Morgan - Electrostatics 2003 (Institute of Physics Conference Series) - Institute of Physics Publishing (2004)

Electrostatics 2003
© 2004 by Taylor & Francis Group, LLC

Other titles in the series
The Institute of Physics Conference Series regularly features papers presented at
important conferences and symposia highlighting new developments in physics and
related fields. Previous publications include:
180 Microscopy of Semiconducting Materials 2003

Papers presented at the Institute of Physics Conference, Cambridge, UK
Edited by A G Cullis and P A Midgley
179 Electron Microscopy and Analysis 2003

Papers presented at the Institute of Physics Electron Microscopy and Analysis Group
Conference, Oxford, UK
Edited by S McVitie and D McComb
177 Optical and Laser Diagnostics 2002

Papers presented at the First International Conference, London, UK
Edited by C Arcoumanis and K T V Grattan
174 Compound Semiconductors 2002

Papers presented at the 29th International Symposium on Compound
Semiconductors, Lausanne, Switzerland
Edited by M Ilegems, G Weimann and J Wagner
173 GROUP 24: Physical and Mathematical Aspects of Symmetries

Papers presented at the 24th International Colloquium, Paris, France
Edited by J-P Gazeau, R Kerner, J-P Antoine, S Métens and J-Y Thibon
172 Electron and Photon Impact Ionization and Related Topics 2002
Papers presented at the International Conference, Metz, France
Edited by L U Ancarani
171 Physics of Semiconductors 2002

Papers presented at the 26th International Conference, Edinburgh, UK
Edited by A R Long and J H Davies
170 Compound Semiconductors 2001

Papers presented at the 28th International Symposium on Compound
Semiconductors, Tokyo, Japan
Edited by Y Arakawa, Y Hirayama, K Kishino and H Yamaguchi
169 Microscopy of Semiconducting Materials 2001

Papers presented at the 12th International Conference on Microscopy of
Semiconducting Materials, Oxford, UK
Edited by A G Cullis and J L Hutchison

Electrostatics 2003
Proceedings of the Electrostatics Conference of the Institute of Physics

held in Edinburgh, UK,
23–27 March 2003
Edited by
Hywel Morgan
New York London

Published in 2004 by Published in Great Britain by
Taylor & Francis Group Taylor & Francis Group
270 Madison Avenue 2 Park Square
New York, NY 10016 Milton Park, Abingdon
Oxon OX14 4RN

No claim to original U.S. Government works
Printed in the United States of America on acid-free paper
10 9 8 7 6 5 4 3 2
International Standard Book Number-10: 0-7503-0949-0 (Print Edition) (Hardcover)
International Standard Book Number-13: 978-0-7503-0949-3 (Print Edition) (Hardcover)
This book contains information obtained from authentic and highly regarded sources. Reprinted material
is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable
efforts have been made to publish reliable data and information, but the author and the publisher cannot
assume responsibility for the validity of all materials or for the consequences of their use.
No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic,
mechanical, or other means, now known or hereafter invented, including photocopying, microfilming,
and recording, or in any information storage or retrieval system, without written permission from the
publishers.
For permission to photocopy or use material electronically from this work, please access
www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC)
222 Rosewood Drive, Danvers, MA 01923, 978–750–8400. CCC is a not-for-profit organization that
provides licenses and registration for a variety of users. For organizations that have been granted a
photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Library of Congress Cataloging-in-Publication Data

Catalog record is available from the Library of Congress
Visit the Taylor & Francis Web site at

http://www. taylorandfrancis.com

Preface
Electrostatics is a diverse and expanding field with an ever increasing range of applications.
This trend is reflected in this conference, the eleventh in a series that has continued to grow
since its inception in 1967. For the first time, the conference was held as part of the Institute
of Physics Congress, which this year was held in Edinburgh. The electrostatics conference
attracted over 80 submissions, and was attended by well over 120 delegates. A large number
of these papers were submitted for publication, and I am pleased to say that 58 papers made
it through the refereeing process to appear in these proceedings. I shall be eternally grateful
to all the reviewers who demonstrated such patience and diligence, and without whose help
this proceedings would not have seen the light of day.
A glance through the proceedings will bear testimony to the diversity of research and
technology in modern electrostatics. The Bill Bright Memorial Lecture was given by Prof.
T. B. Jones, University of Rochester. His talk on Electrostatics and the Lab on a Chip was
also the plenary scientific talk of the Institute’s Congress. Prof. Jones’ obvious enthusiasm
captivated the audience as he highlighted recent advances in the application of electrostatics
in microsystems, particularly for the manipulation of fluids and particles. Several eminent
speakers were invited to provide us with up to date reviews on their research. The first
plenary lecture was given by Dr. Ulrich von Piddol of Physik Technische Bundesanstalt,
Germany, on ignition hazards. He described the general issues associated with electrostatic
detection and hazard prevention, focusing on fires caused by electrostatic discharges.
Dr. Carol Livermore from MIT described how electrostatic machines operate on the
micro-scale, and demonstrated how MEMS technology has lead to a new generation of
electrostatic machines for micro-turbines and motors. Prof. Masao Washizu from Japan
gave an illustrative talk showing how electrostatics can be applied on an even smaller scale
to manipulate and hold single molecules of DNA. Prof. Antonio Castellanos provided a
detailed insight into the effects of electric fields on fluids in micro-devices, where the
scales are different and it is relatively easy to generate extremely high electric fields with
only a few volts. Dr Jaakko Passi from VTT Industrial Systems, Finland presented an overview
of ESD issues pertaining to the use of protective materials and garments routinely used in
the electronics manufacturing and fabrication industry. Prof. Bill Greason from the
University of Western Ontario gave an informative review of measurement methods in
electrostatics. The conference was brought to a close with a lecture by Prof. Peter Castle
from the University of Western Ontario who showed how electrostatics is ever present in our
environment, from lighting conductors to complex electrostatic precipitators for recycling
plastics.
I would like to extend my sincere thanks to the members of the organising committee
and to the papers advisory committee for guidance and suggestions. I would like to thank
Dr Jeremy Smallwood, the conference chairman for all his support and advice, Belinda
Hopley of the Institute of Physics for ensuring that the conference ran smoothly and for
organising the conference dinner, and to both Belinda and Geraldine Coyne for assistance
in editing and compiling the proceedings.
Once again, the conference lived up to expectations. The range of international attendees
from industry and academia, came from as far a field as the USA and Japan, as well as the UK

vi
and Europe. The equally wide range of excellent and stimulating papers, all demonstrate
the continued international interest in electrostatics and its nearly ubiquitous practical
relevance to the modern world. Finally, I would like to thank everyone who attended and
contributed to making this conference such a memorable event.
Prof. Hywel Morgan

Chair, Programme Committee

Contents
Preface v
Bill Bright Memorial Lecture

Electrostatics and the lab on a chip
TB Jones 1
Section 1: Hazards
Electrostatic ignition hazards—occurrence, detection and prevention
U von Pidoll 11
Ignition hazards associated with earthing and bonding of charged conductive objects
K Schwenzfeuer and M Glor 19
An investigation of the electrostatic ignition risks associated with plastic coated metal
G P Ackroyd and S Puttick 25
Flow electrification in transformers: sensor prototype for electrostatic hazard
O Moreau, T Paillat and G Touchard 31
Explosibility of shredder dusts for electrical appliances
M Nifuku, J Gatineau, C Barre, S Horiguchi, H Katoh and M Hatori 37
Section 2: MEMS and Applications

Microscale electric induction machines for power applications
C Livermore, A Forte, T Lyszczarz, S D Umans and J H Lang 45
Micro-machined variable capacitors for power generation
P Miao, A S Holmes, E M Yeatman, T C Green and P D Mitcheson 53
Electrostatic charging of trigger actuated spray devices
L F Gaunt and J F Hughes 59
Unipolar charging and contact discharging of insulating particles on the surface of a
grounded electrode
A Samuila, A Mihalcioiu, A Urs and L Dascalescu 65
The investigation of the ozone productivity of a new discharge type
I Jenei, E Kiss and I Berta 71
Electrostatic forces on ion-charged toner particles
D A Hays and J Q Feng 77

viii
Optimizing the process parameters of injection moulding to minimize the static charge
of polypropylene test rod
M Murtomaa, S Kankaanpää, J Nurmio, M Leino, J Mäkelä, P Järvelä, E Laine and
V-P Lehto 83
Section 3: Bioelectrostatics
Bio-nanotechnology of DNA based on electrostatic manipulation
M Washizu 89
Electrorotation of dense colloidal suspension
K W Yu, J P Huang and G Q Gu 95
The effect of disaccharides on the transport of lipophilic ions in cell membranes studied
by electrorotation
R Reuss, M Horbaschek, J M Endter, U Zimmermann and V L Sukhorukov 101
Cell sorting and separation using dielectrophoresis
D Holmes and H Morgan 107
Dielectrophoretic transport and sorting of particles using an electrode micro-array
B Malyan, J Kulon and W Balachandran 113
AC electrokinetic focussing in microchannels: micro- and nanoparticles
H Morgan, D Holmes and N Green 119
Section 4: Measurements
A wide bandwidth probe for electrostatic discharge measurements
J M Smallwood and G L Hearn 125
Predicting the maximum voltages expected on inhabited cleanroom garments in
practical use
J N Chubb, P Holdstock and M Dyer 131
Contact charging method for the measurement of charge decay in electrostatic
dissipative materials
J Paasi, T Kalliohaka, T Luoma, R Ilmén and S Nurmi 137
A particle charge spectrometer for determining the charge and size of individual dust
grains on Mars
S Fuerstenau and G Wilson 143
Measurement of optical intensity and fluence generated by spark discharges
J C Crager and M N Horenstein 149
Atmospheric ion spectra and the rate of voltage decay of an aspirated cylindrical
capacitor
K L Aplin 155

ix
Spray current dependence on flow rate and conductivity in cone-jet mode

vacuum spraying
K L Smith and J P W Stark 161
Visualization and particle image velocimetry measurements of electrically generated
coherent structures in an electrostatic precipitator model
J Mizeraczyk, M Kocik, J Dekowski, J Podliêski, T Ohkubo and S Kanazawa 167
Section 5: EHD & Numerical Modelling

Electrohydrodynamics in microelectrode structures
A Castellanos, A González, A Ramos, N G Green and H Morgan 175
Electrohydrodynamic atomization of viscous liquids
A Jaworek, W Balachandran, A Krupa, J Kulon and W Machowski 181
Pumping of electrolytes using arrays of asymmetric pairs of microelectrodes subjected
to ac voltages
A Ramos, A González, A Castellanos, N G Green and H Morgan 187
Electrode polarisation, dielectrophoresis and electrorotation
N G Green 193
Numerical simulation of fine particles charging and collection in an electrostatic
precipitator with regular barbed electrodes
L MDumitran, P Atten and D Blanchard 199
The numerical simulation of multi-field wire-plate electrostatic precipitators
D Brocilo, J S Chang and R D Findlay 207
Numerical modelling of dielectrophoretic effect for sub-micron particles manipulation
B Malnar, W Balachandran and F Cecelja 215
Modelling studies of charged particle interactions for a space application
K L Aplin and V P Tarakanov 221
Analytical solutions of surface potential distribution on thin insulators having grounded
backing conductor and their applications to electrostatic characterisation
A Ohsawa and M Ohuchi 227
Towards an improvement of thermal modelling and mathematical deconvolution in
FLIMM
D Marty-Dessus, A Petre, L Berquez and J L Franceschi 233
Section 6: ESD
Electrostatic testing of ESD-protective clothing for electronics industry
J Paasi, S Nurmi, T Kalliohaka, G Coletti, F Guastavino, L Fast, A Nilsson, P Lemaire,
J Laperre, C Vogel, J Haase, T Peltoniemi, G Reina, A Börjesson and J Smallwood 239
Assessment of ESD threats to electronic components and ESD control requirements
J Smallwood and J Paasi 247

x
Calorimetry in the detection of discharge events

Z Kucerovsky, W D Greason and M Wm Flatley 253
Electrical circuit topologies for pulsed corona plasma generation
K Yan, E J M van Heesch, P A A F Wouters, A J M Pemen and S A Nair 261
Martian Regolith simulant particle charging experiments at low pressures in the
presence of corona fields
F B Gross 267
Influence of an insulating flat plate on a DC surface corona discharge at various air
relative humidities
C Louste, E Moreau and G Touchard 273
Detachment and reattachment of a low velocity airflow along an inclined wall actively
controlled by a low frequency square wave corona discharge
A Labergue, L Leger, E Moreau, G Touchard and J P Bonnet 279
Surface corona discharge along an insulating flat plate in air applied to
electrohydrodynamically airflow control: electrical properties
E Moreau, G Artana and G Touchard 285
Optimized geometry of a corona electrode arrangement for water ozonization
I Suarasan, R Morar, L Ghizdavu and L Dascalescu 291
Characterisation of ESD waveform and peak current from charged printed wiring
board to ESD hand tools
T Kalliohaka and J Paasi 297
The role of capacitance in corona-electrode arrangements
C G Noll, N R Greene, S T Ashman and M A Catino 303
Section 7: Environment
Electrostatics and the environment
G S P Castle 309
Measurement methods in electrostatics applications: review and trends
W D Greason 315
Pulsed arc discharges for water treatment and disinfection
H Z Zastawny, H Romat, N Karpel Leitner and J S Chang 325
Electrostatic spray application of decontaminant agents onto the human body as a
bioterrorism countermeasure: process development and evaluation
S E Law, S C Cooper and M A Harrison 331
Atmospheric turbulence and surface atmospheric electricity observations
R G Harrison 337
Dust particles removal by a novel two-stage electrostatic precipitator
A Jaworek, A Krupa and K Adamiak 343

xi
Dust particle removal by wet-type electrostatic scrubber

A Krupa, A Jaworek, T Czech, MLackowski and J Luckner 349
Expert system applications for electrostatic separation processes
L Dascalescu, A Samuila, M Mihailescu, A Iuga and R Köhnlechner 355
Comparative analysis of computer-simulated and experimental sparking voltage of the
wire-plate system
Z Dudzicz 361
Decomposition of diesel particulate materials and nitric oxides using a dielectric barrier
discharge
Y Yamagata, T Matsui, T Ebihara and K Muraoka 367
Questions and Answers 373

1
Electrostatics and the Lab on a Chip
T B Jones
University of Rochester, Rochester, NY 14625, USA
Abstract. At the heart of the laboratory on a chip are chemical or biochemical probes designed
to detect specific molecules or reaction products. For these probes to perform their function, a
microfluidic “plumbing system” is needed to accept samples, then manipulate, dispense, and
distribute tiny liquid volumes on the chip. The most promising mechanisms for handling such
small quantities of liquid, from microliters down to tens of picoliters, are all electrostatic, e.g.,
electrocapillarity, electroconvection, electrophoresis, electro-osmosis, and dielectrophoresis.
Despite years of study of these effects, the body of existing work is an imperfect guide to their
effective exploitation in microfluidic applications. Attempts to harness electrostatic forces in
structures <100 microns in size often encounter unanticipated behavior. Such surprises can be
attributed to dramatic changes in the relative influences of the various competitive forces (e.g.,
viscous shear, surface wetting, and capillarity). In this paper, some of the more promising
avenues of research on electric-field-mediated microfluidics are examined. To explicate why
and when electrostatic forces can be advantageous in the lab on a chip, Trimmer’s bracket
notation is invoked to perform an investigation of the scaling laws for microfluidic systems.
The methodology facilitates examination of the effects of device size on throughput, processing
time, temperature rise, and other important measures of system performance.
1. Introduction
Micro total analysis system (µTAS) technology [1] seeks to replace the conventional
tools of chemists, biological scientists, and medical researchers with programmable, robotic
microsystems capable of performing rapid chemical/biochemical analyses or protocols
using very small inventories of liquid analyte and reagent. Such systems, intended to
replace conventional test tubes, well plates, cell culture dishes, capillaries, as well as liquid
chromatographs, cell cytometers and other cumbersome diagnostic instruments, exemplify
the laboratory on a chip. This revolutionary concept promises massive parallelism plus
sufficient improvements in automation and speed to eliminate or greatly ameliorate serious
bottlenecks in conventionally equipped laboratories that stem from the unavoidable need
to perform large numbers of experiments to satisfy the requirements of statistical significance
of data.
Applications for the laboratory on a chip range widely, from microreactors for the
chemical industry and pharmacological screening systems for drug discovery, to
microbiological diagnostic machines and genetic research tools. In combinatorial chemistry,
the ability to conduct large numbers of reactions under closely controlled conditions
automatically and simultaneously using very small analyte and reagent volumes is
understandably attractive. The use of small liquid inventories (microliters or less) will
significantly reduce mixing and reaction times, to say nothing of cost. Furthermore, being
able to operate with small quantities is attractive in biomedical research where, quite often,

2
only very small inventories of rare cells or substances can be isolated and made available
for testing. Another benefit of these robotic microsystems is that, with minimal need for
human intervention, experiments and tests involving dangerous chemicals and pathogens
can be conducted more safely.
Key sub-systems of the laboratory on a chip include a user interface, sensors,
(bio)chemical probes, and some sort of microfluidic system. The last of these, the microfluidic
system, is the “plumbing infrastructure” essential for the manipulation, dispensing, and
transport of liquids around the chip. A variety of surface and volume force mechanisms are
under consideration to achieve precision fluid control, including capillarity [2], surface
wetting [3], electro-osmosis [4], electroconvection [5], electrowetting [6], and liquid
dielectrophoresis (DEP) [7]. One interesting hybrid µTAS concept would use microfluidics
to transport and dispense liquid aliquots that themselves contain suspended bioparticles
(cells, DNA, etc.). After dispensing small volumes of the particle-bearing liquid at pre-
determined sites on a chip, the DEP force exerted on individual particles within each
droplet would be used to probe the particles, sorting and differentiating them according to
requirements of the process [8], The subject of this paper is the electrostatic forces available
for particle and liquid manipulation in the laboratory on a chip.
2. Electrostatic force scaling for microsystems

During the early years of the VLSI circuit technology revolution, Richard Feynman was
already anticipating MEMS (microelectromechanical systems) and µTAS technology. Such
remarkable prescience was no doubt based on a recognition that the laws of physics do not
exhibit symmetry with respect to changing physical dimensions [9], in other words, that the
relative magnitudes of forces are rearranged when size is changed. One consequence of
these “rearrangements” crucial to MEMS and µTAS is that electrostatic forces become
increasingly important as devices become smaller. William Trimmer recognized the broad
significance of force scaling in engineered microsystems and devised a bracket notation,
which he used for systematic investigations of their scaling laws. He was able to show just
how electrostatic forces effectively dominate over magnetic forces in MEMS devices [10].
In this paper, we employ an approach based on Trimmer’s methods to explore the scaling
advantages of electrostatic forces for the laboratory on a chip using two examples: DEP
manipulation of particles and electric-field-mediated microfluidics.
3. Scaling laws for particle DEP

Dielectrophoresis, the force exerted by a
nonuniform electric field on uncharged,
dielectric particles [11], either attracts particles
to regions of strong electric fields or repels them,
depending on their polarizability relative to the
medium in which they are dispersed, see Fig. 1.
Without loss of generality from the standpoint
of scaling, we restrict attention here to spherical
Fig. 1. Dielectrophoretic (DEP) attraction (or
particles of radius R and relative dielectric repulsion) of particles depends on the relative
constant κp in a fluid medium with κm. The DEP values of particle & suspending liquid dielectric
force may be expressed as a series of multipolar constants. Here, κp1>κm and κp2<κm.

3
contributions, including dipolar, quadrupolar, and higher-order terms [12].
(1)
where Kn=n(2n+1)(κp-κm)/[(n+1)κm+n κp] is a polarization coefficient and ε o is the

permittivity of free space.
The need for a systematic investigation of the scaling rules for DEP is evident from the
observation that the dipolar term depends on R3 while the quadrupolar term is proportional
to R5. First, we define scale factors for the important parameters: α for the electrode
dimensions, β for the particle radius, and γ for voltage. Then, for example, the electric field
E and the del operator ∇ scale as γα”1 and α”1, respectively. Adopting Trimmer’s bracket
notation [10], the scaling laws for the two leading terms in are:
(2)
In this notation, brackets enclosing an algebraic term on the Ihs of an equation represent the
operation of extracting the scaling factors of the term. Subscripts identify any constraints
imposed. On the rhs in brackets are the scale factor “values”. When convenient, a matrixlike
format with columns and rows is used to tabulate different force laws and/or physical
constraints. For example, on the rhs of Eq. (2), the upper row is reserved for the dipole force
term and the lower row is for the quadrupole; to keep track of more multipolar terms, one
simply adds more rows. The value of the bracket notation for MEMS systems is realized
when force laws are combined with various constraints to gain insights into the effect of
reducing physical dimensions, as exemplified below.
3.1 Particle levitation and trapping

The DEP force offers a gentle, controllable means to levitate or to trap particles suspended
in aqueous media using easily fabricated microelectrodes. For example, Fig. 2 depicts a
planar quadrupolar trap for + and -DEP.
Static force balance for a particle levitated or trapped against the gravitational force is:
(3)
where ρp and ρ m are

particle and liquid
medium densities,
respectively, and g- is
the gravitational
acceleration vector.
Consider adjusting the
scale factors while
maintaining the
particle at a fixed, that
Fig. 2. Planar quadrupole particle trap. At the left, particles exhibit +DEP
is, a geometrically and thus collect at field maxima. At the right, particles exhibit “DEP,
similar position forming a loose ensemble near the center where the field magnitude goes
within the structure. to zero, (these photos courtesy of P.Gascoyne, Univ. of Texas at Houston)

4
One can analyze this situation by per-forming the previously introduced bracket operation
on Eq. (3) and equating the terms.
dipole only: γ=α3/2; quadrupole only: γ=α5/2 β”1 (4)
These relations, one each for dipolar and quadrupolar structures, readily reveal one
advantage of reducing the size of DEP levitation structures. Keeping particle size fixed (β
=1) while reducing all electrode dimensions by a factor of 2 (i.e., α=1/2), reduces the
levitation voltage by a factor of 23/2 ~ 2.83 for dipolar electrodes, and by 25/2 ~ 5.66 for
quadrupolar structures. On the other hand, keeping the electrodes fixed (α=1) while
reducing particle size by the factor of 2 (i.e., β=1/2) leaves levitation voltage unchanged
for dipolar structures, but increases it by a factor of 2 for quadrupolar structures. Eq. (5)
summarizes these voltage-scaling rules for fixed particle size, and also includes scaling
rules for electric field E and temperature rise due to Joule heating ∆T for these particle
traps. Refer to the Appendix for a short synopsis of the scaling law governing temperature.
(5)
It is to be recognized that scaling rules do not provide all the information necessary to
decide between dipolar and quadrupolar levitators for a given application; nevertheless,
the exercise does facilitate useful comparisons between the two electrode structures. Of
these two types of particle traps, quadrupolar electrodes seem to have clear advantages
with respect to voltage requirements, maximum electric field, and heating as size is reduced.
These advantages are not fully realized in the control of biological cells using electrode
structures in the 10 to 100 micron size range, but could become very valuable in
nanostructures under development for analysis and manipulation of DNA and other
nanoscale proteins and particles.
3.2 Particle dynamics

Most applications for particulate DEP in the lab on a chip will quite likely involve dynamic
behavior. Thus, one critical dynamic performance measure is the response time t, e.g., the
time required to achieve some targeted level of particle separation or concentration, or the
residence time required in a flow-through analyzer. To investigate, we start with the equation
of motion for the particle. On the micron scale, momentum is negligible and the drag force
dominates. If Stokes’ drag can be used, the equation of motion takes a simple form.
(6)
where µm is the dynamic viscosity of the liquid and v- is particle velocity. Let the scaling
factor for all time variables be τ. Assume that the dipole force dominates, and perform the
bracket operation on Eq. (6) subject to the fixed particle
–1
size condition. Two physical
constraints are considered: fixed velocity v (i.e., ατ =1) and fixed response time t (i.e.,
τ =1). Of these two, fixed response time seems a more relevant performance measure for the
lab on a chip.

5
(7)
Eq. (7) reveals to us that decreasing structure size while keeping the response time t fixed
(i.e., τ =1) achieves significant benefits in reducing voltage requirements, electric field
magnitude, and operating temperature.
4. Scaling laws for E-field-mediated microfluidics

Most electric-field-mediated microfluidic schemes can be categorized as electrochemical
and/or electrohydrodynamic (EHD) in nature. Yet, despite the authoritative works of Levich
[13], Melcher [14], Taylor [15], and others, surprises await those attempting to harness
electrical forces to actuate liquids in <100 micron sized structures. These surprises stem
from the unfamiliar ways that well-known phenomena combine on the microfluidic scale.
The two cases considered here are electro-osmotic flow and electrowetting/DEP actuation.
4.1 Electro-osmotic flow

Ions in a liquid electrolyte close to a metal
wall are subject to strong diffusion effects
that separate the + and – species. This
selective process creates a thin sheath called
the Helmholtz double-layer adjacent to the
boundary. In this layer, ions of one sign
predominate. A tangential DC electric field
imposed along the length of the channel
acts on the ions in this layer and creates a
Fig. 3. Basic description of DC electro-osmotic
pumping force. Electro-osmosis shows effect in a channel. The tangential electric field
promise as a controllable microfluidic imposed by external DC voltage exerts a force
pumping scheme, and Levich’s formula [13, on the free ions in the double-layer adjacent to
pp 474–475] may be used to assess its utility the walls. The resulting shear force pumps the
in microsystems. liquid.
(8)
where ζ is the zeta potential and is the essentially flat velocity for plug flow in the
channel center. Refer to Fig. 3. Scaling Eq. (8) yields the relationship α 2=γτ. The volume
flow of liquid W scales as α3 τ -1. Subject to the condition that the microchannel dimensions
remain large compared to the thickness of the Helmholtz layer, scaling laws for the important
performance parameters of an electro-osmotic pump may be obtained.

6
(9)
It is quite evident from Eq. (9) that fixing volume flow while simultaneously reducing
structure size leads to excessive voltages, electric field magnitudes, and operating
temperature values. On the other hand, these same quantities are reduced if the response
time is fixed as size is decreased.
4.2 Static electrowetting and DEP actuation

Electrowetting using dielectric-coated
electrodes (EWOD) and DEP liquid actuation
are, respectively, the low- and high-frequency
manifestations of the electrostatic force exerted
by a non-uniform electric field on polarizable
media [16]. Figs. 4 and 5 show EWOD and
DEP actuation structures. Because these
mechanisms are capable of rapid manipulation
and movement of relatively large liquid
volumes, they are receiving considerable
attention for use in the lab on a chip. In this
and the next section, respectively, we
investigate static liquid orientation against
gravity and the dynamic transient response of
a liquid. These phenomena are chosen to be
representative models for EWOD and DEP
applications in a microfluidic system. Fig. 4. Splitting of ~200 nanoliter water volume
(EWOD) effect. In the sequence of micrographs
Consider the vertical, parallel-plate into two equal portions using the electrowetting
geometry of Fig. 6. The Structure on the left shown, the individually addressed,
uses a conductive liquid and electrodes coated dielectriccoated electrodes are viewed from
with an insulator of thickness d and dielectric above through a transparent upper electrode
constant κ d, while that on the right uses spaced ~100 microns from the lower electrodes,
(these photos courtesy of C-J Kim, UCLA)
insulating liquid and bare electrodes. liquid
Limiting expressions for static height-of-rise h as defined in Fig. 6 are:
(10)

7
Fig. 5. DEP actuation of water using co-planar electrodes patterned on a substrate and coated with a
dielectric layer (a). Applying AC voltage for ~100 msec, a finger of liquid forms and moves rapidly from
left to right (b). When finger reaches the circular electrodes at right, voltage is removed (c) and capillary
instability pinches off a droplet (d).
where the subscripts “u” and “l” refer to properties of the upper and lower liquids,
respectively. In the analysis to follow, it is convenient to distinguish d and H from the other
electrode dimensions with new scaling factors δ and η, respectively. Performing the bracket
operation on the expressions in Eq. (10), the result is a set of scaling rules for the fixed
height-of-rise constraint, (i.e., η=1).
(11)
Note that for EWOD, the electric field E= V/

2d in the insulating dielectric layer and zero
in the conductive liquid. Thus, there is
virtually no Joule heating. Under the fixed
height stipulation, EWOD does not scale as
advantageously as DEP, though the constant
height constraint is probably not realistic in
a practical situation. Fig. 6. A basic geometry useful in formulating
scaling laws for microfluidic applications of
4.3 Dynamic electrowetting and DEP electrowetting-ondielectric (EWOD) and the
dielectrophoretic (DEP) actuation, (a) EWOD
actuation electrodes are coated with a very thin dielectric
Probably more relevant as a performance layer, d<<D, and operated at low coating becomes
negligible and the height of rise frequency. At
measure for application in the lab on a chip high frequencies, the influence of any
than the hydrostatic relationship of h to V is phenomenon becomes identical to the DEP height-
the transient behavior. Of specific interest is ofrise experiment depicted at the right. (b) DEP
the time required for the liquid to rise to the electrodes require no dielectric layer if the liquid
top of the structure when the voltage is turned is sufficiently insulative. The transient response
of this geometry when voltage is applied
on suddenly. For the purposes of the scaling suddenly serves as an microfluidic device for the
analysis, we employ a simple laminar flow laboratory on a chip. adequate model for the
model for the fluid mechanics. dynamic response of a chip.

8
Fig. 7. The electrostatic droplet transport system shown at

left consists of parallel electrodes arranged like railroad
ties, embedded in a substrate and coated with a dielectric
layer. Voltage is applied to the electrodes in sequence
causing conductive water droplets to move in response to
an electrostatic attractive force. In the micrograph sequence
shown below (a), the electrode pitch is ~200 µm. A pair of
1 µl droplets is transported from left to right (b); they
coalesce at the junction of the two transport structures just
to the right of center (c). (photos courtesy of M. Washizu,
Tokyo University, Japan)
(12)
where the fluid velocity vz ∝ dh/dt and Te is the normal electric stress. Gravity may be
ignored, but surface wetting hysteresis, also ignored here, may be important.
(13)
Combining Eqs. (12) and (13) and then performing the bracket operation, we obtain:
(14)
which is identical to Eq. (11). Again, EWOD actuation has fewer advantages as electrode
structures are scaled down. In particular, reducing the dielectric thickness d provides no
benefit at all.
5. Conclusion
Any laboratory on a chip based on wet chemical probes requires a plumbing system to
dispense and manipulate analyte and reagent liquids. The most promising schemes for
achieving this liquid control exploit electrostatic fields. Similarly, electrostatics can be
used for precise, controllable conveyance, separation, and collection of particles in liquid
suspension. The scaling analyses summarized in this paper reveal unique opportunities for

9
electrostatic forces in microfluidic systems. It is interesting to note that scaling arguments

do not require analytical or numerical solution of boundary value problems or initial value
transients. Only the governing equations are needed; the scaling relations can be extracted
without solving these equations. The value of scaling arguments lies in the ease with which
one can identify classes of force effects and sets of physical constraints that yield either (i)
significant advantage as size is scaled down or (ii) insurmountable difficulties (e.g., higher
required electric fields, excessive operating temperature, etc.). A good example is provided
by DEP liquid actuation. Early experiments used electrode structures with millimeter
dimensions and voltages exceeding 10 kV [17]. In these structures, actuation of even
deionized (DI) water would have been impossible due to excessive Joule heating. However,
when the electrode structure size is reduced by a factor of 20, i.e., α = 1/20, the scaling rules
for both static and dynamic DEP actuation, Eqs. (11) and (14), predict that temperature rise
∆T is reduced by a factor of 400! This prediction is borne out convincingly by experiments
reported with DI water in 10 to 100 micron structures [7]. Furthermore, consistent with the
scaling prediction for voltage V (i.e., γ=1/20), the actuation voltage drops down to 200 to
500 V. Fig. 7 shows another example: an electrostatic droplet transport system that would
never work with electrodes on the scale of millimeters. Other examples of how reducing
electrode structure size favors electrostatic forces abound and, as a result, research laboratories
world-wide are vigorously investigating electric-field-mediated microfluidic systems for
the laboratory on a chip.
Acknowledgment
W.S.Trimmer offered the author encouragement to investigate the scaling laws for µTAS
technology. E.Cummings, P.Gascoyne, C-J.Kim, T.Schnelle, and M.Washizu graciously
provided videos showing examples of electric-field-mediated particle manipulation and
microfluidic systems. The author gratefully acknowledges the Japan Society for the
Promotion of Science, the US National Science Foundation, the US National Institutes of
Health, the Center for Future Health (University of Rochester), and the Infotonics
Technology Center, Inc., for their support.
Appendix
Joule heating is a serious problem in many electric-field-mediated microfluidic systems. It
can lead to unacceptable increases in temperature and therefore must be considered in
microsystem scaling studies. The relationship relating temperature T to Joule heating is:
(A1)
where km and σm are thermal and electrical conductivities, respectively, ρm is mass density,
and cm is specific heat. For the worst-case analysis needed here, the time-dependent term
(∂T/∂t) may be ignored, because the maximum temperature rise occurs in steady-state
conditions. Let χ be the scale factor for temperature. Then, assuming that overall structure
dimensions scale with the electrodes, the scaling operation performed on Eq. (A1) gives
[∆T]=χ=γ 2. This result is used directly to establish the scaling laws for ∆T in Eqs. (5), (7),
(9), (11) and (14).

10
References
[1] See for example the proceedings of the 2002 MicroTAS Conference: Micro Total Analysis
Systems 2002, vols. 1 & 2, (Dordrecht, NL: Kluver).
[2] Harrison DJ, Fluri K, Fan Z, and Seiler K 1995, in Micro Total Analysis Systems 1995, A.Van
Den Berg and P.Bergveld, eds. (Dordrecht, NL: Kluver), 105–115; Kataoka DE and Troian SM
1999 Nature 402, 794–797.
[3] Gau H, S.Herminghaus S, Lenz P, and Lipowsky R 1999 Science 283, 46–49.
[4] Duffy DC, Schueller OJ, Brittain OJA, and Whitesides GM 1999 J. Micromech. Microeng. 9,
211–217.
[5] McBride SE, Moroney RM, and Chiang W 1998, in Micro Total Analysis Systems 98,
D.J.Harrison and A.Van Den Berg, eds. (Dordrecht, NL: Kluver) 45–48.
[6] Sondag-Huethorst JAM and Fokkink LGL 1994 J. Electroanal. Chem. 367 49–57; Pollack MG,
Fair RB, Shenderov AD 2000 Applied Physics. Lett. 77, 1725–1726; Lee J, Moon H, Fowler J,
Schoellhammer T, and Kim C-J 2002 Sensors Actuators A 95, 259–268.
[7] Jones TB, Gunji M, Washizu M, and Feldman MJ 2001 J. Applied Physics 89, 1441–1448; Jones
TB 2001 J. Electrostatics 51–52, 290–299.
[8] Fuhr G and Shirley SG 1995 J. Micromech. Microeng. 5, 77–85; Hughes MP and Morgan H
1998 J. Phys. D: Appl. Phys. 31, 2205–2210; Gascoyne PRC, Vykoukal J, Weinstein R, Gandini
A, and Sawn R, in Micro Total Analysis Systems 2002, vol. 1, (Dordrecht, NL: Kluver), 323–
325; Voldman J, Toner M, Gray ML, and Schmidt MA 2002 J. Electrostatics 57, 69–90; Sano
H, Kabata H, Kurosawa O, and Washizu M 2002, in Technical Digest (IEEE), 15th Int’l Conf. on
Microelectromechanical Systems, 11–14; Cummings EB and Singh AK 2000 Proc. SPIE Conf
on Micromachining and Microfabdication 4177, 164–173.
[9] Feynman R 1965 The Character of Physical Law (Cambridge, MA (USA): MIT Press) 95–96.
[10] Trimmer WSN 1989 Sensors Actuators 19, 267–287.
[11] Pohl HA 1951 J. Applied Phys. 22, 869–871.
[12] Washizu M and Jones TB 1994 J. Electrostatics 33, 187–198.
[13] Levich VG 1962 Physicochemical Hydrodynamics (Englewood Cliffs, NJ (USA): Prentice-
Hall).
[14] Melcher JR 1963 Field-Coupled Surface Waves (Cambridge, MA (USA): MIT Press).
[15] Melcher JR and Taylor GI 1969 Annual Rev. Fluid Mech. 1, 111–146.
[16] Jones TB 2002 Langmuir 18, 4437–4443.
[17] Jones TB, Perry MP, and Melcher JR 1971 Science 174, 1232–1233; Jones TB and Melcher JR
1973 Phys. Fluids 16, 393–400.

11
Electrostatic ignition hazards—occurrence, detection

and prevention
U von Pidoll
Physikalisch-Technische Bundesanstalt (PTB), Braunschweig, Germany
Abstract. This paper is divided into three parts: The first part gives a summary of some recent
fires and their probable creation by electrostatic discharges. The accidents refer to e.g. car fires
during refuelling, sparking caused by hair spray bottles, and fires on flexo printing machines.
The second part describes the testing of complex products for use in explosive atmospheres.
One possible method works with flammable test gases escaping from an earthed electrode of
an ignition probe. A second method, described in detail, is based on measuring the transferred
charge of provoked single and multiple discharges to a test electrode and compares it with the
published threshold value for a particular explosive atmosphere. Both methods yield the same
results if the same discharges are evaluated. The third part deals with new antistatic plastics
surfaces. The materials and methods for producing them are summarized. The antistatic
effectiveness of these materials has to be ensured in quality assurance.
1. Introduction
In recent years electrostatic hazards have become an ever more important ignition source.
The reason for this is quite simple: plastic materials are increasingly used instead of metal
and wood. While wood and metal cannot be charged by rubbing, this is not the case for
insulating plastic materials. As a consequence electrostatic hazards that did not occur in
the past may result with new products made of plastic.
To make a comprehensive statement on present-day knowledge this paper is divided
into three parts. In the first section a compilation of the accidents caused by static electricity
and investigated by the PTB in the last two years is given. As the products for use in
explosive atmospheres have become more complex, the second part of this paper will show
how these complex products may be tested. The third part of this paper will summarize new
ideas, how to avoid electrostatic hazards at minimum cost.
2. Some fires caused by static electricity

2.1. Fires during refuelling of motor cars
Since 1992, series of fires have regularly occurred during refuelling of motor cars which are
attributed to static electricity [1]. The most prominent example lately was the release of a
new minicar in England. 500 new cars had already been delivered when the manufacturer
stopped the production on 3rd September 2001 after two fires had occurred during refuelling
of the new model. According to the German newspaper “Handelsblatt” [2], it was found that

12
the conducting but insulated fuel hose inside the car gets so highly charged by the flowing
fuel that sparkover in the region of explosive fuel vapours occured. The manufacturer
called the cars back and installed an earth cable between fuel hose and car body. With that
modification the problem was solved.
In the last few years only a few fires during refuelling have occurred in Germany. However,
these incidents were almost exclusively produced by personal charging of women. The
German magazine “Der Spiegel” [3], therefore, produced the headline: “Frauen tanken
riskanter” (Women refuel more riskily). According to [3] the reason for this phenomenon is
quite simple: Women usually do not like the smell of gasoline and often are more sensitive
to cold than men. For this reason they frequently leave the filling nozzle during refuelling
to go back inside the car. They then get charged when leaving the car seat and may produce
a spark in the vicinity of escaping fuel vapours when touching the earthed filling nozzle.
From April 2001 to June 2002 eight fires occurred at Japanese filling stations when the
driver opened the tank cap without having touched the refuelling nozzle or any other
earthed part. In every case the relative humidity was less than 50%, and some drivers
reported an electric shock when the fire broke out. For this reason electrostatic discharge is
the most probable ignition source. As some of the accidents were recorded on video these
pictures were later shown on television and in newspapers resulting in a disturbance of the
Japanese population. In the sequel all Japanese car manufacturers decided in June 2002 to
use earthed dissipative tank caps in all their cars as soon as possible [4].
Figure 1. Pictures of a recorded refuelling fire in Japan [4]
Nevertheless this measure would not help prevent the fires occurring when getting
inside the car after the start of the refuelling process. On 23rd September 2002 the American
Petroleum Institute and the Petroleum Equipment Institute of America therefore jointly
published a press release [5] urging all car manufacturers to use only earthed dissipative
tank caps in their cars and additionally print a precise warning in the users manual or close
to the filling cap that getting in the car interior during refuelling is dangerous and may lead
to fire.
As the flash point of diesel is slightly above 55°C, no ignition has been reported for
diesel tank systems up to now. However, modern diesel engines need a diesel return line to
the tank, and new diesel engines produce diesel reflux temperatures of 130°C and more.
There is a great chance that electrostatic discharges or sparks could be produced by electric
devices in diesel tank systems coincidence with hot diesel vapours, especially when the
tank is rather empty. This scenario may lead to the explosion of the fuel tank. For this
reason, some manufacturers only install diesel fuel pumps with built in flame arrestors and
additionally test that the tank system withstands an inner diesel explosion.

13
2.2. Fires caused by metal parts

A manufacturer of hair spray developed a new, flammable product with very promising
properties. When producing and selling the first samples of his product, he became aware
that sparkover occurred when the spray cans were filled or emptied. He immediately stopped
the production to find out the reason for this phenomenon. It was found that the metal cap
crimped to the spray dome was electrically insulated by a thick anticorrosive coating, with
a breakthrough voltage of some thousand Volts. As a consequence the cap got charged by
the small plastic hose charged by the flowing liquid spray material until sparkover to the
earthed dome of the can occurred. The material of the cap was replaced by another metal so
that no anticorrosive coating was necessary and the sparkover disappeared immediately.
Another fire happened with a big plastic container for flammable solvents. The container
was wrapped in an earthed metal grid to suppress brush discharges. The metal grid by all
appearances consisted of welded metal wires. One wire was earthed by an earth cable
attached to it and the area enclosed by a single mesh was clearly less than 100 cm2 as
demanded in [6]. For that reason the fire appeared to be unexplainable at first glance.
Further investigations showed, however, that only the wire with the earth cable attached to
it was made of metal whereas the other wires of the grid were made of welded plastic wires
provided with a metal-like paint so that they practically looked like metal. This arrangement
led to incendive brush discharges from the surface.
Figure 2. Plastic container for flammable liquids wrapped in an earthed metal grid which turned out to
be of insulating plastic material
2.3. Fires occurring at flexo printing machines

In 2001 some fires broke out on flexo printing machines with sleeves (the rotating cylinder
to which the printing block is fastened) of insulating plastic material of 20 cm or more in
thickness. The fires occurred after the ink flow to the printing block had been stopped at the
end of the printing task in order to clean the printing blocks of colour. In some accidents the
printing block was fastened to the sleeve with metal adhesive tape which had no contact to
earth. The ink used was alcohol based and turned out to be conductive. It was found by
experiment that a strong electric field was built up on the printing block after stopping the
ink flow because the paper rubs on the printing block and no more charge dissipation by
the earthed ink occurred.
The results obtained lead to the advice not to use metal adhesive tape for insulating
plastic sleeves, to replace insulating plastic sleeves by conductive sleeves and to reduce
the paper speed to 100 m/min during the cleaning process if insulating sleeves are used.

14
This speed reduces the risk of fire to a certain degree which might be acceptable for a
limited transition period before a replacement by conductive sleeves would take place.
Figure 3. Video sequence of a fire on flexo printing machines using insulating plastic sleeves
3. Testing products for electrostatic safety

In the past, testing of products intended for use in explosive atmosphere concerning their
electrostatic safety was restricted to few requirements: all conductive parts had to be earthed,
and all plastic materials either had to be dissipative or should not exceed a particular
maximum area. However, in the last few years special plastic materials have been developed,
e.g. materials that self-discharge by corona, and which cannot be evaluated according to
the scheme above. Also products with high-voltage electrodes for use in explosive
atmosphere with an automatic high-voltage switch-off actuated when an earthed electrode
approaches have appeared on the market. Accordingly, new direct test methods for these
products had to be developed.
One possible method uses flammable test gases escaping from a spherical earthed metal
test probe. Discharges to the test electrode are provoked under worst case conditions (i.e. in
very dry climate), and the test is passed if no ignition of the test gas has taken place. Care
has to be taken to ensure that the test gas has a somewhat lower minimum ignition energy
than the atmosphere for which the product shall be used to compensate for turbulence
effects in the escaping gas. This evaluation method yields satisfying results with experienced
test persons, but the result is only a simple yes/no statement. For this reason a determination
of the incendivity of a discharge from the recorded electric parameters seems to be an even
more promising alternative test method.
According to the literature [7–9], the product of capacitance C and applied Voltage U
seems to predict under certain circumstances whether or not a discharge from that capacitor
is incendive. This finding was confirmed by a number of ignition experiments in quiescent
atmospheres [10]. An explanation for this behaviour can be given as follows: If the discharge
gap for one minimum ignition volume of an explosive atmosphere is enlarged by a factor of
n, the discharge energy W=0.5CU2 is distributed over n minimum ignition volumes, so
only about the nth fraction of the energy is effective for ignition [11].
Minimum ignition energy for one minimum ignition volume=0.5CU2/n

15
On the other hand, according to Paschen’s law [12], the breakthrough voltage U is nearly
proportional to n if the electrodes used are sufficiently large in diameter (≥25 mm, see
[10,12]). If such electrodes are used, “n” may be substituted by “const U” in the equation
above, and one term of U gets reduced. As a consequence CU=Q=∫ldt becomes the dominant
product for the incendivity of a discharge (Table 1). Care has only to be taken to comply
with the validity of this equation, but as you may set the test conditions for a product as you
like, this is not a problem.
Table 1: Minimum necessary energy Wmin and charge Qmin to get ignition of quiescent fuel/air
mixtures for sparks of different electrode gaps
Tables of Qmin have been published for many substances [10], the measurement equipment
is commercially available (e.g. [13]) and the method including limits for a maximum
permissible Q has been published in European regulations [6,14] (Table 2). As Qmin is the
theoretical value of a capacitively stored charge and the capacitor is usually not completely
discharged, the Qmin values must be rounded down to get the maximum allowed values Qmax
for the experimentally measured Q which is the charge transferred in reality. It was found by
experiment that there is no difference between the incendivity limits for transferred charges
of brush and spark discharges, at least not in the case of gases and vapours.
Table 2: Maximum allowed values for the transferred charges Q of a provoked discharge under
worst case conditions [6, 14]
The validity of the method for brush discharges can be demonstrated as follows: If a
circular insulated plastic surface is charged to the highest possible extend, the incendivity
of a brush discharge from that surface depends only on the size of the area charged. It is well
known from experiments with flammable atmosphere too that if an earthed metal frame

16
forms the boundary of a charged area, this area must be four times greater than without the
frame if it is to have to have the same incendivity. Both facts can be verified by transferred
charges (Fig. 4).
Figure 4. Maximum transferred charge Qmax as a function of a PE plastics surface area in dry
climate
The following limitations of the method must be kept in mind:

• The length of the discharge gap has to be at least be 2 mm for explosion group IIA
according to EN 50014 [15] (e.g. normal organic solvents), 1 mm for explosion group
IIB (e.g. ether, peroxides, epoxides, ethene) and 0.5 mm for IIC (e.g. ethine, hydrogen,
carbon disulphide), otherwise quenching effects will lead to lower discharge incendivity
than expected from the transferred charge Q.
• If single discharges occur, e.g. from insulated plastics surfaces or insulated metal parts
charged by induction, a discharge ball electrode 25 mm in diameter connected to the
input of a coulomb meter may be used. After the discharge has taken place, the ball
electrode has to be moved off the remaining electric field before the display is read.
• If multiple discharges occur, e.g. from high voltage electrodes or textiles with conductive
threads, a rapid oscilloscope (at least 1 Gigasamples/s, 50 input, integrated integration
function) connected to a discharge ball electrode 25 mm in diameter with built-in
shunt (about 0.2 Ω, bandwidth 500 MHz or more) has to be used. The earth cable has to
be directly connected to the shunt to establish good connection between shunt and
earth and be free from sharp bends, e.g. knots.
• The method is not suitable for arc discharges because of energy cumulation of the half
waves.
• Up to now the method has not been tested with discharges from dissipative materials.
• Since the test method is rather new, it cannot be excluded that further limitations will
be detected in the future.
As the method of transferred charges is a direct method, it can be applied to any product.
Moreover, if an oscilloscope is used for recording, the current course gives information
about the type of discharge (Fig. 5). More details about method, equipment, results,
reproducibility and application for testing plastics housings, electrostatic hand spraying
equipment and flexible intermediate bulk containers can be found elsewhere [10,16].

17
Figure 5. Oscillographic shots of different discharge currents (50 nS/division, RShunt=0.23Ω). Top left:
Typical silent corona discharge from a tip (-35 kV, 30 µA). The zero line is shifted towards negative
values and the integral function shows a continuous small negative energy flow. Top right: Typical brush
discharge from a synthetic fibre surface (Q=77 nC). Lower left: Typical spark discharge (Q=48 nC)
from a tip (1 pF at 60 kV via 180 MΩ charging resistor, corresponding to 48 kV eff.) of an electrostatic
hand-held spray gun. Lower right: Typical spark discharge (Q=480 nC) from a standard rolled plastic
capacitor (93.3 pF at 7 kV via 100 MΩ charging resistor). Air capacitors or inductance-free pulse
capacitors show much more damped oscillation and a much higher energy dissipation.
4. New antistatic materials

In the last few years new methods to avoid electrostatic charging of insulated surfaces have
been developed. They include
• structured plastic materials. The plastics surface is structured by grooves and hills so
that only a small area is discharged in a brush discharge. This method has already been
used in the past but has undergone a renaissance due to direct measurement methods to
verify its effectiveness.
• co-extruded plastic materials. Co-extrusion of a thin outer conductive layer and a
thick carrier layer yields a dissipative or conductive plastic material without any
compromise as regards its mechanical or chemical properties.
• plastic materials with embedded metal corona tips. These plastic surfaces will quickly
discharge themselves via their embedded corona tips.
• co-extruded plastics of a positively and a negatively chargeable material. If such a
material is charged by rubbing, a dangerous overall charge can barely be created.

18
• insulated plastic surfaces treated with electrons. Some plastic materials show a strong
reduction in their chargeability after intensive treatment with electrons.
Any of these methods has to be tested under worst case conditions either using a probe with
escaping explosive test gas or by the method of transferred charge. It has, however, also to
be verified by the manufacturer that the material produced has the same electrostatic
properties as the test sample submitted to the test house. The method of transferred charge
allows a simple test procedure which can easily be executed in a manufacturer’s quality
test lab.
5. Conclusions
In the last few years a rapid development in products for use in explosive atmospheres has
taken place. More products appear on the market which have specially treated insulated
surfaces and equipment is sold with high-voltage electrodes with a highvoltage switch-off
actuated when an earthed electrode approaches. These products have to be tested by a
direct test method, e.g. using a test probe with an escaping flammable test gas or by the
method of transferred charge. The method of transferred charge is more suitable for quality
assurance measures of the manufacturers.
However, safe products alone do not assure risk free work in hazardous areas. They have
to be used with care by trained workers too. We observed that this is not always the case and
that some improvements on this topic are necessary to reduce the number of fires and
explosions.
References
[1] von Pidoll U, Krämer H and Bothe H 1997 J. Electrostatics 40&41 523–528
[2] German newspaper Handelsblatt 2001709/04 p 14
[3] German magazine Der Spiegel 2002/30 p 129
[4] Press release Mitsubishi Tech-share Mitsubishi Motor Company June 2002
[5] Press release Stop static American Petroleum Institute and Petroleum Equipment Industry 2002/
09/23 www.api.org
[6] Draft European Report Cenelec prTR 50404:2003 Code of practice for the avoidance of hazards
due to static electricity (new version of Cenelec Report R044–001:1999 Guidance and
recommendations for the avoidance of hazards due to static electricity)
[7] Gibson N and Lloyd F C 1965 Brit. J. Appl. Phys. 1619–1631
[8] Eisfeld D 1984 Gluckauf-Forschungshefte 45/3 116–123
[9] Gibson N and Harper D J 1988 J. Electrostatics 21 27–36
[10] von Pidoll U, Brzostek E and Froechtenigt H R 2002 Determination of the incendivity of electrostatic
discharges without explosive gas mixtures Conference record of the 2002 IEEE Industry
Application Conference 13th—18th Oct. 2002 (Pittsburgh, IEEE)
[11] Lewis B and von Elbe G 1987 Combustion, Flames and Explosions of Gases (New York,
Academic Press)
[12] Gänger B 1953 Der elektrische Durchschlag von Gasen (Berlin, Springer Verlag)
[13] Schnier Elektrostatik, Robert-Bosch Strasse 60, 72810 Gomaringen, Germany.
[14] European Standard EN 13463–1:2001 Non-electrical equipment for potentially explosive
atmospheres—Part 1: Basic methods and requirements
[15] European Standard EN 50014:1997 +AL1999 +A2:1999 Electrical apparatus for potentially
explosive atmospheres—General requirements
[16] von Pidoll U 2002 Rapid determination of the incendivity of discharges from FIBC surfaces
without explosive atmosphere Conference record of the 3rd World Flexible Intermediate Bulk
Container 6th—7th Nov. 2002 (Antwerp, Millennium Conferences)

19
Ignition hazards associated with earthing and bonding of

charged conductive objects
Klaus Schwenzfeuer and Martin Glor

Swiss Institute for the Promotion of Safety & Security, CH-4002 Basle,
Switzerland
Abstract. Earthing a conductive and charged object will produce a spark discharge shortly
before the earth contact will be made. This discharge can be incendive for combustible gasses
and vapours. Effective and less effective measures to avoid the incendivity of this discharge
will be presented.
1. Introduction
Usually a conductive object like a metal drum was earthed before it was used. In this case it
is guaranteed the conductive object will never be charged by static electricity. But special
situations can occur, where it becomes necessary to earth a conductive object
without knowing if it was already charged. Such a situation can happen after an accident. It
is possible that an flammable atmosphere may be present too. In this case it is essential
that an earthing mechanism is known, which can guarantee a spark free contact and thus a
non incendive contact between the earthing clamp and the conductive and charged
object.
The situation described above may typically be present if e.g. the fire brigade has to
empty a tank truck which was involved in an accident and which is still partially filled with
gasoline. Among different fire brigades there are different opinions on how to avoid a spark
during the earthing procedures. Ideas such as “earthing” the truck with a cable not connected
to earth or earthing the truck via a resistance have been proposed in the past. In the following
the efficiency of such measures is evaluated and other measures are proposed.
2. Types of discharge
In the field of static electricity different types of discharges can be distinguished. Depending
on the type of discharge, different values of energy will be released. As a result, the incendivity
of the various types of discharges is different. For the cases described in this paper, the spark
discharge is of great importance. With special geometrical conditions a corona discharge
can become important too and even a brush discharge could occur.

20
2.1 Spark discharge

A spark discharge can occur between two conductors with large radius of curvature. Usually
one of the two conductors is earthed and the other one is charged and insulated from earth.
The discharge appears abrupt with a lightning between the two conductors.
During a spark discharge the whole charge is released. The discharge energy is equivalent
to the stored energy (W) of the charged conductor. With the knowledge of the capacitance
(C) and the potential (U) of the charged conductor, the energy can be calculated according
equation 1.
(1)
A spark discharge is in principle incendive for all combustible gas/air, vapour/air and dust/
air mixtures. In practice, the released energy can amount to approximately 1 J in an
unfavourable situation. Spark discharges can be totally neglected, if all conductive objects
are earthed.
2.2 Brush discharge

A brush discharge can occur when an earthed conductor with a large radius of curvature is
exposed to an electrostatic field of high field strength. The origin of this strong electrostatic
field is usually a highly charged nonconductive surface. The brush discharge consists of
several short lightning flashes radiated from the place with the highest electrostatic field
strength.
The released energy of a brush discharge can hardly be estimated. The experience of
praxis shows, that only flammable atmospheres with a minimum ignition energy of less
than 1 mJ will be ignited. Brush discharges are in principle incendive for most combustible
gas/air and vapour/air mixtures. A brush discharge can be neglected, if high electrostatic
field strengths are avoided.
2.3 Corona discharge

A corona discharge can occur when a conductor with a very small radius of curvature is
exposed to an electrostatic field of high field strength. It is unimportant whether the
conductor is earthed or not. A weak continuous discharge will appear which releases only
a very small amount of energy. Therefore a corona discharge is not incendive for most
combustible gas/air or vapour/air mixtures.
3. Spark discharges during earthing

A spark discharge always occur between two conductive electrodes. Each electrode can be
treated as a capacitor with the capacitance (C). If the capacitor holds the charge (Q), the
potential (U) is reached, according to equation 2.
(2)
Again the stored energy (W) can be calculated with equation 1.

During a spark discharge both electrodes are shorted with a conductive plasma channel.

21
Part of the charge (Q) moves, until both electrodes will reach the same potential. As soon as
the charge flow stops, the plasma channel disappears and both electrodes will remain at the
same potential (U’). The stored energy (W’) of the electrode can be calculated with
equation 3.
(3)
The difference in energy before the discharge (W) and after the discharge (W’) corresponds
to the released energy (∆W).
(4)
3.1 The standard earthing

Normally one electrode is charged and the other one is earthed. In this case the capacitance
of the earthed electrode (CE) is infinite. The whole amount of charge (Q) will flow to earth
and the remaining potential (U’) is zero. The released energy (∆W) is equal to the stored
energy (W).
3.2 The open earth cable

A special situation exists if the earth cable is first connected to the charged object and
afterwards to earth. In this case both electrodes have a finite capacitance. It is assumed that
the charged object has the capacitance (C) and the earth cable, a capacitance (CE). During
a spark discharge between both electrodes, a current will flow through the plasma channel
until both electrodes remain at the same potential (U’)
(5)
With equation 5 a term for the remaining charge (Q’) of the original charged electrode can
be developed.
(6)
In combination with equation 2 this leads to an expression for the final potential (U’) of
the original charged electrode, which is the same as the potential of the unconnected earth
cable after the spark discharge.
(7)
Using equation 4 together with 7 leads to an expression for the transferred energy, which
is drawn within figure 1.
(8)

22
Figure 1:
Relationship between the transferred energy
W and the ratio CE/C of the capacitances
corresponding to equation 8.
This relationship can be discussed for three different cases:
1: The capacitance (CE) is much higher than (C), then the potential (U’ becomes nearly zero
and the transferred energy (∆W) correspond to the totally stored energy (W). This case is
equivalent to the usual situation, where the earth cable is connected to earth.
2: The capacitance (CE) is equal to (C), then the potential (U’) becomes the half of the
original potential (U) and the transferred energy (W) becomes three quarter of the original
stored energy (W).
3: The capacitance (CE) is much lower than (C), then the potential (U’) becomes nearly (U)
and the transferred energy (∆W) becomes nearly zero. This is the situation where an
incendive discharge stays away, however the charged electrode remains charged.
4. Laboratory experiments
To test the incendivity of spark discharges, where a resistance was placed between the non
charged electrode and earth, a spark gap was placed inside a glass tube. The diameters of the
spherical electrodes were 8 mm and the gap 3.7 mm. The volume of the tube was 275 ml. It
was filled with a diethylether/air mixture with a minimum ignition energy of approximately
0.2 mJ. One electrode was charged by brush discharges from a polyethylene stick up to a
potential of 15 kV. The opposite electrode was earthed via a variable resistor. In table 1 the
results were summarized.
Additional tests were performed with the intention to simulate a more realistic situation.
A conductive sphere with a diameter of 60 mm was placed inside a gas flow of a propane/air
mixture respectively an ethylene/air mixture. The conductive sphere was connected to a
high voltage device and an additional capacitor, in order to simulate a charged small
container (50 pF), a 200 Litre tub or a man (200 pF) or a large van (1000 pF).
The high voltage device was disconnected from the conductive sphere after it was
charged to a fixed potential. The charged conductive sphere was then discharged with a
commercial available earth cable (table 2), an earth cable with a special prepared clamp
(including a 1 MΩ resistor, see table 3), or a non conductive glove (table 4). The potential
was increased, starting from 1 kV up to the value where the propane/air mixture or the
ethylene/air mixture was ignited during the earthing process.

23
Table 1: Tests with the ignition chamber
Table 2: Tests with a usual earth cable
Table 3: Tests with an earth clamp including 1 M resistor
Table 4: Tests with a glove
5. Realised and possible solutions

5.1 Open earth cable
As demonstrated in Figure 1, the method of using an “open earth cable” is not reliable,
since such cables will have in most cases a capacitance comparable to the capacitance of
the charged object (truck). The capacitance of the cable also depends on its length and in
practice this has to be large enough to bring the real earth contact out of the hazardous area
generated by the leaking truck.
5.2 Earth cable with high resistance

There already exist earth cables which include inside the clamp a resistor of approximately
1 MΩ. But it is wrong to assume that the user will be protected against incendive discharges
during earthing of a charged object. The experiment and the theoretical background
demonstrate clearly that a resistance of 1 MΩ is not able to prevent incendive sparks.
5.3 Brush discharges during earthing

Another solution could be to touch the charged object first by hand, covered with an glove
and then with an earth cable. In this case the existing charge is carried away by the human
body, before the earth cable will contact the conductive object. If the glove is conductive,
again a spark discharge to an electrode (the human body) connected to earth via a resistance
will occur. This corresponds to the unsatisfactory solution described in section 5.2. If the
glove is non conductive, only brush discharges can occur between the charged object and
the hand.
The behaviour of a brush discharge differs from the behaviour of a spark discharge.
Particularly with a brush discharge, the charged object will not lose the total amount of

24
charge. In addition the energy of a brush discharge may already be incendive for combustible
gas/air and vapour/air mixtures. Therefore this is not a safe solution for discharging an
already charged object.
5.4 Corona discharges during earthing

The considerations above lead to the conclusion, that in all cases a discharge will occur
when an earth cable approaches a charged conductive object. In the presence of a flammable
atmosphere, it is important that the released energy is below the minimum ignition energy.
There exists one very weak discharge type, the corona discharge. It might be incendive for
hydrogen with an optimum concentration, but it is not incentive for all solvent vapours and
most gasses.
It is quite possible to construct a clamp for an earth cable, which includes some conductive
and sharp bristles at the front of the clamp. Each time the clamp approaches a charged
object, a corona current will flow just before the clamp can contact the object. An incendive
discharge will be prevented with a very high probability.
6. Conclusions
An incendive spark discharge cannot be prevented by including a high resistance into the
electrical earth circuit. Experiments have shown that the resulting spark discharges are still
able to ignite solvent vapours despite the modified waveform. Even if the earth cable is first
disconnected from earth, an incendive discharge will appear. The released energy can still
be above the minimum ignition energy of most combustible gas/air and vapour/air mixtures.
Also if the discharge is a brush discharge, it can not be excluded that the released energy is
above the minimum ignition energy of the present flammable atmosphere. Only with the
constructional trick of a corona electrode at the top of the earth clamp, can this hazard be
avoided with a high probability.
Acknowledgements
This work was initiated and sponsored by “Kanton Bern, Oel- Chemiewehr”, Switzerland.

25
An Investigation of the Electrostatic Ignition Risks

Associated with Plastic Coated Metal
G P Ackroyd, S Puttick
Process Hazards Section, Syngenta Agrochemicals, Huddersfield, UK HD2 1FF
Abstract. It is often assumed that, where an electrically insulating plastic is present as a thin
coating on an earthed metal substrate (as is the case with lined pipework and vessels), incendive
brush discharges will not occur. In this situation, the critical thickness for determining whether
a coating is “thin” is generally taken to be 2 mm. However, the experimental work detailed in
this report shows that ignitions of a typical flammable gas/ vapour atmosphere were obtained
by brush discharges from a 0.7 mm thick PFA coating and 0.85 mm thick ECTFE coating. For
the samples examined, the limiting charge transfer for ignitions to occur was approximately
100–150 nC. This corresponded to surface potentials of ~8–10kV (equivalent to a surface
charge density of ~180–260 ˜C/m2).
1. Introduction
The ignition hazards associated with electrostatic discharges from electrically insulating
plastics are well known, and have been discussed in detail in several reports [1, 2, 3].
However, the majority of the work reported has concentrated on the electrostatic properties
of wholly plastic materials. In those situations, the only type of electrostatic discharge
encountered is the relatively low energy brush discharge. In contrast, much of the plastic
used in process industry is present as a lining or backing on pipework and vessels. As long
ago as 1967, work by Heidelberg [4] identified that the presence of a conducting metallic
layer behind an insulating plastic could give rise to a different type of electrostatic discharge.
These were referred to originally as Lichtenberg discharges, but are now more commonly
known as propagating brush discharges. The conditions required for the generation of such
discharges have been investigated in some detail [5, 6, 7].
The fact that two distinct types of discharge can occur is due to the effect that the metal
backing has on the electric field arising from the electrostatically charged plastic. This is
illustrated qualitatively in Figures 1, 2 and 3. With no metal present, only a relatively low
level of charge is required on the surface of the plastic to give rise to a comparatively strong
electric field (Figure 1). Consequently, only a small amount of energy is available to be
released when breakdown of the air occurs. When an earthed metal backing is added to the
system, some of the field generated by the charge on the plastic is drawn towards the metal,
with a consequent reduction in the external field (Figure 2). As a result, the plastic is able to
hold much more charge before the external field becomes sufficiently strong for a discharge
to occur (Figure 3). As a result of the increased quantity of charge present at that point, a
much more energetic discharge (the propagating brush) occurs.

26
Fig. 1. Idealised illustration of the electric Fig. 2. Idealised illustration of the electric
field arising from a uniformly charged fields arising from a charged plastic with a
plastic metal backing
Fig. 3. Idealised illustration of the electric Fig. 4. Idealised illustration of the electric
fields arising from a highly charged thin fields arising from a highly charged thick
plastic coating plastic coating
Previous investigations in this area [5, 6, 7] have tended to centre on the conditions
required for propagating brush discharges to occur, and have given little consideration to
brush discharges. In general, it has been found that thin layers (≤1 mm) readily give rise to
propagating brush discharges. As the plastic thickness increases, the effect of the earthed
metal is lessened (Figure 4) until, at ~8 mm, propagating brush discharges no longer occur
and only brush discharges are observed. Furthermore, it has often been assumed that “thin”
layers do not give rise to simple brush discharges, and that it is necessary to achieve the
high charge densities associated with propagating brush discharges before breakdown can
occur in such systems. A thickness of 2 mm is often quoted as the cut-off point below which
brush discharges are not expected to occur.
2. Experimental Methods
A test procedure was devised that involved charging samples of plastic coated metal to a
pre-determined surface potential, either by triboelectrification with a synthetic mohair
cloth or by spraying on charge from a high voltage supply via a corona discharge apparatus.
Both methods resulted in the charge on the sample being of negative polarity, with the
surface potential being determined by use of a field mill (Figure 5). A hemispherical brass
electrode (diameter 20 mm) connected to an electrostatic voltmeter was then brought towards
the charged sample in order to initiate a discharge. The surface potential was re-measured
following the discharge, and the charge remaining on the sample discharged to a second
voltmeter. The tests were repeated a minimum of 20 times at each level of surface potential.
Tests were also earned out to determine the incendivity of the discharges obtained. The
ignition tests carried out involved discharging the charged sample to an earthed brass
electrode (diameter 20 mm) located in a flammable atmosphere composed of simulated
towns gas (STG) and air (Figure 6). The flammable atmosphere had been calibrated
previously to have an ignition probability of 50% when a 0.2 mJ electrical spark (at a
voltage of ~9 kV, with a system capacitance of ~5 pF) was discharged to the central
electrode.

27
Fig. 5. The equipment used to measure the Fig. 6. The equipment used to determine the
surface potential of the samples incendivity of the discharges
From the measurements, the total amount of charge present on the sample prior to discharge
could be calculated by two methods. Firstly, the total charge can be taken to be simply the
sum of the charge transfer in the measured discharge and the charge removed after discharge.
Secondly, the charge density on the sample can be calculated from the surface potential, as
the field strength and surface charge density for a bi-directional field are related as follows:
σ=ε0ε1E1+ε0ε2E2 (1)
where σ=surface charge density (C/m2)

ε0=permittivity of free space (8.85×10–12 F/m)
εn=relative permittivity of the surroundings on side n (dimensionless)
En=field strength on side n (V/m)
Since the distance to earth through the coating (δ, <1 mm) is very much smaller than that to
the field mill (D, 10 cm), the field strength through the material (E1=V/δ) will be very much
greater than the field strength in the atmosphere (E2=V/D). Thus, provided the relative
permittivity of the coating is relatively low, equation (1) can be simplified to:
σ≈(ε0εV)/δ (2)
where ε=relative permittivity of the coating (dimensionless)

V=surface potential (V, as measured by the field mill)
δ=thickness of the coating (m)
In order to characterise the samples, measurements were also carried out to determine the
volume resistivity, dielectric constant and breakdown strength of the coatings. The volume
resistivities were determined by using the field mill to measure the charge decay rate from
the samples, as the range of the equipment available to the authors was insufficient to apply
the standard tests [8]. The dielectric constant was calculated by measuring the capacitance
of a system containing the sample between two parallel metal plates (CS), and dividing that
value by the calculated capacitance of the same system with a vacuum between the plates.
The breakdown potential was measured by placing an electrode on the top of the sample
and raising voltage applied to it until a sharp increase in the current flow was detected.

28
3. Samples
The samples examined consisted of 1 mm thick (nominal) coatings on 20 cm square mild
steel plates. The plates were coated on both sides, with an exposed stud on one edge to
allow an earthing cable to be connected. Testing showed that the PFA (perfluoroalkoxy)
coating was in fact 0.7mm thick, with a dielectric constant of approximately 1.8 (compared
to a value of ~2 quoted in [9]) and a volume resistivity of ~2×1017 Ωm. The breakdown
strength of the coating was determined as ~20 kV (ie ~29 kV/mm). The coating of ECTFE
(ethylenechloro-trifluoroemylene, commercially referred to as Halar) was 0.85 mm thick,
with a dielectric constant of approximately 2.5 and a volume resistivity of ~2×1017 Ωm.
4. Results
The results of the charge transfer tests are shown in Figures 7–10. As expected, increasing
the surface potential of the sample results in an increase in the amount of charge transferred
in the discharge. Above a certain level, the surface potential is sufficiently high for the
discharge type to change from brush to propagating brush, and the amount of charge
transferred increases dramatically. It was noted that, at low surface potentials, the charging
method did not have a major effect on the charge transferred. However, it did affect the
potential at which propagating brush discharges occurred.
When the samples were charged by contact charging, propagating brush discharges
were obtained from surface potentials of ≥12 kV. In contrast, when charging was by corona
discharge, no propagating brush discharges were observed at surface potentials of up to 20
kV. Some additional tests suggested that propagating brush discharges could occur when
the PFA sample was corona charged to ~25 kV, but they were difficult to reproduce as the
sample tended to break down before a discharge could be obtained. However, as this
investigation was concerned primarily with the ignition risks from brush discharges, a more
detailed examination of this issue was considered to be outside the scope of the project.
Figs. 7–10. The levels of charge transfer from the coatings with relation to surface potential

29
5. Discussion
With regard to wholly plastic materials (ie without an earthed metal substrate), the
relationship between the amount of charge transferred in a single brush discharge and the
risk of ignition has been investigated in several previous papers. Data presented by Gibson
[3] indicate that in such situations, charge transfers of 30–50 nC are required in order for
brush discharges to have a significant probability of igniting a typical flammable gas or
vapour atmosphere (ie one with a minimum ignition energy of ~0.2 mJ). The data presented
in Figures 7–10 show that charge transfers in excess of that level were readily obtained from
the metal backed PFA and ECTFE samples, even at relatively low surface potentials.
If the 30–50 nC charge transfer criterion for incendivity was equally applicable to
plastics with an earthed metal backing then the test results would indicate a high probability
of obtaining incendive discharges from the PFA and ECTFE coatings. However, the ignition
risk from such thin plastic coatings generally is considered to be rather low. This suggests
that either the criterion for ignition is different when a metal backing is present, or that the
probability of obtaining incendive brush discharges from metal backed plastics is relatively
high, contrary to the general view. Therefore, further tests were carried out to investigate
the igniting power of the discharges observed.
Table 1. Ignition frequencies at various surface potentials (propagating brush discharge frequency
included for comparison)
Figs. 11–14. Frequency distribution of charge transfers (with ignition frequency data)

30
These results show that brush discharges obtained from thin plastic coatings of ECTFE
and PFA on an earthed metal substrate can ignite typical flammable gas & vapour
atmospheres. The level of charge transfer needed to cause ignition is clearly higher than the
50 nC required from wholly plastic materials (ie those without an earthed backing).
Comparing the frequency distribution of the charge transfers with the ignition probabilities
(Figures 11–14) suggests that the minimum level of charge transfer required from the
coatings for ignition to occur is approximately 100–150 nC. One possible reason for the
increased charge transfer requirement is that the length of the brush discharges tended to be
relatively short (<5 mm in some cases), so there may have been some flame quenching
effect. However, in view of the limited number of tests carried out so far, further work is in
progress to confirm these figures.
For both the PFA and the ECTFE samples, the minimum level of charge transfer needed
to cause ignition was achieved at a surface potential of ~8–10 kV. Calculating the
corresponding charge densities on the surface of the samples by use of equation {2} yields
values of ~180–260 µC/m2. This is in good agreement with the charge densities calculated
by the summation of the charge transfer values during and after discharge (100–260 µC/
m2). Data presented by Britton [10] confirms that at these values of surface charge density
and charge transfer, ignitions will have been due to brush discharges, rather than propagating
brush or the proposed “transitional brush” discharges.
6. Conclusions
The results of the tests carried out show that it was possible to obtain incendive brush
discharges from a metal sample coated with 0.7 mm of PFA or 0.85 mm of ECTFE. The data
indicate that charge transfers of greater than approximately 100–150 nC were required for
the discharges to ignite typical flammable gas & vapour atmospheres (ie those with minimum
ignition energies of ~0.2 mJ). In the tests, this corresponded to a minimum surface potential
of 8–10 kV (equivalent to a surface charge density of 180–260 µC/m2). However, the data
presented in this report only represent two materials (PFA and ECTFE). Further samples
have been obtained, and the results on those materials will be reported in due course.
References
[1] Gibson N and Lloyd F, Incendivity of Discharges from Electrostatically Charged Plastics, Brit. J.
Appl. Phys., 16 (1965) 1619
[2] Glor M, Ignition of Gas/Air Mixtures by Discharges Between Electrostatically Charged Plastic
Surfaces and Metallic Electrodes, J. Electrostat, 10 (1981) 327–332
[3] Gibson N and Harper DJ, Parameters for Assessing Electrostatic Risk from Non-Conductors—
A Discussion, J. Electrostat., 21 (1988) 27–36
[4] Heidelberg E, Generation of Igniting Brush Discharges by Charged Layers on Earthed Conductors,
Static Electrification Conference 1967
[5] Blythe AR and Carr GE, Characteristics of Propagating Electrostatic Discharges on Dielectric
Films, J. Electrostat., 10 (1981) 321–326
[6] Maurer B et al, Hazards Associated with Propagating Brush Discharges on Flexible Intermediate
Bulk Containers, Compounds and Coated Materials, Inst. Phys. Conf. Ser. No. 85: Section 3
[7] Tolson P, High-Energy Discharges from Plastic Surfaces, J. Electrostat., 22 (1989) 1–10
[8] British Standard BS 2782 Part 2 (1991), Methods of Testing Plastics
[9] Encyclopaedia of Chemical Technology, Kirk Othmer, 4th Ed
[10] Britton L, Avoiding Static Ignition Hazards in Chemical Operations (1999), AIChemE

31
Flow Electrification in Transformers:

Sensor Prototype for Electrostatic Hazard
O Moreau1, T Paillat2, G Touchard2
1
EDF R&D, 1 Avenue du Général de Gaulle, 92141 Clamart France
2
LEA, UMR 6609 CNRS, Groupe Electrofluidodynamique, Université de Poitiers,
Téléport 2, Bd Marie et Pierre Curie, BP 30179, 86962 Futuroscope France
Abstract. As far as Static Electrification diagnosis in transformers is concerned, previous works

have shown that the commonly applied measurements (leakage currents, Electrostatic Charging
Tendency or ageing test for tan˜) were not completely reliable. As part of the research program of
Electricité de France, a prototype of a new sensor dedicated to characterization of the phenomenon,
when pressboard-oil insulation is present, has been developed with the University of Poitiers. To
simulate the pressboard-made channels inside transformer, this device consists in a small pressboard
with facilities for measuring reliable parameters related to the generation (accumulated charge)
and leakage (leakage current) phases of the phenomenon. As it was likely to be installed on a real
transformer, the sensor has been inserted, for its development, in a loop representing at a small-
scale the oil path in a transformer with “on line” conductivity and moisture content monitoring.
This complete setup has made it possible to investigate the parameters behaviour with regard to oil
temperature and velocity during transient and steady states with and without flow. A first electrical
model for the electro-kinetics of the sensor has been deduced from the analysis of the results. This
device seems a promising tool to predict electrostatic hazards and to derive criteria for Static
Electrification diagnosis in transformers.
1. Introduction
The phenomenon of electrostatic charges generation due to a dielectric liquid flow was
observed as far back as the end of the ninetieth century. Moreover, oil industry has been
aware of flow electrification since it had to cope with discharges occurring in industrial
plants. During the seventies in Japan and more recently in the USA electric manufacturers
and utilities began to suspect the static electrification to be responsible for failures in
power transformers after damage surveys revealed some evidences of electrical discharges
(electric tree paths, worm holes, presence of carbon…) on inner pressboards [1].
As soon as a liquid is in contact with a solid wall, the complex solid-liquid initially
neutral becomes polarized under physico-chemical reactions occurring at the interface.
Such a phenomenon leads, on one hand, to a space charge in the oil which can relax in
contact with grounded metallic walls, and on the other hand, to a space charge in the
pressboard which can accumulate according to the leakage paths. As a matter of fact, the
higher leakage impedance the pressboards have, the more electrically stressed they are.
That means that the charge accumulation in the pressboard can only be limited by either a
possible chemical saturation of the interface, or by leakage and discharges along the interface
initiated by gradients of potential becoming too high versus the pressboard dielectric
strength. Although recent works have made the chemistry knowledge improved [1–2], the
complete process still remains partially unknown.

32
Previous studies have shown that the most commonly applied measurements for diagnosis
(leakage currents, Electrostatic Charging Tendency or ageing test for tan˜) are partially
connected to the accumulated charge which can be legitimately considered as relevant to
Static Electrification magnitude [3–4]. As part of a research program of Electricite de
France, a sensor prototype has been developed with Poitiers University to measure reliable
parameters related to the generation and leakage phases of the phenomenon, when
oilpressboard insulation is involved.
1. Sensor Prototype
1.1. Experimental Set-up
An experimental set-up (figure 1) has been built to simulate the oil path along the pressboard
between the windings inside a transformer. It consists in a stainless steel loop where oil
flows through the accumulation duct (4). The vessel (2) inserted in the loop constitutes a
relaxation zone for the remaining charges and may be used either for pressurisation with
gas or for oil recovering under vacuum. Then oil passes through a heat exchanger (3, 6) to
set the temperature and gets back to the pump (5), which can be operated at different flow
rates (1). Oil resistivity and moisture content are measured on line by respectively laboratory-
made conductivity cell and moisture measurement Panametrics.
Figure 1: Experimental set-up

1.2. The sensor
The sensor consists in a 30 cm-length channel of rectangular cross section (3×30 mm) made
of 3 mm thickness pressboard sheets stuck together (9) and inserted in a Teflon frame (10).
Two plane copper electrodes (240×50 mm) have been placed facing the largest external
surfaces of the pressboard duct beyond 2 mm PTFE for electrical insulation (11).
When those electrodes are grounded, it is then possible to measure a mainly capacitive
current (Icapacitive) (7a), related to the charge trapped inside the pressboard (image charge).
In the meantime, the surface potential rises and leads to charge leakage along the

33
oilpressboard interface toward the extremities of the duct. The resulting upstream and
downstream leakage currents can then be collected on the inlet and outlet stainless steel
elements which are insulated from the rest of the loop by PTFE flange coupling (8). All the
currents (7a, 7b and 7c) are measured by pico-ammeters Keithley 610C. This configuration
intends to represent the pressboard parts inside power transformers which have a high
leakage impedance with regard to imposed potential areas (windings and tank).
2.3. Operating process.

As flow electrification phenomenon is very sensitive to solid and liquid purity, a rigorous
operating process is systematically applied to avoid any pollution like moisture for instance.
The loop and pressboard duct are dried by nitrogen gas flow before being submitted to
vacuum for 24 hours. The filling of the equipment is then made also under vacuum by
direct transfer from commercial tanks. Finally oil flows in the loop for about 2 hours with a
bulk pressure on 0.2 bar to impregnate the pressboard duct. In addition, a 24 hours relaxation
period is always applied before starting the experiment campaign. For every oil/pressboard
couple, current measurements are carried out for a temperature cycle (20–10–20–40–60–
80–20°C) and at three laminar flow rates leading to the mean velocities of 40–68–95 cm/s.
This procedure enable us to simulate the different temperature and velocity conditions
encountered in an operating power transformer.
3. Experiment description and analyse

3.1. Experiment description
The oil-pressboard couples usually feature (except for some pathological cases) the following
characteristics:
– After a perturbed transient state due to the pressure wave at starting, the capacitive
current reaches a maximum before declining to zero whereas the leakage currents tend to
a steady state, all of them involving the same time constants: the generated charges
move along the interface toward grounded parts (figure 2).
– The generation process leads to a positive convected charge when the negative counter
charge accumulates inside the pressboard.
– The kinetics of the phenomena is principally influenced by the temperature conditions:
the transient state gets shorter as the temperature increases (figure 3).
– For a set temperature, the capacitive current magnitude increases with the flow velocity,
the kinetic remaining unchanged (figure 3).
Figure 2: Capacitive and leakage currents versus time

34
Figure 3: Influence of temperature and flow velocity
– Downstream current results from leakage negative charges (Idownstream leakage) and
positive convected charges (I close convection) (figure 4) and seems so greatly dependant
on velocity and temperature.
– Sometimes leakage currents feature discontinuities when the flow stops.
Figure 4: Currents related to charge displacements and electrical model
3.2. Derivation of the current source

The analysis of the capacitive and leakage currents makes it possible to assume two
hypothesis concerning the current chemical generator (figure 5):
1. Identical upstream and downstream leakage current tend to show that the pressboard
got uniformly charged with regard to its length. The difference which may appear
during the flow charging process, can be attributed to the “close convection”
contribution. The current generator can then be considered equal to twice the upstream
leakage current.
2. Differences between upstream and downstream leakage current during the discharge
could indicate that the pressboard did not get uniformly charged, making difficult to
estimate the “close convection” current and consequently the generator current during
the charging phase. Nevertheless, assuming that the leakage currents remain in the same
ratio during the charging and discharging phases, the downstream and close convection
current can then be deduced and provide an estimation of the generator current.
3.3. Accumulated charge

It appeared worthy to characterise an oil-pressboard couple with regard to static electrification
with only one relevant parameter instead of the two or three (magnitude, time constants)
required to describe the measured currents. The choice has turned to the accumulated

35
charge inside the pressboard which can be derived from the integration up to infinite time
of the capacitive current.
Figure 5: Derivation of the generator current
To get ride of disturbances which may occur during the transient state, the adopted approach
has been to interpolate the capacitive current with a smooth mathematical function which
will give, after integration up to infinite time, the total accumulated charge inside the
pressboard. The chosen function has been: I(t)=α(1-exp(-βt))exp(-t/τ) where the three
coefficients α, β and τ can be deduced from two points of the experimental curve (Imax, tmax)
and (I1,t1) with t1 chosen greater than tmax (figure 6).
Figure 6: Derivation of the accumulated charge
The accumulated charge for standard oil-pressboard couple features also some typical
behaviours (figure 7):
– The accumulated charge tends to grow as the flow velocity increases, this effect is all the
more noticeable that the temperature is high.
– Whatever the flow rate is, the accumulated charge passes a minimum around 40°C.
Figure 7: Influence of temperature and flow velocity on the accumulated charge

36
4. Electrical model
According to the presumed charge behaviour in the sensor duct, the proposed electrical
model (Figure 4) supposed that the physico-chemical process of separation of charges at
pressboard/oil interface can be considered as a chemical current generator which can be
expressed by [6]:
Igenerator=A0 (1–exp(–kt)) (1)
where t is time, k chemical kinetic constant and A0 the maximum number of negative
charges which may be created by the polarisation process.
Considering the initial and boundary conditions of the charging and discharging phases,
capacitive current and upstream leakage current can be written as:
Icapacitive=α(1–exp(–βt)) exp(–t/τ) (2)
(3)
(4, 5, 6, 7)
τ defines the electrical constant of the model and R is the equivalent resistance of the two
leakage resistances:
(8)
Whence the accumulated charge in the pressboard Q, computed as the integration of

capacitive current up to infinite time, is given by:
(9)
5. Conclusion
Considering that the probability of failures is closely correlated with charge accumulation,
this device seems a promising tool to predict electrostatic hazards in transformers. The
numerous tested oil-pressboard couples have enabled us the constitute an important
database including various generation and leakage behaviors. Inserted in a bypass on a
transformer, the sensor duct would be nearly submitted to the operating conditions
(temperature, moisture, aging). The virtual image of the most constrained inner pressboards
obtained in this way would make it possible to derive a criteria for on line Static
Electrification diagnosis.
6. References
[1] EPRI TR-105019, 1995
[2] G.Touchard, P.Mas, T Paillat, O.Moreau, 1999 Proc. ICDL 396–9
[3] T Paillat, G.Touchard, O.Moreau, 2000 Proc CEIDP 85–8
[4] T.Paillat , L.Onic, O.Moreau, Y.Bertrand, G.Morta, N.Charvet, G.Touchard 2001 Proc IEEE-
IAS Annual Meeting
[5] T.Paillat, N.Charvet, O.Moreau, G.Morta, Y.Bertrand, G.Touchard 2002 Proc CEIDP 180–3
[6] N.Emanuel, D.Knorre 1974 Edition De Moscou,

37
Explosibility of shredder dusts for electrical

appliances
M Nifuku, J Gatineau, C Barre, S Horiguchi, H Katoh, and M Hatori

National Institute of Advanced Industrial Science and Technology (AIST),
Onogawa 16–1, Tsukuba, Ibaraki 305–8569, Japan
Abstract. Paying attention to a fact that a large number (about 19.5 million units which weigh
about 720,000 tons) of major household electrical appliances (refrigerator, air conditioner,
television set and washing machine) are discarded in Japan and that they are regulated to be
recycled now, the authors investigated the explosibility of dusts produced in the recycling
process (mainly polyuremane, metal and plastic dusts). Minimum explosive concentration,
ignition temperature, ignition energy, influence of flammable gas on the dust explosion, Kst
value, etc. were investigated.
The results of the investigation show that the shredder dusts are sensitive to explosion
(minimum explosion concentration around 30–40 g/m3 and ignition temperature about 500–
550 °C). The minimum ignition energy is about 40 inJ. The existence of flammable gas
promotes the explosibility of dust cloud (explosibility increases to about two times). The Kst
value is about 70–75 bar-m/s. These indicate that the shredder dust can be easily ignited and
lead to dust explosion.
1. Introduction
We are using various types of and many electrical appliances in our lives. Accordingly, a
large number (about 19.5 million units which weigh about 720,000 tons, annually) of
major household electrical appliances (refrigerator, air conditioner, television set and
washing machine) are discarded in Japan [1] and they are regulated to be recycled now.
Other electrical appliances (computer, printer, copying machine, heater, etc) are also recycled.
Those electrical appliances are shredded in the recycling process and a lot of combustible
dusts are produced. The dusts are transported and an explosive dust cloud will be produced.
In the case of refrigerator, a lot of insulating material such as polyuremane is used. The
polyurethane material is inflated by cyclopentane which is flammable. When the flammable
gas exists together with the dust clouds, the gas promotes dust explosion.
These recycling processes produce sparks and heat by shredding, collision, etc. Also,
electrostatic discharge can be expected. These process and phenomena have potential
danger to trigger dust explosion. The dust explosion accident had been already reported in
Japan [2]. Similar accidents were also reported [3, 4].
The explosibility is not known well, because the dusts produced in the recycling process
are new in nature and they have not been investigated. Therefore, it is important to
know the explosibility of the dusts. Also, it is important to find ways to prevent dust
explosion.

38
Based upon the background mentioned above, the authors investigated the explosiobility
of dusts producd in the recycling process (mainly polyurethane, metal and plastic dusts).
Minimum explosive concentration, ignition temperature, ignition energy, influence of
flammable gas on the dust explosion, the maximum concentration of oxygen to inert the
dust cloud, Kst value and the maximum explosion pressure were investigated.
2. Experimental
The minimum explosive concentration was measured mainly by using Hartmann explosion
tube. The explosive concentration was calculated by the sample weight supplied and the
volume in the explosion tube (sample weight of the dust divided by the volume of tube).
We attempted to produce as homogeneous a dust cloud (concentration) as possible in the
bottom, central and upper part of the explosion tube. The minimum explosive concentration
is the dividing value between explosive and non-explosive concentration.
The tapping sieve apparatus [5] was also applied to confirm the results by the Hartmann
tube. The tapping sieve apparatus is applied in the Japanese Industrial Standard (JIS) (Z
8818:2002) to measure the minimum explosive concentration. The dust sample is supplied
into the ignition chamber from a sieve on the top. The sieve is tapped and shaken by a
tapping bar. Thus, the dust cloud concentration (dust distribution) in the explosion chamber
is highly improved. The dust concentration is measured by the amount of dust and the
volume in the explosion chamber (sample weight of the dust divided by the volume of
chamber). The minimum explosive concentration is the value between explosive and non-
explosive concentration, also.
The ignition temperature of the dust cloud was measured by the modified version of
Godbert-Greenwald furnace apparatus. The ignition energy was measured by observing the
waveform of the electrical spark in the Hartmann explosion tube. The ignition power
supply was charged up to 1,200 volts and the ignition spark was produced through discharge
resistor. The discharging electric current was regulated by choosing the value of the resistor.
The spark duration was arranged prior to the discharge by adjusting the timing of discharge
stop signal. The dust cloud was ignited under higher dust concentration (a hundred g/m3 or
more). The ignition energy was calculated by the measurements of the discharge voltage,
discharge current and discharging time of the spark waveform on the oscilloscope. The
waveform was acquired from the discharge electrode in the explosion chamber to the
oscilloscope. The minimum ignition energy is the minimum value required to ignite the
dust cloud in the Hartmann explosion tube.
The influence of flammable gas on the dust explosion was investigated by exchanging
the ambient air in the entire system of Hartmann explosion tube to the premixed gas with
specific cyclopentane concentration. The cyclopentane gas was diluted by air.
The maximum concentration of oxygen to inert the dust cloud was measured by
exchanging the entire system of the Hartmann explosion tube to the oxygen-nitrogen
mixture, with specific oxygen concentration.
The Kst value and the maximum explosion pressure were measured by a 20-litre explosion
apparatus [6].
The sample dusts used were polyurethane dust, plastic dust and shredder dust. Plastic
dust contains mainly ABS resin, polystyrene and polypropylene. The shredder dust is the
dust collected at the shredder and the major components are metal dusts (mainly steel, the

39
others are zinc, aluminium, copper, etc.), plastic dusts (ABS resin, polystyrene,
polypropylene, etc.) and polyurethane. Polyurethane is the major component of the shredder
dust. In the recycling process of the refrigerator, the amount of polyurethane dust is about
100 times or more than that of plastic dusts.
3. Results and discussion
3.1. Minimum explosive concentration
The explosion characteristics (relation between dust concentration and explosion

probability) are shown in Figure 1 through Figure 4. The minimum explosive concentrations
of the samples are approximately 30 g/m3 (polyurethane dust and plastic dust) and 40 g/m3
(shredder dust). The effect of particle size is not shown clearly, although the smaller the
particle size, the higher the sensitivity for dust explosion in general. This is because more
metal components are contained in the case of finer dust than the case of coarser dust
sample. Also, electrostatic agglomeration has disturbed the particle size effect. Figure 4
shows that the dusts collected at different processing plant have somewhat lower explosibility
than the ones shown Figure 1 Figure 3.
The minimum explosive concentrations by the tapping sieve apparatus were about 36 g/
m3 (polyurethane dust) and 28 g/m3 (plastic dust). This indicates that the results obtained
by the Hartmann explosion tube are almost identical to the ones by the tapping sieve
apparatus. Therefore, the authors’ data using Hartmann explosion tube can be regarded to
have high accuracy.
Figure 1. Explosion probability of polyurethane Figure 2. Explosion probability of plastic

dust based on particle size distribution. dust based on particle size distribution.

40
Figure 3. Explosion probability of shredder Figure 4. Explosion probability of the dusts

dust based on particle size distribution. collected at different locations (at a plant N).
In Figure 1–Figure 3, the finer dusts (under 63 µm) show a higher value for the minimum
explosive concentration or do not explode at all. The reasons have been mentioned above
in this section (metal components and electrostatic agglomeration).
The minimum explosive concentrations of the samples (about 30–40 g/m3) indicates
that the dusts produced in the recycling process are sensitive to dust explosion.
3.2. Ignition temperature

The ignition temperatures of the shredder dust cloud are shown in Figure 5 and Figure 6.
Although the results are slightly different depending on the recycling plant, the dust cloud
can be ignited at 500–550 degrees Celsius.
Since the shredding process produces heat due to impact, crushing, etc., the temperature
in the shredder can increase. Therefore, the dust in the shredder can be ignited. Monitoring
the temperature in the shredding process will help to reduce the risk of dust explosion.
3.3. Ignition energy

The minimum ignition energies of polyurethane dust, plastic dust and shredder dust were
70, 40 and 200 mJ respectively (the dust cloud concentration was 100 g/m3 for polyurethane
dust and plastic dust, and 150 g/m3 for shredder dust).
The ignition source for dust explosion is produced in the shredding process and the
ignition source will have various characteristics, such as life of heat source, amount of
energy, etc. Based on this, an influence of life of ignition source (i.e. spark duration) on the
ignition energy is shown in Figure 7. It indicates that dust cloud is rather hard to be ignite
at a spark duration of less than about 0.2 ms and that the minimum ignition energy for
polyurethane dust is about 70 mJ at a spark duration of more than about 0.2 ms.

41
Figure 5. Ignition temperature of shredder Figure 6. Ignition temperature of the dusts

dust sample. collected at different locations (at a plant N).
Figure 7. Influence of spark duration on ignition Figure 8. Influence of cyclopantane concen-

energy (sample: polyurethane dust, original tration in the ambient air on the explosibility of
sample, 100g/m3) (spark gap: 5mm). polyurethane dust (original sample).
Since there are almost always many sparks in the shredder (monitored by camera), and the
energy can be regarded as higher than these value, the dust cloud produced in the shredder
could be ignited, as reported [2, 3, 4].
3.4. Influence of flammable gas on the dust explosion

When dust exists together with flammable gas, each of them promotes explosion mutually.
The formation of combustible mixtures around the particle will be promoted and the
combustion of the combustible mixture will become accelerated. It is important to know

42
the characteristics to prevent hybrid explosions. The effect of cyclopentane on the explosion
probability is shown in Figure 8. The explosibility of polyurethane increases roughly with
increasing cyclopentane concentration. The result of plastic dust was similar to the case of
polyurethane. The explosibility of polyurethane dust with cyclopentane over 2000 ppm
seems to remain constant. In practice, the alarm level for cyclopentane is set at 1,500–2,000
ppm. This value is appropriate in regard to the results mentioned here.
3.5. Oxygen concentration to inert dust cloud

Although there are several ways to reduce the risk of dust explosion accidents, it is not
practical to regulate the dust cloud concentration to prevent the dust explosion. This is
because the minimum explosive concentration is usually far below the economical
operational conditions. We have to assume that the shredding process always produces
explosive dust cloud. Under this viewpoint, it will be a good idea to inert the dust cloud.
Figure 9 and Figure 10 show the fundamental characteristics to inert the dust cloud. Explosion
probability decreases with the decrease of oxygen concentration. Polyurethane dust cloud
will not be explosive when the oxygen concentration is below 12.5%.
3.6. Explosion severity

Kst value of polyurethane dust and plastic dust are shown in Figure 11. The values (about 75
and 70 bar-m/s) are not very high in terms of explosion severity. The maximum explosion
pressures of both polyurethane dust and plastic dust are about 4.7 bar. These values (maximum
explosion pressure) are somewhat low and similar to the ones of coal dust, cereal dusts,
etc. [7].
Figure 9. Influence of oxygen concentration Figure 10. Influence of oxygen concentration

on the explosibility of polyurethane dust on the explosibility of plastic dust (original
(original sample). sample).

43
Figure 11. Kst values of polyurethane dust and plastic dust.
4. Conclusions
The results of the investigation show that the shredder dusts are sensitive to explosion
(minimum explosion concentration around 30–40 g/m3 and ignition temperature about
500–550 °C). The minimum ignition energy is about 40 mJ (plastic dust). The existence of
flammable gas promotes the explosibility of dust cloud (explosibility increases to about
two times). Oxygen concentration of 12 % will inert the explosive environment. The Kst
value is about 70–75 bar-ms/s which is not too strong in explosion severity.
These indicate that the shredder dust can be easily ignited and lead to dust explosion.
However, an electrostatic discharge in the recycling process may be somewhat difficult to
trigger a dust explosion but flammable gas can be ignited which might lead to dust explosion.
Therefore, great attention (dust cloud concentration and flammable gas concentration
monitoring, applying venting devices, etc.) has to be paid in the recycling process.
Acknowledgements
The authors express their appreciation to Mitsubishi Electric Corporation for advice, support
and samples in the research. Also, the authors express their appreciation to Dr. H.Matsui
and Dr. M.Yashima, National Institute of Industrial Safety (Japan) for their cooperation to
measure the Kst value and the maximum explosion pressure. (This research was supported
partially by R & D fund of the Ministry of Hearth, Labour and Welfare of Japan.)
References
[1] Nagata K, Ueno K, Terasaki M and Iwata Y 1999 Kaden Risaikuru (Recycling of home
electrical Appliances) (Tokyo: Kougyo Chousa Kai) (in Japanese)
[2] The Mainichi 14 July 2002 p. 24 (in Japanese)
[3] Itgaki H 1998 J. Japan Society for Safety Engineering 37 178–184 (in Japanese)
[4] Shukan Asahi 13 October 2000 132–134 (in Japanese)
[5] Nifuku M, Matsuda T and Enomoto H 2000 J. Loss Prevention in the Process Industries 13
243–251
[6] ASTM E 1226–88 1988 Designation: E 1226–88
[7] Enomoto H (editor and author) 1991 Funjin Bakuhatsu -Kikensei to Bousisaku- (Dust Explosion
-Explosibility Assessment and Ways to Prevent the Explosion-) (Tokyo: Ohmsha Co. Ltd) (in
Japanese)

45
Microscale electric induction machines for power

applications
C Livermore*, A Forte†, T Lyszczarz†, S D Umans‡, and J H Lang‡

*
Department of Mechanical Engineering, †Lincoln Laboratory, and ‡Department
of Electrical Engineering and Computer Science, Massachusetts Institute of
Technology, Cambridge, MA 02139
Abstract. Microelectromechanical systems (MEMS) are being engineered to provide

lightweight, compact power sources for portable devices. This paper presents the development
of a millimeter-scale electric micromotor/generator that is designed to perform Watt-level
electrical-mechanical energy conversion. The device’s operating principle is electric induction;
power conversion is accomplished by the interaction of the voltage on an electric stator with
the image charges that it induces on a nearby spinning electric rotor. The micromotor/generator
is fabricated by silicon micromachining and thin film processing, and it measures just 15 mm
on a side and about 2.5 mm thick. To date, the system has been designed, fabricated, and tested
as a motor at partial actuating voltage, achieving a peak torque of 3.5 µNm and a peak air gap
power in excess of 20 mW, in agreement with theory.
1. Introduction
A team at MIT is developing a family of power MEMS devices for applications like
generating portable electric power, propelling microscale vehicles, and driving miniature
pumps or fans [1]. These devices include miniature gas turbine engines, gas turbine
generators, rocket engines, and electric motors. These devices are designed to operate at
high power densities, and this requires that they maintain high levels of other quantities,
like temperature, frequency, speed, voltage, or current. For example, the micro gas turbine
engine incorporates high speed turbomachinery that rotates at speeds in excess of one
million revolutions per minute and is designed to output tens of Watts of mechanical
power.
This paper focuses on micro-scale electric motors and generators that operate through
electric induction (as opposed to magnetic induction) and are designed for integration into
a complete, button-sized gas turbine generator system. The motor/generators described
here are designed to perform mechanical-electrical power conversion at the Watt level.
Operated as a generator, the present device with its 4 mm diameter rotor is predicted to
output 0.5 W of net electrical power at 15% mechanical to electrical efficiency. A slightly
larger but technologically similar electric induction generator with a rotor about 5 mm in
diameter has been designed to output about 3 W of net electrical power at higher efficiencies
of about 29%. As part of a complete hydrocarbon fuel-burning gas turbine generator system,
such devices have the potential to outperform electrochemical batteries on the critical
metric of energy output per unit weight. To function at these power levels, the devices are

46
designed to operate at high voltages and electric fields (about 300 V and 108 V/m), high
electrical frequencies (a few MHz), and high rotational speeds (about 1 Mrpm).
The potential performance of the electric induction mechanism in microsystems has
been described in the literature [2], and several electric induction devices have been designed,
fabricated, and tested within the microengine project. Torque metrology devices have
achieved torques from 0.2 µNm [3] to 2 µNm at partial actuating voltage. A rotating
micromotor-driven compression system demonstrated torque of 0.3 µNm and power of 0.5
mW at partial actuating voltage [4]. More recently, rotating micromotors have obtained
high levels of air-gap power conversion, more than 20 mW, when operated at partial
voltage [5].
2. Device design
The electric induction actuating principle is illustrated for the case of a motor in Figure 1.
The device has two fundamental components, a 4.2 mm diameter spinning rotor disk and a
fixed stator plate. The stator surface is covered by an array of radial electrodes, arranged
like spokes on a wheel, which are excited with ac voltages to create a traveling wave of
potential around the stator. Facing the stator is the spinning rotor, which is coated with a
thin film of a slightly conducting material. The stator and rotor are separated by a narrow,
3 µm air gap. The electric potential on the stator induces image charges in the rotor film.
During motoring operation the speed of the traveling stator excitation exceeds the rotor’s
rotational speed. As the stator excitation travels along, the image charges follow, conducting
through the rotor film. By design, the conductivity of the rotor film is low enough that the
rotor charges lag behind the stator excitation. This generates a tangential electric field that
pulls on the rotor charges to create a torque, thereby converting electric power into
mechanical power. During generator operation, an external source of mechanical power
spins the rotor at a speed that exceeds the traveling speed of the stator excitation. As a
result, the rotor charges lead the stator excitation rather than lagging behind it. The direction
of the tangential electric field is then reversed when compared to motoring operation,
slowing the rotor and sourcing electrical power into the stator.
Why choose electric induction machines over other types of motor/generators, such as
magnetic induction, variable reluctance, or variable capacitance machines? One reason is
scaling. From Paschen’s curve [6], it is known that a narrow gap can sustain a larger electric
field without breakdown (108 V/m for this geometry) than can a wider gap. The increased
power density makes micro-scale electric induction machines compare more favorably
Figure 1. Schematic diagram of motor operation. A traveling wave of potential on the stator electrodes
induces image charges in the rotor film; the resulting tangential electric field spins the rotor.

47
with magnetic machines than a macro-scale electric induction machine would. A second
advantage of electric induction machines when compared with magnetic machines is material
compatibility with the high temperatures encountered both during fabrication and inside
an integrated gas turbine generator. An electric induction machine requires low-resistance
conductors that are stable at high temperatures (600 C—800 C). A magnetic induction
machine requires magnetic materials with comparably high Curie temperatures, which
presents a greater challenge. A third advantage is minimal control requirements; in other
words, the stator excitation need not be synchronized exactly to the rotor motion. This is
true of all induction machines, and it offers an advantage over other possible types of
motor/generators, like variable capacitance machines. Since the rotor position does not
impact how the machine is operated, the rotor position need not be measured. This is
particularly valuable for these miniature, high power density applications. The same
combination of small size and high rotational speeds that enables high power density in the
microdevice also makes real time sensing and control of the rotor position more difficult.
A schematic cross-section of the motor/generator is shown in Figure 2. The device
comprises a stack of five silicon wafers; the rotor is encapsulated within the stack. In
operation, the rotor is supported axially and in-plane on gas bearings, thin films of pressurized
air that are injected into the narrow gaps between the rotating and stationary components.
Such bearings have previously been demonstrated to support rotation of a similar microdevice
at speeds of well over 1 Mrpm [7]. The rotor, bearings, and auxiliary flow paths are etched
in the silicon wafers. In addition, two of the wafers are patterned with thick and thin films to
create the motor/generator’s electric stator and electric rotor. The electric stator is a planar
array of 786 radial electrodes wired into six phases by a separate layer of interconnections
to form 131 pole pairs. The six phase design limits the voltage across the 4 µm gap between
adjoining phases to a maximum of 300 V when the stator electrodes are excited with a
sinusoid with an amplitude of 300 V. On the side of the rotor facing the stator, the spinning
rotor disk houses a thin layer of moderately boron-doped polysilicon on top of a layer of
oxide insulation; the polysilicon serves as the rotor film. The back side of the rotor houses
a set of turbine blades. The blades can be driven with compressed air to spin the rotor disk
for electrical power generation. Appropriately designed blades could also be driven by the
motor to act as a pump or fan.
Achieving high-power, low-loss performance depends on maximizing the power
conversion capacity and maximizing efficiency. Analytical system modeling was used to
examine the tradeoffs among operating point, conversion capacity, ease of fabrication, and
viscous and resistive losses in the device and external electronics. The results show that
minimizing stray capacitance between the electric components and the substrate is critical
both for power conversion capacity and to minimize electric losses, especially when
operating as an electric generator. System modeling also highlights the importance of
minimizing the resistance of the stator wiring in order to minimize electric losses during the
charging and discharging of the stator’s parasitic capacitance. Requiring low electric losses
creates significant challenges for the device fabrication; however, with higher losses the
motor/generator would not be able to generate net electric power. Finally, models show
that it is imperative to sustain high electric fields of about 108 V/m at high frequencies
between components separated by several micrometers without suffering electric breakdown.
Higher design frequencies enable higher power, and converted power increases as the
square of the operating voltage. However, this demands defect-free lithography and smooth
features on the electric stator; defects contribute to early failure from electric breakdown.

48
For a six-phase generator with the geometry described above, operating with a 300 Vpeak
sinusoidal excitation at 1.5 MHz, 3.2 W of mechanical power into the rotor would produce
0.5 W of electric power out with an overall efficiency of 15%. Viscous losses in the motor
gap account for 1.1 W of expected losses. Dissipation in the rotor film, which reflects the
inherent conversion efficiency of the device, accounts for another 0.5 W. Electric losses in
the stator wiring and external inductor (part of the resonant circuit that excites the generator)
account for another 0.1 and 1.0 W respectively, even in a device designed for minimum
stray capacitance and series resistance. The low efficiency reflects the small power conversion
capacity of this geometry relative to fixed external losses, like the charging and discharging
of external stray capacitance through the external inductors. A factor of two increase in
active area would offer higher overall efficiencies in the 30% range.
3. Fabrication
The motor/generator’s mechanical structures, such as the rotor, turbomachinery, and
bearings, are micromachined from single crystal silicon wafers using photolithography and
double-sided, aligned deep reactive ion etching (DRIE). The fabrication of the bearings
and mechanical structures have been demonstrated previously [7]. The motor/generator’s
electric features are defined on two of the five wafers using thick and thin film processing.
In the fabrication process as designed, the individual wafers are then laminated together
using silicon fusion bonding to form the complete device structure. Figure 3 shows a
photograph of dies from each of the five wafer levels of the micromotor/generator. Each die
is 15 mm on a side; only four dies are visible because two of the levels are already wafer-
bonded together.
The structure and operating point of the final device enable the motor/generator’s
potentially Watt-level output at the cost of a substantial set of fabrication challenges. To
minimize stray capacitance between the electrical structures and the substrate without
creating too much wafer bow, the stator electrodes and rotor film are fabricated on top of 10
µm to 20 µm thick isolated oxide islands that are embedded in the silicon substrate, as
shown in Figure 2. The island concept has been demonstrated in previous work [4, 8]; in
this case the stator islands are instead defined with a liftoff process. The upper surface of the
wafer is covered by an aluminum layer about 1 µm thick; the 20 µm deep etched regions are
Figure 2. Cross-sectional schematic diagram of Figure 3. Fabricated die levels of the micro-
motor showing its five constituent throughetched motor/generator. The electric stator is second
wafers. Also shown is the integration of the electric from the left, and both sides of the rotor are
stator and rotor into the stack. shown below.

49
left bare. A 20 µm thick layer of PECVD TEOS oxide is then deposited on the wafer. Liftoff
is accomplished by a prolonged etch in hydrochloric acid, which removes the aluminum
and dislodges the oxide from the upper wafer surface. The oxide embedded in the etched
regions remains, producing a nearly planar surface.
The stator electrodes and wiring must have low resistivity and enough thermal stability
to tolerate the high temperatures encountered in wafer bonding and engine operation, up to
about 700 C. As a compromise between these requirements, the stator electrodes and
interconnection rings are fabricated as two interconnected 0.3 µm thick platinum levels,
each deposited on top of a titanium adhesion layer. Each stator electrode is 900 µm long,
tapers from 11 µm to 4 µm, and is separated from its neighboring electrodes by just 4 µm.
The two platinum layers are fabricated by liftoff processes on top of the thick oxide islands
in the silicon substrate. The upper platinum layer forms the stator’s electrodes; a separate
underlying layer of platinum interconnection rings groups the electrodes into six phases.
The electrodes and interconnects are separated by a 1 µm thick layer of PECVD TEOS
oxide. Electrical contact between the two layers is established by via contacts etched in the
interlayer oxide. Images of the stator electric structures are shown in Figure 4.
Resistance to electric breakdown is impacted by the fabrication process. A potential
location for breakdown is the 4 µm gap between neighboring electrodes, which must
withstand 300 V at MHz frequencies when the six-phase machine is actuated at the full
design voltage of 300 V. A critical factor in avoiding breakdown between electrodes is the
smoothness of the electrode edges. Because the platinum electrodes are formed by a liftoff
process, they have smoother edges than would a dry-etched structure. The fabricated
structures have been experimentally shown to withstand electric breakdown up to the
design voltage. A sinusoidal potential difference with an amplitude of up to 300 V and a
frequency of 1.75 MHz was applied between neighboring electrode phases. While the
voltage was applied, the electrode structure was examined under an optical microscope to
identify signs of electrode damage. With defect-free lithography, the electrode structure
consistently resisted electric breakdown up to 300 V. However, it is important that the
structure be defect free; debris or other defects between the electrodes can initiate electric
breakdown, which typically halts after the breakdown eliminates the initial flaw.
The micromotor/generator devices are designed to use wafer-level silicon fusion bonding
for the final assembly, and this introduces another fabrication challenge. However, the
motor/generator described here was instead assembled at the die level by clamping the
Figure 4. A) SEM micrograph of stator cross-section, showing electrodes on a recessed oxide island. B)
Optical micrograph of the stator structure, showing electrodes connected by interconnection rings.

50
individual dies together in an aluminum/acrylic package. O-rings create sealed fluidic

connections between the outer ports of the device and the inner surface of the package;
they also provide compression to minimize leakage between the plates.
4. Experimental results
To date, the motor/generator has been operated as a motor. Its torque and power have been
characterized as a function of stator excitation voltage and stator excitation frequency, and
they have been observed to agree with theory. To operate the device, the hydrostatic
bearings are first pressurized to float the rotor on a film of air; the stator is then excited with
a traveling wave of electric potential. The primary observable is the rotor’s rotational
speed. Speed is measured by a fiber optic displacement sensor that monitors the passage of
features that are etched into the rotor. The output of the displacement sensor is converted to
rotational speed by a spectrum analyzer. At a fixed excitation frequency of 200 kHz, the
motor has spun at over 55,000 rpm at a partial actuating voltage of 95 Vpeak fundamental
amplitude.
Two metrics for the motor’s performance are torque and power, which are determined
from the motor’s measured speed and the measured viscous drag on the rotor disk. The
motor operates at no load; its speed is set by the balance between the motor’s actuating
torque and the viscous drag on the spinning rotor. The viscous torque is measured as a
function of rotational speed as described below. The motor’s torque and power are then
extracted from the measured rotational speed using the experimentally determined load
line.
The device’s load line is determined by abruptly removing the stator excitation and
monitoring the rotor’s deceleration with the optical displacement sensor. From these data
and the known rotor inertia, we can extract the decelerating viscous torque. This yields a
fixed viscous torque constant for the device in question of (5.6±1)×10"4 µN·m·s plus a
constant 0.2±0.01 µN-m component of the torque. The constant reflects mechanical rubbing
of a piece of debris in one of the thrust bearings; the debris was observed on subsequent
inspection.
The stator excitation frequency was varied from 50 kHz to 600 kHz while the stator
excitation voltage was set at a fixed value of 45 Vpeak fundamental amplitude. The rotor
speed was measured, and the motor torque and power were extracted as described above.
Figure 5 shows plots of measured torque and power versus electrical frequency. The solid
Figure 5. A) Measured motor torque and B) measured motor power as a function of electrical
excitation frequency at a fixed stator excitation voltage of 45 V.

51
lines are fits of a model [2, 3] to the data; the resistivity of the rotor film and the gap between
rotor and stator are adjustable parameters. The measured motor torque and power both peak
near 200 kHz. The peak reflects a balance between the time scale for rotor charge relaxation,
which is significantly affected by rotor film resistivity, and a time scale related to the
frequency of the stator excitation. At very low stator excitation frequency, rotor charges
relax more quickly than the stator excitation travels along the stator, so the electric field in
the gap has very little tangential component. At very high stator excitation frequency, the
situation reverses, and the rotor charges lag too far behind the stator excitation for effective
power conversion. The frequency of the peak reflects the resistivity of the doped polysilicon
rotor film and corresponds to a rotor film sheet resistance of 800 M /square. This is higher
than the design sheet resistance of 330 M /square. The difference between measured and
design sheet resistance most likely reflects the omission of the final high temperature wafer
bonding anneal, which promotes grain growth in the polysilicon rotor film and decreases
resistivity. The peak amplitudes of the torque and power reflect the size of the gap between
the rotor and the stator; a smaller gap corresponds to a higher electric field and more power
conversion across the gap. The data are consistent with a motor gap of 1.8 µm, as compared
with a design gap of 3 µm. This is consistent with the actual geometry of the device as
fabricated, and it is consistent with the measured viscous torque. With these values for the
motor gap and the rotor film resistivity, the measured torque and power agree well with the
model values.
The stator excitation was then set at a fixed frequency of 200 kHz while the stator
excitation voltage was varied up to 95 Vpeak fundamental amplitude. The resulting torque
and power are plotted in Figure 6. Markers indicate measured data, and the solid lines are
calculations from the model, using the fit parameters from above. At low voltage, the torque
increases quadratically with voltage, as expected. At higher voltages, there is an inflection
point in torque vs. voltage, as the rotational speed approaches the synchronous speed of the
stator excitation. The torque reaches a maximum of 3.5 µNm at a speed of over 55,000 rpm.
The maximum power converted across the air gap exceeds 20 mW, corresponding to a
power density of 16 kW/m3. The measured torque and power again agree well with the
model.
The maximum actuating voltage in these tests was not limited by electric breakdown,
but rather by the high rotational speeds that are necessary to balance a large actuating
torque against viscous drag under no-load operating conditions. Tests at higher rotational
speeds are planned to further measure the performance of the electromechanical device and
Figure 6. A) Measured motor torque and B) measured motor power as a function of stator excitation
voltage at a fixed stator excitation frequency of 200 kHz.

52
the system as a whole. Given the stator’s demonstrated ability to withstand design electric
fields of 108 V/m and the excellent agreement between measured and expected performance
at partial actuating voltage, the motor/generator is expected to reach the Watt-level power
conversion for which it was designed.
5. Conclusions
Millimeter-scale electric induction motors and generators have the potential to perform
Watt-level electrical-mechanical power conversion for applications ranging from portable
electrical power supplies to miniature pumps or blowers. The high power potential of these
devices is enabled by the high electric fields that can be maintained across the microdevice’s
small rotor/stator gap without electric breakdown. A micromotor/generator was designed,
fabricated, and tested, demonstrating an air gap power in excess of 20 mW and torque of 3.5
µNm at speeds in excess of 55,000 rpm when operated as a motor. This performance was
enabled in part by the development of low-loss, high-voltage electric stators, which have
been demonstrated to successfully sustain design voltages.
Acknowledgements
We gratefully acknowledge the contributions of A.Ayon, A.Epstein, L.Frechette, A. Hoelke,
P.Maki, S.Nagle, S.Senturia, C.J.Teo, J.Yoon, and P.Warren. The devices were fabricated at
MIT Lincoln Laboratory and in MIT’s Microsystems Technology Laboratories. This work
was supported by DARPA TTO, the Army Research Office, and the Army Research
Laboratory, managed by Dr. R.Rosenfeld, Dr. T.Doligalski, and Mr. J.Hopkins respectively.
The Lincoln Laboratory portion of this work was sponsored by the Defense Advanced
Research Projects Agency under Air Force Contract F19628-C-0002. Opinions,
interpretations, conclusions, and recommendations are those of the author and are not
necessarily endorsed by the United States Government.
References
[1] Epstein A H et al. 1997 AIAA Paper 97–1773, 28th AIAA Fluid Dynamics Conference, Snowmass
Village, CO
[2] Bart S F and Lang J H 1989 Sensors and Actuators 20 97–106
[3] Nagle S F 2000 Ph.D. Thesis, Massachusetts Institute of Technology
[4] Frechette L G, Nagle S F, Ghodssi R, Umans S D, Schmidt M A, Lang J H 2001 Technical
Digest of the 14th IEEE International Conference on Micro Electro Mechanical Systems, Interlaken,
Switzerland, 290–3
[5] Livemore C, Forte A, Lyszczarz T, Umans S D, and Lang J H 2002 Technical Digest of the 2002
Solid-State Sensor and Actuator Workshop, Hilton Head Isl, SC 251–4
[6] Paschen F 1889 Annal der Physik 37 69–96
[7] Frechette L G, Jacobson S A, Ehrich F F, Ghodssi F, Khanna R, Wong C W, Zhang X, Breuer
K S, Schmidt M A, and Epstein A H 2000 Technical Digest of the 2000 SolidState Sensor and
Actuator Workshop, Hilton Head Isl., SC 43–47
[8] Ghodssi R, Frechette L G, Nagle S F, Zhang X, Ayon A A, Senturia S D, and Schmidt M A 1999
Proceedings of the 10th International Conference on Solid-State Sensors, Sendai, Japan

53
Micro-Machined Variable Capacitors for Power

Generation
P Miao, A S Holmes, E M Yeatman, T C Green and P D Mitcheson
Electrical & Electronic Engineering, Imperial College, London SW7 2BT, UK
Abstract. Variable capacitors are key elements in electrostatic micro-power generators. In

such devices inertial forces are used to do work against the electric field of the capacitor,
thereby converting mechanical energy to electrical potential energy that can be extracted by a
suitable circuit. Applications are envisaged in portable, wearable or implantable electronic
devices where body motion could provide the mechanical energy source. This paper describes
the fabrication and initial testing of a micro-machined variable capacitor for power generation,
with overall dimensions in the meso-scale. The measured capacitance of the device varies from
100 pF to around 1pF as the mass moves from initial to final position, corresponding to a
hundred-fold increase in voltage if the device is operated in constant charge mode. Initial tests
of the capacitor on a vibration system (10Hz) have shown that a periodic high voltage output
of 2.3 kV can be generated if the capacitor is charged by a voltage source of 26 V. This
corresponds to an energy conversion rate of 2.4 µJ per cycle, or 24 µW at a vibration frequency
of 10Hz.
1. Introduction
Most micro-electro-mechanical systems (MEMS) today utilise macroscopic power sources.
This places some limits on the applications of MEMS devices. For example, in some
potential applications, these micro-machined devices have to be completely embedded or
fabricated in structures where external power is not accessible. Miniaturized implantible
medical sensors are one of these applications. Although high-energy batteries are currently
used to power these devices, this solution is less than satisfactory. Batteries are normally
bulky and contain a finite amount of energy, their shelf life is limited and the chemicals
contained in the batteries may be toxic. In search for an alternative power supply for these
applications, micromachined power generators are promising, being of a scale which can
be easily integrated into these MEMS devices. Some of them can be specially designed to
convert ambient energy in the environment into electrical power. Thus these MEMS devices
will become self-powered.
Although miniaturized self-contained power supplies are not a new idea, they have not
attracted much attention until recently. Attempts have been made to design and fabricate
micro-machined power generators utilising thermal energy [1–4], kinetic energy of gas
flow [5] and mechanical energy converted by a piezoelectric element [6] or a permanent
magnet [7–9].
We have designed and fabricated a micro-machined mechanical-electrostatic power

generator. Such an approach was chosen because our particular interest lies in the possible
applications in wearable, carried and medical implant electronics. In these applications,

54
photo-electric energy is normally not available. Although living bodies contain thermal
energy, the extraction possibilities for a micro-scale device are limited, as the thermal
gradients within the body are small, and thermal extraction depends on temperature
difference across the device. Kinetic energy from the volume flow of fluids requires a
comparatively high reservoir fluid pressure, and more crucially, require the device to present
an obstruction to the fluid flow, with obvious safety implications. As for the transduction
mechanism, piezoelectric transducers will suffer from self-discharge at the low frequencies
of these applications, while electromagnetic designs cannot achieve high damping forces
for low speed motion. Also, both piezoelectric and electromagnetic transducers are relatively
complicated to fabricate, particularly at dimensions compatible with body implantation.
The electrostatic transducer is simple in structure, being basically a variable capacitor
formed between two metallic electrodes separated by a dielectric. Tashiro et al [10] reported
an electrostatic power generator that harnesses the motion of a living body. A honey-comb-
type variable capacitor with a capacitance variation of 32–200 nF was used to convert
mechanical energy into electrical energy. An output of 58 µW was reported to be generated
with a constant charging voltage of 24 V and a load of 1.0 MΩ. However this device is
relatively large, at a mass of 0.64 kg.
The key element of an electrostatic micro-machined power generator is a variable capacitor,

which can convert mechanical energy into electrical energy by means of the work done by
an external force against the electric field formed between the capacitor plates. This paper
will present the structure and the results of initials test of such a micro-machined variable
capacitor. Although the external circuit for power extraction from the capacitor is still
under development, the operational principle of the power generator is also discussed.
2. Device structure and operational principle

The initial prototype was fabricated using MEMS techniques, with overall dimensions at a
“meso-scale” between conventional engineering and micro-scale; its cross-section is shown
schematically in Figure 1. A gold proof mass of 4.3 g, supported on a highly flexible
polyimide membrane, is suspended between a silicon top plate and a quartz base-plate. The
very low mechanical Q prevents accurate measurement of the resonance frequency, but we
estimate it at 10–20 Hz. Patterned metal films on the base-plate and the membrane form the
fixed and moving plates of the capacitor, the latter being connected to an external circuit
only at the extremes of its travel where it makes contact with plated contact studs.
Figure 1 Cross-section of the variable capacitor.
Power generation is achieved in this device by pulling apart the base plate and the movable
plate and then extracting energy stored in the electric field. During the input phase, the

55
mass is forced against the contact studs on the base plate, connecting it to a charging
circuit that brings it to a starting potential. In response to motion of the frame, which is
attached to the moving “host” (e.g. person), the mass is accelerated sufficiently in the
opposite direction and moved to the top plate where it is stopped by the contact studs on
the top plate. The energy stored in the capacitor is then extracted. Figure 2 shows the
essential elements of a basic charging and extraction circuit. During the input phase,
transistor Q1 is turned on to transfer energy from the source B1 into the inductor L1. When
Q1 is switched off, this energy is transferred to the variable capacitor C1 through a half-
cycle quasi-resonant action between L1 and C1. During the output phase, the charge is
recovered from the variable capacitor, at a higher voltage, by a second quasi-resonant
action, this time between C1 and L2. This action transfers energy from C1 to L2, and then
from L2 back into B1. The pre-charge operation is initiated by the control circuit while the
moving plate is in contact with the Input terminal, with the charge delivered being governed
by the Q1 on-time as set by the control circuit. The extraction operation occurs
automatically when the moving plate makes contact with the Output terminal.
Figure 2 Simplified schematic of the proposed charging and extraction circuit.

The capacitance measurement of the device was conducted using a PM 6303 RCL meter.
The device was mounted on a fixed stage. The movable plate of the device was attached to
a micrometer which moves the plate with a resolution of 10 ˜m. Capacitance was measured
between the movable plate and the base plate, as a function of the separation of these
plates. The result is shown in Figure 3.
Figure 3 Variation of capacitance throughout the travel distance of the movable plate.
It can be seen that the measured capacitance (crosses) of the device varies from C1=100 pF
to around C2=1 pF as the mass moves from bottom to top, corresponding to a hundred-fold
increase in voltage if the device is operated in constant charge mode. The measured

56
capacitance decreases more slowly than that calculated from an air gap capacitor (dotted
line). Adding fringing field effects to the model, for this plate area and separation, cannot
account for the substantial deviation seen; however, the measured results correspond well
to the model if tilting of the moving plate during travel occurs.
A leakage measurement was also carried out. The capacitor was charged to an initial
voltage V0 at its minimum gap position (input phase) and then disconnected from the
charging source. The movable plate was isolated from both charging source and top contact
for different delay times tD, while held at an intermediate capacitance Cm, and then moved
up to its maximum gap position (output phase), and the output voltage recorded. The
discharge transient amplitude is normalized to the zero delay discharge voltage and plotted
against the delay time as shown in Figure 4. If we assume the device has a leakage path
resistance R, then the discharge transient amplitude can be described by:
Figure 4 Reduction in the discharge transient amplitude against time.
Figure 4 indicates that the time constant for decay of the charge is about 200 s, indicating
a leakage resistance above 1000 G . This suggests that there will be negligible charge
leakage when the device is in operation.
Initial discharge tests of the capacitor were conducted on a vibration system. The variable
capacitor was pre-charged by a current-limited voltage source of 26V connected between
the Input terminal and the Ground terminal, as shown in Figure 5. The device was clamped
to a shaker table (IMV model PET-05A), the motion of which was measured in real time
using a Solartron frictionless inductive displacement transducer. To observe the discharge,
the Output terminal was fed into a virtual earth amplifier with in input resistance of 50 MΩ
and a feedback resistor of 50 kΩ. The amplifier output, representing the capacitor output
voltage scaled by a factor of “1/1000, was observed on an oscilloscope.
Figure 5 Experimental set-up for measuring periodic discharge, with the device mounted on a
vibration testing system.

57
Figure 6 Periodic discharge voltage for 40 Hz operation.
In Figure 6, the periodic discharges generated from the device are seen. The time intervals
of the periodic discharges correspond to the vibrating frequencies of the vibration system.
The low and variable amplitude of the discharges seen in the figure result simply from the
sampling rate being too low to catch the peaks for this multi-cycle measurement. The
discharge test was carried out in the frequency range from 5 Hz to 100 Hz, and the device
was shown to operate reliably across this range.
If the time scale is expanded, a complete discharge transient can be viewed. Figure 7 shows
the discharge transient when the device is vibrating at 10Hz. The amplifier output is -2.3V,
corresponding to a capacitor voltage of 2.3kV. Since the output of the charging source was
set to 26V, this corresponds to an 88.5 fold increase in voltage, which is in reasonable
agreement with the ratio of capacitance increase of 100. The difference may result from
parasitic capacitances in the system. The energy stored in a capacitor is simply ½CV2, and
the maximum energy generated from the device will be the difference between energies
stored in the capacitors at minimum and maximum gap positions. In the present case this
corresponds to an energy conversion rate of 2.4 µJ per cycle, or 24 µW at a vibration
frequency of 10Hz.
Figure 7 Discharge transient of one of the periodic discharges at 10Hz.

58
4. Conclusions and future work

A prototype variable capacitor has been designed and fabricated. The capacitance
measurement shows that the capacitance of the device varies from 100pF to around 1pF,
corresponding to a 100-fold increase in voltage. The leakage test shows that the time
constant is well over 100s, which is much longer than the working cycle of the device,
indicating that charge leakage is negligible in operation. This is certainly necessary for the
achievement of high efficiency in low frequency applications. Discharge tests on a vibration
system proved that the device is operational in a wide frequency range and that a high
output voltage can be generated from the device. Although the variable capacitor is the key
element of the electrostatic power generator, a complete generator will require suitable
power processing electronics. The design of such electronics has been outlined, and
implementation is currently under way. The overall efficiency of the power generator will
of course depend on the associated electronics as well as the electro-mechanical structure
and operation. This will be determined by tests on a fully integrated generator. The existing
generator is a hybrid integrated device; smaller, more highly integrated structures are
planned to meet such applications as medical implants.
Acknowledgements
We are grateful to the Engineering and Physical Sciences Research Council, and to the
European Commission (Disappearing Compute Initiative) for financial support of this
work.
References
[1] Strasser M, Aigner R, Franosch M and Wachutka G 2002 Sensors Actual A: Phys. 97–98 535–
542
[2] Schaevitz S B, Franz A J, Jensen K F and Schmidt M A 2001 Proc. Transducers 01: Eurosensors
XV CD-ROM
[3] Zhang C, Najafi K, Denial L P and Washabaugh P D 2001 Proc.Transducers 01: EurosensorsXV
CD-ROM
[4] Izumida K, Toyota N, Yoshimura M, Muroi Y, Takenaka T and Ikeda K 2001 Proc.Transducers
01: EurosensorsXV CD-ROM
[5] Wiegele T G 1996 Proc.IEEE: MEMS 2308–2313
[6] Glynne-Jones P, Beeby S P, James E P and White N M 2001 Proc.Transducers 01: Eurosensors
CD-ROM
[7] Ching N N H, Wong H Y, Li W J, Leong P H W and Wen Z 2001 Proc.Transducers 01:
Eurosensors CD-ROM
[8] Shearwood C and Yates R B 1997 Electronics Letters 33 1883–1884
[9] Amirtharajah R and Chandrakasan A P 1998 IEEE: Journal of Solid-State Circuits 33 687–695
[10] Tashiro R, Kabei N, Katayama K, Ishizuka Y, Tsuboi F and Tsuchiya K 2002 JSME International
Journal: Serious C 43 916–922

59
L F Gaunt and J F Hughes

Bioelectrostatics Research Centre, Department of Electronics and Computer
Science, University of Southampton, Highfield, Southampton, UK, SO 17 1BJ
Abstract. Electrostatic charging has long been used to improve the efficiency of a range of
sprayed liquids. However, the benefits have not until recently been available for exploitation
by domestic sprays due to the need for a high voltage power supply. A minimum charge-to-
mass ratio (q/m) of 1×10"4 C/kg is generally considered necessary to elicit electrostatic
benefits. This level of charge can now be imparted to liquids atomised from trigger-actuated
spray devices by a passive system, requiring no power supply. Induction charging was
achieved using a triboelectrically charged aluminium electrode. The q/m of the sprayed liquid
was dependent upon the charge residing on the induction electrode. The induction electrode
was electrically isolated and required a charge of between 0.7 and 1.3×10–8 C to deliver a water
spray with a q/m of 1×10–4 C/kg. This level of static charge was readily attained by tribocharging
the aluminium with polythene. Once generated, sufficient charge remained on the electrode
surface to charge successive sprays without the need for regeneration. The performance
advantages for a spray charged in this manner include attraction to and targeting of surfaces
and wrap-around onto surfaces not in the direct line of sight.
1. Introduction
Electrostatic charging is a familiar method used for improving the efficiency of a range of
sprayed liquids, for example in electrostatic crop spraying and painting applications. The
charge is usually imparted either by induction or corona methods, using a high voltage
power supply, or by tribocharging in powder coating systems. Typically a minimum charge-
to-mass ratio (q/m) of 1×10–4 C/kg is considered necessary to elicit electrostatic benefits [1;
2]. Domestic sprays, such as the familiar hand-held, pressure-pack (PP) device have not
previously exploited the advantages of electrostatic spray technology, primarily due to the
necessity for a power supply that adds prohibitive cost and complexity to a simple domestic
device. Recent research, however, has led to the development of a domestic pressure-pack
dispenser that produces an aerosol of droplets that carry this level of charge. Natural charge
exchange processes that occur during atomisation have been enhanced to attain the elevated
q/m values [3]. This has been achieved through the optimisation of formulation and pressure
parameters combined with an actuator incorporating a novel orifice design, which promotes
shearing of the electrical double layer at the liquid/solid interface. This has led to the
development of a domestic insecticide spray with significantly improved insect targeting
characteristics, leading to substantial improvements in insect knockdown and mortality
[2, 4].
Another form of spray device that is commonly encountered in the domestic environment
is the trigger actuated spray. This is used to dispense a variety of water-based detergents

60
and cleaning products such as furniture polish, kitchen and bathroom cleaners and
disinfectants. As the majority of these applications involve targeting a surface, electrostatic
spraying would be beneficial in enhancing product deposition and delivering a more even
coating. A threshold charge-to-mass ratio value of 1×10"4 C/kg would be predicted as a
minimum to achieve these advantages, based on previous research findings [1, 2]. In
attempting to achieve this level of charge, promoting the shearing of the electrical double
layer (as with PP sprays) was found not to be successful. This was believed to be because of
the difference in atomisation characteristics and due to the widespread use of aqueous
formulations. Therefore, an alternative way of producing a charge-on-demand spray was
required. One solution was an induction charging system using an electrically isolated
electrode at high potential, designed such that atomisation of the spray occurred under the
influence of a high electric field. The liquid reservoir was grounded and acted as a counter
electrode. Charged droplets were formed as they broke away from the reservoir. The charge
on the induction electrode could be established by triboelectric charging, which could
occur during the action of squeezing the actuation trigger.
Presented here are preliminary investigations into the level of charge that is required on
such an induction electrode to produce a spray with a q/m above the threshold value of
1×10–4 C/kg, and the q/m of successive sprays.
2. Materials and Methods
2.1. Design of the charge-on-demand trigger spray device

A trigger spray device was designed in order to test the concept of using an induction
electrode to charge a liquid as it was atomised from a conventional trigger pack. A
commercially available trigger pack was modified by covering the outer plastic shroud
with aluminium foil as shown in Figure 1. The foil was extended to the front of the shroud
to create an induction zone around the spray orifice. The distance across this gap was
8.0mm, and the spray orifice was positioned central within this. The foil was positioned so
that it remained electrically isolated from the other parts of the trigger device and the user
during actuation. For some experiments, static charge was deposited on the induction
electrode by contact with a voltage source (Brandenburg Alpha III), typically using a
voltage of between 1 and 8kV DC. The polarity of charge carried by the droplets was always
opposite to that of the induction electrode. The liquid reservoir was grounded via a wire
from the inside of the container and running over the lip to the outside. The induction
electrode was also charged by rubbing with a piece of polyethylene sheet. The charge
typically achieved on the electrode in this way was between 1 and 3×10"8 C, the polarity
always being positive.

61
Figure 1 Schematic drawing of the charge-on-demand trigger spray device
2.2. Measuring charge and charge-to-mass ratio

The charge residing on the induction electrode was measured by lowering the trigger
assembly into a Faraday pail connected to an electrometer (Keithley Instruments, 610C).
Similarly, the charge on the spray from the trigger device was measured by spraying into a
Faraday pail with an aperture wide enough to allow the spray plume to enter. The mean q/
m of the spray was calculated, based on a minimum of 5 replicates.
2.3. Influence of induction electrode potential on spray q/m

The electrical potential of the induction electrode was expected to determine the level of
charge achieved on the liquid during atomisation, as it would determine the intensity of the
electric field. The electrical potential of the induction electrode was controlled by contacting
the electrode with a range of voltages between 1 and 8kV DC. The electrode polarity was
positive, so the spray polarity was negative. The q/m of the liquid was measured, and the
potential restored before another measurement was taken. The test liquids were tap water
and a 5% detergent solution. Statistical analysis (t-test) was performed to identify the
significance in differences between the q/m of the two test liquids.
2.4. q/m of successive charges

As charging the sprayed liquids by induction was not expected to remove the charge from
the induction electrode, successive trigger actuations should be charged. However, the
charge on the induction electrode might be expected to decrease by natural leakage,
resulting in a corresponding decrease in the q/m of the spray. To investigate this effect, the
q/m of tap water was measured for the first, second and third successive trigger actuations.
The charge on the induction electrode was between 1 and 2×10–8 C, achieved by triboelectric
charging with polyethylene sheet.
3. Results and Discussion

3.1. Influence of induction electrode potential on spray q/m
The results in Figure 2 show the relationship between the charge on the induction electrode
and the q/m of the resultant spray for tap water and a 5% detergent solution. As the charge

62
on the electrode was increased, so the q/m of the sprays increased. This was seen for both
liquids, although the q/m of the water was significantly higher (p<0.05) than that of the
detergent for the same level of charge on the induction electrode. The increase in q/m as the
level of charge on the induction electrode increased would have occurred as a result of the
intensification of the electric field at the point of liquid atomisation. Thus, a higher charge
was induced in the forming droplets as they broke away from the grounded liquid reservoir.
The difference in q/m between the tap water and detergent solution could have arisen
through differences in conductivity, viscosity or surface tension; factors that can effect the
charging or atomisation characteristics.
The natural q/m of these liquids without a charge on the induction electrode is also
shown in Fig. 2. The values are low, in the region of 7×10–6 C/kg for water. The charge is
believed to arise from shearing of the electrical double layer that exists at the liquid-solid
interface.
Figure 2 Relationship between the charge on the induction electrode and the charge-to-mass ratio of
water and detergent sprays. Based on 5 replicates, error bars indicate the standard error of the mean, *
indicates statistically significant difference between q/m values (p<0.05).
3.2. q/m of successive charges

Figure 3 shows that charge is induced in successive sprays from the trigger, but that the q/
m of the sprays decreased. This is probably due to the charge carried on the induction
electrode decreasing with time or with successive sprays. The charge on the induction
electrode could be expected to decrease over time due to charge leakage from the surface
and neutralisation. It is not necessarily lost as an outcome of the sprays. The results show
that triboelectric charging of the induction electrode is a suitable method for charging the
spray to the theoretical threshold of 1×10"4 C/kg. The natural q/m of the sprays without the
induction electrode being charged is much lower, but also shows a decrease with successive
actuations. The natural q/m may decrease with successive actuations due to the changes in
the electrochemical characteristics of the interface following each actuation.

63
Figure 3 Charge-to-mass ratio of successive trigger sprays of tap water. Tribocharged induction
electrode charge of between 1 and 2×10–8 C. Based on 5 replicates, error bars show the standard
error of the mean.
4. Conclusions
It has been demonstrated that it is possible to charge sprays produced by trigger actuated
devices by a passive system, not requiring a voltage source. Charge-to-mass ratio values in
excess of 1×10–4 C/kg have been demonstrated on water and detergent solutions. This has
been achieved by atomising the liquid in an electric field created by an electrically isolated
induction electrode at high potential. The minimum charge on the induction electrode for
tap water was between 0.7 and 1.3×10–8C, while for a 5% detergent solution it was between
0.7 and 1.7×10–8C. In these investigations the charge was imparted to the electrode by
contact with a voltage source for reproducibility. Tribocharging of the aluminium against
polyethylene sheet consistently achieved charges of between 1 and 3× 10"8C. This was
shown to achieve water sprays of in excess of 1.2×10"4 C/kg. It is envisaged that in the final
device the triboelectric charging will be generated during the process of trigger actuation
[5]. Work is ongoing to design this contact process into existing trigger devices, and to
optimise the geometry of the charge induction zone.
The electrical potential of the induction electrode influenced the charge-to-mass ratio
of the sprayed liquid. As the electrical potential of the induction electrode was increased,
the q/m of the liquid spray was increased. This was observed for tap water and a detergent
solution, although the q/m of these differed for the same electrode potential. Factors such as
surface tension, conductivity and viscosity will also have considerable influence on the
charge characteristics, through their effect on the atomisation process and the time rate of
charge transfer [6, 7].
The benefits of a charge-on-demand trigger actuated spray device reflect those already
known to industrial applications of electrostatic spraying. These include attraction to the

64
target surface, wrap-around onto surfaces not in the direct line of sight and more even
coverage resulting from space charge effects and electrostatic attraction. As a result, the
efficacy and performance of common domestic sprays could be substantially improved.
Acknowledgements
This work was funded by Reckitt Benckiser Ltd (UK), whose support is gratefully
acknowledged.
References
[1] Singh S, O’Neil BC, Bright SW. 1978 Journal of Electrostatics 4:325–334.
[2] Gaunt LF, Hughes JF, Harrison N. 2003 Journal of Electrostatics 57:35–47.
[3] Fox R, Harrison N, Hughes JF, Whitmore LF. 1999 patent WO 99/01227.
[4] Whitmore LF, Hughes JF, Harrison N, Abela M, O’Rourke P. 2001 Pest Management Science
57 (5):432–436.
[5] Hughes JF, Gaunt LF. Spraying devices. 2001 Patent application GB 0116543.0.
[6] Cross J. Electrostatics: principles, problems and applications. 1987 Adam Hilger, Bristol.
[7] Law SE. Electrostatic atomization and spraying. In: Chang J, Kelly A.J., Crowley J.M., editors.
1995 Handbook of electrostatic processes. New York: Dekker, pp 413–440.

65
Unipolar Charging and Contact Discharging of Insulating

Particles on the Surface of a Grounded Electrode
Adrian Sarauila1, 2, Adrian Mihalcioiu1, 2, Alin Urs1, 2, Lucian Dascalescu1

1
Electronics and Electrostatics Research Unit, LAII-ESIP, UPRES-EA 1219
University Institute of Technology, 4 Varsovie Ave., 16021 Angouleme, France
2
Technical University, 15 C-Daicoviciu Street, 3400 Cluj-Napoca, Romania
Abstract. The study of the unipolar charging of insulating particles has been stimulated by the
recent development of several important electrostatic technologies: precipitation of dusts,
deposition of powders, separation of granular materials. The aim of the present work is to
analyse the discharging conditions of insulating disks in contact with an electrode, after having
been charged in a unipolar corona field. The experiments were carried out on a roll-type
electrostatic separator that simulated the charging/discharging conditions in an industrial unit.
The measured data show that the discharge process is a complex phenomenon, depending of
at least the following factors: the conductivities of the respective bodies, the aspect of the
contact, and the environmental conditions.
1. Introduction
The technological developments in the field of electrostatic processing of particulate matter
(precipitation of dusts, deposition of powders, separation of granular materials [1–5]) have
always stimulated the research of the physical phenomena related to the ionic charging of
conducting and non-conducting particles in D.C. or A.C. electric fields [6–8]. The equations
established by Pauthenier and Moreau-Hanot some 70 years ago [9], continue to be
confidently used for modeling the unipolar charging of single spherical particles moving
freely in a uniform external electric field Eo, where uniform monopolar space charge with
density q exists. This model is accurate enough for the electrostatic precipitators, as in most
industry applications the dust particles are larger than 2 µm in diameter (thermal diffusion
can be neglected) and their volume concentration is low.
Numerical techniques can presently be employed for evaluating the ionic charging of
particles under different conditions, non-amenable to analytical calculations: single
stationary spheres in a uniform electric field [11, 12], single spherical particles moving on
the surface of an electrode [13], one or several insulating cylinders evolving in a
monoionized electric field [14]. The experiments carried out on the unipolar charging of
insulating particles in contact with the rotating roll electrode of a corona-electrostatic
separators validated some of these results [15, 16]. They pointed out an additional effect of
significant importance for the outcome of the electrostatic separation process: the decay of
the charge carried by the particles in contact with an electrode [17].
The aim of the present work is to analyse some of the factors that influence the discharging
conditions of insulating disks in contact with an electrode, after having been charged in a
unipolar corona field.

66
2. Materials and method

The study was carried out on of four types of disk-like insulating particles, described in Fig.
1. The samples II to IV were obtained by classification from a genuine polyamide granular
material provided by a plastic manufacturer. Type III and IV are cylinders terminated by
curved surfaces. In order to obtain type II particles closer in shape to a “standard” disk, glass
paper was used to convert the curved bases of genuine granules similar to those of type III
and IV into rather smooth planes.
Particle charging/discharging conditions in an industrial unit were simulated by using
a laboratory roll-type corona-electrostatic separator (CARPCO, Jacksonville, FL), provided
with a wire-type corona electrode, located at a distance s=50 mm from the surface of the
grounded roll electrode and brought to a positive potential U=25 kV (Fig. 2). In each
experiment, groups of three particles were placed on the surface of the roll electrode, with
their centres located in the vertical plane defined by the corona wire and the axis of the roll
electrode. They were subjected to a corona field for 10 s. Then the roll drive was turned on,
at a speed n high enough to throw off the particles, to be collected in a Faraday pail
connected to the electrometer (Keithley Instruments, Model 6514). The separator was also
provided with a thin metallic wire, the role of which was to remove the particles too tightly
“pinned” on the surface of the roll electrode. After turning off the high voltage, the particles
were maintained in the same position for intervals of time varying from tdisch=0 s to 4 min.
Fig. 1. Schematic representation of the four types of particles employed for this study.
Fig. 2. Schematic representation of the experimental set-up.

67
3. Results
In the first set of experiments, the discharging of type I particles was carried out for various
distances d between adjacent particles. The charge of the particles was found to decrease
with the discharging time tdisch as shown in Fig. 3, where each point was obtained by
dividing the measured charge by the number of new particles collected in the Faraday pail
at each experiment.
Type II particles were employed in the similar set of experiments, the results of which
can be examined in Fig. 4. The data measured during the other two sets of experiments,
involving respectively the particles of types III and IV, are represented in Fig. 5.
Fig. 3. Charge Q of type I particles as function of discharging time tdisch, (the particles were
charged at +25 kV , s=50 mm).
Fig. 4. Charge Q of type II particles as function of discharging time tdisch,

68
Fig. 5. Charge Q of type III and type IV particles as function of discharging time tdisch.
4. Discussion
The study of ionic charging and discharging of particles in contact with an electrode may
give clues for the improvement of the efficiency of industrial electrostatic separation
processes. Previous studies demonstrated that the proximity of other bodies modifies the
distribution of the electric field and hence the conditions of corona charging of insulating
particles, as compared with the case when they are single in a uniform electric field. This is
why the experiments were conducted for different distances between adjacent particles.
The results in Figs. 3 and 4 confirm previous findings: distanced particles acquire larger
amounts of charge than those very close to each other do. This aspect, as well as the
dynamics of particle discharging in contact with the carrier electrode, should be taken into
account in the development of any new electrostatic separation technology.
The charge decay rate of polyamide, which is a very good insulator, is much lower than
that of PVC. The charge of polyamide disks reduces to ½ in more than 4 minutes, against
less than 2 seconds for PVC particles. Thus, the difference in charge decay rate might
become a mean for selectively sorting the constituents of a granular mixture of insulating
materials in contact with a grounded electrode. Judging from the dispersion of the measured
charge values in the reported experiments, the feasibility of such an application of the
electrostatic separation technology depends on the solutions found for reducing the
dispersion of particle size, the non-uniformity of the corona generated by the wire electrode,
as well as the variation in the state of the surface of the carrier electrode and the aspect of
particle-electrode contact.
At roughly the same diameter, the disks in sample IV have a larger lateral surface and
carry a larger amount of charge than those in sample III, but the dynamics of charge decay
seems to be similar (Fig. 6). Nevertheless, more experiments are needed to explore in depth
the effect of size, shape and contact surface geometry on particle discharging conditions.
Indeed, the charge decay rate of type II particles is slightly different from that measured for
the disks of type III and IV, which have a different geometry.

69
As the sets of experiments were done in different environmental conditions, the control
of ambient temperature and relative humidity should be mandatory for any further experiment
on this delicate issue.
5. Conclusion
The discharging of insulating particles in contact with a grounded electrode is a complex
phenomenon, depending on at least the following factors: the surface/volume conductivities
of the respective bodies, the aspect of the contact, and the environmental conditions.
1) A good insulator displays lower charge decay rates than less insulating particles. This
property could be used for separating two sorts of plastics, for instance.
2) Particles of same nature but dissimilar with respect to the aspect of their surface
(geometry, roughness) behave differently in contact with an electrode. The explanation
resides in the relation existing between the charge decay rate and the particle-electrode
contact resistance. When one or more of the constituents of a granular mixture are
highly non-homogeneous in terms of particle size, shape and surface state, the
electrostatic separation is very likely to fail, as particles that are similar in nature will
carry different charges and will evolve along totally different trajectories.
3) The relative humidity of the ambient air can modify the value of the contact resistivity
and hence the discharging conditions. Further investigations are needed to quantify
the effect of particle-electrode contact and of the environmental conditions on the
charge decay rate of insulating granules in situations similar to those encountered in
the various industry applications of the electrostatic separation processes.
References
[1] Hughes J F 1985 Electrostatic Powder Coating (New York: Research Studies)
[2] Crowley J M 1986 Fundamentals of Applied Electrostatics (New York: Wiley)
[3] Cross J A 1987 Electrostatics. Principles, Problems and Applications (Bristol: Adam Hilger)
[4] Masuda S and Hosokawa S 1995 Electrostatic precipitation in Handbook of Electrostatic
Processes (Chang J S, Kelly A J and Crowley J M, Editors) (New York: Marcel Dekker) 441–
480
[5] Iuga A, Morar R, Samuila A and L. Dascalescu L 2001 IEE Proc.- Sci. Meas. Technol. 148 47–
54
[6] Mc Donald J R, Anderson M H and Mosley R B 1980 J. Appl. Phys. 51 3632–3643
[7] Meyle B D and Hughes J F 1983 Electrostatics, Inst. Phys. Conf. Ser. 66 (Oxford: IoP) 155–160
[8] Chang J S 1995 Electrostatic charging of particles in Handbook of Electrostatic Processes
(Chang J S, Kelly A J and Crowley J M, Editors) (New York: Marcel Dekker) 39–49
[9] Pauthenier M and Moreau-Hanot M 1932 J. Phys. Radium 3 590–613
[10] Mizuno A 1981 Proc. Int. Conf. on Electrostatic Precipitation, Monterey 304–325
[11] Masuda S and Washizu M 1979 J. Electrostat. 6 57–67
[12] Inculet I I, Malik N H and Young J A 1983 Electrostatics, Inst. Phys. Conf. Ser. 66 (Oxford: IoP)
99–104
[13] Samuila A, Iuga A, Morar R, Tobazeon R and Dascalescu L 1997 J. Electrostat. 40 377–382
[14] Dascalescu L, Urs A, Dumitran L M and Samuila A 2001 Conf. Rec. IEEE/IAS Ann. Meet.(Chicago:
IEEE) paper 9–3
[15] Dascalescu L, Rafiroiu D, Samuila A and Tobazeon R 1998 IEEE Trans. Ind. Appl. 34 35–42
[16] Dascalescu L, Morar R, Iuga A, Samuila A, Neamtu V and Suarasan I 1994 J. Phys. D: Appl.
Phys. 21 1242–1251
[17] Urs A, Samuila A, Mihalcioiu A and Dascalescu L 2002 Conf. Rec. IEEE/IAS Ann. Meet.
(Pittsburgh: IEEE) paper 26–3

71
The investigation of the ozone productivity of a new

discharge type
I Jenei1, E Kiss2, I Berta1

1
Budapest University of Technology and Economics Budapest, Hungary
2
College of Dunaujvaros, Dunaujvaros, Hungary
Abstract. It is well known that the most popular technology for ozone generation use electric
discharges. It is reasonable to investigate the discharge process systematically. This work
aims to improve the efficiency of ozone generators by using novel electrical discharges. In
order to improve the ozone production, a new discharge type was introduced, the discharge
length was increased by using variable electrode distance or secondary electrodes. The equipment
allows to generation of surface discharge as well as other special types of corona discharges.
By widening the gap, both the discharge length and the ozone generation were increased to a
maximum. The flow rate, the electrode distance and pulse voltage were changed. A new state
of the balance of the ozone generation and recombination processes have been set by special
adjustment of the equipment.
1. Introduction
The more and more popular ozone technologies need increasing volumes of ozone, and
predominantly ozone is made by means of electric discharge. This fact proves the importance
of the improvement of the process. An ozone generator using electric discharge to generate
ozone is made of two parts; the high voltage power supply and the reaction chamber
consisting of the discharge and induction electrodes. At the beginning the power supply
used direct current and that followed the alternating one. The basic arrangement of the
electrodes is made by forming concentric cylinders from the discharge and induction
electrodes. The varying electric field induces corona discharges around the discharge
electrode, which generates ozone. The electrode arrangement that produces the surface
discharge [1] was developed by Professor Masuda in Tokyo. In comparison with concentric
cylindrical electrode arrangement, this surface discharge type generator has a higher specific
volume efficiency, because the specific length of the discharge electrode is longer forming
a special shape on the ceramic plate [3].
A new approach was developed for ozone generation. The surface discharge can be
transformed into a special volume discharge using a novel electrode arrangement. The
present investigation describes the results of the experiments with variable electrode distance
arrangement (Fig. 1a.).

72
2. Experiment
There are two methods to transform the surface discharge into a special volume discharge.
The first one uses a variable electrode distance between the discharge and—insulator
ceramic plate—induction electrodes (Fig. 1a.) to widen the discharge length.
The other method uses secondary electrodes (Fig. 1b.) to produce a longer discharge
length. In this case the applied voltage at the secondary electrodes is particularly important.
The applied pulse drive power supply has two high potential outputs. The influencing
parameters of the discharge are the phase difference between the voltage pulse on the
discharge electrode and voltage pulse on the secondary electrodes, as well as the pulse
voltage ratio between the discharge and secondary electrodes. This method is under detailed
investigation at present. The design of the experiments (Fig.2.) allowed the applied potential
and the flow rateto be varied. [2].
Oxygen and artificial air were used as the feed gases for the experiments, the temperature
was 23°C. The shielded cylindrical polyethylene tube was the reaction chamber, which
contained the discharge and induction electrodes. The electrical parameters were measured
by a Tetronix TDS 380 digital oscilloscope (with signal recording capacity). The ozone
concentration was measured by equipment developed (and calibrated by wet chemical
method) at the laboratory by the first and the second author. The ultra-violet absorption of
the ozone is the basic principle of the ozone measurement.
Fig. 1a.) Electrode arrangement with variable electrode distance, b.) The surface discharge with
secondary electrodes
Fig 2. Experimental layout

73
3. Discussion
The first three experiments characterised the effect of potential change on the ozone
concentration at three different electrode distance and flow intensities. The ozone
concentration has a linear relationship to applied voltage (Figs. 3.). At the higher flow
intensity the ozone concentration is lower, but the specific ozone productivity is the same
within a certain voltage interval. Multiplying the concentration by flow intensity results in
approximately similar values for that voltage interval. If Fig. 6. (0 mm electrode distance)
is compared to Fig, 4. (0.5 mm electrode distance), it is discernable that using a greater
electrode distance the specific ozone productivity increases under the same experimental
condition. But further increasing electrode distance (Fig. 5, 1mm electrode distance) results
in a lower ozone productivity than the previous case (Fig. 4.). The previous three figures
show the effect of electrode distance on ozone concentration at different flow intensities;
the three different electrode settings give different ozone producing capacity.
Fig. 3. Ozone concentration (C) plotted against applied voltage

(electrode distance=0 mm)
Fig. 4. Ozone concentration plotted against applied voltage

(electrode distance=0.5 mm)

74
Fig. 5. Ozone concentration plotted against applied voltage

(electrode distance=1 mm)
Fig. 6. Ozone concentration (C) vs. the electrodes distance
The characteristics of electrode distance effects on ozone concentration were investigated

in detail. The functions (Fig. 6.) indicate, that the surface discharge (the distance between
the electrodes is zero) is the least effective ozone producing discharge type. There is a
certain distance for a particular voltage, which is most effective. This distance grows by
increasing the applied voltage. The increased electrode separation results in the formation
of many streamers [4] along the discharge electrode. In this case, discharge can also be seen
in the volume surrounding the discharge electrode, not only on the surface. The increase of
electrode distance is produces more ozone while the electric field is strong enough to
induce streamers. Another key point is the specific electrical energy consumption. The
energy of discharge can be obtained by multiplying the pulse voltage by the discharge
current. The results are in Fig.7. Using different electrode distances there is no significant
change of electrical energy consumption from 0 mm to 0.6 mm. In the previous investigation,
the ozone productivity was increasing with increased distances in this interval.

75
Fig. 7. The energy (E) consumption plotted against the electrode distance
4. Conclusion
The present research has proved that the surface discharge can be transformed into volume
discharge by operation of small distance between the electrodes. The ozone concentration
is an approx. linear function of the applied voltage over the range being investigated. By
changing the separation of discharge electrode from the surface of the ceramic, the ozone
concentration has a maximum at about 0.6 mm and about 0.9 mm, for 6.5 kV and 8 kV
supply voltages, respectively. The optimum is dependent on the applied voltage. This kind
of discharge shape has higher ozone producing capacity then the surface discharge alone.
In addition with the above mentioned results the electrical energy consumption is
approximately the same. According to the present experiment, the combined discharge
type ozoniser seems to be more effective then both the volume type or the surface type
ones.
References
[1] Masuda S and Kiss E 1987 Electrostatics Conference Oxford On streamer discharges in ceramic-
based ozoniser using high frequency discharge
[2] Márton H and László B and JenQ H 1976 Budapest Az ózon Mq´szaki Könyvkiadó
[3] Kiss E and Masuda S 1985 Streamer Induced Capacity of the Ceramic-Made Surface Discharge
Type Ozoniser Ibid.
[4] Horváth T. 1986 Budapest—Csernatony H.A.: High voltage engineering (In Hungarian)
Tankönyvkiad

77

D A Hays and J Q Feng
Wilson Center for Research and Technology, Xerox Corporation, 800
Phillips Rd., Webster, NY 14580 USA
Abstract. The electrophotographic process for producing copies and prints of documents
requires charged toner particles to be moved by an applied electric field. Toner particles about
10 µm in diameter and consisting of a pigment dispersed in a polymer are usually charged by
the phenomenon of triboelectricity. There has been considerable discussion in the literature
regarding the relative contributions of electrostatic and van der Waals forces to toner adhesion.
If the van der Waals force on a toner particle is reduced with surface additives, the electrostatic
adhesion becomes dominant. However, the measured electrostatic adhesion on a triboelectrically
charged toner particle is much greater than the prediction of an electrostatic image force model
that assumes a uniform surface charge distribution. The enhanced electrostatic adhesion is
attributed to a nonuniform surface charge distribution on triboelectrically charged particles. If
the toner is charged by corona ions, the surface charge density tends to be uniform and the
electrostatic adhesion is significantly reduced. We present recent electric field detachment
measurements of toner ion charged with corona currents in an alternating electric field. The
dependence of the electric field detachment on toner coverage is consistent with model predictions
of toner adhesion enhancement due to fringe electric fields from neighboring charged toner
particles.
1. Introduction
Efficient transfer of charged toner particles (~10 µm) between surfaces with an applied
electric field is important for several process steps in electrophotography. Despite numerous
studies of toner adhesion conducted over decades, the interpretations of measurements
reported in the literature are not consistent. The relative importance of electrostatic and van
der Waals forces is still a subject of debate [1].
When the van der Waals component of adhesion is minimized with surface additives,
the measured particle adhesion of triboelectrically charged toner is observed to increase
with increasing toner charge; implying that the electrostatic component of adhesion is
dominant [2]. However, the measured adhesion is much greater than the prediction based
on an electrostatic image force model for a uniformly charged dielectric sphere. To explain
the enhanced electrostatic adhesion, a theory based on nonuniform surface charge
distribution on triboelectrically charged toner was proposed [3–5].
When irregularly-shaped toner particles with surface additives are charged by corona
ions in air fluidized toner, Christy found that the particle adhesion as measured by electric
field detachment is much less than the adhesion of triboelectrically charged toner [6]. This
suggests that the electrostatic adhesion of ion charged toner particles can be described by
an electrostatic image force model in which irregularly-shaped toner particles are
approximated as dielectric spheres with a uniform surface charge density. Theoretically,

78
the electrostatic image force between a uniformly charged dielectric sphere and conductive
surface can be described by the equation,
(1)
where Q is the particle charge, R is the average particle radius and ˜o is the permittivity of
free space [7]. For a particle of dielectric constant κ=4 (typical for a carbon-loaded polymer),
the polarization correction coefficient, α, is 1.9. When an electric field is applied to detach
the particle, the applied force due to the field, E, is
(2)
where β and γ are polarization correction coefficients. For κ=4, we have β=1.6 and γ=0.99.
When the strength of detachment electric field is low as in the case of ion-charged toner, the
second term on the right side of Eq. (2) can be neglected. When the sum of the forces from
Eqs. (1) and (2) (which gives the net electrostatic force) is greater than the nonelectrostatic
adhesion such as the van der Waals force FNE, particle detachment will occur at a detachment
electric field, Ed, of [8]
(3)
For the ion-charged toners studied by Christy in which Q=13 fC and R=6 µm, the calculated
detachment field from Eq. (3) is 1.0 V/µm with FNE assumed to be negligible. This is in
reasonable agreement with the measured median detachment field of 0.7 V/µm [6]. However,
Christy’s measurements were conducted with approximately a monolayer of toner whereas
Eq. (3) is for an isolated particle. Due to fringe electric fields from neighboring charged
particles, a monolayer of toner with a uniform surface charge density should have enhanced
toner adhesion. About a five-fold increase in the detachment electric field was found by
Shapiro and Hays based on calculations for a hexagonal, close-packed array of uniformly
charged dielectric spheres [7].
The present paper is motivated by a desire to understand the reason for the very low
electric field required to detach a monolayer of toner particles charged by corona ions in a
fluidized bed of toner, as reported by Christy [6]. We report preliminary electric field
detachment measurements using an alternative ion-charging method in which toner in an
air stream is subjected to corona ion currents in an alternating electric field. This particle
charging method has been widely studied by Jaworek and Krupa for electrostatic precipitator
and electrostatic powder coating applications [9]. We constructed a miniaturized version
of this ion charging apparatus and combined it with a toner cloud delivery system.
2. Apparatus for ion-charging toner particles in an alternating electric field

If particles entrained in an air stream are subjected to unipolar corona ions in the presence
of an applied electric field E, each particle of radius R and dielectric constant κ will acquire
a maximum charge given by the Pauthenier equation
(4)
Recent studies by Adamiak et al. describe an apparatus and theoretical analysis of ion
particle charging in an alternating electric field [10].

79
2.1 Ion-charging device

Figure 1 shows a schematic of the ion-charging device used in this study to charge toner
particles. The device is similar to that described by Adamiak et al. [10], but differs in scale
(about an order of magnitude smaller), hardware components and power supplies. The
corona ion generating units are so-called scorotrons widely utilized in electrophotography.
The coronodes consist of two pin arrays with corona emitting points. The gap between the
left and right screens is 8 mm, and the length of the ion-charging zone is 2.9 cm. The
coronodes and screens are connected to power supplies through a network of high voltage
(10 kV) diodes and resistors (1.5 MΩ). A sine-wave generator is connected to a left and right
high-voltage power supplies (HVPS) set at a peak voltage of 8 kV. By connecting the left
HVPS to an inverting input and the right HVPS to a noninverting input, the AC voltage of
the left HVPS is 180° out of phase with respect to the right HVPS. When the left HVPS is at
sufficiently high negative voltage, the left coronodes generate negative ions since the
diode between the coronodes and screen is open-circuited and the diode connecting the
left screen to ground is short-circuited. Meanwhile, there is no corona emission from the
right coronodes and screen since the diodes isolate both elements from ground at the same
potential. This electric field causes negative ions to flow from the left scorotron into the
gap where toner particles are entrained in an air stream. While an electrostatic force tends to
push the charged particles towards the right screen, the polarities of the power supplies are
switched during the next half cycle before the particles deposit on the right screen. During
the next half cycle, the right coronodes emit negative ions when the right screen is at
ground potential while the left coronodes and screen are at a high positive potential. Thus,
the toner particles accumulate additional negative charge as they drift towards the left
screen. With increasing cycles, the particles acquire more charge until the Pauthenier
charging limit is reached.
Figure 1. Ion-charging device for toner cloud in an air stream.

80
2.2 Toner delivery and collection system

Figure 2 shows a schematic of the complete apparatus for delivering, charging and collecting
toner for electric field detachment measurements. Toner is placed in a reservoir that contains
a brush slowly rotated by a motor (M). An air stream entrains toner particles for delivery to
the ion-charging device through a pipe and narrow slit centered over the charging device.
After exiting from the ion-charging zone, the toner enters a rectangular cross-section
collection zone (17 cm long) in which a biased electrode is spaced 1.2 cm from a grounded
toner-collecting electrode. An electrostatic force acting on the charged toner causes
deposition onto the grounded electrode. The grounded electrode consists of a thin brass
sheet with a rectangular aperture. The aperture prevents toner deposition on the perimeter
of the aluminum collecting plate where a dielectric shim is placed for electric field
detachment measurements. Toner is deposited over a rectangular area that is 5.1 cm high
and 6.3 cm wide.
A vacuum is supplied to a plenum under the apparatus to provide airflow through the
ion charging and toner collecting zones. The air speed measured with a hot-wire anemometer
is 0.5 m/s at the entrance of the charging zone and 2.5 m/s at the exit of the collecting zone.
The large differential in air speeds is due to air being drawn in through slots in the plastic
shield of the scorotron devices.
Figure 2. Apparatus for delivering, charging and collecting toner onto a plate.

81
3. Measurements
3.1 Toner charge
For initial measurements, we chose a toner that is similar to that used by Christy so that his
results can be compared to our measurements. The black pigmented, irregularly-shaped
toner with a median volume diameter of 11.4 µm (16%<8.9 µm and 84%<14µm) contained
surface additives to minimize the van der Waals force. The toner was placed in the toner
reservoir, delivered to the charging device in the form of a toner cloud with an air stream,
ion charged in the alternating electric field apparatus, and deposited on the collecting plate
with a bias of VA=”1000 V on the opposing electrode. The charge-to-mass ratio, Q/M, of
toner deposited on the collecting plate was approximately—5 µC/gm for a HVPS peak AC
voltage setting of 8 kV at a frequency of 430 Hz. From Eq. (4), the maximum Q/M is
predicted to be
(5)
where ρ is the toner density of 1.1 gm/cm3. For a peak electric field of E=1 V/µm, R=5.7 µm
and κ=4, the calculated Q/Mmax is -8.5 µC/gm. The calculated value based on spherical
rather than irregular-shaped particles is in reasonable agreement with the measured toner
particle charge considering that no attempt was made to optimize the ion-charging
conditions.
3.2 Electric field detachment

Figure 3 shows typical curves for the cumulative toner detachment versus applied electric
field strength for initial toner densities of 0.07 and 0.39 mg/cm2 on the collecting plate
Monolayer coverage corresponds to 0.76 mg/cm2 for a hexagonal close-packed array of
11.4 µm diameter toner. An aluminum receiving electrode was spaced from the toned
collecting plate by a dielectric shim near the perimeter. The gap between the donor and
receiver electrodes was calculated to be 55 µm from a capacitance measurement. For toner
coverage of 0.07 mg/cm2, the particles are expected to be isolated on the average although
Figure 3. Cumulative toner detachment in toner weight percent versus applied electric field comparing
data for initial donor electrode densities of 0.07 and 0.39 mg/cm .

82
clustering occurs since the toner deposition is somewhat uneven. The median magnitude of
the detachment electric field for a low toner coverage is about 0.5 V/µm. This is quite close
to the calculated value of 0.32 V/µm from Eq. (3) for an isolated sphere with FNE neglected
and Q=-3.9 fC corresponding to 11.4 ˜m diameter toner with a Q/M of—5 µC/gm. If FNE is
taken into account, Eq. (3) can also be used to estimate the upper limit of the van der Waals
force. Thus, we obtain βQEd~3.1 nN, |Fi|~2.0 nN (cf. Eq. (1)), and FNE~1.1 nN for Q=3.9 fC,
Ed=-0.5 V/µm, and R=5.7 µm at κ=4. (The magnitude of the second term on the right side of
Eq. (2) of about 0.2 nN is indeed negligible to a first approximation.) For toner coverage
near a monolayer, the median magnitude of the detachment electric field is about 2 V/µm.
The higher median detachment electric field is consistent with the theory accounting for
the fringe electric fields from neighboring charged particles, which yields a detachment
electric field of 1.7 V/µm [7]. The detachment curves exhibit stepwise detachment behavior
that we attribute to an “unzipping” of neighboring toner.
4. Conclusions
We find that the adhesion of ion-charged toner is significantly lower than that of
triboelectrically charged toner with median detachment electric fields typically at 10 to 15
V/µm. Furthermore, we find that the median electric field detachment of ion-charged toner
depends on surface coverage as expected for uniformly charged particles. The measured
detachment fields are in reasonable agreement with theoretically calculated values. The
higher detachment fields for higher toner coverage are due to fringe electric fields from
neighboring charged particles. Although our results for ion-charged toner are consistent
with the theory for uniformly charged dielectric spheres, it is unclear why Christy measured
a median detachment electric field of only 0.7 V/µm for a monolayer of toner.
Acknowledgements
We would like to thank Prof. Kaz Adamiak for his input regarding the ion-charging device.
We appreciate the support from colleagues Paul Julien, Jack LeStrange and Bill Wayman.
References
[1] Rimai D S, Ezenyilimba M, Goebe W K, Cormeir S and Quesnel D J 2002, J. Imaging Sci.
Technol., 46 200–207
[2] limura H, Kurosu H and Yamaguchi T 2000, J. Imaging Sci. Technol., 44 457–461
[3] Hays D A 1978 Photogr. Sci. Eng., 22 232–235
[4] Lee M H and Ayala J 1985 J. Imaging Technol., 11 279–284
[5] Hays D A 1988 Particles on Surfaces 1: Detection, Adhesion and Removal, Mittal K L, Ed. (New
York: Plenum Press) 351–360
[6] Christy O D 1995 IS&T’s NIP 13: International Conference on Advances in Non-Impact Printing
Technologies (IS&T, Springfield, VA) 176–179
[7] Shapiro Y and Hays D A 1999 Proceedings of the 22nd Annual Meeting of the Adhesion Society
(Panama City, FL) 28–30
[8] Feng J Q and Hays D A 2000 J. Imaging Sci. Technol., 44 19–25
[9] Jaworek A and Krupa A 1989 J. Electrostatics, 23 361–370
[10] Adamiak K, Krupa A and Jaworek A 1995 Electrostatics 1995 Inst. Phys. Conf. Ser. 143
(Bristol and Philadelphia: IOP Publishing) 275–278

83
Optimizing the process parameters of injection

moulding to minimize the static charge of
polypropylene test rod
M Murtomaa1,*, S Kankaanpää2, J Nurmio3, M Leino3, J Mäkelä3, P

Järvelä2, E Laine1, V-P Lehto3
1
Department of Physics, University of Turku, FIN-20014 Turku, Finland
2
Laboratory of Plastics and Elastomer Technology, Tampere University of
Technology, FIN-33101 Tampere, Finland
3
Focus Inhalation, PO BOX 980, FIN-20101 Turku, Finland
Abstract. Electrostatic charges of injection moulded test rods have been studied experimentally.
Charges were generated during the injection moulding process and were measured using a
Faraday pail. Several process parameters such as feeding barrel and mould temperatures,
cooling time, injection speed and holding pressure were varied. Effect of these parameters on
the charge were assessed using a Taguchi matrix. Electric field measurements showed that the
plastic components were uniformly charged. Charge decay measurements were performed to
monitor the charge dissipation. Physical dimensions were measured to ensure the usability of
the products. Samples were also characterized by differential scanning calorimetry. Results
prove that a significant reduction in the static charge was achieved by optimizing the process
parameters using the Taguchi matrix.
1. Introduction
In the injection moulding process, the manufactured plastic components may carry a
significant amount of charge which has undesirable consequences in some applications
e.g. in dry powder inhalers (DPI’s). The charging is believed to be due to the separation of
the double layer formed at the metal/polymer interface [1], or due to slip phenomenon [2,
3]. In high resistivity polymers, the charge may lie in the bulk of the material which may
prevent typical neutralizing procedures such as use of ionizers. Instead of neutralizing the
injection moulded components, the process parameters were adjusted to minimize the
charge of the product. Electrostatic charges on manufactured polypropylene test rods were
measured using a Faraday cup. By adjusting several process parameters such as feeding
barrel and mould temperatures, cooling time, injection speed and holding pressure the
generated charge on the polypropylene test rod could be reduced ten times compared to the
charge values obtained with the initial process parameters. Taguchi matrix was used in
assessing the effect of separate process parameters on the generated charge.

84
Table 1. Adjusted process parameters and their values
2. Materials and methods

Polypropylene (PP White 1491/357020, homopolymer, isotactic, BP Amoco) test rods
were manufactured using Fanuc Roboshot α C30 (Fanuc Ltd. Japan), a fully electric injection
moulding machine with 300 kN clamping force. The screw diameter of the machine was 20
mm with an available maximum injection pressure of 200 MPa and an injection speed of
300 mm/s. In the injection moulding process, plastic pellets are fed from a hopper into a
barrel where a screw pushes the material towards a nozzle. From the nozzle the material is
injected into the mould with a pressure created by the screw. The barrel is heated but most
of the temperature rise in the material comes from interaction and friction with the screw.
After injection, the screw pushes the material into the mould with constant pressure, called
holding pressure, to reduce shrinking effects during cooling. After holding pressure the
screw rotates and moves backwards to prepare a new dosage for injection. In the mould the
material is cooled for a given time. After cooling, the back mould moves and the piece is
pushed away from it. The mould closes again and the cycle continues (Fig. 1). The dimensions
of the test rod were approximately: length 100 mm, width 20 mm (at ends)/10mm (at center)
and thickness 0.8 mm (Fig. 2). Five injection moulding parameters were alternated at two
levels. Parameters and their values are shown in Table 1. Barrel temperature means the
temperature in the nozzle. Mould temperature means the temperature setting in the mould.
During holding pressure the melt is pushed slowly forward with constant pressure to fill the
mould completely. Cooling time is the time the injected material is cooled in the
mould.
Fig. 1. Three stages of injection moulding: injection, plastication (feeding), ejection [4].

85
Fig. 2. Test rod (T), cutting point (C) and sprue (S).
In order to stabilize the temperatures and other process parameters, at least 15 pieces were
manufactured before electrostatic measurements were carried out. After cooling time, test
rods together with the sprues dropped approximately 40 cm onto an earthed aluminum foil.
The piece was then picked up at the sprue and the test rod was cut off with a cutter so that
the rod fell into a Faraday pail for charge measurement. Faraday pail was connected to an
electrometer (Keithley 6514, Keithley Instruments Inc., USA). Minimum of 10 measurements
were performed for each process parameter combination. Thermoanalytical parameters,
such as the glass transition temperature (Tg~”30 °C), the corresponding change in the
specific heat capacity (∆Cp~0.1 J(g °C)”1, the melting peak temperature (Tm~166 °C) and
the melting enthalpy (∆Hm~86 Jg”1) were evaluated under N2 and O2 atmosphere with
differential scanning calorimeters (DSC) Pyrisl (Perkin-Elmer Co., USA) and DSC7 (Perkin-
Elmer Co., USA), respectively.
3. Results
3.1. Charge after moulding
Since changing only one parameter at the time would result in 25 measurements, a different
approach was chosen. Using a Taguchi matrix the number of experiments could be reduced
to eight. One additional measurement was performed for verification. Figure 3 represents
the measured mean charges of ten test rods manufactured with parameters which are presented
in Table 2. From figure 3 it can be seen that the static charges of the test rods were
significantly higher in runs 1 to 4 (–9.57 nC to –12.37 nC) compared to runs 5 to 8 (–1.57
nC to –2.94 nC). On the basis of figure 3, one can already assume that the temperature of the
barrel has the most significant effect on the charge. From run 1 to 4 the barrel temperature
was 180 °C and from 5 to 8 it was 210 °C.
Fig. 3. Mean charges of test rods manufactured with different process parameters according to the
Taguchi matrix.

86
Table 2. Taguchi matrix showing the process parameters used in each run (1 to 8)
Using the charge values shown in Fig. 3 and in Table 2, the response table (Table 3) was
calculated. As expected, Table 3 shows that Tbar had the most significant effect on the
charge. Increasing the temperature from 180 °C to 210 °C reduced the charge more than 9
nC. The cooling time had second highest effect. When the cooling time was increased from
5 s to 15 s charge was reduced approximately 1.3 nC. Optimal process parameters for
manufacturing pieces with low electrostatic charge would be: Tbar 2, Tmould 1, vfeed 1, phold 2
and tcool 2. However, Tmould, vfeed and phold had relative small effect and they can be considered
negligible.
There was supposed to be some charge transfer between the cutter and the test rod. This
effect was tested by neutralizing several moulds with AC ionizer and then separating the
test rod and the sprue. Transferred charge from the cutter to the test rod was found to be -
0.16±0.02 nC. This value was considered small compared to the charge generated during
the injection moulding.
To be able to determine whether the charge was mainly surface charge or bulk charge,
additional tests were performed. In these experiments, the specific charge (charge divided
by mass) of both the test rod and the sprue were measured. Results showed that the specific
charge was similar for both pieces. The test rod earned 8 % smaller specific charge than the
sprue. Because the surface areas of the pieces were different (approximately: test rod: 27
cm2, sprue: 18 cm2) it could be concluded that the charge was mainly bulk charge.
Table 3. The response table

87
3.2. Electric field measurements

After the first Faraday pail measurements, test rods were placed on a neutralized support
and the electric field was measured from both ends of the test rod and from both sides at the
middle of the rod. Measurement was performed using JCI 140 Static Monitor (John Chubb
Instrumentation, UK) the distance to the test rod being 20 mm. Measurements showed that
the negative charge was uniformly distributed and thus, no polarization occurred.
3.3. Charge decay

After measurements, the test rods were placed on a table covered with paper in the way that
all pieces which were made with same parameters were piled on top of each other at inclined
angles. There were two reasons for this: first, additional charging due to contacts with the
paper were minimized and secondly, contact area between the rods was minimized although
a contact with the same material yields only small charge. The charge decay was measured
with Faraday pail after one day, four days, three weeks and two months (runs 1 to 4) of
storage at 21 °C and 30 RH%. These measurements are presented in figure 4.
Figure 4 shows that the charge decayed quite rapidly. After one day approximately half
of the charge had decayed. During the next three days all samples carried less than 2 nC of
charge. From figure 4, it can also be concluded that rods which charged most in the injection
moulding process (runs 1 to 4) carried higher charge after one day than the initial charge of
the ones which were manufactured with higher barrel temperature (runs 5 to 8).
3.4. Physical dimensions

To study the usability of the final products the mass and dimensions were also measured
and these mean values are presented in Table 4. Table 4 proves that no significant variations
were observed due to different process parameters.
Fig. 4. Charge of the test rods after moulding, after one day, four days, three weeks and two months of
storage.

88
Table 4. Mean values of mass and length of the test rods produced with different parameters
3.5. Thermoanalytical studies

The determined thermoanalytical parameters did not exhibit logical trends compared to
the initial charges and the charge decay results. This might be due to large deviations of the
detemined values covering the possible small changes. However, the measurements made
under O2 revealed that the used PP grade undergoes exothermic oxidation starting at
210 °C.
4. Conclusions
It was clearly noticed that the barrel temperature i.e. the melt temperature had the most
significant effect on the generated charge. This is most likely due to the fact that when the
temperature was higher, the viscosity was reduced, and the frictional energy involved in
the contact between the melt and the screw was smaller. Thus, the generated charge was
reduced when the temperature was increased. Also, the evident formation of oxygen rich
surface layer may contribute to the smaller degree of generated charge above 210 °C.
Cooling time had the second highest effect on the charge. When the cooling time was
increased, the injection moulded piece stayed in close contact with the mould for a longer
time and, thus, there was more time for the charge to migrate from the sample to the mould.
Because the specific charges of the test rod and the sprue were similar, and also because the
surface areas were clearly different, it could be concluded that the charge which was generated
mainly in the barrel and the screw was located in the bulk of the test rods.
Electric field measurements showed that the samples were not polarized and that the
charge was uniformly distributed. Charge decay measurements showed that after four days
of storage all samples carried less than 2 nC of (negative) charge. Physical dimensions of
the samples were also monitored to ensure the usability of the products and the results
showed no great variations between different batches. Thermoanalytical studies showed no
clear relationship between the charge, but the observed oxidation at 210 °C might give a
cause for further studies. Finally, results prove that a significant reduction in the static
charge was achieved by optimizing the process parameters using the Taguchi matrix.
References
[1] Taylor D M, Lewis T J and Williams T P T 1974 J. Phys. D: Appl. Phys. 7 1756
[2] Pérez-Gonzáles J and Denn M M 2001 Ind. Eng. Chem. Res. 40 4309
[3] Pérez-Gonzáles J 2001 J. Rheol. 45 845
[4] Johannaber F 1985 Injection Molding Machines; A User’s Guide, 2nd Edition (München,
Germany: Hanser Publishers)

89
Bio-nanotechnology of DNA
based on electrostatic manipulation
Masao Washizu
Department of Mechanical Engineering, The University of Tokyo
7–3–1 Kongo, Bunkyo-ku, Tokyo 113–8656, Japan
Abstract. With the use of electrokinetic effects in micro fabricated electrodes, such as
electrostatic orientation and dielectrophoresis, DNA molecules can be stretched and immobilized
onto a predetermined location on a solid surface. Once done, we can make analyse DNA with
spatial resolution, or perform operations to a targetted position on a targetted molecule. We
have demonstrated that 1) a desired portion of stretched DNA can be mechanically dissected,
picked up, and amplified, 2) enzymatic cutting at a targetted position can be made by pressing
an enzyme-immobilized probe on to a stretched DNA, 3) observation of DNA-protein interaction
can be performed in real-time with an optical microscope. We have also shown that large fluid
shear created by electroosmosis is effective in stretching long (~M base) DNA.
1. Introduction
DNA is a long string-like molecule having a diameter of 2 nm in the double helical structure
of most duplex DNA, and the bases are stacked between the sugar-phosphate backbones at
a spacing of 0.34 nm. Genetic information is recorded in the sequence of bases, i.e. by type
of bases as a function of position.
In conventional biochemistry, DNA is treated as a mass in a solution, and there is no
“spatial resolution”, i.e. no direct access to a particular molecule, nor to a particular position
on the molecule. An example of the difficulties arising from the lack of spatial resolution is
found in “shot-gun” sequencing, which is commonly used for the determination of base
sequence. In this method, DNA is first cut into fragments using restriction enzymes, to a size
small enough that can be analyzed by electrophoresis. During this process, the fragments
are mixed up, and the position of the fragment in the original long DNA strand becomes
untraceable. As a result, the total sequence must be inferred using overlaps among each
fragment, but this process becomes increasingly difficult as the number of fragment increases.
If we can do the sequencing knowing where the fragment comes from, this tedious re-
construction process can be omitted.
In order to realize the analysis which uses spatial information, a method to manipulate
individual molecule is required. The method we have developed is based on electrokinetic
effects [1–2] in micro fabricated structures.

90
2. Electrostatic stretch-and-positioning of DNA

A device for stretching DNA is simply a pair of metal electrodes vacuum-evaporated and
patterned on a glass substrate. DNA solution is fed on the device, which is covered with a
cover slip. As shown in fig. 1a), DNA in water takes randomly coiled conformation due to
thermal agitation. When the electrodes are energized to 1 MV/m (e.g. 100 V across 100
micron) by a high frequency (1MHz) power supply (fig. 1b), DNA polarizes, and the
electrostatic orientation occurs in every part of the DNA strand to stretch DNA to a straight
shape, parallel to the field (fig. 1c). Then dielectrophoresis pulls the strand towards the
electrode edge where the field is most intense, until one molecular end touches the electrode
(fig. 1d, where induced charges have been omitted for clarity). The stretching occurs in a
fraction of a second, and the whole process completes within a few seconds. Thus we have
DNA stretched and aligned with one end anchored on the electrode edge. We named the
process “electrostatic stretch-and-positioning of DNA” [3–4].
A few comments on the method are
1) The DNA end is permanently anchored when an electrochemically active metal is
used as the electrode. Aluminum works better than gold or platinum. Fresh aluminum
works better than old (surface oxidized) aluminum. Old aluminum restores
anchoring ability when it is washed by acid. From these facts, we tentatively
assume that the anchoring is by a covalent bond electrochemically formed with
the metal.
2) When stretching DNA, we need a cover slip to limit the water film thickness to 10–
20 µm, in order to minimize liquid flow caused by the applied field (probably by
charge injection).
3) Mechanism of DNA polarization is that by counter-ion [3].
4) Creation of such a high-intensity field is possible in micro-fabricated electrodes
where the surface-to-volume ratio of the high field region is large, and thus the
temperature rise due to Joule heating can be minimized. Even so, the conductivity
of the medium should be kept as low as possible. We use deionized water, or ImM
buffer concentration.
5) The frequency of the applied field is chosen to be around 1MHz, as a compromise
between electrolysis at lower frequency and reduced DNA polarization at higher
frequency.
Fig. 1 Electrostatic stretch-and-positioning of DNA

91
3. Dissection and acquisition of a targetted portion of DNA

Targetted position of the stretch-and-positioned DNA can be dissected mechanically using
a sharp stylus as a cutting knife. However, picking up the dissected fragment is more
difficult. For this purpose, we have developed a method based on sacrificial etching [5],
which is schematically depicted in fig.2. The device consists of a glass substrate, onto
which a sacrificial layer (made of alcohol-soluble positive photoresist), DNA carrier layer
(gelatin), and a pair of electrodes are deposited. DNA is stretched and immobilized onto the
carrier layer with one of its molecular ends aligned on the electrode edge. Using an AFM
stylus as a knife, the targetted portion of the DNA together with the carrier layer is dissected.
By dissolving the sacrificial layer, the DNA fragment on the piece of carrier is recovered
onto a membrane filter. The carrier piece is then melted to obtain DNA fragments in solution.
Fig.2 Dissection and acquisition of a targetted portion of DNA
An experimental demonstration is performed using λDNA as the sample. In order to prove

that the desired position is dissected and picked-up, three primers are prepared, as depicted
in fig.3 a). They correspond respectively to the sequences near the left end of λDNA (denoted
L), near the right end (R), and approximately at the center (C), all about 1kb in length.
Because DNA is electrically symmetrical, the electrostatic positioning yields a mixture of
one orientation and the other, as shown in fig.3 b). If aligned λDNA is cut, say at 3 µm from
the end, and successfully picked up, the PCR product should contain the sequence L and R,
but not C (fig. 3 b). On the other hand, if the sequence C is detected by PCR, it is an
indication that unwanted DNA fragments are coming in.
Fig.3 c) shows the result of electrophoresis of the PCR product. Lanes C1–C3 are positive
control, starting from 10, 100 and 1000 non-cut λDNA molecules respectively. In these
lanes, the three bands, from top to bottom L, R, and C, are seen. Lanes #1 through #5 are
PCR of the carrier piece. All show the bands from L and R, and not C.
These results demonstrate the successful mechanical dissection and recovery of DNA.
From the brightness of the bands in comparison with that of C3, the number of dissected

92
molecules is estimated to be several thousand per run, which is adequate for use in
biochemical processes, such as sequencing.
Fig. 3 Dissection and acquisition of DNA—experimental result
4. Floating-potential electrode for end-anchoring of DNA

In the mechanical dissection of DNA described above, the molecular structure of the cut
end is unpredictable, or even might be damaged. In biochemistry, DNA cutting is usually
made by an enzyme, which gives a predetermined cutting end depending on the type of
enzyme. However, as DNA and the enzyme are mixed in water solution, the cutting is by
chance, and we cannot specify the cutting position.
In order to realize cutting with a defined molecular position at a targetted position, we
have developed a method that we call “molecular surgery of DNA” [6]. The enzyme is
immobilized on a microparticle having a diameter of 1 µm. The particle is grasped with
laser tweezers under a microscope, and is pressed against the immobilized DNA. By doing
so, one can specify the location of the enzymatic reaction, namely at the contact point.
A special DNA immobilization method had to be developed in order to allow free
interaction between DNA and enzymes. If DNA is adsorbed onto a solid surface, the surface
hinders the proper approach of the enzyme, and the enzyme cannot react. The developed
system consists of a pair of energizing electrodes on a glass substrate, and a few thin strips
of aluminum having no electrical connection, which we call floatingpotential electrodes
(FPE) (fig.4) [6]. The spacing between the floating-potential electrodes is made slightly
smaller than the length of DNA to be immobilized, and the glass surface between the
electrodes is etched down about l˜m. When the outer electrodes are energized, DNA is
stretched and pulled towards the edge of the electrodes. When one molecular end of DNA
touches an electrode, it is anchored. At this moment, the other end extends to the edge of
the adjacent electrode, and is dielectrophoretically pulled-in to be anchored. Because the
glass surface is lower, DNA is held free except at both ends. We hereafter call such DNA
“end-anchored DNA”.
The function of the FPE here is to deform the electrostatic field to create field maxima for
DEP trapping of DNA. These electrodes are better left at a floating potential. When the
electrodes are connected to a power supply with low impedance, charge injection creates

93
jet-like flow at the vicinity of the electrode edge, which hampers the approach of DNA
ends.
Fig.4 c) is a photograph of DNA bridging over the FPE’s. Here, a flow of medium, as
shown by the arrow, is intentionally created, by which DNA strands are bent downward.
This proves that the middle part of the DNA is not adsorbed on the solid surface, and that
the anchoring is strong enough to withstand the hydrodynamic drag.
Fig.4 Floating-potential electrode system for immobilization of DNA
By using this electrode system, molecular surgery of DNA with enzymeimmobilized

particle has been demonstrated [5]. First, a particle having no enzyme is pressed against
end-anchored DNA, which is used as a reference to determine if the DNA cutting is
mechanical or enzymatic. In this case, DNA has to be elongated to 1.5 times its original
length before mechanical breakage occurs. When, DNasel, the enzyme which cuts DNA
regardless of the base sequence, is immobilized on the particle, cutting occurred as soon as
the particle touched the DNA. When a restriction enzyme is used (specifically the 6-base
cutter Hind III in the experiment), contact with DNA does not always result in cutting. This
is as expected, because the particle is most likely to hit locations other than restriction sites.
In this case, the particle is moved along DNA keeping contact. The particle apparently
overlooks some of the restriction sites, but if the scanning is repeated for a few times, DNA
is finally cut, at locations in agreement with the restriction site.
The floating-potential electrode system also provides a powerful method for the dynamic
observation of the DNA-protein interactions in single-molecule level. Passive motion of
enzymes such as restriction enzymes along DNA strands [7], as well as the active motion of
RNA polymerase as it synthesizes RNA [8], have been reported.

94
5. Controlled stretching of longer DNA by electroosmotic flow

The electrostatic stretching of DNA uses a field of 1MV/m at 1MHz. The length of DNA is
0.34 nm per base, so, if 3 million base DNA is to be stretched for example, its length is 1 mm,
and the voltage required will be 1kV, which is impractical. We are developing a DNA
stretching method alternative to that based on electrostatic orientation. It uses d.c.
electroosmotic flow, which is a liquid streaming induced by ion drag in the electrical
double layer at a solid/liquid interface. The double layer is only several nm in thickness, so
very high velocity shear can be created, which can efficiently stretch fiber-like objects, as
depicted in fig.5a). Fig.5b) is a photo of DNA coming out from yeast cells, which is stretched
as long as 200µm [9]. We are now developing applications for optical gene mapping.
Fig.5 Stretching DNA using electroosmotic flow
Acknowledgements
The author would like to thank Dr. Hiroyuki Kabata of Tokyo Univ., Mr. Osamu Kurosawa
of Advance Co., Mr. Yasuhiro Nikaido of Kyoto Univ., and Dr. Takatoki Yamamoto of
Tokyo Univ., for collaborations and discussions. This work is in part supported by BRAIN
(Seiken-Kiko) Research and Development Program for New Bio-industry Initiatives, NEDO
(Proposal-Based R&D Program 97S07–005), the Ministry of Education (Kakenhi
11450103, 12555076, 12030216), Micromachine Center, and Advance Co.
References
[1] H.A.Pohl, “Dielectrophoresis”, Cambridge University Press (1978)
[2] T.B.Jones, “Electromechanics of Particles”, Cambridge University Press (1995)
[3] M.Washizu and O.Kurosawa: “Electrostatic Manipulation of DNA in Microfabricated Structures”,
IEEE Trans. IA, Vol.26, No.6, p.l 165–1172 (1990)
[4] M.Washizu, O.Kurosawa, I.Arai, S.Suzuki and N.Shimamoto,: “Applications of Electrostatic
Stretch-and-positioning of DNA”, IEEE Trans. IA. Vol.31, No.3, p.447–456 (1995)
[5] O.Kurosawa, K.Okabe, and M.Washizu: “DNA analysis based on physical manipulation”,
Proceedings of the thirteenth annual international conference on Micro Electro Mechanical Systems
(MEMS2000), p.311–316 (2000)
[6] T.Yamamoto, O.Kurosawa, H.Kabata, N.Shimamoto, and M.Washizu: “Molecular surgery of
DNA based on electrostatic micromanipulation”, IEEE Trans. IA, Vol.36, No.4, p.1010–1017
(2000)
[7] H.Kabata, W.Okada, and M.Washizu: “Single-Molecule Dynamics of the Eco RI Enzyme using
Stretched DNA: Its Application to In Situ Sliding Assay and Optical DNA Mapping”, Jpn. J.
Appl. Phys. Vol.39, p.7 164–7111 (2000)
[8] H.Kabata, S. Matsumoto and M. Washizu,: “Real-Time Imaging of Translocation of Elongating
T7 RNA Polymerase along DNA”, The 7th Asian Conf. on Transcription (ACTVII), p.8,
no.7008 (2002)
[9] M. Washizu, Y.Nikaido, O.Kurosawa, and H.Kabata: “Stretching Yeast Chromosomes Using
Electroosmotic Flow”, to appear in the Journal of Electrostatics.

95
Electrorotation of dense colloidal suspension
K.W.Yu, J.P.Huang, G.Q.Gu

Department of Physics, The Chinese University of Hong Kong, Shatin, NT,
Hong Kong
When a strong electric field is applied to a colloidal or biological cells suspension, the induced
dipole force will cause an aggregation of the suspended particles in response to the field. In the
case of a rotating electric field, the electrorotation (ER) spectrum can be modified due to the
local field effects arising from the manyparticle systems. To capture the local field effect, we
invoke the Maxwell-Gar net t approximation for the dielectric response. The hydrodynamic
interactions between the suspended particles will modify the spin friction, which is a key to
determine the angular velocity of ER. By invoking the spectral representation approach, we
derive the analytic expressions for the characteristic frequency at which the maximum angular
velocity of ER occurs. We report a detailed investigation of the dependence of the characteristic
frequency and the dispersion strength of ER on various material parameters.
Electrorotation (ER) is the movement of dielectric particles subjected to rotating ac electric

fields [1–3]. The ER phenomenon has been developed into a sensitive tool for noninvasive
studies of a broad class of micr op articles, ranging from single living cells, spores, seeds to
synthetic materials [1, 4–8]. Existing theoretical analysis has been limited to the dilute
limit, in which we can focus on the ER behavior of individual particles by ignoring the
mutual interactions between the particles. However, if the suspension is nondilute, as in the
case of aggregation in a strong electric field, one may not ignore the interactions. As an
initial model, we considered a single pair of touching particles dispersed in a suspension.
The mutual polarization interaction between the particles was captured by the multiple
image method [9]. The results showed that the ER spectrum can differ substantially from
that of isolated particles.
In reality, a dense suspension is often encountered. In this case, the many-body (local-
field) effect can be dominant, which can also change the ER spectrum. Moreover, when we
are measuring the ER spectrum of individual particles, each of which can have a different
local environment. For instance, two particles can approach due to an attractive force
between them. Hence, it is more interesting to include the many-body effects on the ER
spectrum of two approaching particles.
Case I: A pair of touching particles

We first consider an isolated spherical particle with diameter D of complex dielectric
constant dispersed in a suspension medium of where . Here and σ
represents the real dielectric constant and conductivity, f the frequency of external rotating
field, and In this case, the dipole factor b of an isolated particle should be
. To discuss the effect of the multipolar interaction, we consider a

96
pair of touching spherical particles with separation R suspended in a medium. In this case,
based on the multiple image method [9], the effective dipole factor of the pair is given by
[10] where the dipole factor for longitudinal and transverse field cases are
respectively:
Here α satisfies the relation cosh α=R/D.

In the presence of a rotating electric field, the angular velocity of ER (Ω) of a particle is
determined by [10, 11]:
(1)
where F(2, η, E0) is a function of 2 the viscosity η of the medium, as well as the field
magnitude E0. Here Im[…] denotes the imaginary part of […]. For an isolated spherical
particle,
For clarity, we represent b* in the spectral representation [10]. Then we have an exact
transformation in terms of a series of subdispersions [10]:
(2)
where
Here δm (T) (or δm (L)) and fmc(T) or (fmc(L)) are the dispersion magnitude and the characteristic
frequency of the transverse (longitudinal) field case, respectively, in terms of the structure
and material parameters [10, 12].
Case II: A dense suspension
In a dense suspension, the effective dielectric constant of the whole system is given by
the Maxwell-Garnett theory:
(3)
where p is the volume fraction of particles. Note that p=0 denotes the isolated-particle
result. In this system, the dipole factor of the particle is given by: .
In order to obtain the analytic expression for the characteristic frequency, we again resort to
the spectral representation approach, and we have
(4)
where the parameters F1, F2, s1, s2, ∆1, ∆2, fc1, fc2, can be expressed as analytic expressions
of the structure parameter (volume fraction) and various materials parameters. For more
details, we refer the reader to Ref.[13].
Regarding the effective viscosity ηe of the whole system, we apply an analogous
MaxwellGarnett approximation [14]:
(5)

97
FIG. 1: (Case I) The minus dipole factor plotted vs frequency both for an isolated particle and
touching particles
–4
for three different separation ratios R/D at 5=1.1, t=-1/900, 2=80 0, and
σ2= 2.8×10 S/m.
where η1 and η2 are respectively the viscosity of the particle and the host. For hard spheres,
we may take η1 → ∞.
Therefore, for a dense suspension, the angular velocity of ER (Ω) of a particle should be
given by [11, 13]:
(6)
Case III: A pair of touching particles in a dense suspension

For a pair of touching particles in many-particle system, the multiple image formulae for
the effective dipole factors are [15]
for transverse and longitudinal field cases, respectively. In the case of a rotating electric
field, the angular velocity of ER (Ω) for a pair should be [11]:
(7)
with b*=(bT*+bL*)/2. This equation is indeed a nontrivial result since it takes into account
both the multiple image effect and the many-body (local-field) effect.
Numerical results
In Fig.1 (Case I), as R/DH≈1 (e.g., R/D=1.0333), the deviation between the isolated and
touching cases is evident. In other words, the multipolar interaction does play an important
role in the ER spectrum. However, as the separation increases, (say, R/D>2) the effect of the
multipolar interaction may be neglected.
In Fig.2 (Case I), we plot the spectral representation for the two sets of poles
for m=1 to 100. As m increases, increases up to 1/3, while decreases towards 1/3. For
small R/D ratio (say, R/D=1.0333), the longitudinal field plays a more important role in
determining ER spectrum than the transverse field.

98
FIG. 2: (Case I) The pole spectrum for two approaching spheres for several separation ratio R/D for the
longitudinal (open symbols) and transverse (filled symbols) cases. Note that the longitudinal (transverse)
part of the spectrum are on the small (large) sm side of the spectrum. The lines are guides for the eyes.
FIG. 3: (Case II) The spectral parameters F1, F2, s1 and s2, as well as the dispersion parameters,
∆1, and ∆2, plotted versus volume fraction p. Parameters: 2 = 800, 1 = 100 = σ2 = 2.8 × 10–4 S/m,
and σ1 = 2.8 × 10–2 S/m.
FIG. 4: (Case II) The rotating speed versus frequency for different volume fraction p. Parameters:
1 = 100, σ1 = 2.8×10–2 S/m, σ2 = 2.8×10–4 S/m, 2 = 800, E0 = 10kV/m, η2 = 1.0×10–3 kg/(ms).

99
FIG. 5: (Case III) The ER spectrum of a pair of particles (PP) in an effective medium at R/D= 1.033, for
different volume fractions. For comparison, the ER spectrum for a single particle (SP) in an effective
medium is also plotted. Particularly, p=0.00(SP) denotes a single particle in a pure host fluid.
FIG. 6: (Case III) The ER spectrum of a pair of particles in an effective medium at p=0.15, for different
R/D. Note that SP(p=0.00) denotes a single particle in a pure host fluid while SP(p=0.15) in an effective
medium with volume fraction p=0.15.
In Fig.3 (Case II), it is concluded that at low volume fraction the rotating-speed peaks
related to two characteristic frequencies are so close that they can almost be seen as
overlapped. This hence leads to one rotation peak in Fig.4.
Fig.4 (Case II) is plotted for three different volume fraction p. We find that increasing
volume fraction is able to reduce the characteristic frequency and the rotating speed peak.
In Fig. 5 (Case III), we investigate the ER spectrum of a pair of particles (PP) in an
effective medium at R/D=1.033, for different volume fractions. The ER spectrum for a
single particle (SP) in an effective medium is also plotted. Here p=0.00(SP) denotes a single
particle in a pure host fluid. Two ER peaks always appear for a pair of particles in an
effective medium. The peak located at lower characteristic frequency is due to the multiple
image effect. Hence, for a many-particle system, the multipolar interaction also plays an
important role in the ER spectrum. Additionally, increasing the volume fraction may redshift
the peak. Moreover, we find that multiple images have an effect on ER spectrum for the low
frequency region only. For the high frequency region, such effect will be small enough to
be neglected. For the whole frequency range, the many-body (local-field) effect plays a
role. Therefore, the multiple image effect may change the ER spectrum of a pair of touching
particles while the local-field effect may offer a correction.
In Fig.6 (Case III), we investigate the ER spectrum of a pair of particles in an effective
medium at p=0.15, for different R/D. Here SP(p=0.00) denotes a single particle in a pure
host fluid while SP(p=0.15) in an effective medium. Obviously, at large separations, the
multiple image effect may still be neglected even though for many-particle system.

100
To sum up, we investigate the ER spectrum of the particle by taking into account the
multiple image effect, as well as many-body (local-field) effect. In view of the relationship
between electrorotation and dielectrophoresis [16, 17], it is straightforward to extend our
work to deal with the dielectrophoretic spectrum. For the fitting of experimental findings,
we refer the readers to two recent papers [18, 19].
Acknowledgments
This work was supported by the Research Grants Council of the Hong Kong SAR Government
under project number CUHK 4245/01P.
[1] Arnold W M and Zimmermann U 1982 Z. Naturforsch. 37 908
[2] Pastushenko V P, Kuzmin P I, Chizmadshev A Y 1985 Stud. Biophys. 110 51
[3] Gimsa J, Muller T, Schnelle T and Fuhr G 1996 Biophys. J. 71 495
[4] Fuhr G, Zimmermann U and Shirley S 1996 Electromanipulation of cells (ed U. Zimmermann
and G.A. Neil Boca Raton, FL: CRC Press) pp 259–328.
[5] Yang J, Huang Y, Wang X, Wang X B, Becker F and Gascoyne P R C 1999 Biophys. J. 76 3307
[6] Becker F F, Wang X B, Huang Y, Pethig R, Vykoukal J and Gascoyne P R C 1995 Proc. Natl.
Acad. Sci. USA 92 860
[7] Chan K L, Morgan H, Morgan E, Cameron I T and Thomas M R 2000 Biochem. Biophys. Acta
1500, 313
[8] Huang J P and Yu K W 2002 J. Phys.: Condens. Matter 14, 1213
[9] Yu K W and Wan J T K 2000 Comput. Phys. Comrmm. 129, 177
[10] Huang J P, Yu K W and Gn G Q 2002 Phys. Rev. E 65, 021401
[11] Jones T B 1995 Electro-mechanics of Particles (Cambridge University Press, Cambridge)
[12] Lei J, Wan J T K, Yu K W and Sun H 2001 Phys. Rev. E 64, 012903
[13] Huang J P, Yu K W and Gu G Q 2002 Phys. Lett. A 300, 385
[14] Choy T C 1995 Physica A 221, 263
[15] Huang J P and Yu K W 2002 cond-mat/0209505
[16] Wang X B, Pethig R and Jones T B 1992 J. Phys. D: Appl. Phys. 25, 905
[17] Gao L, Huang J P and Yu K W 2003 Phys. Rev. E 67, 021910
[18] Huang J P, Karttunen M, Yu K W and Dong L 2003 Phys. Rev. E 67, 021403
[19] Huang J P, Yu K W, Gu G Q and Karttunen M cond-mat/0212518

101
The effect of disaccharides on the transport of lipophilic

ions in cell membranes studied by electrorotation
R Reuss, M Horbaschek, J M Endter, U Zimmermaim and
V L Sukhorukov
Lehrstuhl für Biotechnologie, Biozentrum der Universität Würzburg, Am
Hubland, D-97074 Würzburg, Germany
Abstract. The disaccharide trehalose, a natural cryoprotectant, is increasingly being exploited

in biomedicine as a very efficient stabilizer of frozen and dry macromolecules, membranes and
whole cells. Valuable insight into the mechanisms by which trehalose protects cells can be
obtained by studying its effects on the electrical and ion transport properties of cell membranes.
In the present study, the binding and translocation of the lipophilic anion dipicrylamine (DPA)
across the membrane of Jurkat lymphocytes (suspended in hypotonic trehalose- or sucrose-
substituted media) were determined by means of electrorotation. From the rotation spectra of
individual cells, not only the passive electrical properties of cell compartments but also the
area-specific concentration of DPA adsorbed to the plasma membrane and its translocation
rate constants were evaluated. The substitution of sucrose by trehalose increased significantly
both the plasma membrane capacitance Cm and the adsorption of DPA to the plasma membrane,
whereas the translocation rate of DPA across the membrane was slightly reduced. The high Cm
value indicates that hypotonic stress did not cause any noticeable loss of membrane folds and
microvilli in the presence of trehalose. The observed changes in the transport of DPA are
consistent with the assumption that trehalose increased the dipole potential of the plasma
membrane.
1. Introduction
Trehalose, a disaccharide of glucose, is found at high concentrations in several organisms
that are capable of withstanding various environmental stress conditions. Yeasts, plants and
some animals naturally synthesize trehalose, which protects cells during extreme dehydration
and cold (de-Araujo, 1996; Crowe and Crowe, 2000). Trehalose has also gained common
attention in biotechnology because of its extraordinary ability to maintain the structural
integrity of frozen and/or dry biomolecules, artificial and natural membranes, and even
whole cells (Woelders et al., 1997; Eroglu et al., 2000; Guo et al., 2000).
Most studies of the molecular mechanisms of the trehalose mediated cryo- and
lyoprotection were performed with model lipid membranes, including liposomes, lipid
bilayers and monolayers (Oliver et al., 1998; Wolfe and Bryant, 1999; Luzardo et al., 2000).
It has been suggested that trehalose mechanically stabilizes frozen and dry membranes via
specific interactions with the phospholipid head groups of the bilayer, by formation of
hydrogen bonds and/or by water replacement (Tsvetkova et al, 1998; Lambruschini et al.,

102
2000). In contrast to artificial membrane systems, relatively little is known about the
interactions of trehalose with cell membranes.
The electrorotation technique (ROT) and related methods (Arnold and Zimmermann,
1982; Jones, 1995; Fuhr et al., 1996; Pethig and Markx, 1997; Sukhorukov et al., 2001;
Hughes et al., 2002) offer the opportunity to study the effects of trehalose and other sugars
on the electrical membrane properties in living cells. For ROT measurements, cells are
freely suspended in a liquid medium whose conductivity and osmolality can be varied over
wide ranges by modifying its salt and sugar contents. Earlier ROT studies have demonstrated
that this method allows us to monitor subtle structural changes in the cell membrane in
response to small variations of medium composition (Sukhorukov et al., 1993; Wang et al.
1994).
In the present study, the effects of trehalose and sucrose on the electrical and transport
properties of the plasma membrane of Jurkat lymphocytes suspended in hypoosmolar
medium were compared by means of electrorotation. We found that the substitution of
sucrose by trehalose increased significantly both the plasma membrane capacitance Cm and
the adsorption of the lipophilic anion DPA to the plasma membrane, whereas the
translocation rate of DPA across the membrane was slightly reduced. The high Cm values
indicate that hypotonic stress did not cause any noticeable loss of membrane folds and
microvilli in the presence of trehalose. The observed changes in the transport of DPA are
consistent with the assumption that trehalose increased the dipole potential of the plasma
membrane.

Human lymphoid Jurkat cells were cultured at 37°C under 5% CO2, using standard
techniques. Before use, the cells were washed with and resuspended in either low osmotic
trehalose or sucrose medium (200 mOsm). In some experiments, 12µM DPA was added to
cell suspensions. The suspension conductivity (σe) was adjusted to 1–6 mS/mby addition
of HEPES-KOH, pH 7.4.
The measurements of the field frequency fcl inducing fastest anti-field rotation of cells
were performed by the contra-rotating field (CRF) technique that was described in detail
previously (Arnold & Zimmermann, 1988). ROT spectra were measured in a microstructured
four-electrode chamber (Fraunhofer-Institut Siliziumtechnologie, ISiT, Itzehoe, Germany)
arranged as a planar array of circular electrodes of 60 µm diameter, 160 nm thickness and 200
˜m electrode spacing. To produce ROT fields, the adjacent electrodes were driven by four 90°
phase-shifted, symmetrical square-wave signals from a pulse generator HP 8130A (Hewlett-
Packard, Boeblingen, Germany) with 2.5–4.8 VPP amplitude over the frequency range from
100 Hz to 150 MHz. To monitor the suspension conductivity during ROT measurements,
two further microelectrodes were incorporated into the ROT chamber. The experimental
setup used here is similar to that described in detail elsewhere (Gimsa et al., 1996).
The ROT spectra were normalized to the field strength of 1 VPP/100 µm. During the 5–10
min required to take a complete ROT spectrum, the medium conductivity increased by
about 10%. The cells were observed using a BX 50 Olympus microscope (Hamburg,
Germany), that was equipped with a CCD video camera connected to a video monitor. Cell
radii were determined with a calibrated ocular micrometer. The electrical cell properties
and transport kinetics of the lipophilic ions were determined from the ROT spectra of
individual cells by applying the previously reported models (Jones, 1995; Sukhorukov

103
and Zimmermann, 1996; Gimsa et al., 1996; Reuss et al., 2002). Computations were performed
with the Mathematica software.
3. Results
Because the membrane conductance Gm can not be deduced accurately from individual ROT
spectra (Gascoyne et al., 1995), the area-specific membrane properties averaged over a large
number of cells (n=300 cells) were first determined by studying the effect of medium
conductivity σe on the characteristic frequency of anti-field rotation fcl. These measurements
were performed at low σe ranging from 1 to 6 mS/m, using the CRF technique. Strong linear
relationships between fcl×a (where a is the cell radius) and σe was observed in both trehalose-
and sucrose-substituted media (Fig. 1, open and filled circles, respectively). Equation 1 was
used for the estimation of the plasma membrane parameters (Arnold and Zimmermann, 1988):
(1)
From the CRF data of cells suspended in the 200 mOsm sucrose solution (Fig. 1, solid line),
the mean Cm=8.3±0.2 mF/m2 and Gm=75±29 S/m2 were extracted. In trehalose medium,
Jurkat cells showed Cm=12.2±0.4 mF/m2 and Gm=58±35 S/m2.
The ROT spectra of individual cells were monitored at a conductivity ranging between
2 and 3 mS/m (Fig. 2). At this low conductivity, a better resolution of the mobile charges
within the plasma membrane can be achieved (see below). The control ROT spectra possessed
three well-resolved peaks: one anti-field peak at about 20 kHz (fcl), and two co-field peaks
at 20 and 50 MHz. Similar ROT spectra were obtained for Jurkat cells in sucrose medium of
the same conductivity (not shown).
Figure 1. Determination of the plasma membrane Figure 2. ROT spectra of control Jurkat cells
capacitance Cm and conductance Gm of Jurkat and those treated with DPA (open and filled
cells suspended in 200 mOsm sugar solutions circles, respectively), measured in 200 mOsm
containing either trehalose or sucrose as the major trehalose. The curves show least-square fits of
osmolyte. The lines show the best fits of Eq. 1 to the single-shell (1) and mobile charge models (2)
the CRF data. by assuming a cytosolic dispersion.
The control spectra were fitted by the single-shell model by assuming a dispersion of the
cytosol (Fig. 2, smooth curve). For most cells (n=20), the regression coefficient ρ ranged
between 0.98 and 0.99, which implies good agreement of the data with the model. In
performing the nonlinear least-square fits, the Cm and Gm values obtained by the CRF
technique were assumed. The fitted electrical parameters of control Jurkat cells suspended
in both media are summarized in Table 1.

104
Table 1. Effects of the sugar composition on the electrical and ion transport properties of
hypotonically-treated Jurkat cells
a
In isotonic sucrose and trehalose media (i.e., at 300 mOsm) the mean radii were 7.2 and 7.1 µm,
respectively.
b
The ζ-potential was determined in DPA free solutions
c
The mean Cm and Gm values for 300 cells in each medium were determined by the CRF technique (see
Fig 1).
d
Cytosolic and mobile charge parameters were derived from the ROT spectra of individual cells, such
as shown in Fig. 2. Each value represents the mean ± SE from about 40 ROT spectra for either sugar.
σi is the cytosolic conductivity. εi and ∆εi are the cytosolic (highfrequency) permittivity and its increment.
fdisp is the characteristic frequency of cytosolic dispersion. Nt is area-specific concentration of DPA in
the plasma membrane and ki its translocation rate across the membrane.
In contrast to control cells, the ROT spectra of Jurkat cells treated with 12 µM DPA
displayed an additional anti-field peak centered at about 2 kHz (Fig. 2, filled circles). This
low-frequency peak is dominated by the relaxation of the lipophilic anions (mobile charges)
adsorbed to the plasma membrane. The co-field peaks were not affected by DPA. Table 1
summarizes the membrane transport parameters of DPA derived by fitting the mobile charge
model to the ROT spectra (Sukhorukov and Zimmermann, 1996). Judging by the Nt and ki
values, trehalose caused an enhanced binding of DPA and its slower translocation across
the membrane, as compared to sucrose medium.
In addition to ROT measurements, the DC electrophoretic mobility of Jurkat cells was
determined by means of the laser-Doppler velocimetry. Judging by the observed ζ-values
(Table 1), the replacement of sucrose by trehalose led to a less negative surface charge of
hypotonically-treated Jurkat cells.
4. Discussion
The effects of the sugar composition in suspending medium on the membrane of Jurkat
lymphocytes reported here can be put into two categories. These are 1) effects on the
transport of the lipophilic anion DPA, which can be related to the electrostatic boundary
potentials at/within the plasma membrane, and 2) changes in the plasma membrane
capacitance, which reflect alterations in the membrane structures such as folds and
microvilli.

105
4.1. Binding and translocation of lipophilic anions

Judging by the Nt values given in Table 1, the substitution of sucrose by trehalose in the
suspending medium resulted in a nearly twofold increase in the adsorption of DPA to the
cell membrane. The increase of Nt was accompanied by a slight decrease of the translocation
rate ki of the anion. Since the ζ-potential in trehalose-substituted medium was less negative
than in the sucrose solution, changes in the electrostatic surface potential ψS of the plasma
membrane could be responsible for the enhanced binding of lipophilic anions to the
membrane produced by trehalose.
In addition to ψS, lipid bilayers and cell membranes also possess a large boundary
potential (positive inside), the so-called dipole potential ψD, which results from the alignment
of dipolar residues of the lipids and associated water molecules within the membrane
(Clarke, 2001). The ψD changes (ψD) produced by the substitution of sucrose by trehalose
can be deduced from the transport parameters of DPA obtained here (Table 1), by applying
Eq. 2 (Cseh and Benz, 1998):
(2)
where kiS and kiT are the translocation rate constants of the anions across the membrane in
sucrose- (superscript “S”) and trehalose-substituted media (superscript “T”), respectively,
with the corresponding partition coefficients βS=NtS/2c and βT=NtT/2c. The concentration
of the anion in the medium c was kept constant (c=12 µM). Using Eq. 2 and the Nt and ki data
given in Table 1, the ∆ψD value (±SE) of +10.6±1.7 mV was obtained, indicating that the
replacement of sucrose by trehalose led to a small increase of ψD in the plasma membrane of
Jurkat cells.
4.2. Plasma membrane capacitance

Both CRF technique and ROT spectra revealed that the Cm value of Jurkat cells suspended
in trehalose medium differs markedly from the Cm value in sucrose solutions of the same
hypotonic osmolality. Earlier studies have shown that higher Cm values indicate an
increasing number of membrane folds and microvilli (Sukhorukov et al., 1993; Wang et al.,
1994). Thus, at normal osmolality (i.e., about 300 mOsm), most mammalian cells exhibit
Cm values above 10 mF/m2, whereas stretching the membrane osmotically usually causes
reduction to a flat-membrane value of about 7–9 mF/m2 due to the disappearance of microvilli
(Sukhorukov et al., 2001). The “flat-membrane” Cm value of 8.3. mF/m2 (Fig. 1) obtained
here for Jurkat cells in hypotonic sucrose medium is in agreement with the Cm of 7.1 mF/m2
reported earlier for the same cell line in 150 mOsm inositol medium (Sukhorukov et al.,
2001). In contrast to sucrose and inositol solutions, Jurkat lymphocytes suspended in low
osmotic trehalose medium exhibited a significantly higher Cm value (12.2 mF/m2, Fig. 1)
than expected for this osmolality. This means that hypotonic stress in trehalose-substituted
medium did not cause any noticeable loss of membrane folds and microvilli.
Since the mean cell radii in trehalose and sucrose media were similar (7.1 and 7.4 µm,
respectively, Table 1), regulatory volume decrease (i.e., slow shrinkage of cells to their
original isotonic volume, following fast hypotonic swelling (Ross et al., 1994)) seems
unlikely to be involved in the observed increase of Cm in hypotonic trehalose medium.
Studies on artificial lipid membranes have demonstrated that trehalose (but not sucrose)

106
intercalates and acts as a spacer between the phospholipid head groups (Luzardo et al.,
2000). Therefore, binding of trehalose to phospholipid membranes is accompanied by a
lateral expansion of the bilayer, which can explain the higher degree of membrane folding
in trehalose substituted medium observed in the present study.
Whatever the explanation, microvilli are essential for contact of the plasma membrane
with the cytoskeletal web and thus for mechanical stability of cells. Moreover, the presence
of folds and microvilli, as a membrane reserve instantly available to repair membrane
injuries (Sukhorukov et al., 1993), can be the reason why trehalose improves survival of
eukaryotic cells exposed to various experimental and environmental stress conditions,
including cryogenic temperatures, desiccation, strong electric fields, etc. (Eroglu et al.,
2000; Guo et al., 2000; Mussauer et al., 2001).
Acknowledgements: This work was supported by grants from BMBF (16SV1329 and
0311346) to U.Z. and from the Deutsche Forschungsgemeinschaft (SCHE 209/17–3) to
V.L.S.
References
Arnold W M and Zimmermann U 1982 Z. Naturforsch. 37 908–915
Arnold W M and Zimmermann U 1988 J. Electrostatics 21 151–191
Clarke R J 2001 Adv. Colloid. Interface Sci. 89–90 263–281
Crowe J H and Crowe L M 2000 Nature Biotechnol. 18 145–146
Cseh R and Benz R 1998 Biophys. J. 74 1399–1408
de-Araujo P S 1996 Braz. J. Med. Biol Res. 29 873–875
Eroglu A, Russo M J, Bieganski R, Fowler A, Cheley S, Bayley H and Toner M 2000 Nature Biotechnol.
18 163–167
Fuhr G, Zimmermann U and Shirley S G 1996 in: Electromanipulation of Cells; U Zimmermann and G
Neil eds. (Boca Raton USA: CRC Press)
Gascoyne P R, Becker F F and Wang X B 1995 Bioelectrochem. Bioenergetics 36 115–125
Gimsa J, Müller T, Schnelle T and Fuhr G 1996 Biophys. J. 71 495–506
Guo N, Puhlev I, Brown D R, Mansbridge J and Levine F 2000 Nature Biotechnol. 18 168–171
Hughes M P, Morgan H and Rixon F J 2002 Biochim. Biophys. Acta 1571 1–8
Jones T B 1995 Electromechanics of Particles (New York: Cambridge University Press)
Lambruschini C, Relini A, Ridi A, Cordone L and Gliozzi A 2000 Langmuir 16 5467–5470
Luzardo M C, Amalfa F, Nuñez A M, Diaz S, Biondi De Lopez A C and Disalvo E A 2000 Biophys. J.
78 2452–2458
Pethig R and Markx G H 1997 Trends Biotechnol. 15 426–432.
Mussauer H, Sukhorukov V L and Zimmermann U 2001 Cytometry 45 161–169
Oliver A E, Crowe L M and Crowe J H 1998 Seed Sci. Res. 8 211–221
Reuss O R, Kurschner M, Dilsky S, Horbaschek M, Schenk W A, Zimmermann U and Sukhorukov V
L 2002 J. Electrostatics 56 419–434
Ross P E , Garber S S and Cahalan M D 1994 Biophys. J. 66 169–178
Sukhorukov V L, Arnold W M and Zimmermann U 1993 J. Membr. Biol. 132 27–40
Sukhorukov V L and Zimmermann U 1996 J. Membr. Biol. 153 161–169
Sukhorukov V L, Kurschner M, Dilsky S, Lisec T, Wagner B, Schenk W A, Benz R and Zimmermann
U 2001 Biophys. J. 81 1006–1013
Tsvetkova N M, Phillips B L, Crowe L M, Crowe J H and Risbud S H 1998 Biophys. J. 75 2947–2955
Woelders H, Matmijs A and Engel B 1997 Ciyobiology 35 93–105
Wolfe J and Bryant G 1999 Cryobiology 39 103–129
Wang X B, Huang Y, Gascoyne P R, Becker F F, Holzel R and Pethig R 1994 Biochim. Biophys. Acta
1193 330–344

107
Cell sorting separation using dielectrophoresis

David Holmes and Hywel Morgan
Bioelectronics Research Centre, Department of Electronics and Electrical Engineering,

University of Glasgow, Oakfield Avenue, Glasgow, G12 8LT, Scotland, UK.
Abstract. In this paper we present a cell separation device that uses a combination of
dielectrophoresis (DEP) and pressure driven fluid flow to separate cells of different types
along the length of an interdigitated micro-electrode array. As the cells enter the device they are
focussed into a thin sheet, mid-way between the upper and lower walls of the flow channel
using negative-DEP. As they flow further through the device the cells pass over a second
electrode array and experience a positive-DEP force pulling them from the mid-plane of the
channel onto the ‘separation’ electrode array. Depending upon the mode of operation it is
possible to concentrate a single cell type from a heterogeneous mixture (with all other cell
types passing through the device) or alternatively separate different cell types along the length
of the device. In the second mode of operation the position along the electrode where a cell
lands is a function of the cell’s size and the dielectric properties of its cellular membrane. Use
of the technique to fractionate leukocyte subpopulations is discussed.
1. Introduction
DEP has been applied to the separation and manipulation of a vast array of bioparticles
since it was first described by Pohl in 1978 [1–5]. The technique has been used for the
separation and manipulation of cells by a number of groups [4, 6–13].
The DEP force can be expressed as [14]:
(1)
where is the effective polarisability, v is the volume of the particle and is the electric
field.
A schematic diagram of the DEP-separator is shown in figure 1. The system consists of two
separate arrays of interdigitated bar electrodes integrated into the one device. When particles
enter the device they are carried in a fluid stream and are distributed randomly throughout
the chamber volume. Using negative DEP forces, the initial electrode array forces the wide
distribution of particles entering the device into a well-defined sheet positioned midway
between the upper and lower channel walls. Particles then enter the second or ‘separation’
electrode array, which is energised such that a positive DEP force acts upon either all the
particles or a desired subpopulation of particles, pulling them out of solution onto the
electrode surface.

108
Figure 1. Schematic of the DEP cell separation chip, showing the focussing and separation
electrode arrays.
The devices were fabricated on glass substrates allowing particles held at the electrodes to
be imaged within the device. All captured cells can subsequently be eluted for further
processing by turning off the electric field (or applying negative DEP) whilst continuing to
flow fluid through the device.
The position at which a particle lands on the separation electrode array depends upon the
forces acting upon that particle (DEP, gravity, fluid flow). For a given set of experimental
conditions (e.g. the frequency and magnitude of the applied electric field, fluid flow rate)
particles with the same size and dielectric properties will follow a similar trajectory and
band at the same position along the length of the separation electrode array. Particles with
differing size and/or dielectric properties follow distinctly different trajectories and band at
different positions along the array.

Separation devices were fabricated on glass microscope slides. The electrode arrays were
patterned using standard photolithography and wet etching techniques. A 100µm deep
flow channel was defined in SU8 photoepoxy, with the channel lid aligned and glued in
place using UV curable glue. Inlet and outlet holes were drilled prior to gluing the lid (see
[15] for further details).
Two electrode sizes were used, with electrode (d1) and gap (d2) sizes d=d1=d2=20µm or
40µm. The precise dimensions of the flow channels were measured, for each device, and
this information was used in computer simulations, the data from which was compared with
experiment. An iterative approach was used to calculate the trajectory of single particles
flowing through the device [15–17]. The channels were viewed using an inverted microscope
in either phase contrast or epi-fluorescence mode and the images were recorded using a
CCD camera onto S-VHS or captured directly to a PC.
Experiments were carried out using mixtures of cultured human monocytes (THP-1 cell
line) and human peripheral blood mononuclear cells (PBMCs). PBMCs were collected
from the buffy coat after density gradient centrifugation of whole blood over a Histopaque-
1077 gradient. Cells were labelled prior to mixing with CellTracker™ dyes (Molecular
Probes); the THP-1 monocyte cell line was labelled using one colour of fluorescent probe
and PBMCs labelled with a different colour.
The magnitude and direction of the DEP force depends upon the relative polarisabilities of
the cells and the suspending media. Cells were therefore resuspended at known
concentrations (~106cells/ml) in a low ionic strength media (dH2O containing Ficoll400
(3.5% w/v), sucrose (9% w/v), glucose (0.1% w/v) with the addition of small amounts of
phosphate buffer) of pH 7.4, osmolality ~290mOs/m and conductivity ~10mS/m.

109

Cell suspensions were fed into the device via a sample column and syringe pump.
Experiments carried out using the THP-1 monocyte cell line showed that the cells were
initially focussed to the central plane of the channel mid-way between the upper and lower
electrode arrays. Variation of the applied voltage or flow rates resulted in altered particle
trajectories and therefore different banding positions along the electrode array.
Figure 2 shows the mean banding position (along the separation electrode array) at which
THP-1 cells were captured and held under the influence of +veDEP. The focussing section
of the both devices used electrodes with characteristic dimensions d=d1=d2=40µm, and
applied peak potential of 1.25V at 20MHz for all experiments. For the experiments involving
the 20µm separation electrode array an applied peak voltage of 1.5V at 100kHz, and
medium conductivity of σm= 28mSm”1 was used. In the case of the 40µm separation electrode
array, the applied peak voltage was 1.25V at 1MHz, and σm=26.9mSm”1.
The dashed lines in figure 2 show the results of simulations based on the experimental
parameters (flow rate, applied voltage, channel dimensions, etc.), and using the dielectric
properties of the THP-1 cells as calculated from crossover data (specific membrane
capacitance Cmem=17.7 ±2.7 mFm”2, cell radius was r=6.4 ±1.0 µm and the membrane
conductance Gmem=38.1 ±69.5 mSm”1). A 2D analytical expression for the DEP force was
used in these calculations [18, 19].
Figure 2. Banding position of THP-1 cells on 20µm and 40µm electrode arrays for different flow
rates. Dashed lines represent the simulation taking d1=d2 and the dotted line shows the simulation
with d1=36µm and d2=44µm.
Figure 2 shows how variation in the fluid flow rate results in a change in the banding
position of the cells. The distribution in cell numbers along the device was determined by
image processing of the bright field images. The median point of the cell distribution was
estimated and plotted for the different experimental parameters. The theoretical position of
the band is in close agreement with the experimental results for the 20µm electrodes. The
discrepancy between the calculated banding position and experiment, for the 40µm
electrodes is due to electrode and gap size not being equal. The electrode and gap sizes
were measured and found to be d1=3 6µm and d2=44 µm respectively.

110
In order to account for the variation in the electrode/gap size, as measured for the larger
sized array, a 1D expression for the DEP force was implemented in the simulation. The
effect of this was to reduce the discrepancy between the calculated and experimental results.
EHD effects may also have a slight disruptive effect on the banding positions of the cells.
Mixtures of cells:
Cell suspensions comprised of 1:2 mixtures of fluorescently labelled THP-1 cells and
PBMCs were fed into the device. A typical set of experimental conditions were: cell
concentration of 1×106 cells/ml, flow rate of 1ml/hr, applied potential of 1.25V at 20MHz
at the focusing electrodes, and 2V at 200kHz applied to the separation electrodes. Figure 3
shows a fluorescence image captured after 0.5ml of cell suspension has flowed through the
device under the above conditions.
Figure 3. Fluorescence image of THP-1 cells and PBMCs banding on the separation electrode array.
The vertical black stripes are the electrodes of the separation array.
From the fluorescence image of figure 3 the mean positions of the two cell types can be
seen to differ. Figure 4 shows a plot of cell numbers versus distance along the length of the
separation electrode array for a similar experiment.
Figure 4. Variation in distance along the separation electrode for mixture of THP-1 cells and PBMCs.

111
The mean banding position of the THP-1 cells and the PBMCs are different. The
distribution of the PBMCs can be approximated with a triple gaussian curve, suggesting
that the PBMC sample may also be separating into its sub-populations along the length of
the electrode array. The gaussian shape of the curves shown in figure 4 is due to the inherent
distribution in size and dielectric properties of individual cells within each of the PBMC
sub-populations.
Fractionation of PBMC sub-populations:
The dashed lines in figure 5(a) show the simulated banding position of monocytes and
Tlymphocyte using the size and specific membrane capacitance values given by Yang et al
[20]. Of the cells present in the PBMC population, monocytes and T-lymphocytes have the
highest and lowest values of specific membrane capacitance. These cells therefore represent
the upper and lower limits of the mean capture position for the PBMC subpopulations.
Figure 5(b) shows an example of the simulation results for particle trajectories of the four
main PBMC sub-populations, and their predicted mean banding position along a 40µm
separation electrode.
Figure 5. (a) Variation in distance along the separation electrode versus applied voltage, dashed lines
show simulation (b) simulated particle trajectories for the different PBMC sub-populations.
4. Conclusion
In this paper we have presented a novel DEP electrode configuration, which relies upon
initially focusing all the cells in the flow stream to the central plane of the flow channel.
Cells then enter a second region where they undergo positive DEP and are attracted the
separation electrode array where they are held. This technique is potentially applicable to
the fractionation of a variety of particle types. To allow collection of the cell fractions we
suggest that the separation electrode could be fabricated with independently addressable
sections, allowing the release and collection of individual bands of cells while retaining
the rest of the cell sample against the fluid flow. We are currently investigating the possibility
of using the system to perform differential blood counts.
5. Acknowledgements
THP-1 cells were a gift from Dr S Robertson at Glasgow Royal Infirmary. Thanks to members
of the Bioelectronics group at Glasgow for blood samples and Dr M Thomas for discussion.

112
References
1. Pohl, H.A., Dielectrophoresis. 1978, Cambridge: Cambridge University Press
2. Pethig, R., Dielectrophoresis: Using inhomogeneous AC electrical fields to separate and
manipulate cells. Critical Reviews in Biotechnology, 1996. 16(4): p. 331–348.
3. Hughes, M.P., H.Morgan, and F.J.Rixon, Dielectrophoretic characterisation and separation of
herpes simplex virus particles. Biophysical Journal, 1997. 72(2): p. MP447– MP447.
4. Muller, T., et al., A 3-D micro electrode system for handling and caging single cells and particles.
Biosensors & Bioelectronics, 1999. 14(3): p. 247–256.
5. Washizu, M., et al., Molecular Dielectrophoresis of Biopolymers. leee Transactions On Industry
Applications, 1994. 30(4): p. 835–843.
6. Yang, J., et al., Cell separation on microfabricated electrodes using dielectrophoretic/gravitational
field flow fractionation. Analytical Chemistry, 1999. 71(5): p. 911–918.
7. Talary, M.S., et al., Dielectrophoretic Separation and Enrichment of Cd34+ Cell Subpopulation
From Bone-Marrow and Peripheral-Blood Stem-Cells. Medical & Biological Engineering &
Computing, 1995. 33(2): p. 235–237.
8. Rousselet, J., G.H.Markx, and R.Pethig, Separation of erythrocytes and latex beads by
dielectrophoretic levitation and hyperlayer field-flow fractionation. Colloids and Surfaces a-
Physicochemical and Engineering Aspects, 1998. 140(1–3): p. 209–216.
9. Markx, G.H. and R.Pethig, Dielectrophoretic Separation of Cells—Continuous Separation.
Biotechnology and Bioengineering, 1995. 45(4): p. 337–343.
10. Huang, Y., et al., The removal of human breast cancer cells from hematopoietic CD34(+) stem
cells by dielectrophoretic field-flow- fractionation. Journal of Hematotherapy & Stem Cell Research,
1999. 8(5): p. 481–490.
11. Heida, T., et al., Viability of dielectrophoretically trapped neural cortical cells in culture. Journal
of Neuroscience Methods, 2001. 110(1–2): p. 37–44.
12. Gascoyne, P.R.C., et al., Dielectrophoretic separation of cancer cells from blood. leee Transactions
On Industry Applications, 1997. 33(3): p. 670–678.
13. Cui, L., D.Holmes, and H.Morgan, The dielectrophoretic levitation and separation of latex beads
in microchips. Electrophoresis, 2001. 22(18): p. 3893–3901.
14. Jones, T.B., Electromechanics of Particles. 1995, Cambridge: Cambridge University Press.
15. Holmes, D., M.Thomas, and H.Morgan, Dielectrophoretic separation/isolation of rare particles/
cell types form a heterogeneous suspension within a microfluidic system, in Micro Total Analysis
Systems 2000, Proceedings. 2000, Kluwer Academic Publ: Dordrecht, p. 115–118.
16. Holmes, D. and H.Morgan. Particle focussing and separation using dielectrophoresis in a
microfluidic channel. in Micro Total Analysis Systems 2001, Proceedings. 2001. Monterey, USA:
Kluwer Academic Publ.
17. Holmes, D. and H.Morgan. Dielectrophoretic chromatography of cells, in NanoTech2002: The
6th Annual European Conference On Micro & Nanoscale Technologies for the Biosciences.
2002. Montreux, Switzerland.
18. Morgan, H., et al, The dielectrophoretic and travelling wave forces generated by inter digitated
electrode arrays: analytical solution using Fourier series (vol 34, pg 1553, 2001). Journal of
Physics D-Applied Physics, 2001. 34(17): p. 2708–2708.
19. Morgan, H., et al., The dielectrophoretic and travelling wave forces generated by inter digitated
electrode arrays: analytical solution using Fourier series. Journal of Physics D-Applied Physics,
2001. 34(10): p. 1553–1561.
20. Yang, J., et al., Dielectric properties of human leukocyte subpopulations determined by
electrorotation as a cell separation criterion. Biophysical Journal, 1999. 76(6): p. 3307–3314.

113
Dielectrophoretic transport sorting of particles using an

electrode micro-array
B Malyan, J Kulon, W Balachandran
Brunei University, Department of Systems Engineering
Uxbridge, Middlesex, UB8 3PH, UK
Abstract. The effect of a shaped electrode geometry array on the dielectrophoretic transport
and characterisation of nanometre-sized particles, including possible biological material, was
investigated. Polarisable particles placed within a non-uniform electric field are subject to
dielectrophoretic force acting upon them. Use was made of an electrode array to investigate the
transport and separation of nanometre-sized particles. Electrode arrays having electrode and
gap dimensions of micrometre size were utilised to produce an electric field of sufficient
magnitude for dielectrophoresis of the submicrometer sized particles to occur. An ac electric
signal was applied to electrodes within the electrode array; the electric field thus created was
switched sequentially along the electrode array at a frequency to track the particles’ movement.
Electrode geometries were investigated to maximise the efficiency of particle transport.
1. Introduction
The transport and the separation of nanometre sized particle, including biological particles,
lends itself to many applications such as may be exploited by laboratory-on-achip,
biofactory-on-a-chip [1], doctor-on-a-chip and DNA chip technologies. Dielectrophoretic
force, acting on a particle, permits the gentle and non-destructive manipulation of that
particle. A polarisable particle placed within a non-uniform electric field can demonstrate
dielectrophoresis. The particles have a tendency to migrate to regions of high field strength,
in the case of positive dielectrophoresis, or to move towards regions of low field strength,
in the case of negative dielectrophoresis. If an ac signal is used to create the non-uniform
electric field, then a polarisable particle placed within the field may also show evidence of
the strength of the dielectrophoretic force acting upon it having a dependency upon the
frequency of the supply signal.
Previous research [2] on the dielectrophoretic properties of sub-micrometre sized latex
spheres has established the frequency dependence of the dielectrophoretic force.
The DEP force in the time-averaged form is given by:
(1)
The DEP force depends on , a factor proportional to the electric field geometry
and on the real part of fCM the in phase component of the particles effective polarizability
[2]. The magnitude of the dielectrophoretic force acting upon a particle differs with the

114
frequency of the applied electric field and the complex permittivity of the particle suspension
medium.
The phenomenon of dielectrophoresis has been used for sub-micron particle manipulation
and characterisation and is also being increasingly applied to a biological area of research.
Cell transport and capture [3], levitation and separation [4] have been demonstrated.
Dielectrophoresis appears to have no effect on cell morphology, cell oxidative respiration
and cell dynamics [5]. There is considerable interest in the manoeuvring and electrical
classification of DNA macromolecules [6,7] and DNA fragment populations arising from
polymerise chain reaction (PCR) technique.
A method is described for the physical separation of a mixture of two sizes of particles
suspended within a liquid medium. The particle, suspension liquid combination was
contained within a micro-channel located above a series of contoured electrodes. Very little
fluid flow movement was permitted inside the walls of the microchannel. Dielectrophoretic
force acting on the sub-populations of particles allowed one of the sub-populations of
particles to be moved along the micro-channel, leaving behind the other particle sub
population. In contrast, the same electrode arrangement was also employed to trap one
particle sub-population size and allow fluid flow, through the micro-channel, to carry away
the other particle sub-population.
2. Experiment
Microelectrodes were used to create the non-uniform field necessary for dielectrophoresis;
Designs were fabricated on a soda lime glass, with a transmittance value of 88% at 375–450
nm and having a glass flatness class of 5µm. The electrodes were constructed from 1 µm
thickness chrome, deposited on the surface of the glass. The width of a single electrode was
30 µm, a gap of 30 µm existed between the profile of the electrode tips and the common
electrode, Figure 1.
Figure 1. Photograph some of the electrode tips and the common electrode (top)
The electrodes, in Figure 1, are shown in black, the particles were manoeuvred along a path
parallel to the top common electrode. A micro-fluidic channel was placed above the chrome
electrodes; the top of the channel was fabricated from glass. The sides of the channel were
produced in a plastic material, the micro-channel having dimensions of 100 µm depth, 250
µm, width and 5 mm in length. The plastic material provided a seal to prevent carrier fluid
loss, and allowed removal and replacement of the micro fluidic channel. The electrodes and
micro-channel assembly were mounted on a Vickers Photoplan optical microscope, having

115
a digital video camera output that was connected to a computer to enable picture capture.
Phase contrast microscopy was used in order to give more contrast to the images of sub-
micrometre sized particles suspended in a fluid medium. In phase contrast microscopy a
phase plate is placed in the light path. Light that is only just refracted passes through the
centre of the plate and is not retarded. Highly refracted light passes through the plate
furthest from centre and is held back another one-quarter wavelength; the microscope field
therefore reveals a darker background. The video output screen width was 640 pixels,
equivalent to a “real” image size of 260 µm in width.
Finite element analysis of part of the electrode array was carried out using Algor, a
commercial software package. The result of the software simulation is shown in Figure 2,
which shows a uniform gradient in the electric field magnitude at the tip of the vertical
central electrode; the central electrode was assumed to have a 30V dc signal supplied to it,
the surrounding electrodes being unconnected.
Figure 2. Finite element simulation of the electric field magnitude

surrounding central electrode tip, supplied with a 30 V dc signal
The shaped, sloping top of the central electrode was designed to give a non-symmetrical
electric field magnitude at the tip of the electrode when an electric signal is applied to
electrode. This non-symmetrical electric field magnitude gradient was designed to assist
the dielectrophoretic movement of particles in one direction along the micro-channel,
Figure 2. In the case of negative dielectrophoresis, in order to move particles along the
track the particles have to move from the central electrode past the next microelectrode.
Particles moving from left to right reach an area of the lowest field magnitude before
passing the next electrode. Particles moving from the central electrode towards the left of
Figure 2 pass the next microelectrode tip before reaching an area of lowest field magnitude.
The carrier fluid used was a 1 mmol potassium chloride solution, conductivity of
approximately 170×10-4 m2Smol-1 at 25 degrees centigrade. The particles used were 0.5
µm, 0.915 µm and 3 µm standard micro-spheres, carboxylate, mono-dispersed polystyrene
latex spheres that contain surface carboxyl groups. The final concentration of the particles
in the suspension medium was 0.25%. A signal frequency of 6.5 MHz was chosen for the
experiment, 20 volts peak-to- peak. This signal gave rise to negative dielectrophoresis,
with all the three particle sizes used. Using a 3 µm diameter micro sphere it was established
that a 6.5 MHz, 10v peak-to-peak signal produced an average particle velocity of
approximately of 1 µm per a second within a distance of 5 µm from an electrode. This

116
velocity dropped to approximately 0.5fiim/s at a signal frequency of 2 MHz, but remained

at 1µm/s at 8 MHz. No fluid pumping within the micro fluidicchannel could be observed
with the 6.5 MHz signal applied to an electrode. Random particle motion, due to Brownian
motion, was observed; this was readily increased when the carrier fluid was heated by a hot
light source. A section of the electrode array track covered with the carrier medium, in
which were suspended 3 µm and 0.914 µm micro-spheres, is shown in Figures 3 and 4. The
tip of the middle electrode shows signs of “wear” due to the applied electrical signal.
Figure 3. A section of the electrode array, with 3 µm and 0.914

µm diameter micro-spheres suspended within the fluid medium.
3. Results
The electrical input signal was applied to the electrode 3 of Figure 4, creating a nonuniform
electric field magnitude around the electrode. The particles in the vicinity of the electrode
experience negative dielectrophoresis; the majority of the larger particles in this region are
seen to migrate left towards areas of low field strength, Figure 5, above electrode 2. The
smaller 0.194 µm particles travel at a slower rate than the 3 µm spheres towards the regions
of lower electric field magnitude, and are left behind.
Figure 4. Signal applied to electrode 3 Figure 5. The 3 µm particles are above 2
The electrical input signal has been switched to electrode 2; the 3 µm micro-spheres are
shown moving away from electrode 2, Figure 6,

117
The 3 µm diameter latex spheres trajectory has placed them at the tip of electrode 1, Figure
7, the particles are seen to “daisy-chain” together.
Fiaure 6. Particles moving from electrode 2 Figure 7. 3 µm Darticles reach tip of electrode
The input signal was then switched to electrode 1, the larger particles move left from the tip
of this electrode, Figure 8. The 0.914 µm diameter micro-spheres are left behind in the
micro-fluid channel, Figure 9. The experiment was repeated, separating 0.5 µm and 0.914
˜m diameter particles populations from each other.
Figure 8. 3 micron spheres moving left of the Figure 9. 0>914 micron particles are now
tip of electrode 1. separated from the 3 micron sized particles.
An additional experiment was carried out using fluid flow to separate the two sizes of
particles from each other.
Figure 10. Fluid flow particle separation

Fluid was allowed to flow through the micro-fluid channel (from left to right) at a rate of 1
µm/s, this carried along the particles suspended within the fluid medium. The 6.5 MHz

118
input signal was then applied to all the vertical electrodes, creating regions of low electrical
field strength between the electrodes. All the micro-spheres tended to migrate towards
these regions of low field strength; however the smaller particles were transported away by
the fluid flow leaving the larger particles trapped between the electrodes, Figure 10.
4. Conclusions
Dielectrophoretic separation of two particle size sub-populations, each subject to negative
dielectrophoresis, was achieved. Additionally one of the particle subpopulations was
transported using dielectrophoresis to a different physical location along an array of profiled
microelectrodes, leaving behind the other sub-population. The distance the particles may
be moved is dependent on the length of the electrode array fabricated. Further work is being
undertaken on the movement and characterisation of DNA.
References
[1] Pethig R, Burt J P H, Parton A, Rizvi N, Talary M S, and Tame J A Development of biofactory-
on-a-chip technology using excimer laser micromachining, J. Micromech. Microeng. 8 (1998)
57–63
[2] Green N G, Morgan H, dielectrophoretic investigations of sub-micrometry latex spheres, J.Phys.
D: Appl. Phys. 30 (1997) 2626–2633
[3] Goater A D, Pethig R, Paton C A and Smith, Single Cyyptosporidium Ocyst Isolation and
Capture using a Travelling Wave Dielectrophoresis Device, Electrostatics 1999, Institute of
Physics conference number 163, Proceedings of the 10th International Conference, Cambridge
[4] Arnold W M, Positioning, Levitation and Separation of Biological Cells, Electrostatics 1999,
Institute of Physics conference number 163, Proceedings of the 10th International Conference,
Cambridge
[5] Archer S, Li TT, Evans T, Britland S T and Morgan H, Cell Reactions to Dielectrophoretic
Manipulation, Biochemical and Biophysical Research Communications 257,687–698, 1999
[6] Morishima K, Arai F, Fukuda T, Matsuura H and Yoshikawa K, Transportation of DNA
Molecule Utilizing the Conformational Transition in the higher order structure of DNA, CCAB
Mini Review 1997
[7] Alsbury C A, Diercks AH, Ger van den Engh, Trapping of DNA by dielectrophoresis,
Electrophoresis, 23, 2658–2666, 2002

119
AC electrokinetic focussing in microchannels: Micro- and

nanoparticles
Hywel Morgan, David Holmes and Nicolas Green
Bioelectronics Research Centre, Department of Electronics and Electrical Engineering,

University of Glasgow, Oakfield Avenue, Glasgow, G12 8LT, Scotland, UK.
Abstract. Dynamic focussing of particles can be used to centre particles within a flow-
through channel. This ensures that all particles pass through the same detection volume. In this
paper we describe a method for focussing nano-particles within a microfluidic channel using
AC electrokinetics. The method differs from other proposed focussing methods in that it
manipulates the particle and not the fluid. Simulations of particle focussing for 50nm and 1 µm
particles in a cylindrical flow channel are presented.
1. Introduction
During the last few years there have been a number of developments in the design and
fabrication of miniaturised particle and molecule handling systems for detection and sorting
applications. For example Fu et al [1, 2] developed a particle sorting device which used DC
voltages to control and deflect the movement of fluid in a microfabricated channel. Particles
were detected in the channels using fluorescence, with the light focussed into a detection
volume of some nL. In order to ensure accurate and rapid detection of the particles in the
channel, it is important to centre the particle stream at the focus of the light beam. The
principle of such a detection system is shown in figure 1, which shows how the light from
a high numerical aperture objective lens is focussed into a diffraction limited spot, and how
particles can be detected using fluorescent light sampled using confocal optics.
Figure 1. Principal of confocal optical detection of particles in a narrow channel.

120
Focussing particles into a narrow beam so that they all move at the same velocity in the
centre of the channel can be performed using two methods: hydrodynamic, or electrokinetic.
In hydrodynamic focussing the fluid sample is surrounded by a sheath flow, which squeezes
the sample stream to form a narrow beam of sample. The ratio of fluid flow rates between the
sample and sheath determines the width of this beam. This principle is shown in figure 2a
and has been used by a number of groups to focus particles in the design of particle sorters
(e.g. [3], [4], [5]).
Figure 2. (a) Sheath flow focussing fluid and particles into the centre of the channel, (b) AC fields
generated by micro-electrodes in the channel can be used to focus particles.
DC electric fields can also be used to move fluids using electro-osmosis. In this method [1,
2], the sample fluid streams are driven along the main arms of a cross-shaped channel under
the influence of an applied DC field. As the sample enters the intersection the fluid streams
meet and the sample stream is focussed into narrow beam.
The main drawback of the hydrodynamic technique is that focussing is generally only in
one dimension and complex fabrication schemes are required to achieve two-dimensional
focussing [5]. There is also the need for accurate control of fluid flow rates. In addition,
because this type of focussing acts on the fluid rather than the particles, small particles and
molecules can diffuse from the sample stream into the sheath. DC electrokinetic methods
require high voltages to move particles and or the fluid. Local control of the fluid is not
possible, so that the switching speed is likely to be governed by the compliance of the fluid
in the channel.
To circumvent these problems, dielectrophoresis can be used to move particles along field
gradients and therefore to focus particles. This has been demonstrated by the group of Fuhr
et al in a microchip cell sorter [6, 7], who showed how particles could be focused in two
dimensions in a fluid stream. A channel is fabricated with pairs of electrodes on the top and
bottom, and using negative dielectrophoresis particles are repelled from the electrodes and
into the channel centre. This principle is shown in figure 2b.
2-D DEP focussing has been combined with simultaneous single particle impedance
measurements within a flow-through chip to produce a particle sorting device [8]. For
effective DEP-based focussing, the particles must experience a negative DEP force, which
in dipole approximation is given by equation (1), [9, 10].
(1)

121
In this expression, v is the particle volume, α the particle polarisability, the del operator
and E the electric field. Note that at the centre of the channel, this equation is incorrect
because the dipole vanishes and only the quadropole moment is responsible for the force
[10, 11]. However, in the analysis presented in this paper only the dipole form will be used
as a first approximation.
Equation (1) shows that the DEP forces scales linearly with particle volume so that it is
relatively easy to focus particles such as cells using low voltages. From simple scaling, it
can be seen that in order to achieve effective focusing of molecules and nano-particles the
magnitude of has to be increased considerably. In addition, Brownian motion is also
a significant factor that must be taken into account when designing systems for nano-
particle manipulation [9]. The DEP force must be of the same order of magnitude as the
typical force produced by Brownian motion, otherwise deterministic movement of the
particle will not occur in the transit time of the particle within the channel.
The situation with more than one particle is different because in this case diffusion forces
come into play. For an ensemble of particles, the spatial distribution is governed by the
different particle fluxes and the steady state distribution in particle concentration is reached
when the diffusion flux is exactly balanced by the DEP flux (assuming that the sedimentation
flux can be ignored). For effective focussing, the flux of particles directed into the centre of
the channels must be great enough to pinch the particle distribution into a tight beam. The
DEP force can be estimated provided the (two dimensional) distribution of is known.
Then the particle distribution can be calculated across the channel.
2. System design
The behaviour of a particle in a focussing device was simulated for a system with the
following design. A channel of square cross-section 10µm×10µm has electrodes positioned
at the channel corners as shown in figure 3, which shows a diagram of the proposed device.
Voltages and frequencies are applied to the electrodes to produce negative DEP.
Figure 3. Diagram showing how a cell can be focussed using DEP and four electrodes at the
corners of a channel.
3. Simulation
To determine the efficiency of the focussing device the values of were determined
across a 2-D slice of the system. FlexPDE™ (PDE Solutions Inc.) was used to solve Laplace’s
equation using voltages of +1V and -1V on each electrode. A vector plot of is shown

122
in figure 4a. Note that the direction of the DEP force vectors push particles away from the
electrode edges at the corners and into the centre of the channel. In figure 4b the magnitude
of E2 is plotted along three different lines; a diagonal from the corner, and a horizontal and
vertical line. It can be seen that there is a large variation in the magnitude of the force at the
perimeter of the channel, for example close to the electrodes the value of E2 is 10 times
larger than in the region between the electrodes. In the central region of the channel the
values are similar.
Figure 4. (a) A vector plot of the DEP force, directing all particles radially into the centre of the
channel, (b) The magnitude of E2, along three different lines from the centre of the 10µm diameter
square channel. The diagonal line has a maximum value of E2 close to the electrodes (as expected). In
the centre, and within 3 µm from the channel all the values of E2 are similar.
4. Focussing single particles

To a first approximation the 2-D geometry shown in figure 3 can be reduced to 1-D by
approximating the square channel to a cylinder. In this case the DEP force can be assumed
to only vary in one dimension, i.e. with radius. Although such a system would be extremely
difficult to construct it allows first order estimates of the force and movement of particles to
be calculated. The radial value of was taken as the diagonal as shown in figure 4b and
the analytical function represented by a polynomial of order 5.
An iterative approach was used to calculate the trajectory of single particles through the
device [12, 13]. Assuming that the steady state velocity of a particle is reached
instantaneously the equation of motion for the particle is
(2)
where m is the mass of the particle and the effect of gravity has been neglected, i.e. there is
no buoyancy force. This equation was solved iteratively to determine the particle velocity
and therefore the time taken for it to reach the centre of the channel. The particle flows
along the channel at the same velocity as the fluid where the fluid velocity (as a function of
radius) is given by:
(3)

123
where Q is the volume flow rate, R the channel radius and r the radial distance from the
centre of the channel. For the simulation, the value of the Clausius Mossotti factor in the
DEP force was set to -0.5 (negative DEP) and the particle trajectory determined for different
particle sizes and flow rates. Brownian motion was neglected in these simple calculations.
The effects of electrohydrodynamic fluid flow (EHD) in such a system have been investigated,
both experimentally and via simulation. EHD effects were found to be negligible due to the
symmetry of the system and their contribution to the present simulations are also neglected.
Figures 5 and 6 shows the distance a particle has to travel along the channel before reaching
a central region defined to be within a factor of R/10, i.e. 0.7µm from the centre of a 14µm
diameter cylindrical channel. Figure 5 shows the focussing data for a 1000nm diameter
particle for different applied voltages and flow rates. Figure 6 shows similar data for a
particle of 50nm in diameter.
Figure 5. Simulation of focussing distance for a 1000nm diameter particle as a function of applied
voltage and flow rate.
Figure 6. Simulation of focussing distance for a 50nm diameter particle as a function of applied
voltage and flow rate.
The data shown in these figures indicates that AC electrokinetic focussing of particles in
the channel is possible. Figure 6 shows that for a 50nm diameter particle, efficient focussing
can be achieved within distances of a few millimetres. Reduction in the flow rate or an
increase in the applied potential would allow the use of shorter channels.

124
Figure 7 shows simulated particle trajectories for a 50nm diameter particle, being focussed
in a micro-channel of diameter 14µm, flow rates of 1µl/hr and 10µl/hr and an applied
voltage of 5V.
Figure 7. Simulated particle trajectories for a 50nm diameter particle being focussed in a
microchannel, flow rates of 1 µl/hr and 10µl/hr and applied voltage of 5V.
5. Acknowledgements
We acknowledge QINETIQ Ltd. for funding and the Royal Society for a fellowship for HM.
References
1. Fu, A.Y., et al., A microfabricated fluorescence-activated cell sorter. Nature Biotechnology,
1999. 17(11): p. 1109–1111.
2. Fu, A.Y., et al., An integrated microfabricated cell sorter. Analytical Chemistry, 2002. 74: p.
2451–2457.
3. Kriiger, J., et al., Development of a microfluidic device for fluorescence activated cell sorting.
Journal of Micromechanics and Microengineering, 2002. 12: p. 486–494.
4. Schrum, D.P., et al., Microchip flow cytometry using electrokinetic focusing. Analytical Chemistry,
1999. 71(19): p. 4173–4177.
5. Sundararajan, N., et al. Three-Dimensional Hydrodynamic Focusing in Poly (dimethyl
siloxane)(PDMS) Microchannels for Molecular Detection, in NanoTech2002: The 6th Annual
European Conference On Micro & Nanoscale Technologies for the Biosciences. 2002. Montreux,
Switzerland.
6. Fiedler, S., et al., Dielectrophoretic sorting of particles and cells in a microsystem. Analytical
Chemistry, 1998. 70(9): p. 1909–1915.
7. Schnelle, T., et al., The influence of higher moments on particle behaviour in dielectrophoretic
field cages. Journal of Electrostatics, 1999. 46(1): p. 13–28.
8. Renaud, P. and S. Gawad. On-Chip Impedance Spectroscopy Flow Cytometry. in NanoTech2002:
The 6th Annual European Conference On Micro & Nanoscale Technologies for the Biosciences.
9. Morgan, H. and N.G. Green, AC Electrokinetics: Colloids and Nanoparticles, ed. R. Pethig.
2003, Baldock, UK: Research Studies Press.
10. Jones, T.B., Electromechanics of Particles. 1995, Cambridge: Cambridge University Press.
11. Schnelle, T., T.Muller, and G.Fuhr, Trapping in AC octode field cages. Journal of Electrostatics,
2000. 50(1): p. 17–29.
12. Holmes, D. and H.Morgan. Particle focussing and separation using dielectrophoresis in a
microfluidic channel, in Micro Total Analysis Systems 2001, Proceedings. 2001. Monterey,
USA: Kluwer Academic Publ.
13. Holmes, D. and H.Morgan. Dielectrophoretic chromatography of cells, in NanoTech2002: The
6th Annual European Conference On Micro & Nanoscale Technologies for the Biosciences.

125
A wide bandwidth probe for electrostatic discharge

measurements
J M Smallwood1, G L Hearn2
1
Electrostatic Solutions Ltd., 13 Redhill Crescent, Bassett, Southampton, SO16
7BQ, UK Email: jeremys@static-sol.com
2
Wolfson Electrostatics, Department of Electronics & Computer Science, University
of Southampton SO17 1BJ, UK. Email: glh@soton.ac.uk
Abstract. A simple wide bandwidth probe has been developed for electrostatic discharge
measurements. The probe allows measurement of discharge current waveforms and charge
transfer. The detailed amplitude-time features of the discharge current waveform are revealed
down to nanosecond timescales.
Calibration of the probe has been demonstrated using nanosecond risetime rectangular pulses
and carefully controlled capacitive discharges. The measured results agree well with values
predicted by theory. Example measurements were made of waveforms from a variety of insulating
materials and FIBC fabrics. The discharge waveform features show considerable variety.
We propose that the new probe will make a valuable contribution to electrostatic discharge
research applied to ESD damage risk in electronics and incendivity studies. They will facilitate
study of the influence of discharge current waveform as well as charge transferred as important
parameters in determining incendivity of electrostatic discharges to flammable mixtures.
1. Introduction
Electrostatic Discharges (ESD) from insulating surfaces, metal objects and other materials
are of concern in ignition of flammable materials [1] and also in assessment of the risk of
damage to electronic components and assemblies during manufacture. These ESD events
are known to be very variable, and can have nanosecond risetimes, durations from
nanoseconds to many microseconds, and peak discharge currents of hundreds of amps.
A simple rugged probe has been designed for measuring low to medium level ESD
events measurements from insulating materials, conductive objects and other small-
tomedium sized electrostatically charged items. The objective was to measure ESD current
waveforms faithfully, so that parameters such as peak current, duration, risetime and fall
time can be investigated. Charge transferred in individual discharges can also be measured
by integration of the discharge current waveform.
The probe uses a passive resistor-transmission line design based on the principles
described by Chubb and Butterworth [2] and Smallwood [3]. The probe tip employs an
earthed 20 mm hemispherical shield, with the ESD current collected on a coaxial wire tip
protruding slightly from the shield centre. The shielded probe design ensures that the
majority of the charge collected in the ESD event passes through the measurement circuit2

126
rather than being locally neutralised at the probe tip. This paper describes the probe design
and its calibration at d.c. and using fast rectangular pulse waveforms.
The probe equivalent circuit is shown in Figure 1. The probe is designed to allow
measurement of actual discharge current waveforms and charge transfer measurements
when used with a wide bandwidth (at least 400MHz) digitising oscilloscope.
Figure 1. Equivalent circuit of probe
The ESD current passes into the sensing wire tip and through a low value current sensing
resistor Rp to the earthed hemispherical shield. The 1Ω resistor is fabricated from a parallel
array of 10 carbon resistors in order to give a low inductance resistance for the ESD discharge
path. The choice of carbon resistors gives a wide bandwidth and low inductance as well as
high immunity to the high pulse power dissipation and peak current levels (tens of amps) to
be expected with ESD events [4].
The voltage developed across the current sense resistor is fed into a 50 Ω transmission
line cable Zc (coaxial cable) via a 50 Ω termination resistor Rs. Rs prevents rereflection of
any reflected high frequency current impulses that might return along the cable in the
direction of the probe, from reflections in imperfections in the transmission line and its
termination. The transmission line is terminated in a 50 Ω impedance input of an
oscilloscope for measurement of the ESD waveform.
2. DC calibration
Basic calibration is performed by measurement of the DC resistance of the probe series and
parallel resistance Rs and Rp using a 4 terminal low resistance meter. The voltage output into
the oscilloscope 50 impedance input, in response to a current input Iin, is given by
With typical values of Rs=50 Ω and Rp=50 Ω, the probe sensitivity Vo/Iin is 0.5 V/A.
It cannot be assumed that the d.c. resistance values represent the impedance of these
components at high frequency, and so these values allow calculation only of the nominal
response. The high frequency pulse response must be confirmed by another method.
3. Pulse calibration
Conventional calibration of the ESD probe in terms of frequency response is not easy, as it
has a low impedance (1 Ω) input and is not mechanically compatible with traditional 50 Ω
measurement connectors and equipment. It is also doubtful whether a frequency response

127
measurement alone is sufficient as it is important that the probe accurately maintain the
waveform of the incoming ESD wave. For these reasons, calibration has been performed by
measuring the response to a short duration, fast risetime rectangular pulses.
The pulses were generated by discharge of a charged length of 50 Ω coaxial cable into
a matched impedance output line (Figure 2). When the relay contact closes, a rectangular
pulse is formed in the output line which has a duration determined by the length of the
pulse forming line, and amplitude of Vp/2. With perfect line impedance matching, the pulse
is rectangular, but in practice mismatches create some deviations from this ideal due to
wave reflections. The pulse output is measured using a fast digital storage oscilloscope.
The pulse is applied to the probe tip via a 47 Ω series resistor Rcal within a special coaxial
test connecting jig. An impedance mismatch occurs at the point of the connection of Rcal to
the output line, and the rectangular pulse is diminished in amplitude and followed by
reflected waves. However, the resulting pulse is sufficient to offer a calibration pulse to the
ESD probe, and is directly compared with the waveform on the probe output line using a
separate channel of the oscilloscope.
Figure 2. Pulse test arrangement
3.1. Pulse amplitude and risetime calibration

The waveforms were measured from a trigger point t=0 at the rising edge of the reference
pulse. There is a time separation between the waveforms measured due to the difference in
travel time of the wave through the ESD probe and its cable, compared to the travel time
through the reference output line. Average values of the waveform peaks were calculated
over the time range +10 to +40 ns for both waveforms. The probe output peak voltage Vo
was compared to the reference pulse value Vref and the input current Iin calculated as
Vref(Rcal+Rp). This simple approach assumes that stray components have an insignificant
effect on the current waveshape appearing at the probe input.

128
Figure 3. Pulse amplitude and risetime calibration waveforms
The results obtained support this assumption to a first approximation. The probe output
response was calculated to be 0.505 V/A.
The risetime of the probe output is measured from the normalised output waveform as
the time for transition between 10% and 90% of the waveform average peak amplitude.
Figure 3a shows the input pulse to the ESD probe, and the output due to the same pulse
measured by the ESD probe.
Figure 3b shows detail of the rising edge of the normalised pulse waveform, from which
the risetime measurement was made. The risetime of Vref was 0.7 ns. The measured risetime
of the probe waveform includes some slowing of the pulse rising edge due to the Rcal and
inductance of the probe input connection fixture. The effect of these has not been quantified.
The measured risetime of the output waveform (1.7 ns) can therefore be considered to be a
conservative estimate of the actual probe performance.
4. Some ESD waveforms measured using the ESD probe

We used the ESD probe to measure some ESD events from a variety of charged objects and
material surfaces. The purpose of these simple measurements was to give an indication of
whether the probe would show differences in the waveshapes, amplitudes, rise and fall
times and durations of these ESD events. Figure 4 shows a typical unidirectional waveform
obtained from a negatively charged PTFE slab surface.
Figure 4. Discharge obtained from a negatively charged PTFE surface

129
Figure 5. Discharge obtained from a metal key (estimated capacitance 5 pF) placed on a negatively
charged PTFE surface
In contrast an oscillatory discharge is obtained from a key placed on the charged surface
(Figure 5). The discharge from the PTFE surface has approximately one tenth the amplitude
(about 0.6 A compared to 6 A) and 10 times the duration of the discharge from the key. The
charge transferred in each discharge, calculated by integration of the current waveform
over the duration of the discharge, was of the same order, being 85 nC from the PTFE surface
and 100 nC in the discharge from the key.
Figure 6 shows a discharge waveform obtained from a 130 pF capacitor charged to a
potential of approximately -15 kV in series with a 7.5 Ω resistor. The very high peak current
(around 60 A) and oscillating waveform typical of such discharges is clearly seen. The
charge transferred in this case was calculated to be 1800 nC, approximately the charge
expected to be stored in the capacitor. An accurate measurement of the stored charge has not
yet been made, as our intention was to illustrate the application of the probe in higher
current ESD measurements. The stored charge is not necessarily equal to the CV product,
where C is the low voltage capacitance, as the capacitance of a high voltage capacitor is
often in practice a function of the applied voltage. Further experiments using high quality
capacitors are planned.
The noise sources on the waveforms have not yet been fully studied. Digitising noise
due to the oscilloscope (Tektronix TDS380, 8 bit resolution) could be appreciable if the
waveform only filled a small fraction of the oscilloscope screen. ESD waveforms are variable
and it is often not easy to predict the amplitude of waveforms that will be obtained. So, the
Y scale used is often larger than would give optimum digitising resolution. Use of a high
resolution instrument would alleviate this effect.
Figure 6. Discharge obtained from a 130 pF capacitor via a 7.5 Ω resistor

130
Ringing (oscillation) is often seen to be present after the fast rising edges of a measured
ESD waveform. Often this is a real feature of the discharge, although any ESD probe
inevitably has an influence on this. The waveform is a function of the ESD source (charged
object), the spark electrical characteristics and the effective load (ESD probe). Some of the
energy prior to discharge is stored in capacitance between the probe shield and the charged
object, and if the discharge impedance is low this can form a low impedance resonant
circuit after breakdown. If the spark impedance is sufficiently high then this ringing can be
expected to be virtually absent. These aspects have not yet been studied.
Ringing can also be due to parasitic resonant circuits within a probe itself, rather than
being part of the ESD waveform. In our experiments the calibration waveforms (risetimes
~1–2 ns) indicate that the ESD probe is free of such problems up to several hundred MHz.
It is possible that ESD waveforms with faster risetimes may excite and reveal higher frequency
resonances within the probe. Analysis of real ESD waveforms will have to critically account
for this possibility.
5. Conclusions
A simple ESD probe has been built and calibrated at d.c. and using fast pulse measurements.
The probe is aimed at measurement of low to medium level ESD events measurements from
insulating materials, conductive objects and other electrostatically charged items. The
probe had an amplitude response of 0.5 V/A and a risetime of 1.7 ns.
Some examples of waveforms are given, obtained from a charged PTFE surface, a key
placed on a charged surface, and a charged capacitor. Clear differences in the peak current
and waveform of these waveforms are observed.
We anticipate that the ESD probe will find application in observing differences in
waveform, risetime, peak current and duration in ESD waveforms obtained in particular
situations. This will allow investigation of the effect of such differences on ignition of
flammable atmospheres, and in assessment of the risk of ESD damage to electronic
components.
References
[1] P Kathiramanathan, M J Toohey, J Haase, P Holdstock, J Laperre, G Schmeer-Lioe Measurements
of incendivity of electrostatic discharges from textiles used in personal protective clothing. J
Electrostatics 49 (2000) 51–70
[2] J N Chubb, G J Butterworth. (1982) Charge transfer and current flow measurements in electrostatic
discharges. J Electrostatics 13 pp209–214
[3] J M Smallwood 1999 Simple passive transmission line probes for electrostatic discharge
measurements. Inst. Phys. Conf. Se. 163 pp363–366
[4] M S DiCapua. (1986) High Speed electric field and voltage measurements. In: Fast electrical and
optical measurements Vol. 1. Ed. J.E.Thompson, L H Luessen. NATO ASI Series E No. 108,
Martinus Nijhoff pp 175–221

131
Predicting the maximum voltages expected on inhabited

cleanroom garments in practical use
J N Chubb—John Chubb Instrumentation, Unit 30, Lansdown Industrial Estate,

Gloucester Road, Cheltenham, GL51 8PL, UK
P Holdstock—British Textile Technology Group, Shirley House, Wilmslow Road,
Didsbury, Manchester, M20 2RB, UK.
M Dyer—William Barnet and Son Inc, PO Box 131, 1300 Hayne Street, ARCADIA,
South Carolina 29320, USA.
Abstract: The paper describes how the maximum surface voltages expected on cleanroom
garments in practical use can be predicted from measurements using corona charging of local
areas of garment fabric.
The surface voltage arising when materials are rubbed depends upon the speed and
pressure of the rubbing action and on two features of fabric performance: the time for charge
dissipation and the capacitance experienced by surface charge. These two features can be
easily measured on sample areas of garment fabrics using corona charging and measurement
of the quantity of charge transferred. The paper describes these measurements and the
studies carried out to establish the relationship between the performance of inhabited garments
and the capacitance loading performance of fabrics.
1. Introduction
This paper describes studies on the voltages that can arise on the surface of cleanroom type
garments when these are rubbed and how maximum values can be predicted by local
measurements on sample areas of the fabrics involved.
Surface voltage is the main factor determining the suitability of materials for avoiding
problems and for the constructive use of static electricity.
The two features of significance are:
1) the peak potentials that may arise
2) the time that significant potentials are present
The quantity of electrostatic charge transferred to materials by rubbing and sliding type
actions is limited by the intensity of the mechanical operation (speed and pressure) and by
the character of the materials involved. To minimise risks it is necessary to limit the
maximum potentials that may arise for the maximum quantity of charge likely to occur in
practice. Surface potentials can be limited to low values:
a) if the timescale for dissipation of charge over the surface and away to earth is
short compared to the time of separation of contacting or rubbing surfaces
b) if the charge on the surface of the material experiences a high ‘capacitance’, as
this depresses the potential that will arise per unit of charge.

132
Either, or both, of these aspects may be used to determine, and to control, the surface
potentials likely to occur in practice. These hence determine whether materials will be
suitable for particular applications and/or will give rise to risks or problems.
2. Surface voltages on inhabited cleanroom garments

2.1 Arrangements for measurements
The fist stage of present studies was to find out what surface voltage arise on inhabited
cleanroom garments in simulation of more extreme examples of normal use activities. A
person (operator) has been clothed in a number of different types of cleanroom garments
and boots and stood on an earth potential metal plate. Electrostatic charge has been separated
at a local position on the surface of the garment by striking with the end of a charge neutral
Teflon rod (‘scuff charging’ [1, 2]). The areas chosen for charging were the upper arm and
the tummy area. These were convenient for striking and for surface voltage measurement.
Surface voltage was measured using an electrostatic fieldmeter (JCI 140) mounted with its
sensing aperture 100mm directly in front of the area struck, so observations related to the
area over which charge was separated. The arrangement gave reasonable accuracy of
measurement, because at 100mm separation a 10% error in distance gives only 5% error in
reading. As the area charged is of limited size the reading by the fieldmeter, set up to show
the voltage on an extended surface, will be an underestimate of the immediate local voltage.
It does however provide a good indication of the electric fields that will arise at nearby
earthy projections, and these may be scaled with distance and projection cross-section.
The upper body of the operator was stabilised by resting an arm on a wooden support
mounted via good quality insulation on a robust tripod support stand. This helped maintain
the surface to fieldmeter separation distance stable for the extended runs (up to 20s) needed
for measuring charge decay times.
Charge transferred to the operator by scuff charging was measured by a virtual earth
charge measurement circuit (JCI 178) connected to the metal plate surface on which the
operator stood. The standing plate was supported over an earthed metal plate by discrete
insulators at the corners to avoid risks of tribocharging at the support insulation or of
signals arising from varying capacitance to nearby charged surfaces.
Measurements of charge quantities and variations of surface voltage were recorded on a
digital storage oscilloscope (Picoscope ADC-212) operated on a Notebook PC.
The garments studied were standard cleanroom coveralls and boots, of normal commercial
design, and were worn over normal shirt and trousers—as is usual practice. Features of
garment fabrics are listed on the left side of Table 1. Some garments were new (original) and
some had been laundered, as would normally apply. Laundering was 5 cycles to ISO 6330
procedure 5A at 40C, followed by a final low temperature tumble dry. Studies included a
normal cotton shirt.
All the studies on inhabited garments were made in a controlled environment of 23°C
and 40%RH in the test laboratory of British Textile Technology Group (BTTG),
Manchester, UK.
2.2 Measurement results

An example of recorded observation of surface voltage signal and charge signal is shown in
Figure 1. Surface voltage signals initially fall sharply negative and then rise to a positive

133
peak before decaying away back towards the initial level. The initial fast negative excursion
of surface voltage occurs because while the initially neutral Teflon rod is close to the
surface rubbed the separated charges are close coupled and have no influence nearby. As
the Teflon rod rises from the surface it gets nearer to the fieldmeter, so observations are
dominated by the charge on the Teflon. As the rod swings quickly away observations show
just the influence of the charge left on the garment surface. The charge signal increases from
its initial ‘zero’ level up to a steady value—the rate determined by 2Hz filtering.
Figure 1: Example of surface voltage and charge signals after scuff charging—cotton shirt (sample
19)
For each instance of striking the upper arm in front of the fieldmeter sensing aperture the
recorded observations were analysed to give values for the initial peak surface voltage, the
quantity of charge transferred and, when recordings allowed, the time for the surfaces
voltage to fall to l/e (37%) of the initial peak value.
Graphs of initial peak voltages against quantity of charge showed that, although there
was quite a bit of scatter, there was a linear relationship between initial peak surface voltage
and the quantity of charge transferred. (This was particularly clear in graphs of some earlier
scuff charging studies [2] on flat area samples of personal protective fabrics). This variation
shows there is a capacitance experienced by surface charge and the slope indicates the
value of this capacitance effect. Charge decay times were difficult to measure on inhabited
garments but it was clear that the initial peak voltages did not relate to decay times—and
certainly not to values of resistivity. The main factor determining the initial peak surface
voltage was the effective capacitance experienced by the charge separated at scuff charging.
3. Relation to characteristics of samples

It seemed plausible that the initial peak surface voltages measured on inhabited garments
might relate to capacitance loading1 values measured with corona charging on sample
areas of the fabrics of these same materials.
1
‘Capacitance loading’ is the relative capacitance experienced by charge on the material compared to that
for a similar distribution and quantity of charge on a thin layer of a good dielectric—where the capacitance
is essentially that of just the spatial distribution of charge and any influence of proximity of nearby earthy
surfaces [2]. The enhanced capacitance to the deposited charge probably arises from coupling of the
deposited charge to some structural feature in the material. This might be a relatively conductive layer or
pattern of threads or a high dielectric constant feature. Coupling may link to nearby earthy surfaces or just
to a larger effective area of material.

134
Corona charging provides the basis of easy to use instrumentation with opportunity to
control the levels and polarity of surface charging [2, 3]. This method of testing has been
shown to give very comparable values to tribocharging [2], to give results that are
reproducible and are not affected by cumulative corona exposure [4].
In making measurements on sample areas it seemed sensible to try to use comparable
quantities of charge as were experienced in the inhabited garment studies. These were in
the range 10–50nC. Results showed that charge decay times were often fairly independent
of the quantity and polarity of charge transferred. However, with capacitance loading the
values varied essentially linearly with quantity of charge and the variation was different
between positive and negative polarity. This linear variation is observed with a wide variety
of materials (not just those including conductive threads)—but the reason for this variation
is not yet clear. Examples of such variations are shown in Figure 2 below.
Figure 2: Variations of capacitance loading with quantity of charge
Examples of variations of capacitance loading with quantity of corona charge are shown
in Figure 2 above. Predicting garment surface voltages on the basis of the variation of
capacitance loading with quantity of charge would give a variation of surface voltage that
rises initially quickly and then progressively more slowly with quantity of charge. This did
not match practical experience. A more appropriate basis has proved to be use of the
minimum extrapolated value for capacitance loading at zero charge level.
When the results of fabric sample measurements were compared to the measurements
made on inhabited garments it was clear there was a definite relationship between the
maximum local surface voltage Vmax (volts) that can arise on an inhabited garment and the
quantity of charge q (nC) transferred by rubbing actions and the value of capacitance loading
measured with corona charging extrapolated to zero charge (CLq=0). The relationship is:
Vmax=f q/(CLq=0)
A value for the factor f around 75 seems the most appropriate to give the maximum voltage
values likely to arise for a given amount of charge transferred on inhabited garments.

135
Table 1 summarizes results for a variety of garments. This Table includes measurements
of charge decay times and a number of resistivity values. It also includes values for the
ratios between the average of 10 experimental surface voltage measurements on inhabited
garments and the predicted values of Vmax.
It is clear that the wide range of local initial peak surface voltages on inhabited garments
can be predicted using capacitance loading values obtained by corona charging
measurements on fabric samples. It is also clear that these initial peak voltages do not relate
to values of charge decay time or to resistivity values [5].
To use the above route to predicting maximum surface voltages it is necessary to choose
some sensible maximum value for the quantity of charge likely to be transferred in rubbing
actions appropriate to the application. Experience in the present work suggests quantities
in the range 10–50nC. The predictions of maximum peak voltage relate to local charging.
As noted earlier (section 2.1), values will be diminished for larger areas charged.
4. Conclusions
The present studies have shown that:
- the initial peak surface voltage per unit quantity of triboelectric charge (V nC-1)
provides a relevant indicator of the practical electrostatic performance of inhabited
garments. Values measured varied over a wide range for fabrics of various designs
and constructions.
- the maximum surface voltages expected on inhabited garments can be predicted
from measurements of ‘capacitance loading’ on sample areas of garment fabric
using corona charging. Measurements are needed of the variation of capacitance
loading with quantity of charge.
- initial peak garment surface voltages do not relate to charge decay times or to
surface resistivity.
- a short decay time (below ¼ s) remains an effective way to generally avoid problems,
- high capacitance loading values are needed to limit initial peak voltages to low
values
Makers of garments will benefit from the ability to predict the performance of inhabited
garments from corona charging measurements on sample areas before manufacture of actual
garments. Benefits will also arise if it proves possible to relate capacitance loading values
of fabrics to features of fabric design and construction.
The method of testing the suitability of materials described, using corona charging for
measurement of charge decay and capacitance loading, has been submitted as a New Work
Item to the International Electrotechnical Commission TC101 Standards committee.
References:
[1] J.N. Chubb “Measurement of tribo and corona charging features of materials for assessment of
risks from static electricity” Trans IEEE Ind Appl 36 (6) Nov/Dec 2000 p1515–1522
[2] J.N. Chubb “New approaches for electrostatic testing of materials” J. Electrostatics 54 (3/4)
March 2002 p233 (Presented at ESA meeting, Brock University, Niagara Falls, May 2000)
[3] J.N. Chubb “Instrumentation and standards for testing static control materials” IEEE Trans
Ind. Appl. 26 (6) Nov/Dec 1990 p1182.
[4] J.N. Chubb “Dependence of charge decay characteristics on charging parameters” ‘Electrostatics
1995’ Inst Phys Confr Series 143 p103
[5] J.N. Chubb, P.Holdstock, M.Dyer “Can cleanroom garments create electrostatic risks?”
Cleanroom Technology, March 2002 p38

136
Table 1: Results of testing on fabrics and inhabited garments
Note: Surface conductive threads (1), (2) and (3) are of different configurations. The ‘core’ conductive threads are all tri-lobal in a polyester sheath.

137
Contact charging method for the measurement of charge

decay in electrostatic dissipative materials
J Paasi, T Kalliohaka, T Luoma, R Ilmén, S Nurmi
VTT Industrial Systems, PO Box 1306, FIN-33101 Tampere, Finland
Abstract. Charge decay time of electrostatic dissipative materials is an important parameter in

evaluating the capability of a material to control static electricity. The measurement of charge
decay time, however, is not a simple task. Several methods exist, but too often the reproducibility
and reliability of results is poor. In this work we have improved a contact charging method,
based on the charging of a sample by a charged electrode, with the aim to increase the
reproducibility and reliability of the technique. Attention was paid to the way of grounding, to
sample dimensions, and to the monitoring of electrostatic potential of the sample. The improved
method is presented together with test results for different kinds of electrostatic dissipative
materials.
1. Introduction
Use of electrostatic dissipative materials is an effective way in controlling static electricity
and risks due to electrostatic discharge (ESD) in many branches of industry. The performance
of electrostatic dissipative materials in managing static electricity is usually evaluated by
the measurement of surface resistivity of the material, because resistance measurement is
simple and the recommended way by many standards, such as ref. [1]. A better parameter for
the electrostatic material evaluation, however, is the charge decay capability of the material,
because what is required for a material used in the control of static electricity is that the rate
at which charge can decay is greater than the rate at which charge is generated. The
measurement of charge decay rate or time is, unfortunately, not so straightforward and easy
as the measurement of resistivity.
Ideal measurement of charge decay rate of a material or product simulates conditions
encountered in practice. That is difficult because charge decay of electrostatic dissipative
or insulating materials depends, not only on the intrinsic material properties, but also on
the way how the charge is generated on the tested material (initial charge distribution) [2],
what is the initial density of the charge generated on or in the tested material [3], how the
material is grounded, and what is the geometric and dimensional arrangement of the system.
All these have lead to measurement techniques which are either so complex that they are
used only in research laboratories or so poorly specified that the reproducibility and reliability
of the results is poor (see ch. 2 for more details). There is a clear call in the industry for a
simple but reliable enough method for easy evaluation of the charge decay performance of

138
electrostatic dissipative materials without a necessity of investment for expensive

instruments. In this work we present a proposal for such a technique.
2. Measurement of charge decay

Existing techniques for the measurement of charge decay can be divided into four categories
depending on the way how the sample under the test is charged: contact, corona, tribo, and
induction charging techniques. In the following we review briefly the different techniques.
Definition of charge decay time of a material varies between different standards and methods
but that is not discussed here.
2.1. Contact charging technique
Probably the most commonly used method is the FTMS 101C method [4] where a sample is
clamped in a frame between two electrodes. A field meter is mounted in front of the sample
in the centre of the frame. The sample is charged and discharged through the metallic frame
connected in turn to voltage source and ground. The method is, in principle, simple but, in
practice, it is very difficult to be sure that a reading of the field meter is a truly sign of an
excess charge on a material. It may be influenced by the polarisation of the material.
Furthermore, in heterogeneous material less conducting parts of the sample may not be
charged at all by this technique.
Another widely used contact charging technique (particularly in industry) applies charge
plate monitor (CPM). In this non-standardised technique a tested material is placed on a
CPM, the material is then charged, and, finally, the material is grounded by placing a finger
of grounded, measuring person on the top of the material. The method is again simply but
the reproducibility and reliability of the method is poor, mainly due to uncertainty related
to the grounding.
A good feature of all contact charging techniques for the measurement of charge decay
is that they simulate situations often encountered in practice and that they are simple and
expensive instruments are not required.
2.2. Corona charging technique

Corona discharge is widely used for uniform charging of test samples. Corona charging
methods can be divided into two categories depending on whether the sample is grounded
during charging or not. Reproducibility of results is good. The technique is widely spread to
research laboratories but not to industry, although corona charging is the only charge decay
measurement method specified in the IEC 61340–5–1 document [1], which has a standard
status in many countries. Disadvantages of the method are related to the fact that a measurement
can be started only after a delay (after the corona brush has given way for the field meter) which
means that the true peak value of the potential is not observed. For some materials, such as
modern BSD-fabrics, that may mean a fundamental error in the result as the initial, conducting
threads dominated, exponential decay turns rapidly to hyperbolic decay [3] due to slow
charge migration in insulating base fabrics. Because of the delay, the measurement may start
only after the initial potential peak has decayed (on the other hand, the measured low initial
peak potential could represent another figure of merit than the charge decay time [5]).
2.3. Tribo charging technique
Tribocharging techniques are also widely used both in research laboratories and in industry
because a sample can be easily charged by rubbing, simulating situations often encountered

139
in practice. Reproducibility of the results, however, is usually poor. There are a few methods
for specific type of samples, such as the prEN 1149–3 method [6] for ESD fabrics, where the
reproducibility of the results is somewhat better.
2.4. Induction charging technique
Charging by induction is less used for charge decay measurements although it simulates
conditions often encountered in practice. Reproducibility of results can be good if the set-
up is carefully designed. Standardisation of induction charging methods is only in progress
[6]. Unfortunately, induction charging methods, while good for research laboratories, may
easily become too complex for industry.
3. Improved contact charging method

Our aim was to develop a simple but reliable method (thus suitable for the industry) for the
charge decay measurements without a necessity for the investment of expensive instruments.
Also the method should simulate conditions encountered in practice. Therefore we choose
contact charging as the starting point and improved the existing technique by the use of a
charge plate monitor (CPM).
The principle and measurement set-up of the improved contact charging method is
presented in Fig. 1. A sample of fixed dimensions (width and length) is placed on the metal
plate of the CPM. One end of the sample is charged, at first, to a predetermined voltage
using the CPM (or alternatively metallic plate+insulator+high-voltage (HV) power supply).
Then the charging plate is disconnected from the power supply and the other end of the
sample is grounded using an electronic HV switch and the voltage of the charged region is
monitored as a function of time by the field meter of the CPM or by an external field meter
or by a non-contact electrostatic voltmeter above the sample. In order to guarantee good
electrical contact between the charging electrode and the sample, a non-conducting and
low-charging plate should be used above the sample as a weight. A PC can be connected to
the field meter for the recording of charge decay curves, or alternatively one could rely just
to the time counter of the CPM.
The improvements we did for the technique are related to the reliability and
reproducibility of results. Grounding point of the sample was moved well apart from the
charging region because in practice a sample is seldom grounded at the same point where
Figure 1 Sketch of the improved contact charging method set-up

140
charge is generated and, thus, the charge has to migrate on or in the sample for some
distance. Secondly, sample dimensions were fixed in order to have the same ground path
length for all samples to be tested, allowing comparison between different materials. Improper
grounding of the sample is often the main reason for poor reproducibility of results.
Therefore, we paid a special attention on the way how the sample is grounded. We found it
necessary to use fast electronic switch in the making of ground connection in order to get
high reproducibility for the results. The sample was connected to the ground wire by
metallic clamps (with conducting rubber when necessary).
4. Test measurements
In the test measurements of the improved contact charging technique one key question was
whether the field meter of the CPM was self-sufficient or whether an additional field meter
above the sample was required. The main problem with using the field meter of a CPM is
that it does not measure directly the sample potential but the potential of the charging
plate. A second key question was how the results correlate with charge decay measurement
done with the corona charging technique of IEC 61340–5–1 [1].
In the tests we used a set-up consisting of a charge plate monitor according to ref. [1] with
an internal field meter (FM) of 100 ms response time, an electrostatic voltmeter (EVM) of 3 ms
response time above the sample, a polycarbonate plate above the sample with a measuring
window for the voltmeter probe. Hundreds of tests with different kinds of samples and conditions
were done. In the measurements presented in this work we used five different kinds of samples:
100% cotton, ESD textile (PES with carbon fibre grid of 3.5 mm ×3.5 mm, standard corrugated
cardboard (E-flute), ESD cardboard (E-flute with carbon black film and sealer on the surface),
and ESD plastic. Sample dimensions were fixed to 15 cm×30 cm. The measurements were
done at 50 % RH and 23°C. The samples were conditioned 72 h prior to measurements. In the
tests the sample was charged up to about 1200 V and charge decay time was determined either
as the duration from 1000 V to 100 V (10%) or to the 1/e value from the maximum (37%).
Possible residual charge on the sample was neutralised before each measurement.
Typical charge decay curves for slow and fast decay time materials are shown in Figs 2
and 3, respectively (100% cotton and ESD plastic). From the figures we can see that for
sufficiently slow decay rates (decay time »100 ms) the field meter reading of the CPM is
Figure 2 Example charge decay curves of a slow decay time material (100% cotton)

141
Figure 3 Example charge decay curves of a fast decay time material (ESD-plastic)
equivalent to the true sample potential measured by the EVM (upper FM). Only for fast
decay materials (decay time <100–200 ms) a field meter above the sample, and faster than
the FM of the CPM, is required for accurate measurements.
In order to compare the improved contact charging technique with other methods we
made a comparison test with the corona charging method and surface resistance measurements
[1]. Also, with an induction charging method, using slightly modified contact charging set-
up of Fig. 1, placing a polycarbonate frame was placed between the charging plate and the
sample. The procedure in the induction charging measurements was the same as in the
contact charging measurements using only the EVM above the sample for the sample
potential monitoring. Results of the measurements are summarised in Table 1. Each result
is an average of 10 measurements with both positive and negative charging/discharging
procedures. From the results we can see that the charge decay times from the contact
charging measurements are in most cases in line with the corona charging and resistance
measurements. Deviations are discussed in section 5.
5. Discussion
The improved contact charging method seems to be simple and reliable method for the
evaluation of charge decay performance of a material. If a charge plate monitor is available,
the field meter of the CPM gives sufficiently good results for industrial needs (for research
purposes a faster field meter would be preferable). This means that expensive new investments
are usually not required, which would make the technique attractive for the industry.
Table 1 Comparison of techniques
*Residual charge of 100-200 V

142
Comparing the measured charge decay times to those observed by corona and induction
charging techniques and to the resistive measurements, the correlation is good for the
tested cardboard and plastic materials. Only for fabrics (cotton and ESD-fabric) the
measurements gave different results. Taking into account that the contact and corona
charging techniques simulate different kinds of practical situations, we cannot say that one
method gives correct results while the other does not. For example, for the ESD-fabrics the
corona method was not able to give any decay time result. The fast initial, conducting
threads dominated decay which was almost over before the measurement started and the
field meter measured mainly the charge remaining on the insulating base fabric with very
long decay time (thousands of seconds). The contact charging technique could not detect
the residual charge on the insulating base fabric because the charge migrated from the
insulating surface via the metal electrode to the grounded, conducting threads of the BSD-
fabric. That led to fast charge decay times and demonstration of good material performance
which was contrary to the results of the corona technique measurements. Thus the techniques
give complementary information on the material performance, necessary for comprehensive
understanding of the electrostatic properties of a material.
Charge decay time of a material is related to the product of the resistance and capacitance
of the measurement system. Therefore the sample dimensions as well as distance to ground
planes should be fixed. If using a CPM, the system capacitance would be around 20 pF for
electrostatic dissipative materials. Thus, when the capacitance is fixed, the results are
determined, in practice, by the resistance of the ground path for the charge and possible
errors and limitations in the electric field (or potential) measurement.
6. Conclusions
We have developed an improved method for simple but reliable testing of charge decay
properties of electrostatic dissipative materials based on charging by a charged electrode.
The capability of the method has been successfully tested in hundreds of measurements
with different kinds of materials used in the control of static electricity in industry and by
comparative tests with other methods for charge decay measurements. The improved contact
charging method seems to be a suitable technique for industrial use as well as in research
laboratories.
Acknowledgements
The work is partially supported by the Finnish National Technology Agency Tekes via the
STAHA programme.
References
[1] International Electrotechnical Commission 1998 Technical Report 61340–5-1 Electrostatics—
Protection of electronic devices from electrostatic phenomena—General requirements
[2] Crowley J M 1990 The electrostatics of static-dissipative worksurfaces J. Electrostatics 24
221–237
[3] Seaver A E 2002 An equation for charge decay valid in both conductors and insulators Proc.
ESA-IEC Joint Meeting 2002 349–360
[4] US Federal Test Method 1982 Standard 101C Method 4046.1
[5] Chubb J 2002 New approaches for electrostatic testing of materials J. Electrostatics 54 233–244
[6] European Committee for Standardization 2001 Draft prEN1149–3 Protective clothing—
Electrostatic properties—Part 3: Test methods for measurement of charge decay

143
A particle charge spectrometer for determining the

charge and size of individual dust grains on Mars
Stephen Fuerstenau and Gregory Wilson

Jet Propulsion Laboratory, California Institute of Technology 4800 Oak Grove
Dr., Pasadena CA 91109
Abstract. Airborne dust is a primary factor influencing climate on Mars. Electrostatic charge
carried by dust particles can play a role in their suspension, settling, and adhesion to spacecraft
surfaces. Furthermore, the possibility for atmospheric discharges on Mars and the likelihood
of associated hazards to robotic exploration remains an open and important question. In order
to understand these aspects of the Martian surface environment, NASA is developing a
portable instrument capable of measuring the charge and size of airborne particles. In the
Particle Charge Spectrometer (PCS) the measurement of charge is based on direct capacitive
coupling to a charge sensitive pre-amplifier. Aerodynamic particle diameter in the range from
2 to 100 micrometers is determined from the particle velocity in the sample inlet tube. Currently
the field-portable device can measure charge on single particles with an error of less than 180
electronic charges and at a rate of over 200 particles per second. The goal of the research is to
provide this measurement capability in an instrument that weighs under a kilogram and will
operate in the Martian atmosphere with just a few watts of power. This paper describes the
instrument and preliminary data from laboratory and field tests. The measurement approach is
presented in the context of earlier particle charge instruments.
1. Introduction and background

1.1 Motivation
As with precipitation particles on Earth, airborne dust grains are theorized to play a dominant
role in the as yet undetermined electrification of Mar’s atmosphere (Eden and Vonnegut
1973, Farrell et al. 1999). Dust devils, the rotating dust-laden plumes of thermally unstable
air that are a frequent phenomenon on Mars (Thomas and Gierasch 1985, Edgett and Malin
2000), have been observed on Earth to induce strong perturbations in local electric fields
(Frier 1960, Crozier 1964, Kamra 1969) due to the electrostatic charge carried by the dust.
The electrostatic charge can play an important role in the mechanisms of particle suspension,
transport, sedimentation, and capture on surfaces, influencing the rates at which dust is
injected and removed from the Martian atmosphere. Knowledge of such rates is crucial to
understanding the Martian global dust cycle and climate. Moreover, atmospheric lightning,
surface charging and discharging, interference with radio communication, and enhanced
adhesion to materials, may all pose significant hazards to future robotic and human

144
exploration missions. In response to these issues NASA has listed the measurement of the
charge on dust particles in the size range from 1 to 100 microns as a priority science
objective. Initially we sought to simply adapt a previous instrument for use on Mars.
However, we quickly arrived at the conclusion that the measurement of charge on suspended
particles in this size range is by no means a routine measurement and remains a largely
unexplored parameter of dust-laden environments on Earth.
1.2 Particle Charge Measurement

Aerosols generated by mechanical means, for example sprays and dust lifted by wind, are
almost invariably highly charged, in contrast to particles that form by condensation such as
fogs and smokes which typically carry only a few unit charges at most. Numerous aerosol
instruments based on mobility analysis rely on this charge to size particles smaller than one
micron. Above this size mobility analysis becomes impractical without the use of high
voltage (>1 kV). Despite the importance of the substantial electrical charge carried by
larger particles to fields such as atmospheric science and powder technology, very few
instruments have been developed for the purpose of surveying charge on individual particles
in the sub pico-coulomb range, ie. 10 to 1,000,000 charges (Brown 1997).
Those methods that do look at individual particles tend to sacrifice speed of analysis for
sensitivity. A well-known case in point is the famous oil droplet experiment in which
Millikan was able to determine the charge on individual particles with the precision of a
single electron (Millikan 1910). This approach requires a microscopic inspection of the
size and mobility, or velocity, of drifting particles. Kunkel employed strobe photography
in a similar set-up to obtain the charge on dust particles in the 5 to 15 µm (Kunkel 1949).
The approach required among other constraints, extreme care with reduction of small
perturbations in air flow and temperature gradients. The E-SPART instrument monitors the
motion of particles in response to oscillating acoustic and electric fields with laser doppler
velocimetry (Mazumder 1991). While the instrument is able to measure size and charge on
particles at count rates of up to 200 particles per second, the equipment is reported to
possess an extreme size and weight, more than 50 Kg. Furthermore, the low pressure
atmosphere on Mars would not sustain potentials required by these devices above a few
hundred volts without electrical discharge.
2. Instrument Description and Function
2.1 Measurement Approach

The PCS instrument is based on direct sensing of the charge on particles as they pass
through a tube electrode that acts as a Faraday cage. As particles enter and exit the conducting
tube they induce a change in potential in the tube, due to image charge forces, that can be
detected with a charge sensitive preamplifier connected to the tube electrode. The magnitude
of the induced signal is proportional to the charge carried by the moving particle and has

145
been analyzed in detail (Weinheimer 1988). The small size of the cylindrical electrode
(1mm×7 mm) aids sensitivity by ensuring that the capacitance is low. This strategy for
measuring charge has previously been applied for single micro-particles with relatively
high amounts of charge (q>10–14 Coulombs) (Vercoulen 1995), and with much lower charge
levels (q<10–17 Coulombs) in experiments to ‘weigh’ individual nanoparticles of biological
interest (Fuerstenau 1995, 2001). The Faraday cage approach to measuring charge on cloud
droplets dates back to at least the 1940’s (Gunn 1949) but most of the numerous airborne
instruments that employed it have exhibited limits of sensitivity not less than 0.1 to 1 pico-
Coulombs.
Figure 1. Photograph of the remote front end of the charge sensitive preamplifier. The coated glass
capillary (0.5 mm I.D. bore) enters and exits from the side of the box and passes through the 7 mm long
tube electrode at the center of the housing. Diagram (right) of bench-top PCS system with Power supply,
sensor head, pulse processing electronics enclosure, and computer.
The PCS instrument has two unique features. Unlike most of the airborne predecessor
instruments, the PCS utilizes a small air pump to sample particles with a high-speed flow.
Particles accelerate in the 100 m/s flow at different rates according to their aerodynamic
diameter. Larger particles require a longer time to accelerate and hence a longer time to
travel the length of the tube electrode. The travel time, which is inherent to each measured
particle event, is a monotonically increasing function of the particle size. The PCS calculates
an aerodynamic particle size based on the measured transit time. A similar velocity-based
particle sizing approach is implemented at the exit of a super-sonic jet nozzle in several
commercially available instruments using laser light scattering (Dahneke 1990). In the
PCS the imposed flow allows information on both charge and size to be determined from
the shape of the amplified pulse. A second feature of the instrument is the use of a glass-wall
capillary to de-couple the flow geometry from that of the electrode, obviating the need for
careful alignment of the Faraday tube with the gas flow. The approach works because
electrostatic charge can be sensed across a dielectric barrier, and can be implemented
without introducing noise, even at the extremely sensitive levels displayed in our device.
The glass-wall method has been used extensively for sensing charge on powder particles
(Gajewski 1975, 1993), but all of these experiments involved clouds of powder and not
individual particles.

146
2.2 Calibration and Performance

The noise level of the charge sensitive preamplifier is currently less than 200 electrons. The
minimum charge detectable is 300–400 electrons and the maximum charge at this sensitivity
is +/-120,000 electrons. The charge equivalent noise is determined by pulsing a known
voltage across a calibrated capacitor onto the gate of the FET which is in contact with the
tube electrode. The pulsed capacitor introduces a known amount of charge onto the
preamplifier input, which in turn induces a measurable voltage pulse at the output inversely
proportional to the feedback capacitance of the amplifier. The procedure results in a measure
of the system output voltage as a function of input charge. The capacitance of the tube
electrode assembly and all feedback elements are kept small in order to reduce noise. It is
estimated that cooling the amplifier will reduce the intrinsic noise of the FET by perhaps
30%. The correlation of particle transit time to particle size is currently based on a simple
model for particle motion in a one-dimensional flow. The size response of the PCS has been
tested with charged particles of known size generated with a vibrating orifice droplet
generator. At present particles of equal size can pass through the tube electrode with velocities
varying by as much as a factor of two. Therefore, the PCS instrument at the moment gives
only a rough indication of particle size. Size measurement of particles in the range from 1
to 200 micrometers is theoretically possible if the particle transit time can be made more
uniform by confining all particles to the center streamlines of the flow.
In the data system is based on analog signal processing. Signals from the preamplifier
are fed to a pulse processing board that further amplifies them and reduces the pulse
amplitude and duration to two analog signal levels. These are relayed on a single channel
to the data acquisition card of a lap-top computer. A ‘LabView’ (National Instruments)
program acquires the information for up to 200 individual particles per second and displays
them after processing in a continuously updated graph of charge vs. particle size. The
prototype system is shown in figure 1.
Figure 2. Charged airborne dust recorded June 2002 at Eloy, Arizona with the PCS in the vicinity of a
dust devil. Data are consistent with a strong negative electric field observed at the time of collection.
Reported diameter is based on particle velocity in the sampling tube. The error in the measured particle
size is currently at least 100 % (see above discussion.)

147
2.3 Sample Data

We have used the PCS in the laboratory to investigate charge on particles from various
agitated dusts and electrostatic sprays. Figure 2 illustrates a data set collected in the desert
during the passage of a nearby dust devil. Noteworthy is the fact that most particles carried
a negative charge, as large as 50,000 e- for some particles. A vertical electric field of over 20
kV per meter was observed at ground level when these data were collected. The true charge
distribution is difficult to determine from this data set because the minimum charge
discrimination level was set fairly high for the negative particles and weakly charged
particles could not be detected. Future observations should reveal if the charge distribution
of recently suspended dust is consistent with that predicted by for triboelectrically induced
charge on dust clouds and if it parallels that for other particle charging processes in the
atmosphere.
3. Future Developments and Applications

We are developing an improved sample inlet system based on dilution. It will increase by
a factor of 10 to 30 the maximum particle concentration sampled by the 1 liter per minute
flow of the PCS. We are also exploring the performance of the system at the 10 mbar
pressure of the thin Martian atmosphere to determine any variation in the sizing capability
of the instrument. Compact versions of the device are being developed for possible
deployment on balloons or aircraft to aid studies of thunderstorm electrification. It is
anticipated that the PCS will find application in agricultural investigations aimed at better
understanding crop dusting and artificial pollination. The charge carried by industrial
powders, such as pharmaceuticals during processing and toners inside working xerography
machines, may eventually be observed with the PCS instrument.
Acknowledgement
The authors are grateful to Mr. Norman Madden for vital contributions to the design of the
amplifier and pulse processing electronics. This research was carried out at the Jet Propulsion
Laboratory, California Institute of Technology, under a contract with the National
Aeronautics and Space Administration.
References
Brown, R.C. “Tutorial Review: Simultaneous Measurement of Particle Size and Particle Charge” Journal
of Aerosol Science, 28, no. 8, pp.1373–1391, 1997
Crozier, W.D., The electric field of a New Mexico dust devil, J. Geophys. Res., 69, 5427, 1964.
Dahneke, B. “Beam forming and sensing apparatus for aerodynamic particle sizing system” United
States Patent No. 04938592 July 7, 1990
Edgett, K.S., and M.C. Malin, Martian dust raising and surface albedo controls: Thin, dark (and
sometimes bright) streaks and dust devils in MGS MOC high resolution images. Lunar Planet. Sci.
XXXI, Abstract No. 1073, Lunar and Planetary Institute, Houston, Texas, March 2000.

148
Farrell, W., Kaiser, M., Desch, M., Houser, J., Cummer, S., Wilt, D., and Landis, G., “Detecting
Electrical Activity from Martian Dust Storms,” Journal of Geophysical Research—Planets, Vol. 104,
No. 2, 3795–3805, Feb. 25, 1999.
Frier, G.D., The electric field of a large dust devil, J. Geophys. Res., 65, 3504, 1960.
Fuerstenau, S.D. and Benner, W.H. “Molecular Weight Determination of Mega-Dalton DNA Electrospray
Ions using Charge Detection Mass Spectrometry” Rapid Communications in Mass Spectrometry, V. 15,
pp 1528–1538 (1995)
Fuerstenau, S., Benner, W., Thomas, J., Brugidou, C., Bothner, B., and Siuzdak, G. “Mass Spectrometry
of an Intact Virus” Angew. Chem. Int. Ed., 40, No. 3, 541–544, 2001.
Gajewski, J.B. “Static electricity and measurement of the parameters of a flow in pneumatic conveyances”
Proceedings of Electrostatics 1975
Gajewski, J.B. Glod, B.J. and Kala W.S. ‘Electrostatic Method for Measuring the Two- Phase Pipe
Flow Parameters’ IEEE Trans, on Industrial Applications Vol 29, No 3, 650–655, 1993
Gunn, R., Electronic apparatus for the determination of the physical properties of freely falling raindrops,
Rev. Sci. Inst., 209 291–296, 1949
Kamra, A.K., Electrification in an Indian dust storm, Weather, 24, 145, 1969.
Kunkel, W.B. and Hansen, J.W. “A dust electricity analyzer” Rev. Sci. Inst., v 21, no. 4, p.308, 1949
Millikan, R.A.“A new modification of cloud method of determining the elementary electrical charge and
the most probable value of that charge”. Phil. Mag., 10, pp. 209–228
Mazumder MK, Ware RE, Yokoyama T, Rubin BJ, Kamp D,

IEEE Transactions On Industry Applications, 27 (4): 611–619 Jul-Aug 1991
Thomas and Gierasch (1985). Dust devils on Mars. Science 230, 175–177.
P.H.W. Vercoulen, Electrostatic processing of particles, Ph.D. Thesis, Delft Univ. of Technology Delft,
The Netherlands, 1995.
Weinheimer, A.J., “The charge induced on a conducting cylinder by a point charge and its application to
the measurement of charge on precipitation” J. Atmospheric and Oceanic Technology, 5, p. 295, 1988

149
Measurement of Optical Intensity and Fluence Generated

by Spark Discharges
J C Crager† and M N Horenstein

Department of Electrical and Computer Engineering, Boston University, 8 Saint
Mary’s Street, Boston, MA 02215, USA.
Abstract. This paper presents the results of an investigation into the optical intensity and
fluence emitted by spark discharges as a function of energy dissipation in the spark gap.
Optical fluence is defined as the integral of light intensity with respect to time. The experiments
reported here involve sparks generated by high-voltage energy stored in fixed capacitors.
These results will be used to develop a theory that relates spark energy to emitted light. The
ultimate goal of the work is to develop a reliable method for measuring and predicting the
energies of sparks that originate on insulating surfaces.
Experiments were conducted in air at atmospheric pressure for capacitive discharges
between two 32-mm diameter brass spheres at stored energies ranging from 5 to 400 mJ, gap
lengths ranging from 0.5 to 3.5 mm, and capacitor values ranging between 1.1 and 4.4 nF.
Simultaneous measurements of discharge current and emitted light intensity were made using
a current transformer and a silicon photodiode detector, respectively. These measurements
indicate that the instantaneous light intensity is nearly proportional to the magnitude of the
discharge current. Under the assumption that the total spark energy is proportional to the
energy stored in the capacitors, the results show that the optical fluence varies almost linearly
with where C is capacitance and Vb is the breakdown voltage of the spark gap. The
relationship between the optical fluence and the product of Vb and the integral with respect to
time of the squared discharge current was also found to be nearly linear. The latter is related to
the electrical power delivered to the spark.
1. Introduction
With the growing use of plastics and other electrically insulating materials, the problems
associated with electrostatic discharges are of increasing concern. The unintentional
generation of static electricity is a common occurrence in industrial as well as in domestic
activities. The manifestations of static electricity in our everyday lives can range from
harmless nuisances like an electrostatic shock upon touching a door knob to severe
production losses in manufacturing facilities (e.g., semiconductor devices, photographic
film, etc.) to major hazards (e.g., explosions) that result in loss of life. Although it is well
known that discharges from electrostatically-charged insulators can have sufficient energy
to ignite combustible mixtures [7], there is little quantitative information on their character
and incendivity—that is, their ability to cause explosions. Thus there is a need to better
† E-mail: jccrager@bu.edu.

150
Figure 1. Experimental setup.
understand the phenomenon of electrostatic discharge in order to develop the appropriate

countermeasures to reduce hazards and production loss.
The work presented here investigates an optical technique for measuring and predicting
the energy dissipated in spark discharges. At present we are studying the light intensity and
optical fluence (light intensity integrated over time) of sparks between conductors generated
by high-voltage energy stored in fixed capacitors. According to [3], when a capacitor is
discharged through a spark gap, more than 95% of the energy stored in the capacitor is
released in the spark. The energy dissipated in these types of sparks is approximately
, (1)
where C is capacitance and V is the charging voltage. Unfortunately, there is no analog to

equation (1) for approximating the energy dissipated in insulator-conductor or insulator-
insulator discharges. It is therefore the purpose of this preliminary study to develop a
theory, backed by experimental results, that relates the emitted light to the spark energy in
order to develop a non-invasive optical method for measuring the energy of sparks, regardless
of their place of origin. In future work, we intend to extend the results of this investigation
to sparks originating on charged insulating surfaces.
2. Experimental methods
Figure 1 shows a diagram of the experimental setup. The spark gap consisted of two 32-mm
brass spheres in air at atmospheric pressure mounted on brass rods that allowed for adjustment
of the gap. The circuit used to energize the spark gap was comprised of a capacitor C
charged slowly through a series resistor R by a high-voltage power supply V0 until breakdown
occurred. Four 1.1-nF capacitors were combined to provide values of capacitance ranging
from 1.1 to 4.4nF. Gap lengths ranged from 0.5 to 3.5 mm with breakdown voltages between
2.7 and 13.3 kV, respectively.
A high-voltage probe and digital multimeter were used to measure the breakdown voltage.
The series resistance R=1GΩ was chosen so that the time constant of the charging circuit,
which was on the order of a few seconds, was much larger than the time it took to discharge

151
Figure 2. Photodiode circuit. Figure 3. Output circuit of CT.
the capacitors when breakdown occurred. Simultaneous light intensity and current
measurements were made, and the respective signals were captured, using a Hewlett-Packard
54616B digitizing oscilloscope (bandwidth: 500 MHz; sampling rate: 2 GSa/s). Optical
measurements were made using a high-speed Si PIN photodiode (Hamamatsu S1223 series)
with a 320 to 1100nm spectral response. The photodetector circuit is shown in Figure 2
where Vr=20V, R1=10kΩ, C=100µF, and R2=50Ω. The discharge current was measured using
a current transformer (CT) having a secondary winding comprised of three turns of 16-AWG
wire wound on a toroidal core (core material: nickel-zinc ferrite, material 61; CT inductance:
0.35µH). A 5–Ω resistor connected in parallel across the secondary terminals of the
transformer formed an RL integrating circuit (Figure 3). The current and light-intensity
signals measured during each discharge event were stored in a computer for subsequent
calculations.
The current and light intensity of sparks were measured using the CT and optical detection
system for stored energies in the range 5 to 400 mJ, breakdown voltages in the range 2.7 to
13.2kV, and capacitance values in the range 1.1 to 4.4 nF. The software programs Agilent
VEE Pro and MATLAB were used for calculations and data analysis. Typical current and
light-intensity waveforms are shown in Figure 4. The damped oscillatory shape of the
current waveform is an artifact of the external circuitry which, to a certain approximation,
behaves like an underdamped series RLC circuit when the spark gap is conducting. Inspection
of the waveforms shows that the instantaneous light intensity varies roughly with the
magnitude of the discharge current [2, 4]. Given the fundamental dependency of power
dissipation on arc current, it is therefore reasonable to assume that the light intensity is in
some way related to the power dissipation in the spark [5].
Optical fluence H is defined by the integral
(2)
where I is the light intensity and τ is the duration of the signal. Figures 5 and 6 show the
dependence of H on capacitance and breakdown voltage, respectively. We have found that
H varies almost linearly with capacitance for fixed breakdown voltages. This result can be
explained by noticing that the durations of the current and light intensity signals (Figure 4)
are roughly the same. Because the duration of the current in a damped RLC circuit increases

152
Figure 4. Typical light and current waveforms.
Figure 5. Fluence dependence on capacitance. Figure 6. Fluence dependence on breakdown

voltage Vb.
linearly with increased capacitance, for fixed breakdown voltages, the duration of the light
intensity and H will also increase linearly with capacitance. As suggested by Figure 7, we
have also found that H increases to the third power of the breakdown voltage for a fixed
capacitance. For the examined energy range, an empirical formula for the optical fluence is
given by
, (3)
where a is an experimentally-determined constant, C is the energy-store capacitance, and

Vb is the breakdown voltage of the spark gap (Figure 7). Figure 8 shows that H is nearly
proportional to the product of Vb and the integral over time of the squared discharge
current. These results are in agreement with those found in [6, 8], where the fluence was
described by the equation
, (4)
where n was said to vary between 2.5 and 4.

153
Figure 7. The relationship between fluence and Figure 8. The relationship between fluence and
0.5CVb3 for gap lengths ranging between 0.5 and the product of Vb and the integral of the current
3.5 mm for four different capacitance values. squared for gap lengths ranging between 0.5 and
3.5 mm for four different capacitance values.
4. Suggested theory
One possible theory that may explain the result given by (3) begins by considering an
established spark channel. Assume that the plasma in the spark channel is both in local
thermodynamic equilibrium (LTE) [1, 8, 9] and is electrically neutral, that is, ne=n+ where
ne and n+ are the electron and positive ion concentrations, respectively. The light intensity
I due to radiative recombination can be described by the rate equation
, (5)
where β is the electron-ion recombination coefficient, A is the cross-sectional area of the

channel, d is the spark gap length, and n=ne=n+ is the electron (positive ion) concentration.
Equation (5) can be written in terms of the electron current density J=en by substituting for
n2, resulting in
. (6)
Here e is the unit electronic charge, υ is the electron drift velocity, and i is the current in the
channel. According to [10], the current density in a spark channel comprised of an LTE
plasma can be expressed in terms of the fraction of ionized molecules f(T) as
J=f(T)E, (7)
where T is the local plasma temperature and E is the electric field in the spark channel. The
fraction of ionized gas molecules f(T), also called the degree of ionization, can be obtained
from the Saha relation [10] and is defined as the ratio of electron concentration to original
molecule concentration. From equations (6) and (7), the light intensity can now be written as
, (8)
where Vch=Ed is the voltage drop across the channel, and iVch defines the instantaneous
power dissipation in the channel.

154
For a fully-developed spark channel in LTE, both the electric field in the channel and
f(T) are assumed to be constant with respect to time. Integrating (8) with respect to time to
obtain the fluence results in
, (9)
where ε is the energy dissipated in the spark column. Since more than 95% of the energy
stored in the capacitors is believed to be released in the spark [3], we can approximate ε
using (1), leading to
, (10)
where C is the energy-store capacitance, and Vb is the breakdown voltage of the spark gap.
Because an increase in Vb is accompanied by an increased current and thus an increased
degree of ionization f(T) [10], then given (10), it is reasonable to assume that
, (11)
which is in agreement with the experimental results described by equation (3).
5. Conclusion
Although the assumption of constant drift velocity in the spark is only approximate, the
theory does suggest that the light intensity is not proportional to just the power dissipation,
but it is proportional to the power dissipation weighted by a function related to the degree
of ionization in the spark channel. The weighting function would explain why the optical
fluence is not directly proportional to the stored energy, but is instead proportional to .
References
[1] Bazelyan E M and Raizer Y P 1998 Spark Discharge (New York: CRC Press) 38–42.
[2] Craggs J D and Meek J M 1946 The Emission of Light from Spark Discharges Proceedings of
the Royal Society of London A 186 241.
[3] Cross J A 1987 Electrostatics: Principles, Problems, and Applications (Bristol: Adam Hilger)
136.
[4] Flowers J W 1943 The Channel of the Spark Discharge Physical Review 64 Numbers 7 and 8
225–235.
[5] Greason W D, Kucerovsky Z, Bulach S, and Flatley M W 1997 Investigation of the Optical and
Electrical Characteristics of a Spark Gap IEEE Trans. Incl. Applicat. 33 Number 6 1519–1526.
[6] Kornetzki V M, Fomin V, and Steinitz R 1933 Messung der Funkenhelligkeit und Funkendauer
Z Tech Phys 14 274–280.
[7] Lüttgens G and Wilson N 1997 Electrostatic Hazards (Oxford: Butterworth-Heinemann).
[8] Meek J M and Craggs J D 1953 Electrical Breakdown of Gases (London: Oxford University
Press) 410–411.
[9] Uman M A 1984 Lightning (New York: Dover) 153–159.
[10] von Engel A 1994 Ionized Gases (New York: American Institute of Physics Press) 82–84,
272–273.

155
Atmospheric ion spectra and the rate of voltage decay of

an aspirated cylindrical capacitor
Karen L Aplira
Space Science and Technology Department, Rutherford Appleton Laboratory,
Chilton, Didcot, Oxfordshire OX1 1 OQX
Abstract. Aspirated cylindrical capacitors are commonly used instruments for atmospheric
ion measurements, more of which are needed for climate and pollution studies. However the
classical theory of one operating mode of this instrument is based on an approximation, which
may introduce errors in ion measurement. A modification to the classical theory of the instrument
is consequently proposed. Applying this modification enables ion mobility spectrum information
to be extracted from the rate of voltage decay of the aspirated capacitor in air. It can also be used
to improve air conductivity measurements.
1 Introduction
The ventilated cylindrical capacitor or Gerdien condenser [1] is a long-established
instrument for atmospheric ion measurements. It consists of a cylindrical outer electrode
containing a coaxial central electrode, with a fan to draw air through the electrodes. With an
appropriate bias voltage applied across the electrodes, a current flows in proportion to the
ion concentration. Early ion measurements inferred the current from the rate of decay of
voltage across the electrodes, using a gold-leaf electrometer [2]. As electronics technology
developed, this technique was superseded by direct measurements of the current.
Contemporary instruments based on this principle deploy modern electronics and computer
control to use both the Current Measurement and Voltage Decay measurement modes for
self-calibration [3,4].
Measurements with modern instrumentation showed that although generally comparable,
the two modes were not always completely consistent [3]. This motivated reconsideration
of the classical principles of the Voltage Decay mode. The modified approach described
here improves conductivity determination and enables ion mobility spectra to be retrieved
from voltage decay measurements. Improved ion measurements are needed for solar-terrestrial
physics, pollution studies, and assessing long-term changes in atmospheric electrical
parameters related to climate change.
2 Classical theory of air conductivity measurement

Atmospheric conductivity is proportional to both the atmospheric ion concentration n, and
the average mobility µ of the air ion population. Molecular ions with µ>0.5 cm2V”1s”1 are
conventionally defined as “small ions” [5]. The unipolar air conductivity σ can be written as
(1)

156
where e is the charge on the electron. Measurements averaged over some years show the
typical shape of the average small ion mobility spectrum [5]. The mean mobility is defined
as the mode of the distribution [6], usually 1.3–1.6 cm2V”1s”1 for positive and 1.3–1.9
cm2V”1s”1 for negative ions, (the difference arises because of the varying chemical composition
of positive and negative ions) [5, 6]. Typical values of atmospheric conductivity are 5–100
fSm-1, depending on pollution levels [7].
To derive a simple expression for atmospheric conductivity, it is convenient to assume
that all atmospheric ions have the same mobility, and (1 is commonly simplified to
σ±=en±µ±. (2)
For a classical Gerdien condenser in Current Measurement mode, if the ions reaching the
central electrode are constantly replenished, air conductivity is proportional to the ion
current measured at the central electrode. (The “conductivity measurement regime” requires
an adequate ventilation speed and bias voltage, and can be verified by a linear response of
measured current to a changing bias voltage). Using (1, and considering the charge arriving
at the central electrode per unit time results in the routinely-used equation for calculating
bipolar atmospheric conductivity from current measurements with a Gerdien condenser,
considered as a classical capacitor with capacitance C [8, 9].
. (3)
In the Voltage Decay mode, a voltage across the electrodes will decay due to the current
i flowing through the air, which has a large resistance R. If the charge on the capacitor is Q,
elementary circuit analysis gives
(4)
from which the familiar expression for the exponential decay of charge from a capacitor to
a voltage at time t, Vt from an initial voltage V0 with time constant τ, is
. (5)
The current of negative ions from a positively charged capacitor can be derived from
Gauss’s Law [8], when the permittivity of free space is ε0 as
. (6)
Since current i is equal to rate of change of charge dQ/dt then
(7)
By analogy with (5, for a Gerdien condenser in Voltage Decay mode σ is given by
. (8)
(8 has been the standard expression for calculating air conductivity from voltage decay
measurements throughout the history of the Gerdien condenser instrument, using τ
determined from a time series of voltage measurements [2,3,8]. (4 assumes that the resistance
of air is constant, to give an exponential rate of decay of charge from a Gerdien condenser.
However, measurements show that a non-exponential decay rate is commonly observed
[3,10], suggesting this approximation may not be universally appropriate.

157
3 Modification to the classical theory of the Voltage Decay mode

The minimum mobility of ion contributing to the air conductivity measurement, the critical
mobility µc, [e.g. 8] is often approximated from the ventilation speed through a Gerdien
condenser u, its length L and electrode radii a and b and the bias voltage V:
(9)
where k is a constant related to the size of the Gerdien condenser
(10)
The functional dependence of µc on the bias voltage has been exploited to compute ion
mobility spectra from the changing ion current at the central electrode [11].
During a Voltage Decay measurement the voltage across the electrodes decreases; from
(9,µc also varies during the decay. The number of ions contributing to the measurement
therefore changes as a function of the decay voltage. This is analagous to voltage decay
through a voltage-dependent resistor R(V) in a parallel RC circuit, and (4 can be rewritten
to include this,
(11)
Solving (11 gives a modified form of the classical decay, in which Vt is implicitly defined:
(12)
R(V) and the shape of the decay curve are related to the shape of the ion mobility spectrum.
If all ions have the same mobility (referred to here as a flat ion mobility spectrum) R(V) is
constant during the decay and (12 reduces to (5, the classical exponential decay. For all
other spectra, (here called variable ion spectra), R(V) is not constant and there is no
theoretical basis on which expect an exponential decay. This is the general case in the
atmosphere [5].
4 Inversion of prescribed mobility spectra
The voltage decays expected from prescribed atmospheric ion mobility spectra can be
calculated by numerical integration of (12 [7] (Figure 1). The simplest ion spectrum is flat
(Figure 1a). As the number of ions in each mobility category is constant, the effective
resistance does not change during the decay, and the voltage decay is exponential (Figure
1b). Another simple shape of mobility spectrum is based on existing average ion spectra [5]
(Figure 1c). The voltage decay from this spectrum (Figure 1d) is not exponential, with only
73% of its variance explained by an exponential model. A typical voltage decay time series
in atmospheric air would only yield an exponential decay to a first approximation.
Figure 2 shows typical voltage decay time series in atmospheric air. The instrumentation was
optimised for ultra-low leakage [3,4], and a calibrator was developed specially for the novel
electrometer used in the experiment [12,13]. Over the entire data set with 39 voltage decays
measured over two weeks, the mean r2 was 0.887 with a standard deviation of ±0.134 (the
range of values was 0.310–0.998). This non-exponential behaviour is consistent with many
independent observations of a variable ion spectrum in atmospheric air [4,5,6,7,8,11].

158
Figure 1 Sample ion mobility spectra with equivalent voltage decays. Ion mobility spectra (left) are
simplified and replaced by four equal mobility bands, with the average mobilities indicated. Concentrations
and mobilities are typical for surface atmospheric levels. Calculated voltage decays (right) assume a
Gerdien condenser with dimensions given in [3]. a) A flat ion spectrum inverts to an exponential decay,
c). b) a Gaussian ion spectrum inverts to a non-exponential decay, d).
Figure 2 Consecutive voltage decay time series measured in atmospheric air on June 12 1998, showing
fits to the classical exponential approximation, and coefficient of determination (r2) values.

159
5 Computing ion spectra from voltage decay measurements

As the deviation of voltage decay curves from the exponential is related to the atmospheric
ion spectrum, the shape of the voltage time series contains spectral information. Every
voltage decay time series could therefore give an ion spectrum within the mobility range
defined by the start and end voltage. Although (12 is not analytically soluble in the general
case, it can be inverted using a finite difference solution, applicable for small changes, to
calculate ion spectra from voltage decay measurements.
5.1 Numerical example

A cubic spline was used to interpolate between the measured values of voltage and time.
They were then resampled at constant mobility resolution with the same number of points
as the initial data set. (12 was then inverted to give a set of average mobilities and ion
concentrations, with the voltage decay measurement, ventilation speed and cylindrical
capacitor dimensions as inputs. If the original voltage decay has N points, the inversion
yields a mobility spectrum with N-1 points. The highest voltage (lowest ion mobility) does
not have a corresponding spectral point because it is used as the initial voltage V0.
A flat ion spectrum with a mean concentration <n>=2000 ions cm-3, a typical value for
clean air, was transformed to obtain an artificial voltage decay using (12. The generated
voltage and time values were then used to invert the data back to an ion mobility spectrum.
<n> calculated using this method was 1999 cm-3 , a mean error of 0.05% compared to the
initial spectrum before transformation to a voltage decay; the maximum error in any spectral
bin was – 0.3%.
Sets of random errors within conservative time and voltage limits of ± 10s and 0.01V
were generated and added to the exponential voltage decay series. These errors were chosen
to simulate the uncertainty associated with manually extracting points from a voltage decay
time series plotted on a chart recorder. These errors are thought to be conservative as electronic
data logging would usually be more reliable, measuring time and voltage up to an order of
magnitude more accurately. The mean error in <n> was + 0.32%, and the maximum error was
+ 2.3%. The inversion procedure can therefore calculate average ion concentrations from
voltage decay time series to ± 0.5 %, and within a given mobility band the error is ± 2.5%.
6 Application of the inversion to calculate conductivity

Ion mobility spectra were calculated from the two example voltage decays shown in Figure
2 using the method described in section 5.1. µc for each interpolated point was calculated
from (9. Mean mobilities <µ> for each integrated strip were then determined by averaging
consecutive critical mobilities. The conductivity σ was calculated by integration across
the measured mobility range as in (1, for both decays, agreeing to 4%. Ion concentrations
were normalised to mean ion concentrations <n> using
(13)
The ion mobility range measured during a voltage decay is only related to the start and end
voltage. As shown in Figure 2, Decay 60 ran for longer than Decay 59, reaching lower
voltages, and therefore <µ> measured during the decay is higher. Since the conductivities

160
were very close for the two samples, the increased <µ> for Decay 60 implied a lower <n>.
Characteristics of the calculated ion spectra are summarised in Table 1.
Table 1 Summary of spectral characteristics from consecutive voltage decay time series
measurements in atmospheric air from June 12 1998
Typical urban ion concentrations are often lower than the values reported in Table 1. This
is likely to be related to errors in absolute values of µc, particularly from estimating
ventilation speed in the cylindrical capacitor in the presence of turbulent fluctuations in
wind speed. Alternative expressions for (9 based on volumetric flow rate exist, but would be
subject to similar errors. Further work is required to improve knowledge of the airflow in the
cylindrical capacitor ion counter.
7 Conclusions
Modifying the classical theory of air conductivity measurement for voltage decays from an
aspirated cylindrical capacitor has two principal advantages. Firstly, ion spectra can be
retrieved from simple voltage time series if the instrument characteristics and ventilation
speeds are known. Secondly, this spectral information can be used to calculate conductivity
directly by integration rather than with simplifying assumptions, such as exponential decay,
which introduce error. Previous ion spectrum measurements have usually required bias
voltage switching and sensitive current sensing, requiring dedicated electronics and signal
processing. The new method described here needs only the simplest logging equipment
and is amenable to remote (e.g. balloon-borne) applications. Additional information can
now be extracted from existing historical data sets.
References
[1] Gerdien H, Phys. Zeitung 6, 800–801 (1905) (in German)
[2] Swann W F G, Terr. Mag. Atmos. Elect 19, 81 (1914)
[3] Aplin K L and Harrison R G, Rev. Sci. Inst 71, 3037 (2000)
[4] Aplin K L and Harrison R G, Rev. Sci. Inst 72, 3467 (2001)
[5] Hõrrak U, Salm J and Tammet H, J. Geophys Res 105(D7), 9291–9302 (2000)
[6] Mohnen V A, Formation, nature and mobility of ions of atmospheric importance In: Dolezalek H.
(ed.), Electrical processes in atmospheres, Springer Verlag, Darmstadt, Germany (1974)
[7] Aplin K L, PhD thesis, The University of Reading, UK (2000)
[8] MacGorman D R and Rust W D, The Electrical Nature of Storms, Oxford University Press, New
York, USA (1998)
[9] Chalmers J A, Atmospheric Electricity, 2nd edition, Pergamon Press, Oxford, UK (1967)
[10] Venkiteshwaran S P, Measurement of the electrical potential gradient and conductivity by
radiosonde at Poona, India, In: Smith L G (ed.), Recent advances in atmospheric electricity,
Pergamon Press, Oxford, UK (1958)
[11] Dhanorkar S and Kamra A K, J. Geophys. Res 98, D2, 2639 (1993)
[12] Harrison R G and Aplin K L, Rev. Sci. Inst. 71, 8, 3231–3232 (2000)
[13] Harrison R G and Aplin K L, Rev. Sci. Inst., 71, 12, 4683–4685 (2000)

161
Spray Current Dependence on Flow Rate and

Conductivity in Cone-jet mode Vacuum Spraying
K L Smith and J P W Stark

Department of Engineering, Queen Mary, University of London, UK
Abstract. Experiments have been performed to investigate the properties of an electrospray in

cone-jet mode under vacuum conditions. The total spray current was measured as a function
of liquid flow rate for various values of solution conductivity. The solution used was Triethylene
Glycol, (TEG), doped with Sodium Iodide, (Nal). Several Nal concentrations were used to
achieve a range of conductivities between 10 and 100 µS/cm. The solution was sprayed from
a stainless steel capillary. Preliminary analysis from this experimental investigation has revealed
a dependence of the current-flow rate, I(Q), characteristics upon conductivity, κ. Our data is
consistent with a power law relationship for I(Q), however the best-fit exponent decreases
with increasing conductivity. Our results also indicate a dependence of the total beam emitted
spray current, I, on the applied voltage, Vapp.
Crucial to our work has been the development of a non-invasive, online system for
continuous monitoring of the low flow rates associated with the high conductivity fluids we
have used. With this flow measurement system, we can measure flow rates in the capillary-
vacuum system to an accuracy of 1nls-1.
1. Introduction
Electrospray techniques have the potential to be developed into micro-propulsion thrusters
for a range of space applications, with many advantages over other systems. These advantages
include lower power per unit thrust and lower overall mass and volume requirements.
The concept of colloid propulsion is not new. Early work on such concepts can be traced
to the early sixties [1–3] when a variety of studies and significant developments were
undertaken in both Europe and the USA. However, much of the work on electrosprays to
date has concentrated on atmospheric sprays at minimum flow rate conditions. The
electrospray models of Fernandez de la Mora [4] and Ganan-Calvo [5] are based largely on
such data. However, for the application of electrosprays in a colloid thruster system, operating
in a high vacuum environment at flow rates other than the minimum stable flow, further
understanding of the properties of electrosprays under these conditions is required.
The most important spray parameters in determining the thrust of a colloid thruster are
the flow rate, Q, the fluid conductivity, κ, the applied electric field strength, E, and the
emitted spray current I. In this study we have used modest conductivity triethylene glycol
(TEG) solutions of Nal with varying concentrations to give a range of fluid conductivities.
These solutions have been sprayed over a range of flow rates and field strengths within the
stable cone-jet regime.

162
2. Experimental Arrangement
Properties of the electro-spray system which have been measured are:
a) Supplied flow rate to the electro-spay emitter Q(nL/s)

b) Emitter spray current, I (nA)
c) Emitter voltage, V(kV)
d) Fluid conductivity, κ(S/m)
A schematic of the experimental arrangement is provided in Figure 1. The emitter used in

all the experimental results reported here, is a stainless steel capillary, having an outside
diameter 0.560 mm and inside diameter 0.305 mm. This was mounted in a standard SGE Ltd
union. The fluid was supplied to this union via a silica capillary pipe. The spray grid, which
provided the electrospray field, was a grounded stainless steel plate with a 6mm diameter
centralised hole, positioned 3mm from the front face of the emitter. The emitter grid system,
shown in figure 1 inset, was contained within a vacuum chamber, figure 1, which was kept
at pressures below 10–3 mbar during spraying. The fluid was fed to the emitter from a fluid
chamber which, following degassing was kept under N2 at pressures between ~10 to
1000mbar. The differential pressure between the fluid reservoir and the spray chamber
determines the volumetric flow rate to the emitter. The emitter and fluid reservoir system
were isolated from ground potential while the emitter was raised to a positive potential of
up to 5kV above ground.
Observation of the electrospray through a view port in the main vacuum chamber was
enabled using a PULNiX TM-1300 monochrome CCD camera in conjunction with a Navitar
12:1 telescopic zoom lens. The overall horizontal and vertical resolution of the imaging
system is 5 µm. This imaging system was used to determine if the electrospray was in cone
jet mode figure 2 and that there was no visible corona discharge. No current was recorded in
the system in the absence of an electrospray, which additionally confirmed that ionization
of the surrounding gas was not occurring.
A Quartz crystal microbalance was used to measure downstream on-axis charge to mass
ratios. These ratios were compared to emitted spray current mass flow rate ratios and were
found to be consistent. An on-line flow rate measurement system has been developed with
a temporal resolution of Is and volumetric flow rate accuracy of Inl/s absolute, with relative
flow fluctuations (stability) of +/- 0.1nl/s. This system measures the pressure drop along a
custom built inert glass lined tube from SGE International Ltd. The tubing has an internal
diameter of 0.3 mm and a length between pressure tappings of 100mm. The pressure at each
tapping is measured by one of a pair of Digiquartz Model 740–23A temperature compensated
quartz crystal pressure transducers, which have an accuracy of ± 0.16Pa. Figure 1 shows
how this system is integrated into the apparatus. The emitter current was measured at +5kV
by a custom built ammeter, which comprised of a two-stage optically isolated system to
safely transmit a signal from high voltage to a data logger at ground potential. A battery
powered high voltage stage, floating at the same potential as the emitter, contains a current
to voltage converter in the transresistance configuration. This was designed to measure
currents up to ±2 µA with a response of ~1mV/nA and a time response of Is. This converter
is followed by a voltage to frequency converter for optical signal transmission. The AD650
voltage to frequency chip was configured for bipolar inputs, giving 50–150 kHz for an
input range of ±2.5 V. A fibreoptic cable connected to a frequency to voltage converter chip

163
at ground potential comprises the second stage, which can then be directly and safely
connected to a PC, where all data was logged automatically.
Figure 1 Schematic of the vacuum chamber with the grid-emitter system position, with inset showing
the grid-emitter system
Figure 2 Electrospray (cone jet mode) of 0.01S/m TEG+Nal

164
3. Results
In figure 3 the current as a function of flow rate is presented for a solution of TEG+Nal with
a conductivity of 0.005S/m at various emitter potentials. Also plotted for each data set are
regression fitted power law relationships for the experimental data.
For comparison the power law relationship given by,
taken, from the reference model [4] is shown. The values for f(ε) are
derived from empirical data, and are also given [4] ;f(ε) is adopted to compare the electrospray
properties from fluids with differing permittivity ε. An alternative model has been developed
by Ganan-Calvo and co-workers [6] which avoids the use of the empirical parameter f(ε).
This alternative model identifies two distinct trends for the current flow rate dependence;
these different behaviours are expressed in terms of an analytical function for the fluid
properties, including the viscosity, but again the exponent is independent of the electric
field. These two behaviours identify the exponent of the flow rate dependence to be either
0.5, as in [4], or 0.25. For the fluid properties in our experiments the appropriate I:Q
relationship is given by:
Where I0=(ε0γ 2/̃)0.5 , and Q0=γ ε0/̃. Q0 is the value identified in reference [5] to be the minimum
stable flow rate. However in the Ganan-Calvo model the minimum flow rate is seen,
potentially to be significantly larger than Q0, and is given by Qmin=Q0(ε˜–1)0.5 From figure 3
we can discern a distinct trend of increasing power law exponent with increasing voltage.
Clearly this trend is not reflected in the reference model relationship. Although the
reference model [4] strictly applies only to flow rates at or near minimum flow rate, our
data reveals a trend not captured by the model. There is a clear dependency of the current on
the applied voltage, with the exponent a in the fitted curves of I(Qa) increasing with voltage.
To reiterate, this data is for a fluid having a single value of conductivity. Recently reported
data of Ku and Kim [7] also reveals this sensitivity of current to applied voltage in vacuum
conditions.
In figure 4 the effect of fluid conductivity on the spray current flow rate relationship,
may be seen. Again assuming a power law dependence I ∝ Qa, where now a, appears to have
some dependence on fluid conductivity. Figure 5 shows a plot of these values of, a, with
changing conductivity. In Hartman’s numerical model [8] in contrast to Fernandez de la
Mora [4], a dependence of, a, with changing conductivity is identified. The observations
of Hartman [8] appear to agree with the preliminary findings of this report, such that there
appears to be a logarithmic dependence of the exponent of the I(Q) curves on conductivity.
However Hartman’s numerical model is confined to conductivities an order of magnitude
lower than those investigated in this research. Further examination of this relationship
needs to be conducted to verify these preliminary results, where overlapping results would
clarify the form, a, with changing conductivity.

165
Figure 3 Emitter current as a function of flow rate for various values of applied voltage for TEG+NaI
κ=0.005S/m
Figure 4 Logarithmic plot of dependence of power exponent, a, on conductivity for various

conductivities of TEG+NaI at 4.1kV

166
Figure 5 Effect of fluid conductivity on spray current flow relationship logarithmic plot of dependence
of power exponent, a, on conductivity for various conductivities of TEG +NaI at 4.1kV
4. Conclusions
The results presented here, indicate that vacuum electrospray characteristics are more
sensitive to electrostatic conditions than previously recorded. Of particular importance is
the sensitivity of the current-flow rate relationship to electrostatic conditions and the fluid
conductivity. Further investigation into the exact nature of the conductivity dependency is
required, for example by extending the range of measurement by using fluids with similar
conductivities to those presented by Hartman [8].
5. References
[1] A.G.Bailey. 1988 Electrostatic spraying of liquids (Research studies press Ltd., England.)
[2] Hendricks CD 1962 J. Colloid & Interface Science, 17 249
[3] Perel, J, Bates T, Mahoney J, Moore RD, & Yahiku AY 1967 AIAA paper 67–728, Colorado
Springs
[4] Fernandez de la Mora J, & Loscertales JG 1994 J. Fluid Mechanics 260 155
[5] Ganan-Calvo AM, J Davila & A Barrero 1999 J. Aerosol Sci. 30 863.
[6] Ganan-Calvo AM, J Davila & A Barrero 1997 J. Aerosol Sci. 28 249
[7] Ku, B.K. & Kim, S.S. 2003 J. Electrostatics 57 109
[8] Hartman,R.P.A et al 1999 J. Aerosol Sci. 30 823

167
Visualization and particle image velocimetry

measurements of electrically generated coherent
structures in an electrostatic precipitator model
Jerzy Mizeraczyk1, Marek Kocik1, Jaroslaw Dekowski1, Janusz PodliDski1,
Toshikazu Ohkubo2, Seiji Kanazawa2 ”
1
Centre for Plasma and Laser Engineering, Institute of Fluid Flow Machinery,”
Polish Academy of Sciences, 80–231 Gdansk, Fiszera 14, Poland
2
Department of Electrical and Electronic Engineering, Oita University, 700
Dannoharu, Oita 870–1192, Japan
Abstract: There is still an interest in improving electrostatic precipitator (ESP) collection of

fine particles (micron and submicron sizes). However, it is not yet clear whether the coherent
structures and flow turbulence caused by the presence of electric field and charge in the ESPs
advance or deteriorate fine particle precipitation process. In this paper results of the laser
visualization and Particle Image Velocimetry (PIV) measurements of the particle flow velocity
field in a wire-to-plate type ESP model with seven wire electrodes are presented. The
measurements were carried out for negative and positive polarity of the wire electrode. The
PIV measurements clearly confirmed formation of the coherent vortex structures (size of
several centimetres) in the ESP model, which interact with the primary flow. The particle flow
pattern changes caused by the strong interaction between the primary flow and electrically
generated turbulence and coherent structures are more pronounced for higher operating voltages
and lower primary flow velocities. The coherent structures for the positive voltage polarity are
more stable and regular than those for the negative polarity due to the nonuniformity of the
negative corona along the wire (tufts). Due to the same reason the intensity of the flow
turbulence is larger for the negative polarity.
1. Introduction
In recent years a special environmental concern is directed towards controlling the emission
of micron and submicron particles in electrostatic precipitators (ESPs), which operating
with high overall efficiency, are not effective in the removal of fine particles. Many of the
fine particles of 1 µm or less in size contain toxic trace elements. Hence, there has long been
interest in improving ESP collection of fine particles.
The motion and precipitation of particles in the duct of an ESP depend on the particle
properties, electric field, space charge and gas flow field. It was shown (e.g. [1–3]) that a
significant interaction between these factors exists, resulting in considerable turbulent
flow patterns in the ESP. However, it is not yet clear whether these turbulent flow patterns
advance or deteriorate fine particle precipitation process. To elucidate the influence of the
electrically generated flow disturbances on the precipitation of fine particles in ESPs more
experimental investigations are needed.

168
In this paper, results of laser flow visualization and Particle Image Velocimetry (PIV)
measurements of particle flow velocity fields in a wire-to-plate type ESP model with seven
wire electrodes are presented. This investigation is expected to be helpful in elucidating
the motion of the fine particles in ESPs.
Figure 1. Experimental set-up.
2. Experiment
The apparatus used in this experiment consisted of an ESP model, a laser flow visualization
set-up and standard PIV equipment for the measurement of velocity field (Fig.1).
The ESP model was a transparent plane-parallel acrylic duct, 160 cm long, 20 cm wide
and 10 cm high. In the middle of the ESP model, seven stainless-steel wire electrodes
(diameter of 0.1 cm, length of 20 cm, 10 cm apart from each other) were mounted in the
acrylic side-walls, parallel to the opposite facing stainless-steel plate collecting electrodes.
The distance between the collecting electrodes (100 cm long and 20 cm wide) was 10 cm.
The applied voltage, either negative or positive, was varied from that of the corona onset
up to 30 kV, which corresponds to a mean electrostatic field strength of 6.0 kV/cm at the
wire electrode (no discharge). The voltage was supplied to each wire electrode through a 10
MΩ resistor. Air seeded with cigarette smoke (size of less than 1 µm in dry air) was blown
along the ESP duct with an average velocity ranging from 0.14 to 0.60 m/s (a flow velocity
of about 0.8 m/s is typical of ESPs).
The laser flow visualization set-up consisted of a pulsed CuBr laser (λ=510.6 and 578.2
nm) used as a light source, a cylindrical telescope to transform the CuBr laser beam into a
laser sheet (1 mm thickness) passing along the ESP model and a video-camera for recording
the flow patterns revealed by the laser sheet in the ESP model.
The PIV equipment was the same as used in [4]. The velocity field maps presented in this
paper are composed of several adjacent velocity fields (from 4 to 9), each of an area of
10×10 cm. All the presented velocity field maps resulted from the averaging of 100
measurements, i.e. the presented velocity maps are time-averaged.
3. Results
In the following text, the externally forced flow of air (seeded with cigarette smoke particles)
along the duct is called a primary flow (not influenced by the electric force), in contrast to
the flow of the seed particles, which may not follow the primary flow when applying the
voltage.

169
When the visualization or PIV method is used for velocity field monitoring in the flow
seeded with particles, in which an electric force does not exist, the obtained velocity field
pattern reflects the motion of the primary flow, assuming that the seed particles follow the
flow. However, when an electric force exists in the flow, the seed particle motion depends
not only on the primary flow motion, but is also influenced by the electric force. Therefore,
the velocity field patterns monitored by both methods (visualization and PIV) in the ESP
model reflect the motion of the seed particles, or, in general, particles present in the ESP
model.
In the presented visualization images, the existence of the seed particles in the illuminating
plane is revealed by the lighter regions, while the absence of the seed particles is detected
as the darker regions. Therefore, the relative seed particle concentration can be qualitatively
perceived by the brightness level recorded in the visualization images.
In this paper, Reynolds number Re is the ratio of the product of the mean flow velocity
and the wire-plate spacing (5 cm) to the dynamic viscosity of the carrier gas (0.15 cm2/s).
NEHD number is the ratio of the conductive electric Rayleigh number to Reynolds number
squared [5].
3.1. Flow velocity field patterns

3.1.1 Negative Polarity. The particle velocity field patterns in the ESP model for two
different time-averaged velocities of the primary flow (0.14 m/s and 0.6 m/s) at constant
negative voltage of 24 kV are shown in Figs. 2 and 3, respectively. They confirm existence
of strong secondary flows of the seed particles, caused by the electric field and charge. The
secondary flow patterns depend on the applied voltage and velocity of the primary flow.
At a relatively low primary flow velocity of 0.14 m/s, (i.e. at a moderate Reynolds
number, Re=460) and operating voltage of 24 kV (i.e. at a high electrohydrodynamic
number, NEHD=30), strong vortices are formed in both the upstream and downstream regions
of the ESP duct (Fig. 2). In this case, the electric force is dominating over the inertial one.
Figure 2. Flow velocity field in the ESP model at a primary flow velocity of 0.14 m/s. Negative polarity,
voltage 24 kV, Re=460, NEHD=30. Only the flow patterns around first three wire electrodes are shown
(on the left side of the third wire electrode ,the seed particles vanished due to efficient electrostatic
precipitation).

170
Figure 3. Flow velocity field in the ESP model at a primary flow velocity of 0.6 m/s. Negative polarity,
voltage 24 kV, Re=2000, NEHD=2.3.
When the primary flow velocity increases, the influence of the electric field force on the
particle motion becomes diminishing, and the vortices disappear. The flow velocity pattern
as indicated by the seed particles is dominated by the primary flow (Fig.3). In this case, the
precipitation of the seed particles is not as efficient as at the lower primary flow velocity
(0.14 m/s, Fig. 2) due to the higher seed particle mass flow rate, shorter residence time and
weaker influence of the electric field. As a result the seed particles are present in the primary
flow after passing all seven wire electrodes, making the PIV measuring possible along the
whole duct of the ESP model (compare Fig. 2, where the PIV measuring was not possible
beyond the third electrode due to the strong particle precipitation).
We found that the electric field influence on the particle velocity field pattern becomes
stronger at higher operating voltage, resulting in stronger coherent vortex structures, also
in the upstream region of the ESP model duct. In particular, relatively large velocity
fluctuations in the flow pattern are observed at negative voltage polarity. They are likely
due to the discharge current which is higher and less uniformly distributed along the wire
electrodes (forming the so-called tufts) than that at positive polarity (at the same voltage).

171
Figure 4. Image of the coherent structures in the ESP model at a main flow velocity of 0.14 m/s. Positive
polarity, voltage 24 kV.
3.1.2 Positive polarity. At positive voltage polarity, the time-averaged particle velocity
fields are astonishingly very similar to those at negative polarity shown in Figs. 2 and 3.
The secondary flow with apparent vortices (as in Fig. 2 for negative polarity) formed at
positive polarity is also more pronounced at lower velocities of the primary flow (lower
Re). It weakens with increasing primary flow velocity, similarly as for positive polarity in
Fig. 3.
At lower primary flow velocity (less than 0.4 m/s) the seed particles are very fast removed
from the carrier gas already in the region of the first 3 wire electrodes, similarly as at
negative polarity. After the third wire electrode the flow is practically free from the seed
particles.
3.2 Coherent flow structures

Fig. 4 shows an image of the coherent vortex structures formed before the first wire electrode
at positive polarity voltage of 24 kV and primary flow of 0.14 m/s. First large vortex
(about 5 cm in diameter) is placed about 7 cm before the first wire electrode. Then this
vortex moves downstream and becomes smaller (about 3 cm in diameter) and several
centimetres upstream from this vortex second, smaller vortex appears (see right upper
corner of Fig. 4). When the second vortex reaches diameter of 2–3 cm, both vortices merge
together (see lower part of Fig. 4) into one larger vortex. Then the cycle starts again.
Duration of each cycle differs slightly and lasts several seconds.
Coherent vortex structure found in Fig. 4 at positive polarity is also present at negative
polarity, but is less stable. Vortices are larger than at positive polarity and often there is only
one large vortex at lower or upper plate electrode.
Fig. 5 shows coherent vortex structures behind the first wire electrode at a positive
polarity voltage of 24 kV and primary flow of 0.6 m/s. The vortices generated around the wire
electrode are convected downstream alternately from the upper and lower side of the wire
electrode, similarly to the so-called Karman vortex street. The vortices (about 5–20 mm in
diameter) electrically generated behind the first electrode are much larger than the vortices
(about 2 mm in diameter) formed behind the first electrode when no voltage is applied.

172
Fig. 6 shows coherent vortex structures behind the first negatively-polarized wire electrode
at voltage of 24 kV and primary flow of 0.6 m/s. Similar to positive polarity, Karman vortex
street structure was generated, but this structure loses its stability before second wire electrode
and acts as flow turbulence
Figure 5. Image of the coherent structures in the ESP model at a main flow velocity of 0.6 m/s. Positive
Figure 6. Image of the coherent structures in the ESP model at a main flow velocity of 0.6 m/s. Negative
4. Conclusions
The presented results for the flow visualization and PIV measurements of the particle
velocity field in the ESP model confirmed that the presence of the electric field and charge
causes a significant change in the particle flow pattern. After applying the voltage, secondary
flow having velocity of several tens of cm/s and the coherent vortex structures are formed in
the ESP model, interacting with the primary flow. This results in strong vortices in the
downstream and upstream regions of the ESP model duct, if the primary flow velocity is
relatively low (below 0.4 m/s). The primary flow disturbances are more pronounced for lower

173
primary flow velocities (low Reynolds number) and/or higher operating voltages (high
NEHD). Under these conditions, the precipitation of the particles in the ESP model is high.
The flow patterns in the ESP model at positive voltage polarity are more stable and
regular than those for the negative voltage polarity. The turbulence intensity is also smaller
for positive voltage polarity. The more turbulent character of the flow patterns at negative
polarity is likely due to nonuniformity of the negative corona discharge, exhibiting the
form of tufts along the wire electrode. On the other hand, the time-averaged velocity flow
fields are astonishingly very similar for both voltage polarities.
Acknowledgement
This research was sponsored by the Foundation for Polish Science (Fundacja na Rzecz
Nauki Polskiej, subsidy 8/2001) and the State Committee for Scientific Research (KBN PB
1756/T10/01/21). Fruitful discussions with Prof. J.S.Chang are thankfully acknowledged.
References
[1] Ohkubo, T, Nomoto, Y, Adachi, T, and McLean, K.J, 1986, J. Electrostatics, 18, 289–303
[2] Atten, P., McCluskey, F.M.J., and Lahjomri, A.C., 1987, IEEE Trans. Ind. Appl., 23–4, 705–711
[3] Chang, C.L. and Bai, H., 2000, Aerosol Sci. Tech., 33–3 (2000), 228–238
[4] Mizeraczyk J., Kocik M., Dekowski J., Dors M., Podlinski J., Ohkubo T, Kanazawa S., and
Kawasaki T., 2001, J. Electrostatics, 51–52, 272–277.
[5] Chang, J.S. and Watson, A., 1994, IEEE Trans. Dielec. Elec. Ins., 1–5, 871–895

175
Electrohydrodynamics in microelectrode structures

Antonio Castellanos1, Antonio González2, Antonio Ramos1, Nicolas G Green3
and Hywel Morgan3
1
Dpto. de Electronicay Electromagnetismo, Universidad de Sevilla, 41012-Sevilla,
Spain
2
Dpto. de Fisica Aplicada III Universidad de Sevilla, 41092-Sevilla, Spain
3
Dpt. Electronics and Electrical Engineering, University of Glasgow, G12 8LT, UK
Abstract. Dielectrophoretic forces have been used to control submicron particles in aqueous
solutions. However, a complete analysis of the particle dynamics must include the motion of
the liquid that drags the particles. In the first part of this work we describe the EHD
equations relevant to microsystems, and their solution for simple microelectrode systems.
This allows to classify the different types of fluid flows, and to determine their domain of
influence as well as their scaling laws. In the second part of the work we analyse the relative
importance of particle displacements in aqueous solutions due to different flows in comparison
to the displacements caused by dielectrophoresis
1. Introduction.
Electrohydrodynamics (EHD) at the micrometre scale is a subject of great interest, among
other reasons, because electrically induced microflows often appear in the manipulation of
colloidal particles by means of electric fields. The control and manipulation of particles of
micron size in microelectrode structures is a well-established technique [1]. In particular,
the dielectrophoretic (DEP) manipulation of bioparticles such as viruses, cells, DNA is of
special interest. Theoretical estimations indicate that to overcome Brownian motion of
sub-micron particles it is necessary to impose high electric fields. However, the experiments
revealed that the liquid was put in motion, and that this motion was a limiting factor far
more important than Brownian motion. The local Joule heating produces temperature
gradients that generate buoyancy forces and charge densities that are moved by the electric
field [2,3]- In addition, recent experimental work has shown that the application of an ac
electric field to a pair of coplanar microelectrodes produces a tangential component close
to the electrodes, that acts in the double layer charge and generates a steady (non-zero time
averaged) fluid flow. This flow has been termed ac electroosmosis, with a tangential velocity
which depends on both the applied potential and frequency [4,5,6]
To manipulate the particles we must know the relative importance of the particle
displacement due to the different fluid flows in comparison to the other displacements
caused by gravity, DEP and Brownian motion.
Bioparticles, that have typical sizes in the range from 0.1 microns up to 10 microns, are
usually in aqueous saline solutions with a conductivity that ranges between 10-4 and 10-1S/
m. Typical system lengths of microelectrodes (interelectrode gaps) used in the DEP
manipulation of bioparticles vary from 1 to 500 microns. Applied signals to these electrode
structures range from 0 to 20 V giving rise to electric fields that can be as high as 5 MV/m.
These applied signals have frequencies that range between 102 to 107 Hz.

176
Here we present some semi-quantitative results obtained for a simplified system of two
infinite half-plane electrodes a distance l apart, in order to gain a first understanding of
these scaling laws. With due precautions these results can be generalised to more complicated
microelectrode shapes by taking into account that different zones may have different
characteristic length scales.
2. Fluid flow: Equations and boundary conditions.

2.1 Electrical equations and boundary conditions
The electromagnetic equations reduce to the quasi-electrostatic limit. In addition for saline
solutions the convection current can be neglected in front of the ohmic current. Assuming
that conductivity σ and permittivity ε are independent on time, the electric field is curl-free
and verifies the charge conservation equation
(1)
Because relative variations in σ and ε are usually small, that implies, in the first
approximation, that E is the gradient of a potential that obeys Laplace’s equation.
For the boundary condition the double layer between the metallic surface and the
electrolyte, which in our case is of nanometre size, can be modelled as a capacitor between
the electrode and the electrolyte. The corresponding boundary condition expresses that the
current that goes into this capacitor equals the increase in the stored charge [6,8]
(2)
For the interface electrolyte/glass the boundary condition reduces to ∂φ/∂n=0 as the effect
of the double layer is negligible. For the upper boundary, we can consider that the points
are at infinity and the electric field and potential vanish at that boundary.
2.2 Mechanical equations

The liquid motion is governed by the Navier-Stokes equations for an incompressible fluid.
The inertial terms are negligible, as for microsystems the Reynolds number is usually very
small. Since we are interested in the time-averaged flow Navier-Stokes equations reduce to
(3)
As stated in this equation the fluid motion is caused by the joint action of gravity and
electrical forces. Here ∆ρm is the change of ρm with temperature. The time average electrical
force contains a Coulombian and a dielectric term [3,9]
(4)
For high frequencies (ω>>σ/ε) the permittivity gradients dominate. For ω<<σ/ε, the
conductivity gradients are usually dominant, since relative variations ∆σ/σ due to
temperature, in electrolytes, are much greater than relative variations in permittivity, ∆ε/ε.
In any of the rigid boundaries, the normal velocity vanishes. The tangential velocity
close to the electrodes is caused by ac electroosmosis and can be calculated generalising
Smoluchowsky formula. The time-averaged expression for the tangential slip velocity is
given by [6,8]
(5)

177
where ∆φ is the voltage drop across the double layer and Et the tangential field just outside
the double layer. The parameter Λ is the ratio of the voltage drop across the diffuse layer to
that across the total double layer. For the rest of the lower boundary the effect of the double
layer is negligible and the tangential velocity vanishes due to viscous friction [6].
2.3 Temperature equation

The heat convection is small compared to heat diffusion in microsystems, and the time-
averaged temperature equation reduces to Poisson’s equation, with the Joule heating as
source
(6)
In the experiments in microelectrodes the boundary conditions for the temperature field are
given by the surroundings and they can be very different between experiments [6]. In order
to make some calculations, we will consider that the electrodes are at room temperature.
Also, the upper and lateral boundaries are going to be considered at room temperature
placed at a distance of the order of the total size of the electrodes.
3. Fluid Flow: Scaling Laws

In our experiments, we have used a simple electrode design consisting of two coplanar
rectangular electrodes 2mm long and 500 µm wide, with parallel edges separated by a 25
µm gap, mounted on a glass substrate. In order to obtain some numbers, the electric field
lines between electrodes are going to be considered semi-circular when needed. In this
case, the electric field is E=V/πr uθ, where V is the voltage amplitude of the applied signal
and r is the distance to the centre of the gap.
3.1 Joule-heating-induced electrothermal fluid flow

When an electric field is applied between the electrodes, the Joule heating produces a
temperature field [3] that generates gradients in σ and ε, giving rise to an electrical body
force, according to (4). Taking into account the different behaviour at low and high
frequencies we obtain, for our microelectrode system,
(7)
Computations using the finite element method give estimations of the maximum velocity
of the same order as the previous expression when r is around 50 µm [9]. That was the
typical size of the convective roll in the computations. This is a value between the
interelectrode gap length (25 µm) and the height of the upper boundary (200 µm).
3.2 Joule heating induced buoyancy.

The same temperature field produces buoyancy. For a typical system length r= 25µm, and
applied voltage V=10 V, the ratio of gravitational to electrical forces, fg/fe~7×10–4, is very
small. As the typical size of the system is increased the order of magnitude of the gravitational
force becomes greater than the electrical force. The transition is for r~300 µm. The
computations in our system give a typical velocity due to buoyancy of the form
I (8)

178
3.3 Ac electro osmosis

The tangential velocity described in (5) produces a motion in the bulk. A simple model
based on an array of resistor-capacitor circuits [4] or a more detailed mathematical analysis
[8] can give reasonable values for the velocity. The circuit model gives
(9)
A typical value of C is ε/λD, the surface capacitance of the diffuse double layer in the Debye-
Hückel model. Experimentally, Λ is around 0.25 for our electrodes made of titanium and a
medium conductivity around 0.002 Sm-1. For an ideal double layer made of a Stern layer
and a diffuse layer, the parameter Λ decreases with conductivity, since the capacitance of
the diffuse layer increases with conductivity. The velocity as a function of frequency has a
maximum, and tends to zero for frequency going to zero or infinity. The maximum velocity
is obtained for Ω=1, that is, for angular frequency ε0=σ/(Cπr). Taking C=ε/λD), the resulting
frequency is several orders of magnitude smaller than σ/ε.
4. Motion of Particles
The velocity of a particle moving in a fluid, assuming that Stokes drag is valid, is given by
(10)
where v is the fluid velocity, F the exerted force, and γ the friction coefficient (γ=6πηa for
a spherical particle). Here, inertia has been neglected, as the characteristic time of
acceleration, τa=m/γ (10–6s for cells and submicron particles), is usually much smaller than
l/v, and therefore, the particle can be considered to move at the terminal velocity.
The DEP force arises from the interaction of a non-uniform electric field and the dipole
induced in the particle. For linear, isotropic dielectrics, the relationship between the dipole
moment phasor p of a spherical particle and the electric field phasor E is given by p(ω)=υα(ω)
E, where υ is the particle volume, α its effective polarisability, and ω is the frequency of the
electric field. The time-averaged force on the particle is given by [3]:
(11)
The first term on the right-hand side is non-zero if there is a spatially varying field magnitude,
giving rise to DEP. The second term is non-zero if there is a spatially varying phase, as in
the case of Travelling Wave DEP. For DEP alone, the velocity induced on a particle is:
(12)
where we have considered only Maxwell-Wagner interfacial polarisation. indicates

the complex permittivity. The part of the expression in brackets is referred to as the Clausius-
Mossotti factor and describes the frequency variation of the DEP mobility and force. This
factor varies between +1 and -1/2 and the particle will move towards (positive DEP) or away
from (negative DEP) regions of high field strength depending on frequency. Assuming a
Clausius-Mossotti factor of order one, an estimation of the DEP velocity magnitude is
(13)

179
Figure 1: Map of the dominant mechanisms in a solution on two co-planar electrodes, for 1kHz (left) and
1MHz (right), in the plane applied voltage versus typical system length. The solution conductivity is
σ=lmS/m and the particle radius is 1 µm.
To this displacement, we must add the effect of gravity on the particle, which depends on the
difference in mass density between the particle and the surrounding liquid. For a spherical
particle and assuming |ρp-ρm| of the order of ρm, an estimation of the gravitational velocity is
(14)
The factor 0.2 could be smaller as many bioparticles have densities close to water density.
Thermal effects also influence colloidal particles through Brownian motion. The
associated displacement has zero average, but a non vanishing variance. If we want to move
a single particle, we must overcome this random Brownian displacement, that shows a rms
(in one dimension) given by
(15)
where kB is Boltzmann’s constant, T is the absolute temperature and t is the period of
observation.

Now we can compare the displacements of particles due to fluid motion, to the effect of DEP
forces, that are expected to control them. Figure 1 and 2 show the domains of influence of
the different forces on the plane applied voltage versus system length. The level curves
represent the magnitude of the particle velocity induced by the dominant mechanism. We
consider an aqueous saline solution at room temperature, with mass density, viscosity, heat
conductivity and permittivity those of water, with (1/σ)(∂σ/∂T)=0.02K-1, (1/ε)(∂ε/∂T)= -
0.04K-1, (1/ρm)(∂ρm/∂T)=-2×10–4K-1. In figure 1 we consider the behaviour for a conductivity
σ=1mS/m, when the applied voltage varies from 0.1V to 100 V, and the typical system size
ranges from lµm to 1cm. For a frequency of 1kHz (left), we see that electrothermal forces,
which go as the fourth power of the voltage, are only dominant for very high values of V1,
while buoyancy in the liquid dominates for large system sizes. For moderate to small sizes
and almost all voltages, the liquid drags the particles through electroosmotic flow. The
DEP forces are never dominant, as we are close to the peak of the electroosmotic velocity,
which is much higher than the DEP one. The situation changes for a high frequency (right).
In this case, the ac electroosmotic flow becomes negligible, and the electrothermal forces
1
For low frequencies, high voltages may not be accessible because of electrolysis at the electrodes.

180
Figure 2: Map of the dominant mechanisms in a solution on two co-planar electrodes, for 1kHz (left) and
1MHz (right), in the plane applied voltage versus typical system length. The solution conductivity is
σ=100mS/m and the particle radius is lµm.
dominate at high voltages. For small system sizes (and then, relatively large particles) the
DEP force is the most important one and the particles can be controlled using electric fields.
For high conductivities (figure 2, for σ=100mS/m) we have an even stronger difference
between low and high frequencies. For frequencies much smaller than σ/ε, like 1 kHz (left)
the voltage drops happen mainly in the double layers, thus reducing the field in the bulk.
It results in the reduction of the volume forces, as the electrothermal ones. The ac
electroosmotic flow is also reduced because of the smaller tangential field near the electrodes.
For 1 MHz (right), the field in the bulk is higher and the behaviour is similar to the case of
low conductivity (fig. 1, right). However, electrolysis may take place in the region of low
frequencies and high voltages, thus making that part of the maps inaccessible.
As the scales for the different forces can vary locally, it is possible that in certain regions
(especially near electrode edges) DEP dominates over other mechanisms, since it goes as r-3.
6. Acknowledgements
This research has been financed by the Ministerio de Ciencia Y Tecnologia del Gobierno
Espanol, Direccion General de Investigacion under project BFM2000–1056.
References
[1] R Pethig, Crit. Revs. Biotech., 16 331 (1996).
[2] T Muller, A Gerardino, T Schnelle, S G Shirley, F Bordoni, G DeGasperis, R Leoni, and G Fuhr,
J. Phys. D: Appl. Phys., 29 340 (1996)
[3] A.Ramos, H.Morgan, N.G.Green, and A.Castellanos. J. Phys. D: Appl. Phys. 31 2338 (1998)
[4] A.Ramos, H.Morgan, N.G.Green and A.Castellanos, J. Coll. Interf. Science, 217 420 (1999)
[5] NG Green, A Ramos, A Gonzalez, H Morgan and A Castellanos, Phys. Rev. E, 61 4011 (2000)
[6] NG Green, A Ramos, A Gonzalez, H Morgan and A. Castellanos, Phys. Rev. E, 66, 026305
(2002)
[7] A Castellanos, Electrohydrodynamics: Ch. 4 (ed. A Castellanos), Springer-Verlag, NY (1998)
[8] A González, A Ramos, NG Green, A Castellanos and H Morgan, Phys. Rev. E, 61 4019 (2000)
[9] NG Green, A Ramos, H Morgan, A Castellanos and A Gonzalez. J. Electrost. 53 71 (2001)

181
Eleetrohydrodynamic atomization of viscous liquids
A Jaworek1, W Balachandran2, A Krupa1, J Kulon2, W Machowski2,3

1
Institute of Fluid Flow Machinery, Polish Academy of Sciences, Fiszera 14, 80951
Gdansk, Poland, jaworek@imp.gda.pl
2
Department of Systems Engineering, Brunei University, Uxbridge UB8 3PH, UK.
3
present affiliation: Visual Information Laboratory, Mitsubishi Electric Europe,
Guildford, U.K.
Abstract. Liquid sprayed by the electrohydrodynamic (EHD) method is atomized only by

electrical forces without additional mechanical energy applied. EHD spraying is an effective
method for generating electrically charged aerosol. The effect of liquid viscosities on the
spraying process was investigated experimentally for distilled water and ethylene glycol
mixtures, with viscosity ranging from ImPas to about 22mPas. When the viscosity increased
the jet diameter and the droplet size decreased in the jet mode of spraying. It was noticed that
transition of the spindle mode to the jet mode takes place for lower voltages as the liquid
viscosity increases.
1. Introduction
Electrohydrodynamic spraying of liquids is an effective method for charged aerosol
production. In electrohydrodynamic spraying, the liquid is atomized only due to electrical
forces and no additional mechanical energy is required. The droplets generated by this
method are electrically charged up to a fraction of the Rayleigh limit allowing to control of
droplet motion by electrostatic fields. DC voltages or AC superimposed on DC voltages are
usually applied as excitation to disintegrate the liquid jet (Huneiti et al., 1998).
The liquid can be electrohydrodynamically atomized to droplets in two main ways: the
first, in which only fragments of liquid are ejected from the nozzle by deforming and
detaching the liquid meniscus, and the second, in which the liquid is elongated into a long
fine jet or a few jets which disintegrates into droplets. The atomization process depends on
the physical properties of the liquid, the nozzle-extractor geometry, the voltage at the
nozzle, and liquid flow rate.
Experimental investigations of the effect of liquid viscosity on the transition between
spraying modes, jet diameter, and the droplet size in the jet mode are reported in this paper.
2. Experiment
The experimental arrangement is shown schematically in Fig. 1. The spray nozzle was
made of a stainless steel hypodermic needle of inner diameter of 0.25 mm, and outer
diameter of 0.4 mm. The grounded ring electrode made of brass wire of diameter of 1 mm,

182
and with inner diameter of 9 mm was positioned as shown. The distance between the tip of
the nozzle and the plane of the ring electrode was equal to 7 mm. A syringe pump was used
to feed the liquid to the spray nozzle. The capillary nozzle was connected electrically to a
high voltage power supply (SPELMANN HV).
Fig.1. Schematic view of the experimental stand
A CCD camera equipped with a long distance microscopic lens was used to observe and
record the spray process by a continuous or stroboscopic illumination of flash time of 30 µs.
The droplet size distribution was measured by MALVERN Mastersizer S.
It is not possible to change liquid viscosity without varying other important liquid
parameters. Mixtures of two liquids of similar surface tension and conductivity but wide
differences in viscosity were used. Distilled water and ethylene glycol were chosen to
prepare mixtures with properties given in Table 1. The viscosity of the mixtures changed in
the range of ImPas to about 22mPas. All measurements were carried out at normal pressure
and ambient air temperature of about 16° C.
Table 1. Physical properties of the liquids sprayed. (@ T=16±1°C).
3. Results
The spindle mode and the jet mode were generated by this spray system. In the spindle
mode, the meniscus takes the form of a thick elongated jet, which detaches as a vast
spindle-like fragment of liquid before a continuous jet can be formed by the electric field.
The spindle can disrupt into several smaller droplets of different sizes on the action of
electrostatic repulsive forces. The spindle mode differs from the dripping mode because no
regular droplets uniform in size are ejected from the meniscus. Bi-modal size distribution
characterises the spindle mode: large droplets with most probable diameters varying in the

183
range from 200 to 500 µm, and smaller satellite droplets, usually finer than 50 µm (Fig.2).
The number of smaller droplets increases with increasing viscosity, especially for high flow
rate of the liquid.
Fig. 2. Modal diameter of the droplets and transition between the spraying modes for liquids of
different viscosity vs. liquid volume flow rate. DC potential applied to the nozzle 10kV. -main
drops, ◊ -satellites.
Fig. 3 Droplet volume size distribution in the jet mode for two volume flow rates.
Fig. 4. The effect of liquid viscosity on the jet and droplet diameter in the jet mode. DC potential of
the nozzle 10 kV.
In the jet mode, a regular axisymmetric cone meniscus is formed, which gradually changes
into a thin jet. The jet undergoes kink or varicose instability, and breaks up into fine
droplets due to electrical and inertial forces. Examples of the size distribution of the droplets
are shown in Fig.3. The diameter of the jet, measured at the same distance of 4 mm from the
nozzle outlet, is shown in Fig.4. It is about 140 µm for distilled water, and decreases with the
increase in liquid viscosity to about 90 µm, for ethylene glycol. It was also noticed that the
jet becomes finer and shorter with increasing voltage. It can be seen from Fig.4 that the most
probable diameter of the droplets decreases with the increase in liquid viscosity.

184
The transition from the spindle to jet mode causes changes in the size distribution
(Fig.2). Initially the droplet mean size in the spindle mode increases with flow rate. For
higher flow rates, the bi-modal distribution disappears, and only fine droplets are produced
in the jet mode. The droplet mean size only slightly increases with flow rate. The graph
referring to the size of the droplets in the jet mode seems to be a continuation of the graph
referring to the satellites in the spindle mode, which results in rapid increase in the number
of generated droplets.
4. Discussion
It was observed that switching of the spindle mode to the jet mode take place when the
flow rate increases. It can be assumed that the transition occurs when the inertial time
constant τp, i.e., the characteristic time of the jet acceleration/deceleration due to inertial
forces (cf. Jaworek and Krupa, 1999)
(1)
is higher than the flow-rate time constant ˜l
(2)
which is the time required for sufficient volume of the liquid to flow into the spray nozzle.
The flow-rate time constant τl referring to the flow rate at which it exceeds the inertial time
constants τρ is indicated on the abscissa of the plot in Fig.2. These values are close to those
for which the switching of the spindle mode to jet mode was determined experimentally.
The viscosity of liquid plays a main role in jet formation in the jet mode. The jet
diameter and its velocity are governed by the balance of the surface tension σ, liquid
viscosity η and electrical shear stress qE. These forces, in a simplified scalar form, are given
by the following equations (cf. Jaworek and Krupa, 1999):
(3)
(4)
where q is the surface charge density on the jet, E is the electric field, and vjz and vjr are the
axial and radial components of liquid velocity. In these equations ∆vjr/∆r and ∆vjz/∆z are
the gradients of the liquid velocity, which are unknown functions of acceleration force.
From these equations it can be concluded that viscosity has only a minor effect on the jet
formation. Viscosity influences the length of the jet by opposing the tangential component
of electric field (cf. Eq.(4)). Viscous force can also counteract the surface tension effect thus
resulting in dynamic decrease in the jet diameter (cf. Eq.(3)). It was also observed that with
increasing voltage, the jet becomes thinner, transporting the liquid with higher velocity.
Fig.4 shows that the increase in viscosity causes a slight decrease in the jet diameter and
in droplet size. The jet diameter was estimated from the photographs. Precise theoretical
determination of the jet diameter from equation (3) and (4) is not possible because the
gradient of the jet velocity along the jet and the surface charge density at the jet surface are
difficult to estimate.
These results are in accordance with those obtained by Hines (1966), who concluded
that the jet diameter decreases as the liquid viscosity increases, but remains in contrast with

185
the dimensional analysis of Fernandez de la Mora and Loscertales (1994). They assert that
the jet diameter is independent of liquid viscosity, and the droplet diameter to jet radius
ratio increases with the increase of liquid viscosity. From our experimental results both of
these quantities decrease with liquid viscosity within the tested range. Smith (1986) also
concluded that the aerosol droplet size increases with increasing viscosity of the liquid, but
his experiments were carried out for not constant flow rate but constant pressure.
Recently, Hartman et al. (2000) also investigated the effect of liquid viscosity on droplet
size and obtained an increase in the primary droplet size but decrease in the secondary
droplet, with viscosity increasing in the range from about 2 to 50 mPas.
Garcia and Castellanos (1995) considering the effect of liquid viscosity on stability and
breakup of a liquid jet concluded that the viscosity has diminishing effect on the drop size
when the electric field increases, whereas the length of the jet is mainly controlled by the
liquid viscosity and the electric field is of less importance. Ogata et al. (1978) determined
from measurements for water-glycerol mixtures that the volume surface mean diameter of
the droplets is proportional to viscosity η12/9. These results suggest that the droplet size
should slowly increase with increasing viscosity, and should increase about 2 times for our
experimental conditions, but this was not confirmed by the measurements (cf. Fig.4).
In a newly published paper, Jayasinghe and Edirisinghe (2002) confirmed experimentally,
for water-glycerol mixtures, that the increase in viscosity causes the jet diameter to decrease.
Watanabe et al. (2003), however, noticed that an increase of liquid viscosity made the
droplets larger, and the length of the spindle at the capillary outlet also increased with
viscosity.
5. Conclusions
The effect of viscosity on electrohydrodynamic spraying of liquids is a more complex
problem than predicted by any theoretical consideration. Experimental results presented in
the literature are not conclusive because viscosity of a liquid can not be changed with
remaining other important parameters constant.
In this paper it was experimentally shown that the increase in liquid viscosity causes the
jet diameter to slightly decrease and the jet length to increase. Also the diameter of the
generated droplets slowly decreases with increasing liquid viscosity. The results were
obtained for a liquid conductivity which varied within the range of one decade only. The
jet mode develops from the spindle mode for lower voltages as the liquid viscosity increases.
It was also noticed that the spindle mode changes to the jet mode when the inertial time
constant is higher than the flow-rate time constant.

186
τρ intertial time constant

ρl liquid density
D0 capilary diameter
σl surface tension
τl flow-rate time constant
Qv liquid volume flow-rate
η liquid viscosity
ri jet radius
vi jet velocity
Acknowledgments
This paper has been supported in part by Brunei University and in part by the Institute of
Fluid Flow Machinery, Polish Academy of Sciences. The scientific cooperation was
supported by British-Polish Research Partnership Programme WAR/341/222.
References
Fernandez de la Mora J., Loscertales I.G., 1994, The Current Emitted by highly Conducting Taylor
Cones. J. Fluid Mech. 260, 155–184
Garcia F.J., Castellanos A., 1995, Effect of Interfacial Electrical Stresses of the Stability of Viscouse
Liquid Jets. Inst. Phys. Conf. Ser. No 143, 301–304
Hartman R.P.A., Brunner D.J., Camelot D.MA., Marijnissen J.C.M., Scarlett B., 2000, Jet break-up in
electrohydrodynamic atomization in the cone-jet mode. J. Aerosol Sci. 31, No.l, 65–95
Hines R.L., 1966, Electrostatic atomization and spray painting. J. Appl. Phys. 37, No.7, 2730–2736
Huneiti Z.A., Balachandran W., Machowski W., 1998, Harmonic spraying of conducting liquids employing
AC-DC electric fields IEEE Trans. Ind. Appl. 34 No.2, 279–285
Jaworek A., Krupa A., 1999, Classification of the Modes of EHD Spraying. J. Aerosol Sci. 30, No.7,
873–893
Jayasinghe S.N., Edirisinghe M.J., 2002, Effect of viscosity on the size of relics produced by electrostatic
atomization. J. Aerosol Sci. 33, No.10, 1379–1388
Ogata S., Hatae T., Shoguchi K., Shinohara H., 1978, The dimensionless correlation of mean particle
diameter in electrostatic atomization. Int. Chem. Eng. 18, No.3, 488–493
Smith D.P.H., 1986, The Electrohydrodynamic Atomization of Liquids. IEEE Trans. Ind. Appl. 22, No
3, 527–535
Watanabe H., Matsuyama T., Yamamoto H., 2003, Experimental study on electrostatic atomization of
highly viscous liquids. J. Electrostatics 37, No.2, 183–197

187
Pumping of electrolytes using arrays of asymmetric pairs

of microelectrodes subjected to ac voltages
Antonio Ramos1, Antonio González2, Antonio Castellanos1, Nicolas G Green3,

Hywel Morgan3
1
Dpto. Electronica y Electromagnetismo, Universidad de Sevilla, 41012 Sevilla,
Spain
2
Dpto. Fisica Aplicada III, Universidad de Sevilla, 41092 Sevilla, Spain
3
Dpt. Electronics and Electrical Engineering, University of Glasgow, G12 8LT, UK
Abstract. Net flow of water driven with an array of asymmetric pairs of microelectrodes
subjected to an ac electric potential has recently been reported. The oscillating electric field
interacts with the oscillating induced charge in the diffuse double layer on the electrodes. A
steady electroosmotic velocity distribution on top of the electrodes is then originated. Net fluid
flow is produced because the slip velocity distribution is anisotropic. This work presents a
theoretical analysis of the pumping phenomena and observations of experimental streamlines
of fluid flow on top of a microelectrode array. The analysis is based upon an electroosmotic
model in ac fields. The electrical equations are solved numerically using the charge simulation
method. The bulk flow generated by the electroosmotic slip velocity is calculated. Experimental
streamlines of the flow induced on top of a microelectrode array are reported and compared
with the theoretical results.
1. Introduction.
There is a requirement in technological applications (especially in biotechnology) for
precise control of fluids in small channels (a process called microfluidics) [1]. Many
techniques have been developed to pump liquids in microsystems, including
micromechanical methods, electroosmosis, electro-wetting, thermocapillary pumping and
electrohydrodynamic pumping. However, all of these systems present inherent drawbacks;
some require external temperature gradients or high applied voltages, others use moving
parts or produce pulsating flow.
Recently, Brown et al [2] have demonstrated pumping of an electrolyte on an array of
asymmetric microelectrodes, energised by a single ac signal of the order of kilohertz and at
low voltage (around 1 volt). The system is, therefore, of interest in the development of
micro-pumps. Figure 1 shows a schematic diagram of the microelectrode system. This type
of pump followed from theoretical arguments regarding unidirectional flow resulting from
applied spatially asymmetric potentials [3]. The physical mechanism responsible for the
flow is electroosmosis in ac fields, where non-uniformities in the field geometry produce a
non-zero time-averaged electroosmotic slip velocity at the surface of the electrodes [4–7].
In this paper, we present experimental observations of streamlines and a theoretical
analysis of the pumping phenomena. The theoretical analysis is based upon the linear
electroosmotic model presented in [6,7].

188
Figure 1: Schematic diagram of the array cross section showing the asymmetric pairs of electrodes
Figure 2: Superimposed succesive video frames showing the experimental streamlines.
2. Experimental set-up.
The experiments were performed on an array of interdigitated microelectrodes mounted on
a glass slide. The electrodes consisted of a large electrode (of width W1=100 µ̃ m) separated
from the small electrode (W2=10 µm̃) by a gap (G1=10 µm̃). The asymmetric pair of electrodes
was repeated leaving a gap between the small electrode in a pair and the large electrode in
the next pair, G2=100 µ m. The periodic array had, therefore, a wavelength of 220 m̃. The
electrodes were 120 nm thick, made of a layer of gold sandwiched between two layers of
titanium. A square glass chamber was constructed around the electrode array and filled with
electrolyte. The electrode/electrolyte system could be observed both from above and from
the side. A microscope objective and camera were pointed horizontally along the electrodes,
so that the electrodes could be imaged in cross section. A concentration of KCl in water of
conductivity 2.1 mS/m was used for the electrolytic solution. An ac potential difference
was applied to each pair of asymmetric electrodes with voltage amplitude from 0 to 6 V, and
frequencies from 1 to 100 kHz. Electrolysis does not occur because of two reasons: the
electrodes are coated with an insulating oxide layer, and the voltage drop is distributed
between two double-layers and the bulk (the voltage across the bulk increasing with
frequency). Fluorescent latex particles of 557 nm in diameter were used as tracers to observe
the fluid flow. The movement of the particles was recorded on video and transferred to a
computer. Successive video frames were superimposed to visualise the fluid flow streamlines.
After applying an ac voltage to the microelectrode array, fluid flow was originated (see
fig. 2). The fluid flow was driven at the electrode surface, with highest velocities observed
close to the gap between each pair of asymmetric electrodes. Net flow resulted from the small

189
electrode to the nearest large electrode (right to left in the figure), going down and up when
the fluid approached the inter-electrode region. Because the chamber is closed at right and
left ends, the fluid recirculates, going from left to right at the top. For this figure, the frequency
and voltage amplitude of the applied signal were 5 kHz and 4 volts, respectively. Velocities
around 20 µ̃m/s were observed at a height from the electrodes of 100 µm. The velocity
increased with voltage, and it was around 50 µm at 6 V for the same position.
3. Formulation of the problem

The system used to generate the fluid flow is shown by the 2D geometry depicted in fig. 1.
An infinite periodic array of asymmetric pairs of electrodes is fabricated onto an insulating
substrate (glass). The electrodes are considered to be infinitely long and thin, so that any
single pair of electrodes can be characterised by widths W1 and W2. The gaps between
consecutive electrodes are G1 and G2 as shown in the figure. The size of a repeating basic
cell is, therefore, L=W1+W2+G1+G2. Above the electrode array there is a solution of KCl,
with conductivity σ and permittivity ε.̃ At the interfaces between the metal and electrolyte,
and glass and electrolyte, double layers are formed. The characteristic thickness of a double
layer is given by the Debye length, λD, and in most cases it is negligibly small (~10 nm)
compared to the other lengths in the system. When an ac voltage is applied to the electrodes,
an electrical current is established in the solution. To a first approximation we assume that
the applied voltage is low enough such that electrolysis of the electrolyte does not occur.
We also assume that the frequency of the applied signal is low enough, i.e. ωλD2/D<<1, so
that the double layer is in quasi-equilibrium. Here D is the mean diffusion coefficient of the
ions and λD2/D=ε /σ the time an ion takes to travel the Debye length by diffusion. For
periods of the applied signal much greater than λD2/D, the ions can equilibrate locally.
Under these conditions the bulk electrolyte behaves in a resistive manner and the double
layer can be considered to behave as an ideal capacitor. Therefore, the electrical potential
φ in the bulk electrolyte satisfies Laplace’s equation
(1)
The boundary condition on the electrode surface describes the charging of the double layer
due to the current in the bulk. For sufficiently low voltage there is a linear relationship
between the surface charge and the voltage drop across the double layer. In this case, the
surface charge conservation equation can be written using phasors as
(2)
where CDL is the capacitance per unit of area of the total double layer (diffuse plus compact
or Stern layers), i is the imaginary unit, φ is the potential just outside the double layer, and
Vj is the potential applied to electrode j. At the interface between the electrolyte and the
glass a similar boundary condition holds, however, in this case the boundary condition can
be simplified to ∂φ/∂y=0.The boundary condition at y→∞ is that the potential tends to
zero. For a bi-dimensional problem, this is a correct boundary condition provided that the
total electrical flux at x=0 is zero over a wave-length. This ensures that the electrical current
cannot extend to infinity.
Owing to electrode polarisation, the electric field in the bulk electrolyte is frequency-
dependent. When the frequency is low, most of the applied voltage is dropped across the
double layer (across the capacitor). Conversely, when the frequency is high, most of the
applied voltage is dropped across the bulk electrolyte. The typical transition frequency can

190
be estimated from simple circuit theory to be ω0~(σ/ω)(λD/l), which is several orders of

magnitude smaller than the charge relaxation frequency σ/ε. Fluid flow due to ac
electroosmosis is observed in the region of this characteristic frequency ω0.
Once the electric potential has been solved, the electro-osmotic fluid velocity at the
surface of the electrodes can be calculated. For diffuse double layers in quasi-equilibrium
on perfectly polarisable metal surfaces, the electro-osmotic slip velocity is given by the
Helmholtz-Smoluchowski formula [7,9]
(3)
Here η is the viscosity of the fluid, ∆φ the potential drop across the diffuse double layer and Ex
is the tangential electric field outside the double layer. For our problem, both ∆φ and Ex are
oscillating functions of time, of frequency ω. The time-averaged horizontal fluid velocity at the
interface between the double layer and the bulk is [6,7]
(4)
where A is given by Λ=Cs /(Cs+Cd) and Cd and Cs are the capacitances of the diffuse and
Stern or compact layers, respectively. For the glass/electrolyte interface, an estimate of the
potential drop across the diffuse double layer shows that this is negligibly small and, from
equation (3), the electroosmotic velocity on the glass is negligible.
To obtain the velocity in the bulk, the Navier-Stokes equations must be solved. For
micro-systems the Reynolds number is usually very small, so that the convection term in
the equations can be neglected. For the time-averaged velocity, and in the absence of
externally applied body forces, the equations reduce to:
(5)
where p is the pressure and u the velocity field, u=u ex+v ey. The boundary conditions for
these equations at y=0 are: (a) the tangential velocity is equal to the slip velocity on the
electrodes, given by eq. (4); (b) the tangential velocity is zero at the glass; (c) the normal
velocity is zero for any x at y=0 (electrodes and glass). Far from the electrodes in the normal
direction, the fluid can be considered to be free of stress, where ∂u/∂y=0 and v=0. This
boundary condition assumes that the upper boundary of any actual device is much higher
than the wave-length of the problem. Given the periodicity of the electrode array we look
for periodic solutions in the x direction. If a net fluid flow occurs, the velocity tends to u=U
ex for y→∞, where U is a constant. Net flow will occur if the average value of the slip
velocity over a wave-length is non-zero, i.e. . This equality can be
proven by solving eq. (5) in the x direction using Fourier analysis. The velocity Fourier
component that produces net flow is the one with zero wave-number, the constant
component.
4. Numerical analysis
The electrical potential can be numerically solved for the periodic array using the surface
charge simulation method. The periodicity of the system is taken into account by considering
the potential generated by an infinite array of parallel source lines. The potential due to a
uniform grid of source lines of unit value, with spacing L along the x axis at positions
x=0,±L,±2L,±3L,… is [10]

191
(6)
where k=2π/L. Taking into account the potential created by a periodic surface charge
distribution of period L placed at y=0 and the relationship between the surface charge and
the normal derivative of the potential at y=0, it can be written for y=0
’ (7)
This equation relates the potential at y=0 to the normal derivative at y=0, and together with
the boundary condition eq. (2), is the governing equation for φ that has to be solved
numerically. We have used a Galerkin formulation in order to obtain a numerical
representation of ∂φ/∂y on each electrode. Subsequently, the numerical slip velocity on
each electrode is calculated from eq. (4).
Once we have calculated the electroosmotic slip velocity on the electrodes, the velocity
in the bulk is obtained through a numerical integration. Let us define the stream function
by the pair of derivatives ∂/∂y=u, ∂/”x=-v. The stream function originated by an
array of velocity source lines with spacing L along the x axis at positions x=x’+nL is [11]
(8)
At y=0 these velocity sources satisfy v(x)=0, Therefore, the stream

function generated by a periodical velocity distribution given by the electroosmotic slip
velocity (eq. (4)) is
(9)

The numerical results indicate that net fluid flow occurs in the direction from the narrow
electrode to the nearest wide electrode, i.e. to the left in fig. 1. This is the direction observed
in the experiments. Figure 3 shows the pumping velocity U as a function of frequency for
the asymmetric array of the experiments of ref. [2]: W1=4.2 µm, W2=25.7 µm, G1=4.5 µm,
G2= 15.6 µm and V0=1V. We have taken a surface capacitance (from the Debye-Huckel
theory) CDL=ε/λD=0.0233 F/m2 and a parameter Λ=1 (no compact layer). The computed
frequency of peak velocity is 2.08 kHz compared to a experimental value of 2.9 kHz. The
numerical maximum velocity is 2.7 times greater than the corresponding experimental
value given in ref. [2]. The existence of a compact layer at the electrode surface would
reduce the predicted velocity and increase the predicted frequency. The experimental
values are obtained if the capacitance of the diffuse layer is set equal to 1.9ε/λD and that of
the compact layer set to 0.9ε/λD.
Figure 4 shows the streamlines computed for our experimental set-up at a non-dimensional
frequency =10. In order to impose the non-slip condition at the upper wall placed at y=h
and zero net flow (recirculation), we have added to the function given in eq. (9) the stream
function =Uh((y/h)3–2(y/h)2). At a height of the order of G1, the flow is very close to that
originated by an array of velocity sources with a direction from right to left in the figure, i.e.

192
Figure 3: Numerical pumping velocity as a function of frequency for the case of ref. [2]
Figure 4: Computed streamlines for W1=100 µm, W2=10 µm, G1=10 µm, G2=100 µm, at ωCDLW1/
σ=10
from the small to the large electrode of a pair. This compares with the experimental streamlines
of fig. 2.
Acknowledgements
The authors would like to thank the Spanish government agency Direction General de
Ciencia y Tecnologia under contract BFM2000–1056, the Royal Academy of Engineering
UK and the Royal Society for funding.
References
[1] Whitesides GM and Stroock A, Physics Today, 54, 42, June (2001).
[2] Brown ABD, Smith CG and Rennie AR, Phys. Rev. E, 63, 016305 (2001).
[3] Ajdari A, Phys. Rev. E, 61, R45 (2000).
[4] Ramos A, Morgan H, Green NG and Castellanos A, J. Colloid Interf. Scl, 217, 420 (1999).
[5] Green NG, Ramos A, Gonzalez A, Morgan H and Castellanos A, Phys. Rev.E, 61, 4011 (2000).
[6] González A, Ramos A, Green NG, Castellanos A and Morgan H, Phys. Rev. E, 61, 4019 (2000).
[7] Green NG, Ramos A, Gonzalez A, Morgan H and Castellanos A, Phys. Rev. E, 66, 026305
(2002).
[8] Hunter RJ, Zeta Potential in Colloid Science. Academic Press, San Diego (1981).
[9] Levich VG, Physicochemical Hydrodynamics. Prentice-Hall, Englewood Cliffs, NJ (1962).
[10] Morse PM and Feshbach H, Methods of Theoretical Physics, Me Graw-Hill, NY (1953).
[11] Ramos A, Gonzalez A, Castellanos A, Green NG and Morgan H, Phys. Rev. E (in the press).

193
Electrode polarisation, dielectrophoresis and

electrorotation
Nicolas G Green
AC Electrokinetics and Electrohydrodynamics Group
Department of Electronics and Electrical Engineering, University of Glasgow,
Oakfield Avenue, Glasgow G12 8LT, Scotland, UK
Abstract. Dielectrophoresis (DEP) and Electrorotation (ROT) are used for the
characterisation, identification, manipulation and separation of a wide range of bioparticles,
such as cells, bacteria and viruses. The particles are suspended in aqueous solutions of ions
and the electric fields are generated by arrays of microelectrodes by the application of AC
potential signals. When a potential is applied across an electrolyte using electrodes, an
electrical double layer forms in the fluid above the electrode surface, screening the potential
from the bulk and reducing the electric field in the bulk. This paper discusses a linear
solution of electrode polarisation in microelectrode geometries and the frequency dependent
effect on dielectrophoresis and electrorotation. The movement of bio-particles, such as cells,
under these conditions is discussed as a function of frequency
1. Introduction
Dielectrophoresis (DEP) is the movement of polarisable particles resulting from the
interaction of a non-uniform electric field and the dipole induced in the particles [1,2].
Electrorotation (ROT) is the rotation resulting from the interaction of a rotating electric
field and the particle [1]. Collectively referred to as AC Electrokinetics, these techniques
have been used for the characterisation, identification, manipulation and separation of a
wide range of bioparticles, such as cells, bacteria and viruses. In practice, the particles are
suspended in aqueous solutions of ions (electrolytes), with a range of conductivities. The
electric fields are generated by arrays of microelectrodes from the application of AC
potential signals [2].
When a potential is applied across an electrolyte using electrodes, an electrical double
layer forms in the fluid above the electrode surface. In many cases, the charging of the layer
can be considered to be analogous to the charging and polarising of a capacitor, and the
mechanism is often referred to as electrode polarisation. The double layer consists of the
counterionic species to the applied potential, attracted to the surface by the potential and
repelled by diffusion. At low frequencies, the charging of the double layer screens the
potential from the bulk, reducing the electric field experienced by the particle in the bulk
[3].
For microelectrode structures and AC potential signals, the charging of the electrical
double layer is a complicated, frequency and geometry dependent problem. The reduction
in the electric field strength leads to a reduction in the dielectrophoretic force. In addition,
the non-uniform field interacts with the double layer to produce a tangential flow, referred

194
to as AC electroosmosis [4]. This paper discusses electrode polarisation in microelectrode

geometries typically used in AC electrokinetic experiments and the frequency dependent
effect on dielectrophoresis and electrorotation. The model is based on a linear model of the
double layer, determined from experimental measurements of the double layer impedance.
The movement of a variety of bioparticles under these conditions is also discussed.
2. Background, theory and numerical simulation

2.1 Dielectrophoresis
Dielectrophoresis is the movement of a particle caused by the interaction of a non-uniform
electric field and the dipole moment it induces in the particle. The time-averaged
dielectrophoretic force is generally written as [1,2]
(1)
where υ is the volume of the particle, is the complex, frequency-dependent effective

polarisability of the particle, Re[…] indicates Real part of and E is the electric field phasor.
This expression is valid if the electric field phasor is only real. If the electric field phasor is
complex, the dielectrophoretic force is given by
(2)
where Im[…] indicates the imaginary part of. This is the expression required for travelling
wave dielectrophoresis, with the travelling wave force given by the second term on the
right hand side of equation (2).
If there is a spatial variation in the phase of the field: if the field phasor is complex, the
travelling wave part of the dielectrophoretic force is non-zero.
2.2 Electrorotation
Electrorotation is the rotation of a particle caused by the interaction of a rotating electric
field and the induced dipole moment. The time-averaged electrorotational torque is
(3)
This torque is also non-zero in a complex field.
2.3 Frequency dependent polarisability

The frequency dependent polarisability in equations (1), (2) and (3) is, for a solid spherical
particle suspended in a fluid of permittivity £m, is given by
(4)

195
Figure 1 Plots of the real (solid line) and imaginary (dashed line) of the Clausius-Mossotti factor for (a)
a solid spherical particle and (b) the two shell model for a biological cell with membrane,
cytoplasm and nucleus.
where and are the complex permittivities of the fluid and particle respectively.
A general complex permittivity is given by where is the permittivity, σ is the
conductivity, ω is the frequency of the applied AC field and i is the imaginary unit.
The term in brackets in equation (4) is referred to as the Clausius-Mossotti factor and
describes the frequency dependence of the polarisability. Figure (1) shows plots of the real
and imaginary parts of the Clausius-Mossotti factor for a solid particle and a biological
cell, modelled using the shell model [1,2].
3. Numerical simulation of the electrical potential

3.1 Polarisation of the double layer: Electrode Polarisation
In this paper, we will assume that the polarisation of the double layer is linear and that it can
be represented by a constant capacitance across the electrode surface. This is an
approximation that will not be valid close to the electrode edge where the field is strongest.
The double layer also consists of two distinct layers, the diffuse layer consisting of a cloud
of counterions which have close to bulk diffusion properties and the Stern layer which
consists of a layer of packed hydrated ions and possibly bare adsorbed ions [3]. Since direct
simulation of electrode polarisation is therefore very complicated, we will use the
experimentally measured impedance of the double layer to calculate the polarisation [4].
3.2 Simulation of potential and electric field

The electrical potential in the bulk, φ , is then a solution of Laplace’s equation with the
boundary condition at the electrode surface given by
(5)
where ZDL is the measured specific impedance of the double layer and Vo is the potential
applied to the electrode. The complex potential φ=φ R +i φ I is solved in the problem space
shown in figure (2), with the symmetry boundary conditions as shown and the above
expression for the boundary condition on the electrode. This is the minimum solution
space for interdigitated bar electrodes [2], assuming that the electrodes are infinitely long

196
to give a two-dimensional problem. The

symmetrical nature of the electrodes means that
the problem space consists of one half of one
electrode and half of the adjacent gap.
The potential was solved using FlexPDE®, a
commercially available Finite Element Partial
Differential Equation solver, for a conductivity of
2×10-3Sm-1 and electrode width of 10µm. The
problem was posed as the simultaneous solution
of the Laplace equation for the real and imaginary
parts of the potential, linked by the boundary
condition, given by equation (5) on the electrode.
The resulting potential at the electrode surface is
shown in Figure (3). The real part of the potential Figure 2 The problem space for the complex
decreases with frequency, but importantly, does potential φ and the boundary
conditions.
not decrease in the same manner across the
Figure 3 Plot of the real (a) and imaginary (b) parts of the complex potential at the surface of the
electrode plotted as a function of frequency and distance from the centre.
Figure 4 Plot of the magnitude of the electric field in the problem space at (a) 100Hz and (b) 100kHz.
The grayscale for the plots is log10(magnitude).

197
electrode surface. The imaginary part is maximum around 1kHz for this conductivity. The
magnitude of the resulting electric field is shown in Figure (4) for two frequencies, 100Hz
(a) and 100kHz (b). At high frequencies, the electric field magnitude is symmetrical about
the edge of the electrode at 5µm. At low frequencies, the electric field magnitude is no
longer symmetrical.
The effect of this change in field distribution can be seen in Figure (5), where at high
frequencies the DEP force (figure 5(a) and the first term in equation (2)) is symmetrical
about the electrode edge, but at low frequencies (figure 5(b)) the DEP force direction is no
longer symmetrical. Note that the magnitude is much smaller (c.f. figure (7)).
In addition to this alteration to the DEP force, the twDEP force is non-zero (second term
in equation (2)). Figure (6) shows the twDEP force component direction for the same two
frequencies, the magnitude of which is highest around 1kHz. This will not affect the solid
sphere (figure 1(a)) since but will affect the cell since the .
Figure 5 Plot of direction of at (a) 100kHz and (b) 100Hz.
Figure 6 Plot of direction of at (a) 100kHz and (b) 100Hz.

198
Figure 7 Plot of magnitude of the DEP force for three positions (indicated by the legend) along
y=2.5µm (a) and y=7.5µm (b), showing the variation in the frequency dependency.
However, this additional force is small

compared to the DEP force component.
What is significant is the fact that at
distances less than the typical size of the
electrodes, the frequency dependence of
the DEP force is spatially dependent as
shown in Figure (7). In addition, there is
a substantial electrorotational torque
acting on particles close to the electrode
edge, as shown by figure (8).
4. Conclusion
It has been demonstrated that the effects
of electrode polarisation in planar Figure 8 Plot of the rotation rate of a typically particle
due to the electrorotational torque. The
microelectrodes cannot be described by
grayscale is log10(rad s-1).
simply multiplying the DEP force by a
fixed frequency dependent factor. The non-uniform nature of the electrode polarisation
leads to a spatially dependent frequency variation as well as travelling wave effects and
strong electrototational torques close to the electrode edge. The results were based on a
linear model of the double layer and a better picture of experimental results would require
more accurate double layer models.
Acknowledgements
The author is a Royal Academy of Engineering Post-doctoral Research Fellow at the
University of Glasgow. The author would also like to thank Professor Hywel Morgan,
Professor Antonio Castellanos, Dr Antonio Ramos and Dr Antonio Gonzalez for valuable
discussions.
References
[1] Jones T.B. Electromechanics of Particles, Cambridge Univ. Press, Cambridge (1995)
[2] Morgan H. and Green N.G. AC Electrokinetics: Colloids and Nanop articles, Research Studies
Press, Baldock, UK (2003)
[3] Lyklema J. Fundamentals of Interface and Colloid Science Acad. Press Ltd, London, (1995).
[4] Green N.G., Ramos A., Gonzalez A., Morgan H. and Castellanos A. Fluid flow induced by non-
uniform AC electric fields in electrolytes on microelectrodes III: Observation of streamlines and
numerical simulation Phys Rev E 66 art no. 026305 (2002)

199
Numerical simulation of fine particles charging and

collection in an electrostatic precipitator with regular
barbed electrodes
L M Dumitran1,2, P Atten1* and D Blanchard1

1
LEMD, CNRS and Joseph Fourier University, BP 166–38042 Grenoble Cedex 9
(France)
2
Laboratory of Electrical Materials, Politehnica University of Bucharest, 313 Splaiul
Independentei, 77206 Bucharest (Romania)
Abstract. The processes of charging and collection of fine particles (≤a few µm) in an
electrostatic precipitator (ESP) are modelled taking into account the secondary gas flow
induced by “barbed” ionizing electrodes. In particular, in configurations characterized by a
rectangular lattice of ionizing point electrodes, the secondary flow in the form of axial rolls can
be vigorous and the resultant gas flow inside the precipitator can be considered as axially
invariant to a first approximation. A numerical simulation of charging, transport and collection
of fine particles is presented, based on the approximation of a gas flow independent of the axial
variable. The three-dimensional character of the distributions of electric field E and ionic space
charge ρ which plays an important role in the charging of particles is taken into account. This
allows simulating the particles trajectories and the dynamics of their charging. A statistical
study shows that all particles of a given size do not get the same charge. The histograms of
charge are presented and discussed. Finally the obtained mean particle charges are compared
with theoretical predictions.
1. Introduction
Electrostatic precipitation is a commonly used technique to clean the gases exhausted in
atmosphere by many industrial factories or plants. But the efficiency of collection of
industrial precipitators (ESPs) is generally poor for fine particles of size ranging from 0.1 to
1 µm which are hazardous for human health. In order to try to improve the removal of fine
particles it is necessary to characterize their properties and behaviour during the precipitation
process. Many methods were proposed to account for the observations and measurements
on laboratory and industrial ESPs. The oldest one is the simple analytical approach of
Deutsch [1] refined by Leonard [2]. More recently numerical models were developed by
several authors (Meroth [3], Medlin [4], Egli et al. [5], etc..). The electrohydrodynamic
effects are retained with different degrees of complexity. However, almost all recent
numerical models consider the simplified case of homogeneous discharges along the
cylindrical corona electrodes, which restricts the influence of the electrical forces on the
gas flow mainly to velocity modulation in a horizontal plane (Oxy—Figure 1-a). Moreover,
these authors generally use the so-called k-ε approach to describe the small-scale gas

200
Figure 1. a) Schematic view of the electrostatic precipitators with barbed ionising electrodes and picture
of axial rolls; b) Convective cell and 2-D computation domain.
turbulence and neglect any secondary flow induced by a non-uniform distribution of

Coulomb force. Using these models is questionable in the case of barbed ionising electrodes
where the corona discharges are localized and the charge density distribution ρ is strongly
non uniform.
In a previous work performed on two laboratory ESPs with a rectangular lattice of point
electrodes [6,7,8], a rather vigorous secondary flow in the form of axial rolls was observed
as sketched in Figure 1-a. These rolls arise from the periodicity along the vertical direction
Oz of the distributions of electric field E, charge density ρ and force density ρẼ. A first
important observation is that the gas flow inside the filter exhibits the same structure in all
cross-sections Oyz. To a first approximation, the secondary flow does not depend on the
axial direction Ox. This two-dimensional (2-D) character of the large scale secondary gas
flow allowed to perform a rather simple analysis [6,9] in a convective cell (Figure 1-b).
A second important observation is that the typical value of the cross-flow velocity
component in the convective cells is clearly higher than the velocity of fine particles
induced by the electric force [8]. This was justified by the numerical approach based on the
2-D model of the convective flow [9]. Therefore the secondary flow can basically influence
the charging and transport of fine particles in precipitators and the collection efficiency.
A first investigation of this question was developed by neglecting the axial modulation
of E and ionic space charge distributions [6] (2-D approach). This study gave an illustration
of some of the processes occurring in ESPs with barbed ionising electrodes. But the 3-D
character of the distribution of ionic density might play an important role in the process of
particles charging. We present here a study retaining realistic distributions of E and ρ.
2. Gas flow model

A basic fact is that the drift velocity of ions (~102 m/s) is much higher than the gas velocity
(~1 m/s); the influence of convection on transport of ions is therefore negligible and the
electric problem for the ions decouples from the fluid dynamical problem. The charged fine
particles, conversely, have a drift velocity with respect to the gas of the order of 10-1 m/s and
the entrainment by the secondary gas flow inside of ESP plays a very important role [6–9].
For the secondary flow, the experimental observations suggest a simplification of the

201
problem by stating that the velocity components depend on y and z only (axial invariance).
Such an approximation might appear drastic but i) the typical entry length for the growth of
v and w from zero to their steady value is limited and ii) a moderate axial modulation of the
velocity components will not qualitatively change the influence of the secondary gas flow
on the transport and collection of fine particles.
The flow retained here is similar to the one computed in [6–9]. The Navier-Stokes
equations of conservation of mass and momentum lead to a 2-D problem in the Oyz plane
for the stream function and the vorticity where the electrical terms come from the axially
averaged values of electric field and ionic charge density. The effect of small scale turbulence
is heuristically taken into account through a turbulent diffusion term. The equations are
solved in the domain shown in Figure 1-b with the adequate boundary conditions, which
gives the velocity components v and w [6–9]. Then u(y,z) is determined [8].
3. Three-dimensional distributions of electric field and ionic space charge

Particles get their charge from the ions surrounding them through the two mechanisms of
field charging and diffusion charging. With barbed ionising electrodes, a noticeable volume
fraction of the precipitator is charge free. In these zones ρ=0 there is no charging; it is
therefore important to determine the 3-D distribution of ionic charge density. By neglecting
the convection and diffusion components of the ionic current density j, the governing
equations for electric potential Φ and ionic space charge ρ are, at steady state :
(1)
(2)
where εo is the vacuum permittivity and K the mobility of ions. Concerning the charge
conservation equation (2), the correct formulation from the mathematical viewpoint requires
to prescribe, as boundary condition, the charge density ρ =ρ0 at the injecting electrodes.
The ρ 0 value has been chosen in such a way that the computed mean current density on the
plates takes the same value than the experimental one. Because of the double periodicity of
the tips (in Ox and Oz directions) and of the associated symmetries, the computation domain
is restricted to the box depicted in Figure 2-a.
Equations (1) and (2) are integrated using the classical SOR and MOC methods
respectively and the solution is obtained by successive approximations [6–9]. Figure 2-b
shows the typical charge density distribution in the horizontal plane z=0 and Figure 2-c
gives the current density on the collecting plate y=d. These figures illustrate that ρ =0 in a
rather large volume and that the ionic current density on the grounded plate is non zero
only in a zone of ellipse shape facing the injecting point; around this zone we have j=0 [7].
4. Particle trajectory and charging

We propose here a lagrangian approach which has the advantage that it is possible to
examine in detail the charging and collection of particles and to highlight the role of the
secondary flow in the separation process. Taking into account the periodicity of the
ionizing electrodes, we have the spatial distributions of E, ρ and of the gas velocity field

202
Figure 2. a) 3-D computational domain (total length of the filter L=100 cm, plate-plate distance 2d=9 cm,
total height h=29 cm, point spacing s=3.8 cm); b) Ionic space charge distribution in the symmetry plane
z=0 (ρ0=1.4.10–3 C/m3 and Vappl=20 kV); c) current density distribution on the grounded plate y=d.
Ug in the whole volume of the electrostatic precipitator. The motion equation for spherical
particles is:
(3)
where vp is the particle velocity, mp and qp are the particle mass and charge and Cu is the
Cunningham correction factor depending on the particle diameter dp. The response time of
particles to changes in the local gas velocity is τ̃p=mpCu/(3πηgdp). For small particles (dp<2
µm) this time is very short (τ̃p<10 µs) so that the acceleration phase is neglected. The
equations for the trajectories of the charged particles are then simply dxp/dt=Ug+KpE where
the particle mobility Kp is proportional to the instantaneous particle charge qp. Unlike
previous lagrangian approaches [3,4], the effect of small scale turbulence is neglected here
and only the contribution of the secondary flow is retained. The particle charge qp is a
function of time and depends on the local values of electric field and ionic space charge. To
determine the evolution of qp along the trajectory we used the equations of the field modified
diffusion model of Lawless [10] taking into account the field and diffusion mechanisms of
charging (see the set of equations in [10]). The main parameter in the charging process is the
saturation charge qps due to field charging only :
(4)
Figure 3 presents examples of trajectories and charge evolution as a function of the non
dimensional time t*. Clearly the gas flow plays a major role, the trajectories roughly looking
like expanding helices. The increase in the mean radial scale of the quasi helices results
from the action of the electric field on the charge carried by the particles.

203
Figure 3. Trajectories and charge evolution for 3 particles of different diameter (Vappl=20 kV, Ug=1.3 m/
s). On the figures of charges, the thin line represents the local saturation charge by field qps while the thick
one is the total charge.
There is no qualitative change in the shape of trajectories when dp passes from 0.1 µm to a
few micrometers; only the expanding rate increases with dp (Figure 3).
Concerning the evolution of qp, Figure 3 illustrates the fact that diffusion charging is the
dominant mechanism for small particles (dp<0.5 µm) whereas field charging is dominating
for dp>1 µm. The particle charge increases mostly by steps and not smoothly. When retaining
the 2-D distribution for E and ρ, the charge increments occur roughly periodically and arise
from the passage of the particles in the zone close to the injecting blade [6] where E and ρ
take high values. Here the quasi-periodic passage of the particles in the vicinity of the axis
y=z=0 most often results in no charge increment because of the zero charge density there,
except very close to the points. Figure 3 clearly shows that most of the peaks of the saturation
charge qps (field charging) do not correspond to an increment in qp. In comparison with the
results obtained in [6], this observation is very significant because it illustrates that the
charging process firstly depends on the ionic charge distribution.

204
Figure 4. Histograms of charge of collected Figure 5. Histograms of charge of non-collected

– –
particles (Vappl=20 kV, Ug=1.3 m/s). particles (Vappl=20 kV, Ug=1.3 m/s).
5 Results and discussion

A statistical study was performed on sets of 105 particles of a given diameter, the initial
position of each particle at the inlet (x=0) being chosen randomly. Due to their different
trajectories and charging events, the particles are not all charged and collected in the same
manner. Figure 4 and 5 show some histograms of charge qp for respectively collected and
non collected particles. For the collected fine particles (dp<1 µm), qp ranges in an interval of
noticeable width. When field charging is dominant (dp>1 µm), the histograms exhibit a
narrower peak (Figure 4). For the non collected particles, the shape only weakly depends on
dp (Figure 5).
It therefore appears that, for fine particles which carry a low charge qp, the dispersion in
qp makes the weak efficiency of collection even weaker. The other fact to be noted is that
collected and non collected particles carry charges of very similar value.
Figure 6. Mean charge of collected and non-collected particles as a function of their diameter dp
–
(Vappl=20 kV and Ug=1.3 m/s).

205
The dominant factor which determines the collection or escape of a particle is the
location at which it enters the precipitator (x=0). In particular, nearly all particles passing
through a central zone around the axis of the convective roll are not collected (see the
example for dp=0.5 µm in Figure 3). The mean charge of collected particles of a given size
is slightly greater than that of non collected ones (Figure 6). For the ESP configuration
considered here, the computed mean charge is up to 30 % higher than the expression given
by Cochet [11] which is based on the mean field value Vappl/d (Figure 6). This arises because
the particles have a finite probability to visit the zones where the field and the ionic charge
density take high values.
Due to the assumptions of the present model, the above results have a value more
qualitative than really quantitative. They concern the case of a marked steady secondary
flow without turbulence. A turbulence of limited intensity would result in a turbulent
diffusivity for the particles affecting their trajectories, but the character of expanding helices
should remain; also there should be no drastic change in the characteristics of particle
charge.
References
[1] White H.J., 1963 Industrial electrostatic precipitation (Wesley Publishing Company, Inc.)
[2] Leonard G., Mitchner M. and Self S.A. 1980 Atmospheric environment 14 1289–1299
[3] Meroth A.M., 1997 Numerical Electrohydrodynamics in Electrostatic Precipitators PhD thesis
Karlsruhe University 1
[4] Medlin A.J. 1998 Electrohydrodynamic Modelling of Fine Particule Collection in Electrostatic
Precipitator PhD thesis, University of New South Wales
[5] Egli W., Kogelschatz U. and Sagi C.J. 1996 Proceedings of 6th ICESP, Budapest Hungary 166–
171
[6] Blanchard D., Dumitran L.M. and Atten P. 2001 Electroaerodynamic secondary flow in an
electrostatic precipitator and its influence on transport of small diameter particles Proceedings of
8th ICESP 1 A1–4
[7] Blanchard D. 2001 Collecte des fines particules et caractérisation des couches de poussière dans
un précipitateur électrostatique PhD. Thesis, Université Joseph Fourier Grenoble, France
[8] Dumitran L.M. 2001 Collection des fines particules dans un dépoussiéreur électrostatique PhD.
Thesis, Université Joseph Fourier Grenoble, France
[9] Blanchard D., Dumitran L.M. and Atten P., 2001 J. Electrostatics 51–52, 212–217
[10] Lawless P.A. and Sparks L.E. 1988 IEEE Trans. Ind. Appl., 24 5 922–927
[11] Cochet R. 1961 La physique des forces électrostatiques et leurs applications Coll. Int. 102, CNRS
Paris 331–338

207
The numerical simulation of multi-field wire-plate

electrostatic precipitators
D Brocilo, J S Chang and R D Findlay

McMaster University, Hamilton, Ontario, Canada
Abstract. In the past, the majority of numerical models for the prediction of dust collection
efficiency of multi-field electrostatic precipitators (ESP) were based on the single-field model,
assuming the same partial collection efficiency in each field. However, no comprehensive
numerical or experimental validation has been conducted for this assumption. In this work, a
three-dimensional model that includes: the corona electrode geometry, the multi-field effect,
Poisson’s electric field, ion, neutral and charged dust density was developed, as well as the
diffusion and field charging for a wide range of dust sizes (10-3 to 10-2 µm). The dust collection
efficiency of multi-field ESP was obtained from a cumulative expression of the modified
Deutsche equation with theoretical drift velocity: (a) in each field of ESP, and (b) based on
the first field value. The numerical results show that the assumption of same particle
collection efficiency in each section of a multi-field ESP is not valid for dimensionless diffusion
Reynolds number (Rai) and electric field number (FE) ratio (Rai/FE) greater than 0.001. The
experimental results obtained from a three-field bench-scale ESP, agree well for higher applied
voltages.
1. Introduction
In the past, the majority of numerical models for the prediction of dust collection efficiency
of multi-field electrostatic precipitators (ESP) were based on the first-field model assuming
the same particle migration velocity in each field of ESP [1], i.e. “cumulative model”.
However, no comprehensive numerical or experimental validation has been conducted for
this assumption. In this work, the “unified model” that takes into account interfering between
fields has been developed. The comparison between models has been conducted in order to
determine limitations for the cumulative model assumptions in terms of the dimensionless
diffusion Reynolds number (Rai), Debye number (Db) and electric field number (FE), where
the ion convection reduced the ion distribution in the first field, the space charge effect on
charged particles distribution in consequent fields, and the section length effect on the
particle charging rate. Additionally, the (Rai /FE) ratio was used as a measure of general and
diffusion ion transfer versus ion transfer due to the electric field. The (Db2/FE) ratio was used
as a measure of space charge effect due to ions or charged particles. Here, diffusion Reynolds
number (Rai=ugL/Di) is defined as a product of gas Reynolds number (Re=ugL/v) and ion
Schmidt number (Sci—v/Di). Ion and charged dust Debye numbers (Dbi=L/λDi; λDi=√ (εkT/
Nioe2 ); Dbd=L/λDd; λDd=√ (εkT/Ndoe2) are a ratio between characteristic length (L) and
Debye length (λD).

208
2. Modelling
The second section calculates main gas flow, electric field, ion, neutral dust, and charged
dust distribution profiles in ESP. The coupling among several parameters is established
based on a dimensionless number. The initial value of electric field was obtained from
Laplace’s equation. After that, the initial number of ions (Nio) was estimated from the total
current and electric field on the surface of discharge electrode, followed by, the calculation
of ion density from the ion transport equation containing the diffusion, mobility, convection
and ion sink terms. The space charge effect, due to ions and charged dust, was evaluated
based on the ion and dust Debye number. In case of the space charge due to the ions/
charged dust, the coupling sub-routine is executed. At first, the coupling routine searches
for the initial number of ions for which the electric field on the surface of the electrode is
approximately 2MV/m. After that the ion/neutral/charged-dust and electric field are coupled
in iterative process until the volume average value of electric field and ion density between
iteration steps is less than 0.001 [2]. In case of multi-field, the ion-density/initial-surface-
charge at the front plane of ith-field equals to the values at the back plane of preceding (i-
1)th-field. The initial surface charge of dust particles in the first field is set to zero. Diffusion
and field charging are implemented based on the particle size (10–3 to 102 m) in terms of
Knudsen number (Kn) [2]. In case of the initial particle charge Qo=eNo, the initial charging
time τdo was obtained for diffusion and field charging.
The last section calculates the total collection efficiency of dust particles based either
on the modified Deutsche’s equation (Mode 1) or neutral and charged dust density at an
inlet and outlet section of ESP (Mode 2). Hence, the MESP code Mode 1 requires only the
electric field and ion density distribution, that are available in case of negligible space
charge effect due to the charged dust particles. The modified Deutsche equation is based on
the cumulative expression as follows:
(1)
(2)
where, pi is the local dust particles penetration, rd is the radius of dust particles, µm is the
local particle migration velocity, µd is the particle migration mobility, Q is the dust particle
surface charge, A is the local collection surface area, E is the cross-sectional averaged
electric field, Cm is the Cunningham slip factor , Qg is the gas flow rate, µg is the gas
viscosity, λ is the mean free path of ions, and k is the dust property correction coefficient
(0.5≤ k≤ 1). The present study was based on k=1.
3. Numerical Results and Experimental Validation

Wire-plate type ESPs containing two and three-fields are evaluated. Figure 2 shows a
sketch of a top view of two-field ESP. Collecting plate-to-plate length (D) was fixed to
5cm and wire-to-wire (W) length was changed in the range from 2D to 4D. The parametric
studies for dust particles from all three charging regimes are analysed at (a) voltage levels
of 18 and 20 kV, and (b) gas flow rates of 0.2 and 1 m/s.

209
Figure 1. Block Diagram of MESP code

210
Figure 2. Sketch of a top view of two-field ESP
Figure 3a and 3b show cross-sectional averaged values of ion density <N> and electric
field <E> along the length of the 1st field for various W/D ratios. The total average current
(I) is increasing with increasing W/D ratio (I(W/D=2)=0.152mA; I(W/D=3)=0.22mA; I(W/
D=4)=0.293mA). The ion density <N> of W/D≥3 geometry ratio is slightly affected by ion
convection at the entrance zone of the 1st field. For W/D≥4 geometry, the cross-averaged
electric field is higher at y*>0 than at y*<0 position.
Figure 3. Local volume averaged values of: (a) ion density <N>, and (b) electric field <E>, along the
length of the 1st field for various W/D ratios. (Main gas velocity Ug=1m/s, and applied voltage
V=18 kV).
The field and diffusion charging contributions to the surface charge of 0.002, 0.02, 0.5 and
2 µm dust particles along the length of the 1st-field are shown in Figure 4. The charge
contribution due to the field charging is increasing in the area close to discharge electrodes
due to the higher electric field in vicinity of the electrode. At the discharge electrode
location, the charge contribution due to the field and diffusion charging is decreased due to
the smaller cross-sectional averaged electric field and ion density at that locations. The
field charging of particles of 2 µm in diameter reaches the saturation after y*=0.8. After
which, the surface charge is increasing mainly due to the diffusion charging. However, the
particle charging rate is very small. Figure 5a and 5b show shows total surface charge and
collection efficiency of various dust particles along the length of two-field ESP.

211
Figure 4. Contributions due to the field and diffusion charging methods along the 1st field of the two-
field ESP for W/D=2, Ug=lm/s, and V=18kV.
Figure 5. Total number of elementary charges on the surface of various dust particles and local collection
efficiencies based on “unified model” along the two-field ESP W/D=4, Ug=lm/s, and V=18kV.

212
Figure 6 shows the difference in the prediction of the total particle collection efficiencies
of two-field ESP: (a) at the end of the 2nd field based on the “unified model”, (b) assuming
that collection efficiency of the 2nd field is same as the collection 2nd . efficiency of the 1st
field (“cumulative model”). The assumption of the same collection efficiency may
underestimate the total collection efficiency from 10 to 33% , depending on the W/D ratio
and dust particle size.
Table 1. Underestimation of the collection

efficiency of the “Cumulative” model compared
to the “Unified” model
Figure 6. Total collection efficiency based on
cumulative and unified model
The experimental results, obtained from a three-field ESP [3], agree well with predicted
collection efficiency of 0.5 µm dust particles at higher applied voltages, as shown in the
Figure 7. At voltages close to the corona on-set voltage, the model overestimates the
collection efficiency. The assumption of the uniform current density along the length of
the wire was not valid for low voltages due to the small electrode misalignment.
Figure 7. The partial dust particle penetration at various negative applied voltages for a three-field ESP.
(dpN=0.5 µm is the mean dust particle diameters from the dust size distribution in terms of the particle
number. Ug=0.21 m/s is the mean gas flow velocity, dw=1.5 mm is the diameter of the discharge
electrode.)

213
4. Concluding remarks
Based on the numerical results, the following concluding remarks were obtained:
(a) The ion convection at the entrance zone of the 1st field was observed for W/D≥3
geometry. The (Rai/FE) ratio is 9.7×10–4. This however has an adverse effect on the
various dust particles. It seems that the particles charged mainly by diffusion charging
are affected more than other particles.
(b) For given operating and geometry conditions, space charge effect due to the ions can
be observed. The ion (Dbi2/Fs) ratio increased with increasing W/D ratio. Dbi2/FE ratio
at W/D geometry of 2, 3 and 4 was 5.08×100–2, 7.4×100–2, and 8.3×10–2, respectively.
Space charge effect due to the charged particles was negligible due to the low dust
loading. In case of W/D=2, V=18kV, and dp=0.1 m, the space charge effect can be
observed for dust loading greater than 2×1010[#pt/m3].
(c) Particles greater than 2 µm reach field saturation charge after the first discharge electrode
even for W/D ratio of two.
(d) The assumption of the same collection efficiency may underestimate the total collection
efficiency of two-field ESP from 10 to 33% , depending on the W/D ratio and dust
particle size.
References
[1] Lawless R.A., Yamamoto T., and Otani Y.; “Modeling of Electrostatic Precipitators and Filters”;
Chapter 22 of Handbook of Electrostatic processes; Marcel Dekker Inc, 1995
[2] Brocilo D., Chang J.S., and Findlay R.D.; “Modeling of Electrode Geometry Effect on Dust
Collection Efficiency of Wire-Plate Electrostatic Precipitators”; ICESP VII Proceedings of 8th
International Conference on Electrostatic Precipitation; Volume I, A4–3
[3] Zukeran A., Looy P.C., Chakrabarti A., Berezin A.A, Jayaram S., Cross J.D., Ito T., and Chang
J.S., “Collection Efficiency of Ultrafine Particles by an Electrostatic Precipitator Under DC and
Pulse Operating Modes”, IEEE Transactions on Industry Applications, Vol. 35, No. 5, September/
October 1999, pp. 1184–1191.

215
Numerical modelling of dielectrophoretic effect for

sub-micron particles manipulation
B Malnar, W Balachandran, F Cecelja

Department of Systems Engineering, Brunei University, Uxbridge, Middlesex
UB8 3PH, United Kingdom
Abstract: In this paper, we present our initial design of a CAD tool for the simulation of the
dielectrophoretic (DEP) behaviour of any given polarizable particle in an AC field generated
by microelectrodes of any desired layout. The finite difference method is used to obtain the 3D
complex potential and temperature distributions. From the given electrical field distribution,
the electrohydrodynamic (EHD) force on a suspending liquid and the DEP force on suspended
particles are calculated. For the DEP force calculation, multipolar dielectrophoretic and
electrorotation theory is used to account for higher order multipolar moments. The DEP and
EHD forces have been calculated and the results presented for the system of microelectrodes
that has been used for dielectrophoretic manipulation of particles.
1. Introduction
When a polarizable particle is subjected to a non-uniform AC field, it will experience a
force due to interaction between the field and the induced dipole across the particle [1].
This effect has been termed dielectrophoresis. The general expression for the instantaneous
value of the dielectrophoretic (DEP) force on the dipole is given by [1]:
(1)
where is the induced dipole and is the electrical field. For a spherical particle, the
induced dipole is given by [1]:
(2)
where εm is the permittivity of a suspending medium, r is the radius of the particle and fCM is
the Clausius—Mossotti factor given by:
(3)
where εp* and εm* are the complex permittivities of the particle and the medium, respectively.
Complex permittivities are defined as ε*= ε - jσ/ω, with ε being the permittivity, σ the
conductivity, ω the radial frequency, and The expression for the DEP force on a
spherical particle is obtained from equations (1)–(3) and is given in a time-averaged form
by [5]:

216
(4)
where Re(fCM) denotes the real part of the Clausius—Mossotti factor.
In recent years it has been shown that microelectrodes of suitable layout can be
successfully used to move biological particles utilizing low AC voltages [5]. This holds a
great promise for future microsystem technology and development of integrated systems
for applications in medicine, food industry etc. [12].
As dielectric properties of biological particles are being investigated and more available
[9], [11], a need for a CAD tool that would predict the DEP behaviour of the particles
becomes apparent. In this paper, we present initial design of that tool. The concomitant
program is written in C++ programming language and all of the necessary calculations are
implemented without the use of external tools. The results are stored in data files and
analyzed using MATLAB graphic tools.
2. Theory
A review of forces that act upon particles suspended in a liquid and subjected to a non-
uniform AC field can be found elsewhere [1], [5]. In our work, we first calculate the 3D
potential and temperature distributions. There are two equations to be solved: (i) the complex
Poisson equation is solved and the potential distribution obtained, (ii) the thermal equation
describing the relation between the temperature and the electric field is solved and the
temperature distribution is obtained. These two equations are solved using the finite
difference method.
The potential distribution is described with the complex Poisson equation [6]:
(5)
where ϕ is the complex potential given as ϕ=ϕR+jϕ1, with ϕR being the real part and ϕI the
imaginary part of the potential. From the potential distribution, the electrical field can be
calculated as:
(6)
For the calculation of the temperature distribution, ERMS is obtained as:
(7)
The temperature distribution is required for the calculation of the conductivity and
permittivity gradients, which are needed for the calculation of the thermal
electrohydrodynamic (EHD) force. The temperature distribution is obtained from the
following equation [5]:
(8)
where λ is the thermal conductivity and T is the temperature. Equations (5) and (8) are
coupled because the conductivity and the permittivity of the suspending liquid are functions
of the temperature and the temperature is a function of the electrical field strength. Therefore,
they need to be solved iteratively. After the potential and temperature distributions are
obtained, the DEP and thermal EHD forces are calculated.
The DEP force on particles given by equation (4) is based on the dipole approximation
of the particle. However, although adequate in many cases, this approximation brakes
down when higher moments must be taken into account to achieve higher accuracy of the

217
DEP force calculation [3], [10]. The generalized expressions for the time-averaged dipole,
quadrupole and octopole DEP forces on a spherical particle with radius r are given in [3].
The relation between the fluid velocity profile due to EHD pumping and the force on the
liquid due to its conductivity and permittivity gradients is described by Navier-Stokes’
equation for the low Reynolds number given by [2]:
(9)
where p is the pressure, η is the viscosity, is the fluid velocity, and is the time-
averaged electrical force density. The buoyancy force is neglected in this equation, and
is given by [2]:
(10)
where ρq is the charge density and ρm is the mass density. The last term is the electrostriction
and it can be omitted from calculation of force density [2]. In time averaged form and with
known gradients of permittivity and conductivity of liquid, the force density is given by
the following equation and is frequency dependant (both the magnitude and the direction
of the force):
(11)
Solution of equations (5)–(8) and (10) is implemented in C++ code. Firstly, the program
reads the input file. It is a simple text file which contains dimensions and boundary
conditions for the electrical and thermal problems, dimensions of a mesh for the finite
difference method, dimensions of the electrodes, and frequency and magnitude of the
applied voltage. When all the data are input into the program, the mesh is generated and the
equation (5) is solved using the finite difference method. The system of the equations in the
finite difference method is solved using the point SOR method. Initial values of the
conductivity and the permittivity are calculated assuming the initial temperature of 300 K
in all of the mesh points. After obtaining the potential distribution and RMS of the electrical
field, the temperature distribution is obtained by solving equation (8), again using the
finite difference method. Then the permittivity and the conductivity are updated utilizing
the obtained temperature solution and equation (5) is solved again. Equations (5) and (8)
are solved iteratively until the desired accuracy of the temperature distribution has been
reached. Then the solution of the DEP force and the EHD pumping force are obtained by
solving equations (4) and (11).
3. Analysis of results for castellated microelectrodes
Castellated planar microelectrodes have been successfully used to trap and sort sub-micron
particles utilizing positive and negative dielectrophoresis [13].
Figure 1. Microelectrodes—top view Figure 2. Cell for the calculations

218
We have adopted this layout of the electrodes to exemplify the solution of the equations
stated above. The electrodes are shown in Figure 1, and because of their symmetry, only the
solution for a cell enclosed by the rectangle in Figure 1 has to be obtained. The cell is
shown in Figure 2. In this case, the electrodes are chosen to be infinitely thin. There are
three main regions of the solution in z-direction: the suspending liquid is enclosed by two
layers of glass. The bottom one is a base for microelectrodes fabrication, and the top one is
a cover with a microchannel.
3.1. The potential and electrical field distribution

The potential distribution given by equation (5) can be solved for the entire cell in Figure
2. However, if the conductivity and permittivity of the liquid are sufficiently larger than the
conductivity and permittivity of glass, we can consider only the liquid for the electrical
problem [4]. In that case, the boundaries for the solution are x=0, x = 60, y=0, y=30, z=1000,
and z=1050 µm. Generally, the boundary conditions are different for the real and the
imaginary part of the potential. For the cell in Figure 2, the boundary conditions for the
planes x=0, y=0, y=30, z=1000, and z=1050 µm are ∂ϕR/∂n=∂ϕI/∂n=0, while at the plane
x=60 µm both ϕR and ϕI are equal to zero.
Figure 3. ϕR [V], z=1002 µm Figure 4. |ERMS| [V/m], z=1002 µm
The real part of the potential is shown in Figure 3 and ERMS is shown in Figure 4. The voltage
applied is 10 V, 1 MHz sine voltage. The conductivity and permittivity of the liquid are
temperature dependent and modelled using the following equations [6]:
σ=4.0 (1+0.22 (T—20°C)), [mS] (12)
εr=78.54 [1—(T—25°C)(4.6*10 –8.86*10 (T—25°C))]
–3 –6
(13)
3.2. The temperature distribution

The temperature distribution given by equation (8) is solved for the entire cell in Figure 2.
For the liquid, λ is assumed constant and equals 0.56 Jm–1s–1K–1. For glass, λ is assumed to
be 0.6 Jm–1s–1K–1.
Boundary conditions for planes x=0, x=60, y=0, and y=30 µm are ∂T/∂n=0, while for
planes z=0 and z=1150 µm the temperature is fixed at 300 K. The electrodes can be modelled
to either have a fixed temperature (defined in the input file), or to be transparent for the heat
flux and thus not to be considered in the thermal problem [2]. In this case, the electrodes are
transparent. The temperature distribution is shown in Figures 5 and 6. Figure 5 shows the
temperature in plane z=1002 µm, while Figure 6 shows the temperature distribution in
plane y=20 µm.

219
Figure 6. T[K], y=20 µm

Figure 5. T [K], z=1002 µm
The conductivity and permittivity distributions in plane z=1002 µm are shown in Figures 7 and 8,
respectively.
Figure 7. [S], z=1002 µm Figure 8. [C/Vm], z=1002 µm
3.3. The DEP force

The x and z components of the dipole force on particles given by equation (4) are shown in
Figures 9 and 10, respectively. The particles assumed are latex spheres 557 nm in diameter.
The conductivity of latex spheres can be approximated as σ=2KS/r [13] where K S is the
surface conductance and r is the radius of the sphere. KS is taken to be 2.32 nS. The relative
permittivity of latex spheres is taken to be 2.55 [13].
Figure 9. DEP force, x [N], z=1002 µm Figure 10. DEP force, z [N], z—1002 µm
It can be seen that the DEP force is positive and the particles are attracted towards the tips
of the electrodes where the force is strongest. The quadrupole and octopole forces are
qualitatively similar, but negligible in comparison with the dipole force in this case and
therefore are not shown.
3.4. The EHD force on liquid

The EHD force on the liquid given by equation (11) is shown in Figures 11 and 12. Figure
11 shows the x-component of the force density and Figure 12 shows the z-component in the

220
plane z=1002 µm. With the known distribution of the EHD force density, the fluid velocity
profile can be calculated using equation (9). However, this is not covered in this paper.
Figure 11. EHD force, x [N/m3], z—1002 µm Figure 12. EHD force, z [N/m3], z=1002 µm
4. Conclusion
Modelling of a known microelectrode structure has been presented and the results correspond
well to the published experimental results [2], [13]. Further work is under way to improve
and enhance the CAD tool. Particles need to be introduced to the computer experiment and
all the relevant forces implemented, such as Brownian motion and inter-particle forces.
Also, fluid velocity profile needs to be calculated to obtain the EHD pumping force on
particles.
5. References
[1] Jones T B 1996 Electromechanics of particles (Cambridge University Press)
[2] Green N G, Ramos A, Gonzalez A, Castellanos A, Morgan H 2001 J. Electrostal 53 71–87
[3] Jones T B, Washizu M 1996 J. Electrostal 37 121–134
[4] Green N G, Ramos A, Morgan H 2002 J. Electrostal 56 235–254
[5] Ramos A, Morgan H, Green N G, Castellanos A, 1998 J. Phys. D: Appl. Phys. 31 2338–2353
[6] Schnelle T, Muller T, Gradl G, Shirley S H, Fuhr G 1999 J. Electrostal. 47 121–132
[7] Green N G, Morgan H 1998 J. Phys. D: Appl. Phys. 31 L25-L30
[8] Müller T, Gradl G, howitz S, Shirley S H, Schnelle T, Fuhr G 1999 Biosensors & Bioelectronics.
14 247–256
[9] Bakewell D J, Ermolina I, Morgan H, Milner J, Feldman Y 2000 Biochim. Biophys. Acta 1493
151–158
[10] Schnelle T, Muller T, Fiedler S, Fuhr G 1999 J. Electrostal. 46 13–28
[11] Markx G H, Davey C L 1999 Enzyme & Microbial Tech. 25 161–171
[12] Malyan B, Balachandran W 2001 J. Electrostal 51–52 15–19
[13] Morgan H, Hughes M P, Green N G 1999 BiophysicalJournal 77 516–525

221
Modelling studies of charged particle interactions for a

space application
Karen L Aplin and Vladimir P Tarakanov*
Space Science and Technology Department, Rutherford Appleton Laboratory,
Chilton, Didcot, Oxfordshire OX11 0QX
* High Energy Research Centre, Moscow, Russia
Abstract. An electron emitter neutralise! to prevent spacecraft charging from the use of
positive ion thrusters is under development at the Rutherford Appleton Laboratory. A particle-
in-cell code, KARAT, has been used to simulate laboratory experiments. Laboratory test
results have been successfully obtained with simulations, for model verification. Neutraliser-
ion thruster interactions in space have also been modelled. An ion-propelled spacecraft rapidly
charges to kilovolt potentials, which can be dissipated by an electron-emitting neutraliser. The
effect of background particle concentrations on neutralisation has also been studied.
1. Introduction
Application of new satellite technology for fundamental physics experiments in space
requires precise adjustment of the satellite’s position, making micropropulsion an important
new challenge in spacecraft engineering. One propulsion method is to accelerate heavy
ions such as caesium (Ce), away from the spacecraft by a high electric field, a technique
known as Field Emission Electric Propulsion (FEEP). A consequence of FEEP is that the
loss of ions from the spacecraft results in a large static charge, reducing propulsion efficiency.
The Rutherford Appleton Laboratory (RAL) has been contracted by the European Space
Agency (ESA) to produce an engineering model of a FEEP neutraliser. A concurrent ESA
contract with a different laboratory is funding development of a FEEP thruster, which
precludes prototype and breadboard testing of the ion thruster and neutraliser together
until both have been constructed as engineering models. Modelling studies are required to
assess issues arising from electron-ion interactions for neutraliser design, and are also of use
to simulate ion thrusters and neutralisers in space, where direct measurements are difficult.
The KARAT particle-in-cell (PIC) code is ideal for such applications. It solves Maxwell’s
equations and equations of motion for axisymmetric 2-D geometries. KARAT has proved
successful in microwave work, and has been used to simulate astrophysical plasmas [1]
and atomic cluster-plasma interactions [2]. In this paper KARAT is used to model the
laboratory test set-up, to increase both understanding of the test results and confidence in
the model output for the space application. Neutraliser function in the space environment
has also been simulated.
2. The KARAT model

KARAT is an electromagnetic PIC code. In this paper the 2-D version for an axisymmetric
system (including three electromagnetic field and momentum components) has been used.
The code also includes inelastic processes for particle-gas reactions. Geometry and particle
characteristics such as mass, energy and charge are defined by the user. The model assumes
that all objects have homogeneous conductivity [3]. For this application, a Ce+ thruster

222
emitting 10 keV ions has been assumed throughout. Particles are emitted from pre-defined
parts of the spacecraft surface and have a Dirac-delta energy distribution (all particles have
the same energy). Particles are shown on the figures as dots representing the PIC-particles
used in model calculations. The ratio between the number of model particles and displayed
particles is defined by a merging factor, M; when M=1 the ratio is 3×109 (M is typically
0.001 in this paper).
3. Spacecraft charging by an ion thruster

A simple model of a spacecraft propelled by a FEEP thruster was set up to observe the
electrical effects of the thruster. The spacecraft is modelled as cylindrical with a radius of
2.5m and a depth of 2m. The boundaries are much greater than the satellite dimensions and
are set at ground potential (as is the spacecraft) at the start of the simulation, Figure 1. The
thruster is assumed to be a uniform Ce+ beam, as described in section 0 above.
Figure 1 Geometric set up of model showing cylindrical satellite in centre. Ions are emitted from the right
hand surface of the shaded region, with the arrow showing the direction of the ion beam.
The selfconsistent KARAT solution of Maxwell’s equations and particle motion show
that the spacecraft charges to potentials of kV in times of order µs (with a typical ion current
of ~10mA), causing back-streaming of the ion beam, Figure 2a. This reduces the effectiveness
of the ion thruster, by peturbing the potential well propelling the ions away from the
spacecraft.
Figure 2 a) Equilibrium ion beam trajectory without a neutraliser. Ions are shown as dots, b) Equilibrium
trajectory of ion beam with neutraliser positioned at r=2.5m, z=10m.

223
Inclusion of a neutralise! emitting a few mA restores the ion beam trajectory, Figure 2b.
In this scale of model the field emitter neutraliser is simply defined as a surface emitting
100 eV electrons. The RAL specification is to neutralise an ion current of 6mA from four
FEEP thrusters. KARAT runs have shown that only partial neutralisation occurs if the
neutraliser current is less than the ion current. If the neutraliser current is equal to the ion
current, neutralisation occurs in microseconds. Although the process can be accelerated if
there is an excess of electrons, the RAL neutraliser is designed to emit 6mA, the same as the
thruster.
The effect of neutraliser location in this model geometry was found to be minimal. However
a key assumption is the homogeneity of conductivity of objects represented in the model.
For an application with varying conductivity, such as a real satellite, it is recommended
that the neutraliser and thruster are as close together as possible. This simple example
demonstrates the necessity for a neutraliser.
4. Laboratory Simulations
4.1 Electron emission technology
Field emission of electrons from silicon nanotips is the core technology used in the neutraliser
project. Space is a relatively new application for field emission, but RAL has already
developed a miniaturised mass spectrometer using field emission for the ROSETTA mission
[4]. RAL also has in-house construction capability at the Central Microstructures Facility,
who fabricate the micromachined devices [5].
Figure 3 (a) Left- array of single crystal silicon field emitters with integrated gates produced at RAL -
CMF. (b) Right—magnified SEM single emitter tip (tip radius <10nm). The aperture shown is 2µm
diameter
Electrons are emitted from the tips (Figure 3a) (held at ground potential) if a positive
voltage of up to 200V is applied to the gate structure (Figure 3a,b) to accelerate the
electrons away from the emitting surface. In space, the electrons are attracted towards the
positive charge of the ion thruster, but in laboratory tests there is no ion thruster so a
Faraday cup at a positive potential of ~300V is used to collect the emitted electrons.
Laboratory measurements take place in a cylindrical vacuum tank at a pressure of 10"6 torr.
Electrons are emitted from a gated field emitter about 1cm away from the Faraday cup.

224
4.2 Modelling the laboratory test environment

The laboratory environment model is on a smaller scale than the satellite model discussed
in Section 3. The gate is represented in KARAT by a fine positively charged grid close to
the field emitters. Electrons are emitted with energy of 5eV, (comparable to the work
function of the material) from a small surface to simulate field emission. 100V is applied to
the gate to accelerate the electrons. The geometry of the cylindrical vacuum tank is readily
established, and currents at the electrodes (i.e. gate and Faraday cup), and particle trajectories
were monitored.
One experiment decreased the Faraday cup (collector) potential to zero to investigate
the effects of running the neutraliser without the ion beam switched on, which is the
proposed initial switch on operation mode of the final neutraliser. The emitted electrons
moved to the gate in preference to the collector. The experiment was simulated in KARAT,
and the modelled results concurred with experimental observations of the gate current
increasing and collected current decreasing as the collector voltage decreased. In both
model and experiment, no current reached the collector when it was at zero potential.
The confidence in the model gained from the general agreement between experiment
and simulation allows KARAT to be used for more detailed study of electron motion.
Modelled electron trajectories gave further physical insight. With a small collector potential,
self-repulsion very quickly prevents any more electrons from reaching the cathode, and an
abrupt change of direction can be seen. At zero collector potential, all the electrons oscillate
between the field emitter and gate.
5. Space Environment Simulations

The successful simulation of laboratory experiments gives credence to model runs based
on the space environment, which are less easily tested. As the neutraliser lifetime may be
affected by bombardment from ions and neutrals, it is important to simulate the effects of
background particle concentrations. Although an electromagnetic model cannot simulate
the physical effects of the neutrals on the delicate field emitter array, KARAT can be used
to predict the location and concentration of the neutral particles. KARAT can also simulate
effects of the background plasma on neutralisation efficiency by modelling interactions
between the charged particles emitted by the thruster and neutraliser.
The two most probable orbits for future neutraliser applications are low earth orbit
(LEO) and geosynchronous orbit (GEO). LEO is within the thermosphere, or ionosphere,
where atomic oxygen is the dominant constituent [6]. GEO is the height at which the
satellite orbits in phase with the earth and makes one rotation every 24 hours.
Table 1 Background particle concentrations in possible neutraliser orbits (from [6] and [7])
The first use of the neutraliser on the ESA SMART-2 satellite, scheduled for launch in
2005, is likely to be in GEO or GEO-like orbit. However the worst case for potential
degradation of the neutraliser is LEO, from the effects of atomic oxygen bombardment. As
future missions may also be in LEO, there is a need to quantify effects of background
particle concentrations in different orbits. It is possible to use the parameters in Table 1 as

225
input plasma and neutral concentrations in KARAT. Inelastic processes are modelled using
a Monte-Carlo method, with one species of neutral particles included; cross-sections of
important ionisation processes are also parameterized [3].
5.1 Low Earth Orbit

In the KARAT simulation the number of particle species was restricted to four, so it is not
possible to directly simulate LEO where five species of neutral and charged particles are
under consideration. The problem is therefore broken down into two cases a) background
and satellite-emitted charged particle interactions b) anthropogenic charged particles and
ambient neutral particles.
a) Charged particle interactions

An orbit height of 500km in the middle of the LEO range was chosen; the background
plasma consists principally of 104–105 cm”3 each of electrons and UV-ionised oxygen
(with a smaller fraction of nitrogen ions, not included in this simulation) [8]. In KARAT,
positive ions were set to have a relative atomic mass of 11. Neutralisation is minimally
affected by the background charge-neutral plasma. However, as the charged particle density
is sensitive to the time of day and solar cycle [8] further simulations are required. Advection
of plasma is also significant in the upper ionosphere [6] and should be included. In
preliminary KARAT runs without a plasma source term, the background plasma self-
neutralised in 20µs, indicating that ionisation and advection processes are not yet adequately
represented.
b) Neutral particle interactions

Neutralisation efficiency is unperturbed by neutral O atoms, though the process is tens of
microseconds slower than in a true vacuum. Some ionisation occurs in the model: after 12.5
µs, a pocket of positive charge developed outside the usual beam path with 108 particles
cm”1 compared to 1010 cm”1 in the beam path. These are likely to have been formed by a
well-known ionisation process between an electron and neutral species [9], producing a
positive ion and two electrons.
5.2 Geosynchronous orbit

GEO was simulated with charged particle concentrations as in Table 1; neutral particles
were not included as concentrations are negligible and poorly defined. The background
electrons migrated towards the ion beam but their peak concentration, located in the densest
part of the ion beam was 106 cm-1 compared to 108 cm-1 for the neutraliser electrons at the
same position. Therefore background electrons contribute a maximum of 1% to the ion
beam neutralisation.
6. Conclusions
KARAT can effectively reproduce spacecraft charging with a FEEP thruster. Use of a 100eV
electron beam as a simulated neutraliser discharges the spacecraft’s negative charge;
1
The model particles move a factor of faster than a particle with a relative atomic mass of M due to
their increased mobility. This speeds up the model runs, and is physically equivalent to using more massive
particles in longer model runs.

226
neutraliser position is unimportant for a uniformly conducting surface. The neutraliser

should match the ion current of 6 mA. Laboratory results were successfully modelled on a
smaller-scale simulation including a positive grid to accelerate electrons away from a low-
energy electron source. In the absence of the ion thruster, electrons are expected to return to
the acceleration electrode (gate)—a necessary result, as the neutraliser will be switched on
before the ion thruster on the satellite. Neutralisers should therefore be built to withstand
the full electron emission current returning to the gate.
Effects of two different orbit conditions on the neutraliser’s electrical behaviour have
been considered. Ambient neutral particles in LEO did not affect neutraliser operation, but
recombination hindered plasma modelling. A source term needs to be included to assess
the effect of background charged particles. GEO background plasma contributed a maximum
of 1% to neutralisation. These preliminary model runs suggest a neutraliser is necessary
and will function effectively in both its prospective orbits.
7. Further Work
It has been suggested that in the lower LEO range the plasma concentrations are so high
that a neutraliser is not needed. Since background electrons do contribute slightly to
neutralisation in GEO, there is a need to extend this study and assess the implications for
neutraliser deployment. To do this, particle source and advection terms need to be defined
to improve these plasma simulations.
Further ion thruster physics is required for a more realistic simulation. Thruster
development work suggests that charged and neutral Ce atoms and clusters are emitted
from the ion thruster, and reactions involving these particles are likely to have the most
significant implications for the neutraliser. One way to improve the simulation of ion
generation from the currently assumed degenerate, uniform ion beam would be to set up a
detailed geometry of the ion thruster. Output of this model such as ion energy and spatial
distribution will then be used to input to larger scale simulations.
Acknowledgement
This project is funded by the European Space Agency.
References
[1] Khodataev Y K, Bingham R, Tarakanov V P and Tsytovich V N, Plasma Phys Rep 22, 11, 932–
942(1996)
[2] Eloy M, Azambuja R, Mendonca J T and Bingham R Phys of Plasmas, 8, 3, 1084 (2001)
[3] Tarakanov V P, User’s Manual for Code KARAT v 8.05, Moscow, Russia (2002)
[4] Kent B J, Huq E, Dominey J N and Morse A D, The use of microfabricated field emitter arrays
in a high precision mass spectrometer for the Rosetta mission, Presented at the 3rd Round Table
on Micro/Nano Technologies for Space, ESTEC, The Netherlands (2000)
[5] Huq S E, Kent B J, Stevens R, Lawes R A, Xu N S and She J C, J. Vac. Sci. Tech B. 19, 3, 988–
991 (2001)
[6] Tribble A C, The space environment: implications for spacecraft design, Princeton University
Press, Princeton, New Jersey (1995)
[7] Wynn-Williams G, The fullness of space, Cambridge University Press, Cambridge (1992)
[8] MacGorman D R and Rust W D The electrical nature of storms, Oxford University Press, New
York (1998)
[9] Smirnov B M, Physics of ionized gases, Wiley, New York (2001)

227
Analytical solutions of surface potential distribution

on thin insulators having grounded backing
conductor and their applications to electrostatic
characterisation
A Ohsawa†1 and M Ohuchi‡
† Physical Engineering Safety Research Group, National Institute of Industrial

Safety, 1–4–6 Umezono, Kiyose, Tokyo 204–0024, Japan
‡ Department of Electronic Engineering, Tokyo Denki University, 2–2
Kandanishiki, Chiyoda, Tokyo 101–8457, Japan
Abstract. Analytical solutions for surface potential on thin insulators having a grounded
backing conductor are obtained by using the model of a distributed resistor-capacitor network.
In this paper, we present the steady-state and transient solutions for disk and rectangular thin
insulators, and apply the solutions to determine the resistance path to ground and charge
relaxation.
1. Introduction
Thin insulators having a grounded backing conductor are often found in industry, such as
floors, sheets on desks or stocking shelves, liners and paints on metal surfaces, etc. The
static charges of the insulators themselves and something on them (e.g. a human body on a
floor) introduce electrostatic hazards and troubles in many fields of manufacturing industries.
To prevent such hazards and troubles, many types of antistatic goods have been developed.
However, there are sometimes problems that some of them are not effective for reducing the
charges. For example materials, of which only surface resistance is as low as specified for
the antistatic, can not reduce the charge [1]. To evaluate antistatic materials we usually
measure their resistances according to standard methods and sometimes charge decay time
is additionally measured. Analytical solutions of resistance and charge decay, therefore,
may be useful to discuss the measured results, and to evaluate and design antistatic materials.
Some analytical calculations have been carried out using a one dimensional model [2, 3]
and in vessels with insulating liners [4].
1
Email: ohsawa@anken.go.jp

228
Figure 1. Equivalent circuit models, (a) equivalent network for disk insulators having a grounded
backing conductor and (b) small segment of equivalent circuit for rectangular insulators on a grounded
conductor.
In this paper, we will solve analytically surface potential distributions and their relaxation
for disk and rectangular thin insulators using distributed resistor-capacitor networks,
although comprehensive dimensionless numerical analyses for square insulators were
presented [5]. Analytical solutions are relatively easy for implementation by using a suitable
mathematical software rather than numerical analysis. The solutions will be applied to
determine the resistance path to ground and charge relaxation that characterise the
electrostatic properties of the insulators.
2. Theory
2.1. The model for disk insulators

2.1.1. Transient solution for disk insulators The model, Ohsawa [5] has developed, is
modified for disk insulators having a grounded backing conductor. Assuming that the
radius of the disk, b. is much greater than the thickness δ, and the surface current at the
edge (r=b) is zero, we can apply the equivalent circuit shown in figure 1(a). Using Kirchhoff
‘s voltage and current laws for the small segment of the equivalent circuit, the equations of
the surface potential ˜(r, t) and the surface current i(r,t) at radius r and time t are derived as
follows.
(1)
(2)
where
(3)

229
Here Rs is the surface resistance per unit length at r, C is the capacitance per unit length at
r, and Gv and Rv are the volume conductance and resistance per unit length at r respectively.
ρs and ρ˜ are the surface and volume resistivities and ε is the permittivity of the disk
insulator. Substituting the derivative of equation (1) into (2) yields the partial differential
equation for the surface potential,
(4)
or
(5)
The boundary condition that the surface current at edge r=b is zero, i(b,t)=0, gives the
boundary condition for the surface potential,
(6)
Letting
(7)
yields,
(8)
Using separation of variables under the boundary condition of equation (6), u(r, t) is given
by
(9)
Hence the transient solution is obtained,
(10)
or
(11)
Here using the orthogonal series of Bessel functions [6].
(12)
where f(r)=υ(r, 0), is the initial condition of the surface potential distribution, J0 and J1 is
Bessel functions of the first kind and αnb is the nth roots of J1. These solutions correspond
to the charge relaxation for the disk insulators. For example, when the initial condition is
υ(r, 0)=V at 0≤r≤b, A0=V and A1=0. Then the transient solution becomes.
(13)
Also when υ(r, 0)=V at 0≤r≤a and υ(r, 0)=0 at a<r≤b, the transient solution is expressed by
(14)

230
2.1.2. Steady-state solution for disk insulators Substitution of the time derivative ∂/∂t=0
into equation (5) gives the equation for steady state
(15)
Since a cylindrical electrode is usually used in the standard measurement of resistance to

ground, we use boundary condition, υ(r)=V at 0≤r≤a, where a(a<b) is radius of the electrode.
Using this condition involving ∂v/∂r|r=b=0, we obtain
(16)
where, I0 and I1 are modified Bessel functions of the first kind and K0 and K1 are modified
Bessel functions of the second kind.
2.2. The model for rectangular insulators

2.2.1. Transient solution for rectangular insulators The small segment of the equivalent
circuit for rectangular insulators is shown in figure 1(b). This model is the same in numerical
analyses [5]. The partial differential equation for the surface potential on rectangular
insulators, υ(x, y, t), is expressed by,
(17)
or,
(18)
where,
(19)
Here, Rs is the surface resistance per unit surface area, Gυ and Rυ are the volume conductance
and resistance per unit surface area, and C is the capacitance per unit surface area. Here we
use coordinates that the centre of the rectangular insulators is the origin point and side
lengths are 2a and 2b in direction x and y, respectively. Using the similar analyses used for
the disk insulators in section 2.1.1 and the boundary conditions of ∂v/∂x|x=”a,a=0 and ∂v/
∂y|y=-b.b=0 which were derived from the perpendicular surface current of zero at the edge, we
obtain the transient solution for rectangular insulators,
(20)
or
(21)
Using Fourier series,
(22)
where, f(x, y)=υ(x, y, Q) is the initial surface potential distribution at time t=0.

231
2.2.2. Steady-state solution for rectangular insulators We consider one quarter section of
0≤x≤ a and 0≤y≤ b due to symmetry and separate it into 4 regions shown in the parentheses
in equation (23). To simplify the calculation of the potential, a cube electrode of side
length 2c is used instead of the cylindrical electrode. Substitution of ∂/∂t=0 into equation
(17) yields the equation for the steady-state surface potential, υ(x,y), on rectangular insulators.
Solving the equation under the boundary conditions of the electrode, and ∂v/∂x|x=a=0 and
∂v/∂dy|y=b=0, we obtain,
23
This is an approximate solution, while it almost agrees with numerical results [5].
3. Applications of the solutions

The model may be applied to not only homogeneous materials but also composite materials,
when their effective resistivities and permittivity can be found. This has been demonstrated
by comparing with experiments [5].
3.1. Resistance path to ground

First we show the resistance to ground for the disk insulators. When a standard electrode is
used, the measured resistance consists of parallel resistances corresponding to right under
the cylindrical electrode, Re=ρυδ/(πa2), and a ring insulator between r=a and b, Rr.
Substituting equation (16) into (1), we obtain the surface current for steady state,
(24)
Applying Ohm’s law, the ring resistance is expressed by
(25)
Hence the resistance path to ground, R, is obtained by

R=ReRr/(Re+Rr). (26)
For the rectangular insulators, the total current toward the backing, it, is expressed by
(27)
where Gυ and υ(x, y) use equations (19) and (23). Hence the resistance path to ground for
the rectangular insulators, R, is
R=V/it. (28)

232
Figure 2. Charge relaxation of disk insulators, (a) An example of relaxation of the surface potential
distribution of a disk insulator, (b) Decays of the surface potentials at the centre of disk insulators having
different ρs/(ρυδ).
3.2. Charge relaxation

Using equation (14) the relaxation of the surface potential distribution on a disk insulator
for a=0.1 m, b=1 m, δ=5 mm, ρs=109 0 Ω and ρυ=108 Ωm is shown in figure 2(a). Here we use
an initial condition, υ(r, 0)=V at 0≤r≤a and υ(r, 0)=0 at a<r≤b. The resistance to ground for
these conditions is 1.06×107 Ωusing equation (26). Figure 2(b) shows the decays of the
potentials at the centre of disk insulators for a=0.1 m, b=1 m, δ=5 mm, and different values
of ρs/(ρυδ).
4. Conclusion
We analytically obtained the steady-state and transient solutions for the surface potential
on disk and rectangular insulators having a grounded backing conductor. They were applied
to determine the resistance to ground and charge relaxation. We showed that both the
resistance and the charge decay greatly depend on ρs/(ρυδ). Since the solutions for the
insulators are expressed by the functions of their surface and volume resistivities, permittivity
and dimensions, and further the calculation is relatively easily given by any mathematical
software rather than numerical analysis, the solutions may be useful for evaluating and
designing antistatic goods.
References
[1] Ono H, Ohsawa A and Tabata Y 2003 J. Electrostal. 57 355
[2] Zahn M 1979 Electromagnetic Field Theory: a problem solving approach (Malabar: Robert E.
Krieger Pub.) p 189
[3] Slowiriski Z 1979 J. Electrostat. 8 59
[4] Jones T B and Chan S 1989 J. Electrostal. 22 199
[5] Ohsawa A 2001 J. Electrostal. 51–52 625
[6] Spiegel M R 1968 Mathematical Handbook of Formulas and Tables (New York: McGraw-Hill)
p 144

233
Towards an improvement of thermal modelling and

mathematical deconvolution in FLIMM
D Marty-Dessus, A Petre, L Berquez and J L Franceschi

Laboratoire de Genie Electrique, Universite Paul Sabatier—CNRS UMR 5003
31062 TOULOUSE Cedex, FRANCE
Abstract. The well-known Laser Intensity Modulation Method (LIMM) allows the
determination of experimental space charge and polarization (“charge function”) profiles in
thin (<100µm) polymers by the detection of a periodic thermally induced current.
An intensity and frequency modulated laser diode is used as a thermal source to illuminate
the sample to be studied. The periodical and local low and large expansion induce a relative
charge displacement within the irradiated volume. The resulting shortcircuit pyroelectric
current can then be detected between the electrodes and processed in order to rebuild the
trapped space charge and polarization. Varying the laser beam modulation frequency, one
can control the thermal wave diffusion length and then emphasize the contribution of the
corresponding depth into the sample to the total pyroelectric signal. At the same time, a bi-
dimensional scanning of the beam and appropriate modelling of the temperature give a
quantitative estimation and lead to a spatial determination of the charge function. We have
developed an evolution of LIMM, called FLIMM (Focused LIMM) in which the laser beam
is focused on the sample in order to obtain a spatial cartography of the space charge or
polarization. In this study, we propose to show the ability of this technique to investigate
charge functions, with a discussion concerning the effect of a three-dimensional modelling
of the thermal gradient, and a comparison of different mathematical deconvolution
procedures used to solve “ill-posed” problems generally associated with the FLIMM
technique.
1. FLIMM method
Based on the LIMM technique [1–4], FLIMM uses an intensity and frequency modulated
laser diode (20mW, λ=780nm) which can be focused on the sample to be studied. The spot
size can be set from lµm to several millimetres. The sample is prepared, by depositing a thin
metal layer (typically 50nm of gold) on its faces constituting two electrodes. It is possible
to increase the photothermal conversion by adding some very thin absorbing layers as
bismuth or carbon (20nm). The working frequencies range between IHz to 100kHz.
As a consequence of the created thermal gradient, the induced periodical and local low
and large expansions cause a relative charge displacement within the irradiated volume.
The lower the frequency, the deeper the irradiated zone. Varying the laser beam modulation
frequency, one can control the diffusion length of the thermal waves and then emphasize
the contribution of the corresponding depth into the sample to the total current signal I(f).
In short-circuit conditions, the expression is given by [5]:

234
(1)
where A, L, p(z), E(z), αx, αε are the surface area of metallic electrodes, the sample thickness,
the pyrolectric coefficient in the direction of the sample thickness Z, the electric internal
field, the thermal dilatation coefficient of material and the temperature dependence
coefficient of the material permittivity respectively. T(z,f) describes the variations of the
thermal gradient with respect to frequency and in the direction of the sample thickness Z.
Because of its weakness (several pA typically), this current must be electronically
wellconditioned by a low-noise wide-bandwidth current-to-voltage converter. The signal
is extracted from noise by means of a lock-in amplifier and recorded via a PC. Finally, a
mathematical treatment allows the reconstruction of the polarization or charge profiles in
the direction of the sample thickness.
2. One and three-dimensional thermal gradient modelling

2.1. One-dimensional modelling
For a spot size beam larger than the thickness of the sample, and in our experimental
conditions, the thermal gradient T(z, f) may be expressed as [6]:
(2)
where j0(Wm-2) is the power beam density, k(Wm-1K-1) the thermal conductivity, η the
absorbance of the electrodes, L(m) the sample thickness and γ the complex wave vector
number.
2.2. Three-dimensional modelling

During the FLIMM procedure, the beam has to be focused at the sample surface. In this case,
the one-dimensional modelling of the alternative temperature gradient is no longer valid.
A three-dimensional approach was then considered (figure 1) [7]. The periodic and volume
heating is supplied by the modulated laser diode with a given penetration depth of the
luminous flow within the sample.
Figure 1. Geometrical problem Figure 2. 3D deconvolution procedure
One carries out a double Fourier transform of the equation of heat diffusion in the two
directions X and Y. The significant simplification resulting from this allows a resolution in
the transformed plane by using Green functions and with the assumptions used, to a simple

235
expression of heat in the Fourier domain. The return in the real plane is performed using a
numerical computation (figure 2).
3. Deconvolution procedure
Knowing the experimental pyrolectric current and the variations in the temperature shapes,
one can build space charge or polarization profiles by using a suitable mathematical
deconvolution.
The FLIMM fundamental expression (1) comprises a Fredholm intregral of the first kind
where r(x) is the unknown function. The kernel g(x, f) is a set of data resulting from the
thermal modelling and I(f) the experimental results. The mathematical treatments of such
equations are generally called ‘ill-posed problems’ because of the extreme sensibility of
the solutions due to external perturbations. Then, their resolution can lead to an infinite set
of solutions within the experimental error domain, or to very large mathematical instabilities.
It is particularly the case in our experimental problem, where the signal-to-noise ratio is
very low.
In order to overcome this problem, many authors [8–12] have already proposed
some mathematical targeted methods. In the considered case, the more suited ones are
called approximation and regularization techniques.

In this part, the influence of the thermal modelling is first enhanced. Then, the influence of
the different mathematical deconvolution procedures used are discussed.
4.1. 1D/3D Temperature comparison

The sample under test is PE, 25µm thick (L), coated by two 50nm gold electrodes. It was
poled under a DC field of 60kV/mm for seven days. The pyroelectric currents were measured
and recorded from the two sides of the sample (z=0 and z=L) (figure 3). The space charge
densities were calculated by a regularization technique, and are plotted on figure 4.
Figure 3. Pyroelectric currents vs. frequency Figure 4. Space charge profiles comparison
The main differences enhanced are located near the surface of the sample (for and
correspond to the highest frequencies used (>1kHz). Actually, two different approaches are
considered: a one-dimensional modelling associated with a surface way of heating the
sample (using suitable boundary conditions), and three-dimensional considerations in the
case of volume heat generation. Comparing in both cases the temperature variations for

236
different frequencies, with respect to the thickness Z (figures 5 and 6), one can notice in
particular that the different values given by the first derivative are high for the
curves corresponding to the 1D modelling, and near zero for the 3D one. These observations
could thus explain the differences observed figure 4.
Figure 5. 1D temperature modelling Figure 6. 3D temperature modelling
4.2. Comparison between approximation and Tikhonov regularization methods

In order to obtain a satisfactory in-depth resolution, the approximation method requires the
measurements of currents to be recorded from both sides of the sample. On the other hand,
the regularization should be powerful enough to give a profile through the whole sample,
which makes it possible to reduce significantly the time of measurement.
Polarization Thickness (µm) Number of Sheets Configuration

P positive 25 Sample A two P-N
N negative 25 Sample B three N-P-N
U unpoled 44 Sample C three U-P-U
Table 1. PVDF sheets Table 2. Configurations of samples under study
Figure 7. Regularization from the two sides Figure 8. Approx. and regularization results
Results obtained with sample A show that inversions with the regularization method
must be performed from the two faces, because of the lack of resolution noted after
approximately the middle of the sample (figure 7). However, even if the two methods seem
to give similar results with a simple sample (figure 8), this is not the case for more complicated
structures (sample B and C, figures 9 and 10).

237
Figure 9 shows the polarization of sample C. The first layer of the sample is a 40µm α-
PVDF sheet, whose polarization is theoretically zero. This result is highlighted by the
approximation method, but not with the regularization one which seems to take into account
by anticipation the following layer by smoothing the solution, which could explain the
shift of the curve. Concerning the middle sheet (poled PVDF), the regularization gives a
quantitative response, whereas the approximation gives just a qualitative one. The fall of
the polarization level could also be explained by a bad transmission of the thermal wave
from one medium to another, due to the coupling material used (1µm Pentadecane). This
has not been taken into account in our thermal models, and further developments are in
progress to describe more precisely the behaviour of thermal waves in multi-layered samples.
Figure 9. Sample C Figure 10. Sample B
Similar results are obtained with sample B (N-P-N structure, figure 10) and lead to the
identical conclusions as in the previous case.
4.3. Comparison between different regularizations methods

Different regularization methods have been used and compared, in particular TSVD
(Truncated Singular Value Decomposition), PPTSVD (Piecewise Polynomials SVD) and
Tikhonov ones.
Figures 11 and 12 show that the PP-TSVD method gives a global idea of the distribution
of polarization in the sample, but could mask some particular effects or behaviours. On the
other hand, the theoretical level of polarization is reached in this case, which constitutes an
interesting point, in particular during a calibration process. This method should thus be
used successfully as an informative and complementary technique to others.
Results with the TSVD inversion are very close to that given by the Tikhonov
regularization one (Fig 11). In our following studies, this last one will be preferred because
the way of choosing the regularization parameter (L-curve or Self-Consistency methods
[13– 15]) seems to be more reliable than in TVSD.

238
Figure 12. Sample C Figure 11. Sample A
5. Conclusion
During the FLIMM procedure, great care must be taken in modelling the periodical
temperature variations inside the sample. This must be done more precisely by taking into
account the focusing of the beam spot and the more or less complex geometries under
study. W are now trying to develop some specific thermal models, more suitable for multi-
layered structures.
Dealing with the mathematical inversion, the best results were obtained using both
approximation and Tikhonov regularization methods. These are complementary in terms
of in depth-resolution, and they will be preferred for further developments.
6. References
[1] Das-Gupta D K and Hornsby J S, 1990, J. Phys. D : Appl. Phys., 23, 1485–1490
[2] Lang S B, 1991, Ferroelectrics, 118, 343–361
[3] Das-Gupta D K, Hornsby J S, Young G M and Sessler G M, 1996, J. Phys. D : Appl. Phys., 29,
3113–3116
[4] Lang S B, 1998, IEEE Trans. Dielec. Electr. InsuL, 5(1), 70–76
[5] Bloss P and Schafer H, 1994, Rev. Set. Instrum., 65, 1541–1550
[6] Bauer S and Ploss B, 1991, Sensors and Actuators A, 25–27, 417–421
[7] Marty Dessus, Berquez L, Petre A, Franceschi J L, 2002, J. Phys. D : Appl. Phys., 35, 3249–
3256
[8] Franceschi J L, Biellmann C, Berquez L and Marty-Dessus D, 2001, Jpn. J. Appl. Phys., 40, 888–
890
[9] Ploss B, Emmerich R and Bauer S, 1992, J. Appl. Phys., 72(11), 5363–5370
[10] Ploss B, Emmerich R and Bauer S, 1992, J. Appl. Phys., 51, 1135–1141
[11] Lang S B, 2001, Integrated Ferroelectrics, 38, 111–118
[12] Hansen C, 1996, Num. Lin. Alg. with Appl., 3(6), 513–524
[13] Hansen C, 1992, Inverse Problems, 8, 849–872
[14] Hansen C, O’Leary D P, 1993, SIAM J. Sci. Comput., 14(6), 1487–1503
[15] Honerkamp J, Weese J, 1990, Continuum Mech. Thermodyn. 2, 17–30

239
Electrostatic testing of ESD-protective clothing for

electronics industry
J Paasi1, S Nurmi1, T Kalliohaka1, G Coletti2, F Guastavino2, L Fast3,

A Nilsson3, P Lemaire4, J Laperre4, C Vogel5, J Haase5, T Peltoniemi6,
G Reina7, A Börjesson8, J Smallwood9
1
2
Dept. of Electrical Engineering, University of Geneva, 1–16145 Geneva, Italy
3
SP—Electronics, PO Box 857, SE-50115 Boras, Sweden
4
Centexbel, av. du parc 38, B-4650 Herve, Belgium
5
STFI e.V., PO Box 1325, D-09125 Chemnitz, Germany
6
Nokia Oyj, PO Box 319, FIN-90651 Oulu, Finland
7
Celestica Italia, via lecco 61,1–20059 Vimercate (MI), Italy
8
Agb-konsult, Ban-grand 19, SE-50467 Boras, Sweden
9
Electrostatic Solutions Ltd., Bassett, Southampton, SO16 7BQ, UK
Abstract. Current standard test methods do not adequately evaluate the performance of
modern electrostatic discharge (ESD) protective garments used to protect ESD sensitive
devices during handling in electronics industry. A new European research project—ESTAT-
Garments—aims to supply the standards body IEC TC101 with a means to assess the
effectiveness of ESD garments and to develop appropriate test methods.
This paper reviews the requirements for ESD garment test and presents results of measurements
of charge transferred and peak ESD current in direct electrostatic discharges obtained from
charged fabrics. Results obtained using triboelectrification and direct contact charging methods
showed no significant differences. Fabrics could be placed in the following order of increasing
charge and peak current corresponding to decreasing surface resistance: carbon core fibre,
carbon surface conducting fibre and stainless steel conductive fibre.
1. Introduction
The evolution of electronics has been made possible by continual reduction of the size of
semiconductor devices. Unfortunately, this size reduction leads to inherent increase in
sensitivity of the components to electrostatic discharge (ESD) damage. Investigations
performed in different parts of the world, see e.g. [1] have shown that about 30–50 % of
all failures in electronic products detected during manufacturing can be attributed to some
kind of electrical overstress, of which ESD is the most important type.
ESD events are highly variable. The waveform characteristics—rise time, peak current,
and duration—are strongly influenced by the electrical characteristics, geometry and
dimensions of the materials in the discharge circuit, the level of initial charge, and the
speed of approach of the contacting “electrodes” [2]. The practical need for assessment of

240
ESD sensitivity of components using standardised electrostatic discharges has led to the
development of three ESD models. Human Body Model (HBM) simulates an ESD from a
charged human body to a device. Machine Model (MM) simulates an ESD from a charged,
floating metal object to a device. Charged Device Model (CDM) simulates the discharge of
charge accumulated on the device itself, to a grounded conductive object. All these models
are based on a capacitor charged to a certain ESD voltage, which is discharged into the
device to simulate the ESD event. Sensitivity of electronic components is given as an ESD
withstand voltage, defined as the largest ESD test capacitor voltage that the device can
stand without damage. [3]
Device manufacturers commonly include on-chip protection networks in 1C designs,
which can often withstand ESD in the kV-range. On the other hand, effective on-chip
protection is not always possible and many discrete components are in the range of 100–
150 V according to the standard HBM ESD test [3,4]. The number of ultrasensitive devices
with ESD withstand voltages below 100 V is increasing, including special rfdevices, flat
panel displays, and magnetoresistive (MR) recording heads.
It is necessary to take ESD protective measures in the manufacturing environment in
order to protect ESD sensitive devices (ESDS) against damage [3,5]. In this study we
concentrate on the risk of ESD due to charged clothing of the manufacturing operators.
Electrostatic charges typically accumulate when the operator is moving, by triboelectric
effects (rubbing or separation of two different materials). Specially designed ESD garments
are worn over the ordinary clothing of the operator to protect ESDS from accumulated
charge on the underlying clothes. These garments should provide shielding against any
surface voltages or voltage transients (ESD) arising from underlying garments. In some
cases the ESD garments also play other important roles such as protecting the electronics
from dust particles originating at the operator (cleanroom clothing).
Current ESD garment test standards [5,6] are mainly based on research performed in the
1980’s for homogeneous materials. They do not allow a proper characterisation of the
modern heterogenous material garment performance [7,8]. Since then the electronics industry
has demanded increasing performance, and at the same time there has been much progress
in the textile industry. The ESD garments in use today are made of composite fabrics where
a grid or stripes of conductive threads are present inside a matrix of cotton, polyester or
mixtures of these materials. The conductive threads are often composite (core conductive
fibres, sandwich type fibres etc.). The garment conductive fibres should be grounded or
electrically connected to the operator’s body, but in many real factories we have observed
that this grounding is not achieved. In this case the ESD garment is free to acquire a voltage
and become itself the source of ESD.
A European research project “Protective clothing for use in the manufacturing of
electrostatic sensitive devices” (ESTAT-Garments) commenced in early 2002. The main
goals of the three-year project are to supply the International Electrotechnical Commission
Technical Committee TC101 with a basis to qualify the effectiveness of clothing used in
safe handling of ESD sensitive devices, and to develop appropriate test methods for the
characterisation of ESD protective garments. The project partners—VTT (FIN), University
of Geneva (I), SP (S), Centexbel (B), STFI (D), Nokia (FIN), Celestica (I)—consist of experts
of electrostatics, electrostatic measurements, textile technology and electronics
manufacturers (end-users of the garments). To achieve the main goals, the project aims to
understand the electrophysical processes leading to ESD damage to sensitive devices from
garment materials.

241
2. Requirements for electrostatic testing of ESD garments

Standard test methods for ESD fabrics and garments are mainly resistive or measure the
charge decay time of the material [5,6]. A good state-of-the-art review of the subject is
given in ref. [8]. Current standard test methods do not satisfactory qualify the effectiveness
of novel BSD-protective garments and ignore potentially important factors such as:
* triboelectric propensity of the fabric,
* the effect on performance of grounding the conductive garment elements,
* the risks introduced by unearthed conductive fibres as a possible source of ESD,
* possible ESD risk arising from charged insulating areas of a heterogeneous fabric,
* the possible penetration of electrostatic fields from underlying normal clothing
through the fabric,
* the influence of grounded operator wearing the garment on the protective
performance of the garment
The use of resistive measurements with core conductive fibre fabrics leads to the rejection
of a fabric or garment, simply because the measuring electrode makes contact only with the
non-conducting surface of the fibres. The charge decay test [5] emphasises the influence of
the insulating base fabric on the material performance. To overcome this problem, it has
been proposed that one should also measure the capacitance experienced by charge on the
surface of the material [9] or the electrostatic shielding factor of the material [10].
Verification of all these factors could require a long list of measurements to evaluate the
garment fabric and the garment as worn (system measurement) [11]. The ESTAT-Garments
project includes basic research to evaluate the importance of such factors.
An ideal ESD garment might never be realised due to contradictory requirements: 1) low
resistance for fast dissipation of charge and to prevent field-induced damage, 2) high
resistance to slow down the charge decay and limit the energy transferred in a direct discharge,
3) total suppression of electrostatic fields from charge under, and on, an ESD garment
surface, and 4) an anti-static material that does not generate a charge when contact is made
to any other material [8]. A good understanding of the electrophysical processes is required
to achieve proper compromises. Evaluation of the garment should ideally be done in a
single, simple test. That, unfortunately, may not be a realistic target.
3. Assessing the risk of damage to electronics with reference to garments

For many components, ESD damage is related to the energy or peak power of the ESD
dissipated within the device. These are related to the peak discharge current, which depends
on the ESD voltage and the characteristics of the ESD source, and the impedance of the
discharge circuit. For voltage sensitive components the critical parameter is electric field
strength inside a device, which is proportional to charge. Hence, ESD peak current, charge
transferred, and device charging have been suggested to be key parameters in assessing
ESD risk [13]. A new way for the assessment of ESD threats to electronic components has
been proposed, based on the use of current and charge thresholds derived from the component
HBM and CDM withstand voltages, respectively. ESD garments may therefore be evaluated
by measuring discharge currents from charged garments (direct discharge) and device
charging due to induction or triboelectrification (CDM ESD).
A ESD failure caused by charged operator or charged clothing can potentially happen in
at least three different ways: by a direct discharge to a device, by a discharge from a charged
device, and by radiation, i.e. by an induced EMI (electromagnetic interference) pulse due

242
to ESD. In the ESTAT Garments studies these risks will be evaluated and either justified or
excluded. This paper, however, only considers direct ESD from the garment material as a
source of ESD risk. Direct discharges from the clothing of the operator, from unearthed
conducting threads of the garment, or from insulating surfaces of the garment fabric are
related to improperly worn, ungrounded or defective, or poorly designed ESD garments,
respectively. In the remainder of this paper we focus on studies of direct ESD discharges
obtained from fabrics used for manufacturing ESD protective garments
4. The experimental arrangement

In the following experiments we have used two types of passive ESD probe to measure the
discharge currents from a number of different discharge events [13]. One probe (SP probe)
was developed at SP in Sweden. The other probe (JS probe) has been developed and described
by Smallwood and Hearn [14,15]. Both probes have been found to give similar results.
Approximately 24cm×24cm samples of the protective fabrics were fastened in a
conducting frame and placed 5cm from a large ground plane (Figure 1). An electrostatic
field meter mounted in the centre of this ground plane enabled measurement of the average
potential of the system test fabric and its supporting frame.
We have included in this study protective fabrics with three different kinds of conducting
threads, namely: Core-conducting carbon fibre (CC), Surface-conducting carbon fibre (SC)
and Stainless steel fibre (SS). The conducting threads are woven into the main fabric in a
square of 10mm and 5mm dimensions. We refer to the different fabrics as CC10, SC10,
SS10, CC05, SC05 and SS05, where the “10” in SS10 stands for the 10 mm square and 05
for the 5mm square, and the letter code (SS etc) defined above.
The major part of these fabrics consists of polyester or cotton-polyester fibres. In these
tests we have not observed any performance difference between these base materials.
All the test fabrics were preconditioned to 23 degrees Celsius and 12% RH for at least 72
hours before the measurements were made under the same conditions. Samples were charged
by triboelectrification or by direct charging via the metal frame. Tribocharging was achieved
by rubbing with a 5cm diameter Teflon disc until the average potential reached a given
value (typical 1000V). The test fabric would be left for a minute to rest before it was
discharged via one of the ESD probes.
Direct charging could only be used on conducting parts of the fabrics. In this case a
voltage was applied to the sample support frame for two minutes before the power supply
was disconnected. The sample was subsequently discharged through an ESD probe.
Figure 1. Experimental arrangement for measurement of ESD from charged fabric surfaces.

243
Figure 2 shows examples of the discharge waveforms obtained from a direct charged sample,
and a triboelectrically charged sample, indicated measured with the SP-probe. The fabric
used in this experiment was the core-conducting CC05. The discharge waveforms are
similar in amplitude and in shape. There is an apparent voltage offset in the case of the
direct charging waveform, due to measurement instrument effects. This was taken into
account in the measured values given below.
Five discharges were made for each charging case with the sample average potential,
and charging potential, fixed at 1000V. The average peak current was 0.33 mA for direct
charging, and 0.27 mA for tribocharging. The average charge transferred in the discharges
was 36pC for direct charging and 31pC for tribocharging. Direct charging and tribocharging
appear to give the approximately the same discharge currents and charge transfer for given
experimental conditions. Similar tests made for the fabrics with surface conducting carbon
fibre and stainless steel fibres indicated that this holds for all the other fabrics tested.
Comparing the different fabrics requires a large number of measurements. Average charge
transfer and average peak currents were calculated from approximately six discharge events.
The direct charging method and the SP probe were used in these experiments. Figure 3
shows the average peak discharge current, and charge transferred in the discharge as a
function of the charging voltage, for the fabrics SC05 and SC10.
While there is some variation in the relationship between the results SC05 and SC10
results, charge transfer and peak current increased rapidly with charging voltage for both
fabrics. The SC05 fabric gave higher charge transfer but lower peak current than SC10 at
most voltages. The charge transferred and peak current also appear to increase with charging
potential for the CC05 and CC10 fabrics and for the SS10 fabric (Figures 4 and 5)
Figure 4 shows the results for CC05 and CC10 fabrics. Below the charging potential of
700V no discharges were obtained. It is unclear whether the measurement system was not
sufficiently sensitive to measure the discharges, or there were no significant discharges
below this threshold. As the conductive fibres have a buried conductive core, it is possible
that a threshold voltage must be exceeded before charge can be transferred to, or from, the
conducting fibre core. It is not clear from these results whether there are significant differences
between the charge transfer and peak current for the two grid sizes for this fabric.
Figure 2. Discharges obtained after direct charging and tribocharging CC05 fabric.

244
Figure 3. a) Average peak discharge current and b) average charge from a discharge as function of the
charging voltage for the SC05 and SC10 fabrics.
The peak current and the charge transferred in a discharge event increased with the
charging potential for all of the fabrics CC05, CC10, SC05, SC10 and SS10. If we compare
the amount of charge and the peak current between the different fabrics for a given charging
potential, we find the following order of the fabrics for decreasing peak current and charge:
SS10, SC10 and CC10. This corresponds to increasing surface resistances of these fabrics
(SS10 has the lowest value).
Examples of discharges taken from the conducting threads of fabrics charged to 2 kV are
given in Figure 6. Figure 6a represents an ESD from stainless steel threads, and Figure 6b
from surface conducting carbon fibre threads. Peak ESD current from surface conducting
carbon threads is about one hundredth of the peak current from stainless steel threads. ESD
peak current from core conductive carbon threads (not shown) is slightly smaller than that
from surface conducting threads.
Figure 4. a) Peak current and b) average charge transferred in discharge events from CC05 and CC10
fabrics as function of the charging potential.

245
Figure 5. a) Average peak current and b) charge transferred in a discharge event as function of the
charging potential for SS10 fabric
According to [13] the 13 A discharge current of the stainless steel thread would be expected
to represent a threat for devices with HBM withstand below 20 kV and the 0.12 A current of
Fig. 2b a risk for devices with HBM withstand below 200 V. No discharges were obtained
from the insulating material between the threads at this charging level.
In order to be able to measure direct discharges from the insulating parts of the ESD
protective fabrics, fabrics have to be charged by triboelectrification (or by ion deposition)
up to several kV. Such high charging level is quite unlike in practice for ESD garment, on
the contrary to normal clothing where surface potential of several kV is easily achieved. It
seems that direct discharges from insulating surfaces of the garment fabric are not a significant
ESD risk. Charged, unearthed conducting threads, however, form a risk for ESD damage of
devices: unearthed stainless steel threads may be a risk for all ESDS and carbon fibre
threads at least for Class 0 devices with HBM ESD withstand voltages below 250 V.
In Figure 6b the discharge from SC fibres continued at a lower level for a significant time
duration after the initial peak. It is possible that this is due to charge draining from the wider
area of linked conductive fibres through the sample over a period of time.
Figure 6. ESD current waveforms from fabri with SC fibre threads. Only the initial part of the
discharge curve is shown.

246
6. Conclusions
A European research project “Protective clothing for use in the manufacturing of electrostatic
sensitive devices (ESTAT-Garments)” is in progress, aimed at giving a basis to qualify the
effectiveness of clothing used for the BSD-safe handling of ESD sensitive devices (ESDS)
and to develop appropriate test methods for the characterisation of such ESD protective
garments.
Measurements have been made of direct ESD from fabric samples which have stainless
steel, surface conducting carbon, or carbon core fibres as conducting elements. Peak current
and charge transferred in the ESD events have been measured to evaluate their significance
as a possible ESD damage risk to sensitive electronic devices. No difference has been found
between discharges from triboelectrically charged, and direct charged, fabrics. Both peak
current and charge transferred in the discharge increased with the fabric surface voltage.
Fabrics could be placed in the order of increasing charge and peak current at a given surface
voltage: carbon core fibre, carbon surface conducting fibre and stainless steel conductive
fibre, corresponding decreasing surface resistances of these fabrics.
Discharges from stainless steel fibres charged to 2kV were about a hundred times greater
amplitude (13 A) than from surface conductive carbon fibres (0.12 A). This amplitude
could represent a significant ESD risk to sensitive devices.
Acknowledgements
This work is supported by the European Commission, Contract No. G6RD-CT-2001–00615.
References
[1] Merril R and Issaq E 1993 Proc. EOS/ESD Symposium EOS-15 (New York: ESDA) 233–237.
[2] Amerasekera A and Duwury C 2002 ESD in silicon integrated circuits (West Sussex: Wiley)
ISBN 0–471–49871–8
[3] ESD Association 1999 Standard ANSI/ESD S20.20
[4] International Electrotechnical Commission. 2002 IEC61340–3-1 Electrostatics—Part 3.1 Methods
for simulation of electrostatic effects—Human Body Model (HBM)—Component testing. ISBN
2–8318–6247–7
[5] International Electrotechnical Commission 1998 Technical Report 61340–5-1 Electrostatics—
Part 5.1: Protection of electronic devices from electrostatic phenomena—General requirements.
ISBN 2–8318–4608–0
[6] ESD Association 1997 Standard ESD STM2.1
[7] Dyer M J 1997 Proc. EOS/ESD Symposium EOS-19 (New York: ESDA) 276–286
[8] Baumgartner G 2000, Consideration for developing ESD garment specifications, Report ESD
TR 05–00 (New York: ESD Association)
[9] Chubb J 2002 J. Electrostatics 54 233–244
[10] CEN 2001 Standard draft prEN 1149–3
[11] SP 2000 SP-Method 2175
[12] Smallwood J M and Paasi J 2003 “Assessment of ESD threats to electronic components and
ESD control requirements” Proc. Electrostatics 2003. loP Conf. Series. Institute of Physics,
London.
[13] Fast L, Paasi J, Kalliohaka T, Borjesson A, Smallwood J. 2003. Direct discharges from ESD
fabrics. 1st Nordic ESD Conference, Karlskoga, Sweden
[14] Smallwood J M, “Simple passive transmission line probes for electrostatic discharge
measurements”, (1999), Inst. Phys. Conf. Se. 163 pp. 363–366
[15] Smallwood J M and Hearn G L 2003 “A wide bandwidth probe for electrostatic discharge
measurements”, Proc. Electrostatics 2003. loP Conf. Series. Institute of Physics, London

247
Assessment of ESD threats to electronic components and

ESD control equirements
Jeremy Smallwood1 and Jaakko Paasi2
1
Electrostatic Solutions Ltd., 13 Redhill Crescent, Bassett, Southampton, SO 16
7BQ,
UK. Email: jeremys@static-sol.com
2
VTT Industrial Systems, PO Box 1306, FIN-33101 Tampere, Finland. Email:
jaakko.paasi@vtt.fi
Abstract. We propose new methods for the assessment of real electrostatic discharge (ESD)
threats to electronic components in modern electronics manufacturing environment, where the
devices are more sensitive to ESD than ever. The work was based on two fundamental
questions: what are the key parameters of ESD that must be controlled in order to minimise
risk of damage, and what are the threshold levels of these parameters, below which ESD risk
is low? As a result of the study, proposals are given for two new ways of assessing risk: by
measurement of peak ESD current, and charge induced on test objects. We suggest how
guideline limits for ESD damage thresholds based on these tests may be derived.
1. Introduction
In modern electronics manufacture ESD (electrostatic discharge) sensitive electronic devices,
such as discrete transistors and integrated circuits, may be at risk from damage from ESD
from objects and materials in their environment. Typical ESD sources include operators,
charged insulators and isolated metal parts in automated handling equipment and manual
assembly, including clothing of operators. Effective ESD control to minimise device failures
due to ESD in modern electronics manufacturing environment requires that real ESD threats
to electronic components are well assessed.
ESD waveforms are highly variable, and ESD sensitivity in components must be assessed
using an ESD model that appropriately simulates the real world ESD waveform. Historically,
this has led to the development of three main component ESD withstand test models:
Human Body Model (HBM), Machine Model (MM) and Charged Device Model (CDM)
[1,2,3]. Standard tests are now listed by bodies such as the ESD Association, JEDEC and the
IEC and are regularly used in qualification of production electronic device designs.
Device ruggedness specified as HBM, MM and CDM withstand voltage may be found
on some modern device data sheets. Device manufacturers commonly include protection
networks in 1C designs, targeted at achieving 2kV HBM, 200V MM, and IkV CDM ESD
withstand voltage thresholds. However there are many devices that are unprotected and
have ESD withstand voltages less than IkV. RF technologies commonly have HBM and
MM withstand in the 30–200 V range. Giant MagnetoResistive (GMR) heads are probably
the most sensitive devices currently available, with ESD withstand voltages in the Volt
region.

248
In practice, these ESD withstand voltages are a poor guide to ESD risk in the manufacturing
environment. Voltage is a consequence of the charge and its environment, rather than
being a fundamental contributor to ESD risk. When a charged printed wiring board is
transported along a conveyor, measurements may show low voltage due to nearby metallic
machine parts. In another part of the machine the same board carrying the same charge may
indicate a much higher voltage [4]. A device ESD withstand voltage indicates the risk of
damage only under a particular set of circumstances with a particular model ESD source.
The risk of ESD damage of devices due to charged insulators near or in contact with a
device cannot be assessed well by using any voltage criterion.
There is a need to identify more realistic and quantifiable parameters than voltage for
assessing ESD risk. The objective of this paper is to propose ESD risk thresholds that are
related to the real device damage mechanisms, in terms of practical measurable parameters
such as peak ESD current and charge storage on ESD sensitive devices.
2. Human Body Model, Machine Model and Charged Device Model ESD withstand
The Human Body Model (HBM) simulates the situation where a charged person touches a
device, with another pin of the device typically grounded. Machine Model (MM) emulates
the situation where a charged metallic object (such as a trolley or charged machine part) is
the ESD source. Both types of event are “2-pin” events where a discharge current enters one
pin and leaves another, passing through device internal circuitry.
The situation is simulated using a simple electronic circuit (Figure 1). Additional
components may need to be added to tailor the waveform to better represent real world
waveforms [2]. A capacitor C is charged to an ESD voltage Vesd. On initiating the ESD event,
the ESD current flows through a circuit resistance R, inductance L and the load device (and
possible spark gap). Typical component values are shown in Table 1. The largest voltage
Vesd that the device can stand without damage is the device ESD withstand voltage.
HBM and MM ESD typically create the same types of damage signatures in the
component, even though the ESD waveforms are very different [2,3]. The peak current
achieved by HBM is achieved at much lower ESD voltage in MM due to the lower circuit
impedance, which consists primarily of stray L and R, and the impedance of the victim
device. For the same ESD voltage, MM ESD currents are about 10–20 times the HBM
value. Perhaps not surprisingly, the ESD withstand voltage for MM is found to be 10–20
times less than HBM. The duration of HBM and MM ESD is typically 100–150 ns.
In the Charged Device Model (CDM) the victim device itself acts as the circuit capacitor.
C is now the device effective capacitance (a variable), and the L and R are stray components
within the discharge path.
Figure 1. A general ESD model.

249
Table 1. Typical ESD model simulation component values
CDM ESD damage typically has different failure signatures than HBM and MM. CDM ESD
occurs when a device becomes charged, and then is brought into contact with a low impedance
ground. The discharge is a “1-pin” event with the discharge current originating in the
device, and limited by the ground path and device internal impedances. A short duration (a
few ns), high current (higher than in MM) waveform results.
Charged device model is of growing importance, especially in automated handling and
assembly systems where a device may become charged by triboelectrification or induction
from nearby fields, and subsequently contact metal machine parts.
3. Component damage due to energy dissipation in internal weak components

In HBM and MM, damage of ESD sensitive devices often occurs because a small internal
region is intensely heated by the passage of the ESD current. The damage region may be a
metallization track, or a transistor junction.
The power dissipation in the damage region is given by
P(t)=I(t)2R(t) (1)
where P(t), I(t) and R(t) are the instantaneous power dissipation, discharge current and
effective device resistance. The power dissipation causes a temperature rise in the region—
damage is caused if the temperature exceeds a certain threshold (e.g. the material melting
point). Modern devices have extremely small (sub micron) internal feature dimensions, and
so only a small energy is required to raise the hot spot to the damage temperature.
Models of this failure mode have been based on the general heat flow equation
(2)
where ρ is the density, Cp is the specific heat, T the temperature of the material, K is the
thermal conductivity, and q(t) is the rate of heating per unit volume of the heat source.
Smith has argued that a general heat equation of this type applies for all ESD failures
including metal, semiconductors and insulating materials [5].
Wunsch & Bell [6] developed the first junction burnout model based on the solution of
the heat flow equation for a planar heat sorce, considering the temperature rise ∆T of the
junction to the fail temperature. Tasca [7] extended the Wunsch-Bell model to wider time
regimes and solved the heat flow equation for a spherical source in an infinite medium. The
reader is referred to Tasca’s original work for details. In Tasca’s solution power Pf required
to raise the spherical region to fail temperature is a function of the pulse duration
(3)
where A=ρCp and V is the volume of of the heat source, B=(KρCp)1/2 and S the surface area
of the heated region, C is a constant describing the steady state condition, T is the
temperature of the device and T0 is the initial temperature. The equation above shows that

250
the power required to failure for a given failure temperature is determined by the duration
of the discharge pulse. For short pulses (t<0.1 µs) the situation is adiabatic (i.e. there is no
or very little heat flow out of the heated region) and the time dependence follows an 1/t
dependence. In this regime, a constant amount of energy AV(T-T0) is required to raise the
spherical region to failure temperature, dependent only on the size of the heated region,
properties of the materials, and the temperature rise to failure. In the regime of intermediate
pulse lengths (0.1 (is to 100 µs) some heat diffuses away from the defect region and the
failure power BS(T-T0) follows a t-1/2 dependence. For long pulse (t>100 µs) energy is lost
to the surrounding material at a constant rate (dependent on device heat sinking) and,
thus, the failure power is constant. In summary, the power required to reach the failure
temperature decreases as t is increased until a steady state is achieved.
4. Proposal of a discharge current threshold for ESD damage

For ESD damage to occur, the internal power dissipation in the weak region must exceed a
threshold value to overcome heat losses from the region. This power dissipation must occur
at some discharge current level, and a current threshold for ESD damage therefore is proposed.
If this current level is exceeded the power dissipation will rapidly exceed the power
required to heat the damage site to the damage temperature (P=f(I)) and providing the
discharge is of sufficient duration, damage will occur. An ESD event having a lower peak
current can be expected to not exceed the power threshold, and therefore not cause damage.
A threshold obtained for a longer duration discharge, where power loss from the heated
region is significant, can be expected to give a safe value for shorter ESD durations.
One question is, how can the threshold be measured? We do not know where the weakest
region in the device may be, or what is its size or effective resistance.
In the HBM ESD withstand test the discharge current is primarily governed in many, if
not most, cases by the series 1500 Ω resistance. The duration of the discharge is about 100
ns, which is likely to be on the threshold of the adiabatic heating regime. We can propose
that the peak current in the HBM ESD event may directly give a reasonable measure of the
ESD current threshold for damage of the device in HBM & MM-like situations. Perhaps
most usefully, the test is already performed on the majority of devices. If the device HBM
withstand voltage VHBM is known, then an appropriate peak ESD current threshold IESDmax
can be calculated from
(4)
Thus a threshold of 100 V HBM gives a peak current threshold of 66.7 mA. Protection of
10 V HBM devices requires a reduction in peak current threshold to 6.7 mA. In general, the
allowable current is 0.67 mA per Volt HBM withstand.
One complication is that a real HBM waveform includes an initial transient due to the
discharge of stray circuit components. This transient will tend to make the device fail at a
lower HBM ESD threshold voltage. This, however, would lead to underestimation of the
ESD current withstand threshold, which increases the margin of safety for ESD risk assessment
when using the calculated peak current as a damage threshold.
In assessing an ESD garment material, or and automated handling line, for ESD risk, the
HBM ESD withstand peak current is therefore a primary specification. The material or
equipment should not be able to source an ESD peak discharge current exceeding this
threshold. A survey of such discharges may be made part of the material or equipment
laboratory qualification procedure.

251
5. Proposal of a charge threshold for ESD damage

Some authors have also found evidence that CDM withstand voltage is related to peak ESD
current [8, 9]. However, the peak current is a result of device and discharge circuit impedance
as well as the ESD voltage. So, it is very much a device package, situation and environment
dependent parameter. It is not easy to see how peak discharge current can realistically be
used as a failure threshold in this case. The HBM values do not apply, as the failure modes
are typically different, for example due to internal oxide layer breakdown.
5.1. Proposal of a charge limit from the CDM test parameters

The CDM peak current is likely to be a function of CDM voltage arising on the device
capacitance. So, it is reasonable to propose a charge threshold could be defined, with ESD
damage risk occurring when charge induced on, or accumulated by the device reaches and
exceeds this value. Charge measurement is relatively easy and a device independent test
can be envisaged. The device CDM voltage withstand VCDM is a standard test result, and
coupled with a device capacitance value Cd could yield a charge threshold value Qth
CdVCDM=Qth (5)
One problem is, what value of Cd should be used? Device capacitance is seldom measured,
and in practice varies with the device position and surrounding materials and objects. If the
capacitance in the CDM test configuration is known, this would be an appropriate value.
More often, the capacitance is not known. The obvious approach is to choose a value at an
extreme of the range, although it is not initially clear whether a high or low value presents
the safest choice. Device capacitance is often in the range 1–30 pF for integrated circuits.
5.2. Specification of a charge threshold

The CDM peak current is determined by the series resistance and inductance of the circuit.
These are stray components of indeterminate value, inherent in the device structure. Two key
questions are, can we justify linking peak current and device charge level, and if so, what
combination of VCDM and Cd could give the highest peak current? We can simplify the
argument by considering each series component in turn. If the series resistance R is sufficiently
large that the effect of the inductance is negligible, the peak current I p is given by
(6)
Clearly this is greatest when VCDM is maximum. Substituting for charge
(7)
Thus the peak current is proportional to the device charge level, and for a given device
charge level, the highest value of Ip is found at minimum device capacitance. Similar result
can be obtained for the case where the resistance is negligible compared with the inductance.
From this analysis, choosing a low value for Cd would give a safe estimate for use in
charge threshold calculation. A value of 1 pF would represent a small device at the lower
end of the possible range. Most devices sensitive to CDM ESD would be expected to have
higher capacitance than this. While a small device in free space could have a lower
capacitance, ESD risk is only present when a device approaches an object and the capacitance
is greater than the free space value.

252
If the CDM withstand voltage VCDM of the most sensitive device expected handled is
known we can use this value, otherwise we might choose a low value of, say, 100 V. The
calculated charge threshold for these estimated values is then 0.1 nC. Induced charge or
triboelectrically generated charge less than this value would not be expected to cause
CDM damage to most devices. In practice a variety of factors in any case reduce the ESD
risk. If a greater margin of safety is required, then the charge threshold may be reduced
accordingly.
The analysis is supported by the results of Brodbeck and Kagerer [8] who have shown
that device charge levels at the damage threshold reduced with reducing device capacitance,
although the peak current at the threshold level remained the same.
6. Conclusions
Two new methods of assessment of ESD risk to sensitive electronic devices have been
proposed here, with proposals of how appropriate limits may be derived from standard ESD
test results. In summary the proposals are that
• Discharge current, power and energy related damage thresholds can be simplified to be
represented by an ESD peak current threshold for damage which covers a wide range of
situations with a margin of safety. The peak discharge current threshold is derived from
component HBM ESD sensitivity data
• A charge threshold can be derived that corresponds to ESD risk to a charged device.
The threshold may be derived from standard CDM ESD sensitivity data.
Further work is in progress to evaluate the usefulness of the approach in practice, applied to
assessment of ESD garments and automated manufacturing facilities. We anticipate that
the proposals may find more general application in many different situations in electronics
manufacture.
Acknowledgements
The work is partially supported by the Finnish National Technology Agency Tekes via the
STAHA programme and by the EC, Contract No. G6RD-CT-2001–00615.
References
[1] Gieser H, Ruge I 1994 Survey on Electrostatic Susceptibility of Integrated circuits Proc. ESREF
Symp. 447–455
[2] Wang A Z H 2002 On-chip ESD protection for integrated circuits (Boston: Kluwer)
[3] Amerasekera A., Duwury C 2002 ESD in Silicon Integrated Circuits 2nd Ed. (New York: Wiley)
[4] Paasi J, Tamminen P, Kalliohaka T, Kojo H, Tappura K 2002 ESD control tools for surface
mount technology and final assembly lines. Proc EOS/ESD Symp. EOS-24 250–256
[5] Smith J S. 1997 General EOS/ESD equation Proc EOS/ESD Symp. EOS-19 59–67
[6] Wunsch D C , Bell R R 1968 Determination of threshold failure levels of semiconductor diodes
and transistors due to pulse voltages. IEEE Trans. Nuc. Sci. NS-15 244–259
[7] Tasca D.M. 1970 Pulse power failure modes in semiconductors IEEE Trans. Niicl. Sci. NS-17
364–72
[8] Brodbeck T, Kagerer A 1998 Influence of device package on the results of CDM tests—Z
consequences for tester characterisation and test procedure Proc. EOS/ESD Symp. EOS-20 320–
327
[9] Reiner J C 1995 Latent gate oxide defects caused by CDM ESD Proc EOS/ESD Symp. EOS-17
311–321

253
Calorimetry in the detection of discharge events
Z Kucerovsky†, W D Greason†, and M Wm Flatley‡†

† Applied Electrostatics Research Centre, University of Western Ontario,
London, N6A 5B9, Canada
‡ Suncor Energy Products, Sarnia, P.O. Box 307, Ontario, N7T 7J3, Canada
Abstract. Optical signals emitted by electrostatic discharges were studied using calorimetry,
determining the accuracy and sensitivity. A special calorimeter has been developed and used
with optical sources, namely a spark gap, and an International Electrotechnical Commission
standard human-body-model ESD generator (3.1×10-4 J, +25 kV, 8-mm dia, 150 pF, 150 Ω).
An optical system consisting of lenses and a mirror collected signal for the calorimeter—an
optical fiber provided inputs for a calibrator comprised of a broadband photodetector and two
spectrometers. The calorimeter’s sensor used a four-thermistor bridge, powered by an a.c.
generator (2 V, 9 kHz); the imbalance signal was processed by a band-pass amplifier, rectified,
integrated, digitized, and computer processed. The system’s response in the visible and i.r.
regions was determined for the threshold optical signals and its long-term stability measured.
The calorimeter responsivity was S=5.85×10 3 V.J -1 and its noise equivalent input
NEI=1.71×10–12 J.
1. Introduction
Early studies of electromagnetic emission, starting with the foundation work of Hertz in
eighteen-nineties, incorporated the spark gap in the electromagnetic wave generator. The
complex physical processes in the spark and arc were studied later by a number of scientists
[1,2,3]. The spark events were usually handled as non-periodic phenomena, as non-periodic
spark pulses are easier to generate [4,5,6]; however, the single event spark makes the study
of its energy content and spectrum challenging [7]. The spark energy content is similar to
that of a pulsed laser in magnitude and duration; consequently, laser methods such as
calorimetry can be used in the measurements of the spark energy [8,9].
The theory and techniques of modern calorimetry used in physical and chemical sciences
have been the subject of recent reviews [10,11,12] and the topic of a number of publications
[13,14,15,16]. In calorimetric measurements, temperature difference between two bodies is
determined, usually with a thermocouple, bolometer or thermistor. In a limited temperature
range, the thermocouple is nearly linear in response, its low mass offering accuracy,
sensitivity and fast response [9]; unfortunately, its signal needs a d.c. or a chopper-stabilized
amplifier. Compared to thermocouples, thermistor resistors are less linear and non-active,
but can be arranged to yield a.c. signal, can be linearized [17] and their yield optimized
[18]. The thermistor’s signal processing amplifier is often complex [19], and the strategy for

254
the thermistor utilization sophisticated (e.g. Blumlein bridge [20]). A modern calorimeter
may incorporate an optical fiber [13] or be optimized for the speed of response [14].
Our work focused on the measurement of the energy contained in the spark’s optical
emission using a calorimeter with a thermistor based thermometer.
2. Theoretical part
Energy εp transferred from a pulsed Gaussian beam to a solid, non-reflecting, target,
is
. (1)
Assuming a target larger than the beam size, the incident and absorbed energies are equal.
If there is no appreciable heat loss, the balance equation for the energy ∆εp received by
the target of the heat capacity Cv with respect to the corresponding temperature increase
∆T is:
∆εp=Cv∆T. (2)
Target’s specific heat being cs and mass m, the temperature increment is
. (3)
Since the temperature increase is proportional to the energy deposited on the target, the
energy contained in the optical beam can be measured with a low-mass thermometer—in
our calorimeter, thermistors have been used.
The room temperature resistance of a typical thermistor may be taken as RT=2 kΩ (at
T=300 K). If its signal is band-limited by an active a.c. filter to a 100-Hz bandwidth
(∆B). it generates Johnson noise
, (4)
which for the given parameters is υ^2=3.3×10-15 V2 (in expression (4), k=1.38×10–23 J·K–1 is
the Boltzmann constant). This relatively low level of noise was deemed reasonable for
measuring the energy of the ESD discharge.
The discharge signal of the IEC class was generated in the standard human-bodymodel
serial circuit consisting of a storage capacitor (C=150 pF) and a resistor (R= 150 Ω).
Neglecting circuit’s inductance, the capacitor voltage υC is
. (5)
Assuming that the resistance of the discharge Rd is time invariant and Rd << R, the
discharge process is dominated by resistance R, yielding an exponential current waveform
, (6)

255
shown in figure 1. The energy transferred from the capacitor to the spark discharge is
. (7)
If, in the first approximation, the spark impedance is taken as constant Rd = 1.0 Ω, the total
energy transferred from the storage capacitor to the spark is
which yields εd=0.31 mJ and εR=46.6 mJ for the values of energy deposited by the pulse
into the spark and into the model network buffer resistor R, respectively (for υc=25 kV,
C=150 pF, R=150Ω, Rd = 1.0 Ω).
Figure 1. Human-body-model ESD recommended waveform (a) and the pulse (b) used for calorimeter
calibration (both defined by the IEC).
3. Experimental part
The optical part of the apparatus is shown in the left part of figure 2. The baseline was a
short optical bench (not shown), which supported a holder accommodating the IEC human-
body-mo del generator, and the optical aligner and calibrator. The middle holder secured
the position of an optical condenser (L1,2). The calorimeter (C) and an optical fiber pick-up
were manipulated by the third holder. Two grating spectrometers (S1000, Ocean Optics)
were available remotely, their signal supplied via optical fibers. A light emitting diode,
placed on the optical axis of the spark gap, was used for calibration and alignment. Five
interchangeable diodes covered the visible and near infrared spectrum (473, 556, 587, 640,
and 880 nm). The condenser lenses (100 mm diameter) yielded an effective solid angle
Ω=0.50 sr for the signal collecting optics. The spark was located at the condenser’s first
focus (124 mm) and the collected radiation focused by the second lens on the calorimeter
or the optical fiber.
A cross section of the calorimeter is shown in the right part of figure 2. The body of the
device was made of copper. The circular entrance aperture (21.4 mm diameter) had a provision
for an optional, quartz window (W) sealed with an O-ring. The window, the input aperture,

256
Figure 2. Optical arrangement of the calorimeter (left) and a cross section of the calorimeter sensing head
(right) (Left figure: C, calorimeter; D, IEC test generator; L1,2 lens; S, grounded sphere) (Right figure: B,
polystyrene baffle; E, electronic parts; T1,4 measuring- T2,3 reference thermistors); W, optical window.
and the heat shield baffle (B) allowed to achieve the F-stop number of one. The two measuring
thermistors (T1,3) were bonded to a black coated sensor, and the two reference thermistors
(T2,3) to the copper body. All the four thermistors were identical in make (Fenwall, 192–
102DET-A01).
For ESD emulation, an IEC human-body-mo del network was used (KeyTek, M2000,
150 pF, 150 Ω, 8 mm dia., +25 kV tip); the instrument satisfied the IEC standard for
electrostatic discharge [21]. Figure 1 shows the test waveforms: (a), IEC recommended; and
(b), provided by the used discharge circuit.
Our thermometer used four thermistors in a Wheat stone bridge, arranged to double the
sensitivity as shown in figure 3. The bridge (T1"4) was a.c.-powered (G, S, 2.0 V, 9 kHz), and
its imbalance voltage processed by a low noise amplifier (A1,2, LT1037) and an integrated
instrumentation amplifier (A3"5, LT1167N), followed by a unit-gain amplifier (A6, LMC662).
A synchronous detector (D, CD4053) was used for the signal rectification, and an integrator
(I, LM358) prepared the signal for digitization (Keithley, digital voltmeter M 2001) and
computer processing.

Experiments were performed to determine the sensitivity and time constant of the calorimeter
using the lEC-class discharge (150 pF, 150 Ω, +25 kV, 6 mm, figure 4). The waveform on the
left shows the calorimeter response, when the full magnitude of the spark optical energy
was allowed to impinge on the calorimeter sensor. The response on the right was produced
with the same alignment and driver setting, but with the optical signal attenuated
approximately ten times. The waveforms show that in the given arrangement, the calorimeter
signal was not limited by the systems noise, but by the drift. The drift is negligible in figure
on the left, but pronounced in the waveform on the right, where the ‘signal-to-drift’ ratio
was lower because less energy reached the calorimeter sensor.

257
Figure 3. Thermistor bridge and the processing electronics (A1,2, low noise amplifiers (LT1037); A3–5,
instrumentation amplifier (LT1167N, comprises resistors r1"6); A6, amplifier (LM358); D, synchronous
detector; G, function generator (a.c. 9 kHz); I, integrator; O, output; R1–8, resistors; S, Bridge’s symmetrical
power; T1,4, sensing—, T2,3, reference thermistors).
Figure 4. Calorimeter responses to an unattenuated (left) and attenuated (right) illumination from a spark
discharge in air (220–610 nm), generated by a 25 kV IEC pulse (note the low level of drift for the
unattenuated signal).
The calorimeter responses were observed with a number of pulsed and continuous optical
sources. Two grating spectrometers (Ocean Optics, S1000, 100 µm entrance slit) were used
to confirm that, within the instruments range, the calorimeter output was invariable with
the power spectrum of the received signal. The instrument’s responsivity, in terms of output
voltage to input energy, was measured to be S=5.85×103V.J”1. The noise equivalent input
was determined to be NEI=1.71×10–12 J, taking the output noise level equal to that of
Johnson’s noise and the signal-to-noise as one.
5. Conclusions and recommendations

The performance of the open aperture calorimeter (figure 4) was limited by its sensing head
drift, caused mostly by the laboratory air currents. A coupling between the optical condenser
and the calorimeter aperture, and the use of an optical window (figure 2) would alleviate
this problem. As the drift time scale is much different than that of the spark signal, the drift
can be also accounted for by the signal processing software.
As used, the system’s spectral sensitivity ranged from the ultraviolet to infrared. Since

258
the spectra of the typical sources are known a priori, optical filtering of the calorimeter
signal can be used to improve the detection limit, eliminating the spurious light and
atmospheric background interference. In an industrial instrument, if it were to be designed,
a microprocessor or microcontroller signal processing could be used to advantage.
The calorimeter responsivity and its noise equivalent input limit were adequate. The
response depended somewhat on the length of the spark; however, for the same electrode
separation, the system’s response to the spark gap and the IEC human-bodymodel standard
circuit were well correlated. The detectivity and speed of response of the calorimeter can be
changed by choosing a different type of thermistors and mass of the sensor.
The calorimeter system, with responsivities from the joule to micro joule levels, was
found to be suitable for the measurements of the emission energy of the spark in studies of
ESD events, and for a nonintrusive determination of the minimum ignition energies (e.g.
MIE=0.016 mJ for a hydrogen-air mixture [22], while the detectivity of the calorimeter is
3.10 µJ). With a suitable optical system and an improvement in the systems speed of
response, a calorimeter may be suitable for the measurements of discharges encountered on
planar objects, such as fabrics [23]. As calorimetry offers reasonable noise-equivalent energy,
the method might be used for the detection of electromagnetic emission occurring during
the pre-breakdown state.
Acknowledgments
The authors are grateful to Mr. Aartsen for lending his expertise to this project, to the
National Science and Engineering Research Council of Canada, for shouldering a part of
the project cost, and to the Suncor Energy, Sarnia, for providing most of the equipment.
References
[1] Y.P.Raizer, Gas Discharge Physics, Springer Verlag, 1991.
[2] A.von Engle, Ionized Gases, Clarendon, Oxford, U.K., second edition, 1965.
[3] W.D.Greason, Z.Kucerovsky, M.W.Flatley, and S.Bulach, IEEE Trans, On Industry Applications,
vol. 34, no. 3, pp. 571–579, May/June 1998.
[4] F.Llevelyn-Jones, Ionization and Breakdown in Gases, Methuen, London, U.K., 1957.
[5] Erich E.Kunhardt, IEEE Trans. Plasma Sci., vol. PS-8, pp. 130–137, September 1980.
[6] C.C.Davis, Lasers and electro-optics fundamentals and engineering, Cambridge University
Press, 1996.
[7] Z.Kucerovsky, I.I.Inculet, and A.K.W.Lee, IEEE Transactions on Industry Applications, vol.
IA-21, no. 1, pp. 17–22, January/February 1985.
[8] Z.Kucerovsky, W.D.Greason, and J.D.Reis, in Proceedings of the 1990 Industrial Automation
Conference in Toronto, IEEE Canada, Ed. IEEE Toronto, Canada, June 19–21, 1990, pp. 6.1–
6.6.
[9] Bojan B.Radak and Branislav B.Radak, Rev. Sci. Instrum., vol. 62, no. 2, pp. 318–320, February
1991.
[10] Y.Kraftmakher, Physics Reports—Review Section of Physics Letters, vol. 356, no. 1–2, pp. 1–
117, January 2002.
[11] V.B.F.Mathot, Journal of Thermal Physics and Calorimetry, vol. 64, no. 1, pp. 15–35, January
2001.
[12] A.Golutvin, Nuclear Instruments & Methods in Physics, Section A—Accelerators, Spectrometers
and Associated Equipment, vol. 453, no. 1–2, pp. 192–198, Oct 11, 2000.
[13] N.J.Garfield. M.A.Howson, and N.Overend, Rev. Sci. Instrum., vol. 69, no. 5. pp. 2045–2049,
May 1998.
[14] B.Fröchte, Y.Khan, and E.Kneller, Rev. Sci. Instrum., vol. 61, no. 7, pp. 1954–1957, July 1990.

259
[15] Grzegorz Bednarz, Brian Millier, and Mary Anne White. Rev. Sci. Instr., vol. 63, no. 8, pp. 3944–
3952, August 1992.
[16] M.Wm. Flatley, “Design and analysis of a calorimeter,” Master’s thesis, The University of
Western Ontario, Faculty of Graduate Studies, London, Canada, March 1999.
[17] Daniel Slomovitz and Jose Joskowicz, Meas. Sci. Technol, vol. 1, no. 12, pp. 1289–1284, Dec
1990.
[18] Sherif Sedky, Paolo Fiorini, Kris Baert, Lou Hermans, and Robert Mertens, IEEE Trans. Electron
Devices, vol. 46, no. 4, pp. 675–682, April 1999.
[19] G.De Geronimo, G.Bertuccio, and A.Longoni, Rev. Sci. Instrum., vol. 67, no. 7, pp. 2643–2647,
July 1996.
[20] P.N.Murgatroyd and M.Belloufi. Meas. Sci. Technol, vol. 1, no. 1, pp. 9–12, January 1990.
[21] “Electromagnetic compatibility…: Electrostatic discharge requirements,” International
Electrotechnical Commission, 1984, ESD Standard.
[22] U.von Pidoll, E.Brzostek, and H-R Froechtenigt, in Conference Record of the 2002 IEEE Industry
Application 37th Annual Meeting in Pittsburgh, October 13–18, 2002. session 07-p4 of CD
ROM, pp. 1 7.
[23] J.Chubb, “Private communication,” February 1, 2003.

261
Electrical circuit topologies for pulsed corona plasma

generation
K Yan, E J M van Heesch, P A A F Wouters, A J M Pemen, S A Nair

EPS Group, Department of Electrical Engineering
Eindhoven University of Technology, 5600 MB, The Netherlands
e-mail: k.yan@tue.nl
Abstract. This paper discusses novel pulsed circuit topologies for pulsed corona plasma
generation. As a general principle, decrease of output impedance of pulsed power source leads
to scale up of corona plasma power. DC superimposed pulse energization provides advantages
of decrease of switching duty, increase of average power, and improvement of matching.
Critical circuit elements include pulse forming networks (PFNs), triggered spark-gap switches
and transmission line transformers (TLTs). Two main circuit topologies are based on single
and multiple switches. The TLTs are used for both impedance matching and synchronization
of the multiple switches.
1. Introduction
Early patents awarded to Siemens and Lodge-Cottrell on gas cleaning by pulsed corona
technique can also be dated back to the 1930’s [1,2], where capacitive type pulsed power
sources with rotary spark-gap switches were used. Within the last 15 years, the techniques
have been widely investigated again for pollution control [3,4]. Lack of cost-efficient
pulsed power sources, however, resulted in wide disappointment from industrial point of
view. The most critical component in development of pulsed power technology is the high-
voltage and large-current switch. Moreover, the source should be immune to repetitive
spark breakdowns inside the reactor. To our knowledge, there is no system reported to
match these requirements. Several recent industrial systems are based on magnetic
compression with pulse duration of around 200–500 ns [5–8], which is almost ten times
longer for primary streamer propagation [9].
For 1–10 kW systems, high-pressure triggered gas-filled spark-gap switches are the most
cost effective because of high hold-off voltage, large possible currents, and small forward
voltage drop. When scaling up the system from 10 to 100 kW, the main issue, however, is
electrode erosion of the switch. It increases significantly when the switching current becomes
larger. In order to solve the problem, a novel circuit topology based on multiple switches
and transmission line transformers (TLTs) was proposed [9]. According to the concept,
this paper proposes new circuit topologies for corona plasma generation. Ultra-short
pulsed power systems have become available enabling technique for industrial
implementations.

262
2. Single-switch pulsed power circuit topology

Figure 1 shows a circuit topology with a single switch for corona plasma generation. With
regard to technical aspects, the system can be divided into seven parts, namely AC/DC/
pulse converter, pulse forming network (PFN), switch S, transmission line transformer (TLT),
coupling capacitor Cdc, corona reactor, and DC charging unit.
Figure 1. A circuit topology for corona plasma generation. The two-stage TLT is constructed with two
cables and two types of magnetic cores, where and hereafter the characteristic impedance of each
cable is defined as Z0. The TLT is connected in parallel at its input (0.5Z0) and in series at the output sides
(2Z0).
The corona reactor is initially charged to a bias voltage. By firing the switch S, energies
stored in the PFN and in the coupling capacitor C dc are transferred to the reactor
simultaneously. After this pulsed energization, the DC charging unit resonantly charges
the reactor and the capacitor Cdc again via the inductor Ldc and the TLT. The characteristic
impedance of the PFN is equal to the TLT input impedance. The DC bias improves matching,
reduces switching duty, and increases total corona power. The TLT is constructed by using
rigid cables and two types of magnetic cores (A and B). The A and B cores are used to
increase the efficiency and to prevent extra energy dissipation inside the switch, respectively.
A drawback of using a PFN is that the input voltage on the TLT is only 50% of the charging
voltage. Multi-stage TLTs are required for voltage multiplication. The advantage of using
a PFN, however, is that a small fall time of the voltage pulse will improve the energy transfer
efficiency and consequently reduce the electrode wear.
Various kinds of switch-mode converters can be used to charge the PFN or a high-
voltage capacitor [10,11]. The three-step resonant charging scheme together with the LCR
trigger method may be the most robust corona circuitry [12,13]. Shorter pulse duration
corresponds to smaller energy dissipation inside streamer channels. Industrial corona plasma
generation requires ultra-short pulsed power systems with low output impedance.
3. Multiple-switch pulsed power circuit topologies

Figure 2 shows an example of a multiple-switch circuit topology. Two identical triggered
spark-gap switches S1 and S2 are used. At the input side of the TLT, the two cables are inter-
locked via the two switches and two PFNs. At the output side, the two circuits correspond
to in series and in parallel outputs, respectively. When the two switches are closed, the two
cables become in parallel at the input side.

263
As discussed for the circuit in Fig. 1, after pulsed energization, the DC charging unit
resonantly charges the reactor and the capacitor Cdc again via two inductors Ldc-i and Ldc–2.
Whenever one of the switches S1 and S2 closes, the corresponding PFN discharges via both
cables. As a result, induced current inside the TLT will overcharge the another switch. And
then, the two switches close simultaneously. No special trigger circuits are required. A
simple LCR trigger circuit can be used to trigger-one of the switches for up to a pulse
repetition rate of 1000 pulses per second (pps) [12,13]. In comparison to the circuit in Fig.
1, these two new circuits produce an identical output power. The switching duty for each
switch, however, is reduced by a factor of two. The switch lifetime would be improved by a
factor of four because the electrode erosion rate is almost proportional to the square of the
switching current [14].
Figure 2. Two multiple PFNs and switch topologies. The two stage-TLTs are connected in parallel at its
input side (0.5Z0). At the output side, the TLT are connected in series for circuit (a), and in parallel
(0.5Z0) for circuit (b), respectively.
In fact, the circuit principle shown in Fig.2 can be used for any number of switches. As
an example, Fig.3 shows three circuit diagrams with four switches and a four-stage TLT.
Although the three circuits give three identical output power, circuits (a) and (b)
respectively lead to voltage or current multiplication, whereas circuit (c) produces both
voltage and current multiplication. Whenever one of the switches closes, the induced
current inside the other cables will overcharge the corresponding spark-gap switches. As
a result, all switches close simultaneously. The four switches equally share the output
power. TLTs are used for impedance matching and synchronization of the four switches.
In comparison with the circuit in Fig.l, the switch lifetime can be improved by a factor of
16 because the switching current for each switch is reduced by a factor of four. An increased
number of switches and reduced output impedance are the basics for scaling up corona
plasma power. As shown in Fig.3, series and parallel connection at the output side of TLTs
can be used to match various corona plasma generations.
The Marx pulse generator has been one of the most popular power sources for various
kinds of pulsed power applications. Increasing the output voltage via a series of multiple
switches reaches high pulsed-power levels. The generator, however, cannot easily be used

264
for large pulsed current generation, which in contrast needs the use of multiple switches in
parallel. With regard to the circuit principle shown in Fig.3, one may easily conclude that
Marx generator indeed can be as the PFN in those circuit topologies. TLTs can be used to
synchronize multiple Marx generators. Figure 4 shows an example with two Marx generators
and a TLT. Each Marx generator consists of two PFNs and two switches. As discussed
before, the four switches close simultaneously, and equally share the output power. One
advantage of integrating the Marx generator with the TLT is that the number of TLT
cables can be significantly reduced.
Figure 3. A circuit topology with four PFNs, four switches and a four-stage TLT. The TLT is connected
in parallel at the input side (0.25 Z0). For circuits a, b, and c, the output impedances are 4Z0, 0.25Z0, and
Z0, respectively.
4. Discussions and future perspectives

Using a TLT and multiple spark-gap switches, cost effective industrial corona plasma
systems can be realized. Table 1 summarizes the main characteristics of the presented
circuits. Increasing the number of switches, one can reduce the electrode erosion rate
approximately to 10–10–10–9 cm3/shot for large pulsed power generation. Pulsed circuit

265
topologies consist of a single and/or multiple circuit elements as shown in Fig.5. For
Zin=Zout (=Zo or ≠ Zo), the TLT is indeed similar to a single cable. For Zin ≠ Zout, the layout of
the TLT can be various forms. The application of inter-connected lines among multiple
circuit elements results in novel circuit topologies.
Figure 4. A circuit topology with four PFNs, four switches and a two-stage TLT. The TLT is connected
in parallel at both input and output sides (0.5 Z0).
Table 1. Characteristics of various circuit topologies
In order to verify experimentally these novel circuit topologies, an experimental set-up

was constructed with three self-triggered spark-gap switches and a three-stage TLT. The
three-stage TLT is constructed with three 2 m long cables (50 Ω) together with magnetic
cores placed on the outside of each cable. At the output side, the three cables are connected
in series, which gives an output impedance of 150 Ω. At the input side, the three cables are
connected in parallel when the three switches are closed, which gives an input impedance
of 16.7 Ω. Three identical 10 nF capacitors are charged simultaneously by a DC power
source. The three spark-gap switches are constructed by using six spherical electrodes
with a diameter of 10 mm. The spark-gap distances are about 1 mm. All of them are used in
air. When the average electric field between the two spark-gap electrodes is increased to
about 30 kV/cm, the three switches close simultaneously. Experiments were carried out
under a pulse repetition rate of 0.5 pps. Experiments show that the three spark-gap switches
are simultaneously closed within 7–10 ns. Detailed discussions are reported elsewhere
[15]. Based on the same principles as discussed above, TLT is also used to synchronize
multiple transient plasmas [16], where the spark-gap switches are used as transient plasma
reactors. Figure 6 shows a typical result with two plasma reactors or two spark switches.
Two plasma reactors or two spark gap switches are synchronized within 2–4 ns.

266
Figure 5. Circuit element.
Figure 6. Typical discharging current

waveforms for two pulsed plasmas.
References
[1] Siemens A.G., 1932, GB375865.
[2] Lodge-Cottrell Limted, 1933, GB386560.
[3] Chang J.S., 2003, Journal of Electrostatics, 57, 3–4, 273–291.
[4] Van Veldhuizen E.M., 2000, Electrical Discharges for Environmental Purposes: Fundamentals
and Applications. Nova Science Publishers, Inc., New York 11743.
[5] Hosokawa S., Sonoyama K., and Yamamoto T., 2001, Proc. of the Third Int. Symp. on Non-
Thermal Plasma Technology for Pollution Control, Korea, p1 53.
[6] Kim Y.S., Choi S.H., Kim J.I., Kim T.H., Yoo J.S., Pack M.S., and Jang G.H., 2001, Proc. of
the Third Int. Symp. on Non-Thermal Plasma Tech. for Pollution Control, Korea, p1 59.
[7] Oh J.S., Cho M.H., Ko I.S., Namkung W., Jang G.H., 1997, 11th IEEE Pulsed Power Conf,
p1091.
[8] Zhao J., Wang B., Ren X., and Zhu Z., 2002, Chinese Academy of Engineering, 4, 74–78.
[9] Yan K., 2001, Corona Plasma Generation, PhD Thesis, Eindhoven University of Technology,
The Netherlands, http://alexandria.tue.nl/extra2/200142Q96.pdf.
[10] Smulders H.W.M., van heesch E.J.M., and van Paasen S.V.B., 1998, Trans IEEE on Plasma
ScL, 26,1476–1484.
[11] Yan K., van Heesch E.J.M., Pemen A.J.M., Huijbrechts P.A.H.J., van Compel F.M., van
Leuken H., and Matyas Z., 2002, IEEE Trans on Industry Applications, 38, 866–872.
[12] Yan K., van Heesch E.J.M., Pemen A.J.M., Huijbrechts P.A.H.J., and van der Laan P.C.T.,
2001, Review of Scientific Instruments, 72, 5, 2443–2447.
[13] Yan K., van Heesch E.J.M., Nair S.A., and Pemen A.J.M., 2003, Journal of Electrostatics, 57,
29–33.
[14] Donaldson A.L., Engel T.G., and Kristiansen M., 1989, IEEE Trans, on Mag., 25, 138–141.
[15] Yan K., Smulders H.W.M., Wouters P.A.A.F., Kapora S., Nair S.A., van Heesch E.J.M., van
der Laan P.C.T., and Pemen A.J.M., 2003, Journal of Electrostatics, 58, 221–228.
[16] Yan K., Nair S.A., Hooijmans M., van Heesch E.J.M., and Pemen A.J.M., 2003, Macro Review,
16, 1, 226–230, The Review of Japan Macro-Engineers Society.

267
Martian Regolith Simulant Particle Charging

Experiments at Low Pressures in the Presence of Corona
Fields
Frank B Gross
Electrical and Computer Engineering
FAMU-FSU College of Engineering
Tallahassee, Florida 32310
Abstract. NASA maintains an ongoing interest in Martian dust particle charging as it impacts
future Mars missions. Particle charging can be used in experiments to help gather information
about Martian saltation, tribo-electric charging, and discharge. Particle charging and adhesion
can also become a nuisance in future Mars missions. The Martian soil has been simulated by
Allen, Jager, Morris, Lindstrom, Lindstrom, and Lockwood [1] and has been called the JSC
Mars-1 Martian Regolith Simulant. Previous tribocharging experiments have been conducted
with the simulant on a few materials by Gross and Grek [2], Gross, Grek, Calle, and Lee [3],
and Gross [4].
In order to collect more information about the extent of dust particle charging at Martian
atmospheric pressures with new materials, experiments were performed in a 7 Torr
environment testing the tribocharging of the JSC Mars-1 Martian Regolith Simulant. An
apparatus was devised that dropped the simulant down a deflection board whose surfaces
were coated with Rulon, Lexan, Teflon, G-10, Acrylic, and Acetate. The particles, under the
force of gravity, bounced, rolled, slid, and were deflected until falling into a Faraday cup for
charge measurement. Additionally, the particles were allowed to fall through either an AC,
+DC, or –DC corona field. The corona fields are intended to test the effects of bombarding
the simulant with positive and/or negative ions to test the enhancement or suppression of
tribo-charging. The ultimate goal of the experiments was to compare which materials of the
set were more electrophobic or electrophilic under the conditions tested. A final graphical
and tabular comparison is given.
1. Introduction
Future NASA Mars missions will rely on knowledge of the electrostatic conditions and
their impact on various candidate materials which might be used with various man-made
devices.
Under previous research studies [2]–[4], 7 Torr pressure experiments were conducted, to
test the contact charging between the Martian soil simulant and the materials of copper,
glass and acetate. The purpose of this recent research study was to test a new group of
materials under corona conditions and 7 Torr pressures to determine which materials were
more or less susceptible to tribo-electric charging in the presence of corona fields. We
evaluated tribo-charging with flouroplastics (Rulon and Teflon), a polycarbonate (Lexan),

268
a glass and carbon fiber laminate (G-10), acrylic, and acetate. We also created the option to
drop the Martian soil simulant through AC or +/–DC corona fields to test the enhancement
or suppression of tribo-charging.
2. Experiment description
2.1 JSC Mars-1 Martian Regolith Simulant

The Martian soil simulant was formulated by the efforts of Allen, Jager, Morris, Lindstrom,
Lindstrom, and Lockwood [1]. The chemical composition of the simulant is shown in table
1 below. The grain size distribution by wt% is given in table 2 below.
Table 2.
JSC Mars–1 Grain Size Distrubution
Table 1.
Chemical Composition of
JSC Mars-1 Simulant
2.2 Apparatus and Procedure

All experiments were conducted under 7 Torr pressures with air. The Martian simulant was
dropped down a particle deflection board whose surface was covered with the various
materials mentioned above. Additionally the particles were dropped through AC, or +/–DC
corona fields. The simulant, deflection boards, and corona field sources were all housed in
a large vacuum bell charge.
We chose to simulate rolling, sliding and impacting by devising a deflection board.
Figure 1 shows a depiction of the deflection board. The surface materials were cut to fit each
slide. The surface dimensions for each slide were 19×38 mm.
Each deflection board is 88mm×340mm in size and contains 16 slides. Each slide is
tilted at a 45° angle. The boards are hung vertically in the bell jar underneath a particle
dropper. The particle dropper is remotely controlled to release the simulant after the bell jar
has been evacuated to 7 Torr pressures. The particles fall down the deflection board and
into a Faraday cup for charge measurement.
Additionally our experiment included the effects of AC and +/–DC corona on simulant
charging. We are including AC corona to possibly neutralize the net charge on the simulant
because AC corona has both positive and negative ion flow. We also have added +/–DC

269
corona because we can produce either positive or negative ions by reversing the polarity. It
was hoped that the +/- ions would enhance simulant charging. In both cases, the simulant is
dropped directly through the field before bouncing down the deflection board.
Figure 1. Typical deflection board apparatus
The exact description of the AC and DC corona devices is given in [4]. The AC corona
was produced by two arrays of opposing steel needles. We adjusted the voltage to be about
80% of the breakdown value at low pressures. The AC ion current flow between electrodes
was 2.7 µA. The voltage necessary to produce a corona field was 500 +/- 5% volts.
The DC corona was produced by one array of steel needles with an opposing ground
plane. In this way we could produce either negative or positive ions by switching the
polarity across the needle system. In the DC case our current flow is either an-ion or cat-ion
flow but both species are not flowing simultaneously. We therefore increased the number
of needles to increase the current and consequently the ion density. The typical ion
currents in the DC corona case were .26 µA. Both plus and minus DC currents were produced.
The voltage across our DC ionizer was 700 +/- 5% volts.
Each experiment was performed with 2 grams of the simulant taken from a drying oven.
After neutralizing the simulant with a ion blower, we placed the simulant sample in the
holder and evacuated the bell jar to 7 Torr pressures. We next applied the appropriate
corona and dropped the simulant down the deflection board coated with the material
under evaluation. The accumulated charge was measured in a Faraday cup and normalized
by 2 grams to get charge per gram values.

270
Figures 2–5 below show the average charge accumulated on the simulant due to each
deflection board under the corona conditions indicated. The error bars indicate the standard
deviation in the measured data. While trying to keep moisture out of the simulant through
a drying oven some moisture was reabsorbed by the simulant before the experiment was
completed and therefore the total charge accumulated was skewed more negatively by the
presence of moisture. The materials are ordered from least charged to most charged and the
order is different for each case.
Figure 6 shows the percentage increase or decrease in charging for each of the three
corona cases. The error bars indicate the standard deviation across the measured data for
AC, +DC, and “DC. The material which charged the simulant the least was Teflon in the—
DC corona case (-0.18 nC/g). The material which charged the simulant the most was Acrylic
in the AC corona case (-5.86 nC/g).
Figure 2. Average charge/gram accumulated on each sample with no corona
Figure 3. Average charge/gram accumulated on each sample with AC corona

271
Figure 4. Average charge/gram accumulated on each sample with +DC corona
Figure 5. Average charge/gram accumulated on each sample with “DC corona
Figure 6. Percent change in charging relative to no corona for various corona cases.

272
4. Conclusions
We simulated the saltation process on Mars by dropping the JSC Mars-1 Martian Regolith
Simulant down deflection boards at 7 Torr pressures. The deflection boards were coated
with Teflon, Rulon, Acetate, Lexan, G-10, or Acrylic. The simulant was dropped down the
boards with and without AC, +DC, and –DC corona fields being present. The corona fields
were created with a system of needles. The typical ion current was 2.7 µA for AC and
0.26µA for +/- DC. The particle charging for each corona case was measured against charge
values with no corona present and percent changes were recorded. The presence of moisture,
due to re-absorption as the simulant was moved from the drying oven, caused charge
measurements to be skewed more negatively. The Teflon was the most electro-phobic
while the Acrylic was most electro-philic. Rulon and Teflon charging was either increased
or decreased due to the presence of the various corona fields. G-10 charging was diminished
with all corona cases. Lexan, Acrylic, and Acetate charging was increased in all corona
cases.
5. References
[1] Allen, C.C., Jager, K.M., Morris, R.V., Lindstrom, D.J., Lindstrom, M.M., and Lockwood, J.P.
(1998) Martian soil simulant available for scientific, educational study, EOS, 79, 405–409.
[2] Gross, F.B., S. Grek, “JSC Mars-1 Martian Regolith Simulant Particle Charging Experiments in
a Low Pressure Environment-Phase I”, Kennedy Space Center/FAMU-FSU College of
Engineering, Contract No. NAG 10–0264, Final Report, October 2000.
[3] Gross, F.B., S. Grek, C. Calle, R. Lee, “JSC Mars-1 Regolith Simulant Particle Charging
Experiments in a Low Pressure Environment”, Journal of Electrostatics, Vol. 53, Pages 257–266,
2001.
[4] Gross, F.B., “JSC Mars-1 Martian Regolith Simulant Particle-Charging Experiments in the
Presence of AC and DC Corona Fields”, Journal of Electrostatics, Vol 58, pp 147–156, 2003.

273
Influence of an insulating flat plate on a DC surface

corona discharge at various air relative humidities
Christophe Louste, Eric Moreau, Gérard Touchard
LEA, UMR 6609 CNRS, Groupe électrofluidodynamique, Université de
Poitiers,
Téléport 2, BD Marie et Pierre Curie, BP 30179,86962 Futuroscope—
Poitiers—France
Abstract. Previous work has shown that a DC surface corona discharge may be produced on
an adapted insulating flat plate. The discharge is generated by a high voltage applied between
two parallel wire electrodes placed close to the surface. This DC surface corona discharge
generates an ionic wind in the vicinity of the surface. It may be used to modify the local
physical properties of the air. This electroaerodynamic effect is utilized to modify an air flow
around an obstacle.
Recent work has proved that different surface discharge regimes like “generalised glow” or
“high spot type” can produce an efficient electroaerodynamic effect. Moreover, the plasma
sheet created at the vicinity of the surface might change the air mechanical properties of the
interface like air viscosity and boundary conditions.
In this experimental study, the electrical properties of the discharge are examined. The
experiments bring out the essential part of the insulating material. Indeed, without surface,
the DC corona discharge is unstable. With an insulating flat plate the discharge is more
stable. Experiments test several insulating materials at different air relative humidity. They
show that the electrical properties of the discharge depend strongly on these two parameters.
1. Introduction
In previous works [1–3–4], experiments have shown that a DC surface corona discharge
may be used to produce a mechanical effect on a fluid flow around an obstacle such as a flat
plate. Consequently, an electroaerodynamic actuator has been developing in our laboratory.
In this actuator, a surface corona discharge is used to produce a mechanical effect on the air.
Its main advantage is that it converts electric energy into kinetic energy without moving
mechanical parts. It has already been proved that it can produce a wind of few meters per
second very close to a surface. So, this actuator is particularly well suited to control air flow
around obstacles or in drag reduction for instance.
Several experimental setups are proposed to produce the surface corona discharge [5–
9]. For instance, Artana et al. uses a wire anode in copper and a strip cathode in aluminium
flushed in a surface in PMMA [4,8]. Roth et al. places several wire electrodes on the both
side of an insulating flat plate [9]. In the experimental setup presented in this article, a pair
of wire electrodes in copper is flushed mounted at the surface of an insulating surface. As
has been previously shown [3], there is a direct relationship between the mechanical effect
and the ionic wind when the velocity of the air stream is less than 20m/s. However the ionic

274
wind characteristics depend on the electrical setup, on the electrode configuration and on
the properties of the ambient gas.
Recent work in our laboratory have shown that when a high voltage is applied between
two wire electrodes; different regimes of surface discharge can be produced. Each regime
has specific electroaerodynamic effects that are studied in our laboratory according to two
approaches. The first one concerns the mechanical effects of the discharge on an air flow
[1–4,10]. The second one concentrates on electrical properties of the discharge. Since there
exists a relation between the ionic wind velocity and the current intensity, this second
approach consist of investigating the influence of the geometrical, electrical and ambient
gas parameters on the discharge current intensity of the discharge. The present study deals
with the second approach and more exactly with the influence of the insulating material
used for producing the surface discharge at various relative air humidity.
2. DC surface discharge in air

Previous work have already shown that with two wire electrodes, five regimes of discharge
can be obtained. They have been previously described in [2–3]. The first regime is the “spot
type” regime. In this regime, only few visible luminous points are observed on the electrodes
and a very low current is measured. When the electric field increases, a thin sheet of blue
ionized air appears between the two electrodes. This blue sheet is very close to the surface.
This is the “generalized glow discharge” regime. The third regime is named “high spot
type” regime. There is no visible plasma sheet. However, a high current may be obtained in
this regime. Many luminous points are visible on two wire electrodes. With a higher electric
field, the filamentary regime is reached. In this regime, the whole current is concentrated in
some filaments which join the two electrodes. The spark regime is the last regime. All these
regimes could be obtained on a PMMA surface. However the ionic wind is too low in the
spot regime to produce a perceptible mechanical effect. The filamentary and the spark
regime are too unstable to be used. So, the “generalized glow discharge” and the “high spot
type” regime are the two only regimes that are used in our actuator. The main inconvenience
of these two corona regimes is that the current intensity depends on the relative humidity of
air.
3. Experimental setup
Figure 1 shows the experimental setup. In order to produce a corona discharge a high
voltage is applied between two parallel electrodes. The anode and the cathode are both a
wire electrode in copper. They have a diameter of 2 mm and a length of 30 cm. The
electrode gap is 4 cm. The insulating plates are cut into rectangular shape having a length
of 30 cm, a width of 10 cm and a thick of 5 mm. 3 insulating materials have been tested:
PMMA, ordinary glass and Pyrex. Surfaces are always cleaned with n-heptane before
experiments. A DC HV power supply DELL (±40 kV, 3.75 mA) is used to obtain a negative
potential from—6 kV to -20 kV at the cathode. A TREK HV power supply maintains a high
positive voltage of +20 kV at the anode. The electrical potential difference is indicated
with a precision of 0.1 kV. Measurements of the discharge current are performed using the
Trek current sensor (DC accuracy ±10µA, output noise 40µA, bandwidth ≈10 kHz). To
control the relative humidity of air, the insulating flat plate is placed into an hermetic box
in PMMA at atmospheric pressure.

275
Figure 1. Side view of the experimental setup
4. Results
This work is the second part of a study on the DC corona discharge established between two
wire electrodes with the same diameter flush mounted on the surface of on an insulating flat
plate in atmospheric conditions. In the first part [11], experiments showed that the insulating
surface stabilized the discharge. Indeed, when a high voltage is applied between two parallel
wires electrodes of 2 mm placed in air (without insulating surface), a corona discharge
could be obtained in certain humidity conditions. But this discharge is very unstable.
When an insulating surface is placed close to the electrodes, the discharge is stabilized. In
this second part, we focus on the discharge current for different insulating materials. Three
insulating materials are tested: PMMA, ordinary glass and Pyrex.
4.1. Surface in PMMA

Figure 2 presents a current -time evolution when a high potential of -20kV is applied at the
cathode and a high potential of 20kV is applied at the anode at t=0s. The slew rate of the
Trek power supply connected at the anode is 380 V/JLIS. The current intensity is measured
by the sensor for 100s. A high spot discharge is produced. The discharge current increases
abruptly (about 1 ms) and reaches a steady state value (0.28 mA) with no overtaking. The
current intensity is very stable. It is an important characteristic of the “high spot type”
discharge regime.
In Figure 3, the current is measured for various voltages at different relative air humidities.
The current is presented in current per meter (discharge current divided by the electrode
length) as a function of the applied electric field (potential difference divided by the
electrode gap). When the electric field is lower than 6 kV/cm, there is no discharge. From 6
kV/cm to 8 kV/cm the current increases linearly with the electric field. The current is low
but some luminous points are visible on the electrodes. It is the “spot type” regime. When
the electric field increases up to 8 kV/cm, a high spot discharge light up. A lot of points are
visible. The current versus electric field remains nearly linear but the slope is clearly more
important. When the relative air humidity is ≤20%, high current value may be obtained and
the discharge is stable the spark regime is reached at about 9.5 kV/cm. On the other hand,
when the air humidity is equal to 30%, sparks appear from 8.6 kV/cm.
Furthermore, for relative air humidity ≥40%, it is impossible to produce a stable discharge
without spark or filaments.

276
Figure 2. Current versus time on a PMMA Figure 3. Current versus electric field for different
surface. air relative humidity values on a PMMA surface.
4.2. Surface in ordinary glass and in Pyrex

Figure 4 and Figure 5 present a typical current time evolution when a potential of -12 kV is
applied on the cathode and a potential of 20 kV is suddenly applied to the anode at t=0s. In
figure 4, the surface is in ordinary glass. The current increases abruptly (0.55 mA) and then
decreases until reaching an asymptotic steady state in about 100 s. In figure 5, the flat plate
is Pyrex. The steady state value of current seems to be reached more quickly in about 20s.
However, the current is clearly more stable on the common glass surface (figure 4) than on
the Pyrex one (figure 5).
Figure 4. Current versus time on a common Figure 5. Current versus time on Pyrex
glass surface. surface.

277
Figure 6 presents a current-electric field characteristic on a Pyrex surface in atmospheric

conditions when the air humidity is equal to 40 %. In figure 6, on a Pyrex surface, two
corona discharge regimes are observed. When the voltage is up to 25 kV, a first regime of
discharge is obtained (Figure 5, regime 1). This regime is also called “spot type regime”.
When the voltage is increased, a thin sheet of blue ionised air between the two electrodes
may be observed very close to the surface (regime 2, Figure 5). It is the “generalised glow
discharge” regime. In this regime, the current is about six times higher than in the “spot
type” regime. A particular point C can also be noticed. It is the breakdown point. Figure 7
shows the current versus electric field at various air relative humidity on a Pyrex surface.
Figure 8 presents the same evolutions on an ordinary glass surface. The maximum current
discharge intensity is obtained when the electric field approaches the disruptive field
value. The maximum current intensity increases when the air relative humidity increases.
This evolution can be observed as well as on a glass or on a Pyrex surface. The disruptive
field value seems to be constant in spite of the air relative humidity variations.
Figure 6. Current-voltage characteristic on a Figure 7. Current versus electric field for

Pyrex surface in atmospheric conditions (40% different values of air relative humidity on
of air relative humidity). a Pyrex surface.
Figure 8. Current versus electric field for Figure 9. Maximum discharge current
different relative humidity of air values on versus relative humidity of air values for
an ordinary glass surface. three different materials.

278
4.3. Comparison between the three material behaviors.

Figure 9 presents the maximum discharge current at different air relative humidity. As a
rule, the steady state current intensity of the glow discharge increases nearly linearly with
the relative humidity of air on the glass surface as on the Pyrex surface (figure 7–8).
Conversely, the maximum current value decreases when the relative humidity increases on
the PMMA surface. These dielectrics have different properties of surface water adsorption.
This factor is probably the most important reason why the different dielectrics used show
the different humidity effects. But, it is very difficult to prove the direct relationship between
the discharge current and surface water adsorption because measuring surface water
adsorption is complex in practice. The surface roughness was also suspected to have an
influence on the initiation of the corona discharge but its effect is probably less important
than the water adsorption because the corona discharge could also be obtained on a rugged
surface, such as wood or paper, or on a smooth surface such as glass.
5. Conclusion
The aim of this study was to obtain an increase in the discharge current and a stable
discharge even with a high level of relative humidity. As has been observed, the surface
material and the relative humidity have a great influence on the current intensity of a dc
surface corona discharge. Stable discharges have been obtained up to 80% of air relative
humidity on a glass surface. The most important part of this work was to show that the
current intensity increases when the relative humidity increases on ordinary glass or on
Pyrex surfaces, but decreases with relative humidity on a PMMA surface. The properties of
surface water adsorption were suspected to be the most important factor in explaining the
relation between the current intensity of the discharge and the relative humidity of air. To
confirm this hypothesis, new experiments will be made. In these experiments, many glasses
with well known concentrations of alumina and boron will be tested.
This study shows that the current intensity of a “high spot type” discharge regime is more
stable than the “generalised glow” one. This is the second part of a more complete
experimental study. In the near future, many geometrical parameters will be tested. Others
insulating material will be tried in order to increase the current intensity.
References
[1] Léger L., Moreau E., Artana G., Touchard G., 2001 J. of Electrostatics, vol. 51–52, 300–306.
[2] Loeb L., “Electrical Coronas”, 1965, Univ. Of California Press, Berkley.
[3] Léger L, Moreau E, Touchard G, 2001 Trans, on Industry Applications, 38 1478–1485.
[4] Artana G, DiPrimio G., Desimone G., Moreau E., Touchard G., 2001 AIAA # 3056.
[5] El-Khabiry S., Colver G, 1997 Physics of fluids, vol. 9, 587–599
[6] Colver G, El-Khabiry S, 1999 IEEE trans. on industriy appl., vol. 35, 387–394.
[7] Corke T.C., Matlis E., 2000 AIAA , #2323.
[8] Artana G., D’Adamo J., Leger L., Moreau E., Touchard G, 2001, AIAA 0351.
[9] Roth J.R., Sherman D, Wilkinson S.P., 1998, AIAA #0328.
[10] Labergue A, Leger L, Moreau E, Touchard G, Bonnet J.P, 2003, Electrostatics
[11] Louste C. Moreau E., Touchard G., 2001 CEIDP, 822–826.

279
Detachment and reattachmeet of a low velocity airflow

along an inclined wall actively controlled by a low
frequency square wave corona discharge
A Labergue, L Leger, E Moreau, G Touchard, J P Bonnet

Laboratoire d’Etudes Aérodynamiques, Groupe Electrofluidodynamique,
UMR 6609 CNRS, Université de Poitiers,
Téléport 2, Bd Marie et Pierre Curie, BP 30179, 86962 Futuroscope
Abstract. This work analyses the ability of an electrohydrodynamic actuator to modify the
dynamic of the detachment and the reattachment of an airflow. The device considered consists
in a pulse square wave corona discharge established between two electrodes flush mounted on
an inclined wall. In this paper, visualizations of low velocity airflows (0.4 m/s) and measurements
by Particle Image Velocimetry at 1.8 m/s are presented. They show that the discharge can
accelerate and reattach the airflow along the wall when this one is naturally detached. More, the
dynamic of the airflow may be controled by the discharge frequency.
1. Introduction
For several years, the ability of actively control airflow along a wall of an obstacle by a DC
surface corona discharge has been studied in our laboratory [1–4] and others [5–6]. This is
electroaerodynamics. In practice, positive ions are produced at the anode and drift to the
cathode under the effect of coulomb forces. Then they exchange momentum with neutral
particles and induces an airflow tangential to the wall. The induced fluid motion is called
ionic wind. The goal of this electrohydrodynamic actuator is to modify the airflow profile
within the boundary layer with the help of the ionic wind induced by the corona discharge.
The long-term objective of the study presented in this paper is to reduce the noise
generated by the primary and the secondary airflows at the propulsion nozzle of a supersonic
plane motor. Because it is expected that the noise is mainly due to the shearing region
between the low velocity primary airflow and the high velocity secondary one (Figure 1a),
the proposed method is to ameliorate the mix between the two airflows in order to generate
a turbulent zone [7]. To enhance the mix, a push-pull effect is expected. During the first
phase, the primary flux is attached to the wall since the secondary airflow is actively
detached by the corona discharge (Figure 1b). The process is inverted during the second
phase (Figure 1c).
Previous experiments have been performed in our laboratory with the help of a blowing
device and showed interesting results [8]. In the present paper, a preliminary experimental
study at low velocity (up to 1.8 m/s) is presented. It considers a naturally detached airflow
along an inclined wall. Our objective is then to determine the reattachement time when the
discharge is established and the detachment time when it is turned off. First, the experimental
setup will be presented. Secondly, flow visualisations and P.I.V. (Particle Image Velocimetry)

280
measurements will allow determining the detachment time and the reattachment time.
Then results concerning the excitation frequency will be discussed.
Figure 1. Schematic representation of the push-pull effect.
Figure 2 shows the profile used for our study. It represents schematically the shape of the
nozzle. Below and above the profile, there are respectively the primary and the secondary
airflows. In this study, the two airflows have the same velocity U0. The electrodes are flush
mounted on the inclined wall as shown in this figure.
The anode is a 0.7 mm diameter wire placed inside a groove, 4 cm downstream of the first
trailing edge. The cathode is a 2 mm diameter wire in a groove. The distance between both
electrodes is 4 cm. A DC high voltage (HV) generated by a power supply DELL (40 kV, 3.75
mA) is applied to the anode. The cathode is connected to a HV amplifier TREK (± 20 kV, 10
mA). With such a distance between both electrodes, the electric field must be about 8 kV/
cm in order to establish the corona discharge. Therefore, the anode is always at +15 kV and
–17 kV are applied to the cathode when the discharge must be established. An accurate
description of the discharge is given in [9].
Figure 2. Side view of the profile.
3. Flow visualizations
The profile is placed in a wind tunnel (free airstream velocity 0≤U0≤8 m/s, test cross-section
1.2×1.2 m2) and the airflow around the obstacle is visualized with an oil smoke filament
illuminated by a two dimensional laser sheet. The visualizations are undertaken with an
airflow velocity U0=0.4 m/s.

281
The airflow reattachment is discribed by four successives steps in Figure 3. At the trailing
edge 1, on the first picture (t=0), the smoke wire is horizontal and there is an important wake
above the inclined wall. This means that the airflow is detached. At t=0, the electric field is
applied. Then the smoke wire inclines itself progressively along the wall. At the last picture
(t=240 ms), it leans against the wall. The airflow is completely reattached. The study of the
detachment shows the same kind of pictures than the reattachment. In this case, at t=0, the
discharge is turned off and the smoke wire detaches itself progressively until reaching an
horizontal position. These visualizations show clearly the dynamics of the reattachment
and the detachment. We can conclude that the discharge has an important effect on the
airflow because it is possible to reattach an airflow which is naturally detached. However,
this set of experiments can not give us the time of the reattachment and detachment with a
good accuracy. Indeed, it is very difficult to determine when begins and finishs the
reattachment and the detachment of the airflow. That is the reason why we are going to
realize P.I.V. measurements in the next part.
Figure 3. Visualizations of the reattachement.
4. Particle Image Velocimetry (P.I.V.)

P.I.V. is a measurement technique which gives the 2D velocity fields from the Fourier
analysis of digital image pairs of smoke particles which follow the airflow and which are
illuminated by a laser sheet. We use the DANTEC system controlled by FlowManager®
software. In our experiments, the time between two image pairs is 2800 µs. The interrogation

282
area is 32×32 pixels with a maximum overlap of 50% (adaptative correlation is used). Each
velocity field is filtered with a peak-validation. Peak validation stores vectors which have
a signal/noise ratio greater than 1.2. These measurements are done in the same wind tunnel.
During all P.I.V. experiments, the flow velocity is U0=1.8 m/s.
4.1. Mean flow fields

Before studying the reattachment and the detachment of the airflow, let us observe the
effect of the discharge on the wake. The two maps (Figure 4) show the areas where velocity
has the same value. The DANTEC system calculates a velocity field each 66 ms (f=15 Hz).
In fact, we have 15 digital image pairs per second. In our case, the airflow is stored for 60 s,
corresponding to 900 (15×60) image pairs. An average of these 900 image pairs is performed
in order to obtain the mean flow field in the case of discharge off (Figure 4a) and in the case
the discharge on (Figure 4b).
When the discharge is off, the velocity in the wake is globally low. Therefore, the flow
is detached. When the discharge acts, the velocity increases in the wake and the flow is
parallel to the inclined wall. The flow is then reattached.
Figure 4. Mean flow field without discharge (a) and with discharge (b).
4.2. Instantaneous flow fields.

The objective of this part is to determine the time of the reattachment and detachment. To
reach it, one must use a parameter which varies as a function of these two conditions. Here
we have considered the mean velocity Umean inside an area placed above the inclined wall
(hashed area in Figure 5). In fact, Umean is computed for each instantaneous flow field during
the phase of reattachment and detachment. Results deduced from ten experiments are
presented in Figure 6.
Before t=0, the discharge is off. The velocity Umean≈1.3 m/s in the wake whereas U0=1.8
m/s. At t=0, the discharge is established. Umean increases up to Uo at t≈0.2 s. Umean=U0 at t=0.2
s means that the airflow is reattached. The reattachment time is then tr≈200 ms. At t=0.53 s,
the discharge is put off. Umean decreases to its initial value 1.3 m/s at t=0.86 s. In consequence,
the detachment time is td≈330 ms. This time is higher than the reattachment time.

283
Let us introduce the adimentional time t*=t×U0/L (L=20 cm is the length of the inclined
wall). We obtain tr*=1.8 and td*=2.97. In this case, U0/L correspond to the convection time
by U0 of the eddy which is created during the reattachement [8]. If tr* and td* are constants
and known for each velocity [8], then tr and td may be calculated whatever U0 and L. From
tr and td, we can deduce the maximum frequency dischage fmax which is:
fmax=1/(tr+td)=U0/(L×(tr*+td*)) (1)
where the discharge is applied during tr and it is off during td (because tr≠td the duty of the
square wave discharge is tr/(td+tr)). fmax is the maximum frequency for which the airflow has
the time to reattach and detach fully. In our case, fmax=17(0.33+0.2)≈1.9 Hz.
Figure 5. Area used for Figure 6. Mean velocity Umean and

Umean Computation. discharge voltage for a whole
reattachment and detachment cycle
versus time at U0=1.8 m/s.
5. Airflow excited by a pulse discharge

Experiments have been carried out to study the flow dynamic with a pulse discharge with
frequencies up to 30 Hz, and for a constant free airflow velocity U0=0.4 m/s. Figure 7 shows
two flows for two different frequencies. In Figure 7a, the frequency is 6.5 Hz. In Figure 7b,
the frequency is 14 Hz. The two pictures show, inside the wake downstream the trailing
edge 2, some small eddies which leave this trailing edge. Their number increases with the
discharge frequency.
Figure 7. Visualizations of airflows controlled by a pulse discharge with a frequency of 6.5 Hz (a) and
with a frequency of 14 Hz (b).

284
Figure 8 shows the number of eddies per second as a function of the discharge frequency.
It shows that each reattachment induces an eddy, for frequencies up to 14 Hz. Indeed, for
one frequency f0, we count f0 eddies during one second. Beyond 14 Hz, it is impossible to
count the eddies number because they are not detached from each other. In this set of
experiments U0 is constant and from (1), fmax=0.4/(0.2×(1.8+ 2.97))=0.4 Hz. For fd≤0.4 Hz,
the airflow has enough time to be fully reattached or detached. But with a such frequency,
the airflow stays reattached and detached during a certain time. If fe≥0.4Hz, the airflow has
not the time to be fully reattached or detached. In this case, the discharge time is too short
to allow the fully reattachement. And the time of discharge off is also too short to allow a
fully detachment. Nevertheless, an eddy is created even if the airflow is never fully reattached
and detached.
Figure 8. Number of eddies versus frequency, UQ=0.4 m/s.
6. Conclusion
P.I.V. measurements confirm visualizations. This two sets of experiments prove that a
corona discharge is able to reattach an airflow which is naturally detached. Results with a
pulse discharge shows that it seems easy to control the airflow reattachment and detachment
with a such device. New experiments with higher velocity airflows will be done soon.
References
[1] Léger L, Moreau E, Artana G, Touchard G 2001 J of Electrostatics 51–52 300–306
[2] Léger L, Moreau E, Touchard G 2001 IEEE Trans, on Industry Applications 38 1478–1485
[3] Léger L, Moreau E, Touchard G 2002 AIAA #2833
[4] Artana G, DiPrimio G, Moreau E, Touchard G 2001 AIAA #3056
[5] Soetomo F, M S 1992 Thesis, Iowa State University
[6] El-Khabiry S, Colver G 1997 Physics of fluids 9 587–599
[7] Bonnet JP, Collin E, Tensi J, Moreau E, Delville J, Touchard G 2002 Brevet français CNRS
0112113
[8] Bonnet JP, Collin E, Tensi J, Moreau E, Touchard G 2002 Premier Collogue National pour la
Recherche Aéronautique sur le Supersonique 258–263
[9] Moreau E, Artana G, Touchard G 2003, Electrostatics

285
Surface corona discharge along an insulating flat plate in

air applied to electrohydrodynamically airflow control:
electrical properties
E Moreau (1), G Artana (2), G Touchard (1)

(1)
Laboratoire d’Etudes Aérodynamiques, UMR 6609 CNRS,
Electrohydrodynamic group, Université de Poitiers, Téléport 2, Bd Curie, BP
30179, 86962 Futuroscope, France.
(2)
Universidad de Buenos Aires, Facultad de Ingeniería, Paseo Colón 850, 1063
Buenos Aires, Argentina.
Abstract. Several studies have shown that a surface corona discharge may be used as an air-
moving actuator in order to control the airflow around an obstacle, such as an airfoil to enhance
lift or to reduce the drag for example. For some years, our laboratory has been working on this
subject, especially in the case of a DC corona discharge. Although efficient aerodynamic
effects have been observed, it is sometimes difficult to control the discharge properties.
Consequently, the present paper deals with experimental work concerning the electrical properties
of a DC surface corona discharge established between two wire electrodes flush mounted on
the wall of a PMMA insulating plate. This study shows that (1) Several discharge regimes
may be obtained as a function of the applied electric field. (2) When the free airstream flows in
the same direction of the ionic wind, the discharge is more stable and its current increases with
the airstream velocity U0. (3) When the gas flows in the opposite direction (Uo < 0), the current
is quite stable. (4) The discharge current is nearly proportional to E/P where E is the electric
field and P the gas pressure. (5) Higher discharge currents may be reached when the negative
polarity is applied to the wire electrode with the upper diameter. (6) The discharge current
decreases when the air humidity (RH) increases. (7) In a range between +20°C and +65°C, the
wall temperature has no influence on the discharge. (8) To obtain a more stable DC corona
discharge, one needs U0 high, RH low and the positive polarity must be applied to the wire
electrode with the lower diameter.
1. Introduction
When a high potential difference is applied between two electrodes, ions are produced
which drift from the injection electrode to the collecting one under Coulomb forces. They
exchange momentum with the neutral fluid particles and induce a fluid motion usually
called ionic wind. In the case of a corona discharge established in ambient air between two
wire electrodes flush mounted on the surface of an insulating material, we have previously
proposed that positive ions were produced at the anode and electrons at the cathode [1–2].
Indeed, very few negative ions are produced at the cathode, and they may be neglected.
Under Coulomb forces, these charges drift to the electrode of opposite sign. Because the
electron mass is negligeable compared to the ion mass, the ionic wind is mainly due to

286
positive ion drift, from the anode to the cathode (Figure 1). For few years, we have been
studying the ability of using this type of actuator to control airflow along the wall of an
obstacle. This is electroaerodynamical control. The goal of this electrohydrodynamic
actuator is to modify the airflow profile within the boundary layer. The main advantage of
this process is that it directly converts electric energy into kinetic energy, without any
moving mechanical part. This process might be used by aeronautic industry to reduce the
drag of a plane or to stabilize the flow in order to avoid unsteadiness which generates
unwanted vibrations, noise and losses.
In previous work [1–5], a great number of experiments such as visualization, particle
imaging velocimetry (PIV) measurements, velocity profile and wall pressure measurements
with different types of obstacles (flat plates, circular cylinders, NACA airfoils) for velocities
up to 30 m/s with associated Re number up to 105, allowed us to show a high acceleration
of the airflow downstream the discharge when the ionic wind acts in the direction of the free
airstream. Indeed, the ionic wind induced by the corona discharge increases the velocity at
close vicinity of the wall, resulting in many cases in an airflow reattachment (Figure 2) or
reduction of the wake size. Although this process is very promising, the main inconvenience
of the DC discharge is that its properties depend highly on the atmospheric conditions of
the ambient air and the electrode geometry. Consequently, the purpose of the present paper
is to study the surface corona discharge as a function of several parameters in order to
enhance its stability. More especially, we observe the influence of several parameters on
the discharge properties. These parameters are the free airstream velocity, the air pressure,
the electrode polarity, the wall temperature and the air humidity.
Figure 1. Schematic representation of the Figure 2. Visualisation of an airflow

electrohydrodynamic actuator. along a flat plate turned off toward the
wall by a DC discharge.
The obstacle is a flat PMMA plate 4 mm thick. The anode is a 0.7-mm diameter copper wire
electrode placed inside a 0.7-mm depth groove, 4 cm downstream from the leading edge.
The cathode is a 2 mm diameter wire placed inside a 2 mm depth groove, 4 cm downstream
from the anode (Figure 1). Electrode length is 22 cm. The DC surface corona discharge is
obtained by application of a high positive potential at the anode (≈+22 kV) and a negative
potential of “10 kV at the cathode. Two DC HV power supplies DELL (±40 kV, 3.75 mA)
are used. In previous work, the wire cathode was sometimes replaced by a plate such as
aluminium foil [4–5].
The wall temperature may be controlled by a flat filament resistor of 150 Watt mounted

287
on the rear side of the flat plate. A temperature sensor (accuracy 0.1°C) is placed on the side
where the discharge is established. The potential difference and the mean value of the
discharge current are indicated by the HV power supplies with a precision of 0.1 kV and 10
µA respectively. More, the alternating component of the discharge current is visualised
with an AC current transformer (ACCT Bergoz, current full scale 1 mA, precision lower than
3 µA, bandwidth of 300 kHz) connected to a TREK oscilloscope. The current transformer
consists of a toroid sensor with inner diameter of 28 mm placed around the positive HV
cable. To modify the free air stream velocity U0, the flat plate is placed in a wind tunnel loop
(U0 up to about 30 m/s). The incidence angle of the obstacle is adjustable and the obstacle
direction may be changed. The velocity U0 is measured with a Pitot tube connected to a
micromanometer. Precision is typically 0.025 m/s. Discharge experiments may be performed
in a chamber where the pressure P is adjustable over a large range (102<P < 105Pa).
3. Electrical properties
3.1. DC Discharge at ambient conditions

To establish the surface corona discharge, a negative potential of -10 kV is applied to the
cathode, and an adjustable positive potential is applied to the anode. Then in ambient air,
positive ions are created at the anode and drift to the cathode with a mean velocity depending
on their mobility. Figure 3 shows the discharge current per unit length I (mA/m) as a
function of the applied electric field E (kV/cm). The current per unit length is the ratio of
the discharge current with the electrode length, (a discharge current of 100 µA with 22 cm
length electrodes corresponds to I ≈0.45 mA/m). In Figure 3, one can see that the current
increases with E=V/d, V being the voltage difference between electrodes and d the distance
separating them. When the electric field between both electrodes increases, several discharge
regimes may be observed. They have been accurately described in [2]. The range of these
different regimes may vary highly as a function of geometric parameters (such as the position
of the electrodes, the distance between the electrodes, the shape of the electrodes, etc.) or
atmospheric parameters (such as ambient air humidity, pressure, etc…). In the experimental
conditions used here (air humidity between 45 % and 55 %), the followings results are
obtained: above the corona-starting voltage (5.25 kV/cm in Figure 3), the first regime is the
“spot type” regime. The discharge is then concentrated in some visible spots on the wire.
The current density is smaller than 0.3 mA/m and the ionic wind is low. When the electric
field is increased, a thin sheet of blue ionized air between the two electrodes is observed
(current values from 0.3 to 0.8 mA/m). For higher currents (from 0.8 to several mA/m), the
“high spot type” discharge is obtained. With the high spot type discharge, one cannot
observe a thin sheet of blue ionized air between the two electrodes. Compared to the
generalized glow discharge, the two main advantages of the high spot type discharge are its
stability and the fact that high current values (and then faster ionic winds [3]) may be
reached. If the electric field increases more, the whole current is concentrated in some
filaments; this is the “filamentary type regime”. Above, sparks appear and the discharge is
very difficult to control. Figure 4 presents a typical discharge current as a function of time
for a mean current of 300 uA (I≈1.4 mA/m and E≈8.6 kV/cm). It shows that there are lots of
current peaks with a magnitude up to 40 µA and a response time τ from 20 µs to 100 µs. A
FFT analysis indicates that the most part of these current peaks have a frequency lower than
10 kHz.

288
Figure 3. Current versus electric field. Figure 4. Discharge current versus time
(U0=0 m/s and E=7.6 kV/cm).
3.2 Influence of the free air stream velocity U0

The local current density in ith direction Ji depends on the local magnitudes of the free
charge density ρc, the ion diffusion coefficient D, the ion mobility µ, the electric field Ei
and the convective velocity ui in ith direction. It may be expressed as:
(1)
In classical electrohydrodynamical theory, the diffusive and convective terms are usually
neglected. This simplification is justified because these terms are largely lesser than the
Coulomb drift term. However in aeronautic applications, convective velocity is high and it
can not be neglected. For instance, in our experiments and considering the ion mobility
µ≈2.0 10–4 m2/Vs, the Coulomb drift velocity µEi is about 150 m/s while ui is comparable to
U0≈30 m/s. Figure 5 presents typical curves of current versus electric field for U0 equal to 0
and 30 m/s. It shows that the discharge current and the corona starting voltage increases
with U0. In Figure 6, square symbols, the electric field E is fixed at 7.5 kV/cm, U0 at 0 m/s
and the current is about 0.4 mA/m. If U0 is increased we can observe that the discharge
current increases almost linearly with the free air stream velocity as predicted by Eq. (1).
Moreover, below 12 m/s the discharge is “high spot” while above 12 m/s it becomes
“glow”. This means that the airflow favours the “glow discharge” in spite of the “high spot”
type discharge. The same behavior is obtained with E equal to 8 and 8.25 kV/cm. In order
to increase the convective term ui, we have inclined down the leading edge of the plate to
an incidence angle α of 10°. As in Figure 5, Figure 7 presents the current-electric field
characteristics for different U0 values. It shows that the slope of the E-I characteristic increases
with U0. Taking into account Eq. 1 and if we consider that the cross-sectional area of the
discharge is constant, then the slope of these characteristics is ρc×µ. Then the slope increase
may be associated to an increase of the space charge density ρc with U0.
Figure 8 presents the current as a function of the free airstream velocity for a equal to 0
and 10 degrees and E=8 kV/cm. As previously indicated by Figure 6 for α=0, it shows that
the discharge current increases with U0. The effect is higher when the plate is inclined
because the convective term ui is higher. On the other hand, if the air stream flows in the
opposite direction to the current, the convective term ui of Eq. 1 should diminish the
current value because it is negative. However when the flat plate is reversed in the wind
tunnel loop, the current does not decrease but increases slightly.

289
This result shows that the influence of the air flow is not restricted to the convective effect.
It indicates that the phenomena of charge generation, charge distribution, charge drift and
convection are strongly coupled. Figure 9 presents the alternative component of i (t) at 30
m/s for α=10°. Compared to Figure 4, Figure 9 shows that the airflow largely increases the
peak repetition and modifies the discharge spectra. This confirms that the airflow modifies
considerably the discharge.
Figure 5. Current versus electric field for 2 Figure 6. Current versus air stream velocity for
air stream velocity values, 3 electric field values.
Figure 7. Current versus electric field for 4 Figure 8. Current versus free air stream velocity
airstream velocities with α=10°. for E=8 kV/cm.
Figure 9. Alternating component of the Figure 10. Current versus electric field for
current (ImeanH”300 µA) versus time for different air pressure values (in 105 Pa).
U0=30 m/s and α=10°.

290
3.3. Influence of the air pressure P

One parameter which may change in aeronautic applications for example is gas pressure.
This is the reason why we studied the discharge behaviour in a large range of pressures
(102<P<105 Pa). For instance, Figure 10 presents the current-electric field characteristics for
different pressure values. It shows that the discharge needs a lower electric field when the
pressure decreases. In fact, as previously observed in the case of a AC discharge [6], it seems
that the current is nearly proportional to E/P. On one hand, when decreasing the pressure
until 104 Pa, the discharge becomes more and more stable and homogeneous. On the other
hand, when the pressure is lower than 104 Pa, the discharge is more and more concentrated
in some visible points on the electrodes such as a filamentary discharge. It is then difficult
to obtain a homogeneous discharge along the surface of the insulating plate.
3.4. Influence of the electrode polarity, the plate temperature and the air humidity
Others parameters have been studied. Briefly, results are as follows: (1) Concerning the
influence of the electrode polarity on the ionic wind velocity, results will be published
very soon. Meanwhile, as in cylindrical configuration [7], the present study has shown that
higher discharge currents (about +20 to +50 %) were reached when the negative polarity
was applied to the wire electrode with the lower diameter (+10 kV at the 2 mm diameter
electrode and ≈-20 kV at the 0.7 mm diameter electrode). This behaviour has been observed
whatever the air pressure. (2) In a range between +20°C to +65°C, the wall temperature has
no influence on the discharge properties. (3) In the case of a PMMA flat plate, the discharge
becomes unstable when the air humidity is higher than 55 %. Below this threshold, the
discharge current decreases when the air humidity increases. More the air humidity promotes
the “glow discharge” in spite of the “high spot type discharge”. On the other hand, other
behaviours may be observed with others insulating materials [8–9].
4. Conclusion
This experimental study has shown that the electrical properties of a DC surface corona
discharge established between 2 flush mounted wire electrodes depend highly on the air
properties, such as its pressure, its velocity and its humidity. In order to obtain a more stable
discharge, one needs a pressure between 104 and 105 Pa, a low humidity (RH< 50 %) and a
high air stream velocity (up to 30 m/s in our case).
References
[1] Léger L, Moreau E, Artana G, Touchard G 2001 J.of Electrostatics 50–51 448–454
[2] Léger L, Moreau E and Touchard G 2002 IEEE Trans.on Ind. Appl. 38 1478–1485
[3] Léger L, Moreau E and Touchard G 2002 1st AIAA Flow Control Conf. paper #2833
[4] Artana G, D’Adamo J, Moreau E, Touchard G 2002 AIAA Journal 40 1773–1779
[5] Artana G, Diprimio G, Desimone G, Moreau E, Touchard G 2001 4th AIAA Weakly ionized
Gases Conf., paper #3056
[6] Moreau E, Léger L and Touchard G 2002 IEEE—CEIDP 272–278
[7] Artana G, Desimone G and Touchard G 1999 Electrostatics Cambridge 147–152
[8] Louste C, Moreau E and Touchard G 2003 Electrostatics Edimbitrg
[9] Louste C, Moreau E and Touchard G 2002 IEEE—CEIDP 822–826

291
Optimized geometry of a corona electrode arrangement

for water ozonization
Ilie Suarasan,1 Roman Morar,1 Lucia Ghizdavu,2 Luciae Dascalescu3
1
High Intensity Electric Fields Research Laboratory, Technical University, 15
C-Daicoviciu Street, 3400 Cluj-Napoca, Romania
2
Chemical Department, Babes-Bolyai University, 3400 Cluj-Napoca, Romania
3
Electronics and Electrostatics Research Unit, LAII-ESIP, UPRES-EA 1219
University Institute of Technology, 4 Varsovie Ave., 16021 Angouleme, France
Abstract. Barrier electric discharges are commonly used for generating the ozone employed
for water treatment. Previous studies demonstrated the feasibility of using corona discharge
for potable or waste water ozonization. The present paper aims at analysing the effect of
corona electrode geometry on the efficiency of ozone generation (amount of ozone/consumed
energy). The experiments were carried out with three types of high-voltage supplies energizing
a test cell consisting of a needle-type corona electrode, facing an horizontal plate (collecting)
electrode, and a Petri dish, containing the treated liquid: potable, mono- or bi-distiled water.
The ozone measurements were correlated to the current-voltage characteristics of the electrode
system, for various values of the following parameters: interelectrode spacings, distance
between adjacent needles of the corona electrode p, high voltage level U. A special device
carrying a current probe was employed for measuring the distribution of corona current
density at the surface of the collecting electrode. The measurements showed that the most
effective ozone generation is obtained for a s/p ratio slightly greater that 1.
1. Introduction
During the last 20 years, various novel methods of ozone generation [1–5] have been
studied as competitors for the classical Siemens water treatment reactors, based on dielectric
barrier discharges [6–9]. Several promising results have been published by those who
investigated the production of active species such as OH, H2O2, and O3 in water and aqueous
solution using the so-called pulsed-streamer corona discharge [10–13]. A combination of
air stripping and pulsed corona has also been tested [14–15]. Other research groups have
studied ozone generation in devices where DC corona is created between an electrode and
the surface of a liquid [16].
In a previous work [17], the authors attempted to evaluate AC corona discharges from
multipoint high-voltage electrodes above a water film as means for direct ozonization of
the respective liquid. The reported results showed that the efficiency of ozone generation
depends on the following factors: (i) number and density of the discharge points of the
corona electrode; (ii) gap length between the discharge points and the surface of the liquid;
(iii) thickness of the liquid layer; (iv) duration of the exposure to the corona discharge; (v)
presence of a dielectric barrier between the collecting electrode and the liquid.
The aim of the present paper is to analyse more in depth the effect of corona electrode

292
geometry on ozone generation, and this for various types of high-voltage supplies: AC, DC
(full-or half-wave rectified). The study was suggested by previously reported results that
show that the efficiency of ozonization is strongly related to the characteristics of the
corona discharge, mainly to the intensity of the corona current. Therefore, an attempt was
made to find the electrode arrangement capable to ensure a high and uniform density of the
corona current in the air-gap above the liquid to be treated.
2. Experimental procedure
The study was carried out on an experimental device composed of an “active” high-voltage
electrode and a grounded plate electrode (Fig. 1). The liquid to be treated (volume: 10 to 70
ml) was contained in a Petri glass vessel of inner diameter Φ=95 mm.
Two types of “active” electrodes were tested. The first consisted of one to three metallic
points, disposed at various separations (p=1–2.5–5–7.5–10 mm) on a cylindrical metallic
support. The second, not presented in the figure, consisted of a multi-point (brush-type)
arrangement, the distance between adjacent spikes being 1 mm. The tungsten wire segments
of 150 ˜m in diameter, the tips of which were employed as coronating points, were attached
to a metallic plate. The distance (s) between the surface of the “treated” liquid and the tips
of the tungsten wire segments could be continuously adjusted from 20 to 40 mm. The
experiments were carried out with the “treated” liquid either connected to the earth (case I)
or at floating potential (case II) with respect to the grounded electrode.
The electrode system was fed from a continuously adjustable custom-designed 0–25
kV, 5 mA, 50 Hz, high voltage supply, associated—in some experiments—with either a
half- or a Miwave solid state rectifier.
The device that was employed for the measurement of the corona current distribution
generated by the various electrode arrangements is presented in Fig. 2. A current probe,
consisting of the exposed end of an enameled copper wire (diameter: 0.4 mm; the thickness
of the enamel coating was estimated to be >10 µm), was embedded in the center of the
collecting electrode (a polished aluminum plate, 140 mm×200 mm, with rounded edges, so
that to avoid “parasitic” discharges). The probe was connected to a nano-ammeter.
The distance between the tips of the needles and the collecting electrode facing them
Fig. 1. Schematic representation of the experimental device;

1—AC high voltage power supply, 25 kV, 5 mA, 50 Hz+half- or full-wave rectifier;
2—Petri vessel containing the liquid to be treated; 3—active electrode; 4—treated liquid;
5—metallic contact between the treated liquid and the grounded electrode;
6—grounded plate electrode (all the dimensions are in mm).

293
Fig. 2. Experimental device for the measurement of the corona current density at the
surface of the collecting electrode.
could be continuously adjusted from 5 to 50 mm. The multiple-needle electrode was attached
to an X-Y positioning system. The Y-axis is defined by the projections of a row of needles
on the surface of the collecting electrode. For a given inter-electrode distance and a fixed
position of the collecting plate, the needles were translated along two orthogonal directions,
so that the current probe could scan a 75 mm×75 mm rectangle.
The quantity of ozone in the “treated” liquid was determined with the iodometric method.
Thus, the generated ozone was absorbed by a neutral or alkaline 2% potassium iodide (KI)
solution (70 ml), which was then acidified by adding 5 ml of H2SO4 solution (2N), for the
delivery of iodine from the complex of KI3. The delivered iodide atoms were visualized
(blue coloration) by using 1 ml of amidon solution (0.5%), and their quantity was determined
by titration with a Na2S2O3 solution (0.00 IN). The details of the method have been given
in [17].
3. Results
In a first set of experiments, carried out on the device presented in Fig. 1, current-voltage
characteristics were obtained for various electrode configurations. An example is shown in
Fig. 3.
The curves in Fig. 4, obtained during this same set of experiments illustrate the variation
Fig. 3. Current-voltage characteristics of the air gap between a two-needle corona electrode and the
liquid, at two values of the distance p between the corona emitting points. The “treated” liquid is
either connected to the earth (case I) or at floating potential (case II) with respect to the grounded
electrode

294
Fig. 4. Variation of the total corona current with the distance between adjacent points.
Fig. 5. Ozone quantity as a function of the inter-points distance, for a discharge gap s=20 mm, an
applied high-voltage U=20 kV, 50 Hz, and a discharge exposure duration of 30 seconds.
of the corona current with the distance p between two adjacent points of discharge. The
relation between this distance p and the quantity of ozone detected in the mass of the liquid
is shown in Fig. 5. The efficiency of ozone generation, defined as the ratio between the
quantity of O3 in water and the intensity of the corona current is calculated in Fig. 6, for
various p. The results given in Figs. 3 to 6 were obtained with the “treated” liquid either
connected to the earth (case I) or at floating potential (case II) with respect to the grounded
electrode
The second set of experiments, carried out on the experimental devices presented in
Fig. 2 enabled the measurement of the corona current density at the surface of the collecting
electrode, for various arrangements. Two sets of typical curves are displayed in Figs. 7
and 8.

295
Fig. 6. Ozone generation efficiency as function of the distance p between adjacent points of the
corona electrode, for a discharge gap s=20 mm, an applied high-voltage 20 kV, 50 Hz, and an
exposure time of 60 s.
Fig. 7. Distribution of the measured current density at the surface of the collecting electrode for
three values of the air-gap s (x is measured from the central axis of the electrode system in a
direction that is perpendicular to the plane containing the three coronating points, p=20 mm).
Fig. 8. Distribution of the measured current density at the surface of the collecting electrode for three
pairs of values (s, U); for p=20 mm. Note that y is measured from the central axis of the electrode
system in the direction defined by the plane containing the three corona-emitting points. Thus, the
central of the three points corresponds to y=0 mm, and an adjacent point is at y=20 mm. The current
density is maximum in the axis of the points and null approximately midway between them.

296
4. Discussion
The correlation between the intensity of the electric current and the amount of ozone is
welldocumented [7]. This justifies the interest of current-voltage characteristics in Fig. 3.
These curves indicate that, at given voltage, the current is slightly higher and the
corresponding quantity of ozone is likely to be larger when the corona emitting points are
less close to each other. The characteristics displayed in Fig. 5 clearly confirm this prediction,
for both cases under study (with the liquid grounded or at floating potential) and for both
2- and 3- point electrode models.
The optimum in terms of efficiency (quantity of ozone obtained for 1 mA of corona
current) is obtained for p smaller than the discharge gap s, as shown in Fig. 6. This finding
is not completely explained by the aspect of the I(p) curves in Fig. 4, and prompted the
second set of experiments, regarding the spatial distribution of the corona discharge. The
aspect of the curves given in Figs 7 and 8 indicate that the distribution of the corona current
density is not uniform at the surface of the liquid. By reducing the distance between the
points, the volume occupied by the corona discharge diminishes and may explain the
diminution of the quantity of ozone, in spite of the fact that, as shown in Fig. 4, the total
current is quasi-constant.
The curves of current density distribution in Figs. 7 and 8 diverge significantly from those
predicted by the widely-accepted Wartburg law. The explanation resides in the presence of
a metallic support in the proximity of and connected at the same high-voltage supply as the
coronaemitting points. The effect of this support is to intensify the field in a larger region of
the discharge gap and hence to expand the region where ozonization takes place. The
development of electrode systems for the direct ozonization of liquids should take into
account these observations and the results of further research on other electrode designs.
References
[1] Yoshida K and Tagashira H 1986 Mem. Kitami Inst.Technol. 1611
[2] Okazaki S, Sugimutsu H, Niwa H, Kogoma M, Moriwaki T and Inomata T 1988 Ozone Sci. Eng.
10 137
[3] Braun D, Küchler U and Pietsch G 1991 J. Phys. D: Appl. Phys 24 564
[4] El-Mohandes M T, Ushiroda S, Kajita S, Kondo Y and Horii K 1992 IEEE Trans. Ind. Appl. 1
1567
[5] Hegeler F and Akiyama H 1997 IEEE Trans. Plasma Sci. 25
[6] Eliasson B, Hirth M and Kogelschatz U 1987 J. Phys. D: Appl. Phys. 20 1421
[7] Kogelschatz U and Eliasson B 1995 Ozone generation and applications. In: Chang, J.S., Kelly,
A., Crowley, J.M. (eds.), Handbook of Electrostatic Processes (New York: Marcel Dekker) 585
[8] Kitayama J and Kuzumoto M 1997 J. Phys. D: Appl. Phys. 30 2453
[9] Yagi S and Tanaka M 1979 J Phys D: Appl. Phys. 12 1509
[10] Wilberg D M, Lang P S, Hochemer R H, Kratel A and Hoffmann M R 1996 Environ. Sci.
Technol. 30 2526
[11] Clements J S, Sato M and Davis R H 1987 IEEE Trans Ind. Appl. 23 224
[12] Sato M, Ohgiyama T and Clements J S 1996 IEEE Trans Ind. Appl., 32 106
[13] Sun B, Sato M, Harano A and Clements J S 1998 J. Electrostatics, 43 115
[14] Ai-Arainy A A, Jayaram S. and Cross, J.D. 1996 Conf. Rec. ICDL (Rome, Italy) 427
[15] Lubucki P, Jayaram S, Cross J D and Al-Arainy A A 1996 Ann. Rep. CEIDP (San Francisco)
730
[16] Goheen S C, Mong G M, Pillay G and Camaioni D M 1994 First Int. Conf. Advanced Oxidation
Technology for Water and Air Remediation, London, Ontario, 1994, pp. 83–84.
[17] Suarasan I, Ghizdavu L, Ghizdavu I, Budu S and Dascalescu L 2002 J. Electrostat. 54 207

297
Characterisation of ESD waveform and peak current from

charged printed wiring board to ESD hand tools
T Kalliohaka and J Paasi

Abstract. Electrostatic discharge (ESD) protective hand tools are used in the manufacturing
of electronics in order to protect sensitive electronics from ESD failures. A hand tool is
commonly made ESD protective by adding resistance into the handle of the tool. This added
resistance reduces the peak current of human body type ESD, greatly reducing the risk of ESD
failure. The situation becomes different when the discharge is from a charged device to the
hand tool. In this situation the resistance between the ESD hand tool and the charged device
can be almost zero leading to the most severe form of ESD when two metal objects at different
electrostatic potentials come into contact. In this paper we investigate both charged device
model (CDM) and human body model (HBM) type of ESD. A charged printed wiring board
(PWB) is discharged by the tip of ESD hand tool in the CDM test, and a charged person is
discharged to a PWB in the HBM test. Different kinds of hand tools are compared: metallic
pliers with electrostatic dissipative handle, metallic tweezers with carbon fibre tip and carbon
fibre tweezers with non-metallic parts etc.
1. Introduction
Choosing a correct hand tool is important when handling very BSD-sensitive devices
(ESDS). In this paper we investigate different kinds of hand tools in both CDM and HBM
types of discharges. A charged PWB with ESDS components must be handled very carefully
with hand tools. In a CDM-type of discharge charge transfer from the PWB through a pin of
ESDS is very rapid and may damage the ESDS. Resistance between an ESDS pin and a hand
tool and the capacitance of the PWB as well as the hand tool determine the peak current of
the CDM ESD pulse. In addition to the resistance and capacitance of the ESDS—hand tool
system, the total amount of transferred charge is influenced also by the capacitance of the
human body and the resistance from the tip of a hand tool through the body to ground.
2. Test set-up
2.1 CDM ESD test

The tests were done using an unpopulated 4-layer PWB with FR4 insulator. The layers of
the PWB were connected so that the PWB has maximum capacitance. Dimensions of the
PWB were 130×200×1.3 mm. The bottom surface of the PWB was connected to ground and
a 16 mm long, 0.6 mm diameter, sharp point needle was attached on top of the PWB with

298
silver loaded solder. A high frequency current transformer was attached so that the needle
goes through the current transformer. The current transformer was Tektronix CT1. In the
measurement, the PWB was at first, charged to 100 volts. Then a person touched the tip of
the needle with the hand tool tip to achieve a discharge from the PWB to the hand tool.
Current waveform was measured with a high bandwidth oscilloscope, Agilent Infmiium
54845A having 1.5 GHz bandwidth and 8 Gs/s maximum sample rate. The PWB layer
structure and the sketch of the measurement system is presented in Fig. 1. The capacitance
of the PWB was measured with Fluke 79III multimeter as 9.4 nF. The charge of the PWB was
measured with Monroe Nanocoulombmeter 284. The measured charge of the PWB was 930
nC when the PWB was powered up to 100 V. The measured capacitance value was very
close to that (9.3 nF) calculated from the measured charge.
In tests, the person was holding a hand tool in the left hand and grounded by a wrist wrap
in the right hand. Resistance-to-ground from the left hand was measured with a Megger
BMM2000. The value was 3.3 MΩ with 100 V measurement voltage. Measurements were
made only for one potential and capacitance value, because the results could be easily
scaled to another level. Also, below certain voltage level discharges are difficult to achieve.
2.2 HBM ESD test

In the HBM type of test the PWB and the test set-up were similar to the CDM ESD test, but
the direction of current was opposite. The measuring person was insulated from ground and
charged to 200 volts. The potential of the PWB was neutralised before a discharge. The
charged person touched the needle by the tip of the hand tool and current waveform was
recorded. Voltage increase on PWB’s surface was measured with TREK 541 electrostatic
voltmeter and it was only 3–4 volts due to the high capacitance of the PWB.
2.3 Resistance of the hand tools

Hand tools for the study were selected to have a wide range of parameters. Resistance from
the handle of the hand tool to the tip was measured, and the order of the tools in Table 1 is
Figure 1 Sketch of the test set-up

299
Description Resistance (Ω)

A Metal tweezers 0
B Snipe nose pliers with ESD handles 110
C Metal tweezers with carbon fibre tip 130
D Carbon fibre tweezers 640
E Snipe nose pliers with ESD handles 800
F Snipe nose pliers with ESD handles 7,6 k
G Snipe nose pliers with insulating handles 500 M
Table 1 Resistance of the tested hand tools
based on these measurements. The measurement method for the hand tool resistance was a
modification of the method represented in [1]: Handles of the hand tool were compressed
between jaws of a table vice. Conductive rubber sheet was placed between the vice jaw
and the hand tool handles to ensure proper electrical connection, and a gold plated brass
plate of 5×10×1 mm was placed between the jaws of the hand tool. The torque used to
tighten the table vice was controlled. Then the resistance between the body of the table
vice and the brass plate was measured with a Megger BMM2000. The lowest resistance
(0 Ω) was for metallic tweezers and the highest for ordinary (non-ESD) snipe nose pliers
(500 MΩ).
3. Results
3.1 CDM ESD test
Discharge measurements were repeated five times for every hand tool. The hand tools were
grounded and the PWB was charged to 100 volts before a discharge. An average of five
discharges was calculated both for the peak current and for the charge (Figs. 2 and 3).
Reproducibility of the results was good, but the charge calculated from the discharge current
waveform was only a small part of the total charge stored in the capacitance of the PWB. In
the measurements the main interest was in the initial peak of the discharge current, and with
that focus, the measurement was not able to cover the entire duration of the discharge.
The peak current is mainly determined by the capacitance of the PWB and the hand tool.
Current transfer from a capacitor to another is very fast due to very small resistance and
inductance. Remaining part of the energy stored in the PWB is transferred to ground through
the person. This remaining part of the charge transfer is slow compared to charge transfer
Figure 2 Peak currents of the CDM ESD test for the hand tools

300
Figure 3 Charge transfer in the CDM ESD test for the hand tools
between the PWB and the hand tool. Charge transfer through the human body is on a
different time scale than the initial fast discharge, and it is not easily measured during the
same measurement and with the same high frequency current probe as the initial peak.
Thus, either the peak current or the charge transfer could be measured “correctly” by a
single measurement. Figures 4 and 5 represent typical CDM ESD waveforms for tools
B and D.
3.2 HBM ESD test

The HBM ESD test for hand tools was quite similar to the CDM test. The main characteristics
of the charge transfer are determined by the capacitance of the hand tool as in the CDM ESD
test. In the HBM case the ratio of the capacitances and the direction of the current were
opposite to the CDM case. The capacitance of the hand tools was very small compared to
the capacitance of the four layer PWB. Therefore the voltage increase in the PWB was only
3–4 volts in spite of 200 V potential in a hand tool and the measuring person. The capacitance
of a hand tool was calculated to be 4 pF with 4 V potential increase in the PWB. Figures 6
Figure 4 CDM ESD waveform for tool B Figure 5 CDM ESD waveform for tool D

301
Figure 6 Peak currents of the HBM ESD test for the hand tools
and 7 represent peak current and charge transfer values of HBM type of discharge. Figures
8 and 9 represent typical HBM ESD waveforms for tools B and D.
Figure 7 Charge transfer in the HBM ESD test for the hand tools
Figure 8 HBM ESD waveform for tool B Figure 9 HBM ESD waveform for tool D

302
4. Discussion
The discharge current waveform in hand tool measurements can be divided into two parts.
The first part is the fast initial charge transfer between the capacitances of a PWB and a hand
tool, and the second, slower part, is the charge transfer where the human body is involved.
With metallic tool tips the duration of the first current peak is in the order of 1 ns or less. The
duration of the second part is 100–300 ns. In a single discharge only part of the total charge
stored in the PWB capacitance is transferred. After the current transfer between the
capacitances is completed, potential difference between the PWB and the hand tool is
small and current may stop flowing. In some of the measured ESD waveforms this sudden
stop of current flow was clearly visible. The potential of the PWB was also monitored in
some of the CDM type of discharges with an electrostatic voltmeter, and the remaining
potential of the PWB varied between 90 and 0 V. So in some cases charge was totally
discharged from the PWB, but in some cases the discharge was only partial leading to a
potential drop of only 10 V. In spite of that, measured charge did not vary much because
only the fast initial discharge was measured.
ESD tolerance values of electronic components are given in volts for different types of
discharge models. If this voltage is known, it is possible to calculate peak current and
charge threshold values for ESDS components [2]. By measuring peak currents and
transferred charge values, in situations where PWB handling with a hand tool is necessary,
it is possible to know how close we are to the ESD withstand level of the components and
what is the risk of ESD damage. If the safety margin is very narrow, it is important to choose
a correct hand tool.
5. Conclusions
The only way to reduce peak current in the first contact discharge is to avoid metal to metal
contact. Among the tested hand tools there were two tools with non-metallic parts at the tip.
Both tweezers had low peak current values if compared to hand tools with metallic tips.
Also the amount of charge transfer was smaller for tweezers with non-metallic tips. Ordinary
pliers with the highest resistance from handles to the tip had the lowest charge transfer due
to the minimum charge transfer between the human body and the hand tool.
Acknowledgements
The authors thank Heta Kojo at VTT for technical assistance. The work is partially supported
by the Finnish National Technology Agency Tekes via the STAHA programme.
References
[1] International Electrotechnical Commission 1998 Technical Report 61340–5–2 Electrostatics—
Protection of electronic devices from electrostatic phenomena—User guide
[2] Smallwood J and Paasi J 2003 Assessment of ESD threads to electronic components and ESD
control requirements, Electrostatics 2003, p. 247–252.

303
The role of capacitance in corona-electrode arrangements
Charles G Nolfa, Nathaniel R Greeneb, Seth T Ashmanb,

Michael A Catinob
a
XiPro Technologies LLC, Hatfield, PA 19440–0332
b
Bloomsburg University, Hartline Science Center, Bloomsburg, PA 17815
Abstract. The surface charge density on a corona electrode is modified by adjustments to its
capacitance with respect to the environment. The capacitance, electric flux per applied potential,
is discussed in terms of concentric spheres and various cylinder and cylinder-plane electrode
arrangements. Experimental data is presented to illustrate the influence of capacitance on
corona onset conditions in the parallel wire-cylinder electrode configuration. Conductive
shields were found to enhance the current transfer from an ionizer to a target surface in closely
spaced electrode arrangements. The shield enhances ionization by increasing the capacitance
and charge density on the corona electrode. At spacings larger than 50–100 mm, the corona
emitter sees a free-space environment that is largely geometry independent.
1. Introduction
The corona-electrode/counter-electrode geometry of typical commercial ionizers is typically
complex and shaped for applications. Since the corona ionizers operate at voltages from
5,000 V to well above 50,000 V, protective shields are added for structural and safety
reasons. The shields are often reported to enhance the uniformity and delivery of charge to
surfaces. Support for this claim is neither obvious nor widely discussed in the research
literature.
More than 99% of ions generated by corona systems pass between the electrodes of
closely spaced electrode assemblies, and consequently, do not participate in intended
applications. The transfer of ions to gas improves with larger electrode spacing, where
electric fields are weaker and exposure of the ions to the gas stream is greater. In many
applications, including electrostatic painting, high-voltage transmission, and room
ionization for ESD control, corona ionization occurs in “free space” and the location of the
counter charge is unclear. Typically, the meaning or location of “ground” or “earth” is often
neglected, avoided, or assigned to an arbitrarily positioned boundary for model closure.
In this work the capacitance C of a corona emitter will be defined through Gauss’s law as
the net electric flux per unit applied potential over its surface: C=⌽/V0 where is
the net electric flux, and V0 is the potential difference between the electrodes.
In the case of concentric spherical electrodes the capacitance tends to a constant value
C∞=4πε0a, where the radius of the outer sphere is taken to infinity and the inner sphere has
a radius a. As the radius of the distant outer sphere is changed, the flux, or electric field,

304
about the sphere then remains essentially constant, yielding reduced sensitivity of corona
processes about the smaller sphere to the geometric environment. There are known feedback
mechanisms from photoionization at counter electrodes, but these generally involve more
closely positioned counter electrodes than those for free-space electrodes of interest here.
Aside from concentric spheres, electrode systems show dependence of the flux on the
shape of the counter electrode. Generally, there is a strong geometric sensitivity for closely
spaced electrodes, and this sensitivity becomes weaker with further electrode separation.
Chow and Yovanovich [1] have shown the local geometric shape has relatively insignificant
influence on the electric fields distant from the smaller electrode. Chakraborty, et al. [2],
use moment methods to estimate capacitance and potential distributions for simple free-
space electrodes. The capacitive environment of nearly “free-space” corona emitters is a
second objective of this research.
In this paper we explore the capacitance of two-dimensional electrode systems, and in
particular the electric fields and fluxes about wires near and distant from various counter
electrodes. Our preliminary studies have shown that local changes in electrode capacitance
will intensify ionization from corona emitters [3]. The relatively simple measurement of
capacitance is then found to be a powerful tool in ionizer development for exploring
corona onset conditions. In a sense, the practical range for influence of the counter electrode
is the overall goal of this work.
2. Analytical models
Four electrode geometries are considered in the present work: coaxial cylinders,
cylinderplane, parallel cylinders of equal radius, and parallel cylinders of unequal radius.
Key variables are the radius of the first cylinder (a), the radius of the second cylinder (c),
and the distance from the axis of the first cylinder to the nearest point on the counter
electrode (d). The nomenclature was defined for the purpose of comparing the influences of
the various counter-electrode arrangements on the first cylinder. Proceeding through the
four electrode geometries opens the corona electrode to a more free space environment and
increasing d illustrates the influence of taking the counter electrode away.
The capacitances per unit length for these electrode sets have been calculated [4].
Coaxial cylinders
Cylinder-plane
Parallel cylinders of equal radius
Parallel cylinders of unequal radius

305
Figure 1 plots C1, C2, and C3 for a 0.18 mm-diameter cylinder (a=0.09 mm) and its counter
electrode as a function of d. As the counter electrode is moved more distant from the
cylinder, the capacitance falls in each case—more rapidly at smaller distances. The
capacitance is highest when the counter electrode surrounds the cylinder, takes on an
intermediate value when the cylinder is parallel to a conductive plane, and is smallest for
parallel cylinder systems. Increasing the size of the electrode (variable a) reduces the
capacitance values, but preserves the trends discussed above. Figure 2 compares the
capacitances C4 of a 5-mm-diameter cylinder as the radius of a second, parallel cylinder is
increased, while keeping the distance d from the center of the first cylinder to the surface of
the second cylinder constant. As the second cylinder’s radius grows, it subtends a greater
solid angle as viewed by the wire, contributing to the wire’s capacitance. At very large radii,
the second cylinder begins to play the role of a planar counter electrode, explaining the
plateaus in the capacitance curves. Indeed, the expression for C4 approaches that of C2 in
the limit of a large second-cylinder radius c.
Fig. 1. Capacitance of a 0.18-mm-diameter cylinder inside a larger cylinder,

near a plane, or near an equal-diameter cylinder.
The estimates of capacitance in Section 2.0 contain an assumption that all electrode
capacitance comes from the defined counter electrode. It is clear from these calculations
that the capacitance changes slowly at distances greater than about 50 mm from the “corona”
electrode and that the capacitance is highest when the counter electrode geometry surrounds
the wire. It is then not surprising that in room environments the capacitance of a typical
corona emitter tends toward a constant value as the counter electrode is moved away. The
corona emitter is small compared to the environment and bears the capacitance of an
isolated sphere or point electrode.
3.1 Experimental arrangement

A series of simplified experiments were performed to illustrate phenomena associated with
cylinder-cylinder and cylinder-plane electrode configurations. The experimental
arrangement in its most complete form is illustrated in Fig. 3. It is described in Ref. 3.

306
Fig. 2. Capacitance of a 5-mm-diameter wire spaced 5, 10, and 20 mm from

a second cylinder of variable radius.
Fig. 3. General arrangement of equipment.
3.2 Capacitance of cylinder parallel to a conductive plane

Figure 4 compares the capacitance of a 4.75-mm-diameter brass cylinder and a 0.18-mm-
diameter Ni-Cr wire (each 250 mm in length) and each oriented parallel to a grounded plane
as a function of their separation d. The capacitances tend to nearly constant values at larger
spacing.

307
3.3 Capacitance of two cylinders above a conductive plane

In the second series of experiments the 4.75-mm-diameter cylinder was placed 119.5 mm
above the plane surface. A second cylinder of equal diameter was placed above and in
contact with the first cylinder. The combined capacitance of these cylinders with respect to
the plane was 5.55 pF. As the second cylinder was raised above the first (keeping its electrical
connection to it), the capacitance of the cylinder pair increased towards a value near twice
that for the cylinders in contact. The capacitances are additive and are those for two free-
space cylinders.
Fig. 4. Capacitance of a 4.75-mm-diameter cylinder and a 0.18-mm-diameter wire

above a flat metal surface.
3.4 Influence of capacitance on corona ionization

The average electric field near a wire of radius a at potential V0 is proportional to its
capacitance per unit length C, relative to its entire surroundings, For fine wires
the actual distribution of this field over the surface of the wire does not strongly influence
corona onset. It is the average intensity of the field, brought on by its capacitance and
attendant surface charge that yields local gas breakdown to ionization.
The experimental arrangement shown in Fig. 3 was used to explore the influence of wire
capacitance on corona ionization. The wire was positioned 110 mm above the grounded
surface, and a cylinder placed beneath it a distance d1=11.7 mm. When a second cylinder is
positioned a distance d2 above the wire and moved towards it, the capacitance of the wire with
respect to the two cylinders increases. In particular, the capacitance between the wire and the
lower cylinder remains constant and the capacitance between the wire and the upper cylinder
increases. Thus, as d2 decreases, with fixed voltage on the wire with respect to the grounded
cylinders, the charge density on the wire increases. This occurs while the charge density on
the lower cylinder remains constant and that on the upper cylinder increases.
Under normal operation as an ionizer, the wire is in corona. The charge density on the
wire then determines ionization with control and stabilization served by the counter
electrodes and space charge. The counter-electrode cylinders in the above arrangement
primarily serve as current-collectors. The increase in charge density brought on by the
increased capacitance raises the electric field near the wire and, thus, the ionization current.

308
From this simplified model, the function of the shield—the second cylinder—is not to
project additional ions to the counter electrode, but rather to enhance current to both the
shield and counter electrode.
3.5 Capacitance and corona onset voltage as predictors of current enhancement

Let Il represent the wire current to the lower cylinder with the upper cylinder absent (d2/d1=
∞). When the upper cylinder is positioned symmetrically above the wire (d2/d1=1), the effect
is to increase the lower cylinder current to Il’. From Fig. 5, this current enhancement is
measured to be 80%.
Fig. 5. Current to cylindrical current collectors with d1=1.17 cm.
4. Conclusions
Conductive shields are found to enhance the current transfer from an ionizer to a target
surface in closely spaced electrode arrangements. The shield enhances ionization by
increasing the capacitance and charge density on the corona electrode. At spacings larger
than 50–100 mm, the corona emitter sees a free-space environment that is largely geometry
independent. Here again, altering the capacitance in the emitter region can increase
ionization.
References
[1] Chow Y L and Yovanovich M M 1982 J. Appl. Phys. 53, 8470–5
[2] Chakraborty C, Poddar D R and Chakraborty A 1993 IEEE. Trans, on EM Compatibility 35,
98–102
[3] Noll C G, Ashman S T, Catino M A and Greene N R 2001 Proc. Ann. Mtg. Electrostatics Society
of America
[4] Page L and Adams N I 1960 Principles of Electricity: An Intermediate Text in Electricity and
Magnetism (Princeton, D. van Nostrand Co.) 89–92

309
Electrostatics and the Environment

G S P Castle
Department of Electrical and Computer Engineering, University of Western Ontario,
London, ON, Canada, N6A 5B9
Abstract. Electrostatic phenomena play a very important role in the environment both in nature
and in various industrial processes used to alleviate pollution. This paper is intended to review
some of these electrostatic phenomena and applications with emphasis on three particular areas;
lightning, electrostatic precipitation, and plastics separation/recycling. Lightning represents the
oldest manifestation of electrostatics and the most spectacular demonstration of the power of
electrical forces in nature. Surprisingly, in spite of the fact that it has been studied for many years,
differences of opinion still exist as to its cause. The current understanding of the mechanisms
and theories are reviewed. Electrostatic precipitation is well known being the first significant
industrial application of electrostatics. It is used extensively to remove particulate contaminants
from air. Current challenges exist in attempting to extend the process to also deal with gaseous
pollutants. These integrated approaches are highlighted including the combined use of
precipitators with scrubbers, pulsed discharge plasma reactors and other advanced oxidation
techniques. Electrostatic techniques have been preferentially used for separating mixtures of
conductive and nonconductive materials for recycling. The current challenge is to try to extend
these principles to separation of dissimilar nonconductive materials such as plastics. One
successful approach to this problem has been developed at the University of Western Ontario.
The system is described and some results of recent industrial separations are reported.
1. Introduction
The topic of electrostatics and the environment covers such a broad field that it is impossible
to do justice to all aspects in the brief space allotted to this discussion. For example,
electrostatic phenomena play very important roles both in nature and in various industrial
processes used to alleviate air, water and soil pollution. Within these general categories
there are a great many specific examples that can be identified ranging from the controversial,
such as the effect of air ions on human health, through to the unusual, such as the use of
electric fields for strengthening unstable soils by extracting water from clays.
Rather than attempting a comprehensive review of all of the areas encompassed by this
topic the intention of this discussion is to highlight three particular areas of general
interest and special importance to the environment; lightning, electrostatic precipitation
and plastics separation/recycling.
2. Lightning
Lightning represents the oldest manifestation of electrostatics and the most spectacular
demonstration of the power of electrical forces in nature. It can even be argued that it is
possibly the most important single factor in our environment as it has been suggested that
energy released through early lightning strikes was instrumental in the creation of life on
earth. In pioneering experiments reported in 1953 [1], Nobel laureate Harold Urey’s student,
S.L.Miller demonstrated that when a gas mixture simulating what is believed to be the
constituents of the early earth’s atmosphere (H2O, H2, CH4, and NH3) was exposed to a
continuous electric discharge for one week, trace amounts of amino acids were found to be

310
produced. These materials are the building blocks of organic life. Clearly this possibility
should be the starting point in any discussion of the interconnection between electrostatics
and the environment.
However as engineers and physicists we are not normally too comfortable in dwelling
on this subject as it is more within the domain of biochemists and perhaps even theologians
and philosophers. We are more concerned with the nature of lightning itself.
Who amongst us has not been both fascinated and awed by the sheer splendour and
power of a full-blown thunder, lightning storm? Since the earliest of times lightning has
been observed and its mechanism puzzled over.
2.1 Importance of Lightning to the Environment
There are multiple reasons why lightning is so important to us in our natural environment.
For a start it can be deadly. Approximately 1000 people die each year from lightning strikes.
(Interestingly, Brazil the world’s largest tropical country accounts for over 10% of these
fatalities [2]). Furthermore lightning represents the largest single cause of power and
telecommunication failures. However, like all things in nature, it is not only something to be
feared. For example although the majority of forest fires are primarily initiated by lightning
and cause much tragedy and economic damage each year, biologists understand that such
destruction is a natural and necessary part of the health and regeneration of forests [3].
From the point of view of electrostatics however the significance of lightning is very clear,
it maintains the electrical equilibrium of the earth. Under normal atmospheric conditions, the
earth is negatively charged and experiences an average “fair weather” electric field of
approximately 100 V/m, in the downwards direction. This is accompanied by a positive ion
current flow of 10-12 A/m2 (approximately 1800 A world wide). [4] On its own, this would
result in a neutralization of the negative charge on the earth within minutes and elimination
of the electric field that we know exists in steady state. Lightning is the mechanism that
stabilizes this phenomenon by replenishing the negative charge on the earth.
As I speak, throughout the world on average, over 1000 thunderstorms are occurring
resulting in 100 lightning strikes per second. These localized electrical discharges allow
positive charge to flow from the earth into the atmosphere resulting in potential equilibrium.
What is the cause of these discharges?
2.2 Mechanism of Lightning
Scientific study began in earnest with the “enlightening” experiments of Benjamin Franklin
in the late 1700’s and has subsequently been studied in detail for many years. However
surprisingly, in spite of this long period of study even today we do not fully understand it.
Differences of opinion still exist as to the causes of this most common natural phenomenon.
What we do know is that in a normal thunderstorm, clouds develop in which the lower
part of the cloud is polarized with negative charge and the upper part with positive charge.
On first thought this does not present a situation that should result in an electric discharge
since the earth is also negatively charged. However, the lower part of a thundercloud will
normally have a much larger negative potential than the earth underneath resulting in a
strong negative electric field that attracts negative electrons towards the earth. This process
has been well studied and results in the well known “step leader”, the sporadic, irregular
passage of fast electrons moving in 50 m spurts, stopping for approximately 50 microseconds,
then continuing in an erratic path to the nearest point on the ground. Once this leader
connects with ground a complete conducting path exists between the cloud and earth. The
leading electrons are the first to flow to ground leaving behind an electron depleted region
consisting of positive ions that then attracts more electrons from further up the leader. This

311
progresses rapidly upwards along the conducting channel so that all of the negative charges
adjacent to the point of origin in the cloud flow rapidly to ground through this “return
stroke”. This takes place very rapidly (typically the return stroke lasts 0.1ms and travels at
a speed of 38× 106m/s) and very energetically (typically consisting of 20 C and a peak
current of 10 kA). The remaining negative charge in the cloud can redistribute very quickly
and travel down the remains of this ionized path (the “dark leader”) creating another return
stroke. This repeats several to tens of times resulting in the liberation of sound and light
energy which we know as thunder and lightning. Of course this is a very simplistic and
incomplete description of this complex phenomenon since it is known that approximately
10% of lightning strikes are actually “positive” lightning and much lightning occurs from
cloud-to-cloud or even internal to a single cloud. Also great controversy exists about the
nature (and according to some even the existence) of “ball” lightning [5]. However perhaps
the most surprising thing about our state of knowledge is the incomplete understanding
about how the charge separation in the cloud comes about in the first place. Many theories
and experiments have attempted to answer this and the two most accepted generally are
based upon either particle interactions or inductive processes.[6] In the former it is believed
that interactions among ice crystal, graupel and liquid water lead to charge separation and
then segregation in the cloud due to differing masses of the respective materials. In the
induction process it is believed that an ambient electric field causes polarization of particles
that then separate. Clearly the latter situation requires a pre-existing electric field, which
then leads to the conclusion that probably both processes are important.
Nature however keeps some secrets well and it would appear that much work remains to
be done to come to a definitive answer to this question.
3. Electrostatic Precipitation
Moving from nature and the environment to the effects of industrialization and the
environment, it is well known that the first significant industrial application of electrostatics
was electrostatic precipitation (ESP). [7] It is currently used extensively both in industry
and domestic applications to control air pollution by removing particulate contaminants
from air. Although we now recognize ESP as one of the most important technologies for air
pollution control, it is ironic to note that the first commercial installation which took place
in California almost 100 years ago was primarily motivated by economic considerations
related to the recovery of valuable catalysts rather than protection of the environment.
3.1 Importance of ESP to the Environment
Currently throughout the world there are many thousands of installations of ESP’s cleaning
effluents from major industrial sources such as coal-fired power plants, smelting plants
etc. In addition, two stage electrostatic precipitators are widely used to remove small
particulates from recycled air in industrial and domestic applications.
Although the principles of ESP technology are well-understood, continual improvements
in terms of collection efficiency, reliability and cost effectiveness have taken place, primarily
in the last 35 years or so. Driven by increasingly stringent environmental regulations major
advances have occurred with the development of techniques to improve the collection for
high resisitivity materials and for very fine particles in the size range less than 10µm. These
changes have included improvements in; the mechanical design of the precipitators, the
aerodynamics of the air flow, the use of conditioning agents to reduce particle resistivity and
induce agglomeration, the use of new power sources that allow intermittent and pulse
energization to improve performance and save energy, and the use of “on line” measurement
and control to maintain optimum performance under varying conditions.[8]

312
Current challenges exist in attempting to extend the process to also deal with gaseous
pollutants such as SOx and NOx. At this time the only feasible way to remove such gases is
to use an ESP to remove the particles followed by absorption in a liquid scrubber. Other
noxious gases such as volatile organic compounds (VOC) and dioxins are currently removed
by incineration. However much ongoing effort is being expended in trying to induce
chemical reactions using such techniques as non-thermal discharge plasmas, electron beam
reactors, and advanced oxidation techniques[9]. Unfortunately, economic and reliability
issues are significant barriers to adoption and much more work will be required before these
become viable alternatives.
4. Plastics Separation/Recycling
As a final example of the importance of electrostatics to the environment, I would like to
highlight a relatively new technology, the separation of plastic waste. It is well known that
electrostatic techniques have been preferentially used for many years for separating mixtures
of conductive and nonconductive materials for recycling. [10] However the current challenge
is to try to extend these principles to the separation of dissimilar nonconductive materials
such as plastics. This process has the major advantage of using dry materials, unlike for
example alternative methods such as flotation.
The importance of the need for this type of recycling can be highlighted by the fact that
in North America paper and paperboard have a recycling rate of over 40%. These materials
are of course renewable resources. Yet it is energy efficient and beneficial to the environment
to recycle them. On the other hand, plastics, which are produced from nonrenewable and
rapidly depleting petrochemical resources, are recycled at an overall rate of only 5%.[11]
The rest is disposed of in landfill or in some other such wasteful or nonproductive manner.
It is true that certain types of plastics such as used in plastic bottles are recycled at a higher
rate. (Post consumer bottle recycling was approximately 22% in the USA in 2001.) However
bottles represent an easily separated and relatively clean product. More generally, post
consumer plastics are mixed with many contaminants and contain several types of plastic.
The separation problem is complicated by the mutual incompatibility of the most widely
used commercial plastics which include Polyvinyl Chloride (PVC), Polyethylene Terephalate
(PET), high and low density Polyethylene (HDPE & LDPE), Polypropylene (PP), and
Polystyrene (PS). All these plastics have special individual properties and as a result find
uses in many different products. Separation purities must be very high in order to enable the
material to be reused in place of virgin material. Thus there is a real demand for a process
that can selectively segregate mixtures of two or more types of plastic efficiently and
economically.
4.1 Principle of Separation
At the IOP Electrostatics 1991 meeting we presented a paper describing the feasibility of
separating certain types of commercial grade plastics using contact charging in a fluidized
bed and dropping them through a horizontally oriented electric field. [12] A disadvantage of
this process is that fluidization requires significant mechanical energy. Since then this work
has been extended to produce a more practical and energy efficient separator involving
simply a rotating inclined cylindrical drum. In this system plastic particles, typically consisting
of two or three different constituents chopped or ground to a size of the order of 5 mm, are fed
through the rotating drum and undergo contact charging with each other and to a lesser extent
with the inside wall of the cylinder. This equipment evolved from a laboratory scale separation
unit that consisted of a 10cm diameter grounded metal charging drum, 1m in length, and
having one end closed. This was mounted above a 3m high separation tower in which the

313
high voltage electrodes were mounted. The electrodes were connected to variable +/- 60kV
power supplies. Depending upon the voltages applied, either symmetric or asymmetric
horizontal electric fields of up to 300kV/m were possible. A series of nine shielded collection
trays having their long axis perpendicular to the electric field were placed at the base of the
tower. In this way both the mass and charge of each collected fraction were measured. The
separation tower acts as a macroscopic mass spectrometer where the greater the charge to mass
ratio (Q/M) the further the particles are deflected from straight vertical fall.
An extensive series of laboratory scale batch experiments were carried out in which a
weighed quantity (typically 200g) of two or three component mixed plastic material was
rotated (typically at 5–20 rpm) for a fixed time (typically 2 minutes) with the drum slightly
inclined upwards. The cylinder was then tilted slightly downwards (about 10 degrees) and
the particles gradually fell out of the drum into the tower and were collected in the trays.
The Q/M and composition of the components was determined either by using different
coloured constituent materials or by x-ray spectrography. The quality of separation was
measured by determining the sample purity (mass of desired component in collected sample/
total mass in collected sample) and the efficiency was measured by the sample recovery
(mass of desired component in collected sample/total mass of initial material). By appropriate
adjustment of the electric field strength and inclusion of samples from various numbers of
the collection trays adjacent to either the positive, negative or grounded field electrodes it
was possible to determine different levels of material purity and recovery. Detailed results
were presented elsewhere [13]. Selected examples are shown in Table 1.
Table 1. Separation results for selected mixtures of virgin plastic mixtures
4.2 Industrial Applications

Plas-Sep Ltd., London, ON, Canada, has subsequently extended the excellent results found in
the laboratory tests into industrial scale applications. A unit module capable of handling a
continuous feed of one tonne/hour uses a charging drum 0.5m in diameter and 2m long. The
fall distance is 4m. Two adjustable separators that collect 3 components, positively charged,
negatively charged and middlings are used in place of the nine collection trays. The polarity

314
of the charge that develops depends upon the materials and various versions of effective
‘Triboelectric Series” have been published for different plastics. However a number of factors
normally complicate the prediction of chargeability. For example, such things as surface
printing dyes, colorants and contaminants from the previous use for the plastic can dramatically
affect the charging polarity and magnitude. However in many practical cases for given
combinations of materials although not predictable a priori the charging properties are
consistent. Two factors affect this. The first is that the fact that the plastics need to be chopped
thus exposing a considerable amount of virgin plastic surface. The second is the need to
ensure that the relative humidity (RH) of the air in which the plastic is treated is controlled.
Normally a RH<50% is satisfactory. Successful and economically attractive separations have
been achieved in post-industrial waste applications for two-component mixtures in the
automobile industry (tail light and bumper assemblies) and the electrical industry (chopped
wire insulation.) Even more significant are recent results that have successfully separated
three component plastic mixtures from dashboard scrap.[14] The technology is also suitable
for separating many types of post-consumer waste plastic material although the necessity to
pre-treat and clean the material makes the economics less attractive.
5. Conclusions
Clearly it is impossible to do full justice to a topic as broad as electrostatics and the
environment in this brief discussion. However it is hoped that the three specific examples
presented here demonstrate the fundamental importance of this topic in nature, pollution
control and conservation of our resources. Much essential work remains to be done both in
understanding and applying electrostatic phenomena in the environment.
6. References
1) S.L.Miller, “A Production of Amino Acids Under Possible Primitive Earth Conditions”, Science,
Vol 117, No 3046, May 15,1953, p528
2) O.Pinto, “Lightning in Brazil Sets Shocking Record”, www.cnn.com/2002/TECH/science/1 1/
05/brazil.lightning.reut/index.html
3) R.W.Gorte, “Forest Fires and Forest Health”, CRS Report for Congress, 95–511 ENR, National
Council for Science and Environment, Washington, DC, July 14, 1995
4) R.P.Feynman, R.B.Leighton and M.Sands, “Chapter 9:Electricity in the Atmosphere”, The Feynman
Lectures on Physics, Volume II, Addison-Wesley Publishing Co., Reading, Mass, sixth printing,
1977, pp9–1–11
5) D.J.Turner, “Ball Lightning and Other Meteorological Phenomena”, Physics Reports 293, Elsevier,
1998, pp 1–60
6) K.Miller et al, “Modeling and Observations of Thundercloud Electrification and Lightning”,
Atmospheric Research 58, Elsevier, 2001, pp89–115
7) H.J.White, “Centenary of Fredrick Gardner Cottrell”, J. Electrostatics, 4, 1977/78, pp1–34
8) K.R.Parker, “Applied Electrostatic Precipitation”, Chapman and Hall, London, 1997
9) H.H.Kim et al, “Low Temperature NOX Reduction Processes Using Combined Systems of Pulsed
Corona Discharge and Catalysts”, Journal of Physics D: Applied Physics, 34, 2001, pp604–137
10) H.Higashiyama and K.Asano, “Recent Progress in Electrostatic Separation Technology”, Particulate
Science and Technology, Vol 16, #1, 1998, pp77–90
11) R&D Magazine, Cahners Business Information, Highlands Ranch, CO, Sept 2002, p54
12) I.I.Inculet and G.S.P.Castle, “Triboelectrification of Commercial Plastics In Air”, Institute of
Physics Conference Series, #118, Oxford, 1991, pp217–222
13) I.I.Inculet, G.S.P.castle and J.D.Brown, “Electrostatic Separation of Plastics for Recycling”,
Particulate Science and Technology, Vol 16, #1, 1998, pp91–100
14) J.D.Brown, “Electrostatic Plastics Separation”, Recycling Technology Newsletter, Natural
Resources Canada, Vol 5, #1, 2000, pp1–2

315
Measurement methods in electrostatics applications:

review and trends
W D Greason
Department of Electrical and Computer Engineering, Faculty of Engineering,
University of Western Ontario, London, Ontario, N6A 5B9, Canada
Abstract. A review of evolving measurement techniques in applied electrostatics is presented
for the period from the 1999 Institute of Physics Conference on Electrostatics, held in
Cambridge, to the present. The main sources referenced include the Journal of Electrostatics,
the Electrostatic Processes Committee contributions to the IEEE Transactions on Industry
Applications and the Conference Proceedings of the Electrical Overstress/Electrostatic
Discharge Symposia.
Measurements in electrostatics applications can vary from the basic, employing standard
instruments with a relatively small number of recordings to the complex, involving many
specialized instruments used to obtain data intensive records of events. Several areas of
interest include: non-invasive measurements employing signature analyses principles which
do not disturb the phenomena under study, investigation of microgap phenomena involved in
magnetoresistive recording heads and MEMS technology, data acquisition and analyses in
sensor intensive applications recording high frequency transient events associated with
electrostatic discharges and the study of charge decay on insulators including the characterization
of spatial charge distribution profiles.
Enhanced measurement methodologies are clearly evolving; our educational institutions are
encouraged to include the subject of measurement of electrostatic phenomena in courses on
electronic instrumentation and measurement.
1. Introduction
A review of measurements related to the following topics is presented: charge, discharges,
medium characterization, ESPs and reactors, probes and transducers, biotechnology, MEMS
and space exploration.
2. Charge
The measurement of surface potential decay on dielectric materials has been described [1–
4]; a space-charge measuring system based on a laser-induced pressure pulse (LIPP) method
along with thermally stimulated discharge current (TSDC) measurements can be applied to
analyze charge distribution in thin films [5,6]. Space charge, electric field and potential in
multi-layered dielectric films can be studied from interfacial charges measured using the
step-electroacoustical (SEA) technique [7,8].
The measurement of bipolar charge in powders [9] and on aerosols [10,11] has been
described. Measurement methodologies for the charge carrier extraction from gas streams
[12], charge distributions of particles in gas-solids pipe flow [13] and streamer inception in
liquids using a differential charge measurement system have been presented [14].

316
Experimental systems to study corotron charging behaviour in electrophotographic

processes have been summarized [15,16]. Apparatus to investigate various aspects of
triboelectrification have been described [17–19]. A reading unit with the ability of non-
contact charge detection was used to demonstrate the feasibility of an electret floppy disk
for digital information storage [20].
3. Discharges
3.1 General
Electrostatic discharges leading to explosions are a frequent cause of industrial accidents.
Experimental approaches to study ignition tests with brush discharges [21], the ignition of
methane-air mixtures by multiple capacitor discharges [22] and measurements of incendivity
of electrostatic discharges from textiles used in personal protective clothing [23] have been
reported. The following are examples of measurement based studies on the general topic of
discharges: bipolar dc corona discharge from a floating filamentary metal particle [24];
electrical discharge occurring from a space charge cloud formed with charged particles
[25]; characterisation of electrostatic discharges from insulating surfaces [26]; NOx
production in spark and corona discharges [27]; correlation between the optical signatures
and current wave forms of long sparks with applications in lightning research [28];
methodology to study the resistance of spark discharges [29].
3.2 Corona
Experimental apparatus and measurement procedures have been described for dc corona
discharges [30–32], corona discharges and a water environment [33–39], pulsed corona
discharges [40–43], corona discharge in the pin-to-plane geometry [44–47] and ozone
generation in ac corona discharge [48].
3.3 ESD
ESD continues to be a major reliability issue in electronic devices and systems. Test methods
and methodologies continue to evolve [49–51]. Measurement of the indirect effect of ESD
offers a non invasive means to characterize the event; detection of the EMI due the radiated
electric and magnetic fields has been studied [52–60] as well as the use of a calorimeter to
detect the optical signal emitted by a spark discharge [61,62]. Other measurement related
topics include: broadband measurement of ESD risetimes [63]; fields on horizontal coupling
planes excited by direct ESD to the vertical coupling plane [64]; interferometric temperature
mapping during ESD stress of protection devices [65]; surface resistivity and charge decay
measurements of materials [66].
3.3.1 GMR Heads
Giant magnetoresistive heads used in the disk drive industry are very susceptible to damage
from low level ESD transients. Procedures for the testing and measurement of ESD
susceptibility have been described and documented as follows: tribocharging and electrical
breakdown at the head-disk interface [67]; ESD sensitivity of GMR heads at variable pulse
length [68]; measurement of current transients [69]; considerations for ESD standards for
measuring and testing of MR heads [70]; electrostatic voltmeter and fieldmeter measurements
on GMR heads [71]; field-induced charged device model testing of MR heads [72].

317
3.3.2 TIP
Transmission line pulsing has become established as a measurement method to characterize
the performance of ESD protection devices. Measurement set ups for various applications
have been reported: TLP ESD testing of GMR heads [73]; TLP testing of low voltage
triggering SCR devices [74]; analysis of the switching behaviour of ESD protection
transistors [75]; breakdown and latent damage of ultra thin gate oxides [76]; verification
of ESD protection device response [77]; TLP of integrated structures [78]. The design of
TLP test sets, calibration and evolving standards have also been described [79–80]
4. Medium Characterization
4.1 Dielectrics
Measurement methods to study the properties of dielectric materials have been reported
and include: interdigital dielectrometry for the non-destructive evaluation of material
properties [81]; high-temperature conduction of plasma-treated films [82]; non-linear
conductivity spectra of antistatic polymer films [83]; influence of electron-beam irradiation
on electrical parameters of dielectric materials [84].
4.2 Powders and Particles

Experimental apparatus and measurement methods have been presented for powder coatings,
fluidized beds, the charging of particles and powders and phenomena related to the charge on
powders and particles. For powder coatings, work reported includes: adhesion measurements
for coatings using drop test rig and virtual oscilloscope [85]; measurement of thickness and
adhesive properties [86]; electroseparation and efficiency of deposition [87]; powder flow
control system with capacitance sensor [88]. Two papers on fluidized beds [89,90] provide
detail on the instrumentation and measurement methods employed. Aspects of the charging
processes include: unipolar charging of aerosol particles in alternating electric field [91];
change in charging characteristics of polymer powder by plasma treatment [92];
characterization of chargeability of biological particulates by triboelectrification [93]. Work
related to the charge on particles and powders include: real-time particle size and charge
distribution analysis [94]; evaluation of charge retention properties of powder paints by
thermally stimulated current spectroscopy [95]; free air beam in an electric field for
determination of charge on powders [96]; trajectories of charged aerosol particles near a
spherical collector [97]; electrostatic measurements on lactose-glucose mixtures [98]; static
charge elimination on pellets in a silo using a nozzle type eliminator [99].
4.3 Sprays
Recent publications include the description of measurement set ups for experiments on the
spraying of liquids [100–102] and ceramic suspensions [103–105]. Some topics of interest
include: development of electrostatic pressure-swirl nozzle for agricultural applications
[106]; electrostatic application of pollen sprays [107]; electrohydrodynamic spraying
characteristics of glycerol solutions in vacuum [108].
4.4 Liquids
Recent papers which discuss measurement principles related to liquids can be broadly
classified as follows: charge generation, liquid characterization, specific measurements.

318
Charge generation has been described for impeller mixing of used transformer oil [109] and
the internal electrification of diesel oil injectors [110]. Liquid characterization processes
include: deformation and break-up of aqueous drops of dielectric liquids in high electric
fields [111]; electrohydrodynamic surface waves of thin oil films generated by needle-
plate barrier discharges under reduced gas pressure [112]; surface tension reduction of
liquid by applied electric field using vibrating jet method [113]. Specific measurement
procedures reported include: power consumption measurements for ac and pulsed dc for
electrostatic coalescence of water-oil emulsions [114]; measurements of surface conductivity
in dielectric liquid [115]; Kerr electro-optic field mapping measurements [116]; the use of
FET models for analysing electro-osmosis [117]. Papers of special interest include liquid
dielectrophoresis on the microscale [118] and electrostatic actuation of liquid droplets for
microreactor applications [119].
5. Large Scale Operations
5.1 ESP
Experimental arrangements and instrumentation for measurement have been described for
pulsed and dc electrostatic precipitator systems [120–123]. Other subjects include: smoke
precipitation by charged water aerosol [124]; effect of particle diameter and corona electrode
geometry on the particle migration velocity in electrostatic precipitators [125];
pseudoelectret filter for micrometer-sized particles in exhaust gases [126]; submicron
charged dust particle interception by charged drops [127].
5.2 Reactors
Non-thermal plasma processing is one of the most promising technologies to remove toxic
gas contaminants in air; several publications describe the experimental set ups, instrumentation
and measurement methods employed in the experiments [128–142]. Other topics involving
the description of measurements include: measurement of hydroxyl radicals in an atmospheric
pressure discharge plasma using laser-induced fluorescence [143] improvement of NOx removal
efficiency using atomization of fine droplets in corona discharge [144]; plasma-assisted
chemical process for NOx control [145]; removal of NF3 from semiconductor-process flue
gases by tandem packed-bed plasma and adsorbent hybrid systems [146]; reduction of NOx
from natural gas combustion flue gases by corona discharge radical injection techniques
[147]; indoor air cleaning using a pulsed discharge plasma [148].
6. Probes/Transducers
The design and characterization of PVDF transducers [149,150] and electrostatic flow
probes [151–155] has been documented. Other probes developed for specific measurement
applications include: vibrating-wire transducers for electrostatic measurements [156];
floating double probe characteristic utilized to measure electron temperature [157]; the use
of Tassicker’s measurement formula in the boundary biased-probe method [158]; visualization
of the distribution of electron traps on polymer surfaces using a scanning micro laser probe
[160].

319
7. Biotechnology
The use of electrostatics in the field of biotechnology is an emerging topic of interest.
Measurement methodologies have been reported in the following work: selective detection
of viable bacteria using dielectrophoretic impedance measurement method [161]; effect of
culture temperature on high-voltage pulse sterilization of escherichia coli [162]; integrated
planar concentric ring dielectrophoretic (DEP) levitator [163]; electrical sterilization of
escherichia coli by electrostatic atomization [164]; enhancing effect of electret on
transdermal drug delivery [165]; molecular surgery of DNA based on electrostatic
micromanipulation [166]; inactivating microorganisms using a pulsed electric field
continuous treatment system [167]; quantitative analysis of DNA orientation in stationary
ac fields using fluorescence anisotrophy [168].
8. MEMS
Microelectromechanical systems (MEMS) technology is in its development stage and will
present measurement challenges in the future. Two applications have been described: a
MEMS field mill made from two surface micro-machined polysilicon thin film layers
deposited over a silicon substrate [169]; electrostatic micromirrors for subaperturing in an
adaptive optics system [170].
9. Space Exploration
NASA has an interest in dust particle charging on Mars; the design and development of
instruments for in situ measurements presents a great challenge. The design of an
electrometer to evaluate the electrostatic nature of the Martian soil and atmosphere has
been presented [171]; an experimental apparatus to evaluate the triboelectrification of a
Martian soil simulant has been reported. Future work will include the development of
particle charge spectrometers to measure the size and charge on individual dust particles
and an instrument to measure electrostatic fields generated by rover interactions with the
Martian surface.
10. Trends and Needs

Instrumentation and measurement are essential components of research based programs. As
new electrostatic based technologies evolve, there is a need for enhanced measurement
methodologies. Non-invasive measurements which do not disturb the phenomenon under
study are of interest since instrumentation loading effects are eliminated. The study of
microgap phenomena involved in magnetoresistive recording heads and MEMS technology
present measurement challenges. Wide bandwidth probes for electrostatic discharge
measurements are needed to characterize current waveforms and charge transfer. There is a
need for integrated measurement devices for sensor intensive applications.
The use of advanced instrumentation and measurement principles in electrostatics
research presents a challenge to educators. More than ever, one must not only know how to
measure a physical parameter but also be able to interpret the results to know what is being
measured. Our educational institutions are encouraged to include the subject of measurement
of electrostatic phenomena in courses on electronic instrumentation and measurement.

320
References
[1] Sahli S, Bellel A, Ziari Z, Kahlouche A and Segui Y 2003 J. Electrostatics 57 169–181
[2] Chubb J 2002 J. Electrostatics 54 233–244
[3] Chubb J 2000 IEEE Trans. Incl Appl. 36 1515–1522
[4] Chubb J 1999 Proc. IOP Conf. Ser. No. 163 329–333
[5] Takashima K and Oda T 2000 IEEE Trans. Incl Appl. 36 76–81
[6] Takashima K and Oda T 1999 IEEE Trans. Incl Appl 35 1192–1197
[7] Motyl E 2001 J. Electrostatics 51–52 522–529
[8] Motyl E 2001 J. Electrostatics 51–52 530–537
[9] Zhao H, Castle G S P and Inculet 11 2002 J. Electrostatics 55 261–278
[10] Kulon J, Hrabar S, Machowski Wand Balachandran W 2001 lEEETrans. Incl Appl. 37472–
479
[11] Kulon J and Balachandran W 2001 J. Electrostatics 51–52 552–557
[12] Noll C G, Miller W and Bracikowski C 2002 J. Electrostatics 54 283–291
[13] Matsusaka S, Umemoto H, Nishitani M and Masuda H 2002 J. Electrostatics 55 81–96
[14] Dumitrescu L, Lesaint O, Bonifaci N, Denat A and Notingher P 2001 J. Electrostatics 53 135–
146
[15] Feng J Q, Morehouse P W and Facci J S 2001 IEEE Trans. Incl Appl 37 1651–1657
[16] Hays D A 2001 J. Electrostatics 51–52 57–63
[17] Greason W D, Kucerovsky Z and Zaharia C 2002 J. Electrostatics 56 385–398
[18] Ohara K, Nakamura I and Kinoshita M 2001 J. Electrostatics 51–52 351–358
[19] Yu Z and Watson K 2001 J. Electrostatics 51–52 313–318
[20] Amjadi H and Franz C J.Electrostatics 50 265–277
[21] Schwenzfeuer K and Glor M 2001 J. Electrostatics 51–52 402–408
[22] Ptasinski L and Zeglen T 2001 J. Electrostatics 51–52 395–401
[23] Kathirgamanathan P, Toohey M J, Haase J, Holdstock P, Laperre J and Schmeer-Lioe G 2000
J. Electrostatics 49 51–70
[24] Asano K, Hishinuma R and Yatsuzuka K 2002 IEEE Trans. Incl Appl. 38 57–63
[25] Higashiyama Y, Sugimoto T and Takahashi M 2001 J. Electrostatics 51–52 381–386
[26] Davidson J L, Williams T J, Bailey A G and Hearn G L 2001 J. Electrostatics 51–52 374–380
[27] Rehbein N and Cooray V 2001 J. Electrostatics 51–52 333–339
[28] Gomes C and Cooray V 1998 J. Electrostatics 43 267–274
[29] Greason W D 1999 IEEE Trans. Ind. App. 35 359–365
[30] Noll C G 2002 J. Electrostatics 54 271–282
[31] Noll C G 2002 J Electrostatics 54 245–270
[32] Asano K, Yatsuzuka K and Yamaki T 2000 IEEE Trans. Ind. Appl. 36 87–92
[33] Suarasan I, Ghizdavu L, Ghizdavu I, Budu S and Dascalescu L 2002 J. Electrostatics 54 207–
214
[34] Higashiyama Y, Yanase S and Sugimoto T 2002 J. Electrostatics 55 351–360
[35] Yamamoto M, Nishioka M and Sadakata M 2002 J. Electrostatics 56 173–187
[36] Sugimoto T, Asano K and Higashiyama Y 2001 J. Electrostatics 53 25–38
[37] Sugimoto T and Higashiyama Y 2001 J. Electrostatics 53 209–219
[38] Fujii T, Aoki Y, Yoshioka N and Rea M 2001 J. Electrostatics 51–52 8–14
[39] Moon J, Kim J and Lee D 1998 IEEE Trans. Ind. Appl. 34 1212–1217
[40] Ono R and Oda T 2002 J. Electrostatics 55 333–342
[41] Yan K, van Heesch E J M, Pemen A J M, Huijbrechts P A H J, van Gompel F M, van Leuken
H and Matyas Z 2002 IEEE Trans. Ind. Appl. 38 866–872
[42] Yan P, Fujiwara M, Zhou Y and Ishida M 2001 J. Electrostatics 51–52 266–271
[43] Kirkpatrick M, Finney W C and Locke B R 2000 IEEE Trans. Ind. Appl. 36 500–509
[44] Kanazawa S, Ito T, Shuto Y, Ohkubo T, Nomoto Y and Mizeraczyk J 2002 J. Electrostatics 55
343–350

321
[45] Kucerovsky Z, Greason W D and Ieta A 2001 J. Electrostatics 50 147–157

[46] Dascalescu L, Samuila A, Rafiroiu D, Iuga A and Morar R 1999 IEEE Trans. Ind. Appl. 35
543–548
[47] Takahashi N, Goldman A, Goldman M and Rault J 2000 J. Electrostatics 50 49–63
[48] Feng R, Castle G S P and Jayaram S 1998 IEEE Trans. Ind. Appl. 34 563–570
[49] Chaine M, Verhaege K, A very L, Kelly M, Gieser H, Bock K, Henry L G, Meuse T, Brodbeck
T and Barth J 1998 EOS/ESD Symposium Proc. EOS-20 301–310.
[50] Carey R E and DeChiaro L F 1998 EOS/ESD Symposium Proc. EOS-20 40–53
[51] Lin D, Pommerenke D, Barth J, Henry L G, Hyatt H, Hopkins M, Senko G and Smith D
1998 EOS/ESD Symposium Proc. EOS-20 29–39
[52] Rudack A C, Pendley M and Levit L 2002 J. Electrostatics 54 95–104
[53] Ptasinski L, Zeglen T and Gajda J 1999 Proc. IOP Conf. Ser. No. 163 225–228
[54] Takai T, Kaneko M and Honda M 1998 J. Electrostatics 42 305–320
[55] Frei S and Pommerenke D 1998 J. Electrostatics 44 191–203
[56] Lin D L, DeChiaro L F and Jon M 1998 J. Electrostatics 44 159–175
[57] Bernier J, Croft G and Lowther R 1998 J. Electrostatics 44 149–157
[58] Jacksen N, Tan W and Boehm D 1998 EOS/ESD Symposium Proc. EOS-20 233–244
[59] Munoz J L, Tan J, Adriano C, Roldan E and Sadie J 2000 EOS/ESD Symposium Proc. EOS-
22 60–64
[60] Jacksen N, Nelsen L, Boehm D and Odom T 2000 EOS/ESD Symposium Proc. EOS-22 193–
201
[61] Kucerovsky Z, Greason W D and Flatley M W 2002 IEEE Trans. Ind. Appl. 38 1162–1167
[62] Kucerovsky Z, Greason W D and Flatley M W 2002 IEEE Trans. Ind. Appl. 38 77–84
[63] Bonisch S, Pommerenke D and Kalkner W 2002 J. Electrostatics 56 363–383
[64] Frei S and Pommerenke D 1998 J. Electrostatics 44 177–190
[65] Furbock C, Pogany D, Litzenberger M, Gornik E, Seliger N, Gobner H, Muller-Lynch T,
Stecher M and Werner W 2000 J. Electrostatics 49 195–213
[66] Paasi J, Nurmi S, Vuorinen R, Strengell S and Maijala P 2001 J Electrostatics 51–52 429–434
[67] Himle J and Wallash A 2002 EOS/ESD Symposium Proc. EOS-24 142–146
[68] Guarisco D and Li M L 2000 EOS/ESD Symposium Proc. EOS-22 322–326
[69] Ogle W and Moore C 2000 EOS/ESD Symposium Proc. EOS-22 217–223
[70] Henry L G and Wallash A 2000 EOS/ESD Symposium Proc. EOS-22 337–342
[71] Pritchard D 2000 EOS/ESD Symposium Proc. EOS-22 499–504
[72] Wallash A J 1996 EOS/ESD Symposium Proc. EOS-18 8–13
[73] Wallash A 2002 J. Electrostatics 56 331–339
[74] Guilhaume A, Galy P, Chante J P, Foucher B, Bardy S and Blanc F 2002 J. Electrostatics 56
281–294
[75] Wolf H, Gieser H and Wilkening W 2000 J. Electrostatics 49 111–127
[76] Wu J, Juliano P and Rosenbaum E 2000 EOS/ESD Symposium Proc. EOS-22 287–295
[77] Hyatt H, Harris J, Alanzo A and Bellew P 2000 EOS/ESD Symposium Proc. EOS-22 111–120
[78] Gieser H and Haunschild M 7996 EOS/ESD Symposium Proc. EOS-18 85–94
[79] Chu C Y and Worley E R 1998 EOS/ESD Symposium Proc. EOS-20 311–319
[80] Barth J, Verhaege K, Henry L G and Richner J 2000 EOS/ESD Symposium Proc. EOS-22
85–96
[81] Mamishev A V, Takahashi A R, Du Y, Lesieutre B C and Zahn M 2002 J. Electrostatics 56
465–492
[82] Takanashi T and Ikezaki K 2002 IEEE Trans. Ind. Appl. 38 846–851
[83] Tajitsu Y 2002 J. Electrostatics 55 299–310
[84] Vila F and Sessler G M 2001 J. Electrostatics 51–52 146–152
[85] Dastoori K and Makin B 2001 J. Electrostatics 51–52 509–514
[86] Dastoori K, Makin B and Telford J 2001 J. Electrostatics 51–52 545–551

322
[87] Barmuta P and Cywinski K 2001 J. Electrostatics 51–52 239–244

[88] Yanagida K, Morita T and Takeuchi M 2000 J. Electrostatics 49 1–13
[89] Alissa Park A, Bi H T, Grace J R and Chen A 2002 J. Electrostatics 55 135–158
[90] Yao L, Bi H T and Park A 2002 J. Electrostatics 56 183–197
[91] Lackowski M 2001 J. Electrostatics 51–52 225–231
[92] Murata Y and Aradachi T 2001 J. Electrostatics 51–52 97–104
[93] Banerjee S and Law S E 1998 IEEE Trans. Incl. Appl. 34 1201–1205
[94] Mazumder M K, Grable N, Tang Y, O’Connor S and Sims R A 1999 Proc. IOP Conf. Ser. No.
163 335–340
[95] Ogiwara T and Ikezaki K 1999 Proc. IOP Conf. Ser. No. 163 325–328
[96] Epping R.H and Kuettner A 2002 J. Electrostatics 55 279–288
[97] Jaworek A, Adamiak K, Krupa A and Castle G S P 2001 J. Electrostatics 51–52 603–609
[98] Murtomaa M and Laine E 2000 J Electrostatics 48 155–162
[99] Kodama T, Suzuki T, Nishimura K, Yagi S and Watano S 2002 J. Electrostatics 55 289–297
[100] Huneiti Z and Balachandran W 2001 J. Electrostatics 51–52 558–564
[101] Davies K and Gerken S 2000 J. Electrostatics 49 139–150
[102] Huneiti Z A, Balachandran W and Machowski W W 1998 IEEE Trans. Ind. Appl. 34 pp. 279–
285
[103] Miao P, Balachandran W and Xiao P 2002 IEEE Trans. Ind. Appl. 38 50–56
[104] Miao P, Balachandran W and Wang J L 2001 J. Electrostatics 51–52 43–49
[105] Balachandran W, Miao P and Xiao P 2001 J. Electrostatics 50 249–263
[106] Laryea G L and No S 2003 J. Electrostatics 57 129–142
[107] Law S E, Wetzstein H Y, Banerjee S and Eisikowitch D 2000 IEEE Trans. Ind. Appl. 36 998–
1009
[108] Ku B K and Kim S S 2003 J. Electrostatics 57 109–128
[109] Wagner J P and Clavijo F R 2000 J. Electrostatics 50 31–47
[110] Romat H and Badri A 2001 J. Electrostatics 51–52 481–487
[111] Eow J S, Ghadiri M and Sharif A 2001 J. Electrostatics 51–52 463–469
[112] Lee T S and Chang J S 2001 J. Electrostatics 51–52 565–570
[113] Sato M, Kudo N and Saito M 1998 IEEE Trans. Ind. Appl. 34 294–300
[114] Lee C M, Sams G W and Wagner J P 2001 J. Electrostatics 53 1–24
[115] Matsubara Y 1999 J. Electrostatics 46 125–130
[116] Gung T J, Ustundag A and Zahn M 1999 J. Electrostatics 46 79–89
[117] Watanabe S, Yamada Y, Hobo Y, Hayashi N, Sumi T, Dykes D and Touchard G 2001 /.
Electrostatics 51–52 455–462
[118] Jones T B 2001 J. Electrostatics 51–52 290–299
[119] Washizu M 1998 IEEE Trans. Ind. Appl. 34 732–737
[120] Elayyan H S B, Bouziane A and Waters R T 2002 /. Electrostatics 56 219–234
[121] Zukeran A, Looy P C, Chakrabarti A, Berezin A A, Jayaram S, Cross J D, Ito T and Chang J
S 1999 IEEE Trans. Ind. Appl. 35 1184–1191
[122] Liang X, Looy P C, Jayaram S, Berezin A A, Mozes M S and Chang J S 2002 IEEE Trans.
Ind. Appl. 38 69–76
[123] Mok Y S, Lee H W and Hyun Y J 2001 J. Electrostatics 53 195–208
[124] Balachandran W, Krupa A, Machowski W and Jaworek A 2001 J. Electrostatics 51–52 193–
199
[125] Jedrusik M, Gajewski J B and Swierczok A J 2001 J. Electrostatics 51–52 245–251
[126] Inculet 11, Castle G S P, Slanina M and Duca M 2002 IEEE Trans. Ind. Appl. 38 64–68
[127] Jaworek A, Krupa A and Adamiak K 1998 IEEE Trans. Ind. Appl. 34 985–991
[128] Oda T J. 2003 Electrostatics 57 293–311
[129] Yamamoto T, Okubo M, Nagaoka T and Hayakawa K 2002 IEEE Trans. Ind. Appl. 38 1168–
1173

323
[130] Kuroki T, Takahashi M, Okubo M and Yamamoto T 2002 IEEE Trans. Ind. Appl. 38 1204–
1209
[131] Okubo M, Tanioka G, Kuroki T and Yamamoto T 2002 IEEE Trans. Ind. Appl. 38 1196–1203
[132] Kim H H, Prieto G, Takashima K, Katsura S and Mizuno A 2002 J. Electrostatics 55 25–41
[133] Song Y, Kim S, Choi K and Yamamoto T 2002 J. Electrostatics 55 189–201
[134] Moon J D, Lee G T and Geum S T 2000 J. Electrostatics 50 1–15
[135] Ogata A, Yamanouchi K, Mizuno K, Kushiyama S and Yamamoto T 1999 IEEE Trans. Ind.
Appl. 35 1289–1295
[136] Ogata A, Shintani N, Mizuno K, Kushiyama S and Yamamoto T 1999 IEEE Trans. Ind. Appl.
35 753–759
[137] Shimizu K and Oda T 1999 IEEE Trans. Ind. Appl. 35 1311–1317
[138] Moon J D, Lee G T and Chung S H 1999 IEEE Trans. Ind. Appl. 35 1198–1204
[139] Okumoto M, Su Z, Katsura S and Mizuno A 1999 IEEE Trans. Ind. Appl. 35 1205–1210
[140] Yamamoto T and Jang B W L 1999 IEEE Trans. Ind. Appl. 35 736–742
[141] Kohno H, Berezin A A, Chang J S, Tamura M, Yamamoto T, Shibuya A and Honda S 1998
IEEE Trans. Ind. Appl. 34 953–966
[142] Okumoto M, Rajanikanth B S, Katsura S and Mizuno A 1998 IEEE Trans. Ind. Appl. 34 940–
944
[143] Ono R and Oda T 2000 IEEE Trans. Ind. Appl. 36 82–86
[144] Ohneda H, Harano A, Sadakata M and Takarada T 2002 J. Electrostatics 55 321–332
[145] Yamamoto T, Yang C L, Beltran M R and Kravets Z 2000 IEEE Trans. Ind. Appl. 36 923–
927
[146] Chang J S, Kostov K G, Urashima K, Yamamoto T, Okayasu Y, Kato T, Iwaizumi T and
Yoshimura K 2000 IEEE Trans. Ind. Appl. 36 1251–1259
[147] Urashima K, Chang J S, Park J Y, Lee D C, Chakrabarti A and Ito T 1998 IEEE Trans. Ind.
Appl. 34 934–939
[148] Mizuno A, Kisanuki Y, Noguchi M, Katsura S, Lee S W, Hong Y K, Shin S Y and Kang J H
1999 IEEE Trans. Ind. Appl. 35 1284–1288
[149] Janiczek T 2001 J. Electrostatics 51–52 167–172
[150] Mazurek B, Janiczek T and Chmielowiec J 2001 J. Electrostatics 51–52 76–81
[151] Gajewski J B 2000 J. Electrostatics 48 279–294
[155] Kacprzyk R and Gajewski J B 2001 J. Electrostatics 51–52 124–130
[156] Struminsky V I 2002 J. Electrostatics 54 301–310
[157] Liu C,Wang J, Yu K, Eliasson B, Xia Q, Xue Band Zhang Y 2002 J. Electrostatics 54 149–158
[158] Jayaram S, Xu X Y and Cross J D 2000 IEEE Trans. Ind. Appl. 36 920–922
[159] Shi X, Yumoto M and Sakai T 1999 IEEE Trans. Ind. Appl. 35 549–553
[160] Murata Y and Kubota K 1999 Proc. IOP Conf. Ser. No. 163 341–344
[161] Suehiro J, Hamada R, Noutomi D, Shutou M and Kara M 2003 J. Electrostatics 57 157–168
[162] Ohshima T, Okuyama K and Sato M 2002 J. Electrostatics 55 227–235
[163] Qian L, Scott M, Kaler K V I S and Paul R 2002 J. Electrostatics 55 65–79
[164] Lee H 2001 J. Electrostatics 51–52 71–75
[165] Cui L, Jiang J, Zhang L, Song C, Zhao W and Lin J 2001 J. Electrostatics 51–52 153–158
[166] Yamamoto T, Kurosawa O, Kabata H, Shimamoto N and Washizu M 2000 IEEE Trans. Ind.
Appl. 36 1010–1017
[167] Qin B, Barbosa-Canovas G V, Swanson B G, Pedrow P D and Olsen R G 1998 IEEE Trans.
Ind. Appl. 34 43–50
[168] Suzuki S, Yamanashi T, Tazawa S, Kurosawa O and Washizu M 1998 IEEE Trans. Ind. Appl.
34 75–83

324
[169] Horenstein M N and Stone P R 2001 J. Electrostatics 51–52 515–521

[170] Horenstein M N, Pappas S, Fishov A and Bifano T G 2002 J. Electrostatics 54 321–332
[171] Buchler M, Cheng L, Orient O, Thelen M, Gompf R, Bayliss J and Rauwerdink J 1999 Proc.
IOP Conf. Ser. No. 163 189–196
[172] Gross F B, Grek S B, Calle C I and Lee R U 2001 J. Electrostatics 53 257–266

325
Pulsed arc discharges for water treatment and

disinfection
H Z Zastawny1, H Romat1, N Karpel Leitner2, J S Chang3
1
Laboratoire d’Etudes Aerodynamiques, Electrohydrodynamic group, Bd Pierre
et Marie Curie, Téléport 2, BP 30179, 86962 Futuroscope-Chasseneuil, France
2
Laboratoire de Chimie de 1’Eau et de FEnvironnement, 40 av. du Recteur
Pineau 86000 Poitiers, France
3
McMaster University, Hamilton, Ontario, Canada
Abstract. Within the framework of a collaboration between our laboratory and McMaster
University, a physical process based on the properties of pulsed arc discharges within water
has been investigated for the treatment and the disinfection of water. The arcs are produced in
a cylindrical tank between two rod-to-rod electrodes. In this article we first present the
experimental device and then electrical results obtained (current and potential measured from
one of the electrodes) for different kinds of water, for varying applied potentials and for
several electrode gaps in the water. We also give the results that we obtained on the pressure
waveforms from piezoelectric pickups placed at two different distances from the arc zone.
Finally we analyse all these results in order to better understand the propagation of the shock
waves produced by the pulsed arc system.
1. Introduction
Most of the actual processes used for the treatment or the disinfection of water and which
are based on chemistry properties generate by-products which are increasingly regulated
by the European Union. Alternative processes have been studied during the last decade by
a certain number of authors in order to find a compromise between the efficiency and the
environmental impact.
Most of the research workers [1–3] have studied the electroacoustical efficiency of an
electrical discharge in water. J.S.Chang et al. [4] have particularly investigated the pulsed
arc discharges in water and showed that an arc could generate a strong shock wave, cavitation
bubbles as well as strong ultraviolet radiation and OH radicals. They tested this process on
zebra mussels empirically and successfully.
This paper focuses on the characterization of the electrical pulsed arc parameters and on
their influence on the generated pressure. Arcs are created in a cylindrical transparent pilot
between two electrodes, one of them being connected to a capacitor via a high voltage
spark gap switch. When this external switch allows current to circulate, all the electric
charge of the capacitor is commuted to a rod-to-rod system in the water, which produces a
plasma in the water and a compulsive pressure pulse [5].

326
2. Experimental arrangements
The experimental device, which is shown in figure 1, comprises a cylindrical tank, a pulsed
arc generator and the measurement system. The tank is made of Plexiglas and two rod-to-
rod type electrodes are placed on its revolution axis. The production of pulsed arcs between
these electrodes requires an electrical system which is composed of a capacitor (50 µF), its
charge circuit and a high voltage spark gap switch. The latter is composed of two rod-to-rod
type electrodes which are separated by air. The level of applied potential is fixed by the gap
between these electrodes. When the potential reaches the disruptive value for the air, an arc
is created in the air and a current can circulate through the air and after within water.
The voltage and the current are measured from the input electrode with respectively a
voltage divider type sensor and a coil type sensor, both linked to a Tektronix oscilloscope
which can record the signals.
Figure 1—Schema of the experimental apparatus
3. Electrical characterization of the discharge for different media

In a first series of experiments we tested different media such as tap water and solutions of
ferric or permanganate ions. For different electrode and spark gap values, we measured the
applied voltage and the discharge current from which we deduced the electrical power.
3.1. Influence of the discharge on the media

First we did tests on a violet solution of permanganate ions (KMnO4) of concentration 0.1
g.L-1. After one hour of discharges at the rate of one arc every 3 seconds, the final concentration
measured by colorimetry techniques was 0.05 g.L-1.
Next we did the same experiment with a solution of ferric ions (FeCl3) whose initial pH
was 3. After one hour of treatment the pH was 5 and the colour of the solution corresponded
to a decrease of 1/3 of the concentration of ferric ions.
These two rudimentary experiments show that arc discharges generate reactions within
liquid containing reducing agents (decrease of the concentration of ferric ions and
permanganate ions). The ferric ion (Fe3+) reduction involves the production of OH and Os
radicals which permit the elimination of some pollutants like pesticides [6].

327
3.2. Electrical parameters

Figure 2 shows a typical recording of the voltage, the current and the electrical power
obtained whatever the media in which arcs are produced, or the electrode gap and the
applied voltage. When the capacitor discharges through the spark gap switch, the potential
of the input electrode rises abruptly and then decreases regularly until breakdown, which
corresponds to the circulation of a weak conduction current in the water (a few hundreds of
amperes is not high compared to the thousands of amperes of maximum current during the
discharge!). When the breakdown occurs, the potential decreases rapidly to zero while the
current assumes the shape of a peak.
Figure 2—Typical voltage, current and electrical power recording
3.3. Parameters influencing on the discharge

The value of the conduction current before breakdown is always higher for the ferric solution
than in the other cases; it is 5% of the value of the current peak observed after the breakdown
within the ferric solution but 1% of it in the case of the permanganate solution. The solution
of ferric ions is more conductive than the permanganate or water.
Figure 3 (a) gives the evolution of the voltage stage time before the breakdown as a
function of the electrode gap between the electrodes in water when Vmax (potential at the
beginning of the stage) is equal to 4 kV. For a fixed electrode gap we can see that the lowest
time necessary before breakdown is for the ferric solution, which is likely to be due to the
conductivity of the liquid. In fact the duration of the stage before the arc discharge depends
on the power injected into the water by the conduction current. For a fixed potential, if the
conductivity of the liquid is high, its resistance is low and the power injected in the liquid
(V2/R) is important. So the system does not need as much time in this case as in the other
ones (with lower conductivity) to initiate the arc.
We also notice that whatever the discharge media, the more the electrode gap increases,
the longer the voltage stage lasts. When the gap increases, the resistance of the water
increases too, and for a fixed potential the power injected decreases. The time necessary is
then higher when the distance between the electrodes increases.

328
Figure 3(b) shows the results obtained for the voltage stage time as a function of Vmax
when the electrode gap is fixed at 1mm. We see that the voltage stage time decreases when
Vmax grows. For the same reason when Vmax increases the power injected increases too and
the time necessary to initiate the process of discharge diminishes (the arc starts when the
energy injected in the water is sufficient to initiate the process).
Figure 3—Voltage stage time as a function of the electrode gap (a) and as a function of Vmax (b)
Figure 4—Electrical energy released per discharge as a function of the electrode gap (a) and as a
function of Vmax (b)
Figure 4(a) gives the mean energy released during the discharge as a function of the gap
between the electrodes in the water for an applied potential Vmax of 4 kV. The mean electrical
energy per discharge has been calculated from the time integral of the electrical power
visualised on the oscilloscope. Whatever the media, this energy decreases when the electrode
gap grows. As shown in figure 5(b) for which the gap is fixed to 1 mm, it increases when the
applied potential Vmax increases.

329
The power released during the discharge is proportional to the applied potential and to the
inverse of the length of the plasma (resistance of the medium). In both graphs of figure 5 we
can see that indeed the energy (the mean power multiplied by the duration of the discharge
which is constant in any case: 50 µs) decreases when the length of the plasma (the gap)
increases and increases when the applied potential increases. We also notice in these graphs
that the energy released is dependant of the solution. The reason for this is not yet understood.
4. Pressure outline for underwater arcs

In order to understand the propagating pressure wave phenomena, pressure measurements
were carried out with arcs produced within water for a fixed electrode gap and for varying
spark gap values. As shown in figure 1, the piezoelectric pressure sensor were placed on the
lateral circular wall of the tank, 100 mm from the arc zone, or on the top of the cylinder, 50
mm from the arc zone.
Figure 5—Typical pressure wave, potential, current recording
A typical visualization of the pressure wave phenomena on the oscilloscope is shown in

figure 5. The presence of the pressure stage before the breakdown is due to electrical
interferences and must not be considered here. The pressure waveform starts 34 µs after the
discharge, as shown in the right-hand graph of figure 5, which corresponds roughly to the
theoretical time of the sound wave propagation within the water [7]. Its evolution shows a
sequence of peaks with an observable time period. This pressure waveform may represent a
spherical shock waveform which includes many reflections on the tank walls. Nevertheless
we must notice on the right-hand side graph of figure 5 that the shape of the first peak of the
pressure is very similar to the one of the current or of the power.
As shown in the figure 6, when the electrode gap is set to 2 mm and the potential Vmax is
equal to 4 kV, the first pressure peak we can observe after the discharge is about 10 bars
when the pressure is measured 100 mm from the arc zone and about 23 bars when it is
measured 50 mm from the arc zone.
Although the distance between the sensor and the arc may not be the only factor
influencing the pressure value and although the sensor does not face the arc similarly in
both positions the first pressure peak always increases with the potential Vmax. The more the

330
Figure 6—Pressure peak as a function of Vma
arc injects energy into the liquid, the higher the first peak is. Further experiments in which
the electrode gap and the sensor position will be varied are planned in the future.
5. Conclusion
Pulsed arc discharge generation involves reactions within polluted water: the presence of
certain oxidants induces the production of useful radicals for the eradication of some
pollutants.
Arc production within a liquid is influenced by the properties of the media, the applied
voltage and the electrode gap. The combination of these various parameters acts on the
setting up of the discharge: the resistance between the electrodes determined by the electrode
gap and the specific conductivity of the environment influences the passage of the
conduction current and induces the presence of a slow decreasing voltage stage.
Pressure wave propagation is one of the mechanisms generated by pulsed arcs. In our
experimental case, these pressure waves may have the shape of a spherical shock wave
which reflects on the pilot walls. In future work, a numerical simulation of this wave type
propagation will be compared to the experimental profile.
References
[1] Roi N A and Frolov D P 1958 Soviet Physics: Doklady 118–121
[2] Gavrilov G N et al. 1977 Soviet Physics: Technical Physics 22 868–870
[3] Okun’ I Z 1971 Soviet Physics: Technical Physics 16 216–226
[4] Chang J S, Looy P C and Urashima K 2000 IEEE Conference on Electrical Insulation and
Dieletric Phenomena 10 105–108
[5] Martin E A 1960 J. Appl. Phys. 31 n°2 225–267
[6] Foret C 2002 DEA Université de Poitiers LCEE
[7] Mortimer B J P, Skews B W, Felthun L T 1998 Shock Waves 8 63–69

331
Electrostatic Spray Application of Decontaminant Agents

onto the Human Body as a Bioterrorism Countermeasure:
Process Development and Evaluation
S Edward Law1, Steven C Cooper2 and Mark A Harrison3
1
Department of Biological and Agricultural Engineering, University of Georgia,
Athens, GA 30602–4435, USA
2
Electrostatic Spraying Systems, Inc., 62 Morrison St., Watkinsville, GA 30677,
USA
3
Department of Food Science and Technology, University of Georgia, Athens, GA
30602–7160, USA
Abstract. Electrostatic-induction charging of conductive liquids has been developed as an

effective method for rapidly applying decontaminant sprays (e.g., antitoxins, disinfectants,
sanitizers, etc.) and medicinal sprays onto the human body as a protective countermeasure
against chemical and biological agents contaminating cutaneous surfaces. Aqueous-based
sprays having typically -16 mC/kg charge-to-mass droplets of ~30 µm volume-median diameter
were electrostatically deposited by action of spacecharge and image-charge electric fields and
quantified fluorometrically for 52 body locations on a test mannequin as well as on a similar
size human subject. Averaged over all body sites, charged spray deposited significantly (α<
0.0001) greater areal density of liquid than did identically dispensed uncharged sprays, providing
~ 1.8-fold increase. The electrodeposition benefit ranged from ∼1.0- to 3.8-fold for specific
body sites, and charged sprays exhibited greater uniformity of deposition over the body (e.g.,
the coefficient of variability was 29% and 40%, respectively, for charged and uncharged spray
applications). Additionally, preliminary microbiological assessment, via deactivation of an
innocuous bacterium (Pseudomonas fluorescens) inoculated onto body target sites prior to
spraying a broad-spectrum quaternary ammonium decontaminant, showed that ~2.38 log10
reduction in colony-forming units was provided by the electrostatically-enhanced
countermeasure vs. only ~0.66 logic reduction for similarly applied uncharged decontaminant,
indicating ~50-fold greater bactericidal efficacy for the electrostatic method.
1. Introduction
Recent world events have prompted reassessment and refinement of civilian-defense
strategies and capabilities for whole populations [1]. In the case of airborne dispersal, there
exists potential need for an effective countermeasure and associated apparatus capable of
rapidly decontaminating chemically and biologically active agents deposited onto
cutaneous surfaces of the human body. System requirements dictate thorough spatial coverage
of the skin with decontaminant yet minimal waste and disposal of the material. Electrostatic
spray application of finely atomized droplets has been shown to meet similar deposition
requirements in numerous applications ranging from finish coatings onto manufactured
goods to chemical and biological pest-control agents onto agricultural crops [2]. This latter

332
system [3] has been developed having spray-charging attributes especially compatible
with the engineering and safety constraints necessary for applications onto humans, namely:
lowvoltage/low-current/low-capacitance electrostatic-induction spray-charging nozzles
operating at 1 kV dc or less providing 10–15 mC/kg spray charge-to-mass values; charged-
spray generation by pneumatic atomization of conductive liquids (e.g., 10–4–10 S/m) into
20–30 µm droplets having residual aerodynamic energy to penetrate the electrified droplets
into Faraday-shielded regions (e.g., underarm skin) with subsequent space-charge-field
deposition; and liquid flowrates in the 1–2 mL/s range.
This paper reports the design and experimental evaluation of a portable, microprocessor
controlled, walk-through booth facilitating electrostatic application of decontaminant (e.g.,
antitoxins, disinfectants, sanitizers, etc.) and medicinal sprays onto the human body. Design
operational conditions are 90 persons/h, 100 mL/person charged spray dispensed, and 20
mL/person liquid-waste disposal. Skin-deposition assessments include fluorescent tracer
documentation of spatial distribution and areal density of spray mass, as well as assessment
of microbiological efficacy of the electrostatic process via deactivation of an innocuous
microorganism applied to the skin prior to spraying an appropriate decontamination agent.
2. Experimental Analysis
The overall experimental facility for electrostatically applying sprays onto test subjects is
depicted by Fig. 1. It consisted of a dielectric (high-density polyethylene) spray chamber
having a grounded mesh-steel floor over a dielectric sump and a grounded aluminum footplate
on which the subject stood. An assembly of three induction-charging spray nozzles oscillated at
approximately 1.13 rad/s in the vertical plane to provide a 14 s sequence often 75° head-to-foot
spray sweeps over the standing body facing the nozzles spaced 1 m away. During a short
interlude the subject was rotated 180° and the spray sequence was repeated. Following a 30 s
quiescent period for residual spray deposition and/or evacuation, the subject exited the chamber.
Not shown is an attached 1.05 m wide×0.65 m long×2.08 m high process-support room housing
ancillary equipment for spray-liquid metering, pneumatic and induction-voltage inputs to the
charging nozzles, mechanical oscillation, and a microprocessor controller for properlysequencing
all operations. Detailed descriptions of additional apparatus, materials, procedures, and statistical
design of the experiment follow.
2.1. Apparatus and Materials

2.1.1. Spray-charging. Spray liquid of 0.1 S/m
conductivity was prepared by combining 1.5
g of fluorescent tracer particles (DayGlo
Corp.—Blaze Orange GT15N), 0.1 gNaCl, and
1.0 mL of Triton X-100 non-ionic surfactant
into 1000 mL of deionized water. The liquid
was pneumatically atomized (186 kPa) at 1.19
mL/s per nozzle into ~30 µm volume-median
diameter spray and inductively charged
to approximately –16 mC/kg by an
embeddedelectrode spray-charging nozzle
(Electrostatic Spraying Systems, Inc.—
MaxCharge™ model) operating at 1.1 kV dc. Fig. 1. Electrostatic spray chamber.
Similar nozzles typically exhibit a 10–15 uC/
kg’V linear spray-charging response vs. input voltage [4].

333
For the microbiological assessment, the spray liquid consisted of a 0.78% v/v tap-water
solution of the quaternary ammonium disinfectant Coverage Plus® (Merck and Co., St.
Louis, MO; EPA Reg. No. 6836–78–1043) containing Di-n-alkyl dimethyl ammonium
chloride and N-alkyl dimethyl benzyl ammonium chloride. Atomization and spray-charging
performance were similar to the tracer spray liquid above.
2.1.2. Target systems. A commercial mannequin was utilized to simulate a human subject
for most spray applications. The polymeric surface was rendered conductive by spray
coatings of a water-based EMF-shielding paint
(Staticveil®, Less EMF Inc., Ghent, NY); surface
resistivity on test cells measured 2 ohms (i.e., ohm/sq)
and typical resistance to earth was <1.5 kΩ from the
head and other body locations. Figure 2 shows the
1.85 m tall male mannequin, erect standing, with arms
modified for 20° outward orientation vis-àvis the
body’s central axis. Circular targets denote some of
the 52 deposition sites evaluated. To facilitate
replicated spray applications and to eliminate
background fluorescence attributable to the
mannequin’s surface, stainless steel flat washers (17.5
mm OD×5.16 mm ID×1.60 mm thick) were affixed to
the surface using flat-head stainless steel screws; an
identical backing washer was installed behind each
target washer to preclude surface contamination from
the mannequin. When installed, the effective target
area of each test washer was 2.07 cm2.
Abbreviated sites (viz., 11) were utilized for the
microbiological assessment and the human subject.
For the former, target washers were autoclaved for
sterilization, inoculated with the bacteria, and installed
with a sterile backing washer onto the mannequin
using sterile forceps. For the latter, double-sided paper
tape, with a copper-tape grounding jumper, attached
a target washer at each of the designated human-body
sites; negligible background fluorescence from the
tape’s adhesive to the target backside was confirmed. Fig. 2. Test mannequin.
2.2. Experimental Procedure

2.2.1. Fluorometric analysis. Spray tracer deposited onto the targets was reclaimed into 10
mL of wash liquid (0.1% Triton X-100 surfactant in deionized water) by 20 min agitation in
15 mL wash vessels [5]. Tracer deposition and its areal concentration (ng/cm2) were
subsequently quantified using a digital fluorometer (Turner model 450–005) exciting the
tracer by quartz-halogen irradiation (λ<440 nm) and measuring emission (λ>535 nm) to
include the tracer’s 602 nm dominant fluorescence. Calibration standards established a
linear (R2=0.999) fluorometer response. Four liquid samples from the wash cell for each
respective sprayed target were fluorometrically analyzed and fluorometric mean values
converted to (Spray-Liquid Deposition Density, nL/cm2)=3.08 (fluorometer reading)+33.55.

334
2.2.2. Microbiological assessment. The ubiquitous, nonsporeforming, Gram negative, aerobic

bacterium Pseudomonas fluorescens was selected for this assessment. A culture was grown in
tryptic soy broth at 28° C for 20–22 h before inoculating 10 uL onto the front surface of each
sterile test washer and drying in a laminar flow hood. Sterile procedures were observed for all
handling of the targets including attachment and detachment from the mannequin. After
spraying and detachment, each washer was placed into a 50 mL sterile centrifuge tube, 10 mL
of sterile 0.1% peptone buffer added, and the tube shaken vigorously for 60 s before a portion
of liquid from each was spiral plated onto plate-count agar plates. The plates were incubated
at 28° C for 48 h before colony-forming units (CFU) of bacteria were enumerated.
2.2.3. Experimental design. Using a randomized complete block statistical design, the two
spray treatments of the (52 body sites)×(2 charge conditions) were completed in a single
day; four day-replications were made. Spray-charge condition was charged or uncharged.
Experimental data were analyzed by SAS General Linear Models procedure using standard
analysis of variance (ANOVA) to identify statistically significant differences in mean values
of deposition density achieved by charged vs. uncharged application treatments onto
individual target sites as well as onto sites averaged over the body both regionally and
wholly. Coefficient of variability values were computed for estimates of deposition
uniformity for charged vs. uncharged spray treatments.

3.1. Deposition Analysis
Table 1 summarizes the mean values (n=4) for areal concentration (nL/cm2) of spray liquid
deposited onto specified target sites for the two spray-charge treatments. Accompanying
ANOVA results indicate which of the 52 treatment means significantly differ from one another.
Figures 3 and 4 illustrate, respectively, the trends in areal deposition density for the uncharged
vs. charged spray-application methods for the mannequin and the human subject.
Table 1. Mean values of spray-liquid deposition density onto designated mannequin and human body
sites for uncharged and charged spray applications, (nL/cm2).

335
3.1.1. Mannequin deposition. The

increases in spray deposition attributable
to electrostatic-force effects are presented
in Fig. 3 as ratios of charged deposition
divided by uncharged deposition at each
target site on the mannequin. These
increases ranged from 0.98-fold for sites
electrostatically shielded from the
oncoming spray stream (e.g., site 36) to a
maximum of 3.82-fold for difficult sites
parallel to the air-velocity vector (e.g., site
37) for which electrostatic forces prove
most beneficial in redirecting the spray
droplets for deposition. Averaged over all
52 target sites, charged spray deposited Fig. 3. Electrostatic enhancement of deposition.
significantly ( α < 0.0001) greater areal
density of liquid (2159 nL/cm2) than did uncharged spray (1211 nL/cm2) for a 1.78-fold
overall electrodeposition benefit. In addition to increasing the whole-body deposition,
electrostatic effects improved the uniformity of surface coverage by spray; charged-spray
applications exhibited a significantly (α=0.0567) lower coefficient of variability of 29% as
compared with 40% for uncharged spray.
3.1.2. Zonal deposition. In addition to whole-body analysis, mannequin target sites were
grouped into the following zones where detailed comparisons showed electrostatic spray
deposition within all zones to be significantly (␣ < 0.01) improved in both quantity and
uniformity: 1) Head/neck (sites #1–7)—1511 nL/cm2, cv 19% ch. vs. 764 nL/cm2, cv. 36%
unch.; 2) Torso (#8–15, 18, 19, 21, 29–32)—2224 nL/cm2, cv 24% ch. vs. 1247 nL/cm2, cv
30% unch.; 3) Appendages including arms/hands (#16, 17, 20, 22–28, 33, 34)—1878 nL/
cm2, cv 25% ch. vs. 1057 nL/cm2, cv 26% unch. and legs/feet (#35–52)—2549 nL/cm2, cv
23% ch. vs. 1490 nL/cm2, cv 38% unch.
3.1.3. Human deposition. Figure 4 graphically compares areal concentrations of spray-
liquid deposited at the 11 human-body sites by the uncharged and charged tracer sprays.
Averaged over all sites, whole-body charged-spray deposition (1261 nL/cm2) exceeded
uncharged-spray deposition (592 nL/cm2) by 2.13-fold.
3.2. Microbiological Assessment

Plate-count results from the unreplicated applications of decontaminant spray indicated
approximately a 47-fold greater reduction in bacteria CFU achieved by charged vs.
uncharged treatments when averaged over all target sites; individual charged-spray sites
exhibited typically only 0.1–9.7%) (median ~1.0%) surviving bacterial counts as
compared with counts for corresponding uncharged-spray sites. Overall bactericidal
efficacy, as expressed by log10 reductions in CFU from the typical 8 log10 value of the
inoculated but unsprayed control targets, was approximately 2.38 log10 and 0.66 log10,
respectively, for charged and uncharged applications of the 0.78% v/v aqueous spray of
broad-spectrum disinfectant.

336
Fig. 4. Charged and uncharged spray deposition achieved on human body.
4. Conclusion
Incorporation of electrostatic forces of attraction significantly increased the overall spray
deposition by ~2-fold onto cutaneous surfaces of the human body and onto a similar test
mannequin. For individual target sites, the electrostatic deposition benefit ranged from
~1.2to 4.3-fold for the human body and ~1.0- to 3.8-fold over the 52 mannequin body sites
evaluated. Additionally, charged spray exhibited a significantly more uniform deposition
pattern as compared with similarly applied uncharged spray. Preliminary microbiological
assessment documented an approximately 50-fold greater bactericidal efficacy for the
electrostatic method which portends well for its further development as an effective
bioterrorism countermeasure.
Acknowledgements
This work was supported in part by funds provided by the Georgia Agricultural Experiment
Stations. Appreciation is expressed to Mr. Patrick Harrell for assistance in fabrication of
experimental apparatus and graphics preparation as well as to Mr. Sean Ireland for assistance
in experimental and statistical analyses.
References
[1] World Health Organization. 2002. http://www.who.int/health_topics/bioterrorism/en.
[2] Law SE 2001 Jour. Electrostatics 51(1):25–42.
[3] Law SE Cooper SC and Law WB 1999 Inst. Phys. Conf. Ser. 163:243–248.
[4] Law SE 1978 Trans. ASAE. 21(6):1096–1104.
[5] Law SE and Cooper SC 2001 IEEE Trans. 1A-37(6):1597–1602.

337
Atmospheric turbulence and surface atmospheric

electricity observations
R G Harrison
Department of Meteorology, The University of Reading,
P.O Box 243, Barley Gate, Reading Berks RG6 6BB U.K.
Abstract. Durable and low maintenance atmospheric electricity sensors are required if
measurements which span long (year to decade) timescales are to be obtained. A passive
spherical sensor, together with a high impedance electrometer, presents one method of
monitoring the vertical Potential Gradient at the surface. In suitable conditions, such a sensor
can monitor the state of the global atmospheric electrical system. Local factors however, such
as air pollution and turbulence, influence the effectiveness of the sensor. In a series of
measurements to investigate its properties, a passive sphere collector has been operated alongside
(1) a long wire antenna and (2) a fine wire thermometer capable of responding to turbulent
fluctuations near to the surface. In the first case the potential of the wire was found to be
3.15±0.23 greater than that of the sphere, and in the second, the average monthly Potential
Gradient tended to an asymptotic value, as the magnitude of the turbulence increased.
1. Introduction
Long-term monitoring of the global climate system is increasingly important, and the
atmospheric electrical system is emerging as a long-neglected one aspect, which may have
an effect on cloud formation [1]. Over 70years of continuous measurements of the vertical
component of atmospheric electric field (or Potential Gradient, PG) at UK sites have recently
been discovered [2], which provide an important resource for such studies. A long-term
decline in the PG has recently been identified [3]. The UK measurements, at the Observatories
of Kew (London), Eskdalemuir (Scotland) and Lerwick (Shetland), span much of the
twentieth century, but ceased in the early 1980s. Although observations have continued at
other sites [4], no systematic measurements in the UK have been made since then. One
limitation is the availability of durable and low maintenance sensors, as it is now increasingly
common for regular environmental measurements to be obtained automatically, rather than
by using stations staffed with observers. A stable and well-characterised sensor is necessary
if the measurements are to be used to study long-term, climate-related changes in atmospheric
electricity.
Measurement of the surface electric field usually use mechanical field machines [5],
(field mills), or a suitable potential probe (collector) [6] connected to an ultra-high impedance
electrometer [7]. Although field machines offer a more rapid time response, they eventually
require maintenance because of wear to moving parts and general deterioration under
atmospheric conditions. In obtaining long-term averages, reliability is generally a more
important consideration than time response, and a collector-based instrument, schematically
represented in figure 1, is more appropriate. The PG is derived from measurement of the
potential V at the height z of the collector, and is typically ~150V.m-1 at 1m under fair

338
weather conditions. The collector acts to acquire the same atmospheric: radioactive sources,
flame probes, water droppers and long wire antennas have all been used for this purpose
[5,6]. The first three such collectors all require maintenance or special precautions, and the
associated apparatus may introduce geometrical distortions into the electric field. A long
wire antenna produces negligible field distortion [8], but requires high-integrity insulation
at each end and a careful tensioning arrangement. Only very small currents are furnished by
such antenna systems, typically of order 10-14A, and electrometer instrumentation is required
for the potential to be measured correctly. An alternative approach is investigated here,
using a small conducting sphere as the collector rather than a long wire antenna. By using
such a sphere with a diameter of a few centimetres, the collecting area is comparable with
that of a long wire antenna and a similar electrometer system. This has the advantage of
requiring only a single mast to support the collector, which presents some mechanical
simplifications in long-term deployment and maintenance. Figure 1 illustrates the operating
principle of this experimental arrangement. Near-surface turbulence mixes ions and charged
particles into the region near to the sensor, which acquires the local potential of the air.
Figure 1 The atmospheric potential determined by a passive spherical sensor exposed at a fixed height
connected to an electrometer voltmeter (left), and (right) practical installation of a sensor operating at
z=1m.
2. Calibration
In order to evaluate its operation in the atmosphere, a spherical collector was deployed
alongside a long-wire antenna system [7], with both sensors at z=1m above the surface. The
spherical collector was constructed from a standard table-tennis ball (radius a= 1.9cm)
coated with a conductive screening paint (Nickel Screening Compound, RS part no 247–
4267). The collector was mounted at one end of a 6.25mm diameter stainless steel rod,
spacing it a horizontal distance of 0.5m from a conducting vertical support mast. The
spacing rod was fixed to the mast using a nylon mounting, into which the rod was fitted
using a PTFE sleeve. The spherical collector assembly was positioned at the midpoint of
the 20m long-wire antenna, mounted at a horizontal distance of 1m from the antenna. Both
the antenna and spherical collector were connected to independent electrometers. Figure 2
shows the data obtained during a summer’s day. The wire potential Vw is always greater than
the sphere’s potential Vs, but the two track each other closely. From many such measurements,
it was established that the ratio k of the wire potential to sphere potential (Vw/Vs) is 3.15±0.23.
when both sensors are operated at the same height. This electrostatic reduction in the

339
potential greatly relaxes the dynamic range required of the electrometer attached to the
sphere, when compared with an equivalent wire, and provides an additional reason to use a
sensor with simple spherical geometry.
Figure 2. Time series of potentials Vs and Vw obtained during a comparison of the passive sphere collector
operating adjacent to a long wire antenna, both at 1m above the surface. The data values are 5minute
averages from IHz samples.
3. Diurnal variations
A failing of a wire antenna for long term use is that, depending on the sensitivity of the
electrometer deployed and the conductivity of the air, physical circumstances may arise in
which there is insufficient current supply to the antenna for it to function correctly. Correct
operation requires turbulent mixing of natural atmospheric ions into the air adjacent to the
antenna, which in turn depends on the micrometeorological conditions. In previous work
[9], it has been found that turbulent mixing provides the conditions for satisfactory operation
of an antenna. One method for quantifying the amount of atmospheric surface layer turbulent
mixing is to use the Richardson number [10] Ri, and satisfactory operation of the passive
wire has previously been found empirically to require Ri<1.0. Routine calculation of Ri
requires considerably more instrumentation, however, so a rather simpler assessment of the
operating regimes is desirable, preferably with only one simple additional measurement.
Similar operating considerations apply to the passive sphere sensor, as the supply of current
is again limited by the effective area of the sensor, and will be influenced by the charge
transport from atmospheric turbulence. An investigation of these limitations was made at
the same time as the calibration in figure 2 was carried out, by installing a sensitive fine
wire Pt thermometer [11] close to the passive sphere sensor, figure 3. Figure 4 shows the
potential on both the wire and sphere sensors as a function of time, and the temperature
variations. It is clear that the potential on the wire collapses at the beginning and end of the
measurements, but that the sphere shows less variation. Short-term variations in temperature
do not lead to large changes in the measurements and the operation of the passive sphere
cannot be reliably inferred from the rate of change of temperature.

340
Figure 3. View of passive sphere collector operating adjacent to a fine wire thermometer, both at z=1m
above the surface.
Figure 4. Variation of potential on both the long wire antenna and passive sphere sensor, during the same
day in July 2002 shown in figure 2. The air temperature, measured with a fine wire thermometer, is also
shown.
4. Turbulence
In a longer series of measurements spanning several months at Eskdalemuir Observatory,
Scotland, in the latter half of 2002, a fine wire thermometer was operated continuously
alongside the passive sphere sensor. The instruments were sampled at IHz, and averages
computed at 5minute intervals. Figure 5 shows the data obtained for a full day, with fair
conditions for the afternoon. Comparing this data with that of figure 4, it is clear that the
operation of the sensor is, again, not related to absolute temperature. Satisfactory operation
of the sensor occurs when there are large and regular fluctuations occurring in the air

341
temperature. This again indicates the importance of adequate turbulent mixing in ensuring
adequate charge exchange with the sensor, as in the case of a long wire antenna.
Figure 5. Time series of air temperature and sphere potential (both at 1m), for instruments at Eskdalemuir
Observatory, Scotland on 30th October 2002.
As the air temperature and its rate of change are not sufficient to identify the different
operating regimes of the passive sphere, a further analysis of the effects of turbulence was
undertaken. The turbulent fluctuations in temperature can also be determined from the
fine wire thermometer, which has a rapid time response [11]. From micrometeorological
theory, the standard deviation of air temperature σT and turbulent temperature scale T* are
related by
(1)
where ψ is a semi-empirical function of the sensor height z, the local displacement height
d and the Obukhov Length L, a function of atmospheric stability [10]. σ T therefore provides
an indication of the turbulence. Turbulence considerably varies between strong mechanical
and thermally generated turbulence in daytime (unstable) conditions, to weak turbulence
in nocturnal (stable) conditions.
Standard deviations of temperature σ T were calculated for 15minute periods, again
using 1Hz data. The average sphere potential measured in the same interval was plotted
against the temperature standard deviation, for a set of measurements extending over one
month, Figure 6. It is clear that sphere’s potential is sensitive to the amount of turbulent
mixing and that σ T does provide a readily measured indicator of the controlling turbulence
scales. Using a logarithmic fit, the asymptotic value of the equivalent potential at 1m can
be estimated.

342
Figure 6. The average equivalent potential at 1m (kVs) using data from one month, found by selecting the
values of Vs when the surface layer turbulence, as measured by the standard deviation of temperature σT,
is greater than the specified minimum value on the x-axis.
5. Discussion
Long term measurements of atmospheric potential gradient, derived from a point measurement
of potential require simple and low maintenance instrumentation. A passive sphere sensor,
connected to a sensitive electrometer, provides measurements of potential when there is
sufficient turbulent mixing for sufficient charge exchange with its surroundings to occur.
As a result it will only operate for a fraction of the day, however this will not present a
significant problem if a long-term average is sought. The practical requirement of fair
weather conditions is another limitation always present, and therefore monthly averages
are typically derived. Measurements of standard deviations of temperature have been
demonstrated to provide an independent assessment of the operation of the sensor, although
further characterisation of the micrometeorological effects is desirable.
Acknowledgement
Assistance was given by the British Geological Survey and Met Office at Eskdalemuir.
6. References
[1] Carslaw K S, Harrison R G and Kirkby J 2002 Science 298, 5599, 1732–1737
[2] Harrison R G 2003 Weather 58, 11–19
[3] Harrison R G, 2002 Geophys Res Lett, 29(7) DOI 10.1029/2002GL014878
[4] Märcz F 1990 Ann. Geophysicae 8, 525–530
[5] Chalmers J A 1967 Atmospheric Electricity, (2nd edition, Pergamon Press, New York,)
[6] Israel H 1973 Atmospheric Electricity (Israel Programme for Scientific Translations)
[7] Harrison R G 1997 Rev. Sci. Inst 68, 3, 1599
[8] Crozier W D 1965 J. Geophys. Res. 68, 3451
[9] Barlow J F and Harrison R G 1999 In Christian H J (ed) Proc 11th International Conference on
Atmospheric Electricity, Guntersville, Alabama (NASA/CP-1999–209261, 575,)
[10] Kaimal J C and Finnigan J J 1994 Atmospheric boundary layer flows: their structure and
measurement (Oxford University Press)
[11] Harrison R G and Pedder M A 2001 Rev Sci Inst, 72, 2, 539

343
Dust particles removal by a novel two-stage electrostatic

precipitator
A Jaworek1, A Krupa1, K Adamiak2
1
Institute of Fluid Flow Machinery, Polish Academy of Sciences, Fiszera 14, 80–
952 Gdañsk, POLAND
2
Department of Electrical and Computer Engineering, University of Western Ontario,
London, Ontario, CANADA N6A 5B9
Abstract. A two-stage electrostatic precipitator is a gas-cleaning device in which the charging

and precipitation processes are separated. This device can be used to avoid back-corona
discharge, when dust of high resistivity has to be removed from an exhaust gas. A bench-top
size laboratory model of the two-stage electrostatic precipitator was investigated experimentally.
The charging stage was accomplished using the alternating electric field charger, a device that
allows charging the particles to higher levels than conventional DC-corona chargers. The
precipitation stage was created by a set of five parallel plates, two of which were maintained at
high potential and the remainder grounded.
1. Introduction
In order to avoid difficulties in collecting fly ash of high electrical resistivity, the charging
and collection processes can be separated and accomplished in two different stages. Such a
device is known as two-stage electrostatic precipitator. First, the particles are passed through
the charging stage, where they are charged in a mono-ionised electric field, similar to a
conventional electrostatic precipitator, where the voltage has to be reduced to eliminate the
back-corona discharge. Next, the particles enter the precipitation stage, which is usually
formed by a set of parallel plates with every second plate maintained at high potential and
remainder grounded. Such a set of electrodes is free of the back-corona discharge due to
lack of the ionic current.
The two-stage electrostatic precipitator was considered first by Masuda and Hosokawa
[1] to control the emission of high resistivity dust. The boxer charger was used in their
experiments as the charging stage [2,3]. The collection section was made of 4 electrodes
spaced at 150 mm. Different types of precharger were developed and tested in two-stage
electrostatic precipitators since: a nozzle charger equipped with corona electrodes, through
which the particles flowed with high velocity to prevent them from precipitation [4], corona-
triode charger [4], and a quadrupole precharger with four passive rods placed at a square
corners and a discharge electrode at the centre of this arrangement [5].
In this paper, the alternating electric field charger [6–8] is proposed and tested for the
first time as a precharger in a two-stage electrostatic precipitator. The particles are charged
by an ionic current in the alternating electric field, and the charge imparted to the particles
is close to the Pauthenier limit, i.e., is much higher than that available in other techniques.
Regardless of numerous investigations on the two-stage electrostatic precipitators, many
questions remained unanswered. One issue is the re-entrainment of the dust particles from

344
the collection electrodes due to lack of cohesive forces, which are present in standard
electrostatic precipitator because of the ionic current flowing through the dust layer. This
problem was, hitherto, not investigated and is considered in this paper.
2. Experimental
The schematic view of experimental model of the two-stage electrostatic precipitator with
alternating electric field precharger is shown in Fig.1. The precharger and collection
electrodes were mounted in a duct of square cross section of 160×160 mm, made of clear
acrylic plastic. The precharger consisted of two parallel grids made of 9 stainless steel rods,
2 mm in diameter spaced 20 mm from each other, and two discharge electrodes placed
outside of the grids. Each discharge electrode was made of a brass sheet with 98 stainless
steel pins 0.6 mm in diameter and 5 mm long, protruding 5 mm from the sheet. The distance
between the grids was 50 mm, and from each grid to the tip of the pins behind 45 mm. The
precharger electrode system was connected to two high voltage transformers (ZWAR-ABB,
Poland, model UMZ-24) connected in anti-phase, through the circuit made of diodes and
resistors, as shown in Fig.1. The ionic charging current is emitted by the discharge electrodes,
and flows through the grids to the charging zone between the grids. The alternating electric
field existing within the charging zone causes the particles to oscillate, but prevents them
from migrating to the charger electrodes, thus decreasing their precipitation at this stage.
Fig.1. Schematic diagram of the experimental set-up of two-stage electrostatic precipitator with the
alternating electric field precharger (not in scale).
The electrodes in the collection section were made of a brass sheet with 1 mm thickness.
Five parallel plates 130 mm high and 160 mm long downstream were placed in the duct.
The space between them was 30 mm. Two plates were maintained at high potential from the
high voltage power supply (SPELMANN HV model SL600W/30kV/P), and other three
were grounded. A flow straightener was mounted at the inlet of the channel to eliminate the
air vortices. At the outlet of the duct a collection grid, made of fine phosphor-bronze mesh
of holes of about 40 µm, was used to measure the electric charge carried by dust particles,
which escaped the precipitator.
Dust from the last stage of an electrostatic precipitator installed in the Lambton Power
Station, Ontario, Canada of mean particle diameter of about 10 µm was used in the
experiments. The particles were fluidised by compressed air and injected into the channel,
where they were charged by the alternating-electric-field charger. The airflow through the

345
duct was forced by a suction fan placed at the outlet of the duct. The trajectories of the
particles in the collection stage were recorded with a CCD digital camera (SONY DCR-
TRV525). The space between the collection electrodes of the precipitator was illuminated
by a light beam of about 2 mm in thickness.
The gas velocity in the channel was
measured by a hot wire anemometer (Mini
Anemometer Series 490, Kurtz Instruments
Inc.). The size distribution of the particles
was measured by Malvern Particle Size
Analyser model 2600. The Sauter mean
diameter of the particles D3,2 was equal to
about 10 µm (Fig.2). The dust volume
concentration within the duct measured by
the Malvern was in the range from 0.0004 to
Fig.2. Fly ash particle size distribution 0.0014 %.
The obscuration was measured behind the collection section by means of the Malvern
Particle Size Analyser. The obscuration ratio, shown in Fig.3, is the relation of obscuration
measured for the voltage applied to the collection section, to the obscuration when the
voltage was off and the dust could freely flow through the precipitator. Because the fluidiser
does not work smoothly and the particle concentration changes with time, the switch-on/
switch-off procedure was repeated a few times, and the ratio in Fig.3 is a mean value of the
obscurations measured during these periods. The obscuration ratio decreases with increasing
voltage applied to the collection electrodes and ratio can be a measure of the precipitator
collection efficiency. The collection efficiency can be estimated to 60–70% for fly ash
particles, and over 90% for MgO, if the voltage is sufficiently high.
Fig.3. Obscuration ratio at the outlet of the two-stage electrostatic precipitator for fly ash (a) and MgO
(b) particles.
The particle migration velocity to the collection electrodes results from the balance of
Stokes drag force:
Fs=πη6rvy (1)
and electric field force:
Fe=QE (2)
where: Q is the particle charge, E the electric field between the electrodes, η-the gas viscosity,
and r -the particle radius.

346
Assuming the AC charger is supplied by a voltage of 14 kVrms, the Pauthenier charge limit
for the fly ash particles of 10 µm in diameter is 2.3 fC (assuming that its dielectric constant is
that of SiO2—the main component of the fly ash), and for MgO is 2.73 fC. The velocity,
transversal to the gas flow, for the particle charged to the Pauthenier limit is 0.38 m/s and 0.54
m/s, respectively, for the collection electrode voltage equal to 12 kV. This means all particles
should be collected over the first few centimetres of the collection section.
The problem of particle re-entrainment was also studied. It was noticed that at higher
voltages (10–12 kV) the fly ash particles begin to jump between the collection electrodes.
Examples of the dust particle motion between the electrodes are shown in Figs.4 and 5.
Each figure is a combination of a few consecutive frames of the video recordings. The
particles of MgO (Fig.5) detach as agglomerates, while the fly ash particles are much
numerous, but because they are single they are hardly visible (Fig.4). The electric force is
sufficiently high to overcome the cohesion forces and some particles are removed from the
dust layer due to charge obtained by conduction. These particles flow to the opposite
electrode, where they can be recharged. After a few ‘jumps’ they can ultimately leave the
interelectrode space. The re-entrainment of the particles is also evidenced by increasing
obscuration measured behind the collection stage. Using the trajectories of the particles re-
entrained from the collection electrodes their charge can be estimated using the equation:
(3)
where α is the angle between the particle trajectory and the gas velocity vector, d—the
electrode spacing. It is assumed that no cluster is detached from the collection electrodes, but
only a single fly ash particle of mean size 10 µm. For these particles the charge Qr is 1.4 fC.
The phenomena are quite different for particles of high cohesiveness, like MgO. The
particles are easily agglomerated and cannot be re-entrained even for higher voltages, close
to the discharge onset level. Only rare jumps were observed, and usually large agglomerates
were involved (Fig.5). The particle re-entrainment did not have a noticeable effect on the
current, probably because the number of both positively and negatively charged re-entrained
particles was balanced.
Fig.4. Trajectories of re-entrained fly ash Fig.5. Trajectories of MgO particles between
particles between collection electrodes. the electrodes of the collection section.
From the recorded trajectories of the particles between the collection electrodes it can be
seen that the dust particles, or dust cluster, detach from the electrode due to the electrostatic
force as soon as it is re-charged to the opposite polarity. Then, they flow to the opposite
electrode. After impact on this electrode the following situations are possible:

347
1. The particle rebounds off the electrode, switching the polarity of its charge, and with
opposite charge returns to the flow. The particle can also be recharged by oppositely
charged particles and as neutralised returns to the flow, as shown in Fig.5;
2. The particle rebounds off the electrode with opposite charge, and returns directly to
the opposite electrode;
3. The particle impacts the electrode and sputters a new particle oppositely charged, but
remains itself on the electrode surface.
The current at the precipitator outlet was
also determined to estimate charged particle
concentration leaving the precipitator. The
ratio of the currents flowing to the
collection grid with and without voltage
applied to the collection electrodes for the
fly ash particles is presented in Fig.6. The
current ratio decreases with increasing
voltage, but for the voltage above about
14 kV it sharply increases because of an
Fig.6. Current ratio at the precipitator outlet vs.
collection electrode voltage for fly ash particles
electrical discharge from the collection
electrodes.
4. Discussion
Two-stage electrostatic precipitators were proposed 20 years ago to increase the collection
efficiency of fine particles, especially with high resistivity, and to prevent back-corona
discharge. However, many experiments show (cf. Table 1) that this is not a highly effective
device. Usually, the collection efficiency is lower than 90% mainly because of, as it was
shown in our experiment, low cohesiveness of the collected dust. Higher collection
efficiencies were obtained only for highly cohesive particles, mainly DOP (cf. Table 1) or
with simultaneous spraying of the water aerosol (Yoo et al. [13]).
Table 1. Two-stage electrostatic precipitators—comparison of the results

348
From the previous experiments it can be also deduced, and this remains in agreement with
our results, that the method of particle charging is less important than the particle collecting
technique. Bart et al. [10] subjected the particles to oscillatory motion in order to promote
their agglomeration in the collection section by using AC electric field, but the collection
efficiency in this type of precipitator was not very high. Chang et al. [5] succeeded with
high collection efficiency for particles of diameter larger than 0.1 µm, because the collection
stage was of the wire-plate geometry with discharge current reduced to prevent the back-
corona discharge. However, no significant effect of precharging on the collection efficiency
was observed for particles smaller than 0.1 µm.
5. Conclusions
From the experimental results presented in this paper it can be concluded that the cohesive
forces play a fundamental role in two-stage electrostatic precipitators. For dust of high
cohesiveness (MgO) the collection efficiency is very high and only seldom re-entraining
particles were observed. For particles of low cohesiveness (fly ash) strong re-entrainment
was recorded, which was mainly caused by the electric forces in the collection stage. A
charge conducted to a particle from the electrode causes this particle to be detached from
one electrode and attracted to the opposite. The zigzag motion of particles between the
collection electrodes and their flow off the precipitator were observed. This implies that the
charge on the particle is the same but of opposite polarity and is acquired by induction
processes. However, the mechanism of dust particle rebounding is unclear. The angles of
impact and rebounding are usually equal and full understanding of the role of mechanical
and electrical forces requires further investigations.
Two-stage electrostatic precipitation can be effective only if the cohesiveness of the
dust will be increased. Moisturising of the dust layer may be one of possible remedies.
Acknowledgements.
This paper has been supported by the Natural Sciences and Engineering Research Council
(NSERC) of Canada. The authors wish to thank Mr. Randall M. Van Hooren from the Ontario
Power Generation for providing the precipitator dust samples. The authors are grateful to Mr.
Marian Jaworski from the Department of Mechanical and Materials Engineering, University
of Western Ontario for his technical help, which have made these investigations possible.
References
1. Masuda S., Hosokawa Sh., IEEE Ind. Appl. Soc. Cof. Rec. (1982), 1094–1101
2. Masuda S., 2nd Int. Conf. Electrostatic Precipitators, Kyoto, Nov. 1984, 177–85
3. Masuda S., Washizu M., Mizuno A., Akutsu K., Conf. Electrostatic Precipit., Leura, 21–24 Aug. 1978
4. McLean K.J., IEE Rev. 135 (1988), Pt.A, No.6, 347–361
5. Chang J.S., Looy P.C., Webster C., Berezin A.A., Zukeran A., Ito T., 7th Int. Conf. Electrostatic
Precipitation. 20–25 Sept 1998, Kyongju, 620–627
6. Jaworek A., Krupa A., J. Electrostat. 23 (1989), 361–370
7. Adamiak K., Krupa A., Jaworek A., Inst. Phys. Conf. Ser. No. 143, Bristol 1995, 275–278
8. Lackowski M., J. Electrostat. 51–52 (2001), 225–231
9. Hautanen J., Janka K., Keskinen J., Lehtimäki M., Kivisto T., J. Aerosol Sci. 17 (1986), No.3, 622–626
10. Bart S.F., Melcher J.R., Ehrlich R.M., Ind. Eng. Chem Res. 27 (1988), No.l, 123–131
11. Asano K., Higashiyama Y., Yatsuzuka K., Nishimura R., 6th Int. Conf. Electrostatic Precipitation.
18–21 June 1996, Budapest, 57
12. Asano K., Choi Ch., Yatsuzuka K., Lim H., Polish-Japanese Symp. Non-Thermal Plasma
Processing of Water and Air. 29–31 May 2000, Sopot, 101–104
13. Yoo K.H., Lee J.S., Oh M.D., Aerosol Sci. Technol. 27 (1997), No.9, 308–323
14. Yoo K.H., Lee H., Choi M., Lee J.S., Oh M.D., 7th Int. Conf. Electrostatic Precipitation. 20–25
Sept 1998, Kyongju, 93–98
15. Kawada Y., Jindai W., Zukeran A., Ehara Y., Ito T., Takahashi T., Takamastu T., 7th Int. Conf.
Electrostatic Precipitation. 20–25 Sept 1998, Kyongju, 76–83
16. Kubo T., Kawada Y., Zukeran A., Ehara Y., Ito T., Takahashi T., Kawakami Takamatsu T., J.
Aerosol Sci. 30 (1999), Suppl.l, 793–794
349
Dust particle removal by wet-type electrostatic

scrubber
A Krupa1, A Jaworek1, T Czech1, M Lackowski1, J Luckner2
1
Institute of Fluid Flow Machinery, Polish Academy of Science
P.O. Box 621, PL-80952 Gdansk, Poland, jaworek@imp.gda.pl
2
Military Institute of Chemistry and Radiometry
Al. Gen. A. Chrusciela 105, 00–910 Warsaw, Poland
Abstract. Cleaning of exhaust gases from micrometer size particles by charged aerosol was
investigated experimentally. The aerosol was generated by electrospraying using a multinozzle
spray system. The dust particles were charged in a dc corona discharge charger. The dust was
deposited on the droplets due to Coulomb forces. The collection efficiency for fly ash particles
of mean size of 5 µm was higher than 90%, and water consumption lower than 0.08 1/m3.
1. Introduction
The efficiency of removal of dust particles smaller than a few micrometers in diameter by
conventional electrostatic precipitators decreases with decreasing particle size. The electric
charge on such particles cannot be too large and, as a result, the air drag force can be
stronger than the electric force that causes the particles to pass the precipitator. Additionally,
small fly ash particles can be easily re-entrained from the collection electrode because the
cohesive forces are very weak. Conventional wet scrubbers are also ineffective in removal
fine particles from exhaust gases due to low inertial forces on the dust particles, which can
flow around the droplet following the stream lines.
The removal of small particles can be enhanced when the dust particles and the scrubbing
droplets are electrically charged to the same or opposite polarities [1–5]. The charged
droplets sweeping the precipitation chamber act as small collectors, attracting the particles
due to Coulomb force. This method can be of much higher overall collection efficiency as
compared to the conventional electrostatic precipitators or inertial scrubbers.
This paper presents the experimental results of measurements of the collection efficiency
of a laboratory scale electrostatic scrubber. Usually, mechanical (rotary, pneumatic) atomisers
with induction charging were tested as a source of charged droplets. In this paper we use an
electrohydrodynamic atomiser for generation charged spray.
2. Experimental
The measurements were carried out in the experimental system shown schematically in
Fig.1. The channel of rectangular cross section of 160×160 mm was made of plexi glass.
The particles were charged in a corona discharge between a needle electrode and a grid. The
discharge electrode was made of a stainless steel needle 0.9 mm in diameter and 5 mm long.
The grid was made as a set of stainless steel rods 2 mm in diameter, and spaced at 20 mm

350
each other. The distance between the needle tip and the grid was 20 mm. The needle was at
the potential of -12 kV or +13 kV relative to the grid. The ionic current emitted by the
discharge electrode flows to the grid. The dust particles, which are injected through 6 holes
placed around the needle, are charged by this ionic current.
Fig. 1. Schematic view of an experimental arrangement for electrostatic scrubbing

investigations.
An electrospray system was used as a source of charged droplets. The spray system was
constructed from four stainless steel capillaries 0.5 mm i.d., 0.7 mm o.d, and 12 mm long,
and grounded electrodes placed in the plane of the nozzle tips (Fig.2a). The ground electrodes
were ring-shaped, and they were made of brass wire 3 mm in diameter. The inner diameter of
each ring was 28 mm. The distance between the spray nozzles was 30 mm. The nozzles were
connected to a high voltage power supply (SPELMANN HV model SL600W/30kV/P). A
photograph of the spray plume generated by the system, taken out of the channel, is presented
in Fig.2b. The droplets size distribution was measured by an optical particle size analyser,
model AWK made by K&K (Poland). The voltage at the capillaries was adjust to the
magnitude at which the precession mode (cf.[6]) of spraying was obtained. This mode is
characteristic in that the whole jet rotates regularly around the capillary axis, and its end
also rotates around the mother jet. For this mode, the spray plume is most wider and the
droplets are relatively small, that can be advantageous for the scrubbing applications. The
size distribution of distilled water droplets is shown in Fig.3. The number and volume mean
sizes were of about 80–90 µm in diameter (Fig.3).
At the outlet of the duct a collection two-layer paper-fibrous filter was used to collect the
dust escaping the scrubber. The filter was equipped with a metal grid made of fine mesh to
measure the charge carried by dust particles. The mesh was made of phosphor-bronze wire
with holes of about 40 µm. At the inlet of the channel a flow straightener was mounted to
eliminate the vortices. The air flow velocity was controlled by the hot wire anemometer TSI
model 8455.
A grounded guard electrode was placed between the spray system and the collection
grid to eliminate the leakage current from the high voltage electrodes, which could disturb
the measurements of the current from the grid.

351
Fig.2. Scheme of the four-nozzle spray system (a), a photograph of the spray plume of the four-
nozzle spray system (voltage to the nozzles 10 kV) (b).
Fig.3. Number and volume size distributions of the spray generated in the four-nozzle spray system
for voltage of 10 kV to the nozzles.
Dust from the last stage of an electrostatic precipitator of mean particle size of about 5
µm was used in the experiments. The dust size distribution was determined using Laser
Aerosol Particle Size Spectrometer model LAP 320 of TOPAS GmbH. The mean size of dust
collected on the filter at the duct outlet, i.e., that escaping the electrostatic scrubber was 6
µm. The size distribution of both samples of the dust are shown in Fig.4.
The dust/air mixture was generated in a specially designed fluidised bed by compressed
air, and injected into the channel through holes in the corona charger. The air flow through
the duct was forced by a suction blower placed at the outlet of the duct.

352
Fig.4. The size distribution of the dust used in the experiments and that escaping the scrubber.
The collection efficiency of the electrostatic scrubber was estimated by weighting the dust
injected to the scrubber and that collected on the paper tissue. The total mass min of the dust
injected to the scrubber is the difference between the dust weight in the fluidiser before and
after each run of the experiment. The mass mcoll of the dust collected by the paper tissue was
determined by weighting the tissue before and after the run. Two consecutive runs, 6
minutes each, were necessary for estimation the collection efficiency. The first run was for
the scrubber operating without the spray and corona charger, and the second one with the
charger and spray system turned on. The collection efficiency was estimated from the
equation:
(1)
where: mcoll/on and mcoll/off is the mass of the dust collected on the tissue with and without the
spray, respectively, and min/on and min/off is the mass of the dust injected to the channel, for
each case respectively.
The dust loading in the experiments was about 1 g/m3, and mean particle concentration
was estimated to the order of magnitude of 109 particles per cubic meter. The water
consumption ranged from 0.03 to 0.08 1/m3 depending on gas velocity. The specific charge
of the dust particles estimated from the total current from the metal mesh and the mass of the
collected dust varied between 10 and 20 mC/kg, irrespective of gas velocity. The variations
in specific charge were probably caused by changes in spraying and charging conditions,
including changes in air humidity and uncontrollable corona discharge current or
tribocharging. The charge on a particle of mean size of 6 µm determined from the measurements
of the current collected by the mesh at the channel outlet can be estimated to 2–4 fC.
The collection efficiency determined in these experiments is shown in Fig.5. It can be
noticed that charging the droplets and dust particles result in more efficient removal of the
particles by the scrubber. Charging the droplets and the dust to opposite polarities allows
the particles to be collected on the droplets while traversing the aerosol shower. Charging
the dust and droplets to the same polarity cleans the gas more effectively than opposite
charging. This result can be explained by repulsion of dust particles by the aerosol shower,

353
and precipitation them on the duct walls. The cleaning of the gas from fine particles is
more efficient for lower velocity. The collection efficiency by charged aerosol for uncharged
dust is also shown in Fig.5 for reference. The collection efficiency drops very quickly with
gas velocity increasing. Further increase of the collection efficiency can be achieved by
using a two-stage scrubber, that however, results in higher water consumption.
Fig.5. Collection efficiency of the electrostatic scrubber.
4. Conclusions
The results of experimental investigation of removal of dust particles by electrostatic
scrubber have been presented in the paper. The results show that removal of dust particles
by water droplets can be increased by electrical charging of the droplets and particles to the
same or opposite polarities. The equipment utilising electrostatic forces operates more
effectively for lower gas velocities than that in which inertial collection is dominant. The
water consumption in this type of cleaning device is also very low, it can be lower than 100
ml/m3, that is much lower than that in conventional inertial scrubbers. The collection
efficiency was up to 90% for dust and droplets charged to the same polarity and gas
velocity of about 0.2 m/s. This method can be applied for industrial or flue gas cleaning. In
an industrial scale equipment the spray system could be installed in wider channel but of
similar height.
Acknowledgements
This paper has been supported by the State Committee for Scientific Research of Poland
(KBN Grant No. 1440/T10/2000718).
References
1. Kraemer H.F., Johnstone H.F., Collection of Aerosol Particles in Presence of Electrostatic Fields.
Ind. Eng. Chem. 47 (1955) No 12, 2426–2434

354
2. Schmidt M., Theoretische und experimentelle Untersuchungen zum Einfluss elektrostatischer

Effecte auf die Nassentstaubung. Dissertation. Universitat Karlsruhe 1993 (in German)
3. Metzler P., Weiss P., Büttner H., Ebert F., Electrostatic Enhancement of Dust Separation in a
Nozzle Scrubber. J. Electrostatics 42 (1997), No. 1–2, 123–141
4. Jaworek A., Krupa A., Adamiak K., Submicron Charged Dust Particle Interception by Charged
Drops. IEEE Trans. Ind. Appl. 34 (1998), No.5, 985–991
5. Adamiak K., Jaworek A., Krupa A., Deposition Efficiency of Dust Particles on Water Droplets in
Electrostatic Scrubbers. IEEE Ind. Appl. Soc. Annual Meeting, St. Louis, Oct. 12–15, 1998,
1919–1926
6. Jaworek A., Krupa A., Classification of the modes of ehd spraying. J. Aerosol Sci. 30 (1999),
No.7, 873–893

355
Expert System Applications for Electrostatic Separation

Processes
Lucian Dascalescu,1 Adrian Samuila,1 Michaela Mihailescu,2
Alexandru Iuga,2 Reiner Köhnlechner3
1
Electronic&Electrostatics Research Unit, LAII-ESIP, EA 1219
University Institute of Technology, 4, av. Varsovie, 16021 Angouleme Cedex,
France
2
Technical University, 15, C. Daicoviciu street, 3400 Cluj-Napoca, Romania
3
Hamos GmbH, 17 im Thai, Penzberg, Germany
Abstract—This paper aims to analyze the structure of an expert system that has been developed
in order to capture expertise in the field of electrostatic separation processes and to make it
available to all those involved in their industrial application. A knowledge base was created
and is continuously updated. In addition to the basic rules that were formulated for guiding the
choice of the most appropriate electrostatic separation technique, an experimental design
assistant has been integrated in the expert system. This enables tuning of the operating
parameters in accordance with the characteristics of the granular mixture to be processed. The
second block of the expert system proved to be effective in transferring the expertise in trouble
shooting and diagnosis from research or supervisory staff to process operators.
1. Introduction
The application of expert systems for process diagnosis, operator support and training in
mineral processing, as well as in chemical and metallurgical industry, started in the early
1980’s [1]-[6]. By the end of the 1990’s, electrostatic separation techniques had become a
mature field of engineering [7–11], which justified the authors’ attempt to prototype a
dedicated expert system [12]. The objective was to capture expertise and to make it available
to all those involved in the industry application of these techniques. The development of
the expert system applications took into account the various methods of knowledge
acquisition described in the literature [13–15].
The aim of the present paper is to analyze the peculiarities of that expert system for
electrostatic separation processes. Three applications are discussed: (1) choice of equipment;
(2) optimization of the operating parameters; (3) trouble-shooting guide.
2. Representation of knowledge on electrostatic separation processes

The key issue of knowledge representation consists in the definition of the structure
employed to organize the acquired expertise in a form suitable for computer processing
[16]. The choice of MERISE formalism [17] for representing the knowledge on electrostatic
separation technology and equipment was related to the peculiarities of the two applications
that were addressed at this stage of development of the expert system.

356
Fig.1. MERISE conceptual model of the knowledge on the composition of the various granular
mixtures, on the types of electrostatic separators and the respective electrode configurations.
The first application that was analyzed focused on the choice of the most appropriate
electrostatic separation equipment for the processing of a given granular mixture. The
corresponding MERISE conceptual model reflects the relations that exist between six
types of entities: ‘MIXTURE’, ‘CONSTITUENT’, ‘MATERIAL’, ‘SIZE CLASS’,
‘SEPARATOR’ and ‘ELECTRODE’. Each entity has several properties, one of them serving
as identifier. The examples in Table 1 show that each occurrence of the entity
‘CONSTITUENT’, for instance, is characterized by a different value of the identifier
‘CodeConstituent’. Each occurrence of the entity ‘MATERIAL’ of which a ‘CONSTITUENT’
is «MADE OF » has a different identifier ‘CodeM’ and contains a list of physical properties
that are relevant for the electrostatic separation techniques. These values are either those
found in the literature or measured by the researcher (for new types of plastics or new
minerals).
One of the difficulties in the development of this part of the expert system is the multitude
of configurations that should be taken into account. The example given in Fig. 2 illustrate
the way in which the significance of the properties ‘Diml’, ‘Dim2’, … of the entity
‘ELECTRODE’ have been defined in connection with the ‘Type’ of the electrode that was
considered for that application: DUAL WIRE, … Similarly, the properties ‘Posl’, ‘Pos2’, …
of «CONFIG» define the position of the electrode by taking into account the ‘Type’ of the
‘SEPARATOR’ (which is CORONA with ROLL CARRIER ELECTRODE) and the ‘Type’ of
the ‘ELECTRODE’: ‘Pos1’=␣, ‘Pos2’=h, ‘Pos3’=␤.
Table 1
Examples of occurences of the entity ‘CONSTITUENT’

357
Fig. 2. Schematic representation of the electrode system of a roll-type electrostatic separator.
The relation PROCESSED associates a given ‘MIXTURE’ to an existing ‘SEPARATOR’,

while the relation «CONFIG.» defines the corresponding ‘ELECTRODE’ arrangement. For
instance, the ‘MIXTURE’ encoded RECYCL01 can be PROCESSED on a ROLL ‘Type’
‘SEPARATOR’, ‘CONFIG.’-urated as in Fig. 2. In practice, the user has to indicate to the
expert system the properties of the ‘CONSTITUENT’ of a ‘MIXTURE’ to be PROCESSED
and the associated ‘SizeClass’. If these information correspond to a ‘MIXTURE’ that is
already in the knowledge base, the type of recommended ‘SEPARATOR’ and of its optimal
‘ELECTRODE’ configuration are indicated.
In all the cases when the characteristics of the solids to be separated in a new application
do not correspond to any of the repertory mixture, the expert system is able to guide the user
through the selection process, following the rules synthesized in Table 2. Thus, for the
separation of a granular mixture containing large insulating particles, the expert system
recommends the use of two corona electrodes.
Table 2
Selection of the electrostatic separator for a given granular mixture

358
Fig. 3. Recommended operating parameters for the electrostatic separation

of a 50% Cu—50% PVC granular mixture (chopped electric cable and wire wastes).
Regarding the second application targeted by the present work, namely the prescription
of optimum operating parameters, the first prototype of the expert system [12] enabled only
the fitting between a new application and a similar electrostatic separation process already
existing in the knowledge base (Fig. 3). Taking into account the wide variety of different
materials that have to be treated by an electrostatic separator operating in an industrial
environment, this part of the system proved to be ineffective if not associated with a program
capable to guide the optimization of the process by using experimental design techniques.
In its present form, the expert system provides recommendations on the domain of
variation of the various factors and constraints to an independent program of experimental
design [18]. This program proposes the electrostatic separation tests that should be carried
out for identifying the model of the process and the optimum setting of the input variables
(high voltage level, position of the electrodes, feed rate, etc.). Hamos GmbH, Penzberg,
Germany, is currently using this procedure for establishing the optimum operating parameters
for the newly developed electrostatic separation technologies.
3. Trouble-shooting and diagnosis of electrostatic separation processes

For the expert system to be able to diagnose the operation of an electrostatic separation
process, it has been provided with a list of symptoms as the starting point for the knowledge
base query aimed at the identification of the problem that might have generated the
malfunction. The elaboration of rule sets on which the most probable diagnostic may be
established for each symptom has been an extremely complex activity. The data regarding
the failure histories of several types of industrial electrostatic separators have been collected
and analyzed. The conclusions have been corroborated with the knowledge on the physical
phenomena that govern the operation of the respective type of separator. The diagnostic
and trouble-shooting procedures have finally been validated by a 6-month “in-field” survey
of an industrial process.
The example given in Fig. 4 illustrates this approach in the case of industrial roll-type
corona-electrostatic separators employed for the processing of chopped electric wire and

359
cable wastes. More than 75% of the malfunctions registered for this type of application are
evidenced by the fact that the metal collected in the dedicated compartment has lower
purity than the one expected under normal operating conditions. This symptom indicates
a relative decrease of the “pinning effect”, due to the image force Fi, in comparison with the
“detaching effect”, due to the centrifugal force Fc. As Fi is proportional with the square of
the electric charge Q, the value of which depends on the strength of the corona field, the
first action to take is check if the supply voltage U of corona electrodes has the prescribed
value. The average value of the voltage Uav across the inter-electrode gap may decrease if
the frequency of spark discharges exceeds the upper recommended limit. These discharges
are normally caused by conducting particles passing through the electric field and which
reduce the dielectric strength of the air gap. This symptom can indicate either that the
concentration of metal particles in the granular mixture is higher than presumed, or that
their dimensions are larger than the prescribed values.
A detailed explanation of this and other symptoms is presented elsewhere [12]. The
survey of an industrial process at RIPS-ALCATEL, France, indicated that in 42% of the
cases the malfunction was due an accidental modification of the size characteristics of the
constituents (particle sizes are larger than the ones taken into account when the optimum
parameter values were established), in 36% of the cases the disturbing factor was a change
in the conductivity of insulating particles (material humidity is higher than the normal
value; conditioning temperature of the granular mixture is lower than the prescribed value),
in 15% of the cases an increase of the material feed flow was incriminated (the mono-layer
presentation of the material on the surface of the cylindrical electrode is no more assured),
all the other causes representing only 7% of the registered events.
In each of the above-mentioned cases, the operator followed the advice of the expert
system: correct the operating parameters of the grinders; increase the temperature at which
the drying or the thermal conditioning of the material is accomplished; diminish the material
feed flow. For the considered process, the expert system was able to indicate the values at
which the respective parameters should be set. In other situations, it is very likely that
statistical process control and experimental tools should be employed for effective diagnosis
and trouble-shooting. Efforts are presently made to integrate this kind of tools in this block
of the expert system.
Fig. 4. Expert system diagnosis of a metal purity problem.

360
4. Conclusions
The experience accumulated by the authors since 1999, when the present work was started,
suggests that the expert systems are likely to become the next wave of computer applications
to hit the mature electrostatic technologies, such as precipitation of dust, spraying of
powders, or sorting of granular mixtures. The industry application of the prototype of such
an expert system for electrostatic separation processes clearly has demonstrated how
efficiently the knowledge of many experts can be integrated and efficiently stored, in a
flexible structured data base. The modularized nature of the rule-based expert system enables
the continuous updating of the knowledge base.
The learning tools of the expert system contribute to a diminution of the training costs
of the new operators for industrial electrostatic separators. The efficient trouble shooting
assistance provided by the expert system offers faster solutions for most of the current
operation problems. Even if the advantages of such an expert system have been clearly
demonstrated, and most of the difficulties related to the development of a fully-operational
expert system for electrostatic separation techniques have been surpassed, further research
is needed to expand the field of expertise covered by this tool.
References
[1] Weiss S M and Kulikowski C A 1984 A Practical Guide to Designing Expert Systems (London:
Chapman and Hall).
[2] Hayes-Roth F (Ed) 1986 A Guide to Expert Systems (Ontario: Addison-Wesley).
[3] Harmon P, Maus R and Morrissey W 1988 Expert Systems. Tools and Applications (New York:
Wiley).
[4] Benchimol G, Levine P and Pomerol J C 1990 Systemes experts dans I’entreprise. (Paris: Edition
Hermes)
[5] Laguitton D and Leung J 1989 Advances in expert system applications in mineral processing in
Processing of Complex Ores (New York: Pergamon) 565–574
[6] Proceedings of AIPAC’89 (Advanced Information Processing in Automatic Control) 1989 (Oxford:
Pergamon)
[7] Inculet I 1986 Electrostatic Mineral Separation (New York: Wiley)
[8] Higashiyama Y and Asano K 1998 Paniculate Science and Technology 16 77–90
[9] Knoll F S and Taylor J B 1985 Mineral and Metallurgical Processing 2 106–114
[10] Dascalescu L, Morar R, Iuga A, Samuila A and Neamtu V 1998 Paniculate Science and Technology
16 25–42.
[11] Kwetkus B A 1998 Paniculate Science and Technology 16 55–68
[12] Mihailescu M, Samuila A, Iuga A and Dascalescu L 1999 Conf. Rec. IEEE/IAS Ann. Meet.
(Pheonix: IEEE Press)
[13] Frenzel L E 1987 How to develop an expert system in MacLaren D (Ed) Understanding Expert
Systems (Indianapolis: Howard W. Sams & Co) 149–182
[14] Hart A 1986 Knowledge Acquisition for Expert Systems (Toronto: McGraw-Hill).
[15] Harris C A, Woo E, Hall M and Miller R 1989 Knowledge acquisition techniques for expert
systems in Processing of Complex Ores (New York: Pergamon) 529–536
[16] Hodouin D and Flament F 1989 Frames for the representation of knowledge in Processing of
Complex Ores (New York: Pergamon) 553–564
[17] Tardieu H, Rochfeld A and Colletti 1998 La methode MERISE. Tome 1: Principes et outils (Paris:
Editions d’Organisation)
[18] Mihailescu M, Samuila A, Urs A, Morar R, Iuga A and Dascalescu L 2002 IEEE Trans, Ind.
AppLM 1174–1181.

361
Comparative analysis of computer-simulated and

experimental sparking voltage of the wire-plate system
Z Dudzicz
Department of Electrical Engineering and Automatic Control, Technical University
of Opole. 45–233 Opole, Sosnkowskiego 31. Poland
Abstract. The long term purpose of this paper is to develop new design for zones of charging
and depositing of dust particles in electrostatic precipitators and dust measuring instruments.
For this purpose, experimental voltage-current (U-I) characteristics of a basic (highvoltage)
wire-plate system are first constructed. Particular attention is paid to analysis of sparking
voltage in the wire-plate system. Results of measurements are computer processed and analyzed
by means of the Matlab package. On the other hand, computer simulations for selected UI
characteristics are developed and their results compared to the experimental ones.
1. Introduction
A wire-plate system is the fundamental construction element of almost all types of
electrostatic precipitators and other facilities operating under high-voltage corona discharge,
such as electrostatic separators, band-pass filters and feeders of charged dust particles.
Experimental analyses are very time-consuming (and expensive), hence application of
computer simulations to determine selected sparking voltage and U-I characteristics can be
of great importance when designing electrofiltration facilities. Effective combination of
experimental studies and simulation analysis of electroprecipitation phenomena enables
the development of new filter designs.
2. Experimental analysis of sparking voltage versus electrode distance for a wire-

plate system
Analysis of sparking voltage, initial voltage of corona discharge and U-I characteristics
aims at testing the electrical strength of interelectrode space as well as the dedusting
efficiency of electrostatic precipitators. A wire-plate system is used in small two-zone dust
collectors, where it operates as a charging zone in which the particle charging process
occurs. It thus affects dedusting efficiency. The experimental analysis is made for the
following wire diameters (in mm): 0.2, 0.55, 0.8, 1.2, 1.5, 2.0 and 3.0, with wire length of
0.58 m. An aluminium plate shaped as a 500 mm diameter disc is used as a plane electrode.
A measuring part is protected by a protection ring of 551–571 mm, in diameter which plays
the role of the Rogowski profile. A measuring disc along with the ring is fastened onto the

362
Figure. 1. Sparking voltage of arrangement wire-plate as a function a wire diameter.
insulator plate. Both the disc plate and the discharge electrodes are fixed onto insulation
supports in a way that ensures a continuous control of the inter-electrode distance. The
following wires are used as discharge electrodes.
1. Kanthal wire with diameter of 0.2 mm, used in photocopying printer KS-2.
2. Steel wire with of 0.55, 0.8, 1.2, 1.5 mm with chemical composition; 99 % Fe, 0.1 % Cu,
0.3 % Mn, 0.1 % Ni.
3. Wire made of brass as rod brass with diameter of 3.0, 4.0 mm.
Voltage was measured by electrostatic kilovoltmeter. Current values were measured by

magnetoelectric microampermeter LG-1. The measuring system was supplied with high DC
with negative polarity by means of ABK-70 cable power supply apparatus. All measurements
were recalculated to normal conditions 20° C and 760 mm Hg.
Fig 1 shows the results of empirical measurements of sparking voltage as a function of

electrode distance. As the distance increases, the value of sparking voltage, Up is also
growing. This curve may be approximated by straight lines. One curve contains up to 8
measuring points. Each point on the characteristics represents an average value from five
experimental measurements. The average value differs insignificantly from individual
measuring values. These differences reached a maximum of 3–4 % of the average value. As
an example two random cases are given below:

363
20.5 average, measured values 20.2, 19.8, 19.6, 21.2, 21.5

17.0 average, measured values 16.3, 16.5, 17.0, 17.5, 17.5
Figure 2. Sparking voltage of wire-plate set-up as a function of wire diameter.
A dashed line with angle inclination of 45 degree represents a constant value of electric
field intensity, equal to 10 kV/cm. It is an average value measured for a non-uniform
electric field. This value allows us to estimate the electric strength of this arrangement of
electrodes. The curves Up=f(h) can be approximated by means of a straight line. For distances
below h=25 mm the route curves are increased slightly. Next follows a flex point, which in
the next curves moved towards lower values. At h>25 mm, this route becomes linear. The
angle of inclination of the curves is smaller by about 10 degrees than the angle of the
broken line (dashed line), that corresponds to the average value of electric field strength.
The increase in h value from 5 to 35 mm causes an increase value of sparking voltage by
about 20.12 kV. All curves represent similar routes and change directions at the same
points. Distances between curves are to a large degree constant, only upper sections of
these characteristics are more diversified. Along with the increase of h for each curve, the
average value of Up also increases. Each Up=f(h) curve representing a wire of larger diameter
is placed higher than the previous one.
3. Sparking voltage experimental research of a wire-plate arrangement as function of

wire discharge electrode diameter
Analysis of the influence of the change in sparking voltage as a function of the wire
discharge electrode diameter appears very interesting. The intention has been to determine
the mode in which the electric strength of the interelectrode zone changes. The characteristics

364
performed on the basis of experimental measurements are shown in Fig. 2. All characteristics
have very similar routes. Along the first section, the curve represents increased Up values
similar to saturation curve. Within the diameter range of 1.5, 2.0 mm an inflexion point
occurs and from this point onwards the route is similar to a straight line. These curves
compared to those in Fig. 1 are evenly drawn on the picture. The curves are placed more
evenly with nearly equal distances between each other. When the distance is constant the
value of sparking voltage, Up increases proportionally with increasing wire electrode
diameter. This increase in Up which occurs with increase in wire diameters from 2.0 to 4.0
mm amounts to the following values of h; 5.5, 9.4. 10.5, 19.1, 10.9, 10.8 kV. The average
values for the electric field intensity for the same curves are; 19.0, 9.4, 7.0, 5.1, 4.4, 3.6
kV/cm.
A relationship between the sparking voltage and wire diameter has been observed,
however the increase is a minor one. Inclination of a single curve was restricted to the limits
of 15–20 degrees. A more significant increase has been observed for curves of bigger
distances h. The highest value of electric field intensity growth Ep, occurs for wire kanthal
of 0.2 mm. Next, the Ep value strongly decreased below the average value which amounted
to 7.35 kV/cm. Decrease in sparking voltage Ep is compared best to the decrease in E value.
4. Computer simulation
Computer simulation used for this paper concerned values which were calculated from
empirical measurements and theoretical computer simulation on the basis of partial
measurements only. In this case the computer modelling and simulation relate closely to
partial selected results of measurements. Pure theoretical simulation of phenomena of electric
partial discharges and correlation to measurement of sparking voltage or voltage-current
characteristics value is very risky. Such processes occur inside non-uniform electric field
and additionally run in the polyphase environment.
The computer simulation is a tool to process experimental measurement results. On the
basis of partial measurements only, we can simulate sparking voltage, voltage-current
characteristics saving time and labour otherwise required for full scale experimental
measurements. When partial measurements are completed, we know Up=f(h) at d=constant
and Up=f(d) at h=constant, and the angle of inclination, and are able to employ computer
simulation for further curves for other h and d values. Such accurate approximation of
curves is enough for technological purposes. Another solution is to apply partial
approximation, that is to split a curve into sections and to approximate subsequent sections
separately [3].
The area on Fig.2 can be divided into two different regions. In the first one the saturation
curve increases gradually and the second is similar to a straight line. For these regions
different methods of calculation of Up=f(d) were applied. It was assumed that the curves
were distributed uniformly in the region between the highest and the lowest curve.
Comparison of experimental results to theoretical calculated values proves consistency at
10% accuracy. Such accuracy can be accepted only for evaluation of probable sparking
voltage level in industrial applications.

365
Sparking voltage as a function of electrode distance can be approximated by straight lines

at 5–10% accuracy.
5. Comparison of the research results with Lewitow’s investigations.

As indicated by Lewitow’s research, with the same arrangement of wire-plate, the whole
range of wire electrode diameters can be divided into two separate areas. In the first area
the sparking voltage does not depend on the wire diameter of discharge electrode. The
average electric field intensity, at the moment a breakdown spark occurs, is always constant
and equals 20 kV/cm. Thus, sparking voltage Up changes proportionally to electrode
distance [1].
Up= Ep (h- r) (1)
This area is called the constant region gradient. The constant region gradient occurs only
for wire diameters d< 2–3 mm. This area is wider for greater distances of electrodes, h . The
research was carried out for h within the range of 20–100 mm, and wire diameter of 0.8 mm
to 10 mm.
In the second part of the wives for d> 3mm, the sparking value, Up decreased. On the
basis of the experimental measurements a generalised curve has been obtained.
Up/Uo—f(h/r) (2)
This area has been called the similarity region. For similarity region an empirical equation
has been found, to calculate the value of Up sparking voltage [1].
Up/Uo=(0.306 h/r—0.642) ln (h/r) (3)
When the diameter increases above 10 mm, another renewed increase in sparking voltage
may be expected as the distribution of the electric field became slightly non-uniform. The
critical values of diameter, where the diameter moves from the constant gradient region to
the similarity region, may be obtained by employing the aforementioned equations [1,2].
Comparison of the author’s and Lewitow’s measurements points to significant differences.
Constant gradient region for wire diameter below 3 mm has not been achieved. Our
investigations were carried out for smaller range of distances h=4–40 mm. While Lewitow’s
distances range was h=20–200 mm. Our investigations were carried more accurately. The
value of strength intensity is constant but by two levels lower. The Up relationship increased
accordingly. No sparking voltage decrease has been noted in any other similarity region.
6. Conclusions
As the initial value of corona discharge depends on wire diameter, therefore the initial
value of electric field intensity, Eo and sparking Ep have to depend on discharge electrode
wire diameter. As indicated by the author’s earlier research, values of field intensity, both
initial Eo and spark Up, reached very high values within the range of a few hundred kV/cm.

366
The most interesting processes of corona breakdown take place at the surface of discharge
electrode of small diameter.
The process of corona breakdown depends strongly on wire diameter and the voltage
value. For very small distances, corona discharge does not occur and the value of the
breakdown sparking voltage is equal to the initial voltage of this arrangement. The intensity
of corona discharge and breakdown increased together with the decrease in size o f diameter
of the discharge electrode. At very thin wires, the value of strength intensity is so high that
it disturbs strongly the average value of the electric field intensity [3, 4].
Research carried out and the discussion described above point to the following conclusion:
1. Characteristics of sparking voltage and voltage-current can be successfully investigated

employing computer-aided experimental research.
2. Within the determined regions 5 <h<40 mm and diameter 0.2<d<4.0 mm, simulation
method can be used to produce curves, using partial measurements.
3. The non-uniform field with corona discharge can be successfully approximated by
means of curves and straight lines.
The investigations were supported with funds of the Scientific Research Committee within
the research project No. 4 T10A 031 22.
References
[1] Osnowy elektrogazodinamiki dispiersnych sistem Energia Moskwa 1974
[2] Dymowyje elektrofiltry. Energia. Moskva 1980
[3] Dudzicz Z 2002 Elektrofiltry 2002. Kraków. Komputerowa symulacja rozkladu pola elektrycznego
przestrzeni mi´dzyelektrodowej elektrofiltru. p.78–84
[4] Bachtajew Sz A Koronnyj razrjad na mikroprowodach, Nauka Kazachskoj SSR, 1984

367
Decomposition of diesel particulate materials and nitric

oxides using a dielectric barrier discharge
Yukihiko Yamagata, Takafumi Matsui, Takashi Ebihara, Katsueori Muraoka
Interdisciplinary Graduate School of Engineering Sciences, Kyushu University,
Kasuga Fukuoka 816–8580, Japan
Abstract. We have demonstrated the usefulness of a trapping-filter trapping-low temperature

burning scheme for diesel participate materials (DPM) aftertreatment in an actual diesel exhaust
gas using a dielectric barrier discharge (DBD). The DPM trapped by a mesh electrode were
decomposed to CO and CO2 molecules using the DBD. Simultaneous decomposition of the
DPM and nitric oxides in diesel exhaust gas is also presented.
1. Introduction
Emissions of diesel participate materials (DPM), nitric oxides (NOx) and hydrocarbons
(HC) from diesel engines and other combustion processes are becoming more and more
strictly regulated from health concern. Therefore, there have been many attempts to reduce
these emissions using diesel participate filters, catalysts, absorption/thermal-desorption or
by combinations of these effects [1–5]. However, because of lack of sufficient action and/or
excessive cost, promising techniques are yet to be explored.
Recently, thermodynamic non-equilibrium gas discharges, such as packed-bed discharge
[6], corona discharge [7] and dielectric barrier discharge (DBD) [2–5,8], have attracted
much attention for these purposes. For example, packed-bed discharges have been recognized
to be able to decompose NOx and volatile organic compounds (VOC). However, it is not
suitable for use in high gas flow rate due to large pressure loss caused by the structure.
Moreover, most of the electrical energy is consumed for excitation and ionization of
predominantly existing nitrogen and oxygen molecules instead of the decomposition of
these materials, because these hazardous materials are exhausted in very small concentrations
of a few tens to hundreds ppm. Therefore, the decomposition efficiency using thermodynamic
non-equilibrium gas discharges for these hazardous materials is very low. In order to extract
the potential of these discharges for decomposition, densification or localization of these
hazardous materials is prerequisite.
We have proposed a new decomposition technique for environmentally hazardous
materials with low concentration [3,4,9], and achieved successful decomposition of VOC
[9]. This technique is based on the combination of a DBD with densification/localization
and a trapping-filter or honeycomb-structured adsorbents. This combination does not spoil
the high gas flow rate, and the DBD can easily generate non-equilibrium plasmas on a large
discharge space. We consider that this concept is the most suitable for the decomposition of
hazardous materials in very small concentrations.
In this paper, we demonstrate this technique for the decomposition of DPM and NOx
exhausted from an actual diesel engine.

368
Figures 1 and 2 show the illustration of the cross section of a discharge reactor and the
experimental set up for the decomposition of DPM and NOx in an actual diesel exhaust gas,
respectively. The box shape electrode, which one side opened, was made from a stainless
steel mesh (mesh size of 2 µm) and placed in a discharge reactor. This mesh electrode acts as
a trapping filter of DPM and a grounded electrode. A powered-electrode covered on both
sides by a set of mica sheets (thickness of 1 mm) was inserted in the mesh electrode with a
separation of 1 mm from the each mesh surface. The DBD was generated between the
powered-electrode and the mesh electrode by applying a high voltage (up to 10 kV) with a
frequency of 1 to 10 kHz. Figure 3 shows the photograph of the DBD generated at 7 kV (3
kHz). As can be seen from this photograph, the DBD was uniformly generated in the reactor.
Part of the exhaust gas from a diesel engine (0.273 l, 3600 rpm, maximum power 6.0 PS)
was fed into the discharge reactor. The gas flow rate was changed from 1 to 5 l/min. As the
Fig. 1 Illustration of the cross section of a discharge reactor.
Fig. 2 Experimental set up for the decomposition of DPM and NOx in an actual diesel exhaust gas.

369
Fig. 3 Photograph of the DBD generated at 7 kV (3 kHz).
exhaust gas passed through the mesh electrode as a trapping filter, DPM were trapped near
the surface of the mesh electrode. After sufficient accumulation of DPM over the filter, the
DBD was generated between the powered-electrode and mesh electrodes to decompose the
trapped DPM. Using this configuration, the DPM are localized (condensed) in a limited
space, which ensures an effective decomposition of the DPM.
The differential pressure across the mesh electrode increases when a large amount of
DPM is trapped, while the pressure decreases when the DPM are decomposed by DBD. So
the differential pressure is an important parameter to indicate the degree of DPM trapping.
Thus, we measured the differential pressure, the gas concentration, gas temperature and the
smoke density in order to understand the performance of this reactor.

In order to check the reliability of the mesh electrode as a trapping filter for the DPM, we
measured the smoke density before and after the mesh electrode. Figure 4 shows that the
smoke density changes as a function of load power for the diesel engine. The smoke density
before the mesh electrode increases with increase load. As for the 100 % load, the DPM
density was estimated to be 98 µg/l. On the other hand, the smoke density after the mesh
electrode is nearly zero at any load of the power. This result suggests that the mesh electrode
traps almost all of the DPM. As can be seen in Fig. 3, the DBD is uniformly generated
between the powered-electrode and the mesh electrode. From these results, this electrode
configuration is suitable for the decomposition of the DPM trapped by the mesh electrode.
The DBD was periodically generated in order to demonstrate the trapping and the
decomposition of the DPM using the discharge reactor. Figure 5 shows the temporal changes
Fig. 4 Smoke density changes as a function of load power for the diesel engine.

370
Fig. 5 Temporal changes of a differential pressure across the mesh electrode, the smoke density
before the mesh electrode, and the gas temperature in the discharge reactor.
of a differential pressure across the mesh electrode, the smoke density before the mesh
electrode, and the gas temperature in the discharge reactor. Each DBD was sustained at 7 kV
with 3 kHz for 10 min, and the load of the diesel engine was 100 %. The differential pressure
increased initially with the time passage due to the trapping of the DPM by the mesh
electrode. On the other hand, the differential pressure rapidly decreased as soon as the DBD
was generated, and was kept at a low value while the DBD was sustained. This feature
occurred periodically, as seen with the DBD treatment in Fig. 5.
The DPM mainly consists of carbon and HC. As the DPM are exposed to the DBD, the
DPM should be decomposed into CO2 and H2O through reactions with surrounding oxidizing
atoms and molecules, such as oxygen and NO2. Figures 6 (a) and (b) show the concentrations
of CO2 and CO molecules after the discharge reactor, respectively. These concentrations
were measured at several times for the non-DBD and subsequently-occurred DBD periods.
The CO2 and CO concentrations for the DBD periods are larger than that for non-DBD
periods although there are some fluctuations in the concentrations. Furthermore, water
drops were observed on the discharge reactor walls after several discharge treatments. This
Fig. 6 Concentrations of CO2 and CO molecules after the discharge reactor.

371
is caused by the DPM decomposition into H2O. These experimental results suggest the
DBD can effectively decompose the DPM into CO, CO2 and H2O and regenerate the mesh
electrode for the next DPM trapping.
Figure 7 shows the temporal change of NQx concentration after the mesh electrode. The
DBD was periodically generated in the same experimental conditions as for Fig. 5. As soon
as the DBD was generated, the NQx were also decomposed simultaneously by the DBD.
After the rapid decrease, the NCbc concentration slightly increased with time. At the same
time, the DPM might be already decomposed by the DBD because the differential pressure
was low, as seen in Fig. 5. Thus, the NOx concentration change seemed to be interlocked
with the DPM decomposition. A nitric monoxide (NO), which is a main content in diesel
exhaust gas, is easily oxidized to NO2 in the discharge following reaction (1). Then the
DPM is directly oxidized by NO2 following the reaction of (2).
NO+O→NO2 (1)
NO2+DPM→CO2+N2 (2)
These results suggest that the combination of the DBD with a densification/localization
technique may be able to achieve simultaneous decomposition of DPM and NOx from
diesel exhaust gas.
In order to exploit a more advanced idea that may revolutionarily change the aftertreatment
of diesel exhaust gases, we have now been trying to consider new scheme. Here, NOx and HC
are to be adsorbed in adsorbents, similarly as VOC described above, and DPM are collected to
the adsorbent surface by the principle of the electrostatic precipitator. Then, a barrier discharge
is formed between the parallel electrodes, inducing desorption and destructions of the NOx,
HC and DPM simultaneously. Further results will be published soon.
Fig. 7 Temporal change of the NOx concentration behind the mesh electrode.

372
4. Conclusion
We have demonstrated a successful decomposition of the DPM and NOx in an actual diesel
exhaust gas using the combination of a DBD with densification/localization using a
trapping—filter. The DBD can effectively decompose the trapped DPM into CO, CO2 and
H2O, and regenerate the mesh electrode for the next DPM trapping. The NOx were
simultaneously decomposed by the DBD. These results suggest the usefulness of this system
for the diesel exhaust gas aftertreatment.
Acknowledgements
This work has been partially supported by Grant-in-Aid for Scientific Research from the
Ministry of Education, Science, Sports, Culture and Technology of Japan.
References
[1] Hawker P, Hüthwohl G, Henn J, Koch W, Lüdders H, Lüers B and Stommel P 1998, Society of
Automotive Engineering 980189 374–380.
[2] Chae JO, Hwang JW, Jung JY, Han JH, Hwang HJ, Kim S and Demidiouk VI 2001, Phys.
Plasmas 8 1403–1410.
[3] Muraoka K, Yamagata Y and Ebihara K 2001, Proc. 1st Korea-Japan Joint Symposium on Energy
and Environment (KIER, Korea, Dec. 13–14, 2001) 3–4.
[4] Yamagata Y, Matsui T, Ebihara T and Muraoka K 2002, Proc. 2nd Japan- Korea Joint Symposium
on Energy and Environment (Kyushu Univ., Fukuoka, Japan, Oct. 15, 2002) 29–30.
[5] Eliasson B and Kogelschatz V 1991, IEEE Trans. Plasma Sci. 19 309.
[6] Yamamoto T, Ramanathan K, Lawless PA, Ensor DS, Newsome JR, Ramsey GH and Plaks N
1992, IEEE Trans. Indust. Applicat. 28 528–534.
[7] Kim HH, Takashima K, Katsura S and Mizuno A 2001, J. Phys. D-Appl Phys. 34 604–613.
[8] Goto N, Kudo S, Motoyama H and Ohyama S 2002, Jpn. J. Appl. Phys. 41 L64–L66.
[9] Urae H, Yamagata Y, Muraoka K, Yamada K, Yamauchi H and H. Okano H 2002, Trans. IEE
Japan, 122-A 965–976 (Japanese).

373
Question and Answers
Section 1: Hazards
Ignition hazards associated with earthing and bonding of charged conductive objects
K Schwenzfeuer and M Glor 19
Question (John Chubb)
Have you considered the use of a wood earthing stick as a simple way of removing charge
from a conducting body without risk of a spark discharge? Obviously the wood should not
be too wet or too dry—although a wet surface will go into a spray discharge before contact
and so act like a corona discharger.
Butterworth et al reported some studies at an earlier Electronics Conference on the surface
resistivity range needed to avoid occurrence of spark type discharges.
Author’s Reply (Schwenzfeuer)
A piece of wood was not tested, but we did discharge tested with a glove. The surface of this
glove was so called “charge dissipative”. What happened was, that a brush discharge
occurs during the approach and this discharge was still incendive for ethylene. It must be
assumed, that a piece of wood would behave similar and an incendive brush discharge
could not be excluded totally.
Section 2: MEMS and Applications
Microscale electric induction machines for power applications

C Liver more, A Forte, T Lyszczarz, S D Umans and J H Lang 45
Question (M Horenstein)
How do you deal with the issue of bearing friction?
Author’s Reply (Carol Livermore)
The rotor is supported both axially and in-plane on gas dynamic bearings. The rotor does
not contact its housing; rather, they are always separated by a thin film of pressurized air.
There are substantial viscous losses associated with the relative motion of the components
when they are separated by micron-scale gaps. Experiments indicate that our Couette flow
model is an accurate description of these losses, and we simply fold the losses into our
system level models. For relatively larger microgenerators, viscous losses can equal 50% of
the input mechanical power, which is undesirable.

374
Question (Keith Davies)

The small gaps preclude gas electrical breakdown but do you feel that field emission might
eventually limit the possible electric stress?
Author’s Reply (C Livermore et al)
I am not certain what the ultimate limit will be. I have tested stators for damage under high
voltage conditions by applying a phase-to-phase ac voltage while inspecting the electrode
structure under a microscope. (The gaps between neighbouring electrodes are 3 ˜m—µ4 m).
If there is a defect in the lithography, such as an asperity, breakdown occurs and electrode
material is visibly removed. For a small number of defects, the breakdown is a local event
that does not significantly impair stator functioning. More widespread breakdown, such as
would occur with rough polysilicon electrodes, does impair functioning and leads to a
detectable increase in current.
In the absence of widespread initial flaws, we have maintained 300V phase-to-phase at
MHz frequencies without observable damage or current rise.
We have also applied up to 150V peak-to-peak voltage to the stator electrodes with a 2 mm
gap between stator and rotor. Upon disassembly no stator damage was observed.
Question (Edward Law)
As you move from mm-size gas to µm-size gaps, does the classical Newtonian concept of
gas viscosity hold strictly true? Are further refinements beyond the usual Cunningham
correction factor required?
Author’s Reply
I have not considered further refinements; you raise a point that is worth looking into.
I have not to date noticed a discrepancy between observed viscous torque constant and that
which is expected from a Couette flow with the classical gas viscosity.
Micro-machined variable capacitors for power generation

P Miao, A S Holmes, E M Yeatman, T C Green and P D Mitcheson 53
Question (Lucian Dascalescu)
Have you been able to evaluate to what extent gravity affects the operation of your device?
In other words is the device able to operate equally well in horizontal and vertical position?
Author’s Reply (P Miao)
We have tested the device at both positions. We did not see any significant difference. This
is because the gravitational force is relatively small compared to the acceleration force of
the vibration table. The mass is very light ~ 4 g).

375

L F Gaunt and J F Hughes 59
Question (Muntomaa)
How did you measure the change of the spray, and were all the droplets collected.
Author’s Reply (L F Gaunt)
We collected the spray droplets in a Faraday Pail to measure their charge. As the spray
plume is not very divergent, all droplets were collected.
Ejection of a charged spray leaves a counter charged in the can. What happens to it?
The counter charge accumulating in the liquid reservoir is grounded through the user.
This can be achieved by co-extension of a conducting strip in the plastic bottle.
Question (J M Chubb)
What advantages does your tribo charging offer compared to piezo electric units that are
widely used for gas igniters?
Our technique offers just one solution and others may also be conceived.
However, cost is an important consideration, and the cheapest technique must be pursued.
Question (Prof Bacachandran)
How did you determine the threshold Qfm value?
Did you calculate the electric field strength at the atomization region?
What is the typical value of this field?
Author’s Reply (L F Gaunt, J F Hughes)
The threshold 2/m value was found experimentally, using the wrap-around phenomenon as
an indicator of electrostatic forces coming into play. As the droplet size distribution is wide,
this represented a threshold value relating to the specific spray.
During the design stage for the novel bigger spray, the electric field strength at he atomization
region will be modelled in order to maximise it.

376

D A Hays and J Q Feng 77
Might toner rounding due to wear in a fluidized bed cause the adhesion of the ion charged
toner to be less than the adhesion of toner ion charged in the alternating electric field?
Author’s reply (D A Hays)
One can only speculate why the toner adhesion measurements reported by Christy from the
Moore Research Center at Grand Island, NY were lower than expectations. It doesn’t seem
possible that toner would be rounded since the interparticle forces in fluidized beds are
actually quite gentle. It seems more likely that since the toner was simultaneously charged
and deposited onto a roll, most of the ion charging occurred after the toner deposition. If the
ion charging was predominantly on the caps (top) of the toner, the electrostatic adhesion
will be much less than expectations for uniformly charged toner.
Section 3: Bioelectrostatics
Bio-nanotechnology of DNA based on electrostatic manipulation

M Washizu 89
Question (David Bakewell)
What was the conductivity of the solution when restriction enzymes and RNA polymerase
were used in your experiments?
Author’s Reply (M Washizu)
We cannot stretch DNA in buffer solutions with high conductivity. We use de-ionized water
for stretching, and after immobilizing DNA, we replace the solution to the buffer that is
optimum for RNA synthesis.
Question (Gary Stevens)
Do you use particles smaller than 1mm, or AFM stylus, or even carbon nano-tube to
immobilize enzymes and do DNA surgery?
Author’s Reply (M Washizu)
1) If you use small particles in laser manipulation, many particles will be trapped and
coagulated at the focal spot.
We use particles whose diameter is about equal to the beam spot size.
2) Electrostatic DNA stretching is done in a thin layer of water sandwiched by a microscope
slide and a cover slip. It is inaccessible by mechanical stylus. Of course you can
remove the cover slip, but DNA may be broken by hydrodynamic force, and drying
takes place very quickly.

377
Cell sorting and separation using dielectrophoresis

D Holmes and H Morgan 107
Question (M Horenstein)
How do you know the polarization versus frequency for the individual cells so that you can
choose the proper operating frequencies?
Author’s Reply (D Holmes)
Generally, for the experiments shown here we have used all types of known dielectric
properties. Either measured by electro rotation in our lab or values taken from the literature.
However, one could imagine using a mixture of particles, some known (control particles)
and other particles of unknown properties. The relative position of the particles on the
separation could then be used to predict the properties of the unknown cells.
Question
You showed separation of two particle sites (3 µm &. 0.914 µm). What is the expected
resolution for particle separation based on size difference) using your system, and also what
is the minimum particle site you can manipulate using this technique?
Author’s Reply (D Holmes)
Particles as small as DNA macromolecules have been manipulated using this system. Work
is being undertaken on particle size separation by altering switching times.
Section 4: Measurements
A wide bandwidth probe for electrostatic discharge measurements

J M Smallwood and G L Hearn 125
Question (Mark Zaretsky)
Have you optimized the shield and tip geometry for creating and monitoring a discharge?
Author’s Reply (J M Smallwood)
The probe has a 20 µm hemispherical tip to the shield, because this is a commonly used
probe size in conventional spherical ESD probes.
The ESD probe central conductor protrudes slightly from the shield hemisphere profile.
This is to increase the probability that the ESF current occurs to the probe conductor rather
than the shield. Of course there is a trade-off here. If the tip protrudes too much then at high
fields corona discharge may occur or several ESD smaller events may occur instead of one
larger event. The protrusion must be kept small to reduce this effect.

378
Contact charging method for the measurement of charge decay in electrostatic

dissipative materials
J Paasi, T Kalliohaka, T Luoma, R Ilmén and S Nurmi 137
1) Have you made, do you plan to make, any comparison of your contact charging method
to tribo charging?
2) I would also express my concern about the validity of contact charging as a generally
applicable method. Comments are included on our Website (www.sci.co.uk) comparing the
merits and limitations of various methods of charge decay measurement. A major problem
with contact charging is that relatively insulating components on the surface may not be
effectively charged. The response to earthing the sample shows mainly the fastest decay
route within the material, not at the surface.
Author’s Reply (Jaakko Paasi)
1) We have done some comparative tests but good comparison is difficult because of low
reproducibility of tribocharging.
2) I fully agree with Dr. Chubb that the contact charging method fails in fully characterising
the charge decay behaviour in some materials but the situation is similar with all existing
methods. We can always find such materials which will not be fully characterised by using
a single method. Therefore, complementary measurements using two or three different
methods should be recommended for complex materials.
A particle charge spectrometer for determining the charge and of individual dust grains
on Mars
S Fuerstenau and G Wilson 143
To concentrate your particle flow along the axis of your sampling tube, I would commend
use of a clean air sheath with a streamline flow. We did this in the monitor for airborne fibres
we developed, and reported at an earlier Electrostatics Conference.
I wondered about using time varying electric fields to get aerodynamic/inertia size
information for particle sizing. This might be a longitudinal field or better a lateral field
with examination of the relative signals seen on side vs other side sensing. This would be
more compatible with the time response needed.
Author’s Reply (Fuerstenau)
We are planning to use a clean air concentric flow in order to confine particles to the center
streamlines of the flow. Your sheath flow design may be very appropriate. The inertia and
drag forces on the relatively large particles are large compared to the electric forces.
Measuring variations in the particle velocity dye to the imposition of an electric field will
probably be problematic.

379
Question (R G Harrison)
The triboelectric dust charging leads to a negatively skewed particle charge distribution.
Will additional sources of ionisation, e.g. cosmic rays, lead to a more symmetrical charge
distribution on Mars?
Author’s Reply (S Fuerstenau et al)
Cosmic rays in Mars’ atmosphere will generate small ions, of both polarities. These will
discharge particles tending to push the particle charge distribution towards zero and therefore
towards the normal symmetrical charge distribution for the particles. However, to the extent
that dust is a source of net charge in the atmosphere, the introduction of charge pairs from
cosmic ray ionization is not expected to change the balance of charge in atmosphere on the
whole. Cosmic ray ionization will shift the charge distribution on the particles but will not
remove any net charge asymmetry due to the introduction of triboelectrically charged
particles.
Measurement of optical intensity and fluence generated by spark discharges

J C Crager and M N Horenstein 149
Question (W D Greason)
The current probe measures all of the discharge current but it is not known what
percentage of the light output is captured by the photodetector.
Your comments would be appreciated.
Author’s Reply (Crager & Horenstein)
It is true that the current probe measures the total current and the photodetector measures
only a portion of the light emitted. Since it is difficult to capture all of the light, we are
assuming here that the light has a constant radiation pattern. Therefore, we have been
measuring the relative light emission from the spark gap under the assumption that a
change in the current would be accompanied by a relative charge in the emitted light.
Question (Ulrich von Pidoll)
I think the quality of the capacitor used in the discharge circuit is very important concerning
the energy actually dissipated in the discharge. This effect is the greater, the greater the
capacitance and the greater its inductance. Have you made a test series with a nearly ideal
capacitor (air capacitor, pulse capacitor for MHz)?
Author’s Reply (Joseph C Crager and Mark Horenstein)
No we have not performed tests with nearly ideal capacitors. We are, however, currently
running experiments with smaller capacitors (in the 100pF range). This is a good suggestion
and we will take it into consideration.

380
Section 6: ESD
Electrostatic testing of ESD-protective clothing for electronics industry

J Paasi, S Nurmi, T KaUiohaka, G Coletti, F Guastavino, L Fast, A Nilsson, P Lemaire, J
Laperre, C Vogel, J Haase, T Peltoniemi, G Reina, A Borjesson and J Smallwood 239
Would you not expect that when a garment surface is rubbed the charge will spread quickly
over a large area of the garment and will then have a large area for the charge to dissipate
through the undergarments to the earthed body underneath? In this case surely seam-seam
conductivity bonding is not too critical?
Author’s Reply (Paasi et al)
Charge can dissipate from the garment through normal clothing to the earthed body of the
operator under proper conditions (good contacts between the ESD garment normal clothing
and the operator’s body; sufficient conductivity of the normal clothing materials is also
required. In a general situation, such a dissipation path is not guaranteed. Therefore, good
conductivity over the garment seams is a necessity.
Why is the MBM damage threshold set at 2KV when real body voltages can be greater than
10kV?
Author’s Reply (J Smallwood)
I’m not aware of how this target level was determined. I understand that this is a take-off
between the economic benefit of providing more rugged devices and thus preventing
excessive ESD losses, and the additional cost and possible performance losses associated
with providing a higher level of ESD protection.
Question (N L Allen)
You referred to hazards in terms of voltage, but surely the important criteria are charge and
stored energy?
Author’s Reply (J Paasi)
I used voltage just because it is common practice of the field (ESO risks in electronics
industry). Clearly, voltage is a poor guide in defining risks, because for a given charge,
voltage depends on the capacitance of the object (which can vary more than an order of
magnitude). Charge and energy would be much better guides for risk assessment. Even
more better would be charge and peak ESO current, as will be shown in the following talk/
paper by Smallwood and Paasi.

381
Section 7: Environment
Electrostatics and the environment

G S P Castle 309
Question (Mark Horenstein)
With regard to lightning you pointed out that the earth’s “leaky capacitor” would quickly
discharge were it not for the charge regeneration mechanism. If that regeneration mechanism
begins with the negative side of the polarized cloud discharging to ground, is there any
speculation as to how the remaining positive charge migrates to the ionosphere? One
would think that the positive charge would migrate in the fairweather field down to earth.
Author’s Reply (G S P Castle)
The positive charge in the upper part of the cloud does not migrate as High as the ionosphere.
It is driven upwards in the cloud by thermodynamically induced updrafts. At a height of
about 50km, the air is conductive enough that it essentially an equipotential. This positive
charge produces the “fairweather field” of about 100V/m directed downwards and yes, left
alone it would cause positive charge to migrate to earth and discharge the earth in a matter
of minutes. It is the localized regeneration due to thousands of nearly simultaneous lightning
strikes across the globe that replenishes the negative charge on the earth and the positive
charge in this layer.
What is the mechanism by which lightning discharges get started at the top end to get the
leader started? I understand that discharges start with emission of VHP that suggest discharges
between individual charged droplets may be the initiating mechanism.
I have no special insights into the microscopic details of this initiation. However it seems
reasonable that localized field perturbations leading to microdischarges between charged
droplets could initiate the stepped leader.
Comment (J Smallwood)
A few years ago I was involved in an EC funded project that developed belt-based
electrostatic separators and produced similar results to yours, from post consumer plastic
waste streams. My view is that the main problems for viable recycling are that; 1) virgin
material is so cheap that it is difficult to have a viable recovery process without subsidy, 2)
the market distrusts recycled material and prefers to buy virgin material, so we have a
problem of market education.
I agree with both your observations, we have experienced similar reluctance on the part of
plastics users for the reasons you cite. However I believe that both concerns are becoming

382
less of a constraint. Certainly in the long term virgin plastic prices can only increase and the
viability becomes more obvious. However, even at current prices profitability has been
demonstrated because of the low energy consumption of the process and the high purities
and recoveries that are now possible. In addition, legislative encouragement to recycling is
becoming more common in jurisdictions throughout the world.

Morgan - Electrostatics 2003 (Institute of Physics Conference Series) - Institute of Physics Publishing (2004)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Morgan - Electrostatics 2003 (Institute of Physics Conference Series) - Institute of Physics Publishing (2004)

Uploaded by

Copyright:

Available Formats

Electrostatics 2003

© 2004 by Taylor & Francis Group, LLC

180 Microscopy of Semiconducting Materials 2003

179 Electron Microscopy and Analysis 2003

177 Optical and Laser Diagnostics 2002

174 Compound Semiconductors 2002

173 GROUP 24: Physical and Mathematical Aspects of Symmetries

171 Physics of Semiconductors 2002

170 Compound Semiconductors 2001

169 Microscopy of Semiconducting Materials 2001

© 2004 by Taylor & Francis Group, LLC

Proceedings of the Electrostatics Conference of the Institute of Physics

New York London

© 2004 by Taylor & Francis Group, LLC

© 2004 by Taylor & Francis Group, LLC

Library of Congress Cataloging-in-Publication Data

Visit the Taylor & Francis Web site at

© 2004 by Taylor & Francis Group, LLC

© 2004 by Taylor & Francis Group, LLC

Prof. Hywel Morgan

© 2004 by Taylor & Francis Group, LLC

Bill Bright Memorial Lecture

Section 2: MEMS and Applications

© 2004 by Taylor & Francis Group, LLC

© 2004 by Taylor & Francis Group, LLC

Spray current dependence on flow rate and conductivity in cone-jet mode

Section 5: EHD & Numerical Modelling

© 2004 by Taylor & Francis Group, LLC

Calorimetry in the detection of discharge events

© 2004 by Taylor & Francis Group, LLC

Dust particle removal by wet-type electrostatic scrubber

© 2004 by Taylor & Francis Group, LLC

Electrostatics and the Lab on a Chip

© 2004 by Taylor & Francis Group, LLC

2. Electrostatic force scaling for microsystems

3. Scaling laws for particle DEP

© 2004 by Taylor & Francis Group, LLC

contributions, including dipolar, quadrupolar, and higher-order terms [12].

where Kn=n(2n+1)(κp-κm)/[(n+1)κm+n κp] is a polarization coefficient and ε o is the

3.1 Particle levitation and trapping

where ρp and ρ m are

© 2004 by Taylor & Francis Group, LLC

3.2 Particle dynamics

© 2004 by Taylor & Francis Group, LLC

4. Scaling laws for E-field-mediated microfluidics

4.1 Electro-osmotic flow

© 2004 by Taylor & Francis Group, LLC

4.2 Static electrowetting and DEP actuation

© 2004 by Taylor & Francis Group, LLC

Note that for EWOD, the electric field E= V/

© 2004 by Taylor & Francis Group, LLC

Fig. 7. The electrostatic droplet transport system shown at

© 2004 by Taylor & Francis Group, LLC

electrostatic forces in microfluidic systems. It is interesting to note that scaling arguments

© 2004 by Taylor & Francis Group, LLC

© 2004 by Taylor & Francis Group, LLC

Electrostatic ignition hazards—occurrence, detection

2. Some fires caused by static electricity

© 2004 by Taylor & Francis Group, LLC

Figure 1. Pictures of a recorded refuelling fire in Japan [4]

© 2004 by Taylor & Francis Group, LLC

2.2. Fires caused by metal parts

2.3. Fires occurring at flexo printing machines