JOURNAL OF LATEX CLASS FILES, VOL. XX, NO.
XX, NOVEMBER 2017 1
Communication
The Radiation Efficiency Cost of Resonance Tuning
Lukas Jelinek, Kurt Schab, Member, IEEE, and Miloslav Capek, Member, IEEE
AbstractExisting optimization methods are used to calculate
the upper-bounds on radiation efficiency with and without the
arXiv:1712.02613v1 [physics.class-ph] 15 Nov 2017
constraint of self-resonance. These bounds are used for the
design and assessment of small electric-dipole-type antennas. We
demonstrate that the assumption of lossless, lumped, external
tuning skews the true nature of radiation efficiency bounds when
practical material characteristics are used in the tuning network.
A major result is that, when realistic (e.g., finite conductivity)
materials are used, small antenna systems exhibit dissipation
factors which scale as (ka)4 , rather than (ka)2 as previously
predicted under the assumption of lossless external tuning.
Index TermsAntenna theory, optimization methods, power
dissipation, energy efficiency.
I. I NTRODUCTION
R ADIATION efficiency is a parameter of paramount im-
portance for electrically small radiators since it signifi-
cantly limits their performance [1]. Techniques for radiation
efficiency maximization [2], [3] or attempts to set physical
bounds [4][9] on this parameter therefore naturally accom-
pany developments in antenna technology.
In most cases, it is desirable to design a small antenna to
be tuned (i.e., resonant) at a specified frequency. This can be
accomplished either by designing the antenna itself to be self-
resonant or through the use of an external tuning network. Al-
though the high cost of resonance tuning in radiation efficiency
was recognized long ago [3], many recent works assume that
tuning can be done in lossless manner [4][7]. In fact, careful
review reveals that this is a common assumption in many
standard textbooks [1], [10], [11] as well. This assumption,
however, leads to physical bounds which are unachievable by
realistically tuned antenna designs [12][14].
Recently, the effect of resonance tuning has once more
been taken into account by two different paradigms. The
first approach [15], [16] used full-wave treatment of optimal
currents on arbitrarily shaped lossy surfaces. The second em-
Manuscript received November XX , 2017; revised November XX, 2017.
This work was supported by the Czech Science Foundation under Project
15-10280Y. Kurt Schab was supported by a grant from the Intelligence
Community Postdoctoral Research Fellowship Program. All statements of fact,
opinion, or analysis expressed are those of the author and do not reflect the
official positions or views of the Intelligence Community or any other U.S.
Government agency. Nothing in the contents should be construed as asserting
or implying U.S. Government authentication of information or Intelligence
Community endorsement of the authors views.
L. Jelinek and M. Capek are with the Department of Electromagnetic Field,
Faculty of Electrical Engineering, Czech Technical University in Prague, Tech-
nicka 2, 16627, Prague, Czech Republic (e-mail: lukas.jelinek@fel.cvut.cz,
miloslav.capek@fel.cvut.cz).
K. Schab is with the Department of Electrical and Computer Engineering,
Antennas and Electromagnetics Laboratory, North Carolina State University,
Raleigh, NC, USA (e-mail: krschab@ncsu.edu).
ployed spherical wave expansion [8] reaching analytic resonant
bounds for spherical geometries.
The purpose of this communication is to show that effect
of resonant tuning on radiation efficiency of electrically small
antennas can precisely be evaluated for arbitrary surface
current supports. Furthermore, we demonstrate that for electri-
cally small antennas resonance tuning using realistically lossy
materials leads to an unpleasant quartic frequency scaling
of dissipation factor. This opposes the optimistic quadratic
frequency scaling predicted by the bounds derived for systems
externally tuned by lossless lumped circuits.
This paper is organized as follows. Sections II and III
introduce the necessary mathematical tools and restricting
assumptions. Section IV compares self-resonant and non-
resonant radiation efficiency bounds. The bounds are com-
pared to realistic designs in Section V.
II. P HYSICAL A SSUMPTIONS
Throughout the paper several restrictive assumptions are
made in order to obtain mathematically tractable problem:
Time-harmonic steady state is assumed within convention
F (t) = Re {F () exp (jt)} with angular frequency .
An antenna and potential tuning element are made solely
of a resistive sheet of given surface resistance Rs . No
other material bodies are allowed.
Throughout the paper, the surface resistance Rs = 0.01 ,
which approximatively corresponds to a thick copper
cladding at frequency f = 1.4 GHz, is used. In cases
presented in this paper, all results scale with the value
of Rs to a high degree of precision.
Apart from the above mentioned simplifications, the radiator
is assumed to be tuned to resonance at a given frequency.
III. M ATHEMATICAL T OOLS
The assumption of an antenna as an arbitrary surface S
in R3 allows us to employ the electric field integral equa-
tion [17]
n (E s (J ) + E i ) = Rs J on S, (1)
where E s is the scattered electric field, E i the incident
electric field, J a surface current density, Rs the surface
resistivity, and n a unitary normal to the surface S. The surface
is further discretized into triangles and Rao-Wilton-Glisson
(RWG) basis functions { n (r)} are used to represent surface
current density as [18]
X
J (r) In n (r) (2)
n
JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. XX, NOVEMBER 2017
and to recast (1) into its matrix form [17]
(Z + Rs ) I = V, (3)
where I is a vector of expansion coefficients, V the excitation
vector [17], Z = R + jX the impedance matrix [17] and
the basis-function overlap matrix [15].
With the help of aforementioned matrix formulation the
complex power [19] can be written as
1 H
Prad + jPreact I (R + jX) I, (4)
2 `/2
and the cycle mean power lost as heat [19] can be written as
Rs H
Ploss I I. (5)
2
The radiation efficiency can be written as
1 region. Solid lines correspond to the external tuning by a lumped lossless
= , (6)
1+
with = Ploss /Prad being the dissipation factor [20].
IV. O PTIMAL C URRENT M AXIMIZING R ADIATION
E FFICIENCY
The classical procedure to find a current distribution that
maximizes radiation efficiency is to solve [20], [21]
Rs In = n RIn (7)
and take the current corresponding to its lowest eigenvalue.
The resulting current distribution minimizes dissipation fac-
tor and thus maximizes radiation efficiency .
The solution to (7) is not necessarily self-resonant. If
resonance is required, the dissipation factor obtained in (7)
can only be achieved if there exists a lossless lumped element
that can tune the current to resonance without affecting dissi-
pation [1], [4][7], [10], [11]. On planar regions this method
generates constant current density which is a trivial solution to
radiation efficiency maximization [7]. Otherwise the method
generally tries to make the current distribution as uniform as
possible. This solution tacitly neglects the effect of resonance
tuning and is for several canonical shapes depicted in Fig. 1
as a function of electrical size ka. Note that these curves scale
as (ka)2 for ka < 1.
The additional constraint on self resonance can be incorpo-
rated as
minimize IH I,
subject to IH RI = 1, (8)
H current density suggests that structures from Fig. 3, which
I XI = 0,
and its solution can directly be found by a procedure described
in [15], or by convex relaxation technique described in [16].
The results generated by (8) are depicted by dashed lines
in Fig. 1 for the same problems previously considered. The
difference between solid and dashed curves in Fig. 1 at
small electrical sizes shows the efficiency cost of resonance
tuning. The tuning cost is most easily described by a change
from (ka)2 frequency scaling to (ka)4 scaling, which
agrees well with the findings [8], [9] on a spherical shell.
Results presented in Fig. 1 show that this phenomenon is of
general nature for many electrically small objects. Another
2
101 ext. self-
cylinder
100 sphere
rectangle
arb. shape
101
102
103 `/2
a
10 4 `/ `
10 5
101 100
ka
Fig. 1. Bound to dissipation factor for selected shapes (cylinder, sphere,
rectangle, and an arbitrary shape; depicted in insets) of the current carrying
element (ext.), while dashed lines correspond to self-resonant bounds
(self-) .
interesting phenomenon shown in Fig. 1 is the joining of
the marked and unmarked curves at larger electrical sizes
(ka > 1). This shows that at high electrical sizes the resonant
current distribution tends toward being uniform as it relies
less and less on tuning by a magnetic-dipole-like current, thus
coinciding with the solution to (7).
The self-resonant current maximizing radiation efficiency
on a rectangular support (Fig. 1, rectangular marks) is shown
in Fig. 2a. This current shape is approximately optimal in the
full frequency range of Fig. 1 and resembles a combination
of electric-dipole-like and magnetic-dipole-like currents as
suggested in [8], [9]. In fact, if the optimal current is evaluated
on a spherical shell at small electrical sizes (Fig. 1, triangular
marks), it precisely leads to a resonant combination of TE10
and TM10 spherical modes [8], [9].
V. C OMPARISON OF B OUNDS WITH R EALISTIC D ESIGNS
A. Self-Resonant Antennas
The optimal current density depicted in Fig. 2a is difficult
to realize as a driven antenna current in practice, especially
when a design is restricted to have only one localized feed. The
method described in [15], however, yields also all local optima
of the problem in (8). The local optimum with the second
lowest dissipation factor is depicted in Fig. 2b. The depicted
resemble a Julgalt pastry [22] and a Palmier pastry [23], could
be good candidates for approaching the radiation efficiency
bound. That this is the case is shown in Fig. 4 although it must
be admitted that neither of the structures approach the bound
closely (having dissipation factor at least six times higher than
the bound). The fact that the geometry in Fig. 3a exhibits high
radiation efficiency was already shown in [13, design PMD2].
Though the designs presented here are not necessarily the
optimal antenna geometries for attaining maximum radiation
efficiency, we point out that both designs follow the (ka)4
trend predicted for self-resonant efficiency bounds, and not the
(ka)2 trend predicted by (7).
JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. XX, NOVEMBER 2017 3

(a) (b) (a) (b)

Fig. 2. Surface current densities minimizing dissipation factor for rectangular
region with side ratio 1:2. The figures corresponds to the two lowest minima
of (8). Achieved values of dissipation factor are = 0.140 and = 0.193
for the left and right panels, respectively. The used electrical size is ka = 0.3,
but the current shape is practically unchanged for electrical sizes ka < 1.
Fig. 3. Structures mimicking the shape of current density from Fig. 2b. As
in the case of optimal current densities, the meander layout is restricted to a
rectangle with edge length ratio approximately 1:2. These particular designs
resonate at approximately ka = 0.355 with = 0.517 (Julgalt pastry, left
panel) or at ka = 0.335 with = 0.462 (Palmier pastry, right panel).

103
B. Antennas Externally Tuned by Realistic Components bowtie & spiral inductor
102 fat dipole & spiral inductor
Next, we show that antennas externally tuned by realistic
components do not surpass the self-resonant bound and in 101
fact stay well above the self-resonant spiral meanders when 100
restricted to the same rectangular support. A fat dipole antenna
101

and a bowtie antenna will be used as particular designs,

see insets in Fig. 5. These are chosen for their space-filling
properties, inspired by the uniform current predicted by (7).
At ka < 1 both antennas are capacitive with input impedance
ZA = RA + jXA and XA < 0. The real part of the antenna
impedance can be split as RA = Rrad + Rloss , where Rrad and
Rloss account for radiation and ohmic losses, respectively.
Assume that the antennas are tuned to resonance by series
lumped inductors with impedance ZL = RL (1 + jQL ), where
QL is their Q-factor. It is then straightforward to show [3] that
the total dissipation factor of the system including the tuning
inductor reads
 
|XA |
= A 1 + , (9)
QL Rloss
where A denotes the dissipation factor of the antenna alone. In
this way, we understand the externally tuned efficiency bounds
to be attainable only when components with infinite Q-factor
(i.e., lossless) are available. Any finite Q-factor in the tuning
network must increase loss and will make the lowest bound
in (7) unreachable.
Imagine now that one would desire the total dissipation
factor of an externally tuned antenna to be that of the self-
resonant bound from Fig. 4 (circular marks). The relation in (9)
can then be used to extract the Q-factor of the inductor QL
that would be necessary to achieve this goal. The resulting
102
103 rectangle:
10 4 ext. self- (abs. bound) Palmier pastry
self- (2nd local) Julgalt pastry
105 1
10 100
ka
Fig. 4. Comparison of dissipation factors corresponding to the self-resonant
bound and the externally tuned bound of a rectangular plate, with dissipation
factors of self-resonant spiral meanders from Fig. 3. Data are also compared
to bowtie antenna and fat dipole antenna which were tuned by spiral inductors
made of the same resistive sheet as the antenna proper but being ten times
smaller in electrical size, i.e., kaL = ka/10.
Q-factors QL for the fat dipole and bowtie antennas are
depicted in Fig. 5.
The question now stands if this Q-factor is achievable by
realizable inductors. The negative answer is supported by an
example of a planar spiral inductor and a helical inductor,
which Q-factors are depicted in Fig. 6. In both cases, the
inductors are made of the same material as the antenna. To
approximate the assumption of lumped tuning, let us suppose
that the inductors are at least ten times smaller in electrical size
than the antenna, i.e., kaL = ka/10. In this case the required
JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. XX, NOVEMBER 2017
105
bowtie
fat dipole
`/2
104
`
QL
103
ka = 0.2 `/2
QL = 1500
`
102 1
10 100
ka
Fig. 5. Q-factor of a tuning lumped inductor with which the corresponding
antenna will exhibit the same radiation efficiency as the self-resonant bound
for a rectangular support of the same size.
tuning Q-factors from Fig. 5 are at least an order of magnitude
higher than the realizable Q-factors from Fig. 6. For example,
we observe in Fig. 5 that at ka = 0.2, the bowtie requires
an inductor Q-factor of approximately 1500 to reach the self-
resonant efficiency bound. However, in Fig. 6 it is apparent
that a planar spiral inductor with size kaL = ka/10 = 0.02
can only obtain a Q-factor as high as 53 while a helical
inductor can only reach 130, both far below this requirement.
It is also worth mentioning that higher Q-factors are achieved
by stepping out of the planar geometry, as demonstrated by
the Q-factors of a helical spiral being approximately double
those of planar spirals. Being thus fair the radiation efficiency
bounds for planar supports should only be compared with
tuning by planar inductors. Such comparison is made in Fig. 4
where the bowtie and fat dipole antennas are tuned by planar
spiral inductors. In both cases the dissipation factors of the
externally tuned antennas are well above the self-resonant
spiral meanders.
One may ask whether using physically large inductors
or superconducting materials in the matching network is a
way around the dissipation factor scaling with (ka)4 . If
matching components are made larger while still being con-
sidered approximately lumped (e.g., kaL = ka/8 rather than
kaL = ka/10 in the preceding analysis), then the above anal-
ysis holds, though the precise numerical results may slightly
change. If, however, the tuning network is made similar in size
to the antenna itself, then it becomes appropriate to include
the spatial support of the tuning network into the derivation of
the efficiency bounds. In this way, the tuning network becomes
part of the current optimization problem in (8) where it may
even be leveraged as a source of radiation. In any case the
dissipation factor of the system will scale as (ka)4 .
If high conductivity materials are used for either the match-
ing network or antenna structure, the formulation of both
radiation efficiency bounds remains unchanged, save for the
construction of the loss matrix Rs . However only in the
limiting case that the matching network conductivity goes to
infinity is the (ka)2 trend possible in a resonant system.
4
spiral turns helix turns ka = 0.02
2 4 QL = 130
4 9
9 13
102 linear approx.
QL
ka = 0.02
QL = 53
101
103 102 101
kaL
Fig. 6. Q-factors of planar spiral inductors and helical inductors for varying
numbers of turns. Inductors were made with a spacing between strips equal
to one half and one third of the strip width for the spiral and helical geometry
respectively. The diameter of the helical inductor has been made slightly
bigger than the inductor length, in order to achieve higher quality factor [24].
The radius aL describes a sphere circumscribing the inductor.
VI. C ONCLUSION
It has been shown that radiation efficiency bounds which
assume external tuning by lossless lumped elements are overly
optimistic and that much tighter self-resonant bounds can
easily be calculated. It was further demonstrated that selected
self-resonant antennas can approach this bound and that their
radiation efficiency surpass that of the non-resonant antennas
tuned by realistic reactances. An important conclusion is that
when resonance tuning is demanded, an unpleasant frequency
4
scaling of dissipation factor (ka) must be assumed for
electrically small antennas, rather than the previously predicted
2
(ka) scaling.
R EFERENCES
[1] K. Fujimoto and H. Morishita, Modern Small Antennas. Cambridge
University Press, 2013.
[2] G. S. Smith, Radiation efficiency of electrically small multiturn loop
antennas, IEEE Trans. Antennas Propag., vol. 20, pp. 656657, 1972.
[3] , Efficiency of electrically small antennas combined with matching
networks, IEEE Trans. Antennas Propag., vol. 25, pp. 369373, 1977.
[4] K. Fujita and H. Shirai, A study of the antenna radiation efficiency for
electrically small antennas, in IEEE Antennas and Propagation Society
International Symposium, 2013, pp. 15221523.
[5] A. Karlsson, On the efficiency and gain of antennas, Prog. Electro-
magn. Res., vol. 136, pp. 479494, 2013.
[6] K. Fujita and H. Shirai, Theoretical limitation of the radiation efficiency
for homogenous electrically small antennas, IEICE T. Electron., vol.
E98C, pp. 27, 2015.
[7] M. Shahpari and D. V. Thiel, Physical bounds for antenna radiation
efficiency, 2016, eprint arXiv: 1609.01761. [Online]. Available:
https://arxiv.org/abs/1609.01761
[8] C. Pfeiffer, Fundamental efficiency limits for small metallic antennas,
IEEE Trans. Antennas Propag., vol. 65, pp. 16421650, 2017.
[9] V. Losenicky, L. Jelinek, and M. Capek, Comments on fundamental
efficiency limits for small metallic antennas, pp. 12, eprint arXiv:
1711.00849. [Online]. Available: https://arxiv.org/abs/1711.00849
[10] W. L. Stutzman and G. A. Thiele, Antenna theory and design. John
Wiley & Sons, 2012.
[11] J. L. Volakis, C. Chen, and K. Fujimoto, Small Antennas: Miniaturiza-
tion Techniques & Applications. McGraw-Hill, 2010.
[12] S. R. Best and D. L. Hanna, A performance comparison of fundamental
small-antenna designs, IEEE Antennas Propag. Mag., vol. 52, no. 1, pp.
4770, Feb. 2010.
JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. XX, NOVEMBER 2017 5

[13] S. R. Best and J. D. Morrow, On the significance of current vector

alignment in establishing the resonant frequency of small space-filling
wire antennas, IEEE Antennas Wireless Propag. Lett., vol. 2, pp. 201
204, 2003.
[14] S. R. Best, Electrically small resonant planar antennas, IEEE Antennas
Propag. Mag., vol. 57, no. 3, pp. 3847, June 2015.
[15] L. Jelinek and M. Capek, Optimal currents on arbitrarily shaped
surfaces, IEEE Trans. Antennas Propag., vol. 65, no. 1, pp. 329341,
Jan. 2017.
[16] M. Gustafsson, M. Capek, and K. Schab, Trade-off between
antenna efficiency and Q-factor, Electromagnetic Theory Department
of Electrical and Information Technology Lund University Sweden,
Tech. Rep., 2017. [Online]. Available: http://lup.lub.lu.se/record/
48c12514-eba2-4493-b8a6-c8777051a002
[17] R. F. Harrington, Field Computation by Moment Methods. Wiley
IEEE Press, 1993.
[18] S. M. Rao, D. R. Wilton, and A. W. Glisson, Electromagnetic scattering
by surfaces of arbitrary shape, IEEE Trans. Antennas Propag., vol. 30,
no. 3, pp. 409418, May 1982.
[19] R. F. Harrington, Time-Harmonic Electromagnetic Fields, 2nd ed. Wiley
IEEE Press, 2001.
[20] , Antenna excitation for maximum gain, IEEE Trans. Antennas
Propag., vol. 13, no. 6, pp. 896903, Nov. 1965.
[21] M. Uzsoky and L. Solymar, Theory of super-directive linear arrays,
Acta Physica Academiae Scientiarum Hungaricae, vol. 6, no. 2, pp.
185205, December 1956.
[22] Julgalt, Wikipedia, The Free Encyclopedia. [Online]. Available:
https://sv.wikipedia.org/wiki/Lussekatt
[23] Palmier, Wikipedia, The Free Encyclopedia. [Online]. Available:
https://en.wikipedia.org/wiki/Palmier
[24] H. Nagaoka, The inductance coefficients of solenoids, Journal of the
College of Science, vol. 27, pp. 133, 1909, article 6.