ERRATUM
The June
1994
issue of
Pure
and
Applied
Chemistry
(Volume
66,
No.
6)
included the articles ‘Theo-
retical basis of non-equilibrium near atmospheric pressure plasma chemistry’ by
A.
A.
Fridman
and
V.
D.
Rosanov, and ‘Modelling of dielectric barrier discharge chemistry’ by
B.
Eliasson,
W.
Egli and
U. Kogelschatz. Unfortunately, the front pages
of
these two papers were transposed during produc-
tion
so
that the body of both papers
was
printed out
of
position within the issue, and given the wrong
page numbers and running headlines.
Both articles are reproduced in their entirety on the following pages with the correct page numbers
and running headlines.
A
correct version
of
the contents page is
also
reproduced. These pages should
be
used to replace those in the faulty printed copies.
Blackwell Scientific Publications apologizes to the authors for this error, and to readers who will
have experienced some confusion.
Pure
&
Appl.
Chsm.,
Vol.
66, No.6,
pp.
1275-1286, 1994.
Printed
in
Great
Britain.
0
1994
IUPAC
Modelling
of
dielectric barrier discharge chemistry
B. Eliasson, W. Egli and
U.
Kogelschatz
ABB Corporate Research, CH-5405 Baden, Switzerland
Abstract
The discharge physics and plasma chemistry
of
dielectric barrier discharges (silent
discharges) is discussed. Numerical models describing electrical breakdown,
microdischarge formation and the ensuing free radical chemistry are presented.
Applications ranging from ozone generation, excimer
UV
lamps and C02 lasers
to
surface treatment and pollution control are described. Experimental and modelling
results on C@ hydrogenation in dielectric barrier discharges are also presented.
INTRODUCTION
Dielectric barrier discharges or barrier discharges, frequently also referred to as silent discharges, have been
utilized for more than a century. Early investigations originally started by Siemens (1) in 1857 concentrated
on the generation of ozone by subjecting a stream of oxygen or
air
to the influence of a dielectric barrier
discharge (DBD). Ozone and nitrogen oxide formation in DBDs became an important research issue for
many decades (2-4). The names of Warburg
(5,6)
and Becker
(7,s)
in Germany, Otto
(9)
in France, of
Briner et
al.
(10) in Switzerland, of Philippov and
his
group
in
Russia (1
I),
of
Devins (12) in the United
States, Lunt (13) in England and Fuji et al. (14) in Japan, should be mentioned in this connection. More
recent references
can
be
found in a book by Samoilovich et a1.(15) and in some review papers by Eliasson
and Kogelschatz (16-19). As of today, ozone generation is still the major industrial application of DBDs
with thousands of installed ozone generating facilities based on this principle. For this reason the dielectric
barrier discharge is sometimes also referred to as the ozonizer discharge.
In recent years novel applications of DBDs with respect to pumping of C02 lasers (20-22) and excimer
lamps (23-28), flue gas treatment (29-31), surface modification (32-34), pollution control (17,
35-38),
H2S decomposition (39) and C02 hydrogenation (40,41) have been proposed. All these applications take
advantage of a mature technology originally developed for ozone generators. Its main advantage is that
large gas flows at about atmospheric pressure can be subjected to nonequilibrium plasma conditions
with
only negligible increase of the enthalpy of the feed gas. With all these diverse applications in mind the
dielectric barrier discharge configuration will be discussed in a very general sense as a plasmachemical
reactor. Special discharge properties will
be
described as well as charged particle reactions and the free
radical chemistry initiated by the gas discharge. Several results are presented concerning the modelling of
dielecmc barrier discharge chemistry in different gases.
THE
DIELECTRIC BARRIER DISCHARGE
In this section the electrode configurations and typical operating conditions of DBDs are described. When
the electrical field in the discharge gap is high enough
to
cause electrical breakdown a large number of
microdischarges are observed if the pressure is of the order of 1 bar. This is a typical pressure range for
ozone generation, excimer formation as well as for flue gas treatment. Plasma formation and electrical
conductivity is restricted to these microdischarges while the space in between is not ionized and serves
only
as a background reservoir for the energy dissipated in the microdischarges and for the lqng-lived species
created. Fig. 1 shows a frame
from
a video taken at very low power density. It shows microdischarges
in
an annular discharge gap between
two
quartz cylinders. Visual access was provided by a wire mesh used
as
1275
1276
B.
ELIASSON,
W.
EGLl
AND
U.
KOGELSCHATZ
the outer electrode. The gap spacing was
5.5
mm, the operating frequency
200
kHz and the exposure time
of the video frame 16 ms.
SILENT
DISCHARGE
I
GEOMETRY
Fig.
2
shows schematically a microdischarge in a discharge gap bordered by one dielectric and
one
metal
electrode, a configuration normally encountered in ozone generators. In other applications like
C02
lasers
and excimer lamps both electrodes may
be
covered with dielectrics. The preferred materials for the dielectric
barrier are glass or quartz, in special cases also ceramics, enamel or even polymer coatings. Besides the
planar configuration sketched in Fig.
2
also annular discharge gaps between cylindrical electrodes and
dielectrics are used in many technical applications.
The discharge gap itself has a typical width of a few
mm:
about
1
mm
for ozone generators, a few mm for
excimer lamps, and up to
50
mm
for CO, lasers.
To
initiate a discharge in such a discharge gap filled with a
gas at atmospheric pressure, voltages in the range of a few kV
are
required. Since the current has to pass
the dielectric in the form of a displacement current (capacitive coupling), alternating voltages are required to
drive a DBD. Although traditionally the line frequency was the obvious choice, modem DBD equipment
often runs at higher frequencies. Ozone generators use thyristor-controlled inverters generating square-
wave currents in the
kHz
range, while excimer lamps and
C02
lasers use transistorized switch-mode power
supplies with essentially sinusoidal output voltages. Their frequencies can
go
up to about
1
MHz.
The gas can either flow through the DBD (ozone generation,
C02
conversion,
H2S
disposal, pollution
control) or it can
be
recirculated
(C02
lasers) or even fully encapsulated (excimer lamps).
In
most high-
power applications efficient cooling of at least one
of
the electrodes is used.
BREAKDOWN PHENOMENA AND DISCHARGE PHYSICS
At atmospheric pressure electrical breakdown
in
a large number of statistically distributed microdischarges
is the normal situation for most gases in DBD configurations. It should be mentioned, however, that
discharge plasmas which
are
apparently homogeneous can
be
obtained in the same electrode configurations
if the pressure is lowered. These discharges
are
normally referred to as
RF
discharges and have found
widespread applications in the semiconductor industry for plasma etching and plasma deposition
procedures. Even at elevated pressures close to 1 bar fairly homogeneous discharges can
be
obtained if the
discharge is pulsed, if large amounts of diluents like He or Ne are added
(32, 34,42,43)
or if special
additives are used
(44).
The normal appearance
of
dielectric barrier discharges at elevated pressure, however, is that shown in Fig.
1. It is characterized by a large number of short-lived microdischarges. Each microdischarge has
an
almost
cylindrical plasma channel of typically
100
pm
radius and spreads into a larger surface discharge at the
dielectric surface.
By
applying an electric field larger than the breakdown field local breakdown
in
the gap
is initiated. At the given conditions propagating electron avalanches quickly produce such a high space
charge that self-propagating streamers
are
generated (16). This condition is typically met before the initial
Modelling of dielectric barrier discharge chemistry
1277
electron avalanche reaches the opposite electrode. The field enhancement at the streamer head, moving
much faster than the electron drift velocity, is reflected at the anode and travels back to the cathode where,
within a fraction of 1 ns, an extremely thin cathode layer is formed. At this time a conductive channel,
which can be charaterized as a transient glow discharge, bridges the gap. At atmospheric pressure electron
densities of 1014 to 1015 cm-3 are reached. The thickness of the cathode layer reaches only a few pm
(45,
46). Charge accumulation at the dielectric surface results in a local reduction of the electric field which
chokes the current flow in a microdischarge typically within less than 100 ns. For a given gas, the choking
effect depends on the extension of the surface discharge and the properties of the dielectric barrier. If the
external voltage is still rising, additional microdischarges will preferentially strike at other locations with
higher electric fields.
Thus,
the dielectric serves a dual purpose. It limits the amount of charge and energy
imparted to an individual microdischarge and, at the same time, distributes the microdischarges over the
entire electrode area. Typical charges transported by individual microdischarges are of the order of nC,
typical energies of the order of
pJ.
It is important to realize that some control of the plasma characteristics is possible by making use of special
gas properties, adjusting the pressure or changing the electrode geometry or the properties of the dielectrics.
The radius rma of the microdischarge column, for example, depends on the gas density n and on the gas
properties. rma is proportional to the reciprocal value of the product of the gas density n and the derivative
of the effective ionization coefficient
E
with respect to the reduced electric field E/n (at breakdown) (47).
Fig.
3
shows an for different gases. The effective ionization coefficient
2
is obtained from the ionization
coefficient
a
and the attachment coefficient
q.
Its value corresponds to
(a
-
q).
The sharp decay of the
curves of
02
and C02 below breakdown in Fig.
3
is due to electron attachment in these electronegative
gases. From these curves it is evident that the radii of the microdischarge channels can
be
grouped iu the
following order:
oxygen
c
carbon dioxide
<
air
c
nitrogen
c
xenon
c
helium.
EFFECTIVE IONIZATION
COEFFICIENT
E/n
=
(a-q)/n
10
100
1000
Eln
(Td)
1
SILENT
DISCHARGE
CURRENT PULSES
I
0
40
80
120
TIME
(
nsec
)
Fig.
3:
Effective ionization coefficient indifferent Fig.
4:
Microdischarge current pulses in different
The shape of the current pulse of a microdischarge is influenced by the gas properties and by
the
relative
overvoltage applied to the discharge gap. Fig.
4
shows current pulses in different gases, all calculated for
1 bar. The reduced fields were
230
Td, 155 Td, 155 Td and 100 Td for N2, 02,
(2%
and Xe respectively.
Fig.
5
shows three current pulses in CO;? at different voltages. Surprisingly enough, the transported charge
in these
three
pulses is practically identical.
The total charge
Q
in a microdischarge depends on the gas properties and can be influenced by the gap
spacing and by the properties of the dielectric.
Q
is proportional to the width of the discharge gap d, and to
the quantity &/g (E: relative permittivity,
g:
thickness of dielectric). Recently, the latter relation was
experimentally checked to hold up to extreme &-values of about 1000
(48).
Contrary to what one might
expect,
Q
does
not
depend on the gas density n (16).
gases. gases.
1278
B.
ELIASSON,
W.
EGLl AND
U.
KOGELSCHATZ
SILENT DISCHARGE
CURRENT PULSE IN
COq
I
mA)l
,
TIME
f
nsec
1
DISSOCIATION
OF OXYGEN AND CARBON DIOXIDE
e
+
C02
-CO+
0
+
a
0
x
P
10-12
10 100 l0OC
E/n
(Td)
Fig.
5:
Mcrodischarge current pulses in
COZ
at
different overvoltages.
MICRODISCHARGE
CHEMISTRY
Fig.
6:
Dissociation rate coefficient
in
02
and
CR.
The initial phases of microdischarge formation are characterized by electron multiplication, space charge
formation and ionization, dissociation and excitation processes initiated by energetic electrons. The ionic
and excited atomic and molecular species initiate chemical reactions that finally result in the synthesis of a
desired species (e.g. ozone, excimers, methanol) or the destruction of pollutants (e.g.
H2S,
NOx,
SOx,
VOCs etc.). The resulting chemistry can
be
dominated by charged particle reactions in which case the term
plasma chemistry would adequately describe the situation. This situation occurs in many low pressure
discharges.
In
the majority of DBD applications, however, most charged particles decay before any major
chemical changes happen.
In
this case it is more appropriate to speak of a free radical chemistry involving
neutral species like atoms, molecular fragments and excited molecules. In any case, discharge activity and
energy dissipation occurs only within the microdischarges and sets the initial conditions for the ensuing
chemical reactions. Thus, a correct description
of
the physical processes during breakdown and
microdischarge formation is prerequisite for a detailed understanding of DBD chemistry.
In many cases the first step is a dissociation of the initial species by electron collisions. Fig.
6
shows the
rate coefficients for this dissociation process for the two important gases
02
and
C02
as a function of the
reduced electric field. The dissociation
of
02
(49-53)
has been investigated in connection with ozone
OXYGEN
FRACTIONAL ENERGY LOSSES
1
uu
DISSOCIATION
(
6.0
ov
+
8.4
OV)
YO)
80
60
40
20
0
1
10 100
1000
Eln
(Td)
CARBON
DIOXIDE
FRACTIONAL
ENERGY
LOSSES
loor
EXCITATION
8o
VIBRATION
.OF
\/7
DISSOCIATION
(6.9~~
+ll.Jev)
ELASTIC
1
10
100
1000
Eln
(Td)
Fig.
7:
Fractional energy losses in
02.
Fig.
8:
Fractional energy losses in C02.