x = ui / 2kTi /mi and erf(x) is the error function‡ . For ions with high streaming
velocity ui  vth,i the ion current reduces to
!
2 2eφ
Ii = πa ni eui 1− , (2.11)
mi u2i
which is obtained by replacing the thermal ion energy kTi in Eq. (2.2) by the kinetic
energy of the drifting ions mi u2i /2. The effect of streaming ions is further discussed in
Sec. 2.4.3.

2.2.4. Collisions
Ion-neutral collisions considerably modify the ion current onto the dust since the ion
loses its kinetic energy in that collision and is subsequently accelerated towards the dust.
Further, the presence of collisions will lead to ions trapped in the electrostatic potential
well performing Keplerian orbits around the dust [14]. In plasma discharges in noble gases
ion-neutral collisions are predominantly charge-exchange collisions where an ion transfers
its charge to a formerly neutral atom

A+ + A −→ A + A+ .

In a weakly collisional regime one might estimate that every ion-neutral collision in a
sphere of a certain radius R0 around the dust leads to a collection of this ion. The
probability of such a collision is roughly R0 /`mfp when independent collisions can be
assumed. Here, `mfp = 1/(nn σin ) is the mean free path for ion-neutral collisions, nn is the
‡
The error function is given by
Zx
2
erf(x) = √ exp(−y2 ) dy
π
0
2.3. Other Charging Currents 13

neutral gas density and σin is the ion-neutral (charge-exchange) collision cross section.
The thermal current of ions through a sphere of radius R0 is simply I = πR20 ni evth,i
yielding a collisional current of
R0
Icoll = πR20 ni evth,i
`mfp
Hence, the total ion current can be assumed [9] to be the sum of the ion collection current
in Eq. (2.2) and the collision current
R3
!
eφp
2
Ii = πa ni evth,i 1− + 2 0 . (2.12)
kTi a `mfp
A reasonable size R0 would be where the ion-grain interaction energy is of the order of
the ion thermal energy, i.e. using a Debye-Hückel interaction of shielding length λD
kTi Zd e −R0 /λD
= e .
e 4π0 R0
To a good accuracy, one finds [15] that (R0 /λD )3 = 0.1(a/λD )2 (eφp /kTi )2 (thereby making
also use of the capacitance model in Sec. 2.4.2.). Combining this all, the ion charging
current including collisions can be written as
 !2 
eφp eφp λD 
Ii = πa2 ni evth,i 1 − + 0.1 . (2.13)
kTi kTi `mfp

The effect of collisions on the particle potential will be further discussed in Sec. 2.4.4..

2.3. Other Charging Currents

Two other possible charging mechanisms are described here which are considered im-
portant mostly for astrophysical situations [16]. These are charging by UV radiation
and secondary electron emission. Both processes lead to electron emission from the dust
particle and can thus be treated as a positive current to the particle.
Other possible charging mechanisms, like thermal emission, field emission, sputtering,
or ion-induced electron emission are neglected, since they are usually not important for
dusty plasmas in the laboratory or in space.

2.3.1. Photoelectron Emission

The absorption of UV radiation releases photoelectrons and hence constitutes a positive
charging current. The magnitude of the current depends on the photoemission yield η
which is specific for the dust material (η ≈ 1 for metals and η ≈ 0.1 for dielectrics). In
14 2. Charging of Dust Particles

many cases the released photoelectrons can be described by a Maxwellian distribution

with a temperature Tν . The charging current by photoemission is then given by [16]

Iν = ηπa2 eK φp ≤ 0
!
2 −eφp
Iν = ηπa eK exp φp > 0 , (2.14)
kTν
where K is the flux of photoelectrons. Photoemission by UV radiation might be important
for dust grains in astrophysical situations near stars. For example, this effect is made
responsible for dust layers floating above the surface of the Moon.

2.3.2. Secondary Electron Emission

Secondary electron emission from the impact of energetic electrons might also be impor-
tant under certain conditions in astrophysical situations. The secondary electron emis-
sion coefficient δ(Ee ) describes the number of released electrons per incoming electron.
It strongly depends on the energy of the impinging electron Ee and is described by the
empirical formula
s !
Ee Ee
δ(Ee ) = 7.4δm exp −2
Em Em
that has the maximum yield of δm at Ee = Em . Here, δm and Em are material-dependent
quantities that are in the range of δm = 0.5 . . . 30 and Em = 0.1 . . . 2 keV. Since the
maximum yield δm can be much larger than 1, many more electrons can be released for
each incoming electron. This may even lead to positively charged dust grains.
Meyer-Vernet has calculated the current from secondary electron emission: Integrating
the secondary yield over a Maxwellian distribution of incoming electrons leads to the
following expression of the charging current
s !
2 8kTe eφp
Is = 3.7δm πa ne exp F5,0 (Em /4kTe ) for φp < 0 (2.15)
πme kTe
s ! !
2 8kTe eφp eφp eφp
Is = 3.7δm πa ne 1+ exp − F5,B (Em /4kTe ) for φp > 0
πme kTe kTe kTs
q
with Ts being the temperature of the released secondary electrons, B = 4eφp /Em and
Z
∞
2 +u)
F5,Y (x) = x2 u5 e−(xu du ,
Y

which is an integral that can be evaluated numerically.

2.4. The Particle as a Floating Probe 15

Figure 2.2: OML currents onto a dust particle for different values of electron and ion
temperatures. The intersection of electron and ion current gives the floating potential
(the negative electron current is plotted as positive, here). Other parameters used here are
a = 4.7 µm and ni = ne = 1·109 cm−3 , which influence only the absolute magnitude of the
currents, but not the floating potential. As an example, the floating potential φfl ≈ −5 V
is indicated for the Te = 3 eV and Ti = 0.03 eV.

2.4. The Particle as a Floating Probe

2.4.1. OML Currents
As mentioned above, the floating potential of the particle is determined from the condition
that all currents to the particle vanish, see Eq. (2.1). In laboratory discharges usually
only the collection currents need to be considered. The OML electron and ion currents
as a function of particle potential φp are shown in Fig. 2.2. The floating potential is then
found from the intersection of the electron and ion currents.

Te /Ti 1 2 5 10 20 50 100
H -2.504 -2.360 -2.114 -1.909 -1.700 -1.430 -1.236
He -3.052 -2.885 -2.612 -2.388 -2.160 -1.862 -1.645
Ar -3.994 -3.798 -3.491 -3.244 -2.992 -2.660 -2.414

Table 2.1: Normalized floating potentials eφfl /kTe in quasineutral plasmas ne = ni in

Hydrogen, Helium and Argon for different electron-to-ion temperature ratios.
16 2. Charging of Dust Particles

Figure 2.3: Normalized floating potential of a dust particle for reduced electron density
ne /ni .

Analytically, the floating potential φfl is obtained by equalizing the electron and ion
currents resulting in
s !
eφfl mi Te ne eφfl
1− = exp . (2.16)
kTi me Ti ni kTe
This equation can be easily solved numerically for φfl for given values of the plasma
parameters. In Table 2.1 the calculated floating potentials are shown for various discharge
conditions. For the astrophysically important case of the isothermal (Te = Ti ) hydrogen
plasma the well-known Spitzer value φfl = −2.5kTe /e is approached. Under typical
laboratory discharges in heavier gases with Te  Ti a good rule-of-thumb approximation is
φfl ≈ −2kTe /e. However, for electron energy distributions with even a small suprathermal
electron component the floating potential will be decisively different.
If the electron density is reduced with respect to the ion density ne < ni , e.g. in the
sheath of a discharge, the electron charging current is reduced and the particle attains a
more positive floating potential. This is shown in Fig. 2.3 where the floating potential
is shown for reduced electron density ne /ni = 0 . . . 1. For ne /ni = 1 the above value is
retrieved (see Table 2.1), for ne /ni → 0 even positively charged dust is found.

2.4.2. The Capacitance Model

After having determined the floating potential on the particle (at least in the case of pure
OML currents) the charge on the particle has to be found. Therefore, the dust particle is
2.4. The Particle as a Floating Probe 17

considered as a spherical capacitor of capacitance C. The particle charge is then given by

Qd = Zd e = Cφfl , (2.17)

where Zd is the number of elementary charges on the dust. In the vacuum case the
capacitance of a sphere is

C = 4π0 a .

In an ambient plasma with shielding length λD the capacitance is

a
 
C = 4π0 a 1 + ,
λD
which for the typical case of a  λD reduces to the vacuum value.
That means that a particle of a = 1 µm radius attains Zd = 695 elementary charges
per volt floating potential. With the above mentioned rule-of-thumb estimation of the
floating potential φfl = −2kTe /e an approximate formula of the dust charge is given by

Zd = (−)1400aµm Te,eV (2.18)

with Te,eV being the electron temperature in electron volts and aµm the particle radius in
microns. We will often treat Zd as a positive number and deal with the fact that the dust
(usually) is negatively charged by a minus sign for the elementary charge.

2.4.3. Floating Potential with Streaming Ions

For streaming ions, the floating potential is obtained from the equality of the electron
current and the ion current for drifting ions according to Eq. (2.10). For this case, the
floating potential is shown in Fig. 2.4. One can see that for small drift velocities up
to the thermal velocity the floating potential is nearly unchanged and equal to the case
without drift. Then, first a decrease of floating potential is observed (by about a factor
of two) and after that a strong increase. Thus the particle first becomes more negative
with respect to the OML case. The minimum floating potential is observed when the ions
have approximately Bohm velocity ui ≈ vB = (kTe /mi )1/2 . With further increased drift
velocity the particles become less negatively charged. The first decrease of the potential
is due to the reduction of the collection cross section (1 − 2eφ/mi u2i ) with increasing
drift velocity. The increase of the potential at high drift velocities finally is due to the
increased ion flux ni eui . Moreover, it is seen that the approximated ion current according
to Eq. (2.11) already works for relatively low drift velocities ui > vth,i .
18 2. Charging of Dust Particles

Figure 2.4: Particle floating potential as a function of the streaming velocity of the ions.
The floating potential has been normalized to the electron temperature, the drift velocity
to the ion thermal velocity. The classical OML result is also indicated. The solid line
is the full Eq. (2.10), the dashed line is the approximation of Eq. (2.11). Parameters:
Te /Ti = 100 in argon.

2.4.4. Floating Potential with Collisions

In the case of collisions, the particle floating potential is derived by balancing the electron
current and the collisional ion current according to Eq. (2.13). Figure 2.5 shows the
floating potential of a dust particle in a collisional plasma background as a function of the
mean free path `mfp . The constant OML result eφfl /kTe = −2.414 (compare Tab. 2.1)
is shown as the dotted line. The red solid line indicates the influence of collisions using
the collisional ion current of Eq. (2.13). For low collisionality (large mean free path) the
OML result is retrieved. With shorter mean free path the particle potential becomes
substantially less negative, the charge number substantially reduces. This is due to the
collision-enhanced ion current to the particle.
For comparison, the results of a more sophisticated calculation and fitting procedure
by Zobnin et al. [17] are also shown (see also [18]). Zobnin revealed two effects: first, the
floating potential depends on particle size (which is not found for OML or the collisional
current). Second, at very small mean free path the particle potential again becomes more
negative. This is due to the effect that there the ion motion towards the dust is effectively
hampered by the very frequent ion-neutral collisions (hydrodynamic limit). However, the
2.4. The Particle as a Floating Probe 19

Figure 2.5: Particle floating potential as a function of ion mean free path. Parameters:
ni = ne = 2 · 108 cm−3 , Te = 3 eV, Ti = 0.03 eV in argon. The dotted line is the OML
result. The solid red line indicates the result from the collisional ion current according to
Eq. (2.13). The blue lines indicate the results of more sophisticated calculations by Zobnin
et al. [17] for different particle sizes.

floating potential derived from our collisional ion current of Eq. (2.13) is in good agreement
with the more sophisticated results down to mean free paths of the order of millimeter or
even sub-millimeter for particle sizes in the micrometer range. Hence, collisions effectively
reduce the particle floating potential and dust charge (up to a factor of 2) compared to
the OML results.

2.4.5. Floating Potential with Secondary Electron Emission

If secondary electron emission is important, as for cosmic grains, the floating potential is
determined from the condition

Ie + Ii + Is = Itot = 0 .
20 2. Charging of Dust Particles

Such a situation is shown in Fig. 2.6, where the secondary electron emission current
Is together with the OML charging currents Ie and Ii are shown as a function of the
particle potential φp . Under the conditions chosen here the total current vanishes for
three values of the floating potential. The floating potential thus is not a unique value,
but is multivalued [19].

Figure 2.6: Secondary electron emission current Is , the OML charging currents Ie and Ii
as well as the total current Itot as a function of grain potential. The total current vanishes
for three values of the floating potential, where the middle root is unstable, but the two
extreme roots are stable. Hence, positive and negative grains can exist under the same
plasma conditions. Parameters are as indicated in the inset.

The two outer values of the floating potential φp1 and φp3 are stable roots, the middle
φp2 is unstable. The first stable root is always negative φp1 < 0, the second stable is
always positive φp3 > 0. For these two outer roots

dItot
<0 .
dφ φp1,3

Thus, if a positive charge +∆Q, e.g. an ion, is added to the dust at the floating potential
φp1,3 the particle potential gets more positive φp1,3 → φp1,3 + ∆Φp which results in a neg-
ative total current to the particle that compensates the positive charge +∆Q. Therefore,
2.5. Temporal Evolution of the Particle Charge 21

φp1,3 are stable roots. Correspondingly, φp2 is unstable. Any fluctuation of the particle
charge near φp2 will switch the potential to either φp1 or φp3 .
That means that in the same plasma negatively as well as positively charged dust
grains can exist. The oppositely charged particles can then immediately coagulate in
the plasma. This process might have an enormous influence on the growth mechanism of
planetesimals in astrophysical dusty plasmas where secondary electron emission is present.

2.5. Temporal Evolution of the Particle Charge

So far, the particle charge has been considered as static. However, the temporal behavior
of the dust charge can have a significant influence on the dynamic properties of the dust.
The time scale for OML charging by ions τi is found from Eq. (2.1) by considering the
ion charging current only
s !
dQ 8kTi eφp
= Ii = πa2 ni e 1− .
dt πmi kTi
Replacing the particle potential using the capacitor model this results in the differential
equation
s !
dQ 8kTi Qe
= πa2 ni e 1− . (2.19)
dt πmi 4π0 akTi
This equation has the solution
kTi
Q(t) = Q0 e−t/τi + 4π0 a
e
with the charging time scale
kTi 1
τi = 4π0 a 2
. (2.20)
e πa eni vth,i
This time scale can be interpreted as the charging time τ = RC of the particle capacitance
C = 4π0 a through the resistance R = U/I, where U = kTi /e is the typical electric
potential of thermal ions and I = πa2 ene vth,i is the typical thermal OML ion current to
the particle. Or, alternatively from Eq. (2.7), R = dU/dI at the floating potential. The
electron charging time is given by the same formula where the ion quantities are replaced
by those of the electrons:
kTe 1
τe = 4π0 a 2
. (2.21)
e πa ene vth,e
Due to the higher electron mobility it follows that τe  τi , and the overall charging time
is dominated by the slower ions. Typical charging times in laboratory discharges are of
the order of microseconds (see also Fig. 2.7 and 2.8). This is still typically much faster
than the dynamical time scale of the particles, so the particle charge should always be in
equilibrium with the plasma conditions. However, we will find some exceptions from that
reasoning in the following.
22 2. Charging of Dust Particles

Figure 2.7: Temporal evolution of a dust particle in the rf sheath. The inset is a mag-
nification of the oscillating particle charge. Parameters: a = 4.7 µm, kTe = 3 eV,
kTi = 0.03 eV, ni = 1 × 109 cm−3 , and the electron flooding is assumed as α = 25 %
of the rf period. The equilibrium charge then is Zd = 8610 negative elementary charges.
After [20].

2.5.1. Charging in the rf Sheath

In laboratory investigations on colloidal plasmas the particles are usually trapped in the
sheath of rf discharges, where the (time averaged) electric field force on the particle
balances gravity (see Sec. 3.7.). In the rf discharge, the electrons are able to follow the
instantaneous rf voltage (which typically oscillates at 13.56 MHz), whereas the ions and
the dust grains only react to the time averaged fields. In a simplified picture, the electron
component of the plasma can be described as oscillating back and forth between the
electrodes leading to a periodically increasing and collapsing sheath at the electrodes. A
dust particle trapped in the sheath therefore periodically “sees” a quasineutral plasma
environment, when the sheath is flooded by electrons, and a pure ion sheath, when the
sheath is expanded. Hence, the electron charging current to the dust particle will also be
modulated by the rf frequency, thus
Q̇ = Ii + Ie for 0 < t < αT
Q̇ = Ii for αT < t < T
where α = 0 . . . 1 is the fraction of the rf period T = 75 ns when the dust particle is in
the “flooded” sheath (see Fig. 2.7a).
The temporal behavior of a dust particle in the sheath is shown in Fig. 2.7b for a par-
ticle immersed into the sheath starting from zero charge. In accordance with the charging
time scale, see Eq. (2.20), the steady state charge is reached after a few microseconds.
Due to the periodic oscillation of the electron component, the dust particle charge is
also modulated with the applied frequency of 13.56 MHz leading to a sawtooth behavior
2.5. Temporal Evolution of the Particle Charge 23

of the dust charge. In the inset one can see the steep electron charging towards more
negative charge numbers and the slower ion decharging. After long times the particle
charge is modulated by few hundred elementary charges at a mean particle charge of a
few thousand. Hence, the charge modulation is a few percent at a high frequency. These
modulations should not affect the particle dynamics on the time scale of the dust particle
motion.

2.5.2. Stochastic Fluctuations

Besides the periodic modulation of the dust charge due to the rf sheath, the particle
charge experiences stochastic fluctuations due to the discreteness of the charge carriers
– electrons and ions. This effect has been investigated by Cui and Goree as well as
Matsoukas and Russel [21, 22] simulating the collection of individual electrons or ions
according to the OML currents. The probability of collecting an ion pi or electron pe is
taken proportional to the ion (electron) current to the particle. The collection of electrons
and ions is treated as a stochastic process. At the floating potential, obviously pi = pe .
In Fig. 2.8 the temporal behavior of the dust charge due to stochastic charging processes
starting from zero charge is shown.
In the beginning at zero charge the probability of collecting an electron is much,
much higher than collecting an ion due to the higher mobility of the electrons. The
particle gets negatively charged. After a certain time, the charging time according to
Eq. (2.20), the probability of electron or ion collection becomes more and more equal

Figure 2.8: Temporal behavior of the dust charge by stochastically collecting individual
electrons and ions. The inset is a magnification of the dust charge fluctuations around
equilibrium Zeq = 8610. Parameters are the same as in Fig. 2.7.
24 2. Charging of Dust Particles

to each other. After having reached the equilibrium value the dust charge fluctuates
around this value with an amplitude of about 100 elementary charges. The equilibrium
value found in the stochastic approach is the same as for the static approach according
to Eq. (2.16). It has been shown from q simulations and analytical calculations that the
charge fluctuations are δZrms ≈ 0.5 Zeq . These fluctuations are small and quite high-
frequent so that an influence on the particle dynamics is not to be expected for micron
sized particles. However, nanometer sized grains have equilibrium charges of the order of
Zeq = 1. Charge fluctuations can cause the particles to be charged even positively. Then
one could also find positively and negatively charged dust in the same plasma which can
can lead to fast particle coagulation. This process might be of interest for particle growth
in technological discharges (see Chapter 9).

2.6. Influence of Many Particles and Electron Depletion

So far, only the charging of single dust particles has been considered. Now, we would like
to investigate the behavior of many dust particles. When the dust density nd is high, a
considerable amount of electrons is bound to the dust. Thus, the density of free electrons
in the plasma might be drastically reduced due to the presence of the dust which in turn
influences the charging behavior of the dust [24].
For a quantitative description we consider a situation as described in Fig. 2.9a) where
a finite dust-containing plasma region is embedded in an infinite dust-free plasma. In the
outer plasma, we find the undisturbed plasma density ni,0 = ne,0 , the dust density is zero
(nd = 0) and also the plasma potential ψ is defined as zero. In the dust-containing plasma
cloud the plasma potential ψ differs from that in the outer plasma due to the presence
of the charged dust particles [Fig. 2.9c)]. The electron and ion density in the dust cloud
is assumed to adjust to the change in potential with a Boltzmann factor, which might be
justified in astrophysical situations where Te ≈ Ti ,
eψ
ne = ne,0 exp( )
kTe
eψ
ni = ni,0 exp(− ) .
kTi
The influence of the dust becomes noticeable when the quasineutrality condition

ne e = ni e − nd Zd e = ni e − nd Cφfl (2.22)

is affected by the dust. The negative charge is either due to free electrons or due to
electrons bound on the dust particles. If the dust density is high then there are not enough
electrons present to charge the particles to their single-particle floating potential. The
particles thus acquire a more positive potential. This effect is called electron depletion.
2.6. Influence of Many Particles and Electron Depletion 25

Figure 2.9: a) Sketch of the dust cloud in an in extended prstine plasma. b) Particle
floating potential φfl and cloud potential ψ as a function of the Havnes parameter P.
Parameters: Te /Ti = 100 in Helium. c) A one-dimensional section through the dense
dust cloud. The particle potential φ and the cloud potential ψ in different regimes of the
parameter P. After [23].
26 2. Charging of Dust Particles

Hence, the quasineutrality condition has to be coupled to the charging equation. Using
^ and plasma potential ψ
the normalized dust floating potential φ ^ with

^ = eφfl
φ ^ = eψ
and ψ
kTe kTe
the quasineutrality condition and the charging equation are written as
!
^ − exp − Te ψ ^ − 4π0 a kTe nd φ
 
exp ψ ^ = 0 (2.23)
Ti | e {ze n }
e,0

P
! s " ! #
Te ^ Te mi Te ^  
^ = 0 .
1− φ − exp 1 + ψ exp φ (2.24)
Ti Ti me Ti

The terms containing exp(ψ) ^ and exp(Te /Ti ψ)

^ are the normalized electron and ion den-
sities, respectively. The Eqs. (2.23) and (2.24) have to be solved simultaneously for ψ ^
and φ,^ i.e for the floating potential of the dust and the local plasma potential in the dust
cloud. The dust influence is described in terms of the so-called Havnes parameter
nd Zd 4π0 a kTe nd nd
P= = = 695 Te,eV aµm ,
ne,0 e e ne,0 ne,0
where Te,eV is the electron temperature in electron volts and aµm is the particle radius in
microns. Figure 2.9b) illustrates the influence of the electron depletion with increasing
Havnes parameter P. The single particle limit is obtained for P → 0. Near P = 0.1
the floating potential on the particle increases considerably towards more positive values
and approaches φ ^ fl → 0 for P → ∞. This reflects the electron depletion. There are not
enough electrons present as to charge all dust particles to the full single-particle potential,
thus, the available electrons are distributed among the dust particles, i.e. the free electron
density ne = 0 and Zd = ni /nd .
 27

3. Forces and Trapping of Dust Particles

In this part we will discuss the main forces acting on dust particles in a plasma discharge.
These forces are gravity, electric field force, ion drag force, thermophoresis and neutral
drag. After analysis of the forces under plasma conditions the trapping of dust particles
in the laboratory, under microgravity and in plasma processing devices will be described.
We are then in a position to derive a method for the charge measurement in the sheath
of a plasma and to discuss vertical oscillations in the plasma sheath.

3.1. Gravity
The gravitational force simply is

~Fg = md ~g = 4 πa3 ρd ~g , (3.1)

3
where ~g is the gravitational acceleration and ρd is the mass density of the dust grains.
Since this force scales with a3 it is the dominant force for large particles in the micrometer
range and becomes negligible for particles in the nanometer range.

3.2. Electric Field Force

Obviously, the electric field force due to an electric field ~E
~FE = Qd~E = 4π0 aφfl~E (3.2)

is the governing force for charged particles. With the applied capacitor model this force
scales linearly with the particle size. In the plasma sheath strong electric fields prevail
that exert an electric field force that is large enough to levitate large micron-sized grains
against gravity. In the plasma volume, due to quasineutrality, only small electric fields
exist. Thus, in the plasma volume, particles can be trapped only when gravity is not
important as for nanometer particles or for large particles under microgravity conditions.

3.2.1. Shielding
One subtle problem should be discussed in more detail, here: Since in a plasma the
(negative) dust particle is surrounded by a (positive) shielding cloud of ions (see Fig. 3.1a),
one might think that the electric field force acts individually on the dust and on the
shielding cloud in opposite directions, thereby canceling each other. In other words: the
entire system of dust and shielding cloud is neutral and thus no net electric field force
28 3. Forces and Trapping of Dust Particles

Figure 3.1: Electric field force and polarization forces: a) “right” and b) “wrong” picture
of the dust particle and its shielding cloud under the influence of an external field E.
Formation of a dipole moment on the dust particle induced by an electric field (c) or by
directed charging (d). Polarization of the dust cloud due to an ion streaming motion.

should exist, see Fig. 3.1b. If this reasoning was true there would be no action of the
electric field force on the dust particle.
Hamaguchi and Farouki [25] have discussed that point in detail: To understand the
behavior of the dust one has to carefully distinguish between the source and the effect.
The shielding cloud is formed due to the presence of the electric field of the dust particle
(and the external field). The shielding cloud reacts according to these fields. Hence, the
shielding cloud is a response to the presence of the dust. When the dust moves from a
place A in the plasma to a place B, the shielding cloud is not pushed from A to B by the
dust. Rather, the ion cloud decays at A and re-forms at B. The shielding cloud is not
“attached” to the dust by a force and thus there is no counterforce from the cloud onto
the dust.
Consequently, the full force of Eq. (3.2) acts on the dust particle, the dust particle is
moved by this force and the shielding cloud is formed according to the electric fields near
the dust. The shielding cloud does not hinder the action of the electric field force on the
dust. This reasoning holds for spherically symmetric shielding clouds. If the spherical
symmetry of the shielding is distorted by the external field, additional polarization forces
exist, that will be described now.
3.2. Electric Field Force 29

3.2.2. Polarization Forces

Polarization forces arise from dipole moments on the particles or the shielding cloud. In
general, the polarization force on a dipole is
~Fdip = ∇(~
~ p · ~E) , (3.3)
where ~p is the dipole moment. Dipole moments on the particles can either be influenced
by an external electric field or generated by directed charging processes (for dielectric
particles).
The dipole moment on a dielectric particle by an external electric field ~E (see Fig. 3.1c)
is given by
 − 1~
~p = 4π0 a3 E ,
+2
where  is the dielectric constant of the dust material. Due to the a3 scaling the force
due to an induced dipole becomes relevant only for extremely large particles.
Dipole moments may also arise from a directed charging (Fig. 3.1d) due to streaming
ions where one side of the particle is hit more frequently by ions than the other side. For
the case of a particle in the plasma sheath the top side is expected to be more positive
than the bottom. Note, that the dipole moments by directed charging and the induced
dipole are opposite.
Hamaguchi and Farouki [25] have analyzed a situation which is close to that found in
a plasma sheath. In the sheath there is a difference in electron and ion density which is
sustained by an electric field ~E. Consequently, the plasma densities can change over the
size of the shielding cloud, thus a spatially dependent shielding length λD (r) is considered
in the model (see Fig. 3.1e). In the lower sheath of a discharge the electric field usually
points downward from the positive plasma towards the negative electrodes. Near the
plasma edge the electron and ion densities are larger than at the electrode. Thus, the
shielding length on the top side λD1 of a dust particle in the sheath is smaller than that
on the bottom λD2 .
The authors have solved the Debye-Hückel equation for the electrostatic potential φ(r)
in this situation
φ(r) Qd
∆φ(r) − = δ(r)
λD (r)2 0
and found the electric force on the dust particle including the contribution of the distorted
shielding cloud as
~ D (r)
Q2d ∇λ
FE = Qd~E − . (3.4)
8π0 λ2D
The force has two components. The first one is just the usual electric field force as in
Eq. (3.2). The direction of the force is determined by the charge. The second term is due to
30 3. Forces and Trapping of Dust Particles

the deformation of the shielding cloud. It is always in the direction of decreasing shielding
length λD . The total force might thus be increased or decreased by the deformation.
According to Hamaguchi and Farouki the shielding length to be used in Eq. (3.4) depends
on the ion streaming velocity ui . For subthermal ion drifts (i.e., ui < vth,i ) ions can
contribute to shielding and the relevant shielding length is the ion Debye length λD = λD,i .
For supersonic ion streams (ui  vth,i ) the ions are too fast to contribute to shielding
and thus the relevant shielding length is the electron Debye length λD = λD,e (see also
Sec.7.9.). In most cases, however, the polarization forces are negligible, except for very
large particles.

3.3. Ion Drag Force

The next two forces to be discussed are drag forces which arise from a relative motion of
a plasma species relative to the dust particle. In the case of the ion drag force there is
an ion stream relative to the dust particle and the interaction is electrostatic, whereas for
the neutral drag atoms or molecules of the neutral gas background drift past the dust and
interact by direct collisions. The number of particles interacting with the dust per time
interval dt is given by dN = nσ~vrel dt, where n is the density of the streaming species,
σ is the cross section for interaction and ~vrel is the relative velocity (see Fig. 3.2). Then,

Figure 3.2: Sketch to determine the number of plasma particles per unit time moving
relative to the dust particle and interacting with it. Here, σ indicates the cross section for
the relevant interaction of the plasma species with the dust.
3.3. Ion Drag Force 31

the drag force is given from the momentum transfer ∆p per time interval as

~Fdrag = dN∆p = ∆pnσ~vrel . (3.5)

dt
Now, we start with the ion drag force. The ion drag is due to streaming ions. The
streaming motion can arise from ambipolar diffusion or supersonic motion in the sheath.
The ion drag consists of two parts, the collection force ~Fdir due to ions directly hitting the
dust and the Coulomb force ~FCoul due to Coulomb scattering of the ions in the electric
field of the dust particle (see Fig. 2.1). The ion drag force will be discussed in two steps
using, first, a simplified model for a qualitative understanding of the processes before,
second, a more quantitative description is given. The ion drag force can have significant
contributions in dusty plasmas. For example, the ion drag force is made responsible for
the formation of the “void” (see Sec. 3.7.).

3.3.1. Qualitative Discussion: Barnes model

In this first qualitative model, we note that the collection part of the ion drag force is just
due to those ions which are also responsible for the ion charging of the dust and is thus
given by
!
~Fdir = πa2 mi vs ni ~ui 1 − 2eφfl . (3.6)
mi v2s

Here, mi vs is the momentum transfer of the ion at the mean velocity vs = (u2i + v2th,i )1/2
given by the ion thermal velocity vth,i and the ion drift velocity ui . The interaction cross
section is that for ion collection defined in Eq. (2.5).
The Coulomb force is exerted by those ions which are not collected by the dust, but
are deflected in the electric field of the dust grain. From plasma physics textbooks it is
known that
λD
σ = 4πb2π/2 ln Λ = 4πb2π/2 ln
bπ/2

is the cross section for ion-electron Coulomb collisions, where bπ/2 = e2 /(4π0 mi v2 ) is the
impact parameter for 90◦ deflection and ln Λ is the Coulomb logarithm. The Coulomb
logarithm is due to truncation of the intrinsically infinite Coulomb collision cross section
at the Debye length λD , i.e. Coulomb collisions outside the screening cloud are regarded
unimportant. A cut-off towards small impact parameters is not necessary since electrons
and ions are considered point-like.
For collisions with finite-sized dust particles the above Coulomb cross section has to
be modified. The minimum collision parameter is bc since ions with b < bc are absorbed
32 3. Forces and Trapping of Dust Particles

Figure 3.3: Calculated ion drag force on a double-logarithmic scale. In (a) the total force
and contribution of the ion collection and the Coulomb scattering are shown using the
Hutchinson/Khrapak model. (b) Comparison of the qualitative Barnes model (using λD,e )
with the sophisticated Hutchinson/Khrapak model.

by the particle and contribute to the direct force. Integrating the Coulomb collision cross
section from bc to λD yields [26]
 1/2
λ2D + b2π/2
σ= 4πb2π/2 ln   .
b2c + b2π/2

with the slightly modified 90◦ collision parameter

Qd e aeφp
bπ/2 = 2
= .
4π0 mi vs mi v2s
In the last equation we have used the capacitor model to replace the dust charge Qd .
Thus, finally, the Coulomb force on a dust particle from streaming ions is
 1/2  
λ2D + b2π/2 a2 e2 φ2p λ2D + b2π/2
~FCoul = 4πb2π/2 mi ni vs ~ui ln  2  = 2π n ~
u
i i ln   (3.7)
bc + b2π/2 mi v3s b2c + b2π/2

In this force calculation, still only ion trajectories within one Debye length λD around
the dust are considered. An open question so far is which Debye length has to be used,
the electron or the ion Debye length. As a crude estimate Eq. (3.7) with the electron
Debye length λD,e can be used [27].
3.3. Ion Drag Force 33

Finally, the total ion drag force is just the sum of the direct and the Coulomb collision
force
~Fion = ~Fdir + ~FCoul (3.8)

and is directed along the ion streaming motion. The ion drag force is shown in Fig. 3.3.
The ion drag shows a pronounced S-shape. For small drift velocities the force first in-
creases, then decreases and finally increases again. The first maximum is obtained when
the drift velocity is of the order of the thermal velocity of the ions. For small drift veloci-
ties only the Coulomb force needs to be considered. The increase of the force at very low
velocities is due to the increase of the Coulomb logarithm (due to the decrease of bπ/2 )
with velocity. After the maximum the Coulomb force decreases approximately as 1/u2i .
For ui  vth,i the collection force becomes dominant and increases as u2i .

3.3.2. Quantitative Discussion: Hutchinson/Khrapak model

There are, however, some difficulties associated with the Barnes model. First, for the
highly charged dust grains also Coulomb collisions outside the Debye sphere have to be
included [28] since the ions feel the electrostatic potential of the dust far further than just
one Debye length, hence the Coulomb logarithm has to be modified. As already became
obvious in the above discussion a second problem is associated with the correct Debye
length at the different ion streaming velocities. Finally, especially at low ion streaming
velocities one has to consider that the ion velocity is a combination of the ion thermal
velocity and the drift velocity which is adequately described by a shifted Maxwellian
distribution function.
The collection force with a shifted Maxwellian ion distribution is then written as (see
e.g. [29, 30])

Fdir = πa2 mi ni v2th,i H(u) , (3.9)

q
where H(u) which is a function of the normalized ion drift velocity u = ui / 2kTi /mi and
the normalized particle potential χ = −eφp /kTi is given by ∗
√ "  √ #
π  2
  π 1
H(u) = u 2u2 + 1 + 2χ e−u + 4u4 + 4u2 − 1 − 2(1 − 2u2 )χ erf(u) 2 .
8 2 u

The function scales as H(u) → (π/4)u2 = u2i /v2th,i for u  1, hence Fdir = πa2 mi ni u2i
which corresponds to the Barnes expression in Eq. (3.6) at high drift velocities. Hence,
as shown in Fig. 3.3(b), for u  1 the two models agree.
∗
The different prefactor given here compared that in Ref. [30] is due to the different normalizations of
the thermal velocity.
34 3. Forces and Trapping of Dust Particles

Similarly, the Coulomb force is now given by

a2 e2 φ2p
FCoul = 32 ni G(u) ln Λ (3.10)
mi v2th,i
√ 2
with the Chandrasekhar function G(u) = [erf(u) − (2/ 3)ue−u ]/(2u2 ). For strong ion
drifts u  1 the Chandrasekhar function scales as G(u) → 1/(2u2 ) = (π/16)v2th,i /u2i
yielding the same functional form as the Barnes expression in Eq. (3.7).
However, the strongest difference to the qualitative discussion above lies in the de-
termination of the Coulomb logarithm ln Λ. It has been shown by Khrapak et al. and
Hutchinson [29, 30] that the Coulomb logarithm in the modified form
bπ/2 + λs
ln Λ = ln (3.11)
bπ/2 + a
with a velocity dependent screening length already yields quite accurate results. The
screening length
λ2D,e
λ2s = + a2 (3.12)
1 + 2kTe /(mi v2s )
interpolates between the linearized Debye length at thermal ion velocities and the electron
Debye length at high ion streaming velocities. This is reasonable since at higher streaming
velocities the ions cannot participate in shielding (see Sec. 7.9. and also Sec. 3.2.2.).
Extensive numerical calculations by Hutchinson [30] have revealed that for evaluat-
ing bπ/2 = aeφp /(mi v2eff ) in the Coulomb logarithm of Eq. (3.11) and for the velocity
dependent screening length λs in Eq. (3.12) an effective ion velocity of the form
  q 3 
kTi ui / kTe /mi
v2eff = 2 + u2i 1 + 
  
q
mi

0.6 + 0.05 ln µ + (λD,e /5a)( Ti /Te − 0.1)

should be used. Here, µ ist the atomic mass number of the ion.
The Hutchinson/Khrapak model yields highly accurate quantitative results for the ion
drag force. For a particular case the total ion drag and its collection and Coulomb con-
tribution using this model is shown in Fig. 3.3(a) as already discussed above. Fig. 3.3(b)
shows the comparison of the Hutchinson/Khrapak and the Barnes model indicating sub-
stantial differences of the order of a factor of 2 and more, especially for the slow ion drift
velocities that are relevant in experiments.
As complicated as the above discussions and equations already seem, they still refer to
the case of collisionless ion trajectories. Simulations of the ion drag including ion-neutral
collisions indicate a certain influence of collisions on the ion drag force [31, 32]. A “simple”
treatment of collisions is, however, not available.
3.4. Neutral Drag Force 35

3.4. Neutral Drag Force

The neutral drag is a friction force due to neutral gas atoms hitting the dust particle.
Neutral gas atoms or molecules that hit the dust lead to a slowing of the dust particle
motion (when the dust is moving). Using Eq. (3.5) the force on a moving dust grain with
velocity ~vd is readily given as [33]

~Fn = −δ 4 πa2 mn vth,n nn~vd , (3.13)

3
where mn , nn and vth,n are the mass, the density and the thermal velocity of the neutral
gas atoms, respectively. The cross section for interaction is just the geometrical particle
cross section πa2 . The momentum transfer of the neutral gas onto the dust is of the
order of mvth,n and slightly depends on how the gas atoms are reflected from the particle
surface, e.g. diffuse, by specular reflection, isotropic etc. These differences are attributed
to the parameter δ which lies in the range between 1 (for specular reflection) and 1.44
(for diffuse reflection). Experiments [34] on plastic microspheres in a plasma revealed a
value of δ = 1.26 ± 0.13 and δ = 1.44 ± 0.19 using two different methods of measuring
the drag force.
This neutral drag force was calculated by Epstein in 1924 for the analysis of the friction
force in Millikan’s famous oil drop experiment. We will often use that neutral drag in the
form of a friction force , i.e.

~Fn = −md β~vd 8 p

with β = δ . (3.14)
π aρd vth,n
Here, β is the (Epstein) friction coefficient and linearly depends on the gas pressure p.
The friction coefficient is inversely proportional to the particle radius a which means
that in relation to their mass smaller particles experience stronger damping than larger
particles.

3.5. Thermophoresis
The thermophoretic force acts on a dust particle due to a temperature gradient in the
neutral gas. In an (over)simplified picture, it can be argued that neutral gas atoms from
the “hotter” side hitting the dust grain have a larger momentum and thus exert a stronger
force than atoms from the “colder” side. This leads to a force towards colder gas regions.
However, the complete picture is more difficult and will not be discussed here. Following
a rigorous treatment from gas kinetic theory, the thermophoretic force is found to be
2
~Fth = − 32 a kn ∇T
~ n (3.15)
15 vth,n
36 3. Forces and Trapping of Dust Particles

with Tn being the temperature of the neutral gas and kn the thermal conductivity of the
gas. The thermophoretic force is considered to be important for sub-micron particles
due to heating of the gas by the plasma discharge. It has been intentionally applied for
levitation of particles using strong temperature gradients by Rothermel et al. [35] and to
the formation of 3D Yukawa balls (see Sec. 8.4.).

3.6. Laser Forces

In dusty plasmas focused laser beams are used to manipulate dust particles. The origin
of the laser-particle manipulation goes back to the pioneering works of Ashkin [36] who
has confined particles in optical traps. These works also have led to the techniques of
laser-cooling of ions which is needed to form, e.g. Bose-Einstein condensates.
The laser beam interacts with the dust particle and pushes the particle in the direction
of the beam. The origin of this laser force in dusty plasmas can be due to two sources,
the radiation pressure or the so-called photophoretic force.
The force exerted by radiation pressure is just the momentum transfer of the laser
photons pph that hit the dust particle. The radiation pressure can be written as [36]
dpph dNph h ∆Nph hν I
Prad = = = = . (3.16)
Adt Adt λ A∆t c c
Here, Nph is the number of photons hitting the dust particle and λ and ν are the laser
wavelength and frequency, respectively. The intensity I of the laser beam is just the
number of photons of energy hν per time interval focused onto the cross section A
Nphoton hν
I= .
A∆t
The force exerted be the radiation pressure then is
I
Frad = γ Ad (3.17)
c
where Ad is the geometric cross section of the dust particle Ad = πa2 and γ is a coefficient
that takes the kind of interaction of the photons with the particle into account, i.e. γ = 2
for total reflection of the photons or γ = 1 for pure absorption.
For transparent particles, the radiation force also has a component perpendicular to
the beam when the intensity profile of the beam is taken into account (see Fig. 3.4).
Typically, a laser beam has a profile that has maximum intensity in the center and then
gradually decreases outward. For a dust particle in the laser beam, this means that more
photons go through the side of the particle that is closer to the center of the beam (the
upper side in Fig. 3.4) compared to the other side (the lower side). The photons are
deflected by the particle due to refraction. Since more photons on the upper side of the
particles are deflected downwards than vice versa, an upward force (towards the center)
3.6. Laser Forces 37

Figure 3.4: Scheme of the laser-particle interaction due to radiation pressure.

arises. This means, the particle is “sucked” into the beam in the direction perpendicular to
the beam. This of course helps in the application of laser manipulation in dusty plasmas.
This “trapping” of particles in the beam has been applied in the optical trap experiments
of Ashkin and has also been observed in dusty plasma experiments.
The second mechanism that yields a force from laser beam interaction is much more
difficult to analyze quantitatively. There it is assumed that the laser heats the illuminated
particle surface. Similar to the thermophoretic force discussed above, neutral particles
that impinge on the “hot” side of the dust are reflected at higher velocities than on the
cold side. This leads to a force away from the “hot” side of the particle.
Now, the temperature distribution across the particle is very difficult to determine.
When the particle is strongly absorbing, the photons are absorbed at the illuminated
surface. The hot side is the illuminated side and the particles are pushed in the direction
of the beam. If the particle is less absorbing the photons may be absorbed near the back
side of the particles. Then the back side is heated and the photophoretic force is opposite
to the laser beam. Under idealized conditions (i.e. the particle is much smaller than
the mean free path of the gas and the photon energy is absorbed on the front side) the
photophoretic force can be written as [37]
πa3 pI
Fphoto = (3.18)
6 (pavth,n + κT )
where p is the gas pressure, T is the gas temperature and κ is the thermal conductivity
of the particle.
From the simple models, the photophoretic force should exceed the radiation force
considerably. However, this is difficult to judge due to the very complicated nature of the
photophoretic force. Thus, a final answer whether radiation pressure or photophoretic
forces dominate under the conditions of the dusty plasma experiments cannot be given,
here.
Irrespective of the exact mechanism, lasers have been successfully applied in dusty
plasmas to drive and manipulate particles (see e.g. [38] and references therein). And in
many cases, the radiation pressure force was sufficient to explain the particle motion.
38 3. Forces and Trapping of Dust Particles

Figure 3.5: Strength of the various forces as a function of dust particle radius. The
employed parameters are: ρd = 1500 kg/m3 , Te = 2 eV, φfl = −4 V, E = 1000 V/m, ni =
1015 m−3 , ui = vth,i = vth,n , ∇Tn = 200 K/m, kn = 0.016 kg m s−3 (Ar), vth,n = 400 m/s
(Ar).

3.7. Compilation of Forces, Dust Levitation and Trapping

After the description of the relevant forces on the dust particles, now the question is how
these forces can lead to particle levitation and trapping. Therefore, the strength of these
forces as a function of particle size is calculated under the typical experimental conditions
of a low-power Ar discharge (Fig. 3.5). For micron-sized dust particles (right panel) the
dominant forces are gravity and electric field force. The thermophoretic force is already
quite large when assuming a moderate temperature gradient of 200 K/m† . The ion drag
force is smaller than the electric field force and gravity under the chosen conditions. Hence,
for dust particles well above 1 micron in diameter and with small temperature gradients
only electric field force and gravity are important. (Please note, that the electric field is
considered as constant, here. In a discharge, the dust will move to a position, where the
electric field force is large enough to balance the other forces.) The neutral drag is of
interest only for moving particles and not for the identification of stable equilibria. Thus,
the neutral drag is not considered, here.
For nanometer-sized dust particles (left panel of Fig. 3.5) gravity is negligible, as is
for micron-sized dust under microgravity conditions. Then, ion drag force becomes the
dominant force which has to be balanced by the electric field force.
†
For a 3 cm electrode gap this corresponds to a temperature difference of 6◦ C, only.
3.7. Compilation of Forces, Dust Levitation and Trapping 39

Figure 3.6: Trapping and levitation of dust particles: a) large particles, gravity is domi-
nant, b) small particles or microgravity conditions, gravity is unimportant

Consequently, these relevant forces dominate the mechanism of particle trapping in

plasma discharges. For large particles in the micrometer range where the dominant forces
are gravity and electric field force a force balance is only obtained in the sheath of the lower
electrode, where the upward electric field force is strong enough to balance the particles
against gravity. Since the electric field in the sheath increases towards the electrode
there typically is only a single position where electric field force and gravity balance.
Thus, horizontally extended, but vertically restricted dust arrangements are possible (see
Figs. 3.6a and 3.7a).
For smaller particles in the nanometer range or for large particles under microgravity
conditions gravity is not important. Smaller electric fields are sufficient to levitate and
trap the particles. Thus, particles can be trapped in the plasma bulk. There, the electric
field force is pointing into the plasma bulk for negatively charged particles, whereas ion
drag (and also the thermophoretic force) usually point outward. Thus, trapping of dust
particles should be possible in the entire plasma volume and three-dimensional extended
dust clouds should be formed (see Fig. 3.6b). However, it is found that for nanometer
particles or under microgravity conditions large regions without dust particles, so-called
“voids”, exist which are assumed to be due to the interaction of the ion drag and electric
field force [39, 40, 41], see Fig. 3.7b,c). Recently, 3D spherical dust clouds have been
formed in the laboratory by a combination of thermophoretic forces (to balance gravity)
and a confinement by electric forces using dielectric materials (glass), see Sec. 8.4.
40 3. Forces and Trapping of Dust Particles

Figure 3.7: a) Plasma crystal with two layers in the sheath of a gas discharge (top view
and side view), (b) cross section through an extended 3D dust cloud of microspheres under
microgravity conditions, and (c) dust clouds of nanometer-sized dust in a silane discharge
(the dust appears as white clouds).

3.8. Vertical Oscillations

We have seen that large, micron-sized particles are trapped in the plasma sheath due to
a balance of electric force and gravity. However, the charging processes are often difficult
to describe when all processes like the oscillating rf-sheath, ion-neutral collisions and all
the various charging currents are taken into account. Here, vertical oscillations provide
a useful and essential tool to measure the charge of the dust particles. Additionally, the
oscillations allow to extract the electric environment of the plasma sheath.
In the sheath where the particles are trapped the electric fields are generally spatially
dependent, i.e. E = E(z). The electric field usually increases monotonously from the
plasma-sheath boundary, where the electric field is small, towards the electrode where it
is largest. The dust charge will also be spatially dependent due to the space-dependent
densities and velocities of electrons and ions in the sheath. Typically, a unique equilibrium
position z0 exists where the particles are trapped (see Fig. 3.8c), i.e.

Q(z0 )E(z0 ) = md g . (3.19)

The equation of motion for a particle in the vertical direction (relative to the equilibrium
position) is then given by
Q(z)E(z)
z̈ + βż + = Fext , (3.20)
md
where β is the friction coefficient describing the neutral gas drag [Eq. (3.14)] and Fext are
other external forces applied to the particle. To solve this equation of motion, one has to
3.8. Vertical Oscillations 41

consider that the dust charge generally depends on the plasma conditions and therefore
is itself a dynamic variable. In the following, we will investigate vertical oscillations of
dust particles in the sheath for various situations.

3.8.1. Linear Resonance and Charge Measurement

In the most simple picture, vertical resonances can be treated in the following way: For
small deviations from the equilibrium position the particle charge can be assumed to
be spatially constant Q(z) = Q0 and the electric field as linearly increasing E(z) =
E0 + E1 (z − z0 ). Here,
∂E e
E1 ≡ = (ni − ne )
∂z 0
is the slope of the electric field (neglecting any horizontal variations). When the charge
density difference ni − ne is constant (the so-called “matrix” sheath model) the slope of
the electric field is constant. (see Fig. 3.8c). Such a linear electric field model is supported
by a number of simulations of rf discharges [42] and theoretical analysis [43]. Under these
assumptions the equation of motion reads
Q0 E1
z̈ + βż + z = Fext .
md

Figure 3.8: a) Scheme of the experimental setup in a typical experiment on complex

plasmas. The particles are illuminated by vertical and horizontal laser sheets. The particle
motion is recorded from top and from the side with video cameras. b) Electron micrograph
of the MF particles typically used in the experiments. c) Trapping of the particles in the
sheath of an rf discharge. See text for details.
42 3. Forces and Trapping of Dust Particles

This equation is just that of a damped harmonic oscillator where the microspheres are
trapped in the harmonic potential well [44, 45]
1 1
md ω2res (z − z0 )2 = Q0 E1 (z − z0 )2 (3.21)
2 2
with the resonance frequency of
Q0
ω2res = E1 . (3.22)
md
With external periodic excitation the resulting resonance curve is known from the damped
oscillator as
1
A(ω) = A0 q , (3.23)
(ω2res − ω2 )2 + β2 ω2

where A0 /ω2res ist the oscillation amplitude of the particle for ω → 0. The measurement
of the vertical resonance of a trapped particle gives us a handle on the determination of
the charge-to-mass ratio and allows to determine the particle charge Q0 , if the mass of
the particles is known [44, 45].
The assumption of a constant space charge density e(ni − ne ) = 0 E1 models the
linearly increasing electric field (with the slope E1 ) in agreement with simulations of the
rf sheath [42]. This allows the connection of the sheath electric field to the ion density
measured by Langmuir probes in the bulk plasma. For a high voltage sheath the time
averaged electron density ne = αni is just the fraction α of the rf period where the particle
“sees” a quasineutral environment due to the periodic rf oscillation of the plasma sheath,
as discussed in Section 2.5.1.. For the condition of micron-sized particles in an rf sheath
a rough estimate is α ≈ 1/3 [45]. This results in the following form of the resonance
frequency used to analyze the experiments
Qni (1 − α)e
ω2res = . (3.24)
0 md
The charge measurements have been performed using monodisperse MF (melamine
formaldehyde) microspheres (see Fig. 3.8b), which are perfectly spherical and have a very
low mass dispersion. Therefore, the particles have the same charge and are trapped at
the same height in the sheath. A number of vertical resonance measurements have been
made using these particles [45, 46]. The vertical oscillations were driven by applying a
very low-frequent modulation of the electrode rf voltage, see Fig. 3.9a. In doing so, the
sheath width is modulated and the particle is forced to oscillate vertically in the trapping
potential well.
From a frequency scan, a resonance in oscillation amplitude was obtained near 20 Hz
for dust particles of 2a = 9.47 µm diameter, see Fig. 3.9b. The measured data points are
fitted with a resonance curve according to Eq. (3.23). From the fit the parameters A0 ,
3.8. Vertical Oscillations 43

Figure 3.9: Measuring the charge on MF microspheres. a) Experimental setup for excita-
tion of resonances by rf voltage modulation and laser manipulation. b) Resonance curves
obtained for a 9.47 µm MF particle for excitation by electrode voltage modulation and by
laser pressure.

the amplitude at very low frequencies, β and ωres are obtained (Here: A0 = 5ω2res a.u.,
β = 26 s−1 and ωres /(2π) = 19.9 Hz). Applying this technique under different plasma
conditions, from Eq. 3.22 the corresponding particle charges are found to be about 10 000
elementary charges and the floating potential is about 3 V (see Fig. 3.10). Estimations
based on OML charging according to Eq. (2.16) result in charges that agree with the
measured values within a factor of 2-3. It is seen that the dust charge slowly increases
from about 6000 to 11000 elementary charges with decreasing pressure (120 to 40 Pa).
This cannot be explained by pure OML charging, since the defining quantity, the electron
temperature, is nearly constant (Te = 2.2 eV) in that pressure range. Rather, it is the
charge reduction due to the ion-neutral collisions as described in Sec. 2.4.4.
The width of the measured vertical resonance peak is determined by the neutral gas
drag on the particle and is in quantitative agreement with the Epstein [33] friction coef-
ficient β in Eq. (3.20). For the above mentioned experiment (at a gas pressure of 70 Pa)
the expected friction coefficient is calculated to be in the range between β = 20 s−1 and
29 s−1 , depending on the parameter δ, the measured value is β = 26 s−1 .
An alternative, non-invasive technique to manipulate dust particles is by means of a
focused laser beam as described in Sec. 3.6. in more detail. Here, one should note that
the laser beam pushes the particle in the direction of the beam. By switching the laser
“on” and “off” the vertical resonance curve of can be excited and measured (see Fig. 3.9).
The resonance excited by the laser technique is nearly identical with the electrode manip-
44 3. Forces and Trapping of Dust Particles

Figure 3.10: Measured dust charge as a function of discharge pressure. The uncertainty
in the measured values is due to the uncertainty of the ion matrix sheath model, i.e. the
uncertainty in the value α. From [46, 47].

ulation. With laser excitation additional spurious resonances at ωres /2, ωres /3 etc. are
excited due to the square wave excitation (laser “on” and “off”) compared to the sinusoidal
excitation at the electrode. Since the non-invasive laser technique and the electrode mod-
ulation give almost identical results it can be concluded that applying a small-amplitude
potential modulation to the electrode does not lead to a severe disturbance of the plasma
sheath environment.
Recently, the sensitivity of the vertical resonance method has been considerably im-
proved by taking into account the phase relation between the exciting force Fext and the
particle oscillations [48].

Other Charge Measurements

Other in-situ charge measurements make use of the particle-particle interaction by the
analysis of wave propagation or particle-particle collisions. These methods require a simul-
taneous measurement of the shielding length which also defines the interaction strength.
The methods and their results will be presented in Sec. 5.1., but it should already be men-
tioned here that the charges obtained from particle-particle interaction are very similar
to those of the resonance method.
In ex-situ charge measurements particles are dropped through a discharge into a Fara-
day cup where their charge is measured. In such experiments the influence of different
charging mechanisms like electron beams or UV radiation was investigated, see Fig. 3.11.
Indeed, with increased UV radiation positive dust is observed.
3.8. Vertical Oscillations 45

Figure 3.11: (right) Scheme of the experimental setup. The particles fall from the dropper
into the Faraday cup, where the dust charge is measured. When the particles are irradiated
by a strong UV source (left) the particles will charge positively. With a photoemitting
cathode present, the particles are charged negatively due to the electrons released by this
cathode (middle). From [49].

3.8.2. Parametric Resonances

The next type of vertical oscillations is given by parametric resonances. Parametric
resonances occur when the confining potential is periodically modulated. A paradigm for
parametric resonances is the children’s swing.
Parametric resonances in dusty plasmas can be excited by placing a wire in the plasma
sheath close to the dust particles. When applying sinusoidal electric potentials to the wire
vertical oscillations are driven. The resulting resonance curve is shown in Fig. 3.12. It is
seen that for small excitation voltages a single vertical resonance at ωres as in the linear
case is driven. At higher amplitudes, however, a second resonance at 2ωres suddenly
appears. This second resonance is a clear indication of parametric resonance. Thus,
driving the particles at double frequency also leads to large vertical particle oscillations
[50].
Parametric resonances occur when the strength of the external confinement is modu-
lated periodically. The equation of motion then reads

z̈ + βż + ω2res (1 + h cos ωt) = 0 , (3.25)

46 3. Forces and Trapping of Dust Particles

Figure 3.12: Parametric excitation of the vertical resonance. (a) Resonance curves with
the appearance of the second resonance at higher excitation voltages. (b) Amplitude of the
second resonance versus excitation voltage and (c) critical excitation voltage as a function
of gas pressure.

where h is the modulation depth. That means that the strength of the confinement and
thus the resonance frequency ωres changes periodically with a frequency ω that generally
is different from ωres . In our case, the resonance frequency ωres is due to the confinement
of the particle in the sheath by gravity and electric field force as discussed above. The
modulation at ω is due to the electrostatic potential on the wire.
This equation is known as Mathieu’s equation in mechanics. It is known, that para-
metric resonances occur when the modulation frequency ω is close to ωres or 2ωres . When
friction is present (i.e. β > 0) the second resonance occurs only when the modulation
depth exceeds a threshold value. This is also seen in Fig. 3.12 where a certain excitation
amplitude is needed to excite the second resonance. This excitation threshold also in-
creases with increasing discharge pressure, i.e. gas friction. Such a behavior is expected
from a parametric oscillator.
Following that reasoning, the occurrence of parametric resonances means that the
external confinement is modulated and disturbed by applying an electrostatic potential
to the wire, a situation that is not observed by modulation of the rf voltage. Thus, wire
3.8. Vertical Oscillations 47

Figure 3.13: a) Calculated particle potential in the sheath using the OML model and
a standard sheath model for the electron and ion densities and velocities , b) Nonlinear
resonance curve for a 9.47 µm particle. Note the hysteresis in the resonance for increasing
and decreasing frequency. c) Linear electric field and position dependent particle potential
profiles that will lead to the observed nonlinear resonance for three different particle sizes.
From [51].

excitation has to be used with great care in dusty plasma experiments.

3.8.3. Nonlinear Resonances

Now, we like to turn to nonlinear vertical particle oscillations. As outlined in Eq. (3.20) a
position-dependent charge and/or higher than linear terms in the electric field mean non-
linear modifications to the harmonic potential well. Obviously a non-harmonic potential
well leads to nonlinear resonances. The electric field is only linear when the difference in
charge densities of electrons and ions is constant (matrix sheath). If that does not hold
the electric field will vary in a nonlinear way.
The dust charge in the sheath can be position-dependent due to two counteracting
effects, see Fig. 3.13a). As an illustration, the charge is calculated using the OML model
with streaming ions according to Eq. 2.10 together with a simple dc model of a space
charge sheath to account for varying electron and ion densities and velocities. In the
presheath the ions are accelerated to Bohm velocity. At the sheath edge (z ≈ 0.5 mm)
the ion velocity further increases by acceleration in the sheath electric field. This leads
to a less efficient decharging of the dust grains by the ions and thus a higher (negative)
potential, i.e. charge, on the particle. On the contrary, deep in the sheath the electron
density is considerably reduced leading to smaller electron current and, hence, less negative
dust particles, see also Fig. 2.3.
To analyse nonlinear oscillations, we assume that Q(z) and E(z) can be expanded in
48 3. Forces and Trapping of Dust Particles

polynomials according to

E(z) = E0 + E1 z + E2 z2 + E3 z3 . . .
Q(z) = Q0 + Q1 z + Q2 z2 + Q3 z3 . . . (3.26)

Then, the equation of motion becomes nonlinear (up to 3rd order is considered, here,
corresponding to a potential well of 4th order) with

z̈ + βż + C1 z + C2 z2 + C3 z3 = Fext . (3.27)

The coefficients of nonlinearity are given by

1
C1 = = (Q0 E1 + Q1 E0 )
m
1
C2 = = (Q0 E2 + Q1 E1 + Q0 E2 )
m
1
C3 = = (Q0 E3 + Q1 E2 + Q2 E1 + Q3 E1 ) (3.28)
m
The particle potential can then be written as
1 1 1
V(z) = mC1 z2 + mC2 z3 + mC3 z4 . . . (3.29)
2 3 4
Here, C1 corresponds to the (modified) linear resonance frequency, compare Eq. (3.21),
C2 to an up/down asymmetry of the vertical potential well and C3 to a weaken-
ing/strengthening of the potential well with larger oscillation amplitudes. Nonlinear
resonance curves have been excited in experiments at quite low gas pressure and high
excitation amplitudes. In experiments of Ivlev et al. [52] the nonlinear oscillations were
driven by applying a voltage to a wire placed below the dust particle. In contrast, Zafiu
et al. [51] used the rf voltage modulation at the electrode.
The measured resonance curve shows distinct features of nonlinearity, see Fig. 3.13b.
First, the resonance curve is not smooth. At certain excitation frequencies there is a jump
in the oscillation amplitude from small to large values. Second, the resonance curve is bent
towards lower frequencies and, third, there is a pronounced hysteresis in the resonance
curve when going from smaller to larger frequencies or the opposite way.
The main resonance frequency is about 13 Hz, here, which corresponds to that of
the linear resonance under these discharge conditions. The bending of the resonance
towards lower frequencies is a clear indication that the potential well gets weaker with
increasing amplitude, i.e. C3 < 0. (For C3 > 0 a bending towards higher frequencies
would be expected.) This is easily seen from the following reasoning: Near the resonance
the particle has a large oscillation amplitude and “feels” also the nonlinearity C3 . When
the potential gets weaker with increasing amplitude, the particle feels weaker restoring
forces and the resonance frequency is then shifted towards lower frequencies. At lowered
3.8. Vertical Oscillations 49

Figure 3.14: Self-excited vertical oscillations due to a position dependent charge and
delayed charging.

frequencies then again higher oscillation amplitudes can be sustained that further decrease
the resonance frequency. In this way, a bending towards lower frequencies is observed.
By comparing the measured resonance curve with calculated resonances using the
equation of motion (3.27) the coefficients of nonlinearity can be determined. These co-
efficients can then be related to position dependent dust charges or electric field profiles
that may range from linear via parabolic to cubic. When using different particle sizes
different regions of the sheath can be probed resulting in a more or less consistent set of
parameters Qi and Ei . Zafiu et al. found best agreement among experiments with different
particle sizes for a linearly increasing electric field and a position dependent dust charge
see Fig. 3.13c. As mentioned above, the position dependent dust charge is due to the
increased ion stream and the reduced electron density in the sheath. Position-dependent
dust charges have been reported recently in experiments using hypergravity conditions
[53].
The idea of a time and space dependent charge is supported by the observation of self-
excited vertical oscillations, see Fig. 3.14. There vertical particle oscillations have been
observed that grow in time without external drivers [54]. The growth time is of the order of
10 s. The oscillations reach a large amplitude until the particles drop from the discharge.
Such oscillations are only possible if a source of energy is provided that can overcome
the energy loss by friction with the neutral gas. A possible energy gain mechanism can
lie in the combination of a position-dependent dust charge and finite charging times.
When during the oscillation the dust particle has an instantaneous charge Qt (z) which is
different from the equilibrium
R charge Qeq (z) the restoring force F = Qt (z)E(z) and thus
the restoring energy F ds is different from the equilibrium situation. The difference is
50 3. Forces and Trapping of Dust Particles

illustrated by the shaded area in Fig. 3.14. Careful analysis [52] of this delayed charging
has shown that indeed such a mechanism could overcome the energy loss by gas friction
at low gas pressures. These self-excited oscillations and the nonlinear resonances clearly
identify the dust charge as a dynamical variable.
 51

4. Coulomb Crystallization
After identifying the basic mechanisms of charging and trapping of dust particles in a
discharge, we now like to investigate the many-particle interaction of the dust in view of
crystallization of the dust ensemble.

4.1. The One-Component Plasma (OCP)

At first, we like to investigate how systems of many charged particles behave under their
mutual influence. The simplest situation that can be considered is that of point charges
immersed in a homogeneous neutralizing background. As early as in the 1930s, Eugene
Wigner [55] has discussed the situation of ions in a “sea” of electrons to study the crys-
talline order in metals. In dusty plasmas, we obviously identify the point charges with the
dust particles and the neutralizing background with the plasma ions. In such a system
with a homogeneous background the point charges interact by pure Coulomb repulsion.
The background is only necessary to ensure overall charge neutrality.
Such a system of pure point charges is described by only a single parameter, the
Coulomb coupling parameter
Q2 1
Γ= (4.1)
4π0 bWS kT
that describes the electrostatic interaction of neighboring dust particles in terms of their
thermal energy. Here, T is the temperature of the point charges and Q is their charge.
As a measure of the interparticle distance the Wigner-Seitz radius bWS = (3/4πn)1/3 is
defined (n is the density of the point charges). For example, in a simple cubic structure
with particle separation b the Wigner-Seitz radius is bWS = 0.62 b.
A charged particle system is said to be strongly coupled when Γ > 1, i.e. when the
electrostatic energy of neighboring particles exceeds the thermal energy. For a usual
electron-ion plasma Γ is much smaller than unity, e.g. one finds Γ = 8 · 10−3  1 for ions
at Ti = 0.03 eV and ni = 1 · 109 cm−3 and even less for electrons due to their (typically)
higher temperature.
From simulations of charged-particle systems it was found that the point charges
arrange in ordered crystalline arrangements when the coupling parameter exceeds a critical
value of Γc = 168 ± 2 [56]. For Γ < Γc the particles are in a fluid (= liquid or gas-like)
state. Note, that a purely repulsive system has only a solid-fluid transition, there is no
liquid-gas transition. A liquid-gas transition requires an attractive part in the interaction
potential.
So, if the Coulomb energy by far dominates the thermal agitation the particles are
forced to crystallize, a process known as Wigner or Coulomb crystallization. In 3D, the
52 4. Coulomb Crystallization

Figure 4.1: Crystal structures in 3D (a) bcc,(b) fcc and Figure 4.2: Crystal struc-
(c) hcp ture in 2D: hexagonal struc-
ture

particles arrange in a bcc lattice (see Fig. 4.1) which is the minimum energy configuration
for pure Coulomb interaction.
The minimum energy configuration is determined from the sum of the electrostatic
energies between all particles, the so-called Madelung energy,

1 X Q2
N
U= , (4.2)
2N i6=j 4π0 rij

where rij is the relative distance between particle i and j. From detailed calculations the
Madelung energy is found for pure Coulomb interaction as

Ubcc = −0.895 929 255 682 for the bcc structure

Ufcc = −0.895 873 615 195 for the fcc structure
Uhcp = −0.895 838 120 459 for the hcp structure

in units of Γ = Q2 /(4π0 bWS ). The energies for the different lattice types are very close
to each other, but the bcc structure is the one with the lowest energy (see Fig. 4.1 for the
different lattice types).
In 2D systems crystallization takes place at the critical value Γc = 125 [56]. In 2D,
the Wigner-Seitz radius is analogously defined as bWS = (πn)−1/2 . The minimum energy
configuration is the hexagonal structure as shown in Fig. 4.2. This is the structure you
would expect from densely covering a table with coins.

4.2. Yukawa Systems

When shielding by the ambient plasma electrons and ions is taken into account the in-
teraction potential between the point charges is usually described by the Debye-Hückel
4.2. Yukawa Systems 53

potential (also named Yukawa potential, especially in the field of complex fluids)
Q r
 
φ(r) = exp − , (4.3)
4π0 r λD
where λD is the Debye shielding length
!−1/2 s
1 1 0 kTe,i
λD = + 2 with λDe,i = . (4.4)
λ2D,e λD,i ne,i e2
Besides the Coulomb coupling parameter Yukawa systems are characterized by a second
parameter, the so-called screening strength, κ = bWS /λD which is the Wigner-Seitz dis-
tance bWS in units of the Debye length. The OCP-limit is obtained again for infinite
screening length, i.e. for κ → 0.
The phase diagram of the Yukawa system is shown in Fig. 4.3. First, it is seen that
the critical value for the solid-fluid transition strongly depends on the screening strength,
i.e. Γc = Γc (κ). The melting line increases almost exponentially with κ. Due to the
exponential shielding a much higher Coulomb coupling parameter is required to enter the
crystalline regime. For κ = 0 the OCP value of Γc = 168 is retrieved [58, 57].

Figure 4.3: Phase diagram of the Yukawa system in the Γ -κ plane. The melting line
increases almost exponentially with the screening strength. In the solid phase two different
crystal structures, bcc and fcc are found. After [57].
54 4. Coulomb Crystallization

The solid phase itself shows two different crystalline structures: for lower values of
κ the bcc structure as in the OCP is found. For stronger screening the fcc structure is
obtained. For increasing screening strength the Yukawa interaction becomes more and
more like a hard-sphere potential. The packing density for spheres in the bcc structure
is 68 % whereas it reaches 74 % in the fcc structure (the fcc structure is what one would
obtain from stacking oranges into several layers). Thus at higher screening an increased
packing density becomes more favorable.

4.3. Coulomb Crystallization in Dusty Plasmas

After the general introduction to strongly coupled systems we now investigate under which
conditions Coulomb crystallization in dusty plasmas will be possible. The following ideas
have already been presented in 1986 by Ikezi [59], years before plasma crystals have been
discovered experimentally. Ikezi’s model uses a few simplified assumptions which however
capture the main features of the problem. First, the melting line of Yukawa systems is
approximated by an effective Coulomb coupling parameter
Z2d e2
!
bWS
Γc,eff = exp − = Γ exp(−κ) = 168 ,
4π0 bWS kTd λD
where it is assumed that the melting line exactly increases exponentially with increasing
κ (Sometimes, the above definition is used as the coupling parameter, but usually Γ and
κ are treated separately). Above Γc,eff = 168 crystallization is achieved in Ikezi’s model.
Since λD ≈ λD,i from Eq. (4.4) only ion screening is considered. The second assumption
is concerned with the charging of the dust. For the dust charging the rule-of-thumb
expression according to Eq. (2.18) is used. The dust charge, however, is limited when all
electrons are bound to the dust (electron depletion), i.e.
ni
Zd,lim = e .
nd
Now, under which parameters of ion and dust density, Wigner crystallization in dusty
plasmas can occur? The ion density affects the screening length λD and the dust charge
limit. High ion density means high dust charge limit, but strong screening. The dust den-
sity also affects the dust charge limit and the interparticle distance b. High dust densities
mean low maximum dust charge, but small interparticle distance which results in high
coupling. Both parameters thus have counteracting effects. From these considerations,
Wigner crystallization should be possible in the dark area ABC in Fig. 4.4. The boundary
of this region is dominated by different mechanisms. From A to B, the dust density is
nearly constant. Also the charge Zd is constant, there are no depletion effects due to the
low dust density. On the entire upper boundary from A to C via B the particle is charged
to its single particle value (here a = 10 µm is assumed). From A to B, the Debye length
is much larger than the interparticle distance and screening effects are not important.
4.4. Crystallization in Bounded Systems 55

Figure 4.4: Existence diagram of Wigner crystals in dusty plasmas. In the dark area
Coulomb crystallization should be possible for a particle of 10 µm radius. In the total
shaded area crystallization can occur for arbitrarily sized particles. Note the wide loga-
rithmic scale on both axes. After [59].

Near B the situation changes. The ion density becomes so high that the Debye length
is now of the order of the interparticle distance and screening becomes dominant. Thus
the boundary bends towards much higher dust densities and thus smaller interparticle
distances until point C is reached. On the boundary from C to A there are relatively
large dust densities and low ion densities. Here, depletion effects become dominant. The
charge on the dust is determined by the available free electrons thus limiting the coupling
parameter.
Although this is a quite crude model it shows that Coulomb crystallization in dusty
plasmas is possible in a range of ion and dust densities that is several orders of magnitude
wide. For typical plasma discharges with ni = 109 to 1010 cm−3 plasma crystals should
exist for dust densities in the range from nd = 103 to 105 cm−3 .

4.4. Crystallization in Bounded Systems

In the previous sections, the formation of Wigner crystals in extended 3D (and 2D)
systems has been discussed. However, as we have seen, the plasma crystals in the experi-
ments (see Fig. 3.7) are extended in the horizontal plane, but are confined in the plasma
sheath by quite strong vertical fields. Similar confined systems are also found in ion trap
experiments.
56 4. Coulomb Crystallization

Figure 4.5: Layer formation and crystalline structure in a confined system: Vertical
position of the individual crystal layers. The respective crystal structure is indicated by
the three inserts at the top, where the structure is viewed from top. Different shades of
the circles denote different layers. After [60].

The situation of confined OCP systems has to capture features of both 2D and 3D
systems with competing planar hexagonal or volume bcc/fcc structures. To study possible
crystal structures in bounded systems let us consider a system of point charges which is
extended in the horizontal x − y plane, but is confined by a parabolic potential in the
vertical z-direction [60]. When the number of point charges is not too high the particles
will arrange in a monolayer 2D crystal with hexagonal structure as discussed above. When
4.5. Structural Information 57

the density of charges is increased (or, equivalently, the vertical confinement is weakened)
the monolayer system jumps to two layers, three layers and so on. This is an expected
behavior of repulsive particles. It is interesting to note here that the crystal structure
of the multi-layer system changes between square, bcc 110 and hexagonal (see Fig. 4.5).
Thus the possible crystal structures of the infinite 3D systems are also found in a system
with only a few layers.
The scenario is not much different if screening between the charges is taken into ac-
count [61, 62]. There, similar jumps to multi-layer systems with similar crystal structures
are observed. However, the exact transition points, of course, depend on the screening
strength.
Structures of systems confined in all spatial dimensions are described in Chapter 8.

4.5. Structural Information

To characterize the degree of order of the point charges and to define a system as fluid or
solid it is necessary to have quantitative measures. Here, such techniques are presented
with special emphasis on 2D systems, but they can be easily extended to 3D systems.

4.5.1. Wigner-Seitz Cells

One of the basic techniques to characterize a system of point charges is given by the
Wigner-Seitz cell analysis. There, the 2D plane is covered with polygons where each
particle is in the center of its own cell. The construction is as follows: for each particle
the perpendicular bisector of the connection to neighboring particles is determined. The
bisectors define a closed polygon around the particle of interest (see Fig. 4.6a). The main
advantage of this procedure is that it determines which particles in the vicinity of the
reference particle are the neighboring particles. Thus, the number of nearest neighbors is
also determined. For a 2D system, the mean number of neighbors is 6. However, particles
with 5 or 7 neighbors might exist which are termed defects. The relative abundance of
6-fold polygons is already a good qualitative hint whether a structure is crystalline or not.
A structure with more than 90 % of hexagons can be called crystalline (see Fig. 4.8).

4.5.2. Pair Correlation Function

The pair correlation function g(r) is the probability to find a particle at a distance r from
a chosen particle. This is then averaged over all (chosen) particles. One finds that
g(r) → 1 for r → ∞ ,
since one will always find a particle at large distances, and
g(r) → 0 for r → 0 ,
58 4. Coulomb Crystallization

Figure 4.6: a) Sketch of the Wigner-Seitz cell construction. b) Definition of the angle θ

since the particles cannot come infinitely close to each other.

The pair correlation function gives the mean interparticle distance as well as the near
and far order of the particle system. An example is shown in Fig. 4.7 for simulated OCP
systems at different coupling strengths. The first peak in g(r) is due to the neighboring
particles, further peaks due to overnext neighbors and so on. For a hexagonal structure√
you would expect a peak at the interparticle distance b, the next peaks would be at 3b
and 2b.
The more pronounced the peaks are and the more peaks can be identified the higher
is the order of the system which also can be seen from Fig. 4.7. For small values of the
coupling parameter (Γ ≈ 10) there is only a very small hump at the nearest neighbor
distance and the pair correlation function is nearly flat. This means that there is no
strong order in the system and one can find particles at all distances larger than the

Figure 4.7: Pair correlation function g(r) for various coupling parameters Γ of the OCP
system. The nearest neighbor distance is found to be at r ≈ 1.7bWS . From [56].
4.5. Structural Information 59

interparticle distance. When Γ is increased the nearest neighbor peak grows and also the
peaks of overnext neighbors (and so on) grow, indicating increased order of the system.
This can also be easily seen in the comparison of solid, fluid and gas-like states in Fig. 4.8.
One might expect that at very high coupling parameters Γ  Γc the particles should be
at the exact lattice sites and the pair correlation function should collapse to single sharp
δ-peaks. This is indeed true for 3D systems, but not for 2D. It was derived from very
basic principles that there are always long-range fluctuations in 2D systems that destroy
order over very long distances. The form of g(r) can be analyzed more quantitatively
which is of interest for the analysis of phase transitions in 2D systems, but that will not
be explicated here.

4.5.3. Structure Factor

In condensed matter physics, the pair correlation function usually cannot be determined
directly. Rather, the structure factor is determined from scattering of x-rays or neutrons.
In dusty plasmas, we are able to measure the pair correlation function directly and to
calculate the structure factor in order to compare with condensed matter experiments.
The structure factor S(q) is just the Fourier transform of the pair correlation function
Z
S(~q) = 1 + nd (g(r) − 1) exp(i~q · ~r) d~r , (4.5)

where ~q is the wave vector of the scattered radiation. In 2D, this can be written as
Z
2π
1 X 1 1 X
N N
S(q) = exp(iqrij cos ϕ) dϕ = J0 (qrij ) ,
N ij 2π N ij
0

where ϕ is the angle between ~r and ~q and J0 is the zero-order Bessel function. The
calculated structure factor is also shown in Fig. 4.8.
A large first peak in the structure factor means long-range periodicity of the pair
correlation function. Thus, a peaked structure factor corresponds to long-range order.

4.5.4. Angular Correlation Function

Finally, the angular correlation will be discussed, here. Until now, only the translational
order has been characterized. In contrast, the angular correlation function measures how
the bond angles between nearest neighbors are oriented relative to each other as a function
of distance between the bonds. The angular correlation is defined as
g6 (r) = hexp(6i[θ(r) − θ(0)])i , (4.6)
where θ is the angle of a nearest neighbor bond relative to a fixed axis (see Fig. 4.6b).
The factor 6 takes into account the presumed hexagonal order of the system and h·i
60 4. Coulomb Crystallization

denotes the average over the particle arrangement. When g6 (r) is close to 1 the nearest-
neighbor bonds at a relative distance r are oriented along the same direction. The bond
orientational order is destroyed by defects (particles with 5 or 7 neighbors). Behind defects
the lattice orientation is different from the starting point. A value of g6 close to 0 means
that the bonds are randomly arranged (no correlation), a value of g6 = −1 describes
anti-correlation: the bonds have a relative orientation which differs by 30◦ which is the
maximum difference in angle for a 6-fold symmetry. The angular correlation function of
our model systems is also shown in Fig. 4.8. The angular correlation function also plays
a large role in the identification of melting processes in 2D.

4.5.5. Example
The different techniques to characterize a particle arrangement is illustrated for three
model systems (see Fig. 4.8) that reflect crystalline, fluid (Γ > 1) and gas-like (Γ <
1) order (it should be noted again, here, that for a system with only repulsive forces
like OCP or Yukawa there is only a solid-fluid transition). In the solid state the pair
correlation function is very pronounced. Pair correlation can be observed at least up to
7 interparticle distances. Correspondingly, the structure factor is large and sharp. The
angular correlation function is large and decays only slowly with distance. In the fluid
state, one can identify the nearest-neighbor and overnext-neighbor peak in g(r). The
structure factor thus is smaller and not that sharp. Also the angular correlation rapidly
decreases. The gas-like state exhibits no correlation at all. The pair correlation is flat
(compare Fig. 4.7 at Γ = 10), the structure factor is flat and there is no angular correlation,
even for the smallest distances. This reflects the gas-like characteristics of this particle
arrangement.

4.6. 3D Crystals
The previous discussions were related to some extent to 2D crystals (or crystals with a
few layers) since they are easily produced in laboratory experiments. Nevertheless, three-
dimensional dust crystals have been observed under microgravity conditions aboard the
ISS [63] where the particles are not forced into the space charge sheath, but remain in
trapped in the plasma volume. Figure 4.9 (a) shows the reconstruction of the 3D positions
of about 10 000 particles forming an ordered arrangement. The order is illustrated by the
pair correlation function in the inset indicating the strong peak of nearest neighbors and
two subsequent peaks.
The identification of local order is much more difficult in 3D than in 2D due to the much
more possibilities of particle arrangements. In this experiment, it was tried to identify
the local crystal structure. For that purpose, for each particle of the crystal a local order
parameter q4 and q6 was calculated that accounts for local 4- and 6-fold order using the
4.6. 3D Crystals 61

Figure 4.8: Structure of solid and fluid and gas-like structures, (a) Wigner-Seitz cell
construction. For the solid and liquid state the defects are shaded (light grey: 5-fold
defect, dark grey: 7-fold defect). (b) Pair correlation function g(r), (c) structure factor
S(q) and (d) angular correlation function g6 (r).
62 4. Coulomb Crystallization

Figure 4.9: (a) 3D dust crystal under microgravity conditions aboard the ISS. The inset
shows the pair correlation function before and after full crystallization (green and red
curve, respectively). (b) Color-coded (from dark to light) is the number of particles with
local order parameters q4 , q6 indicating local 4- and 6-fold symmetry. From [63].

order properties of spherical harmonics [63]. The abundance of particles in the crystal
with local order parameters q4 , q6 is given in Fig. 4.9 (b) where also the order parameters
of ideal hcp and fcc lattices (and also bcc which is not shown here) are indicated (compare
also Fig. 4.1). It is seen that quite a substantial fraction of the particles are located close
to the ideal hcp and fcc order parameters indicating high crystallinity (either as hcp or fcc
lattice). However, also a certain fraction of the particles are randomly oriented showing
that the particles are not fully crystallized.
In the plasma volume the forces acting on the particles are much smaller than in
the sheath. Hence also the interparticle forces are usually much smaller since they have
to compete with much smaller confining forces. Hence, such 3D crystals are usually
much softer than those in the sheath (although those are already very soft). Hence,
crystallization occurs on a very, very long time scale. Hence, formation (and observation)
of 3D crystals is an interesting and difficult problem.
 63

5. Dust Particle Interaction

In the previous chapters we have seen how a plasma crystal of microspheres trapped in the
sheath of the discharge is formed. We have assumed that the interaction is described by
a Coulomb or Debye-Hückel repulsion. Whether the interaction can really be described
by such a shielded Coulomb potential will be discussed first for particles in the same
horizontal layer. The vertical interaction will be discussed afterwards. We will learn
that the interaction among dust particles trapped in the sheath of a discharge is dra-
matically different in the horizontal and in the vertical direction due to the dust particle
environment.

5.1. Interaction in the Horizontal Plane

A close inspection of a 2D plasma crystal trapped in the sheath of a plasma discharge (see
Fig. 3.7a) reveals very interesting details. In the horizontal plane, the particles arrange in
a hexagonal structure (see inset), where each dust particle has six nearest neighbors. We
have seen, that such a hexagonal structure is a minimum energy configuration in mono-
layer or few-layer systems for a repulsive Coulomb or Debye-Hückel interaction between
the dust particles.
A direct experimental determination of the horizontal interaction potential V between
microspheres has been achieved from central collisions of two particles by Konopka et
al. [64]. For this purpose two microspheres (2a = 8.9 µm) are trapped in the sheath of
an rf discharge. The horizontal confinement was due to a copper ring placed onto the
electrode. The copper ring leads to an additional horizontal parabolic confinement (of
strength ω0 ) which was confirmed by the analysis of the motion of a single particle. One
of the two microspheres was sitting at rest in the center of the confining potential, the
other was pulled “uphill” by a positively biased probe wire and was subsequently released
(see Fig. 5.1a).
The equation of motion for the two particles reads (using the relative coordinate xr
and the reduced mass µ = m1 m2 /(m1 + m2 ) = md /2)
∂V(xr )
µẍr + µβẋr + µω20 xr = − , (5.1)
∂x
where V is the interaction potential to be determined. The integration of the equation of
motion gives
Zt
1 1 1
2 2
V(xr ) = md ẋr + md ω0 xr + md β ẋ2r (τ)dτ ,
2
(5.2)
4 4 2
0
64 5. Dust Particle Interaction

Figure 5.1: Konopka’s two-particle collision experiment: a) Scheme of the experimental

arrangement, b) electrostatic interaction energy as a function of relative particle distance
xr . Symbols: experimental results, Lines: Best fits of a Debye-Hückel (Yukawa) potential.
The two curves A and B correspond to different discharge conditions (see text). After
[64].

where the the first term is the kinetic energy of the two-particle system (using the reduced
mass), the second describes the potential energy in the horizontal parabolic confinement
and the third accounts for losses by friction with the neutral gas. The interparticle
potential is measured from the relative particle positions and velocities which are easily
determined from video data. This enables to reconstruct the interaction potential with
good accuracy from the particle trajectories.
The experiments [64] show that the so obtained electrostatic energy directly reflects
the interaction potential between the microspheres that can be described very accurately
by a Debye-Hückel (Yukawa) type interaction
Z2d e2 xr
 
V(xr ) = Zd eφ(xr ) = exp − , (5.3)
4π0 xr λD
as shown in Fig. 5.1b, where a Debye-Hückel interaction was fitted to the experimental
results. From the fit the following values of the charge and shielding length have been
obtained: Zd = 13 900, λD = 0.34 mm (for case A at high discharge voltage Upp =
233 V), Zd = 17 100, λD = 0.78 mm (case B, low voltage Upp = 64 V). As expected
the Debye length increases for lower plasma density (i.e., lower discharge voltage). The
particle charges are in the range of the expected values from the charging theory and are
comparable to the measurements using the resonance technique. The screening length is
of the order of the electron Debye length in these discharges. This might be expected.
However, that point has to be discussed in some more detail later.
The horizontal interaction can also be derived from wave experiments. The dispersion
relation of waves contains the particle-particle interaction. Wave dispersion, and thus
5.2. Vertical Interaction 65

particle interaction, will be discussed later in Chapters 6 and 7.

5.2. Vertical Interaction

When returning to the plasma crystal in Fig. 3.7, the side view reveals that the particles
of different vertical layers are located directly atop of each other. They are vertically
aligned. This unusual ordering is practically always observed for two or more vertical
layers and has been recognized early after the discovery of the plasma crystal [65, 45].
The vertically aligned strings move as an entire chain in the crystal. This demonstrates
that in the vertical direction additional attractive forces must be present. Possible sources
and mechanisms for the formation of additional attractive forces will be described now.
Clearly, the observed vertical alignment of the microspheres certainly is not a minimum
energy configuration for purely repulsive particle interactions. The fundamental reason
for the difference between horizontal and vertical interactions lies in the fact that the
main external forces on the microspheres – electric field force and gravity – act in the
vertical direction. Due to gravity the microspheres are trapped in the non-neutral, non-
equilibrium environment of the sheath which strongly influences the interaction between
the microspheres.
For the very large and rod-like particles used in the experiment of Mohideen et al. [66]
and Annaratone et al. [67] dipole moments on the particles might explain the observation
of particle alignment along the electric field. However, these dipole effects are much too
small to overcome the repulsion between the particles used in the experiments, here (see
also Sec. 3.2.2.). Hence, other mechanisms have to be taken into account.
The ions in the sheath are the most obvious candidate as the source for the generation
of attractive forces. Different mechanisms of ion-mediated attractive forces have been put
forward in a number of theoretical models and simulations. Here, we will now discuss
the formation of attractive forces from the viewpoints of a “wave” model and a “particle”
model.

5.2.1. The wakefield potential (wave model)

In this class of models, the attraction is due to wakefield formation downstream the
particles by a (supersonic M > 1) ion flow [68]. Here, it is assumed that the dust particles
excite ion-acoustic waves of all frequencies in the ion stream. The ion-acoustic waves
superpose to form an attractive resonance below the dust particles where the ion density
is increased with respect to the background level. In this approach the interaction between
the ion flow and a single dust particle is described with the help of the dielectric function
. Here, the ion flow velocity vi0 is assumed to be in the z direction with M = vi0 /vB
being the Mach number of the flow (vB is the Bohm velocity). The electrostatic potential
66 5. Dust Particle Interaction

Figure 5.2: a) The ion wake potential in an ion flow. Downstream of the particle an
oscillating potential is formed with alternating attractive (φc > 0) and repulsive (φc < 0)
regions. After [68]. (b) Spatially resolved wakefield in the direction of the ion flow (z) and
perpendicular to the flow (ρ). The potential contours are indicated (solid lines: negative
potentials, dashed lines: positive potentials). The shaded area has a positive potential and
is thus attractive for a second negative dust particle. Note the different axis scaling of z
in panels (a) and (b).

around the dust particle is given from linear response theory in general terms by [69]
Z
Zd e 1
φ(~r) = 3 2
ei~q·~r d~q , (5.4)
8π 0 q (~q, ω − qz vi0 )
where ~q (q = |~q|) is the wave vector of the excited ion acoustic waves and (~q, ω) is the
dielectric response of the plasma. Eq. (5.4) is just the Fourier notation of the particle
potential and the dielectric function describes the response of the plasma species (electrons
and ions) to the electrostatic potential. Here, we use a moving frame with the velocity of
the streaming ions. The frequencies are therefore Doppler shifted ω → ω − qz vi0 (qz is
the z component of the wave vector).
The plasma dielectric response is given by
1 ω2p,i
(~q, ω − qz vi0 ) = 1 + − , (5.5)
q2 λ2D,e (ω − qz vi0 )2
5.2. Vertical Interaction 67

where the second term on the RHS describes the electron shielding and the third term is
the ion response. For the electrons the low-frequency limit of the dielectric response is
used due to the high mobility of the electrons. For the ions the (Doppler shifted) high
frequency limit is taken.
After some algebra the inverse of the dielectric function is found as
q2 λ2D,e
"
ω2s
#
1
= 1 + , (5.6)
(~q, ω − qz vi0 ) 1 + q2 λ2D,e (ω − qz vi0 )2 − ω2s

where ωs = qvB /(1 + q2 λ2D,e )1/2 is the frequency of oscillations in the ion flow. For
comparison, the dielectric response of a Coulomb potential with screening by electrons
(i.e. a Debye-Hückel potential) is simply given by
1 q2 λ2D,e
= .
(~q, ω) 1 + q2 λ2D,e
By substituting the dielectric function in Eq. (5.4) the total potential can be written
as the sum of two potentials
φ(~r) = φD (~r) + φc (~r) (5.7)
where
Zd e r
 
φD (~r) = exp −
4π0 r λD,e
is the usual Debye shielding potential and
Z
Zd e q2 λ2D,e ω2s
Φc (~r) = ei~q·~r d~q . (5.8)
8π3 0 q2 1 + q2 λ2D,e (ω − qz vi0 )2 − ω2s
Using cylindrical coordinates (x, y, z) → (ρ, ϕ, z), an approximate solution on the vertical
axis ρ = 0, i.e. behind the dust particle, is given by [68]
 √ 
Zd e 2 cos z λD,e M2 − 1 q
D,e M − 1.
Φc (ρ = 0, z) = for z > 3λ 2 (5.9)
4π0 z 1 − M−2
According to this model an oscillating ion wake potential φc downstream of the dust
particle is created with an alternating sequence √ of regions with enhanced positive and
negative potential (see Fig. 5.2a). For z < 5λD,e M2 − 1 the potential is negative due
to the presence of the negative dust particle at z = 0. The potential (and the cor-
responding√ion density) then “overshoots” and forms an attractive potential between
5 < z/λD,e M2 − 1 < 8. This wake potential is attractive to a second negatively charged
particle when φc > 0. This ion wake provides the attractive force necessary to explain the
vertical ordering of the particles. This mechanism is similar to Cooper pairing in super-
conductors [70], in that the dust particle polarizes the surrounding medium, the plasma,
68 5. Dust Particle Interaction

which in turn leads to attraction of other particles. In addition, it is reasoned that other
dust particles will arrange in the areas of positive potential defined by the first particle.
The vertical scales introduced by this collisionless model are different from those in the
experiment: The model predicts that the lower √ particle would be found in the maximum
of φc which is approximately at z = 6λD,e M2 − 1 ≈ 3000 µm when λD,e = 500 µm is
assumed. This is far from the observed vertical distances of z = 400 to 600 µm.
Using a more advanced dielectric function including collisions (with the ion-neutral col-
lision frequency νi ) and Landau damping (through the application of a shifted Maxwellian
velocity distribution fi0 ) via
Z
1 ω2p,i ~q∂fi0 (~v)/∂~v
(~q, ω − qz vi0 ) = 1 + 2 2 − 2 d~v , (5.10)
q λD,e q qz vi0 − ω − iνi
a more realistic wakefield potential is derived [69]. This wakefield potential is shown in
Fig. 5.2b) in a plane parallel and perpendicular to the flow. The attractive potential
maximum is shifted more closely to the particle and is now found at about z ≈ 2λD,e
which agrees more closely with the experiment. The Landau and collisional damping also
leads to a rapid decay of the oscillations along the flow so that generally only a single
potential maximum behind the dust particle is obtained. Further, wake is V-shaped due
to the Mach cone of the ion acoustic waves excited by the grain.
The above discussed wave model included the excitation of linear waves by a single
dust particle. In a region near the dust the linear description might become invalid and
further the effects introduced by the presence of a second particle in the wakefield are not
addressed (which has been resolved in a recent article [71]).

5.2.2. The ion focus (particle model)

The problem of attraction will now be investigated on the microscopic particle (ion) level
[73, 74, 32, 75]. For this purpose, the experimental conditions are closely mimicked in
simulations: A two-layer plasma crystal in the sheath of an rf discharge is considered;
the electric field and electron distribution are taken into account as time-averaged, but
spatially dependent quantities.
The ion trajectories are calculated according to the following equation of motion in
the electric field of the sheath Esheath (z) and of the dust particles
X
md~¨r = e Esheath (z) −
 
∇φi (~r) ,

where φi (~r) is the particle potential of particle i according to Eq. (4.3) and the sum is
over all particles in the upper and lower layer. In the ion motion also collisions are taken
into account. This is done the following way: The above equation of motion is solved
between the collisions of the ions. The time between two collisions is chosen randomly
in such a way that a defined ion mean free path λmfp is ensured (see Fig. 5.3a). In the
5.2. Vertical Interaction 69

Figure 5.3: a) Ion density distributions calculated from the ion trajectories in the sheath
for a vertically aligned pair of microspheres. Charge exchange collisions are included in
this simulation as can be seen by abrupt changes in the particle trajectories. b) Ion density
distribution for vertically aligned pairs (left, same as (a), but without the trajectories)
and horizontally shifted crystal layers (right). The particle positions are indicated by the
arrows. After [72, 73].

collision the ion loses its kinetic energy and starts again with a random thermal velocity.
This is done until the ion hits the electrode or a dust particle. It is also assumed that the
ions enter the plasma sheath with Bohm velocity.
The advantage of following the ion trajectories is that from the principle of actio =
reactio the force exerted on the ion by the dust particle is the same force that acts on the
dust by the ions. Thus the force on the dust due to the ions can be directly determined
from the ion trajectories.
This experiment-related approach ensures that the forces on the particles can be calcu-
70 5. Dust Particle Interaction

lated under very realistic conditions. Figure 5.3 shows the ion density distribution derived
from the ion trajectories for the exact vertically aligned situation (δx = 0) and for the case
that the lower layer is displaced by a quarter of the interparticle distance (δx = 0.25b).
It is seen that the shielding ion cloud around each particle is extended downstream due
to the ion flow. The ions are deflected below the dust by the Coulomb collisions with
the particles. That results in a region of enhanced ion space charge density (“ion focus”)
below each dust grain. The positive space charge is the reason for the attractive force
on the particles. In this approach only a single attraction region is found in agreement
with the more sophisticated wave model [69]. Hence, the wakefield (wave model) and the
ion focus (particle model) can be considered as complimentary descriptions of the same
phenomenon.

5.2.3. Non-reciprocal attraction

The ion trajectory simulations with a second dust particle, however, allow a deeper insight
into the attraction [73]. Namely, one also finds surprisingly that the ion cloud of the upper
particles is independent of the position of the lower particle. The upper half of Fig. 5.3b
looks the same whether the lower particle is directly beneath the upper particle or whether
it is shifted.
In a somewhat simplified picture the behavior can be described as following: Due to
the supersonic ion flow the information on the location of the lower particles cannot be
conveyed upstream. Since there is no reaction on the upper particles from the lower, the
interaction between the particles is non-reciprocal: the lower particles are attracted by the
ion focus of the upper, but there is no attractive force acting on the upper particles. This
is an apparent violation of Newton’s third law actio = reactio when looking only at the
dust particles. Although each single ion reacts with the dust particles by actio = reactio
the collective ion population does not. This is an outcome of the supersonic ion flow that
introduces a broken symmetry into the system. In other words, the steady ion flow puts
energy into the system, thus, it is an open system where Newton’s third law does not
necessarily hold [73].
The vertical particle interaction must therefore be described by a non-reciprocal at-
tractive force. Due to the non-reciprocity of the interaction, we cannot assign a potential
to this kind of interaction. We thus have non-conservative forces. Moreover, all techniques
that rely on energy arguments, like OCP and Yukawa systems, cannot be directly applied
to this situation.
The strength of the force acting on the lower particle in the horizontal direction Fx
has been calculated from the ion trajectories as a function of the horizontal displacement
δx. It is the restoring force that drives the lower particle back into the vertically aligned
position. The strength of the force is shown in Fig. 5.4b. It is seen that the attractive,
restoring force from the ion focus is decisively larger than the repulsion from the upper
5.2. Vertical Interaction 71

particles. It is also seen that the force is linear in δx for small displacements, i.e.

Fx = −katt δx ,

which allows a linear stability analysis of the entire system (which is done in Sec. 5.4.).

Figure 5.4: a) “Schweigert” model of the particle interaction derived from the simulation
in Figure 5.3. The ion focus is mimicked by a positive point charge beneath a dust particle.
The attractive force acts only on the lower particle. b) Strength of the attractive force
(horizontal component) as a function of horizontal displacement. Symbols are results
from the simulation, the solid line is from the model where the ion density distribution is
replaced by a positive point charge. The dashed line indicates the linear behavior for small
displacements. The repulsive force from the upper layer is shown for comparison. From
[72].

This interaction can be translated into the “Schweigert” model presented in Fig. 5.4a.
The distorted ion cloud around the dust, i.e. the ion focus, can be mimicked very ac-
curately by replacing the ion cloud with a positive point charge Zf at a distance d − df
directly below the upper particle (see Fig. 5.4b). Typical values are Zf = 0.5Zd and
df = 0.4b. The force exerted by this point-charge ion-focus exactly matches the forces
determined from the full simulation, see Fig. 5.4b. This model system has the advantage
that it can be analyzed quantitatively in view of stability and phase transitions.
72 5. Dust Particle Interaction

5.3. Measurement of the Attractive Force

This peculiar type of particle interaction can be verified, also quantitatively, from exper-
iments [76, 77, 78]. The non-reciprocal attractive forces can be probed experimentally
in a very simple experiment. For this purpose, only two single particles are immersed
into the plasma (see Fig. 5.5a). The first has a diameter of 3.47 µm (and a mass of
m1 = 3.31 × 10−14 kg) and the second one of 4.81 µm (m2 = 8.82 × 10−14 kg). Due to
their different masses the particles are trapped at different heights in the sheath, where
the force balance is fulfilled for each of the particles. Thus, the vertical position of the
particles is practically fixed. In the horizontal plane, the microspheres can move freely
under the influence of the mutual interaction. From the response of one particle to the
motion of the other particle the interaction between them can be directly extracted.
As in the case of the excitation of the vertical resonance, the focused beam of a manip-
ulation laser (690 nm, 40 mW) is used to push the upper and lower particle individually
and in a defined manner.
Without any laser force applied the two single microspheres are found vertically aligned
as in the case of a plasma crystal. When now the upper particle is pushed by the radiation
pressure of the laser beam both upper and lower particle move in the same way, their
horizontal position is (nearly) identical, i.e. the particles stay vertically aligned (see
Fig. 5.5b). That proves that the lower particle experiences an attractive force mediated
by the upper particle.
When, however, the lower particle is pushed the two particles are separated horizon-
tally, the alignment is broken (see Fig. 5.5c). The lower particle can be pushed far away
from the upper (t = 1 − 4 s). When the laser is switched off again, the lower particle
approaches the aligned position below the upper one, whereas the upper (although being
the lighter particle) does not move towards the lower. In contrast, shortly before the
lower particle reaches the aligned position (around t = 6 s) the upper particle experi-
ences the repulsive force from the lower and moves away from the lower. This definitely
demonstrates that the upper particle does not feel an attractive force mediated by the
lower.
This confirms the fact that Newton’s third law actio = reactio seems to be violated
under these conditions. This is an outcome of the supersonic ion flow that introduces a
preferred direction into the system. In addition, the ion stream introduces a source of
free energy, thus the dust particle system and its environment is an open system where
Hamiltonian dynamics does not hold. One of the main outcomes in connection with the
theoretical analysis presented is that the interaction forces between the dust particles is
non-reciprocal.
From the laser interaction with the two-particle dust system the horizontal attractive
force component can also be derived quantitatively, see Fig. 5.6. By pushing the lower
particle with the laser, the horizontal attractive force is balanced by the radiation pressure
of the laser beam. By calibrating the laser force Flaser using experiments on single particles
5.3. Measurement of the Attractive Force 73

Figure 5.5: Experiment on the non-reciprocity of the attractive force between two dust
particles. a) Scheme of the experimental setup. Horizontal position of upper and lower
particle when b) the upper particle is pushed and c) when the lower particle is pushed.
After [77].

we have determined the attractive force quantitatively. With increasing laser force the
lower particle is shifted horizontally by a displacement ∆x away from the aligned position.
Using the model of the positive point charge that mimics the ion focus the attractive
force on the lower particle can be calculated. The horizontal component of the Coulomb
force between the lower dust particle with charge Z2 and the ion focus with charge Zf is
74 5. Dust Particle Interaction

Figure 5.6: Measurement of the attractive force between two dust particles. The measured
magnitude of the attractive force (symbols) allows to derive the strength of the ion focus.
The solid line is the calculated attractive force according to Eq. (5.11). After [78].

given by
Zf Z2 e2 ∆x Zf Z2 e2 ∆x
Fatt = = , (5.11)
4π0 r2f2 rf2 4π0 (∆x2 + d2f )3/2

where rf2 = (∆x2 + d2f )1/2 is the distance between the ion focus and the lower particle. In
Fig. 5.6 the measured forces are shown in comparison to the above equation. The charge
on the upper and lower particle was measured as Z1 = 2200 and Z2 = 5900, respectively,
from the resonance method. Adjusting the values of Zf and df in Eq. (5.11) agreement
is found for Zf = 0.8Z1 and df ≈ 400 µm. These values are in good agreement with the
simulations.
This “Schweigert” model [73] of the ion focus with the non-reciprocal attraction due
to the formation of the ion focus has hence been verified qualitatively and quantitatively
by experiments. Moreover, this model is also able to explain the stability of plasma
crystals and phase transitions from the ordered, solid phase to an unordered, fluid phase
as described in the following. Moreover, this model has also been successfully applied to
study mode-coupling instabilities [79].
5.4. Oscillatory Instability of the Vertical Alignment 75

Figure 5.7: Forces in a two-layer plasma crystal.

5.4. Oscillatory Instability of the Vertical Alignment

After having shown that the vertical interaction between the microspheres is indeed given
by a non-reciprocal attractive force the question arises under which conditions the verti-
cally aligned plasma crystal observed in the experiment is stable or unstable. Here, this
analysis is performed for the simplest case of a two-layer chain of particles, see Fig. 5.7.
There, the (reciprocal) repulsive forces between the negative dust particles and the (non-
reciprocal) attractive forces on the lower layer particles towards the ion focus of the upper
particles are indicated. From such a plasma crystal model the equations of motion in the
horizontal plane for the nth particle in the upper (index 1) and lower (index 2) layer are

(n) (n) k1 
(n−1) (n) (n+1)
 Z2d e2 
(n) (n)

ẍ1 + βẋ1 = x1 − 2x1 + x1 + x1 − x2 (5.12)
md 4π0 md d3
(n) (n) k2 
(n−1) (n) (n+1)
 Z2d e2 
(n) (n)

ẍ2 + βẋ2 = x2 − 2x2 + x2 + x2 − x1
md 4π0 md d3
Zd Zf e2  (n) (n)

− x − x . (5.13)
4π0 md d3f 2 1

Here, a small horizontal elongation x from the vertically aligned position is assumed. In
addition, d and df are the vertical distance between the lower and upper layer and lower
layer and ion focus, respectively, and Zf the (positive) charge of the ion focus.
The first term on the RHS is the repulsive interaction between the particles of the
same layer (upper or lower). This determines the horizontal oscillation frequency of the
particles in the respective plane. The second term is the repulsion between upper and lower
aligned particle (both taken as pure Coulomb forces for simplicity, here). The attraction
76 5. Dust Particle Interaction

by the ion focus at the same horizontal position (n) appears only in the equation of
motion for the second layer due to the non-reciprocity of the attractive force. The final
point to be addressed, here, is that in the lower layer the intralayer repulsion is assumed
to be possibly different from that in the upper layer, which means k2 < k1 . We will
address this question in more detail below. The model presented here is simplified, it has,
however, all the necessary ingredients to describe the physical mechanisms and to explain
the experimental findings. A complete analysis taking into account a full 2D two-layer
plasma crystal with hexagonal order and screened interaction is not much more difficult
[73], but also does not include significant differences.

5.4.1. Solution of the Coupled Oscillations

To solve the equation of motion, first, we take the usual solution for waves on a linear
(n)
chain xk = xk exp(inqb − iωt), where q is the wave vector and ω is the frequency of
the wave. n b is the equilibrium position of the n-th particle in the chain. The equations
then become
k1  iqb  Z2d e2
−ω2 x1 − iβωx1 = x1 e − 2 + e−iqb + (x1 − x2 ) (5.14)
md 4π0 md d3
k2  iqb  Z2d e2
−ω2 x2 − iβωx2 = x2 e − 2 + e−iqb + (x2 − x1 ) (5.15)
md 4π0 md d3
Zd Zf e2
− (x2 − x1 ) . (5.16)
4π0 md d3f
From trigonometric relations it is found that
!
iqb −iqb 2 qb
e −2+e = 2 cos qb − 2 = 2 (cos qb − 1) = −4 sin .
2
Introducing the notations
Z2d e2
ω2d =
4π0 md d3
Zf d3
ξ =
Zd d3f
k1,2
ω21,2 = 4 ,
md
where ωd is – apart from some multiplicative constants – the dust plasma frequency which
is more rigorously defined in Sec. 6.1. and ξ is the relative strength of the attraction
compared to the repulsive force. Since Zf ≈ 0.5Zd and df ≈ 0.4d the relative strength
is ξ ≈ 8 which indicates that the attraction by the ion focus is much stronger than the
5.4. Oscillatory Instability of the Vertical Alignment 77

repulsion by the upper particle. Finally, ω1,2 is the frequency of the interactions in the
upper and lower layer.
The equation of motion for the upper and lower layer particle then reads
" !#
2 qb
ω + iβω − ω21 2
sin x1 = −ω2d (x1 − x2 ) (5.17)
2
" !#
qb
ω2 + iβω − ω22 sin2 x2 = −(ξ − 1)ω2d (x1 − x2 ) , (5.18)
2
or, finally,
" ! #
2 qb
ω + iβω − ω21 2
sin + ω2d x1 = ω2d x2 (5.19)
2
" ! #
2 qb
ω + iβω − ω22 2
sin − (ξ − 1)ω2d x2 = −(ξ − 1)ω2d x1 . (5.20)
2
This is now the standard form of the equations describing the interaction between the two
layers. The non-reciprocal attraction can still be easily identified from the term containing
the parameter ξ. The non-reciprocity becomes obvious from the fact that ξ only appears
in the equation for the lower particle, but not for the upper.
This set of equations is of fourth order in ω, but can be solved analytically by multi-
plying the two equations yielding
 
λ2 − λ ω̃21 + ω̃22 + ω̃21 ω̃22 = −(ξ − 1)ω4d , (5.21)

where we have used the following abbreviations

λ = ω2 + iβω
!
2 2 2 qb
ω̃1 = ω1 sin − ω2d
2
!
qb
ω̃22 = ω22 sin2 + (ξ − 1)ω2d
2
Thus, the fourth-order equation actually is bi-quadratic. The solution is
s
ω̃21 + ω̃22 (ω̃21 + ω̃22 )2
ω2 + iβω = λ1,2 = ± − ω̃21 ω̃22 − (ξ − 1)ω4d
2 s
4
ω̃21 + ω̃22 (ω̃21 − ω̃22 )2
= ± − (ξ − 1)ω4d (5.22)
2 4
78 5. Dust Particle Interaction

5.4.2. General Analysis of the Instability

Here, we should step back and have a look at this solution (5.22). We do not want to
stick to the exact details of the equation, but to its general form
q
λ=D± E2 − F .
There are two possibilities, namely that either the parameter λ is real or complex, i.e. that
the expression under the square root is either positive (E2 > F → λ = λreal ) or negative
(E2 < F → λ = A ± iB). First, the case of a real λ = λreal will be discussed. Then, we
have
ω2 + iβω − λreal = 0
and we would get the complex solution
s
β β2
ω = −i ± λreal − .
2 4
This means that the solution for x = xk exp(inqb−iωt) is a damped harmonic oscillation.
There is nothing interesting about that. It does not lead to any instability. In this case,
the vertical alignment is stable, both particles return with damped oscillations to their
aligned state x1,2 = 0.
On the other hand, when λ = A±iB is complex, the situation is dramatically different.
We get the same formal solution as above, namely,
s
β β2
ω = −i ± A− ± iB , (5.23)
2 4
but the imaginary contribution below the square root changes the solution drastically.
There are four roots to this equation: a first pair is ω1,2 = ±ωr,1 − iωi,1 , where the
imaginary part is always negative. This also corresponds to a damped oscillatory motion,
as above. The second pair of roots is ω3,4 = ±ωr,2 + iωi,2 , where the imaginary part ωi,2
can be either negative or positive. The change from negative to positive imaginary part
ωi,2 can be achieved by changing the value of β.
A solution with positive imaginary part of ω corresponds to an oscillatory instabil-
ity with (exponentially) growing oscillations. Thus, such a solution would describe an
instability of the vertically aligned pair of particles. The threshold value of the friction
coefficient β∗ for this change is found for the condition that the imaginary part is exactly
zero, i.e. ωi,2 = 0. We then have the condition that
ω2r,2 + iβ∗ ωr,2 = A + iB
which results in
√ B
ωr,2 = A and β∗ = √ . (5.24)
A
5.4. Oscillatory Instability of the Vertical Alignment 79

The four roots of Eq. (5.22) are shown in Fig. 5.8. As described, two of the roots always
have a negative imaginary part, even for β = 0. These would correspond to damped
oscillations around the vertically aligned position. The two other roots cross the ωimag =
0-line at the finite friction constant β∗ . These correspond to oscillations with growing
amplitudes rather than damped amplitudes. They thus are unstable oscillations. These
unstable solutions are found in the entire range of the friction constant 0 < β < β∗ .
This means that unstable oscillations occur even though there is still friction of the dust
particles with the background gas. Thus energy has to be constantly supplied to the
oscillation of the vertically aligned dust system. We will identify the energy source below.

Figure 5.8: Roots of the instability equation (5.22) for assumed (but realistic) values
of A = 4400 s−2 and B = 1600 s−2 . When the imaginary part of the solution is larger
than zero unstable oscillations with growing amplitude occur.
√ Note, that this threshold is
∗ −1
reached for finite values of the damping constant β = B/ A = 24 s .

5.4.3. Conditions for the Instability

Now, we have to identify under which conditions a complex λ = A ± iB is obtained in
Eq. (5.22) which leads to the unstable oscillations. To obtain a complex λ the square root
must be negative. A necessary (but not sufficient) condition obviously is that
ξ>1 .
80 5. Dust Particle Interaction

This means that the attractive force due to the ion focus (which is represented by the
parameter ξ) has to be larger than the repulsion by the upper particle. This is expected
since without domination of attraction a vertical alignment would not be found in the
first place.
From a more detailed analysis (that is not done here) one finds the more stringent
conditions that

ξ > 2 and ω21 > ω22 i.e. k1 > k2 .

First, the attraction must exceed the repulsion by at least a factor of two and, second,
the spring constant reflecting the interlayer interaction must be weaker for the lower
layer than for the upper. This second point can be understood from the fact that the
lower layer particle does not only feel the attraction from the ion foci of the vertically
aligned upper particle, but also the attraction from the other neighboring particles. The
attraction to the neighboring upper particles then weakens the repulsion from the lower
layer neighboring particles which results in k2 < k1 .
The model presented here is simplified wherever possible, but it retains the main
mechanisms. More sophisticated models exist that derive the above mentioned mechanism
from a straightforward mathematical analysis. There, the somewhat ad hoc explanation
of k1 > k2 is taken into account in a much more proper way. But in order not to focus
on the mathematical details a “bare-bone” model is discussed here that allows to grasp
the main ingredients of the instability.
From the physics point of view, this instability can be explained as follows: The
streaming ions act as a source of free energy. The energy of the beam is transferred
to the plasma crystal due to the Coulomb collisions of the ions with the dust. Above
the critical value of damping the energy transferred by the ions can be dissipated by the
friction of the dust particles with the neutral gas. Below the threshold the frictional losses
cannot compensate the energy input and the ion energy piles up to drive the oscillatory
instability.
On the level of the individual dust particles it looks like this: the lower layer particle
feels an attractive force due to the ion wakefield of the upper particle. Hence it moves
towards a position where it is aligned with the upper particle. However, the attractive
force is non-reciprocal which means that the upper particles does not feel the attraction.
In contrast, the upper particle feels the (screened) Coulomb repulsion from the negative
charge of the lower layer particle. Hence, the upper particle tries to move away from the
lower one. But the lower tries to follow the upper. The lower wants to be close to the
upper, but the upper wants to escape. Due to the neighboring particles in the same layer,
this type of unstable situation turns into an oscillatory instability. The neighbors of the
upper particle provides a “cage” for the upper particle where it can basically oscillate at
the dust plasma frequency. The lower particle tries to follow the oscillations of the upper,
the upper tries to escape, and so, the oscillatory instability develops.
5.5. The Phase Transition of the Plasma Crystal: Experiment 81

Summarizing, the linear stability analysis shows that a critical value of the friction
coefficient β∗ > 0, i.e. the gas pressure in the discharge, exists below which both upper
and lower layer perform horizontal oscillations about the vertically aligned equilibrium
position with exponentially increasing oscillation amplitude. These are short wavelength
modes on the linear chain with qb > 1 with a frequency near the dust plasma frequency
ωpd . Moreover, from the stability analysis the relative phase and oscillation amplitude
between particles of upper and lower layer can be derived. Above the critical pressure,
the vertical alignment is found to be stable. This instability is directly connected to the
frictional damping of the dust particles and not to a change of discharge parameters with
changing pressure.
This theoretical description agrees very well with experimentally observed oscillations
in two-layer plasma crystals near the melting transition [72], see Fig. 5.9. The calculated
values for oscillation frequency, relative phase and amplitude as well as the pressure
threshold for the onset of these oscillations are within a factor of two of the measured
ones which gives this model a high credibility.

5.5. The Phase Transition of the Plasma Crystal:

Experiment
We now will see what the effects of the above mentioned instability mechanism are. In
the following, we will investigate the experimental melting transition of plasma crystals.
The experiments have been performed with two-layer plasma crystals which are subject
to the above mentioned instability. Mono-layer crystals do not show a melting scenario
as described below.
Experimentally, the melting transition is observed when the gas pressure in the dis-
charge is reduced [46, 80]. At high gas pressures (118 Pa in this particular experiment)
well ordered crystalline structures are found (Fig. 5.10a). The particles do not move con-
siderably, they stay in their respective Wigner-Seitz cells. At reduced pressure the particle
arrangement undergoes a transition to a liquid and, finally, to an almost gas-like state.
During this transition, at first, stream line particle motion around crystalline patches sets
in, that gradually turns into a more and more irregular particle motion. This transition
is also seen in the correlation function g(r) and g6 (r), see Fig. 5.10c,d. The translational
as well as the orientational ordering is strongly decreased from the ordered state at high
pressures to the completely disordered, gas-like state at 39 Pa.
The transition from the ordered to the liquid state is accompanied, and even driven,
by the horizontal oscillations of the vertically aligned pairs that were discussed in the
previous section. One can easily observe these oscillations around the equilibrium position
by video microscopy. They are not visible in the trajectories since these are averaged over
10 seconds. The oscillations are, nevertheless, described in Fig. 5.9.
From the thermodynamic point of view, the transition is driven by a dramatic increase
82 5. Dust Particle Interaction

Figure 5.9: Comparison of experimental and theoretical oscillatory instability. a) Fre-

quency of particle oscillations and growth rate of the instability, b) ratio of oscillation
amplitudes Al /Au and c) phase difference of lower and upper particle. Symbols denote
experimental values from oscillations during the melting transition and the lines corre-
spond to the theoretical values from the stability analysis. d) One experimental example
of horizontal oscillations of a vertically aligned pair. After [72].

of dust temperature from room temperature at 118 Pa to Td = 50 eV at 39 Pa, see

Fig. 5.10b. This high temperature is far above the energies of all other plasma species.
This dramatic increase of dust temperature cannot be explained by simple changes in
the discharge conditions with reduced pressure. In this pressure range of interest the
densities of electrons and ions change by a factor about 3 and the electron temperature
is nearly unaffected. These small changes cannot be made responsible for the change of
dust temperature by a factor 1000.
Naturally, the previously described instability mechanism based on the non-reciprocal
5.5. The Phase Transition of the Plasma Crystal: Experiment 83

Figure 5.10: a) Trajectories of the dust particles over 10 seconds for decreasing discharge
pressure, b) temperature of the dust particles as a function of discharge pressure. A
temperature below 0.7 eV could not be detected due to the limited optical resolution, c)
pair correlation and d) orientational correlation function versus pressure. After [46].
84 5. Dust Particle Interaction

attraction is the source of energy input to the microspheres. The oscillatory instability
sets in at about 80 Pa under the conditions of this experiment (see Fig. 5.9). This is
exactly the pressure when fluid particle trajectories set in. The energetic oscillations are
becoming more and more irregular and turn into a chaotic motion of the particles which
can be interpreted as heating the dust particles to the high temperatures mentioned above.
One should note that very similar phase transitions have been observed in krypton at
a lower gas pressure, but at nearly the same values for the frictional damping constant β
[80].

5.6. The Phase Transition of the Plasma Crystal:

Simulations and Theory
To identify the heating mechanism and melting dynamics of plasma crystals the melt-
ing transition was investigated in simulations in which the experimental conditions were
mimicked in great detail [81]. Some of the interesting questions are how the instability
mechanism drives the transition and how the oscillatory particle motion is randomized to
heat the crystal.

Figure 5.11: a) Particle trajectories for different friction coefficients during the melting
transition. b) Particle energy of upper and lower layer versus friction coefficient. The
points of which the trajectories are shown in a) are indicated by arrows. The two curves
correspond to the energy of upper and lower layer, respectively. From [81].

In the simulations an ideal hexagonal two-layer plasma crystal is considered. The

5.6. The Phase Transition of the Plasma Crystal: Simulations and Theory 85

negatively charged particles interact via a screened Coulomb repulsion and the particles
of the lower layer experience the non-reciprocal, attractive force from the ion clouds of
the upper particles. Here, the full horizontal dependence of the attractive force according
to Fig. 5.3c is taken into account. In the simulation only horizontal displacements ~ρjk =
(xjk , yjk ) from the equilibrium positions are taken into account (k = 1, 2 indicates upper
and lower layer, respectively, and j denotes the number of the particle in that layer):
d2~ρjk d~ρjk 1 ~ 1 ~
+ β = Fjk + FL . (5.25)
dt2 dt md md
The force acting on the particles ~Fjk consists of two parts. The first describes the repulsion
due to the other dust particles in the same or in the other layer. The second part describes
the attractive force on the lower particles due to the ion focus (replaced by a single point
charge). In addition, a Langevin force FL is applied to the particles to give them a finite
temperature (room temperature) in the crystalline state at high pressure. This Langevin
force is kept constant throughout the melting process leading to no additional heating.
This model is exactly the same as that used for the linear stability analysis in Section
5.4. with the exception of the Langevin force and the consideration of the full horizontal
dependence, and not only the linear part for small elongations. As in the experiment,
the gas pressure, i.e. the friction constant β, was slowly reduced in the simulation. The
resulting energy of the dust particles as a function of gas pressure is shown in Fig. 5.11
together with the particle trajectories for different values of the friction coefficient.
For the melting transition of an ideal crystal a two-step melting scenario is obtained:
Starting at high pressures the well ordered crystal is found. With decreasing friction the
oscillatory instability described in Section 5.4. sets in at about βin = 0.165ωpd . This
leads to a dramatic increase of the dust kinetic energy from room temperature to about
10 eV. However, that does not lead to the melting of the crystal, instead, a hot crystalline
state is found, here. With further reduction of friction (β∗ = 0.12ωpd ) the transition to
a liquid state is observed.
The maximum energy of the dust particles and the overall melting scenario in the
simulation agrees well with that obtained from experiments. However, some differences
in comparison to the experiment are found in the simulation. First, the experimental
melting transition takes place over a broader range of β than the simulation. Second, the
two-step melting is not clearly observed in the experiment and, third, in the experiment
characteristic stream line particle motions around crystalline patches are seen that are
not present in the simulation.
This discrepancy is resolved when also defects in the plasma crystal are taken into
account [82]. The crystal with defects shows a melting scenario that is very similar to the
experimental: the increase of dust temperature with reduced pressure is slower, in overall
quantitative agreement with the experiment. In addition, the two-step melting transition
is “smeared out” over a broader range of β so that the melting is gradual and no sharp
transition can be assigned. Finally, the particle trajectories show the streamline particle
86 5. Dust Particle Interaction

motions that are characteristic for the experiment. The melting of a crystal with defects
is shown in Fig. 5.12. Compare this simulated phase transition with the experimental one
in Fig. 5.10. One can observe a deep agreement between the experimental and simulated
phase transition. Since the simulations are only based on the non-reciprocal attraction as
discussed above and no other heating mechanisms are taken into account this agreement
gives strong confidence in the existence and relevance of these non-reciprocal forces.

Figure 5.12: a) Particle trajectories for different friction coefficients during the melting
transition. In the upper left panel the 5-fold defect is marked with a pentagon. b) Particle
energy versus friction coefficient. Open circles (◦): Two-step melting of a crystal with
defects (simulation), squares ( ): Experimental results. After [82].
 87

6. Waves in Weakly Coupled Dusty Plasmas

As another example of collective motions in dusty plasmas, we now turn to the topic
of waves. For dusty plasmas, a number of wave modes exist that have been predicted
theoretically and that will not be discussed entirely in these Lecture Notes. Instead, we
will focus on a few fundamental wave types that have been observed in experiments. For
a more detailed overview on waves the reader is referred to dedicated monographs [1, 3].
In general, two categories of waves can be identified, namely those which do not require
strong coupling of the dust particles and those which rely on the strong coupling. In the
first category, we find, e.g., the dust-acoustic (DAW) and dust ion-acoustic wave (DIAW).
The dust lattice wave (DLW) with its different “polarizations” requires an ordered dust
arrangement on lattice sites and thus belongs to the second category. Here, we start with
the discussion of the weakly coupled waves, the DAW and the DIAW. The DAW is a
wave where the dust particles are the moving species, the DIAW considers the dust as
immobile, but the dust influences the wave propagation.

6.1. Dust-Acoustic Waves

The DAW is a complete analog to the ion-acoustic wave: The ion-acoustic wave is an
electrostatic ion wave where the ions provide the inertia. In the DAW, the inertia is given
by the dust.
The dust acoustic wave [83] is a very low-frequent wave with wave frequencies of
the order of the dust plasma frequency ωpd which, due to the high dust mass, is much
less than the ion plasma and electron plasma frequency (ωpi , ωpe ). Analogously to the
electron and ion plasma frequency the dust plasma frequency is defined as
s
Z2d e2 nd0
ωpd =  ωpi , ωpe , (6.1)
ε 0 md
where nd0 is the equilibrium (undisturbed) dust density.
Before the mathematical details of the wave motion are described, the wave mechanism
is explained (see Fig. 6.1). Consider a wave-like disturbance of the dust density. The dust
density disturbance is associated with an enhanced negative space charge due to the
negative dust charge. This dust space charge Zd nd will then be immediately shielded by
the ambient plasma ions and (to a lesser extent) electrons. In contrast to the ion-acoustic
wave, here the shielding is by both electrons and ions. However, the shielding is not
complete, the dust charge density fluctuations nd Zd are larger by a very small amount
compared to that of electrons and ions ni − ne . However, this is enough to form negative
and positive space charge regions which lead to electric fields that further drive the wave.
88 6. Waves in Weakly Coupled Dusty Plasmas

Figure 6.1: Comparison of the wave-like dust density disturbance with the ion and electron
density fluctuations. It is seen that the dust density fluctuation Zd nd is slightly larger than
that of electrons and ions ni − ne .

The electric field is maximum near the zero-crossing of the dust density perturbations.
Therefore, the electric field force pushes the fluctuations in the direction of the electric
field and thus the wave propagates.
For the derivation of the DAW the equation of continuity, the momentum equation
and Poisson’s equation for the dust species are used which can be written as
∂nd ∂
+ (nd vd ) = 0 (6.2)
∂t ∂x
∂vd ∂vd kTd ∂nd Zd e ∂φ
+ vd + γd = − βvd (6.3)
∂t ∂x md nd ∂x md ∂x
∂2 φ e
2
= − (ni − ne − Zd nd ) . (6.4)
∂x ε0
There are a few small differences to the usual ion-acoustic wave: the momentum equation
(6.3) includes friction with the neutral gas (βvd ), and Poisson’s equation (6.4) includes
all three charged species, electrons, ions and dust. To solve these equations, the dust
density and velocity as well as the electron and ion densities are considered as fluctuating
quantities with

nd = nd0 + ñd
vd = 0 + ṽd
φ = 0 + φ̃ .
6.1. Dust-Acoustic Waves 89

The fluctuations are considered to be wave-like, i.e. proportional to exp(iqx − iωt). The
electrons and ions are assumed to have a Boltzmann distribution, namely
! !
eφ̃ eφ̃
ne = ne0 exp ' ne0 1+ = ne0 + ñe
kTe kTe
! !
eφ̃ eφ̃
ni = ni0 exp − ' ni0 1− = ni0 + ñi
kTi kTi
Here, ne0 and ni0 denote the equilibrium (undisturbed) values of the electron and ion
density. For the undisturbed densities the quasineutrality condition is fulfilled, i.e.
ni0 = ne0 + Zd nd0 ,
where the dust is assumed to be negatively charged and adds to the electron charge
density.
The above three equations can then be written as
−iωñd + iqnd0 ṽd = 0 (6.5)
kTd Zd e
−iωṽd + iqγd ñd = iqφ̃ − βṽd (6.6)
md nd0 md
e
−q2 φ̃ = − (ni − ne − Zd nd ) . (6.7)
0
Here, as usual, we have used that for the wave-like fluctuations the spatial and temporal
derivations can be replaced by the products with the frequencies and wave vectors (∂/∂t →
−iω; ∂/∂x → i q). The last equation then becomes
!
e e eφ̃ eφ̃
−q2 φ̃ = − (ni0 − ne0 − Zd nd0 ) − ni0 + ne0 − Zd ñd .
0 0 kTi kTe
The first term is zero due to quasineutrality. Applying the quasineutrality again to replace
ne0 = ni0 − Zd nd0 and using the relative dust density  = nd0 /ni0 we get
e2 ni0
!
2 Ti eZd q
−q φ̃ = 1 + [1 − Zd ] φ̃ + nd0 ṽd ,
0 kTi Te 0 ω
where we have used the first equation to replace ñd with ṽd . From that we can write
eZd q λ2D,i
φ̃ = − nd0 ṽd 2 2 .
0 ω q λD,i + 1 + TTei [1 − Zd ]
Inserting this expression for φ and the first equation into the equation of motion we yield
after a few rearrangements
 
kTd kTi 1
ω2 + iβω = γd + Z2d    q2 . (6.8)
md md 1 + Ti
(1 − εZd ) + q2 λ2D,i
Te
90 6. Waves in Weakly Coupled Dusty Plasmas

Figure 6.2: (a) Dispersion relation of the dust-acoustic wave without damping. The solid
line is the full dispersion relation, the dotted line indicates the acoustic limit with the
dust-acoustic velocity. (b) Dispersion relation with small friction (β = 0.1 ωpd ) and (c)
with large friction (β = 0.5 ωpd ). Here, the solid line refers to the real part of the wave
vector and the dashed line to the imaginary part. Note, that in (b) and (c) the axes have
been exchanged compared to (a).

This is the full dispersion relation of the DAW. It contains a number of effects. The first
term in the brackets is the dust thermal velocity and the second contains the influence of
the electron and ion drive on the dust inertia. This is more clearly seen under the typical
assumption of cold dust (Td = 0) and cold ions (Ti  Te ). Then, the dispersion relation
simplifies to
ω2pd q2 λ2D,i ω2pd q2 b2
ω2 + iβω = = , (6.9)
1 + q2 λ2D,i κ2 + q2 b2
which is the same as for the ion-acoustic wave where the ion properties are replaced by
those of the dust and the electron properties by those of the ions. The second expression
has been obtained by introducing the screening strength κ = b/λD,i . Thus the wave
frequencies (and wave speeds) decrease with increasing κ.
The dispersion relation of the DAW is shown in Fig. 6.2a. For large wave numbers
2 2
q λD,i  1 the wave is not propagating and oscillates at the dust plasma frequency
ωpd . For small wave numbers q2 λ2D,i  1 the wave is acoustic ω = qCDAW with the
dust-acoustic wave speed
s
kTi 2
CDAW = Z . (6.10)
md d
As for the ion-acoustic wave, the wave speed is determined by the temperature of the
lighter species (Ti ) and the mass of the heavier (md ). The dust-acoustic wave speed also
includes the contribution of the dust charge Zd and the relative dust concentration . It
6.1. Dust-Acoustic Waves 91

is also interesting to note that the governing shielding length is the ion Debye length λD,i
as the ions are the oppositely charged fluid that shields the repulsion between the dust
particles.
We now like to analyze the influence of friction on the wave motion. Before doing so we
like to remind to how instabilities and waves are treated in “usual” plasma physics. There,
the instability condition of a wave that is proportional to exp(iqx−iωt) is obtained when
the imaginary part ωI of the wave frequency ω = ωr + iωi becomes larger than zero.
Naturally, the wave vector q is a real value.
In contrast, when waves are excited in a frictional medium the wave frequency ω
has to be taken as a real value and, consequently, the wave vector has to be treated as
complex q = qr + iqi , where the real part qr = 2π/λ is related to the wave length λ and
the imaginary qi = 1/L to the damping length L in the system. The damping length L
is the distance where the wave amplitude is reduced to 1/e. In this situation, we have to
determine the real and the imaginary part of the wave vector for each value of the wave
frequency. Thus, it is more convenient to plot the dispersion relation as qr (ω) and qi (ω)
instead of the “usual” dispersion ω(q).
Figure 6.2b,c shows the DAW dispersion for small and large values of the friction
coefficient β. For small friction the real part of the wave vector behaves similarly to
the case of no damping. Close to ω = ωpd the wave vector turns over and decreases

Figure 6.3: Observation of the DAW in a dc discharge. (a) The DAW is seen as regions
of high and low dust density in scattered light. (b) Measured dispersion relation of the
DAW. From [84, 85].
92 6. Waves in Weakly Coupled Dusty Plasmas

dramatically towards zero again. In this range the imaginary part of the wave vector
jumps from small values, i.e. low damping, to large values. For ω > ωpd an overcritically
damped DAW is found. This region with ω > ωpd was not accessible in the case of no
damping. With damping present, this region can be entered, however only overcritically
damped as an evanescent wave.
For larger friction constants (Fig. 6.2c) the wave speed ω/q increases and the max-
imum observable wave number decreases drastically. Moreover, the real and imaginary
part of the wave vector are comparable over the entire range: the DAW is found to be
strongly damped throughout.
Dust acoustic waves have been observed experimentally in weakly [84, 85] and strongly
coupled dusty plasma systems [86]. In the weakly coupled system [84, 85], a dc discharge is
driven between an anode disk and the chamber walls. The dust particles are accumulated
from a dust tray placed below the anode region. The dust is found to form dust density
waves with a certain wavelength and frequency (see Fig. 6.3a). By applying a sinusoidal
voltage on the anode the wave can be driven and the dispersion relation is obtained
(Fig. 6.3b). The wave shows a linear, acoustic dispersion in agreement with the DAW at
long wavelengths.
In a different experiment [86], dust-acoustic waves have been driven in a plasma crystal
by a sinusoidal voltage on a wire close to the crystal. The propagation of the wave in the
crystal was observed by video cameras and the corresponding wave length and damping
length are derived. The measured dispersion relation was found to be in close agreement
with a damped DAW, although the system is strongly coupled. Compare the measured
dispersion relation in Fig. 6.4b) with that of the calculated in Fig. 6.2c).

6.2. Ion-Flow Driven Dust Acoustic Waves

In a number of experiments under different conditions (dc or rf discharges in the laboratory
or under microgravity) self-excited dust acoustic waves appear (e.g. [84, 87, 88, 89, 90, 91]
to list a few), see also Fig. 6.5 and also Fig. 6.3. The ambipolar or sheath electric field that
confines the dust usually also drives an ion flow. Hence, essentially every dust cloud in the
experiment is penetrated by streaming ions. As the two-stream instability in “ordinary”
plasmas excites slow space charge waves, here in a dusty plasma, the ion current excites
the dust acoustic waves.
The dispersion relation of such a driven dust acoustic wave can be written in simplified
form as
1 ω2pi ω2pd
(ω, q) = 1 + − − =0 , (6.11)
q2 λ2D,e Ωi (Ωi + iνi ) − q2 v2th,i ω(ω + iβ)
where

Ωi = ω − qui
6.2. Ion-Flow Driven Dust Acoustic Waves 93

Figure 6.4: a) Wire excitation of the DAW. The wave is driven by a sinusoidal voltage on
the wire close to a plasma crystal. b) Measured dispersion of the DAW: real and imaginary
part of the wave vector as a function of wave frequency. From [86].

is the Doppler-shifted frequency that the drifting ion “see”. Hence, the ion term includes
the drifting ions as well as ion-neutral collisions in form of the collision frequency νi . For
the electrons, again, we have taken the low-frequency limit since we are interested in the
very low-frequency dust acoustic waves [compare Eq. (5.5)]. For the dust, temperature
is neglected (i.e. vth,d = 0), but friction with the neutral gas is included by the Epstein
friction coefficient β.
The dispersion relation has, in general, 4 complex roots for the frequency ω = ωr +
iωi , but we are interested only in the very low frequency limit ω ≤ ωpd  ωpi , ωpe
(since we are dealing with self-excited waves here we analyze the situation with real and
imaginary wave frequencies instead, as above, with complex wave vectors). Figure 6.6
shows the calculated real and imaginary parts of the frequencies for the ion-flow driven
waves according to Eq. (6.11) in comparison to the undriven case according to Eq. (6.9).
First, it is seen that the dispersion for the two cases look quite different. The attainable
real wave frequencies are higher for the driven case. They can even exceed ωpd . The
imaginary part of the frequency for the driven wave is positive for a large range of wave
vectors q indicating that the wave is (exponentially) growing in time. Hence, the driven
waves are unstable despite the presence of friction of the dust particles with the neutral gas
(as accounted for by β). Only for very short waves (qλD,i ≈ 1) wave damping occurs. The
most unstable wave is found for qλD,i ≈ 0.5 with ωr ≈ ωpd under the chosen conditions
94 6. Waves in Weakly Coupled Dusty Plasmas

(which are similar to that of Ref. [88]).

In contrast a wave without the ion-flow drive under the same conditions shows a slow
increase of the wave frequency (real part) from zero towards the dust plasma frequency
ωpd . The imaginary part is almost constantly negative at −β/2 indicating the wave
damping by neutral gas friction. Very long waves (q → 0) are completely damped.
Hence, the (almost) omnipresent ion flow can excite dust acoustic waves against the
neutral gas damping. The wave dispersion is accordingly modified and quantitatively
differs from the simple undriven case. It has also been shown that the dust waves can be
driven under an angle relative to the ion motion [90].

6.3. Dust Ion-Acoustic Wave

The next wave type to be discussed is the dust ion-acoustic wave. The dust ion-acoustic
wave is the usual ion-acoustic wave where also the dust has been taken into account.
In contrast to the DAW where the dust is the moving species, the dust particles are
considered as immobile in the DIAW. The wave frequencies are of the order of the ion
plasma frequency ωpi  ωpd . The influence of the dust lies only in the reduction of the
free electron density since a certain fraction of the electrons is attached to the dust. We

Figure 6.5: Snap shots of naturally excited DAWs. (a) DAW in a dc discharge in a glass
tube, (b) DAW in a dc anodic plasma, (c) DAW in an rf discharge under microgravity.
From [89, 88, 90].
6.3. Dust Ion-Acoustic Wave 95

apply the equation of continuity, the momentum equation and Poisson’s equation for the
ions
∂ni ∂
+ (ni vi ) = 0 (6.12)
∂t ∂x
∂vi ∂vi e ∂φ
+ vi = (6.13)
∂t ∂x mi ∂x
∂2 φ e
= − (ni − ne ) . (6.14)
∂x2 ε0
These equations are exactly those which are used to derive the dispersion relation of the
ion-acoustic wave. In comparison to the equations used to describe the DAW, here, in the
momentum equation the kinetic pressure of the ions and the friction force are neglected.
Poisson’s equation does not include the dust since the dust is immobile. The electrons
are treated as Boltzmann distributed
eφ
ne = ne0 exp( ) .
kTe

Figure 6.6: Dispersion relation of the dust acoustic wave excited by an ion flow (solid
lines) according to Eq. (6.11). In comparison the dispersion of the undriven DAW accord-
ing to Eq. (6.9) is also shown (dashed lines). The red and blue lines indicate the real and
imaginary part of the wave frequency for a real wave vector q, respectively.
96 6. Waves in Weakly Coupled Dusty Plasmas

The only place where the dust properties enter is the quasineutrality condition

ni0 = ne0 + Zd nd0 .

Thus, here, the undisturbed electron and ion densities are different since a fraction of the
electrons is bound on the dust. The equations are solved analogously as for the DAW.
The dispersion relation of the DIAW is then given as

2ω2pi λ2D,e q2 
ni0 kTe

q2
ω = = (6.15)
1 + q2 λ2D,e ne0 mi 1 + q2 λ2D,e
which is that of the pure ion-acoustic wave with the additional factor of ni0 /ne0 > 1.
The dispersion relation of the DIAW is shown in Fig. 6.7 in comparison to the usual
ion-acoustic wave. The sound speed of the DIAW (ω/q for q → 0) is larger than that of
the DIAW by the additional factor (ni0 /ne0 )1/2 , namely
s s
ni0 kTe
CDIAW = . (6.16)
ne0 mi
Thus, with increasing dust charge density and thus reduced electron density the speed of
the DIAW will increase in comparison to the pure ion-acoustic wave. The DIAW has the
same maximum frequency, the ion plasma frequency ωpi . The DIAW can be obtained
from the ion-acoustic wave by rescaling the qλD,e -axis since for the DIAW the electron
Debye length differs from that of the IAW due to the reduced electron density.
Experimentally, the DIAW has been observed in a Q-machine plasma, where dust has
been immersed [92]. It was found that the wave speed of the IAW increases when dust is
present (see Fig. 6.8). The increase of the wave speed has been taken as an indication of
the existence of the DIAW.

6.4. Other Wave Types

Theoretically, a number of additional waves types are discussed for weakly coupled dusty
plasmas. Among those are the dust-cyclotron wave and the dust ion-cyclotron wave.
Others are dust-Alfvén waves or the dust-Whistler waves.
The dust cyclotron wave and dust ion-cyclotron wave are again the exact twins of the
ion-cyclotron waves. In the dust-cyclotron wave the dust is the mobile species. The ions in
the “usual” ion-cyclotron wave are replaced by the dust and both electrons and ions take
the role of the electrons. Consequently, the wave dispersion relation of the dust-cyclotron
wave reads as
kTi 2
ω2 + iβω = ω2cd + Z2d q
md
6.4. Other Wave Types 97

Figure 6.7: Dispersion relation of the DIAW in comparison to the usual ion-acoustic
wave(IAW).

where ωcd = Zd eB/md is the cyclotron frequency of the dust at the magnetic field strength
B. Obviously, it is the same dispersion relation as for the DAW, only shifted by the
cyclotron frequency. The “usual” ion-cyclotron wave dispersion is known to be
kTe 2
ω2 = ω2ci + q
mi
with the ion cyclotron frequency ωci = eB/mi . So, the ion-acoustic wave speed of the
ion-cyclotron wave is replaced by the dust-acoustic wave speed for the dust-cyclotron
wave.
Similarly, in the dust ion-cyclotron wave the dust is immobile and the ion-cyclotron
dispersion includes only the term ni0 /ne0 > 1 to account for the electrons bound on the
dust. The dispersion then is
ni0 kTe 2
 
ω2 = ω2ci + q
ne0 mi
Similar to the dust ion-acoustic wave, dust ion-cyclotron waves have been driven in a
magnetized Q-machine plasma [92]. There, an increased amplitude of the ion-cyclotron
wave is observed with increased dust charge density (see Fig. 6.9). This is not a direct
measurement of the increased phase speed as in the case of the DIAW, but it demonstrates
that the presence of dust allows an easier excitation of the ion-cyclotron wave.
Thus, the situation is completely analogous to the DAW/DIAW wave dispersion.
98 6. Waves in Weakly Coupled Dusty Plasmas

Figure 6.8: (a) Experimental setup for the observation of the DIAW. The dust is im-
mersed into the plasma by a rotating dust “drum”. (b) Measured velocity of the DIAW
with increasing dust charge density ZD . From [92].

Figure 6.9: Wave amplitude of the dust ion-cyclotron wave with increasing dust density.
The wave amplitude is normalized to the case of no dust. From [92].

Finally, a number of dust-influenced waves has been discussed theoretically. However,

experimental measurements of such waves are rare. In addition, nonlinear properties of the
6.4. Other Wave Types 99

various wave types have also been analyzed theoretically. Again, only a few experiments
are available. We thus limit the presentation of dust waves in weakly coupled systems to
the examples mentioned here.
100 7. Waves in Strongly Coupled Dusty Plasmas

7. Waves in Strongly Coupled Dusty Plasmas

In this section, dust lattice waves (DLW) will be discussed. As the name suggests, the
dust lattice wave requires the particles to be ordered in a crystal lattice. We thus need a
strongly coupled dust system. The dust lattice waves will be treated here mostly for the
2D case.
In 2D systems, the dust lattice wave has three different “polarizations”. Compres-
sional, shear and transverse modes will be discussed (see Fig. 7.1). For the compressional
(longitudinal) mode, the particle motion is along the wave propagation direction and

Figure 7.1: Compressional wave, shear wave and transverse wave.

7.1. Compressional Mode in 1D 101

leads to compression and rarefaction of the dust. In the shear mode, the dust motion is
perpendicular to the wave propagation but inside the 2D crystal plane. The transverse
mode also describes particle motion perpendicular to the wave propagation, but here the
dust motion is an out-of-plane motion and thus requires the consideration of the vertical
confinement of the dust. These three wave types have been observed in the experiment
and will be presented in the following.

7.1. Compressional Mode in 1D

The dispersion relation of the 2D-DLW will be illustrated starting from the simpler model
of a 1D chain of dust particles. On a linear chain the dust particles have equidistant
equilibrium positions Xn = nb, where b is the interparticle distance. Neighboring dust
particles are considered to be connected by springs of spring constant k. These “spring
constants” are of course due to the repulsive interaction between the dust particles as
described below. The force on the n-th particle is F = k(xn−1 − xn ) due to the spring to
the left neighbor and F = k(xn+1 − xn ) due to the right neighbor. The equation of motion
for the n-th particle then is
md ẍn − md βẋn = k(xn−1 − 2xn + xn+1 ) , (7.1)
where xn is the elongation of the n-th particle from its equilibrium position Xn . Here,
also friction with the neutral gas is included. Using the ansatz for waves on a linear chain
xn = A exp(inqb−iωt) with wave vector q and wave frequency ω the equation of motion
becomes
−md ω2 − imd βω = k(eiqb + e−iqb − 2) = 2k (cos qb − 1)
and the dispersion relation (see also Sec. 5.4.)
!
2 k qb
ω + iβω = 4 sin2 (7.2)
md 2

Figure 7.2: Sketch of a linear dust arrangement with longitudinal particle displacements.
The grey circles indicate the equilibrium positions in the chain, the bluish the instantaneous
positions in the traveling wave.
102 7. Waves in Strongly Coupled Dusty Plasmas

is obtained. This is the well-known dispersion relation of waves on a linear chain which
is also very familiar in condensed matter physics.
Now, the spring constant k has to be related to the repulsive interaction between the
dust particles. From mechanics it is known that the spring constant is just the second
derivative of the interaction potential. This yields for a Debye-Hückel interaction at the
interparticle distance b
d2 φ Z2d e2 −κ  2

k= = e 2 + 2κ + κ , (7.3)
dx2 x=b
4πε0 b3
where the screening strength κ = b/λs has been used. The corresponding dispersion can
then be written as
Z2d e2
!
2 qb
 
2 −κ 2
ω + iβω = e 2 + 2κ + κ sin , (7.4)
πε0 md b3 2
Finally, the dispersion relation can be extended to include also the influence of many
neighbors. Therefore, simply the “springs” to all other neighbors at distance `b have to
be considered yielding [93]
X∞
md ẍn − md βẋn = k(`b)(xn−` − 2xn + xn+` ) .
`=1

The full 1D dispersion relation then is given by [93]

1 2 X e−`κ 
∞ !
2 `qb

2 2 2
ω + iβω = ωpd 2 + 2`κ + ` κ sin , (7.5)
π `=1
`3 2

Figure 7.3: Dispersion relation of the 1D dust lattice wave for different values of the
screening strength κ at zero damping.
7.1. Compressional Mode in 1D 103

where the dust plasma frequency

Z2d e2
ω2pd = (7.6)
ε0 md b3
has been introduced for the strongly coupled case by identifying nd ≈ b−3 in (6.1). It
should, however, be noted that the dust plasma frequency is used here only because of
the mathematical similarity. It does not correspond to a collective oscillation of the dust
particles relative to a stationary background. Rather this frequency is related to the
Debye frequency in solid-state physics.
The dispersion relation of the compressional 1D dust lattice wave is shown in Fig. 7.3
for different values of κ. The dispersion relation has approximately the form of a sine
function. For small wave numbers (long wavelengths) the wave is acoustic (ω ∝ q),
whereas the dispersion bends over for larger wave numbers and finally reaches a maximum
where ∂ω/∂q ≈ 0. The maximum wave number is given by qb = π which corresponds
to a wave length λ = 2b. This wavelength is the shortest possible wave length on a string
of particles and corresponds to 180-degree out-of-phase motion between nearest neighbor
particles. As for the DAW, the wave frequencies decrease with increasing κ.
The sound speed of the 1D compressional DLW is given by
X e−`κ
∞
s v
∂ω ω Z2d e2 u
t 1
u
c = lim = lim = (2 + 2`κ + `2 κ2 ) = c0 f(κ) , (7.7)
q→0 ∂q q→0 q ε0 md b 4π `=1 `

Figure 7.4: Dispersion relation of the 1D dust lattice wave for different values of the
damping strength β at κ = 1.
104 7. Waves in Strongly Coupled Dusty Plasmas

where
s
Z2d e2
c0 = (7.8)
ε0 md b
is a measure of the sound speed of dust lattice waves in plasma crystals and

1 X e−`κ
∞
v
u
u
f(κ) = t (2 + 2`κ + `2 κ2 ) (7.9)
4π `=1 `

is a function of κ only and takes into account the crystal order.

The effect of damping is shown in Fig. 7.4 where real and imaginary part of the wave
vector are shown as a function of wave frequency. An analytical solution exists when only
nearest neighbor interaction (` = 1) in Eq. (7.5) is taken account. Otherwise the solution
has to be found numerically. Fig. 7.4 shows the case of many neighbor interaction. It can
be seen that with increased damping the maximum real wave vector is reduced. In the
same way, the imaginary part (corresponding to the inverse damping length) increases.
Similar to the dispersion relation of the DAW, in the DLW the imaginary wave vector
grows dramatically when the wave frequency comes close to the maximum wave frequency
ωmax ≈ 0.8ωpd (compare Fig. 7.3). Heavily damped waves are found when the wave enters
a frequency regime that was not accessible in the frictionless case where ω > ωmax .

7.2. Dust Lattice Waves in 2D

The dispersion relation of 1D waves can be extended to the case of a two-dimensional
lattice in a straightforward manner for the compressional mode as well as for the shear
mode [94, 95].
There, one has to take into account the hexagonal order of the 2D lattice. For the
compressional wave, the dispersion relation then reads

1 2 X e−r̃κ  2 
∞ !
2 `qb
 
2 2 2 2
ω + iβω = ω ` 3 + 3r̃κ + r̃ κ − r̃ (1 + r̃κ) sin , (7.10)
2π pd `,m=1 r̃5 2
√
where r̃ = `2 + m2 . The term after the sum is just the second derivative of the Yukawa
potential in x-direction at the lattice site (`, m). Here, it is assumed that the wave
propagates in x-direction (along a~ 1 in Fig. 7.5). Thus, the force (related to the derivative)
and thus the particle motion is in the same direction as the wave. Hence, this describes
a compressional (longitudinal) wave mode.
For the shear wave one also has to take the second derivative, but in the y-direction
(along a~ 2 ), when the wave is propagating in the x-direction. The force and the motion
are in the y-direction whereas the wave propagates in the x-direction. Hence, the wave
is a shear wave. (Strictly speaking, this type of wave can be termed transverse wave.
7.2. Dust Lattice Waves in 2D 105

Figure 7.5: Hexagonal lattice with the lattice vectors a1 and a2 . The expression in
brackets (`, m) gives the 2D particle position in units of the interparticle distance b.

However, we like to reserve the term “transverse wave” for the out-of-plane wave discussed
in Sec. 7.7.)
That yields
1 2 X e−r̃κ  2 
∞ !
2 `qb
 
2 2 2 2
ω +iβω = ω m 3 + 3r̃κ + r̃ κ − r̃ (1 + r̃κ) sin .(7.11)
2π pd `,m=1 r̃5 2

The shear wave character is seen from the fact that the term after the sum includes m
(denoting the y-direction) whereas the sin-expression includes ` (denoting the x-direction).
The computed dispersion relation of the compressional and shear 2D dust lattice wave
is shown in Fig. 7.6. The compressional mode has a form that is very similar to the 1D-
case: For long wavelengths qb  1 the dispersion is acoustic. For shorter wave lengths
the compressional mode becomes dispersive and attains a maximum near qb = π. In
contrast, the shear mode is nearly acoustic for all wavelengths. Thus, the shear mode is
only little dispersive.
It is readily seen here, that the sound speed of the compressional mode is much larger
than that of the shear mode. The sound speeds of the compressional and shear mode are
given by [94]
ccomp = c0 fcomp (κ) cshear = c0 fshear (κ) (7.12)
where
1 X e−r̃κ
∞
v !
15 9
u
u
fcomp = t (1 + r̃κ) + r̃2 κ2
4π r̃ r̃ 8 8
106 7. Waves in Strongly Coupled Dusty Plasmas

and

1 X e−r̃κ
∞
v !
3 3
u
u
fshear = t − (1 + r̃κ) + r̃2 κ2
4π r̃ r̃ 8 8

where the summation is over all possible distances in the hexagonal lattice r̃. Comparing
the sound speed of the compressional 1D and 2D waves one finds only slight differences,
e.g. the factor 15/8 instead of 2 and 9/8 instead of 1. These factors are the effect of the
hexagonal lattice structure. The sound speed of the shear mode contains a negative term
which substantiates that the sound speed of the shear mode is smaller than that of the
compressional.
This dispersion holds for finite values of the screening strength κ. For pure Coulomb
interaction κ = 0 the sum in the compressional dispersion relation (7.10) would diverge.
√
Thus, for pure Coulomb interaction it is found that ω ∝ q for long wavelengths (and
thus the sound speed c = ω/q → ∞ for q → 0).
Moreover, the wave dispersion has been discussed for a compressional wave propagat-
ing along a~ 1 and a shear wave along a ~ 2 . In the general case there can be any arbitrary
~
angle θ of the wave propagation q relative to the lattice orientation a ~ 1 which has an
influence on the dispersion relation [95]. However, that influence only manifests in the
very short wavelength regime qb > 2.5 when the exact position of the nearest neighbors
is probed by the wave.

Figure 7.6: Dispersion relation of the 2D dust lattice wave for different values of the
screening strength κ (without damping). The solid lines represent the compressional mode,
the dashed lines the shear mode.
7.3. Compressional 1D Dust Lattice Waves: Experiment 107

Figure 7.7: a) Experimental setup for the excitation of 1D dust lattice waves in a chain of
dust particles. b) Snap shots of the dust chain oscillation. The first particle shows strong
“blooming” when hit by the laser beam. One can see that a wave is propagating into the
dust chain. After [96].

7.3. Compressional 1D Dust Lattice Waves: Experiment

Compressional dust lattice waves in 1D and 2D systems have been identified and measured
by Homann et al. [96, 97]. Here, we start with the presentation of the 1D results.
A linear chain of dust particles has been realized by placing a rectangular barrier on
the electrode (see Fig. 7.7a and Sec. 8.1.). Waves in the dust chain have been excited
by focusing a laser beam onto the first particle. The laser beam is then switched “on”
and “off” periodically and a wave is launched into the linear chain of dust particles. The
oscillations are clearly seen in the video snap shots (Fig. 7.7b). It is also obvious that
the amplitude of the wave decreases with distance in the chain due to friction with the
neutral gas. The wave motion is then analyzed directly from the video data in terms
of phase and amplitude. Therefore the oscillations of each particle in the chain relative
to its equilibrium position is identified (see Fig. 7.8a). It is immediately seen that the
phase of the oscillation progresses from one particle to next. Correspondingly also the
oscillation amplitude decreases. The phases and amplitudes of all particles are then
derived and analyzed as a function of equilibrium position (Fig. 7.8b,c). The observed
linear dependence of the phase with particle position directly reflects the wave length
λ and thus the real part of the wave vector qr = 2π/λ. The amplitude of the wave is
found to decrease exponentially with distance from the first particle. This determines the
damping length L and correspondingly the imaginary part of the wave vector qi = 1/L.
Thus, the wave in the chain propagates as

x = A exp (iqx − iω t) = A exp (i(qr + iqi )x − iω t)

= A exp (iqr x − iω t) exp(−qi x) ,
108 7. Waves in Strongly Coupled Dusty Plasmas

where A is the oscillation amplitude of the first particle and x measures the distance from
the first particle. The wave vector qr and qi depend on the wave frequency ω.
By measuring the imaginary and real part of the wave vector in such a way for dif-
ferent wave frequencies, the entire dispersion relation is identified. Fig. 7.9 shows the
wave vectors as a function of wave frequency. The measured dispersion relation is then
compared with the theoretical dispersion relation of the dust-acoustic wave according to
Eq. (6.9) and of the dust lattice wave according to Eq. (7.5). It is clearly seen that the
DAW does not match the measured dispersion relation whereas the DLW is in very good
agreement. Thus, the 1D dust chain does not exhibit DAW-like wave motion in contrast
to the finding of Section 6.1. where extended systems with DAW dispersion have been
found. Here, the dispersion is clearly of DLW type.
The measured dispersion relation also allows to determine the screening strength κ.
This is demonstrated in Fig. 7.9 where the measured dispersion is shown in comparison
with theoretical curves for a range of κ-values. Best agreement between experiment and
theory is obtained for a screening strength of κ = 1.1, reasonable agreement is found in

Figure 7.8: a) Particle oscillation relative to the equilibrium position as a function of

time. The numbers denote the particle in the chain, where the first particle is the one hit
by the beam. b) Phase and c) amplitude of the particle oscillation derived from a) as a
function of position in the chain for different excitation frequencies. After [96].
7.3. Compressional 1D Dust Lattice Waves: Experiment 109

the range between κ = 0.7 and 1.5, thus

κ = b/λs = 1.1 ± 0.4 . (7.13)

This means that the screening length λs in the dust system is of the order of the inter-
particle distance b (or the other way round). In this case

λs ≈ b = 930 µm .

This value of the screening length is of the order of the electron Debye length. We will
return to the discussion of the screening problem below.

Figure 7.9: a) Measured real and imaginary wave vectors as a function of wave frequency.
The measured dispersion relation is compared to the theoretical dispersion relation of the
DAW and DLW. b) Measured dispersion relation in comparison with theoretical DLW
dispersions for various values of κ. In this Figure, normalized units ω/ωpd (using the
measured dust charge Zd ) and qb (using the measured interparticle distance b) have been
used. After [96].
110 7. Waves in Strongly Coupled Dusty Plasmas

7.4. Compressional 2D Dust Lattice Waves: Experiment

Experiments which are similar to those described above have been performed to measure
the dispersion relation of the compressional DLW in two dimensions [97]. In the 2D case,
the laser beam of an Argon ion laser was expanded into a line focus and directed onto
the first row of particles in a 2D plasma crystal. By periodic modulation of the laser
power a plane wave was launched in the plasma crystal (see Fig. 7.10a). Again, the wave
motion of the dust was analyzed in terms of the phase and amplitude (Fig. 7.10b,c) as a
function of distance from the excitation region. Similarly, the phase dependence directly
reflects the wavelength λ and the amplitude decrease the damping length L for a given
excitation frequency. In that way the dispersion relation of the wave has been measured
(Fig. 7.10d,e). The measured dispersion was found to show good agreement with the
2D DLW dispersion. From this comparison the screening strength is determined to be
κ = 1 ± 0.3. So, also in 2D the interparticle distance b and shielding length λs are
comparable and are found to be close to the electron Debye length.

Figure 7.10: (a) Scheme of the experimental setup for the excitation of 2D dust lattice
waves. (b) Phase and (c) amplitude of the dust particle motion as a function of distance
from the excitation region for an excitation frequency of 2.8 Hz. (d,e) Real and imaginary
wave vector as a function of frequency. The symbols denote the experimental data. The
lines indicate the dispersion relation of the 2D DLW for various values of the screening
strength κ. After [97].
7.5. Shear 2D Dust Lattice Waves: Experiment 111

7.5. Shear 2D Dust Lattice Waves: Experiment

Shear dust lattice waves have been observed by Nunomura et al. [98] also using laser beam
excitation. There, a short pulse of a narrow laser beam excites a row of particles along the
direction of the beam (see Fig. 7.11). The velocity pulse created by the beam propagates
in a direction perpendicular to the beam. The dust particle motion and pulse travel
direction are perpendicular, thus a shear wave is observed here. The outward velocity
of the beam is much smaller than for a compressional pulse and in agreement with the
acoustic velocity of the shear wave, see Eq. (7.12).
Recently, Nunomura et al. [99] have investigated shear and compressional waves in
great detail by laser excitation. There, they also have studied the wave propagation along
different lattice orientations and found reasonable agreement with the theoretical DLW
dispersion relation.

7.6. Mach Cones

When an object moves through a medium with a velocity faster than the wave speed in
that medium a V-shaped disturbance, the Mach cone, is excited. This phenomenon is

Figure 7.11: Shear dust lattice waves. (a) Dust particle velocity vectors at certain time
steps after a laser beam pulse. The initial laser beam pushed the particles in the central
region from right to left. (b) Velocity profiles perpendicular to the beam direction. The
central bar indicates the excitation region. From [98].
112 7. Waves in Strongly Coupled Dusty Plasmas

well known, e.g. from the sonic boom behind a plane at supersonic velocity. Similarly,
Mach cones can be observed in dusty plasmas using objects faster than the acoustic speed
of the DLW.
The formation of a Mach cone is illustrated in Fig. 7.12e. The supersonic object moves
at a velocity V that creates disturbances on its way at any instant. The disturbances
propagate through the medium at the sound speed c. Since c < V the front of the
disturbance lags behind the object, thus forming the well-known V-shaped front, the
Mach cone. The Mach cone has an opening angle µ that satisfies the relation
c
sin µ = . (7.14)
V
Thus from a single measurement of the opening angle µ the sound speed of the DLW is
readily obtained from which other parameters like particle charge or screening strength
can be determined.
Mach cones in dusty plasmas have first been observed by Samsonov et al. [100, 101].
There, dust particles which accidentally are trapped below the actual 2D plasma crystal
are found to move at large, supersonic, speeds at low gas pressure. The disturbance by
these lower fast-moving particles excites a Mach cone in the upper plasma crystal.
In a different experiment [102] Mach cones in plasma crystals have been generated
using the focal spot of a laser beam that was moved at supersonic speeds V through the
crystal using a moving galvanometer scanning mirror (Fig. 7.12a). The laser technique
allows the formation of Mach cones in a repetitive and controllable manner.
Figure 7.12b shows a video snap shot where the Mach cone created by the laser spot
is easily seen. The opening angle here is about 45◦ . A more detailed picture is obtained
when investigating the absolute values of the dust particle velocities. The particle speeds
are shown in a gray-scale plot in Fig. 7.12c. Here, a strong first Mach cone is easily seen.
However, additional secondary and tertiary Mach cones are also observable. The first
strong cone is just the expected behavior for the Mach cone as described above. The
additional features arise from the dispersive nature of the DLW at shorter wavelengths
[103]. The laser spot creates wave disturbances at all wavelengths. Due to dispersion
short wavelength waves travel at a different velocity than those at long wavelengths. This
makes the picture described in Fig. 7.12e more complicated: Like the wave pattern of a
moving ship (which is not a Mach cone as described above), the secondary and tertiary
Mach cones can be interpreted as interference patterns of the waves launched by the
moving laser beam.
From the measurement of the opening angle µ of the first Mach cone at various laser
spot velocities the Mach cone relation is verified (see Fig. 7.12d). The sound speed
measured here is about 20 mm/s. This is again a very small value which demonstrates
that the dynamic processes in dusty plasmas occur on a long time scale.
The Mach cone in Fig. 7.12 is a compressional Mach cone due to excitation of compres-
sional waves. Shear Mach cones by the excitation of shear waves have been demonstrated
7.6. Mach Cones 113

Figure 7.12: Mach cones in dusty plasmas. (a) Scheme of the experimental setup. (b)
Video snap shot of the dust crystal. The laser spot moves from right to left. A V-shaped
disturbance is clearly observable. (c) Gray-scale map of the dust particle velocities. (d)
Test of the Mach cone relation. Plot of 1/ sin µ of the measured cone angle µ as a function
of laser spot velocity V. From this, the sound speed is measured to be c = 19.9 mm/s. (e)
Sketch of the Mach cone formation.

by Nosenko et al. [104]. Shear Mach cones are observed at much lower laser spot velocities
V due to the much smaller acoustic velocity of the shear waves.
Mach cones are discussed as a diagnostic tool: Mach cones are assumed to be observ-
114 7. Waves in Strongly Coupled Dusty Plasmas

able in the rings of Saturn by the Cassini spacecraft after its arrival at Saturn in 2004
[105]. In Saturn’s rings, large boulders moving in Keplerian orbits have supersonic speeds
relative to the smaller dust particles which move at speeds determined by their electro-
static interactions with Saturn’s plasma environment (see Sec. 10.5.). The observation of
Mach cones would allow detailed studies of the plasma conditions in the rings.

7.7. Transverse Dust Lattice Waves

All the above mentioned wave types and experiments have dealt with particle motions
in the plane of the 2D plasma crystal. We now turn to the transverse dust lattice wave.
There, an out-of-plane (vertical) particle motion perpendicular to the wave propagation
is expected. Such vertical displacements are stabilized against the Coulomb repulsion of
the particles by the vertical confinement potential. The external confinement is essential
for this type of wave. Otherwise the wave would not be stable.
For the dispersion of the transverse DLW vertical displacements zn in a 1D chain of
particles are considered (see Fig. 7.13). The equation of motion then reads

md z̈n − md βżn + md ω20 zn = kz (zn−1 − 2zn + zn+1 ) , (7.15)

where ω0 is the strength of the vertical confinement [compare Eq. (3.22)] and kz is the
vertical “spring” constant. Using δz = zn − zn−1  b, the spring constant is given by
δz q
δz δz Z2d e2 −κ
2 2
Fz = F(r) = F( δz + b ) √ ≈ F(b) = e (1 + κ) δz = kz δz ,
r δz2 + b2 b 4πε0 b3
where Fz is the vertical component of the Coulomb force F(r) = −∂φ(r)/∂r. The fact
that here the spring constant involves the first derivative of the particle interaction φ
(in contrast to the second derivative for the compressional and shear wave) states that
the equilibrium situation is essentially unstable. For repulsive interaction the transverse
elongations would grow indefinitely. The transverse oscillations are only stabilized by the
additional counterforce due to the external confinement.

Figure 7.13: Sketch of a linear dust arrangement with transverse particle displacements.
7.7. Transverse Dust Lattice Waves 115

Figure 7.14: Dispersion of the transverse dust lattice wave for different values of κ without
damping. The frequency scale, here, is normalized to ω0 , not ωpd . Also note the limited
frequency scale.

Following the above procedure, the dispersion relation of the transverse DLW is given
by [106]
!
2 1 qb
ω + iβω = ω20 − ω2pd e−κ (1 + κ) sin2 . (7.16)
π 2

One can see that the influence of the vertical confinement ω20 is necessary to yield a
stable dispersion relation.∗ It is interesting to note that this wave is a backward wave
(∂ω/∂q < 0), that is phase and group velocities move in opposite directions. In addition,
the wave frequency approaches a finite value for long wavelengths (ω → ω0 for q → 0).
Such a type of wave is called “optical wave” in analogy to waves known from condensed
matter. For q → 0 this is a synchronous oscillation of all particles at the resonance
frequency of the vertical potential well used for the determination of the dust charge (see
Sec. 3.8.).
The calculated dispersion relation is shown in Fig. 7.14. Transverse dust lattice waves
have been observed by Misawa et al. [107]. In their experiment, a linear chain of dust
particles shows vertical oscillations (see Fig. 7.15) which propagate along the chain. From
∗
Otherwise ω2 ∝ −ω2pd sin2 (qb/2) yielding a positive imaginary value for ω which indicates expo-
nential growth.
116 7. Waves in Strongly Coupled Dusty Plasmas

Figure 7.15: Transverse dust lattice waves. (a) Still image of a 1D particle chain, (b)
Grey scale image of the vertical displacement of the dust particles in the chain. The wave
is seen to propagate backwards. (c) Measured dispersion relation of the transverse DLW.
From [107].

the time traces it is immediately seen that the wave is a backward wave (negative slope
in the space-time diagram). The authors have measured a part of the dispersion relation
where a finite frequency is found for q → 0 and the dispersion also has a negative slope,
as expected for the transverse DLW. However, the overall agreement of the measured and
the theoretical dispersion is not very satisfying.

7.8. Dispersion Relation from Thermal Particle Motion

Recently, a method was developed by Nunomura et al. to determine the dispersion relation
of the various wave modes from the thermal Brownian motion of the dust particles in the
plasma crystal [108, 109]. This powerful technique allows the measurement of the entire
dispersion from a single video sequence.
To derive the dispersion relation from the Brownian motion, first the particle velocities
~v(~r, t) in the 2D particle plane are determined. From that the Fourier components of the
velocity
ZT ZL
2
^v(~q, ω) = ~v(~r, t)e(−i~q·~r+iω t) d~r dt (7.17)
TL
0 0
7.9. A Note on Shielding 117

are determined. Here, L and T are the length and period over which the particle motion
is integrated. The compressional mode is then obtained by taking the components of ~vk~q,
only. The shear mode is derived from the components of ~v ⊥ ~q. This integral is evaluated
for all wave vectors ~q and frequencies ω. The square of the value of this integral is then
proportional to the intensity of the wave at the chosen values of ~q and ω.
Figure 7.16 shows the wave energy density of the Fourier components as a function of q
and ω for the compressional and shear mode of an actual experiment. Following the above
procedure the wave energy is large along distinct lines indicating the dispersion relation
of the two wave modes. The dispersion relations obtained from this Brownian motion
technique are in very good agreement with the theoretical predictions. This demonstrates
the ability of this method to derive the dispersion relation from a single video sequence.
This method has also been applied to a linear chain of particles to measure the trans-
verse mode dispersion (Fig. 7.16 bottom). Again good agreement with the theoretical
dispersion is obtained. This powerful technique has also been applied to finite systems in
the following chapter.

7.9. A Note on Shielding

At this point, a few words on the screening of dust particles in the sheath of rf discharges
are appropriate. The above mentioned wave experiments (as well as the experiments on
finite clusters in the following chapter) have revealed that the screening strength κ =
b/λs is of the order of 1. This means that the shielding length λs is of the order of

Figure 7.16: Dispersion relation of the modes determined from the thermal Brownian
motion. Top: compressional wave, middle: shear wave, bottom: optical wave. Note that
in these figures the wave vector extends from qb = −2π to qb = +2π, thus qb = 0 is in
the center. After [108, 109].
118 7. Waves in Strongly Coupled Dusty Plasmas

the interparticle distance b. The interparticle distance and thus the screening length is
typically a few hundred microns for the experiments with micron-sized particles. Direct
collision experiments in Sec. 5.1. also have revealed a screening length of the same size.
We now would like to discuss the observed screening lengths in some more detail.
The observed shielding of a few hundred microns is close to the electron Debye length
s
0 kTe
λs ≈ λD,e =
ne e2
for typical plasma conditions of ne around 108 to 109 cm−3 and Te around 2 to 5 eV. In
comparison, the ion Debye length
s
0 kTi
λD,i =
ni e2
is about 60 to 100 µm due to the much smaller ion temperature Ti ≈ 0.05 eV. The
combination of these two Debye lengths, the linearized Debye length, is given by
1 1 1
= + .
λ2D λ2D,i λ2D,e
Since the shorter length scale dominates the screening properties the linearized Debye
length is very close to the ion Debye length λD ≈ λD,i  λs . The observed screening length
should be of the order of the linearized or ion Debye length if the ions are responsible for
screening. Obviously, however, only the electrons contribute to screening, here.
This apparent paradox is resolved when ion streaming motion is taken into account.
The experiments presented here have all been performed in the sheath of a plasma dis-
charge. There, the ions stream with Bohm velocity vB towards the electrode. Since the
ion flow velocity is supersonic, one can argue that the ions cannot contribute to shielding.
Khrapak et al. and Hutchinson [29, 30]) have given an expression for the shielding length
with ions drifting at a velocity ui , compare Eq. (3.12), as
λ2D,e
λ2s = , (7.18)
1 + kTe /(kTi + (1/2)mi u2i )
which is also shown in Fig. 7.17. For low drift velocities ui → 0 the screening length
λs = λD whereas for high drift velocities λs = λD,e . Hence, when the ion streaming velocity
is below the ion thermal velocity the appropriate screening length is the linearized Debye
length. If the ion streaming velocity is increased the ions cannot contribute to shielding
since then the ion motion is dominated by the drift and not by the dust potential. For large
drifts ui  vth,i the appropriate screening length is the electron Debye length λD,e . Since
the experiments are performed in the sheath where ui ≥ vB  vth,i electron screening is
dominating.
7.9. A Note on Shielding 119

1
λDe

0.8
λs/λDe

0.6

0.4

0.2 λD≈ λDi

0 −2 −1 0 1 2
10 10 10 10 10
ui/vth,i

Figure 7.17: Screening length λs (in units of the electron Debye length λD,e ) as a function
of ion drift velocity ui (in units of the ion thermal velocity).
120 8. Finite Dust Clusters

8. Finite Dust Clusters

So far, waves in extended (1D and 2D) systems as well as crystalline structures in extended
2D and 3D systems have been presented. Now, we would like to draw the attention to
collective effects in finite systems. Such Coulomb clusters are systems of a small number
of charged particles trapped in an external potential. As early as 1904, such systems
have been analyzed by J.J. Thomson as a model for the structure of the atom [110].
Nowadays, the structure and the dynamic properties of finite systems are easily modeled
on a PC. Coulomb clusters are particularly appealing because of the interplay between the
Coulomb repulsion among the particles and the confinement due to an external potential.
This interplay determines the structure of a cluster and also its dynamics. Structure
and dynamics dramatically depend on the exact particle number N, and the formation
of highly symmetric configurations with “magic” particle numbers is observed in a wide
variety of situations in the classical and quantum world.
Examples of charged-particle clusters of present-day interest are found as ions in Paul
and Penning traps, ions in storage rings of accelerators, electrons on the surface of liq-
uid helium, electrons in quantum dots, brown dwarf stars, charged particles in colloidal
suspensions [111] or, as will be discussed here, charged dust particles in dusty plasmas
[112, 113, 114].

8.1. Formation of Finite Dust Clusters

Finite dust Coulomb clusters are formed by trapping only a small number of dust particles
into a confinement potential provided by a combination of forces described in Chapter 3.
The confinement potential energy V depends on the peculiarities of the setup and the
exploited forces (see Fig. 8.1), but can often be considered as harmonic, i.e.
1 1 1
V(xi , yi , zi ) = md ω2x x2i + md ω2y y2i + md ω2z z2i
2 2 2
1  
= md ω20 x2i + αy y2i + αz z2i . (8.1)
2
Here, ωx,y,z is the confinement strength in x, y, z direction and αy,z = ω2y,z /ω2x is the
relative strength of the confinement with respect to the confinement in x direction (ω0 =
ωx ). By changing the relative confinement strengths αx,y various confinement geometries
from 1D to 2D and 3D can be realized. How this is done in the experiment will be
demonstrated below.
8.2. Structural Transitions in 1D Dust Clusters 121

Figure 8.1: Sketch of the confinement of clusters in an rf discharge. In the vertical

direction a very strong confinement is provided by the electric field force and gravity. In the
horizontal direction a barrier on the electrode provides a shallow horizontal confinement.
The force due to that potential energy Fpot balances the Coulomb repulsion FCoul of the
particles.

8.2. Structural Transitions in 1D Dust Clusters

Linear (1D) dust clusters are easily generated by placing a rectangular metal barrier (e.g.
6 mm height and 5 × 40 mm2 inner dimension) onto the lower electrode (see Fig. 8.2). A
small number of particles (N = 1 to 20) are then dropped into the space charge sheath
above the barrier [115], see also Fig. 8.1.
Vertically the particles are strongly confined due to the balance of electric field force
and gravity (i.e. αz  1). Horizontally the particles are confined due to the barrier on
the electrode. Since the elongation of the barrier is much larger in y-direction than in
x-direction the particles are only weakly confined along y. Thus, αy  1 and the particles
arrange along y in a linear 1D dust cluster.
Figure 8.2b) shows the arrangement of N = 4 to N = 18 particles in the barrier. It
is seen that for N = 4 to N = 9 the particles strictly arrange in a linear arrangement.
When, however, the particle number is increased from 9 to 10 a zigzag transition in the
center of the chain occurs. For 18 particles a zigzag structure is seen nearly throughout
the entire chain.
The reason for that is easily understandable: the confinement in the y-direction com-
presses the chain along its extension. The interparticle distance is smallest in the central
part of the chain. When more and more particles are inserted into the chain the compres-
sion increases until it is easier for the central particles to make a transverse excursion (in
the x-direction). Then, the force along y between neighboring particles due to compres-
sion exceeds the force in x from the confinement [115, 116].
122 8. Finite Dust Clusters

Figure 8.2: a) Scheme of the experimental setup for the confinement of 1D dust clusters.
b) Snap shots of the 1D dust cluster for N = 4, 9, 10 and 18. A structural transition
in the cluster is seen by increasing the particle number. From 9 to 10 particles a zigzag
transition occurs. From [115].

8.3. Structure of 2D Finite Dust Clusters

In 2D clusters, a small number of dust particles N = 1 to 1000, say, are trapped in the
sheath above the lower electrode. Vertically, the particles are as usual strongly confined
by the electric field force and gravity. Here, an additional weak parabolic confinement (of
strength ω0 ) is applied in the horizontal plane. Such a confinement can be realized by a
shallow circular parabolic trough in the electrode (see Fig. 8.3). Hence, here, ω2x = ω2y 
ω2z (αy = 1, αz  1).
Under the interplay of the horizontal parabolic confinement and their Coulomb repul-
sion the particles arrange in concentric shells (see Fig. 8.4). The 2D clusters are sometimes
said to form a “periodic table”. The structure and their dynamic properties dramatically
depend on the particle number N. When changing from 5 to 6 particles (or from 15
to 16) a new shell opens up. Like in atomic and nuclear physics there exist “magic”
particle numbers of high dynamic stability, e.g. the N=19 (1,6,12) cluster. The notation
(N1 , N2 , N3 , . . .) refers to N1 particles in the inner ring, N2 in the second and so on. These
clusters will be analyzed below in more detail.
These finite 2D clusters can be described in terms of their total energy

1 XN
Z2d e2 X exp(−rij /λD )
N
2 2
E = m d ω0 ri + , (8.2)
2 i=1
4πε0 i>j rij

where ~ri = (xi , yi ) is the 2D position of the i-th particle in the horizontal plane and
rij = |~ri − ~rj |. The strength of the horizontal confinement is denoted by the horizontal
8.3. Structure of 2D Finite Dust Clusters 123

Figure 8.3: a) Scheme of the experimental setup for the confinement of 2D dust clusters.
b) Snap shots of the 2D dust cluster for N = 34 and 145. From [117].

resonance frequency ω0 . The first term is the potential energy due to the confinement in
the horizontal plane and the second is the Coulomb repulsion of the particles.
The equilibrium structure of these systems is derived from the minimum of the total
energy [118]. Experimentally they have been observed, e.g. by Juan et al. [112] and
Klindworth et al. [113]. The observed cluster structures are in perfect agreement with the
theoretical predictions. We will illustrate that in the following.

8.3.1. Example of 2D Cluster Configurations: Configuration (0,6)

The structure of a 2D cluster configuration will be illustrated using N = 6 particles under
Coulomb interaction (λD → ∞). We will study two possible configurations, namely the
(0, 6) and (1, 5) configuration where either 6 particles are on a single ring or 5 particles
are on the outer ring with a central particle (see Fig. 8.5). We assume that the particles
are ordered in angular position and that only radial displacements can occur.
For the (0, 6) configuration the potential energy is given by

1 X N
1
Epot = md ω20 r2i = md ω20 Nr26 ,
2 i=1
2

where r6 is the distance of the 6 particles from the center of the configuration. For the
Coulomb energy we have to determine the distances between the different particles: They
are rij = r6 for 1 − 2, 2 − 3, 3 − 4, 4 − 5, 5 − 6 and 1 − 6 (since they form equilateral
124 8. Finite Dust Clusters

Figure 8.4: “Periodic table” of finite 2D clusters with N = 1 to N = 25 particles. These

images are taken from the experiment.

√
triangles); in addition rij = 3r6 for 1 − 3, 1 − 5, 2 − 4, 2 − 6, 3 − 5 and 4 − 6; finally,
rij = 2r6 for 1 − 4, 2 − 5, 3 − 6. These are the 15 possible combinations for i > j at N = 6.
8.3. Structure of 2D Finite Dust Clusters 125

Figure 8.5: Two possible configurations of a 6-particle cluster: (a) (0,6) and (b) (1,5).
The numbers denote the different particles.

Thus, the Coulomb energy is

Z2d e2 X 1
N
Z2 e2
!
Z2 e2
!
6 6 3 6 3
ECoul = = d +√ + = d 6+ √ + .
4πε0 i>j rij 4πε0 r6 3r6 2r6 4πε0 r6 3 2
| {z }
10.97
The total energy is then given by
1 Z2 e2
E = md ω20 Nr26 + d 10.97 (8.3)
2 4πε0 r6
The distance r6 is undefined at the moment. It must be evaluated from the condition that
r6 is the equilibrium distance. For r < r6 the Coulomb energy dominates, for r > r6 the
potential energy. The equilibrium is found from
∂E Z2d e2 Z2d e2 2 10.97
=0 → md ω20 Nr6 − 10.97 = 0 → r36 =
∂r 4πε0 r26 4πε0 md ω20 2N
Inserting this into Eq. (8.3) yields
 !2 1/3  v s 
Z2d e2 md ω20  10.972 2N 
u
3
u
3
E6 =  N t + 10.97
4πε0 2 (2N)2 10.97
 !2 1/3
Z2 e2 md ω20 
= 16.95  d .
4πε0 2
126 8. Finite Dust Clusters

When measuring the distances and energies in units of

" #1/3  !2 1/3
Z2d e22 Z2d e2 2
m d ω0 
r0 = and E0 =  (8.4)
4π0 md ω20 4π0 2

the equilibrium distance and energy are found as

s
3 10.97
r6 = = 0.97 and E6 = 16.95 .
2N
In these normalized units the energy and the distances do not depend on the physical
parameters like particle charge, mass etc. They only define the absolute values of the
energy and the absolute size of the cluster, but they do not have an influence on the
structure, on the dynamics and so on. Structure and dynamics are determined only by
the particle number N.
When we analyze the confinement near the equilibrium position we can develop the
energy into the different orders, like
1 ∂2 E 2
E(δr) = E6 + δr + . . . .
2 ∂r2
The term linear in the displacement from the equilibrium position δr = r − r6 is zero since
r6 is the equilibrium position where ∂E/∂r = 0. The second order term results in
∂2 E Z2 e2 2
!
1 1 2
= md ω20 2N + 10.97 d 3
= md ω20 2N + 10.97
∂r2 r=r6
2 4πε0 r6 2 10.97/2N
1
= md ω20 6N . (8.5)
2
Thus it is found that the 6 particles of the (0,6) configuration are experiencing a har-
monic confinement. When identifying this confinement with a harmonic potential well of
resonance frequency ω for N particles we find
1 ∂2 E 2 1
δr = Nmd ω2 δr2 → ω2 = 3ω20 . (8.6)
2 ∂r2 2
√
This means that the 6 particles experience a potential well of strength 3ω0 . Since
we have considered only radial displacements in this analysis the corresponding particle
motion is a coherent radial oscillation of all 6 particles. Thus, the cluster grows and
decreases in size periodically. Therefore, this type of motion is called breathing mode.
The breathing mode is found to be at the fixed frequency given by Eq. √ (8.6). Moreover,
it is found that the breathing mode always has the frequency ω = 3ω0 independent
of particle number N. This is a general result that holds for pure Coulomb interaction
between the particles. When a screened interaction is taken into account the frequency
of the breathing mode increases with increasing screening strength.
8.3. Structure of 2D Finite Dust Clusters 127

8.3.2. Example of 2D Cluster Configurations: Configuration (1,5)

The alternative configuration consists of one central particle surrounded by another 5.
The potential energy is readily found as

1 X N
1
Epot = md ω20 r2i = md ω20 (N − 1)r25 ,
2 i=1
2

where r5 is the distance of the 5 outer particles from the center. The Coulomb energy is

Z2 e2 X 1
N
Z2 e2
!
Z2 e2
!
5 5 5 5 5
ECoul = d = d + + = d 5+ + .
4πε0 i>j rij 4πε0 r5 1.18r5 1.90r5 4πε0 r5 1.18 1.90
| {z }
11.87
Here, the distance from the central particle to the other 5 is r = r5 , between particles
2 − 3, 2 − 6, 3 − 4, 4 − 5, and 5 − 6 it is r = 1.18r5 and between 2 − 4, 2 − 5, 3 − 5, 3 − 6,
and 4 − 6 it is r = 1.90r5 .
The equilibrium distance is found as
∂E Z2d e2 Z2d e2 2 11.87
=0 → md ω20 (N − 1)r5 − 11.87 → r35 =
∂r 4πε0 r25 2
4πε0 md ω0 2(N − 1)
Inserting this into the total energy, one obtains in normalized units

r5 = 1.059 and E5 = 16.82 .

Hence, the cluster configuration (1, 5) has a lower energy than (0, 6) and thus is energet-
ically favored. This is also seen in Fig. 8.4 where the (1, 5) configuration is found for 6
particles.
Finally, the radial confinement is given by
∂2 E 1 Z2 e2 2
= md ω20 2(N − 1) + 11.87 d
∂r2 r=r5
2 4πε0 r35
!
1 2
= md ω20 2(N − 1) + 11.87
2 11.87/2(N − 1)
1
= md ω20 6(N − 1) . (8.7)
2
Since here only the outer N − 1√= 5 particles take part in the breathing mode the same
breathing mode frequency ω = 3ω0 is found also for the (1, 5) configuration.
128 8. Finite Dust Clusters

8.4. Structure of 3D Finite Dust Clusters

It is also possible to confine 3D dust clusters under laboratory conditions. The confinement
potential is isotropic in 3D and does not lead to void formation. The confinement has been
achieved using the combined interaction of thermophoretic levitation and a horizontal
boundary due to glass walls (see Fig. 8.6). A glass box is placed onto the lower electrode.
The dielectric walls charge negatively and therefore provide an inward electric force on
the negative dust particles for horizontal confinement. The lower electrode is heated and
induces a temperature gradient in the neutral gas. The temperature gradient provides an
upward thermophoretic force that, at least partially, compensates the gravitational force
on the particles [119]. By tuning the electrode temperature and plasma properties a 3D
isotropic confinement potential is formed to confine the dust particles (ω2x = ω2y = ω2z =,
i.e. αy = αz = 1).
The clusters in this three-dimensional confinement also arrange in highly ordered struc-
tures [121]. The systems form concentric spheres, arranged in an “onion shell” structure,
for which the name “Yukawa ball” has been established [114, 122, 123, 120]. Like 2D
finite clusters where the particles arrange in concentric rings, 3D clusters have defined
spheres. Since a 3D structure is difficult to visualize, the structure is most clearly seen
in cylindrical coordinatesq(ρ, z), where the horizontal coordinates (x, y) are mapped onto
the radial position ρ = x2 + y2 and plotted versus the vertical coordinate z. An ex-
perimental dust cluster is shown in Fig. 8.7. The cluster consists of N = 190 particles
and is arranged in 4 concentric shells in a (2, 21, 60, 107) configuration, i.e. 2 particles
are on the innermost shell, followed by 21, 60 and 107 particles on the second, third

Figure 8.6: a) Scheme of the experimental setup for the confinement of 3D dust clusters.
Thermophoretic levitation is used to (partially) compensate the gravitational force. b)
Scheme of the 3D stereoscopic imaging unit. From [119, 120].
8.4. Structure of 3D Finite Dust Clusters 129

Figure 8.7: 3D dust cluster in an experiment with N = 190 particles. a) ρ-z plot. The
cluster is seen to have 4 shells. b) Wigner-Seitz-cell analysis of the outer (4th) shell and
c) of the 3rd shell. Pentagons are marked blue, hexagons are green. Defect polygons with
more than 6 nearest neighbors are colored red. After [114].

and forth shell, respectively. The structure within each shell is visualized in Fig. 8.7b,c)
where a Wigner-Seitz-cell construction has been performed on the individual shells. The
shells consist of hexagons and pentagons (like the famous C60 buckyballs or an ordinary
footballs). A certain number of pentagons is needed to ensure the curvature of the sphere
and to form closed shells. But also some defects, i.e. particles with more than 6 nearest
neighbors are found.
A selection of smaller clusters (N < 100) is shown in Fig. 8.8. There, a 3D recon-
struction of the structure is shown. Also these clusters arrange in nested spherical shells
ranging from one shell with central particle (N = 17) to three shells (N = 91). From
this 3D bond structure also the highly ordered arrangement is readily seen. Analogously
to the 2D case, also here a “periodic table”-like construction of the clusters is observed.
Typically, a new shell opens up, when the inner shell has 12 particles [124, 125].
However, there is an interesting difference between 2D and 3D clusters concerning the
structural properties and screened particle interaction. In 2D, the observed cluster struc-
tures (i.e. occupation numbers of the different rings) is nearly independent of the particle
interaction: pure Coulomb, Yukawa or even logarithmic interaction potentials (almost)
always yield the same cluster structure [126]. In contrast, in 3D, the occupation number
for fixed total particle number N of inner shells is higher for Yukawa interaction com-
pared to the pure Coulomb case. Consequently, outer shells have lower particle numbers
for Yukawa interaction than for Coulomb interaction. Hence, already the structure of 3D
clusters reflects the shielding strength [122, 123, 120] which is found from the experiments
130 8. Finite Dust Clusters

Figure 8.8: Cluster configurations reconstructed from single video snap shots with N =
91, 52, 31, and 17. The Yukawa balls consist of concentric shells with the configurations
(4, 25, 62) for the N = 91, (11, 41) for the N = 52, (5, 26) for the N = 31, and (1, 16) for
the N = 17 cluster. From [120].

to be of the order of κ ≈ 1. Therefore, the term “Yukawa ball” is fully justified.

8.5. Normal Modes in 2D Finite Clusters

After the discussion of the static equilibrium properties of finite systems we like to delve
into the dynamic properties of these systems. The breathing mode discussed above was
a first example in that direction.
The dynamic properties of finite systems cannot be described in terms of wave disper-
sions as in the case of extended systems. First, only a certain number of wavelengths can
fit into finite systems. Moreover, due to the boundaries of finite systems, where phase
jumps and reflections can occur, the distinction between shear and compressional modes
is not possible. The modes that occur in finite systems have both compressional and shear
contributions. For small systems the boundaries are important and dominate the dynamic
properties of the system. Thus, the dynamic properties of finite clusters are described by
their normal modes which replace the dispersion relation of infinite systems [127]. We
will treat the problem for the 2D case, here, for simpler visualization. The extension to
3D systems is easily done.
The starting point is the total energy, compare Eq. (8.2),
X
N X
N
exp(−κrij )
E= r2i + , (8.8)
i=1 i>j
rij
8.5. Normal Modes in 2D Finite Clusters 131

Figure 8.9: The 6 eigen modes of a N = 3 cluster with the corresponding mode frequencies
ω2 (in units of ω20 /2).

where we have used normalized distances and energies according to Eq. (8.4). The first
term is the confinement potential and the second the screened interaction. The screening
strength is in these finite clusters accordingly defined as κ = r0 /λD .
For multi-particle systems it is known from theoretical mechanics that the normal
modes are obtained from the dynamical matrix (for 2D systems)
∂2 E
!
∂2 E
 ! 

 ∂xi ∂xj ∂xi ∂yj 

A=  !  , (8.9)
∂2 E
!
∂2 E
 
 
∂yi ∂xj ∂yi ∂yj
that contains the second derivative of the total energy with respect to the particle coor-
dinates xi and yi . These second derivatives are similar to that in Eq. (7.10) where we
have seen in the determination of the DLW that the effective “spring constant” of the
interaction among the different particles is related to the second derivative of the inter-
action potential. The dynamical matrix A is just the generalization of this concept for
multi-particle systems in more than one dimension.
In extended form, the 4 elements that constitute the dynamical matrix A are them-
132 8. Finite Dust Clusters

selves 1N × 1N matrices that contain the possible combinations of i and j, namely, e.g.
∂2 E ∂2 E
 
 ... 
2
!  ∂x1 ∂y1 ∂x1 ∂yN 
∂E 
.. .. .. 
= . . . .
 
 
∂xi ∂yj 
2 2


∂E ∂E 
...
 
∂xN ∂y1 ∂xN ∂yN
Now the eigen value problem
     
x1 x1 x1
 .   .   . 
 .   .   . 
 .   .   . 
xN xN xN
     
2
     
A  = λ  =ω   (8.10)

 y1 


 y1 


 y1 

 ..   ..   .. 

 . 


 . 


 . 

yN yN yN
has to be solved∗ .
The eigen values and eigen vectors of A describe the normal mode oscillations of
the finite clusters. The eigen values λ are the oscillation frequencies ω2 and the eigen
vectors (x1 , . . . , xN , y1 , . . . , yN )T describe the mode oscillation patterns. The matrix A is
2N × 2N, thus there are 2N eigen modes for a system with N particles in two dimensions
(consequently, in 3D, the matrix A is 3N × 3N with 3N eigen modes).
This is demonstrated for the simple case of N = 3 particles in Fig. 8.9 where the 2N
eigen modes are presented. Modes that occur in any cluster are the two sloshing modes
(i.e. oscillations of the entire cluster in the horizontal confining potential, modes number
5 and 6), the rotation of the entire cluster (mode number 2) and the breathing mode
(i.e. coherent, purely radial motion of all particles, mode number 1). For the 3-particle
cluster also two “kink” modes are found (mode number 3 and 4). Naturally, the mode
frequency for the two sloshing modes (5,6) is ωslosh = ω0 since the cluster oscillates in
the confinement as a whole, for the rotation (2) of the entire cluster ωrot = 0 since there
are no restoring forces and the cluster is always in equilibrium. The breathing mode (1)
has the interesting property
√ that for pure Coulomb interaction, i.e. κ = 0, its frequency
always is ωbreath = 3ω0 independent of particle number (see Sec. 8.3.1.). For shielded
interaction (κ > 0) the frequency of the breathing mode increases and slightly depends on
particle number N (see Fig. 8.10). Also the frequency of the “kink” mode increases with
κ. Of course, the frequencies of the sloshing mode and cluster rotation are independent
of κ and stay at their frequencies ωslosh = ω0 and ωrot = 0, respectively.
∗
This is easily seen from a simple analogy: Starting from the equation of motion for a simple spring

mẍ = −kx → −ω2 x = −(k/m)x

assuming an oscillatory solution x → x exp(−iωt). Now, for the many particle case k/m is replaced by
8.6. Modes from Thermal Particle Motion 133

Figure 8.10: Evolution of the mode frequency ω2 (in units of ω20 /2) of the 6 eigen modes
of the 3 particle cluster. After [117].

8.6. Modes from Thermal Particle Motion

As for the case of the waves, the normal modes can be extracted experimentally from the
thermal Brownian motion of the dust particles [117]. Likewise, first, also for the finite
clusters the velocity of the Brownian motion of all particles ~vi (t) is determined from a
video sequence. Then, the contribution of the thermal motion to each of the eigen modes
is determined from the projection of the thermal velocities onto the eigen mode pattern
by
X
N
f` (t) = ~vi (t) · ~ei,` ,
i=1

where ~ei,` is the eigen vector for particle i in mode number `. The function f` (t) is
the contribution of the thermal motion to the eigen mode ` in the time domain. For
the second derivatives, i.e the dynamical matrix A, and x becomes the vector of particle positions in x, y.
134 8. Finite Dust Clusters

comparison, in the wave analysis (Sec. 7.8.), the Brownian particle velocites were projected
onto the chosen wave vector ~q. For the clusters, the eigen mode pattern takes the role of
the wave vectors.
Finally, the spectral power density of each mode `
ZT 2
2
S` (ω) = f` (t)eiω t dt (8.11)
T
0

of f` is calculated. The spectral power density is the square of the Fourier transform of
f` (t). The power spectrum contains the contribution of the thermal motion to each of the
2N eigen modes in the frequency domain and is proportional to the energy stored in the
modes. It can be shown that
Z
∞ Z
∞
2 1 1
S` (ω) dω = hv` i , thus md S` (ω) dω = md hv2` i = E`
2 2
0 0

is the energy stored in mode number `.

For the 3-particle cluster the power spectrum obtained from the thermal motion of the
three particles is shown in Fig. 8.11. Darker colors represent higher spectral power density.
The measured spectrum is then also compared with the theoretical mode frequencies, i.e.
the eigen values of the dynamical matrix. Since the theoretical mode frequencies depend
on ω0 and κ, the theoretical mode frequencies were fitted to the experimental data by
changing these two values. It is seen, that the observed power spectrum is in very good
agreement with the expected mode frequencies.
This technique can be applied to clusters of any size and has been demonstrated for
clusters with several hundred particles. From the analysis of the mode spectra the best-
fit parameters ω0 and κ are derived. Note, that only these two parameters are fitted
to match 2N modes, thus giving the obtained values a high credibility. Then, from the
measured absolute values of cluster size or interparticle distance and by comparison with
the scaling parameter r0 = [2Z2d e2 /(4π0 md ω20 )]1/3 also the particle charge can be directly
extracted. The particle charge was found to be in the presented experiments about 10 000
elementary charges on a 2a = 9.5 µm particle and the screening strength κ was found in
the range between 0.5 and 2 [117].

8.7. Stability
The claim was made that the power spectrum contains the complete dynamic information
on the cluster. This is illustrated here for the very basic dynamic property, the stability
against perturbations.
In order to perturb a cluster configuration a force against the restoring forces of the
cluster need to be applied. If the forces to induce perturbations are large then the system
8.7. Stability 135

Figure 8.11: (a) The 6 eigen modes of a N = 3 cluster with the corresponding mode
frequencies ω2 (in units of ω20 ). (b) Spectral power density of the 3 particle cluster derived
from the thermal motion of the particles. The white dots indicate the best-fit theoretical
values of the mode frequencies. After [117].

is more stable than if only small forces are required. In other words, if the restoring forces
of the cluster are large against an applied perturbation the cluster is stable. Since the
restoring forces directly determine the frequencies of the various modes it is seen that the
mode with the lowest eigen frequency is the one that determines the stability of the entire
cluster. The easiest way to disturb the cluster is along the mode with the lowest eigen
frequency.
The rotational mode always has an eigen frequency of 0 and thus is always the mode
with the lowest eigen frequency. However, a rotation of the entire cluster does not change
the cluster configuration and thus is not really a perturbation of the cluster. So, the
rotation mode is excluded from our following analysis and we seek for the eigen mode
with the lowest frequency besides cluster rotation.
As an example we take the 19-particle cluster (see inset in Fig. 8.12a), which is a “magic
number” configuration (1, 6, 12) due to its hexagonal symmetry of inner and outer ring.
The particle number in inner and outer ring is commensurable. The inner and outer ring
are locked into each other like the teeth of a tooth-wheel. One would expect that this
cluster configuration is very stable, justifying the notion of “magic number”. The power
spectrum to investigate the stability of the cluster is shown in Fig. 8.12b. One sees that
there is a large gap between zero frequency (rotation) and the lowest eigen frequency at
about 0.9 Hz. This is indeed a very large frequency gap and shows that the lowest eigen
mode already has a quite high frequency which means high restoring forces and thus a
high stability.
The corresponding mode patterns are shown in Fig. 8.12a. The already established
136 8. Finite Dust Clusters

Figure 8.12: a) Selected modes of a cluster with N = 19 particles together with a video
snap shot of that cluster. b) Measured mode spectrum of this 19-particle cluster. After
[117].

breathing mode, center-of-mass mode and rotation are indicated. Two modes, that in-
dicate the modes with the lowest frequency are the vortex-antivortex mode and the in-
nershell rotation. The vortex-antivortex mode consists of two vortices, a clockwise and
a counter-clockwise rotation within the cluster. This vortex-antivortex pair describes the
pattern which results in the easiest disturbance of the 19-particle cluster. Similarly low
frequencies are found for the intershell rotation where we find a differential rotation of
inner and outer rings. Vortex-antivortex formation and intershell rotation very often are
the lowest frequency mode in 2D clusters.
For comparison, the mode spectrum of a N = 20 particle cluster ist shown in Fig. 8.13.
This cluster has a configuration of (1, 7, 12) which is not a magic number configuration,
since the number of particles in the inner and outer ring are not commensurate. Hence,
inner and outer ring cannot interlock. Consequently, the intershell rotation is found
at extremely low frequency of about 0.1 Hz, which is drastically less than for the 19-
particle cluster. This demonstrates that the 20-particle cluster is very unstable against
this intershell rotation. In contrast, the breathing mode of this cluster is found nearly at
the same frequency as for N = 19 indicating that the plasma and confinement parameters
of these two clusters are nearly identical.

8.8. Phase Transitions

As a final example of modes in 2D clusters, the phase transition by reducing the gas
pressure is revisited. The mechanism of the phase transition has been introduced in
8.8. Phase Transitions 137

Figure 8.13: Video snap shot of a cluster with N = 20 particles together with the measured
mode spectrum. The breathing mode and the intershell rotation are marked by B and I,
respectively. After [117].

Sec. 5.6.
We have seen there that the ion flow in the plasma sheath results in the formation
of positive ion space charges (“ion focus”) beneath the dust particles. In addition, the
attraction arising from the positive space charge can only be communicated downstream
the ion flow. This leads to the formation of the vertically aligned pairs of particles. With
reduced gas pressure the vertical alignment becomes unstable (“Schweigert instability”)
and gives rise to growing oscillations until a fluid state is reached.
Here, we modify the situation in that we investigate a 2D finite system (with about
40 particles) instead of an extended system [128]. In addition, only one single particle
is placed in the layer below the actual cluster (see Fig. 8.14). So there is only a single
vertically aligned pair of particles. The advantages of such a system are that, first, the
heating effect can be definitely attributed to the single lower-layer particle, and, second,
the full dynamics in terms of the normal modes of the cluster is accessible from the mode
spectra (the mode spectra are obtained only for the cluster of the upper particles, the
lower particle is used as a heat source, only).
As in the case with the extended system, the phase transition is induced by reduction
of gas pressure. With decreasing gas pressure, the single lower layer particle starts to
oscillate about its vertically aligned equilibrium position due to the instability arising
from the non-reciprocal attraction. Due to the mutual Coulomb repulsion, the oscillating
lower particle heats the upper particles. At the lowest gas pressures (below 8 Pa) the
lower particle is still oscillating below upper layer particles, but from time to time this
138 8. Finite Dust Clusters

Figure 8.14: Scheme of the experimental setup to drive a phase transition in finite 2D
clusters. A single particle is confined below the actual cluster. By reducing the gas pressure
oscillations of the lower particle are excited due to the Schweigert instability.

particle may jump from one upper particle to another thus heating different particles.
For illustration of the melting transition of the dust cluster, the particle trajectories
are shown in Fig. 8.15(a). At the highest gas pressure (12 Pa) the particles only slightly
move around their equilibrium positions, at reduced pressure (10 to 11 Pa) the oscillations
become visible from the circular particle trajectories in the cluster center. At even lower
pressure the particles start to exchange equilibrium positions which is an indication of
melting (at 8 Pa an exchange has nearly occurred, at 6 Pa frequent exchanges take place).
The dynamics of the cluster melting process is visible in great detail by analysis of the
power spectra, see Fig. 8.15(b). For 12 Pa, the spectral power density of the individual
modes is concentrated around a quite narrow band of frequencies that closely follows the
mode theoretical frequencies of a solid cluster with a particle charge of Z = 11 000 ± 1000
and a screening length of λD = (1000 ± 500) µm.
With reduced gas pressure (10 and 11 Pa), the spectrum changes completely. All
modes show a maximum at the same frequency fu = 4 Hz which is apparent from the dark
horizontal band in the spectrum. This frequency corresponds to the unstable oscillations.
The dominance of this frequency in all modes is surprising. In addition to this dominant
8.8. Phase Transitions 139

frequency, the underlying mode structure of the crystalline state is still faintly observable
in the spectrum. From the mode spectra it is seen that the unstable oscillations appear
in all the modes although only a single vertically aligned pair exists in this cluster. The
pair starts their unstable oscillations at 11 Pa. Upper and lower particle start to oscillate
around their equilibrium positions. Due to the Coulomb repulsion between the particles
the oscillation is communicated to all particles in the upper layer and is thus visible in all
the modes. It is interesting to note, here, that a single aligned pair is sufficient to drive
the entire system of 40 particles into the liquid state. That means, that the oscillations of
the vertical aligned pair “pump” so much energy into the system that the entire cluster
melts in spite of the still relatively large frictional damping.
Below 8 Pa the situation changes again. The spectrum becomes broad for all modes
and the close relation to the solid-state mode frequencies is lost. From this, it becomes
obvious that the cluster is in a liquid state.
The analysis of the dynamic properties allows to fully substantiate the findings from
the extended two-layer system. Here, the situation is simplified as far as possible and
detailed information about the onset, the frequencies, the energies and the mechanisms
is visualized.

Figure 8.15: Phase transition of a 2D cluster with one additional particle in the lower
layer: a) particle trajectories, b) mode-resolved power spectra, c) mode-integrated power
spectra. After [128].
140 9. Technical Applications of Dusty Plasmas

9. Technical Applications of Dusty Plasmas

After discussing various fundamental aspects of dusty plasmas the next two chapters deal
with more “applied” questions. In this chapter, the role of dust in processing plasmas is
discussed, in the following chapter dusty plasmas in astrophysical situations are presented.
We start with the description of particle growth in low-temperature plasmas which is
followed by the dust removal in processing plasmas and potential issues of dust in fusion
devices as well as application of dusty plasmas to the formation of “new” materials.

9.1. Particle Growth Mechanisms

In technological applications the dust particles are not introduced into the plasma, rather
the dust particles grow (by wanted or unwanted means) in the discharge itself. This
is because typical plasma processes operate in reactive gases, like silane, hydrocarbons,
fluorocarbons or organo-silicons. These gases are required to deposit thin films on sub-
strates or to etch into silicon layers, photoresists or protective films. For example, Silane,
SiH4 , is used to deposit thin (amorphous) silicon films. This is needed in computer chip
manufacturing or in solar cell production to make use of the properties of silicon as a
semiconductor. In the plasma, the silane molecules are dissociated into SiH0...3 which is
deposited onto suitable substrates. There a hydrogen-rich silicon film is deposited. Hy-
drocarbons (methane CH4 , acethylene C2 H2 or similar) are used to make diamond-like
carbon films. There it is intended to grow carbon films that have the properties of di-
amond, especially in view of hardness. These films are used as covering layers to resist
against wear. In addition, HMDSO (Hexamethyldisiloxane) Si2 O(CH3 )6 is used as barrier
layers in PET bottles or Tetra packs. HMDSO provides SiO2 -layers that prevent the
carbon-dioxide in beer, lemonade etc. to diffuse through the walls of the container.
All these reactive gases have the tendency to polymerize under plasma conditions.
The particle growth mechanisms will be discussed in some more detail, here.
As a model system, we will study the growth mechanism in silane, since it is one of the
most investigated systems and is of technological relevance. Silane, SiH4 , has the same
structure as methane (CH4 ), where the 4 hydrogen atoms are on the edges of a tetraeder
and the silicon atom is sitting in the center. Silane can provide a certain insight into the
mechanisms of particle growth. Other reactive gases certainly behave differently due to
the different chemical properties involved, but silane shall serve as a paradigm here.
In a plasma discharge (typically rf discharge), the particles grow from molecules (of Å
size) to a few hundred nanometer in diameter on the time scale of seconds, see Fig. 9.1.
The formation of dust particles works best for high discharge powers and higher gas
pressure. The growth is not homogeneous, but one can identify different phases of particle
9.1. Particle Growth Mechanisms 141

Figure 9.1: Particle size and particle density in a silane discharge as a function of time.
From [2].

growth:
1. Cluster formation: The particles grow from molecules to clusters of a few nanometer
in size. This phase is dominated by plasma chemistry.
2. Agglomeration phase: The particles grow rapidly from a few nanometer to 50
nanometer, say. Correspondingly, the density of particles drops dramatically.
3. Accretion phase: One finds a slow increase in particle size and roughly constant
particle density. This phase is dominated by powder dynamics.
The last two phases are observable in Fig. 9.1. These measurements have been obtained
using laser light scattering. However, this technique is sensitive only down to particle
sizes of about ten nanometer. The initial growth phase, i.e. cluster formation, has to be
attacked by different means, e.g. mass spectrometry.
In the following, we will describe the different phases in some more detail.

9.1.1. Cluster Formation

In the cluster formation phase the dust particles grow from the molecule size to clusters of
a few nanometer. Nanometer sized particles contain of the order of 100 atoms/molecules.
For such small particles (molecules), chemical processes dominate the growth. Possible
chemical reactions can be summarized in the form
Sin Hm + SiH4 −→ Sin+1 Hm 0 + (H,H2 ) products (9.1)
142 9. Technical Applications of Dusty Plasmas

Sin H+ +
m + SiH4 −→ Sin+1 Hm 0 + (H,H2 ) products (9.2)
Sin H− −
m + SiH4 −→ Sin+1 Hm 0 + (H,H2 ) products (9.3)

where reaction pathways are described for neutral, positively charged and negatively
charged molecules, respectively.
In this phase, the particles can be measured by mass spectrometry. With modern
mass spectrometers, particle masses up to 2000 amu can be measured with high temporal
resolution. That means, that clusters with about 60 Si-atoms can be measured with such a
device (Si has a mass of 28 amu). Additionally, mass spectra of positively charged particles
(ions), negative ions or neutrals can be discriminated by applying opposite voltages at
the entrance orifice of the mass spectrometer to repel the unwanted species. The result
of such a measurement is shown in Fig. 9.2.

Figure 9.2: Mass spectrum of negative, positive ions and neutral molecules in a silane
discharge. From [2].

One can easily see that negative ions can be identified up to masses of larger than
1000 amu, whereas positive ions can be seen only up to 400 amu, and neutrals extend
only to 100 amu. This finding clearly suggests that the reaction pathway is dominated by
negative ions. This can be understood from the fact that the plasma potential is typically
positive with respect to the walls. Thus, negative particles, like electrons and the negative
silane molecules are trapped in the plasma. Positive ions are readily forced to the walls
or electrodes. Thus negative molecules have longer residence times in the plasma and can
thus dominate the cluster phase. A further and important reason seems to be the fact
that the reaction chain with negative molecules is also favored from chemistry.
When looking closely at the negative ions in Fig. 9.2, one easily sees a periodic structure
in the mass spectrum. The peaks correspond to clusters with 1, 2, 3 . . . silicon atoms with
some bound hydrogen atoms. In the mass spectrum silicon clusters up to 34 silicon atoms
are easily identified. Probably, even larger clusters exist in the discharge, but have not
been measured with the mass spectrometer in this particular experiment.
9.1. Particle Growth Mechanisms 143

Figure 9.3: Relative concentration of [H]:[Si] in a molecule (cluster) as a function of the

number n of silicon atoms in the cluster. After [2].

In view of the negative clusters the favored reaction pathway is

SiH− −
3 + SiH4 −→ Si2 H5 + H2 (9.4)
Si2 H− −
5 + SiH4 −→ Si3 H7 + H2 (9.5)
...
−
Sin H2n+1 + SiH4 −→ Sin+1 H−
2n+3 + H2 (9.6)

where SiH−3 is the precursor of this reaction chain. This ion is formed by dissociative
attachment, i.e. an electron attaches to the SiH4 molecule and a hydrogen ion is removed
from the molecule, namely

SiH4 + e− −→ SiH−
3 +H .

Following this reaction chain, the ratio of hydrogen to silicon in a cluster then should
be [H]:[Si] = (2n + 1)/n which is indeed found from the mass spectra for small Si-clusters
with up to n = 5 or 6 silicon atoms (see Fig. 9.3). For large silicon clusters with n > 10
the mass spectra show a concentration of hydrogen to silicon that is very close to [H]:[Si]
= 4:3. This means that there are equal number of Si–Si and Si–H bonds. Thus, at free
bonds randomly hydrogen or silicon is attached. From infrared absorption spectroscopy
one knows that there are no double bonds in Si clusters.
The structure of various silicon clusters is shown in Fig. 9.4. The structure of small
clusters (n < 10) is, of course, dominated by the geometry of their chemical bindings.
Larger clusters (n > 20) are elongated ellipsoids, whereas yet larger clusters (n ≈ 30)
144 9. Technical Applications of Dusty Plasmas

√
become more and more spherical. The radius of a n = 30 cluster is about 3 30 ≈ 3 Si–Si
distances wide (a Si–Si bond is about 5 Å). Hence, the cluster is of about 1.5 nm radius
(3 nm diameter).

Figure 9.4: Structures of small Sin clusters. (a) n = 2 to 10 and (b) n = 21 to 30. From
[2].

9.1.2. Agglomeration Phase

In the agglomeration phase, the particles rapidly grow from a few nanometer to several
ten nanometers. At the same time, the density of clusters dramatically drops by several
orders of magnitude. However, it can be estimated from the measurements (see Fig. 9.1
in the time interval 50 to 60 seconds) that the overall volume of the clusters does not
change very drastically. From that, it is induced that the large number of small particles
agglomerate (stick together) to form a smaller number of larger clusters (see also Fig. 9.5).

Generally speaking, the agglomeration phase is not well understood so far. The prob-
lem lies in the fact that clusters of nanometer size are difficult to measure. For smaller
clusters one can use, e.g., mass spectrometry (see above), for larger clusters of tens of
nanometers light scattering techniques are available. In addition, this growth phase is
very fast which makes time resolved measurements difficult.
Thus, one is now in the need for a model to explain that the nanometer particles stick
together to form larger clusters. The rapid time scale of agglomeration makes chemical
9.1. Particle Growth Mechanisms 145

Figure 9.5: Electron micrographs of particles grown in an argon-silane discharge a)

nanometer sized particles (end of cluster phase), b) and c) agglomerated particles of 45 nm
and 100 nm in diameter, respectively. One can see that the particles consist of smaller
agglomerated “balls” of about 10 nm in diameter. From [2].

reactions for this process unlikely. Two alternative agglomeration schemes are discussed,
the charged particle agglomeration and the neutral agglomeration model.
The neutral agglomeration model is taken from aerosol science. In that model, particles
that collide with each other on their random thermal motion have a certain probability
to stick together. From that model, the particle radius rp is found to increase in time t
as
rp = r0 (1 + Ct)2/5
and the density np drops as
np = n0 (1 + Ct)−6/5
where C is a constant taking into account sticking coefficients, mean free path and further
more. The parameters r0 and n0 are the particle radius and density at the start of the
agglomeration. Such a model fits the experimental results quite well.
The charged particle model brings charging processes into play. In the starting phase
of the agglomeration the particle density is very high (more than 109 cm−3 in Fig. 9.1).
The electron density in these discharges is of the order of 108 to 109 cm−3 , thus there
are much more clusters (dust particles) than electrons. The electrons are dramatically
depleted (see Sec. 2.6.). The average charge on the dust particles is therefore very small,
146 9. Technical Applications of Dusty Plasmas

about 0.1 elementary charges (that means that for a certain fraction of time, e.g. 1 s, the
particle has a charge of 1e and for another 9 s the particle is neutral). In that case the
neutral agglomeration model seems to be reliable.
When the particle density however drops to about 107 cm−3 , electron depletion is not
so dramatic as at the start. Then the particles can have a charge that would correspond to
the single particle case. Particles of nanometer size would attain mean charges of the order
of a few elementary charges, there are however strong fluctuations of the particle charge
for small particles (see Sec. 2.5.2.). In that case, the particles are negatively charged on
average, but at certain times they can become neutral or even positive due to the random
collection of electrons and ions. The presence of positive and negative charged particles
at the same time in the plasma strongly enhances the agglomeration rate due to their
Coulomb attraction.
In addition, the role of UV photodetachment and secondary electron emission is cur-
rently under investigation and might play a considerable role for nanometric dust particles.

9.1.3. Accretion Phase

After the rapid growth in the agglomeration phase it is found that the particles still grow
further in size (see Fig. 9.1 at t > 60 s). The growth rate in this phase is often found to
be close to the growth rate for the corresponding thin-film deposition.
At this stage, the particles have typical sizes of tens of nanometers. They are neg-
atively charged and trapped in the plasma bulk (see Sec. 3.7.). Thus, effectively, the
particles behave like small substrates on which additional layers of silicon-hydrogen films
are deposited.
The growth of these particles can be continued until the particles reach the micrometer
range. The particles are subject to the forces described in Chapter 3. When the particles
have grown to several micrometer size the particles experience stronger forces and may
eventually drop out of the discharge.

9.2. Technological Impacts of Dusty Plasmas

The technological implications of dusty plasmas can be generalized as twofold: dusty
plasmas can have “good” and “bad” impacts on technology. Surface processing technolo-
gies are obviously concerned with particle-induced failures and it was recognized only
until the end of the 1980’s that (dusty) plasma processes could be involved in such dust
contamination problems. Nowadays, dust particles are not only considered as unwanted
pollutants, since a number of useful applications have emerged from dusty plasmas. Here,
we like to illustrate some of the “good” and “bad” properties of dusty plasmas and we
will start with the older problem of contamination by dust.
9.2. Technological Impacts of Dusty Plasmas 147

Figure 9.6: Electron micrographs of a) a trench structure on a silicon wafer. Trenches are
used to manufacture capacitors which are buried vertically in the substrate to minimize the
occupied wafer space. b) Killer particle lying across several conductor paths. This particle
destroys the functionality of the integrated circuit. From Surface Technology Systems Plc.

9.2.1. Surface Contamination by Dust

Plasma processes are widely used in technological applications such as plasma etching
and sputtering or thin-film deposition. It is commonly said that 70 % of the numerous
manufacturing steps for computer chip production are plasma processes. Although the
plasma conditions are chosen in such a way that dust particle formation is unlikely they
are optimized for highest growth rates of the film to be deposited. They are thus close
to conditions where growth can occur. In these processing plasmas particles can grow
either due to plasma polymerization as described above or particles etched or sputtered
from the substrate can arrive in the plasma and can be trapped there (see Fig. 1.3 in
the Introduction). When such dust particles are deposited onto the substrate during the
discharge cycle or during switch off these particles can do severe damages to the surface
properties and can easily destroy integrated circuits. They can cause short circuits or
may clog up wafer trenches (see Fig. 9.6).
Roughly speaking, such so-called “killer particles” are dangerous when they have about
half the structure size. Current technology works with 300 mm wafers with structure sizes
of 32-45 nm. Thus, presently, particles of 20 nm and larger pose a serious problem. As
mentioned earlier, particles of that size can just be detected by light scattering, even
smaller particles will fall below the detection limit. This causes a severe diagnostic prob-
148 9. Technical Applications of Dusty Plasmas

lem in the future when even smaller structures will be fabricated.

Current technologies to prevent particles from falling onto the substrates make use
of the forces that have been discussed in Chapter 3. One possibility is to drill circular
grooves in the electrode around the wafer. The grooves then finally lead to the vacuum
port. The grooves disturb the sheath potential in such a way that the particles are
preferably trapped above the groove and not above the wafer. The grooves duct the dust
particles to the vacuum port where the dust is sucked into the vent line.
Additional techniques directly use a gas puff that is blown across the wafer prior to
plasma switch-off. The dust particles are blown into regions where they can do no harm
to the wafer and when the plasma is switched off the particles settle onto the walls or
outer regions of the electrode. For permanent removal of dust by gas flow one of the
electrodes has many small holes through which the gas enters the discharge volume, a so-
called showerhead configuration. On the other side of the chamber the gas is constantly
evacuated. Thus, the discharge works with a constant gas flow through the chamber (most
technological applications use discharges with permanent gas flows to remove sputter
products and keep the discharge under constant conditions on the molecular level). The
gas flow in the chamber is then directed in such way that possible dust particle are
constantly washed out of the discharge.
Another way is to heat different parts of the chamber. The thermophoretic force
drives the particles to colder plasma regions. Thus by heating the regions where no dust
is wanted the particles are removed.
Which technique is applied in a specific discharge strongly depends on the exact con-
ditions and a general statement is not possible here. But one should keep in mind that
gas drag or thermophoretic forces scale with the particle radius as a2 , the electric force
proportional to a. Thus, the above described techniques might not work efficiently when
smaller and smaller dust particles have to be removed.

9.2.2. Flue Gas Cleaning

Another very important application is very similar to the question discussed above: the
removal of dust particles in flue gases, e.g. of coal power plants. The soot particles from
the burning process have to be removed very efficiently (more than 99 % of the particles
have to be removed), the removal has to be done with low electrical power (otherwise the
entire power plant would not work efficiently) and finally large volumes of soot-laden gas
have to be treated.
This task can be accomplished by flue gas cleaning using corona discharges, so called
electrostatic precipitators. A corona discharge works at air pressure. There a high negative
voltage is applied to a wire (similar to a Geiger-Müller counter). The gas near the wire
is ionized and the electrons attach to the dust particles. The negatively charged dust
particles drift to the positive walls where the dust is collected. From time to time the
dust is removed by hammer strokes to the wall. This process is indeed very effective and
9.2. Technological Impacts of Dusty Plasmas 149

Figure 9.7: Electrostatic precipitators used in power plants to clean the flue gas from
dust particles. Left: Schematic view, right: installed device (Siko Engineering).

efficient and fulfills the above requirements.

Such a flue gas cleaning device is very large (see Fig. 9.7) and typically cleans 5000 m3
per hour of flue gas using only several tens to a few hundred kilowatts of electrical power
(compared to the 500 MW produced by the power plant).

9.2.3. Dust in Fusion Devices

Dust in fusion devices occurs due to the interaction of the plasma with the plasma facing
components (like graphite or CFC (carbon fibre composite) tiles) of the vessel. Dust
can be produced by thermal overload of the surfaces leading e.g. to brittle destruction
of carbon, to melt layer loss of metals or to desintegration of codeposited layers, see
e.g. [129, 130, 131]. Alternatively, dust can grow from sputtered carbon as in the low-
temperature plasmas described above. Examples of dust particles collected in existing
devices are shown in Fig. 9.8. Particles in active plasmas have also been seen from laser
scattering experiments and with fast cameras.
The problems that are associated with dust in fusion devices are the following. First,
dust can lead to difficulties with the vacuum vessel and pumping. Second, diagnostics
might be covered by dust. For example, mirrors for optical diagnostics of the plasmas
might be “blinded” by the dust. More important, material eroded from the wall could
enter the core plasma, e.g., in some tokamaks tungsten is used as a plasma facing compo-
nent. As a high-Z material (the atomic number of tungsten is Z = 74), tungsten, when
150 9. Technical Applications of Dusty Plasmas

Figure 9.8: a) REM micrographs of dust collected from fusion devices. From [130, 131].

eroded from the wall into the core plasma, leads to strong bremsstrahlung losses and
thus unwanted cooling of the plasma. Finally, in a real deuterium-tritium fusion plasma,
the radioactive tritium can be chemically bound to carbonaceous dust particles (In the
edge plasma the plasma conditions are not too different from those of low-temperature
plasmas). The dust therefore adds to the radioactive inventory of the fusion device. For
an existing tokamak, the density of carbonaceous dust in the edge plasma has been cal-
culated from simulations (see Fig. 9.8). There it has been found highest in the divertor
regions where the plasma is relatively cold.
These issues are not problematic for existing devices. The potential implications for
safety and operational performance in fusion devices such as ITER are under investigation.

9.2.4. Formation of New Materials

Plasmas in reactive, particle-forming gases are well suited to form new materials. On the
one hand, as we have seen, particles can be easily trapped inside the plasma and their
positions can be controlled by the various forces acting on them. On the other hand, the
chemically active species allow to grow particles or to modify their surface properties.
When using plasmas to modify the surface properties of particles or to grow thin films
on substrates one exploits the drastically different temperatures of the various species.
The electrons are usually the hottest species with a temperature of a few eV. They are thus
able to excite or ionize atoms and molecules, or to dissociate molecules into chemically very
active species. Thus, the various radicals or other chemical active species are generated
by the electrons. Thus, plasmas provide similar chemically active species as in hot wet-
chemical reactants.
9.2. Technological Impacts of Dusty Plasmas 151

The ions and the neutral gas are typically at room temperature or at slightly elevated
temperatures. In low-pressure plasmas the species with the highest density is by far the
neutral gas. Thus, also the substrate and the particles are kept at low temperatures which
is essential for the (thermal) stability of the substrate or the particles.
Consequently, in such reactive discharges one can have the advantage of high-
temperature chemistry without the thermal stress on the particles or the substrates. This
opens up the road to a large variety of new materials.
Coating of particles can be technologically applied to give the particles desired optical
or chemical surface properties, e.g. [132]:

• Iron particles are coated with an optical black surface. These particles then can be
used as toner particles which can be handled by magnetic fields which would result
in a new type of copy machine.

• Particles are coated with catalytic material. Due to the very large surface area of
the particles, such systems provide very efficient catalysts.

• Particles surfaces can be modified so that medical and pharmaceutical agents can
attach to the surface of the particles. Again, the very large surface area of the
particles leads to an efficient and controlled way to apply the medical drugs.

• Fluorescent particles are coated with a thin layers that keep the particles stable
against bombardment from plasma particles. Such fluorescent particles can then be
used in as the fluorescent layer in light tubes.

9.2.5. Polymorphous Silicon Films

As a final example for the technical potential of dusty plasmas, the deposition of silicon
films with special properties is discussed, namely thin films in solar cell technology.
Solar cells that do not achieve a very high efficiency are often installed in electronic
devices like pocket calculators or watches. They can be easily deposited with plasma pro-
cesses and are quite cheap. These films are deposited in silane discharges on appropriate
substrates under conditions where dust particle formation does not occur. This thin sili-
con film that is responsible for the opto-electric properties of the solar cell is amorphous
and a lot of hydrogen is incorporated in the film. These films are denoted as a-Si:H, where
the “a” stands for amorphous. The efficiency of such films is of the order of 4 to 6 %,
only.
In contrast, solar cells for power production (on the roof of houses or in large solar
cell fields) require high efficiency. This is only achieved by crystalline silicon. There,
efficiencies above 15 % are standard and up to 22 % have been achieved. However, solar
cells from crystalline silicon are very expensive. They are formed from molten silicon at
high temperatures and subsequently slow cooling.
152 9. Technical Applications of Dusty Plasmas

Figure 9.9: Electron micrograph of a silicon film at 50◦ C in its as-deposited state (left)
and after annealing at 425◦ C for one hour. From [133].

Figure 9.10: Evolution of the efficiency of solar-cells made from a-Si:H and pm-Si films.
From [133].

Here, an interesting new development has emerged in the last recent years [133]. Solar
cells have been deposited in silane discharges just like those that are used to deposit a-
Si:H films. However, the discharges are operated under plasma conditions (gas pressure,
9.2. Technological Impacts of Dusty Plasmas 153

gas flow, discharge power, substrate temperature etc.) that are very close to powder
formation. Previously, such conditions have been avoided since the formation of dust at
thin film deposition was considered as very unfavorable.
Operating the discharge close to powder formation means that no particles larger than
10 nanometers are formed. Thus, in-situ diagnostics will characterize the discharge as
dust-free since particles below 10 nanometers cannot be detected. However, it is found that
close to the conditions of detectable dust formation already particles of a few nanometers
in size are formed. These particles are then incorporated into the film before the particles
start to grow in the agglomeration and accretion phase.
Such nanometer clusters are indeed incorporated in the film as can be seen in the
micrograph in Fig. 9.9. The dust particles of nanometer size are crystalline and in the
film they appear as tiny crystalline patches in the otherwise amorphous silicon matrix.
The circles and arrows hint at those crystalline regions. After annealing for one hour the
film has much larger crystalline regions. The dust particles serve as a crystalline nucleus
where the crystallization can start from. The resulting film is neither crystalline nor
amorphous and consists of some larger crystalline patches in a still amorphous matrix.
The film is then termed as polymorphous silicon (pm-Si). So, these new type of silicon
film is due to the incorporation of nanometer-sized dust.
So, what are the properties of the pm-Si films? Since the film is somewhat crystalline
and somewhat amorphous, the efficiency of solar cells made from these films (9 to 10 %) is
considerably higher than that of the a-Si:H films (see Fig. 9.10). What is even more, the
efficiency stays constant over time, whereas the a-Si:H films tend to degrade, i.e. their
efficiency drops with time. So, these dust-incorporated films have decisively improved
properties, but essentially use the same technology as the a-Si:H film deposition. Only
the parameters of operation are somewhat different and close to the particle formation
threshold. It should be noted, here, that the incorporation of larger particles does not
lead to improved films.
So with small changes to the manufacturing process a large improvement in the effi-
ciency of solar cells is achieved due to the application of dusty plasmas.
154 10. Astrophysical Dusty Plasmas

10. Astrophysical Dusty Plasmas

The field of dusty plasma physics has started from the investigations of dusty plasmas in
astrophysical situations, from the study of comets, interstellar clouds or the rings of the
Great Planets.
In space, dust particles can collect electric charges by various processes, as discussed
in Chapter 2, like electron and ion collection, secondary emission, photoelectric charging,
sputtering etc. The particles react to the magnetic and electric fields in the plasma envi-
ronment of planets, the wind of charged particles coming from the Sun and gravitational
forces. The combination of these processes give rise to a description of dust processes in
astrophysical situations that is sometimes referred to as “gravito-electrodynamics”.
Dusty plasmas represent the most general type of space plasmas. In fact, it is difficult
to find astrophysical plasmas that do not include dust particles (maybe except the fusion
plasmas in the interior of stars). There is a vast amount of possibilities where astrophys-
ical and extraterrestrial dusty plasmas naturally occur. When starting near the Earth’s
surface we find the so-called Noctilucent Clouds and Polar Mesospheric Summer Echoes
(PMSE). In our solar system we find dust in the rings of the Great Planets like Jupiter and
Saturn, in cometary tails and in the zodiacal light. Interstellar dust-containing hydrogen
or molecular clouds also add to this list.
From the vast amount of different astrophysical situations we will restrict here to the
formation of spokes in the rings of Saturn and dust particle “orbits” at Jupiter and Saturn.
These are situations where experimental data are already available (from the spacecraft
Galileo at Jupiter) or where they will become available in the near future (Cassini at
Saturn, see http://www.esa.int/SPECIALS/Cassini-Huygens/). In addition, a few words
on the noctilucent clouds will be given here first [134].

10.1. Noctilucent Clouds and PMSE

Noctilucent clouds are “nightly shining” clouds that are observable in the summer months
(May to September) at high-latitude regions (50 − 80◦ ), like Northern Europe. They can
be seen long after sunset. They are thus located at high altitudes of 80 − 110 km above
the surface of the Earth where the sun, although being already below the horizon, still
illuminates the clouds (see Fig. 10.1).
The clouds are thus at the lower edge of the ionosphere. They are assumed to consist
of water or methane ice particles which are charged by the plasma of the ionosphere.
Surprisingly, in the summer months the atmosphere at these altitudes is extremely cold
(about 110 K) and is much colder than during winter times (this is the so-called summer
anomaly). This low temperature allows the water and the methane to freeze under the
10.1. Noctilucent Clouds and PMSE 155

Figure 10.1: Photo of a noctilucent cloud. From Wikipedia.

conditions of the upper atmosphere preferably on small condensation nuclei. This explains
why the clouds are preferably visible in the summer months.
Noctilucent clouds have first been reported in 1885, a few years after the big explosion
of the Krakatoa volcano in 1883. It is not clear whether noctilucent clouds have been
present before this explosion, but have not been reported, or whether the Krakatoa has
triggered the existence of the noctilucent clouds by blowing large amounts of dust into
the atmosphere. Nevertheless, the occurrence and strength of noctilucent clouds have
increased since their discovery. It is sometimes argued that the increased methane release
on Earth leads to the increased occurrence of noctilucent clouds.
The “dust” particles in the noctilucent clouds can be observed by sounding rockets
launched into the ionosphere, by lasers fired into the atmosphere (so-called LIDARs, i.e.
LIght Detection And Ranging) and by backscatter from radar signal radiated into the
atmosphere (see Fig. 10.2). With LIDAR technique the light scattered from particles
in the atmosphere is detected at different places on the Earth surface and the height of
he scattering particles is found from triangulation. LIDARs are sensitive to dust parti-
cles larger than approximately 30 nm due to the use of light scattering (see comments
in Chapter 9). Radars do not “see” the dust particles directly, but the radar signal is
scattered from the electron and ion clouds around the dust particles. Radar backscatter
156 10. Astrophysical Dusty Plasmas

Figure 10.2: Simultaneous occurrence of PMSE and noctilucent clouds over Northern
Norway. Color-coded is the radar backscatter signal as a function of time and height. The
contour lines denote the noctilucent cloud position determined from the LIDAR. From
[135].

usually is an indication of smaller dust particles of 1-20 nm size. Often during summer
months one observes radar backscatter in the mesosphere (up to 100 km altitude), the
above mentioned PMSEs. It is reasonable to assume that the PSMEs are related to the
presence of noctilucent clouds. Indeed, there are numerous observations that substantiate
this reasoning, see Fig. 10.2. However, often one sees noctilucent clouds and no radar
backscatter, or vice versa. This might due to the fact that the presence of PMSEs and
noctilucent clouds rely on different particle sizes, larger than ≈ 30 nm for the noctilu-
cent clouds and smaller than ≈ 30 nm for radar backscatter. The physics behind the
noctilucent clouds and PMSEs is, however, not fully understood so far.

10.2. Dust Streams From Jupiter

We now turn to dusty plasmas in our solar system and will discuss dust properties near
Jupiter and Saturn [136].
Jupiter is the largest planet of our solar system (its mean radius is RJ = 69 911 km)
and the one with the highest mass (mJ = 1.898·1027 kg). One would therefore expect that
due to gravitational attraction all dust in its vicinity would be “sucked” towards Jupiter.
Nevertheless, the Ulysses spacecraft has discovered a high-speed dust flux away from the
Jovian system in 1992. This dust stream has been investigated in much more detail by
the Galileo spacecraft since 1996. Galileo had a dust detector on board that was able to
10.2. Dust Streams From Jupiter 157

Figure 10.3: Left: Orbits of the Galileo spacecraft from 1996 to 2003. The black circles in
the center represent the orbits of the 4 Jovian moons Io, Europa, Ganymed and Callisto.
Right: Measured dust impact rate (color coded) on Galileos path near the inner moons.
High impact rates are found near the moon Io. From Max-Planck-Institute Heidelberg,
E. Grün, http://www.mpi-hd.mpg.de/dustgroup/galileo/folien/jupstream.html.

measure the velocity and the mass of the arriving dust particles and was calibrated for
dust particles in the size range 30 nm < a < 100 nm. The dust detector measured a
certain fraction of quite large particles (> 1 µm). The majority of the particles were very
small (smaller than the calibration range, i.e. a < 30 nm) and very fast. These particles
were generally directed away from Jupiter.
On its path, Galileo measured the dust particles arriving at the dust detector (see
Fig. 10.3). Very high impact rates were found close to the moon Io. Io is an active volcanic
moon and is closest to Jupiter. So, Io was already very early considered to be the source
of the dust particles emerging from the Jovian system. Dust from the volcanic eruptions
is sent into the magnetosphere of Jupiter and is transported outward. To describe this
outward motion the dust particle motion has to be modeled quite accurately.

10.2.1. Model
The equation of motion of a dust particle is given by
Q  ~ + ~Ec − GmJ 1 ~r
~¨r = d ~r˙ × B

(10.1)
md r3
158 10. Astrophysical Dusty Plasmas

where G is the gravitational constant. Hence, electric field force, the Lorentz force and
gravitational forces have to be considered. The magnetosphere would rigidly co-rotate
with the planet at the rotation frequency ΩJ if the magnetosphere was perfectly conduct-
ing. B~ is the magnetic field of Jupiter and ~Ec = (~r × Ω
~ J) × B
~ is the co-rotational electric
field. Other forces can be neglected, here.
As the grain moves through the plasma the particle charge is determined by the
different currents to the particle (see Eq. (2.1))
dQd X
= Ii
dt i

where typically electron and ion collection, secondary emission and photoelectron emission
are considered.
The magnetic field B ~ (and thus the co-rotational electric field ~Ec ) around Jupiter is
quite accurately known and is used to calculate the Lorentz force on the dust particles. It
is important to note, here, that the magnetic field axis and the rotational axis of Jupiter
have a relative angle of about 10◦ . Thus during Jupiter’s rotation the magnetic field
precesses, i.e. has a “tumbling” motion.
The equation of motion for the dust particles is then solved numerically. Dust particles
of various sizes are considered to originate from Io’s orbit and their motion through the
Jovian magnetosphere is followed in the simulation.

10.2.2. Model and Experiment

To illustrate their behavior the orbits of particles with fixed charge are considered, see
Fig. 10.4. It is seen that negatively charged dust (of fixed charge Zd ) moves on closed
orbits. Particles of 80 nm radius are found to have the largest excursions which are still
very small compared to the size of the Jovian system. Positively charged dust is found
on closed orbits for particles smaller than 20 nm or larger than 120 nm. Positive dust
particles in the intermediate size range have no closed orbits and may leave the system.
These are the particles that are responsible for the high-speed stream away from Jupiter.

In more sophisticated calculations taking the charging equation into account the tra-
jectories of 10 nm particles emerging from the orbit of the moon Io have been calculated
(see Fig. 10.5). One can see that indeed particles are emitted outwards, away from Jupiter.
The particles leave Io and form a comet-tail like swarm that spirals outward. One can
also see that the particles leave the equatorial plane and acquire positions quite far above
and below the equatorial plane (upper right panel). This is due to the inclination of the
magnetic field axis relative to Jupiter’s rotation axis. So, the dust particles form a pat-
tern around Jupiter like a ballerina skirt. The particles on their way usually have positive
charges due to photoelectron emission.
10.3. Dust Orbits at Saturn 159

Figure 10.4: Orbits of dust particles with fixed potential. (a,b) Negatively charged dust
with a surface potential of −10 V (corresponding to a charge of Zd = 7 · a with parti-
cle radius a given in nanometers). The considered dust particles range between 20 and
300 nm. (c,d) Positively charged dust at +10 V for particles in the size range between
10 and 180 nm. The trajectories are a projection onto the plane of the rotation axis (the
rotation is around x = 0) and y is the distance above or below the equatorial plane of
Jupiter. From [136].

Finally, to check the reliability of the simulation results the flux of outward-moving
dust particles is calculated and compared to the experimental data collected by the dust
detector on Galileo. There the number of dust impacts as a function of orientation relative
to Jupiter is shown. This comparison is shown in Fig. 10.6. One can see a remarkable
agreement between these two curves where the maxima at 5 and 17 h (local time) and
the minimum at 10-14 h is modeled very accurately. This gives a lot of confidence in the
accuracy of the calculations.
To summarize, the observed dust stream of particles away from Jupiter is believed to
be due to positively charged grains of about 10 nm radius that emerge from the volcanic
moon Io.

10.3. Dust Orbits at Saturn

The situation of dust particle dynamics at Saturn is very similar to that at Jupiter.
Therefore, similar dust streams are expected also for Saturn. There, the source of the
160 10. Astrophysical Dusty Plasmas

Figure 10.5: Orbits of dust particles of 10 nm radius. Upper panel: “top view” and
“side view” of particle motion at Jupiter. The particles are considered to be emitted from
Io’s orbit (the dark ring in the upper left panel). Lower panel: same for the situation at
Saturn. The particles are emitted due to disturbances of the magnetosphere by the moons
Dione, Helene and Rhea. From [136].

dust particles can be the moons Dione, Helene and Rhea which orbit in the outskirts
of the magnetosphere of Saturn (their distances from Saturn are 6.3RS and 8.7RS with
RS = 58 232 km being the radius of Saturn, mS = 568 · 1024 kg). These moons are not
volcanic, but dust may enter the Saturnian magnetosphere by meteroid bombardment.
The simulated dust transport for Saturn is also shown in Fig. 10.5. The particle motion
is purely radially outward. The particles stay in the equatorial plane. This is due to the
fact that the magnetic field axis of Saturn is closely aligned with the rotation axis.
The aligned magnetic field also gives rise to a new phenomenon that is not observable
at Jupiter. Since the aligned situation is very symmetric there exist stable particle orbits
that are high above the equatorial plane and that never cross this plane, so-called “halo”
orbits, see Fig. 10.7. Particles in these orbits are kept at such positions due to the balance
of three forces: the centrally inward gravitational force, the radially outward centrifugal
force and the electromagnetic force that points upward under a certain angle (see inset
of Fig. 10.7). Such particles might be observed by the spacecraft Cassini that has arrived
at Saturn in Summer 2004. At the end of its nominal 4-year mission Cassini will explore
high inclination orbits and might then prove the existence of these particles.
10.4. Spokes in Saturn’s Rings 161

Figure 10.6: Flux of Jovian stream particles as a function of local time from simulation
(upper panel) and experimental data collected from 1995 to 2002 (lower panel). From
[136].

Figure 10.7: “Halo” orbits of dust particles above the equatorial plane at Saturn. The
inset shows the action of the three responsible forces, gravity, centrifugal force and elec-
tromagnetic forces (Cover of Geophysical Research Letters, May 2001).

10.4. Spokes in Saturn’s Rings

Next, the formation of spokes will be discussed. We have mentioned this example as a
paradigm for dusty plasma physics in astrophysical situations in the Introduction. And,
indeed, the problem of spoke formation can only be treated when the various processes
in dusty plasmas are taken into account (like charging, forces etc.).
162 10. Astrophysical Dusty Plasmas

It is interesting to note that at least two different theories exist to explain the formation
of spokes and a definite answer cannot be given at the moment. It is expected that also
here the spacecraft Cassini will allow a deeper insight into the spoke formation.
Spokes are known since the early 1980’s from photos of the spacecrafts Voyager 1 and
2 and they have been rediscovered by Cassini in 2005 (Interestingly, the spokes were not
visible when Cassini arrived at Saturn in early 2004). The spokes appear in the B-ring
of Saturn which extends from 91 975 km to 117 507 km and is the most opaque (“solid”)
ring of Saturn and one of the most prominent rings. Spokes are radial features in the
B-ring, see Fig. 1.1a). The spokes appear dark when viewed in backscattered light and
bright in forward scattered light. This suggests that the spokes consist of sub-micron
particles which have such scattering properties. Spokes come into existence in less than
5 minutes and disappear after about 5 hours. The radial elongation of the spokes is a
few 104 km, their width is between 200 and 1000 km. The spokes are therefore a very
dynamical phenomena and are unlikely to be explained by only gravitational effects.
Here, we like to illustrate the problem of spoke formation using the model of Goertz
and Morfill [137]. In this model it is assumed that spokes become visible when dust
particles are lifted above ring plane. In the ring the small dust particles are not visible
due to the larger rocks and boulders (of 10 cm to 10 m size). Above the ring plane the
particles become visible and show the scattering properties as described above. To be
lifted above the ring plane the dust particles must be accelerated. In this model the
acceleration is due to a perpendicular (vertical) electric field. Now the question arises
how such an electric field appears and why its appearance is only sporadic (spokes do not
appear always and everywhere).
Under normal conditions the plasma density in the ring plane is small (ne ≤ 104 cm−3 )
and the surface potential of the dust is slightly positive on the sun-facing side, the mean
charge of sub-micron dust is expected to be much smaller than one elementary charge.
The dust is essentially neutral. In the shadow, it is expected to be around −6 V since
there is no photoemission. The ring plane is an equipotential line. Due to the small dust
and the small electric fields at the ring surface dust particles cannot be lifted above the
ring plane.
It is then assumed that a local increase of plasma density can change this equilibrium
situation drastically. Such a local plasma density increase can be due to meteoretic
impacts into the ring plane. In the plasma with higher density the surface potential of
the dust becomes negative (around -6 V). Furthermore, the equipotential lines become
compressed below the local plasma (see Fig 10.8). This can accelerate dust particles above
the ring plane where they become visible.
Having now elevated dust particles there is the need to explain the radial motion
of the spokes. There, one has to look into the rotation of the different species. The
higher-density plasma cloud is coupled to the magnetosphere of Saturn. Since the entire
magnetosphere of Saturn rotates at the same frequency ΩS as Saturn also this plasma
cloud will rotate at this speed. The dust particles however will rotate around Saturn
10.4. Spokes in Saturn’s Rings 163

Figure 10.8: Model for the levitation of dust above the ring plane and the formation of
spokes. Under a locally increased plasma density (dark grey) the equipotential lines are
compressed leading to an increased electric field. The particles are lifted above the ring
plane. There, the particles move on Kepler orbits and accumulate on one side of the local
plasma disturbance and are then transported radially (see text). The view in this Figure
is radially away from Saturn. After [137].

on Kepler orbits with the Keplerian frequency ΩK > ΩS .∗ Thus, the plasma cloud and
the particles will rotate at different speeds, so the dust is accumulated on one side of the
plasma cloud (see Fig. 10.8). The particles cannot leave the plasma cloud due to confining
electric fields at the plasma cloud boundary.
Since the plasma cloud is quasineutral and the negative particles accumulate on one
side an electric field arises. This electric field together with the magnetic field of Saturn
(which is nearly vertical near the ring plane) leads to an E × B drift in the radial direction.
Thus, the plasma cloud with the particles will stretch radially and radially elongated dust
structures are formed, which is seen as the spokes.
The particles lifted above the ring plane should have sizes in the range of 100 to
300 nm. Smaller particles need much more time to charge up (the charging time is
indirect proportional to particle size, see Eq. 2.20). Larger particles have too large mass
and are not accelerated fast enough. This size range is just the one expected from the
scattering of the sunlight. Thus the model makes a number of predictions which are in
agreement with the findings from the observations.
In a different model, Bliokh and Yaroshenko [138] explain the spoke formation by
∗
The particles in the B-ring rotate faster than Saturn. Saturn performs one revolution in 10.2 h, the
Kepler period at a distance of 100 000 km from Saturn is about 9 h.
164 10. Astrophysical Dusty Plasmas

Figure 10.9: Simulated Mach cone in Saturn’s rings. From [139].

density waves in a multi-stream situation. This approach also allows to account for the
very many narrow rings and gaps within the B-ring. Although such dust density waves
are probably not strong enough to fully explain the spoke formation they represent an
intriguing mechanism. This demonstrates that the question of spoke formation is not
settled yet.

10.5. Mach Cones in Saturn’s Rings

Finally, we like to address an interesting diagnostic method that might be able to measure
plasma properties in the rings of Saturn. This idea was first proposed by O. Havnes in
1995 [105].
There, the effect of Mach cone formation in Saturn’s rings is considered. As mentioned
above, large boulders of centimeter to meter size move at Keplerian velocities around
Saturn. At a radial distance of R = 100 000 km the Keplerian velocity is around 19 km/s.
Small dust particles might be coupled to the magnetosphere of Saturn that co-rotates
with the planet. At that distance this velocity is around 17 km/s. Thus there is a large
relative velocity of larger boulders and small dust particles. This velocity difference might
even be larger than the sound speed of dust waves in the ring. Thus, Mach cones might
emerge in Saturn’s rings. The kind of wave that might be excited there are dust-acoustic
waves. Since the dust is expected to be only weakly coupled dust-acoustic waves might
be launched. These are the weakly coupled dust waves with the smallest sound speed and
thus will be preferably excited by boulders that move with supersonic speed relative to
the smaller dust particles. Simulated Mach cones of dust-acoustic waves in the rings of
10.5. Mach Cones in Saturn’s Rings 165

Saturn are shown in Fig. 10.9.

These Mach cones will be very similar to those observed in strongly coupled dust
systems as described in Section 7.6. The main difference is that in strongly coupled
systems we find Mach cones due to dust lattice wave excitation whereas in Saturn’s rings
we expect Mach cones of dust-acoustic waves. As for the DLW Mach cone, the wave
pattern of the DAW Mach cone shows a complicated structure with fine details due to
the dispersion of the DAW and other effects like spatially and temporally varying dust
densities or dust charges. From the Mach cones and its detailed structure the properties
of the dusty plasma inside the rings cane be derived. Again, it is anticipated that Mach
cones might be observed by the Cassini spacecraft.
166 11. Summary

11. Summary
In these Lecture notes an overview over the various effects in dusty plasmas has been given.
Fundamental properties of dusty plasma like particle charging, interaction potentials and
forces on the dust have been included. Waves in weakly and strongly coupled dusty
plasmas have been discussed and the extension towards normal modes in finite systems
has been presented. Finally, a rough description of applications of dusty plasmas in
technology has been given, before some examples from the origin of dusty plasma physics,
namely astrophysical situations, have been mentioned.
To summarize, the main properties of dusty plasmas compared to “usual” plasmas are
compiled below again:

• Dusty plasmas are at least three component plasmas (electrons, ions and dust). In
this sense, dusty plasmas are somewhat comparable to negative ion plasmas.

• However, the typical charge on the charge carriers (dust) are of the order of 10 000
elementary charges which leads to strong coupling on the one hand and to strong
reactions to electric fields on the other hand.

• The dust charge is variable and depends on the local plasma parameters. The
charging time of particles is finite. Thus, the charge becomes a dynamic variable
and can lead to novel dynamic phenomena.

• The dust mass is by orders of magnitudes larger than that of electrons and ions.
Thus the dominant time scale is that of the dust plasma frequency ωpd which is by
orders of magnitude smaller than that of electrons and ions leading to convenient
time scales for the observation of dynamic processes in laboratory discharges. Also
the separation of time scales leads to new types of waves and dynamical phenomena.

• The slow time scales allow that electrons and ions contribute to shielding which
should result in different shielding scales.

• The dust size is not negligibly small leading to surface phenomena and forces which
are unimportant in “usual” plasmas.

Due to all of these unique properties of dusty plasmas a number of new phenomena
occur in dusty plasmas like new force, new types of waves, crystallization processes, phase
transitions, observation of processes on the kinetic level and many more. We hope that we
have clarified the origin of these phenomena and that we have demonstrated why dusty
plasmas have become one of the very interesting fields in plasma physics.
167
168 11. Summary

List of symbols
Symbol Definition

md , me , mi , mn Mass of dust particles, electrons, ions and neutrals

Td , Te , Ti , Tn Temperature of dust particles, electrons, ions and neutrals

nd , ne , ni , nn Density of dust particles, electrons, ions and neutrals

λD,e , λD,i , λD Electron, ion Debye length and linearized Debye length λ−2 −2 −2
D = λD,e + λD,i

q
8kTα
vth,e , vth,i , vth,n Thermal velocity of electrons, ions and neutrals: vth,α = πmα

ui Ion drift velocity

a Dust particle radius

b Nearest neighbor distance in plasma crystals

φp , φfl Dust particle potential φp and floating potential φfl

Qd = (−)Zd e Dust particle charge Qd and charge number Zd

τe , τi Dust charging time scale due to inflow of electrons and ions

β Epstein friction coefficient

ω0 Frequency of trapping potnential

bWS Wigner-Seitz radius bWS = (3/4πn)1/3 (3D) or bWS = (πn)−1/2

r0 Size parameter of finite systems r0 = (2Z2d e2 /(4π0 md ω20 ))1/3

κ Screening strength κ = bWS /λD (for extended systems)

κ = r0 /λD (for finite systems)
r
Z2d e2 nd
ωpd Dust plasma frequency ωpd = 0 md
(for weakly coupled systems)
r
Z2d e2
ωpd = 0 md b3
(for strongly coupled systems)
BIBLIOGRAPHY 169

Bibliography
[1] F. Verheest, Waves in Dusty Space Plasmas (Kluver Academic Publishers, Dor-
drecht, 2000). 1, 87

[2] Dusty Plasmas, edited by A. Bouchoule (John Wiley & Sons, Chichester, 1999). 1,
141, 142, 143, 144, 145

[3] P. K. Shukla and A. A. Mamun, Introduction to Dusty Plasma Physics (Institute of

Physics Publishing, Bristol, 2002). 1, 87

[4] P. Bliokh, V. Sinitsin, and V. Yaroshenko, Dusty and self-gravitational plasma in

space (Kluwer Academic Publ., Dordrecht, 1995). 1

[5] G. S. Selwyn, J. Singh, and R. S. Bennett, J. Vac. Sci. Technol. A 7, 2758 (1989). 3

[6] H. M. Mott-Smith and I. Langmuir, Phys. Rev. 28, 727 (1926). 7

[7] V. R. Ikkurthi, K. Matyash, A. Melzer, and R. Schneider, Phys. Plasmas 15, 123704
(2008). 11

[8] J. Allen, B. M. Annaratone, and U. deAngelis, J. Plasma Phys. 63, 299 (2000). 11

[9] M. Lampe, V. Gavrishchaka, G. Ganguli, and G. Joyce, Phys. Rev. Lett. 86, 5278
(2001). 11, 13

[10] C. T. N. Willis, M. Coppins, M. Bacharis, and J. E. Allen, Plasma Sources Sci.

Technol. 19, 065022 (2010). 11

[11] J. Allen, R. Boyd, and P. Reynolds, Proc. Phys. Soc. 70, 297 (1957). 11

[12] R. V. Kennedy and J. E. Allen, J. Plasma Phys. 67, 243 (2002). 11

[13] E. C. Whipple, Rep. Prog. Phys. 44, 1198 (1981). 12

[14] J. Goree, Phys. Rev. Lett. 69, 277 (1992). 12

[15] S. A. Khrapak et al., Phys. Rev. E 72, 016406 (2005). 13

[16] C. K. Goertz, Rev. Geophys. 27, 271 (1989). 13, 14

[17] A. Zobnin, A. D. Usachev, O. F. Petrov, and V. E. Fortov, Phys. Plasmas 15,

043705 (2008). 18, 19

[18] S. Khrapak and G. Morfill, Contrib. Plasma Phys. 49, 148 (2009). 18
170 BIBLIOGRAPHY

[19] N. Meyer-Vernet, Astronom. Astrophys. 105, 98 (1982). 20

[20] T. Nitter, T. K. Aslaksen, F. Melandsø, and O. Havnes, IEEE Trans. Plasma Sci.
22, 159 (1994). 22

[21] C. Cui and J. Goree, IEEE Trans. Plasma Sci. 22, 151 (1994). 23

[22] T. Matsoukas and M. Russell, J. Appl. Phys. 77, 4285 (1995). 23

[23] I. Goertz, F. Greiner, and A. Piel, Phys. Plasmas 18, 013703 (2011). 25

[24] O. Havnes et al., J. Geophys. Res. 92 A3, 2281 (1987). 24

[25] S. Hamaguchi and R. T. Farouki, Phys. Rev. E 49, 4430 (1994). 28, 29

[26] M. S. Barnes et al., Phys. Rev. Lett. 68, 313 (1992). 32

[27] M. D. Kilgore, J. E. Daugherty, R. K. Porteous, and D. B. Graves, J. Appl. Phys.

73, 7195 (1993). 32

[28] S. Khrapak, A. V. Ivlev, G. Morfill, and H. Thomas, Phys. Rev. E 66, 046414
(2002). 33

[29] S. A. Khrapak, A. V. Ivlev, S. K. Zhdanov, and G. E. Morfill, Phys. Plasmas 12,

042308 (2005). 33, 34, 118

[30] I. H. Hutchinson, Plasma Phys. Control. Fusion 48, 185 (2006). 33, 34, 118

[31] L. Patacchini and I. H. Hutchinson, Phys. Rev. Lett. 101, 025001 (2008). 34

[32] V. R. Ikkurthi, K. Matyash, A. Melzer, and R. Schneider, Phys. Plasmas 16, 043703
(2009). 34, 68

[33] P. S. Epstein, Phys. Rev. 23, 710 (1924). 35, 43

[34] B. Liu, J. Goree, V. Nosenko, and L. Boufendi, Phys. Plasmas 10, 9 (2003). 35

[35] H. Rothermel et al., Phys. Rev. Lett. 89, 175001 (2002). 36

[36] A. Ashkin, Phys. Rev. Lett. 24, 156 (1970). 36

[37] A. M. Ignatov and S. G. Amiranashvili, Phys. Rev. E 63, 017402 (2001). 37

[38] A. Melzer, Plasma Sources Sci. Technol. 10, 303 (2001). 37

[39] J.-L. Dorier, C. Hollenstein, and A. Howling, J. Vac. Sci. Technol. A 13, 918 (1995).
39
BIBLIOGRAPHY 171

[40] G. E. Morfill et al., Phys. Rev. Lett. 83, 1598 (1999). 39

[41] J. Goree, G. Morfill, V. Tsytovich, and S. V. Vladimirov, Phys. Rev. E 59, 7055
(1999). 39

[42] P. Belenguer et al., Phys. Rev. A 46, 7923 (1992). 41, 42

[43] E. Tomme, D. Law, B. M. Annaratone, and J. Allen, Phys. Rev. Lett. 85, 2518
(2000). 41

[44] A. Melzer, T. Trottenberg, and A. Piel, Phys. Lett. A 191, 301 (1994). 42

[45] T. Trottenberg, A. Melzer, and A. Piel, Plasma Sources Sci. Technol. 4, 450 (1995).
42, 65

[46] A. Melzer, A. Homann, and A. Piel, Phys. Rev. E 53, 2757 (1996). 42, 44, 81, 83

[47] A. Homann, A. Melzer, and A. Piel, Phys. Rev. E 59, 3835 (1999). 44

[48] J. Carstensen, H. Jung, F. Greiner, and A. Piel, Phys. Plasmas 18, 033701 (2011).
44

[49] A. Sickafoose, J. Colwell, M. Horanyi, and S. Robertson, 367 . 45

[50] H. Schollmeyer, A. Melzer, A. Homann, and A. Piel, Phys. Plasmas 6, 2693 (1999).
45

[51] C. Zafiu, A. Melzer, and A. Piel, Phys. Rev. E 63, 066403 (2001). 47, 48

[52] A. Ivlev, U. Konopka, and G. Morfill, Phys. Rev. E 62, 2739 (2000). 48, 50

[53] J. Beckers et al., Phys. Rev. Lett. 106, 115002 (2011). 49

[54] S. Nunomura, T. Misawa, N. Ohno, and S. Takamura, Phys. Rev. Lett. 83, 1970
(1999). 49

[55] E. Wigner, Trans. Faraday Soc. 34, 678 (1938). 51

[56] S. Ichimaru, Rev. Mod. Phys. 54, 1017 (1982). 51, 52, 58

[57] S. Hamaguchi, R. Farouki, and D. H. E. Dubin, Phys. Rev. E 56, 4671 (1997). 53

[58] M. O. Robbins, K. Kremer, and G. S. Grest, J. Chem. Phys. 88, 3286 (1988). 53

[59] H. Ikezi, Phys. Fluids 29, 1764 (1986). 54, 55

[60] D. Dubin, Phys. Rev. Lett. 71, 2753 (1993). 56

172 BIBLIOGRAPHY

[61] H. Totsuji et al., Phys. Lett. A 221, 215 (1996). 57

[62] H. Totsuji, T. Kishimoto, and C. Totsuji, Phys. Rev. Lett. 78, 3113 (1997). 57

[63] B. Klumov et al., Plasma Phys. Control. Fusion 51, 124028 (2009). 60, 62

[64] U. Konopka, G. Morfill, and L. Ratke, Phys. Rev. Lett. 84, 891 (2000). 63, 64

[65] J. H. Chu and L. I, Phys. Rev. Lett. 72, 4009 (1994). 65

[66] U. Mohideen et al., Phys. Rev. Lett. 81, 349 (1998). 65

[67] B. A. Annaratone et al., Phys. Rev. E 63, 036406 (2001). 65

[68] S. V. Vladimirov and M. Nambu, Phys. Rev. E 52, R2172 (1995). 65, 66, 67

[69] M. Lampe, G. Joyce, and G. Ganguli, Phys. Plasmas 7, 3851 (2000). 66, 68, 70

[70] M. Nambu, S. V. Vladimirov, and P. K. Shukla, Phys. Lett. A 203, 40 (1995). 67

[71] I. H. Hutchinson, Phys. Rev. Lett. 107, 095001 (2011). 68

[72] A. Melzer et al., Phys. Rev. E 54, R46 (1996). 69, 71, 81, 82

[73] V. A. Schweigert et al., Phys. Rev. E 54, 4155 (1996). 68, 69, 70, 74, 76

[74] V. A. Schweigert et al., JETP 88, 482 (1999). 68

[75] W. J. Miloch, M. Kroll, and D. Block, Phys. Plasmas 17, 103703 (2010). 68

[76] K. Takahashi et al., Phys. Rev. E 58, 7805 (1998). 72

[77] A. Melzer, V. Schweigert, and A. Piel, Phys. Rev. Lett. 83, 3194 (1999). 72, 73

[78] A. Melzer, V. Schweigert, and A. Piel, Physica Scripta 61, 494 (2000). 72, 74

[79] L. Couëdel et al., Phys. Rev. Lett. 104, 195001 (2010). 74

[80] H. Thomas and G. E. Morfill, Nature 379, 806 (1996). 81, 84

[81] V. A. Schweigert et al., Phys. Rev. Lett. 80, 5345 (1998). 84

[82] I. V. Schweigert, V. A. Schweigert, A. Melzer, and A. Piel, Phys. Rev. E 62, 1238
(2000). 85, 86

[83] N. N. Rao, P. K. Shukla, and M. Y. Yu, Planet. Space Sci. 38, 543 (1990). 87

[84] A. Barkan, R. L. Merlino, and N. D’Angelo, Phys. Plasmas 2, 3563 (1995). 91, 92
BIBLIOGRAPHY 173

[85] C. Thompson, A. Barkan, N. D’Angelo, and R. L. Merlino, Phys. Plasmas 4, 2331

(1997). 91, 92

[86] J. B. Pieper and J. Goree, Phys. Rev. Lett. 77, 3137 (1996). 92, 93

[87] H. Prabhakara and V. Tanna, Phys. Plasmas 3, 3176 (1996). 92

[88] J. D. Williams, J. Edward Thomas, and L. Marcus, Phys. Plasmas 15, 043704
(2008). 92, 94

[89] S. Ratynskaia et al., Phys. Rev. Lett. 93, 085001 (2004). 92, 94

[90] A. Piel et al., Phys. Rev. Lett. 97, 205009 (2006). 92, 94

[91] T. M. Flanagan and J. Goree, Phys. Plasmas 17, 123702 (2010). 92

[92] R. L. Merlino, A. Barkan, C. Thompson, and N. D’Angelo, Phys. Plasmas 5, 1607

(1998). 96, 97, 98

[93] F. Melandsø, Phys. Plasmas 3, 3890 (1996). 102

[94] F. M. Peeters and X. Wu, Phys. Rev. A 35, 3109 (1987). 104, 105

[95] X. Wang, A. Bhattacharjee, and S. Hu, Phys. Rev. Lett. 86, 2569 (2001). 104, 106

[96] A. Homann et al., Phys. Rev. E 56, 7138 (1997). 107, 108, 109

[97] A. Homann, A. Melzer, R. Madani, and A. Piel, Phys. Lett. A 242, 173 (1998).
107, 110

[98] S. Nunomura, D. Samsonov, and J. Goree, Phys. Rev. Lett. 84, 5141 (2000). 111

[99] S. Nunomura et al., Phys. Rev. E 65, 066402 (2002). 111

[100] D. Samsonov et al., Phys. Rev. Lett. 83, 3649 (1999). 112

[101] D. Samsonov, J. Goree, H. Thomas, and G. Morfill, Phys. Rev. E 61, 5557 (2000).
112

[102] A. Melzer, S. Nunomura, D. Samsonov, and J. Goree, Phys. Rev. E 62, 4162 (2000).
112

[103] D. H. E. Dubin, Phys. Plasmas 7, 3895 (2000). 112

[104] V. Nosenko, J. Goree, Z. W. Ma, and A. Piel, Phys. Rev. Lett. 88, 135001 (2002).
113

[105] O. Havnes et al., J. Geophys. Res. 100, 1731 (1995). 114, 164
174 BIBLIOGRAPHY

[106] S. V. Vladimirov, P. V. Shevchenko, and N. F. Cramer, Phys. Rev. E 56, 74 (1997).

115

[107] T. Misawa et al., Phys. Rev. Lett. 86, 1219 (2001). 115, 116

[108] S. Nunomura et al., Phys. Rev. Lett. 89, 035001 (2002). 116, 117

[109] B. Liu, K. Avinash, and J. Goree, Phys. Rev. Lett. 91, 255003 (2003). 116, 117

[110] J. J. Thomson, Philos. Mag. 39, 237 (1904). 120

[111] Strongly Coupled Coulomb Systems, edited by G. J. Kalman, J. M. Rommel, and

K. Blagoev (Plenum Press, New York, 1998). 120

[112] W.-T. Juan et al., Phys. Rev. E 58, 6947 (1998). 120, 123

[113] M. Klindworth, A. Melzer, A. Piel, and V. Schweigert, Phys. Rev. B 61, 8404 (2000).
120, 123

[114] O. Arp, D. Block, A. Piel, and A. Melzer, Phys. Rev. Lett. 93, 165004 (2004). 120,
128, 129

[115] A. Melzer, Phys. Rev. E 73, 056404 (2006). 121, 122

[116] T. E. Sheridan and K. D. Wells, Phys. Rev. E 81, 016404 (2010). 121

[117] A. Melzer, Phys. Rev. E 67, 016411 (2003). 123, 133, 134, 135, 136, 137

[118] V. M. Bedanov and F. Peeters, Phys. Rev. B 49, 2667 (1994). 123

[119] O. Arp, D. Block, M. Klindworth, and A. Piel, Phys. Plasmas 12, 122102 (2005).
128

[120] S. Käding et al., Phys. Plasmas 15, 073710 (2008). 128, 129, 130

[121] R. W. Hasse and V. V. Avilov, Phys. Rev. A 44, 4506 (1991). 128

[122] M. Bonitz et al., Phys. Rev. Lett. 96, 075001 (2006). 128, 129

[123] D. Block et al., Phys. Plasmas 15, 040701 (2008). 128, 129

[124] P. Ludwig, S. Kosse, and M. Bonitz, Phys. Rev. E 71, 046403 (2005). 129

[125] S. Apolinario, B. Partoens, and F. Peters, New J. Phys. 9, 283 (2007). 129

[126] Y.-J. Lai and L. I, Phys. Rev. E 60, 4743 (1999). 129

[127] V. A. Schweigert and F. Peeters, Phys. Rev. B 51, 7700 (1995). 130
BIBLIOGRAPHY 175

[128] R. Ichiki et al., Phys. Rev. E 70, 066404 (2004). 137, 139

[129] J. Winter and G. Gebauer, J. Nucl. Mater. 266-269, 228 (1999). 149

[130] J. Sharpe, D. Petti, and H.-W. Bartels, Fusion Engineering and Design 63-64, 153
(2002). 149, 150

[131] S. I. Krasheninnikov, A. Y. Pigarov, R. D. Smirnov, and T. K. Soboleva, Contrib.

Plasma Phys. 50, 410 (2010). 149, 150

[132] H. Kersten et al., New J. Phys. 5, 93 (2003). 151

[133] P. Roca i Cabarrocas, in Advances in Dusty Plasmas, edited by Y. Nakamura, T.

Yokota, and P. Shukla (Elsevier, Amsterdam, 2000), p. 213. 152

[134] O. Havnes, in Dusty Plasmas in the new Millenium, edited by R. Bharuthram, M. A.

Hellberg, P. Shukla, and F. Verheest (AIP Conference Proceedings, Melville, New
York, 2002), Vol. 649, p. 13. 154

[135] V. Nussbaumer et al., J. Geophys. Res. 101, 19161 (1996). 156

[136] M. Horanyi, in Dusty Plasmas in the new Millenium, edited by R. Bharuthram,

M. A. Hellberg, P. Shukla, and F. Verheest (AIP Conference Proceedings, Melville,
New York, 2002), Vol. 649, p. 22. 156, 159, 160, 161

[137] C. K. Goertz and G. Morfill, Icarus 53, 219 (1982). 162, 163

[138] P. V. Bliokh and V. V. Yaroshenko, Sov. Astron. 29, 330 (1985). 163

[139] A. Brattli, O. Havnes, and F. Melandsø, Phys. Plasmas 9, 958 (2002). 164
Index
acethylene, 140 1D, 121
aerosol, 145 2D, 122
angular correlation function, 59 3D, 128
Ashkin, 36 magic number, 135
normal modes, 130–139
Bohm velocity, 17, 47 periodic table, 122, 129
Boltzmann factor, 10, 24, 89, 95 phase transition, 137
breathing mode, 126 stability, 134
Brownian motion, 116, 133 zigzag transition, 121
capacitance, 17, 21 dust ion-acoustic wave, 94
Cassini spacecraft, 160 dispersion relation, 96
charge, 17, 43 experiments, 96
position dependent, 49 dust ion-cyclotron wave, 97
charging, 6 dust lattice wave, 100
charging time, 21, 23 1D, 101
collsions, 12 compressional, 104
energy barrier, 11 experiments, 107–111
fluctuations, 23 shear, 105
impact parameter, 7, 31 sound speed, 103, 105
in rf discharge, 22 transverse, 114
OML, 7–17 dust plasma frequency, 87, 103
photoelectron current, 13 dust removal, 148
secondary electron current, 14, 19 dust-acoustic wave, 87
collisions, 63 dispersion relation, 90
Coulomb cluster, 120, see dust cluster experiments, 92
Coulomb coupling parameter, 51 wave speed, 90
Coulomb crystallization, 51, 54 dust-cyclotron wave, 96
cross section dynamical matrix, 131
Coulomb, 31
electrostatic precipitators, 148
electron collection, 10
Epstein friction, 35
geometric, 8, 35, 36
ion collection, 8 Faraday cup, 44
floating potential, 6, 15, 16, 26
Debye length, 11, 29, 30, 32, 118
force
Debye-Hückel equation, 29
collection, 31
Debye-Hückel potential, 53, 64
Coulomb, 31
dipole moment, 29, 65
drag, 30
drift velocity, 12, 17, 30
electric field, 27
dust cluster, 120

176
INDEX 177

friction, 35 phase transition, 81–86

gravity, 27 defects, 85
ion drag, 31 dust cluster, 137
neutral drag, 35 plasma potential, 24, 26
photophoretic, 37 Poisson’s equation, 88, 95
polarization, 29 polar mesospheric summer echoes, 154
radiation pressure, 36 polymorphous silicon, 153
thermophoresis, 35
quasineutrality, 24, 42, 89, 96
Galileo spacecraft, 156
gravito-electrodynamics, 154 Saturn, 160
growth, 140 screening strength, 53, 132
sheath model, 41
halo orbits, 160 matrix sheath, 41
Havnes parameter, 26 silane, 140
HMDSO, 140 silicon clusters, 142
spectral power density, 134
Jupiter, 156 Spitzer, 16
Kepler, 12, 163 spokes, 161
killer particle, 147 strong coupling, 51
structure factor, 59
Langevin, 85
thermal velocity, 9
Langmuir, 6, 7
laser manipulation, 36, 43, 72, 107, 110, vertical alignment, 65
111 instability of, 75–81
vertical resonance
Mach cone, 111, 164
linear, 42
Madelung energy, 52
nonlinear, 47
mass spectrometry, 142
parametric, 45
Mathieu’s equation, 46
self-excited, 49
Maxwellian distribution, 9, 14
methane, 140 wakefield, 65
MF particles, 42 waves
microgravity, 38 dust ion-acoustic, see dust ion-
acoustic wave
noctilucent clouds, 154
dust ion-cyclotron, see dust ion-
non-reciprocal attraction, 70, 72
cyclotron wave
OCP, 51 dust lattice, see dust lattice wave
OML current, 9 dust-acoustic, see dust-acoustic wave
one-component plasma, see OCP dust-cyclotron, see dust-cyclotron
wave
pair correlation function, 57 Wigner-Seitz cell, 57
178 INDEX

Wigner-Seitz radius, 51

Yukawa potential, 53
Yukawa system, 53