Enhancement of the electrical stability of ZnO varistors by a novel immersion process

Article abstract of DOI:10.1111/j.1551-2916.2011.04453.x

A novel immersion process was carried out to improve the electrical stability of ZnO varistors under a dc electrical field. It was performed by dipping the conventionally sintered ZnO varistor samples in a bismuth nitrate solution with a subsequent heat treatment. The results showed that this process could improve the electrical stability and the nonlinearity of the ZnO varistors simultaneously. All the immersed specimens exhibited a reduced leakage current and a stable nonlinearity in the dc degradation test. The optimum concentration of the immersion solution was found to be 1.0-2.0M. The microstructure and the electrical performance of the immersed ZnO varistors were investigated. This process generates an effective penetration of molten Bi₂O₃ from the surface into the interior, which compensates the near-surface bismuth loss that resulted from its vaporization during sintering. The extraneously penetrated molten Bi₂O₃ plus inherently molten Bi-rich phase facilitate the reconstruction of an intergranular layer and fills more grain boundaries with the γ-Bi₂O₃ phase. It is the abundant γ-Bi₂O₃ intergranular phase that improves the electrical stability by restraining the oxygen desorption at the grain boundary. The enhanced nonlinearity is related to the modification of the defect structure and the component distribution in the intergranular region.

Full text of DOI:10.1111/j.1551-2916.2011.04453.x

J. Am. Ceram. Soc., 94 [9] 2939–2945 (2011)
DOI: 10.1111/j.1551-2916.2011.04453.x
r 2011 The American Ceramic Society
ournal
J
Enhancement of the Electrical Stability of ZnO Varistors by a Novel
Immersion Process
Lei Meng, Liaoying Zheng, Lihong Cheng, Zheng Yao, and Guorong Liw
Key Laboratory of Inorganic Functional Materials and Devices, Shanghai Institute of Ceramics, Chinese Academy of
Sciences, Shanghai 200050, China
Slavko Bernik
Department for Nanostructured Materials, Jozef Stefan Institute, Ljubljana SI-1000, Slovenia
˘
Slovenia Center of Excellence NAMASTE, Ljubljana 1000, Slovenia
A novel immersion process was carried out to improve the elec-
trical stability of ZnO varistors under a dc electrical field. It was
performed by dipping the conventionally sintered ZnO varistor
samples in a bismuth nitrate solution with a subsequent heat
treatment. The results showed that this process could improve
the electrical stability and the nonlinearity of the ZnO varistors
simultaneously. All the immersed specimens exhibited a reduced
leakage current and a stable nonlinearity in the dc degradation
test. The optimum concentration of the immersion solution was
found to be 1.0–2.0M. The microstructure and the electrical
performance of the immersed ZnO varistors were investigated.
This process generates an effective penetration of molten Bi2O3
from the surface into the interior, which compensates the near-
surface bismuth loss that resulted from its vaporization during
sintering. The extraneously penetrated molten Bi2O3 plus inher-
ently molten Bi-rich phase facilitate the reconstruction of an
intergranular layer and fills more grain boundaries with the
c-Bi2O3 phase. It is the abundant c-Bi2O3 intergranular phase
that improves the electrical stability by restraining the oxygen
desorption at the grain boundary. The enhanced nonlinearity is
related to the modification of the defect structure and the com-
ponent distribution in the intergranular region.
layer abounds in electron trap states,6 which are responsible for
the nonlinear current–voltage behavior of the ZnO varistor.
Considerable attention has been focused on the dc degrada-
tion of ZnO varistors, because it remains the primary restriction
of ZnO varistors’ application in dc high-voltage power trans-
mission and urban rail transport. When subjected to a contin-
uous electric field, ZnO varistors exhibit a continuous decrease
in the nonlinearity exponent and an increase in the leakage cur-
rent, which finally leads to thermal runaway of the varistor. Al-
though many studies have looked at degradation under an ac7
and pulsed electrical field,8 dc degradation is still difficult to
overcome due to the severe migration of metastable ions under
the constant electrical field. Several methods have been pro-
posed to improve the dc electrical stability of ZnO varistors. Fan
and Freer9 reported that certain Ag-doped ZnO Varistors
showed obviously modified electrical stability but with a re-
duced nonlinear exponent. Wu and Shyu10 have pointed out
that the addition of B–Si–Pb-based glass could improve the dc
degradation behavior of ZnO varistors by decreasing unstable
zinc vacancies, but it also led to a deterioration of the nonlinear
exponent; a heat treatment at 6001–7001C after sintering11–14
was an effective method to improve the electrical stability of a
ZnO varistor by diffusing out unstable zinc interstitials prefer-
entially. Iga et al.13 and Takemura et al.15 reported that the
g-Bi2O3 phase, which transitioned from the b-Bi2O3 phase when
annealed around 6001–7001C, could effectively enhance the elec-
trical stability of ZnO varistors. However, a deteriorated non-
linear exponent11,13 was also found. Almost all these methods
were accompanied by a deterioration of the nonlinearity expo-
nent. Consequently, more efforts are required to enhance the
electrical stability of ZnO varistors without deteriorating their
electrical properties.
I. Introduction
ZnO varistors have been extensively used as protection devices
against transient voltage surges in electrical and electronic sys-
tems because of their excellent nonlinear current–voltage prop-
erty and their high energy-absorption capability. At present,
commercial varistors tend to be ZnO–Bi2O3-based varistors. Re-
garded as the primary ingredient, Bi2O3 plays a very important
role in constructing the intergranular potential barrier.1–3 Bis-
muth mainly exists at the grain boundary because of the larger
ion radius of Bi31 (B0.114 nm), compared with that of Zn21
(B0.074 nm). Bi2O3 is prone to forming a liquid phase during the
sintering due to a solid-solution reaction with ZnO at about
7401C4 and its low-melting point (B8251C).5 Thus, it can pref-
erentially dissolve the nonlinearity-forming additives and distrib-
ute them homogeneously around the ZnO grain. When cooling
down, the Bi-rich intergranular phase is formed and mainly con-
centrates at the multigrain junctions, with the additives segregat-
ing uniformly at the grain boundaries. This Bi-rich intergranular
Considering the important contribution of Bi2O3 to nonlinear
behavior and the remarkable effectiveness of heat treatment in
improving electrical stability, there appears to be an opportunity
to solve this issue by controlling the bismuth content and the
Bi2O3 phase simultaneously. The strategy to simply increase the
content of Bi2O3 in the raw material is not feasible due to its
high Bi2O3 volatilization16 during sintering and negative effect
on the electrical properties of ZnO varistors. Takemura et al.15
confirmed that a slight increase in the bismuth content of the
raw material would have a detrimental effect on the nonlinearity
exponent and the electrical stability of ZnO varistors, even if it
was annealed at 6001C for 2h.
L. M. Levinson—contributing editor
In this paper, a novel immersion process is proposed. This
process involves increasing the bismuth content of the presin-
tered ZnO varistors via a postprocessing and control of the
Bi2O3 phase transition by heat treatment. The influence of this
process on the microstructure and the electrical property of ZnO
varistors will be investigated, and a possible mechanism for the
improvement of these properties will also be discussed.
Manuscript No. 28510. Received August 25, 2010; approved January 17, 2011.
This work was supported by the NSFC of China (No. 50675218), the Innovation Project
of Shanghai Municipality (No. 07XI-023), and the 973 Project of China (No.
2009CB623305).
wAuthor to whom correspondence should be addressed. e-mail: grli@sunm.shcnc.ac.cn
2939

2940
Journal of the American Ceramic Society—Meng et al.
Vol. 94, No. 9
II. Experimental Procedure
The average grain size (d) was determined using the linear inter-
cept method.17
(1) Preparation of Samples
The current–voltage (I–V) characteristics of the samples were
recorded using an Advantest R8240 Digital Electrometer
(Suzhou, China) with a dc source at room temperature. The
breakdown electrical field (E1 mA) was measured at a current
density of 1.0 mA/cm2 and the leakage current density (JL) was
measured under 0.75 E1 mA (75% of the breakdown electrical
field) at room temperature. In addition, the nonlinear exponent
(a) is defined by the following equation:
A commercial formula was adopted to fabricate ZnO varistor
samples using a conventional ceramic manufacturing process.
The raw materials consisted of 93.7 mol% ZnO, 0.7 mol%
Bi2O3, 1.0 mol% Sb2O3, 1.0 mol% Cr2O3, 0.5 mol% Co2O3,
0.5 mol% MnCO3, and other additives including SiO2, NiCO3,
B2O3, and Al(NO3)3 ꢀ 9H2O. The raw materials were mixed by
ball milling them in deionized water, dried and pressed into disks
(roughly 20 mm in diameter and 2 mm in thickness) with
polyvinyl alcohol as a binder. After sintering at 11801C for
1 h, the pellets were lapped to 16.8 mm in diameter and 1.0 mm
in thickness. These bismuth-containing samples were designated
as the BC samples. All of the samples used in this paper are
listed in Table I. Another batch of samples was prepared with all
the ingredients above, except Bi2O3, to observe the thermal
diffusion of the bismuth during the heat treatment, which will
be discussed in Section III(1). These bismuth-free samples are
referred to as the BF samples.
The immersion process was carried out as follows: the BC
and BF samples were immersed in bismuth nitrate solutions of
different concentrations (1.0, 2.0, and 3.0M) at 501C for 3 h. The
solution was held in a 2 L beaker, which was placed in a silicone
oil-bath pot. A sample holder was used to prevent the immersed
samples’ surfaces from making contact with the bottom of the
container. The BC samples immersed in the 1.0–3.0M solutions
were designated as the BCIX (X 5 1–3) samples. The bismuth
nitrate solution was prepared by dissolving the Bi(NO3)3 ꢀ 5H2O
in ethylene glycol. After the immersion, the samples were im-
mediately oven dried at 1501C for 1 h. The samples were then
heat treated in air at 8501C for 2 h and then furnace cooled
(before this, heat treatments at temperatures from 7001C to
9501C with an interval of 501C were performed, and the heat
treatment at 8501C for 2 h was demonstrated to show the best dc
stability). A batch of samples without immersion was also heat
treated at the same time as the immersed samples in order to
investigate whether the immersion had an effect or whether it
was solely the heat treatment. These samples are referred to as
the BCH samples.
log J2 ꢁ log J1
a ¼
(1)
log E2 ꢁ log E1
where E1 and E2 are the electric fields corresponding to J1 5 0.1
mA/cm2 and J2 5 1 mA/cm2.
The dc accelerated degradation tests were carried out in three
stages: 0.85 E1 mA/130711C/14 h in the first stage, 0.90 E1 mA
/
130711C/14 h in the second stage, and 0.95 E1 mA/150711C/14 h
in the third stage. The variation of the leakage current with time
under each degradation condition was measured at an interval
of 1 min. The degradation rate coefficient KT is calculated
according to the equation below18
IL ¼ IL þ KTt1=2
(2)
0
where IL is the leakage current of the sample at time t and IL is
the initial leakage current at time t 5 0.
0
The capacitance–voltage (C–V) characteristics of the samples
were measured before and after the dc degradation tests at room
temperature. The measurements were performed using a preci-
sion RF Impedance Analyzer (Agilent 4294A, Hewlett-Packard,
Santa Clara, CA) at 1 kHz using a dc bias varying from 0 to
40 V. According to the Schottky grain-boundaries barrier
model, the related microparameters can be derived from the
C–V characteristics based on the following equations19:
ꢀ
ꢁ
2
ꢂ
ꢃ
1
1
2
ꢁ
¼
fb þ Vg
(3)
(4)
Cb 2Cb0
qesNd
The densities of the BC, BF, and BCIX samples were deter-
mined using the conventional Archimedes method. Silver paste
was coated on both faces of the samples and the silver electrodes
were formed after heating at 5001C for 1h. The area of the elec-
trodes was approximately 1.77 cm2.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2eNdfb
Ns ¼
q
where Cb is the capacitance per unit area of a grain boundary,
Cb0 the value of Cb when Vg 5 0, q the electron charge, es the
permittivity of ZnO(es 5 8.5e0), fb is the barrier height, the Vg is
the applied voltage per grain boundary, Nd is the donor con-
centration in the depletion layer, and Ns is the density of the
interface states at the grain boundary.
(2) Measurement
The crystal phases were identified by X-ray diffraction (XRD,
Bruker AXS D8 Advance, Frankfurt, Germany) using CuKa ra-
diation generated at 40 kV–40 mA. The microstructures and the
compositions of the ZnO varistor samples were determined using
an Electron Probe Micro Analyzer (EPMA, JXA-8100, JEOL,
Tokyo, Japan) in the back-scattered mode. For the examination,
the surfaces and cross-sections of the BC, BF, and BCIX samples
were ground and polished to a mirror-like surface, and then
chemically etched in a 10M NaOH solution for about 3 min.
III. Results and Discussion
(1) Microstructure
Figure 1 shows the EPMA micrographs of the BC and BCIX
samples in the back-scattered mode. All of the samples show
four typical phases of ZnO varistors: ZnO, Zn7Sb2O12, Zn2SiO4,
and the Bi-rich intergranular phase. The fine spinel particles (B
in Fig. 1(b)) locate around the ZnO grains. The dark Zn2SiO4 (C
in Fig. 1(b)) scatters sporadically among ZnO grains. The white
Bi-rich intergranular phase (D in Fig. 1(b)) distributes at mul-
tigrain junctions. The BC sample shows a small and nonuniform
distribution of the Bi-rich phase at the grain boundary. Com-
pared with the BC sample, the amount of Bi-rich intergranular
phase for BCIX samples apparently increases and the Bi-rich
intergranular phase distributes more homogeneously at multi-
grain junctions. It seems that the amount of the Bi-rich inter-
granular phase increases with increasing concentration of the
immersion solution. The average grain size and the density of
the specimens are shown in Table II. Neither the heat treatment
nor the immersion processes changed the grain size and the
Table I. The Varistor Samples Prepared by Different
Processes in this Paper
Samples
Process
BC
BF
Sintered with all the ingredients
Sintered with all the ingredients except Bi2O3;
immersed in 2.0M bismuth nitrate solution;
8501C heat treated
BCH
Sintered with all the ingredients; 8501C heat
treated
BCIX (X 5 1, Sintered with all the ingredients; immersed in
2, 3)
1.0-, 2.0-, and 3.0M bismuth nitrate solution,
respectively; 8501C heat treated

September 2011
Enhancement of the Electrical Stability of ZnO Varistors
2941
Fig. 1. Electron Probe Micro Analyzer micrographs of (a) BC, (b) BCI1, (c) BCI2, (d) BCI3 samples observed in the back-scattered mode. The ZnO,
Zn7Sb2O12, Zn2SiO4, and Bi-rich intergranular phases are marked by A, B, C, and D, respectively, in (b).
density significantly, with the density of the samples being about
95% of the theoretical density (ZnO, 5.68 g/cm3).
BF sample is dried immediately after the immersion, a layer of a
certain substance adheres to the surface of the sample with a
thickness of about 20 mm; according to the EDS analysis, it
should be the bismuth-based compound. A visual examination
showed that the two parallel surfaces of the immersed samples
are uniformly covered by a layer of yellow precipitation, most
likely bismuth subnitrate, when taken from the solution. This
layer was formed by the hydrolysis of the bismuth nitrate with
the crystal water lost from Bi(NO3)3 ꢀ 5H2O, in accordance with
Four different phases were also found in the XRD patterns,
as illustrated in Fig. 2, which is in agreement with the EPMA
analysis. The Bi38CrO66 (JCPDS Card No. 36–1238) refers to
the cubic g-Bi2O3 phase with some Cr, Zn, Sb, and Ni present,
as revealed by the EDS analysis (not shown here). Note that the
g-Bi2O3 is adopted to represent the Bi38CrO66 for the discussion
in the remainder of the text. The diffraction peaks were nor-
malized by dividing by the strongest (101) peak of the ZnO for
comparison. Firstly, the peak of the g-Bi2O3 is detected, while
the peak of the b-Bi2O3 still exists for the BCH and BCIX sam-
ples. This suggests that the phase transition from b-Bi2O3 to
g-Bi2O3 indeed occurred during the heat treatment, and some of
the b-Bi2O3 was retained to room temperature when cooled
down. Most importantly, the g-Bi2O3 peak of the BCIX samples
becomes stronger than that of the BCH sample. The BCI2 sam-
ple has the strongest peak of g-Bi2O3.
As the inherently contained bismuth leads to a difficulty in
directly observing the bismuth diffusion for BCIX samples, it is
reasonable to observe that of BF samples instead due to the their
similarity of density. Figure 3 shows the cross-section micro-
graph of one BF sample. It is clear (see Fig. 3(a)) that when the
Table II. Average Grain Size and Density of Different
Samples
Samples
BC
BF
BCI1
BCI2
BCI3
d (mm)
10.0
5.49
9.91
5.43
9.81
9.89
5.42
10.2
NM
r (g/cm3)
NMw
Fig. 2. XRD patterns of (a) BC, (b) BCH, (c) BCI1, (d) BCI2, (e) BCI3
samples.
wNot measured.

2942
Journal of the American Ceramic Society—Meng et al.
Vol. 94, No. 9
(a)
(b)
(c)
A
B
C
Fig. 3. The back-scattered Electron Probe Micro Analyzer micrograph (left) and corresponding point EDS analysis (right) for the cross-section of one
BF sample (a) the natural cross-section just dried after the immersion (b) the natural cross-section after the heat treatment (c) one interior part of the
polished and etched cross-section after heat treatment.
Eq.(5). The precipitation preferentially forms on the solid
surface where the energy of nucleation is low, according to the
heterogeneous nucleation theory of chemical precipitation. The
nitric acid formed will dissolve a little of the surface substance of
the varistors to make the grain boundaries or pores more
‘‘open,’’ which might contribute to the penetration of molten
Bi2O3 during the heat treatment. With a higher concentration,
the thickness of bismuth subnitrate precipitation, i.e., the quan-
tity of the diffusion source will increase and more of the nitric
acid will form. An optimum concentration of immersion solu-
tion will be helpful, while an excessive concentration would
cause compositional gradient and be detrimental to the varistor
sample due to severe dissolution of the varistor component.
in Fig. 4. For the BC sample (Fig. 4(a)), a scarce and nonuni-
form distribution of a white Bi-rich phase is found in the region
of 0–300 mm away from the surface. By contrast, the BCIX
samples show a uniform and rich distribution of the Bi-rich
phase in the near-surface region, except the BCI3 sample. Con-
siderable accumulation of the Bi-rich can be are observed in the
surface of BCI3 sample. It is most likely that the vaporization of
the Bi-rich phase during sintering leads to the absence of the Bi-
rich phase in the near-surface region for the BC sample. For the
BCIX samples, the immersion process compensates the near-
surface bismuth loss by the diffusion of molten Bi2O3. However,
an excessive diffusion source, originating from a high immersion
concentration, might cause a problem of a compositional gra-
dient from the surface to the interior as a BCI3 sample.
It is reasonable to conclude that during heat treatment, the
molten Bi2O3, the combination of extraneously penetrated mol-
ten Bi2O3 with an inherently molten Bi-rich phase, will induce
the reconstruction of the grain-boundary layer. On the one
hand, the molten Bi2O3 would firstly dissolve the grain-bound-
ary layer with several phases present. The mass transportation
of molten Bi2O3, that is, dissolution–reprecipitation process,
makes the distribution of grain-boundary composition more si-
multaneous. On the other hand, the molten Bi2O3 will dissolve
some additive elements, especially Cr, and ultimately forms the
Bi-rich intergranular phase (i.e., Bi38CrO66 phase) when cooled
down.
BiðNO3Þ3þH2O ¼ BiONO3 # þ2HNO3
(5)
Figure 3(b) indicates an effective bismuth diffusion process
with a diffusion depth of about 200 mm from the surface into the
interior. The ultimately formed white Bi-rich phase distributes
uniformly at the multigrain junctions, as shown by Fig. 3(c).
Based on these observations, the bismuth diffusion process
should be reasonably postulated as follows: when heat treated,
the salt precipitation would firstly decompose into Bi2O3 below
5001C, and then the Bi2O3 will melt into a liquid phase at about
its melting point (B8251C). The molten Bi2O3 readily penetrates
down into the open channels of the ZnO varistor matrix, such as
open pores, grain boundaries, and triple points of ZnO grain.
The molten Bi2O3 could further form a continuous Bi2O3 liquid
film along the grain boundary and wet the ZnO grains. It is most
probably this diffusion process that leads to the increased
amount of the g-Bi2O3 for the BCIX samples, as described by
the XRD pattern. As reported by the Barsoum et al.,20 this
penetration is not thermodynamically, but kinetically con-
trolled. The molten Bi2O3 will penetrate ZnO grain boundaries
at a rate that depends on the grain-boundary composition. The
change of the XRD pattern in the present paper confirms that
certain chemical reaction or dissolution definitely occurs be-
tween the molten Bi2O3 and the solid grain-boundaries’ phase.
The near-surface phase distribution of BC and BCIX samples
is shown by the polished and etched cross-section micrographs
As shown by Fig. 3(b), the extraneous penetrated molten
Bi2O3 could only reach a finite depth (about 1/5 of the total
thickness) near the surface in this paper; in other words, it could
only improve the distribution of the Bi-rich phase in the regions
approaching the surface. However, it is exactly the near surface
modification that compensates bismuth loss due to vaporization
in sintering and makes more grain boundaries decorated by
Bi-rich phase.
(2) The Electrical Characteristics
(A) The I–V Characteristics: Figure 5 shows the I–V
characteristics of the different varistor samples (in terms of the
electrical field–current density, E–J). The BCIX samples exhibit

September 2011
Enhancement of the Electrical Stability of ZnO Varistors
2943
Fig. 4. The polished and etched cross-section Electron Probe Micro Analyzer micrographs of (a) BC (b) BCI1 (c) BCI2 (d) BCI3 samples in the back-
scattered mode.
a flatter nonlinear curve between 10ꢁ5 and 10ꢁ3A/cm2 than the
BC samples. The nonlinearity exponent (a), the breakdown field
(E1 mA), and the leakage current density (JL) are listed in Table
III. The results show that the a of all the immersed samples ob-
viously increases, although the E1 mA declines slightly compared
with the BC samples. This indicates that the nonlinearity expo-
nent can be effectively improved by the immersion process. The
BCH samples also show an increased a and a slightly changed
E1 mA. The JL shows little change among the BC, BCH, and
immersed samples.
ment, the temperature of which (8501C) is much higher than the
commonly reported 6001–7001C. As illustrated in Section III(1),
this heat treatment at B8501C facilitates the reconstruction of
the intergranular layer, which leads to a homogeneous Bi-rich
phase distribution, as shown by Fig. 1, and the modification of
the defect chemistry in the vicinity of the intergranular layer.14
Dai and colleagues21 and Inada22 confirmed that the nonlinearity
can be improved by a heat treatment at 8001–9001C.
(B) The dc Degradation Tests: Figure 6 shows the time
dependence of the leakage current for the samples subjected to
three different degradation conditions. In the first stage, the
leakage current of all the BCIX samples increases at the very
beginning and then decreases slowly, while the leakage current
of the BC and BCH samples increases with time. The uptrend
for the BCH sample is less than the BC sample. In the second
stage, the leakage current of BCH sample increases sharply with
increasing time as in the BC sample, while the BCIX samples still
show a stable the leakage current. In the third stage, the vari-
ation of leakage current exhibits a trend similar to the first one.
Table IV summarizes the degradation rate coefficient KT
of different samples in each stage. The variation of a for the
Considering the similarity of the values for the BCIX and
BCH samples, the enhanced a could be related to the heat treat-
Table III. Electrical Characteristics of Different Varistor
Samples
Samples
a
E1 mA (V/cm)
JL (mA/cm2)
BC
38.1
42.6
42.6
42.2
42.5
3610
3740
3460
3440
3460
0.9
BCH
BCI1
BCI2
BCI3
1.02
1.33
1.48
1.35
Fig. 5. The I–V characteristics of ZnO varistors samples used in the
present study (in terms of electrical field–current density, E–J).

2944
Journal of the American Ceramic Society—Meng et al.
Vol. 94, No. 9
(b)
(c)
(a)
Fig. 6. The dc degradation behavior of ZnO varistor samples under (a) the first stage: 0.85 E1 mA/130711C/14 h (b) the second stage: 0.90 E1 mA
/
130711C/14 h (c) the third stage: 0.95 E1 mA/150711C/14 h.
samples through the dc degradation tests is shown in Fig. 7. All
of the BCIX samples show a uniquely negative KT, i.e. a reduced
leakage current, and a stable nonlinearity through the three degra-
dation stages, while the BC and BCH samples exhibit a continu-
ously increased leakage current and a reduced nonlinear exponent.
These sufficiently illustrate that the immersion process can effec-
tively improve the dc electrical stability of the ZnO varistor and its
prominent advantage over simple heat treatment at 8501C.
(C) The C–V Characteristics: The C–V characteristics’
measurement was performed to investigate the possible mecha-
nism for the improvement of both the electrical stability and
nonlinearity by the immersion process. The calculated micropa-
rameter values (fb, Nd, and Ns) before the dc degradation tests
and their deviation after the first dc degradation stage are pre-
sented in Table V. The fb of 41.5 eV in present paper does not
match the one of o1 eV determined by current–temperature
characteristics.12,23 Ronfard-Haret24 reported that the voltage
dependence of the contact capacitance can lead to unrealistic fb
values. Alim et al.25 mentioned the mismatching barrier height
determined from C–V and I–T measurements and illustrated
that the C–V measurement yielded a frequency-dependent built-
in potential and electrical barrier height, etc. in device-related
parameters. However, the C–V analysis is the commonly
adopted method to analyze the microparameters and fb values
of 41.5 eV using the C–V method have been reported in some
published literatures.14,18,23 These confirm that a C–V measure-
ment is feasible and meaningful to analyze the relative change
and variation trend of microparameters.
It is most probably the slightly enhanced barrier that leads to
a decreasing leakage current with time and a stable nonlinearity
for the BCIX samples in the dc degradation test. Because of high
activation energy, the oxygen vacancies remain as the main do-
nor in the ZnO grains, with Zinc interstitials preferentially
diffused out after heat treatment. Based on the oxygen-desorp-
tion theory,27,28 the oxygen vacancies can also migrate toward
the grain boundaries under a continuous dc field and react with
the adsorbed oxygen located at the grain interface, as shown by
the following equation:
VꢂO þ Oa0 d ¼ VOꢃ þ 1=2O2 " ðgÞ
(7)
However, the abundant g-Bi2O3 at the intergranular layer can
prevent the oxygen from diffusing out and consequently restrain
the chemical reaction above. Eda et al.29 reported that the g-
Bi2O3 has a lower oxygen-ion mobility than the b-Bi2O3. Wang
and Li21mentioned that the g-Bi2O3 has a smaller interstitial
volume and a more compact structure than the b-Bi2O3. Thus, it
is more difficult for oxygen to migrate in theg-Bi2O3 than in the
b-Bi2O3. This might result in the accumulation of oxygen
around the grain interface and the oxygen can capture an elec-
tron to form adsorbed oxygen (see Eqs. (8) and (9)) due to its
high electron affinity.3,13,28 As a result, the Nd and Ns values of
the BCIX samples are not reduced, with a stable barrier formed
1=2O2ðgÞ þ VꢃO ¼ OꢃO
OꢃO þ e0 ¼ O0ad
(8)
(9)
Before the degradation tests, compared with the BC samples,
the fb, Nd, and Ns of the BCH and the BCIX samples decline.
This could be related to the Zinc interstitials’ migration. Ac-
cording to the ion-migration theory proposed by Gupta and
Carkson,26 the zinc interstitials in the deletion layer can be
diffused out via a heat treatment and permanently removed by
reacting with the native zinc vacancies at the grain interfaces, as
shown by Eq. (6). This process causes a reduction in the Nd and
Ns values, which finally leads to a decline in the barrier height
The less g-Bi2O3 amount of the BCH samples leads to
inadequate restraining against oxygen-ion migration, resulting
in their slightly decreased parameters after degradation. The
drastic decline of the parameters after degradation for the BC
samples results from the migration and removal of zinc intersti-
tials under a continuously dc electrical field as shown by Eq. (6).
With the analysis of microstructure and electrical perfor-
mances, although the near-surface nonuniform distribution of
Znꢂi þ V0Zn ¼ Znꢃi þ VꢃZn
(6)
After the degradation tests, compared with the values before
the tests, the fb, Nd, and Ns of the BCIX samples increase and
the BCI2 samples show the maximum increment, while the ones
of the BC and BCH samples decrease obviously.
Table V. The Microparameters of Varistor Samples Before dc
Degradation Test and their Deviation after the First dc
Degradation Stage
fb (eV)
Nd (1017cmꢁ3
)
Ns (1012cmꢁ2
)
Table IV. Degradation Rate Coefficients (KT) of Different
Samples in each dc Degradation Stage
Samples Before After (Dfb) Before After (DNd) Before After (DNs)
BC
BCH
BCI1 1.66
BCI2 1.75
BCI3 1.55
2.97
2.30
ꢁ1.85
ꢁ0.16
0
10.17
10.12
2.89
2.87
2.85
3.04
2.64
ꢁ1.37
ꢁ0.09
10.10
10.42
10.12
5.03
4.42
3.74
3.96
3.48
ꢁ2.79
ꢁ0.22
10.06
10.46
10.21
Samples
BC
BCH
BCI1
BCI2
BCI3
First stage (mA/h1/2
)
60.0 23.3 ꢁ3.19 ꢁ3.36 ꢁ2.30
50.0 20.0 ꢁ2.67 ꢁ1.73 ꢁ0.20
510 80 ꢁ8.28 ꢁ14.32 ꢁ10
Second stage (mA/h1/2
)
Third stage (mA/h1/2
)

September 2011
Enhancement of the Electrical Stability of ZnO Varistors
2945
defect structure and the component distribution in the inter-
granular region, which is caused by the reconstruction of the
intergranular layer.
References
1P. R. Emtage, ‘‘The Physicals of Zinc Oxide Varistors,’’ J. Appl. Phys., 48 [10]
4372–84 (1977).
2K. Eda, ‘‘Conduction Mechanism of Non-Ohmic Zinc-Oxide Ceramics,’’
J. Appl. Phys., 49 [5] 2964–72 (1977).
3T. K. Gupta, ‘‘Application of Znic Oxide Varistors,’’ J. Am. Ceram. Soc., 73 [7]
1817–40 (1990).
4W. Onreabroy, N. Sirikulrat, A. P. Brown, C. Hammond, and S. J. Milne,
‘‘Properties and Intergranular Phase Analysis of a ZnO–CoO–Bi2O3 Varistor,’’
Solid State Ionics, 177, 411–20 (2006).
5D. Dey and R. C. Bradt, ‘‘Grain Growth of ZnO during Bi2O3 Liquid-Phase
Sintering,’’ J. Am. Ceram. Soc., 75 [9] 2529–34 (1992).
6R. A. Winston and J. F. Cordaro, ‘‘Grain-Boundary Interface Electron Traps
in Commercial Zinc Oxide Varistors,’’ J. Appl. Phys., 68 [12] 6495–650 (1990).
7M. Takada, H. Yoshino, and S. Yoshikado, ‘‘Relationship between Grain
Boundaries and Electrical Degradation of ZnO Varistors,’’ Electr. Eng. Jpn., 159
[3] 1–9 (2007).
Fig. 7. The variation of a for the samples through the dc degradation
tests.
8C. G. Shirley and W. M. Paulson, ‘‘The Pulse-Degradation Characteristic of
ZnO Varistors,’’ J. Appl. Phys., 50 [9] 5782–89 (1979).
9J. Fan and R. Freer, ‘‘The Electrical Properies and d.c. Degradation Char-
aceristics of Silver Doped ZnO Varistors,’’ J. Mater. Sci., 28, 1391–5 (1993).
10J.-M. Wu and J.-J. Shyu, ‘‘Electrical Properties and Degradation Improve-
ment of ZnO Varistors Doped with PbO–B2O3 Based Glasses,’’ J. Mater. Sci., 24,
1881–88 (1989).
the Bi-rich phase is observed in the sample with a higher immer-
sion concentration, the negative effect of the compositional
gradient on the overall electrical performances has not yet been
found. All of BCIX samples show good dc stability and similar
electrical behavior; however, considering the possibly negative
influence of nonhomogeneous component distribution, the opti-
mum concentration of immersion solution was determined to be
1.0–2.0M. The effect of compositional gradient on the perfor-
mance of thick ZnO varistors samples is not clearly understood
and needs further investigated. Nevertheless, it needs to be em-
phasized that the key feature of the immersion process is the
near-surface compensation of Bi2O3 with more grain boundaries
having the g-Bi2O3 phase.
11M. Takada and S. Yoshikado, ‘‘Effect of Thermal Annealing on Electrical
Degradation Characteristics of Sb–Bi–Mn–Co-added ZnO Varistors,’’ J. Eur.
Ceram. Soc., 30, 531–38 (2010).
12T. K. Gupta and W. D. Straub, ‘‘Effect of Annealing on the ac Leakage
Components of the ZnO Varistor.I. Resistive Current,’’ J. Appl. Phys., 68 [2] 845–
50 (1990).
13A. Iga, M. Matsuoka, and T. Masuyama, ‘‘Effect of Heat-Treatment on Cur-
rent Creep Phenomena in Nonohmic ZnO Ceramics,’’ Jpn. J. Appl. Phys., 15 [9]
1847–48 (1976).
14C. S. Chen, ‘‘Improvement on the Degradation of Microwave Sintered ZnO
Varistors by Postannealing,’’ J. Mater. Res., 13 [6] 1560–67 (1998).
15T. Takemura, M. Kobayashi, Y. Takada, and K. Sato, ‘‘Effects of Bismuth
Sesquioxide on the Characteristics of ZnO Varistors,’’ J. Am. Ceram. Soc., 69 [5]
430–36 (1986).
16M. Peiteado, M. A. D. L. Rubia, M. J. Velasco, F. J. Valle, and A. C.
Caballero, ‘‘Bi2O3 Vaporization from ZnO-Based Varistors,’’ J. Eur. Ceram. Soc.,
25, 1675–80 (2005).
IV. Summary and Conclusions
17M. I. Mendelson, ‘‘Average Grain Size in Polycrystalline Ceramics,’’ J. Am.
Ceram. Soc., 52, 443–6 (1969).
ZnO varistors with an improved dc electrical stability were ob-
tained by a novel immersion process. All the immersed samples
showed a reduced leakage current with time (KT 5 ꢁ2.3 to ꢁ3.4
mA/h1/2) under three continuously accelerated dc degradation
conditions. At the same time, the enhanced nonlinearity expo-
nent (a 5 B43) of the varistors showed good stability in the dc
degradation test. The sole heat treatment at 8501C shows an
unsatisfactory dc degradation characteristic, although it can
modify the nonlinearity.
The immersion process uniquely produces a layer of bismuth
subnitrate precipitation adhering to the surfaces of the ZnO va-
ristor samples and subsequently an effective penetration of mol-
ten Bi2O3 from the surface into the interior. The penetration of
molten Bi2O3 compensates the near-surface absence of Bi2O3
resulting from its vaporization during sintering. During the heat
treatment, the extraneously penetrated molten Bi2O3 plus inher-
ently molten Bi-rich phase leads to the reconstruction of an
intergranular layer and fills more grain boundaries with the
g-Bi2O3 phase.
18J. Fan and R. Freer, ‘‘The Role Played by Ag and Al Dopants in Control-
ling the Electrical Properties of ZnO Varistors,’’ J. Appl. Phys., 77 [9] 4795–800
(1995).
19K. Mukae, K. Tsuda, and I. Nagasawa, ‘‘Capacitance-vs-Voltage Character-
istics of ZnO Varistors,’’ J. Appl. Phys., 50 [6] 4475–76 (1979).
20M. W. Barsoum, A. Elkind, and F. A. Selim, ‘‘Low Breakdown Voltage
Varistors by Grain Boundary Diffusion of Molten Bi2O3 in Zno,’’ J. Am. Ceram.
Soc., 79 [4] 962–66 (1996).
21Z. L. Wang and S. T. Li, The Manufacturing and Application of ZnO Varistors
Ceramics, pp 160–88. Science Press, Beijing, 2009.
22M. Inada, ‘‘Effects of Heat-Treatment on Crystal Phases, Microstructure, and
Electrical Properties of Nonohmic Zinc Oxide Ceramics,’’ Jpn. J. Appl. Phys., 10
[8] 1439–46 (1979).
23S. N. Bai and T. Y. Tseng, ‘‘Influence of Sintering Temperature on Electrical
Properties of ZnO Varistors,’’ J. Appl. Phys., 74 [1] 695–703 (1993).
24J. C. Ronfard-Haret, ‘‘Electric and Luminescent Properties of ZnO-Based
Ceramics Containing Small Amounts of Er and Mn Oxide,’’ J. Lumin., 104, 1–12
(2003).
25M. A. Alim, S. Li, F. Liu, and P. Cheng, ‘‘Electrica Barriers in the ZnO Va-
ristor Grain Boundaries,’’ Phys. Stat. Sol. (A), 203 [2] 410–27 (2006).
26T. K. Gupta and W. G. Carkson, ‘‘A Grain-Boundary Defect Model for In-
stability/Stability of a ZnO Varistor,’’ J. Mater. Sci., 20, 3487–500 (1985).
27E. R. Leite, J. A. Varela, and E. longo, ‘‘A New Interpretation for the De-
gradation Phenomenon of ZnO Varistors,’’ J. Mater. Sci., 27, 5325–29 (1992).
28S. L. Jiang, T. T. Xie, H. Yu, Y. Q. Huang, and H. B. Zhang, ‘‘Studies on the
Degradation Properties of the ZnO-Based Ceramic Film Varistors,’’ Microelec-
tron. Eng., 85, 371–74 (2008).
The abundant g-Bi2O3 at the grain boundary contributes to
the formation of a stable intergranular barrier by effectively in-
hibiting the oxygen desorption. This is critical to enhance the
electrical stability under a dc electrical field. The improvement
in the nonlinearity can be attributed to the modification of the
29K. Eda, A. Iga, and M. Matsuoka, ‘‘Degradation Mechanism of Non-Ohmic
Zinc Oxide Ceramic,’’ J. Appl. Phys., 51 [5] 2678–84 (1980).
&