IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 48, NO.
3, MARCH 2000
The Foster Reactance Theorem for Antennas and
Radiation Q
Wen Geyi, Perry Jarmuszewski, Member, IEEE, and Yihong Qi, Member, IEEE
Abstract—The calculation of antenna has been an interesting
and a controversial topic for years. In this paper, we first give a rig-
orous study of antenna by introducing a complete description of
the complex power balance relation for an antenna system. Using
the complex Poynting theorem, we have shown that the antenna is
essentially equivalent to a one port lossy network. The Foster re-
actance theorem is usually stated for a lossless network. The main
purpose of this paper is to determine whether the Foster reactance
theorem holds for antennas. By making use of a complex frequency
domain version of the Poynting theorem, we have shown that the
Foster reactance theorem is valid for an antenna. Finally, the Foster
reactance theorem for the antenna has been applied to demonstrate
the widely held assumption 1 , provided 1, where
stands for the fractional bandwidth of an arbitrary antenna.
Index Terms—Antenna input impedance, energy storage,
factor.
I. INTRODUCTION
T HE quality factor or is used to describe the high-input
reactance and narrow bandwidth of small antennas. The
of a system is usually defined as [1]
(1)
where is the time average energy stored in the system and
is the average dissipated power. In most applications of this
definition, the is evaluated at the resonant frequency. In this
case the can be expressed as
(2)
where is either the average stored electric or magnetic
energy. For a nonresonant antenna, it is tacitly assumed that
the antenna system should be tuned to resonance. Resonance is
achieved by adding a capacitive or inductive energy storage ele-
ment depending on whether the stored energy is predominantly
magnetic or electric. In this case, is chosen as average stored
magnetic energy or the average stored electric energy
in the near-field zone around the antenna, whichever is larger,
and is the total radiated power. The parameter defined by (2)
has been referred to as antenna .
The evaluation of antenna can be traced back to the clas-
sical work of Chu [2] who derived the theoretical value of for
an ideal antenna enclosed in an imaginary sphere. Chu’s anal-
ysis is based on the spherical TE and TM mode expansions pro-
Manuscript received January 4, 1999; revised September 23, 1999.
The authors are with Research In Motion, Waterloo, ON, N2L 3W8 Canada.
Publisher Item Identifier S 0018-926X(00)02447-9.
0018–926X/00$10.00 © 2000 IEEE
Authorized licensed use limited to: Princeton University. Downloaded on May 27, 2009 at 10:44 from IEEE Xplore. Restrictions apply.
401
duced by sources distributed on the surface of the imaginary
sphere. The major shortcomings of the Chu method are that it
is restricted to spherical modes and requires several approxima-
tions. Collin and Rothschild also presented a method for eval-
uating antenna . Their method is based on recognizing that
the total stored reactive energy can be calculated by subtracting
the radiated field energy away from the total energy in the field.
Such method has been successfully used by Fante and reexam-
ined by McLean to study antenna [5], [6].
Although the concept of antenna was proposed many years
ago, its definition, calculation, and relation to system bandwidth
have long been controversial topics [7]–[14]. Some authors even
questioned the applicability of the complex Poynting theorem in
the calculation of antenna [12]–[14].
In this paper, we try to give a more rigorous discussion of
antenna . In Section II, we present a complete description of
the complex power balance relation for electromagnetic fields,
hopefully removing the misunderstanding associated with the
complex Poynting theorem. To complete this paper we give
a derivation of the equivalent circuit of an antenna di-
rectly from the complex Poynting theorem. It is well known
that the Foster reactance theorem holds only for a lossless net-
work. From the point of view of circuit theory, an antenna is
a one-port network with loss. The loss represents the radiated
power from the antenna and this loss prohibits the direct use
of the Foster reactance theorem. A formula for the frequency
derivative of the input reactance of any linear, passive, time-in-
variant electromagnetic system has been derived by Rhodes [8],
[9]. The formula consists of five terms and has several frequency
derivatives of field quantities. So it is very difficult to figure
out how the frequency derivative of the input reactance changes
with the frequency. In Section IV, we make use of a complex
frequency domain approach to demonstrate that the Foster reac-
tance theorem (i.e., the frequency derivative of the input reac-
tance of the antenna is always greater than zero) still holds for
an arbitrary antenna system. Antenna engineers have used the
fact that the antenna fractional bandwidth is approximately
equal to the reciprocal of antenna for a high system, but
no explicit demonstration has been given. Wheeler has pointed
out that the common expression for bandwidth in terms of
is neither logical nor helpful in clear exposition [15]. It is be-
lieved that such a relationship must exist for any complete radi-
ating system, just as surely as it does for a nonradiating system.
Another common belief is that the relation cannot be proven.1
Rhodes has shown that the reciprocal relationship between
and does hold for a planar dipole [7]. Fante has tried to give
1D. R. Rhodes, “Author’s reply,” IEEE Trans. Antennas Propagat., vol. 15,
pp. 568–569, July 1967.
402 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 48, NO. 3, MARCH 2000

a study of antenna of general ideal antenna based on the fre-

quency derivatives of Maxwell’s equations, which has been the
standard way of deriving the Foster reactance theorem. How-
ever, such a study, if applied to an antenna system, may yield a
surface integral of the product between field and its frequency
derivative and require an assumption that either the terminal
voltage or the current is a constant [5]. An exact analysis for
such an integral is essentially difficult. In Section V, we apply
the Foster theorem for antenna system to give a simple demon-
stration of the widely held assumption .
Fig. 1. A volume V containing source region V .

II. POYNTING THEOREM AND EVALUATION OF USING FIELD

EQUATIONS As mentioned by McLean [6] and Grimes [13], some authors
implicitly define reactive power to be the imaginary part of the
The differential form of the complex Poynting theorem for
surface integral of the complex Poynting vector. This is a mis-
time harmonic field in an isotropic medium is given by
leading definition. Actually (10) shows that the surface integral
(3) of the real part of the Poynting vector is independent of the sur-
face as long as it encloses the source region . Equations
where is the complex Poynting vector (8) and (9) show that the surface integral of the imaginary part
and are the magnetic and electric of the Poynting vector depends on the integration surface in
field energy densities. the near field region (in the far field region it becomes zero).
Let be the volume occupied by the electric current source Using (4), (10), and (7) we get
and be the surface surrounding , take the integration
of the imaginary part of (3) over a volume containing , as
shown in Fig. 1, we obtain

(4)
(11)
where is the boundary of . Choosing , we have

The above relation indicates that the complex power flowing out
of is equal to the radiation power plus the reactive power out-
(5) side . This expression seems to be the most general form for
the Poynting theorem. Most of the writers directly explain the
and if we choose , where is the region enclosed by third term on the second line of (11) as the reactive energy stored
a sphere with radius , where is sufficiently large so that around the antenna. In Collin’s and Rhode’s work [3], [7], the
it lies in the far field region of the antenna system, we get second term of the right-hand side of (11) was missed. Collin
and Fante’s work [4], which is based on Collin’s method, have
been questioned by other authors [12]–[14]. These authors even
(6) proposed that the complex Poynting theorem is intrinsically an
insufficient basis for a full description of power in a radiated
Since is a real vector in the far-field region, we have field and resorted to the time dependent Poynting theorem. Such
a misunderstanding is actually caused by an improper explana-
(7) tion of the power balance relation (11).
To clarify this point, let ( ) and ( ) de-
From (4), (5), and (7), we obtain note the stored (radiated) electric field and magnetic field energy
densities, respectively, we can then define
(8)

(12)
(9)
These calculations are physically appropriate since density is a
summable quantity. It is readily seen from (9) that
Take the integration of the real part of (3) over the volume
in the far field zone since the complex Poynting vector becomes
containing the source region , we obtain the radiated power
real as approaches . This observation indicates that the
electromagnetic field energy and the magnetic field energy for
(10)
the radiated field are identical everywhere. To be explicit, we

Authorized licensed use limited to: Princeton University. Downloaded on May 27, 2009 at 10:44 from IEEE Xplore. Restrictions apply.
GEYI et al.: FOSTER REACTANCE THEOREM FOR ANTENNAS AND RADIATION 403

cite the far-field expression generated by an arbitrary current

distribution as follows:

(13)
(18)
where is an observation point in the far-field region and
is a unit vector along . From the above expression, we easily The antenna can be calculated as
obtain

(14) if
(19)
Hence, we have if

The above calculation has been used by McLean to calculate

(15)
directly from the fields of the TM spherical mode. This
and the total energy of the radiated fields is simply twice the approach is actually equivalent to Collin’s method.
electric or magnetic energy density of the radiated fields. Math-
ematically, (15) holds everywhere. Therefore we obtain III. THE ANTENNA AS A ONE-PORT DEVICE
In the following, we choose in such a way that is co-
incident with the antenna surface (except for a portion where
crosses the antenna feed plane), as shown in Fig. 2.
Taking the integration of (3) and using the divergence the-
orem over the region we get
(16)

from (8), (9), and (12). Here, and stand for the total
stored magnetic and electric energy in the volume surrounding
the radiator. From (12) the total stored reactive energy can be
expressed as (20)

where is the conductivity of the medium in . If we

assume the antenna surface is perfectly conducting,
vanishes everywhere on except over the input terminal .
For a single-mode transmission line we have

(21)

where and are equivalent voltage and current at the feeding

(17) plane respectively. Introducing (21) into (20) and using the fact
that , we find
where is the radius of the sphere , and is the wave ve-
locity. Both terms in the curved bracket are divergent as
, but it can be shown that the net term in the curved bracket is (22)
convergent [5]. The following has been used in deriving (17)
where

(23)
So the stored electric and magnetic field energy will be given by
is the power loss.
The antenna impedance is defined by

(24)

Authorized licensed use limited to: Princeton University. Downloaded on May 27, 2009 at 10:44 from IEEE Xplore. Restrictions apply.
404 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 48, NO. 3, MARCH 2000

Fig. 3. Impedance of an arbitrary antenna.

(27)

All the frequency-domain quantities can be recovered by letting

in (27). From (27) a relation similar to (3) can be obtained
in the region outside
Fig. 2. An arbitrary antenna and its equivalent circuit.

and by using (16), we can rewrite (22) as

(28)

If we take the integration of (28) over the connected region

(25) , as shown in Fig. 2, we find the relation

where
(29)

where

(26)
(30)
Hence, the equivalent circuit for the antenna is shown in Fig. 3.
The above quantities have no physical interpretations except
IV. FOSTER REACTANCE THEOREM FOR ANTENNAS AND when . Following a discussion similar to the previous
EVALUATION OF ANTENNA USING EQUIVALENT CIRCUITS section, we have
The Foster reactance theorem is a very important tool for
the synthesis of networks. In standard textbooks, the Foster re-
actance theorem is usually stated for a loss-free network. It
has been shown that an antenna system is equivalent to a one
port lossy network ( ). So the question may be raised
whether the Foster reactance theorem could apply for an antenna (31)
system. In the following section, we will attempt to answer this
question. First, we introduce a complex frequency for a single-mode transmission line. Define
and all calculations will be confined to the complex frequency and substitute it
plane. For clarity, all quantities in the complex frequency plane into (29) to get
will be embellished using this symbol to distinguish them from
the corresponding quantities in the frequency domain. Taking
the Laplace transform of the time-domain Maxwell’s equations (32)
in a lossless medium we have
If is sufficiently small so that we can make a first
order approximation and derive directly

Authorized licensed use limited to: Princeton University. Downloaded on May 27, 2009 at 10:44 from IEEE Xplore. Restrictions apply.
GEYI et al.: FOSTER REACTANCE THEOREM FOR ANTENNAS AND RADIATION 405

from the Maxwell equations, defined in the complex plane as and the definitions of and are similar to and , re-
shown in (33) at the bottom of the page. So we have spectively. Since is an analytic function its real and imag-
inary parts satisfy the Cauchy–Riemann conditions, i.e.,
(34)
(41a)
where , previously defined in (10), is radiated power in the
frequency domain which is independent of . Substituting (34)
into (32) we obtain
(41b)

By direct calculation we have

(35)

The impedance in the complex frequency plane can then be ex-

pressed as

(42)

where use of (17) has been made. From (41a) and (42) we obtain
(36)

If is sufficiently small, we can make the first order approx- (43)

imation which is independent of . So (36) can be
rewritten as
Now we decompose the complex impedance
of its real and imaginary parts, i.e.,
which is the Foster reactance theorem for an antenna system.
A formula for the frequency derivative the of input reactance
of any linear, passive, time-invariant electromagnetic system has
been derived by Rhodes [8], [9]. This formula contains four in-
tegrals of frequency derivatives of field quantities, making it dif-
(37)
ficult to obtain any useful information from it.
The above discussion can be easily generalized to an antenna
into the sum array system with input ports. In this case (35) should be
changed to

(38)

where

(39) (44)

where and are the voltage and current vectors at

the antenna array terminals. Introducing the impedance matrix
of the antenna array

(40) (45)

(33)

Authorized licensed use limited to: Princeton University. Downloaded on May 27, 2009 at 10:44 from IEEE Xplore. Restrictions apply.
406 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 48, NO. 3, MARCH 2000

and following a similar discussion we have where or whichever is larger. Substituting (51)
and (52) into (53) we have

(46)
(54)

where either the or value is chosen to give the higher .

(47) For a single antenna system (54) is simplified to

Instead of (42) we have

(55)

The term obtained from is usually larger than that

(48) obtained from for a high system [19]. If is large, it
is assumed that it is related to the frequency bandwidth of the
for a lossless system. So the Foster reactance theorem for the antenna as follows. Consider the antenna to be resonated with a
antenna array will take the form suitable reactive network at the frequency of interest . The fre-
quency at which the normalized input is equal to times
its value at resonance, is the half-power point. At this point the
(49) average power delivered from the antenna is one-half that de-
livered at resonance. Let be defined as the frequency incre-
ment, is the distance in frequency between the half-power
The above derivation is also applicable to a bounded one port points (3-dB points) of the antenna. The fractional bandwidth
lossless microwave system where [17]. The advantage is then defined as
of such an approach is that we do not need to assume that either
the terminal voltage or the current is a constant, as most of the (56)
textbooks do to get the same result as (43) or (49).
Now the antenna can be calculated from the corresponding
equivalent circuit using the Foster reactance theorem. For sim- If the is high the following relationship, which holds for a
plicity, we assume that the antenna system is lossless ( nonradiating system, is assumed
). The impedance of the antenna array for this case will be
(57)
(50)
This widely held assumption has never been explicitly proven
From (47) and (49) we obtain the stored magnetic and electric for distributed systems. As pointed by Wheeler [15], the
field energy as follows common expression for the bandwidth in terms of is
neither logical nor helpful in clear exposition. Most writers
state that if is large compared with unity, it is equal to
. Rhodes demonstrated that this relation does hold for the
planar dipole by computing and bandwidth independently
(51) [7] and he firmly believed that such a relation must exist for
any complete radiating system, just as surely as it does for
nonradiating systems.
The radiated power of the antenna array is
A discussion of the relation between and was given
by Fante. His approach was based on the frequency derivatives
of Maxwell’s equations, which has been the standard way of
deriving the Foster reactance theorem. For an antenna system,
however, a surface integral between the product between field
(52)
and its frequency derivative must appear in the derivation and
an assumption that either the terminal voltage and the current
So the antenna will be given by is a constant must be made. An exact analysis for such an inte-
gral is essentially difficult [4]. In the next section, we will give
a simple demonstration of this observation using the Foster re-
(53) actance theorem for an antenna system.

Authorized licensed use limited to: Princeton University. Downloaded on May 27, 2009 at 10:44 from IEEE Xplore. Restrictions apply.
GEYI et al.: FOSTER REACTANCE THEOREM FOR ANTENNAS AND RADIATION 407

V. ANTENNA AND ANTENNA BANDWIDTH Introducing the above into (63) and using (58), we find
In this section, we consider a high system. Let denote
one of the resonant frequencies of a single antenna system; then
by definition we have

(58)

We rewrite (40) at the resonant frequency as follows:

(65)

Here or , whichever is larger.

(59)
VI. CONCLUSION
Here we assume that is sufficiently small so we can make the
first-order approximation Although the concept of antenna was proposed many
years ago, its definition, calculation and relationship to system
bandwidth have long been controversial topics. A recent
discussion focuses on the applicability of the well-known
complex Poynting theorem to the calculation of antenna
[1], [12]–[14]. In this paper, we have shown that the complex
By (41b) we have Poynting theorem can be used in the calculation of antenna
and one does not need to revert to the time domain Poynting
theorem as suggested by references [12]–[14].
(60)
The central theme of this paper has been to show that the tra-
ditional Foster reactance theorem holds for an arbitrary antenna
Thus, as one moves off resonance, can be written as system using the Poynting theorem in the complex frequency
domain. The Foster reactance theorem has then been used to
demonstrate the widely held assumption is valid for
(61)
an arbitrary antenna system, provided .

The half-power points occur when REFERENCES

Q
[1] R. E. Collin, “Minimum of small antennas,” J. Electromagn. Waves
Applicat., vol. 12, pp. 1369–1393, 1998.
(62) Q
[2] L. J. Chu, “Physical limitations of antenna ,” J. Appl. Phys., vol. 19,
pp. 1163–1175, 1948.
[3] R. F. Harrington, “Effect of antenna size on gain, bandwidth, and effi-
ciency,” J. Res. Nat. Bureau Stand., vol. 64D, pp. 1–12, Jan. 1960.
so that the fractional bandwidth can be written Q
[4] R. E. Collin and S. Rothschild, “Evaluation of antenna ,” IEEE Trans.
Antennas Propagat., vol. AP-12, pp. 23–27, Jan. 1964.
[5] R. L. Fante, “Quality factor of general idea antennas,” IEEE Trans. An-
(63) tennas Propagat., vol. AP-17, pp. 151–155, Mar. 1969.
[6] J. McLean, “A re-examination of the fundamental limits on the radiation
Qof electrically small antennas,” IEEE Trans. Antennas Propagat., vol.
44, pp. 672–676, May 1996.
[7] D. R. Rhodes, “On the stored energy of planar apertures,” IEEE Trans.
So the problem is to evaluate . To do this, Fante em- Antennas Propagat., vol. AP-14, pp. 676–683, Nov. 1966.
ployed the frequency derivative of Maxwell’s equations and ex- [8] , “Observable stored energies of electromagnetic systems,” J.
tended the treatment of Harrington’s work [19, pp. 394–396]. Franklin Inst., vol. 302, no. 3, pp. 225–237, 1976.
[9] , “A reactance theorem,” in Proc. Roy. Soc. London, vol. 353, 1977,
As a result, he obtained an integral defined on an infinitely large pp. 1–10.
surface for which an asymptotic analysis is very difficult. In ad- [10] G. V. Borgiotti, “On the reactive energy of an aperture,” IEEE Trans.
dition, he had to assume that either the terminal voltage or the Antennas Propagat., vol. AP-15, pp. 565–566, July 1967.
current is a constant. In fact, this can be easily done using the [11] R. E. Collin, “Stored energy Q
and frequency sensitivity of planar
aperture antennas,” IEEE Trans. Antennas Propagat., vol. AP-15, pp.
Foster reactance theorem. From (51) we get 567–568, July 1967.
[12] C. A. Grimes and D. M. Grimes, “The Poynting theorem and the poten-
tial for small antennas,” Proc. IEEE Aerosp. Conf., vol. 3, pp. 161–176,
1997.
[13] D. M. Grimes and C. A. Grimes, “Power in modal radiation fields: Limits
of the complex Poynting theorem and the potential for electrically small
(64) antennas,” J. Electromagn. Waves Applicat., vol. 11, pp. 1721–1747,
1997.

Authorized licensed use limited to: Princeton University. Downloaded on May 27, 2009 at 10:44 from IEEE Xplore. Restrictions apply.
408
[14] Q
, “Radiation of dipole-generated fields,” Radio Sci., vol. 34, no.
2, pp. 281–296, 1999.
[15] H. A. Wheeler, “Small antennas,” IEEE Trans. Antennas Propagat., vol.
AP-23, pp. 462–469, July 1975.
[16] E. A. Guillemin, Introductory Circuit Theory. New York: Wiley, 1953.
[17] R. B. Adler, L. J. Chu, and R. M. Fano, Electromagnetic Energy and
Transmission and Radiation. New York: Wiley, 1960.
[18] R. F. Harrington, Time Harmonic Electromagnetic Fields. New York:
McGraw-Hill, 1961.
[19] , Field Computation by Moment Methods. Piscataway, NJ: IEEE
Press, 1993.
Wen Geyi was born in Hunan, China, in December
1962. He received the B.Eng., M.Eng., and Ph.D. de-
grees in electrical engineering from Xidian Univer-
sity (formerly Northwest Telecommunications Insti-
tute), Xian, China, in 1982, 1984, and 1987, respec-
tively.
From 1988 to 1990, he was a Lecturer in the Radio
Engineering Department, Southeast University, Nan-
jing, China; from 1990 to 1992 he was an Associate
Professor in the Institute of Applied Physics, Univer-
sity of Electronic Science and Technology of China,
Chengdu; and from 1992 to 1993 he was a Visiting Researcher in the Department
of Electrical and Computer Engineering, University of California at Berkeley.
Since 1992 he has been a Full Professor in the Institute of Applied Physics,
University of Electronic Science and Technology of China. He was a Visiting
Professor in the Electrical Engineering Department of University of Waterloo
from February 1998 to May 1998. From 1996 to 1997 he was the Vice Chairman
of the Institute of Applied Physics, University of Electronics of China and has
been the Chairman of the Institute since 1997. He is a member of the edito-
rial board of the Journal of Electromagnetic Wave Propagation (China) and a
member of the editorial board of Science and Engineering Computation Series.
He has authored two books and has more than 100 publications. His current
research interests include applied mathematics and physics, microwave theory
and techniques, and antennas and propagation.
Dr. Wen is a member of editorial board of IEEE TRANSACTIONS ON
MICROWAVE THEORY AND TECHNIQUES. He is a council member of China
Society of Computational Physics.
Authorized licensed use limited to: Princeton University. Downloaded on May 27, 2009 at 10:44 from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 48, NO. 3, MARCH 2000
Perry Jarmuszewski (S’91–M’92) received the
B.Sc. degree in electrical engineering from Queen’s
University, Kingston, Canada, in 1992, and the
M.Eng. degree from McMaster University, Canada,
in 1997.
In 1994, he joined Research in Motion (RIM), Wa-
terloo, ON, Canada, as an RF Design Engineer and in
1996 he became Director of RF Engineering at RIM.
He is an inventor in three patents. His current inter-
ests include low-power transceiver circuits for mobile
data communications and applied electromagnetics.
Yihong Qi (M’92) received the B.Eng. degree in
electronics engineering from the National Univer-
sity of Defense Science & Technology (formally
Changsha Institute of Technology), China, in 1982,
the M.Eng. degree in Space Radio Technology
from the Chinese Academy of Space Technology,
China, in 1985, and the Ph.D. degree in electronics
engineering from Xidian University, China, in 1989.
From 1985 to 1986, he worked as an Assistant En-
gineer at Xian Institute of Space Radio Technology,
China. From 1989 to 1992 he was a Postdoctoral
Fellow at Southeast University, Nanjing, China. He worked as an Associate
Professor at Southeast University from 1992 to 1993. From 1993 to 1995
he was a Postdoctoral Fellow at McMaster University, Canada. He has been
working for Research in Motion (RIM) as an Electromagnetic Researcher
since 1995. He has published more than 50 papers in technical journals
and conferences in the area of antennas and wireless communications and
coauthored two books. He was an inventor in three patents.
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 55, NO. 3, MARCH 2007 1013

Comments and Replies

Comments on ‘The Foster Reactance Theorem for
Antennas and Radiation ’
J. Bach Andersen and S. Berntsen

In [1], the authors present a proof that Foster’s theorem [2] is valid
in general for antennas, which is a surprising result. Foster’s theorem,
as it is generally understood, expresses that the gradient of reactance
and susceptance with frequency is always positive for a lossless net-
work. That this should be valid for any antenna with radiation losses
is intuitively wrong and goes against numerous experimental and nu-
merical results, as well as other theoretical derivations [3], [4]. The
purpose of this comment is to show that the proof in [1] is erroneous.
First a brief look at Foster’s theorem [5], [6], for which it is often for-
gotten that it must be valid for both reactance and susceptance. In the
original Foster’s theorem [2] there is no mentioning of reactance gra-
dients, except that this part was proven and published earlier by Zobel
[7], so “Zobel’s reactance theorem” would be a more appropriate name.
Foster [2] showed that for a lossless circuit there is an alternating series Fig. 1. Reactance and susceptance versus frequency for a series resonance for
of poles and zeros on the real frequency axis, and a zero for reactance lossless and lossy case.
corresponds to a pole for susceptance. Foster’s theorem says that the
gradient of both quantities versus frequency always is positive. This is
illustrated in Fig. 1 which shows the case of a pole at zero frequency and In [1, Eq. 37], (slightly modified to make the point clear) the
a zero at ! = 1 for the reactance, and vice versa for the susceptance, a impedance of a general antenna is expressed as follows:
series connection of a unit capacitor, a unit inductor, and a resistor

ZA (s) = RA (s) + jXA (s)

2P 4[Wm + We ] 4j! [Wm 0 We ]
1 rad
Z = r + j! + : (1)
j! =
j I j2 + jI j 2 +
jI j 2 (3)

in the complex frequency plane s = a + j! . Wm and We are the

For the resistance equal to zero there are positive gradients every-
stored magnetic and electric energies in a volume around the antenna,
where for both reactance and susceptance. For a series resistance of
extended into the complex plane, and similarly for the radiated power
r = 0:2, the reactance is of course unaffected, but a dramatic shift
in susceptance which is now zero at the resonance and a negative gra-
P rad , all being real functions of s. For  = 0, the expression reduces
to the standard expression for the impedance. In [1], it is assumed that
dient. It is easy to show that at resonance the slopes of reactance and
susceptance have different signs. Thus, it is already clear at this stage
ZA is an analytic function of s and, therefore, Cauchy-Riemann equa-
tions may be used by finding the derivative for  = 0 and to derive
that Foster’s theorem in general cannot be valid for a circuit with loss,
the following inequality [1, Eq. 43] under the assumption that P rad is
which is what an antenna is. Adler et al. [8] do show that a function
independent of 
like (1) is analytic in s in the complex frequency plane

@RA @XA 4(Wm + We )

>0
Z (s) = r + s + 1 (2) @
=
@!
=
I2jj (4)
s

and apply Cauchy-Riemann to show that the reactance gradient is pos-
itive, but this is too simple a model to cover all antenna cases. Note that
Z is a function of s only and that r must be a real constant. Y (s) =
1=Z (s) is of course also analytic, but as indicated above with a nega-
tive gradient at resonance.
Manuscript received March 6, 2006; revised May 19, 2006.
The authors are with Aalborg University, DK 9100 Aalborg, Denmark
(e-mail: jba@cpk.auc.dk).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TAP.2007.891877
which is Foster’s theorem for a lossless network, in [1] assumed valid
for antennas as well. The problem with the proof is that ZA , as given
in (3) is not an analytic function of s and, therefore, Cauchy-Riemann
equations may not be applied. All the powers and energies involve com-
plex conjugations, so they are not analytic functions of complex fre-
quency, except if assumed to be real constants, which may not be true.
Even if the powers and energies were assumed constants, a problem
remains, since  and j! in (3) are not analytic functions of s.
For further illustration of the gradients of reactance the impedance
of a thick dipole has been computed by the well-established finite-dif-
ference time-domain (FDTD) method, and the result is shown in Fig. 2.
Both the positive and negative gradients are clearly seen.

0018-926X/$25.00 © 2007 IEEE

1014 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 55, NO. 3, MARCH 2007
Fig. 2. FDTD solution of a thick dipole fed by a transmission line. The refer-
ence plane is removed from the antenna.
REFERENCES
[1] W. Geyi, P. Jarmuszewski, and Y. Qi, “The Foster reactance theorem
Q
for antennas and radiation ,” IEEE Trans. Antennas Propag., vol. 48,
no. 3, pp. 401–408, Mar. 2000.
[2] R. M. Foster, “A reactance theorem,” Bell Syst. Tech. J., vol. 3, pp.
259–267, 1924.
Q
[3] A. D. Yaghjian and S. R. Best, “Impedance, bandwidth, and of an-
tennas,” IEEE Trans. Antennas Propag., vol. 53, no. 4, pp. 1298–1324,
Apr. 2005.
[4] L. Li, H.-X. Liu, Y. Shi, and C.-H. Liang, “Study of generalized res-
onance in multi-antenna system and generalized Foster reactance the-
orem,” Progress Electromagn. Res., vol. PIER 52, pp. 255–276, 2005.
[5] C. G. Montgomery, R. H. Dicke, and E. M. Purcell, Principles of Mi-
crowave Circuits. New York: McGraw-Hill, 1948, sec. 5.24.
[6] R. E. Collin, Foundations for Microwave Engineering. New York:
McGraw-Hill, 1968, pp. 153–155.
[7] O. J. Zobel, Bell Syst. Tech. J., pp. 5–9, Jan. 1923.
[8] R. B. Adler, L. J. Chu, and R. M. Fano, Electromagnetic Energy Trans-
mission and Radiation. New York: Wiley, 1965.
0018-926X/$25.00 © 2007 IEEE
Reply to “Comments on ‘The Foster Reactance Theorem
for Antennas and Radiation ’”
Wen Geyi
The comments proposed by the authors of [1] will be shown to be
superfluous.
The Basic Facts About the Foster Reactance (Susceptance) The-
orem: The first question of [1] is essentially related to how the Foster’s
theorem should be presented. The Foster theorem is usually summa-
rized in two different parts, one is Foster reactance theorem and another
is Foster susceptance theorem. For a lossless network, both Foster re-
actance theorem and Foster susceptance theorem hold, i.e., the slope
of reactance and susceptance are positive simultaneously. Note that the
lossless condition is just a sufficient condition for Foster reactance (sus-
ceptance) theorem to hold but not a necessary condition. The Foster
reactance (susceptance) theorem can be generalized to a lossy network
in numerous situations. For example the Foster reactance theorem does
hold for a simple series RLC circuit or any lossy network that consists
of a resistor serially connected to any lossless network. A lossy net-
work that satisfies the Foster reactance theorem may not satisfy the
Foster susceptance theorem and vice versa. The reactance curves and
susceptance curves shown in [1, Fig. 1] are just a demonstration of the
above facts. The present author fails to see any points of this demon-
stration. Since the element values of the equivalent RLC circuit of an
antenna depend on frequency [2]–[4], the above demonstration, which
is based on a RLC circuit with constant element values, does not apply
to the antenna at all. For example both the reactance curve and suscep-
tance curve of a rectangular aperture antenna have a positive slope [5],
[6] and the above demonstration cannot not explain this phenomena. It
must also be mentioned that Adler et al. [2] have never used a simple
RLC circuit with constant element values to study a lossy electromag-
netic network as [1] has claimed. Instead all the element values of the
equivalent RLC circuit have been assumed to be functions of frequency
[2, Eq. 6.88].
It has been shown in [4] that the slope of the reactance of an ideal
antenna (without ohmic loss) excited by a narrow-band signal (so that
only the dominant mode is propagating in the feeding line) is given by
@X 4 W~ m + W~ e
@!
= I2
j j
: (1)
Since the electromagnetic stored energy is always positive, the above
equation indicates that the slope of the reactance of an ideal antenna
is always greater than zero if the terminal current is not zero. When-
ever the terminal current approaches to zero, the slope of the reactance
@X=@! approaches to infinity, which usually happens at the frequency
points where the reactance curve abruptly changes from positive in-
finity to negative infinity. Note that an ideal antenna is still equivalent
to a lossy network, where the loss stands for the radiated energy into
free space. This equivalent loss is physically different from the usual
ohmic loss where the energy is dissipated into heat, which is why the
Foster reactance theorem still holds for an ideal antenna.
On the Analyticity of the Antenna Input Impedance: The second
question of [1] is that the input impedance of antenna is not analytic
since all the stored energies involve complex conjugations. It is a
common knowledge that the input impedance of an arbitrary lossy
electromagnetic network is analytic on the open right half of complex
Manuscript received July 11, 2006; revised November 13, 2006.
The author is with Research In Motion, Waterloo, ON N2L 5Z5, Canada
(e-mail: gwen@rim.com).
Digital Object Identifier 10.1109/TAP.2007.891878
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 55, NO. 3, MARCH 2007 1015

Fig. 1. Reactance of a dipole (L = 125 mm, D = 75 mm, d = 5 mm, Fig. 2. Reactance of the coaxial aperture antenna with infinite flange.
a = 0:5 mm).

frequency plane to insure that the network is physically realizable [2]. From the above equation, Li et al. have derived that [8] (for a single
The antenna is equivalent to a lossy network and its input impedance antenna system)
must also be analytic on the open right half of complex frequency
domain. To prove this, let us rewrite [1, Eq. 3] as
@X W
~m + W ~e
S 1 n @@! ds 0 2R @I :
4 4 E
= +
[2P
rad
+ 4sW ~ m + 4s3 W
~ e] @! jI j 2 jI j 2
@!
(5)
ZA (s) = : (2) S
jI j
2

To get rid of the conjugation s3 , one may introduce a new quantity

In deriving this formula, the far field pattern has been assumed to have
one component only so that one can write E1 = uE jE1 j ej
~ 0 (s ) = s 2 W
W e
~ e (s) used in [2, Eq. 6.87]. This relation has been
(this was mistakenly written as E1 = jE1 j ej in [8]) with
validated in [7, Eq. 15]. Now (2) can be extended to an analytic function uE being a unit vector. For this reason, (5) holds only for a specific
of s by replacing all complex conjugations s3 in W ~ m (s ) , W~ 0 (s) with
e class of antennas. The last two terms in (5) cannot be explained
0s, and j! (resp. 0j!) by s (resp. 0s) in P and jI j . Thus, (2)
rad 2
physically. Indeed one can easily show that the last two terms on
becomes (still use same notations) the right-hand side of (5) actually cancel each other. For example
the far field patterns for the dipole and loop antenna are given by
2P
rad
(s) + 4sW
~ m (s ) +
~ (s)
4W
s
E = jI [cos(201 kl cos ) 0 cos(201 kl)]=2 sin  [11, Eq. 4-62a]
Z A (s ) = : and E = akIJ1 (ka sin )=2 [11, Eqs. 5-54b], respectively, and
[I ( s ) I ( 0s)] (3)
all these patterns only have a single component, and, thus, (5) holds.
Note that (2) and (3) are identical when  = 0. The antenna impedance When the above expressions are substituted into (5) and the rela-
given by (3) is now an analytic function and thus the Cauchy-Riemann tionship R = (2= jI j2 ) S S 1 nds is used, one finds that the last
conditions apply. If  is assumed to be a very small number, one may two terms in (5) cancel each other. Therefore, [8] provides another
have the Taylor expansion E(s) 1 E(0s) = jE(j! )j2 + jQ(! ) + independent proof that Foster reactance theorem holds for an arbitrary
o(), where Q(!) is a real function of ! . Similar expansions hold ideal antenna. A number of numerical calculations have shown that
for H(s) 1 H(0s) and I (s)I (0s). Considering these expansions and the Foster reactance theorem also holds for those antennas whose far
neglecting the higher order terms than  in (3) lead to RA (; ! ) = field patterns have more than one component [6].
2P
rad
(! ) + 4  [ W ~ e (! )] = jI (! )j2 , from which and the
~ m (! ) + W It should be mentioned that all field quantities in (4) and (5) must be
understood as the Fourier transform of the corresponding field quanti-
Cauchy-Riemann condition at  = 0, one immediately gets the Foster ties in the time domain since the frequency derivatives must be defined
reactance theorem for an ideal antenna [4, Eq. 43]. in a frequency continuum. Especially, the terminal voltage V and cur-
Comments on the Latest Work About the Foster Reactance The- rent I are the Fourier transform of its time-domain counterpart, and
orem for Antennas: The third question of [1] is that some recent pub- they depend on frequency [12]. This fact has been ignored in [9].
lications have shown that a dispersive term involving the gradients of Why are Some Numerical and Experimental Results Inconsistent
far-fields should be added in (1) as shown in [8] and [9]. To make things With Foster Reactance Theorem?: The fourth question of [1] is that the
clear let us cite a well-known formula from Fante’s paper [10, Eq. 16] Foster reactance theorem for antenna is against some experiments and
as follows: numerical results, and the FDTD simulation for a thick dipole shows
I 3 @V +V
3 @I = 1 @ E1 2 H3 0 @ H1 2 E3 1 nds both positive and negative slope [1, Fig. 2]. To answer this question, one
@! @! r2 @! 1 @! 1 must understand that some antenna impedance may have singularities
S
depending on the position of the reference plane, where the reactance
~m + W
+j 4 W ~e : (4) curve abruptly changes from positive infinity to negative infinity so that
1016 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 55, NO. 3, MARCH 2007
the Foster reactance theorem holds. It should be noted that all numer-
ical methods cannot accurately model these singularities due to numer-
ical errors, and thus a negative (but very steep) slope may occur. The
FDTD results given in [1] are from an inaccurate simulation. In fact the
same thick dipole has been recalculated using the moment method and
it turns out that the Foster reactance theorem holds very well, as shown
in Fig. 1. In this example, the reference plane must be far away enough
to guarantee a single mode assumption at the reference plane since the
parallel feeding lines are exposed to the dipole in this situation. Fig. 2
shows another example which is based on an analytical solution [13].
In this case the position of reference plane is not critical (i.e., the Foster
reactance theorem holds even if the reference plane is right at the an-
tenna aperture) since the influence of the flange on the field distribution
inside the coaxial cable is negligible. For the same reason all antenna
measurements cannot accurately handle singularities and in addition
the ohmic loss will be introduced in the measurements. Therefore, the
negative slope may appear in the measured reactance curves, and the
Foster reactance theorem only holds approximately.
It must be emphasized that the Foster reactance theorem only holds
for an ideal antenna with the feeding line connected. The modeling of
wire antennas is usually based on a number of approximations. The
most dramatic approximation is that the feeding line has been replaced
by a delta gap. This kind of approximation is questionable and cannot
be checked experimentally as discussed in [6] since all practical an-
tennas involves a feeding line, and the feeding line itself contributes
significantly to the value of antenna impedance. When the delta gap is
used for calculating the impedance of thick wires, one should limit the
calculation to the low frequency range since the delta gap is only valid
for thin wire.
It should also be mentioned that the Foster reactance theorem holds
only in the frequency range between the cutoff frequency of the dom-
inant mode and the cutoff frequency of the first higher order mode of
the feeding line as a single-mode assumption has been used at the an-
tenna terminal in our proof [4, Eq. 21]. When the operating frequency
is higher than the cutoff frequency of the first higher order mode, the
feeding line is equivalent to a multiple-transmission line system and the
antenna becomes a multi-port device. In this case, the Foster reactance
theorem fails. Some discussions about the single-mode excitation can
be found in [14].
In summary, the comments by authors of [1] are based on several
misunderstandings about the antenna input impedance. Especially they
fail to note that all circuit parameters for a microwave network depend
on frequency [13]. The Foster reactance theorem for ideal antennas
has been extensively checked by numerical simulations, and so far no
counter example has been found.
REFERENCES
[1] J. B. Andersen and S. Berntsen, “Comments on “The Foster Reac-
Q
tance Theorem for Antennas and Radiation ,” IEEE Trans. Antennas
Propag., vol. 55, no. 3, pt. II, pp. 1013–1014, Mar. 2007.
[2] R. B. Adler, L. J. Chu, and R. M. Fano, Electromagnetic Energy and
Transmission and Radiation. New York: Wiley, 1960.
[3] W. Geyi, “Calculation of element values of antenna equivalent circuit
(Invited),” in Proc. ISAP, Seoul, Korea, 2005, pp. 1029–1032.
[4] W. Geyi, P. Jarmuszewski, and Y. Qi, “The Foster reactance theorems
Q
for antennas and radiation ,” IEEE Trans. Antennas Propag., vol. 48,
pp. 401–408, 2000.
[5] I. C. Jan, R. F. Harrington, and J. R. Mautz, “Aperture admittance of a
rectangular aperture and its use,” IEEE Trans. Antennas Propag., vol.
39, pp. 423–425, 1991.
[6] W. Geyi, “New magnetic field integral equation for antenna system,”
Progress Electromagn. Res., vol. 63, pp. 153–170, 2006.
Q
[7] ——, “A method for the evaluation of small antenna ,” IEEE Trans.
Antennas Propag., vol. 51, pp. 2124–2129, Aug. 2003.
[8] L. Li, H. X. Liu, Y. Shi, and C. H. Liang, “Study of generalized res-
onance in multi-antenna system and generalized Foster reactance the-
orem,” Progress Electromagn. Res., vol. 52, pp. 255–276, 2005.
0018-926X/$25.00 © 2007 IEEE
Q
[9] A. D. Yaghjian and S. R. Best, “Impedance, bandwidth and of an-
tennas,” IEEE Trans. Antennas Propag., vol. 53, pp. 1298–1324, 2005.
[10] R. L. Fante, “Quality factor of general idea antennas,” IEEE Trans.
Antennas Propag., vol. AP-17, pp. 151–155, 1969.
[11] C. A. Balanis, Antenna Theory: Analysis and Design, 2nd ed. New
York: Wiley, 1997.
[12] W. Geyi, “A time-domain theory of waveguide,” Progress Electro-
magn. Res., vol. 59, pp. 267–297, 2006.
[13] N. Marcuvitz, Waveguide Handbook. London, U.K.: Peter Pere-
grinus, 1993.
[14] G. Goubau, Electromagnetic Waveguides and Cavities. London,
U.K.: Pergamon Press, 1961.
Comments on “Waveguide Miniaturization Using Uniaxial
Negative Permeability Metamaterial”
Fan-Yi Meng, Qun Wu, Bo-Shi Jin, Hai-Long Wang, and Jian Wu
Abstract—It is shown that in the paper by S. Hrabar, J. Bartolic, and Z.
Sipus, the negative permeability of the modified split ring resonators, which
are used to obtain the miniaturized waveguide, is not excited, and there is
no negative permeability metamaterial.
Index Terms—Backward-wave, metamaterials, negative permeability,
uniaxial anisotropy, waveguide.
In the paper cited [1], S. Hrabar, J. Bartolic, and Z. Sipus stated that
the rectangular waveguide can be miniaturized utilizing the modified
split ring resonators (MSRRs) based on the left-handed metamaterial
theory. They measured the transmission property of the electromag-
netic wave in the miniaturized waveguide filled with MSRRs, as shown
in [1, Fig. 8], and found out that there is a passband below the cutoff
frequency of the TE10 mode. They claimed that the reason for this phe-
nomenon, is that the TE10 mode can propagate in a form of backward
wave in such waveguide below its cutoff frequency, because the un-
axial negative permeability property of the MSRRs will be excited and
the left-handed metamaterial will function in this case. In this Com-
ment we show that, on the contrary, unaxial negative permeability of
MSRRs is not excited and there is no left-handed metamaterial at all.
We have numerically simulated the measurement system for the
miniaturized waveguide in [1], creating its simulation model with CST
MW Studio simulation tools. The simulation model consists of the
miniaturized waveguide loaded with MSRRs and two feeding J-band
waveguide, which are interfaced with the miniaturized waveguide.
Two waveguide ports are set on the J-band waveguides. Moreover, the
fixed impedance of 50 Ohm is used as the normalized impedance in
Manuscript received March 22, 2006; revised June 26, 2006. This work was
supported in part by the Natural Science Foundation of China under Grant
60571026, and in part by the National Key Laboratory of Electromagnetic
Environment under Grant 514860303.
F.-Y. Meng, Q. Wu, B.-S. Jin, and H.-L. Wang are with the Department
of Electrical and Communication Engineering, Harbin Institute of Technology,
Harbin, Heilongjiang 150001, China (e-mail: blade@hit.edu.cn).
J. Wu is with the National Key Laboratory of Electromagnetic Environment,
Beijing 102206, China.
Digital Object Identifier 10.1109/TAP.2007.891880