You are on page 1of 8

2644 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO.

10, OCTOBER 2010


Mechanisms of Impulse Breakdown in Liquid:
The Role of Joule Heating and Formation
of Gas Cavities
Vladimir M. Atrazhev, Vladimir S. Vorobev, Igor V. Timoshkin, Member, IEEE, Martin J. Given, Member, IEEE,
and Scott J. MacGregor, Member, IEEE
AbstractThe impulse dielectric behavior of insulating liquids
is of signicant interest for researchers and engineers working in
the eld of design, construction, and operation of pulsed power
systems. Analysis of the literature data on transformer oils shows
that potentially there are several different physical processes that
could be responsible for dielectric breakdown by submicrosec-
ond and microsecond impulses. While for short submicrosecond
impulses ionization (plasma streamer) is likely to be the main
breakdown mechanism, for longer impulses, the thermal effects
associated with Joule heating start to play an important role.
This paper provides a theoretical analysis of the latter mechanism
in dielectric liquids of different degrees of purity stressed with
high-voltage (HV) impulses with duration sufcient to cause local
heating, evaporation, and formation of prebreakdown gas bubbles.
The proposed model is based on the assumption that dielectric
breakdown is developed through percolation channels of gas bub-
bles, and the criterion of formation of these percolation chains is
obtained. To test the developed model, the breakdown eld-time
characteristics have been calculated for the liquid with chemical
composition close to that of transformer oils but with known
thermodynamic characteristics (n-hexane). Its dielectric strength
has been obtained as a function of externally applied pressure and
temperature. The analytical results show good agreement when
compared with the experimental data available in the literature.
Index TermsDielectric liquids, impulse breakdown, volttime
characteristics.
I. INTRODUCTION
T
HE DIELECTRIC properties of liquids, such as mineral
oils, liquid hydrocarbons, and ester uids, have been
studied for several decades, and a signicant number of re-
search papers that focused on the breakdown and prebreakdown
behavior of liquids stressed with high voltage (HV) have been
published [1][5]. This interest in different aspects of the
dielectric characteristics of liquids results from the widespread
applications of liquid dielectrics in power and pulsed power
Manuscript received October 25, 2009; revised January 22, 2010; accepted
March 8, 2010. Date of publication May 3, 2010; date of current version
October 8, 2010. This work was supported by the Royal Society through the
award of the Joint International Grant in 2007.
V. M. Atrazhev and V. S. Vorobev are with the Theoretical Department,
Joint Institute for High Temperatures, Russian Academy of Sciences, Moscow
127412, Russia (e-mail: atrazhev@yandex.ru; vrbv@mail.ru).
I. V. Timoshkin, M. J. Given, and S. J. MacGregor are with the Department of
Electronic and Electrical Engineering, University of Strathclyde Royal College
Building, G1 1XW Glasgow, U.K. (e-mail: igor.timoshkin@eee.strath.ac.uk;
m.given@eee.strath.ac.uk; S.j.macgregor@eee.strath.ac.uk).
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TPS.2010.2046337
technologies. In conventional power systems, such as power
transformers, the liquids are normally stressed with dc or low-
frequency ac voltages. Certain pulse power applications, for
example, ultrawideband sources of electromagnetic radiation,
the generation of impulse plasmas for drilling and mineral
processing, and medical applications including sterilization
and lithotripsy, use pulses in the microsecond to millisecond
regime, generally driven by Marx and Blumlein generators.
However, there are a range of modern pulsed power applica-
tions where the requirements are such that the liquid insulators
must be capable of withstanding impulses with magnitudes
higher than tens of megavolts per centimeter with rise times
shorter than 1 ns. For example, HV impulses with duration
of hundreds of picoseconds are used for the generation of
subnanosecond electron beams in pulsed X-ray systems and for
the generation of hot plasmas.
It is known that the breakdown voltage (breakdown eld
E) of dielectric materials is a function of the duration t of
the applied electrical eld and such a dependence is normally
called the voltage or eld-time (V -t or E-t) characteristic.
Typically, the breakdown eld of dielectric materials, including
liquids, increases as the pulse duration is reduced: the shorter
the voltage impulse, the higher the electric eld the dielectric
materials can withstand. Knowledge of the eld-time behavior
of the insulating liquids is therefore important from an engi-
neering point of view as this information helps in the design
and optimization of pulsed power systems, such as advanced
compact HV generators and modulators.
Fig. 1 shows the E-t data for transformer oils available in the
literature [6][13].
As expected, the breakdown strength of the oil decreases with
an increase in the duration of the applied HV impulses. The
time scale on this graph covers seven orders of magnitude, from
10
10
to 10
3
s, and the data cannot be tted with a single
analytical function.
In [13], Martin reported an empirical scaling rule for the
impulse breakdown strength for transformer oil. According to
this author, experimental E-t data for transformer oil stressed
with impulses with effective duration between 150 and 300 ns
can be tted by the following scaling law:
E t
2/3
(1)
where E is the electric eld (in megavolts per centimeter),
and t is the duration of the voltage pulse between 63% and
0093-3813/$26.00 2010 IEEE
ATRAZHEV et al.: MECHANISMS OF IMPULSE BREAKDOWN IN LIQUID: JOULE HEATING AND GAS CAVITIES 2645
Fig. 1. E-t characteristics of transformer oils. The open triangles and squares
are from [6], the solid squares are from [7], the star is from [8], the solid circles
are from [9] and [10], the open circles are from [11], the solid diamonds are
from [12], and the solid line is from scaling law (1) [13]. Dash-dotted line is
from tting of data [9][11] with (2), the dotted line is from tting of data [12]
with (2), and the dashed line is from tting of data [7] with (2).
TABLE I
PARAMETERS A AND B IN (2) FOR SELECTED EXPERIMENTAL DATA SETS
100% of the crest voltage. This equation gives a straight line
in a logarithmic scale for the E-t curve shown in Fig. 1 by a
solid line. The open triangles and squares in Fig. 1 present data
from [6], the closed triangles present data from [7], the solid
circles present data from [9] and [10], the open circles present
data from [11], the solid star is taken from [8], and the solid
diamonds are from [12]. All of the breakdown data have been
measured for transformer oils.
Following the phenomenological scaling rule given in [13],
the analytical tting of selected experimental data has been
conducted with the power curves. The straight dashed lines in
Fig. 1 show the result of this tting procedure based on the
following equation:
E = a t
b
(2)
where E is in megavolts per centimeter, and t in nanoseconds.
The short stroke line indicates the analytical tting of the data
from [12] (solid diamonds), the dashed-dot line shows the t of
the data from [7] (solid squares), and the dashed line is a t of
the data given in [9][11] (open and closed circles). Parameters
a and b for all these cases are listed in Table I.
There is 10% and 23% difference between the values
of parameters b for the data given in [7] and [9][11] and the
coefcient 2/3 provided in [13], respectively. The coefcient
b for the data in [12] is 67% lower from the coefcients
b for the data given in [7]. The chemical composition of
these transformer oils, the different degrees of purity, and the
different experimental conditions could make contributions to
this difference in the slopes (coefcients b) of the log E- log t
curve. However, the potentially dominant factor responsible
for such difference could be the signicantly longer duration
of voltage impulses (0.110 s) used in the tests described in
[12] as compared with the shorter impulses (0.3100 ns) used
in other studies, for example, [6] and [7]. Looking at Fig. 1,
one may conclude that different physical mechanisms may be
responsible for the dielectric behavior of liquids at different
time scales. While for shorter impulses, i.e., 0.3100 ns, the
formation and propagation of the ionization wave (streamer)
could be the main mechanism of breakdown, and for longer
impulses, the development of gas bubbles due to Joule heating
and the evaporation of the liquid could become the dominant
process. This paper is aimed at a theoretical analysis of the
latter breakdown mechanism in dielectric liquids of differ-
ent degrees of purity. While transformer oils are of interest
for engineers and researchers from a practical point of view,
these liquids have quite a complex chemical composition, and
their thermodynamic properties are not readily available in the
literature, which makes difculties for theoretical modeling.
In this paper, n-hexane, a hydrocarbon liquid with chemical
composition similar to that of mineral transformer oils and with
well-known thermodynamic and physical characteristics, has
been chosen as a model liquid to compare the analytical results
with experimental data.
The effect of the hydrostatic pressure, temperature, and im-
pulse duration on the dielectric strength of hydrocarbon liquids
has been studied experimentally by many researchers, for ex-
ample, [16][18]. In [17], the authors showed that the electric
strength of liquid hydrocarbons has a weak dependence on the
duration of the voltage impulses if these impulses are longer
than several s: for example, the breakdown eld for n-hexane
increases by only 15% with a decrease in the pulse duration
from 4.5 to 1 s. However, for shorter impulses (< 0.1 s),
the breakdown strength of dielectric liquids increases more
signicantly with decreasing pulse duration, following the be-
havior described in (2). For example, setting the coefcient
b = 0.5114 (Table I, third column), the breakdown eld of
the transformer oil increases by 130% with a decrease in the
pulse duration from 0.1 to 0.02 s.
The decrease in the duration of the applied impulses and
the growth in the breakdown eld elevate the probability of
the development of an ionization wave (streamer) in the liquid
without its transformation into gaseous state. The qualitative
estimation of the breakdown eld in liquid n-hexane stressed
with short impulses that do not lead to boiling has been
conducted in [15]. This estimation provides the breakdown
value for the liquid E
br
(Liq) 2 MV/cm, which is half of
the breakdown value obtained by extrapolation of the Paschen
curve for gaseous n-hexane into the region of liquid densi-
ties E
br
(Gas) 4 MV/cm. This theoretical estimation of the
breakdown eld, i.e., E
br
(Liq) 2 MV/cm, is lower than the
experimental breakdown values for transformer oils stressed
with submicrosecond impulses, i.e., 10 MV/cm, as shown
in Fig. 1. Potentially, this increase in E
br
(Liq) with further
decreases in the impulse duration into the submicrosecond
2646 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO. 10, OCTOBER 2010
region can be explained by the increase in the velocity of the
ionization wave (streamer) propagating in the liquid [5]; this
effect starts to be signicant for impulses shorter than 100 ns. If
the streamer velocity is close to the drift velocities of electrons,
i.e., 10
4
m/s, then it takes 50 ns for the streamer to bridge a
0.5-mm interelectrode gap (experimental conditions described
in [8]). Hence, the dielectric breakdown in liquid stressed with
shorter impulses requires a higher streamer velocity, which will
necessitate higher electric elds.
The breakdown behavior of a pure liquid hydrocarbon
(n-hexane) also indicates that different physical processes may
be involved in the shorter and longer time scales. The weak
dependence of the breakdown behavior on temperature and
pressure using short impulses agrees with an ionization mecha-
nism as impact electron ionization in the dielectric liquids does
not depend on the external pressure and temperature.
The observed dependence on pressure and temperature for
longer impulses must require some additional mechanism, lead-
ing to the hypothesis that in the longer time scale the gas
cavities may be formed by Joule heating of the liquid, which
will result in eld enhancement and gas breakdown inside these
bubbles. An experimental observation that supports this bubble
formation mechanism is based on the measurements of the
breakdown voltage U
br
for gaseous and liquid n-hexane as a
function of the product of the interelectrode spacing d and the
gas density N, i.e., U
br
(Nd), i.e., the Paschen curve. According
to the Townsend ionization mechanism, the dielectric strength
of a vaporized liquid or gas increases with an increase in
their density N or external pressure p [14]. The externally
applied electric eld accelerates electrons, which potentially
can gain sufcient energy to cause ionization of molecules.
This ionization process can be described by the Townsend
ionization coefcient (E/N, N), which is proportional to the
gas density N. /N is a strong function only of the ratio
E/N. If the ionization process leads to the formation of an
electron avalanche causing breakdown, then the product of the
gap length d and the Townsend coefcient is constant. In the
planeplane electrode geometry, this condition can be written as

N
_
U
br
Nd
_
Nd = const. (3)
This equation describes the breakdown voltage U
br
as a func-
tion of the product Nd, i.e., U
br
(Nd), which is an analytical
description of the Paschen law. The Paschen curves for high val-
ues of Nd (above the Pashen minimum) for liquid and gaseous
n-hexane are shown in Fig. 2. Experimental points starting from
Nd 3540 10
22
m
2
and higher show a deviation from the
similarity law [see (3)] for U
br
(Nd), which potentially can
be attributed to the effects of local eld enhancement on the
cathode surface due to mircoprotrusions [27].
Line 2 in Fig. 2 shows the analytical values of U
br
(Nd),
which have been calculated by taking into account ionization
processes in a dense media [15]. The breakdown voltages
measured in supercritical gaseous n-hexane stressed with 10-s
voltage impulses [16] and in liquid n-hexane stressed with
shorter impulses [17], [18] are also given in Fig. 2, and these
experimental voltages are close to the analytical values. How-
ever, the magnitudes of breakdown voltages measured in liquid
Fig. 2. Breakdown voltage of n-hexane as a function of parameter Nd.
Curves: (1) extrapolation of low-density gas data [14] and (2) calculation
for high-density medium ionization [15]. Points are experimental data for
(3) liquid, 10 s pulse [16]; (4) liquid, 1.2 s pulse [17]; (5) liquid, 0.5 s
pulse [18]; and (6) supercritical dense gas, 10 s pulse [16].
n-hexane stressed with 10-s impulses [16] are lower than
those predicted by the theory of ionization in liquids: these
experimental breakdown data are similar to U
br
for dense gases
rather than liquids. This observation supports the suggestion
that for longer impulses, the electrical breakdown occurs in
a liquidgas mixture. This conclusion also follows from the
pressure dependence of E
br
for longer pulses discussed in [17].
The present investigation includes the analyses of a cavity-
based breakdown mechanism in dielectric liquids, assuming
that vapor bubbles are formed in liquids stressed with HV
impulses of length greater than 1 s. Joule heating leads to
the evaporation of the liquid; as a result, gas-lled bubbles
are formed, and the external electric stress E can trigger the
dielectric breakdown of the suspension of bubbles. The analysis
of bubble formation is conducted by applying the theory of
homogeneous nucleation in superheated liquids. The bubble
concentration, their radii, and the expansion times have been
estimated. The electric breakdown of the network of bubbles
lled with vapor is described using percolation theory and the
Paschen characteristics for the vapor [19]. Analytical expres-
sions that link the breakdown strength of the insulating liquids
with their temperature, external pressure, interelectrode gap,
and pulse duration have been obtained, and these analytical re-
sults have been compared with the experimental data available
in the literature.
II. PREBREAKDOWN PROCESSES IN PURE LIQUIDS
STRESSED WITH HV PULSES
A. Heating of Pure Liquid by Conduction Current
In the case of nanosecond and subnanosecond impulses,
the main breakdown mechanism is through the formation and
ATRAZHEV et al.: MECHANISMS OF IMPULSE BREAKDOWN IN LIQUID: JOULE HEATING AND GAS CAVITIES 2647
propagation of an ionization wave (streamer). However, in the
longer microsecond time scale, Joule heating starts to play a
dominant role and results in the development of a gasliquid
mixture. In this latter case, which is considered in this paper,
dielectric breakdown progresses through the gas bubbles in the
liquid. It is assumed that the liquid is stressed with HV impulses
with duration sufcient to cause Joule heating resulting from
electronic currents emitted from inhomogeneities on the cath-
ode surface. In the theory for simplicity, such emission spots
are modeled as planes.
The space-charge limited current emitted into a liquid is
governed by the continuity equation, whereas the electric eld
is governed by the Poisson equation
j = en
e
W = const,
dE
dx
=
en
e

r
. (4)
The space-charge limited current density can be expressed
as a linear function of the eld
j = (E E
c
)

0

r
2

W
d
(5)
where d is the electrode separation,
r
2 is the dielectric per-
mittivity of the liquid, and E
c
10
8
V/m is the eld strength
at the cathode [21]. Experimental data on the eld emission
currents in liquid n-hexane have been published [22], [23],
and it is possible to make an estimation of the current density.
For the case with a eld of magnitude E = 2 10
8
V/m and
an interelectrode distance of d = 200 m, which corresponds
to the experimental conditions described in [17], the current
density is j = 0.6 10
6
A/m
2
, assuming that the electron drift
velocity is W = 10
4
m/s, which is close to its gas kinetic value
of 2 10
4
m/s [20] in a eld with magnitude of E 10
8
V/m.
This current density results in the prebreakdown current value
of 0.2 mA for the cathode emission spot with linear dimension
of 10 m.
The value of j allows an estimation of the time that is
required to heat the liquid from an initial temperature T
0
to its
boiling temperature T
b
(p) to be made

h

c
p
N
l
(T
b
(p) T
0
) k
jE
1.3 s. (6)
Here, N
l
= 4.6 10
27
m
3
is the number density of liquid
n-hexane, c
p
= 24 is its specic heat capacity, and k is the
Boltzmanns constant. The boiling temperature of n-hexane
T
b
(p) varies with the external pressure from 342 K for p =
0.1 MPa to 480 K for p = 2.0 MPa. According to (6), the time
that is required to heat n-hexane up to its boiling point, for the
current density derived using (5), is in the microsecond scale.
Assuming a cylindrical geometry for the conductive channel
between electrodes, the heat conduction equation can be
written as
c
p
N
l
dkT
dt
= jE +
d
2
kT
dr
2
. (7)
Energy losses due to thermal conduction can be neglected if
jE (kT/L
2
). It is possible to estimate linear dimensions
Fig. 3. Evaporation (p, T) diagram for liquid n-hexane. The solid line is for
the boiling temperature [24]. The vertical line shows the temperature intervals
corresponding to heating from T
0
= 290 K to the boiling point T
b
and to the
spinodal point T
s
at p = 0.5 MPa.
of the current emission spot L that satisfy the criterion
L
_
kT
b
jE
_
0.5
0.6 m (8)
where = 8 10
21
(m s)
1
is the thermal conductivity of liq-
uid n-hexane normalized by the Boltzmanns constant. Hence,
the thermal conduction losses in n-hexane could be neglected
for emission spots larger than 0.6 m.
B. Evaporation of Pure Liquid and Development of Bubbles
As the temperature of the liquid in the Joule heating zone
reaches its boiling value T
b
(p), the formation of vapor bubbles
becomes possible. This formation process requires a specic
time during which the liquid continues to be heated by the
electronic current. In the case of a pure liquid without any
impurities that can act as nucleation sites, this heating process
can continue beyond the boiling temperature of the liquid,
which means that the liquid could be heated up to the limit
of its absolute thermodynamic stability, i.e., the spinodal line
in p-T graph. Such superheated liquid is metastable, and its
phase transition due to uctuations into a thermodynamically
stable gaseous state takes place instantaneously, i.e., signif-
icantly faster than all the other processes considered in this
model. This type of evaporation from the metastable state
(referred to as a burst-like or explosive boiling process [28])
takes place at temperatures close to the spinodal temperature
T
s
, which does not depend on the applied electrical eld or
electric current. The spinodal temperature of liquid n-hexane
T
s
is close to its critical temperature T
s
0.95 T
cr
480 K
and is a weak function of the external pressure, as shown
in Fig. 3.
The burst-like boiling process of liquid evaporation and the
formation of bubbles is determined by the molecular transition
from the liquid into the gas state and starts from the develop-
ment of critical bubbles with radii of
R
c
(T
s
) =
2
p
b
(T
s
) p
10
9
m (9)
2648 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO. 10, OCTOBER 2010
where 2 10
3
N/m is the surface tension of liquid
n-hexane, and p
b
(T
s
) 2.5 MPa is its saturated vapor pressure
corresponding to the spinodal temperature.
The energy required for this process k(T
s
-T
b
) is supplied by
the surrounding superheating liquid. The critical concentration
of the bubbles N
b
is determined by the heating time
h
as
N
b
=
c
p
N
l
T
b
(p)k
q

vap

= 1.3 10
21
m
3

vap
=
(2MT
s
k)
0.5
8R
c
(T
s
)
(10)
where
vap
5 10
13
s is the characteristic evaporation time
(time required to create the bubble at the spinodal line condi-
tions), q 4.8 10
20
J is the boiling energy per molecule, and
M = 1.4 10
25
kg is the molecule mass.
The rst fraction in (10) for N
b
determines the concentration
of the vapor molecules; however, the total concentration of the
bubbles N
b
is signicantly smaller because (
vap
/
h
) 1.
The higher energy c
p
N
l
(T
s
T
b
)k accumulated in the su-
perheated liquid is spent on the generation of vapor molecules,
resulting in an increase in the dimension (radius R) of the
bubbles, but not in their number.
The number of vapor molecules is N
b
(4R
3
/3)N
vap
and the
vapor density at the boiling temperature T
b
is N
vap
= p/T
b
.
The energy that is required to generate a single vapor molecule
is q, so the total energy balance is
c
p
N
l
(T
s
(p) T
b
) k = N
b
4R
3
3
p
T
b
q. (11)
Using (10) and (11), the nal radius of the bubble, as the
liquid cools down to its boiling temperature during a short
period of time
gr
10
11
s, can be estimated as
R =
_
3(T
s
T
b
)k
4p

vap
_
1/3

= 10
7
m. (12)
The bubbles grow due to the evaporation of the liquid
molecules into their gas interior. This process stops as soon as
the temperature of the surrounding liquid decreases from the
spinodal value T
s
down to its boiling value T
b
. As a result of
this burst-like evaporation, a foam (liquid enriched with vapor
bubbles) is developed. The vapor concentration in the foam
4R
3
N
b
/3 is higher than unity, which means that this liquid
is transformed into its microdispersed state which has a lower
breakdown strength, close to that of the vapor.
The main feature of this breakdown mechanism is the heating
of the liquid up to its spinodal temperature, which leads to
the development of a large number of vapor cavities. This
mechanism provides a nominal pressure dependence of the
breakdown eld; however, it cannot explain the change in
breakdown voltage with the external pressure that is observed in
the experiments. Potentially, this model will provide adequate
results for liquids pressurized up to such a degree that T
b
(p)
becomes close to T
s
.
III. PREBREAKDOWN PROCESSES IN PRACTICAL LIQUIDS
WITH IMPURITIES STRESSED WITH HV IMPULSES
In practice, the dielectric liquids will have different degrees
of purity and may have relatively high concentrations of mi-
croscale or nanoscale impurities. In this practical situation, lo-
cal evaporation in the stressed liquid starts on these impurities,
which become nucleation centers for the vapor bubbles. In this
situation, the overcritical heating to the spinodal temperature
before bubbles are formed does not need to occur: the boiling
process starts as soon as the temperature of the liquid reaches
its boiling value T
b
(p). In this model, the number density of the
vapor bubbles is equal to the number density of the nucleation
centers, i.e., the number density of impurities in the liquid N
p
.
When the liquid is being heated by the conduction current, the
molecules can evaporate into the gas cavities, resulting in an
increase in the radius of these prebreakdown bubbles R(t).
A percolation channel of bubbles can form between the
electrodes if the total volume of gas bubbles within the liquid
exceeds the percolation threshold. For a 3-D geometry, this
threshold occurs when the volume of gas bubbles is 1/3 of the
volume of the liquid [29]. The vapor inside these bubbles has
a lower dielectric permittivity and a lower breakdown strength
compared with the bulk liquid. Therefore, the dielectric strength
of the liquid with the formed percolation channel is lower
compared with the dielectric strength of the liquid with no gas
cavities leading to breakdown. This criteria can be expressed in
terms of R(t) and N
p
as
4
3
R(t)
3
N
P
=
1
3
. (13)
This allows the conditions for the dielectric breakdown
through a percolation channel to be formulated. Joule heating of
the liquid is inuenced by several factors, including the current
density, the electric eld strength, and the impulse duration.
The formation of the continuous percolation channel in the
impulse electric eld requires heating of the liquid from its
initial temperature T
0
up to the boiling temperature T
b
(p),
followed by the development of bubbles. From the energy point
of view, the generation of a volume of vapor that is sufcient
for the development of the percolation chain requires an energy
density of
4
3
R(t)
3
N
P
N
V
q = 0.3N
V
q (14)
where q is the boiling energy per molecule, and N
v
is the vapor
density.
Using (14), it is possible to write the energy balance equation
assuming that all the energy delivered into the liquid by an
external electric source through the impulse conduction current
(Joule heating) is sufcient for the local heating and evaporation
of the liquid and for the formation of the continuous percolation
channel as
j(E
br
)E
br
= c
p
N
l
(T
b
(p) T
0
) k + 0.3qp/T
b
(p) (15)
where j is the electric current density, E
br
is the eld strength
enough for breakdown, is the impulse duration, and T
0
and p
are the initial temperature and pressure in liquid.
ATRAZHEV et al.: MECHANISMS OF IMPULSE BREAKDOWN IN LIQUID: JOULE HEATING AND GAS CAVITIES 2649
Fig. 4. Field-time characteristics for liquid n-hexane for different tempera-
tures: (1) 213 K; (2) 290 K; and (3) 313 K. Solid lines are calculations using
(16); points are experimental data from [17].
The rst term in the right-hand side of (15) describes the
energy required for heating the liquid to its boiling temperature,
and the second term in the right-hand side of this equation
describes the energy required for evaporation of the liquid.
It is assumed that the vapor in the bubbles satises the ideal
gas condition N
v
= p/T
b
(p). An analytical expression for the
breakdown eld can be obtained using (5) and (15) as
E
br
=

_
2d
_
c
p
kN
l
(T
b
(p) T
0
) +
0.3qp
T
b
(p)
_

r
W
+
E
2
c
4
+
E
c
2
.
(16)
Equation (16) has been used in the analysis of the dielectric
behavior of liquid n-hexane. The eld time E-t characteristics
of this liquid have been obtained at three temperatures, i.e.,
213 K, 290 K, and 313 K, for impulses with duration from
0.5 to 6 s. The analytical results are shown in Fig. 4 together
with the experimental data published in [17]. It can be seen
that the present model [see (16)] adequately describes the main
feature of liquid n-hexane E-t curves: the breakdown eld
E
br
decreases with an increase in pulse duration. For shorter
impulses, the higher breakdown strength can be explained by
the necessity of dissipating higher levels of electrical power
to provide sufcient energy for heating and evaporation to
form a percolation channel within the duration of the im-
pulse. Equation (16) has also been used in the analysis of
the breakdown eld of liquid n-hexane as a function of the
externally applied pressure (for pressures up to 2 MPa) for
pulses with durations of 2 and 4.5 s. Fig. 5 shows the results
of this analysis: the breakdown eld of liquid n-hexane in-
creases with the increase in the external pressure at xed initial
temperature.
This increase in the breakdown strength can be explained by
the growth of the boiling temperature with external pressure
T
b
(p), which means that a larger amount of Joule energy is
required to heat the liquid to its boiling point. Stronger electric
elds are required for breakdown in pure liquids as from the
present model the pure liquid should be heated to its spinodal
Fig. 5. Breakdown strength of liquid n-hexane as a function of the external
pressure for different pulse duration. Sold lines show calculations by (16) for
2- and 4.5-s impulses; points show experimental data from [17] for the same
pulse duration: (1) = 2 s; and (2) = 4.5 s. E
s
is the breakdown strength
for pure liquid n-hexane stressed with 2-s impulse.
temperature T
S
= 480 K, which is higher than the boiling
temperature.
The type of breakdown that occurs, either through the for-
mation of a percolation channel from bubbles initiated at impu-
rities or the burst-like boiling at the spinodal temperature, will
depend on whether the time to form a percolation channel
ch
is
signicantly shorter than the time required to heat the liquid to
its spinodal temperature:
ch

h
. The growth of the bubbles
is restricted by the inertia and thermal conductivity of the liquid.
The latter process is faster than the Rayleigh expansion of the
bubble, and the time
R
, which is required for the bubble to
reach a radius R, can be written as [26]

R
= 1.83R
_
MN
l
/p. (17)
As discussed above, to satisfy the breakdown condition [see
(16)], the volume of the bubbles should be 0.3 of the liquid
volume, so the concentration of the impurities in the liquid
N
p
required to trigger the percolation breakdown mechanism
instead of the burst boiling mechanism can be estimated as
N
p

0.9
4

_
_
MN
l
p
cr
_
0.5
1

h
_
3
10
11
m
3
. (18)
According to [17], this concentration of impurities may exist
in liquid n-hexane, which suggests that the reported breakdown
behavior can be attributed to the growth of bubbles, leading to
the formation of a percolation channel.
IV. CONCLUSION
This paper has analyzed the physical processes that can
occur in liquid dielectrics stressed with HV impulses and that
may result in their dielectric breakdown. For longer impulses
with duration of 0.51 s and above, Joule heating could
be a dominant factor, which results in local boiling of the
2650 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO. 10, OCTOBER 2010
TABLE II
ABBREVIATIONS AND UNITS
liquid. This model is supported by experimental data on the
dielectric behavior of liquids with different external pressures
and pulse durations. This paper discusses two cases: 1) the
case of pure liquid and 2) the case of a liquid with impu-
rities, which could become nucleation centers. In the former
case, the burst-like boiling of the liquid can be triggered by
thermodynamic density uctuations, and this type of boiling
process occurs when the liquid is heated by the conduction
current up to its spinodal temperature, where the liquid becomes
thermodynamically unstable. The time interval that is required
for such burst boiling is very short, i.e.,
gr
0.01 ns, and the
burst-like boiling model describes neither the time dependence
of the breakdown strength E
br
(t) nor its pressure dependence
E
br
(p), which have been observed in the experiments. In the
case of a liquid containing impurities, it is assumed that a local
boiling process takes place, where these impurities play the
role of primary nucleation centers. In this model, the liquid
is heated only up to its boiling temperature. The gas bubbles
are formed on the nucleation centers, and they are lled with
saturated vapor with a density signicantly lower than the den-
sity of the surrounding liquid and a lower dielectric permittivity
and breakdown strength. For a eld strength of 1 MV/cm,
electron emission into the liquid plays an important role in
the formation of these prebreakdown gas cavities. In this case,
the condition for the breakdown eld can be obtained based
on Joule heating of the liquid up to its boiling temperature. If
the gas volume produced exceeds 0.3 of the liquid volume,
a percolation channel formed from the gas-lled bubbles pro-
vides a path for the complete breakdown of the gap between
the electrodes. Analytical calculations conducted in this paper
show that this model describes the temperature and pressure
behavior E
br
(t) and E
br
(p) of the breakdown voltage of liquid
n-hexane stressed with HV impulses with duration in the range
from 1 to 6 s for pressures up to 2 MPa [17]. According to
the present model, the breakdown strength of liquid n-hexane
increases with an increase in the external pressure and tends to
the dielectric strength of pure n-hexane. Analytical E-t curves
agree well with the experimental data for the microsecond time
scale: the breakdown voltage of liquid n-hexane increases with
a decrease in the pulse duration, and a higher eld magnitude
is required to compensate for shorter impulses to deliver the
energy sufcient for the local evaporation of the liquid.
APPENDIX
See Table II.
ACKNOWLEDGMENT
V. M. A. and V. S. V. would like to thank the University
of Strathclyde for support during their stays in Glasgow, and
M. J. G. would like to thank the Joint Institute for High
Temperatures for support during his stays in Moscow.
REFERENCES
[1] A. H. Sharbaugh, J. C. Devins, and S. J. Rzad, Progress in the eld
of electrical breakdown in dielectric liquids, IEEE Trans. Elect. Insul.,
vol. EI-13, no. 4, pp. 249276, Aug. 1978.
[2] R. Tobazeon, Liquid breakdown and its relation to gas breakdown,
in Gaseous Dielectrics VI, L. G. Christophorou and I. Sauers, Eds.
New York: Plenum, 1991, pp. 159169.
[3] R. Tobazeon, Prebreakdown phenomena in dielectric liquids, IEEE
Trans. Dielectr. Elect. Insul., vol. 1, no. 6, pp. 11321147, Dec. 1994.
[4] Y. V. Torshin, Prediction of breakdown voltage of transformer oil from
predischarge phenomena, IEEE Trans. Dielectr. Elect. Insul., vol. 10,
no. 6, pp. 933941, Dec. 2003.
[5] A. Denat, High eld conduction and prebreakdown phenomena in dielec-
tric liquids, IEEE Trans. Dielectr. Elect. Insul., vol. 13, no. 3, pp. 518
525, Jun. 2006.
[6] J. Mankowski, J. Dickens, and M. Kristiansen, High voltage subnanosec-
ond breakdown, IEEE Trans. Plasma Sci., vol. 26, no. 3, pp. 874881,
Jun. 1998.
[7] J. Mankovski, L. Hateld, and M. Kristiansen, Nanosecond breakdown
in liquid dielectrics, in Proc. 11th Int. Conf. High Power Particle Beams
(BEAMS), Prague, Czech Republic, Jul. 1996.
[8] J. M. Lehr, F. J. Agee, R. Copeland, and W. D. Prathe, Measurement of
the electric breakdown strength of transformer oil in the sub-nanosecond
regime, IEEE Trans. Elect. Insul., vol. 5, no. 6, pp. 857861, Dec. 1998.
[9] K. A. Zheltov and A. N. Petrenko, Subnanosecond high-current electron
accelerator, Instrum. Exp. Tech., vol. 28, pt. 1, no. 5, pp. 10121015,
1986.
[10] K. A. Zheltov, V. I. Matakov, A. N. Petrenko, V. F. Shalimanov, and
T. V. Chernyaeva, Pulse generator of monochromatic X-radiation, Prib.
Tekh. Eksp., vol. 5, pp. 183185, 1981.
[11] A. A. Vorobev, U. Y. Ushakov, and V. V. Bagin, Electrical strength of
liquid dielectrics stressed with nanosecond impulses, Electrotekhnika,
vol. 7, pp. 5557, 1971.
[12] K. C. Kao and J. P. C. McMath, Time-dependent pressure effect in
liquid dielectrics, IEEE Trans. Elect. Insul., vol. EI-5, no. 3, pp. 6468,
Feb. 1970.
ATRAZHEV et al.: MECHANISMS OF IMPULSE BREAKDOWN IN LIQUID: JOULE HEATING AND GAS CAVITIES 2651
[13] J. C. Martin, Comparison of breakdown voltages for various liquids un-
der one set of conditions, in J.C. Martin on Pulsed Power, T. H. Martin,
A. H. Guenther, and M. Kristiansen, Eds. New York: Plenum, 1996.
[14] A. E. Heylen and T. J. Lewis, The electric strength of hydrocarbon
gases, Brit. J. Appl. Phys., vol. 7, no. 11, pp. 411415, Nov. 1956.
[15] V. M. Atrazhev, E. G. Dmitriev, and I. T. Iakubov, The impact ionization
and electrical breakdown strength for atomic and molecular liquids,
IEEE Trans. Elect. Insul., vol. 26, no. 4, pp. 586591, Aug. 1991.
[16] A. H. Sharbaugh and P. K. Watson, The electric strength of hexane vapor
and liquid in the critical region, J. Appl. Phys., vol. 48, no. 3, pp. 943
951, Mar. 1977.
[17] K. C. Kao and J. B. Higham, The effect of hydrostatic pressure, tempera-
ture and voltage duration on the electric strength of hydrocarbon liquids,
J. Electrochem. Soc., vol. 108, pp. 522528, 1961.
[18] R. W. Crowe, A. H. Sharbaugh, and J. K. Bregg, Electric strength and
molecular structure of saturated hydrocarbon liquids, J. Appl. Phys.,
vol. 25, no. 12, pp. 14801484, 1954.
[19] J. S. Mirza, C. W. Smith, and J. H. Calderwood, Sparking potentials
of saturated hydrocarbon gases, J. Phys. D, Appl. Phys., vol. 4, no. 8,
pp. 11261133, Aug. 1971.
[20] L. G. Cristophorou, R. P. Blaunstein, and D. Pittman, Mobilities of
thermal electrons in -electron containing organic gases, Chem. Phys.
Lett., vol. 22, pp. 4147, 1973.
[21] K. Dotoku, H. Yamada, S. Sakamoto, S. Noda, and H. Yoshida, Field
emission into nonpolar organic liquids, J. Chem. Phys., vol. 69, no. 3,
pp. 11211125, Aug. 1978.
[22] P. K. Watson and A. H. Sharbaugh, High-eld conduction currents in
liquid n-hexane under microsecond pulse conditions, J. Electrochem.
Soc., vol. 107, pp. 516521, 1960.
[23] W. F. Schmidt, Electronic conduction processes in dielectric liquids,
IEEE Trans. Elect. Insul., vol. EI-19, no. 5, pp. 389418, Oct. 1984.
[24] NIST. [Online]. Available: http://webbook.nist.gov/chemistry
[25] O. Lesaint and G. Massala, Positive streamer propagation in large oil
gap: Experimental characterisation of propagation modes, IEEE Trans.
Elect. Insul., vol. 5, no. 3, pp. 360370, Jun. 1998.
[26] L. Rayleigh, On the pressure developed in a liquid during the collapse of
a spherical cavity, Philos. Mag., vol. 34, no. 199, pp. 9498, 1917.
[27] G. A. Mesyats, Pulsed Power. Berlin, Germany: Springer-Verlag, 2004.
[28] V. P. Skripov, Metastable Liquids. New York: Halsted, 1974.
[29] C. Lorenz and R. Ziff, Precise determination of the critical percolation
threshold for the three dimensional swiss cheese model using a growth
algorithm, J. Chem. Phys., vol. 114, no. 8, pp. 36593661, 2001.
Vladimir M. Atrazhev was born in Moscow,
Russia, in 1949. He received the degree from
Moscow State University, Moscow, in 1972 and
the Ph.D. degree from the Institute for High Tem-
perature, Russian Academy of Sciences, Moscow,
in 1979.
From 1972 to 2002, he was a Scientic Researcher
with the Theoretical Department, Institute for High
Temperature, Russian Academy of Sciences. In
2002, he became a Lead Researcher with the
Institute for High Energy Densities, Joint Institute
for High Temperatures. Since 1992, he has been a Visiting Researcher with
the Laboratoire dElectrostatique et de Materiaux Dielectriques, CNRS, and
Fourier University, Grenoble, France. He is currently with the Theoretical
Department, Joint Institute for High Temperatures, Russian Academy of
Sciences. His research interests include transport properties of electrons in
strong electric eld in dense media, properties of nonequilibrium plasma and
electrical discharges, electric breakdown in dense gases, and liquid dielectrics.
Vladimir S. Vorobev received the Ph.D. and D.Sci.
degrees from the Institute for High Temperatures,
Russian Academy of Sciences, Moscow, in 1967 and
1979, respectively.
He is the Head of the Theoretical Department,
Joint Institute for High Temperatures, Russian Acad-
emy of Sciences, Moscow, Russia. He became a
Professor of mathematics in 1985. He specializes
in the thermophysics and electrophysics properties
of condensed matter and plasma. He is the author
of numerous publications in peer-reviewed journals
and the coauthor of the book on physics of low-temperature plasma. His
research interests include similarity laws in thermodynamic, phase equilibrium
in external electrical eld, uctuation theory of phase transitions, and similarity
laws for metastable states in gases and liquids.
Igor V. Timoshkin (M07) received the degree
in physics from Moscow State University (MSU),
Moscow, Russia, in 1992 and the Diploma and
Ph.D. degrees from the Imperial College of Science,
Technology and Medicine (ICSTM), London, U.K.,
in 2001.
After graduation from MSU, he was a Researcher
with the Moscow State Agro-Engineering Univer-
sity, Moscow. Then, he was with the Institute for
High Temperatures, Russian Academy of Sciences,
before moving to ICSTM in 1997. He joined the
Department of Electronic and Electrical Engineering, University of Strathclyde,
Glasgow, U.K., in 2001 as an Academic Visitor, where he became a Research
Fellow in 2002 and a Lecturer in 2006. His research interests include properties
of solid and liquid dielectric materials, electronics of plasma discharges in
condensed media, practical applications of electrohydraulic and high-power
ultrasound pulses, biodielectrics, and effects of electromagnetic elds on bi-
ological objects.
Martin J. Given (M99) received the degree in
physics from the University of Sussex, Brighton,
U.K., in 1981 and the Ph.D. degree in electronic
and electrical engineering from the University of
Strathclyde, Glasgow, U.K., in 1996.
He is currently a Senior Lecturer with the Depart-
ment of Electronic and Electrical Engineering, Uni-
versity of Strathclyde. His research interests include
ageing processes and condition monitoring in solid
and liquid insulation systems, high-speed switching,
and pulse power applications.
Scott J. MacGregor (M95) received the B.Sc. and
Ph.D. degrees from the University of Strathclyde,
Glasgow, U.K., in 1982 and 1986, respectively.
He became a Pulsed Power Research Fellow in
1986 and a Lecturer in pulsed-power technology in
1989. In 1994, he became a Senior Lecturer, with a
promotion to Reader and Professor of High Voltage
Engineering, in 1999 and 2001, respectively. Since
January 2010, he has been the Dean of the Engineer-
ing Faculty, University of Strathclyde. His research
interests include high-voltage pulse generation, high-
frequency diagnostics, high-power repetitive switching, high-speed switching,
electronic methods for food pasteurization and sterilization, generation of high-
power ultrasound (HPU), plasma channel drilling, pulsed-plasma cleaning of
pipes, and stimulation of oil wells with HPU.

You might also like