You are on page 1of 5

Progress In Electromagnetics Research Symposium Proceedings, Moscow, Russia, August 1923, 2012 631

Elliptical Split Ring Resonator: Mathematical Analysis, HFSS


Modeling and Genetic Algorithm Optimization
M. Ramaraj, S. Raghavan, Sumanta Bose, and Swadhyaya Kumar
National Institute of Technology, Trichirappalli 620 015, India
Abstract In this paper, a novel equivalent circuit model and mathematical analysis is pro-
posed to estimate the resonant frequency of an Elliptical Split Ring Resonator (ESRR) for a
range of major to minor axis ratios. Performance enhancement is achieved by employing MAT-
LAB based Genetic Algorithm (GA) to optimize the physical parameters of the ESRR by iterative
convergence to the desired resonant frequency. HFSS EM CAD Tool is used to model the op-
timized ESRR & plot its S-parameters for a range of frequency sweep to validate its Negative
Refractive Index Material (NRIM) property.
1. INTRODUCTION
Split Ring Resonators (SRR) are common in the metamaterial design as they exhibit negative
permittivity and permeability for frequencies close to their resonant frequency. It is a major
engineering concern to estimate the resonant frequency due to the existence of negative value of
permeability over a very narrow band of frequency. Analysis of Circular and Square SRR are
commonplace, with relevant literature available [1]. However, the analysis of Elliptical SRR is only
on a simulation level yet [2]; and no substantial mathematical analysis exists, to the best of the
authors knowledge. ESRRs are of special interest as they can be easily incorporated with microstrip
antennas to get highly directional beam patterns because of their enhanced NRIM properties. This
work covers a 3-fold analysis: Mathematical formulation, Performance Optimization & EM CAD
Simulation.
2. RESONANT FREQUENCY: MATHEMATICAL FORMULATIONS
2.1. ESRR Geometry & Mathematical Modeling
Figure 1(a) shows the schematic geometry of a ESRR with dimensions indicated. Fig. 1(b) shows
the equivalent circuit model of the ESRR, forming a L-C network. The inductance is due to the
gap between the rings and the capacitance is due to the rings and the gaps in the rings itself.
When a magnetic eld is applied along the z-axis, an electromotive force appears around the
ESRR which induces a current passing from one ring to the other through the gaps and the structure
behaves like an L-C network having resonant frequency f
o
expressed as:
f
o
=
1
2

L
net
C
net
(1)
where L
net
and C
net
are the net-eective inductance and capacitance of the equivalent L-C network.
(a) (b)
Figure 1: (a) Schematic diagram of a ESRR; (b) Equivalent L-C circuit network.
632 PIERS Proceedings, Moscow, Russia, August 1923, 2012
2.2. Net Capacitance Calculation
Considering an axis through the vertically opposite gaps g
1
and g
2
of the ESRR, we denote the
capacitance of its upper and lower half rings with C
u
and C
l
. Due to zero skew rotation, C
u
must
equal C
l
and is given by:
C
ring
= C
u
= C
l
=
_
l g
2
_
C
pul
(2)
where l is the eective perimeter (Appendix A); g = g
1
= g
2
(assumption) is the gap width; and
C
pul
is the per-unit-length capacitance (Appendix B) of the ESRR. For an accurate estimation
of the resonant frequency, the eect of the gap capacitance is also computed using parallel plate
capacitance approximation. For g = g
1
= g
2
, the gap capacitance C
g1
must equal C
g2
and is given
by :
C
gap
= C
g1
= C
g2
=

o
c h
g
(3)
where c is the width of the conductor and h is the depth of the substrate, as shown in Fig. 1(a).
It is evident that each of the half ring capacitor is in parallel with the gap capacitor and the two
such systems are in series. Hence, the net-equivalent capacitance of the ESRR, C
net
, after some
algebraic manipulations can be expressed as :
C
net
=
C
ring
+ C
gap
2
(4)
2.3. Net Inductance Calculation
Maxwell showed that the self-inductance of a conducting loop is equal to the mutual inductance
between a pair of laments spaced orthogonally apart at a distance termed as the geometric mean
distance (g.m.d.) of the area of the Section [3]. It can be computed using the general Neumann
expression [4],
M =
_
s
_
s

ds ds

cos
r
(5)
where s and s

are the pair of lament loops; r is the distance from the center of any element on
one loop to the center of any other element on the other loop; and is the angle between s and s

lament loops. Cooke showed that the Neumann double integral can be simplied into two sets of
innite series [5], and that the self-inductance of an elliptical loop can be expressed as the innite
series:
L = 8 E () a
_
log
e
16 E () a
h
2 0.4375
2
0.1445
4
. . .
_
(6)
where a is the semi-major axis; h is the g.m.d. of the cross-section; E () is the complete elliptical
integral of the second kind (Appendix B); and =
_
a
2
b
2
_
/
_
a
2
+ b
2
_
where b is the semi-minor
axis of the ESRR. This formula holds good with an error 0.175% for 0 < < 1/3, which is the
case for most practical designs. For > 1/3, other closed form expressions [5] must be considered.
For computational convenience, Basheno deduced a simpler expression to compute the self-
inductance of any closed loop conductor using semi-empirical methods [6]. The self-inductance of
any closed loop conductor is :
L
net
= 0.002l
_
log
e
2l


_
2log
e
l

S
+
_
+

4
_
(7)
where l is the perimeter of the ESRR (Appendix A); and S = ab is the area of the ESRR; is half
of the thickness of the cross-section of the conductor (= c/2 for ESRR); and is the permeability
of the conductor, taken to incorporate the internal linkage of the ux in the cross section of the
conductor. The parameter is a constant which depends on the geometry of the closed loop. In
case of ESRR,

= 0.034, when eccentricity e = 1/

2. For other values of e, the corresponding


is tabulated in [7].
2.4. Resonant Frequency Computation
The resonant frequency f
o
of the ESRR can be computed using eq
n
...
(1) by substituting C
net
and
L
net
from eq
n
...
(4) and eq
n
...
(7) respectively. Clearly, f
o
is a strong function of the geometrical and
electrical parameters.
Progress In Electromagnetics Research Symposium Proceedings, Moscow, Russia, August 1923, 2012 633
3. GENETIC ALGORITHM OPTIMIZATION
3.1. Genetic Algorithms Steps
Step 1: Create an initial population of all the parameter forming a binary sequence known as
chromosomes.
Step 2: Rank the chromosomes according to their tness using tness function and select the
best few.
Step 3: Perform Cross-Breeding and Mutation to get the next generation ospring with new
characteristics.
Step 4: Repeat Step 2 and Step 3 for a number of generations to get the overall best candidate
chromosomes.
Step 5: Extract the optimal valued chromosomes and plot them to get the least error and
best candidates.
It should be noted that the initial population must be large enough so that there is no domination
of gene. This ensures satisfactory convergence to obtain the optimal values.
3.2. Implementation of Genetic Algorithm for ESRR Optimization
The Genetic Algorithm optimization of the physical parameters of the ESRR is carried out in
MATLAB, with the numerical data tabulated in Table 1.
3.3. Genetic Algorithm Optimized Results
Genetic Algorithm optimization of ESRR in MATLAB yields the optimized parameter values, that
are tabulated in Table 2. They are best suited to match the desired frequency and minimize the
error.
The optimized population swarm of the 4 parameters viz. a, b, c & d for minimum relative
error between the desired and the calculated frequency are plotted in Figure 2.
Table 1: Numerical range and value of parameters for MATLAB simulation.
Parameter Range/Value Parameter Range/Value
Major Axis 5 mm < 2a < 11 mm 1 (FR4 Substrate)
Minor Axis 2 mm < 2b < 8 mm
r
4 (FR4 Substrate)
Conductor Width 0.01 mm < c < 0.2 mm Desired Frequency 22 GHz
Conductor Spacing 0.01 mm < d < 0.3 mm % Crossover 45%
Split Gap g = 0.3 mm % Mutation 4%
Substrate Depth h = 0.5 mm Total Population 800
Table 2: GA optimized parameter values for best performance of ESRR.
Parameter Best Value Parameter Best Value
Major Axis 2a = 5 mm Conductor Width c = 0.2 mm
Minor Axis 2b = 2 mm Conductor Spacing d = 0.3 mm
634 PIERS Proceedings, Moscow, Russia, August 1923, 2012
Figure 2: Plot of relative error in frequency vs. range of a, b, c & d.
Figure 3: Scattering parameters (S
11
& S
21
) of ESRR modelled in HFSS over a frequency sweep of 050 GHz.
4. HFSS MODELING OF ESRR: NRIM PROPERTY VALIDATION
The ESRR has been modeled in Finite Element Method (FEM) based Electromagnetic (EM) CAD
Tool HFSS (High Frequency Structure Simulator) using the optimized physical parameter values of
Table 2. The Scattering Parameter plot of the ESRR generated using HFSS is shown in Figure 3.
The crossovers of S
11
& S
21
at multiple frequency bands clearly indicate the enhanced NRIM
property of the ESRR [2].
5. RESULTS & CONCLUSION
The Elliptical Split Ring Resonator (ESSR) geometry was discussed, followed by a detailed mathe-
matical analysis by considering its geometrical dimensions and electrical specications such as the
and
r
of the substrate in itself. A novel closed form expression for the resonant frequency was
derived considering all the associated and aecting parameters. It is noteworthy that a designer
can start with the Circular SRR and then deviate the semi-major axis, a and the semi-minor axis,
b away from the radius value of the CSRR. This gives the designer 2 tweaking parameters, viz.
a & b, to accurately design an ESRR resonating at certain desired resonant frequency. Varying
the a & b essentially varies the C
net
(due to varying l in C
ring
Equation (2)), and also L
net
(due to varying l & S in L
net
Equation (7)); but there are no changes in C
gap
and C
pul
as
these do not depend on a & b. However these may be varied by changing c, d, g & h.
The optimum geometrical dimensions of the ESRR were estimated for a desired resonant fre-
quency. MATLAB based Genetic Algorithm (GA) Optimization was used to successfully converge
the values of geometrical dimensions of the ESRR for a particular desired resonant frequency value
by minimizing the frequency estimation error. The optimum geometrical dimensions value has been
tabulated with remarks.
The CAD model of the ESRR was simulated on a commercially available electromagnetic sim-
ulator, High Frequency Structure Simulator (HFSS) based on the Finite Element Method elec-
Progress In Electromagnetics Research Symposium Proceedings, Moscow, Russia, August 1923, 2012 635
tromagnetic mode solver. The S-parameters (S
11
& S
21
) of the ESRR were plotted for a range
of frequency sweep. The cross-cuttings in the S
11
& S
21
plot give the indication of the NRIM
property of the ESRR. An ESRR loaded microstrip antenna can be used eciently for wireless
communication in the microwave regime due to its NRIM property.
APPENDIX A
The perimeter l of an ellipse with semi-major axis a and semi-minor axis b can be expressed
as an innite series, called the Gauss-Kummer Formula, utilizing the hypergeometric function F ()
where:
l = (a + b) F
_

1
2
,
1
2
; 1;
_
a b
a + b
__
= (a + b)
_

n=0
_
1/2
n
_

_
a b
a + b
_
2n
APPENDIX B
The per-unit-length capacitance C
pul
=
o

r
+1
2
_

E(

1
2
)
E()
, where =
d
d+2c
and E () is the
complete elliptical integral of the second kind, dened as E (k) =
_
/2
0
_
1 (ksin )
2
d.
REFERENCES
1. Noginov, M. A. and V. A. Podolskiy, Tutorials in Metamaterials, CRC Press, May 2010.
2. Sharma, V., A microstrip metamaterial split ring resonator, Int. J. of Physical Sciences,
Vol. 6, 2011.
3. Maxwell, J. C., A Treatise on Electricity and Magnetism, Clarendon Press, Oxford, 1881.
4. Jeans, S. J. H., The Mathematical Theory of Electricity and Magnetism, Cambridge University
Press, 1925.
5. Cooke, N., Self-inductance of the elliptical loop, Proc. of the Inst. of Elec. Engrs., Vol. 110,
No. 7, 1963.
6. Basheno, V. J., Abbreviated methods for calculating the inductance of irregular plane poly-
gons of round wire, Proc. Inst. Radio Engrs., 1927.
7. Grover, F., Inductance Calculations: Working Formulas and Tables, Dover Publications, New
York, 1946.