04a

DIELECTRIC LOADED ANTENNAS
L. SHAFAI
University of Manitoba
Winnipeg, Manitoba, Canada
1. INTRODUCTION
A transmit antenna converts the energy of a guided wave
in a transmission line into the radiated wave in an
unbounded medium. The receive antenna does the re-
verse. The transmission lines such as waveguides and
coaxial and microstrip lines use conductors mostly to
conne and guide the energy, but antennas use them to
radiate it. Because the radiated energy is in unbounded
region, phase control is often used to direct the radiation
in the desired direction. Dielectrics play an important role
in this process, and this article discusses a few represen-
tative cases. An important antenna parameter is its
directivity, which is the measure of its control over the
energy ow. To increase the directivity, the antenna size
must be increased, and the inuence of dielectrics on their
performance changes considerably. Thus, in this article,
the use of dielectrics in antenna applications is divided
into two categories of large high-gain and small low-gain
antenna applications.
In high-gain antenna applications, reectors and
lenses are used extensively [1]. They are passive and
operate principally on the basis of their geometry. Conse-
quently, they are relatively low-cost, reliable, and wide-
band. Reectors are usually made of good conductors, and
thus have lower loss, and because of their high strength,
can be made light. But reectors suffer from limited scan
capability. Lenses, on the other hand, because of their
transparency, have more degree of freedom, specically,
two reecting surfaces and the relative permittivity or
refractive index. They also do not suffer from aperture
blockage. However, lenses have disadvantages in large
volume and high weight.
In microwave antenna applications, lenses have nu-
merous and diverse applications, but in most cases they
are large in size with respect to wavelength. Thus, physi-
cal and geometric optics apply, and most of the lens design
techniques can be adopted from optics to microwave
applications. The aperture theory and synthesis techni-
ques can also be used effectively to facilitate designs. In
addition, the use of an optical ray path in lens design
makes the solution frequency-independent. In practice,
however, the lens size in microwave frequencies is nite
with respect to the wavelength, and the feed antenna is
frequency-sensitive. Thus, the performance of the lens
antenna also becomes frequency-dependent.
Natural dielectrics at microwave frequencies have re-
ective indices larger than unity and for collimation,
require convex surfaces. However, articial media using
guiding structures, such as waveguides, are equivalent to
dielectrics with refractive index less than unity, and result
in concave lenses. They are usually dispersive, resulting
in variation of the refractive index with frequency, and
have narrower operating bandwidths.
In small antennas dielectrics are used often to improve
the radiation efciency and polarization of the antennas,
such as waveguides and horns. This is important in
telecommunication applications, where polarization con-
trol is required to implement frequency reuse and mini-
mize interference, especially in satellite and wireless
communications. Horn antennas and reector feeds are
examples that incorporate dielectrics or lens loading to
improve performance [2].
Another area of important dielectric use is insulated
antennas in biological applications and remote sensing
with buried or submerged antennas. The use of dielectric
loading eliminates direct RF energy leak into the lossy
environments, and ensure radiative coupling into the
target objects. Often a full-wave analysis is needed to
provide a proper understanding of their resonance prop-
erty and coupling mechanism to the surrounding media.
The dielectric loading is also used for antenna minia-
turization. Low-loss dielectrics with medium to high re-
lative permittivities are now available and are used
increasingly to reduce the antenna size. A number of
important areas include dielectric-loaded waveguides
and horn, and dielectric resonator and microstrip anten-
nas. By aperture loading of waveguides and small horns,
excellent pattern symmetry and low cross-polarization
can be obtained, which are essential features of reector
and lens feeds. In addition, the dielectric loading reduces
the size of the antennas and makes them useful candi-
dates for multibeam applications, using reectors and
lenses. Miniaturization of the antenna is also an impor-
tant requirement in wireless communications. Microstrip
patch or slot antennas with high-relative-permittivity
substrates play an important role in this area, and their
derivatives are used in most applications. In dipole and
monopole cases the dielectric loading is external and used
for size reductions.
Finally, dielectric loading can also be used for gain
enhancement, without shaping them such as lenses. Pla-
nar dielectrics can be used as radomes or covers to protect
the antennas. By proper selection of the radome para-
meters, the antenna gain can also be increased signi-
cantly, while protecting the antennas from the
environment.
2. DIELECTRIC LENS ANTENNAS
In optical terms, a lens produces an image of a source
point at the image point, and lenses could be located
anywhere in the space. As an antenna, this property
means that the source and image points are focused at
each other and the lens has two focal points. In turn, these
focal points signify locations in the space, where rays
emanating from the lens arrive at equal phases. This
D
893
property provides a mathematical relationship for describ-
ing the lens operation, and therefore its design.
To simplify the mathematics, the lens conguration is
assumed to be rotationally symmetric, and the focal points
are placed on its axis. A further simplication can be made
for antenna applications, where the image point moves to
innity; that is, the lens focuses a nearby source point, on
its axis, to another axial point at innity. In such a case,
all rays leaving the lens travel parallel to its axis, and
their phasefronts are planes normal to the lens axis. This
is shown in Fig. 1, where e
r
is the relative permittivity of
the lens material, and n

e
r
p
is its refractive index.
To design the lens, one is required to determine the
geometry of its two faces, front and back, or the coordi-
nates x
1
, y
1
, and x
2
, y
2
of points P
1
and P
2
(Fig. 2). There
are four unknowns to be determined. The equality of the
phase on the phasefronts requires that the electrical
length between the focal points and the phasefronts be
independent of the pathlengths. This provides one equa-
tion. Two other equations can be obtained from the ray
optics at the lens interface points P
1
and P
2
, namely, the
Fermat principle of minimum pathlengths. This enforces
the well-known Snell law of refraction at the lens surface
points. An additional relationship must be generated from
the required lens properties, to enable a unique solution
for the lens design.
To enforce the invariance of the ray pathlength, the
central ray passing through points A, B, and C is selected
as the reference and its length from S to C is compared
with that of the ray passing through points P
1
, P
2
, and P
3
.
This provides the following equation
SP
1
nP
1
P
2
P
2
P
3
SAnABBC 1
or
r
1
nr
3
L
1
F nTL
0
2
where in terms of P
1
and P
2
coordinates each length is
given by
r
1
x
2
1
y
2
1
1=2
r
3
x
2
x
1
2
y
2
y
1
1=2
L
1
x
3
x
2
L
0
x
3
F T
3
and F and T are the lens focal length and axial thickness,
and are therefore constant lengths dening the lens.
Enforcing the Fermat principle at points P
1
and P
2
results in differentiation of the pathlength in Eq. (1) in
terms of its variables x
1
, y
1
and x
2
, y
2
and setting it to zero.
This provides the slope of the lens surface proles at each
point P
1
and P
2
.
At point P
1
one obtains
d
dx
1
r
1
nr
3
L
1

d
dx
1
F nT L
0

d
dx
1
L
0
4
where the constants F and T are dropped and after
simplications one nds
dy
1
dx
1
x
1
r
3
x
2
x
1
nr
1
y
2
y
1
nr
1
y
1
r
3
5
At point P
2
, a similar differentiation in terms of x
2
gives
dy
2
dx
2
x
2
x
1
n r
3
y
2
y
1
n
6
Equations (2), (5), and (6) are three fundamental equa-
tions to design the required lens. Without another rela-
tionship, x
1
may be selected as the independent variable.
Then others, namely, x
2
, y
1
, and y
2
, become dependent
Feed antenna
Convex lens
r
n > 1
n =
r
Feed antenna
Concave lens
r
n < 1
(a)
(b)
Figure 1. Geometry of lens antennas showing the feed and
inuence of lens on ray direction.
894 DIELECTRIC LOADED ANTENNAS
variables to be determined in terms of x
1
. The solutions
give the lens proles in rectangular coordinates. If the
lens proles in polar coordinates are required, Eqs. (2), (5),
and (6) can be obtained in terms of r
1
,y
1
and r
2
,y
2
, the
polar coordinates of P
1
and P
2
. Differentiating Eq. (2) in
terms of y
1
and y
2
gives
dr
1
dy
1
nr
1
r
2
siny
2
y
1
r
3
nr
2
cosy
2
y
1
r
1
7
and
dr
2
dy
2
nr
1
r
2
siny
2
y
1
r
2
r
3
sin y
2
r
3
sin y
2
nr
2
r
1
cosy
2
y
1
8
where use is made of the following polar coordinate
relationships:
x
1
r
1
cos y
1
y
1
r
1
sin y
1
x
2
r
2
cos y
2
y
2
r
2
sin y
2
r
3
jr
1
r
2
j r
2
1
r
2
2
2r
1
r
2
cosy
2
y
1
1=2
9
Solutions of Eqs. (7) and (8) give the lens proles in polar
coordinates, which are often more compact in form. Also,
for some simple lens congurations, they result in well-
known and easily recognizable parametric equations of
the conic sections, generalizing the solution.
2.1. Examples of Simple Lenses
The lens design becomes considerably easier, if one of the
lens surfaces is predetermined. This eliminates one of the
differential equations, as the surface prole is already
known. Among many surfaces to select the simpler ones
are the planar and spherical surfaces, with the planar
normal to the lens axis. Such selections give simple prole
equations. The planar surface is described by a constant
x coordinate and the spherical one by a constant polar
coordinate r. These simplications also assist in solutions
of the other lens prole, for which an analytic solution can
also be determined. Since either of the lens proles can be
predetermined as planar or spherical, four possible solutions
exist. Only two, however, result in simple conical sections.
If the second surface S
2
is assumed to be planar, normal
to the lens axis, the rays arriving from the right-hand side,
parallel to the lens axis x, enter the lens unaffected and
change direction only after the rst lens surface S
1
. Then
they focus at S; that is, only the S
1
surface of the lens
collimates the beam. Looking from the left side, spherical
rays originating from the focal point S, enter the lens at S
1
and become parallel to its axis. Thus, after leaving the lens
at S
2
, since they are normal to S
2
, their direction remains
unchanged. In this case, the active surface S
1
of the lens is
a hyperbola in cylindrical lens, and hyperboloid in rota-
tionally symmetric lens.
If the surface S
1
is spherical, it becomes inactive, since
the focal point is a point source and rays emanating from
(r
1
,
1
)
S
1
S
2
r
1
2
Phase front
P
2
(x
2
, y
2
)
F
S
y
C
A
T
x
D
B
r
,n
Source
point
P
1
(x
1
, y
1
)
(r
2
,
2
)
(X
3
, 0)
r
2
r
3
L
1
L
0
P
3
Figure 2. Geometry of a lens indicating
ray and surface coordinates.
DIELECTRIC LOADED ANTENNAS 895
it constitute spherical waves. Thus, when S
1
is predeter-
mined as a spherical surface, they enter the lens unaf-
fected. Their collimation is done entirely by the lens
second surface S
2
. Its surface is again a conic section,
and its cross section is elliptic. In the other two cases, both
surfaces S
1
and S
2
of lens participate in beam collimation
and consequently are interdependent and more complex.
2.1.1. Lens with Planar S
2
. On S
2
, x is constant and
slope is innite (Fig. 3) and the surface is dened by
x
2
F T
y
2
y
1
10
A consequence of this is L
1
L
0
in Eq. (2), and using
Eq. (10) it becomes
r
1
nr
3
F 11
which, using Eqs. (10), becomes a function of x
1
and y
1
. It
can be solved directly to yield the prole of S
1
as
y
2
1
n
2
1x
1
F
2
2n 1Fx
1
F 12
or in polar coordinates
r
1
n 1F
n cos y
1
1
13
They represent rectangular and polar equations of a hy-
perbola, which is the lens prole on S
1
. They can also be
used to determine the lens thickness on the axis. For this,
one can use two extreme rays, passing through the lens tip
and the axis. The equality of the electrical lengths gives
F nTr
1
y
1 max

D
2
_ _
2
F T
2
_ _
1=2
14
A solution of this equation gives the lens thickness T as
T n1
1
n1D
2
4n 1
F
2
_ _
1=2
F
_ _
15
and
y
1 max
cos
1
1
n
_ _
tan
1
D=2
F T
_ _
16
Equation (16) shows that, for a given dielectric, the lens
aperture angular size is limited by its refractive index n.
In other words, with common dielectrics there is a limit on
the compactness of the lens. That is, the focal length F
cannot be reduced beyond the limit specied by Eq. (16).
2.1.2. Lens with Planar S
1
. In this case, both lens sur-
faces contribute to the beam collimation. Its surface can be
determined similar to the case (in Section 2.1.1) by enfor-
cing x
1
F and innite slope for S
1
(Fig. 4). The results
are [3]
x
1
F
x
2
fn1TF
2
y
2
1
1=2
n
2
1y
2
1
n
2
F
2
1=2
n
2
FF
2
y
2
1
1=2
g
n
2
F
2
y
2
1
1=2
n
2
1y
2
1
n
2
F
2
1=2
y
2
y
1
1
x
2
F
n
2
1y
2
1
n
2
F
2
1=2
_ _
T
1
2
n 1
1
4F
2
D
2
1=2
2F
17
x
D
n
y
S
F T
1
r
1
S
1
S
2
Hyperboloid
(r
1
,
1
)
P
1
(x
1
, y
1
)
Figure 3. Geometry of lens with a planar surface S
2
.
x
y
D
n
S
F
T
1
S
1
S
2
(r
1
,
1
)
P
1
(F, y
1
) (r
2
,
2
)
P
2
(x
2
, y
2
)
Figure 4. Geometry of lens with a planar surface S
1
.
Note that, since the beam collimation is due to both
surfaces, the coordinates of S
2
are now dependent on those
of S
1
.
2.1.3. Lens with One Spherical Surface. When S
1
is a
spherical surface, all spherical wave originating at the
focal point S pass through it unaffected. The second sur-
face, S
2
, collimates the beam. The geometry is shown in
Fig. 5, and S
2
is an ellipse given as
r
2
n 1R
n cos y
2
18
where RFT and other parameters are as dened in
Fig. 5. Its equation in rectangular coordinates has the form
y
2
x
2
n 1R
n
_ _
2
x
2
2
_ _
1=2
and
T
1
2
n 1
1
2F 4F
2
D
4
1=2
19
y
2 max
cos
1
1
n
_ _
the last equation again sets a limit for the peak angular
aperture of the lens for a given dielectric material.
When the surface S
2
is assumed to be spherical, then
both lens surfaces participate in collimating the beam.
The inner surface S
1
can be obtained from [3]
n
2
r
2
2
r
2
1
2r
1
r
2
cosy
1
y
2
n 1Tr
2
cos y
2
r
1
2
n
2
r
1
siny
1
y
2
sin y
2
n 1T r
2
cos y
2
r
1
20
T
4n 1F
2
n 3D
2
4n 1n 3
2
_ _
1=2
F
n 3
3. EFFECT OF LENS ON AMPLITUDE DISTRIBUTION
The lens equations (1) to (6) were based on the ray path
analysis, or in antenna terms, the phase relationships.
The amplitude distributions were not considered. In prac-
tical applications, however, the amplitude distributions
are also important and will inuence the aperture ef-
ciency of the lens, sidelobe levels, and cross-polarization.
To state it briey, a uniform aperture distribution gives
the highest directivity, but has high sidelobes because of
its high edge illumination. Sidelobes can be reduced by
tapering the eld toward the edge. Excessive tapering,
however, rapidly reduces the lens directivity. It is there-
fore useful to know the inuence of the lens on the eld
amplitude as well.
Assume that A(y) is the angular dependence of the
wave amplitude radiating from the focal points and A(r),
with r r siny, the amplitude distribution of the colli-
mated beam. Then, using the conservation of power, and
neglecting the reection at the lens surface, the following
amplitude relationships can be obtained [1].
Hyperbolic lens of case in Section 2.1.1
Ar
Ay
1

1
F
n cos y
1
1
3
n 1
2
n cos y
1
_ _
1=2
21
Elliptic lens of case in Section 2.1.3
Ar
Ay
1

1
F
n cos y
1
3
n 1
2
n cos y
1
1
_ _
1=2
22
An inspection of these equations shows that in Eq. (21) the
amplitude ratio decreases with y
1
; that is, after leaving
the lens the eld is concentrated near its axis. The
amplitude, in fact, drops to zero at the angle y
1 max
, given
by Eq. (16). This lens, therefore, enhances the eld taper
of the source and is a good candidate for low-sidelobe
applications. However, its aperture efciency will be low.
Incontrast, the amplitude ratio in Eq. (22) increases withy
1
;
that is, this lens corrects the amplitude taper of the source
and enhances the aperture efciency, but in the process,
raises the sidelobe levels. Thus, it may be used in applica-
tions in which the aperture efciency is more critical than
the sidelobe levels.
For most common dielectrics the refractive index is
n1.6, (i.e., e
r
2:55). For these materials the limit of the
aperture angle is y
1max
51.31. Within this limit the
amplitude ratios of Eqs. (21) and (22), normalized to axial
values, are shown in Table 1. The amplitude tapering of
hyperbolic lenses is clearly evident. A 351 lens adds
another 10 dB to the aperture eld taper, and beyond
401, the lens is practically useless. For large-angle-lens
applications, higher-dielectric-constant materials must be
used. Table 1 also shows the amplitude enhancement of
elliptic lens. A 351 lens improves the aperture eld uni-
formity by as much as 6.3 dB. It increases rapidly there-
after and yields about 10 and 20 dB improvements for lens
angles of 451 and 501, respectively. These amplitude
enhancements, however, must be accepted as theoretical
S x
D
n
T
S
y
F
1
Spherical
S
1
S
2
(F,
1
)
P
1
(x
1
, y
1
)
(r
2
,
2
)
P
2
(x
2
, y
2
)
Figure 5. Geometry of lens with a spherical surface S
1
.
limits, since at these wide angles the lens surface
reectivity will reduce the practically attainable levels.
Surface matching layers must be used to minimize the
reections.
3.1. General Lens Design
In the general lens of Fig. 2, both surfaces are proled and
participate in collimating the beam. Thus, a more versa-
tile lens can be obtained. However, Eqs. (1)(6) showed
that there are at least four unknown coordinates, x
1
, y
1
and x
2
, y
2
, to be determined. But, the optical relationships
provided only three equations, which are not sufcient to
determine uniquely the coordinates of both surfaces S
1
and S
2
. Another relationship must be generated, which
may be imposed on the amplitude distribution A(r), to
control the directivity or sidelobes. Alternatively, one may
impose conditions on the aperture phase errors. An im-
portant case is the reduction of phase errors due to the
source lateral defocusing. This will allow beam scanning
without excessive degradation in efciency and sidelobe
levels. In most cases, however, the problem is too complex
for analytic solution and a numerical approval must be
used.
4. ABERRATIONS
The term aberration, which originated in optics, refers to
the imperfection of lens in reproduction of the original
image. In antenna theory, the performance is measured in
terms of the aperture amplitude and phase distributions.
The phase distribution, however, is the most critical
parameter and inuences the far eld signicantly. It is
therefore used in evaluating the performance of aperture
antennas such as lenses and reectors. With a perfect lens
and a point source at its focus, the phase error should not
exist. However, there are fabrication tolerances, and mis-
alignments can occur that will contribute to aberrations.
Even without such imperfections, lens antennas can suffer
from aberrations. Practical lens feeds are horn antennas
and small arrays. Both have nite sizes and deviate from
the point source [2]. This means that part of the feed
aperture falls outside the focal point, and rays emanating
from them do not satisfy the optical relationships. Thus,
on the lens aperture, the phase distribution is not uni-
form. Similar situations also occur when the feed is moved
off axis laterally to scan the beam. Again, aperture phase
error occurs as a result of pathlength differences. A some-
what different situation arises when the feed is moved
axially, front or back. In this case, the phase error is
symmetric, as all the rays leaving the source with equal
angles travel equal distances and arrive at the aperture at
an equal radial distance from the axis, that is, on a
circular ring. But the length of the ray increases, or
decreases, with radial distance on the aperture. The phase
error is, therefore, quadratic on the aperture and reduces
the aperture efciency, while raising the sidelobes.
The general aberration (i.e., the lens aperture
phase error) can depend implicitly on both feed and
lens coordinates and be difcult to comprehend.
However, like all other phase-error-related problems, it
can also be represented as the pathlength difference with
a reference ray. For rotationally symmetric rays, the
natural reference is the axial ray. The pathlength differ-
ence can then be obtained by a Taylor-type expansion of
the general ray length in terms of the axial one. For small
aberrations the rst few terms in the expansion will be
sufcient to describe the length accurately. In terms of the
aperture polar coordinates r and f, the expansion
becomes
Lr; f L
axial
ar cos fbr
2
1 cos
2
f gr
3
cos f
23
where a, b, and g are constants indicating the magnitude
of each phase error. The leading term is linear in r and f,
then becomes quadratic, cubic, and so on, and the magni-
tude of each depends on the nature of imperfection caus-
ing the phase error. The even terms are caused by either
an axial defocusing or an axially symmetric error. The odd
terms can be due to a lateral displacement of the feed, or
asymmetric errors.
The effects of each error can be investigated by its
introduction in the aperture eld and determining the far
eld using a Fourier transformation or diffraction inte-
gral. For one-dimensional errors (i.e., r x and f0), the
effect can be understood easily, and has been investigated
by Silver [1]. The rst term is linear and in a Fourier
integral shifts, the transform variable. It thus causes a tilt
of the beam, but the gain remains the same. Using Silvers
notation, if f x is the aperture distribution and g(u) the
far eld, that is, its Fourier transform with a linear phase
error, one nds, with no phase error
g
0
u
a
2
_
1
1
f x expjux dx 24
Table 1. Amplitude Distributions
a
for the Hyperbolic and Elliptic Lenses of Figs. 3 and 5
Amplitude ratio
Ar
Ay
1
Ray Angle y
1
(degrees)
0 10 20 30 35 40 45 50
Hyperbolic lens equation [Eq. (21)] Relative value (dB) 1.0 0.928 0.733 0.466 0.328 0.196 0.084 0.008
0.0 0.65 2.70 6.64 9.70 14.17 21.5 41.75
Elliptic lens equation [Eq. (22)] Relative value (dB) 1.0 1.060 1.26 1.69 2.06 2.67 3.17 9.25
0.0 0.51 2.01 4.55 6.29 8.54 10.03 19.33
a
Where n1.6, e
r
2.55, y
1max
51.31.
and with phase error
gu
a
2
_
1
1
f x expjux ax dx g
0
u a 25
where upa=l sin y and a is the aperture length. Equa-
tion (25) shows that the beam peak is moved from the y 0
direction to y
0
calculated by
u a 0
or
y
0
sin
1
al
pa
_ _
26
A quadratic phase error is symmetric on the aperture
and does not tilt the beam, but reduces its gain. For small
values of b, it can be calculated analytically [1] and is
given by
gu
a
2
_
1
1
f x expjux bx
2
dx
a
2
g
0
u jbg
0 0
0
u
27
where g
0 0
0
u is the second derivative of g
0
(u). Because of
this phase error, the gain decreases progressively with
increasing b, and eventually the beam bifurcates and
maxima appear on either side of the axis. It also raises
the sidelobe levels. Figure 6 shows typical pattern degra-
dation due to this error.
The next important phase error is the cubic one that
has odd power dependence on the aperture coordinate.
However, this error not only tilts the beam but also
reduces the gain, and asymmetrically affects the sidelobes,
raising them on one side while reducing them on the
opposite side. Its effect is therefore a combination of the
effects of linear and quadratic phase errors. For small
errors its far eld is given by [1]
gu
a
2
_
1
1
f x expjux dx
3
dx
a
2
g
0
u dg
0 0 0
0
u
28
where g
0 0 0
0
u is the third derivative of g
0
(u). For a few
small phase errors the far elds of this phase error are
shown in Fig. 7. They show clearly the beam tilt, the gain
loss, and raising the sidelobes toward the beam tilt. They
are known as coma lobes, after the corresponding aberra-
tion in optics. Also, because this phase error causes more
severe pattern degradation than others, it should be
eliminated, especially as it manifests mostly in beam
scanning. Feed lateral displacements to scan the beam
can readily cause coma lobes. Fortunately, a number of
lens surface modications have been found to reduce the
effects of this error [3].
5. ZONED LENSES
So far, the equations used for lens designs equalized the
ray path lengths. The frequency of operation, or its
wavelength, did not enter the equations. Thus, in princi-
ple, they should function at all frequencies. However, the
directivity of a lens depends on the lens aperture size D,
and is often used for high gain applications. This results in
large lens sizes in wavelength, and at microwave frequen-
cies, in large physical sizes, both the aperture diameter D
and thickness T. It can, therefore, become excessively
heavy and difcult to use. Since the thickness of the lens
can be several wavelengths, it can be reduced along the
ray path in multiple wavelengths without altering the
relative phase change. The process starts at the edge,
where the thickness is zero. Moving down toward the axis,
the thickness increases progressively until it becomes one
wavelength. This thickness can be made zero without
Figure 6. Effect of quadratic phase error on the far-
eld pattern, reducing the mainbeam and raising
the sidelobes.
altering the phase. The process can be continued K times
until one arrives at the lens axis. In practice, one must
maintain a small thickness t
m
to provide adequate
mechanical strength, the value of which will depend on
the lens size, the material strength, and the application
type.
With zoned lenses, and neglecting t
m
, because the
thickness does not exceed one electrical wavelength, its
thickness is limited to l/(n1). Including the minimum
thickness t
m
, the total thickness is limited to t
m
l/(n1)
regardless of the number of zones. The pathlengths in
wavelength, however, are not equal. With K zones, the ray
path at the edge will be longer by a length equal to
(K1)l. This causes the frequency dependence of lens
operation, limiting its bandwidths. Enforcing the com-
monly used Silver criterion for this aperture phase error
[1] (i.e. 40.125l), the useful bandwidth of a lens with K
zones can be calculated from [1]
Bandwidth
25
K 1
percent 29
Equation (29) is valid for small variations of l and uniform
aperture distributions. For taper distributions, the effect
of phase errors is smaller and the actual bandwidth can
exceed that of Eq. (29).
Zoning the lens can cause one additional and severe
problem due to shadowing. Two adjacent rays from the
focus can travel through two separate zones, resulting in a
dark ring zone on the aperture. This occurs in the trans-
mit mode, and causes a loss of directivity and increased
sidelobe levels. In the receive mode, the energy falling on
the shadow zones never reaches the lens focus and dif-
fracts into the space, again causing reduction of gain and
increased noise temperature. Figure 8 shows the geometry
of a three-zone lens and shadowing due to R
1
and R
2
rays.
Zoning without shadowing is also possible, but should
be done on the nonrefracting surface of the lens. In a
hyperbolic lens, this should be done on the planar
backsurface. Shadowing will be eliminated, but phase
errors still occur at the transition lines due to diffraction
effects.
6. REFLECTION FROM LENS SURFACE
Because the wave impedance in air and the dielectric
medium of lens are different, reections occur for all the
rays. The reection coefcient depends on both the wave
polarization and the angle of incidence, namely, the angle
of ray with the local normal on the lens surface. Neither
can be avoided. With a linearly polarized wave, the
relative polarization, with respect to the plane of inci-
dence, changes from perpendicular to parallel, as the ray
direction rotates on the lens surface. However, their
reection coefcient behaves differently. For perpendicu-
lar polarization, it increases progressively with the inci-
dence angle but for parallel polarization, it decreases
initially, and after vanishing at the Brewster angle, in-
creases rapidly. Consequently, incidence angles must be
kept small, less than 301, to minimize the polarization
effects on the lens aperture distribution.
The surface reection effects can be reduced,
when warranted, by utilizing an impedance-matching
layer between the lens and free space. At normal and
small angles of incidence, the refractive index of the
matching layer can be found using a quarter-wavelength
transformer rule. It is the geometric means of the refrac-
tive index of the lens dielectric and that of air. In practice,
a different dielectric material may be used as the match-
ing layer, or it may be synthesized by preferentially
removing a fraction of the dielectric material from the
lens surface, such as drilling l/4 holes or cutting grooves
[4]. However, care must be taken to determine their
polarization effects.
The surface reections also inuence the impedance
mismatch at its feed. The problem is most severe in cases
where the lens surface is coincident with one of the
Figure 7. Effect of cubic phase error on the far-eld
pattern, causing beam tilt and asymmetry in the
sidelobes.
equiphase surfaces, namely, the wavefront. Then, the
entire reected wave travels back to the feed, the degree
of which depends on the lens refractive index. At normal
incidence, since the reection coefcient is |R|(n1)/
(n1), the reected power is unacceptably large for all
common dielectrics, and a matching surface should be
used. In the event that a matching layer cannot be used,
the reection effects on the feed can be minimized by
lateral defocusing of the feed, or retuning of the feed over a
narrow bandwidth.
7. LENSES WITH no1
Lens equations (1) to (6) were developed without specify-
ing the value of the refractive index, and are therefore
valid for no1 cases as well. However, the lens surface
becomes inverted. For instance, the hyperbolic lens equa-
tion [Eq. (13)] for no1 modies to
r
1
1 nF
1 n cos y
1
30
and the lens surface becomes elliptical, concave toward
the focus, similar to Fig. 1b. On the inner region a
minimum thickness t is required to provide mechanical
strength. Zoning is also possible and will cause shadowing
when incorporated on the actively refracting surface. The
bandwidth limitations due to n remains the same as the
dielectric lenses with n41. However, the lens media for
no1 such as metal plates and waveguides are usually
frequency-sensitive and exhibit narrower bandwidths.
8. CONSTRAINED LENSES
The function of a lens is to modify the phasefront
of an incident wave, say, from spherical to planar.
In practice, this may be accomplished by means
other than the dielectric lenses. In most general
cases, the lens surfaces consist of a plurality of receiving
and radiating elements, interconnected by processing
elements. The received signals on one surface are modied
in amplitude and phase and reradiated from the elements
of the next surface. In passive designs, the interconnection
is due to transmission lines, such as parallel plates,
waveguides, and even coaxial lines. The design process
is similar to the dielectric lenses and is governed by the
pathlength equation. Snells law, however, is not satised
at all surfaces, and the problem of surface reection and
transmission must be solved using the wave equation.
Nevertheless, lenses can be designed with similar
surfaces, but with inverted curvature, as the dielectric
lenses [3].
The simplest case uses parallel plates, with spacing a,
between 1l and 0.5l. When the electric eld is parallel to
the plates, a non-TEM waveguide mode is excited and has
a wavelength lp given, in terms of the free-space wave-
length l, by
lp
l
1
l
2a
_ _
2
_ _
1=2
31
Figure 8. Geometry of a zoned
lens with shadowing effects.
which can be used to dene an equivalent refractive index as
n
l
lp
1
l
2a
_ _
2
_ _
o1 32
In cylindrical lenses, when the plates and electric eld are
normal to the cylinder axis, Snells law of refraction
governs the transition between the lens and outside
media. But when they are parallel to the cylinder axis,
the incident rays are constrained to pass between the
plates and Snells law is not satised [1].
An example of the rotationally symmetric constrained
lens is the planarelliptic surface lens of Eq. (30).
It is usually zoned to reduce its size and weight [4]. Other
useful transmission media are the rectangular and square
waveguides, operating in TE
10
or TE
01
modes. The wave-
guide dimensions must be such that only these modes can
propagate and higher-order modes are suppressed. The
square waveguide can be used for circularly polarized
applications, otherwise, must be avoided to reduce cross-
polarization.
9. INHOMOGENEOUS LENSES
In the lenses studied so far, the refractive index n was
constant and the shape was proled to satisfy the ray path
condition. On the other hand, if the lens shape is kept
xed, then another parameter, such as the refractive
index, must be allowed to change in order to help in
collimating the beam. This is achieved in a family of
lenses, the most important ones of which are spherical
in shape, such as Luneberg lens, Maxwells sh-eye, and
Eaton lenses. Their spherical shape provides a perfect
three-dimensional symmetry, useful in applications such
as the wide-angle scanning. They also have only a radial
inhomogenity, making them both physically and electri-
cally symmetric.
9.1. Luneberg Lenses
The term Luneberg lens refers to a family of lenses with
two axial foci. They can be both outside the lens or one
inside and the other outside. The most useful case, how-
ever, is the lens with one focus on its surface, while the
second one is at innity. Thus, an axial point on the lens
surface is focused to an axial point at innity, on the
opposite side of the lens. The refractive index of this lens is
given by
nr 2
r
a
_ _
2
_ _
1=2
33
where a is the lens radius and r is the radial distance of a
point inside the lens. At the origin, the refractive index is
no
2
p
, and on its surface it becomes unity. Both are
practically signicant. The refractive index values and
variations are in reasonable range, and the lens can be
synthesized. Moreover, the unity of its refractive index on
the surface eliminates the impedance mismatch and,
consequently, the surface reections. The geometry and
ray paths of this lens are shown in Fig. 9, with a feed
horn on its surface. Scanning the feed on its surface
scans the radiated beam, without alteration. The scan
limit is set only by the mechanical limitation of the feed
horn motion. With a spherical conducting cap on its sur-
face the lens also acts as a perfect reector (i.e., a back-
scatterer; Fig. 10). The main difculty with this lens is its
fabrication problems. Multilayer shells are normally used
to synthesize the refractive index inhomogenity. Figure 11
shows one case, where 10 layers are used to construct an
18-in. diameter lens. While the approximation to a con-
tinuously variable refractive index is reasonable, the wave
scattering at the layer transitions reduces the lens ef-
ciency.
With the abovementioned refractive index, the
Luneberg lens performance is ideal at the geometric
optics limits, when the lens diameter in wavelength is
large. At microwave frequencies, the wavelength is large
and the lens diameter in wavelength may not be large.
Its performance, namely, directivity, and sidelobe levels
deteriorate rapidly. In such cases, the refractive index
prole can be modied to improve its performance. This
can be done by determining the excitation efciencies of
various spherical modes and calculating its far eld and
directivity [5]. The new dielectric permittivity prole is
dened as
e
r
n
2
2B A
2
r
a
_ _
2
34
the constant parameters A and B are determined to
maximize the gain. Three different cases are identied
and investigated. Their refractive index proles are shown
Feed antenna
Lens
Figure 9. Typical ray paths in a Luneberg lens.
in Fig. 12. For case (a), AB1, which is the ordinary
Luneberg lens. For prole b, A1 and B40, and the
dielectric prole is the same as Luneberg lens with a
constant increase given by (2B 2). In Fig. 12 prole b is
for B1.1. For prole c, B1 and Ao1, which increases
the lens permittivity at its surface (A
2
0.95 in Fig. 12).
For prole d A
2
B, and the dielectric prole falls between
proles a and b: (A
2
B1.1 in Fig. 12). It gives a lens
permittivity close to unity at its surface, which will
improve its impedance match to free surface. Proles b
and c give larger refractive indexes and are expected to
perform better at lower frequencies. This is investigated
Figure 10. A passive Luneberg lens reector re-
turning the incident rays.
Figure 11. Multilayer spherical shell construction of a Luneberg
lens.
2.4
2.2
2.0
1.8
D
i
e
l
e
c
t
r
i
c

c
o
n
s
t
a
n
t
1.6
1.4
1.2
0 0.2 0.4
Radial distance (r/a)
0.6 0.8 1.0
b
d
a
c
Figure 12. Refractive index of modied Luneberg lenses to
improve low-frequency performance.
using the spherical harmonics, and the results for the
directivity, sidelobe levels, and beamwidths are shown in
Table 2.
9.2. Constant n Spherical Lens
The difculty with fabrication of the inhomogeneous
lenses encouraged investigators to search for quasi-ideal
spherical lenses with constant refractive index. An inter-
esting case is a lens with e
r
3, studied earlier by Beke
and Farnell [6] and more recently by Mason [7]. With a
Huygen source feed at its surface, the computed phase
distribution across its aperture, for different relative
permittivities, is shown in Fig. 13. For e
r
3, the phase
error is below 101, across about 60% of the aperture. It
remains within acceptable range for gain calculated over
at least 70% of the aperture, resulting in excellent gain
performance over a wide range of frequencies. The only
drawback seems to be the excitation of internal modes at
their resonance. Their effect reduces with the loss tangent
of the dielectric material.
10. DIELECTRIC-LOADED HORNS
Horn antennas are among the most useful and versatile
antennas. They have a relatively simple shape and are
easy to fabricate and use. They are used as test antennas,
feeds for reector and lens antennas, or independently
as communication antennas. Because of their diverse
applications, their electrical specications vary consider-
ably. As test antennas, they are used as gain standards
and required to have good polarization isolation in
the principal E and H planes. Rectangular horns are
commonly used for this application to simplify the
polarization denition and gain calculation. As a feed
for reector and lens antennas, the requirements are
signicantly different. While having a nite aperture
size, they must behave as a point source, have small
sidelobes and backlobes to minimize power spillovers,
and have negligible cross-polarization in the entire radia-
tion zone. To achieve such stringent requirements, their
design must be precise and an accurate solution must be
known to assess their performance. This is more so with
circular horns and consequently have found more
widespread applications as feeds than rectangular ones.
Electromagnetic analysis, however, has shown that
conventional smooth-walled horns cannot provide
radiation patterns with acceptable polarization purity
and low spillover. Corrugated horns are developed for
these applications, but are costly and narrowband.
Dielectric loading of the horn has been shown to improve
the performance and in certain applications may be used
to replace corrugated ones.
In applications where horn antennas are used as in-
dependent communication antennas, gain and aperture
efciency may be the fundamental parameters to optimize.
However, to obtain high gain, the horn aperture size must
increase, which also increases the aperture phase errors.
The phase errors can be maintained low by using small
Table 2. Performance Parameters of Modied Luneberg Lens
Luneberg Lenses Modied Luneberg Lenses
AB1 A1 A
2
B
Diameter
(l)
B
Value
Gain
(dBi)
Beamwidth
(deg)
First
Sidelobe
Level (dBi)
Gain
(dB)
Beamwidth
(deg)
First
Sidelobe
Level (dB)
Gain
(dBi)
Beamwidth
(deg)
First
Sidelobe
Level (dB)
2 1.4 14.79 30.2 14.41 17.56 23.5 17.15 16.85 24.0 14.79
4 1.16 20.761 15.1 16.05 22.70 13.0 16.9 22.0 13.25 16.1
6 1.1 24.34 9.8 16.9 25.75 9.0 16.97 25.17 9.1 16.4
8 1.075 26.90 7.3 17.1 27.98 6.7 17.01 27.56 7.0 16.6
10 1.04 28.78 5.8 16.35 29.35 5.5 15.97 29.26 5.6 16.81
0.8 0.7 0.6 0.5
Normalized effective radius
0.4 0.3 0.2 0.1 0
60
50
40
30
20
10
0
N
o
r
m
a
l
i
z
e
d

p
h
a
s
e

(
d
e
g
)
10
20
30
40
50
60
c
r
= 1.5
c
r
= 2.0
c
r
= 2.5
c
r
= 3.0
c
r
= 3.5
c
r
= 4.0
c
r
= 6.0
Figure 13. Phase across aperture of a constant n spherical lens.
cone angles, but this increases the horn size. A convenient
solution is then to use a lens at the horn aperture to
reduce or eliminate the phase errors, by collimating the
beam. Consequently, compact high-gain horns can be
designed with controlled aperture phase and amplitude
distributions, to improve the aperture efciency and horn
gain. Alternatively, lenses can be used to suitably modify
the aperture distribution in both amplitude and phase to
shape the radiation patterns.
In this section, initially the dielectric-loaded and lens-
corrected horns will be discussed. Then, the use of
dielectric in small antennas such as waveguides, microstrip
antennas, and dipoles, will be considered, to improve their
operation in specic applications.
10.1. Dielectric Loading
Historically, dielectric-cone loading inside smooth-walled
conical horns was used by Clarricoats et al. [8], and Lier
[9] to simulate the effect of corrugations. Corrugated
horns, with quarter-wavelength corrugation depths, can
support hybrid HE
11
mode. This mode radiates with low
cross-polarization and can be designed to have negligible
sidelobes. Introduction of the cone dielectric, with an
airgap as shown in Fig. 14 inside a smooth-walled horn,
was also shown to support hybrid modes and improve the
performance. Clarricoats et al. [8] used low-dielectric-
constant materials, such as foams with a relative permit-
tivity of 1.13. But, in Liers work [9], solid-dielectric cones
with a relative permittivity of 2.5 was used, again showing
good performance. Both groups of investigators also ana-
lyzed these dielectric-loaded horns using modal expan-
sions, and studied the effects of the airgap, horn
permittivity, aperture diameter, are angle, and the throat
region. Airgap size was found to be strongly dependent
on the aperture diameter, and both are dependent on
the dielectric permittivity. The airgap size generally in-
creases with the horn diameter, and for a given diameter
there is a minimum relative permittivity of dielectric to
support the hybrid mode to minimize the cross-polariza-
tion. Both are angle and the throat region have similar
inuences. Large are angles, and asymmetric throat
region design, excite higher-order modes and thus
increase cross-polarization.
A variation of the conical dielectric-loaded horn is
shown in Fig. 15. Its wall is proled. A large are angle
near its throat reduces its axial length and results in a
compact horn. Then, its small are angle near the aper-
ture improves the cross-polarization. The prole is de-
scribed by the following equation
rz r
th
3Dr 1
2z
3L
_ _
z
L
_ _
2
Dr r
ap
r
th
35
where r
ap
and r
th
are the horn radii at the antennas
aperture and throat. A prole horn of this type was
designed and optimized. Its performance is compared
with the linear horn in Table 3. Its cross-polarization is
improved by 4 dB. The effect of length reduction on the
performance of the abovementioned prole horn is also
shown in Table 4. It shows that the performance remains
steady and comparable to a linear horn for length reduc-
tions by as much as 22%.
10.2. Lens-Corrected Horns
In high-gain horns, the aperture diameter in wavelength
is large, and the horn length can be excessive, unless its
are angle is made large. But the combination of large
aperture size and large are angle can cause severe
(a)
(b)
Figure 14. Geometry of a dielectric-loaded horn showing two
possible dielectric insertions. Figure 15. Geometry of a dielectric-loaded prole horn.
aperture phase error. This problem can be remedied by
using a lens at the horn aperture. Figure 16 shows three
possible options. These simple lenses and others, includ-
ing zoned lenses, may be used, and would correct the horn
aperture phase distributions. But each lens will have
different inuences on the aperture amplitude distribu-
tion. The properties of the rst two lenses were investi-
gated earlier, and Table 1 showed their effect on the
amplitude distribution. Type (a) (in Fig. 16) increases
the amplitude taper according to Eq. (21) and will reduce
both sidelobes and the aperture efciency. Type (b) will
compensate for the amplitude taper, and according to
Eq. (22), the lens permittivity can be used to control the
aperture distribution, and thus the horn efciency and the
pattern sidelobes. For type (c), an analytic expression is
not available and a numerical procedure must be used.
However, as was indicated earlier with respect to this lens,
both surfaces help in collimating the beam, but its second
surface is similar to type (b) lens and its inuence on the
aperture distribution will be similar as well. With a
hybrid-mode horn, corrugated or dielectric-loaded, the
resulting aperture distributions for different lens relative
permittivities are shown in Fig. 17, which shows that for e
r
around 1.22, the aperture amplitude distribution is nearly
uniform.
11. DIELECTRIC-LOADED WAVEGUIDES
Waveguides have small aperture size and are not as
efcient radiators as horns. Part of the energy leaks out
and induces current on the outside wall, which radiates
laterally and backward, causing large backlobes. The
wave impedances of waveguide modes are also different
from the free-space intrinsic impedance, and strong reec-
tions can occur on the aperture, causing poor input im-
pedance match. These problems can be partly overcome by
aring the waveguide at its aperture. However, similar
and even better performance can be obtained by loading
the waveguide by a short section of a dielectric. The size
and shape of the dielectric constant provide several para-
meters that can be used to shape the radiation patterns
and tailor them to the desired specications. Table 5
shows the results for three different end loadings, and
the type of performance variations one could achieve [2].
Two other examples are shown in Figs. 18 and 19, with
combinations of dielectric and cavity loadings [2]. In
Fig. 18, the end geometry is optimized for nearly perfect
pattern symmetry, with negligible cross-polarization. Fig-
ure 20 shows its copolar and cross-polar radiation pat-
terns. In Fig. 19, the combination was again optimized for
a heavily shaped radiation pattern, again with negligible
cross-polarization in the forward direction. It is an ideal
feed for deep parabolic reectors with small f/D0.25. It
provides high aperture efciency of 81% due to its front
pattern null, very low cross-polarization, and extremely
Table 3. Performance
a
of Dielectric Loaded Linear and
Proled Horns
Parameter Linear Horn Proled Horn
3 dB beamwidth (deg) 14.8 13.7
10dB beamwidth (deg) 26.9 24.8
Directivity (dBi) 22.1 22.5
Efciency (%) 61.8 68.1
Peak cross polar (dB) 32.2 36.0
VSWR 1.04 1.03
a
Where R
th
1.14 cm, r
up
27.7 cm, L30.9 cm, e
r
1.13, airgap1.2 cm.
Table 4. Performance of Prole Horn with Length
Reduction
Length (cm)
Peak
Cross-Polarization
(dB)
3dB
Beamwidth
(deg) Efciency (%)
30.9 36.0 13.7 68.1
27.5 36.8 13.8 64.4
24.0 31.6 14.0 57.1
15.0 27.6 16.1 32.0
0
max
0
max
F
F
(r, z)
T
D
j
z
r
F T
0
(a)
(b)
(c)
Figure 16. Three examples of lens types for loading horn
aperture.
low noise temperatures due to small f/D, the focal length:
diameter ratio.
12. MICROSTRIP AND DIELECTRIC RESONATORS
Microstrip antennas are discussed in a separate article
(see MICROSTRIP ANTENNAS), and usually consist of a con-
ducting patch separated from a ground plane by a di-
electric substrate. They are low-prole and increasingly
popular antennas for practically any type of application.
Their radiation patterns, however, are asymmetric with
unequal E- and H-plane patterns. But, with careful opti-
mization, the pattern symmetry can be achieved to
minimize cross-polarization. Figure 21 shows a case of
stacked patches with a side choke for equalizing the
principal-plane pattern, low backradiation, and cross-
polarization. Similar performance can also be obtained
using a dielectric resonator in lieu of a microstrip patch.
The dimensions of the dielectric resonator are related to
the wavelength by
d
1:841l
4np
16
pd
1:841h
_ _
2
_ _
1=2
36
The excited mode is the TM
110
mode, and produces radia-
tion similar to that of a microstrip patch. In Fig. 22, the
resonator and the cavity are optimized for symmetric
pattern in the principal planes to reduce the cross-polar-
ization. They are shown in Fig. 23, with excellent sym-
metry. Both the microstrip and resonator antennas can be
used as efcient reectors and lens feeds with high
aperture efciency and low cross-polarization.
13. INSULATED ANTENNAS
Practically all antennas have conducting parts, but in
certain families of antennas, especially small resonant
18
12
6
0
d
B
6
14
1.1
1.2
1.23
1.5
30 20
Horn alone
0, deg
10
Figure 17. Aperture amplitude distribution for a lens corrected
horn 301 semiare angle hybrid-mode horn, type c lens.
Figure 18. Geometry of a dielectric cavity-loaded waveguide
feed.
Table 5. Performance of Dielectric-Loaded Waveguide with Shaped Dielectrics
Half-Beamwidths
3dB 10dB
Geometry
Peak Cross-Polarization
0y90
(dB) Gain (dBi) E plane H plane E plane H plane

a
60
0.519z
33.95 8.28 36.82 36.18 71.47 72.51
b
0.1z
60
0.6z
24.74 8.11 37.21 38.32 73.42 71.35
c 0.6z
0.619z
24.43 13.47 19.43 20.25 33.13 35.17
d0.6l, e
r
2.5
ones, the conduction current radiates directly. Typical
examples are the wire antennas and microstrip antennas
that are often half-wavelength resonators. In wire anten-
nas, the current is excited by the applied voltage directly
on the wire, which radiates in the surrounding space. In
microstrip antennas, the currents are both on the patch
and its ground plane, which are separated by a dielectric
substrate. For this reason, only the patch current is
exposed to the surrounding medium. However, in either
case, the physical constants of the medium are excessively
lossy, and can short-circuit the antenna current and
prevent its operation. In practice, this problem can occur
in remote sensing and biological applications. In the
former case, the antennas may be buried underground,
or submerged in sea and ocean waters that have high
electrical conductivities. In the latter case, the antennas
are implanted into various types of body tissues that can
have excessively high conductivities. In such cases, to
ensure antenna operation, the conduction currents must
be insulated from the surrounding conducting medium. A
simple but effective method is to use a thin dielectric
coating on the antenna conductor carrying the radiating
currents. The coating will provide insulation between the
0
R
e
l
a
t
i
v
e

p
o
w
e
r

o
n
e

w
a
y
,

d
B
8
16
24
32
40
180 135 90 45 0
0, deg
45 90 135 180
Figure 21. Geometry and radiation patterns of a stacked micro-
strip feed, with peripheral choke to minimize back radiation.
0
8
16
24
32
40
R
e
l
a
t
i
v
e

p
o
w
e
r

o
n
e

w
a
y
,

d
B
0 36 72 108 144 180
0, deg
X
Z
0
Figure 19. Geometry and radiation pattern of a shaped dielectric
and cavity loaded waveguide feed.
0
0 60 120 180
10
R
e
l
a
t
i
v
e

p
o
w
e
r

(
d
B
)
20
30
40
0
E-plane
H-plane
cross-p.
Figure 20. Radiation patterns of the waveguide feed of Fig. 18,
showing perfect pattern symmetry and near negligible cross-
polarization.
Z
d
h
D
H
3t
Figure 22. Geometry of a dielectric resonator antenna, showing
a conducting cavity of height H and diameter D, with a dielectric
disk of height h and diameter d.
conducting antenna and the medium, thereby eliminating
the conduction current. The excitation energy will then
transfer into the pointing vector, leaving the antenna.
The behavior of the insulated antennas in a medium of
complex permittivity differs considerably from that in free
space, and should be analyzed carefully. For instance,
consider a conventional dipole of length 2h, as shown in
Fig. 24. The wire is a good conductor and has a diameter of
2a, insulated by a cylindrical dielectric region of diameter
2b and propagation constant k
1
, located in an innite
exterior region of k
2
. With a thin-wire approximation,
the dipole current can be represented by a sinusoidal
distribution of the form described in Ref. 10. The time
factor is assumed to be exp(jot)
Iz
jV sin k
L
h jzj
2Z
ca
cos k
L
h
37
where
k
L
k
1
1
H
2
0
k
2
b
k
2
bH
2
1
k
2
b ln
b
a
_
_
_
_
1=2
38
Z
ca
B
1
k
L
2pk
1
ln
b
a
_ _
39
B
1
om
0
k
1
40
k
1
om
1
e
1
1=2
41
and H
0
(2)
and H
1
(2)
are Hankel functions of zero and rst
order. Note that with a perfect insulation dielectric k
1
is
real but k
2
is complex due to the presence of Hankel
functions in Eq. (38). It reduces to k
1
when b, the radius
of the insulation, becomes innitely large. In view of
Eq. (38), the dipole current distribution, input impedance,
as well as the radiation resistance, and the resonance
frequency can depend strongly on the radius b and
propagation constant k
1
, and k
2
, the propagation constant
of the exterior region. The latter may not be fully known,
or constant, during the application because of variations
in moisture content and other variables. Thus, the insula-
tion parameters should be selected appropriately to mini-
mize the dependence of k
l
on k
2
.
14. MEDICAL AND BIOLOGICAL ANTENNAS
Another area in which insulated antennas play an im-
portant role is the biological and medical applications.
They can be noninvasive (i.e., not penetrating the body) or
invasive. In either case, the properties of insulated anten-
nas can be signicantly different from those in free space.
Thus, care must be taken in their design and analysis to
ensure adequate power transfer to the right tissue. Non-
invasive radiators are often dielectric-loaded waveguides
and horns, discussed in the previous section. The dielec-
tric loading in this case is used to improve impedance
matching and coupling to the body. Their design is not
signicantly different from those of other dielectric-loaded
waveguides, except that the end shaping must prevent
hotspots and improve penetration.
Microstrip antennas and arrays are other types of
radiators suitable for noninvasive applications. However,
their resonance property and power coupling to the body
can be sensitive to the extent and nature of contact to the Figure 24. Geometry of an insulated dipole antenna.
0
R
e
l
a
t
i
v
e

p
o
w
e
r

(
d
B
)
10
20
30
40
180 90 0
0
90 180
E-plane
H-plane
cross-p.
Figure 23. Radiation patterns of the dielectric resonator an-
tenna.
skin. Dielectric coating over the radiating patch or slot can
insulate the antenna and minimize the bodys inuence.
This is due to the fact that, in microstrip antennas, the
resonance depends on the effective dielectric constant, and
not the actual substrate permittivity. With single-layer
substrates of thickness h, this effective permittivity, for a
conductor linewidth of W, is given by
e
eff
e
r
1
2

e
r
1
2
_ _
1
12h
w
_ _
1=2
42
However, it can change signicantly by introducing a
higher permittivity layer over the substrate. Conse-
quently, in biological applications, where the tissue rela-
tive permittivity can be excessively high due to the water
content having e
r
D80, the nature of the proximity or
contact with the body can alter e
eff
signicantly [11]. Since
microstrip antennas are narrowband, or at best not wide-
band, the efciency of their radiation and coupling to the
body can be deteriorated. The effect can be reduced by
introducing a superstrate layer over the microstip an-
tenna, to control the relative permittivity variations.
Invasive-type radiators can produce more uniform and
controllable heating patterns, but they require implanta-
tion in the tissue. The most convenient types are the
insulated needle radiator, basically the end of the coaxial
line. However, this type of antenna can generate strong
currents on the outer coaxial conductor and cause tissue
heating behind the antenna. An improvement can be
obtained by introduction of a quarter-wavelength choke
over the coaxial conductor to form a sleeve antenna. Their
analysis and sensitivity study can be carried out similar to
the insulated dipole antennas. Figure 25 shows the geo-
metry of needle and sleeve antennas.
15. NRD WAVEGUIDE ANTENNAS
A nonradiative dielectric (NRD) waveguide, shown in Fig.
26, consists of a dielectric slab sandwiched between two
parallel conducting plates. It is known to have low resis-
tive losses at high frequencies and useful structure for
design of low-loss circuit components, such as couplers,
lters, and even ampliers [12]. If an antenna can be
designed using an NRD guide, it will be low-loss, and can
be integrated with other NRD components. The resulting
system of circuitantenna combination will therefore be
compact and provide high operating efciency, useful in
many high-frequency applications. However, in an NRD
the eld propagates within the dielectric slab and is cut off
between the parallel plates outside the dielectric, where it
attenuates exponentially. To cause radiation, therefore,
one must expose its guided eld to the external region.
This can be accomplished in three different ways: (1) by
cutting a slot on one of the conductors over the dielectric
slab, (2) by terminating the parallel plates a short distance
from the dielectric slab, and (3) by terminating the
dielectric slab in free space.
In the rst antenna type, cutting an aperture or slot on
one of the conducting plates over the dielectric disturbs its
guided eld and causes the radiation [12,13]. However, the
slot also causes discontinuities in the NRD guide that
excites the dominant parallel-plate mode, which is not
low-loss. In the second antenna type, the attenuating eld
between parallel plates leaks outside and radiates like a
leaky-wave antenna [14,15]. This is a continuous radia-
tion along the guide, and its radiation beam squints with
frequency. The third antenna type radiates from the open
end of the guide, and provided its termination geometry
does not differ from that of the guides cross section, it does
not generate other modes [16]. It is therefore a well-
behaved antenna and its radiation is due primarily to
the guided mode of the dielectric. Its operation is dis-
cussed below.
Figure 26 shows the cross-sectional and end views of a
conventional NRD guide [12]. The principal transverse
eld distributions are shown in Fig. 27, and in region I,
which is the dielectric region, these elds are given by
E
y
1
K
2
e
r
b
2
y
jom
0
e
0
e
r
cos
px
a
cosb
y
ye
jb
g
z
43
H
x
2
jb
g
m
0
cos
px
a
cosb
y
ye
jb
g
z
44
Figure 25. Implantable radiator types: (a) needle radiator;
(b) sleeve antenna. Figure 26. Geometry of NRD guide.
Between the parallel plates, in region II, the correspond-
ing elds are
E
y2
cos
b
y
b
2
_ _
k
2
a
2
y
jom
0
e
0
cos
px
a
_ _
e
ayjyjb=2
e
jb
g
z
45
H
x2
jcos
b
y
b
2
_ _
b
g
m
0
cos
px
a
_ _
e
a
y
jyjb=2
e
jb
g
z
46
The guided eld parameters in these equations are com-
puted from the following transcendental equations:
b
2
g
k
2
0
e
r
pa
2
b
2
y
47
k
2
0
e
r
pa
2
a
2
y
48
b
y
tan
1
2
b
y
b e
r
a
y
49
k
2
b
2
y
k
2
0
a
2
y
50
Equations (43) to (46) can be used to dene the equiva-
lent electric and magnetic currents on the truncated
aperture of Fig. 26, which can then be used to determine
the radiation integrals [17]. The resulting far radiated
elds in the principal planes are
E
yjp=2
E
0
y
1 z
1
cos y G1 z
1
cos y
b
2
sinb=2k siny b
y
b=2
b=2ksiny b
y
b=2
_
sinb=2k siny b
y
b=2
b=2k siny b
y
b=2
_
E
0
y
1z
2
cos y
G1 z
2
cos y
2e
r
cosb
y
b=2
a
2
y
k
2
sin
2
y
_ _
a
y
cos
b
2k sin y
_ _
k siny sin
b
2k sin y
_ _ _ _
51
E
jjp=2
E
0
j
cos y z
1
Gcos y z
1
1
b
y
sin
b
y
b
2
_ _
_
cos y z
2
Gcos y z
2
2e
r
a
y
cos
b
y
b
2
_ __
coska=2 siny
ka=2 siny
2
p=2
2
_ _
52
with
Z
g1
k
2
e
r
b
2
y
ke
r
b
g
Z
Z
z
1
53
Z
g2
k
2
a
2
y
kb
g
Z
Z
z
2
54
where Z is the free-space impedance. The parameter G is
the reection coefcient of the NRD guide radiating
aperture, calculated using its wave impedance in
Eqs. (53) and (54) and the free-space impedance.
An example of the radiation patterns is shown in Fig. 28.
The dielectric region dimensions are 15 10.16 mm, and
the relative permittivity is 2.55. At 9.5 GHz the impedance
of the open-ended NRD guide is z 0.5940j0.5520,
which gives a reection coefcient of G 0.1203
j0.3380. Other propagation parameters are shown in
Table 6 [16]. Figure 28 shows an excellent agreement
between the computed and measured radiation patterns,
indicating that the single-mode operation of the NRD
guide for this antenna type is a good assumption. The
radiation patterns are also well dened and have a beam
peak in the forward direction. The antenna is therefore a
suitable candidate for single-element use or high-gain
array applications.
16. ANTENNA MINIATURIZATION
Antenna miniaturization is an important issue in certain
communication and mobile applications. Dielectric load-
ing can be a useful tool for such applications, because for
natural dielectrics the relative permittivity of the material
is larger than unity. The velocity of electromagnetic waves
Region II Region II
Region I
Region II Region II
E
y
y y
b/ 2
a/ 2 a/ 2
b/ 2
x
Figure 27. Fields in dielectric and air regions of
NRD guide.
and their wavelengths within the dielectric medium,
therefore, reduce by the square root of the relative per-
mittivity. Since the waveguide and antenna dimensions
are in terms of the wavelengths, they also become smaller
inside dielectrics. Thus, the phenomenon can be used
effectively for reducing the dimensions of the antennas.
In certain applications, such as microstrip antennas, this
is common knowledge, and for the case of dielectric
resonators, it was discussed in Section 12. For other
applications the case must be handled accordingly. An
important issue is realization of the fact that the normal
wavelength reduction by the square root of the relative
permittivity occurs only inside innite dielectric regions.
In other cases, the effect is less and consequently an
effective dielectric constant e
eff
is normally dened. For
microstrip substrates this has been determined analyti-
cally, and was provided in Eq. (42). In most cases, however,
the problem must be solved numerically. Here the case of a
dielectric-coated monopole is discussed. Monopoles have a
simple geometry and are important antenna candidates,
but must be installed vertically, making them a tall
structure. A reduction of their length can be very desirable
in many applications.
Figure 29 shows the geometry of a dielectric coated
monopole [18]. The dielectric material is a cylindrical
ceramics of relative permittivity of e
r
38, covering a
conducting tube of diameter d
in
and height 6.4 mm. The
conguration is placed over an innite ground plane and
solved numerically using a nite-element method to de-
termine its resonant frequency, impedance, return loss,
and the bandwidth. The computed results are compared
with measurements in Fig. 30 for the resonant frequency,
and in Fig. 31 for the return loss.
Aside from a small frequency shift, the agreement is
good. Figure 30 shows that, as the coating thickness
increases, the effect of the dielectric increases. This is
manifested in the form of reduced resonant frequency as
the monopole diameter is decreased, while keeping the
coating diameter constant. In the limit of smallest mono-
pole diameter, the resonant frequency is about 4.7 GHz, a
reduction of about 2.4, as compared to its resonant fre-
quency in free space. The effective permittivity is there-
fore about 5.8, which is much smaller than 38, the actual
relative permittivity of the dielectric material. For larger
effective permittivity values, one must either increase the
dielectric thickness or increase its length beyond the
monopole end. Since both of these solutions result in an
increased size of the overall structure, a compromise must
be made in any practical application.
17. GAIN ENHANCEMENT
Dielectric loading, when used judiciously, can be an effec-
tive means for increasing the gain of an antenna. The
focusing effect of dielectric lenses was discussed in Section 2.
This section addresses the case of planar dielectric
loading. Lenses are not desirable solutions in some
applications because of the high cost of fabrication
or manufacturing tolerances at millimeter-wave frequen-
cies. Simple planar dielectric layers are more desirable to
90
75
60
45
30
15
0
0
15
30
45
60
75
90
0 10 20 30 40
Relative amplitude (dB)
Calculated
Measured
E-Plane pattern H-Plane pattern
Figure 28. Measured and calculated radiation pat-
terns of NRD guide.
Table 6. Propagation Constants of NRD Guide
k
0
198.97
a
y
138.40
b
y
205.44
b
z
121.98
8.0mm
6.4mm
d
in
Ceramics (c
r
= 38)
Resonator
Solder
Ground plane
Connector
Figure 29. A cross section of the dielectric-coated antenna.
use. They are natural in microstrip structures and cost-
effective forms in most radome applications. When used
properly, they can increase the antenna gain in proportion
to their relative permittivity, compared to adjacent
regions.
Figure 32 shows the geometry of a typical dielectric
covered region. A planar dielectric layer of thickness t in
region II is placed over another layer of thickness H in
region I, which is over a conducting ground plane. Inside
region I an antenna element is represented by an electric
current I
0
, parallel to the ground plane and at a distance
h. This problem was investigated by Jackson and
7
6
5
4
1 2 3
d
in
(mm)
4 5
R
e
s
o
n
a
n
t

f
r
e
q
u
e
n
c
y

(
G
H
z
)
Measured
Calculated
Figure 30. Comparison of the calculated and measured resonant
frequencies of antenna in Fig. 29.
Measured
Calculated
0
5
10
R
e
t
u
r
n

l
o
s
s

(
d
B
)
15
20
4.5 5.0
Frequency (GHz)
5.5
Figure 31. Return loss of the dielectric-coated monopole antenna
(d
in
3.2mm) plotted against frequency.
Antenna
ground plane
h
t
H
n
1
=
n
2
=
z
I
o
x
Region II
Region I
2
Figure 32. Antenna embedded in a two-layer
planar dielectrics.
r t
H
D
microstrip
patch
microstrip
patch
dielectric
radome
cavity
Figure 33. Geometry of a radome-covered cavity antenna.
Alexopoulos [18] using a transmission-line approximation.
They have shown that the optimum thicknesses for
maximizing the gain are given by
n
1
H
l
0
m
2
; m1; 2; 3; . . . 55
n
2
t
l
0
2n 1
4
; n1; 2; . . . 56
n
1
h
l
0
2p 1
4
; p1; 2; . . . 57
and for this optimum thickness relationships the gain is
approximately
Gain8
H
l
0
e
2
e
1
_ _
58
This indicates that, if region I is an air medium, then
the antenna gain increases proportionally to the relative
permittivity of region II. For example, if the relative
permittivity of region II is selected to be 100, then the
antenna gain will increase by about 20 dB, over its gain
without the presence of the dielectric layer. However, this
gain increase is at the expense of the antenna gain
bandwidth, which decreases inversely in terms of the
relative permittivity.
The antenna spacing of Eq. (57) is not a hard rule and
can be relaxed without a signicant gain degradation.
This allows the use of the other two parameters in
Eqs. (55) and (56) to design radomes for gain enhancements.
In an earlier study [19], the author placed a microstrip
antenna at the bottom of a cavity that had a diameter
D2l. A radome was used to cover the cavity. From
Eqs. (55) and (56) the optimum radome thickness is a
quarter of a dielectric wavelength, and the optimum
cavity height must be half-wavelengths. Figure 33 shows
the antenna geometry, and Figs. 34 and 35 provide the
variation of its gain with the cavity and radome thick-
nesses. The microstrip antenna gain without the radome
was about 6 dBi, and the computed peak gains, in Figs. 34
Figure 34. Variation of the cavity gain with its
height H, D2.0l,

e
r
p
t 0:25; e
r
10.
Figure 35. Dependence of the cavity gain on
radome thickness, D2.0l, H0.55l, e
r
10.
and 35, are about 15.7 dBi, indicating an increase of about
9.7 dB, which is in good agreement with the prediction of
Eq. (58). A set of measured gain patterns, in the principal
E and H planes, are shown in Fig. 36. They conrm the
predicted gain enhancement.
BIBLIOGRAPHY
1. S. Silver, Microwave Antenna Theory and Design, Peter
Pereginus, London, 1984.
2. A. D. Oliver, P. J. B. Clarricoats, A. Kishk, and L. Shafai,
Microwave Horns and Feeds, Peter Pereginus, London, 1994.
3. Y. T. Lo and S. W. Lee, Antenna Handbook, Theory Applica-
tions and Design, Van Nostrand Reinhold, New York, 1988,
Chapter 16.
4. R. C. Johnson and H. Jasik, Antenna Engineering Handbook,
2nd ed., McGraw-Hill, New York, 1984.
5. M. Barakat and L. Shafai, Studies on certain modied Lune-
berg lenses, IEE Proc. 130 (Part H)(5):363368 (Aug. 1983).
6. G. Beke and G. W. Farnell, A homogeneous dielectric sphere
as a microwave lens, Can. J. Phys. 34:790803 (1956).
7. V. B. Mason, The Electromagnetic Radiation from Simple
Sources in the Presence of a Homogeneous Dielectric Sphere,
Ph.D. dissertation, Univ. Michigan, 1972.
8. P. J .B. Clarricoats, A. D. Oliver, and M. Rizk, A dielectric
loaded conical feed with low cross-polar radiation, Proc. URSI
Symp. EM Theory, Spain, Aug. 1983, pp. 351354.
9. E. Lier, A dielectric hybrid mode antenna feed, a simple
alternative to the corrugated horn, IEEE Trans. AP-34:
2129 (1986).
10. R. W. P. King, S. R. Mishra, K. M. Lee, and G. S. Smith, The
insulated monopole: admittance and junction affect, IEEE
Trans. Anten. Propag. AP-23(2):172177 (March 1975).
11. I. J. Bahl and S. S. Stuchly, Analysis of a microstrip covered
with a lossy dielectric, IEEE Trans. Microwave Theory Tech.
MTT-28:104109 (Feb. 1980).
12. J. A. G. Malherbe, The design of a slot array in nonradiating
dielectric waveguide, Part I, theory, IEEE Trans. Anten.
Propag. AP-32(12):13351340 (Dec. 1984).
13. A. Sanchez and A. A. Oliner, A new leaky waveguide for
millimeter waves using nonradiative dielectric (NRD) wave-
guide, Part I: Accurate theory, IEEE Trans. Microwave Theory
Tech. MTT-35(8):737747 (Aug. 1987).
14. J. A. G. Malherbe, A leaky-wave antenna in nonradiative
dielectric waveguide, IEEE Trans. Anten. Propag. AP-36(9):
12311235 (Sept. 1988).
15. J. A. G. Malherbe, Radiation from an open-ended nonradia-
tive dielectric waveguide, Microwave Opt. Technol. Lett.
14(5):266268 (April 1977).
16. R. E. Colin and F. J. Zucker, Antenna Theory, Part I, McGraw-
Hill, New York, 1969.
17. N. Bamba et al., Finite-element analysis of dielectric coated
antenna, Int. Symp. Antennas and Propagation, Sapporo,
Japan, Sept. 1992, Vol. 2, pp. 433436.
18. D. R. Jackson and N. G. Alexopoulos, Gain enhancement
methods for printed circuit antennas, IEEE Trans. Anten.
Propag. AP-33(9):976987 (Sept. 1985).
19. L. Shafai, D. J. Roscoe, and M. Barakat, Simulation and
experimental study of microstrip fed cavity antennas, Int.
Symp. Antennas and Applied Electromagnetics, ANTEM96,
Aug. 1996, pp. 549554.
Figure 36. Measurement of principal plane patterns of cavity antenna, D4 cm, H0.9cm,
t 1.27mm, e
r
10.2, f 17.2GHz.
DIELECTRIC MEASUREMENT
R. BARTNIKAS
IREQ/Institut de Recherche
dHydro-Quebec
Varennes, Quebec
Canada
Dielectric measurements are concerned with the charac-
terization of solid, liquid, and gaseous insulating materi-
als over a wide range of DC and AC conditions at different
frequencies temperatures, eld strengths, and pressures,
under differing environments. The frequency range cov-
ered extends downward from the power frequency of 50 or
60 Hz through the ultralow frequency range from 10
2
to
10
6
Hz to DC and upward into the audiofrequency (AF),
radiofrequency (RF), and microwave ranges and, nally,
into the optical region for optically transparent dielectrics.
It can be appreciated that a variety of specimen cells are
required to suit the nature of the test and to act as con-
tainment vessels or holders for the specimens undergoing
evaluation. The test methods and specimen containers
used over the lower frequency spectrum differ substan-
tially from those employed over the higher frequency spec-
trum (4300 MHz), because, at lower frequencies, the
dielectric specimen behaves as a lumped circuit element,
as opposed to its distributed parameter behavior over the
higher frequency region, where the physical dimensions of
the specimen become of the same order as the wavelength
of the electric eld. This delimiting difference necessarily
requires other test procedures to be utilized at high fre-
quencies, and constitutes perhaps the main reason for the
bifurcation and the unfortunate, but often attending, iso-
lation of the two elds of endeavoreven though the aim
over the lower and upper frequency regions is identical,
namely, the characterization of dielectric materials.
Space does not permit a detailed description of all the
dielectric measurement procedures and, consequently,
only a cursory presentation is made. Nor is it possible,
within the given constraints, to delve into the various di-
electric conduction and loss mechanisms in order to dis-
cuss the interpretative aspects of the measurement
methods. Accordingly, the presentation is necessarily con-
ned to a concise description of the most common methods
of dielectric measurement employed currently. Wherever
feasible, the methods given attempt to comply with the
general guidelines of those specied in national and in-
ternational standards, such as those by ASTM (American
Society for Testing and Materials) and IEC (International
Electrotechnical Commission), in order to put methods
forward that are universally accepted and have withstood
the test of time. The dielectric measurement methods pre-
sented here will deal principally with those of dc conduc-
tivity, dielectric constant and loss as a function of
frequency, and voltage breakdown or dielectric strength.
1. DC CONDUCTIVITY MEASUREMENTS
1.1. Volume Resistivity (q
v
)
Insulating materials employed on electrical equipment
usually characterized by a high insulation resistance
and thus provide an isolating medium between adjacent
components that are maintained at different potentials. In
certain applications, such as for capacitor components,
bushings, and cables, they must exhibit extremely low
leakage current. In other applications, where partially
conducting polymers are of interest, the insulation resis-
tance values are substantially reduced. Insulation resis-
tance measurements, which are generally carried out
under DC conditions, yield not only data on the electrical
conduction characteristics of a material, but may also pro-
vide an indication of the uniformity or impurity content of
the insulating material. It is thus of considerable practical
interest to classify the various insulating materials in
terms of their DC insulation resistance, which can then
be related to their DC electrical conductivity. The DC con-
ductivity s
DC
of an insulating or dielectric material, is
more fundamental property, as it bears a direct relation-
ship to the conduction mechanisms taking place in the
dielectric. It is dened as [1]
s
DC
J
DC
E
1
where J
DC
is the DC leakage or conduction current den-
sity in A/cm
2
and E is the direct electrical eld in V/cm; the
units of s
DC
are S/cm. If it is assumed that the DC con-
ductivity arises from a drift of singly charged carriers e in
the eld direction, having a charge concentration n per
cm
3
and a mobility of m in cm
2
V
1
s
1
, then Eq. (1) may
be expressed as
s
DC
emn 2
The measured DC volume insulation resistance, R
v
, is
related to the DC volume resistivity of dielectric r
v
by
r
v
A
d
R
v
3
where A is the area of the measuring electrodes in cm
2
and
d denotes the thickness of the dielectric specimen in cm;
by definition, the DC conductivity is inversely related to
the DC volume resistivity as
s
DC
1=r
v
4
such that the units of volume resistivity are in O cm.
There are various specimen holder electrode systems
and measurement techniques available for determining
the volume insulation resistance R
v
in terms of which the
volume resistivity r
v
may be computed, employing Eq. (3)
[24]. For illustrative purposes, only the most prevalent
ones in use will be considered. Figure 1 depicts a typical
three-terminal electrode system with the dielectric speci-
men held between circular parallel-plane metallic elec-
trodes. The electrodes are usually made of stainless steel,
with the low-potential (guarded) electrode of diameter D
1
,
having a diameter size less than the high-potential elec-
trode, whose diameter D
3
is equal to that of the guard ring
electrode. The separation between the latter and the low-
potential guarded electrode is equal to g, such that gr2d,
916 DIELECTRIC MEASUREMENT
where d denotes the thickness of the dielectric specimen.
The gap g between the low-potential and guard electrodes
must be sufciently large to prevent leakage over the sur-
face of the dielectric from inuencing the volume resist-
ivity measurement; this is particularly important with
high-input-impedance electrometers. A value of g 2d is
most expedient, since it permits the measurement of both
volume and surface insulation resistance with an identical
electrode conguration.
The fringing of the ux lines essentially extends the
guarded electrode edge into the gap region bounded by the
measuring or low-potential electrode and the guard ring.
Hence the area, A, in Eq. (3) is not the geometrical area
of the low-potential electrode, but is approximately given
[2] by
A
pD
1
g
2
4
5
The determination of specimen thickness d in Eq. (3), does
not present itself as a trivial problem [5]. Exact parallel-
ism between the two opposite sides of a solid dielectric
specimen is difcult to achieve, in practice. With polymers,
it is common to make several thickness measurements
along the specimen surface, either with a micrometer or a
dial gauge, and then determine the average value of d.
With most polymeric materials, the dielectric specimens
will tend, in general, to conform to the surface of the mea-
suring electrodes. However, with hard materials, the op-
tically at electrodes will generally not be contiguous with
every portion of the surface of the dielectric. In such cir-
cumstances, the three-terminal electrodes must be either
paint or vapor deposited on the rigid surfaces of the spec-
imen. For this purpose, silver or aluminum is frequently
employed, although aluminum is less desirable, because of
its propensity to form nonconducting oxide lms. Alterna-
tively, tinfoil electrodes may be utilized, in conjunction
with a minute thickness of silicone grease, applied to en-
sure their adhesion to the specimens surface. When liquid
dielectrics are evaluated, permanently mounted three-ter-
minal electrodes are employed, in conjunction with a cell
container into which the liquid specimen submerges the
measurement electrodes.
Figure 2 portrays a schematic three-terminal circuit
diagram for the measurement of the volume insulation
resistance R
v
. Perhaps one of the most important consid-
erations in the measurement of R
v
is the time at which,
following the application of the electric eld, the actual
measurement is made. When the voltage is suddenly ap-
plied across the specimen, the observed initial charging
current is associated with the polarization of the dielec-
tric; both the induced and permanent dipoles in the di-
electric become aligned in the direction of the electrical
eld. Once this very rapid process is completed, the cur-
rent commences a monotonic decline with time, as surplus
free charge carriers are gradually swept out of the dielec-
tric by the electric eld. The nature of these charge car-
riers and their mobility are directly associated with the
structure of the dielectric material. If the dielectric has an
open structure, such as glass, the charge carriers may be
ions; similarly, in a dielectric liquid such as an oil, where
electrolytic contamination may be the source of the charge
carriers, ions may be also responsible for the conduction
current. In polymers, where the latitude of ionic motion is
greatly restricted, the conduction process is frequently
governed by electrons. Ideally, the R
v
value should be
measured when the conduction or the so-called leakage
current attains a constant value, which is a function of
the dielectric under test. For example, in a polymer, the
value of constant current may be achieved when all excess
free electrons have been removed from the dielectric and
the residual leakage current is entirely determined by the
trapping and detrapping rates of the electrons at the var-
ious traps (principally shallow traps). Thus, the number of
migrating electrons at any one time approaches a constant
value when an equilibrium is attained between the time
each electron resides trapped in a well and the time it is
free to migrate before it again becomes re-trapped. Since
the complexity of the conduction process virtually ensures
that different dielectrics are characterized by different
times necessary for the leakage current to attain a con-
stant value, it has been agreed ad arbitrium that all in-
sulation resistance measurements should be made
following a one-minute application of the electrical eld.
Since the volume resistivity r
v
of good insulating
materials falls in the range from 10
12
to 410
18
Ocm, the
leakage current I
l
for such materials must be measured
with a picoammeter, as indicated in Fig. 2. The guard
circuit improves the accuracy of the measurement by re-
ducing the inuence of the leakage resistance. The effects
of the coaxial cable resistance connected across the DC
power supply can be greatly decreased by shunting the
input of the coaxial cable to its shield, by means of an op-
erational amplier with unity gain; this feature is often
incorporated in commercially available electrometer/
ohmmeter instruments [6].
The volume insulation resistance R
v
, in addition to be-
ing contingent on the time of the voltage application, is
D
3
D
1
D
2
Dielectric
High-potential
electrode
Low-potential
electrode
Guard ring
g
d
Figure 1. Three-terminal electrode system for the measurement
of volume resistivity (after ASTM D257) [3].
DIELECTRIC MEASUREMENT 917
also a function of the applied voltage V; it is temperature-
dependent as well. Thus, the value of V and the temper-
ature must be specied; in general, the values of 100 and
500V are most commonly employed [2,3]. Following one
minute of voltage application, the value of R
v
then calcu-
lated from
R
v
V
I
l
6
In the measurement of R
v
an accuracy of 5% may be
readily obtained. However, as the volume resistivity r
v
is
subsequently obtained in terms of Eq. (3), the accuracy of
the r
v
value is somewhat degraded, as a result of errors
inherent in the measurement of the specimen thickness d
and the estimation of the electrode area A [refer to Eq. (5)]
when compensation for the eld fringing effects is made.
1.2. Surface Resistivity (q
s
)
Surface resistance R
s
of solid insulating materials is, to a
large extent, determined by the state of cleanliness or
contamination of the surface of the dielectric under test. It
is, as well, a strong function of surface moisture, particu-
larly if the moisture lm contains electrolytic impurities
either intrinsic to the liquid lm itself or as a result of
solid ionic contaminants originally present on the solid
dielectric surface. Surface resistance is thus a measure of
the materials propensity to surface contamination and
constitutes a useful indicator as concerns the surface
tracking resistance of insulators when subjected to elec-
tric elds. It is common practice to condition the speci-
mens prior to measurement in a dry atmosphere, before
performing the actual measurement at 50% relative hu-
midity.
Surface resistance measurements may be carried out,
either with two- or three-terminal electrode systems, al-
though three-terminal electrodes are usually employed to
eliminate stray leakage effects. The units of surface resis-
tance are ohms or ohms per square. The latter refers to the
arrangement of the electrodes, the conguration of a
square, on the surface of the specimen, as depicted in
Fig. 3 [2].
It is evident from the electrode arrangement in Fig. 3
that the surface resistance measurement also includes a
contribution of the volume resistance. The magnitude of
this contribution diminishes as the surface conductivity
becomes increasingly greater than the volume conductiv-
ity. The procedure followed in measuring the surface re-
sistance R
s
is identical to that of R
v
. The surface resistance
is given by
R
s
V
I
sl
7
where I
sl
is the surface leakage current and V is the volt-
age across the high-potential (H) and low-potential (L)
electrodes. In Fig. 3, G represents the guard electrodes
and g denotes the separation between the guard (G) and
low-potential (L) electrodes. The surface resistivity r
s
in
ohms or ohms per square, is then determined from [2]
r
s
y
x
R
s
8
where y denotes the length of the low-potential electrode
(L) and x is the separation between the high- (H) and low-
(L) potential electrodes. The electrodes may be applied
with silver paint; alternatively, silver or aluminum elec-
trodes may be deposited upon the surface under vacuum.
Frequently, tinfoil electrodes are utilized with an extreme-
ly thin-layer lm of silicone jelly applied on their under-
side, in order to provide adhesion on the specimens
surface.
Another approach is to employ the circular three-ter-
minal electrode system of Fig. 2, but with the connections
changed as portrayed in Fig. 4. Note that with this
arrangement, the high potential is applied to the circular
electrode (H) encompassing the center electrode, which
Top view of solid
dielectric specimen
V
Coaxial cable Coaxial cable
H
x
g
g
y
Stable
dc
power
supply
G
G
L
Picoammeter A
I
sl
Figure 3. Schematic circuit diagram for the measurement of surface resistance with a three ter-
minal electrode arrangement on the dielectrics surface [2].
Dielectric
Picoammeter
Coaxial cable
I
l
Coaxial cable
Three-terminal
electrode system
Stable dc
power
supply
V
+
Voltmeter A
Figure 2. Schematic circuit diagram for a three-terminal mea-
surement of the volume insulation resistance.
acts as the low-potential electrodes (L), while the upper
electrode is connected to ground. In contradistinction to
Fig. 2 (for volume resistivity measurements), the gap dis-
tance gZ2d; in analogy to Fig. 3, g is equivalent to the
electrode separation distance, x. With circular electrode
symmetry, the surface resistivity becomes [2,3]
r
s
pD
1
g
R
s
9
where D
1
is the diameter of the low-potential electrode (L).
The diameter of the upper grounded electrode (G) may be
equal to or greater than that of the encompassing circular
high-potential electrode (H).
2. PERMITTIVITY AND LOSS MEASUREMENTS ON
LUMPED CAPACITANCE SPECIMENS
Under alternating voltages, dielectric materials are em-
ployed either as supports to insulate electrical components
from each other and ground, or as dielectrics in capacitors.
Some applications require dielectrics of low loss and low
dielectric constant, while in others, high-dielectric-
constant materials are desirable, to provide the highest
possible capacitance for a given physical size. Thus two of
the most important electrical properties of dielectric ma-
terials, in terms of which their use and application suit-
ability at either low or high frequencies are assessed, are
those of dielectric loss and dielectric constant.
The capacitance C of a parallel-plate capacitor contain-
ing a dielectric material having a relative real permit-
tivity, e
0
r
, may be expressed as
Ce
0
r
C
0
10
where C
0
is the capacitance in vacuo and is given by
C
0
e
0
A
d
11
where A is the area of the capacitors plates in cm
2
, d the
thickness of the dielectric, and e
0
the permittivity in vacuo
equal to 8.854 10
14
F/cm. By definition, e
0
r
is equal to
the ratio e
0
=e
0
, where e
0
is the real value of the permittivity.
Frequently, the relative real value of the permittivity e
0
r
is
simply referred to as the dielectric constant. The occur-
rence of loss in dielectrics, which may be associated
with the migration of free charge carriers, space charge po-
larization, or the orientation of permanent dipoles, is man-
ifest externally by a phase shift between the electric eld
(E) and the displacement (D) vectors [1]; consequently, the
permittivity, e, becomes a complex quantity of the form
e e
0
je
0 0
12
where e
00
denotes the imaginary value of the permittivity.
The total current density vector J through the dielectric,
composed of the leakage current density J
l
and cap-
acitative or displacement current density J
c
vectors, may
be expressed in terms of e
0
and e
00
as
J J
l
J
c
oe
0 0
joe
0
E
13
The phasor relationship between the current density
vectors J
l
, J
c
, and J is delineated in Fig. 5a with its cor-
responding RC equivalent circuit in Fig. 5b, in terms of
which the dissipation factor (tan d) of the dielectric spec-
imen may be dened as
tand
J
l
J
c
I
l
I
c
14
where I
l
and I
c
are the corresponding current vectors.
From Eqs. (14) and (15), it follows that
tand
e
0 0
e
0
15
Since the AC conductivity s
AC
is by definition, equal to
J
l
/E, then, in terms of Eq. (13), we obtain
s
AC
oe
0 0
16
and
tand
s
AC
oe
0
17
The AC conductivity s
AC
must be distinguished from its
DC value s
DC
because it may include permanent dipole
V
H
g
Coaxial cable
Coaxial cable
L
G
H
Stable
dc
power
supply
+
A
Dielectric
I
sl
Figure 4. Schematic circuit diagram of a three-terminal circular
electrode arrangement for the measurement of the surface insu-
lation resistance [3,6].
J
c
J
l
C
V
(a) (b)
R
J I
I
l
I
C
Figure 5. Current density phasor relationship in a dielectric
(a) with its corresponding equivalent parallel RC circuit (b).
orientation losses, as well as frequency-dependent space
charge polarization controlled carrier migration processes,
which do not arise under DC conditions. It is readily ap-
parent from the equivalent-circuit diagram, which repre-
sents the lossy part of a dielectric by an equivalent
resistance, that
tand
I
l
I
c
1
oRC
18
where Vis the applied voltage vector and I
l
is given by V/R
and I
c
by joCV. It must be borne in mind that, the parallel
equivalent RC circuit representation in Fig. 5b is valid
only at one particular frequency, since both R and the ca-
pacitance C of the specimen are functions of frequency, as
well as temperature and electric eld. It must be further
emphasized that, whereas some dielectric measurement
circuits view the dielectric specimen as a parallel equiva-
lent circuit with a large equivalent parallel insulation re-
sistance R, others consider the dielectric as a series RC
circuit, where the series resistance R
s
5R. The tan d value
for the series RC circuit representation becomes
tand oR
s
C 19
It is apparent that one can derive the primary dielectric
parameters of s
AC
, e
00
, and e
0
from the measured values of C
and tan d by meant of Eqs. (10), (15), and (17).
2.1. Measurements at Low Frequencies (from 10
6
to 10 Hz)
In studies related to the identication of charge carriers
and space charge effects, it is desirable to carry out mea-
surements in the frequency range between 10
6
and
10 Hz. For measurements below 10
l
Hz, it is common
practice to apply a rapid risetime voltage step pulse across
the specimen and subsequently observe the form of the
charging or decay current. The arrangement for this mea-
surement is very similar to that of the volume resistivity
measurement in Fig. 2, with the exception that a switch is
employed in conjunction with the DC power supply to
abruptly apply a voltage step across the specimen [7,8].
Since the total charging current comprises all the fre-
quency components contained within the voltage excita-
tion step, Fourier transformation procedures can be
utilized to derive the individual current distributions at
the discrete frequencies. This procedure may be utilized
irrespective of whether the specimen is charged or dis-
charged. The relative real and imaginary permittivities e
0
r
and e
0 0
r
, respectively, may be expressed in terms of the
resulting current as
e
0
r
o
1
C
0
V
_
1
0
it cos ot dt
C
1
C
0
20
and
e
0 0
r
o
1
C
0
V
_
1
0
it sinot dt
G
oC
0
21
where V is the magnitude of the voltage step, C
N
repre-
sents the lumped capacitance of the specimen at innite
frequency, and G is the DC conductance. Practical impli-
cations impose the upper and lower integration limits on
Eqs. (20) and (21); the lower limit is xed by the risetime of
the electrometer employed (usually about 1 s) and the up-
per limit by the smallest value of current that the elect-
rometer can measure (about 10
16
A) in the presence of
extraneous noise. A numerical procedure is normally fol-
lowed, to carry out these types of measurement [9]. For
each frequency of measurement, the computer performs a
numerical integration between the two integration limits
to determine the values of e
0
r
and e
0 0
r
.
An automated precision time-domain reectometer
procedure is available that permits rapid measurements
down to 10
4
Hz with an accuracy of 0.1% and a resolu-
tion of 10
5
in the tan d value [10,11]. Its schematic circuit
diagram is depicted in Fig. 6.
Positive and negative voltage steps are applied across
the specimen and the reference capacitors, C and C
ref
, re-
spectively. The operational amplier, in conjunction with
the feedback capacitor C
f
constitute a charge detector,
providing an output that is proportional to the net charge
injected [Q
ref
Q(t)/C
f
] by the two opposite-polarity volt-
age steps of amplitude DV and DV, respectively. As the
voltage across the specimen changes from 0 to DV, the
charge Q(t) owing through the specimen is determined
from
Ct
Qt
DV
22
where C(t) denotes a time-dependent capacitance. Hence
the complex capacitance of the specimen C
*
(o), as a func-
tion of frequency, may be expressed as
C
o C
0
o jC
0 0
o
_
1
0
Ct
0
expjotdt
23
where C(t)
0
is the time derivative of C(t). The relative real
and imaginary permittivities, e
0
and e
00
, are then deduced
from
C
0
o
_
1
0
Ct cos ot dt C0 C
ref
24
and
C
0 0
o
_
1
0
Ct sinot dt 25
where C
0
(o) and C
00
(o) are the real and imaginary capac-
itances corresponding to e
0
r
C
0
and oe
0 0
r
C
0
, respectively;
C(0) is the initial capacitance and the integration is
carried out for all times following the application of the
voltage step at t 0. The minimum and the maximum
feasible measurement times t
max
and t
min
determine the
minimum and maximum measurement frequencies o
min
/
2p and o
max
/2p, respectively. The entire measurement is
completed in less than one cycle at o
min
/2p.
A most useful instrument, which is frequently em-
ployed in the range from 10
2
to 10
2
Hz and occasionally
up to 10
4
Hz, is the ThompsonHarris bridge [12,13]. A
two-terminal specimen cell is utilized in conjunction with
the bridge, so that a correction must be made to take into
account fringing effects at the electrode edges. The sche-
matic circuit diagram of the bridge, portrayed in Fig. 7,
incorporates a specimen biasing feature [14], which is in-
cluded to permit the determination of the depth of charge
traps in the dielectric bulk and adjacent to the measuring
electrodes.
The highly regulated frequency generator used in con-
junction with the ThompsonHarris bridge must provide
exact in-phase and quadrature voltage outputs of Vand jV,
respectively. Operational ampliers delineated in Fig. 7
provide the necessary voltage, phase, and impedance re-
lationships. The capacitive current of the specimen is bal-
anced by the injection of an out-of-phase voltage bV
across a variable capacitor C
c
; injection of a quadrature
voltage of ajVacross C
R
compensates for the conduction or
leakage current in the specimen conductance G
x
. Balance
of the bridge is achieved by manipulating the capacitors
C
C
and C
R
and observing the null point, in terms of the
Lissajous gures displayed on the long-persistence oscil-
loscope. At balance, the conductance of the specimen G
x
is
given by
G
x
oaC
R
26
and the capacitance of the specimen C
x
is
C
x
bC
c
27
+
C
x
R
x
C
r
C
c
Two-phase
generator
Voltage
follower
Voltage
follower
Voltage
follower
dc bias
Specimen
Inverter
Variable
persistance
oscilloscope
V
V j
x y
V
s
V
s
V
s
Figure 7. ThompsonHarris low-frequency bridge with specimen bias control feature [14].
+
Generator
Specimen
Operational
amplifier
Amplifier
and filters
Disk
storage
Recorder
Printer
Clock
Computer
Oscilloscope
A/D
converter
+V
C
C
ref
C
V
f
Figure 6. Schematic circuit diagramof time-domain systemfor measurements down to 10
4
Hz [10].
from which the dissipation factor of the dielectric (tan d) is
obtained as
tand
G
x
oC
x
aC
R
bC
c
28
Note that a and b are dimensionless quantities, represent-
ing the fraction of the voltage, V
s
injected across C
R
and
C
c
, respectively.
Not indicated in Fig. 7 is a zero-offset feature, which is
utilized in routine measurements to compensate for the
DC coupling circuitry, in order to prevent erratic shifts in
the Lissajous gures while balancing is being carried out.
The accuracy of the bridge is 0.1% with resolution ordi-
narily better than 0.1%.
Frequency response analyzer methods may also be em-
ployed for low-frequency measurements. These computer-
ized techniques perform adequately well within the range
from 10
4
to 10
4
Hz [15].
2.2. Power and Intermediate-Frequency Methods
(10 Hz1MHz)
A considerable portion of the electrical insulating materi-
als manufactured for use in electrical apparatus and ca-
bles are evaluated within the frequency range from 50 Hz
to 1MHz, employing primarily bridge type circuits. A fur-
ther substantial portion of tests at high voltages are car-
ried out at xed power frequencies of 50, 60, and 400 Hz.
The bridge circuits designed for power frequency applica-
tions, where measurements are normally made as a func-
tion of voltage, differ significantly from those involving
measurements as a function of frequency at low voltage.
Since most of these tests are performed by means of either
Schering or transformer ratio arm bridges, the discussion
here will be essentially conned to these types of bridges.
A common low-voltage Schering bridge arrangement,
which employs the parallel substitution technique recom-
mended in ASTM D150 [16], is depicted in Fig. 8 for the
case where measurements are carried with a two-terminal
specimen cell.
The capacitance C
3
is selected such that its negligibly
small dielectric losses are approximately equal to those of
the intrinsically low-loss standard capacitor C
s
. Note that
the Schering bridge views the specimen as an equivalent
series R
x
C
x
device, so that the variable arm composed of
the parallel combination of R
1
and C
1
must be capable of
compensating for the losses in the small series resistance
R
x
of the specimen. The null detector, which is normally
an amplier, is tuned to the frequency of the measure-
ment, f o/2p. Balance is rst obtained by an adjustment
of the capacitors C
1
and C
s
, with the specimen disconnect-
ed. Then, with the specimen placed in parallel with the
standard capacitor C
s
, the bridge is rebalanced. The spec-
imen capacitance C
x
is thus determined from
C
x
C
0
s
C
0 0
s
29
and the dissipation factor from
tand
x
oR
1
C
0
s
C
0 0
1
C
0
1
C
0
s
C
0 0
s
30
where C
0
1
; C
0
s
, and C
0 0
1
; C
0 0
s
denote the values of the vari-
able capacitors C
1
and C
s
at balance with the specimen
disconnected and reconnected, respectively. The substitu-
tion technique eliminates the errors introduced by the
coupling effects of the various stray capacitances, but it
does not circumvent errors arising from connecting lead
inuences.
The procedures for the correction of lead and stray ca-
pacitance effects have been standardized and are explicit-
ly enumerated in ASTMD150 [16]. It is the inductance, L
s
,
and the resistance R
s
the leads, which contribute to the
apparent increase of the capacitance DC and the dissipa-
tion factor D tan d in accordance with the relations [16]
DCo
2
L
s
C
2
31
and
Dtand oR
s
C 32
where C is the true capacitance of the specimen; it is to be
emphasized that, as the skin effect increases with fre-
quency, the lead resistance R
s
increases significantly with
the square root of the frequency, f o/2p. A standard
practice for assessing the effect of the leads is to perform
a measurement on a miniature sized capacitor, where the
latter is rst directly connected to the bridge terminals
and then inserted across the far end of the leads. The dif-
ference between the two readings permits the calculations
of DC and D tan d.
The appearance of an edge capacitance C
e
and a ground
capacitance C
g
will lead to an increase in the measured
apparent capacitance
C
a
CC
e
C
g
33
R
1
R
2
C
1
C
s
C
x
R
x
C
3
D
Figure 8. Low-voltage Schering bridge, employing the parallel
substitution technique in accordance with ASTM D150 [16].
and an apparent dissipation factor
tand
a
Ctand
C
a
34
The relative real and imaginary permittivity e
0
r
and e
0 0
r
will then be given by
e
0
r
C
C
0
C
a
C
e
C
g
C
0
35
and
e
0 0
r
C
a
tand
a
=C
0
36
For the normal type of specimen dimensions, where the
parallel-plane cylindrical electrodes are smaller than the
diameter of the specimen, the capacitance in vacuo C
0
with edge effect correction, may be expressed empirically
in pF as [16]
C
0
0:006954
D
2
d
37
where d is the thickness of the specimen and D the diam-
eter of the electrodes in mm. Exact formulas for the edge
correction may be found in Ref. 17; the use of the exact
formulas does not result in a significant difference for the
correction. As with all dielectric measurements, the accu-
racy attainable is contingent not only upon the accuracy of
the observed capacitance and tan d values, but also on the
stray and edge effects of the electrodes employed, as well
as the calculated interelectrode vacuum capacitance C
0
. In
general, the permittivity is determinable to within 71%
and the tan d value to within 7(5%0.0005) [16].
The circuit of a power frequency high-voltage Schering
bridge, portrayed in Fig. 9, represents essentially an in-
verse arrangement of its sibling low-voltage bridge equiv-
alent. The lower bridge arms of R
4
, C
4
, and R
3
constitute
the balancing elements, while the upper arms of the series
representation of the specimen, R,C, together with the
standard capacitor, C
s
, which have a high impedance in
comparison with the lower resistive arms, assume the
major portion of the voltage drop. This arrangement pro-
vides the bridge with an inherent safety feature, since the
lower arms where balance manipulation of the bridge is
carried out, remain at low potential. Figure 9, which rep-
resents the classical Schering bridge circuit, delineates
also the guard circuits balancing elements R
5
and C
5
, ar-
ranged in accordance with the so-called Wagners Earth
method. The guard circuit, which is implemented in order
to eliminate the stray capacitance to ground, necessarily
entails the use of a three-terminal measurement proce-
dure. The solid dielectric slab specimen is placed in a
three-terminal cell of the type depicted in Fig. 1, or if the
specimen is a liquid dielectric, a concentric coaxial elec-
trode cell [18] may be employed. Frequently, the specimen
undergoing test may be a high-voltage power cable or sta-
tor bar, whose high-voltage terminated ends must also in-
corporate a guard circuit [19]. The standard capacitor C
s
,
which must be partial-discharge-free up to the maximum
measurement voltage, is normally a 100pF compressed-
gas-lled unit with negligible dielectric loss. Note that the
low-voltage arms are enveloped in grounded shields; the
shield, screening the low-potential electrodes of the spec-
imen and standard capacitor, including the detector that
normally comprises an amplier tuned to the power fre-
quency, eliminates the stray capacitances to ground and
between the components themselves. Thus, any capaci-
tance current, which may develop between the detector
and the high-voltage portion of the bridge, ows directly to
ground via the auxiliary bridge arm of R
5
and C
5
. Since
the latter are interposed between the shield and the
bridge ground, their manipulation balances the guard or
shield circuit. The switch SW, shown in Fig. 9, permits the
necessary independent balancing steps for the bridge
guard circuit and the bridge itself. At balance, the capac-
itance of the specimen [2] is given by
C
C
s
R
4
R
3
38
and for equal self-inductances inherent with the resistive
elements R
3
and R
4
, the dissipation factor reduces to
tand oR
4
C
4
39
Since high-voltage Schering bridges are normally
designed to operate at one xed power frequency, the di-
als of R
3
and C
4
are calibrated to read directly the capac-
itance and tan d values of the specimen, respectively. It is
to be emphasized that, under high voltage conditions,
should the specimen under test undergo partial discharge,
then the indicated tan d value will reect the power losses
due to partial discharges, in addition to the dielectric loss-
es occurring in the solid, liquid, or solidliquid insulating
system of the specimen [20].
R
3
R
5
R
4
C
5
C
4
C
s
SW
Specimen
R,C
D
Standard
capacitor,
Figure 9. Classical power frequency circuit of high-voltage Sche-
ring bridge with Wagners Earth [2].
Present high-voltage xed power frequency Schering
bridges employ a driven or active guard technique for bal-
ancing of the guard circuit in lieu of the classical Wagners
Earth method. In this approach, the guard circuit and the
detector, D, are maintained automatically, at the same po-
tential, by means of a unit gain operational amplier,
whereby only a single balance step is required for the
bridge. This feature, together with other improvements in
the Schering bridge, is well exemplied in the Tettex pre-
cision Schering bridge, which has been designed for use on
thin dielectric specimens up to 2kV; its circuit is depicted
in Fig. 10.
The bridge is limited in voltage, since now the lower
arms are capacitive in nature, in order to attain a higher
sensitivity as the stray capacitances are thus greatly re-
duced. The capacitance of the specimen, C, is obtained by
an adjustment of C
4
to yield
C
C
s
C
3
C
4
40
and the value of tan d at the null is obtained by adjust-
ment of R
3
; the already low dielectric loss standard
capacitor C
s
is articially reduced to zero, such that the
tan d value of the specimen becomes
tand oR
3
C
3
G
3
oC
3
41
In the measurements carried out with Schering bridges,
the capacitance and tan d values of the dielectric speci-
mens are obtained in terms of the resistance and capac-
itance elements of the bridge. Hence, the precision and
accuracy of the measurements are determined by the ac-
curacy of these resistances and capacitances themselves.
Precise dielectric measurements may also be performed by
means of an inductively coupled voltage divider, utilizing
a transformer arrangement, thereby circumventing some
of the accuracy and stability constraints associated
with resistive and capacitive elements [2,22]. Perhaps
one of the nest precision/accuracy commercially
available transformer ratio arm bridges for variable-fre-
quency measurements in the range from 10 Hz to
100 kHz is that of Gen Rad, under the designation type
1621 transformer ratio arm bridge. Its schematic circuit
diagram is delineated in Fig. 11. A 12-digit readout of the
specimen capacitance C
x
with a 10 ppm accuracy is pro-
vided within the range from 10
7
to 10 mF. A basic accu-
racy of 0.1% is attainable for conductance G
x
measurements within the range of 10
10
10
3
mSthat
is, a tan d value of 10
7
at 1 kHz may be determined
with a four-gure resolution.
Operational
amplifier
Specimen
R,C
Standard
capacitor
Guard
Bridge
R
3
G
3
C
4
C
3
C
s
+1
D
Figure 10. Schematic diagram of precision Tettex Schering
bridge for measurements at power frequency [21].
N
1 C
1
C
2
C
n
R
1
R
2
R
n
C
A
C
B
G
X
C
X
N
2
D
1
N
1
2
N
1
2
N
1
2
N
1

n
N
1
n
N
1
Figure 11. Basic circuit of Gen Rad
type 1621 precision transformer ratio
arm bridge [23].
The capacitances C
A
and C
B
shunting the transformer
winding and the specimen, respectively, do not introduce
any error into the measurement, since the former produc-
es only a reduction of the voltage across the specimen,
while the latter causes only a decrease in the sensitivity of
the detector D. In balancing the bridge, the multiple
tapped transformer principle is utilized in the course of
the resistive and capacitive decade adjustments. Accord-
ingly, the balance equation at the null of the bridge is
given by
N
2
G
x
joC
x
N
1
b
1
G
1
b
2
G
2
b
n
G
n
joa
1
C
1
a
2
C
2
a
n
C
n
42
Equating the real and imaginary terms yields the capac-
itance of the specimen
C
x
N
1
N
2
a
1
C
1
a
2
C
2
a
n
C
n
43
and the conductance
G
x
N
1
N
2
b
1
G
1
b
2
G
2
b
n
G
n
44
The dissipation factor of the specimen as a function of
frequency is obtained as
tand
G
x
oC
x
b
1
G
1
b
2
G
2
b
n
G
n
oa
1
C
1
a
2
C
2
a
n
C
n
45
Note that the transformer ratio arm bridge views the
dielectric specimen as an RC parallel equivalent circuit.
A computer-controlled automatic transformer ratio
arm has been developed for measurements at power fre-
quencies under high-voltage conditions [24]. The bridge
circuit is delineated schematically in Fig. 12, in which the
Specimen
Ground
Protective
ground
Standard
capacitor
Screen 1
Screen 2
ADC Microcomputer
Printer
Data and dialog
tan
tan
range
range
Null
detector
C.T. R
C
N
1
N
2
N
2
N
1
C
x
N
4
N
3
G
1
G
2
N
4
tan
N
3
balance
balance
90

I
1
x,y
Figure 12. Automated power frequency transformer ratio arm bridge with computer control for
measurements at high voltages [24].
coarse and ne balances are obtained by variation of the
current comparator windings N
1
,N
2
and N
3
,N
4
, respec-
tively. The currents in N
3
and N
4
are controlled by the
multiplying analog-to-digital converters (ADCs), a and b,
respectively, and are proportional to the current in the
standard high-voltage capacitor C
s
. The ampere turns
equation for the balance condition at an applied voltage
V across the specimen when the current I is equal to zero
in the winding of the null detector N
1
is given by
VG
x
joC
x
N
1
joC
s
VN
2
aRG
1
N
4
jbRG
2
N
3
46
Equating the real and imaginary terms leads to the ap-
proximate expressions for the parallel, equivalent capac-
itance C
x
and dissipation factor tan d of the dielectric
[18,24]:
C
x
N
2
aC
s
N
1
47
tand
bRG
2
1
a
N
2
48
The automation of the power frequency transformer ratio
arm bridge results in a reduction of the accuracy of the
tan d measurement from 7110
7
to 71 10
5
.
2.3. Radiofrequency Methods (1200 MHz)
Bridge techniques become unsuitable for measurements
at frequencies beyond 1MHz, because of the onset of in-
ductance effects and, as a consequence, within the radio-
frequency region of 1200MHz, resonance rise (Q meter)
or susceptance variation methods must be employed
[2,18]. Within these frequencies, three-terminal tech-
niques become inapplicable cable and measurements
must be carried out using two-terminal specimen holders.
The basic Q-meter circuit is shown in Fig. 13, in which
the capacitance and tan d values of the specimen are de-
termined in terms of a variable standard capacitor C
s
and
the quality factor Q of the circuit. The coil L denotes a
range of shielded xed-inductance coils that are employed
to establish resonance of the circuit with the specimen
(G
x
,C
x
) inserted and removed. By definition, the Q value of
the circuit is equal to the ratio of the peak voltage V
0
across the oscillator to that across the inductance, V
L
such
that
V
0
V
L
1
o
2
L
2
R
2
_ _
1=2
1Q
Q
49
for Qb1. The voltmeter (V) of the Q meter is calibrated to
read the Q values directly, since Q is given by V/IR. Rep-
resenting the values of C
0
s
,Q
0
and C
s
,Q as those obtained
with the specimen disconnected and connected, respec-
tively, yields the capacitance C
x
and tan d values of the
specimen as
C
x
C
0
s
C
s
50
and
tand
1
Q
x
51
where
Q
x
C
0
s
C
s
C
0
s
Q
0
Q
Q
0
Q
52
When the measurements are carried out at high fre-
quencies, a stiff short copper connecting wire should be
employed between the specimen and the high-potential
terminal of the Q meter, so that when disconnected, its
same geometrical position, and conguration, a short
Shielded coil
Specimen
C
x
L
V
A
R
I
C
x
G
x
Figure 13. Q-meter circuit.
distance removed from the high-potential terminal, may
be maintained, to ensure negligible change in the stray
effects of the two positions of the connecting wire. With
parallel-plane micrometer electrode specimen holders, ac-
curacies of 7(0.1%0.02 pF) and 7(2%0.00005pF), for
the respective capacitance and tan d values of solid dielec-
tric specimens may be achieved [16]. For solid specimens,
an excellent precision reproducibility of 70.05% and
75 10
5
for C
x
and tan d, respectively, may be obtained
by means of liquid displacement-type specimen holders
[25]. A uid displacement cell for use at 1 MHz, consisting
of a xed-plate, two-terminal, self-shielding capacitor, in
which the edge and ground effects are taken into account,
is depicted in Fig. 14.
The cell is most frequently employed for measurements
at 1MHz on polyethylene, although cell designs for fre-
quencies up to 100MHz are available. The uid that is
used in conjunction with polyethylene specimens is sili-
cone with a kinematic viscosity of 1.0cSt (110
6
m
2
/s) at
231C, whose dielectric constant matches that of polyeth-
ylene, and whose tan d value between 100 kHz and 1MHz
is negligibly small (about 5 10
5
). The separation be-
tween the xed measuring electrodes of the cell design in
Fig. 14 is 1.5270.05 mm, thereby restricting the specimen
thickness 1.27 mm, in order to allow for the formation of a
nite liquid lm thickness on both sides of the solid
dielectric specimen adjacent to the central (high-
potential) and outer (ground potential) plate electrodes
or terminals. Two identical sizes (68.3100mm) of poly-
ethylene sheet or slab specimens are employed, and mea-
surements are made on the specimens inserted in the
silicone uid and then on the silicone uid itself with the
specimens removed.
The real value of the permittivity or dielectric constant
of the polyethylene specimen e
0
is obtained from the
relation
e
0
e
0
l

DC
C
0
d
0
d
_ _
53
where e
0
l
is the real value of the permittivity of the silicone
uid at the measurement temperature, d
0
denotes the
electrode separation, d represents the average thickness
of the two specimens, and C
0
is the capacitance in vacuo of
the double-plated capacitor within the uid displacement
cell, given by
C
0
2
e
0
A
d
0
_ _
54
where e
0
is the permittivity of free space, A is the area of
the center capacitor plate or electrode; the value of DC is
obtained from
DCC
2
C
1
55
where C
2
is the measured capacitance with the two solid
dielectric specimens immersed in the silicone uid and C
1
the corresponding value with the two specimens removed.
The dissipation factor, tan d, of the two polyethylene
specimens is dened by
tand tand
l
tand
c
tand
l
d
0
=d 56
where tan d
l
the dissipation factor of the silicone uid
itself and is given by
tand
l
C
T
Q
c
Q
1
=C
l
Q
0
Q
1
57
where C
T
represents the total capacitance of the tuned
Q-meter resonant circuit prior to the connection of the
specimen cell, Q
0
denotes the quality factor of the circuit
at resonance prior to the connection of the ungrounded
lead to the cell terminal, and Q
1
is the quality factor of the
measuring circuit at resonance following the connection of
the lead to the terminal of the cell containing the silicone
uid only, and C
l
the capacitance of the silicone uid
determined from the relation
C
l
C
0
1
C
1
58
where C
0
1
is equal to the capacitance reading following the
connection of the leads to the circuit terminals before the
connection of the ungrounded lead to the cell terminal.
The value of the dissipation factor tan d
c
obtained with the
polyethylene specimens immersed in the cell, is deter-
mined from
tand
c
C
T
Q
0
Q
2
=e
0
l
C
0
DQQ
0
Q
2
59
where Q
2
is the quality factor with the two solid specimens
inserted in the cell.
The susceptance variation method originally propound-
ed by Hartshorn and Ward [26] with subsequent rene-
ments [27,28] is perhaps the most common method
utilized for permittivity and loss measurements over the
frequency region extending from 100kHz to 200MHz. The
technique is based on the half-power point measurements
of voltage across an LC resonant circuit, with the solid or
liquid specimen inserted and removed from the test cell. A
Central plate
terminal
Center plate
(ground)
Outer plate
(ground)
Outer plate
Teflon
insulator
Overflow
Thermometer
(ASTM 23C)
Figure 14. Fluid displacement cell with a xed electrode se-
paration equal to 1.5270.05mm after ASTM Test Standard
D1531 [25].
modied susceptance variation circuit and a cross-section-
al prole view of the associated micrometer adjustable
holder for solid specimens are depicted in Figs. 15 and 16,
respectively. For liquids, the specimen holder is similar to
that depicted in Fig. 16, except that two parallel concave
electrodes are employed to permit containment of the
liquid specimen.
The micrometer adjustable electrode system depicted
in Fig. 16 portrays a solid dielectric specimen between
plane-parallel electrodes. In the measurement procedure,
resonance is rst established with the specimen inserted
between the electrodes and the maximum value of the
voltage e
1
of the ACDC converter is recorded. Thereafter,
the specimen is removed and the separation of the elec-
trodes is reduced until resonance is reestablished; this
resonance point is characterized by a larger output voltage
e
o
of the ACDC converter. The capacitance of this air gap
spacing is numerically equal to the capacitance of the
specimen C
x
and is obtained directly from the calibrated
reading of the main micrometer setting. Manipulation of
the main micrometer head, in conjunction with the small
vernier or incremental capacitor, yields the half-power
points of the resonance curve; the resulting width of the
resonance curve is equivalent to a capacitance change,
designated as DC
0
. It is the square law detection feature
of the instrument that relates the DC
0
value directly to
the recorded change of reading in the incremental capac-
itor. Hence the dissipation factor of the specimen [27] is
given by
tan d
e
0
e
1
_ _
1=2
1
_ _
DC
0
2C
x
60
The real value of the permittivity e
0
r
is normally obtained
in terms of the thickness of the specimen d
s
and the quan-
tity, Dd
e
0
r
d
s
d
s
Dd
61
where Dd represents the decrease in separation of the
main electrodes in air required to restore the same capac-
itance as that obtained with the specimen placed between
the electrodes. An accuracy of 1% is achievable on permit-
tivity measurements and tan d may be determined to
within 71.0 10
6
.
3. PERMITTIVITY AND LOSS MEASUREMENTS ON
DISTRIBUTED PARAMETER SPECIMENS
Dielectric specimens behave as distributed parameter sys-
tems when the wavelength of the electromagnetic eld
becomes comparable to or is less than the physical dimen-
sions of the specimen. The transition from lumped to dis-
tributed parameter behavior occurs generally within the
frequency range from 300MHz to 600 MHz. The high-fre-
quency dielectric measurements represent a vast area of
endeavor, which involves the use of resonant cavities of
cylindrical and rectangular shapes, waveguides, or trans-
mission lines, including quasioptical procedures, as well
as optical methods requiring the use of spectrometers and
interferometers. Since it would not be feasible within the
space constrains to cover, even in a cursory manner, all
test method variations over the millimeter and submilli-
meter wavelengths, the test procedures followed over this
range of frequencies will be illustrated by a number of
widely used and representative test methods.
3.1. Reentrant Cavity Method (300600 MHz)
The reentrant cavity measurement technique constitutes,
in essence, an extension of the HartshornWard method; it
uses the same specimen cell arrangement, with the
Coupling coils
Recorder
acdc
converter
Adjustable
electrodes and
specimen
Incremental
capacitor
Potentiometer
Figure 15. Schematic circuit diagram
of modied HartshornWard suscep-
tance variation circuit [27].
Micrometer screw
Bellows
Grounded electrode
Solid specimen
High-potential
electrode
High-potential electrode
Grounded terminal
Vernier (incremental)
capacitor
Figure 16. Micrometer adjustable electrode for use in conjunc-
tion with the susceptance variation circuit [26].
exception that the inner walls are silver-plated, and the
oscillator signal is admitted into the cell cavity via a cou-
pling loop with a detector loop situated on the opposite
wall of the concentric coaxial cell cavity [29]. The reen-
trant cavity is calibrated as a wavemeter, with the main
micrometer adjustable specimen capacitor acting as the
prime frequency control device and the vernier capacitor
as an incremental control device (refer to Fig. 16). As in
the case of the HartshornWard technique, the dielectric
parameters are determined in terms of the width of the
resonance curve with the specimen inserted between the
measurement electrodes and then removed.
3.2. Coaxial LineWaveguide Methods (500 MHz50 GHz)
Waveguide or transmission-line methods are based on the
shorted coaxial-line technique developed by Roberts and
von Hippel [30]. Although for low loss dielectric solids and
liquids, the technique yields optimum performance for mi-
crowave frequencies up to 50GHz; the method has been
used up to 95 GHz [31]. The connement of the electrical
eld within the hollow waveguides circular or rectangular
geometry eliminates stray capacitance and inductance
effects. A standing-wave pattern results within the wave-
guide, from a reection of the incident wave at the short-
circuit termination adjacent to where the solid specimen is
inserted as depicted in Fig. 17. When liquid specimens are
tested, the waveguide is mounted in a vertical position
[18]. Figure 17 indicates the position of the electrical
nodes (position of the interference minima), with the
width of the nodes, Dx, as indicated at the 3-dB points.
In terms of Dx, the voltage standing-wave ratio, abbrevi-
ated as VSWR, or r, may be expressed as
r l
gs
=pDx 62
where l
gs
is the wavelength of the slotted coaxial line; it is
along the slot that a traveling probe is displaced to deter-
mine the VSWR. Since the value of Dx changes when the
specimen is removed from the waveguide, the VSWR (r)
also changes accordingly.
Perhaps one of the most common shorted coaxial trans-
mission line arrangements in use is that described in
ASTM D2520 [32], which is suitable for temperature-con-
trolled measurements up to 16501C, when utilized in con-
junction with a platinum alloy with 20% rhodium as the
material for the specimen holder; its schematic diagram is
delineated in Fig. 18. A micrometer head in the slotted
waveguide section is capable of measuring node width
distances to within 70.0025mm. The traveling probe has
an adjustable depth control and the detector is of the
square-law type that constitutes a requirement for the
VSMR meter. The setting of the isolator is xed at 30 dB,
and the square-wave modulator provides a constant fre-
quency of 1 kHz; the isolator or attenuation pad prevents
frequency pulling between the generator and the remain-
der of the circuit. The lateral dimensions of the solid spec-
imen are selected to be 0.0570.025 mm less than those of
the transmission line. The rectangular waveguide is op-
erated in the fundamental TE
10
mode, which is analogous
to the TEM mode of a cylindrical waveguide, in which the
electrical eld is radial and the magnetic eld concentric
with the coaxial geometry. The cutoff wavelength l
c
in the
TE
10
mode is equal to 2athat is, twice the width a of the
rectangular guide. Thus, the wavelength with an empty
holder at the required test temperature is given by
l
2
gh
l
2
0
l
2
c
l
2
0
2a
2
63
where l
0
is the wavelength of the radiation in free space
and is equal to c/f, where c is the velocity in free space and
f is the frequency.
Traveling
detector
VSWR
meter
Specimen
Short
E
2
x
0
d
s
x
2
x
2
x
2
g
2
3 dB level
Figure 17. Standing voltage wave pattern in a short-circuited

coaxial waveguide containing a solid dielectric specimen [2].
Square
wave
modulator
Traveling
probe
VSWR
meter
Generator Isolator
Frequency
counter
Slotted
line
section
Cooling
sink
Temperature
isolation
section
Platinum test
section
Tube furnace
Figure 18. Schematic diagram of short-circui-
ted rectangular waveguide with a temperature-
controlled test specimen section (after ASTM
D2520) [32].
With the specimen of thickness d
s
inserted adjacent to
the short in the waveguide, the impedance of the line at
the specimenair interface [33] is given by
Z
in
jom
0
=g
2
tanhg
2
d
s
64
where m
0
is the permeability of the nonmagnetic dielectric
material, which is identical to that in free space. Assum-
ing negligible losses in the walls of the waveguide, the
propagation constant of the coaxial waveguide g
2
contain-
ing the specimen is given by
g
2
2pl
2
c
e
0
r
l
2
0

1=2
65
The load impedance at a phase distance j from the ob-
served electrical node for the value of the VSWR r given by
Eq. (62), (34) is
Z
meas
f m
0
l
g
1 jr tanj=r j tanj 66
where l
g
is the wavelength of the guide and is equal to
2p/b
2
; here b
2
is the phase coefcient of the waveguide
with the specimen inserted; j is the corrected phase dis-
tance, dened by
j2pN=2 d
s
=l
gh
x
2
x
1
l
gs
67
where N represents the smallest integer for which j is
positive, x
2
is the position of the traveling detector with
the specimen inserted, as indicated in Fig. 17, and x
1
is the
equivalent distance with the specimen removed.
Equating the impedances Z
in
and Z
meas
yields
tanhl
2
d
s
g
2
l
gh
1 jr tanj
2pjr j tanj
68
Equating the real and imaginary terms yields the rela-
tive real value of the permittivity e
0
r
of the specimen as
e
0
r
b
2
=2p
2
l
2
c
=l
2
0
69
and, for low-loss specimens [34], the dissipation factor
simplies to
tand
Dx1 l
2
0
=e
0
r
l
2
c
1 tan
2
j
d
s
f1 tan
2
b
2
d
s
tanb
2
d
s
=b
2
d
s
g
70
where the width of the node Dx that would be measured at
the face of the specimen is given by
Dx Dx
2
Dx
1
71
The principal factors affecting the accuracy of the mea-
surements are associated with the assumption that losses
at the walls of the waveguide are negligible and that the
nite airgap between the solid specimen and the walls of
the waveguide does not inuence appreciably the results;
evidently, the latter error does not arise with liquid spec-
imens [18]. However, accuracies of 71% for e
0
r
and
7200radians for the loss angle d are achievable.
3.3. Resonant Cavity Methods (about 500MHz60 GHz)
A resonating cavity may be viewed as a transmission line,
which is shortened at both ends that are separated by an
arbitrary multiple of one-half the operating wavelength.
The insertion of a dielectric specimen into the cavity alters
the wavelength and, as a consequence, the change in the
quality factor Q of the cavity with the specimen inserted
and removed can be used to derive the dielectric param-
eters of the specimen. Since resonant cavities have intrin-
sically high values of Q, they constitute an effective means
for measuring low-loss dielectric materials. The specimens
may have different geometrical congurations such as
spheres, sheets, disks, rods, and so forth, and may ll
the cross section of the beam, if necessary. The required
specimen size becomes smaller as the cavity size dimin-
ishes with frequency, thereby also necessitating a redesign
of the cavity with each octave increase in frequency. For
frequencies above 60 MHz, the reduced cavity sizes, irre-
spective of their shape, rapidly approach a practical limit.
Although open resonant cavities or interferometers may
exceed substantially the frequency of 60 MHz [35] their
applicability is conned to specimens having dielectric
constants in excess of 5.
A coaxial waveguide shorted at one end becomes a res-
onant cavity when shorted at both ends. It may be reso-
nated either by varying the frequency of the externally
applied eld or by varying the radial or axial dimensions
of the cavity itself. A very widely used rectangular micro-
wave cavity design for operation in the transverse electric
eld, TE
1ON
mode, is depicted in Fig. 19. Note that, in the
mode designation code, the rst subscript denotes the
number of half-waves across the short-circuited wave-
guide, the second subscript refers to the number of half-
waves from top to bottom of the waveguide, and the third
subscript represents the odd number of half-waves along
the waveguide. The closed cavity arrangement in Fig. 19 is
identical to that given in the test method described in
ASTM D2520 [32]. It is of paramount importance that the
diameter of the iris holes in the transmitting and detect-
ing ends be small to achieve high Q values. The particular
design of the shown resonant cavity is intended for use
with solid rod-shaped specimens, which are held suspend-
ed between the top and bottom holes that are drilled into
the waveguide (refer to Fig. 19). The resonant frequency of
Shorting plate
Solid rod
dielectric
specimen
Electric
field vector
w
h
d
Iris hole
(dia. = h/2.2)
Figure 19. Closed rectangular resonant cavity for tests on solid
rod-shaped dielectric specimens (after ASTM D2520) [32].
the specimen is dened by
f
0
151=w
2
N=d
2
1=2
72
where d is the physical length of the closed cavity and w
its width (both in cm), N denotes the odd number of half-
waves along the cavity, and the resonant frequency is in
GHz. It is palpably evident, from Eq. (72), that higher test
frequencies require closed cavities with increasingly
reduced physical dimensions.
The measurements may be carried out either by means
of the traditional VSWR meter utilizing a point-by-point
approach or, for more rapidly obtainable results, a fre-
quency sweep generator may be employed, as portrayed in
the schematic test circuit of Fig. 20, in accordance with an
IEC method [36]. The latter method may be computerized,
in order to minimize errors, by recording simultaneously
dual outputs from the signal generator and the cavity.
The measurements are carried out with the empty cav-
ity and then with the specimen inserted. The quality
factor Q
0
of the empty cavity is given by
Q
0
f
0
Df
0
73
where the half-power bandwidth of the empty resonant
cavity is
Df
0
f
02
f
01
74
and f
01
and f
02
are the lower and upper frequency half-
power (3 dB) points; the 3 dB points are established by
means of a variable precision attenuator. When the spec-
imen is inserted into the cavity, the quality factor Q
s
be-
comes
Q
s
f
s
Df
s
75
where f
s
the new resonant frequency of the closed cavity
containing the specimen and the half power bandwidth is
Df
s
f
s2
f
s1
76
where f
s2
and f
s1
are respectively, the upper and lower 3 dB
point frequencies. The value of the relative real permit-
tivity e
0
r
and the dissipation factor tan d may now be de-
termined from
e
0
r
V
0
f
0
f
s
2V
s
f
s
_ _
1
_ _
77
and
tand
V
0
4V
s
1
Q
s
1
Q
0
_ _ _ _
V
0
f
0
f
s
2V
s
f
s
_ _
1
_ _ 78
where V
s
and V
0
are respectively, the volumes of the spec-
imen and the empty cavity. Note that the measured quan-
tities are not contingent on the dimensions of the closed
cavity. An accuracy to within 0.5% for the permittivity and
approximately 5% for the dissipation factor are attainable.
The specimen size and location within the cavity plays an
important role; these two parameters inuence the mag-
nitude of the difference between Q
s
and Q
0
on which the
precision and accuracy depend. A high Q value for the
cavity is thus important, since the 3 dB point frequencies
become more clearly dened.
3.4. Quasioptical and Optical Methods (303000 GHz)
Dielectric measurements at microwave frequencies in ex-
cess of 60 GHz become increasingly arduous, as a result of
the unduly small size of resonant cavity required. The
difculty is circumvented by employing for the microwave
frequency range the same methods as those that are uti-
lized in lightwave optics; such procedures are commonly
referred to as quasioptical or free-space techniques. In
analogy to an optical spectrometer, the collimator in a
quasioptical microwave-type instrument consists of a par-
abolic reector connected to a microwave generator, with
the plane-wave source directed toward the dielectric spec-
imen [37]. The latter is in sheet form and is mounted upon
an object table, as in the case of an optical spectrometer.
Another parabolic reector (substituting an optical tele-
scope), connected to a detector, receives the signal, which
is either reected from or transmitted through the sheet
specimen. Thus, the resulting attenuation in the path be-
tween the transmitting and receiving parabolic reectors
constitutes a measure of the dielectric loss in the inter-
vening dielectric sheet [38].
Quasioptical techniques also include the use of optical
cavity resonators, which are suitable for measurements
within the millimeter and submillimeter wavelengths of
the electromagnetic spectrum. This differs from the usual
closed cavity microwave resonator, in that the length of
the resonator corresponds to a length number of wave-
lengths, while the specimen (in sheet form) assumes only a
small fraction of the overall length [2]. There are three
types of quasioptical resonator: the classical FabryPerot
interferometer, the confocal resonator, and the semiconfo-
cal resonator. The confocal resonator has the advantage
that the electric eld is more conned to the axis of the
Generator Isolator Attenuator
Directional
coupler
Matching
device
Matching
device
Resonant
cavity
Frequency
meter
Sweep
frequency
generator
Oscilloscope
Figure 20. Microwave closed resonant cavity measurement sys-
tem, using a sweep frequency generator technique (after IEC
Publ. 377-2) [36].
resonator, resulting in Q values generally higher than 10
5
and lower diffraction losses. The Q of a semiconfocal res-
onator is approximately equal to half that of the confocal
resonator.
For illustrative purposes, the measurement procedure
followed with a quasioptical confocal resonator [39], de-
lineated in Fig. 21, will be described, which has been suc-
cessfully used at frequencies up to 343GHz. As with any
resonant cavity, the resonant frequency of the quasi-opti-
cal cavity is perturbed by the insertion of the specimen.
The specimen is intentionally mounted at an angle, y, to
the vertical axis of the cavity, in order to eliminate stand-
ing-wave phenomena. The angle permits the waves re-
ected from the airdielectric interface to escape from the
resonator. The resonance is restored by reducing the dis-
tance between the two mirrors by an amount equal to D.
The real value of the index of refraction n
0
of the specimen
(39) is then given by
n
0
e
0
r
1=2
1
D
d
s
j
b
0
d
s
_ _
79
where d
s
is the thickness of the dielectric specimen, b
0
is
the phase factor in free space and is equal to 2p/l
0
, and the
angle j is dened by
j tan
1
sin2n
0
b
0
d
s
n
0
1
n
0
1
_ _
cos 2n
0
b
0
d
s
_
_
_
_
80
For tan d measurements, the length of the empty qua-
sioptical confocal resonator must be adjusted to an odd
number of half-wavelengths at the resonant frequency for
which the Q value is to be determined. The specimen must
then be inserted at a position vertical to the axis of the
resonatorthat is, with y 0, the Q value is maximized as
the escape of the power from the resonator is minimized.
With the resonant frequency restored as each mirror is
moved inward a distance, D=2, the quality factor, Q
s
, with
the specimen inserted is then determined. The expression
for the dissipation factor [39] follows as
tand
b
0
bD d
s
1
sinb
0
D d
s
b
0
D d
s
_ _
Q
s
Z
2
e
0
r
2
b
s
d
s
1
sinb
s
d
s
b
s
d
s
_ _
1
Q
s
Q
0
l
0
=2p
Z
2
e
0
r
2
b
s
d
s
1
sinb
s
d
s
b
s
d
s
_ _
_
_
_
_
81
where b
s
is the phase constant in the dielectric medium
and is equal to 2p/l
s
; the value of Z is
Z
n
0
2
ot
2
b
0
x
1
=n
0
2
1 ot
2
b
0
x
1
_ _
82
where x
1
denotes the distance from the reector to the di-
electric sheet.
In the frequency range from 300 to 3000GHz (wave-
lengths of 1mm100 mm), true optical measurement tech-
niques are employed. As this wavelength range overlaps
the infrared region, infrared sources and detectors are
utilized. If a broadband radiation source is employed, the
component measurement frequencies, appearing at the
output of an interferometer, are selected by means of a
computer in terms of their Fourier components. Broad-
band radiation sources require more sensitive detectors; it
is for this reason that laser sources, although monochro-
matic, appear to be more popular.
Figure 22 depicts an arrangement for the measurement
of dielectric absorption at optical frequencies, utilizing a
laser source [40]. The attenuation of the transmitted sig-
nal is obtained by comparing the amplitude of the signal
transmitted through the specimen V with a monitored in-
cident signal V
m
. The method entails the use of different
specimen thicknesses d
s
, which requires adjustment of the
polarizing attenuator, in order to maintain a constant
transmission loss.
The dissipation factor is related to the absorption coef-
cient a
p
, which obtained from the relation [40]
ln Aa
p
d
s
constant 83
the units of a
p
are in nepers per cm; A is the reading of the
attenuator, which is equal to (cos j)
4
; here j is the central
polarizer angle of the attenuator. From the nature of
Eq. (83), it is apparent that the absorption coefcient a
p
can be obtained directly from a linear plot of in lnAversus
the specimen thickness d
s
. The imaginary part of the in-
dex of refraction is equal to ca
p
/4pf, where f is the fre-
quency. Hence, the relative real value of the permittivity is
given by
e
0
r
n
0
ca
p
4pf
_ _
2
84
Mirror
Mirror
Dielectric
l
l
d
s
Figure 21. Quasioptical confocal resonator arrangement [39].
and the relative imaginary value of the permittivity is
e
0 0
r
cn
0
a
p
4pf
85
The dissipation factor, which is equal to the ratio e
0 0
r
=e
0
r
is
then
tand
8pfn
0
ca
p
4pfn
0
2
ca
p
2
86
The foregoing approach is based on transmission tech-
niques, but laser source reection arrangements are also
available. It should be observed that there are a number of
variations in the types of interferometers available for di-
electric measurements, including the classical Michelson
interferometer, which, in conjunction with a broadband
radiation source, is suitable for measurements up to
3000 GHz. Laser, refraction measurements, based on the
MachZehnder approach, may also be employed to derive
dielectric data. A comparison of the various optical mea-
surement techniques at a large number of laboratories in-
dicates that, whereas the real value of the index of
refraction, n
0
, may be determined to an accuracy of 1%,
the errors in the measurement of the absorption coef-
cient, a
p
, may be as high as 37% [41].
4. VOLTAGE BREAKDOWN STRENGTH MEASUREMENTS
Voltage breakdown strength measurements are carried
out on insulating materials to determine whether these
materials can withstand certain operating stresses with-
out failure. Since voltage failure is frequently initiated at
fault sites within solid insulating materials, the dielectric
strength serves as an indicator of the homogeneity of the
material. In liquid dielectrics, low dielectric strength val-
ues may be associated with moisture content, electrolytic
contamination, and a high particle content. With gases for
which the dielectric strength is a definite function of the
composition of the gas (pure or mixture), dielectric
strength data may be used to detect contamination from
other gases, as well as determine the breakdown charac-
teristics of intentionally combined gas mixtures.
The dielectric strength of insulating materials is highly
contingent on the geometry of the test electrodes utilized.
Sharp accentuated electrode edges lead to electrical eld
concentrations at the edges, which cause initiation of volt-
age breakdown of the material at voltages substantially
below those that can be achieved under more uniform
electrical eld conditions. Thus voltage breakdown data
are inextricably associated with the specific geometry of
the test electrodes employed.
The true value of the breakdown strength or, more spe-
cifically, the intrinsic breakdown strength of the dielectric
is obtained when the applied electric eld is perfectly uni-
form. A uniform eld can be achieved by means of Rogow-
skiRengier prole electrodes; however, the application of
such recessed-type electrodes to solid specimens requires
the embedding of the electrodes into the solid dielectric by
means of a suitable molding process when plastic dielec-
trics specimens are tested. Several relatively simple re-
cessed electrode systems have been developed, which do
not entirely produce a uniform eld, but that improve the
electrical eld conguration appreciably, thereby permit-
ting the attainment of dielectric strengths approaching
the intrinsic value. One such simplied arrangement is
depicted in Fig. 23 [42].
Phase
sensitive
amplifier
Detector
Beam divider
(polyethylene terephthalate)
Polarizing
attenuator Specimen
Polymer
lens
Polymer
lens
Golay
cell detector
Golay
cell detector
Laser
radiation
Modulator
(15 Hz)
Filter
Differential
amplifier
Amplifier
Indicator
V
m
V
d
s
In (cos )
D
Figure 22. Dielectric absorption
measurement system at optical fre-
quencies with a laser radiation
source [40].
The recess in the rigid solid dielectric may be machined
to form a highly stressed region in the specimen, with the
electrodes vapor deposited on the dielectric to preclude
any air gaps between the electrodes on the specimen.
Alternatively, with nonporous solid dielectrics, conducting
silver paint may be applied. Should sparkover occur at the
edges prior to dielectric breakdown, the entire electrode
assembly may be immersed in a mineral or silicone oil
bath, provided the solid specimen is a nonporous material.
The conductivity s
m
and dielectric constant e
0
m
of the im-
mersing medium must be selected such that under DC test
voltages [43]
s
m
E
m
> s
s
E
s
87
where E is the electric eld and the subscripts m and s
refer, respectively, to the oil medium and the solid speci-
men. Under alternating test voltages, we obtain
e
0
m
E
m
> e
0
s
E
s
88
If the liquid is partially conducting, then
e
0
m
E
m
sec d
m
> e
0
s
E
s
sec d
s
89
where d is the loss angle.
Although the intrinsic strength of a dielectric material
provides information on the maximum breakdown
strength attainable for that material and thus is used to
ascertain the nature of the mechanism responsible for the
breakdown, it is, per se, of little consequence in practice.
In fact, the intrinsic breakdown strength is usually one to
several orders of magnitude higher than the electrical
breakdown stress obtained with regular parallel-plane
electrodes, or with the various electrical insulation con-
gurations existing in different electrical apparatus. For
this reason, the type of electrodes used in standard routine
breakdown tests on materials are relatively simple to use,
and are designed to provide reproducible results primarily
for comparison purposes.
4.1. Electrode Systems for Routine Breakdown Tests
on Solid Specimens
Present practice in assessing the breakdown strength and
the quality of solid, liquid, and gaseous dielectric materi-
als for use in electrical apparatus and cables involves the
use of a number of electrode systems, in accordance with
national and international standards. For solid materials,
the most commonly employed electrode system is the one-
inch or 25-mm two-cylindrical electrode equal-diameter
system portrayed in Fig. 24 [44]. The edges of the elec-
trodes are rounded to a radius of 3.2mm, to minimize
stress enhancement. In all dielectric strength tests, the
thickness of the specimen must be specied, because the
voltage stress, at which breakdown occurs, increases with
a reduction of the specimens thickness. Thus, although
very thin solid dielectric lms may break down at low
voltages, the corresponding breakdown stresses are ap-
preciably higher than those for thick lms of the same
material, even though the latter may undergo breakdown
at much higher applied voltages.
It is evident that equal-diameter electrodes systems
must be concentric. This requirement may be circumvent-
ed by the use of two electrodes with different diameters,
in accordance with IEC Publication 243, as depicted in
Fig. 25 [45]. Note that the IEC (International Electrotech-
nical Commission) standard species dielectric specimen
thicknesses equal to or less than 3.070.2mm. If tapes of
reduced width are tested, then rod electrodes of the
geometry delineated in Fig. 26 are utilized. When
Specimen with
recessed section
Spherical electrode
HV
Grounded cylindrical
electrode
Evaporated
metallic electrodes
Figure 23. Spherical HV electrode with recessed solid speci-
men [42].
25 mm
25 mm
25 mm
H.V.
2 mm
Brass or
stainless steel
electrode
Dielectric
specimen
3.2 mm
Figure 24. Equal-diameter electrode system for dielectric
strength measurement on sheet materials (after ASTM D149) [44].
Brass or
stainless steel
electrode
Dielectric
specimen
H.V.
25 mm
15 mm
75
+
1 mm
25
+
1 mm
3
+
0.2 mm
3 mm
3 mm
Figure 25. Unequal-diameter electrode system for dielectric
strength measurements on sheet materials (after IEC Publica-
tion 243) [45].
breakdown tests are carried out on thin inorganic lms
with application to electron devices, miniature counter
electrodes are vapor deposited onto the surface of the
specimen. For the evaluation of embedding compounds
or greases, the standard procedure of ASTM D149 re-
quires hemispherical electrodes, having an equivalent di-
ameter of 12.7 mm [44].
The foregoing described electrode systems for solid di-
electric specimens are suitable for tests under AC power
frequency, DC, and impulse conditions. The electrode sys-
tems, for routine determination of the dielectric strength
of liquids, differ from those described for solids. Routine
acceptance tests on oils of petroleum origin for electrical
apparatus and cables are carried out with an oil cup con-
taining parallel-plane polished brass electrodes, with an
interelectrode spacing of 2.570.01mm. The electrodes
have a diameter of 25 mm and a thickness of 3 mm; they
are square at the edges and are separated from the inner
wall of the oil test cup by a distance of not less than
13 mm. The oil test cup assembly is shown in Fig. 27 [46].
The electrodes within the cell must be cleaned with a dry
hydrocarbon solvent following each breakdown test; par-
ticular care must be taken to remove any carbonization
deposits on the electrodes, and the electrodes must be re-
polished should any pitting of the surface manifest. Prior
to admitting the liquid specimen into a cleaned test cell,
the latter must be rinsed by the same liquid to remove any
residues of the cleaning compound.
It is palpably evident from the geometric conguration
of the square-edge electrodes in Fig. 27, that electrical
stress enhancement occurs at the edges of the electrodes
and that, therefore, breakdown is likely to occur there. For
lower viscosity dielectric liquids (o19 cSt or mm
2
/s at
401C), test electrodes, with the geometrical contour de-
picted in Fig. 28, have been found to be particularly effec-
tive in detecting decreases in the breakdown strength as a
result of cellulose ber contamination and absorbed mois-
ture [47]. These electrodes are normally referred to as
VDE (Verband Deutscher Electrotechniker)-type elec-
trodes. Measurements of dielectric strength are performed
with electrode separations of either 1 or 2mm, with a
gentle downward oil ow at the electrodes created by
means of a rotating impeller located beneath the elec-
trodes in the test cell.
Since oil-lled and impregnated electrical power equip-
ment is subjected to lightning and switching impulses, it is
important to assess the quality of the oil in terms of its
impulse breakdown strength. Under nonuniform electri-
cal eld conditions, the dielectric strength of the oil is
contingent on the polarity of the impulse in contradistinc-
tion to negligible differences observed under uniform
elds. For this reason, nonuniform eld electrode systems
are frequently utilized for impulse tests. The electrodes
may typically consist of either two 12.7 mm diameter brass
or steel spheres or, for highly nonuniform elds, one such
sphere and a steel needle point with a 0.06 mm radius of
curvature at the needle tip [48].
The breakdown strength of gases is normally deter-
mined under quasiuniform AC eld conditions. Typical
electrodes utilized for this purpose consist of a sphere-to-
plane geometry, wherein the electrical eld is uniform
directly underneath the sphere adjacent to the plane, be-
coming increasingly less uniform as the separation
between the sphere and the plane increases. With a
sphere-to-plane geometry, electrical breakdown tends to
always occur in the uniform eld regionthat is, at the
point where the separation between the sphere and the
plane is least. The high-potential sphere electrode may
Brass or stainless
steel cylindrical rods
0.8 mm
radius
Dielectric tape
specimen
Figure 26. Cylindrical rod electrodes for dielectric strength mea-
surements on thin narrow plastic tape or other narrow specimens
(after ASTM D149) [44].
Brass disk electrodes
Plastic
container
Gap
adjustment
Oil
25 mm
25 mm
3 mm
2.50
.01 mm
Figure 27. Parallel-plane square-edge electrode system for di-
electric strength measurements on mineral oils (after ASTM
D877) [46].
4 mm
radius
Brass spherically
capped electrodes
25 mm
radius
13 mm
36 mm diameter
Figure 28. VDE electrode system for dielectric breakdown
strength measurements on low-viscosity liquids [47].
be of steel with a diameter of 0.75 in. or 19.1 mm, and the
ground potential electrode may be a cylindrical brass
plane with a 1.5inch or 38.1mm diameter [49]. The tests
are performed at 251C at a standard pressure of 760torr.
4.2. Voltage Breakdown Test Conditions and Procedures
The presence of lethal voltages in breakdown voltage tests
necessitates strict adherence to high-voltage safety prac-
tices. Since the breakdown voltage may be a function of
the ambient temperature, pressure, and humidity, de-
pending on where solid, liquid, or gas specimens are test-
ed, these parameters should be recorded at the time of the
test; solid insulating materials should be conditioned prior
to the breakdown test, so that they may reach thermal and
moisture equilibrium with the environment. For more
lossy solid and liquid specimens, the application of intense
alternating electrical elds may result in cumulative heat
generation due to dielectric losses, thereby leading to a
thermal instability induced breakdown. Solid specimens
may contain gas cavity inclusions, within which intense
recurring partial discharges at elevated alternating elds
may cause rapid deterioration of the adjacent solid insu-
lation, thus leading to conspicuously lower breakdown
strengths. Both thermal and discharge mechanism asso-
ciated breakdowns account for the lower observed AC
breakdown strengths, as opposed to those measured un-
der DC and impulse conditions. Where the breakdown
strength is controlled by the thermal and partial dis-
charge mechanism, the breakdown process is a strong
function for the time of voltage application. Accordingly,
the rate of voltage rise in any voltage breakdown test is an
important parameter.
For solid dielectrics, the rate of ac sinusoidal voltage
rise is xed usually at 500 V/s, though more rapid or slow-
er rise rates may also be used. Breakdown or rupture of
the dielectric is indicated by an audible voltage collapse
across the specimen, as well as a visual burn at the tip of
the breakdown. In order to minimize stress-induced aging
effects in the insulation undergoing the voltage break-
down test, ASTM D149 stipulates that the duration of a
short-time breakdown must not exceed 20s. In the past, a
voltage step test was employed, whereby the voltage was
raised in steps; at each step it was maintained for a preset
time, prior to the next-step increment in voltage, until the
ensuence of dielectric breakdown eventthat is, an
abrupt voltage collapse across the specimen. The use of
the step procedure was required in the absence of voltage
sources with automatically regulated rate of voltage rise
controls.
In DC dielectric breakdown strength determinations on
dielectric material specimens, a single rate of voltage rise
of 500V/s is employed [50]. Under direct voltages, the ini-
tial breakdown event produces a minute channel in the
volume of the solid dielectric, whose trace is not readily
discernible. Reapplication of the direct voltage results in
successively lower breakdown voltages, which conrm
that a DC breakdown has already occurred. Also, the ad-
ditional damage and burning produced within the break-
down channel renders it more visible.
Impulse tests on solid dielectric specimens are per-
formed by increasing the peak voltage of the impulse
gradually, from an initial peak value of 0.7 times the an-
ticipated breakdown voltage [45,50]. The lightning im-
pulse is simulated with an impulse waveform having a
time to peak of 1.2ms and a decay time of 50 ms to 50% of its
initial peak value. Impulse breakdown is indicated by a
voltage collapse at any point of the impulse waveform [2];
the peak voltage value of this impulse wave is considered
as the impulse breakdown voltage. Location of the actual
breakdown channel caused by an impulse may require, as
in the DC case, several reapplications of the voltage pulse,
to cause additional carbonization within the breakdown
channel.
In the measurement of dielectric strength of liquid
specimens at AC power frequencies using the parallel-
plane square-edge electrodes, a xed voltage rise of 3kV/s
is generally specied [46]. To avoid pitting of the test elec-
trode surfaces, the short-circuit current at breakdown in
the specimen is not permitted to exceed 10 mA/kV. For
tests with the same electrode system under direct voltag-
es, the same rate of voltage rise should be adequate. When
the VDE-type electrodes are employed for low-viscosity
liquids at power frequency, a much lower rate of voltage
rise of 0.5kV/s is preferred. Impulse breakdown tests per-
formed on dielectric liquids are often carried out with both
the simulated lightning impulse of the 1.2 by 50 ms form
and a switching surge impulse form with a 100ms risetime
to peak and a decay time 41000 ms. The impulse break-
down tests are carried out either at positive or negative
polarities; often the measurements may be performed at
both polarities. The measurement sequence at either po-
larity is begun at a voltage substantially below the ex-
pected impulse voltage breakdown level. Normally, three
impulse waves are applied at each selected impulse volt-
age test level; it is an accepted practice to traverse at least
three test levels prior to breakdown, with a xed mini-
mum time interval between each voltage level test. ASTM
D3300 recommends a time interval of 30 s. The peak im-
pulse voltage at breakdown is measured oscillographically
across a calibrated resistive voltage divider. Whenever
needle electrodes are employed, the geometry of the nee-
dle tip is altered, due to the energy released by the break-
down spark; this necessitates a change of the needle
electrodes after each breakdown event.
Routine voltage breakdown strength measurements on
insulating gases are normally performed under AC power
frequencies, using a standard rate of voltage rise of 500V/s
[49]. The breakdown strength of gases is a function of gap
spacing and gas pressure; since the value of the latter
varies with the ambient temperature, both the pressure
and temperature must be recorded for breakdown results
obtained with a xed gap setting.
It should be emphasized that, when the breakdown
voltages are determined for solid, liquid, and gas speci-
mens, the gap length or specimen thickness must be stat-
ed in each case. Even when the value of the voltage
breakdown strength is provided in the units of voltage
per unit specimen thickness, the specimen thickness must
still be specied, because the breakdown strength is a
function of the specimen thickness. Also, the dispersion in
the voltage breakdown data requires some form of statis-
tical analysis. Breakdown strength data ordinarily refer to
a mean measured value on 10 specimens. Alowratio (about
0.1) of the standard deviation to the mean value, derived
from the ten measurements, is usually considered as an
indicator of an acceptable probable error in the test results.
5. CONCLUDING REMARKS
The foregoing presentation of the measurement of con-
ductivity, permittivity, and dielectric loss, and dielectric
strength of electrical insulating materials has attempted
to provide a concise cursory description of a number of the
most common measurement techniques in use. Space lim-
itations prevented a discussion on the various conduction
and breakdown mechanism and their inuence on the
measured quantities. For a more in-depth discussion on
the mechanisms involved and their determining inference
on the measured quantities obtained with a variety of dif-
ferent test methods, the reader is referred to other liter-
ature [2,18,20,42].
BIBLIOGRAPHY
1. R. Bartnikas, Dielectrics and insulators, in R. Dorf, ed., The
Electrical Engineering Handbook, CRC/IEEE Press, Boca Ra-
ton, FL, 1997, Chap. 55.
2. R. Bartnikas, ed., Engineering Dielectrics, Vol. IIB, Electrical
Properties of Solid Insulating Materials: Measurement Tech-
niques, STP926, ASTM, Phildelphia, 1987.
3. ASTM D257, Standard Test Methods for D-C Resistance or
Conductance of Insulating Materials, Annual Book of ASTM
Standards, Vol. 10.01, Philadelphia, 1997.
4. IEC Publication 93, Methods of Test for Volume Resistivity
and Surface Resistivity of Solid Electric Insulating Materials.
5. ASTM D374, Standard Test Methods for Thickness of Solid
Electric Insulation, Annual Book of ASTM Standards, Vol.
10.01, Phildelphia, 1997.
6. Low Level Measurements, 4th ed., Keithley Instruments,
Cleveland, OH, 1993.
7. B. V. Hamon, An approximate method for deducing dielectric
loss factor from direct current measurements, Proc. IEE
99:151155 (1952).
8. H. Block et al., Treatment of data in step-response, dielectric
relaxation measurements, J. Chem. Soc. Faraday Trans. II
68:18901896 (1972).
9. H. St-Onge, Electrical conduction in 3-percent carbon lled
polyethylenePart I: Low eld results, IEEE Trans. Electric
Insul. EI-11:2027 (1976).
10. P. J. Hyde, Wide-frequency range dielectric spectrometer Part
II, Proc. IEE 117:18911901 (1970).
11. F. I. Mopsik, Precision time-domain dielectric spectrometer,
Rev. Sci. Instrum. 55:7987 (1984).
12. A. M. Thompson, A bridge for the measurement of permit-
tivity, Parts B and C, Proc. IEE 103:705709 (1955).
13. W. P. Harris, Operators Procedures Manual for the Harris
Ultra-Low Frequency Impedance Bridge, National Bureau of
Standards Report 9627, Washington, DC, 1968.
14. R. J. Kriegler and R. Bartnikas, Dielectric loss and current-
voltage measurements in sodium-contaminated Si-SiO
2
Cr
structures, IEEE Trans. Electron. Devices ED-20:722731
(1973).
15. J. Pugh, Proc. 4th Int. Conf. Dielectr. Mater. Meas. Appl., IEE
Conf. Publ. 239, Lancaster, UK, Sept. 1013, 1984, pp. 247249.
16. ASTM D150, Standard Test Methods for AC Loss Character-
istics and Permittivity (Dielectric Constant) of Solid Electrical
Insulating Materials, Annual Book of ASTM Standards,
Vol. 10.01, 1998.
17. H. J. Wintle and S. Kurylowicz, Edge corrections for strip and
disc capacitors, IEEE Trans. Instrum. Meas. IM-34:4147
(1985).
18. R. Bartnikas, ed., Engineering Dielectrics, Vol. III, Electrical
Insulating Liquids, Monograph 2, ASTM, Philadelphia, 1994.
19. R. Bartnikas and K. D. Srivastava, eds., Power Cable Engi-
neering, Sandford Fleming, Waterloo, Ontario, 1987.
20. R. Bartnikas and E. J. McMahon, eds., Engineering Dielec-
trics, Vol. I, Corona Measurement and Interpretation, STP
669, ASTM, Philadelphia, 1979.
21. A.G. Tettex, Prazisions-Verlustfactor und Kapazitats MeB-
rucke Typ 2821, Prospekt 2821, Blatt 103, Zu rich 1980.
22. O. Petersons, A self-balancing high voltage capacitance
bridge, IEEE Trans. Instrum. Meas. IM-13:216224 (1964).
23. Bridges/Standards, Gen Rad Bulletin JN 4240-681, Gen Rad,
Boston, MA.
24. P. Osvath and S. Widmer, A high-voltage high-precision self-
balancing capacitance and dissipation factor-measuring
bridge, IEEE Trans. Instrum. Meas. IM-35:1923 (1986).
25. ASTM D1531. Test for Dielectric Constant and Dissipation
Factor of Polyethylene by Liquid Displacement Procedure,
Annual Book of ASTM Standards, Vol. 10.01, 1997.
26. L. Hartshorn and W. H. Ward, The measurement of the per-
mittivity and power factor of dielectrics at frequencies from
10
4
to 10
8
cycles per second, J. IEE 79:597609 (1936).
27. W. Reddish et al., Precise measurement of dielectric proper-
ties at radio frequencies, Proc. IEE 118: 255265, (1971).
28. K. A. Buckingham and J. W. Billing, Proc. 3rd Int. Conf.
Dielectric Measurement Applications, Birmingham, England,
Sept. 1013, 1979, pp. 392395.
29. T. V. L. Parry, The measurement of permittivity and power
factors of dielectrics at frequencies from 300 to 600c/s, part
III, J. IEE 98:303311 (1951).
30. S. Roberts and A. von Hippel, A new method for measuring
dielectric constant and loss in range of centimeter waves,
J. Appl. Phys. 17:610616 (1946).
31. W. B Bridges, M. B. Klein, and E. Schweigh, Measurement of
dielectric constant and loss tangent of thallium mixed halide
crystals KRS-5 and KRS-6 at 95GHz, IEEE Trans. Microwave
Theory Tech. MTT-30:286292 (1982).
32. ASTM D2520, Test Methods for Complex Permittivity of Solid
Electrical Insulating Materials at Microwave Frequencies and
Temperatures to 16501C, Annual Book of ASTM Standards,
Vol. 10.01, 1998.
33. W. B. Westphal, Distributed circuits, in A. von Hippel, ed.,
Dielectric Materials and Applications, Wiley, New York, 1954,
Chap. 2, Sect. A2.
34. T. W. Dakin and C. Works, Microwave dielectric measure-
ments, J. Appl. Phys. 18:789796 (1947).
35. A. C. Lynch, Transmission methods for measurement of di-
electric loss, Conf. High Frequency Dielectr. Meas., Natl. Phys-
ical Laboratory, Teddington, UK, March 2729, 1972.
36. IEC Publication 377-2, Measurement of Permittivity and Loss
at Frequencies above 300MHz.
37. A. H. Sharbough and S. Roberts, Dielectric measurement
procedures, in K. Lark-Horowitz and V. A. Johnson, eds., Sol-
id Slate Physics, Vol. VI, Part B, Academic Press, New York,
1959, Chap. 7, Sect. 71.
38. W. Culshaw, Aspectrometer for millimetre wavelengths, Proc.
IEE, Part IIA, 100:514 (1953).
39. J. E. Degenford, A quasi-optic technique for measuring di-
electric loss tangents. IEEE Trans. Instrum. Meas. IM-
17:413417 (1968).
40. J. Chamberlain, Submillimetre-wave techniques, Conf. High
Frequency Measurement, Natl. Physical Laboratory, Ted-
dington, UK, March 2729, 1972.
41. J. R. Birch et al., An intercomparison of measurement tech-
niques for the determination of the dielectric properties of
solids at near millimeter wavelengths, IEEE Trans. Micro-
wave Theory Tech. 42:956965 (1994).
42. J. K. Nelson, in R. Bartnikas and R. M. Eichhorn, eds.,
Engineering Dielectrics, Vol. IIA, Electrical Properties of
Solid Insulating Materials: Molecular Structure and Electrical
Behavior, STP 783, ASTM, Philadelphia, 1983.
43. S. Whitehead, Dielectric Breakdown of Solids, Clarendon
Press, Oxford, 1953.
44. ASTM D149, Test Method for Dielectric Breakdown Voltage
and Dielectric Strength of Electrical Insulating Materials at
Commercial Power Frequencies, Annual Book of ASTM Stan-
dards, Vol. 10.01, 1997.
45. IEC Publication 60, 243, Parts 13, Electric Strength of In-
sulating MaterialsTest Methods.
46. ASTM D877, Standard Test Method for Dielectric Breakdown
Voltage of Insulating Liquids Using Disk Electrodes, Annual
Book of ASTM Standards, Vol. 10.03, 1997.
47. ASTM, D1816, Standard Test Method for Dielectric Break-
down Voltage of Insulating Oils of Petroleum Origin Using
VDE Electrodes, Annual Book of ASTM Standards, 1997,
Vol. 10.03; see also VDE (Verband Deutscher El-
ektrotechniker) Specication 0370.
48. ASTM D3300, Standard Test Method for Dielectric Break-
down Voltage of Insulating Oils of Petroleum Origin under
Impulse Conditions, Annual Book of ASTM Standards,
Vol. 10.03, 1997.
49. ASTM D2477, Standard Test Method for Dielectric Strength
of Insulating Gases at Commercial Power Frequencies, Annu-
al Book of ASTM Standards, Vol. 10.03, 1997.
50. ASTM D3755, Test Method for Dielectric Breakdown Voltage
and Dielectric Strength of Solid Electrical Insulating Materi-
als under Direct Voltage Stress, Annual Book of ASTM Stan-
dards, Vol. 10.02, 1997.
DIELECTRIC PERMITTIVITY AND LOSS
D. K. DAS-GUPTA
University of Wales
Bangor, Wales
Dielectrics can be dened as materials with high electrical
resistivities that conduct virtually no electricity at low DC
electric elds. A large group of materials, including gases,
liquids, semiconductors, ceramics, and organic and inor-
ganic polymers, are classied as dielectrics. There are,
however, no perfect dielectric materials.
The study of the electrical properties of dielectrics aris-
es from their practical need for efcient electrical insula-
tion requirements for long operational life. Many
dielectric materials are classied by their electrical break-
down strength, dielectric loss, permittivity, and polariza-
tion, and these macroscopic properties are related to their
atomic and molecular structures. Although dielectrics are
widely employed in diverse applications (e.g., capacitors,
cables, transformers, motors), the study of dielectrics has
progressed very little since the early investigation of fer-
roelectric phenomena. However, the advent of microelec-
tronics and complex control devices and components in
defense and industrial applications has made dielectric
research important in its own right.
The present article briefly reviews the electrostatic
concepts that lead to time- and frequency-dependent di-
electric phenomena together with the models of dielectric
relaxation behavior in various materials. It also includes
some explanations for the dielectric aging of insulating
materials under high elds in humid environments.
1. DIELECTRIC POLARIZATION
1.1. Static Field
When an electric eld is applied to a dielectric material,
three processes can occur:
1. A steady ow of direct current (due to the DC con-
ductivity s
0
) may occur if free charges are capable
of moving throughout the volume without restraint.
2. Bound charges can form dipoles by aligning with the
eld and provide polarization. On removal of the
eld the dipoles may return to their original random
orientations with the help of thermal energy, giving
rise to dielectric relaxation.
3. Electronic and ionic charges may hop through the
defect sites. These charges are neither free nor
bound, and they give rise to an intermediate form
of polarization, which involves nite charge storage.
Dielectrics may broadly be divided into nonpolar and
polar materials. In nonpolar materials in an external eld
a dielectric polarization occurs when the positive and neg-
ative charges experience an electrical force that causes
them to move apart in the direction of the external eld.
As a result, the centers of positive and negative charges no
longer coincide. The molecules are then said to be pola-
rized, and each molecule forms a dipole and acquires a
dipole moment p, dened thus
pe dl 1
where e is the electronic charge and dl the displacement
(B10
10
10
11
m in magnitude) between the two charge
centers. Note that dl is a vector that points from the neg-
ative to the positive charge. Such dipoles are called in-
duced dipoles. On removal of the eld, the charges are
redistributed and the dipole moment vanishes.
938 DIELECTRIC PERMITTIVITY AND LOSS
With polar dielectrics, which lack structural symmetry,
the charge centers of opposite polarities do not coincide for
a molecule even in the absence of an electric eld. Howev-
er, these molecular dipoles may be randomly distributed,
thus summing to a zero dipole moment over any macro-
scopic volume element [19]. In the presence of an appro-
priate electric eld, the molecules may align themselves in
the eld direction and thus provide a net dipole moment.
Macroscopically, the electric eld in a dielectric is de-
scribed [2] by the electric eld strength E (V/cm) and the
electrical displacement density, also known as the electric
ux density, D (C/m
2
), where both D and E are vector
quantities. Now the polarization can be dened as the
dipole moment per unit volume, namely
P
N
i 1
p
i
=Dn 2
and is also a vector quantity. It should be noted that the
normal component of P at the surface equals the surface
charge density per unit area. These three vectors, D, E,
and P, in a material medium other than vacuum, are
related thus
De
0
EP 3
or
DeE 4
where
e e
0
e
r
5
where e
0
is the permittivity in free space (8.8510
12
F/
m) and e
r
is the relative permittivity (dimensionless) or
the dielectric constant of the material, which takes into
account the polarization effect and is dened as
e
r
C
C
0
6
where C
0
is the capacitance of a capacitor with a vacuum
between two conductors and C the capacitance when the
same region is lled with the dielectric. e
r
is independent
of the shape or size of the conductors and is entirely a
characteristic of the particular dielectric medium. Table 1
[4] gives the values of e
r
for static or low-frequency (o1-kHz)
elds of several materials. e
r
, which is a macroscopic and
directly measurable parameter, is connected with the mi-
croscopic structure of a dielectric material and with its
polarization behavior.
From Eq. (3), we have
De
0
1
P
e
0
E
_ _
E
e
0
1 wE
e
0
e
r
E
7
where
e
r
1 w and w
P
e
0
E
8
and w is the dielectric susceptibility. Thus the parameter w
also provides a link between the macroscopic properties
and the atomic molecular theory of dielectric materials.
We may also write a general relation between P and E
thus (8)
PewEhigher terms in E 9
where the higher terms in E are applicable to the phe-
nomenon of hyperpolarization. It should be noted that w is
the ratio of bound charge density to free charge density of
a capacitor. A measurement of e
r
and hence w provides the
magnitude of the polarization P of a dielectric material at
any particular eld E.
One of the most useful methods of determining P is
to measure the current density J(t) as a function of time,
as [8]
Jt
dD
dt
10
It may be shown [3,8] that for noninteracting dipoles, w
is given by
w0
NP
2
3e
0
kT
11
Table 1. The Relative Permittivity of Some Solid
Dielectrics at 251C
Dielectric e
r
Vacuum 1 (by definition)
Air (atmospheric pressure, 01C) 1.0006
Amber 2.7
Borosilicate glass 4.0
Corning glass 0010 6.68
Corning glass 0014 6.78
Pyrex glass 45
Quartz (fused) 3.8
Diamond 5.5
Porcelain 5.5
Marble 1015
Mica 611
Steatite 6
Polyethylene 2.25
Polyvinylchloride (PVC) 6
Epoxy resin 3.611
Rubber 34
Neoprene 7
Beeswax (white) 2.65
Beeswax (yellow) 2.73
Parafn wax 2.3
Barium titanate 1200
Source: Ref. 4.
DIELECTRIC PERMITTIVITY AND LOSS 939
where w(0) is the static susceptibility in the zero-frequency
limit, N the number density of polarizable molecules, k the
Boltzmann constant, and T the temperature. w is a dimen-
sionless and scalar quantity in an isotropic medium.
2. MICROSCOPIC CONCEPTS OF POLARIZATION
We shall consider here the three cases of electronic, ionic,
and orientational polarizations.
An isolated neutral atom in an electric eld acquires a
dipole moment when an external electric eld produces a
separation of the charge centers of opposite polarities.
This is known as the electronic polarization, and it pro-
vides an induced dipole moment
p
ind
a
e
E 12
where a
e
is the electronic polarizability of an atom and is
given by
a
e
4pe
0
r
3
0
13
where r
0
is the radius of the spherical of an electron cloud
surrounding an atomic nucleus. The molar polarizability
P of a monoatomic gas is given by
PN
0
4p
3
r
3
0
14
where N
0
is the number of molecules in a gram molecule
(Avogadros number). The lowest polarizability belongs to
the noble gases with their completely lled outer electron-
ic shells, which screen the nuclei from the effect of the ex-
ternal electric eld. For hydrogen, with r
0
0.53
10
10
m, a
e
is 1.6610
41
F/m
2
. Hence for a eld E of
10
5
V/m
1
, aE10
36
C/m. The length l of this induced di-
pole is p/e 10
17
m (where e is the electronic charge),
which is indeed an extremely small distance compared
with atomic dimensions [6].
Ionic polarization occurs in ionic substances, such as
alkali halides, whose molecules are formed of atoms that
have excess charges of opposite polarities. In an external
eld the relative positions of the positive and negative ions
of a molecule may shift, thus introducing the dipole mo-
ment in addition to the induced electronic polarization.
The ionic polarization p
i
is given by
p
i
a
i
E 15
where a
i
is the ionic polarizability of the molecule, which
arises from the ionic displacement. The alkali halides (ha-
lides of the group I elements) have the highest polariz-
abilities, possibly because of the single electron in their
outermost shells. Table 2 [6] provides a list of values of
contribution of ions to the molar polarization of typical
alkali halides.
The third type of polarization, known as the orienta-
tional polarization, is associated with permanent dipoles
in dipolar materials that possess a dipole moment even in
the absence of an externally applied electric eld. However,
such a moment may not be observed macroscopically, as
the thermal energy will randomize the dipoles so that the
average moment will be zero over a small physical volume.
On an application of an external electric eld the dipoles
will experience a torque, which will orient them in the
eld direction so that the average dipole moment will no
longer be zero. It may be shown [3] that the orientational
polarizability a
0
is given by
a
0
p
2
3kT
16
By adding the three polarizabilities mentioned above,
the total polarization P can now be written as the sum of
the three components
Pp
e
p
i
p
0
N a
e
a
i
p
2
o
3kT
_ _
E 17
where N is the number of contributing molecules or par-
ticles per unit volume.
Of course, not all atoms or molecules need display
each of these three types of polarizability. Only the
orientational polarization is temperature-dependent.
Equation (17) is known as the LangevinDebye formula,
and we have
w e
r
1
P
e
0
E

N
e
0
a
e
a
i
p
2
o
3kT
_ _
18
Thus a measurement of e
r
as a function of temperature
may help to distinguish the orientational polarization con-
tribution from the sum of the components a
e
and a
i
, which
are practically independent of temperature.
Figure 1 shows (9) a plot of e
r
1 as a function of 1/T for
the molecule of methyl amine (CH
5
N). The intercept for
the line at 1/T0 and its slope are approximately 8
10
4
and 0.6K
1
, respectively. From Eqs. (8), (9), and
(17), we have (9)
a
e
a
i
810
4
e
0
N
% 6 10
4
F=m
2
19
Table 2. Ionic Polarization as a Fraction of the Total
Polarization for Alkali Halides, and (in Parentheses)
the Ionic Polarization of Each Compound Relative to that
of LiF
F Cl Br I
Li 0.68 (1.0) 0.53 (3.22) 0.49 (1.96) 0.40 (2.09)
Na 0.65 (1.12) 0.50 (3.38) 0.44 (1.82) 0.39 (2.13)
K 0.65 (1.95) 0.49 (4.25) 0.46 (2.38) 0.38 (2.34)
Rb 0.62 (2.19) 0.52 (5.14) 0.44 (2.46) 0.38 (2.62)
Source: Ref. 6.
and
p
2
o
0:6e
o
3k
N
% 4:2 10
30
C=m 20
The further separation of a
e
and a
i
is not possible using
this technique alone.
Table 3 gives the electric dipole moments of some mol-
ecules. The commonly used unit of dipole moment is the
debye; 1 D3.3310
30
C/m.
Space charge (interfacial) polarization generally arises
from a presence of electrons and/or ions that have limited
macroscopic motions in the bulk of a dielectric material.
Eventually these charge carriers are localized at lattice
defect sites, metalelectrode interfaces, impurity centers,
and voids. As a result, the electric eld in the dielectric
may become distorted, thus producing an apparent
increase in the dielectric constant. Space charge polariza-
tion is particularly evident in multiphase and inhomoge-
nous dielectrics, and its effect is dominant, particularly at
low frequencies, in practical dielectrics such as impreg-
nated paper, polymers, and sintered ceramics.
The study of dielectric polarization and susceptibility
in liquids and solids is more complicated than in gases
because of the interactions between the atoms and mole-
cules in the condensed phase. These atoms and molecules
will still exhibit electronic, ionic, and orientational polar-
izations. However, the effective local eld E
1
on an atom or
molecule in a liquid or a solid dielectric may not be the
same as the externally applied eld E. It is difcult to cal-
culate the effective local eld E
1
in the condensed phase
except for the most symmetric crystals.
Since Pe
0
(e
r
1) E, for the simplest case of a cubic
crystal, the Lorentz equation for the local eld E
l
(6) is
E
1
e
r
2E=3 21
and PNaE
l
, where a is the total polarizability and N the
number of molecules per unit volume. Hence
Pe
r
2NE=3 22
and
Na
3e
0
e
r
1
e
r
2
23
Equation (23) is the ClausiusMossotti equation, which
relates the microscopic property of the polarizability a
with the macroscopic property of the relative permittivity
or dielectric constant e
r
. Now we have
N
N
0
r
M
24
where N
0
is the Avogadro number, M the molecular
weight, and r the density. Substituting Eq. (24) into
Eq. (23), we obtain the molar polarizability per mole:
N
0
a
3e
0
M
r
e
r
1
e
r
2
25
Equation (25) should be used with caution, as it does not
take dipolar interactions into account properly. Equation (23)
may, however, be used to calculate the electronic
polarizability a
e
from the measured values of e
r
for dilute
gases, for which e
r
E1 and e
r
2E3. For such cases, we
have
a
r
e
0
e
r
1
N

e
0
w
N
26
T
1/T (K)
Methylamine
0
0.001
0.002
0.003
600 300 100C
0 0.001 0.002 0.003
c
r

1
Figure 1. A plot of e
r
1 as a function of 1/T for a molecule of
methylamine (CH
5
N) [9,10].
Table 3. Electric Dipole Moments of Some Molecules
Dipole moment
Molecule (10
30
C/m) (D)
HCl 3.5 1.05
CsCl 35.0 10.5
H
2
O 6.2 1.87
D
2
O 6.0 1.80
NH
3
4.9 1.47
HgCl
2
0.0 0
CCl
4
0.0 0
CH
4
O 5.7 1.71
Table 4 shows that the polarizability of the element ar-
gon does not vary significantly [9] between its dilute gas
and liquid states. That may not be true for other gases
with more extensive electronic structure when condensed
to liquid or solid form.
The polarizabilities of a
e
and a
i
may be determined in-
dependently for ionic crystals in the solid state. The rel-
ative permittivities e
r
of ionic crystals are frequency-
dependent. For an applied eld at low frequencies the val-
ue of e
r
will be dependent on both a
e
and e
i
, whereas in the
optical frequency range the lattice ions will not be able to
follow the applied eld and e
r
will only be e
e
. Table 5 shows
the static (low-frequency) and optical (high-frequency)
values of e
r
for some ionic crystals [9]. The difference be-
tween the values of e
r
in the second and the third columns
is the contribution of the ionic polarization alone, whereas
those in the third column characterize the contribution
due to a
e
.
The behavior of orientational polarization in dipolar
molecules in the gas and liquid phases may be quite large
if rotation of the dipoles is possible [10]. For such a case,
the polarizability will have contributions from a
e
, a
i
, and
a
0
, where the a
0
contribution will be temperature-depen-
dent, where e
r
increases with decreasing temperature.
However, as the temperature is lowered and the materi-
al solidies, the value of e
r
will drop abruptly when the
molecules can no longer rotate, and thus rotation cannot
contribute to the polarization. Figure 2a [9,11] illustrates
such behavior of e
r
for nitromethane. It may be observed
that at 244K, nitromethane freezes and e
r
drops abruptly
from 45 to just under 5. At this temperature a
0
is zero for
nitromethane and its polarizability arises from the a
e
and
a
i
contributions, which are independent of temperature.
However, there are some solids, such as HCl, that do not
show this type of behavior. For HCl in the liquid state e
r
is
large and increases with decreasing temperature, indicat-
ing rotational behavior of the molecules (see Fig. 2b).
However, below 165K, where HCl freezes, e
r
still contin-
ues to increase [9,12] because of the increase in the den-
sity of the material. At 100K, the molecular rotation
nally ceases and a
0
virtually becomes zero. The polariza-
tion contribution at this temperature in HCl originates
from a
e
and hindered rotation [9,12].
3. DIELECTRIC LOSS
3.1. Time-Dependent Dielectric Response
The dielectric behavior has been represented in the pre-
vious section by three vectors, D, E, and P, which are as-
sumed to be collinear in space and in the same phase in
time. However, neither of these two assumptions is nec-
essarily valid. We shall only consider the nature of the di-
electric behavior with time for nonpolar materials and
those containing permanent dipoles. Regarding the spa-
tial collinearity, extensive treatment of crystal symmetry
is necessary and will not be discussed here.
The time-dependent dielectric response may be synthe-
sized [8] from three fundamental time dependences of the
electric eld: the delta function d(t), the step function 1(t),
Table 4. Polarizability of Argon
Form T (K) Pressure (atm) e
r
a
e
(10
40
F/m
2
)
Gas 293 1 1.000517 1.83
Liquid 83 1 1.53 1.86
Source: Ref. 9.
Table 5. Static and Innite-Frequency Capacitivity of
Some Ionic Crystals
e
r
Material Static Optical
LiF 9.27 1.90
LiC 11.05 2.68
NaCl 5.62 2.32
KCl 4.64 2.17
RbCl 5.10 2.18
NaI 6.60 2.96
Source: Ref. 9.
180 220 260 300
60 80 100 120 140 160 180
T (K)
Melting temperature
HCI
f = 300 Hz
T (K)
50
40
30
20
10
0
Nitromethane
f = 70,000 Hz
Melting
temperature
c
r
50
40
30
20
10
0
c
r
(a)
(b)
Figure 2. The behavior of e
r
versus T for completely hindered (a)
and partly hindered (b) rotation of dipoles in the solid: (a) nitro-
methane [9,11]; (b) hydrogen chloride [9,12].
and the harmonic function sin o(t) or cos o(t), where o is
the angular frequency ( 2pf). Equation (3) may now be
represented thus
Dt e
0
EPt
where the rst term on the right-hand side provides the
instantaneous free-space contribution and the second the
delayed polarization. We dene a dielectric response func-
tion f(t). The polarization response to a delta function ex-
citation of strength E Dt is [8]
Pt e
0
EDtft 27
where E is the electric eld, acting over a time period DT.
From the principle of causality, we have
ft 0 for to0 28
In the absence of any permanent polarization, we obtain
lim
t!1
ft 0 29
Furthermore, from the principle of superposition, we
have [8]
Pt e
0
_
t
1
ft tEt dT 30
Equation (30) implies that the magnitude of polariza-
tion at a time t in a dielectric will depend on its past value;
in other words, the material has a memory. On an appli-
cation of an elementary step function eld E1(t), the
dielectric polarization is given by
Pt e
0
E
_
t
0
ft dt 31
The charging current I
c
(t) is given by [8]
I
c
t
dDt
dt
s
0
E
e
0
dEt
dt

dPt
dt
s
0
E
32
e
0
Edt ft s
0
E 33
where the delta function d(t) represents the instantaneous
free-space response of the step function eld, followed by
the polarization current dP(t)/dt of the material. s
0
is the
DC conductivity, if any, of the dielectric at innitely long
time. Thus, we have [8]
P1 e
0
E
_
1
0
ftdt e
0
w0E 34
where P(N) is the polarization with a steady electric eld
E after an innitely long time when the polarizing ele-
ments tend to be oriented along the eld direction. On
removal of this step function eld, a depolarization cur-
rent I
d
(t) will follow as the thermal agitation randomizes
the orientation of the dipoles with time. For this latter
case there will be no contribution of s
0
at E0.
3.2. Frequency-Dependent Dielectric Response
The polarization response to a harmonic eld is known as
the frequency-domain response. Taking the Fourier trans-
form of both sides of Eq. (34), we get
Po e
0
woEo 35
where P(o) and E(o) are the Fourier transforms of the
time-dependent polarization and eld, respectively. w(o) is
the frequency-dependent complex susceptibility, and it is
the Fourier transform of the time-dependent response
function f(t):
wo w
0
o iw
0 0
o
_
1
0
fte
iot
dt 36
The real part w
0
(o) provides the magnitude of polariza-
tion in phase with the harmonic driving eld E(o) and
does not contribute to the power loss, whereas the imag-
inary part w
00
(o), which is in quadrature with the eld, is
referred to as the dielectric loss. w
0
(o) and w
00
(o) may be
represented as odd and even functions of frequency, re-
spectively:
w
0
o
_
1
0
ft cos ot dt 37
w
0 0
o
_
1
0
ft sin ot dt 38
In terms of permittivity, we may write
Do eoEo e
0
1w
0
o iw
0 0
o Eo 39
For zero frequency, namely, the static case, we have
w
0
0
_
1
0
ft dt 40
and
w
0 0
0 0 41
Equation (36) shows that both w
0
(o) and w
00
(o) are func-
tions of the dielectric response function f(t), and these two
parameters are Hilbert transforms of each other, through
what are referred to as the KramersKronig relations:
w
0
o
1
P
C
_
1
1
w
0 0
o
x o
dx 42
w
0 0
o
1
P
C
_
1
1
w
0
x
x o
dx 43
where C denotes the Cauchy principal value of the inte-
gral. For the static case
w
0
0
2
p
_
1
1
w
0 0
x d ln x 44
Equation (44) indicates that the variation of the dielec-
tric parameters with frequency, specifically, the dielectric
dispersion, is an essential property of dielectric materials
[8]. It also shows that any mechanism that can lead to a
strong polarization in a dielectric material must also lead
to large losses in some frequency range. In other words, it
is not possible to have a loss-free dielectric with a nite
susceptibility [8]. In most dielectrics the loss is significant
only in limited frequency ranges. Figure 3 [8] shows sche-
matically two nonoverlapping loss processes at the low
frequencies and a resonance process in the optical fre-
quency range. In a limited frequency region we may dene
a high-frequency permittivity e
Na
, accounting for all
the processes occurring at higher frequencies; thus [see
Eq. (39)]
eo e
1a
e
0
w
a
0
o iw
a
0 0
o 45
from which we get
e
a
0
o e
1a
e
0
w
a
0
o 46
and
e
a
0 0
o e
0
w
a
0 0
o 47
For an alternating voltage the frequency-dependent
complex capacitance C(o) is
Co C
0
o iC
0 0
o 48
where C
0
(o) and C
00
(o) are the real and imaginary parts of
the complex capacitance. The loss angle d is the angle be-
tween the electric eld and the dielectric polarization. The
loss tangent
tan d
C
0 0
o
C
0
o

e
0 0
o
e
0
o
49
is independent of the geometry of the dielectric material.
The existence of the polarization with respect to the
eld leads to the energy dissipation in the dielectric. Now
the power dissipation P per unit volume is
PI
phase
E
where I
phase
is the part of the current in phase with E, and
is given by
PoE
2
e
0
e
0
tan d oE
2
e
0
e
0 0
50
Table 6 gives (6) typical values of the permittivity and
loss factor of various dielectric materials at room temper-
ature for different frequencies. Generally polar materials
have larger permittivities and loss tangents than do non-
polar materials. For many liquids the frequency at which
maximum energy loss occurs at room temperature is ap-
proximately 1000MHz (wavelength lE0.3 m), as shown
[1] for three typical liquids in Table 7, where t is the
relaxation time ( 1/f).
Another type of energy loss occurs in a resonance ab-
sorption process at very high (i.e., IR, visible, and UV)
frequencies. Although the real and imaginary parts of the
complex permittivity vary in a manner similar to that for
dipole relaxation, the origin of the energy loss is different
in this process. At optical frequencies the permittivity of
the dielectric is due almost entirely to the electronic po-
larization. In the absence of any external eld a vibrating
electron of charge e and mass m is elastically bound to its
nucleus by a restoring force, and its equation of motion is
m
d
2
x
dt
2
kx 0 51
where k is the restoring-force constant and x is the dis-
placement of the electron. This equation represents a sim-
ple harmonic motion, and its solution is
x x
0
sin o
0
t A 52
where o
0
(k/m)
1/2
, A is the integration constant, x
0
the
amplitude of oscillation, and o
0
the natural resonance an-
gular frequency of the oscillation. When an external al-
ternating electric eld is applied to this system, the
resulting motion is a forced oscillation, represented by
m
d
2
x
dt
2
mo
2
0
x e E cos ot 53
where E is the amplitude of the eld E and o is its fre-
quency. Clearly the response of the oscillating system will
now depend on both o and o
0
. The oscillations might be
expected to build up without limit when oo
0
, although
they are expected to be small at frequencies far away from
c
r
(
0
)
c
r
c
r1
c
r1
,c
r
1
logo
o
p1
o
p2
o
p3
Figure 3. Schematic diagram of the frequency dependence of the
real and imaginary parts of the complex susceptibility, showing
three processes; the last one is a resonance process [8].
o
0
. However, at resonance (i.e. oo
0
), the oscillations
will in fact be limited (damped) by the emission of elec-
tromagnetic radiation by the oscillating charges, which
dissipates energy. It may be shown that in a resonance
absorption process [6]
e
0
o e
0
Ne
2
m
o
2
0
o
2
o
2
0
o
2
r
2
o
2
54
and
e
0 0
o
Ne
2
m
ro
o
2
0
o
2
r
1
o
1
55
where r is a constant of the material, called the dissipation
constant. These quantities have the form shown in Fig. 3
at very high frequencies. For o0, which is the static
case, we have
e
0
0 e
0
Ne
2
mo
2
0
56
e
00
(o) approaches zero for both oco
0
and o{o
0
, and it
goes through a maximum value of Ne
2
(1/mro). Again e
00
(o)
represents an energy loss and the power loss P is given
again by
Poe
0 0
oe
0
E
2
57
As the characteristic values of o
0
for electron clouds are
very high, the resonance absorptions and their corre-
sponding energy losses occur at very high frequencies in
the IRUV range.
For pure nonpolar dielectrics, whether solid, liquid,
or gas, the polarization is of an essentially electronic
nature. Some polar materials with a highly symmetric
structure, like carbon tetrachloride (CCl
4
), may also
Table 6. Dielectric Properties of Materials
Relative permittivity e
r
Material Direction f 60Hz 100kHz 1 MHz 100MHz Loss tangent tan d
Crystals
Rutile, TiO
2
8
c
170 170 10
4
>
c
90 90 2 10
4
Aluminum oxide, Al
2
O
3
8
c
10.6 10.6 10.6
>
c
8.6 8.6 8.6
Lithium niobate, LiNbO
3
8
c
30 0.05
>
c
75
Ceramics
BaTiO
3
1600 15010
4
Alumina 4.58.5 0.00020.01
Steatite 5.57.5 0.00020.004
Rutile 14110 0.00020.005
Porcelain 68 0.0030.02
Polymers
Polytethylene 2.3 2.3 10
4
10
3
Polypropylene 2.1 2.5 10
4
PTFE 2.1 23 23 2 10
4
Polystyrene 2.55 5 10
5
PVC 36 35 3.5 3.0 10
4
Polycarbonate 2.8 3 10
2
Polyester 45 0.02
Nylon 66 3.5 3.33 3.16 0.02
Glasses
Pyrex 46 0.0080.025
Quartz 4 2 10
4
Vycor 3.8 9 10
4
Miscellaneous
Mica 5 3 10
4
Neoprene 6.3
Source: Ref. 6.
Table 7. Typical Relaxation Times of Three Liquids
Material Temperature (1C) t (10
11
s)
H
2
O 19 1
CH
3
OH 19 6
C
2
H
5
OH 20 13
Source: Ref. 9.
exhibit electronic polarization. The presence of electronic
polarizability may be veried with the Maxwell relation,
e
1
n
2
, where n is the refractive index of the dielectric.
Table 8 compares the e
1
and n
2
values for a few marginally
nonpolar materials [13,14].
4. MODELS OF DIELECTRIC RELAXATION
4.1. Models
The rst model of the dielectric relation is due to Debye
[3]. According to this model, the susceptibility function
w(o) for noninteracting polar molecules is given by [7]
wo /
1
1io=o
p
58
where o
p
is the angular frequency at which the maximum
loss peak occurs. The real and imaginary parts of w(o) are
w
0 0
o /
1
1o
2
t
2
59
and
w
0 0
o /
ot
1o
2
t
2
60
The corresponding time-domain response f(t) follows
the exponential function (15)
ft / e
t=t
61
The loss peak occurs here at oo
P
1/t. Figure 4
shows the dependence of w
0
(o), w
00
(o), and f(t) of Eqs.
(59)(61) [16] in loglog representation. The loss peak is
symmetric about o
p
, and its width at half-height is 1.144
decades on the frequency scale. The Debye behavior has
been observed in gases and in some polar liquids. The re-
laxation behavior of water and deuterium oxide closely
approximates that of the Debye form [1719]. However, it
is generally nonexistent in solids.
To account for the departure of the observed dielectric
behavior, the following empirical expressions have been
proposed. The ColeCole equation (20) is
wo /
1
1 io=o
p
1a
62
where a is a tting parameter in the range 0oar1.
Equation (62) provides a broader and symmetric
relaxation spectrum than the Debye type. Furthermore,
for o4o
0
, w
0
(o) and w
00
(o) show parallelism in the loglog
plot.
Table 8. A Comparison of e
0
and n
2
Values for Several
Nonpolar Materials
Material n
2
e
0
Frequency of
measurement
of e
0
(Hz)
Hydrogen (liquid, 2531C) 1.232 1.228
Diamond 5.66 5.68
Nitrogen (liquid, 1971C) 1.453 1.454
Oxygen (liquid, 1901C) 1.491 1.507
Chlorine (liquid) 1.918 1.910
Bromine 2.66 3.09
Parafn (liquid) 2.19 2.20 10
3
Benzene 2.25 2.284 10
3
Polystryrene 2.53 2.55 10
2
to 10
10
Polyethylene 2.28 2.30 10
2
to 10
10
Carbon tetrachloride 2.13 2.238
PTFE 1.89 2.10 10
2
to 10
10
Source: Refs. 13,14.
10
0
10
0
10
0
10
1
10
2
10
3
10
0
10
1
10
2
10
3
10
4
10
1
10
1
10
0
10
1
10
2
10
1
10
2
e
t
/
t
o
o
2
o
1
(o)
(o)
o/o
p
t /t
(a)
(b)
Figure 4. (a) The ideal Debye response in the frequency domain,
with its characteristic frequency dependence of w
0
(o) po
2
and
w
00
(o) po
1
above the loss peak. (b) The corresponding time-
domain response, which is purely exponential, is plotted here in
the somewhat unfamiliar loglog representation [16].
The DavidsonCole equation has the form [21]
wo /
1
1io=o
p
b
63
where b is yet another curve-tting parameter in the
range 0obr1. Equation (63) provides asymmetric relax-
ation proles at oro
0
, whereas w
0
(o) and w
00
(o) remain
parallel at o4o
0
.
The FuossKirkwood model [22] for the imaginary part
of the susceptibility is
w
0 0
o /
2o=o
p
g
1o=o
p
2g
64
Another relaxation model is given by
wo /
1
s 1
GDs
s 1!
expiDp=2
o
D
o
D
p
_ _
s
65
This is an expansion into the frequency domain of
the KolrauchWilliamsWatts function [15] of time:
exp[ (t/t)
D
]. The parameter D in Eq. (65) has no physical
significance and is not based on the physics of dielectric
interactions.
So far the models have had only one tting parameter,
namely, a for the ColeCole equation, b for the Davidson
Cole equation, g for the FuossKirkwood equation, and D
for the KolrauchWilliamsWatt equation. The model due
to Havriliak and Negami [23,24], the rst one with two
parameters, is given by
wo /
1
1 io=o
p
1a
b
66
It should be stressed again that the tting parameters a
and b in this equation have no physical significance.
A classical form of presentation of the dielectric data is
to plot w
0
(o) or e
0
(o) against w
00
(o) or e
00
(o), namely, the so-
called ColeCole plot [20]. Figure 5 shows the shapes of
the Debye, ColeCole, and DavidsonCole equations for
the susceptibility functions in ColeCole plots. It has been
shown [20] that with the Debye model, a graph of w
0
(o)
against g
0
(o) over the entire frequency range will be a
semicircle and w(N) or e
N
is obtained from the intercept at
the horizontal axis (see Fig. 5a). Thus the relaxation time t
may be obtained from the slope of a straight line from the
Log (o)
Log (o)
Log (o)
Log (o)
Log o
Log ( )o
o
o
[1
(1[) /2
(0)
(0)
0.4
0.2
3 2.1 2.0
1.0
0.3
0.1 0.01
0
10
3
10
1
10
30
(
ix
)
1
[
0.2 0.4 0.6 0.8 1.0
c
(0)/2 =
m
(0)/4
o
o
0 (0)
0
0 c
c(0)
c(0)
o
2
o
1
o
1:
o
:1
o
p
o
p
o
p
= 1/t
o
p
= 1/t
1/t
z
0
z>z
0
z>>z
0
:/2
c
c
c c
/2
(a) (b)
(c)
Figure 5. The frequency dependence of the
real and imaginary parts of the susceptibility
and the ColeCole presentation for (a) Debye,
(b) ColeCole, and (c) DavidsonCole systems [8].
origin to a point on the semicircle for which o is known.
Now the ColeCole relaxation model provides a symmetric
but broader relaxation spectrum, and the corresponding
ColeCole plot is still a semicircle. However, its center is
depressed below the w
0
or e
0
horizontal axis (see Fig. 5b)
with the angle ap/2 between the radius of the circle and w
0
or e
0
axis. There is no molecular interpretation of this fac-
tor a, and it has been interpreted as a spreading factor of
the actual relaxation time about a certain mean value.
The magnitude of a must lie between zero and unity. The
ColeCole plot for the DavidsonCole model is a skewed
plot (see Fig. 5c), representing a severe departure from the
Debye relaxation behavior.
The HavriliakNegami function [23,24] with two pa-
rameters, a and b [Eq. (66)], appears to provide the best
results for the tting of the measured dielectric data for
most materials. However, none of these mathematical
models that invoke a distribution of relaxation energies
(25) or times offer any physical interpretation of material
properties [2629] .
It has been suggested that a dielectric loss spectrum
may be regarded as a mathematical summation of a dis-
tribution function g(t) of Debye responses corresponding
to a distribution of relaxation times [30]; thus
wo
_
1
0
gt
1 ior
dt 67
The distribution functions are always positive, and
curves of w
00
(o) or e
00
(o) can be formed from them by the
superposition of many single relaxation-time or frequency
curves [31]. It has been shown [32] that the product of the
elapsed time and the depolarization current is a convolu-
tion of the distribution function of relaxation frequencies
with a weight function of an asymmetric bell shape. A
similar relationship is also shown to exist for the imagi-
nary part of the permittivity. The same work [32] also
proposes a deconvolution procedure to determine the dis-
tribution function of relaxation frequencies from experi-
mental data. A distribution of relaxation times from the
frequency dependence of the real part of the complex per-
mittivity has also been made with the inverse Fourier
transformation [33]. As stated earlier, however, no distri-
bution of relaxation times that can claim physical reality
can be associated with relaxation systems in condensed
matter [8,26,27].
A two-parameter model for the complex susceptibility
function w(o), known as the universal relaxation law, has
been proposed [8,16], which states that all solid dielectrics
follow fractional power laws in frequency. It is of interest
to note that w(o) may be expressed by a simple empirical
expression [16,23]
wo A1ix
m
n1=m
68
where the exponents m and n lie between zero and unity
and x is the normalized frequency. Equation (68) indicates
that the experimental state of dielectric susceptibility can
be tted with two power-law exponents. The Debye func-
tion is a limiting form of this equation for m1 and n0.
For the symmetric loss peak at o
p
and x 1, we have
m1n. Furthermore, the ratio w
00
(o)/w
0
(o) decreases as
m and 1 n become smaller, thus providing broader peaks
as in the case of the ColeCole function. This leads to the
universal law, characterized by two fractional power laws
in frequency respectively below and above the loss peak
frequency o
P
[8,16]
w
0 0
o tan
mp
2
_ _
w0 w
0
o / o
m
for o{o
p
69
for o{o
p
and
w
0 0
o cot
np
2
_ _
w
0
o / o
n1
for oco
p
70
where the exponents are in the ranges 0ono1 and
0omo1.
As a result, in the high-frequency range of the loss
peak, the ratio of the imaginary to the real part of the
complex susceptibility is a frequency-independent con-
stant:
w
0 0
o
w
0
o
cot
np
2
_ _
71
Hence, in a loglog plot w
00
(o) and w
0
(o) appear as par-
allel lines for oco
p
. It should be noted that for the Debye
process this ratio is ot and thus increases linearly with
frequency, which is consistent with the idea that the pro-
cess is a viscous phenomenon in which the dielectric loss
is linearly related to the angular velocity [16].
For the low-frequency part of the loss peak (ooo
p
), we
have [8,15]
w
0 0
o
Dw
0
o
tan
mp
2
_ _
72
where Dw
0
(o) w(0) w
0
(o) is known as the dielectric de-
crement and is the extent to which the polarization at any
particular frequency falls short of the value of the equi-
librium polarization in a static eld. Equations (69) and
(70) may be represented by the empirical law combining
the two fractional power laws above and below o
p
[8]:
w
0 0
o /
1
o=o
p
m
o=o
p
1n
73
The Fourier transforms of fractional power laws corre-
late the frequency-domain dielectric parameters with
their time-domain behavior thus (16):
o
m
/ t
m1
for tct 74
o
n1
/ t
n
for t{t 75
In the carrier-dominated low-frequency dispersion
(LFD) or quasi-DC (QDC) systems, mobile charge carri-
ers, such as ions and electrons, act as polarizing species
and provide a broad dielectric response [16,3436]
in which no loss peak is observed. The LFD (or QDC)
relaxation is characterized by two independent processes,
belowand above a certain critical frequency o
c
, which may
be represented by Eq. (70). The real and imaginary parts
of the complex dielectric susceptibility steadily increase
with decreasing frequency for small values of n
2
, at fre-
quencies less than o
c
. This is followed by a at loss be-
havior above o
c
with n
1
E1 [8,16,28]. The frequency o
c
plays a role analogous to that for o
p
in a dipolar system.
Figure 6 shows [16] typical behavior of w(o) for the LFD
(or QDC) system. Figure 7 shows schematically the typical
time-domain behavior of a dipolar LFD (or QDC) system
together with the at loss response corresponding to n-1
[15]. Note that the at loss behavior is the limiting case of
the dielectric response that occurs in low-loss materials
with a very small value of the ratio w
00
(o)/w
0
(o). The value
of n
2
can never be zero, and hence n cannot actually have a
value of 1, although nearly at loss behavior has been ob-
served experimentally.
There are few examples of solids, including single crys-
tals of ferroelectrics, that show pure Debye relaxation be-
havior. Avariety of solids (viz., low-loss dielectrics, dipolar
materials, semiconductor p-n junctions, and biological ma-
terials) are known to show dielectric dispersions that may
be tted with the universal fractional power law [Eq. (73)].
Furthermore, dipolar systems exhibit loss peaks, whereas
the carrier-dominated systems exhibit LFD (or QDC)
behavior [8,34,35].
A stochastic model for the universal dielectric disper-
sion has also been proposed [3739]. This probabilistic
model is based on the assumption that individual dipoles
and their environments interact during the process of
relaxation and the dielectric response function is given
by [37]
ft f
0
ao
p
o
p
t
a1
1 ko
p
t
a
76
where f
0
is a constant of the relaxation function f(t), and
k is a positive real number. In the short-time limit this
function is
ft % o
p
t
a1
o
p
t
n
77
where n1 a and 0ono1. The corresponding long-time
limit is
ft % o
p
t
a k=k
o
p
t
m1
78
where ma/k and 0omo1 if aok. The exponents m
and n of the universal fractional law [8] are thus related
by [37]
m
1 n
k
79
where k41n and 0ono1. If 1nokr1, then
1 nrmo1, and this is observed in most analyzed exper-
imental results. For k1 we have m1n, and this is
the ColeCole response. For k 1n we have m1,
which is the DavidsonCole response. If k41, then
0omo1, which is observed only in a small number of
analyzed data [21,33,38]. In this case, k-0 and the
WilliamsWatts response is observed [15,40]. The
probabilistic model [3739] thus suggests a relation
between the empirical parameters m and n, dening the
low- and high-frequency regions of the complex dielectric
susceptibility. It has been suggested [37] that the param-
eter (k) may be related to the waiting-time distribution of
the relaxing dipoles, which may follow a Weibull distribu-
tion, namely
Rs expks
d
80
where R(s) is the waiting-time distribution, k is a positive
real number, and 0odo1. It has been shown [37] that for
a particular waiting-time distribution function, the
10
5
10
1
10
4
10
3
10
2
10
1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
(o)
(o)
o (s
1
)
10
1
10
2
10
3
10
4
10
5
Figure 6. The frequency dependence of a system dominated by
LFD or QDC with n
1
0.8 at high frequencies and n
2
0.5 at low
frequencies. The crossover point is deliberately shifted to high
frequencies to show the LFD or QDC region [16].
10
3
10
2
10
1
10
0
10
1
10
2
10
1
10
0
10
1
10
2
10
3
10
4
10
5
10
3
x
x
1.6
x
0.2
x
0.8
(o)
(o)
(Dipolar)
LFD
charge carriers
f
(
x
)
e
(:)
Debye
x
1
(flat loss)
Figure 7. The time-domain response f(t/t) of typical dielectric
systems, including the Debye exponential response; the dipolar
response with n0.8, m0.6; the carrier-dominated LFD re-
sponse with n
1
0.8, n
2
0.2; and the at loss with n1 [16].
solution for f(t) can be obtained in a simple analytical
form:
ft ft; 1 n; k; 0o1 no1 and k > 0 81
The relatively recent model [41,42] based on a cluster
theory is perhaps the most sophisticated approach to the
explanation of relaxation phenomena observed in imper-
fect materials. The theory has been derived in the frame-
work of quantum mechanics and takes into account the
manybody interactions present in condensed matter.
The dipoles in the condensed phase may be regarded as
connected with other dipoles through their morphological
structure, and it is unlikely that they can act indepen-
dently as in the Debye model. Both solids and liquids are
composed of spatially limited regions possessing partially
regular structural order, and such regions may be called
clusters [41]. In any material many clusters may exist, and
in the presence of coupling between them an array may
form displaying partial long-range order. Absence of cou-
pling in the limit may lead to a cluster gas. In contrast,
systems with strong coupling between these arrays will
produce an almost perfect crystal. The model also consid-
ers two kinds of interactionsintracluster and interclus-
ter exchangesand each of these makes its own
contributions to the nal behavior of the complex suscep-
tibility function.
A dipole in the intracluster motion may rst relax ex-
ponentially (d
t/t
) as suggested in the Debye model. In
doing so, it will affect the eld experienced by other neigh-
boring dipoles in the cluster. These neighboring dipoles, in
turn, may also relax exponentially, thereby affecting the
eld experienced by the rst dipole, and so on. As a result,
the overall effect will be a process with an exponential
single dipole relaxation of the form e
t/t
and concomitant
t
n
behavior for the relaxation of the cluster dipole mo-
ment. The intercluster exchange will have a range larger
than that for the intracluster motion, and its origin will be
in dipoles near the edge of the cluster interconnecting to a
neighboring cluster [29,36,40]. It has been shown [41] that
with the intracluster motion and with the progressive in-
volvement of an increasing number of elements with the
progress of time, a fractional power law (i.e., o
n1
behav-
ior) for the susceptibility function may be obtained. Fur-
thermore, the parameter n (0ono1) is related to the
average cluster structure. Highly ordered structure has
n values approaching unity, thus indicating an existence
of completely correlated clusters. On the other hand, n-0
would signify a large degree of disorder, and the limit n0
would yield Debye-like relaxation behavior.
The intracluster-coupled mode may change to an inter-
cluster mode as the spatial extent of the coupling (wave-
length) increases beyond the cluster size. The
mathematical derivation of the susceptibility function for
the intercluster exchanges is similar to that of the intra-
cluster motion, as the intercluster exchanges are now the
perturbation of an ideal state. The result is also a frac-
tional power law [41], giving an o
m
behavior for the sus-
ceptibility function. Once again the value of m is in the
range 0omo1, and m represents the degree of structural
order, this time on the larger scale of the cluster, namely,
the degree of ordering in the cluster array. Hence, m-0
indicates an almost ideal lattice structure, whereas m-1
may give rise to a wide distribution of clusters. The intra-
cluster motion and the intercluster exchange mechanisms
are schematically represented in Fig. 8 [28].
For the intracluster motion the susceptibility function
is given by [41]
wo / 1i
o
o
p
_ _
1n
2
F
1
1 n; 1 m
o
p
o
p
io
_ _
82
where
2
F
1
is the Gaussian hypergeometric function. It
should be noted that the asymptotic limits of Eq. (82) are
the universal relaxation law [i.e., Eqs. (69) and (70)] [8].
The DissadoHill quantum-mechanical model [41] de-
scribes a QDC phenomenon as a partial conduction pro-
cess that is equivalent to the LFD phenomenon [8]
described above. In the QDC process considerations sim-
ilar to those for the dipoles are given to systems containing
charge carriers. The difference between a QDC process
and DC conduction at low frequencies is that the latter
phenomenon is characterized by
wo !constant 83
and
w
0 0
o /
s
dc
o
84
where s
dc
is the frequency-independent DC conductivity.
For the high frequencies, the MaxwellWagner interfacial
polarization effect [1] may be used to predict a limiting
E
(a)
(b)
Dipoles
Clusters
E
Figure 8. Schematic diagram of (a) intracluster motion and
(b) intercluster exchange mechanism of DissadoHill model of
dielectric relaxation [29,36].
behavior of the form
w
0
o / o
2
85
w
0 0
o / o
1
86
and
w
0 0
o
w
0
o / o
87
The DissadoHill model [41] suggests that the motion of
all charge carriers within a cluster of correlation length is
cooperative, that is, that the motion of a charge carrier to a
neighboring site is limited to the vacancy of such sites and
by other charges surrounding it. The model [41] divides
the response into high-frequency (short-time) behavior
above a critical frequency o
c
, where intracluster motion
occurs, and low-frequency (long-time) behavior below o
c
,
where intracluster motion exchange occurs. The intraclus-
ter motion, which is analogous to the ipping of dipoles, is
now replaced by the hopping of charges between available
sites within a correlation length x, which reduces the
overall polarization of the cluster. The high-frequency re-
sponse has the same functional form as for the dipoles:
o
n1
, 0ono1. Again the physical meaning of the expo-
nent n is the average degree of structural ordering within
a cluster, and small values of n will correspond to irreg-
ularities in a cluster such as might occur when an inter-
stitial ion or a dislocation is present. The parameter n may
also be related to the entropy density per cluster constit-
uent. The value of n may be independent of temperature
for thermally stable cluster structuring [41].
In the intercluster exchange there is a physical trans-
port of charges between the clusters. The charge motion is
no longer correlated with the available sites of the donor
cluster, but rather with those of the acceptor cluster. For
this case the susceptibility function is shown to be a frac-
tional power law of the form o
p
, with 0opo1 [41]. A
small value of p indicates a set of clusters that are almost
identical to each other, while a large value of p is associ-
ated with a broader distribution of clusters in which in-
tercluster exchanges can carry the effective charge
through many clusters over a long distance. In the pres-
ence of both the intracluster hopping and intercluster
charge transport, the susceptibility function of the system
is given by [42]
wo /
o
c
o
c
io
_ _
1n
2
F
1
1 n; 1 p; 2 n;
o
c
o
c
io
_ _
88
The asymptotic forms of w(o) at high and low frequen-
cies with respect to o
c
are [42]
w
0
o / w
0 0
o % w
0
o
c
o
o
c
_ _
p
for o{o
c
89
w
0
o / w
0 0
o / w
0
o
c
o
o
c
_ _
n1
for o > o
c
90
Once again, it may be noted that the asymptotic values
of this model [41,42] are the same as those of the universal
law model [8,15]. The relations between the exponents n
and p of these two models are
p1 n
2
91
nn
1
92
where n
1
and n
2
refer to the values of the parameters of
the universal law above and below o
c
, respectively [41]:
w
0
o / w
0 0
o / o
n
2
1
for o{o
c
93
w
0
o / w
0 0
o / o
n
1
1
for o > o
c
94
Summarizing the above, it appears that all dielectric
materials commonly investigated have the following char-
acteristics in terms of the indices n and m (41):
n0, m1 express the Debye limit of an ideal liquid
with independent cluster constituents in the system.
n1, m0 occurs in an ideal crystal with no internal
relaxation and zero loss.
For real liquids n-0, m-1, and the average clusters
are weakly bound.
For plastic crystals, waxes, and viscous liquids, n %
1
2
and m %
1
2
. These materials have clusters with re-
stricted structural range.
For solids with interstitial impurities and ferroelec-
trics, n-0 and m-1. Ferroelectrics have weakly
bound clusters of dipole reversals, thus yielding a
small value of n.
For imperfectly crystallized materials with topograph-
ical impurities, glasses, and vitreous polymer sys-
tems, n-1 and m-0.
It may be noted that nm1 will occur only when the
intra- and intercluster displacements lie along the same
coordinates, as in LennardJones liquids [43] and hydro-
gen-bonded systems [44].
The cluster model [41,42] is in many ways the most
rigorous description of relaxation of defects in a dielectric
system, and it offers an ab initio derivation of the entire
spectral shape of the frequency dependence of the suscep-
tibility function.
Table 9 lists the theoretical concepts of dielectric relax-
ation models, discussed above.
4.2. Electric Equivalent Circuits for Dielectric Loss
A dielectric capacitor can be represented by an electrical
circuit where the dielectric loss is reproduced mainly by
an equivalent resistance R in series or in parallel with the
capacitor and, occasionally, an inductance. A Debye sys-
tem can be represented, for example, by a resistance and a
capacitance in series, while non-Debye behavior of dielec-
tric susceptibility may be constructed with more complex
circuits.
For such cases, the concept of a universal capacitor [8]
has been proposed, and the resulting frequency depen-
dence of the dielectric parameters is
wo / C
n
o bio
n1
b sin
np
2
i cos
np
2
_ _
o
n1
95
It should be noted that for nonideal dielectric responses,
the circuit elements will have frequency-dependent
dispersive properties. Figure 9 shows schematic represen-
tations of simple circuit combinations of ideal, frequency-
independent elements and some forms of presentation of
dielectric data. The frequency response of lossy capacitors
of the type represented by Eq. (89) is shown in Fig. 10
[8,36,45]. The association of universal capacitors and dis-
persive circuit elements is schematically represented in
Table 10 [29].
4.3. Relaxation Behavior in Materials
The relaxation phenomena have been studied for a wide
range of materials, from covalent, ionic, and van der Waals
crystals at one extreme through glasses, liquids contain-
ing suspensions, solid synthetic polymers, and p-n junc-
tions at the other [41].
The permittivity of nonpolar gases at normal pressure
is close to unity, and the ClausiusMossotti equation [23]
adequately describes its variation with moderate density
changes [46]. At high pressures the molar polarization of
gases deviates from the ClausiusMossotti equation. The
molecular polarizability is enhanced by the attractive forc-
es between the molecules, whereas the repulsive forces
decrease it.
Centrosymmetric molecules do not possess dipole
or octupole moments, but quadrupole moments may
be present in some gases, such as hydrogen, carbon
dioxide, carbon disulde, oxygen, nitrogen, benzene, and
ethylene. Tetrahedral molecules, on the other hand,
have zero dipole and quadrupole moments (e.g., methane
and carbon tetrachloride) [46]. The presence of higher
dipole moments in a molecule induces moments on its
neighbors and produces deviations from the Clausius
Mossotti equation. Polar gases display temperature de-
pendence of the orientational polarization, and their
dielectric loss spectra follow the Debye relaxation behav-
ior in which partial orientation of the permanent dipoles
occurs under an externally applied eld. Polar gases
absorb energy in the microwave region through two pro-
cesses: rotational absorption and unquantized molecular
collision. The high-frequency dielectric properties of gases
have been well reviewed [47,48] and will not be discussed
further here.
There is as yet no exact theory of liquids, which have
been treated either as dense gases or as disordered solids.
Table 9. Theoretical Concepts of Relaxation Models
Function Equation Parameters
Debye
w
0 0
o /
1
o=o
p
1
o=o
p
(26) a 0 b 1
ColeCole
w
0 0
o /
1
o=o
p
a1
o=o
p
1a
(27) 0oao1 b 1
DavidsonCole
w
0 0
o /
1
o=o
p
1
o=o
p
b
(28) a 0 0obo1
HavriliakNegami
w
0 0
o /
1
o=o
p
a1
o=o
p
b1a
(29) 0oao1, 0obo1
Jonscher, Dissado, and Hill (dipolar peak)
w
0 0
o /
1
o=o
p
m
o=o
p
n1
(21) 0omo1 0ono1
Jonscher, Dissado, and Hill (QDC process)
w
0
o / w
0 0
o / o
n21
for o{o
c
(24) 0on
2
o1 0on
1
o1
w
0
o / w
0 0
o / o
n21
for oco
c
(25)
Weron (stochastic model)
ft f
0
ao
p
o
p
t
a1
1 ko
p
t
a
1 k=k
(41) m(1n)/k 1 nomo1
The dielectric relaxation in polar liquids (dilute solutions)
with spherical dipolar molecules can be interpreted in
terms of the orientation of individual dipoles. In the Debye
process, it is assumed that a spherical dipolar molecule
obeys Stokes law, which states that the relaxation time is
proportional to the shear viscosity of the liquid and to r
3
,
where r is the radius of the sphere. However, the relax-
ation time must depend on the viscosities of both the sol-
vent and the solute. The molecular radius calculated from
the relaxation time with the Debye model is usually too
small. Improved t to the relaxation behavior of liquids
may be obtained with empirical formulas (ColeCole [20],
DavidsonCole [21], HavriliakNegami [23,24]) and the
universal law [8].
Circuit Z Y C (o) comments
Resonance
Debye
L R
R 1/R
1/R
1/R
R
R
R
R
C
C
C
C
C
C
C
C
1
C
2
C
0
C
0
< C
s
C
c
1
c
1
c
c
c
2
c
2
c
C
s
C
G
G
G 1/G
1/G
1/G
G
G
0
G
1
G
1
G
2
G
b
"Leaky"
capacitor
Series
barrier
Diffusion
Effect of
nonzero o
0

/4
o
o
o
o
o
0
o
0
0
o
o
o
o
o
o
o
1
o
2
o
1
2
2
o
1
2
c
1
2
o
1
2
o
2
2
c
1
c
2
o
2
o
1
>>
C
C+C
C
s
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
Figure 9. Schematic representation of simple circuits
formed as a combination of ideal, frequency-indepen-
dent elements (ah) and some forms of presentation of
dielectric data (i, j) [8,36,45].
The intermolecular forces in associated liquids are
stronger and perhaps more directional in some cases than
in other liquids. Water is probably the most important as-
sociated liquid. The dielectric relaxation behavior of water
agrees well with the ColeCole model [20] with a
0.0270.007 [45,49]. It has been suggested [50] that the
kinetic process responsible for the dielectric relaxation in
water is cluster formation. Water is composed of uctuating
clusters of bonded molecules with unbonded molecules be-
tween them. Individual molecules are able to move fre-
quently from one cluster to another, and their dipole
orientation will depend on the number of hydrogen bonds
they form [45,50]. It should be noted that the clustering is a
random process and that it is not possible to subdivide wa-
ter molecules into groups that remain the same over a pe-
riod of time longer than the average relaxation time, E9.6
10
12
s, which is perhaps related to the OH stretching
vibration at 1.1010
13
Hz. This vibration is affected by the
hydrogen bonding. Alcohols have a wide distribution of re-
laxation times, which tend to follow the empirical Cole
Cole [20] and DavidsonCole [21] models. The dielectric
properties of liquids have been comprehensively reviewed
elsewhere [45,50] and will not be discussed further here.
A perfect alkali halide ionic crystal such as NaCl can be
polarized only by perturbing its thermal vibrations. How-
ever, in practice all crystals contain dislocations, speci-
cally, polarizable aws, which do not always distort the
lattice, particularly when the ionic radii are similar [45].
The dielectric relaxation behavior in such materials is
complicated by the presence of their ionic and electronic
conductivities. For these materials the relaxation time t
tends to be long (E1 s) at room temperature, and it obeys
t Ae
Ea=kT
96
where E
a
is the thermal activation energy and A is a con-
stant. It is of interest to note that the mechanical relax-
ation time of these materials is often half the dielectric
one, neglecting electrostatic interactions. This implies
that the shear modes of polarization relax twice as fast
as do the tensile ones [45]. The dielectric behavior of alkali
halides with divalent cations has been reviewed exten-
sively by Meakins [51].
Organic semicrystalline and amorphous polymers are
practical electrical insulating materials that consist of
macromolecules. Such molecular solids have both cova-
lent and van der Waals bonds, which facilitate molecular
motion in comparison with entirely covalently bonded sol-
id dielectrics. The activation process in these materials
also follows an Arrhenius relationship of the form
of Eq. (90) except at the glass transition temperature
T
g
. The relaxation process at T
g
is approximated by the
WilliamLandelFerry relationship [52],
tT t
0
exp
C
1
T T
g
C
2
T T
g
_ _
97
log o
1/o
o
n1
1/o
n+1
log o
log o
o
1
o
2
o
3
C
v
G
v
C
0
C
n
C
n
C
C
C
C
log C
log C
log C
R
R
1/R
o
B
10
5
0
0
0
20 15 10 5
4 4
5
S
l
o
p
e
1
S
l
o
p
e

n
b
S
lo
p
e

1
+
n
b
Slope 1+n
y
n = 0.3
0.5
0.7
0.8
0.9
0.99
1
1
1
1
1
1
(b) (a)
(c)
Figure 10. The frequency response of circuits involving universal
lossy capacitors of the type C
n
B (io)
n1
: (a) response of a series
combination of C
n
with a resistor R; (b) calculated frequency de-
pendences for a range of values of the exponent n; (c) the response
of the seriesparallel circuit shown, with two universal capacitors,
one of which corresponds to a series barrier region, while the other
forms the equivalent of a volume region with its parallel conduc-
tance G
v
. The values of parameters assumed in the calculation are
as follows: B
v
1, G
v
10
6
, n
v
0.85, B
s
1, n
s
0.4. At very
low frequencies the volume behavior is dominated by the conduc-
tance, and the response is that of a series combination of C
b
and
G
v
, which is therefore closely similar to that seen in (a) [8].
Table 10. Electrical Analog Equivalent Circuits
Ciruit Function
Debye
[Eq. (52)]
Dipolar relaxation
[Eq. (67)]
QDC process
[Eqs. (87), (88)]
Flat loss
[Eq. (87)]
Source: Ref. 29.
where t
0
is a constant, and C
1
and C
2
are also constants
with values E17 and E51, respectively [46]. The relax-
ation time t decreases with increasing temperature, as
may be observed in isochronal plots of depolarizing cur-
rent against temperature [53].
The relaxation behavior of polymers is related to sev-
eral complex physical parametersshear modulus, heat
capacity, permittivity, and refractive indexwhich exhibit
transitions with increasing temperature [54] (see Fig. 11).
In an amorphous polymer the principal transition is the
glass transition at a temperature T
g
, which is labeled as
the a transition at T
a
in Fig. 11. Above T
g
the free volume
decreases to a critical value, thus severely restricting the
segmental motions of the polymer chains. In a semicrys-
talline polymer there will be an additional transitional
phenomenon at the melting temperature T
m
. There are
other secondary transitions, b and g in order of decreasing
temperature (i.e., Ta4Tb4Tg). For example, in polyeth-
ylene, the a, b, and g relaxations at 1kHz occur at 77, 13,
and 1131C, respectively. The a relaxation is attributed to
motions in the crystalline phase, and the b relaxation
arises from primary motions of the chain branches in the
amorphous phase. The g relaxation may be associated with
a combination of processes including defect migration and
the reorientation motion in the amorphous phase [55]. The
a, b, and g relaxations in polypropylene occur at 80, 0, and
801C. Table 11 gives the glass transition temperatures
T
g
of some common polymers [54].
4.4. Experimental Evidence of Frequency Response and a
Comparison with the Cluster Models
Although ideal Debye response in ferroelectric single crys-
tals has been observed [56], there exist, in general, very
few examples of such responses in condensed matter. Al-
though water may be regarded as a classic dielectric, its
dielectric behavior displays a broadened relaxation peak
that departs from a true Debye relation [8,57]. Near-
Debye relaxation responses have also been observed in
silicon p-n junctions [8]. It may not be appropriate to
discuss experimentally observed dielectric dispersion data
with the ColeCole, DavidsonCole, and Havriliak
Negami models, which are basically empirical in nature.
However, relaxation spectroscopy can provide consider-
able information on dielectric materials from the mea-
surement of the shape of the loss peak as well as the
relaxation rate and amplitude. The shape of a loss peak is
clearly characterized by the parameters m and n of the
DissadoHill [41,42] and universal law [8] models. This
procedure has been employed to demonstrate the presence
of cluster structure in (1) the viscous liquid produced from
the glassy state above a glass transition [58], (2) plastic
crystal phases [59], and (3) ferroelectrics [60,61]. The clus-
ter size becomes strongly temperature-dependent in fer-
roelectrics near the Curie temperature [61]. The
amplitude and the relaxation rate are related [6062].
These considerations also hold true for liquid crystals
[60,61,63]. Figures 12a and 12b show the observed dielec-
tric response of poly-r-benzyl-L-glutamate (PBLG) and
poly-r-methyl-L-glutamate (PMLG), respectively [42].
The loss peaks in both cases are broad with values of n
and m in conformity with the cluster model [41]. Table 12
gives the values of shape indices n and m for PBLG in
different states, from which it may be observed that as the
local order decreases in solution, the value of n decreases,
and that of m increases [41,63]. These examples cover
some typical cluster structures with different values of m
and n [41].
It is suggested that the quantum-mechanical cluster
model provides explanations for the relaxation dynamics
in materials that show non-Debye susceptibility behavior
over a wide frequency range. The cluster model shows that
the free energy of a cluster is held constant and its entropy
T
Temperature
P
Q
P
h
y
s
i
c
a
l

q
u
a
n
t
i
t
y
[
[
:
:
Figure 11. Schematic diagram of the temperature dependence of
complex properties of polymers [54].
Table 11. Glass Transmission Temperature T
g
of
Common Polymers
a
Polymer T
g
(1C)
PE 90, 35
PP 10
Polymethylpentene 30
PS 95
PAN 105
PVC 85
PVF 20, 45
PVDC 15
PA 6 50
PA 66 90
PA 610 40
PMMA 105
POM 90, 10
Poly(phenylene oxide) 210
PC 150
PETP 65
CA 105
NR 75
CR 45
NBR 20
Source: Ref. 54.
a
These are approximate values; where two temperatures are given, the
assignment of the glass transition remains doubtful. T
m
is independent of
chain length for high-molar-mass polymers, but falls somewhat as the
chains become very short.
evolves at the expense of its internal energy (i.e., en-
thalpy), resulting in a power-law relaxation process.
5. APPLICATION OF DIELECTRIC SPECTROSCOPY IN
DETECTING AGING IN INSULATING POLYMERS
5.1. Dielectric Aging and Treeing
Polymers experience aging when subjected to a mechan-
ical or electric stress over an extended period of time. The
aging produces irreversible deterioration of physical,
chemical, and dielectric and other electrical properties,
which may eventually lead to electrical breakdown of an
insulating polymer. It must be stressed that physical and
chemical aging may occur independently without the ap-
plication of an external electric eld. However, the aging
process may be accelerated by the eld in conjunction with
other factors.
The mechanisms for electrical breakdown have been
extensively reviewed [64]. The chemical aging models
have also been reviewed [65] and will not be discussed in
detail here. The present section provides in brief the re-
sults of a study of aging of polyethylene under an AC eld
in humid environment by dielectric spectroscopy.
Dielectric aging in a dry environment at moderate to
high electric eld appears to begin mostly at imperfections
in materials where the local eld tends to be enhanced. At
such locations, treelike electrical channels may form and
propagate due to the occurrence of partial discharges.
Space charges play a significant role in the initiation
and growth of electrical trees [66]. Water trees in poly-
meric insulators with AC elds in humid environments
may arise from microphase separation in partially oxi-
dized polymers as a result of eld-induced electrochemical
processes [67]. It has been shown that water trees in cross-
linked polyethylene consist of tracks of hydrophilic
carboxylate salts in the amorphous phase of the polymer
[6870]. The dielectric aging and the water tree growth
incorporate electrochemical processes following the elect-
rophysical process of water and ion diffusion in the poly-
mer [68]. It has been suggested that the electrochemical
degradation of polyolens associated with aging and water
treeing involve ve fundamental steps: (1) electrolysis of
water, in which oxygen and hydrogen peroxide radicals
are formed, as both are oxidizing agents; (2) initiation of
degradation; (3) catalysis of degradation by metal ions; (4)
chain scission, resulting in the formation of ketones and
carboxylate ions; and (5) conversion of ketones to carboxy-
late ions [71]. Electric-eld-driven oxidation has also been
proposed by other workers [64,7274].
The electrooxidation occurs in the local eld direction,
and water tree tracks are formed by chain scission in the
amorphous regions of the polymer. The track region is hy-
drophilic. As a result, water molecules in the polymer ma-
trix condense to form liquid water in the track, which then
transports ions to provide further oxidation at the tip of
the track. Thus a track propagates itself in a similar man-
ner to that of a self-propagating electrical tree or a gas
breakdown channel, although at a different rate [71].
It may thus be expected that aging and its progress due
to the electrooxidation of a polymer in a humid environ-
ment may be detected by a study of its dielectric behavior
over a wide frequency range.
5.2. Evidence of Aging in Frequency Response
Figure 13 shows the frequency response of the real and
the imaginary parts [w
0
(o) and w
00
(o), respectively] of the
complex susceptibility w(o) of unaged and crosslinked
polyethylene (XLPE) cable samples and samples AC-
aged (6 kV/mm, 50 Hz) for up to 6000h in water at room
temperature [36,75]. It may be observed from the tted
response that there are three relaxation processes: (1) a
high-frequency (HF) loss peak at B510
5
Hz, (2) a
10
5
10
1
10
2
1
1
10
10
9
10
Frequency (Hz)
10
5
10
9
10
Frequency (Hz)
(
o
)
10
1
10
2
10
(
o
)
333 323 315 308 296 K
299 275 257
246 229 K
(a)
(b)
Figure 12. Master curves for the dielectric response of oriented
lms of (a) PBLG and (b) PMLG. The theoretical spectral shape in
the plots has been determined with the values (a) m0.28,
n0.87; (b) m0.24, n0.92. Plot (a) is scaled at 333K; (b), at
299K. In both, the small magnitude of the dispersion has limited
the accuracy with which the real part of the susceptibility could
be determined for the higher frequency values [42].
Table 12. Spectral Shape Indices Observed in Dielectric
Response of PBLG in Different Physical States
Sample m n
Solid, orientationally ordered lm 0.28 0.87
Solid, prepared by the Leuch method 0.42 0.81
Solution in benzene with e-caprolactam 0.78 0.50
Solution in trans-1,2-dichloroethylene
containing NN-formidemethylamide
1.0 0.50
Solution in puried ethylene dichloride 0.78 0.49
Solution in ethylene dichloride 0.61 0.49
Solution in dioxan 0.44
Solution in dioxan with DMF 0.76 0.51
Solution in dioxan 0.81 0.54
medium-frequency (MF) loss peak at B1Hz, and (3) a low-
frequency (LF) loss peak at B10
4
Hz. It is suggested that
the HF loss peak is due to bound water containing ions. It
has been stated that in principle there are two relaxations
in water: the uctuations in polarization and the dissoci-
ation of water into ions. The latter relaxation occurs in the
gigahertz range, whereas the former one may be observed
at B10
5
Hz. For example, for a solution of MgSO
4
in water
at 201C [76], the following chemical reactions of the elec-
trolyte may occur, each possessing its own relaxation char-
acteristic:
Mg
2
OH
MgOH
SO
2
4
H
HSO
4
The rst process is slower than the second, for which
the relaxation peak occurs at B210
5
Hz, which is in
agreement with the location of the HF peak in Fig. 13
[45,76]. The second chemical reaction is more rapid and is
outside the experimental range of Fig. 13. The HF peak
(Fig. 13) is observed to be fairly independent of the aging
time. It has also been shown that the diffusion coefcient
of water vapor in polyethylene is E1.410
6
m
2
/s and is
independent of electrical stress [77,78]. Furthermore, po-
lar impurities in polyethylene have been alleged to attract
water [70], which will be bound in the polymer. In view of
these observations, the origin of the observed HF peak
(Fig. 13) may be attributed to the ions in bound water, as
stated before [28,29,36].
XLPE cable samples contain crosslinking byproducts
(such as cumyl alcohol and acetophenon, as well as anti-
oxidants), which may diffuse out of the cable with the pro-
gress of time. In addition, antioxidants react chemically
with the oxidation products in the sample. The MF loss
peak at B1 Hz appears to increase slightly (Fig. 13) with
continued aging. It also becomes broader, overlapping
with the LF peak. It is suggested that the MF peak may
originate from the presence of the polar moieties discussed
above [28,29,36].
The LF loss peak (Fig. 13), occurring at 10
4
Hz,
changes significantly with aging. It may be noticed that
the magnitude of this peak at rst rises sharply, up to an
aging time of 1000h. Subsequently it decreases progres-
sively, although its magnitude is still greater after 6000 h
of aging than that of the unaged sample. Furthermore, the
LF loss peak becomes broader with increasing aging time.
The LF loss peak amplitude increases initially because
of the formation of free radicals. It may be argued that a
competitive process involving the production of polar moi-
eties due to electrochemical reactions and injected space
charges establishes itself with increasing aging time.
Eventually, the space charge component becomes domi-
nant as the polymer becomes more conductive. The relax-
ation loss behavior thus shows the presence of intracluster
interaction in the MFHF region and of intercluster
charge motion in the low-frequency region; the latter phe-
nomenon becomes dominant with continued aging [29,36].
Figure 14 [28,36,75] shows a possible electrical equiv-
alent circuit for the unaged and electrically aged XLPE
samples in a humid environment. The dielectric relax-
ation behavior of the unaged XLPE cable sample of Fig. 13
may be represented by a parallel-connected network of (1)
three series-connected frequency-dependent resistances
R
1
, R
2
, and R
3
and (2) three dissipative capacitances C
1
,
C
2
, and C
3
, giving three dipolar peaks in the LF, MF, and
HF regions, respectively. The QDC response in the LF re-
gion with progressive aging has been taken into account
by removing the resistance R
3
from the circuit. The ob-
served broadening of the MF peak with aging will cause
R
2
to diminish with aging, although it will still have a
nonzero value. The values of R
1
and C
1
for the HF peak
should not change significantly, as the HF peak remains
unaffected by eld aging. Figure 14 also incorporates the
very high-frequency capacitance C
N
and G
0
in parallel.
The latter parameter represents any DC conduction mech-
anism in the dielectric [75].
10
1
10
2
10
3
10
4
10
0
10
6
10
4
10
5
10
3
10
2
10
1
10
0
10
1
10
2
10
3
10
4
10
5
10
6
Frequency (Hz)
Unaged
1000 h
2000 h
4000 h
6000 h
Calculated data
Figure 13. Dielectric behavior of

XLPE cable samples, unaged and AC-
aged (6 10
6
Vm
1
, 50Hz, room tem-
perature, water) up to 6000h [36].
It is thus suggested that the dielectric spectroscopy,
particularly in the LF range, may be a convenient tool in
identifying aging [79]. Furthermore, the Debye relaxation
model [3] and the intracluster and intercluster manybody
interaction model [41,42] may provide explanations for the
relaxation behavior observed at a molecular level.
BIBLIOGRAPHY
1. C. J. F. Bottcher and P. Bordewijk, Theory of Electric Polar-
ization, 2nd ed., Vol. 1, Elsevier, Amsterdam, 1973, Chaps. 1,
2, pp. 990.
2. H. Frohlich, Theory of Dielectric Constant and Dielectric Loss,
Clarendon, Oxford; 1958, Chap. 1, pp. 114.
3. P. Debye, Polar Molecules, Dover, New York, 1929, Chap. 5.
4. A. A. Zaky and R. Hawley, Dielectric Solids, Routledge, Kegan
& Paul, London; 1970, Chaps. 1, 2, pp. 225.
5. A. R. Von Hippel, ed., Dielectric Materials and Applications,
MIT Press, Cambridge, MA; 1954, Chaps. 13, pp. 336.
6. M. C. Lovell, A. J. Avery, and M. E. Vernon, Physical Proper-
ties of Materials, Van Nostrand Reinhold, New York; 1976,
Chap. 8, pp. 153184.
7. A. Schonhals, Dielectric properties of amorphous polymers, in
J. P. Runt and J. J. Fitzgerald, eds., Dielectric Spectroscopy of
Polymer Materials, Washington: American Chemical Society
(ACS), Washington, DC, 1997, Chap. 3, pp. 81106.
8. A. K. Jonscher, Dielectric Relaxation in Solids, Chelsea
Dielectric Press, London; 1983, Chap. 2, pp. 1361.
9. C. A. West and R. M. Thomson, Physics of Solids, McGraw-
Hill, New York; 1970, Chaps. 18, 19, pp. 388421.
10. R. Sa nger, O. Steiger, and K. Ga chter, Temperatureffect der
Molekularpolarisation einiger Gase und Da mpfe, Helv. Phys.
Acta 5:200210 (1932).
11. C. Smyth and W. Walls, Dielectric investigation of nitrometh-
ane and chloropicrid, J. Chem. Phys. 3, 557559 (1935).
12. C. Smyth and C. Hitchcock, Dipole rotation and the transition
in the crystalline hydrogen halides, J. Am. Chem. Soc.
55:18301840 (1933).
13. J. C. Anderson, Dielectrics, Chapman & Hall, London; 1964,
Chap. 7, pp. 8397.
14. L. Hartshorn and J. A. Saxton, Handb. Phys., 26:640 (1958).
15. G. Williams and D. C. Watts, Non-symmetrical aspects of
multiple dielectric relaxation in behaviour arising from a sin-
gle empirical decay function, Trans. Faraday Soc. 66:8085
(1970).
16. A. K. Jonscher, Universal Relaxation Law, Chelsea Dielectric
Press, London; 1996, Chap. 1, pp. 144.
17. J. M. Alison, A Dielectric Study of Lossy Materials over the
Frequency Range of 4-82 GHz, Ph.D. thesis, Univ. London,
1990.
18. R. M. Hill and A. K. Jonscher, The dielectric behaviour of
condensed matter and many body interpretation, Contemp.
Phys. 24:75110 (1983).
19. J. M. Alison and R. J. Sheppard, A precision wave-guide sys-
tem for the measurement of complex permittivity of lossy liq-
uids and solid tissues in the frequency range 29 GHz-90 GHz,
1: The liquid system for 29-45 GHzan investigation in wa-
ter, Meas. Sci. Technol. 1:10931098 (1993).
20. K. S. Cole and R. H. Cole, Dispersion and absorption in di-
electrics, 1. Alternating current characteristics, J. Chem.
Phys. 9:341351 (1941).
21. D. W. Davidson and R. H. Cole, Dielectric relaxation in glyc-
erol, propylene, and n-propanol, J. Chem. Phys. 12:14841490
(1951).
22. R. M. Fuoss and J. G. Kirkwood, Electrical properties of sol-
ids. VIII. Dipole moments in polyvinyl chloride diphenyl sys-
tems, J. Am. Chem. Soc. 63:385394 (1941).
23. S. Havriliak and S. Negami, A complex plane analysis of
a-dispersion in some polymer systems, J. Polym. Sci. C 14:
99117 (1966).
24. S. Havriliak and S. Negami, A complex plane representation
of dielectric and mechanical relaxation processes in some
polymers, Polymer 8:161210 (1967).
25. J. R. Macdonald, Transient and temperature response of a
distributed, thermally activated system, J. Appl. Phys.
34:538552 (1963).
26. B. Gross, Electret researchstages in development, IEEE
Trans. Electric. Insul. EI-2I(3):249269 (1986).
27. B. Gross, Distribution functions in linear viscoelastic theory,
J. Appl. Phys. 62:27632770 (1987).
28. D. K. Das-Gupta and P. C. N. Scarpa, Polarization and
dielectric behaviour of ac-aged polyethylene, IEEE Trans.
Dielectrics Electric. Insul. 3:366374 (1996).
29. D. K. Das-Gupta and P. C. N. Scarpa, Modelling of dielectric
relaxation spectra of polymers in the condensed phase, IEEE
Electric. Insul. Mag. 15:2332 (1999).
30. G. C. Garton, The distribution of relaxation times in dielec-
trics, Trans. Faraday Soc. A 42:5560 (1946).
31. N. G. McCrum, B. E. Read, and G. Williams, Phenomenolog-
ical theories of mechanical and dielectric relaxations, in
Anelastic and Dielectric Effects in Polymer Solids, Wiley,
New York; 1967, Chap. 4, pp. 102237.
AC-aged Unaged
AC-aged Unaged
G
0
G
0
LF HF
MF
LF HF
MF
R
1
R
2
R
3
C
3
C
2
C
1
R
1
R
2
C
2
C
2
C
1
C
Progressive
aging
Figure 14. Analog equivalent circuits of
unaged and AC-aged XLPE cable samples
[28,36,75].
32. C. J. Dias, Determination of a distribution relaxation fre-
quency, Phys. Rev. B 53:1421214222 (1996).
33. K. Liedermann, The calculation of a distribution of relaxation
times from the frequency dependence of the real permittivity
with the inverse Fourier transformation, J. Non-cryst. Solids
175:2130 (1994).
34. A. K. Jonscher, Dielectric response of polar materials, IEEE
Trans. Electric. Insul. 25:622629 (1990).
35. A. K. Jonscher, The universal dielectric response and its
physical significance, IEEE Trans. Electric. Insul., EI-19:
567577 (1992).
36. P. C. N. Scarpa, Polarization and Dielectric Behaviour of AC
Aged Polyethylene, Ph.D. thesis, Univ. Wales, 1995.
37. K. Weron, A probabilistic mechanism hidden behind the uni-
versal power law for dielectric relaxation, J. Phys. Condens.
Matt. 3:91519162 (1991).
38. K. Weron and A. Jurlewickz, Two forms of self-similarity as a
fundamental feature of the power-law dielectric response,
J. Phys. A Math. Gen. 26:395410 (1993).
39. A. Weron, K. Weron, and W. A. Wyoczynski, Relaxation func-
tions in dipolar materials, J. Stat. Phys. 78:10271038 (1995).
40. J. T. Bender and M. F. Shlesinger, Derivation of the
Kohlrauch-Williams/Watts decay law from activation energy
dispersion, Macromolecules 18:591592 (1985).
41. L. A. Dissado and R. M. Hill, A cluster approach to the struc-
ture of imperfect materials and their relaxation spectroscopy,
Proc. Roy. Soc. Lond. 390:131180 (1983).
42. L. A. Dissado and R. M. Hill, Anomalous low frequency dis-
persion, J. Chem. Soc. Faraday Trans. 2 80:291319 (1984).
43. R. M. J. Cotterill and J. C. Tallon, Melting and the liquid
glassy state, J. Chem. Soc. Faraday Disc. 69:241260 (1980).
44. H. P. Schwan, R. J. Sheppard, and E. H. Grant, Complex per-
mittivity of water at 251C, J. Chem. Phys. 64:22572258
(1976).
45. V. V. Daniel, Dielectric Relaxation, Academic Press, London;
1967, Chap. 7, pp. 95109.
46. N. E. Hill, W. E. Vaughan, and M. Davies, Dielectric Proper-
ties and Molecular Behaviour, Van Nostrand Reinhold, Lon-
don; 1969, Chaps. 35, pp. 191461.
47. K. H. Illinger, Dispersion and absorption of microwaves
in gases and liquids, in J. B. Birks and J. Hart, eds., Progress
in Dielectrics, Academic Press, London; 1962, Vol. 4, pp.
37101.
48. H. G. Sutter, Dielectric polarization in gases, in M. Davies,
ed., Dielectric and Related Molecular Processes, Chemical So-
ciety, London; 1972, Vol. 1, Chap. 3, pp. 6499.
49. E. H. Grant, T. J. Buchanan, and H. F. Cook, Dielectric be-
haviour of water at microwave frequencies, J. Chem. Phys.
26:156161 (1957).
50. J. B. Hasted, Dielectric properties of water and of aqueous
solutions, in M. Davies, ed., Dielectric and Related Molecular
Processes, Chemical Society, London; 1972, Vol. 1, Chap. 5,
pp. 121161.
51. R. J. Meakins, Mechanism of dielectric absorption in solids,
Progr. Dielectr. 3:151202 (1961).
52. M. L. Williams, R. F. Landel, and J. D. Ferry, The temperature
dependence of relaxation mechanisms in amorphous poly-
mers and other glass-forming liquids, J. Am. Chem. Soc.
77:37013707 (1955).
53. D. K. Das-Gupta and R. S. Brockley, A study of absorption
currents in polypropylene, J. Phys. D Appl. Phys. 11:955962
(1978).
54. C. Hall, Polymer Materials: An Introduction for Technologists
and Scientists, 2nd ed., Macmillan Education, London; 1989,
Chap. 2, pp. 3454.
55. R. T. Baily, A. M. North, and R. A. Pethrick, Molecular Mo-
tions in High Polymers, Clarendon, Oxford; 1981.
56. K. Deguchi, E. Okaane, and E. Nakamura, Effects of deute-
ration on the dielectric properties of ferroelectric CsH
2
PO
4
, 1.
Static dielectric properties, J. Phys. Soc. Japan 51:35693574
(1969).
57. P. R. Mason, J. B. Hasted, and L. More, The use of statistical
theory in tting equations to dielectric dispersion data, Adv.
Mol. Rel. Proc. 6:217232 (1974).
58. M. Shablalch, R. M. Hill, and L. A. Dissado, Dielectric exam-
ination of glass-forming system, J. Chem. Soc. Faraday Trans.
2 78:639655 (1982).
59. M. Shablakh, L. A. Dissado, and R. M. Hill, Structure and
dielectric relaxation mechanisms in the cyclic alcohols, cyclo-
pentanols for cyclo-octanol, J. Chem. Soc. Faraday Trans. 2
79:369417 (1983).
60. L. A. Dissado and R. M. Hill, Dielectric behaviour of materials
undergoing dipole alignment transitions, Phil. Mag. B 41:
625642 (1980).
61. L. A. Dissado and R. M. Hill, Dynamic scaling and the rst
order character of ferroelectric transitions, J. Phys. C 16:
40234039 (1983).
62. P. C. Hohenberg and B. I. Halperin, Theory of dynamic critical
phenomena, Rev. Mod. Phys. 49:435479 (1977).
63. L. A. Dissado and R. M. Hill, Examination of the dielectric
susceptibility of poly-r-benzyl-L-glutamate, J. Chem. Soc.
Faraday Trans. 2 78:8193 (1982).
64. L. A. Dissado and J. C. Fothergill, Electrical Degradation and
Breakdown in Polymers, Peter Peregrinus, London; 1992,
pp. 74116.
65. L. Reich and S. A. Stivala, Elements of Polymer Degradation,
McGraw-Hill, New York; 1971, pp. 1275.
66. T. Hibma and H. R. Zeller, Direct measurement of space
charge injection from a needle electrode into dielectrics,
J. Appl. Phys. 59:16141620 (1986).
67. H. R. Zeller, Thermodynamics of water treeing, IEEE Trans.
Electric. Insul. EI-22:677681 (1987).
68. R. Ross and J. J. Smit, Composition on growth of water trees
in XLPE, IEEE Trans. Electric. Insul. 27:519530 (1992).
69. E. F. Steenis, Water Treeing in Polymer Cable Insulations,
KEMA Scientific Technical Report, 1990, Vol. 8, pp.
149208.
70. E. F. Steenis and F. H. Kruger, Water treeing in polyethylene
cables, IEEE Trans. Electric. Insul. 5:9891028 (1990).
71. J. J. Xu and S. A. Boggs, The chemical nature of water treeing:
theories and evidence, IEEE Electric. Insul. Mag. 10(5):2937
(1994).
72. H. J. Henkel et al., Relationship between the chemical struc-
ture and the effectiveness of additives in inhibiting water-
trees, IEEE Trans. Electric. Insul. EI-22:157161 (1987).
73. A. Garton et al., Oxidation and water tree formation in ser-
vice-aged XLPE cable insulation, IEEE Trans. Electric. Insul.
22:405412 (1987).
74. R. J. Densley et al., Water treeing and polymer oxidation,
Conf. Record, Int. Symp. Electrical Insulation, IEEE Publica-
tion 90-CH2727-6, 1990, pp. 178182.
75. P. C. N. Scarpa, A. Svatik, and D. K. Das-Gupta, Dielectric
spectroscopy of polyethylene in the frequency range of 10
-5
Hz
to 10
6
Hz, Polym. Eng. Sci. 36:10721080 (1996).
76. M. Eigen, G. Kurtze, and K. Tamm, Zum Reaktions-
mechanismus der Ultraschallabsorption in Wa ssrigen Elect-
rolytlosungen, Electrochemistry 57:103118 (1957).
77. J. C. Chan and S. M. Jaczek, The moisture absorption of
XLPE cable insulation under simulated service condition,
IEEE Trans. Electric. Insul. EI-13:194197 (1978).
78. H. Li, The Association of Ions and Electrical Properties with
Water Treeing in Low Density Polyethylene, Ph.D. thesis,
Univ. Strathclyde, UK, 1993.
79. D. K. Das-Gupta, Conduction mechanisms and high eld ef-
fects in synthetic insulating polymers, IEEE Trans. Electric.
Insul. 4:149156 (1997).
DIELECTRIC RESONATOR ANTENNAS
H. K. NG
K. W. LEUNG
City University of Hong Kong
Kowloon, Hong Kong SAR
1. INTRODUCTION
1.1. Background
Traditionally, a dielectric resonator (DR) was used as an
oscillator or a lter [1], which was treated as a source of
energy storage rather than as an antenna. The DR was
rst proposed as an effective radiator in 1983 by Long
et al. [2]. They demonstrated the radiation capability of
the DR with different shapes [3,4]. Henceforward, DRs
were widely accepted as efcient antennas.
Researchers look for a compact, low-loss, and low-cost
antenna. A DR is simple in construction; it consists of
dielectric material and, therefore, has no metallic loss.
This is a prominent feature especially for operation in the
millimeter-wave region, where the radiation efciency of
conventional metallic antennas is usually limited by me-
tallic loss. In addition, the wavelength inside the DR is
smaller than that in free space by a factor of 1=

e
r
p
, where
e
r
is the dielectric constant of the DR. Therefore, by
increasing e
r
, it is possible to obtain a smaller antenna.
For many years, researchers have studied different DR
shapes, such as cylindrical [2,5], rectangular [3], hemi-
spherical [4], triangular [6,7], spherical cap [8], and
cylindrical ring [9,10]. Various shapes of the dielectric
resonator antenna (DRA) are shown in Fig. 1. In general,
in fundamental mode a DRA radiates like a magnetic
dipole, a functions that is independent of antenna shape.
Although DRAs are well suited for high-frequency appli-
cations, only a few of them have been examined through
limited theoretical work.
Early studies of the DRA concentrated on linear polar-
ization (LP) [1116]. Figure 2 shows the coaxial-probe-fed
DRA, which was widely used for excitation of LP DRAs.
However, it introduces ohmic loss and large probe self-
reactance at high frequencies. Furthermore, a hole must
be drilled in the superhard DR to accommodate the probe.
This inadvertently creates undesirable airgaps [17] be-
tween the probe and the DR, causing measured results to
deviate from the theoretical design. More recently, an LP
DRA has been investigated using a new excitation scheme
[18]: the conformal strip excitation (Fig. 3). The strip is
simply cut from an adhesive conducting tape and then
mounted on the DRA surface. Since the strip does not
penetrate inside the DRA, the strip length can be adjusted
very easily. This greatly facilitates postmanufacture trim-
mings of the antenna. Moreover, using this excitation
method, the undesirable airgap between the feeding probe
and the DRA [19] can be avoided. This article focuses on
the proposed conformal-strip-fed DRA. We will introduce
an angular displacement for the strip to have one more
degree of freedom for matching the impedance.
The circularly polarized (CP) DRA [2024] has been a
very attractive topic because, it allows a more exible
orientation between the transmitting and receiving an-
tennas than does the LP system. In addition, CP elds are
less sensitive to the propagation effect than are LP elds,
and the CP system is therefore widely used in satellite
communications. A simple and straightforward CP DRA
design utilizes a quadrature feed [5,25,26], but this sub-
stantially increases the size and complexity of the feed
network. Petosa et al. [27] shifted the complexity from the
feed network to the DRA, but the cross-shaped DRA may
(a) (b) (c)
(d) (e) (f)
Figure 1. Various DRA congurations: (a) cylindrical; (b) rec-
tangular; (c) hemispherical; (d) triangular; (e) spherical cap; (f)
cylindrical ring.
Coaxial
probe
Ground plane
Hemispherical DRA
Figure 2. Conguration of coaxial-probe-fed DRA.
960 DIELECTRIC RESONATOR ANTENNAS
not be available in the commercial market. To avoid these
problems, a parasitic patch has been applied to the con-
formal-strip-fed DRA [20,24]. The use of a parasitic patch
on the DRA was rst considered by Li et al. [28] and later
by Chen et al. [29]. In their work, the parasitic patch was
placed on top of a DRA, mainly to tune the operating
frequency instead of exciting CP elds. More recently, the
use of a parasitic patch for the excitation of CP elds has
been extended to aperture-coupled sources [30,31]. In this
article, we will also demonstrate a CP DRA with a para-
sitic patch.
1.2. Methodology
In the analyses, the mode-matching method is used to nd
the various Greens functions, from which the integral
equations for the conformal strip and parasitic patch
currents are formulated. The equations are then solved
using the method of moments (MoM).
To speed the numerical computation, the impedance
integrals are evaluated either using newly obtained re-
currence formulas or by direct analytical integration. It is
well known that evaluating self-impedance integrals nu-
merically is difcult because of the singularity problem of
Greens functions. Around the singular points, excessive
modal terms of the DRA Greens functions are required for
accurate calculation of the functions. However, the ampli-
tudes of high-order Hankel functions are so large that
they are difcult to handle numerically. In addition, the
integrands will uctuate very sharply around these points
and therefore very dense sampling points will be required
to evaluate the integrals accurately, leading to consider-
able programming effort and computation time. The
singularity problem was previously solved using
the recurrence approach [18]. Since low-order integrals
can be evaluated analytically, all the impedance integrals
can be calculated without the need for any numerical
integration. The same approach will be used in this
article. The recurrence formulas [18] can be used only
for a constant latitude current around the equatorial
plane. Although they can be generalized for pulse-mode
basis functions at any latitude angles, new recurrence
formulas for piecewise sinusoidal (PWS) basis functions
are discussed in this article instead to accelerate the MoM
convergence rate.
2. CONFORMAL-STRIP-EXCITED DRA
2.1. Introduction
The conformal strip described in Ref. 18 was placed along
a meridian plane of the hemispherical DRA. Virtually, the
strip length is the only strip parameter used for tuning the
impedance (the width of a slender strip has a relatively
small effect on the input impedance). We extend the
previous theory [18] to include an arbitrary angular strip
displacement y
0
. The strip begins at the base of the hemi-
sphere and then bends back, as shown in Fig. 4. By doing
so, we can desirably obtain one more degree of freedom in
designing the DRA. The effect of strip displacement on
input impedance was investigated, and it was found that
impedance matching can easily be achieved by varying
this parameter. The far elds of the new conguration
were also studied. Measurements were carried out to
verify the calculations, and reasonable agreement
between theory and experiment was found.
2.2. Theory
To simplify the formulation, the coordinate system shown
in Fig. 4 is used. With this coordinate system, the un-
known strip current can be expanded using sinusoidal
functions and, thus, the related MoM integrals can be
evaluated in closed form. A hemispherical DRA of radius a
and dielectric constant e
r
is excited by a conformal strip of
length l and width W. The conformal strip has angular
displacement y
0
from the z axis.
2.2.1. Greens Functions. Nevels and his collaborators
[33,34] have presented the modal Greens function as the
sum of particular and homogeneous solutions. The poten-
tials associated with the source in unbounded regions
correspond to the particular part, whereas the boundary
discontinuities are taken into account by the homoge-
neous part. As both the excitation strip and the parasitic
element are on the DR surface in the present problem, it is
unnecessary to use the previous approach. In the follow-
ing formulation, the elds are assumed to vary harmoni-
cally as e
jot
. The eld and source points are denoted by
r
*
r; y; f and r
*
0
r
0
; y
0
; f
0
, respectively. The Green func-
tions (r r
0
a) of E
f
due to a
^
ff-directed point current is
Hemispherical DRA
Conformal
strip
Ground plane
Coaxial
aperture
Figure 3. Conguration of conformal-strip-fed DRA.
Ground
plane
Hemispherical
DRA
Conformal
strip
l
a
z
x
y
W
Coaxial
aperture
0
0
c
r
Figure 4. Conguration of the conformal-strip-excited hemi-
spherical DRA with an angular strip displacement y
0
.
DIELECTRIC RESONATOR ANTENNAS 961
found as follows
G
E
y
J
f
jZ
0
2par
1
n0
2n1
nn1
n
m1
m
n m!
nm!
sinmf f
0
.
1
D
TE
n
d
dy
0
P
m
n
cos y
0
P
m
n
cos y
sin y
F
n
_
1
D
TM
n
P
m
n
cos y
0
sin y
0
d
dy
P
m
n
cos yC
n
_
1a
G
E
f
J
f
jZ
0
2par
1
n0
2n1
nn1
n
m1
n m!
nm!
cos mf f
0
.
1
D
TE
n
D
m
d
dy
0
P
m
n
cos y
0
d
dy
P
m
n
cos yF
n
_
m
2
D
TM
n
P
m
n
cos y
0
sin y
0
P
m
n
cos y
sin y
C
n
_
1b
where
F
n
^
HH
2
n
k
0
a
^
JJ
n
kr roa
^
JJ
n
ka
^
HH
2
n
k
0
r r > a
_
1c
C
n
^
HH
0
2
n
k
0
a
^
JJ
0
n
kr roa
^
JJ
0
n
ka
^
HH
2
0
n
k
0
r r > a
_
1d
It should be mentioned that the functions F
n
and C
n
have different forms for the E eld inside (roa) and
outside (r4a) the DR. All other symbols were dened in
Refs. 35 and 36.
2.2.2. MoM. Enforcing the boundary condition that the
total tangential electric elds should vanish on the strip
surface, we have
E
s
E
i
0 2
where E
s
and E
i
are dened as the scattered and im-
pressed elds due to the current density J
fs
and the
excitation source, respectively. Using a delta gap source
model, the impressed eld can be written as V
0
=adf.
Then Eq. (2) can be rewritten as
1
W
__
S
0
G
E
f
J
f
J
fs
fdS
0
V
0
a
df 3
where S
0
is the strip surface. Let If J
fs
W be the strip
current, which is expanded as follows using the MoM
If
N
q 1
I
q
f
q
f 4
where I
q
are unknown expansion coefcients to be deter-
mined, and f
q
(f) are PWS basis functions given by
f
q
f
sink
e
h ajf f
q
j= sin k
e
h ajf f
q
joh
0 elsewhere
_
5
where h2l/(N1), f
q
( l qh)/a, and k
e
e
r
1=2
_
k
0
are the PWS-mode half-length, the center-
point of the qth expansion mode, and the effective wave-
number at the DRAair interface, respectively. The
expansion coefcients are found via the matrix equation
[Z
pq
][I
q
] [f
p
(0)], where Z
pq
are given by
Z
pq
ja
2
Z
0
4pW
2
1
n1
2n1
nn1
^
JJ
n
ka
^
HH
2
n
k
0
a
D
TE
n
_
n
m0
2
D
m
.
n m!
nm!
Y
1
n; m
2
Fp; q; m
^
JJ
0
n
ka
^
HH
2
n
0
k
0
a
D
TM
n
n
m1
2m
2
.
n m!
nm!
Y
2
n; m
2
Fp; q; m
_
6
Y
1
n; m
_
y
0
W=2a
y
0
W=2a
dP
m
n
cos y
dy
sin y dy 7
Y
2
n; m
_
y
0
W=2a
y
0
W=2a
P
m
n
cos y dy 8
Fp; q; m
2k
e
acos k
e
h cos mf
h
mk
e
amk
e
a sin k
e
h
_ _
2
cos mf
p
f
q
9
where
f
h
h
asiny
0
10
Since recurrence formulas exist for the integrals Y
1
and
Y
2
, the integrals can be calculated very easily and quickly.
Consequently, the solution (3) is a regular modal series
that can be implemented very straightforwardly. As the
solution can be calculated without need for any numerical
integration, it is computationally very efcient. After the
strip current is obtained, the input impedance can be
readily calculated from
Z
in
1
2

N
q1
I
q
f
q
0
11
where the factor of 2 accounts for the image effect of the
ground plane. From the strip current the radiation can be
found easily.
2.3. Results and Discussion
To verify the theory, a hemispherical DRA of a12.5 mm
and e
r
9.5 was measured using an HP8510C network
analyzer. A conformal strip of length l 12.0 mm and
width W1.2mm was cut from an adhesive conducting
tape. For the calculation part, two basis functions and 60
modal terms were used. Figure 5 shows the measured and
calculated input impedances for y
0
801, and reasonable
agreement between theory and experiment is observed.
The measured and calculated resonant frequencies are
3.52 and 3.58GHz, respectively, with error 1.7%. The
results are consistent with the source-free value of
3.68 GHz [36].
Figure 6 displays the calculated input impedance for
y
0
601, 701, and 801. It is observed that the impedance
increases with y
0
, showing that y
0
can be used as a new
parameter to match the impedance.
The measured and calculated far elds at 3.56 GHz for
l 12.0 mm, W1.2mm, and y
0
801 are shown in Fig. 7,
where reasonable agreement between theory and experi-
ment is observed. The elds are of a broadside mode, as
expected. It is found that the copolarized eld is stronger
than the cross-polarized eld by more than 16 dB in the
broadside direction (y
0
01), which is sufcient for many
practical applications. Note that the H-plane cross-polar-
ized eld is not symmetric about the center, due to the
angular displacement of the excitation strip.
3. CIRCULARLY POLARIZED DRA EXCITED BY A
CONFORMAL STRIP WITH A PARASITIC PATCH
3.1. Introduction
Thus far, the work on CP DRAs has been mainly experi-
mental. The rst theoretical work was carried out by
Esselle [21], who used the nite-difference time-domain
(FDTD) method to study the CP rectangular DRA. The
FDTD method can handle a large class of problems, but
since it is purely numerical, a very long computation time
is required. Moreover, no physical insights into the pro-
blem can be obtained through the numerical formulation.
In this section, a CP DRA excitation method that employs
a single parasitic patch is presented [24]. The conformal-
strip-excited hemispherical DRA [18] is used for the
demonstration. Nevertheless, the CP technique can be
used with other excitation methods and DRAs.
3.2. Theory
The conguration of the DRA is shown in Fig. 8, where a
hemispherical DRA of radius a and dielectric constant e
r
is
Frequency (GHz)
I
n
p
u
t

I
m
p
e
d
a
n
c
e

(
)
Resistance
Reactance
Theory
Experiment
2.6 2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4
40
20
0
20
40
60
80
100
120
140
160
Figure 5. Measured and calculated input impedances for the
conguration a12.5mm, e
r
9.5, l 12.0mm, W1.2mm, and
y
0
801. (From Leung and Ng [32], r 2001 John Wiley & Sons,
Inc.)
Resistance
Reactance 20
0
20
40
60
80
100
120
140
3.2 3.4 3.6 3.8
40
160
3 4
I
n
p
u
t

I
m
p
e
d
a
n
c
e

(
)
Frequency (GHz)
0
0
= 80
0
0
= 70
0
0
= 60
Figure 6. Calculated input impedances for y
0
601, 701, and 801,
with a12.5mm, e
r
9.5, l 12.0mm, W1.2mm. (From Leung
and Ng [32], r 2001 John Wiley & Sons, Inc.)
Theory
Experiment
0
30 30
60 60
90
Co-pol.
X-pol.
(a) E-Plane
90 90
(b) H-Plane
0
30
60
30
60
90
Co-pol.
X-pol.
Figure 7. Measured and calculated co- and cross-
polarized eld patterns at f 3.5 GHz, with a
12.5mm, e
r
9.5, l 12.0mm, W1.2 mm. (From
Leung and Ng [32], r2001 John Wiley & Sons, Inc.)
fed by the conformal strip of length l
1
and width W
1
. The
parasitic patch of length l
2
and width W
2
is displaced at f
0
from the excitation strip. Since the patch current involves
both y and f components, in this case, integrals of
associated Legendre functions will inevitably encoun-
tered. Therefore, a convectional coordinate system is
used in this section.
Image theory is used to remove the ground plane. The
equivalent imaged conguration, shown in Fig. 9, is a
spherical DRA with the strip and patch lengths doubled:
L
1
2l
1
and L
2
2l
2
.
3.2.1. Greens Functions. Since the excitation strip is
assumed to be slender, it has only a longitudinal current.
For the parasitic patch, however, the length-to-width ratio
is arbitrary. Therefore, both the latitude and azimuthal
currents have to be considered. The current distributions
are shown in Fig. 9. In the formulation, the superscripts A
and B refer to the excitation strip and the parasitic patch,
respectively. The
^
yy-directed current J
A
y
ows on the sur-
face of the excitation strip. For the parasitic patch, the
^
yy- and
^
ff-directed patch currents are dened as J
B
y
and J
B
f
,
respectively.
The
^
yy- and
^
ff-directed E-eld Greens functions due to a
^
ff-directed point current J
f
have been obtained in (1a) and
(1b). Now, two more Greens functions in (12a) and (12b)
are found that are due to a
^
yy-directed point current J
y
:
G
E
y
J
y
jZ
0
2par
1
n0
2n1
nn1
n
m0
n m!
nm!
cos mf f
0
.
m
2
D
TE
n
P
m
n
cos y
0
sin y
0
P
m
n
cos y
sin y
F
n
_
1
D
TM
n
D
m
d
dy
0
P
m
n
cos y
0
d
dy
P
m
n
cos yC
n
_
12a
G
E
f
J
y
jZ
0
2par
1
n0
2n1
nn1
n
m1
m
n m!
nm!
sin mf f
0
.
1
D
TE
n
P
m
n
cos y
0
sin y
0
d
dy
P
m
n
cos yF
n
_
1
D
TM
n
d
dy
0
P
m
n
cos y
0
P
m
n
cos y
sin y
C
n
_
12b
3.2.2. MoM. With the Greens functions presented
above, the strip and patch currents can be solved by using
the MoM. Enforcing the boundary condition that the total
E eld should vanish on the conducting excitation strip,
we have
A
E
y
J
y
B
E
y
J
y
B
E
y
J
f
E
i
0 13
a
W
1
Grounded parasitic patch
Hemispherical DRA
Conformal
excitation
strip
Coaxial
aperture
Ground plane
l
1
l
2
z
x
y
[
0
[
0
W
2
Figure 8. Conguration of conformal-strip-
excited DRA with a parasitic patch. (From
Leung and Ng [24], 2003 r IEEE.)
2l
1
2l
2
W
1
W
2
J
0
A
J
0
B
J
[
B
Figure 9. Equivalent geometry of the DRA conguration.
where E
i
is the impressed E eld on the excitation strip.
Equation (13) can be rewritten in terms of Greens func-
tions,
__
S
A
G
E
y
J
y
J
A
y
dS
0
__
S
B
G
E
y
J
y
J
B
y
dS
0
__
S
B
G
E
y
J
f
J
B
f
dS
0
E
i
0
14
where S
A
and S
B
are the surfaces of the excitation strip
and parasitic patch, respectively. For simplicity, the delta
gap source is used again to model the excitation,
E
i
V
0
=ady, where V
0
is the excitation voltage and is
set to unity for convenience. Let I
A
y
J
A
y
W
1
be the excita-
tion strip current and I
B
y
J
B
y
W
2
, I
B
f
J
B
f
L
2
; then (14)
becomes
1
W
1
__
S
A
G
E
y
J
y
I
A
y
dS
0
1
W
2
__
S
B
G
E
y
J
y
I
B
y
dS
0
1
L
2
__
S
B
G
E
y
J
f
I
B
f
dS
0
1
a
dy
15
The next step is to expand the currents using the MoM
as follows
I
A
y
y
N
1
p
1
1
I
yA
p
1
f
yA
p
1
y 16
I
B
y
y
N
2
p
2
1
I
yB
p
2
f
yB
p
12
y 17
I
B
f
f
N
3
p
3
1
I
fB
p
3
f
fB
p
3
f 18
where f
yA
p
1
y, f
yB
p
2
y, and f
fB
p
3
f are PWS basis functions
given by
f
yA
p
1
y
siny
A
h
jy y
A
p1
j
sin y
A
h
; ajy y
A
p
1
joh
1
0 elsewhere
_
_
19
f
yB
p
2
y
siny
B
h
jy y
B
p2
j
sin y
B
h
; ajy y
B
p
2
joh
2
0 elsewhere
_
_
20
f
fB
p
2
f
sin k
e
h
3
ajf f
B
p
3
j
sin k
0
h
3
; ajf f
B
p
3
joh
3
0 elsewhere
_
_
21
in which
y
A
h
h
1
a
; y
A
p
1
p
2
L
1
2a
p
1
y
A
h
; h
1
L
1
N
1
1
22
y
B
h
h
2
a
; y
B
p
2
p
2
L
2
2a
p
2
y
B
h
; h
2
L
2
N
2
1
23
f
B
h
h
3
a
; f
B
p
3
f
0
W
2
2a
p
3
f
B
h
; h
3
W
2
N
3
1
24
and k
e
e
r
1=2
_
k
0
is the effective wavenumber at the
DRAair interface. By employing the Galerkin procedure,
we multiply both sides of (15) by f
yA
q
1
y (1rq
1
rN
1
) and
integrate the result over the strip surface, yielding
1
W
2
1
N
1
p
1
1
I
yA
p
1
__
SS
A
__
S
0
S
A
f
yA
q
1
yG
E
y
J
y
f
yA
p
1
y
0
dS
0
dS
1
W
1
W
2
N
2
l 1
I
yB
p
2
__
SS
A
__
S
0
S
B
f
yA
q
1
yG
E
y
J
y
f
yB
p
2
y
0
dS
0
dS
1
W
1
L
2
N
3
n 1
I
fB
p
3
__
SS
A
__
S
0
S
B
f
yA
q
1
yG
E
y
J
f
f
fB
p
3
y
0
dS
0
dS
_
p=2 l
1
=a
p=2l
1
=a
f
yA
q
1
ydydy 25
Similarly, two more integral equations can be obtained
by enforcing the following boundary conditions on the
parasitic patch:
A
E
f
J
y
B
E
f
J
y
B
E
f
J
f
0 26a
A
E
y
J
y
B
E
W
J
y
B
E
y
J
f
0 26b
Applying the Galerkin procedure again, we multiply the
two equations by f
yB
q
2
y (1rq
2
rN
2
) and f
yB
q
3
f
(1rq
3
rN
3
), respectively. Totally, there are three sets of
equations with three sets of unknowns, which can be
solved via the following matrix equation
Z
AA
0
yy
p
1
; q
1
N
1
N
1
Z
AB
0
yy
p
1
; q
2
N
1
N
2
Z
AB
0
yf
p
1
; q
3
N
1
N
3
Z
BA
0
yy
p
2
; q
1
N
2
N
1
Z
BB
0
yy
p
2
; q
2
N
2
N
2
Z
BB
0
yf
p
2
; q
3
N
2
N
3
Z
BA
0
fy
p
3
; q
1
N
3
N
1
Z
BB
0
fy
p
3
; q
2
N
3
N
2
Z
BB
0
ff
p
3
; q
3
N
3
N
3
_
_
_
_
.
I
yA
p
1
N
1
1
I
yB
p
2
N
2
1
I
fB
p
3

N
3
1
_
_
_
V
A
q
1
N
1
1
0
N
2
1
0
N
3
1
_
_
_
_
27
where, for X,YA or B and a,by or f, we obtain
Z
XY
0
ab
p; q
1
G
X
a
G
Y
b
__
SS
X
__
S
0
S
Y
f
aX
p
aG
E
a
J
b
f
bY
q
b
0
dS
0
dS
28
V
A
q
1
_
p=2l
1
=a
p=2l
1
=a
f
yA
q
1
ydydy 29
with G
A;B
y
W
1;2
and G
B
y
L
2
. For the voltage matrix V
A
q
1
,
the center element is equal to unity but all the other
elements are zero. After the current vector I
yA
p
1
is obtained
from (27), the input impedance can be calculated from
Z
in
g=
N
1
p
1
1
I
yA
p
1
f
yA
p
1
0, where g 1 for the equivalent
spherical structure and
1
2
for the original hemispherical
structure. The other two current vectors I
yB
p2
, I
fB
p3
, to-
gether with I
yA
p
1
, will be used to calculate the radiation
elds of the antenna. To this end, the key step is to
evaluate the various impedance elements Z
XY
0
ab
, which
will be discussed next.
3.2.3. Evaluation of Z Matrix. Numerical evaluation of
Z
XY
0
ab
will be very difcult when XY, since the eld and
source points may coincide in this case, causing the
singular problem to occur. Using the previous approach
[18], all Z
XY
0
ab
can be calculated without the need for any
numerical integration. Thus, the computation time and
programming effort are greatly reduced. To begin with,
the Greens functions (1) and (12) are substituted into (29)
to get nine different expressions:
Z
AA
0
yy
p
1
; q
1

ja
2
Z
0
2pW
2
1
1
n1
.
n
m0
r
TE
n; m
.
m
2
.
Y
A
1
p
1
; n; m
_
Y
A
1
q
1
; n; m
.
F
1
m; f
A
; f
A
n
m0
r
TM
n; m
D
m
.
Y
A
2
p
1
; n; m
Y
A
2
q
1
; n; m
.
F
1
m; f
A
; f
A
_
30
Z
AB
0
yy
p
1
; q
2

ja
2
Z
0
2pW
1
W
2
1
n1
.
n
m0
r
TE
n; m
.
m
2
.
Y
A
1
p
1
; n; m
_
Y
B
1
q
2
; n; m
.
F
1
m; f
A
; f
B
n
m0
r
TM
n; m
D
m
.
Y
A
2
p
1
; n; m
Y
B
2
q
2
; n; mF
1
m; f
A
; f
B
_
31
Z
AB
0
yf
p
1
; q
3

ja
2
Z
0
2pW
1
L
2
1
n1
.
n
m0
r
TE
n; m
.
m
.
Y
A
1
p
1
; n; m
_
Y
B
4
n; m
.
F
2
q
3
; m; f
A
; f
B
n
m0
r
TM
n; m
.
m
.
Y
A
2
p
1
; n; m
Y
B
3
n; m
.
F
2
q
3
; m; f
A
; f
B
_
32
Z
BA
0
yy
p
2
; q
1

ja
2
Z
0
2pW
1
W
2
1
n1
.
n
m0
r
TE
n; m
.
m
2
.
Y
B
1
p
2
; n; m
_
Y
A
1
q
1
; n; m
.
F
1
m; f
B
; f
A
n
m0
r
TM
n; m
D
m
Y
B
2
p
2
; n; m
Y
A
2
q
1
; n; m
.
F
1
m; f
B
; f
A
_
33
Z
BB
0
yy
p
2
; q
2

ja
2
Z
0
2pW
2
2
1
n1
.
n
m0
r
TE
n; m
.
m
2
.
Y
B
1
p
2
; n; m
_
Y
B
1
q
2
; n; m
.
F
1
m; f
B
; f
B
n
m0
r
TM
n; m
D
m
Y
B
2
p
2
; n; m
Y
B
2
q
2
; n; m
.
F
1
m; f
B
; f
B
_
34
Z
BB
0
yf
p
2
; q
3

ja
2
Z
0
2pW
2
L
2
1
n1
.
n
m0
r
TE
n; m
.
m
.
Y
B
1
p
2
; n; m
_
Y
B
4
n; m
.
F
2
q
3
; m; f
B
; f
B
n
m0
r
TM
n; m
.
m
.
Y
B
2
p
2
; n; m
Y
B
3
n; m
.
F
2
q
3
; m; f
B
; f
B
_
35
Z
BA
0
fy
p
3
; q
1

ja
2
Z
0
2pW
1
L
2
1
n1
.
n
m0
r
TE
n; m
.
m
.
Y
B
4
n; m
_
Y
A
1
q
3
; n; m
.
F
2
p
1
; m; f
B
; f
A
n
m0
r
TM
n; m
.
m
.
Y
B
3
n; m
Y
A
2
q
3
; n; m
.
F
2
p
1
; m; f
B
; f
A
_
36
Z
BB
0
fy
p
3
; q
2

ja
2
Z
0
2pW
2
L
2
1
n1
.
n
m0
r
TE
n; m
.
m
.
Y
B
4
n; m
_
Y
B
0
1
q
2
; n; mF
2
p
3
; m; f
B
; f
B
n
m0
r
TM
n; m
.
m
.
Y
B
3
n; m
Y
B
0
2
q
2
; n; m
.
F
2
p
3
; m; f
B
; f
B
_
37
Z
BB
0
ff
p
3
; q
3

ja
2
Z
0
2pL
2
2
1
n1
.
n
m0
r
TE
n; m
D
m
Y
B
4
n; m
_
Y
B
0
4
n; m
.
F
3
p
3
; q
3
; m; f
B
; f
B
n
m0
r
TM
n; m
.
m
2
.
Y
B
3
n; m
Y
B
0
3
n; m
.
F
3
p
3
; q
3
; m; f
B
; f
B
_
38
where
r
TE
n; m
2n1
nn1
.
^
JJ
n
ka
^
HH
2
n
k
0
a
D
TE
n
.
n m!
nm!
39
r
TM
n; m
2n1
nn1
.
^
JJ
n
ka
^
HH
2
n
0
k
0
a
D
TM
n
.
n m!
nm!
40
F
1
m; f
X
; f
Y
_
f
X2
f
X1
_
f
Y2
f
Y1
cos mf f
0
df
0
df
41
F
2
p; m; f
X
; f
Y
_
f
X2
f
X1
_
f
Y2
f
Y1
sin mf f
0
f
fY
p
fdf
0
df
42
F
3
p; q; m; f
B
_
f
B2
f
B1
_
f
B2
f
B1
f
fB
p
f cos mf f
0
f
fB
q
f
0
df
0
df
43
Y
X
1
p; n; m
_
y
X
p
y
X
h
y
X
p
y
X
h
P
m
n
cos yf
yX
p
ydy
44
Y
X
2
p; n; m
_
y
X
p
y
X
h
y
X
p
y
X
h
dP
m
n
cos y
dy
sin y f
yX
p
ydy
45
Y
B
3
n; m
_
p=2l
2
=a
p=2l
2
=a
P
m
n
cos ydy 46
Y
B
4
n; m
_
p=2 l
2
=a
p=2l
2
=a
dP
m
n
cos y
dy
sin y dy 47
and
f
A1
W
1
=2a; f
A2
W
1
=2a; f
B1
f
0
W
2
=2a;
and f
B2
f
0
W
2
=2a
48
The integrals will be evaluated by either using the
recurrence technique or analytical integration. This will
be discussed in the following paragraphs.
3.2.3.1. Y
X
1
p; n; m and Y
X
2
p; n; m. To evaluate
Y
X
1
p; n; m and Y
X
2
p; n; m (XA or B) analytically, the
absolute sign of the PWS functions f
yA
p
y (19) and f
yB
p
y
(20) is rst removed by breaking each integral into two
parts. After several mathematical manipulations, the
following results are obtained
Y
X
1
p; n; m
1
sin y
X
h
siny
X
h
y
X
p
I
1
n; m
cosy
X
h
y
X
p
I
1
0
n; m
siny
X
h
y
X
p
I
2
n; m
cosy
X
h
y
X
p
I
2
0
n; m
49
Y
X
2
p; n; m
1
sin y
X
h
siny
X
h
y
X
p
I
3
n; m
cosy
X
h
y
X
p
I
3
0
n; m
siny
X
h
y
X
p
I
4
n; m
cosy
X
h
y
X
p
I
4
0
n; m
50
where
I
1;2
n:m x
1 x
2
p
P
m
n
x
_ _
x
2
x
1
Km; n
2
2n1
.
1
2n3
K
_
n2; m2
1
2n3

1
2n 1
_ _
Kn; m2
1
2n 1
Kn 2; m2
_
51
I
1;2
0
n; m
nn m1
2n12n3
Kn2; m1
n1nm
2n12n 1

nn m1
2n12n3
_ _
.
Kn; m1
n1nm
2n12n 1
Kn 2; m1
52
I
3;4
n; m
n m1
2n1
Kn1; m
mn
2n1
Kn 1; m
53
I
3;4
0
n; m
1
2n1
fKn1; m1
Kn 1; m1g
54
in which
Kn; m
_
x
2
x
1
P
m
n
x
1 x
2
p dx 55
with f x
x
2
x
1
f x
2
f x
1
. x
1
cosy
X
p
y
X
h
, x
2
cos y
X
p
for I
1;3
; I
1;3
0
and x
1
cos y
X
p
, x
2
cosy
X
p
y
X
h
for I
2;4
; I
2;4
0
.
For K(n,m), the recurrence formulas with arbitrary x
1
, x
2
have been obtained [37] as follows
Kn1; m
2n1
n1n m1
1 x
p
P
m
n1
x
_ _
x2
x1
nnmKn 1; m
56
Kn; m1 2P
m1
n
x
x2
x1
nmn m1Kn; m1
57
with initial values given by
K0; 0 sin
1
x
x2
x1
; K1; 0 1 x
2
x2
x1
;
Kn; 1 P
n
x
x2
x1
58
3.2.3.2. Y
B
3
n; m and Y
B
4
n; m. The integrals Y
B
3
n; m
and Y
B
4
n; m were studied in [18, Eqs. (11) and (12)], and
the results are given by
Y
B
3
n1; m
1
n1n m1
2n1 1 1
nm
_ _

1 x
2
b
_
P
m
n
x
b
nnmY
B
3
n 1; m
_
59
Y
B
3
n; m2 21 1
nm
P
m1
n
x
b
nm1n mY
B
3
n; m
60
where
x
b
cos
p
2
l
2
a
_ _
sin
l
2
a
61
and the initial values are
Y
B
3
n; 0 Y
B
3
n; 1 1
n
1P
n
x
b
62
The integral Y
B
3
n; m can be expressed in terms of
Y
B
4
n; m, which is given by
Y
B
4
n; m 1
nm
1
1 x
2
b
_
P
m
n
x
b
1
2n1
nmY
B
3
n 1; m
mn 1Y
B
3
n1; m
63
3.2.3.3. F
2
p; m; f
X
; f
Y
. The integrals F
2
p; m; f
X
; f
Y
are zero when m0. When m 6 0, the integration can be

performed analytically. The result is given by
F
2
2k
e
a
sin k
e
h
.
cos mf
X2
f
p
cos mf
X1
f
p
cos k
e
hcos mf
B
h
mk
e
amk
e
a
64
The last integral F
3
p; q; m; f
B
is given by (9). Now all
the integrals can be evaluated analytically. Note that Z
XY
0
ab
are simply regular series that can be implemented easily
in a straightforward manner.
3.2.4. Evaluation of Radiation Field Patterns. The far
eld of the antenna comprises radiation from the excita-
tion strip current I
A
y
and parasitic patch current compo-
nents I
B
y
; I
B
f
. The electric far eld is given by
E
y;f
r; y; f
jZ
0
a
2pW
1
.
e
jk
0
r
r
N
1
p
1
1
I
yA
p
1
.
E
A
0
yy;yf
p
1
; y; f
jZ
0
a
2pW
2
.
e
jk
0
r
r
N
2
p
2
1
I
yB
p
2
.
E
B
0
yy;yf
p
2
; y; f
jZ
0
a
2pL
2
.
e
jk
0
r
r
N
3
p
3
1
I
fB
p
3
.
E
B
0
yf;ff
p
3
; y; f
65
in which
E
X
0
yy
p; y; f
1
n1
j
n
2n1
nn1
j
^
JJ
n
ka
D
TE
n
_

n
m0
m
2
n m!
nm!
P
m
n
cos y
sin y
Y
X
1
p; n; mF
4
m; f
^
JJ
n
0
ka
D
TM
n
n
m0
1
D
m
n m!
nm!
dP
m
n
cos y
dy
Y
X
1
p; n; mF
4
m; f
_
66
E
X
0
yf
p; y; f
1
n1
j
n
2n1
nn1
j
^
JJ
n
ka
D
TE
n
_

n
m0
m
n m!
nm!
dP
m
n
cos y
dy
Y
X
2
p; n; mF
5
m; f
^
JJ
n
0
ka
D
TM
n
n
m0
m
n m!
nm!
P
m
n
cos y
sin y
Y
X
5
p; n; mF
5
m; f
_
67
E
B
0
fy
p; y; f
1
n1
j
n
2n1
nn1
j
^
JJ
n
ka
D
TM
n
_

n
m0
m
n m!
nm!
dP
m
n
cos y
dy
Y
B
3
n; mF
6
p; m; f
^
JJ
n
0
ka
D
TE
n
n
m0
m
n m!
nm!
P
m
n
cos y
sin y
Y
B
3
n; mF
6
p; m; f
_
68
E
B
0
ff
p; y; f
1
n1
j
n
2n1
nn1
j
^
JJ
n
ka
D
TM
n
_

n
m0
m
2
n m!
nm!
P
m
n
cos y
sin y
Y
B
4
n; mF
7
p; m; f
^
JJ
n
0
ka
D
TE
n
n
m0
1
D
m
n m!
nm!
dP
m
n
cos y
dy
Y
B
4
n; mF
7
p; m; f
_
69
where, for XA or B
F
X
4
m; f
_
f
X2;X2
f
X1;X1
sin mf f
0
df
0
70
F
X
5
m; f
_
f
X2;X2
f
X1;X1
cos mf f
0
df
0
71
F
B
6
p; m; f
_
f
B2
f
B1
sin mf f
0
f
fB
p
f
0
df
0
72
F
B
7
p; m; f
_
f
B2
f
B1
cos mf f
0
f
fB
p
f
0
df
0
73
The integrals F
X
4
m; f (70) and F
X
5
m; f (71) can easily
be evaluated in a single step. For F
6
p; m; f and
F
7
p; m; f, the results were obtained previously [38,
p. 302] and are summarized here:
F
X
4
m; f
1
m
cos mf f
X2
cos mf f
X1
mO0
0 m0
_
_
_
74
F
X
5
m; f
1
m
sin mf f
X2
sin mf f
X1
mO0
0 m0
_
_
_
75
F
B
6
p; m; f
2k
e
a
sin k
e
h
.
sin mf f
p
cos k
e
h cos mf
B
h
mk
e
amk
e
a
76
F
B
7
p; m; f
2k
e
a
sin k
e
h
.
cos mf f
p
cos k
e
h cos mf
B
h
mk
e
amk
e
a
77
It should be mentioned that, by using
~
HH1=Z
0
^ rr
~
EE,
the magnetic far elds can be found easily.
3.3. Results and Discussion
3.3.1. Convergence Check. The convergence checks for
the modal solution and MoM were done. In the following
calculations, 60 modal terms and N
1
N
2
5, N
3
3 were
used.
The numerical stability of the integrals Y
X
1;2
p; n; m
(XA or B) that utilize the new recurrence formulas has
also been studied. Although these results are analytically
exact, their numerical accuracy may be unsatisfactory
because of the nite precision of the computer. It was
found that the stability of the results decreased with
increasing order and degree of P
m
n
x, and therefore the
worst cases of Y
X
1;2
p; 60; 60 are considered, where p is a
dummy parameter in the stability check. The usable range
for the PWS half-mode angles y
A; B
h
as a function of the
PWS center angles y
A; B
p
is shown in Fig. 10. Only the range
01ry
p
A, B
r901 is shown because the results are symmetric
for 901ry
p
A, B
r1801. Since 01ryr1801 for the adopted
coordinate system, only the region y
A; B
h
y
A; B
p
(below the
straight line y
A; B
h
y
A; B
p
) should be considered for
0
y
A; B
p
90
. It is seen that Y
X
2
has a slightly smaller
stable region than does Y
X
1
and, hence, the overall stable
region is limited by Y
X
1
. The shaded area in the gure
shows the overall stable region, outside which the results
will become unstable. Note that care has to be taken for
y
A; B
p
o46
, as the half-mode angles y

A; B
h
in this case should
not be too small or unstable results will be obtained. In
this article, all results are calculated within the stable
region, and thus the accuracy is ensured. For Y
B
3
n; m (46)
and Y
B
4
n; m (47), the recurrence results [18] are very
stable and can therefore be used directly.
(degree)
20 30 40 50 60 70 80 90
0
10
20
30
40
50
60
70
80
90
Unstable
region
Stable region with error < 0.01%
(p,n,m)
(p,n,m)
1
X
2
X
0
p
A,B
(
d
e
g
r
e
e
)
0
h
A
,
B
0
h
A,B
0
p
A,B
=
Figure 10. The range of PWS-mode half-angle y

A; B
h
for stable
calculations of Y
X
1
p; n; m and Y
X
2
p; n; m, with n60 and
m60. (From Leung [37], r 2001 IEEE.)
135 140 145 150 155 160 165 170
0
20
40
60
80
100
120
140
7.7
7.8
7.9
8
8.1
8.2
8.3
8.4
8.5

P
a
r
a
s
i
t
i
c
-
s
t
r
i
p

l
e
n
g
t
h

l
2

(
m
m
)
P
h
a
s
e

d
i
f
f
e
r
e
n
c
e

b
e
t
w
e
e
n
E
0

a
n
d

E
[

(
D
e
g
r
e
e
)
Parasitic-patch location [
0
(Degree)
Figure 11. Phase difference between far-eld E
y
and E
f
as a
function of parasitic patch location, with |E
y
||E
f
|, e
r
9.5,
a12.5mm, l
1
14mm, W
1
1.2 mm, and W
2
2 mm. (From
Parasitic-patch width W
2
(mm)
P
a
r
a
s
i
t
i
c

p
a
t
c
h

l
e
n
g
t
h

l
2

(
m
m
)

0 1 2 3 4 5 6 7 8
7
7.5
8
8.5
[
0
= 156
[
0
= 157.4
[
0
= 158
0 1 2 3 4 5 6 7 8
88
89
90
91
92
F
a
r
-
f
i
e
l
d

p
h
a
s
e

d
i
f
f
e
r
e
n
c
e
b
e
t
w
e
e
n

E
[

a
n
d

E
0

(
D
e
g
r
e
e
)
[
0
= 156
[
0
= 157.4
[
0
= 158
Parasitic-patch widthW
2
(mm)
(a)
(b)
Figure 12. (a) Far-eld phase difference as a function of parasitic
patch width, with |E
y
||E
f
|, e
r
9.5, a12.5mm, l
1
14mm,
W
1
1.2 mm; (b) parasitic patch length as a function of parasitic
patch width, with |E
y
||E
f
|, e
r
9.5, a12.5mm, l
1
14mm,
and W
1
1.2 mm. (From Leung and Ng [24], 2003 r IEEE.)
3.3.2. Circularly Polarized DRA. It is well known that
the far elds E
y
and E
f
have to be equal in amplitude but
different in phase by 901 for ideal CP elds. The angular
positions of the patch f
0
, patch length l
2
, and patch
width W
2
are tuned to meet these two requirements. The
roles of the parasitic patch and excitation strip will be
discussed.
3.3.2.1. Effects of Parasitic Patch. To demonstrate the
results, a DR of dielectric constant e
r
9.5 and radius
a12.5 mm is used. The excitation strip has length
l
1
14 mm and width W
1
1.2mm. Figure 11 shows the
far-eld phase difference between E
y
and E
f
as a function
of the parasitic patch position f
0
with jE
y
j jE
f
j. The
condition jE
y
j jE
f
j is maintained in the curve by adjust-
ing the length l
2
of the parasitic patch. With reference to
the gure, the 901 phase difference is obtained when
f
0
157.41. The corresponding patch length l
2
is also
given in the gure, where it is found that l
2
should be
equal to 7.96mm in order to excite CP elds.
The effect of the patch width W
2
on the phase difference
is shown in Fig. 12(a) for different f
0
values. It is found
that the phase difference only increases slightly with W
2
.
Again, the condition jE
y
j jE
f
j is maintained by adjusting
l
2
, shown in Fig. 12(b). It is seen that in order to keep
jE
y
j jE
f
j, l
2
should decrease when W
2
increases. Note
that the length-to-width ratio (l
2
/W
2
) is almost the same
for different f
0
. From the results, it can be deferred that
the patch location is more important than the patch width
in the CP design.
3.3.2.2. Effects of the Excitation Strip. The effect of the
conformal strip length l
1
on the phase difference for
different W
2
is shown in Fig. 13(a). The corresponding
patch length l
2
for keeping jE
y
j jE
f
j is shown in
Fig. 13(b). It can be observed from Fig. 13(a) that by
increasing l
1
by 60% (from 10 to 16 mm) the phase angle
is reduced by only 4.2% (from 911 to 87.21) for W
2
4mm.
This shows that the strip length l
1
has only a small effect
on CP operation. Moreover, with reference to Fig. 13(b),
the patch length l
2
remains almost unchanged as l
1
varies,
meaning that virtually no adjustments of the parasitic
patch dimensions are required to maintain CP operation.
These are very favorable results, as this suggests that
once the antenna generates CP elds, the length of the
excitation strip can be varied alone in order to change the
Excitation strip length l
1
(mm)
10 11 12 13 14 15 16 17
87
88
89
90
91
W
2
= 2 mm
W
2
= 4 mm
W
2
= 6 mm
F
a
r
-
f
i
e
l
d

p
h
a
s
e

d
i
f
f
e
r
e
n
c
e
b
e
t
w
e
e
n

E
[

a
n
d

E
0

(
D
e
g
r
e
e
)
P
a
r
a
s
i
t
i
c
-
p
a
t
c
h

l
e
n
g
t
h

l
2

(
m
m
)

10 11 12 13 14 15 16
7
7.2
7.4
7.6
7.8
8
8.2
8.4
W
2
= 4 mm
W
2
= 2 mm
W
2
= 6 mm
Excitation strip length l
1
(mm)
Figure 13. (a) Far-eld phase difference as a function of excita-
tion strip length, with |E
y
||E
f
|, e
r
9.5, a12.5mm, W
1
1.2 mm, and f

0
157.41; (b) parasitic patch length as a function of
excitation strip length, with |E
y
|E
f
|, e
r
9.5, a12.5mm, W
1
1.2 mm and f
0
157.41. (From Leung and Ng [24], 2003 rIEEE.)
3 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8
0
20
40
60
80
100
120
140
160
l
1
= 14 mm
l
1
= 12 mm
l
1
= 10 mm
I
n
p
u
t

r
e
s
i
s
t
a
n
c
e

(
)
Frequency (GHz)
(a)
3 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8
40
20
0
20
40
60
80
100
120
140
I
n
p
u
t

r
e
a
c
t
a
n
c
e

(
)
l
1
= 14 mm
l
1
= 12 mm
l
1
= 10 mm
Frequency (GHz)
(b)
Figure 14. Calculated input impedance as a function of fre-
quency with 0-dB AR at f 3.56GHz: e
r
9.5, a12.5mm,
W
1
1.2 mm, and () l
1
10mm, l
2
7.89mm, W
2
2.73mm,
f
0
158.51; (- - -) l
1
12mm, l
2
7.88mm, W
2
2.73mm, f
0
158.51; (- - -) l
1
14mm, l
2
7.91mm, W
2
2.26mm, f
0
157.41: (a) input resistance; (b) input reactance. (From Leung

and Ng [24], 2003 r IEEE.)
input impedance, without signicantly disturbing CP
operation. This greatly facilitates the impedance tuning
of the CP antenna. It is noted from Fig. 13(b) that the
wider the patch width W
2
is, the smaller the patch length
l
2
is required.
Figure 14 shows the calculated input impedance for
l
1
10,12,14 mm. For each l
1
, the parameters of the para-
sitic patch are adjusted to give the 0-dB AR at f
3.56 GHz. With reference to the gure, the input impe-
dance can be altered by changing l
1
and, thus, the
impedance matching can be achieved for practical designs.
The parameters of the parasitic patch for each case are
shown in the gure caption. Note that the patch sizes and
locations for different l
1
are almost the same, which is
consistent with the discussion above. Since it is futile to
change the strip and patch parameters iteratively, the
design process is very simple.
3.3.3. Measured and Calculated Results. To verify the
theory, an experiment was carried out using a hemisphe-
rical DR of dielectric constant e
r
9.5 and radius a
12.5 mm. The excitation strip has length l
1
14 mm
and width W
1
1.2mm, whereas the parasitic patch
has length l
2
7.9 mm, width W
2
2.2mm, and angular
position f
0
157.41. Both the excitation strip and
parasitic patch were cut from a conducting adhesive
tape. Measurements were done with an HP8510C
network analyzer, and the reference place was set
at the coaxial aperture by using the port extension.
To reduce possible errors introduced by airgaps between
the ground plane and DR, the DR was rst put on the
adhesive side of a conducting tape [18]. Figure 15 shows
the measured and calculated input impedance for the
DRA. The measured and calculated resonant frequencies
(zero reactance) are 3.25 and 3.27 GHz with only 0.62%
error. The return loss (minimum |S
11
|) was also studied.
It was found that the measured and calculated matching
frequencies (|S
11
|o20 dB) are operated at 3.42 and
3.43 GHz, respectively, which are higher than the
zero-reactance values. A similar phenomenon was
observed in Ref. 36.
Figure 16 shows the measured and calculated ARs as a
function of frequency. Again, reasonable agreement be-
tween theory and experiment is obtained. The ripple in
the measured result is caused mainly by the nite ground-
plane diffraction. It is found that the calculated 3-dB AR
bandwidth is 2.4%, which is typical for a singly fed DRA.
The measured and calculated xz and yz plane radia-
tion patterns are shown in Figs. 17a and 17b, respectively.
Again, reasonable agreement between theory and experi-
ment is obtained. As expected, the radiation elds are of a
broadside mode. It is observed that the antenna is oper-
ated in a right-hand CP mode, with more than 27dB
difference between the right-hand and left-hand elds in
the broadside direction (y 01). It should be mentioned
that the antenna can be operated in a left-hand CP mode
by symmetrically displacing the parasitic patch on the
opposite side, that is, by changing f
0
from 157.41 to 202.61.
3 3.1 3.2 3.3 3.4 3.5 3.6
Frequency (GHz)
I
n
p
u
t

i
m
p
e
d
a
n
c
e

(
)
Theory
Experiment
Resistance
Reactance
40
20
0
20
40
60
80
100
120
140
160
Figure 15. Measured and calculated input impedance for the
DRA, with e
r
9.5, a12.5mm, l
1
14mm, l
2
7.9 mm, W
1
1.2 mm, W
2
2.2 mm, and f
0
157.41. (From Leung and Ng [24],
2003 r IEEE.)
3.2 3.25 3.3 3.35 3.4 3.45 3.5 3.55 3.6 3.65 3.7
Frequency (GHz)
0
5
10
15
20
25
30
35
A
x
i
a
l

r
a
t
i
o

(
d
B
)
Theory
Experiment
Figure 16. Measured and calculated ARs as a
function of frequency in the broadside direction,
with e
r
9.5, a12.5mm, l
1
14mm, l
2
7.9mm,
W
1
1.2 mm, W
2
2.2 mm, and f
0
157.41. (From
4. CONCLUSION
Conformal-strip-excited LP and CP hemispherical DRAs
has been studied theoretically and experimentally. In the
analysis, the MoM has been applied to nd the strip
currents. To reduce computation time and complexity,
the integrals associated with the associated Legendre
functions have been evaluated by virtue of recurrence
formulas. Linearly Polarized DRA has been introduced
as an angular strip displacement to facilitate impedance
matching. By adding a parasitic patch on the sidewall of
the DRA, a CP DRA can be excited. The parasitic patch
has been displaced at 157.41 from the excitation strip to
give the optimum AR around the design frequency. It has
been found that the AR and input impedance of the
antenna are controlled predominantly by the parasitic
patch and excitation strip, respectively. Because each
part can be designed independently, the design of the CP
DRA is quite an easy task.
Acknowledgements
This work was supported by a grant from the Research
Grant Council of the Hong Kong Special Administrative
Region, China (Project CityU 1178/01E).
BIBLIOGRAPHY
1. D. Kajfez and P. Guillon, eds., Dielectric Resonators, Artech
House, Norwood, MA, 1986.
2. S. A. Long, M. W. McAllister, and L. C. Shen, The resonant
cylindrical dielectric resonator antenna, IEEE Trans. Anten.
Propag. AP-31:406412 (May 1983).
3. M. W. McAllister, S. A. Long, and G. L. Conway, Rectangular
dielectric resonator antenna, Electron. Lett. 19:218219
(March 1983).
4. M. W. McAllister and S. A. Long, Resonator hemispherical
dielectric antenna, Electron. Lett. 20:657658 (Aug. 1984).
5. G. Drossos, Z. Wu, and L. E. Davis, Circular polarized
cylindrical dielectric resonator antenna, Electron. Lett.
32:281283 (Feb. 1996).
6. A. Ittipiboon, R. K. Mongia, Y. M. M. Antar, P. Bhartia, and M.
Cuhaci, Aperture-fed rectangular and triangular dielectric
resonators for use as magnetic dipole antennas, Electron.
Lett. 29:20012002 (Nov. 1993).
7. H. Y. Lo, K. W. Leung, K. M. Luk, and E. K. N. Yung, Low
prole equilateral-triangular DRA of very high permittivity,
Electron. Lett. 35:21642166 (1999).
8. K. W. Leung, K. M. Luk, and E. K. Yung, Spherical cap
dielectric resonator antenna using aperture coupling, Elec-
tron. Lett. 29:15301531 (Aug. 1993).
9. R. K. Mongia, A. Ittipiboon, P. Bhartia, and M. Cuhaci,
Electric-monopole antenna using a dielectric ring resonator,
Electron. Lett. 29:15301531 (Aug. 1993).
10. K. W. Leung, K. Y. Chow, K. M. Luk, and E. K. N. Yung,
Excitation of dielectric resonator antenna using a soldered-
through probe, Electron. Lett. 33:349350 (Feb. 1997).
11. Y. M. M. Antar and Z. Fan, Theoretical investigation of
aperture-coupled rectangular dielectric resonator, IEE Proc.
Microwave Anten. Propag. 143:113118 (April 1996).
12. N. C. Chen, K. L. Wong, and K. W. Leung, Input impedance of
inclined printed slot antennas and inclined-slot-coupled di-
electric resonator antennas, Microwave Opt. Technol. Lett.
12:4750 (May 1996).
13. K. W. Leung, K. Y. Chow, K. M. Luk, and E. K. N. Yung, Offset
dual-disk dielectric resonator antenna of very high permittiv-
ity, Electron. Lett. 32:20382039 (Oct. 1996).
14. K. W. Leung, Z. N. Chen, K. M. Luk, and E. K. N. Yung,
Aperture-coupled dielectric resonator antenna with a thick
ground plane, IEEE Trans. Anten. Propag. AP-46:12421243
(Aug. 1998).
15. H. Y. Lo, K. W. Leung, K. M. Luk, and E. K. N. Yung, Low
prole equilateral-triangular dielectric resonator antenna
of very high permittivity, Electron. Lett. 35:21642166
(Dec. 1999).
16. K. Y. Chow and K. W. Leung, Theory and experiment of the
cavity-backed slot-excited dielectric resonator antenna, IEEE
Trans. Electromagn. Compat. 42:287290 (Aug. 2000).
17. G. P. Junker, A. A. Kish, A. W. Glisson, and D. Kajfez, Effect of
an air gap around the coaxial probe exciting a cylindrical
(Feb. 1994).
18. K. W. Leung, Conformal strip excitation of dielectric resona-
tor antenna, IEEE Trans. Anten. Propag. AP-48:961967
(June 2000).
19. G. P. Junker, A. A. Kishk, A. W. Glisson, and D. Kajfez, Effect
of fabrication imperfections for ground-plane-backed dielec-
tric resonator antennas, IEEE Anten. Propag. Mag. 37:4046
(1995).
20. R. T. Long, R. J. Dorris, S. A. Long, M. A. Khayat, and
J. T. Williams, Use of parasitic strip to produce circular
(a)
(b)
30
60
90
0
30
60
90
30
30
0
60 60
90 90
Right
hand
Left hand
Right
hand
Left hand
Theory
Experiment
0 40302030 0 40302030
Figure 17. Measured and calculated right-
hand and left-hand eld patterns at
f 3.52GHz, with e
r
9.5, a12.5mm,
l
1
14mm, l
2
7.9mm, W
1
1.2mm,
W
2
2.2mm, and f
0
157.41: (a) xy plane;
(b) yz plane. (From Leung and Ng [24], 2003
r IEEE.)
polarization and increased bandwidth for cylindrical dielectric
resonator antenna, Electron. Lett. 37:406408 (March 2001).
21. K. P. Esselle, Circularly polarized higher-order rectangular
(Sept. 1996).
22. C. Y. Huang, J. Y. Wu, and K. L. Wong, Cross-slot-coupled
microstrip antenna and dielectric resonator antenna for
polarization, IEEE Trans. Anten. Propag. AP-47:605609
(1999).
23. M. T. Lee, K. M. Luk, E. K. N. Yung, and K. W. Leung,
Microstripline feed circularly polarized cylindrical dielectric
resonator antenna, Microwave Opt. Technol. Lett. 24:206207
(March 2000).
24. K. W. Leung and H. K. Ng, Theory and experiment of
circularly polarized dielectric resonator antenna with a para-
sitic patch, IEEE Trans. Anten. Propag. AP-51:405412
(March 2003).
25. R. K. Mongia, A. Ittipiboon, M. Cuhacim, and D. Rosecoe,
Circularly polarized dielectric resonator antenna, Electron.
Lett. 30:13611362 (Aug. 1994).
26. K. W. Leung, W. C. Wong, K. M. Luk, and E. K. N. Yung,
Circular-polarised dielectric resonator antenna excited by
dual conformal strips, Electron. Lett. 36:484486 (March
2000).
27. A. Petosa, A. Ittipiboon, and M. Cuhaci, Array of circular-
polarised cross dielectric resonator antennas, Electron. Lett.
32:17421743 (Sept. 1996).
28. Z. Li, C. Wu, and J. Litva, Adjustable frequency dielectric
resonator antenna, Electron. Lett. 32:606607 (March 1996).
29. Z. N. Chen, K. W. Leung, K. M. Luk, and E. K. N. Yung, Effect
of parasitic disk on a coaxial probe-fed dielectric resonator
antenna, Microwave Opt. Technol. Lett. 15:166168 (June
1997).
30. H. K. Ng and K. W. Leung, Excitation of CP aperture-coupled
dielectric resonator antenna with a parasitic patch, IEEE
Antennas and Propagation Soc. Int. Symp. Digest, Boston,
July 2001, Vol. 4, pp. 202205.
31. K. W. Leung, W. C. Wong, and H. K. Ng, Circularly polarized
slot-coupled dielectric resonator antenna with a parasitic
patch, IEEE Anten. Wireless Propag. Lett. 1:5759 (2002).
32. K. W. Leung and H. K. Ng, Dielectric resonator antenna fed
by a displaced conformal strip, Microwave Opt. Technol. Lett.
29:185187 (May 2001).
33. R. D. Nevels, The annular aperture antenna with a hemi-
spherical center conductor extension, IEEE Trans. Anten.
Propag. AP-35:4145 (Jan. 1987).
34. R. D. Nevels and J. E. Wheeler, Radiation from a dielectric
coated hemispherical conductor fed by a coaxial transmission
line, IEEE Trans. Electromagn. Compat. 31:1620 (Feb.
1989).
35. K. W. Leung, K. M. Luk, K. Y. A. Lai, and D. Lin, Theory and
experiment of coaxial probe fed dielectric resonator antenna,
IEEE Trans. Anten. Propag. AP-41:13901398 (1993).
36. K. W. Leung, K. M. Luk, K. Y. A. Lai, and D. Lin, Theory and
experiment of an aperture-coupled hemispherical dielectric
resonator antenna, IEEE Trans. Anten. Propag. AP-43:1192
1198 (1995).
37. K. W. Leung, Analysis of zonal and rectangular slot on a
conducting spherical cavity, IEEE Trans. Anten. Propag.
49:17391745 (Dec. 2001).
38. K. M. Luk and K. W. Leung, eds., Dielectric Resonator
Antennas, Baldock, Hertfordshire, England; Philadelphia:
Research Studies Press Ltd., 2003.
DIELECTRIC RESONATOR FILTERS
S. BILA
D. BAILLARGEAT
S. VERDEYME
P. GUILLON
IRCOM
Limoges, France
Because microwave ltering is an important function re-
quired to keep a number of different systems in working
order, the specications of a lter are varied. We can,
however, try to classify some of them as presented in
Table 1. The constrains are electrical, mechanical, ther-
mal, and commercial.
The objective in this article is to show the advantages of
the dielectric resonator (DR) technique to satisfy some of
these functions, along with its disadvantages, in compar-
ison with some other well-known solutions.
DRs are suitable for bandpass ltering. DR lters are
classied as three-dimensional (3D) devices, in opposition
to two-dimensional (2D) planar ones.
The main advantages of 2D solutions are their
relative bycompact dimensions, their easier integration
in circuit or module environment, and their well-known
design and manufacturing procedures. They are, however,
limited in their applications to the processing of low pow-
er, sizable relative bandwidth signal, in relation to the
poor unloaded quality factor of localized microwave ele-
ments or planar resonators. Some solutions are proposed
to restrict losses, such as applying supraconductors or ac-
tive-element techniques, but they remain inadequate to
replace 3D devices, in particular for high-power require-
ments.
In the class of 3D devices, designers have rst chosen
waveguides or metallic empty cavities to satisfy their very
narrow bandwidth ltering requirements. However, since
the mid-1980s, high-dielectric-constant materials, having
low loss tangent and good thermal stability, have become
available. The DR solution has been preferred for a num-
ber of applications, in particular spatial ones. This tech-
nique allows us to reduce significantly the cavities and
waveguide device sizes, for equivalent electrical and in-
creased thermal performances. Some average ratios can
be given for dual-mode resonators (DR compared with
cavity):
1 : 4 in volume
1 : 2 in mass
Moreover, the DR shape and the mode in which it is
excited can be chosen to give a response to particular re-
quirements, as we will see later in this article. A number
of DR shapes and lter topologies have, however, been
proposed. Our work here is limited to the presentation of
the most popular ones.
In this article, we present some characteristic
parameters of DR lters, which are generally introduced
974 DIELECTRIC RESONATOR FILTERS
Next Page

04a

Загружено:

Сведения о документе

Исходное описание:

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

04a

Загружено:

Авторское право:

Доступные форматы

DIELECTRIC LOADED ANTENNAS

(dB) Gain (dBi) E plane H plane E plane H plane

Figure 17. Standing voltage wave pattern in a short-circuited

Figure 13. Dielectric behavior of

are zero when m0. When m 6 0, the integration can be

, as the half-mode angles y

Figure 10. The range of PWS-mode half-angle y

1.2 mm, and f

157.41: (a) input resistance; (b) input reactance. (From Leung

Вам также может понравиться