A

Fractal Geometries of
Complementary Split-Ring Resonators
Vesna Crnojevi-Bengin, Member, IEEE, Vasa Radoni, and Branka Jokanovi, Member, IEEE

AbstractComplementary split-ring resonators are typically

used as negative-permittivity particles in microstrip left-handed
structures. In this paper, novel complementary split-ring
resonators that use square Sierpinski fractal curve are proposed
and studied in detail. It is shown that the application of fractal
geometries results in significant miniaturization of the
metamaterial unit cell, in comparison with conventional and
equivalent meander structures. Multiple fractal complementary
split-ring resonators are also analyzed. The influence of different
geometrical parameters and the order of the fractal curve on the
performances are investigated, as well as the efficiency of
excitation of the particles. When used in the design of lefthanded transmission lines, fractal complementary split-rings
improve frequency selectivity in the upper transition band, when
compared to other non-fractal topologies.
Index TermsMetamaterials, split-ring resonators, fractals,
microwave filters.

I. INTRODUCTION
Recently, revolutionary results have been obtained in the field
of metamaterials, artificial structures that exhibit
electromagnetic properties generally not found in nature.
Metamaterials are designed using sub-wavelength particles,
whose size is typically smaller then one tenth of the
propagating signal wavelength. Due to this fact, quasi-static
analysis can be performed and the concept of artificial
effective media can be applied. The structures obtained using
this approach can be considered as a continuous medium with
effective parameters, namely effective dielectric permittivity
and effective magnetic permeability. By a proper choice of
the type and geometrical arrangement of constituent subwavelength particles, the effective parameters of
metamaterials can be made arbitrarily small or large, or even
negative.
Double-negative or left-handed media, that simultaneously
exhibit negative values of permittivity and permeability in a
certain frequency range, were first theoretically studied by
Manuscript received September 21, 2007. This work was supported in part by
the Ministry for Science and Environmental Protection of Republic of Serbia
under Contract 401-00-213/2006-01/05, and by EUREKA project number
E!3853.
V. Crnojevi-Bengin and V. Radoni are with the Faculty of Technical
Sciences, University of Novi Sad, Serbia (phone: 381-21-485-2553; fax:
381-21-475-0572; e-mail: bengin@uns.ns.ac.yu, vasaradonic@eunet.yu).
B. Jokanovi is with the IMTEL-Komunikacije, Belgrade, Serbia (e-mail:
branka@insimtel.com).

Veselago, [1]. However, the first structure that exhibits

negative permittivity by decreasing the plasmon frequency
into microwave range was proposed in mid nineties, [2].
Shortly afterwards, a particle called split-ring resonator, that
provides negative permeability at microwave frequencies, was
introduced, [3]. By superimposing these two structures, the
existence of left-handed metamaterial was experimentally
proved in 2001, [4].
An array of split-ring resonators exhibits the extreme values
of effective magnetic permeability in the vicinity of
resonance, namely highly positive/negative in a narrow band
below/above the quasi-static resonant frequency of the rings.
Although having a narrow frequency range with negative
permeability, the configurations using split-ring resonators
have driven a lot of attention, [5], [6]. An array of split-ring
resonators exhibits filtering properties, and, when properly
polarized, can inhibit signal propagation, thus offering an
effective way to reject a frequency band in the vicinity of its
quasi-static resonance, [7].
In the microstrip technology, split-ring resonators can only
be etched on the upper substrate side, next to the host
transmission line. In order to enhance the coupling, the
distance between the line and the rings should be as small as
possible. Therefore, square or rectangle geometries are
typically used instead of the originally proposed circular ones.
The microstrip line loaded with split-ring resonators is a
single-negative medium, and therefore exhibits a stop-band
characteristic.
With the aim of further miniaturization, other subwavelength particles have been recently proposed, such as the
broadside coupled split-ring resonator, [8], and the spiral
resonator, [9], [10], as well as multiple geometries, namely
the multiple split-ring resonator, and the multiple spiral
resonator, [11], [12]. Although solutions utilizing multiple
spiral resonators exhibit the highest potential for
miniaturization, they also suffer from high insertion loss.
Applying the Babinet principle, a complementary structure
was proposed in [13], namely the complementary split-ring
resonator. In the microstrip technology, complementary splitring resonators are etched in the ground plane, beneath the
microstrip, with their axes parallel to the vector of the electric
field, thus contributing to the negative effective dielectric
permittivity of the structure. In order to obtain the left-handed
behaviour, effective negative permeability has to be
introduced to the structure. This is achieved by periodically
etching capacitive gaps in the conductor strip.
In this paper, novel sub-wavelength particles are presented
based on the application of fractal geometries to

complementary
split-ring
resonators,
[14].
The
electromagnetic behaviour of the proposed particles is
investigated from the point of view of the size reduction,
performances and coupling to the host transmission line.
In Section II, the possibilities of application of twodimensional fractal curves in sub-wavelength particle design
are analyzed. The definition of fractal dimension is given and
it is described why square Sierpinski fractal curve has been
chosen to substitute conventional square complementary splitring resonator geometry. The configuration of the proposed
square Sierpinski complementary split-ring resonators is also
shown.
In Section III new metamaterial unit cells based on the
application of square Sierpinski complementary split-ring
resonators and gaps etched in the microstrip are presented.
The influence of different geometrical parameters to the
performances of the proposed unit cells is analyzed in detail,
and compared with other similar but non-fractal geometries.
The efficiency of excitation of square Sierpinski
complementary split-ring resonator by axial electric field is
analyzed as a function of geometry of the ring. Unit cells that
use multiple square Sierpinski complementary split-ring
resonators are investigated for the first time, as well as the
influence of the fractal curve order on performances of the
proposed unit cells. The electrical model of the unit cell and
the procedure for circuit parameter extraction are given, and
circuit parameters are extracted for all structures. Different
behaviour of conventional square and fractal rings is
illustrated by dispersion diagrams.
In Section IV, the proposed unit cells are applied to the
design of left-handed transmission lines. The simulation and
measurement results are given, and the performances of all
fractal-based structures are compared with those obtained by
configurations that use conventional square complementary
split-ring resonators with the same dimensions. The
conclusion is given in Section V.

miniaturized antennas [18]-[21], frequency selective surfaces,

[22], high-impedance surfaces, [23], [24], left-handed
metamaterials, [25], and radio-frequency identification, [26].
In the metamaterial designs, a bulk effective medium is
formed by embedding a great number of identical subwavelength fractal-shaped inclusions within a host medium.
Also, an extended class of space-filling wire structures based
on grid-graph Hamiltonian paths and cycles has been
investigated, [27]. Possible practical applications of the
results mentioned above are envisaged in the design of thin
absorbing screens, and ground-planes for antennas. However,
they are not directly applicable to the split-ring design,
because, in the latter case, some additional characteristics of
the fractal curves are required.
The fractal curves are characterized by fractal, i.e. noninteger, dimension. The dimension of every fractal curve is
the number between 1 and 2, and can be understood as a
measure of the space-filling ability of the fractal curve. The
dimension D can be determined as a logarithmic ratio between
the number of self-similar segments obtained from one
segment after each iteration, k, and the number of segments
obtained from one segment in each iteration, r:

D=

log k
log r

(1)

A great number of fractal curves are known today. Trying

to determine which fractal curve would suit best given
application, some criterion has to be introduced. With the aim
of further miniaturization, the fractal dimension is chosen as
the most important characteristic. The higher the fractal
dimension, the better fractal curve fills the given area,
therefore achieving higher compactness. Three fractal curves
are known that have fractal dimension equal to 2 (i.e. the
maximum value): Peano, Hilbert and square Sierpinski
fractals, shown in Fig. 1.

II. FRACTAL CURVES

The fractal curves have been known since the end of 19th
century, when Peano constructed a continuous curve that
passes through every point of a given region, [15]. The fractal
curves are generated in an iterative manner by successive
repetition of one geometrical shape with the other (that often
is a collection of scaled copies of the first shape). After each
iteration, a fractal curve of the higher order is obtained, longer
then the previous one, which better fills the area in which it is
generated. This space-filling property offers high potentials
for miniaturization of passive microwave circuits, because,
theoretically, the application of fractal curves allows the
design of infinite-length lines on a finite substrate area.
Our recent results have shown that the application of fractal
curves results in very compact high-Q microstrip resonators
and filters, [16], [17], which outperform all other non-fractal
solutions. This is due to the increase of the overall length of
the microstrip line on a given substrate area as well as to the
specific line geometry.
The fractal geometries have also been utilized for the
design of wire structures used in applications such as

(a)

(b)

(c)

Fig. 1. Fractal curves with the dimension equal to 2: (a) Peano, (b) Hilbert,
and (c) square Sierpinski.

Since Hilbert and Peano fractals are open curves that have
ends on the opposite sides, they are convenient for usage in
structures such as end coupled microstrip resonators, [16], or
metasurfaces, [23]-[25], [27]. However, in the case of splitring resonators, some specific characteristics of the fractal
curve are required. In order to ensure the best performances of
a split-ring, a trade-off is needed between increased
inductance and capacitance of the particle (i.e. maximized line
length) and its geometry. The geometry of the particle has to

allow for efficient excitation, i.e. the particle has to have an

unoccupied area in its middle section.
Square Sierpinski fractal consists of squares that fill the
given area. It resembles Minkowski fractal, but has ratio
between inner and outer squares equal to 2:3. Furthermore, in
the case of square Sierpinski fractal, inner and outer squares
overlap, thus creating a denser structure. Square Sierpinski
fractal curves of the first three orders are shown in Fig. 2.

chosen to be the minimal achievable, i.e. equal to 100 m, to

enhance the coupling.
To facilitate insertion of more then one concentric square
Sierpinski complementary split-ring inside the structure,
while keeping the overall dimensions of the particle fixed, gap
g has to be reduced. This also results in a decrease of length
of insets d. Square Sierpinski complementary split-ring
resonators obtained in this manner are shown in Fig. 4, for the
case N=2, 3 and 4, respectively, where N denotes number of
concentric complementary rings.

Fig. 2. Generation of square Sierpinski fractal curve first three iterations.

(a)

Sierpinski fractal curve of the second order with the

relevant dimensions is shown in Fig. 3(a), where a, b and g
denote size of its segments. According to the rule of square
Sierpinski fractal generation, the following relations are
obtained:
b = 3 g

(2)

a = 2 g

A closed curve can be designed that follows the outer

perimeter of the square Sierpinski fractal curve that can
straightforwardly be used as a substitute to the conventional
square complementary split-ring resonator.

(a)

(b)

(c)

Fig. 4. Multiple square Sierpinski complementary split-ring resonators with

N concentric rings: (a) N=2, (b) N=3, (b) N=4.

III. LEFT-HANDED UNIT CELLS THAT USE SQUARE SIERPINSKI

COMPLEMENTARY SPLIT-RING RESONATORS
A. Configuration
In order to obtain left-handed behaviour, two particles need
to be combined in a unit cell: complementary split-ring
resonator that will provide negative effective permittivity and
a gap in the microstrip that will provide negative effective
permeability. In order to achieve high magnetic coupling
between the line and the ring at resonance, complementary
split-ring resonators are etched in the ground layer under the
gaps. The proposed unit cell is shown in Fig. 5, where both
top (dark grey) and bottom (light grey) conductive layers are
shown. The unit cell was simulated on a 1.27 mm thick
Taconic CER-10 substrate, with r= 9.8 and dielectric loss
tangent equal to 0.0025. The outer dimensions of a single
square Sierpinski complementary split-ring resonator are
equal to 5.1 x 5.1mm, i.e. g/15 x g/15 on a given substrate.

(b)

Fig. 3. (a) Square Sierpinski fractal curve of the second order, (b) Single
square Sierpinski complementary split-ring resonator.

The square Sierpinski complementary split-ring resonator,

designed to follow the outer perimeter of the fractal curve, is
shown in Fig. 3(b), where d denotes length of insets. The
etched line having minimal width achievable by the
conventional PCB technology is used, equal to 100 m. It is
well known, [3], that by the usage of two concentric rings
with slits on the opposite sides, resonant frequency of the
structure can be significantly reduced. In order to investigate
influence of the number of concentric rings, multiple (N2)
square Sierpinski complementary split-ring resonators will
also be analyzed. The separation between concentric rings is

Fig. 5.

Left-handed unit cell that consists of N=2 square Sierpinski

complementary split-ring resonator and a gap in the host microstrip line.

B. Influence of Different Geometrical Parameters

In order to analyze the influence of different geometrical
parameters on the performances of square Sierpinski

complementary split-ring resonator, lengths of insets d are

varied. Fig. 6 shows two extreme cases: d=1800 m is the
original square Sierpinski complementary split-ring resonator,
i.e. the one whose dimensions correspond to the rule of
generation of the fractal curve, and d=600 m corresponds to
the minimum achievable inset. Furthermore, to highlight the
specific features of fractal curves, the proposed unit cell is
compared with a cell that uses conventional square
complementary split-ring resonator (CSRR) as well as with a
cell that uses meander shaped complementary split-ring
resonator (M CSRR), shown in Fig. 6(c). Meander
complementary split-ring resonator has been designed to
occupy precisely the same area as the original square
Sierpinski complementary split-ring resonator, as well as to
have exactly the same circumference of the rings. The
performances of the proposed structures are determined using
EMSight, EM simulator in Microwave Office and the
comparison of simulation results for the lossy case is given in
Table I, where fr denotes resonant frequency of the unit cell,
BW is a 3 dB bandwidth, s210 and s110 are insertion loss and
reflection at resonant frequency, QL is loaded and QU is
unloaded quality factor.

(a)

(b)

(c)

Fig. 6. Analyzed topologies: (a) square Sierpinski complementary split-ring

resonator, d=1800 m, (b) square Sierpinski complementary split-ring
resonator, d=600 m, and (c) meander complementary split-ring resonator
with the same circumference and the same area of the rings as the resonator
depicted in (a).
TABLE I
SIMULATION RESULTS FOR UNIT CELLS THAT USE N=2 SQUARE SIERPINSKI
COMPLEMENTARY SPLIT-RING RESONATORS (SS CSRR) WITH DIFFERENT
LENGTHS OF INSETS, COMPLEMENTARY SPLIT-RING RESONATOR (CSRR)
AND MEANDER COMPLEMENTARY SPLIT-RING RESONATOR (M CSRR)

d [m]
fc1 [GHz]
BW [MHz]
BW [%]
s210 [dB]
s110 [dB]
QL
QU

1800
1.34
97.7
7.29
-5.02
-8.95
13.72
20.01

SS CSRR
1200
1.51
126
8.34
-3.86
-11.2
11.98
20.35

600
1.72
177.3
10.31
-2.9
-14.3
9.7
19.91

CSRR
2.03
256.6
12.64
-2.23
-20.2
7.91
19.7

M CSRR
1.57
127
7.38
-3.83
-12
12.36
21.1

Since decreasing d results in shorter circumference of the

square Sierpinski complementary split-ring resonator, its
resonant frequency increases, due to the reduced inductance
and capacitance of the particle. However, in all cases a
significant reduction of the resonant frequency in comparison

to the conventional complementary split-ring resonator can be

observed, ranging from 15% to 34%, for different length of
inset d. Furthermore, all fractal structures exhibit higher
values of the quality-factor then the conventional ring.
The comparison of unit cells that use meander
complementary split-ring resonator and the original (d=1800
m) square Sierpinski complementary split-ring resonator,
gives the evidence to improved performances of the fractal
geometries in comparison with similar but non-fractal ones.
Although both structures have exactly the same line length
and occupy precisely the same area, resonant frequency of the
unit cell that uses square Sierpinski complementary split-ring
resonator is approximately 15% lower, due to the specific
shape of the fractal curve.
Therefore, it can be concluded that fractal geometries with
the original dimensions perform better in terms of
miniaturization than the modified ones, as well as better then
similar non-fractal structures. Although square Sierpinski
complementary split-ring resonator with insets equal to 600
m shows significantly better insertion loss then its
counterpart with maximized insets, it is less interesting from
the point of view of miniaturization. That is why in the
remaining part of this paper, only fractal structures with the
original dimensions will be considered, modified only to the
minimal extent to allow the insertion of inner concentric
rings. However, it should be kept in mind that fractal rings
can achieve significantly lower insertion loss if insets smaller
then the original ones are used, at the price of somewhat
increased dimensions of the unit cell.
C. Efficiency of Excitation
Complementary split-ring resonator is excited by axial
electric field that penetrates the area in the centre of the ring.
In the case of square Sierpinski complementary split ring
resonator, this area is partially occupied by the ring. In this
section, the influence of fractal geometry on the efficiency of
excitation is analyzed.
Two unit cells with similar geometries are simulated: one
that uses the proposed square Sierpinski complementary splitring resonator, and the other that uses quasi square Sierpinski
complementary split-ring resonator. In the latter case, two
opposite insets are rotated outwards, Fig. 7(b), so that the size
of the homogenous metallic segments within the ring is
increased for approximately 15%. It is commonly believed
that such structure will allow more efficient excitation and
result in stronger coupling between the host microstrip line
and the ring.
The simulation results shown in Table II reveal that the
resonant frequency is almost unchanged, which is expected
due to the fact that both structures have approximately the
same inductance and capacitance. However, the other
performances are almost unchanged as well: the unit cell with
quasi square Sierpinski complementary split-ring resonators
exhibits very slightly wider 3 dB bandwidth, somewhat
smaller insertion loss in the pass-band, and lower quality
factors. All these can be attributed to enhanced excitation, i.e.
better coupling between the ring and the microstrip, which is

the direct result of increased area of the homogenous metallic

segments within the ring. However, all differences are so
small that we concluded that the shape of the fractal ring does
not significantly influence the excitation. The increased
insertion loss that fractal rings exhibit in comparison to the
conventional ones (Table I) is not due to deteriorated
excitation, but is the consequence of the significantly
increased length of the fractal ring. This conclusion is also
supported by additional experiments performed with similar
structures, such as square Sierpinski complementary split-ring
resonator rotated for 45 degrees in respect to the original one.
In all cases, almost negligible differences in performances
were observed although area available for excitation was
significantly changed.

(a)

(b)

Fig. 7. Unit cell that uses: (a) square Sierpinski complementary split-ring
resonator, (b) quasi square Sierpinski complementary split-ring resonator.
TABLE II

compared, due to considerably increased capacitance of the

unit cell. Adding more then two concentric complementary
split-rings changes the resonant frequency only for a few
percents, owing to the specific space-filling property of the
fractal curves. However, when the number of concentric rings
is increased, the second harmonic is significantly shifted
towards the higher frequencies, up to more then three times
the first resonance, thus resulting in a much wider stop band.
TABLE III
SIMULATION RESULTS FOR UNIT CELLS THAT USE MULTIPLE SQUARE
SIERPINSKI COMPLEMENTARY SPLIT-RING RESONATORS
N
1
2
3
4
fc1 [GHz]
1.94
1.34
1.33
1.36
BW [MHz]
159.5
97.7
94.4
121.7
BW [%]
8.2
7.3
7.1
8.9
-4.27
-5.02
-5.04
-4.43
s210 [dB]
-3.24
-8.95
-8.25
-9.21
s110 [dB]
3.08
2.86
3.47
4.25
fc2 [GHz]
fc2/fc1
1.58
2.13
2.61
3.11
12.19
13.71
14.09
11.22
QL
19.48
20.02
20.52
17.54
QU

E. Influence of the Fractal Curve Order

Unit cells with N=1 and N=2 square Sierpinski
complementary split-rings of the third order have been
simulated for the lossy case, to analyze the influence of the
order of the fractal curve. The overall dimensions of square
Sierpinski complementary split-ring resonators are
unchanged, i.e. equal to 5.1 x 5.1 mm. Both structures are
shown in Fig. 8, while their performances are compared in
Table IV.

SIMULATION RESULTS FOR UNIT CELLS THAT USE SQUARE SIERPINSKI

COMPLEMENTARY SPLIT-RING RESONATORS (SS CSRR) AND QUASY SQUARE

SIERPINSKI COMPLEMENTARY SPLIT-RING RESONATORS (Quasi SS CSRR)

fc [GHz]
BW [MHz]
BW [%]
s210 [dB]
s110 [dB]
QL
QU

SS CSRR

Quasi SS CSRR

1.35
97.5
7.22
-5.01
-8.95
13.85
20.23

1.32
97.6
7.39
-4.86
-9.72
13.52
20.08
(a)

D. Unit Cells That Use Multiple Square Sierpinski

Complementary Split-Ring Resonators
The influence of additional inner concentric rings of square
Sierpinski
complementary
split-ring
resonator
on
performances of the unit cell is analyzed in Table III for the
lossy case.
In the case of the conventional complementary square splitring resonators, adding multiple split-rings significantly
reduces resonant frequency, due to the increased inductance
of the structure, [11], [12]. In the case of fractal curves, this
holds only when the structures with N=1 and N=2 rings are

(b)

Fig. 8. Unit cells that use (a) N=1 and (b) N=2 concentric square Sierpinski
complementary split-rings with the fractal curve of the third order.

By comparing the results for N=1 from Table III and Table
IV, it can be seen that the application of the fractal curves of
the higher order results in the reduction of the resonant
frequency of almost 20%, and therefore allows the design of
more compact unit cells. When the number of concentric rings
N is increased, the proposed unit cell scales down to g/20 x
g/20. The increased losses are direct result of the increased

line length, but at the same time the second harmonic is

shifted at more then four times the first resonance.
TABLE IV

inductance L is estimated using a transmission line calculator,

and the capacitance Cg is determined based on the value of
series reactance.
The transmission zero frequency,

SIMULATION RESULTS FOR UNIT CELLS THAT USE N SQUARE SIERPINSKI

COMPLEMENTARY SPLIT-RINGS OF THE THIRD ORDER

N
fc1 [GHz]
BW [MHz]
BW [%]
s210 [dB]
s110 [dB]
fc2 [GHz]
fc2/fc1
QL
QU

1
1.56
58.2
3.7
-5.56
-6.38
3.86
2.47
26.89
37.24

fz =

2
1.14
57.7
5.1
-8.85
-3.76
4.8
4.21
19.69
22.64

f0
1
f
1 ,
Lr = z 2
(2f 0 ) C c

Cr =

2 Lr C r

(6)

f0
1
fz

-20
s21 [dB]

Values of the elements of equivalent circuit are obtained

from the frequency response of the unit cell, for the lossless
case. Two specific frequencies are used in the process: the
resonant frequency of the complementary split-ring, f0, and
the transmission zero frequency fz, at which the impedance of
the shunt branch is equal to zero.
At the resonant frequency of the complementary split-ring,
,

Cc

(5)

thus giving dependence of Lr and Cr from Cc. The response of

the electrical model is fitted to the one obtained from fullwave simulations by changing value Cc, using Variable Tuner
tool available in Microwave Office. To validate the parameter
extraction method and the obtained values of circuit elements,
the responses are compared in Fig. 10, for both fractal and
conventional unit cells with two concentric rings (N=2). It can
be seen that the electrical model describes electromagnetic
behaviour of the unit cell very accurately, and in a wide range
of frequencies, up to the second harmonic. The second
harmonic exists in electromagnetic responses due to the
distributed nature of the structures. A more complicated
electrical model could be developed that would accurately
predict the second resonance too. However, since the second
harmonic is not of interest for the current application, the
simplified model shown in Fig. 9 is used.

Fig. 9. Equivalent circuit of the proposed unit cell.

(4)

2 Lr (C r + C c )

is also easily obtained from full-wave simulations. Equations

(3) and (4) can be rewritten in the following form:

F. Electrical Model and Parameter Extraction

The proposed unit cell can be modelled by the equivalent
circuit shown in Fig. 9 regardless of the type of the
complementary split-ring resonator used (conventional or
Sierpinski) or of the number of concentric rings. This model
is valid under the assumption that the size and the distance
between the adjacent rings are both electrically small, which
is the case. The complementary split-ring resonator is
modelled by the parallel resonant circuit (with inductance Lr
and capacitance Cr), electrically coupled to the host microstrip
line through the line capacitance Cc. The microstrip line is
modelled by the inductance L, while the capacitance Cg
models the series gap in the microstrip.

f0 =

em CSRR
el CSRR

-40

em SS CSRR
el SS CSRR

(3)

-60
0

f [GHz]

the impedance seen from the input port is given by the

impedance of the output port, equal to 50, enlarged for the
reactive impedance of the series branch. This frequency can
easily be determined from the Smith Chart. The line

Fig. 10. Comparison of responses obtained from full-wave electromagnetic

simulations (em) and electrical simulations (el) of model shown in Fig. 9, for
unit cells that use square Sierpinski complementary split-ring resonator (SS
CSRR) and complementary split-ring resonator (CSRR) with two concentric
rings.

Table V contains the extracted values of circuit elements

for unit cells that use multiple square Sierpinski
complementary split-ring resonators. To illustrate the
differences which occur when fractal geometry is used,
extracted parameters for conventional multiple square
complementary split-ring resonators that occupy the same
area as their fractal counterparts are shown in Table VI,
together with the extracted values for the meander
complementary split-ring shown in Fig. 6(c). In both tables, N
denotes the number of concentric rings.
TABLE V
EXTRACTED PARAMETERS FOR UNIT CELLS THAT USE MULTIPLE SQUARE
SIERPINSKI COMPLEMENTARY SPLIT-RING RESONATORS DESIGNED WITH
FRACTAL CURVES OF THE SECOND AND THE THIRD ORDER

Order
N
L [nH]
Cg [pF]
Cc [pF]
Lr [nH]
Cr [pF]

2
2.49
0.329
2.49
2.786
4.095

II
3
2.37
0.354
3.13
2.361
5.044

III
4
2.3
0.389
3.64
2.132
5.222

1
2.11
0.296
2.36
1.114
8.542

2
2.08
0.304
2.63
2.326
7.434

TABLE VI
EXTRACTED PARAMETERS FOR UNIT CELLS THAT USE MULTIPLE SQUARE
COMPLEMENTARY SPLIT-RING RESONATORS AND N=2 MEANDER
COMPLEMENTARY SPLIT-RING RESONATOR WITH THE SAME DIMENSIONS AS

N
L [nH]
Cg [pF]
Cc [pF]
Lr [nH]
Cr [pF]

ITS FRACTAL COUNTERPART.

SQUARE COMPLEMENTARY SPLIT-RINGS
3
4
2

2.22
0.325
2.52
1.774
2.63

2.08
0.338
2.81
2.282
2.47

1.87
0.341
2.71
2.815
2.23

MEANDER
2
2.48
0.327
2.47
2.2
3.75

By comparing results in Tables V and VI (square

complementary split-rings) for N=2, it can be seen that the
application of fractal geometry increases the inductance of the
split-ring for approximately 57% and its capacitance for
more then 55%, for the fixed overall dimensions of the ring.
This is due to the significant increase of the rings
circumference. Host microstrip line inductance, gap
capacitance and coupling capacitance remain almost
unchanged. A comparison between N=2 fractal and meander
geometries reveals significant advantage of the former:
although both complementary split-rings occupy exactly the
same area and have exactly the same circumference of the
rings, square Sierpinski exhibits both larger inductance and
larger capacitance, due to the specific shape of the fractal
curve. Again, this gives evidence to superior performances of
the fractal geometries in comparison with non-fractal ones.
Different behaviour of fractal and non-fractal structures can
be observed for higher values of N. In the case of
conventional square complementary split-rings, the
inductance Lr significantly increases with N, which is
expected due to the increased total length of the rings. In the

same time, the capacitance Cr decreases slightly due to

reduced dimensions of the most inner ring, [7]. On the other
hand, quite the opposite effect is observed in the case of
fractal ring geometry: the inductance Lr decreases and the
capacitance Cr significantly increases with N. This can be
explained in the following manner. Although the
circumference of each ring is increased due to its fractal
shape, the total inductance is reduced because of the existence
of line segments with opposite currents in each ring (such as
insets with length d, Fig. 3(b)). However, the existence of
insets does not influence the resulting capacitance between
the rings: fractal concentric rings with almost doubled
circumference result in more then two times bigger
capacitance, when compared to the conventional square case.
This conclusion also holds for complementary split-rings
that use fractal curve of the third order: although the total line
length is significantly increased in comparison to the rings
that use fractal curves of the second order, the inductance Lr is
almost unaffected. Again, this is explained by the existence of
a great number of parallel line segments with opposite
currents that reduce the total inductance. The observed
reduction of the resonant frequency is mainly due to the
increased capacitance of the structure.
Different behaviour of conventional square and fractal rings
is also noticeable from the dispersion diagrams shown in Fig.
11. The dispersion relation has been calculated from the
equivalent circuit:

C
cos l = 1 + c
2C g

(1 2 LC g )(1 2 L r C r )
1 2 Lr (C c + C r )

(7)

Although both structures are unbalanced and operate in

somewhat different frequency bands, it can be seen that
increasing the number of concentric rings N has different
influence on the dispersion: it increases the stop band in the
conventional case and decreases it in the fractal case.
Furthermore, the left-handed pass band is almost unaffected
by changing N in the fractal case.
IV. LEFT-HANDED MICROSTRIP LINES THAT USE SQUARE
SIERPINSKI COMPLEMENTARY SPLIT-RING RESONATORS
A. Left-Handed Microstrip Lines That Use Unit Cells With
N=2 Square Sierpinski Complementary Split-Ring
Resonators
One possible application of the proposed unit cells is in the
design of left-handed transmission lines. Such transmission
lines can be characterized as narrow band pass filters with a
sharp transition in the lower band edge. However, in the case
of conventional square complementary split-ring resonators,
they exhibit poor frequency selectivity in the upper transition
band.
In Table VII the simulation results for the lossy case are
compared for left-handed microstrip lines that use unit cells
with N=2 square Sierpinski complementary split-ring
resonators with insets equal to d=1800 m and d=600 m,

and conventional N=2 complementary split-ring resonators.

The structures are characterized in terms of a band pass filter,
where fc denotes central frequency, BW is a 3 dB bandwidth,
s210 is insertion loss in the pass band, QL is loaded and QU is
unloaded quality factor.

Bl [rad]

3.14

N=2
N=3
N=4
0.00
1

10

By comparing results obtained for the conventional square

complementary split-ring resonators with those corresponding
to the original fractal geometries (d=1800 m), it can be seen
that the application of square Sierpinski complementary splitring resonators reduces central frequency for more than 35%,
for the same overall dimensions of the unit cell, at the price of
reduced bandwidth and increased insertion loss. However, if
the fractal geometry with d=600 m insets is used instead,
reduction of resonant frequency of approximately 17% is still
achieved, while other performances are preserved. This
demonstrates the potential that fractal geometries have for
miniaturization.
In order to validate simulation results, left-handed
microstrip lines that use square Sierpinski complementary
split-ring resonators with the original dimensions and the
conventional square complementary split-ring resonators were
fabricated in standard PCB technology on a 1.27 mm thick
Taconic CER-10 substrate. Photographs of top and bottom
layers of both fabricated structures are shown in Fig. 12.

f [GHz]

(a)

Bl [rad]

3.14

N=2
N=3
N=4
0.00
1

10

(a)

(b)

f [GHz]

Fig. 12. Top (upper) and bottom (lower) sides of fabricated structures using:

(b)

(a) conventional square complementary split-ring resonators, (b) square

Fig. 11. Dispersion diagrams of left-handed unit cells that use: (a) square
Sierpinski complementary split-ring resonators with fractal curve of the
second order, and (b) conventional square complementary split-ring
resonators.
TABLE VII
COMPARISON OF SIMULATION RESULTS FOR LEFT-HANDED TRASMISSION
LINES THAT USE SQUARE SIERPINSKI COMPLEMENTARY SPLIT-RING
RESONATORS (SS CSRR) AND CONVENTIONAL SQUARE COMPLEMENTARY
SPLIT-RING RESONATORS (CSRR) WITH TWO CONCENTRIC RINGS (N=2)

SS CSRR

SS CSRR

d [m]

1800

600

CSRR
-

fc [GHz]

1.365

1.765

2.123

BW [MHz]

69.5

164

170.95

BW [%]

5.1

9.29

8.05

s210 [dB]

-12.8

-6.48

-5.62

s110 [dB]

-8.95

-14.3

-20.2

QL

19.6

10.76

12.4

QU

20.7

13.88

17.1

Sierpinski complementary split-ring resonators.

The simulation and measurement results for both structures

are shown in Fig. 13. A good agreement with simulations can
be observed, except for a shift in frequency approximately
equal to 10% that have occurred in both cases. Since
manufacturer specifications for substrate material allow r
variations in the range +/- 0.5 as well as variations of
substrate thickness, this can be explained by the discrepancy
between the actual and the simulated values of the dielectric
constant and substrate thickness. The measured insertion
losses correspond well to the simulated ones.
Fig. 14 shows the results of measurements performed up to
6GHz, for both structures. It can be seen that, unlike the
conventional one, the configuration that uses square
Sierpinski complementary split-ring resonators shows sharp
transition on both sides of the pass band. Furthermore, it
suppresses the second harmonic below 22dB, thus creating a
wide and deep stop band.

S11
-10

Conventional
CSRR

s21, s11 [dB]

S21
-20

Sierpinski
CSRR

-30

-40

-50
1,0

1,5

2,0

2,5

3,0

f [GHz]

Fig. 13. Simulation (dotted line) and measurement (full line) results for lines
that use square Sierpinski complementary split-ring resonators and
complementary split-ring resonators.

(a)
(b)
Fig. 15. Top (upper) and bottom (lower) sides of the fabricated left-handed
transmission lines that use unit cells with square Sierpinski complementary
split-ring resonators: (a) N=3, (b) N=4.

S11

N=2 measured
N=2 simulated
N=3 measured
N=3 simulated
N=4 measured
N=4 simulated

-10

-20

S21

-20

s21 [dB]

s11, s21 [dB]

-10

-30

-30

-40

-40
-50
1,0

1,5

2,0

2,5

3,0

3,5

4,0

4,5

5,0

5,5

6,0

-50

f [GHz]

1.0

1.2

1.4

1.6

1.8

2.0

f [GHz]
Fig. 14.

Measurement results for the lines that use square Sierpinski

(a)
0

complementary split-ring resonators (full line) and complementary split-ring

resonators (dotted line) in wider frequency range.

-2

s11 [dB]

B. Left-Handed Microstrip Lines That Use Unit Cells With

Multiple Square Sierpinski Complementary Split-Ring
Resonators
In order to validate simulation results, left-handed
microstrip lines that use multiple square Sierpinski
complementary split-ring resonators were fabricated.
Photographs of top and bottom layers of the fabricated
structures with N=3 and N=4 concentric square Sierpinski
complementary split-ring resonators are shown in Fig. 15.
The measurement and simulation results of the proposed
left-handed microstrip lines are compared in Fig. 16. A good
agreement can be observed, except for a shift in frequency
approximately equal to 10% that occurs in all cases. As in the
previous case, this can be explained by the discrepancy
between actual and simulated values of the dielectric constant
and substrate thickness.

-4

N=2 measured
N=2 simulated
N=3 measured
N=3 simulated
N=4 measured
N=4 simulated

-6

-8
1.0

1.2

1.4

1.6

1.8

2.0

f [GHz]
(b)
Fig. 16. Simulation (dotted lines) and measurement (full lines) results for
left-handed

microstrip

lines

that

use

N=2,3,4

square

Sierpinski

complementary split-ring resonators: (a) insertion loss, (b) return loss.

10

Fig. 17 shows the measurement results up to 6 GHz, for all

proposed left-handed lines. It can be seen that all
configurations show sharp transition on both sides of the first
pass-band. As N increases, second harmonics are shifted
towards higher frequencies. Furthermore, all structures
successfully suppress a frequency band in the vicinity of the
second harmonic of the quasi-static resonance of the rings,
positioned at approximately 2fc1. In that way, wide and deep
stop bands in the transmission characteristics are created.
0

S11

-10

s21, s11 [dB]

-20

S21
-30

-40

REFERENCES

N=2
N=3
N=4

-50

-60
1

geometries is observed and explained: by adding concentric

rings, the capacitance of the structure becomes dominant,
instead of the inductance. Dispersion diagrams further
illustrate different behaviour of the lines that use multiple
conventional and fractal complementary split-rings.
The efficiency of excitation of square Sierpinski
complementary split-ring resonator by axial electric field is
also analysed. Although the centre of the proposed unit cell is
partially occupied by the ring, this does not significantly
influence the excitation. The increased insertion loss that
fractal rings exhibit in comparison to the conventional ones is
not due to deteriorated excitation, but is the consequence of
the significantly increased length of the fractal ring.
When the order of the fractal curve used for complementary
split-ring resonator design is increased, while the overall
dimensions are fixed, the resonant frequency can be lowered
for 44% in respect to the conventional case. In this way, even
more compact unit cells with the dimensions g/20 x g/20
can be designed.

f [GHz]

Fig. 17. Measurement results for the proposed left-handed microstrip lines
that use N=2,3,4 square Sierpinski complementary split-ring resonators in
wider frequency range.

V. CONCLUSION
In this paper, novel complementary split-ring resonator
geometries are presented, that use square Sierpinski fractal
curves of the second and the third order. The topologies with
different number of concentric rings are also analyzed, as well
as the influence that several geometrical parameters have on
the performances. It is shown that fractal geometries with the
original dimensions perform better in terms of miniaturization
then the modified ones, and better then similar but non-fractal
structures.
Simulations and measurements show that the application of
fractal geometries significantly lowers resonant frequency of
the structure, therefore revealing high potential that fractal
topologies have for the unit cell miniaturization. Due to the
unique shape of the fractal curve, inductance of the double
split-ring is increased for approximately 57% and its
capacitance for more than 55%, for the same overall
dimensions of the ring, in comparison to the conventional
case. Furthermore, the improved frequency selectivity in the
upper transition band is achieved.
Since fractal curves fill the space in an optimal manner,
adding concentric split-rings changes the resonant frequency
only for a few percents. However, the second harmonic is
significantly shifted towards the higher frequencies when the
number of concentric rings is increased, thus resulting in
much wider stop bands. The proposed structures with multiple
rings successfully suppress frequency bands positioned at
approximately 2fc1. In this case, behaviour specific to fractal

[1] V. G. Veselago, The electrodynamics of substances with

simultaneously negative values of and , Soviet Physics Uspekhi, vol. 10,
pp.509514, 1968.
[2] J. B. Pendry, A. J. Hoden, W. J. Stewart, and I. Youngs, Extremly low
frequency plasmons in metallic mesostructures, Physical Review Letters,
vol. 76, num. 25, pp. 4773-4776, 17 June 1996.
[3] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart,
Magnetism from conductors and enchanced nonlinear phenomena, IEEE
Trans. on Microwave Theory and Techniques, vol. 47, no. 11, pp. 2075-2084,
November 1999.
[4] R. A. Shelby, D. R. Smith, S. Schultz, Experimental verification of a
negative index of refraction, Science, vol. 292, pp. 77-79, 2001.
[5] R. Marqus, J. Martel, F. Mesa, and F. Medina, Left handed media
simulation and transmission of EM waves in sub-wavelength SRR-loaded
metallic waveguides, Phys. Rev. Lett., vol 89, pp. 183901-03, 2002.
[6] F. Martn, F. Falcone, J. Bonache, R. Marqus, and M. Sorolla,
Miniaturized coplanar waveguide stop band filters based on multiple tuned
split ring resonators, IEEE Microwave Wireless Comp. Lett., vol. 13, pp.
511-513, December 2003.
[7] J. Baena, J. Bonache, F. Martin, R. Marqus, F. Falcone, T. Loptigei,
M. Laso, J. Garcia, I. Gil, M. Portillo, and M. Sorrola, Equivalent-circuit
models for split-ring resonators and complementary split-ring resonatos
coupled to planar transmission lines, IEEE Trans. on Microwave Theory and
Techniques, vol. 53, no. 4, pp. 1451-1461, April 2005.
[8] R. Marqus, F. Medina, and R. Idrissi, Role of bianisotropy in negative
permeability and left handed metamaterials, Phys. Rev. B, Condens. Matter,
vol. 65, pp. 144 441144 446, April 2002.
[9] J. Baena, R. Marqus, F. Medina, and J. Martel, Artificial magnetic
metamaterial design by using spiral resonators, Phys. Rev. B, Condens.
Matter, vol. 69, pp. 14 40214 402, January 2004.
[10] F. Falcone, F. Martn, J. Bonache, M. Laso, J. Garca-Garca, J. Baena,
R. Marqus, and M. Sorolla, Stopband and band pass characteristics in
coplanar waveguides coupled to spiral resonators, Microwave Opt. Technol.
Lett., vol. 42, pp. 386388, September 2004.
[11] V. Crnojevi-Bengin, V. Radoni, and B. Jokanovi, Left-handed
microstrip lines with multiple complementary split-ring and spiral
resonators, Microwave Opt. Technol. Lett., vol. 49, no.6, pp.1391-1395, June
2007.
[12] F. Bilotti, A. Toscano, L. Vegni, Design of spiral and multiple split-ring
resonators for the realization of miniaturized metamaterial samples, IEEE
Trans. Antennas Propag., vol. 55, no. 8, pp. 2258-2267, August 2007.
[13] F. Falcone, T. Lopetegi, M. A. G. Laso, J. D. Baena, J. Bonache, M.
Beruete, R. Marqus, F. Martn, and M. Sorolla, Babinet principle applied to
metasurface and metamaterial design, Phys. Rev. Lett., vol. 93, pp. 197
401(1)197 401(4), November 2004.
[14] V. Crnojevi-Bengin, V. Radoni, and B. Jokanovi, Complementary
Split Ring Resonators Using Square Sierpinski Fractal Curves, Proc. Of

11

European Microwave Conference EuMC 2006, paper #1052, EuMCPoster159, September 2006.
[15] G. Peano, "Sur une courbe qui remplit toute une aire plaine," Math.
Ann., 36, 1890.
[16] V. Crnojevi-Bengin, Novel compact microstrip resonators with
multiple 2-D Hilbert fractal curves, Microwave and Optical Technology
Letters, vol. 48, no.2, pp.270-273, February 2006.
[17] V. Crnojevi-Bengin, and . Budimir, Novel 3-D Hilbert microstip
resonators, Microwave and Optical Technology Letters, vol. 46, no. 3, pp
195-197, August 2005.
[18] K.J. Vinoy, K.A. Jose, V.K. Varadan, and V.V. Varadan, Hilbert curve
fractal antenna: A small resonant antenna for VHF/UHF applications,
Microwave Opt. Technol Lett., vol. 29 , pp. 215219. 2001.
[19] J. Anguera, C. Puente, E. Martnez, and E. Rozan, The fractal Hilbert
monopole: A two-dimensional wire, Microwave Opt. Technol Lett., vol. 36,
pp. 102104. 2003.
[20] J. Zhu, A. Hoorfar, and N. Engheta, Bandwidth, cross-polarization, and
feed-point characteristics of matched Hilbert antennas, IEEE Antennas
Wireless Propag Lett., vol. 2, pp. 25, 2003.
[21] J. Zhu, A. Hoorfar, and N. Engheta, Peano antennas, IEEE Antennas
Wireless Propag Lett., vol. 3, pp. 71-74, 2004.
[22] E.A. Parker, and A.N.A. El Sheikh, Convoluted array elements and
reduced size unit cells for frequency-selective surfaces, IEE Proc.
Microwave Antennas Propag, vol. 138, pp. 1922, 1991.
[23] J. McVay, N. Engheta, and A. Hoorfar, High-impedance metamaterial
surfaces using Hilbert-curve inclusions, IEEE Microwave and Wireless
Components Letters, vol. 14, no. 3, pp 130-132, March 2004.
[24] J. McVay, A. Hoorfar, and N. Engheta, Peano high-impedance
surfaces, Radio Sci, 40, RS6S03, 2005.
[25] J. McVay, N. Engheta, and A. Hoorfar, Numerical study and parameter
estimation for double-negative metamaterials with Hilbert-curve inclusions,
Proceedings of 2005 IEEE Antennas Propagation Society International
Symposium, Washington, DC, vol. 2B, pp. 328331, July 2005.
[26] J. McVay, A. Hoorfar, and N. Engheta, Space-filling curve RFID tags,
Proceedings of IEEE Radio and Wireless Symposium, San Diego, CA, pp.
199202, January 2006.
[27] V. Pierro, J. McVay, V. Galdi, A. Hoorfar, N. Engheta, and I. M. Pinto,
Metamaterial inclusions based on grid-graph Hamiltonian paths,
Microwave Opt. Technol. Lett., vol. 48, no. 12, pp. 25202524, December
2006.
Vesna Crnojevic-Bengin (SM96, M06)
received Dipl. Ing. degree in Telecommunications
and Electronics from the Faculty of Technical
Sciences, University of Novi Sad, Serbia in 1994,
and MSc degree from the Faculty of Electrical
Engineering, University of Belgrade, Serbia in
1997. She received PhD degree in electronics and
microwave engineering from the University of
Novi Sad in 2006, where she is currently an
Assistant Professor.
Dr Crnojevic-Bengin is the recipient of the YU
MTT Award for Scientific Contribution in 2005. Her main research interests
include application of fractal curves in the design of microwave passive
devices and metamaterials.

Vasa Radonic (S06) was born in Novi Sad,

Serbia in 1979. He received the B.Sc. and M.Sc.
degrees from the Faculty of Technical Sciences,
University of Novi Sad, Serbia, in 2004 and 2007
respectively, where he is currently working
towards the Ph.D. degree in the field of
metamaterials.
Mr. Radonic is employed as Research Assistant
at the Faculty of Technical Sciences, University of
Novi Sad, Serbia.

Branka Jokanovic (M89) received Dipl. Ing.

degree from the Faculty of Electrical Engineering,
University of Belgrade in 1977, and MSc and PhD
degrees at the same University in 1988 and 1999,
respectively. She was one of the founders of the
Yugoslav IEEE MTT-S Chapter and its
chairperson in the period 1989-2000. She also
initiated foundation of the Yugoslav Association
for Microwave Techniques and Technology in
1992, and publishing of Microwave Review,
Serbian journal of the national importance in
1994, where she served as the editor for six years. B. Jokanovic is currently
Head of Research Department of IMTEL Communications responsible for the
development of a new generation of microwave links based on
nanotechnology. She is the president of Scientific Council of IMTEL
Communications and member of the Advisory Board of RATEL (Republic
Agency for Telecommunications).
Dr Jokanovic is the recipient of the IMTEL Institute Award for Scientific
Contribution in 1996, the IEEE Third Millennium Award in 2000 and YU
MTT Distinguished Service Award in 2005. She has authored or co-authored
about 100 conference, letter and journal papers, and one book. Her current
research interests include novel metamaterials for millimeter-wave and ultrawideband (UWB) applications and nanostructured materials for broadband
wireless systems.