Content uploaded by Liang Zhao
Author content
All content in this area was uploaded by Liang Zhao on Oct 12, 2022
Content may be subject to copyright.
An experimental and theoretical investigation into the “worm-hole” effect
Liang Zhao, Jiancang Su, Xibo Zhang, Yafeng Pan, Limin Wang et al.
Citation: J. Appl. Phys. 114, 063306 (2013); doi: 10.1063/1.4818446
View online: http://dx.doi.org/10.1063/1.4818446
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v114/i6
Published by the AIP Publishing LLC.
Additional information on J. Appl. Phys.
Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors
An experimental and theoretical investigation into the “worm-hole” effect
Liang Zhao,
a)
Jiancang Su, Xibo Zhang, Yafeng Pan, Limin Wang, Jinpeng Fang, Xu Sun,
Rui Li, Bo Zeng, and Jie Cheng
Science and Technology on High Power Microwave Laboratory, Northwest Institute of Nuclear Technology,
P. O. Box 69 Branch 13, Xi’an, Shannxi 710024, China
(Received 11 May 2013; accepted 30 July 2013; published online 12 August 2013)
On a nanosecond time scale, solid insulators abnormally fail in bulk rather than on surface, which
is termed as the “worm-hole” effect. By using a generator with adjustable output pulse width and
dozens of organic glass (PMMA) and polystyrene (PS) samples, experiments to verify this effect
are conducted. The results show that under short pulses of 10 ns, all the samples fail due to bulk
breakdown, whereas when the pulse width is tuned to a long pulse of 7 ls, the samples fail as a
result of surface flashover. The experimental results are interpreted by analyzing the conditions for
the bulk breakdown and the surface flashover. It is found that under short pulses, the flashover
threshold would be as high as the bulk breakdown strength (E
BD
) and the flashover time delay (t
d
)
would be longer than the pulse width (s), both of which make the dielectrics’ cumulative
breakdown occur easily; whereas under long pulses, that E
f
is much lower than E
BD
and t
d
is
smaller than sis advantageous to the occurrence of the surface flashover. In addition, a general
principle on solid insulation design under short pulse condition is proposed based on the
experimental results and the theoretical analysis. V
C2013 AIP Publishing LLC.
[http://dx.doi.org/10.1063/1.4818446]
I. INTRODUCTION
For solid insulation structures, two factors may lead to a
catastrophic failure: surface flashover and bulk breakdown.
Generally, on a microsecond time scale, the electric field
threshold of the surface flashover is lower than that of the
bulk breakdown, and insulator failures are mostly caused by
surface flashover. Therefore, lots of theories and various
methods are developed to enhance the surface flashover
threshold on a solid/vacuum
1–8
interface or a solid/liquid
9–13
interface. However, on a nanosecond time, insulators mostly
fail due to bulk breakdown rather than surface flashover. Only
several literatures reported this abnormal phenomenon,
among which, Roth and Chantrenne first observed a bulk
breakdown trace in a vacuum insulator ring in the PITHON,
which they termed as the “worm-hole” effect due to a worm-
hole appearance;
14,15
afterwards, Zhao et al. observed the
same phenomenon on a coaxial vacuum insulator in the accel-
erator TPG700.
16
Aside from those results on vacuum/ solid
interfaces, there were also bulk breakdown traces on the trans-
former oil/polymer interfaces, as reported by Wang.
17
No matter where this effect happens, the time scale for
these experiments is nanosecond. Recently, there were
also similar traces observed on the insulators used in ultra-
wide-band (UWB) generators at Northwest Institute of
Nuclear Technology (NINT), the pulse width of which is
600 ps, and the photos of the failed insulators are shown in
Fig. 1. From this figure, one can clearly see wormhole-like
breakdown traces on the insulator surfaces. In this paper,
the term of the “worm-hole” effect is adopted, which is to
denote the phenomenon that solid dielectrics are prone to
breakdown in bulk rather than on surface on a nanosecond
time scale.
According to Fig. 1as well as the experimental results
aforementioned, a “rough” conclusion can be drawn that the
bulk breakdown is the main factor leading solid insulation
structures to fail on short time scales, rather than the surface
flashover. Here, the word “rough” is used, which means that
direct comparisons with the same insulator profile under
different pulse widths are not reported. In addition, there are
little mechanisms for the so-called “worm-hole” effect.
To present a reasonable explanation as well as to derive
useful suggestions to avoid the occurrence of the “worm-
hole” effect under short pulses, further research is needed. In
this paper, an experimental and theoretical investigation for
the “worm-hole” effect is presented, which is formulated in
five sections. Section II is mainly devoted to experiments,
which compare the failure patterns of the same test samples
under different pulse widths. Section III is devoted to the the-
oretical analysis from the perspective of surface flashover
threshold and flashover time delay. Based on the theoretical
analysis as well as the experimental results, a general princi-
ple on solid insulation design is suggested, which is presented
in Sec. IV. Section Vis for the conclusions in this paper.
II. EXPERIMENTAL VERIFICATION OF THE
WORM-HOLE EFFECT
Two sets of experiments were designed and conducted
to compare the failure patterns of dielectrics on different
time scales.
A. Experimental setup
The schematic diagram of the experimental setup is
shown in Fig. 2, which mainly comprises of a nanosecond
a)
Author to whom correspondence should be addressed. Electronic mail:
zhaoliang0526@163.com. Tel: þ86-29-84767621.
0021-8979/2013/114(6)/063306/6/$30.00 V
C2013 AIP Publishing LLC114, 063306-1
JOURNAL OF APPLIED PHYSICS 114, 063306 (2013)
pulse generator, TPG200, a transmission microscope, and
a set of control and diagnostic system. The TPG200 is a
Tesla-type generator,
18,19
which can produce trapezoidal
pulses with a width of 10 ns, a rise and fall time of 3 ns, and
a maximum amplitude of 300 kV. By shortening the gas-gap
switch of TPG200, quasi sine-wave pulses with a full width
at half maximum (FWHM) of 7 ls can also be produced.
The transmission microscope is specially designed with
resolution of 0.7 lm, which can record the images at a rate
of 15 frames per second. The control and diagnostic system
includes a Rogowski coil, a voltage divider, an oscillograph,
and a PC.
A working cycle of the experimental setup is as follows:
(1) a trigger signal is launched to the TPG200 via the PC and
a microsecond or nanosecond pulse is then generated and
imposed on the test sample; (2) the current and voltage
waveforms on the sample are recorded on the oscillograph
via the Rogowski coil and the voltage divider, and then
stored in the PC; (3) the microscopic image of the test sam-
ple are also recorded and transmitted to the PC simultane-
ously via the transmission microscope.
B. Electrodes and test samples
The electrodes are composed of a truncated cone and a
plate, both of which are made of copper. The front of the
cone is a circle with a diameter of 1 mm, which is parallel to
the plate to produce a local quasi-uniform electric field. The
plate is cylindrical with a radius of 30 mm. The test samples
are made of two types of polymers: PMMA (organic glass,
dielectric constant e
r
¼3.6–4, and luminousness 92%) and PS
(polystyrene, e
r
¼2.4–2.6, and luminousness 90%). The sam-
ples are cubes with a size of 2225 mm
3
(thickness (d)
width length), which is to meet both the requirements of
observation and insulation.
During the test, the electrodes and a test sample were all
immersed in clear transformer oil (e
r
¼2.2–2.3) in order to
create a comparable insulation circumstance for the surface
and the bulk of the sample. Fig. 3shows the field distribution
on the mid cross-section of a PMMA sample under an applied
voltage of 100kV. Based on this figure, the field distributions
along the ‘inner line’ and the ‘surface line’ are respectively
obtained, as shown in Fig. 4. It is seen that the field distribu-
tion along the surface (surface line) and in the bulk (inner
line) of the test sample are basically equal to each other, with
only an average a relative deviation less than 10%. Therefore,
a comparable insulation circumstance is created (Figure 4).
C. Experimental results
Two sets of experiments are designed and conducted,
the first was under a short pulse with width 10 ns, and the
second was under a long pulse of 7 ls. In the short-pulse
experiments, the test procedure was as follows: (1) Fixed the
applied voltage as U; (2) Launched a pulse; (3) Observed the
test sample via the microscope, if any bulk breakdown or
surface flashover occurred, stopped to shot the pulse; if not,
continued to impose the pulse on the test sample until the
sample failed. The applied voltage was respectively set as
170 kV, 130 kV, and 100 kV to observe the different failure
patterns, and the time interval between each pulse was about
1 s. The test results are listed in Table I, which shows that all
the samples failed in bulk breakdown.
By tuning the output pulse width to 7 ls, the long-pulse
experiments were also conducted. Taking into account that
FIG. 2. Schematic diagram of the experimental setup.
FIG. 3. Electric field distribution on the mid cross-section of the test sample
(y ¼0, U¼100 kV)
FIG. 1. Photo of the failed insulators due to bulk breakdown under sub-
nanosecond pulses with pulse width of 0.6 ns.
063306-2 Zhao et al. J. Appl. Phys. 114, 063306 (2013)
the surface flashover would cause the insulation to fail and
that a flashover often occurs in one pulse with certain ampli-
tude, we gradually increased the voltage from low to high to
find the exact flashover voltage. The experimental results are
listed in Table II, which shows that even thought the flash-
over voltage is different for the two types of polymers, the
samples are all failed due to surface flashover. That the flash-
over voltage of PS is a little higher than that of PMMA is
because e
r
of PS is more closely matched to that of the trans-
former oil. Similar results under nanosecond pulses in a
uniform field can also be seen in Ref. 11.
It is worth mentioning that the number of the test sam-
ples for surface flashover in Table II is relatively smaller
than that for the bulk breakdown in Table I. This is because
lots of researches have been conducted on the surface flash-
over in oil and it is widely accepted that flashover can cause
the insulation to fail under long pulses. It is also worth men-
tioning that the number of the test sample at low voltage (U)
in Table I(short pulse) is relative smaller than that at high
voltage. This is because the pulse number (N
L
) is inverse
proportional to U
8
, and a small decrease of Uwould result in
a considerable increase of N
L
. For example, when Uis
decreased from 170 kV to 100 kV, N
L
is increased from 10
2
to about 10
4
, the latter of which means a huge work due to
one by one count in our experiments.
27
So tests of the sam-
ples at low voltage are repeated only by 1–3 times.
To further explore the failure patterns, the typical failure
images for each type of samples under different test condi-
tions are compared, as shown in Figs. 5and 6. From the two
groups of images, it is clearly seen that the bulk breakdown
traces are coherent, opaque, and punctured; whereas the sur-
face flashover traces are incoherent, transparent, and random.
This is probably due to the place where the failure occurs.
For the bulk breakdown, the traces are inside polymers,
where the decomposition products such as carbon, short-
chain molecules and small gas molecules are confined, which
make the traces opaque; whereas for the surface flashover,
the traces are on the interfaces, where the products can be
easily diffused to the oil, which makes the trace transparent.
It is noted that the bulk breakdown traces are totally consist-
ent with the wormhole appearance described in Refs. 14–17.
With the experimental results in Table I, Table II,
Figs. 5and 6, a conclusion can be drawn that solid insulation
structures are prone to breakdown in bulk under short pulses,
whereas they tend to suffer surface flashover under long
pulses.
III. THEORETICAL ANALYSIS
A. Failure condition from the perspective of flashover
threshold
As aforementioned, on microsecond time scale, the sur-
face flashover threshold (E
f
) is generally lower than that of
the bulk breakdown (E
BD
). However, when the time scale is
decreased to a nanosecond time, this conclusion would not
hold true. Fig. 7summarizes the experimental data of E
f
dependent on pulse width (s) on a PMMA/transformer oil
interface.
20,21
From this figure, it is seen that in a microsec-
ond pulse width range, E
f
is lower than 0.5 MVcm
1
;
whereas in a pulse width range smaller than 10 ns, E
f
is as
high as 1 MVcm
1
. The latter is close to E
BD
of polymers
under the same pulse width, which can be seen in Ref. 18.
Such a high E
f
would have influences on the applied field
FIG. 4. Electric field distributions along the inner and the surface lines.
TABLE I. Failure results of PMMA and PS under a short pulse width of
10 ns.
Sample
type
Test
voltage/kV
Number of
test samples
Failure
pattern
Pulse numbers
until failure
PMMA 170 9 Bulk 300
PMMA 130 2 Bulk 1000
PMMA 100 3 Bulk 10 000
PS 170 2 Bulk 328
PS 130 2 Bulk 2000
PS 100 1 Bulk 57 453
TABLE II. Failure experimental results under a long pulses of 7 ls.
Sample type
Failure
voltage /kV
Number of
test samples Failure pattern
PMMA 80 2 Surface
PS 85 3 Surface
FIG. 5. Comparison for the failure images of PMMA. (a) 10 ns/100 kV/
N¼8733; (b) 7 ls/80 kV/N¼1, where Nrepresents the pulse number until
failure occurs.
063306-3 Zhao et al. J. Appl. Phys. 114, 063306 (2013)
(E
op
), leading to the acceleration of the processes of electron
emission and the degradation of polymers.
Taking a 2 mm cylindrical PMMA test sample as an
example, the E
BD
(PMMA) j
s¼10 ns
is 1.6MVcm
1
according
to Ref. 18, whereas the E
f
(PMMA) j
s¼10 ns
is about 1.2
MVcm
1
from Fig. 6. In practical application, a safe factor
(b
s
) which is defined as E
BD
/E
op
or E
f
/E
op
is usually set equal
to or larger than 2. If b
s
¼2, since E
op
is smaller than E
BD
,
E
op
j
s¼10 ns
is set as 0.6 MVcm
1
. Under a long pulse condi-
tion of 7 ls, the E
f
is about 0.4 MVcm
1
. So, the corre-
sponding E
op
is only 0.2 MVcm
1
. According to the famous
Fowler-Nordheim formula:
22
jðTÞ¼1:54106E2
/m
exp 6:83107/1:5
m
EhðyÞ
gðE;/m;TÞ;
(1)
where j(T) is the current density in A/cm
2
;/mthe work func-
tion of the cathode in eV; h(y) and ythe middle parameters,
which are that h(y)¼0.956–1.06y
2
and y¼3.8 10
4
ffiffiffi
E
p=/m; the influence due to the variation of E
op
on electron
emission under different pulse widths can be calculated out.
Fig. 8shows the curves of j(T) dependent on E
op
for different
/m. From this figure, it is seen that j(T) for a copper
electrode (/m¼4.5 eV) under 10 ns is approximately
increased by 10 orders compared with that under 7 ls. The
increase of j(T) accelerates the creation of the ‘free radicals’,
which are the small unpaired molecules in polymers.
23
Fig. 9
shows the density of free radicals (j
D
) dependent on E
op
in a
wide electric field range.
24
From this figure, it is seen that j
D
is increased approximately by 7 orders when the pulse width
is tuned from 7 ls to 10 ns. The cluster of free radicals is ad-
vantageous to the formation of the discharge channel in poly-
mers.
25
Once a discharge channel emerges in an insulator, it
would grow gradually as the pulse number increases. In this
way, a bulk breakdown event may occur and a wormhole
trace appears. However, under long pulses, since E
op
is much
lower than that under short pulses, the rate for the creation of
free radical is much slower, and the cumulative breakdown
is developed with a much slower rate accordingly. When E
op
is gradually increased, the probability for the surface flash-
over is increased considerably, which makes the test samples
fail in surface flashover.
FIG. 6. Comparison for the failure images of PS. (a) 10 ns /170 kV/N¼328;
(b) 7 ls/85 kV /N¼1.
FIG. 7. General tendency of E
f
versus son the PMMA/transformer oil
interface.
FIG. 8. j(T) versus E
op
for different work function, where Tis set as the
room temperature of 300 K.
FIG. 9. Density of free radical versus E
op
in a wide field range. The raw ex-
perimental data (solid line) are obtained in Ref. 24.
063306-4 Zhao et al. J. Appl. Phys. 114, 063306 (2013)
B. Failure condition from the perspective of flashover
time delay
With the conception of flashover time delay (t
d
), the dif-
ferent failure patterns for the test samples in short and long
pulses can further be explained. t
d
is defined the time interval
from the time to apply a field to the time when a flashover
event occurs. Fig. 10 shows a general tendency of E
f
versus
t
d
for a PMMA/transformer oil interface under nanosecond
pulses. From this figure, it is seen that t
d
decreases rapidly
as E
f
increases. In addition, t
d
is generally proportional
to E
f1.4
.
On short time scales, as the pulse width decreases, t
d
would be longer than s, which may make the flashover hard to
occur. The underlying reason lies in that the plasma produced
by electron multipactor can not cover the distance between
cathode and anode timely. However, as the pulse number
increases, the discharge channel inside the polymers can
increase gradually as aforementioned. Consequently, a bulk
breakdown event occurs. Fig. 11 compares the general tenden-
cies of E
f
versus sand E
f
versus t
d
. From this figure, it is obvi-
ously seen that a critical time interval, t
c
, exists. When the
time interval is shorter than t
c
,t
d
would be larger than s,which
may prevent the occurrence of surface flashover, whereas
when the time interval is longer than t
c
,swould be longer
than t
d
, which allows the occurrence of surface flashover.
It is considered that the critical time interval is merely a
conceptional parameter, which may be affected by many fac-
tors, such as dielectric profile, dielectric type, and liquid
type, etc., and that the specific value of this parameter is
hard to calculate out.
IV. SUGGESTIONS ON SOLID INSULATION DESIGN
Based on the experimental results and the theoretical
analysis, a general principle on solid insulation design under
short pulse condition is proposed. On a nanosecond or sub-
nanosecond time scale, since E
f
is as large as E
BD
and the
flashover time delay is larger than the pulse width, the focus
for solid insulation design should be on how to prevent
the dielectrics’ inner bulk breakdown, rather than the
conventional surface flashover. For a specific insulation
structure, if the lifetime, N
L
, is required to be no smaller than
a given value, then by modifying the dielectric thickness, d,
or the applied voltage, U, this requirement can be realized
with the following formula:
27
NL¼d7E1
U
8
;(2)
where E
1
is the E
BD
for a unit dielectric thickness, which
represents the dielectric type. The units for a set of d,E
1
and
Uare, respectively, mm, kV, and kVmm
1
or cm, MV, and
MVcm
1
. Still taking a 2 mm PMMA sample as an exam-
ple, since E
1
(PMMA)j
d¼1mm, s¼10 ns
¼173 kVmm
1
,ifN
L
is
required to be larger than 1 10
4
, then Ushould be no
greater than 100 kV. Similarly, for a PMMA sample to suffer
pulse number greater 1 10
4
, if the sustained voltage is fixed
as 200 kV, then the sample should be thicker than 4.4 mm,
accordingly.
V. CONCLUSIONS AND REMARKS
In summary, the famous “worm-hole effect” that solid
dielectrics are prone to fail due to bulk breakdown on short
time scales is experimentally verified by conducting the fail-
ure experiments using the same PMMA and PS test samples
under pulse widths of 10 ns and 7 ls, respectively. In
addition, by summarizing the tendency of E
f
versus son a
PMMA/transformer oil interface and by fitting the experi-
mental data of t
d
versus E
f
in the literatures, it is concluded
that on a nanosecond time scale, E
f
is as large as E
BD
and t
d
is larger than s, both of the two which are responsible for the
occurrence of the “worm-hole effect.”
The theoretical analysis for the “worm-hole effect” is
presented only from the perspective of surface flashover
threshold and flashover time delay in this paper. As a matter
of fact, on a nanosecond scale, the electric field rise-up rate,
@E/@t, would be greater than that on a microsecond time
scale by three orders for the same applied voltage generally.
In addition, the polarization mechanism of polymers on a
FIG. 10. Fit for the tendency of t
d
versus E
f
in a log-log coordinate system.
The raw experimental data are from Ref. 26.
FIG. 11. Comparison of the general tendencies of E
f
versus sand E
f
versus
t
d
in a log-log coordinated system.
063306-5 Zhao et al. J. Appl. Phys. 114, 063306 (2013)
nanosecond time scale is different from that on a microsec-
ond time scale.
28
Whether the two differences have influen-
ces on the occurrence of the “worm-hole effect” should also
be studied in future.
ACKNOWLEDGMENTS
The author gratefully acknowledges the contributions of
Q. Ge and Q. Lin for their help on the experiments.
1
H. C. Miller, IEEE Trans. Electr. Insul. 24, 765 (1989).
2
H. C. Miller, IEEE Trans. Electr. Insul. 28, 512 (1993).
3
O. Milton, IEEE Trans. Electr. Insul. EI-7, 9 (1972).
4
C. Chang, J. Y. Fang, Z. Q. Zhang, C. H. Chen, C. X. Tang, and Q. L. Jin,
Appl. Phys. Lett. 97, 141501 (2010).
5
C. Chang, H. J. Huang, G. Z. Liu, C. H. Chen, Q. Hou, J. Y. Fang, X. X.
Zhu, and Y. P. Zhang, J. Appl. Phys. 105, 123305 (2009).
6
C. Chang, G. Z. Liu, C. X. Tang, C. H. Chen, H. Shao, and W. H. Huang,
Appl. Phys. Lett. 96, 111502 (2010).
7
P. Yan, T. Shao, J. Wang, W. Gao, W. Yuan, R. Pan, S. Zhang, and G.
Sun, IEEE Trans. Dielectr. Electr. Insul. 14, 634(2007).
8
R. A. Anderson and J. P. Brainard, J. Appl. Phys. 51, 1414 (1980).
9
M. P. Wilson, M. J. Given, I. V. Timoshkin, S. J. MacGregor, M. A.
Sinclair, K. J. Thomas, and J. M. Lehr, IEEE Trans. Plasma Sci. 38, 2611
(2010).
10
M. P. Wilson, S. J. MacGregor, M. J. Given, I. V. Timoshkin, M. A.
Sinclair, K. J. Thomas, and J. M. Lehr, IEEE Trans. Dielectr. Electr. Insul.
16, 1028 (2009).
11
M. P. Wilson, I. V. Timoshkin, M. J. Given, and S. J. MacGregor, IEEE
Trans. Dielectr. Electr. Insul. 18, 1003 (2011).
12
A. Sharma, S. Acharya, K. V. Nagesh, R. C. Sethi, U. Kumar, and G. R.
Nagabhushana, presented at the IEEE 31st International Conference on
Plasma Science (2004).
13
W. L. Huang and Z. X. Cheng, IEEE Trans. Dielectr. Electr. Insul. 17,
1938 (2010).
14
S. Chantrenne and P. Sincerny, in Procedings of the 12th IEEE
Inernational Pulsed Power Conference (Monterey, California, 1999),
pp. 1403–1407.
15
I. S. Roth, P. S. Sincerny, L. Mandelcorn, M. Mendelsohn, D. Smith, T. G.
Engel, L. Schlitt, and C. M. Cooke, in Procedings of the 11th IEEE
International Pulsed Power Conference (IEEE, Baltimore, Md, 1997), pp.
537–542.
16
L. Zhao, J. C. Peng, Y. F. Pan, X. B. Zhang, and J. C. Su, IEEE Trans.
Plasma Sci. 38, 1369 (2010).
17
J. Wang, P. Yan, S.-C. Zhang, and G.-S. Sun, presented at the Proceedings
of the 15th IEEE International Pulsed Power Conference (2005).
18
L. Zhao, G. Liu, J. Su, Y. Pan, and X. Zhang, IEEE Trans. Plasma Sci. 39,
1613 (2011).
19
L. Zhao, J. C. Su, X. B. Zhang, and Y. F. Pan, IEEE Trans. Dielectr.
Electr. Insul. 19, 1101 (2012).
20
A. Sharma, K. V. Nagesh, R. C. Sethi, U. Kumarm and G. R. Nagabhushana,
presented at the IEEE International Vacuum Electronics Conference
(2002).
21
B. B. Bakin and V. Y. Ushakov, Electronics (Russia) 4, CTP,76 (1972).
22
H. Bluhm, Pulsed Power Systems (Springer, Karlsruhe Germany, 2006),
Chap. 2.
23
Z. Li, Y. Yin, X. Wang, D. M. Tu, and K. C. Cao, J. Appl. Polym. Sci. 89,
3416 (2003).
24
K. C. Kao, Dielectric Phenomenon in solid (Elsevier academic press,
Amsterdam, 2004), Chap. 8.
25
H. K. Xie and K. C. Kao, IEEE Trans. Electr. Insul. EI-20, 293 (1985).
26
G. J. Li, “Experimental Study on Dielectric Surface Flashover under
Nanosecond Pulse in Transformer Oil,” Master of Engineering Graduate
(University of the Chinese Academy of Sciences, Beijing, 2006),
Chap. 4.
27
L. Zhao, J. C. Su, X. B. Zhang, Y. F. Pan, L. M. Wang, X. Sun, and R. Li,
IEEE Trans. Plasma Sci. 41, 165 (2013).
28
L. Zhao, J. C. Su, Y. F. Pan, and X. B. Zhang, Chin. Phys. B 21(3),
033102 (2012).
063306-6 Zhao et al. J. Appl. Phys. 114, 063306 (2013)
