See discussions, stats, and author profiles for this publication at: [Link]

net/publication/279896134

ZINC OXIDE VARISTORS - A REVIEW.

Article in American Ceramic Society Bulletin · April 1986

CITATIONS READS
360 6,120

2 authors, including:

Lionel Levinson
Eckerd College
65 PUBLICATIONS 3,281 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

ONR Advanced Interconnect View project

All content following this page was uploaded by Lionel Levinson on 08 May 2018.

The user has requested enhancement of the downloaded file.

ZnO Varistor Technology

Lionel M. Levinson, Vartek Associates LLC, Schenectady, New York

I. INTRODUCTION

Metal oxide varistors are novel ZnO-based ceramic semiconductor devices with highly nonlinear
current-voltage characteristics similar to back-to-back Zener diodes. Over the last 30 years these
varistor devices have become the preferred approach to protecting electronic, electrical and power
distribution and transmission circuits from the destructive voltage levels induced by lightning
impulses or switching surges.

The varistors are produced by a ceramic sintering process that provides a structure largely
comprised of conductive ZnO grains surrounded by thin electrically insulating barriers. In this
chapter the physical properties of the resulting three-dimensional series-parallel junction network are
discussed, and the novel electrical properties are interpreted in terms of the material microstructure
and grain-to-grain conduction mechanism.

The applications of ZnO varistors are predominantly in the field of circuit overvoltage
protection, although some uses of the material as an active circuit element have been described [1,2].
Overvoltage protection is a necessity for both electronic circuits and in the electrical power
distribution and transmission industries. As a specific example of the need for this type of device one
need only consider the fact that circuits and systems containing low-cost, high-reliability solid-state
components are typically designed for use on 120-V ac power lines. It is generally accepted,
however, that proper design practices must take into account the fact that solid-state components
cannot, in general, withstand the amount of overvoltage imposed on the circuitry by typical power
system transients. This drives the requirement for voltage surge protection.

The magnitude of this type of problem is graphically illustrated in Table 1, which gives surge data
from a survey on the nature of surge voltage transients in residential power circuits [3]. The data,
collected from a variety of homes in urban and rural locations in New York, Florida, and South
Carolina, indicate that surges in the 1000-V range, and above, are prevalent with a frequency of the
order of several surges per day. Such transients usually result from switching surges or lightning
storms and can play havoc with unprotected solid-state components connected to a power line.
Product reliability therefore requires that the solid-state component be provided with adequate
protection from these transient overvoltages.

In this chapter the status of ZnO varistors is assessed. The ZnO varistor, often called a “metal oxide
varistor,” was initially developed in Japan [5] and became available in the United States in the early
1970’s under the trade name GE-MOV varistor largely for protection of consumer and industrial
equipment designed to work in the voltage range below 1000 volts. As an interesting aside it may be
noted that work on the non-linear electrical properties of ZnO ceramics had been underway in Russia
in the late 1950’s [6,7] but the application potential of the these materials was first recognized by
Matsuoka and his co-workers in Japan [5] about ten years later. An intensive effort was then
launched to develop the understanding and application capabilities of ZnO varistor materials and this
established the basis for the use of this material as the predominant overvoltage surge protective
device.

Varistor ceramics and their applications have been the subject of numerous reviews and the reader
wishing to obtain more detailed information is referred to these quite detailed discussions [8-13].
The most recent review by Clarke [13] provides a thorough description of our current state of
understanding. Gupta’s review [10] provides a discussion of the criteria important to the use of
varistors in high voltage power distribution and transmission applications

This chapter is structured as follows: After a description of the electrical characteristics of the
varistor, the general microstructure and fabrication of the devices are discussed. Varistors have an
interesting and rich behavior as circuit elements, and the effect of frequency and voltage on the
varistor impedance is briefly described. A review of our current understanding of the mechanism of
the highly nonlinear current-voltage characteristic of ZnO ceramic varistors follows. This
understanding requires that we address both the behavior of a single grain-grain junction and the
complexities introduced by variability in the grain structure. Both these issues are discussed.

As with any electronic device, usage in the real world requires that one take into account
possible degradation and failure effects and we material related to these issues is presented. Finally,
some illustrative examples of ZnO applications are featured.

II. VARISTOR ELECTRICAL CHARACTERISTICS

The electrical characteristics of ZnO varistors are similar to back-to-back Zener diodes but with
much greater voltage-, current-, and energy-handling capabilities. Figure 1 shows an oscillograph of
the symmetrical current-voltage characteristics of this material. The device can be regarded as
“insulating” up to a certain field, the breakdown field FBR, above which it becomes conducting.

Table 1 Surge Voltages Observed in a Variety of Urban and Rural Homes in New York, Florida,
and South Carolina

Location Most Duration (µsec Average Highest Most severe

number frequent or cycles) Number of Crest (V) surge
surge crest surges per
(V) hour
duration
(V) (V) (µsec or
cycles)
1 300 10 µsec 0.07 700 10 µsec
2 500 20 µsec 0.14 750 20 µsec
3 300 1 cycle 0.05 600 1 cycle
4 300 2 cycles 0.2 400 2 cycles
5 250 1 cycle 0.01 400 1 cycle
6 800 1 cycle 0.03 1800 1 cycle
7 300 10 µsec 0.1 1200 4 cycles
8 1500 1 cycle 0.2 1500 1 cycle
9 2000 1 cycle 0.4 2500 1 cycle
10 1500 1 cycle 0.15 1500 1 cycle
11 1400 1 cycle 0.06 1700 1 cycle
12 600 3 cycles 0.05 800 3 cycles
13 200 15 µsec 0.1 400 30 µsec
14 1000 4 cycles 0.1 5600 1 cycle
Source: Ref. 3.

A more complete examination of the device behavior requires the display of current and voltage
on a logarithmic scale. Figure 2 gives current I versus applied voltage V for a typical varistor
measured at 77K and for a range of temperatures near 300K. Note that there is an enormous
variation in current (a factor of 1011), while the applied voltage varies about a factor of 3. From
Figure 2 it is also evident that there is wide range of current (10–6 to ≈102 Amp) over which the log
current-voltage relation is approximately linear; that is, we can write
a
J 1  F1 
=  (1)
J 2  F2 

where J is the current density passing through the varistor, α is the measure of non-linearity, and
F is the field applied to the device.

Equation (1) is the so-called empirical varistor power law equation. If α = 1, we would have an
ohmic device. As α→ ∞ , we have a perfect varistor (i.e., the current density varies infinitely for
small changes in the applied field). The exponent α is 25 to 50 or more in a typical ZnO varistor.
Actual values of α, breakdown field, and off-state resistance in typical devices are available
elsewhere [8-13] and will not be specified here.
At very low and very high currents, Equation (1) is no longer a good representation of the
varistor characteristic. The low-current region has been denoted “prebreakdown” and corresponds
physically to the transport of very low level currents through the varistor at voltages below the
breakdown region. The very high current deviations from the power law Equation (1) are believed to
result, at least in part, from an ohmic series resistance derived from the ZnO grains in the varistor.
This region has been labeled “upturn” in Figure 2.
The use of ZnO varistors to protect sensitive equipment from transient overvoltages is in
principle extremely simple; the varistor is directly connected across the power line in parallel with
the load to be protected. This is indicated schematically in Figure 3. Care is taken to choose a
varistor having a breakdown voltage slightly greater than the maximum design voltage applied to the
system to be protected. In normal operation the varistor is insulating (i.e., it operates in the
prebreakdown region of Figure 2). If a transient is incident so that the total voltage applied to the
load rises above the varistor breakdown voltage, the varistor current rapidly increases along its
characteristic current-voltage curve (Figure 2), becoming a conducting shunt path for the incident
transient pulse.
The varistor response [14] to ≈ 500 psec rise time pulses is shown in Figure 4a and 4b. In Figure
4a, the applied pulse voltage is 50V and only the capacitive charging current is observed. The decay
of the varistor voltage at the end of the pulse is associated with a negative varistor discharge current
(not plotted in the figure). In Figure 4b, the pulse voltage is 150V, which is above the breakdown
voltage of the varistor. In this case, the capacitive charging current is almost hidden by the increased
conduction current and the varistor voltage shows a peak that then decreases with increasing time.
There is no indication in this curve of any delay in the initiation of the conduction current once the
varistor voltage has risen above some minimum value. Since the varistor has the ability to “clamp”
the incoming pulse, the varistor “turn on” time is less than the rise time of the pulsing equipment, ≈
5 x 10-10 sec. The initial peaking of the varistor voltage and its subsequent decrease with time has
been denoted the “overshoot” effect and represents a time-dependent modification of the varistor
conduction process that is inherently very fast.
The existence of the overshoot effect has an important device implication. The protective voltage
quoted is usually that determined using standard measuring procedures (for example, pulses of 8
µsec rise time) where the voltage overshoot, Vmax – Vdc is small. However, for faster rise time, high-
current pulses, the protective level may be appreciably poorer. The magnitude of the overshoot
varies from device to device and depends on the varistor formulation.
III. VARISTOR MICROSTRUCTURE AND FABRICATION

ZnO varistors are highly complex multicomponent oxide ceramics whose electrical behavior
depends both on the ceramic microstructure of the device and on detailed processes occurring at the
ZnO grain boundaries. The primary constituent of the varistor is ZnO, typically 80 mol% or more.
As a result of the varistor processing, the ZnO in ZnO varistors is semiconducting, with ρZnO ≤ 1 Ω-
cm at room temperature. In addition to the ZnO, the varistor contains smaller amounts of a number
of other metal oxide constituents (thus the name “metal oxide varistor"). A typical early composition
given by Matsuoka [5] contains 97 mol% ZnO, 1 mol% Sb2O3, and 1/2 mol% each of Bi2O3, CoO,
MnO, and Cr2O3. The varistor constituents are often indicated by the cations alone, and thus a
formulation containing Bi2O3 is said to contain Bi. Current commercial varistor mixes may contain
as many as eight or more metal oxide additives. The reasons for this are predicated on the empirical
approach taken to maximize varistor performance in all of the areas required for product acceptance.
For example, if the coefficient of nonlinearity or exponent, α, was the only important device
parameter, acceptably high coefficients of nonlinearity could easily be achieved in the simplest of
varistor systems comprised of a single varistor forming ingredient such as Bi or Pr and one or two
varistor performance ingredients such as Co, Mn, or Ni. However, a number of other considerations
such as stability, conduction uniformity, power dissipation, maximum surge current, energy
absorption capability, grain resistivity, and so forth, in addition to exponent, must be optimized to
obtain overall long-term reliability as well as electrical performance. Unfortunately, this concert of
requirements necessitates a proliferation of varistor ingredients. Thus, while we have developed a
reasonably precise theoretical model for varistor conduction, commercial improvements proceed
mainly by “educated” trial-and-error activity in mix selection and processing schedule development.
There are, of course, good reasons for this approach. The varistor behavior depends both on the
detailed ceramic microstructure of the device and on the related chemical and physical processes
occurring at the ZnO grain boundaries during high-temperature sintering. Most conduction models
ignore this chemical and microstructural complexity, and while they can indeed account for a wide
variety of observed varistor phenomena, they are usually not specific in the role of the individual
chemical additives. Hence they cannot be used to optimize overall performance by the a priori
selection of process and ingredients. The above notwithstanding it is useful to group various
additives to ZnO varistor devices by the predominant effect they have on the device behavior. A
non-exhaustive list is as follows:

1. Nonlinearity inducers – Bi, Pr.

2. Nonlinearity enhancers – Co, Mn, Sb, Cr
3. Grain growth retardants – Sb, Si
4. Grain growth enhancers – Ti, Al, Ba
5. Stability enhancers – Ba, B, Cr, Ag, Li, K
6. ZnO conductivity enhancers – Al, Ga, In

The above list should be treated with caution since some additives may have more than one effect
depending on other additives present. In addition, the amount of additive (or inadvertent impurity
present) can be critical. For example, additions of Li in the 10-100 ppm level can enhance stability
while larger amounts of lithium can dramatically reduce the degree of nonlinearity, especially in the
high current region (Li is known to make ZnO resistive)

Fabrication of ZnO varistors follows standard ceramic techniques. A simplified flow diagram
used in varistors for electronic component protection is given in Figure 5. The ZnO and other
constituents are mixed, for example, by milling in a ball mill. The mixed powder is dried and pressed
to the desired shape. The resulting pellets are sintered at high temperature (typically, 1100 to
1400°C). The sintered devices are then electroded, often (but not invariably) with a fired silver
contact. Leads are attached by solder and the finished device may be potted in epoxy and tested to
meet required specifications.
For larger single disk volumes—for example, the 3-in.-diameter, l ⅜ in.-thick disk used in high-
voltage arrester applications—the basic flow diagram of Figure 5 is also followed but some of the
steps are more complex. For example, the milling procedure is carried out in a series of steps to
ensure proper additive dispersal throughout the disk volume. The additives are milled separately
before addition of the ZnO. The resulting powder is mixed in a shear blender and spray dried. The
mix is then calcined at temperatures in the range 800 to 900°C to prereact the ingredients. The
calcined mix is again milled and spray dried prior to pressing. The electroding procedure is also
comprised of a series of steps since in practice the disks are stacked one upon the other for arrester
6
insertion on lines of up to 10 V or more. The disks must be lapped prior to electroding to achieve
maximum disk-disk contact. The electrodes are often plasma-sprayed aluminum for maximum disk
face conductance and to allow high-temperature operation. The disk rim may also be passivated with
an insulating glass or ceramic layer at this time. The final step is an anneal at about 600°C to
improve electrical stability. Disk assembly in the arrester or other protective device, which may have
a large number of disk columns in parallel, will not be discussed here.

The microstructure of fired varistor is depicted schematically in Figure 6. The varistor is idealized as
comprised of a set of conducting grains, size d, surrounded by an extremely thin (few nm) adsorbed
layer. Transmission electron microscopy [15-18] and Auger Electron Spectroscopy [19], indicate
that bismuth is segregated to the ZnO grain boundaries. The form of the segregation and whether it
exists as a film , as isolated segregated ions, , or perhaps in some other form has been the focus of
various high-resolution electron microscopy studies [15,18]. At this point it appears likely that the
form of segregation probably depends on the particular formulation and processing used for the
varistor device under study. Current flows between the electrodes as indicated. Typical grain sizes
are d ~ 10 µm, and the grain resistivity ρZnO is about 1 Ω-cm.

The actual microstructure of a ZnO varistor is considerably more complex than the idealized
depiction of Figure 6. In Figure 7, a photomicrograph of a polished and etched section of a typical
varistor is shown. Three phases—grains, intergranular phase, and particles—are evident. While the
microstructures of varistors exhibit considerable variation from one formulation to another, they
typically exhibit the characteristics of a ceramic prepared by liquid-phase sintering. In this case the
ceramic consists of large ZnO grains with a bismuth-rich second phase at the triple junctions [15, 20-
22].

The grains are the predominant phase in Figure 7 and consist of small (~ 20 µm) conducting ZnO
microcrystals. The straight lines evident in some of the ZnO grains correspond to twin boundaries
delineating different ZnO crystal planes having different etch properties. The whitish areas lying
largely between the ZnO phase at triple points have been labeled “intergranular material” and are
bismuth-rich. Actually, the true intergranular barrier relevant to varistor action is not visible in
Figure 7. Varistor action is controlled by depletion layers lying within the ZnO grains at grain-grain
interfaces. These potential barriers are formed at the grain boundaries when the varistor is cooled
from its sintering temperature. Barrier formation is believed to occur in the temperature range of
450°–700°C (or on subsequent annealing). Indeed heat treatment or annealing somewhere in this
temperature range is often a key process step in the manufacture of varistors for high voltage
applications. The necessary annealing step was discovered serendipitously at General Electric in the
mid-1970’s and is often an essential requirement to obtain the high degree of stability necessary for
high voltage varistor applications in the utility industry. The electronic processes occurring during
the annealing stage are presumably related to dopant or trap rearrangement at the ZnO grain
boundary. The details of these processes remain essentially unknown, however.
 In addition to the ZnO grains and intergranular material, a third phase, labeled “particles,” is
visible in Figure 7. X-ray studies [20] have shown that this phase has a spinel-type structure with the
approximate formula Zn7Sb2O12. These spinel particles are insulating and play only a secondary role
in determining device properties. Their presence is a result of the particular composition and
processing of the varistor of Figure 7. Other varistor formulations have other (or no) such secondary
phases. For example, depending on the composition, processing and amount of additives, the
microstructure also can contain particles of pyrochlore.

As will be discussed later, an electrically insulating boundary region, thickness t ~ 1000 Å, is

also formed at each ZnO grain boundary. This boundary or “depletion” region is wholly within the
ZnO grain and is similar to the Schottky barrier commonly associated with semiconductor interfaces.
.

To analyze varistor behavior, it is useful (although an oversimplification) to represent the

varistor microstructure by the block model shown in Figure 8. This model presumes that the device
is assembled of conducting ZnO cubes of size d separated from each other by an insulating barrier
region of thickness t. It should be emphasized that the insulating barrier is not a separate phase but is
a representation of the back-to-back depletion layers at the ZnO grain boundaries.

Referring to Figure 7, we note that in this sample the varistor grain size d is about 20 µm. For
this varistor material the macroscopic average breakdown field FBR specified at a current density of
1 mA/cm2 is around 130 V/mm. It follows that the macroscopic breakdown voltage per intergranular
barrier, vg , is

vg = FBR d = 2.6 V per barrier (2)

The foregoing value of breakdown voltage per grain barrier is characteristic of ZnO varistors
when the grain breakdown voltage is evaluated by measurements on bulk samples. We actually
expect this bulk value to be lower than the true breakdown voltage per grain because the current
always seeks the easiest path (i.e., the path with fewest barriers between the electrodes). The number
of grains for the actual current path is lower than the average number of grains between the
electrodes, or, equivalently, we should increase the value of d used to estimate vg. Indeed,
microscopic data on varistor behavior of single grain junctions (presented below) give vg ≈ 3.5 V per
grain for varistor materials.

It should also be emphasized that an important and significant feature of the behavior of ZnO
varistors is that macroscopic breakdown voltages of 2 or 3 V per barrier are observed for a wide
variety of ZnO varistor materials. Substantial variations in device processing and composition have
relatively minor effects on vg.

From the schematic of Figure 8 it is clear that the electrical characteristics of ZnO varistors are
related to the bulk of material (i.e., the device is inherently multijunction with varistor action shared
between the various ZnO grain boundaries). This implies that tailoring the device breakdown voltage
VBR is simply a matter of fabricating a varistor with the appropriate number of grains, n, in series
between the electrodes. Thus, to achieve a given breakdown voltage, one can change the varistor
thickness (for fixed grain size) or one can vary the grain size to increase the number of barriers, n,
keeping the device thickness constant. In either case
 Dv g
VBR = nv g = (3)
d

where D is the electrode spacing and d is the ZnO grain size. Typical varistor values for protection
of equipment on 120-V ac power lines are VBR = 200 V, d = 20 µm, D = 1.6 mm, and n = 80.

The block model (Figure 8) approach to understanding the behavior of ZnO varistors is equally
useful in developing a simple equivalent circuit representation of this material. How this comes
about is outlined in the following section.

IV. VARISTOR EQUIVALENT CIRCUIT

Measurement of the dielectric constant ε of ZnO varistors typically gives values ε ~ 1000. This high
value, while perhaps initially surprising, is quite easily understood in terms of the varistor model of
Figure 8 and the equivalent circuit of Figure 9. Since t<<d (Figure 8), it is clear that the volume
between the varistor electrodes is largely occupied by conducting ZnO grain material. In fact, the
thickness of insulating dielectric lying between the sample electrodes is not D but Dt/d. Therefore, if
εg is the dielectric constant of the ZnO in the depletion layers, we expect the varistor capacitance C
to be given by

A
C = ε gε0
Dt / d
d A
= ε gε0 (4)
t D

That is, the effective dielectric constant is increased by a factor of d/t.

Measured values of the effective dielectric constant are ~ 1000 [23], implying d/t ~ 100 since εg
~ 10 for ZnO. Taking d ~ 10 µm [24], we find t ~ 1000 Å. This value represents the thickness of the
depletion layer at the ZnO grain boundary [25-27]. The depletion layer largely controls the varistor
low-voltage capacitance. The thickness of any non-ZnO material between the grains is about two
orders of magnitude less [15-19].

Returning to Figures 8 and 9, it is reasonable to describe the varistor intergranular barriers by a

parallel capacitor (Cp) and resistor (Rp) pair. Rp is clearly voltage-dependent. In addition, both Rp and
Cp vary with frequency and temperature (see below). In series with Rp and Cp is the small resistance
rZnO of the ZnO grains. We can usually ignore rZnO except for very high varistor currents or very
high varistor frequencies [28].
In Figures 10 to 12 we plot the measured dielectric constant ε, dissipation factor D = tan δ, and
parallel resistivity ρp of a commercial varistor. The high dielectric constant evident in Figure 10 is
consistent with our previous discussion and reflects the two-phase nature of ZnO varistors. ε
decreases somewhat, from 30 to 105 Hz, with a sharper dispersive drop evident in the range 105 to
107 Hz.
In Figure 11 the dissipation factor D is peaked in the vicinity of 300 kHz. This peak is
reminiscent of a (broadened) Debye resonance, and some discussion of the origin of the resonance
peak can be found in reference 23. It is interesting to note that in the range 30 to 108 Hz, the
dissipation factor varies by only a factor of 10. This has immediate consequences for the frequency
dependence of the parallel resistivity, since

1
ρp = (5)
εε 0ωD
Thus, noting from Figures 10 and 11 that to a first approximation D and ε are constant, we expect ρp
~ 1/ω (i.e., the varistor resistivity should drop as the inverse frequency).
This behavior is illustrated in Figure 12, where we plot the parallel resistivity ρp versus
frequency. From the figure we have ρp ≈ 109 Ω-cm at 30 Hz and ρp ≈ 103 Ω-cm at f = 108 Hz,
implying that the varistor resistivity drops 6 decades when the frequency increases by 6.5 decades.
Clearly, Rp of Figure 9 cannot be regarded as a simple frequency-independent leakage path.
Superimposed on the general drop in Rp with frequency we observe a dispersive anomaly in the
region 105 to 107 Hz, as expected from Figures 10 and 11, and Eq. (5).
In addition to the frequency dependence of Cp, D = tan δ, and Rp, measurement indicates that
these quantities also vary with temperature. Relevant data and an analysis of the results are given in
refs. 23 and 28.
Finally, it is important to emphasize that the data of Figures 10 to 12 were obtained from
measurements at voltages very low compared with the varistor breakdown voltage VBR. In this
voltage region the varistor is an essentially linear circuit element. As the applied voltage approaches
VBR, more complex phenomena can be anticipated, and an example of such behavior is given by the
voltage dependence of the capacitance described in reference 29.

V. MECHANISM OF VARISTOR BEHAVIOR

Any complete theory of the electrical behavior of ZnO varistors must account for features that
depend on the varistor ceramic microstructure and also account for the detailed behavior of a single
ZnO grain-intergranular barrier heterojunction. Generally, the simplified block model of Figure 8
has proved adequate for understanding the implications of varying varistor microstructures, although
the idealized representation of the ceramic of Figure 8 should certainly be used with caution.
Certainly the disorder associated with different arrangements of grain structures must affect the
varistor behavior and we will discuss this is more detail below.

Our current understanding of the phenomena associated with a single ZnO grain-intergranular barrier
junction is incomplete, and perhaps this imperfect understanding is not unexpected since it is
inordinately difficult to obtain well-controlled electrical or chemical data on an irregular grain
interface imbedded between two grains in a ceramic.

We shall discuss in turn

a. Single Grain Boundary Phenomena
b. Effects of Microstructure
c. ZnO Grain Resistivity
d. Degradation and Failure

a. Single Grain Boundary Phenomena

An appreciation of the behavior of a single grain boundary can be obtained by a more detailed
examination of Figure 2. Three regions can be distinguished in these curves. In the breakdown
region, the current is a highly nonlinear function of the applied voltage for many decades of current
and is essentially temperature independent. The dependence of current on voltage in this region of
the current-voltage curve is often described by the empirical relation given in Eq. (1), where α can
achieve values in the range 50 to 100 near 10–3 A/cm2. At very large currents, the curve exhibits an
upturn. This feature is not an intrinsic property of the breakdown mechanism but is associated with
the finite resistivity of the ZnO grains themselves; the upturn represents the voltage drop in the
grains. In the prebreakdown or leakage region at low voltages, the I–V characteristic is close to
linear. In many cases, the temperature dependence of the current can best be described [30] in terms
of an activation energy.

I = I 0 e − eφ / kT (6)

where eφ ≈ 0.6 to 0.8 eV near 300 K.

An interesting and important feature of the breakdown characteristic is that it is relatively
insensitive to the details of chemical composition and processing within reasonable limits. It is, of
course, possible to prepare samples using somewhat arbitrary ingredients in which the nonlinearity is
minimal (α < 10) and a clear, sharp breakdown characteristic is not observable. However, for
compositions that contain a varistor-forming ingredient (heavy elements such as Bi, Pr, and Ba, with
large ionic radii) and at least one or two varistor performance ingredients (generally transition-metal
elements such as Co, Mn, and Ni), and for sintering temperatures in the range 1100 to 1350°C, the
breakdown behavior illustrated in Figure 2 with α ≈ 50 can be achieved.
The data of Figure 2 relate to measurements on a bulk varistor, i.e. the data are averaged over
many thousands of grain boundaries. However, for typical commercial varistors, the gross ceramic
microstructure does not dominate the conduction process. In Figure 13, I–V data are shown for bulk
varistor material, ≈0.15cm thick, for the as-sintered varistor surface obtained using surface
electrodes spaced ≈0.10 cm apart, and for a single grain-boundary junction [29]. The single grain
junction measurements were obtained using microelectrodes evaporated and photolithographically
defined on carefully prepared varistor surfaces. A pair of electrodes separated by a junction is shown
in Figure 14. The measured voltage is 100 times larger for the bulk and surface electrode data than
that given on the ordinate scale. The shapes of these curves are virtually identical, showing that the
electrical properties are determined solely by the behavior of the individual grain-grain junctions.
The varistor voltage itself is determined by the number of grain boundaries between electrodes.
However, we do not mean to imply here that the uniformity of the microstructure is not an important
device consideration. We small discuss this in more detail in Section V(b) below.

A number of theoretical models have been proposed that explain grain-grain conduction and other
varistor phenomena with reasonable credibility [25-27, 29, 31-36]. These models are based on the
double depletion layer concept for the region of closest grain-grain contact, and this is the generally
accepted starting point for any physical model of the varistor junction. The origin of the double
depletion layer can be understood by considering the formation of a grain boundary when two
identical semiconducting grains are joined. The grain boundary is assumed to be comprised of the
same semiconducting material as the grain but also contains defects and dopants. As a result, its
Fermi level is different from that of the two grains, and it also has additional electronic states
because of the defects and dopants within the bandgap. Electrons flow to the grain boundary, where
they are trapped by the defects and dopants. The result of this electron flow is that the electrons
trapped at the grain boundary act as a sheet of negative charge at the boundary, leaving behind a
layer of positively charged donor sites on either side of the boundary thereby creating an
electrostatic field with a barrier at the boundary. The magnitude of the potential barrier can be
calculated by solving the Poisson equation for the potential. A schematic picture of the grain
boundary potential barrier is given in Figure 15.

This double depletion layer model is adequate to describe the leakage region of the varistor
characteristic and its temperature dependence. However, it was realized at an early date [29] that this
simple approach based only on electron participation in current transport cannot account for the very
large non-linearities (nonlinearity exponent α much greater than 50) exhibited by varistors in the
breakdown region. The participation of minority carriers in the conduction process was first
postulated by Mahan et al.[29] and validated by Pike et al. [37] who observed band-gap
electroluminescence at 3.2eV in some varistor compositions. The greater the luminescent intensity,
the higher the varistor nonlinearity. Electroluminescence was observed in three different varistor
compositions, containing Bi, Bi + Mn, and Bi + Mn + Sb additives, respectively. The observed
luminescent intensity is proportional to the square of the varistor current, which is consistent with a
model of hole creation by impact ionization near the ZnO grain boundaries. Impact-ionized holes
have also been invoked to explain the unusual phenomenon of negative capacitance observed in ZnO
varistors when subjected to bias voltages in the vicinity of breakdown [34].

This effect of holes is shown schematically in Fig. 16. Since the depletion layer is very thin (about
100 nm), the electric field in this region can reach very high values (around 10 MV/cm) when the
applied voltage is in the breakdown range. Under such high electric fields, some of the electrons
crossing the barrier gain sufficient kinetic energy that they can produce minority carriers (holes) by
impact ionization of the valence states and acceptor states within the depletion region. The hole
produced by this process diffuse back to the grain boundary and compensate for part of the trapped
negative interface charge. This lowers the potential barrier at the grain boundary. This impact
ionization hole creation process can result in very large nonlinearity coefficients. A mathematical
model of this process is given in reference [38].

Most of the available theoretical models are phenomenological in their construction in that the
physical origin of the surface state and bulk donors describing the varistor barriers are not
considered. For example, in one phenomenological model [33], the surface state density and doping
profile are extracted from C-V and I-V data. While it may be possible in this model to assign a set of
such parameters to each and every varistor sample and thus account for materials and processing
variations in terms of these parameters, they are not, of themselves, independently reasoned or
justified and are thus of limited predictive utility.

b. Effects of Microstructure

As with all ceramics we expect the microstructure of ZnO varistors to be irregular. The idealized
schematic of Figure 8, depicts conduction between the electrodes taking place through a set of
identical parallel paths. If in fact all the grain boundaries were identical electrically and the shape
and size of all the grains the same, then the electrical characteristics of the bulk varistor would be
same as that of an individual grain boundary. In reality, small perturbations in the ceramic
microstructure - for example, in the number of grain boundaries in the various parallel conduction
paths - can cause the current flow to vary considerably from place to place over the electroded faces
of the varistor disk. This behavior can drastically affect varistor performance, especially in the high
nonlinearity region, even in cases where the individual grain-grain barrier electrical characteristics
are “optimal.”

In general we can expect variations in the nature of the grain boundary as well as variations in the
arrangement of these grain boundaries. For example, by measurements on large numbers of
individual grain boundaries, Einzinger [39] has shown that there can be considerable variation in the
breakdown voltage from one boundary to another (Figure 17). His measurements also showed that
some boundaries were ohmic, some nonlinear, while others were insulating. This type of data,
together with observations regarding the complexity of varistor microstructures, have led to the
realization that varistors are far from uniform.

The need to account for microstructural disorder has been apparent from the very early days of
varistor technology. For example, low voltage varistors operate in the 10 volt range. Since the
breakdown voltage of an individual grain boundary in ZnO varistors is ~3.5 V, low-voltage varistors
are typically only a few grains thick. This places severe constraints on microstructural uniformity in
the device. Clearly, modest variations in grain size will lead to significant preferential channeling of
current in the regions where the grains are larger (fewer grains boundaries between the device
electrodes). Intuitively one can convince oneself that the result of microstructural non-uniformity
will be to degrade the varistor performance by increasing the leakage and decreasing the
nonlinearity.

Modeling the detailed effects of microstructural non-uniformity on varistor behavior is difficult due
to the extremely non-linear nature of the device. However, the advent of more powerful computers
has permitted the varistor behavior to be simulated and the results of extensive computations have
been published by the groups working with Clarke [40-43] and Mahan [44-46]. In modeling the
current flow the microstructure is described in terms of an equivalent electrical network of nonlinear
resistors, then disorder is introduced into the network. The spatial distribution of current flow is than
determined in response to an applied voltage.

As expected intuitively, the simulations reveal that disorder in the form of variations in grain size,
grain-grain contact area, or grain boundary barrier height decrease the attainable nonlinearity, cause
a rounding of the current-voltage characteristic in the vicinity of the breakdown voltage turn-on, and
decrease the breakdown voltage itself. The simulations provide insight on the spatial distribution in
current flow and current localization in particular. This is described in detail in reference [13]. As
might be expected, localization is most pronounced within the nonlinear region. The individual grain
boundaries adjust their conductivities in response to the applied voltages and we observe a spatial
redistribution of the principal current paths through the varistor as the voltage is increased or
decreased. Experimental evidence for inhomogeneous current flow through a varistor comes from
thermal imaging of the cross section of a varistor under electrical load [46]. The high-current paths
are visualized by the local increase in temperature caused by Joule heating. The data show that the
electrical energy is primarily dissipated at the grain boundaries and that the current localization paths
are typically only one or two grains wide.

The ability to model and observe the effects of microstructural non-uniformity in varistors has
enabled manufacturers to enhance the energy handing capability of their material by tailoring
manufacturing processes to produce material with superior uniformity.

c. ZnO Grain Resistivity

In the upturn region (figure 2) the ZnO grain resistivity is an important determinant of device
performance. Conduction within the grain is ohmic and in its simplest representation, Figure 9, adds
a voltage (per unit length) of value

Vg = Jρg (7)

to the breakdown voltage, where ρg is the grain resistivity. This voltage produces the upturn shown
in Figure 2 at high currents and is the feature that limits the performance of metal oxide varistors in
high-current surge-suppression applications. The circuit voltage protective level is not simply the
grain boundary breakdown voltage but is higher because of the voltage drop in the grains
themselves.
While pure ZnO is an insulating semiconductor with an ≈ 3 eV band gap, sintered ZnO is
reasonably conducting. As can be estimated from Figure 2, at room temperature ρg is around 1 Ω-
cm, while ρg is 10 or 100 times higher at 77 K. The grain resistivity can be controlled to a certain
extent by dopants such as Al or Li [47-49]. The simple concept that grain resistivity, ρg, solely
determines the upturn, is however, only approximate. For example, one can attempt to evaluate ρg by
linearly extrapolating the varistor breakdown characteristic (in a log-log plot) into the upturn region.
The difference between observed varistor voltage and extrapolated voltage serves to define a voltage
drop ∆Vg, which should be associated with the voltage drop across the ZnO grain interior.
Unfortunately, if we then compute ∆Vg/I, we find that this quantity is not independent of current I. In
fact, ∆Vg/I decreases with increasing current and is not a very good measure of rg.
Other measurements of the grain resistivity are also of interest and provide additional insight into
the interpretation of upturn data. At high frequencies [28], f > 108 Hz, the intergranular capacitance
becomes an effective short circuit, and the equivalent circuit for the varistor shown in Figure 9 can
be represented as a resistor rg whose resistivity is the ZnO grain resistivity. The value of rg can then
be determined by high-frequency impedance measurements. Infrared optical techniques [50] also
provide a method of measuring the ZnO grain resistivity since the reflectance of ZnO has a sharp
minimum whose wavelength position depends on the free electron density and thus on the ZnO grain
resistivity. Both of these techniques give values for the grain resistivity ρg, which are lower than
those obtained from upturn measurements by perhaps a factor of 5 or more [28].
The grain resistivity can be increased by doping with Li, which presumably introduces deep
levels in ZnO compensating the shallow donors responsible for grain conduction. Doping with Al,
Ga, or In, on the other hand, reduces the grain resistivity. Doping varistors with Al is now common
practice, and grain resistivities of ≈0.1 Ω-cm are achieved in commercial varistor materials.
Although ZnO resistivities below this value are achieved in single-crystal studies [51], efforts to
reduce the resistivity below ≈0.1 Ω-cm by conventional means in varistor materials have not met
with practical success. One of the problems encountered in varistor grain doping is the fact that
leakage conduction is also affected by the use of grain dopants such as Al. Thus, for example, the
voltage clamp ratio, V1 KA/V0.1 mA, which is one figure of merit for varistor surge protection
capability, may have its optimum (lowest) value at an Al doping concentration that does not
minimize the grain resistivity. The connection between Al doping and leakage conduction is made
somewhat clearer by secondary ion mass spectroscopy (SIMS) depth profiles showing that the
element Al tends to segregate at grain boundaries rather than just within the ZnO grains themselves
[52]. Thus, the grain boundary is also doped by Al, and apparently such doping does indeed increase
leakage conduction.

d. Degradation and Failure

Device degradation has been an issue affecting the application of ZnO varistors from the advent of
the use of the technology as a surge protective device. Degradation usually refers to the steady
increase in leakage current when the varistor is subjected to a to a constant dc or ac voltage or to a
series of pulses [53-59]. It is important to note that the magnitude and temperature dependence of the
leakage conduction can be very sensitive to the materials and processing schedules used in the
sample preparation. However, degradation as used here refers to the fact that some electrical,
mechanical, chemical, and environmental stresses can tend to increase leakage conduction [59]. An
example of the effect of dc electrical overstress on the I-V characteristics of a ZnO varistor is shown
in Figure 18. The leakage conduction after stress has increased over four orders of magnitude and is
now polarity dependent.

Degradation phenomena in ZnO varistors exhibit several general features [59]. First, the rate of
degradation (rate of increase of leakage conduction) can be strongly influenced by very small
changes in the chemical makeup and/or processing of the varistor, as well as by heat treatments well
below the sintering temperature. Second, degradation associated with a diverse set of stresses
(electrical, mechanical, environmental, etc.) can often be rapidly reversed by annealing at ≈300°C or
lower. Third, and perhaps most striking, the degradation process affects mainly the varistor leakage
and prebreak-down conduction and not the varistor behavior for voltages above the breakdown
voltage. This behavior is clearly indicated in Figure 18. Therefore, changes in barrier height,
depletion layer thickness, intergranular layer trap density, distribution, etc. which are associated with
the degradation process should be such as to leave unchanged the highly nonlinear breakdown
conduction process.

In the early 1970’s it was found by serendipitous experimentation that in some commercial varistor
formulations, the extent of continuous ac voltage-induced degradation can be minimized by low
temperature (400-600 C) annealing. Also, some varistor formulations are less susceptible to
degradation than others. Moreover, after degradation, the original current-voltage characteristics can
often be restored by annealing the degraded varistor at relatively low temperatures (~200°C).
Annealing is accompanied by the generation of current (thermally stimulated currents), indicating
that charge is trapped during the degradation process and can be released from relatively shallow
traps in the degraded material by thermal activation [60]

Two principal mechanisms for degradation have been proposed. Gupta and Carlson [61] have
proposed a form of electromigration, in which interstitials ions (e.g. Zn) in the depletion layer
diffuse preferentially in the direction of current flow. On reaching the grain boundary, the ions
combine with the defects defining the electrostatic barrier, thereby lowering the potential barrier and
correspondingly increasing the leakage current. A second mechanism proposes that the grain
boundary interface states are associated with chemisorption of oxygen and that, during degradation,
oxygen desorption occurs. Although the mechanism primarily responsible for degradation has not
yet been definitively established, it has been shown that small additions of monovalent ions such as
sodium and potassium to the varistor formulation can enhance the stability to degradation[62]. The
beneficial effect of these ions has been attributed to their occupying interstitial sites and thereby
blocking the migration of zinc interstitials.

In addition to issues related to long term degradation, system application of ZnO varistor devices
requires that the material be used under conditions the do not lead to device failure induced by
excessive pulse energy or thermal energy input. In practice two main types of device failure are
commonly observed. One is a so-called puncture mode, in which a hole penetrates through the
varistor with distinct signs of material melting, blackening and vaporization. In the other form of
failure the varistor fractures into two or more pieces but with no melting. Both failure modes occur
above a threshold energy, as shown in Figure 19. It is generally found that fracture mode failures are
associated with short pulse widths while long pulses failures result in punctures [63].

The puncture mode of failure is generally assumed to occur as a result of thermal runaway. The
current becomes so high locally that the ZnO heats sufficiently to cause even more current to
channel the through the local hotspot. Since the grain boundary leakage increases rapidly with
temperature, the increased current caused by local heating along the localization path results in
further current localization and heating. This positive feedback loop continues with melting and
grain boundary electrical shorting the ultimate result.

When varistors are subjected to extremely short (less than 10 microsecond) high current density
pulses (Figure 19), varistor failure tends to occur by fracture of the material. This short pulse
behavior is generally regarded as adiabatic in nature since the associated thermal diffusion length is
less than the ZnO grain size. Material fracture morphology and fracture mode observed depends on
the shape and size of the varistor. Tall varistors used for lightning arresters often fracture near the
varistor block center with the fracture surface parallel to the electrode. It is fairly clear that this type
of fracture is induced by a shock wave generated as a result of the very high current pulse.
Both puncture and fracture failures can be lessened if the varistor ceramic is uniform, defect free and
has good mechanical strength.

VI. VARISTOR APPLICATIONS

ZnO varistors have now been in widespread use for more than 30 years. The technology has
supplanted generally less useful prior approaches to voltage surge protection such as silicon carbide
arresters, gas discharge tubes and to some extent semiconductor-based surge suppressors such as
Zener diodes. A significant advantage of ZnO varistors as a transient protective device derives from
the ceramic nature of the material. Since the material is polycrystalline with energy absorption
occurring essentially at the grain boundaries distributed throughout the volume of the material, ZnO
varistors are inherently able to absorb more energy than single-junction protective devices such as
Zener diodes.

More than one billion ZnO varistor devices are in service. They protect circuits with voltages
ranging from around 10 volts to millions of volts. The larger devices are capable of surviving
transient currents greater than 100,000 amps and can absorb energies greater than 10,000 joules
(energy absorption capability > 200 J/cm3 is routinely available).

Consumer-purchased surge-suppressor electrical outlets based on ZnO varistor technology are

commonly used to protect computers and other high value electronic equipment. A simple example
of this type of unit shown in Figure 20. Three surge-protected power points are available. In figure
21 we show a variety of shapes and sizes for ZnO varistors sold into the residential and industrial
markets for protecting electrical equipment operating at voltages from a few tens of volts to around
1000 volts. Details of voltage and energy ratings, application notes etc are available from many
manufacturers in the USA, Europe and Asia.

A second useful feature of ZnO varistors deriving from the ceramic nature of the material is the
ability to configure a particular device to conform to system constraints. We note some applications
of ZnO varistors to illustrate the versatility of the material. Figure 22 shows a miniature ZnO varistor
sleeve [64] fabricated as a tube that can fit around a connector pin as small as size 22. These devices
fit inside available multipin cable connectors and provide a low-inductance compact configuration
for protection of sophisticated electronics against other electrical transients.

Figure 23 illustrates a miniature “chip” varistor designed to be directly mounted on a printed

circuit board using surface-mount technology. The absence of device leads reduces inductance and
permits dense device packing on the circuit board. Devices of this shape are often made in a multi-
layer configuration (similar to multi-layer capacitors) which facilitates the fabrication of lower
voltage varistor devices.

In addition to the protection of electronic equipment, a major application of ZnO varistors is for
the protection of electric power distribution and transmission systems [65]. In these applications, a
surge suppressor is required to function reliably on systems with voltages ranging up to a megavolt
and to absorb transient energies in the megajoule range. Large volumes of varistor material are
needed to meet these requirements. In Figures 24 and 25, a large station arrester containing hundreds
of ZnO varistor disks, each > 100 cm3 in volume, is depicted. The advantages to the power system
designer of using ZnO varistor transient protection technology is extensively described elsewhere
[66]. It is interesting to note, however, that ZnO varistor technology has significantly simplified the
design and increased the reliability of electrical power system protection against transient surges.

VII. CONCLUSIONS

ZnO varistors have found acceptance as a simple, cost-effective way to protect electrical systems
from transient voltage surges. Device operation is controlled by grain-boundary effects at the ZnO-
ZnO interface. By virtue of their ceramic nature, they can be fabricated into a variety of sizes and
shapes, and this feature facilitates a high degree of user flexibility. Varistors are inherently
multijunction grain-boundary devices and any transient surge energy to be absorbed when the device
acts in a protective mode is distributed among the many ZnO grain junctions. The multijunction
feature of varistor behavior is the heart of the high-current and energy-absorption capability of the
material.

REFERENCES

1. D.E. Castleberry, IEEE Trans. Electron Devices, ED-26, R1123 (1978); D. E.

Castleberry and L. M. Levinson, 1980 SID International Symposium Paper 18.5.
2. R. G. Dosch, B. A. Tuttle, and R. A. Brooks, J. Mat. Res., 1[1], 90-99 (1986)
3. F.D. Martzloff and G. D. Hahn, IEEE Trans. Power Apparat. Syst., PAS-89 (6), 1049–
1056 (July–Aug. 1970).
4. R.W. Fox, Electron. Des. 22, 52 (May 24, 1974).
5. M. Matsuoka, Jpn. J. Appl. Phys., 10, 736 (1971).
6. Kh. S. Valeyev, V. A. Knayazev, and N. G. Drozdov, Electrichestro, 4, 72(1964).
7. M. S. Kosman and E. G. Pettsol’d, Uch. Zap. Leningr. Gos. Pedagog Inst. im A. I.
Gertsena, 207, 191 (1961).
8. L. M. Levinson and H. R. Philipp, Am. Ceram. Soc. Bull., 65 [4] 639–46 (1986).
9. R. Einzinger, Ann. Rev. Mater. Sci., 17, 299–321 (1987)
10. T. K. Gupta, J. Am. Ceram. Soc., 73 [7] 1817–40 (1990).
11. G. Blatter and F. Greuter, Semicond. Sci. Technol., 5 [1] 111–37 (1990).
12. G. E. Pike, pp. 731–54 in Materials Science and Technology, Vol. 11. Edited by M. V.
Swain. VCH, Weinheim, Germany, 1994.
13. David R. Clarke, J. Am. Ceram. Soc., 82 [3] 485–502 (1999).
14. H.R. Philipp and L. M. Levinson, Adv. Ceram., 1, 309 (1981)
15. D. R. Clarke, J. Appl. Phys., 49 [4] 2407–11 (1978).
16. D. R. Clarke, J. Appl. Phys., 50 [11] 6829–32 (1979).
17. W. D. Kingery, J. B. Vander Sande, and T. Mitamura, J. Am. Ceram. Soc., 62 [3–4] 221–
22 (1979).
18. Y.-M. Chiang, J.-R. Lee, and H. Wang, pp. 131–47 in Ceramic Microstructure:Control
at the Atomic Level. Edited by A. P. Tomsia and A. Glaeser. Plenum Press, New York,
1998.
19. W. G. Morris and J. W. Cahn, pp. 223–34 in Grain Boundaries in Engineering Materials.
Edited by J. L. Walter, J. [Link], and D. A. Woodford. Claitors, Baton Rouge, LA,
[Link] ZnO-Sb2O3 Ceramics,” J. Am. Ceram. Soc., 75 [11] 3127–28 (1992).
20. J. Wong, J. Appl. Phys., 46 [4] 1653–59 (1975).
21. A. T. Santhanam, T. K. Gupta, and W. G. Carlson, J. Appl. Phys., 50 [2] 852–59 (1979).
22. M. Inada, Jpn. [Link]. Phys., 17 [1] 1–10 (1978).
23. L. M. Levinson and H. R. Philipp, J. Appl. Phys., 47, 1117 (1976).
24. L. M. Levinson and H. R. Philipp, J. Appl. Phys., 46, 1332 (1975).
25. W. G. Morris, J. Vac. Sci. Technol., 13, 926 (1976).
26. J. D. Levine, Crit. Rev. Solid State Sci., 5, 597 (1975).
27. J. Bernasconi, H. P. Klein, B. Knecht, and S. Strassler, J. Electron. Mater., 5, 473 (1976).
28. L. M. Levinson and H. R. Philipp, J. Appl. Phys., 47, 3116 (1976).
29. G. D. Mahan, L. M. Levinson, and H. R. Philipp, J. Appl. Phys., 50, 2799 (1980).
30. H. R. Philipp and L. M. Levinson, J. Appl. Phys., 50, 383 (1979).
31. J. Bernasconi, S. Strassler, B. Knecht, H. P. Klein, and A. Menth, Solid State Commun.,
21, 867 (1977).
32. P. R. Emtage, J. Appl. Phys., 48, 4372 (1977).
33. P. L. Hower and T. K. Gupta, J. Appl. Phys., 50, 4847 (1979).
34. G. E. Pike, Mater. Res. Soc. Proc., 5, 369 (1982).
35. F. Greuter, G. Blatter, M. Rossinelli, and F. Stucki, in Ceramic Transactions: Advances
in Varistor Technology, Vol. 3, p. 31, L. M. Levinson, ed., American Ceramic Society,
Columbus, Ohio, 1989.
36. F. Greuter, and G. Blatter, Semicond. Sci. Technol. , 5, 111 (1990).
37. G. E. Pike, S. R. Kurtz, P. L. Gourley, H. R. Philipp, and L. M. Levinson, J. Appl. Phys.,
57, 5512 (1985).
38. G. Blatter and F. Greuter, Phys. Rev. B: Condens Matter, 34 [12] 8555–72 (1986).
39. R. Einzinger, Appl. [Link]., 3 [1] 390–408 (1979).
40. Q. Wen and D. R. Clarke, pp. 217–30 Ceramic Transactions, Vol. 41, Grain Boundaries
and Interfacial Phenomena inElectronic Ceramics. Edited by L. M. Levinson and S.-I.
Hirano. AmericanCeramic Society, Westerville, OH, 1994.
41. A. Vojta, Q. Wen, and D. R. Clarke, Comput. Mater. Sci.,6 [1] 51–62 (1996).
42. A. Vojta and D. R. Clarke, J. Appl. Phys., 81 [2] 985–93(1997).
43. C-W. Nan and D. R. Clarke, J. Am. Ceram. Soc., 79 [12] 3185–92 (1996).
44. M. Bartkowiak and G. D. Mahan, Phys. Rev. B: Condens. Matter, 51 [16] 10825–32
(1995).
45. M. Bartowiak, G. D. Mahan, F. A. Modine, and M. A. Alim, J. Appl. Phys., 79 [1] 273–
81 (1996).
46. H. Wang, M. Bartowiak, F. A. Modine, R. B. Dinwiddie, L. A. Boatner, and G. D.
Mahan, J. Am. Ceram. Soc., 81 [8] 2013–22 (1998).
47. K. Eda, in Materials Research Society Symposium Proceedings: Grain Boundaries in
Semiconductors, H. J. Leamy and G. E. Pike, eds., Elsevier, New York, 1982.12 (1997).
48. T. Myoshi, K. Maida, K. Takahashi, and T. Yamaryaki, Adv. Ceram., 1, 309 (1981).
49. W. G. Carlson and T. K. Gupta, J. Appl. Phys., 53, 5746 (1982).
50. H. R. Philipp and L. M. Levinson, J. Appl. Phys., 48, 1621 (1977).
51. G. Heiland, E. Mollwo, and F. Stockman, in Solid State Physics, Vol. 8, F. Seitz and D.
Turnbull, eds., Academic Press, New York, 1959.
52. H. R. Philipp, G. D. Mahan, and L. M. Levinson, Report ORNL/Sub/84-17457/1,
Subcontract 86X-17457C, July 1984 for DOE under Contract DE-AC05-840R 21400.
53. K. Eda, A. Iga, and M. Matsuoka, Jpn. J. Appl. Phys., 18 [5] 997–98 (1979).
54. C. G. Shirley and W. M. Paulson J. Appl. Phys., 50 [9] 5782–89 (1979).
55. K. Eda, A. Iga, and M. Matsuoka, J. Appl. Phys., 51 [5] 2678–84 (1980).
56. T. K. Gupta, W. G. Carlson, and P. L. Hower, “Current Instability Phenomena in ZnO
Varistors,” J. Appl. Phys., 52 [6] 4104–11 (1981).
57. W. Moldenhauer, K. H. Bather, W. Bruckner, D. Hinz, and D. Buhling, Phys. Status
Solidi A, 67, 533–42(1981).
58. M. Hayashi, M. Haba, S. Hirano, M. Okamoto, and M. Watanbe, J. Appl. Phys., 53[8]
5754–62 (1982).
59. H. R. Philipp and L. M. Levinson, pp. 1–21 in Advances in Ceramics, Vol. 7, Additives
 and Interfaces in Electronic Ceramics. Edited by M. F. Yan and A. H. Heuer, American
Ceramic Society, Columbus, OH, 1983.
60. K. Sato and Y. Takada, J. Appl. Phys., 53 [5] 8819 (1982).
61. T. K. Gupta and W. G. Carlson, J. Mater. Sci., 20, 3487–500 (1985)
62. T. K. Gupta and A. C. Miller, J. Mater. Res., 3 [4] 745–54 (1988).
63. K. Eda, J. Appl. Phys., 56 [10] 2948–55 (1984).
64. Aviation Week and Space Technology, May 20, 1985
65. E. C. Sakshaug, J. S. Kresge, and S. A. Miske, IEEE Trans. Power Apparat. Syst., PAS-
96, 647–656 (1977).
66. Application Guide, Tranquell Station Surge Arresters, General Electric Co., GE Industrial
Systems, Fort Edward, NY, 12828. See the web sites
[Link] and
[Link]
Figure 1 Current-voltage trace (actual photo) of a typical ZnO varistor
Figure 2 Current-voltage characteristics of a metal oxide varistor at 77K and for a small range of
temperatures near 300K. The exponent α equals the inverse slope of the curve and is a measure of device
nonlinearity
Figure 3 Typical application of ZnO varistor as a transient protective element
Figure 4 Voltage response (dashed line) and current response (dotted line) of a varistor chip
in a 50-Ω circuit to ≈ 500-psec rise-time pulses of magnitude (a) 50 V and (b) 150 V.
The solid line indicates the response without the chip in the circuit.
Figure 5 Simplified flow diagram for the fabrication of ZnO varistors
used to protect electronic equipment.
Figure 6 Schematic depiction of the microstructure of a ZnO varistor with grains
of conducting ZnO, average side d. Bismuth is segregated to the ZnO grain
boundaries. Electrodes are attached and current flows as indicated. .
Figure 7 Optical photomicrograph of a polished and etched section of a typical varistor.
The various features are discussed in the text.
Figure 8 Schematic depiction of the microstructure of a ZnO varistor.
Grains of conducting ZnO, average side d. Bismuth is segregated to the ZnO grain boundaries.
Electrodes are attached and current flows as indicated
Figure 9 Simple equivalent circuit representing a metal oxide varistor as a capacitance in parallel
with a voltage-dependent resistor. Cp and Rp are the capacitance and resistance of the intergranular layer,
respectively; rg is the ZnO grain resistance. For low applied voltages Rp behaves as an ohmic loss.
Figure 10 Variation with frequency of the room-temperature dielectric constant of a ZnO varistor.
Figure 11 Dissipation factor D = tan δ versus frequency at room temperature.
Figure 12 Parallel resistivity ρP of a ZnO varistor at room temperature.
Figure 13 Current-voltage characteristics of commercial varistor material obtained using (1) microelectrodes
placed across a single grain junction, (2) electrodes placed 0.1 cm apart on the as-sintered varistor surface,
and (3) electrodes placed on the opposite faces of a varistor disk ≈0.15cm thick. Voltages measured in (2) and (3) are
100 times that indicated on the ordinate scale.
Figure 14 Photomicrograph of evaporated Al electrode configuration on the surface of a ZnO varistor.
The electrodes have ohmic contact to two ZnO grains separated by a single grain barrier. The measured
current-voltage curve is given in Figure 13.
Figure 15 Double depletion layer at a ZnO grain boundary. Electrons move from the grain into traps
at the grain boundary, inducing the electron barrier shown. EF and EC denote the Fermi level and conduction
band, respectively (adapted from Pike[12])
Figure 16 Band diagram for a grain barrier under an applied voltage in the breakdown region
Hot electron create holes by impact ionization. These holes move to the grain boundary where they
decrease the barrier height. The hole creation process and consequent barrier height drop produces
the extremely high non-linearity observed in ZnO varistors (adapted from Pike [12] and Clarke [13]).
Figure 17 Variation of breakdown voltage across single grain boundaries.
The data were obtained using micro-contacts. (Adapted from Einsinger [39])
Figure 18 The effect of dc electrical overstress on the current-voltage
characteristics of a ZnO varistor.
Figure 19 Current and pulse Duration map for varistor failure modes.
(Adapted from Reference [63]
Figure 20 Typical Consumer Surge-protected Electrical Outlet
Figure 21 Various shapes and sizes for ZnO varistors sold into the residential and industrial
markets for protecting electrical equipment operating at voltages from a few tens of volts to around 1000 volts
Figure 22 Tubular varistor sleeve mounted on connector pin. The pins are in turn placed
inside the shell of multipin cable connectors.
Figure 23 Surface-mount chip varistor. These leadless devices can be directly soldered
onto a printed circuit board or ceramic hybrid assembly.
Figure 24 Large power station arrester containing hundreds of ZnO varistor
disks each > 100 cm3 in volume
View publication stats

Figure 25 ZnO varistor components used in the construction of the high voltage station arrester
shown in Figure 24. The ZnO varistor station arrester discs are surrounded by a silicone rubber collar
and pressed against the side of a porcelain insulating housing to enhance thermal transfer