476

IEEE

TRANSACTIONS

Analysis

and

the

ON MICROWAVE

THEORY

Evaluation

Complex

Permittivity

and

E.

AbstracfTheorv
and exuerhnentaf results are wesented to
show the possibility of using a resonant post technique for characterizing dielectric and magnetic materials at microwave frequencies.
Results of the temperature dependence of the relative dielectric
constant of nonmagnetic materials with e, varying from 4 to 60 are
presented and also loss tangent measurements at room temperature.
The complex permittivity and permeability of a number of garnet
varying
from 0.25
materials has also been measured with ~ yMJo
to 0.8. The measured real part of the permeability is in good agreement with the theoretical predktions of Sch16mann and the hnaginsw Psrt of the permeability agrees with measurements by Green
et al. on similar materials.

in

materials

with

the

production

dielectric

of

constants

gaps.

Cavity

containing

12 GHz

frequency

bands and capable of measuring

the dielectric
constant
to within
+ 0.5 percent and an error of less than 5)( 105
in the loss tangent.
Conventional
permittivity
been

of materials

exhaustively

Westphal
erally,

techniques

[2], and

the methods

for

measuring

at microwave

reviewed
Shestopalov

by

the complex

frequencies
Montgomery

and

can be subdivided

Yatsuk

[3].

into

have
[1],
Gen-

1) methods

which depend on the standing

wave field within
the
dielectric;
2) methods which depend on waves reflected
from the dielectric;
3) methods which depend on transmitted waves; and 4) resonance methods. The choice of
method, or combination
of methods, will depend on the
frequency,
the value of e, and tan 8, the amount
of
material

available,

the technique
surements.
The methods

the accuracy

is required

techniques
less than

by an (exact

EO1Ocylindrical
require

solution

However,

prepared

errors in the methods

required,

for research

and whether

or routine

mea-

1) are only suitable

for liquid

dielectrics,
while the methods of 2) are subject to considerable
errors in measuring
the complex
voltage
reflection
coefficient
as the dielectric
constant
becomes
large. The methods of 3) are suitable
for the measurement of large values of dielectric
constant, but have the
disadvantage
that they require carefully
prepared specimens, i.e., when used in a waveguidey
the specimens
have to be well polished and carefully
fitted to avoid air

suitable

this range

of the resonator

these

specimens

devised

methods

also

and are subject

to measure

to

the parameters

of the resonator,
with and without
the specimen.
The method described in the present paper is one of
a number of resonator configurations
in which the stored

niques

described

is negligible
and, hence,
than the resonator
tech-

in the preceding.

[4] used the dielectric

to measure

Bell

resonance

low loss tangents,

for measuring
and Gerritsen

and

Rupprecht

of a sufficiently
but

large

the method

is

the real part of the permittiv[5] show that by detecting

the density of modes over a frequency

interval,
the real
part of the permittivity
can be calculated
to within
a
10 percent accuracy using the RayleighJeans
formula.
Solving
plest

the boundary-value

problem

shapes of dielectric

approximations
real part

resonators

to rule out

of the permittivity

this

for even the siminvolves

technique

sufficient

as far as the

is concerned.

Two

methods

which restrict most of the stored energy to the dielectric

and allow
the experimental
configuration
to closely
approximate
the analytical
model are
by Hakki
and Coleman
[6], and Cohn
The Hakki
and Coleman
configuration,
sidered in the present paper, consists
dielectric
post sandwiched
between two
ing plates. Using the TEO.Z modes, the
gap

between

the

dielectric

negligible

and

the

since the electric

those suggested
and Kelly
[7].
which is conof a cylindrical
infinite conducteffect of the air

conducting
fields

plates

tend

to zero

at this point.
The fields outside
the dielectric
decay
very rapidly
and, hence, the use of conducting
plates of
finite dimensions
still provides a good approximation
to
the theoretical
model. In their paper, Cohn and Kelly
discuss the Hakki
and Coleman
method for measuring
high dielectric
constants
using the TEO1l mode, and also
propose a second method which also overcomes
the air
gap
at
the
dielectric-to-conductor
interface.
This
method
utilizes
a dielectric
post fitted in a cutoff circular

Manuscript
received November 6, 1969; revised February 27,
1970. This work was sponsored by the Department
of the Army.
The author is with the Lincoln Laboratory,
Massachusetts Institute of Technology, Lexington, Mass.

are only

10, although

as in the case of the HO1l and

cavities.

precisely

becomes
comprising

constants
the specimen,

ment

2 to

1970

Permeability

can be extended

not suitable
ity. Sabisky

the

8, AUGUST

of Measuring

perturbation

for dielectric

sample

varying
between 3.0 and 120.0, and 10SStangents
less than 5 X 104, have produced
a need for a measurespanning

NO.

COURTNEY

technique

MTT-18,

energy outside
the dielectric
leads to a higher filling factor

INTRODUCTION

developments

microwave

VOL.

Insulators

WILLIAM

ECENT

TECHNIQUES,

of a Method

Microwave

I.

AND

waveguide

circular
guide.

electric
Both

accuracy

whose
mode

methods

dimensions

are such

that

is cut off in the air region

offer

of measurement

similar

possibilities

and sensitivity

the

of the

as far as

are concerned,

COURTNEY

although
nique

: MEASURING

it appears
is limited

5.08 when
examines
offer

that

The structure

477

paper

it appears

to

size and frequency

<

THEORYDIELECTRIC
is essentially
length

a shorted

dielectric

of a cylindrical

by

placing

conducting

and

turning

the transmission

characteristic

equation

a resonator.
modes

The

is well

known

wave-

dielectric

is shorted

normal

INSULATORS

than

present

since

each end of the rod

into

tech-

greater

The

in the sample

i.e., a finite

waveguide

mode.

OF MICROWAVE

can be covered.
II.

guide,

TEOII

PERMEABILITY

waveguide

constants

dielectrometer

flexibility

AND

the circular

dielectric

the

the radial

more

range

that

to

using

PERMITTIVITY

rod

plates

at
line

for

the

(see [9]).

It isl
d??? (a)

Jm(cx)

Km(e)

+ L1(s)

1[
1
m[:++a[$+a

[ aJm(a)

@Km@)

01

Km(p)

d.(a)

2.0

r
6.0

[e, (lA/2L)2]l/2,

[(JV2L)

1]/,

L is the length

tiple

of the cylindrical

for some of the lower

order

of the dielectric

that

2, in a form
waveguide
the resonance

similar
cavity

specimen.

The

modes

is shown

mode

in Fig.

1,

is the second low-frequency

of the modes is required,

i.e., the TEO1l resonance,

mode and if wide separation

a large

value

of (lI/L)

should

be used,
The equations

for a and P in (2) can be rewritten

for

the TEO1l mode

&

(m+

= J12(aI)/[JI(aI)

G(d,)

where al, (31are the first roots of the characteristic

with

m = O, 1=1,

and jl

is the resonant

equa-

frequency.

Hence, by measuring
the frequency
of the TEOH mode
and knowing
the dimensions
of the specimen,
the real
part of the dielectric
constant
can be found.
The loss tangent has been shown by Hakki
man [6] to be, for the TEOll mode,

and Cole-

1 The notation of Hakki and Coleman [6] has been retained in

this paper.

[Ko(&)K2(f?1)

Tf 1P

()

R.=
where

al and (31 are again

teristic

equation

quency

of the TEo1l

shorting

plates,

with

(4)

(5)

F(al)G(/31)],

(6)

Jo(cw,(aJ],

(7)

~12(&)]/~1(61)

(8)

be evaluated

1[2

(9)

the first

roots

from

of the charac-

m = O, 1= 1, fl is the resonant
mode,

a is the conductivity

and QO is the unloaded

tric post resonator.

Using

(4) through

the measured

Q, and the specimen

freof the

Q of the dielec(9), the tan ~~can

resonant

frequency,

the

dimensions.

II 1. THEORYMAGNETIC

()

tion

B,

and

(3)

B1),

[1+

F(a,)

unloaded

e, = 1.0 +

&m,

B=

chart

to the mode chart for a cylin[1 O]. The mode chart shows

of interest,

2-

speci-

assuming a dielectric
constant of 15. The diameter
times
the frequency
squared, i.e., (Df) 2 cm2/s2 is plotted
as a
function
of the ratio of the diameter
to length squared
(D/L)
drical

L&

A
F(aJG(L?I),
%

A = I +

. . ., etc., corresponds
to the mulin the cavity
along the axial

half-wavelengths

direction

I
14.0

where

where -T~(a), K~(~) are Bessel functions

of the first and
second kind, respective y, A is the free-space wavelength,
men, andl=l,2,3,

A
=
Qo

tan8,

(2)
P = :

I
10,0

Fig. 1. Mode chart for a dielectric post resonator. The abscissa is

the ratio of the diameter to the length of the dielectric post. The
ordinate is the diameter multiplied
by the frequency.

D is the diameter,

(D/L)

where
a = ~

(1)

In their paper, Hakki and Coleman

[6] also suggested
that it should be possible to use the TM 011mode to measure the
described
ments

permeability,
provided
the specimen
can be
by a scalar permeability.
Recent
measureby Green
et al. [11] and theoretical
work by

Schlomann
[11] have indicated
that
this is a good
approximation
in a demagnetized
sample when T47rM,/
a <0.8. Since the TEo1l mode is easier to identify,
it is
proposed that this mode, rather than the TM U1 mode,
be used for both the dielectric
and magnetic
measure-

IEEE

478

TRANSACTIONS

ON MICROWAVE

THEORY

AND

TECHNIQUES,

AUGUST

1970

ments on ferrite
materials.
This has the added advantage that it does not reintroduce
the (air-gap
problem
that will arise with the TMOU mode. The characteristic
equation

for the TEO1l mode

becomes
Ko(&J

To(rrm)
._
Jo(a~)

C%.

(lo)

I.@m
K{(@m)

where
wDf~
am = ~
~Df~
~

pm=

where

pr is the

frequency
roots

[W.

[(c/2 Lfm)

relative

(11)

(c/2 Lam)],

(12)

1]/,

permeability,

f~

the

of the TEO1l mode, and am and &

of the characteristic

equation.

It is shown in the Appendix

tangent
is given by
II

tan 0. = = H(a.)
P.

resonant

are the first

that

the magnetic

A.
B.tan&
Q.

loss-

Fig. 2.

Exploded

view of the dielectric post resonator

sample and probes in position.

showing

the

(13)
}

where

(14)
Jo(a~)

+ J1(GJ

)1

J12(a~) Jo(a~)Jz(cx~)

A.=

1 + ~ F(a.)G(@.)
@

(15)

where F(a~), G(@~) are given by (7) and (8) with am and
& replacing
al and /3,, and Q~ is the unloaded
Q of the
TEOII
and

mode

in

dielectric

the

demagnetized

properties

can now

case. The

magnetic

be separated

as fol-

lows.
1) Apply a large magnetic
field to saturate
the ferrite
specimen and reduce the magnetic
losses close to zero.
The ferrite specimen is now equivalent
to the dielectric
resonator

of Section

measured

after

2) The
resonant
is again

I 1, and, hence, e, and tan 8. can be

identifying

sample

is now

frequency
measured.

and

the

new

and unloaded
Q- of the TEOII mode
Since e? and tan 8~ are now known,

DIELECTRIC

from

the theory

of

MEASUREMENTS

The construction
of the sample holder is shown in
Fig. 2. The center and bottom
plates are the two shorting plates which
turn the dielectric
rod transmission
line

into

raised
samples;
plated,

a resonant

or lowered
both
Coupling

structure.

The

center

to accommodate

the

center

(GHz)

Fig. 3. Usable range of radial dielectrometer with 0.085-inch diameter probes and 2.5-inch diameter conducting plates. The
dielectric constant, .,, as a function of the resonant frequency of
the TEOU mode for maximum
samples.

and minimum

values of dielectric

the TEO1l mode.

demagnetized

then P, and tan & can be calculated

Section III.
IV.

FREOUENCY

and

to the sample

bottom

plate

various
plates

is achieved

can be

lengths

of

are gold-

by the right-

angle E-field probes, both of which can be moved in and

out radially,
with respect to the sample,
to vary the
coupling
coefficients.
Placing the specimen exactly in the center is not critical; it merely needs to be placed approximately
symmetrical
with the two probes. Since the unloaded
Q of
the TEOII mode must be measured, the coupling
of each
probe is reduced until no change is detected in the resonant frequency
or 3-d B linewidth
of the resonance.
If
the sample is isotropic,
the TEOI1 mode can be identified
by finding

the second low-frequency

by the mode chart

TEOnZ mode
center plate.

in Fig.

is achieved
As the plate

1. Further

mode,
proof

as indicated
that

this is a

by raising
and lowering
the
is raised, the TM modes move

COURTNEY

: MEASURING

PERMITTIVITY

TABLE

AND

PERMEABILITY

OF MICROWAVE

INSULATORS

479

TABLE

II

HOT-PRESSED
ALUMINA

DIELECTRIC MEASUREMENTS

Material

Loss
Tangent
(x Iv)

Fused quartz

3.83

2.0

Frequency
(GHz)

Diameter
(in~he5)

(Inches)

11.789

0.700

0.350

9.75

0.76

9.73

0.84

9.905

0.701
0.530

0.175
0.265

16.17
16.11
16.10

0.76
0.89
0.62

9.497
7.723
5.848

0.700
0.530
0.700

0.175
0.265
0.350

30.40

2.8

5.633

0.530

0.265

B*

42.4

4.5

4.775

0.530

0.265

c*

63.7
63.8
63.8
63.7t

3.2
2.9
3.1

2.131
2.220
2.059
6.190

1.125
0.999
1.249
0.400

0.443
0.443
0.443
0.150

Alumina

99.7 percent
Magnesium
titanate

12.223

Sample
Number

Length

* Composition
not given by manufacturer.
t Temperature
dependence was measured with
cavityno
loss tangent measurements were made.

it

;:2
6.5
6.6
6.6

9.812
9.809
9.800
9.800
9.804

where
this

that
would

the
to

the

to

higher

remains

stationary.

constant

is known

frequencies
In

most

three conditions

be identified

unambiguously.

usable

built

TEOII
the

so that

enables

mode

dielectric

the frequency

can be predicted.

of the above

The

the

situations,

approximately

of the TEOI1 resonance

eter

while

A combination

for the value

of the particular

values

dimensions

radial

is shown

dimensions.

are determined

by the diameter

3.0

and

dimensions
by
Two specimens

dimension

one was material

constant

nitride
at

limits

of the

C of Table

of 63.8 and

resonating

The

1,

a frequency
with

9.224

of

aligned

0.530 inch, length

along

0.265 inch, and

to within
1.6 inches,

diameter

the

0.25 inch,

by the present

with

C-axis

aligned

1/2. The first specimen

technique;

Bureau

the second

of Standards.

agreement

between

the

dielectric

error in the value of the conductivity

the metal wall losses could account

shown
quite
[13],

in Table

9.0 GHz.
the

It

Results

has been pointed

out

in the theory

of Section

II

that the measurement

of the loss tangent
depends only
on measuring
the dielectric
constant
and the unloaded
Q of the resonator.
However,
this implies using a value
of the bulk conductivity
tained
from some other

of the conducting
plates,
source, and substituting

value into the theoretical

expression
It would be preferable
if a technique
to measure
the surface resistance

obthis

for the loss tangent.

could be developed
of the conducting

constants

is

1. The

values

for

fused

The

values

dielectric

constant

quoted
and

by the
loss

agree

manufacturer

tangent

Temperature
For

these

of

the

for
99.7

titanate
are
tan 8.= 104,
control
runs.

on the consistency
of the measurements,
were cut from the same bar of hot-pressed

alumina,
The results are shown in Table
facturers
catalog
values
are e,= 9.9
2.5x 10-5,
B.

quartz

well with measurements

reported
by Bussey et al.
i.e., 6,=3.83,
tan s3,=l.2x
10- at a frequency
of

As a check
five samples

Temperature

at

were

This correction
has been applied to three specimens of
magnesium
titanate
in Section V.
Measurements
on a number
of other materials
are

temperature
Room

results

used to calculate
for the differences.

percent
dense alumina
and magnesium
E,=9.25,
tan ~.=104,
e,= 15.85, and
respectively,
for the particular
quality

A.

the

well within
the accuracies of the two types of measurement, while in the case of the loss tangents,
a 20 percent

GHz.

at 6.14 GHz.

along

was measured

was measured

The

smallest
sample tested so far is a specimen
of single
crystal
rutile,
diameter=
0.375
inch,
length=
0.125
inch, which had a dielectric
constant
of 87.0 at room
and resonated

is at present

was a rod of length

a dielectric
The

use

accuracy

1/2; the second specimen

The

3 [12],

an

the axis of the disk

dielectrom-

in Fig.

of specimen

the second was boron

of

C-axis

tan 8, = 1.5 X 105 A 40 percent.

a dielectric

constant

the

e, = 9.41 + 0.4 percent

dielectrometer:

2.06 GHz;

of the loss tangent

2 (rod)

laboratory

and

The

sample

were used to test the maximum

with

known,

resistance.

tan 6. = 3.6 X 105

of the shorting
plates and the minimum
the diameter
of the coupling
probes.
radial

specimen,

E. = 9.415

and minimum

maximum

a stanclard

1 (disk)

range

time,

such

outside
the capability
of presently
known
microwave
techniques,
so the following
alternative
was used. Two
samples were cut from a boule of single crystal sapphire:

the National

where the dielectric

constant
is plotted
as a function
of
the resonant frequency
of the TEOH mode for the maximum

At the present

sample

in this

0.265
0.265
0.265
0.265
0.265

of performing

is accurately

surface

Length
(inches)

0.5302
0.5303
0.5305
0.5305
0.5305

be to obtain
the

rod axis to within

the TEO1l mode to

way

loss tangent

measure

required

only

one a disk of diameter

rapidly

Diameter
(inch,es)

being used in the experiment.

appears

experiment

an aluminum

Frequency
(GHz)

9.92
9.92
9.94
9.94
9.93

;
3
4
5

plates

A+

Loss
Tangent
(xlO)

Coefficient

measurements,

II. The manuand tan 8,=

Measurements
the

cavity

was

placed

in

an oven in which
the temperature
was stabilized
to
within
+ 2.0C and the resonant
frequency
and linewidth of the TEo1l resonance measured when the mode
became stationary
at each value of temperature.
Figs.

IEEE

TRANSACTIONS

ON MICROWAVE

THEORY

AND

TECHNIQUES,

AUGUST

1970

10.18[
64.0

Hot-Pressed Alumina (sample 6)

D = 0.530 inch; L = 0.265 Inch

10.14
630

t
1008

D-7 /

62,0

/%%i+3

10.06
t

04
1/
/

61.0
Material
C
D = 0,400 inch; L = 0.150 inch
Temperature
Coefficient
= -543

PPM At 20C

60.0~

9.98

.
a

30

10

50

Gold -Plotin~ I Removed

Gold -Ploted

70

90

110

13

130

TEMPERATURE
TEMPERATURE

(C)

(C)

Fig.

Fig. 4.

Dielectric
constant versus temperature
for material C of
Table I. Temperature
coefficient=
(1/,,) (de,/dZ) at 20C.

7. Dielectric
constant
versus temperature
for hot-pressed
alumina! showing the effect of an air-gap between the sample and
conducting plate.

1~
>

u
fi
3
~

!17S6

f
~
z

t 1,784
0

s
o
03

D = 0,700

inch;

Temperature
41.51
to

30

50

70

TEMPERATURE

Fig. 5.

90

110

162

for material

Fig. 8.

)-

Tifanote

1~
D = 0530
inch; L = 0.265 inch
Temperature
Coefficient
= 116 PPM At

I
50

PPM at 20 C
I
I
I
I
90
ito

= +32

I
70

TEMPERATURE

B of Table I.

Magnesium

I
30

(C)

Dielectric constant versus temperature

16,3

11,776
10

130

0.350 inch

L =

Coefficient

20

:
g

(C)

Resonance frequency of the TEoll resonance versus

temperature for fused quartz.

Hot- Press6d Alumina

(sample
2 )
D = 0.530 inch; L = 0.265 inch

10.05

Temperaf.re

Coefficient

= 129 PPM At 20 C

16,i

TEMPERATURE

Fig. 6.

9.901
10

(C)

Dielectric constant versus temperature

for magnesium titanate.

I
30

I
30

I
70

TEMPERATuRE

Fig. 9.

I
90

I
flo

(c)

Dielectric constant versus temperature

for hot-pressed alumina.

I
130

COURTNEY

: MEASURING

PERMITTIVITY
TABLE

AND

PERMEABILITY

OF MICROWAVE

III

V. ACCURACY

MEASURED AND CORRECTED VALUES

TEMPERATURE COEFFICIENTS (TC)
Fr~~uez~y

,, at ~OoC

For

specified

Measured
TC

Cory:ted

the

and

error

11.79

3.823

9.80

9.93

+130

Hot-pressed alumina

32

14

+116

7.73

16.10

+116

.5.64

30.4

434

450

4.77

42.4

170

187

6.19
2.22

63.8
64.1

543
517

562
537

Magnesium

titanate

Single crystal
1 C-axis

6.14

rutile

Polycrystalline

rutile

87.0

723

98.0

3.14

99

741

813

829

D and
MHz,

tive

since

avoided
bulk

70

ture

Loss tangent
alumina

9 show

range

25

inch

these

results

Table

III

measured

120C
to

150

be

in the tempera-

a number
change with

corrected

for

gives the measured

results

value

of specimens.

for a number

of

[15 ], which

agrees very

closely

with

The
alumina

of Loss Tangent

loss tangents
as a function

of fused

with

this

Bady

the value

and

[14]

consistent
is varied.

results

can

in

nearest

0.3 percent.

I to make

The first

an extreme

If we assume

and

error

some esti-

sample

is instruc-

case which
that

can be

the value

too high

of the

for use in (4)

of 2.0X 105 is involved

3.6 X 10-5 in the

in conductivity,

rises to 6.0X
by proper

10s, but

choice

first

in practice,

of sample

For

in the first
this

dimensions
3.0X

in the

result.

the error

to tan 86 is less than

can be limited

can be
so that

105. This

to 2 X 106 for a 30 percent

corerror

if the TEOSSmode is used in the measure-

However,

a criticism

obtained

of the

with

American

Society

technique

a higher

to compare

of the

use of the

higher

linewidth
the

mode.

measurements

for Testing

and

order

the requirements
Materials

present

on the

between

the

[14 ] perturba-

method.

We chose an

example
where the perturbation
formulas
have some
validity,
i.e., a tan 6, M104 and c,x1O. Then, for the
perturbation
technique
(16)

hot-pressed

are shown

be obtained

where

Q, is the unloaded

rod inserted,
and the
Thus,

6(Af)

frequency

with

the sample

Q of the empty

of the

order

cavity,

of 9.0 GHz.

103
= ~
X 9 X 109 = 0.018 X 10CHZ.

(17)

Thus, the difference

in the two linewidths
is 18 kHz.
For the present method,
(4) can be rewritten

in Fig.

as the

Q of the cavity

Q. is the unloaded

resonant

of 130

10. Over the temperature

range involved,
little change
in the loss tangent
was detected,
but the results show
that quite
temperature

to the

change.

and the corrected

Temperature

quartz

of temperature

inch

measured

materials.

at 20C obtained
in the present measurements.
specimens were cut from the same bar of alumina.

C. Variation

the value

j: 0.0005

temperature,

obtained
a measured
value of 132 ppm at 20C for
hot-pressed
alumina
using a fully-filled
cavity technique
ppm
Both

with

of

in Table

an error

results

accuracy
tion

and hot-pressed

results

for

dimensions

have

130

temperature.

the measured

to

Since the specimen

versus

varies

order modes is that they limit

the measurements
to
above 12,0 GHz for sample dimensions
smaller
than
D = 0.53 inch, L = 0.265 inch. For engineering
purposes,
the advantage
of flexibility
with reasonable
accuracy
obtained
using the TEO1l mode outweigh
the increased

(C)

of fused quartz

the

constant

is 20 percent

1, Thus,
two

the correction

110

TEMPERATURE

4 through

it represents

in conductivity

90

of

measurement,

tolerances

involved.

It is interesting
50

dimensions

and (9), then the new values of loss tangent would read
0,4 X 104, 0.70 X 104, and 0.45X 104, in the order of

rection

lo-,~

Fig. 10.

samples

conductivity

accuracy

30

the

frequency

frequency

in practice.

avoided

L = 0.350

for

is less than

titanate

sample

c Hot-Pressed
Alumino
(sample
2)
D = 0.530
inch; L = 0.265
inch

10

the

mates of the error

a 30 percent

on

of the

dielectric
cases,

the error

ments,

0 Fused Quartz
D =0.700
inch;

most

L and

nesium

last

@n-

MEASUREMENTS

It has been pointed out earlier that the measurement

of the loss tangent
depends on using the value of the
bulk conductivity
for gold in the expression
for tan ~e
[i.e., (4) ]. We can use the results on the three mag-

Table

In

DIELECTRIC

tolerances
accuracy

in the

of cr. In
Fused quartz

OF THE

481

OF

specimen
Material

INSULATORS

tan b.=

11

,
() Qd
Qo

since the constant

A is close to unity
the unloaded
Q of the system if there

(18)

(1.015) and QO is
were no dielectric

IEEE

482

TRANSACTIONS

ON MICROWAVE

losses. The QO is close to the value of a high-quality

resonator
used in the perturbation
technique,
so the
comparison
should be valid. In this case,
8(Aj)
so that the
times larger
present

difference
in linewidths
than in the perturbation

method,

loss tangent,

(19)

= 0.9 X 106 Hz,

to avoid

errors

it is necessary

in this case is 50
method.
In the

greater

than

to measure

105 in the

the linewidth

to

within
0.1 MHz. Adding
to this error, the one due to
possibly
a 30 percent
error in the conductivity,
the
total error in the loss tangent
is less than 4.0X 105.
The difference
in the loss tangents
of two materials
with
sured

the same nominal

with

much

dielectric

greater

constant

accuracy.

can be mea-

In this

case,

in the resonant

()
~

the resonance

Iinewidths

samples.
to within

efficients.
However,
it should
do this sufficiently
accurately
105 in loss tangents

extremely

overcome

it was stated

the

dielectric
constant
of this statement,

air-gap

measurements.
a sample
of

that

problem

inherent

in

sample 6, was prepared,

and its dielectric
constant
as
a function
of temperature
was measured.
The flat
surfaces of the disk were then gold-plated
and the measurements repeated. The gold-plating
was then removed
and the measurements
again repeated.
The three sets
7. The two sets of
agree to within

0.06 percent, which is within

the limits of the frequency
and dimensional
tolerances
of the measurements.
The
difference
between
these values
and the gold-plated
measurements
was 0.3 percent,
which
is larger than
expected
if the dimensions
of the sample
have not
changed during the gold-plating.
A change of 0.3 percent
represents
a change in resonant
frequency
of about 12
MHz,
which
can easily be measured.
There are two
possible factors
first is the fact

contributing
to this discrepancy.
The
that the sample has not been polished

so that the surface finish of the flats at the conducting

plates is of the order of 100 ~in. By measuring
the gap
required for a frequency
shift up to 100 MHz, it is found
that a 100-~in gap should produce at most a frequency
shift of 3.5 MHz. The second contribution
is due to possible overlap
of the gold film at the edges of the disk
since these were not completely
masked during evaporation and plating.
A change in effective
length
of the
sample from 0.265 inch to 0.2645 inch produces a shift

to humidity,

exception
even though
ratio to the atmosphere

11.0 MHz

1970

so that

and this

for all three

tangent
method

are
is no

it offers less surface-to-volume

than others. It is not sufficient

to clean and heat the sample and then keep it in a dessicator

until
the measurement
is made. If there is a
level of humidity
greater than 50 percent in the laborait is possible

to watch

the

linewidth

broadening

with time as soon as it is taken out of the dessicator

and placed in the cavity.
Measurements
on the samples
this

report

higher

in

gave

a room

with

a loss tangent
without

two

air-conditioning.

VI.
A number

field

Even

the

times

than

in

hot-pressed

alu-

had values
compared

of
to

GARNET MEASUREMENTS

of disks
material

were
It

required

to an insignificant

cut

to provide

frequencies.

intensities

to four

air-conditioning

mina, which is a very dense material,

1.4 X 104
without
air-conditioning
7.0 X 106 with air-conditioning.

different

this technique

To test the validity

hot-pressed
alumina,

of measurements
are graphed in Fig.
measurements
without
gold-plating

sensitive

each garnet

can be detected.

In the Introduction,
would

be possible to arrange to
that changes of less than

of about

AUGUST

cent. The measured temperature

coefficient
conditions
was 130 ppmflC
at 20C.
Most
measurements
of dielectric
loss

room

0.01 MHz and, hence, a difference

of 108 could be detected in materials
with loss tangents
of the order of
104. This is an optimistic
figure since it implies
that
one sample can be removed
and the second one placed
in position
without
any change in the coupling
co-

frequency

TECHNIQUES,

rections,
the results agree to within
0.4 percent
and
there appears to be no need to polish the samples since
a 100-win finish produces an error of less than 0.1 per-

in

since the B term in (4) will be the same for both

It is easy to measure

(20)

AND

the above effects can account for the discrepancy

in the
measurements.
However,
even neglecting
these cor-

tory,
tan 8.(1) tanti.(2)

THEORY

from

the

same

measurements

was found
to reduce

that

bar

the

magnetic

the magnetic

level were quite

large,

of

at three
effects

so one sample

of each garnet material

was measured in a field of 40
kOe. The value of the dielectric
constants
for the four
garnet materials
are shown in Table IV. This value of
dielectric
constant
was used to predict
the asymptotic
high-field
value
of the resonance
frequency
of the
TEO1l resonance

for the remaining

The

of the

remainder

samples

measurements

in each series.

were

made

with

magnetic-field
intensities
equal to 13.2 kOe. The errors
involved
in this procedure
will be discussed in the next
section.
The measurement
procedure
was carried out as outlined in Section II 1. The disk samples were annealed,
then cleaned by washing in dilute nitric acid, rinsed in
distilled
water,
rinsed in acetone,
rinsed in isopropyl
alcohol, and finally
heated to 120C for approximately
half an hour. When the sample had cooled to room
temperature,

the

resonance

frequency

and

3-dB

line-

width
of the TEu1
resonance
of the demagnetized
sample were measured.
The resonator
was then placed
between
the magnet
pole-pieces
and the high-field
values of the resonance frequency,
and the 3-d B linewidth
measured.
The dielectric
results are tabulated
in Table V and the magnetic
results graphed
in Figs.
11 and 12. The value of the dielectric
constant
for each
sample

within

the same series is constant

been used to predict

field

resonance.

the asymptotic

value

since this has

of the high-

COURTNEY

: MEASURING

PERMITTIVITY

TABLE

AND

PERMEABILITY

OF MICROWAVE

INSULATORS

IV

TABLE

DIELECTRICCONSTANTOFGARNETS
(Applied field =40 kOe)
Diameter
(inches)

Material

Y3Fe,01,
[email protected]
y~,6@0,~10.4Mn0.0*Fe4.

K%)

a801Z

3 Fe4,9 M*0,4

2,6 *0.4

2,3 d0.7

Y2,6 Gdo,4~.sz

Diameter
(inches)

(inches)

Y2.~GdO.TAIO,lFei,eOlt
c,= 15.24

0.701
0.531
0.701

0.350
0.265
0.175

6.003
7.928
9.775

1.47
1.46
0.96

y2 .&do

0.701
0.531
0.701

0.350
0.265
0.175

6.065
8.012
9.880

1.25
1.21
1.81

YsMnO.lFeA,gOlt
6,=15.15

0.701
0.531
0.701

0.350
0.265
0.175

6.025
7.956
10.326

0.93
1.0
1.66

Y,,,Gdo.4A10,~Mn0,0,Fei,5~Ol,
,,=15!01

0.6005
0.6005
0.6004
0.5003

0.400
0.300
0.200
0.167

6.049
7.059
9.222
111.050

1.74
1.29
1.22
1.80

Material

~ggy

0.350
0.350
0.350
0.350
0.400

5.927
6.003
6.065
6.025
6.049

DIELECTRIC PROPERTIES

Fre-

0.701
0.701
0.701
0.701
0.601

Yz.8Gd0 TAb.4Fe4..@lZ
Y1.6Gd0. ddo,E2Fe4.4s0,2

483

15.66
15.24
14.94
15.15
15.01

4AI0 mFt?&.ao,z

6,=14.94

Fre-

Length

042

k to,4

Mnoo20,2

~04

e4.6

{2

e44s

042

The solid line in Fig. 11 is computed

cal expression derived by Green et al.
mental

points

theoretical

agree

i.e., 747riV/w

<0.8.

some

experimental

results

y47rAZ/U.
quite

in Fig.

03

04 05
(wM/w)

06

07

0.0 09

0 Y3Fe49 Mno , 0,2

40-2

26 Gd0,4 J04

2,3 G0.7 A 20.4 e4.6

n002

e458

42

2,6 G0,4A10,L52

%.6 d0,4

JO

4.460{2

52 e4.480iz J ~~~

A Y3Fe5 0{2 ! J. J. Green al al)

percent

where

purposes

of

Green

12, where

agreement

of

the

the theory

is

of comparison,
et al.

[11]

the imaginary

between

In a magnetic
tensor

OF THE

the

field

GARNET

of 40 kOe,

properties

of the

In

this

configuration,

it

of

the

measurement

of

have

part

of

as a function

of

two

is

methods

MEASUREMENTS

the

mode

splitting,

permeability,

is estimated
the

due

disappears.

that

dielectric

the

accuracy

constant

is com-

parable
to that obtained
with dielectric
materials,
i.e.,
SO. 5 percent. To estimate the error involved
in using a
field of 13.2 kOe, the resonance frequency
of the TEOU
resonance was plotted as a function
of applied magnetic
for the three
samples
of Ya,aGd0,TA10,AFeA,6012. The
results are shown in Fig. 13. The asymptotic
value for

.:
/-,~,1
OY

042

three

(~) has been plotted

ACCURACY

to the
Fig. 11. Relative permeability
as a function of the ratio ~47rM/w
The solid line is given by the theoretical expression on the graph.

The

For

from a theoreti[11 ]. The experi-

satisfactory.

VII.
0.2

within

in the range

valid,

the permeability

04

to

expression

been included

07~
o

tan 6
X104

1%:7

the lowest

This

resonance

gave a value

the material,

totic

which

resonance

was measured

in a field of 40 kOe.

of 15.24 for the dielectric

for

constant

was then used to predict

the

other

two

samples.

of

the asympThe

error

involved
in measuring
the dielectric
constant
with a
field of 13.2 kOe is shown in Table VI. The results for

P
i 0-3

/
/

AD

/
f

one material
have been shown since they are typical
for the garnets
measured.
To maintain
the error less
than one percent for the garnet materials,
i.e., materials
with 47rll < 2000G, e,~l 5.0, the following
constraints
are necessary:
1)

ratio

of

diameter

2) resonant

40-4
0.4

04

length

(D/L)

of the

<2.0,

TEOII

resonance<

9.0

GHz,

3) magnetic-field

0.2

to

frequency

06

0.6

intensity

X15

kOe.

{,0

A further

((JJM/aJ)

Fig. 12. The imaginary part of the permeability as a

function of 74zM/w

is

(D/L)

resonance.

constraint
>1.5

For

on the

dimensions

of the

sample

to avoid
degeneracy
with
the HE1lz
frequencies
above 9.0 C,Hz, the accu-

IEEE

484

TABLE

VI

ERROR IN DIELECTRIC

~&-

Material

CONSTANT
Dielectric
Constant

Length

(inches)

TRANSACTIONS

inches)

H=

13.2

High

kOe
Yt,SGd0,7A10,AFeA601Z O. Tol
47rM = 786 G
0.531
0.701

15.26
15.26
15.26

:::
3.5

THEORY

apply

to nonmagnetic

For

the real

MATERlALY23Gdo7

between

the

I
PREDICTED

D E 0.53i

INCHES

L = 0.265

INCHES

70

obtained

using

4.5 percent

in the worst

the difference
the

field of 13.2 kOe and the asymptotic

results

values

discussed

for

A demagnetized

still

shows

lent

to the splitting

the

sample

mode

still

sample

splitting,
k

has

dimensions

a field

From

in zero applied

approximately
of 13.2 kOe,

off-diagonal

permeability.

measurements

using

equivaand

components

the measurements

hence,
of

ever,
H = 40 kOe

the real part

to frequency
this
D = 0.701

INCHES

L = 0.350

INCHES

an error

can result

of the permeability

tangents

for

is more sensitive

the dielectric
of 2.0X

constant,

and

105 in p when

of 103.
to obtain

and error involved

FIELD

than

in a difference

It is possible
APPLIED

an estimate

in p by rewriting

of the sensitivity
(13)

( kOe)

Fig. 13. The resonant frequency of the TEoll resonance as a function

of applied field for three different
disks of Y.z.&do. 7A1o. lFe4. 6012.

of one percent
from
with

procedure

would

can be maintained

these results.
47riM>

2000

For

ferrites

G and

be necessary.

by plotting

of the spinel

6,=12.0,

It should

the

the

above

be pointed

out

(21)
where

Q~ is the unloaded
and

Note

to a good approximation,

factors

that

Qd the

Q of the TEO1l resonance

zero field

cancel

unloaded

netic loss measurement

ity error which applies

P will

proximately

cancels

out the metal

be possible

to obtain

an accuracy

the

two

values

in dielectric
of applied

loss tangents

field

was

be-

less than

3 X 10-e for samples with a diameter-to-length

ratio of
less than 2.0.
The larger dielectric
loss tangent for the samples with
a diameter-to-length
ratio of 4.0 is due to remaining
magnetic
losses, since these samples were not measured
in fields

of 40 kOe.

value of conductivity,
dielectric
loss tangent

Besides

not

knowing

the

correct

the chief source of error

in the
is the fact that a number of sam-

high

the metal
and hence,

in

fields.

wall

loss

the mag-

is not subject to the conductivto the dielectric

loss measure-

in Table

V, the difference

Q in very

out of this equation

that this extra effort is well justified

since it should lead
to an error of less than one percent, even above X-band
frequencies,
By designing
the resonant
system as an
integral
part of a permanent
magnet,
flux densities
of
the order of 20 kG could possibly
be achieved,
which
should provide
sufficient
accuracy
for most magnetic
materials.
For the dielectric
loss tangents
measurements,
the
error between
measurements
at 13.2 kOe and 40 kOe
is quite negligible.
For all the measurements
reported
tween

of

a field of 13.2 kOe and the asymp-

changes

p is of the order

m,~o

resonance

the

of Green

totic values is the same as for the dielectric

loss tangent,
i.e., the change in tan & is of the order of 3 X 106. How6.0 -

structure

to

for the dielectric

et cd. [11], this would probably

contribute
less than 0.5 percent to the value of p,.
The difference
in the magnetic
loss

0
g
K

racy

amounts

case. If the sample

the limits

field

tensor

u
;
g

1970

measurements.

of the permeability,

values

permeability.

>
g 80

AUGUST

constant,
then the error can be maintained
to within
one percent.
One error it is not possible to estimate
accurately
is that due to the assumption
of a scalar

Alo,4Fe460,2

TECHNIQUES,

dielectric

part

are held to within

,OJ

AND

ples are cut and there is a possibility

of an intrinsic
difference
across the material
used to fabricate
the
disks.
Except
for this difference,
similar
arguments
apply to the garnet dielectric
loss tangent errors as they

F&Id

15.50
15.51
15.80

0.350
0.256
0.175

Percent
Difference

ON MICROWAVE

ment.
For
p = 10-5, W,= 0.9, frequency=
10 GHz,
47rJl= 1800 gauss (i.e., for pure yttrium-iron-garnet),
the difference
in linewidths
to be measured is
f(Af)
Thus, ~ values much
measured.
For an error
be 2.7X

= 0.1 MHz.
less than 105 could easily be
of 3.o percent in p,, the error in

107. Since

the measurement

cent in p even when I..Lis less than

VIII.

wall

of ~ ap-

losses, it should

of less than

one per-

10s.

CONCLUSIONS

The results of this program

have shown that the real
parts of the complex permittivity
and permeability
of a
demagnetized
ferrite can be measured with an error of
less than one percent.
The dielectric
loss tangent
can
be measured

with

an error

of less than

5 X 10-5.

The

COURTNEY

: MEASURING

limitation

on

tangent
this
or

the

PERMITTIVITY

measurement

is calibrating

the

method

is so sensitive

conductivity,

the

being
tion.

developed
The

imaginary

of

metal

the

wall

low-loss

the

possibility

part

of the

PERMEABILITY

dielectric

losses,

to changes

very

offer

AND

loss

but

since

in dielectric

loss

dielectric

485

(25)

materials

of such

(26)

a calibrais not sub-

permeability

ject to the conductivity

error and it should be possible
to measure p values of less than 106 with an accuracy

and the energy

losses as
(27)

of one percent.
Some further
coupling

work

probes.

is required

For

on broad-banding

frequencies

up

to

the

8 GHz,

the

probe system shown in Fig. 2 is sufficient

but in the
X-band
region,
problems
arise due to high-frequency
bypass

capacitance

the sample

length

connectors

and

effects.

This

is less than
probe

is especially

0.2 inch.

cables

true

Using

overcomes

(28)

when

miniature

this

it can be shown

problem,

but the structure

is not so rugged as compared
with
using larger connectors
and semirigid
copper coaxial
inputs
(O. 141-inch
diameter)
unless
stainless
steel
cables

were

probes

of varying

to be used.

give matching
bypass

length

through

program,

effects

solution
holes

would

in the

ble and allows

range

become

of

of interest,

plates,

structure

a wide variation

to
when

troublesome.

be to introduce

bottom

the sides, but the probe

a number

were used and manipulated

in the frequency

capacitance

alternative

In this

the

rather

than

An

fields for the TEOII

have been given
resonator
[6]. They

Am

= ~=
fir

B.

tan

be

Q.,

(29)

H(a.),
)

where the terms Am, B~, H(wJ,

and Q~ are defined in
Section III.
If there are no magnetic
losses, then (29)
reduces

to the dielectric

Coleman

case considered

by Hakki

and

[6].

probes
ACKNOWLEDGMENT

from

dimensions.

A~~ETSDIX
The

I!

tan&

of Fig. 2 is more flexiin sample

that

mode in a dielectric
post
by Hakki
and Coleman

The

author

ing

the

the

manuscript,

wishes

problem

to thank

and,

along

and

Mrs.

with

measurements

at 40 kOe

ities

of the

Bitter

Massachusetts

Institute

were

National

for suggest-

J. A. Weiss,

J. Reid

The

Francis

D. H. Temme

for

the

made
Magnet

for

reading

computations.
using

the

facil-

Laboratory,

of Technology.

are, in the dielectric,

REFERENCES

Alkly

H,=j

Jl(klt-)

(22)

COS yZ

Techniques
of Tficrozoave Measurements.
[1] C. G. Montgomery,
New York: McGraw-Hill,
1951, ch, 10.
[2] W. B. Westphal, in Dielectric
Materials and Applications,
A. R,
Von Hipple, Ed.
New York: Wiley, 1954, pp. 63-122.
[3] V. P. Shestopalov and K. P. Yatsuk, Soviet Phys, Usp., vol. 4,
pp. 617-636,

JanuaryFebruary

1962.
Measurement
of small dielectric
losses in material
with a large dielectric
constant
at microwave
frequencies.
IRE
Trans.
Microwave
Theory
and Techniques,

[4] R. O. Bell and G. Rupprecht,

and in the air-space

E@ =

sin ~z

A2kjKl(k2r)

(23)

Hz=j

The

unloaded

A ,k,2

@#o

Ko(k2r)

Q of the resonator

sin ~z.

is given

by
(24)

where UC.v an d U.i.

in the rod and in

correspond
to the energy
the air-space,
respectively,

stored
and

WW.ll, and We.. are the energy losses in the conducting

walls and the dielectric
rod, respectively.
Writing

the stored

energy

as

vol; MTT-9, pp. 239242, May 1961,

[5] E. S. Sabisky and H. J. Gerritsen, ~. A@@l. P?zys., vol. 33,
p. 1450, 1961.
[6] B. W. Hakki and P. D. Coleman, A dielectric resonator method
of measuring inductive capacities in the millimeter range, IRE
Trans. h(icrowave
Theory and Techniques,
vol. M TT-8, pp. 402410, July 1960.
measurement of high[7] S. B. Cohn and K. C. Kelly, Microwave
Trans. Microwave Theory
dielectric-constant
materials, IEEE
and Techniques,
vol. MTT-14, pp. 406410, September 1966,
[8] C. P. Hartwig, private communication.
New York:
[9] C. C. Johnson, Field and Wave Electrodynamics.
McGraw-Hill,
1965, pp. 173-175.
Engineers
Handbook
and Buyers Guide.
Ded[10] The Microwave
ham, Mass.: Horizon House, 1966, p. 145,
[11] J.. J. Green, E. Schlomann, F. Sandy, and J. Saunders, Rome
Au- Develop. Ctr., Rome, N. Y., Tech. Rept. RADC-TR-6993, February 1969.
[12] C. P. Hartwig, private communication.
[13] H. E, Bussey, J. E, Gray, E, C. Bamberger, 13. Rushton, G.
Russell, B. W. Petley, and D. Morris, International
comparison
Trans. Instrumentation
and
of dielectric measurements, IEEE
Measurement, vol. IM-13, pm 305-311, December 1969.
[14] I. Bady, private communi~a~ion.
J. Afipl. Phys., VO1. 40, pp. 1420-1421T 1969.
[15] ,
[16] Complex dielectric constant of nonrnetalhc magnetic materials
at microwave frequencies, American Society for Testing Materials, Tech. Rept. ASTM-C525-63T,
1968.