IET Microwaves, Antennas & Propagation

Research Article

BWDO algorithm and its application in ISSN 1751-8725

Received on 6th March 2018
Revised 26th March 2019
antenna array and pixelated metasurface Accepted on 1st April 2019
E-First on 10th May 2019
synthesis doi: 10.1049/iet-map.2018.5051
www.ietdl.org

Prakash Ranjan1, Santosh Kumar Mahto2 , Arvind Choubey1

1Department of Electronics and Communication Engineering, National Institute of Technology Jamshedpur, Jamshedpur, India
2Department of Electronics and Communication Engineering, Indian Institute of Information Technology Ranchi at NIT Jamshedpur,
Jamshedpur, India
E-mail: [email protected]

Abstract: This article introduces a novel, binary version of wind driven optimisation (WDO) called binary WDO (BWDO), which
can be specifically used for solving binary valued problems such as antenna array and metasurface synthesis. Validation of
proposed algorithm is done with standard benchmark functions. To demonstrate the efficiency of BWDO, examples of thinned
antenna array and metasurface synthesis are considered. Thinned antenna array problem is solved by formulating single and
multi-objective optimisation problems which achieve minimum SLL of −22.30 dB and −20.28 dB respectively, as compared to
−13.2 dB by periodic array. Further, the proposed BWDO algorithm is interface with Ansys-HFSS and three design examples of
metasurfaces such as MPA and AMC are further considered to exemplify the proficiency of BWDO. Simulation of optimised
MPA exhibits five discrete bands of high absorptions at 8.7, 9.7, 14.6, 15.3 and 16.5 GHz with simulation reflection co-efficient
value of −11.46, −10.00, −12.17, −16.19 and −20.58 dB which is less than −10 dB reflection under normal incidence of wave.
Synthesised single and dual band AMC has in-phase reflection bandwidth of 0.58 and 0.34, 0.16 GHz at central frequency of
5.21 and 5.07, 6.49 GHz respectively. The measured results almost matched with the respective simulated results with slight
variation due to fabrication tolerance.

1 Introduction This paper introduces a modified optimisation method called

binary wind driven optimisation (BWDO) algorithm. WDO
Nature-inspired optimisation techniques are widely used in the algorithm is inspired by the movement of wind in the earth
field of electromagnetics, ranging from synthesis of the antenna atmosphere and employs a continuous-valued vector representation
array to metamaterial-based various devices. In antenna array of solution vector [11, 12]. Hence, the standard WDO may not be
synthesis, in order to minimise the side-lobe level (SLL), the array efficient in solving many electromagnetic problems, where a binary
is made aperiodic by altering the position and excitation amplitude string is used to represent the solution vector (e.g. synthesis of a
[1–5]. This antenna array thinning is a complex problem and few pixelated unit cell of metasurface). To overcome this limitation,
analytical methods [6], numerical algorithm [7] and probabilistic two suitable modifications are done in original WDO which lead to
approach [8] have been reported to obtain lower peaks of the SLL. the proposed novel BWDO. The proposed BWDO algorithm is
Also, soft computing optimisation algorithms such as genetic validated with six standard benchmark optimisation functions
algorithm (GA) [9] and particle swarm optimisation (PSO) [4] are which prove its robustness in solving optimisation problems.
employed recently for the same and their reported results Furthermore, the single and multi-objective array thinning design
demonstrate improved performance compared with conventional problems have been simulated. The radiation patterns of single and
methods [10–12]. multi-objective optimal array designs obtained using BWDO are
In addition to antenna array synthesis techniques, applications 76 and 63% filled aperture and achieve minimum SLL of −22.30
of various optimisation algorithms for the synthesis of and −20.28 dB, respectively.
metamaterials based devices are noteworthy [12, 13]. Various Synthesis problems of metasurfaces such as MPA and AMC are
metamaterial structures that have been successfully optimised solved by interfacing the BWDO algorithm running in MATLAB
include metamaterial perfect absorbers (MPAs) [14], artificial with Ansoft-HFSS which is a commercially available finite
magnetic conductors (AMCs) [15] and many more applications element solver. Three design examples such as five-band MPA,
covering a wide range of frequencies [16]. MPA and AMC are single and dual-band AMCs are synthesised. The simulation results
artificial composite materials having interesting properties of demonstrate the efficiency of BWDO-HFSS interface techniques in
nearly perfect absorption and in-phase reflection for incident of order to synthesise various unit-cell structures using pixelated
waves, respectively [17]. Being a planar periodic structure of its design strategy. Three design examples of metasurfaces such as
unit cell, frequency selective surfaces (FSSs) have developed the MPA and AMC are further considered to exemplify the proficiency
keen interest of research for synthesising MPA and AMC. In of BWDO. Simulation with a numerical simulator of optimised
general, two methods are widely exercised for the synthesis of MPA exhibits five discrete bands of high absorptions at 8.7, 9.7,
FSS. The first method is based on ‘geometry tailoring’, which is 14.6, 15.3 and 16.5 GHz with simulation reflection coefficient
expensive as well as time-consuming because of the fact that it also values of −11.46, −10.00, −12.17, −16.19 and −20.58 dB under
requires human experience and intuition with immense simulation normal incidence of the wave. Synthesised single and dual-band
experiments [18, 19]. The second method utilised optimisation AMCs have in-phase reflection bandwidths of 0.58 and 0.34, 0.16
techniques for the synthesis of unit-cell structure in the form of GHz at the central frequencies of 5.21 and 5.07, 6.49 GHz. The
‘pixelated unit cell’ [20, 21]. With the progress of very large synthesised metasurfaces are fabricated using commercially
computing capacity of the computer, the pixelated unit-cell available FR4 substrate. Measured results almost matched with the
structure provides the versatility synthesising desired FSS but it respective simulated results with slight variation due to fabrication
requires binary optimisation technique. tolerance.

IET Microw. Antennas Propag., 2019, Vol. 13 Iss. 9, pp. 1263-1270 1263
© The Institution of Engineering and Technology 2019
 17518733, 2019, 9, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-map.2018.5051, Wiley Online Library on [20/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Remaining paper is organised as: BWDO algorithm is where rmn, t is defined as the threshold value. The probability under
introduced in Section 2. Validation of BWDO using various the threshold value that the nth bit becomes equals to ‘1’ and ‘0’
benchmark functions is shown in Section 3. The simulation results binary values, respectively, is
for the syntheses of single and multi-objective antenna array
syntheses using BWDO is discussed manifestly in Section 4. P xmn, t = 0 = 1 − S V mn, t = S −V mn, t . (7)
Section 5 illustrates each of the synthesised models of MPA and
AMC separately. This section also discusses the obtained and
simulation results along with their corresponding measured results.
Conclusion and future scope have been described in Section 6. P xmn, t = 1 = S V mn, t (8)

2 BWDO algorithm It is perceived from (4) and (6) that single bit binary will have a
BWDO is a suitable modification of a new nature-inspired, higher probability of flipping its binary state if it is not equal to the
heuristics, global evolutionary optimisation technique based on the corresponding bit in the global best.
atmospheric motion of air particle called WDO algorithm which is For instance, V mn, t − 1 is the velocity of mth air parcel at the (t –
described in [11]. The velocity of the air packet is given as 1)th iteration. Assuming its nth bit and global best are ‘0’ and ‘1’,
respectively. The velocity of air parcel of the iteration can be
Vmn, t = 1 − α Vmn, t − 1 − g xmn, t−1
calculated as
1 1
+ RT − 1 xopt mn, t − 1 − xmn, t − 1 V mn, t = 1 − α V mn, t − 1 − g × 0 + RT −1 × 1−0
i (1) i
otherdim c × V mn, t − 1 (9)
c Vcur + > V mn, t − 1 .
+ i
i

where Vmn, t − 1, Vmn, t denote initial velocity and new velocity, Since the intermediate variable S x is monotonically increasing
functions, we have
respectively. The position vector can be expressed by
S V mn, t > S(V mn, t − 1), (10)
xmn, t − 1 = xm1, t − 1 , xm2, t − 1, …, xmn, t − 1 (2)
It means that the nth bit of air parcel position tends to be its ‘1’
where xmn, t − 1 ∈ 0, 1 , ‘m’ represents the position of air parcels
binary state with a greater probability of S V mn, t and the
and ‘n’ represents binary bit strings.
The Hamming distance of two given lines of code is the number probability that the bit gets its ‘0’ binary state is reduced by
of points at which the lines’ binary code values are different. S V mn, t − S(V mn, t − 1).
Hence, Hamming value for a single bit binary value can have At the tth iteration, if the bit is not flipped, the velocity
three possible values increases at the (t + 1)th iteration to flip the bit with a greater
probability.
xoptmn, t − 1 − xmn, t − 1 The probability of the nth bit to be flipped into 1 with its
velocity V max is given as
1, if xoptmn, t − 1 = 1, xmn, t − 1 = 0
(3)
0, if xoptmn, t−1 = xmn, t − 1 1
P xmn, t = 1 = = 1, (11)
−1, if xoptmn, t−1 = 0, xmn, t − 1 = 1 1 + e− Vmax

The value of velocity V mn, t is a continuous number in the range of and the probability that it remains ‘0’ is
[ − V max, V max], whereas the value of the position of an air parcel is
P xmn, t = 0 = 1 − P xmn, t = 1
a binary number.
In BWDO, the position of air parcel (xmn, t) can take binary 1 (12)
=1− = 0.
values ‘0’ or ‘1’ and V mn, t reflects its possibility. The intermediate 1 + e− Vmax
variable S(V mn, t) is sigmoid limiting transformation (increasing
monotonically) and expressed as Now, let the obtained velocity be V mn, t = 6 < V max, the maximum
and minimum probabilities that the bit flipped to ‘1’ and remains
1 ‘0’ are 0.9975 and 0.0025, respectively. One of the advantages of
S(V mn, t) = (4)
1 + exp( − V mn, t) BWDO is that no boundary handling techniques are required since
the air parcel position xmn, t is always binary valued in the space
within the limitation of S V mn, t ∈ − −V max, V max and expressed domain.
as
3 Validation with standard benchmark functions
1
= 0, if V mn, t → − V max The BWDO algorithm is validated with six standard benchmark
1 + exp V mn, t
functions (unimodal and multimodal) and satisfactory results have
1 been attained. Here, 20 bits of a binary number is used to represent
S V mn, t = , if V mn, t = 0 (5)
2 a continuous-valued number, and thus a continuous-valued
1 problem becomes converted into a binary one. The simulation
= 1, if V mn, t → V max
1 + exp −V mn, t study has manifested that the BWDO algorithm discovers the
optimum values f x* at its best solution vector (x*) of each
The value of S V mn, t is evaluated by (5). Consider the mth air function as mentioned in Table 1. The learning characteristics are
parcel xmn, t, the position of the nth bit is upgraded as shown in Fig. 1.
It can be seen that the BWDO algorithm effectively determines
0 if rmn, t ≥ S V mn, t the solution of standard benchmark functions with minimum
xmn, t = (6) computational cost. The parameters used for BWDO algorithm are
1 if rmn, t < S V mn, t . the same as given in [11]. In the next section, thinned antenna array
problem is solved by formulating single and multi-objective

1264 IET Microw. Antennas Propag., 2019, Vol. 13 Iss. 9, pp. 1263-1270
© The Institution of Engineering and Technology 2019
 17518733, 2019, 9, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-map.2018.5051, Wiley Online Library on [20/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 1 Best solution vector (x*) of each standard benchmark function and optimum values f x*
Sl. no. Function name (description) Description and properties
1. De Jong's function (unimodal) n
f x = ∑ xi − 5 2

i=1

−5.12 ≤ xi ≤ 5.12,
f x* = 0, x* = 5, 5, …, 5
2. axis parallel hyper-ellipsoid function (unimodal) n
f x = ∑ i . xi2
i=1

−5.12 ≤ xi ≤ 5.12,
f x* = 0, x* = 0, 0, …, 0
3. rotated hyper-ellipsoid function (unimodal) n i
f x = ∑ ∑ x2j
i=1 j=1

−65.53612 ≤ xi ≤ 65.536,
f x* = 0, x* = 0, 0, …, 0
4. Rastrigin's function (multimodal) n
f x = 10n + ∑ xi2 − 10 cos 2πxi
i=1

−5.12 ≤ xi ≤ 5.12,
f x* = 0, x* = 0, 0, …, 0
5. Ackley's function (multimodal) 1
n

n i∑
f x = − a . exp −b . xi2
=1
n
1
n i∑
−exp cos cxi + exp 1
=1

a = 20, b = 0.2, c = 2π
−32.768 ≤ xi ≤ 32.768,
f x* = 0, x* = 0, 0, …, 0
6. Schwefel's function (multimodal) n
f x = ∑ −xi sin xi
i=1

−500 ≤ xi ≤ 500,
f x* = − 418.9829, xi = 420.9687, 420.97, …, 420.987

optimisation problems which demonstrate the efficiency of the

proposed BWDO algorithm.

4 Linear antenna array synthesis using BWDO

The design of thinned array with maximum SLL suppression is a
discrete complex problem. The proposed BWDO is used to design
100-element thinned array having minimum SLL and results are
compared with those obtained by GA in [9] and binary PSO in [4].
The 100-element linear array is discredited by a uniform grid of
λ/2. It is also assumed that the array is symmetrical and their
elements are isotropic in nature.
The array factor of such an array is given as
N
2π
AF θ = 2 ∑ xi cos d cos θ + ∅i , (13)
i=1
λ i
Fig. 1 Convergence graph of fitness value versus number of iterations
where λ and di are the wavelength and Euclidean distance to phase obtained using BWDO algorithm for different benchmark functions
reference, respectively. Here, ∅i and xi represent the phase and the
where Δθi = θui − θli and [θli, θui] is the region of SLL suppression
excitation amplitude of the ith antenna element, respectively. This
antenna array problem can be solved by formulating a single or and M is the number of suppressed SLL.
multi-objective optimisation problem, where the optimisation goal Initial parameters play an important role in the rate of
is to attain maximum SLL suppression. convergence for solving complex multidimensional and
multimodal problems such as array synthesis [11, 22] and array
thinning as reported in [4, 9].
4.1 Single-objective thinned array design
In this problem, the excitation amplitudes of array elements are
To achieve the maximum SLL suppression, the fitness function is randomly initialised to ‘0’ or ‘1’ with equal probability. The
given as optimal design configuration obtained at the 98th iteration is given
as: { xi} = {1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1
M θui 111111001111110111111011010111011110
pressure I =
i=1
1
∑ Δθ ∫
i θli
AF 2 dθ (14) 1 0 1 1 1 0 1 0 1 1 0 1 0 1 0 1 0 0 1 0 0 1 0 1 1 1 0 0 1 1 0 0 1 1},
where ‘1’ and ‘0’ represent that an element is turned on and off,

IET Microw. Antennas Propag., 2019, Vol. 13 Iss. 9, pp. 1263-1270 1265
© The Institution of Engineering and Technology 2019
 17518733, 2019, 9, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-map.2018.5051, Wiley Online Library on [20/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 4 Convergence characteristics of 100-element thinned array design
Fig. 2 Normalised pattern of the 100-element aperiodic array antenna
Table 2 Performance comparisons of single and multi-
objective thinned array designs
Starting value, Converging value, Iterations
dB dB
single objective 0.728 0.228 98
multi-objective 0.869 0.117 105

determine the trade-off between the peak of SLL and the number of
elements to achieve the desired SLL in a thinned array design
problem. This is a multi-objective optimisation problem and two
pressure functions as given in (15) are used to determine the
minimum SLL and the element number, respectively.
The optimal design configuration obtained at the 105th iteration
is given as: { xi} = {1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 0 1 1 1
111111011111001110011011010100100110
011010101010000010100110110101010100
0 1}
M θui
1
pressure II = ∑ Δθ i
∫ AF 2 dθ, and
i=1 θli

100
pressure III = 2 ∑ xi (15)
i=1

The radiation pattern of the optimal array design obtained using the
BWDO (63% filled aperture) is compared with uniformly spaced
(λ/2) periodic array (100% filled aperture), and BPSO (66% filled
aperture) in Fig. 3a along with its zoomed radiation patterns with
elevation angles from 85° to 95° in Fig. 3b. The BWDO has
achieved minimum SLL of −20.28 dB as compared with −13.2 dB
by the periodic array. The simulation results demonstrate improved
performance of BWDO as compared with the periodic array. The
Fig. 3 Radiation patterns of 100-element aperiodic array antenna near field in SLL obtained by BWDO is −40 dB, which is better
optimised thinned array with minimum SLL and minimum number of than that obtained by the GA and BPSO. However, a far field in
elements SLL obtained in this case is almost equal to the SLL obtained by
(a) Radiation pattern with elevation angles from 0° to 180°, (b) Zoomed radiation GA in [9] and BPSO in [4].
patterns with elevation angles from 85° to 95° The convergence characteristics of single and multi-objective
thinned array designs obtained using the proposed BWDO
respectively. It can be seen that more elements are turned on near algorithm is shown in Fig. 4 and summarised in Table 2. The
the centre as compared with the edge. simulation results demonstrate the improved performance of the
The radiation pattern of the optimal array design obtained by BWDO as compared with GA and PSO.
the BWDO (76% filled aperture) along with uniformly spaced Thus, the proposed BWDO algorithm was successfully
(λ/2) periodic array (100% filled aperture) and BPSO (77% filled implemented for single and multi-objective thinned array syntheses
aperture) is shown in Fig. 2. The BWDO-based method achieves problems. In the next section, we will illustrate three examples of
minimum SLL of −22.30 dB as compared with −13.2 dB by the metasurface syntheses using pixelated FSS.
periodic array. However, SLL obtained, in this case, is slightly less
than that observed by GA in [9] and BPSO in [4]. 5 Metasurface synthesis
4.2 Multi-objective thinned array design The proposed BWDO is interfaced with electromagnetic finite
element method (FEM) solver for development of the unit-cell
In array thinning problem, the SLL increases while turning off too structure of metasurface. This is carried out through a unique
many array elements. In this example, BWDO is employed to

1266 IET Microw. Antennas Propag., 2019, Vol. 13 Iss. 9, pp. 1263-1270
© The Institution of Engineering and Technology 2019
 17518733, 2019, 9, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-map.2018.5051, Wiley Online Library on [20/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 6 Experimental setup
(a) Experimental setup outline, (b) Anechoic chamber with the experimental setup

optimisation technique. Being eight-fold symmetry, the structure

becomes polarisation insensitive and also reduces the optimisation
problem from 256 dimensions to 36 only.
In general, some of the authors utilise optimisation techniques
through the pixelated unit-cell structure to generate FSS structures
that have taken no note of the fact that the practical realisation of
two units of square pixels touched diagonally cannot be done [12,
20, 23].
Toward this end, this problem has been encountered by
removing a very small part of copper at any such cases and also
included the same in the simulation. The detailed description of the
experimental setup followed by the proposed metasurface is given
below.

5.2 Experimental setup

All the proposed metasurfaces are designed with easily accessible
substrate FR4 (thickness 1.6 mm and relative permittivity 4.4) and
fabricated using printed circuit board technique. Each of the unit
cells has dimensions of 16 mm × 16 mm × 1.67 mm. An array of a
minimum of 12 × 12 unit cells specimen has been built for each of
the proposed metasurfaces. The reflected S-parameter is measured
within the anechoic chamber using two horn antennas as shown in
Fig. 6. These horn antennas are connected to a vector network
analyser (Anritsu VNA Master MS2038C, 5 kHz–20 GHz) to get
measured S-parameter.

5.3 Proposed metasurfaces

Two examples of metasurface (five-band MPA and AMCs)
synthesis are used to illustrate the BWDO efficiency.

5.3.1 Five-band MPA: MPA consists of two metallic layers

Fig. 5 Distinctive scheme for the synthesis of metasurfaces separated by a dielectric substrate. The top layer is a developed
(a) Flowchart of metasurface synthesis, (b) Unit cell pixelated into 16 × 16 mini grids periodic of its unit-cell structure, whereas bottom layer is
with eight-fold symmetry completely covered with copper ground. Any incident power
cannot transmit through due to the entirely covered copper bottom
method of linking MATLAB with Ansys-HFSS which is a layer.
commercially available FEM solver. Fig. 7a manifests the synthesised unit-cell geometry of MPA
which has five bands of powerful reflection dips at 8.7, 9.7, 14.6,
5.1 Design methodology 15.3 and 16.5 GHz with simulation reflection coefficient values of
−11.46, −10.00, −12.17, −16.19 and −20.58 dB. Fig. 7b shows the
Flowchart for the synthesis of unit cell of metasurfaces is shown in fabricated model of the proposed five-band MPA. Fig. 7c shows
Fig. 5a. During the course of optimisation, BWDO which runs in simulation result along with measured results. The measured result
MATLAB uses Ansys-HFSS to calculate the cost function of each has reflection dips at 8.8, 9.4, 14.4, 15.2 and 16.4 GHz with
generated structure. Synthesis of the desired unit-cell structure of reflection coefficient values of −9.97, −9.23, −11.20, −10.40 and
metamaterial absorber can be done by dividing the whole unit-cell −16.06 dB. Slight discrepancy between simulation and
area into 256 units of small square-shaped pixels as shown in measurement results occur due to fabrication tolerance.
Fig. 5. Each pixel represents the presence of copper with its value Performance comparison of the proposed absorber is done in
‘1’. The dimension of each pixel is taken as 1 mm×1 mm. Hence, Table 3 with already reported multiband absorbers. It depicts that
pixelated metamaterial absorber synthesis can be viewed as a the proposed absorber has better design methodology with
binary optimisation problem which can be solved by using BWDO comparative unit-cell size but the highest thickness.
IET Microw. Antennas Propag., 2019, Vol. 13 Iss. 9, pp. 1263-1270 1267
© The Institution of Engineering and Technology 2019
 17518733, 2019, 9, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-map.2018.5051, Wiley Online Library on [20/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 7 Five-band MPA
(a) Unit-cell geometry, (b) Fabricated model, (c) Simulated and measured results

Table 3 Comparison of the proposed five-band MA with reported multiband absorbers

Ref. Number of absorption Design methodology Unit-cell size, Relative unit-cell Thickness, mm Relative thickness
peaks mm size
[24] 2 hit and trial 10 0.51λ 0.5 0.025λ
[25] 3 hit and trial 18 1.425λ 1 0.079λ
[26] 3 hit and trial 9.87 0.393λ 1 0.039λ
[27] 4 hit and trial 23 0.702λ 1 0.031λ
[28] 4 hit and trial 20 0.661λ 1.1 0.037λ
[29] 4 hit and trial 10 0.78λ 1.5 0.065λ
[30] 5 hit and trial 20 1.039λ 1.5 0.078λ
proposed work 5 in pixel form, BWDO 16 0.88λ 3.2 0.176λ
λ refers to the wavelength of the highest absorption frequency.

5.3.2 Artificial magnetic conductors: In the design of AMC, the bandwidth of 0.4 GHz which approximately matched with the
spacing between two unit cells is kept at 0.2 mm. AMC consists of corresponding simulated result.
two metallic layers separated by a dielectric substrate. The top Fig. 9a manifests the developed unit-cell geometry of dual-band
layer is a developed periodic metallic unit-cell structure, whereas AMC. Its fabricated model is shown in Fig. 9b. The simulation
bottom layer is copper ground. Also, as explained earlier, at the shows that it is having zero-crossing phases at 5.07 and 6.49 GHz
point of two diagonally connected copper pixels, a small square- with bandwidths of 0.34 and 0.16 GHz. Fig. 9c shows simulation
shaped copper (0.25 mm × 0.25 mm) is removed and the same has results along with measured results of the same. The measured
been incorporated in the simulation. Two examples of AMC (single result shows that it has zero-crossing phases at 5.15 and 6.55 GHz
and dual band) synthesis are also illustrated using the proposed with bandwidths of 0.2 and 0.19 GHz which approximately
BWDO to show the efficiency of the algorithm. matched with the corresponding simulated results.
Fig. 8a manifests the developed unit-cell geometry of single- All the above three examples of metasurface synthesis show
band AMC. Its fabricated model is shown in Fig. 8b. The that BWDO algorithm can be effectively used to synthesise various
simulation shows that it is having zero-crossing phase at 5.21 GHz types of metasurfaces.
with a bandwidth of 0.58 GHz. Fig. 8c shows simulation result
along with measured results of the same. The measured result
shows that it has a zero-crossing phase at 5.3 GHz with a

1268 IET Microw. Antennas Propag., 2019, Vol. 13 Iss. 9, pp. 1263-1270
© The Institution of Engineering and Technology 2019
17518733, 2019, 9, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-map.2018.5051, Wiley Online Library on [20/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

1269
IET Microw. Antennas Propag., 2019, Vol. 13 Iss. 9, pp. 1263-1270
(a) Unit-cell geometry, (b) Fabricated model, (c) Simulated and measured results

(a) Unit-cell geometry, (b) Fabricated model, (c) Simulated and measured results

© The Institution of Engineering and Technology 2019

Fig. 8 Single-band AMC

Fig. 9 Dual-band AMC

Conclusion
6
 17518733, 2019, 9, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-map.2018.5051, Wiley Online Library on [20/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
The binary version of WDO is proposed and its validation is done [13] Chen, P.Y.: ‘Synthesis design of artificial magnetic metamaterials using a
genetic algorithm’, Opt. Express, 2008, 16, (17), pp. 12806–12818
with standard benchmark functions. Thinned array design has been [14] Gingrich, M.A., Werner, D.H.: ‘Synthesis of low/zero index of refraction
formulated as single as well as multi-objective optimisation metamaterials from frequency selective surfaces using genetic algorithms’,
problems and has been effectively solved using this optimisation Electron. Lett., 2005, 41, (23), pp. 1266–1267
algorithm. Compared to GA, PSO and conventional-based [15] Kern, D.J., Werner, D.H.: ‘A genetic algorithm approach to the design of
ultra-thin electromagnetic band gap absorbers’, Microw. Opt. Technol. Lett.,
approaches, the obtained simulation results exhibit improved SLL 2003, 38, (1), pp. 61–64
suppression with a reduced number of array elements. To [16] Ge, Y., Esselle, K.P.: ‘GA/FDTD technique for the design and optimisation of
exemplify the proficiency of BWDO, three different unit-cell periodic metamaterials’, IET Microw. Antennas Propag., 2007, 1, (1), pp.
structures of metasurface (one MPA and two AMCs) are 158–164
[17] Bossard, J.A., Werner, D.H.: ‘Meta-materials with custom emissivity
synthesised using a technique of interfacing MATLAB with Ansys- polarization in the near-infrared’, Opt. Express, 2013, 21, (3), pp. 3872–3884
HFSS. Synthesised unit-cell structures are fabricated. Measured [18] Vardaxoglou, J., Yiannis, C.: ‘Metamaterial arrays and applications: FSS,
results approximately matched with the respective simulated EBG & AMC structures’. Int. Workshop on Antenna Technology: Small
results. Antennas, Novel EM Structures and Materials, and Applications (iWAT),
Sydney, NSW, 2014, p. 269
[19] Zhao, M., Yu, X., Wang, Q., et al.: ‘Novel absorber based on pixelated
7 References frequency selective surface using estimation of distribution algorithm’, IEEE
Antennas Wirel. Propag. Lett., 2015, 14, pp. 1467–1470
[1] Harrington, R.: ‘Side lobe reduction by nonuniform element spacing’, IEEE [20] Song, Y.C., Ding, J., Guo, C.J.: ‘A semi-analytical numerical method for fast
Trans. Antennas Propag., 1961, 9, (2), pp. 187–192 metamaterial absorber design’, AIP Adv., 2015, 5, (9), pp. 097–108
[2] Ishimaru, A.: ‘Theory of unequally spaced arrays’, IEEE Trans. Antennas [21] Zhao, Y., Chen, F., Chen, H., et al.: ‘The microstructure design optimization
Propag., 1962, 10, (6), pp. 691–702 of negative index metamaterials using genetic algorithm’, Prog. Electromagn.
[3] Kurup, D.G., Himdi, M., Rydberg, A.: ‘Synthesis of uniform amplitude Res., 2011, 22, pp. 95–108
unequally spaced antenna arrays using the differential evolution algorithm’, [22] Mahto, S.K., Choubey, A.: ‘A novel hybrid IWO/WDO algorithm for
IEEE Trans. Antennas Propag., 2003, 51, (9), pp. 2210–2217 interference minimization of uniformly excited linear sparse array by
[4] Jin, N., Rahmat-Samii, Y.: ‘Advances in particle swarm optimization for position-only control’, IEEE Antennas Wirel. Propag. Lett., 2015, 59, pp.
antenna designs: real-number binary single-objective and multi-objective 250–254
implementations’, IEEE Trans. Antennas Propag., 2007, 55, (3), pp. 556–567 [23] Wang, L., Wang, T., Nie, Y., et al.: ‘Synthesis design of metamaterial
[5] Kumar, V., Sahoo, A.K.: ‘Grating lobe and side lobe suppression using multi- absorbers using a genetic algorithm’. IEEE Int. Symp. Signals Systems and
objective optimization techniques’. IEEE Int. Conf. Communications and Electronics (ISSSE), 2010, pp. 1–4
Signal Processing (ICCSP), 2015, pp. 0247–0251 [24] Luo, H., Cheng, Y.Z.: ‘Ultra-Thin dual-band polarization-insensitive and
[6] Bevelacqua, P.J., Balanis, C.A.: ‘Minimum side lobe levels for linear arrays’, wide-angle perfect meta-material absorber based on a single circular sector
IEEE Trans. Antennas Propag., 2007, 55, (12), pp. 3442–3449 resonator structure’, J. Electron. Mater., 2018, 47, (1), pp. 323–328
[7] Lo, Y., Lee, S.: ‘A study of space-tapered arrays’, IEEE Trans. Antennas [25] Bhattacharyya, S., Ghosh, S., Srivastava, V. K.: ‘Triple band polarization-
Propag., 1966, 14, (1), pp. 22–30 independent metamaterial absorber with bandwidth enhancement at X-band’,
[8] Oliveri, G., Massa, A.: ‘Bayesian compressive sampling for pattern synthesis J. Appl. Phys., 2013, 114, (9), p. 094514
with maximally sparse non-uniform linear arrays’, IEEE Trans. Antennas [26] Asgharian, R., Zakeri, B., Karimi, O.: ‘Modified hexagonal triple-band
Propag., 2011, 59, (2), pp. 467–481 metamaterial absorber with wide-angle stability’, AEU-Int. J. Electron.
[9] Haupt, R.L.: ‘Thinned arrays using genetic algorithms’, IEEE Trans. Commun., 2018, 87, pp. 119–123
Antennas Propag., 1994, 42, (7), pp. 993–999 [27] Chaurasiya, D., Ghosh, S., Bhattacharyya, S., et al.: ‘An ultrathin quad-band
[10] Mahto, S.K., Choubey, A.: ‘A novel hybrid IWO/WDO algorithm for nulling polarization-insensitive wide-angle metamaterial absorber’, Microw. Opt.
pattern synthesis of uniformly spaced linear and non-uniform circular array Technol. Lett., 2015, 57, (3), pp. 697–702
antenna’, AEU-Int. J. Electron. Commun., 2016, 70, (6), pp. 750–756 [28] Zheng, D., Cheng, Y., Cheng, D., et al.: ‘Four-band polarization – insensitive
[11] Bayraktar, Z., Komurcu, M., Werner, D.H.: ‘Wind driven optimization metamaterial absorber based on flower-shaped structures’, Prog.
(WDO): a novel nature-inspired optimization algorithm and its application to Electromagn. Res., 2013, 142, pp. 221–229
electro-magnetics’. IEEE Int. Symp. Antennas and Propagation and CNC- [29] Agrawal, A., Misra, M., Singh, A.: ‘Oblique incidence and polarization
USNC/URSI Radio Science Meeting – Leading the Wave APS/URSI, 2010, insensitive multiband metamaterial absorber with quad paired concentric
pp. 0–3 continuous ring resonators’, Prog. Electromagn. Res., 2017, 60, pp. 33–46
[12] Bayraktar, Z., Komurcu, M., Bossard, J. A., et al.: ‘The wind driven [30] Mao, Z., Liu, S., Bian, B., et al.: ‘Multi-band polarization-insensitive
optimization technique and its application in electromagnetics’, IEEE Trans. metamaterial absorber based on Chinese ancient coin-shaped structures’, J.
Antennas Propag., 2013, 61, pp. 2745–2757 Appl. Phys., 2014, 115, (20), p. 204505, 28

1270 IET Microw. Antennas Propag., 2019, Vol. 13 Iss. 9, pp. 1263-1270
© The Institution of Engineering and Technology 2019