(3)
which holds for tuned antennas under the sufficient conditions that
'
0 0
( ) X e and
'
0 0
( ) R e do not change
greatly over the bandwidth (conditions that hold if
0
FBW ( ) 1
V
e , which can always be satisfied if  is
chosen small enough).
In this paper, we would like to analyze mutual coupling effect on the bandwidth enhancement of
antenna arrays. It is known that mutual coupling effect in an antenna array is commonly modeled as a
change in the driving impedance of the elements and it is usually referred to as mutual impedance
variation [4]. For the simplest array that consists of two identical elements with equal excitation, we can
easily obtain the elements drivingpoint impedances as follows
1 11 12 d
Z Z Z = + , (4)
where
1d
Z ,
11
Z and
12
Z represent the elements drivingpoint impedance, the elements self impedance
and the mutual impedance, respectively.
5
For an array with all elements excited simultaneously, it is not the elements self impedance but the
drivingpoint impedance that should be matched. If the array is tuned at a frequency
0
e (normally,
0
e is
not equal to the resonant or antiresonant frequency
0
e of elements due to mutual coupling effect.), the
conductance
1 0 0
( )
d
X e equals zero and the driving impedance is
1 0 0 110 0 120 0 1 0 0
( ) ( ) ( ) ( )
d d
Z Z Z R e e e e = + = . (5)
Substituting (5) into (3), we obtain the fractional matched VSWR bandwidth
0
FBW ( )
V
e for the array as
0 0 0 0
0
' ' '
0 0 0 0 110 0 120 0
4 ( ) 4 ( )
FBW ( )
( ) ( ) ( )
d d
V
d
R R
Z Z Z
 e  e
e
e e e e e
~ =
+
. (6)
It is shown that from (3) and (6), provided we match the array with
1 0 0 0 0
( ) ( )
d
R R e e ~ and appropriately
tune impedances with
' '
110 0 120 0
( ) ( ) Z Z e e ~ , the matched VSWR bandwidth
0
FBW ( )
V
e of the array would
be greatly enhanced.
Through the above analysis, we show mutual coupling effect on array bandwidth could be analyzed in
the view of the mutual impedance. However, the analysis associated with mutual impedance is usually
quite complicated, especially for a larger array. The expressions of elements opencircuit voltage are
usually needed to calculate the mutual impedance, while array elements port voltage is difficult to be
measured for most practical configurations, even to be defined in some situations. On the other hand, S
parameters are a versatile tool to analyze microwave networks. Furthermore, the matched VSWR
bandwidth is naturally defined using the reflection coefficient. Therefore, we would rather investigate the
mutual coupling effect on an arrays bandwidth in the view of Sparameters. Therefore, we go back to
the definition of matched VSWR bandwidth, ( ) e e
+
, which is determined by
2
0
( ) e o
I = . (7)
Expanding the lefthand side of (7) in a Taylor series about
0
e , we find
2
' 2
0 0
( ) ( ) e e o
I A ~ (8)
6
under the condition
0 0
( ) 0 e I ~ and the assumption that the
3
[( ) ] O e
A are
'
0 0
( )
o
e
e
A ~
I
, (9)
which results in the fractional matched VSWR bandwidth
0
FBW ( )
V
e
0
'
0 0 0 0 0
2
( )
( )
V
FBW
e e e e o
e
e e e e
+ +
A A
= = ~
I
, (10)
Equation (10) holds true for tuned antennas under the sufficient condition that
'
0
( ) e I does not change
greatly over the bandwidth. It also reveals that the fractional matched VSWR bandwidth for a tuned
antenna is inversely proportional to the absolute value of the derivative of the reflection coefficient with
respect toe . Note that (10) is a general equation for antennas that can be used to investigate mutual
coupling effect on bandwidth in terms of Sparameters.
2.2 Active reflection coefficient of arrays
Fig. 1 shows a twoelement array fed through a 3dB Tbranch power divider, which can be
considered as the cascaded connection of a threeport network Tbranch power divider) with a twoport
network. The incident and reflected waves are labelled at the ports as a, a
1
, a
2
, b, b
1
, and b
2
. The
relationships between incident and reflected waves can be established by the Smatrix of these two
networks, so that the reflection coefficient of the overall network is derived.
For a lossless 3dB Tbranch power divider [15], we know
1 1
2 2
1 1 1
1 1 2 2
2
1 1 1
2 2
2 2
2
0
b a
a b
a b
=
. (11)
For the antenna array, we consider only passive and identical elements, and then we have
1 1 11 12
12 11 2 2
b a S S
S S b a
=
. (12)
7
Due to the symmetry of the array and the power divider, it can be easily derived from (11) and (12),
1 2
1
2
a a a = =
and
1 2 11 12
1
( )
2
b b a S S = = + .
Finally, we obtain the reflection coefficient of the overall network at the input port of the divider as
11 12
b
S S
a
I = = + . (13)
Equation (13) shows that the reflection coefficient I of the overall network is the superposition of
11
S and
12
S for the twoantenna array.
We further discuss a linear array with 4 identical elements fed through a 4way power divider which is
shown in Fig. 2, with the incident and reflected waves labeled at the ports. Similarly, considering the
system as the cascaded connection of a fiveport network (power divider) with a fourport network, we
can write the relationship of those input and output signals,
1 1 1 1
2 2 2 2
3 1 1 1 1
1 1 2 4 4 4 4
3 1 1 1 1
2 2 2 4 4 4 4
3 1 1 1 1
3 3 2 4 4 4 4
3 1 1 1 1
4 4 2 4 4 4 4
0 b a
a b
a b
a b
a b
(14)
and
1 11 12 1
2 12 22 23 2
3 23 22 12 3
4 12 11 4
0 0
0
0
0 0
b S S a
b S S S a
b S S S a
b S S a
=
. (15)
The fact that the array is symmetric and array elements are identical is considered in (15) and only the
mutual coupling between adjacent elements is taken into account.
8
Figure 1: A twoelement array fed through a Tbranch power divider.
Figure 2: A fourelement array fed through a 4way power divider.
Because
1 4 2 3
, a a a a = = and
1 4 2 3
, b b b b = = , equations (14) and (15) are simplified as
1 1 1
1 1 2 2 2
1 1 1
2 2 2 2 2
0 1 1 b a
a b
a b
=
(16)
and
1 1 11 12
12 22 23 2 2
b a S S
S S S b a
=
+
. (17)
Let
'
22 22 23
S S S = + , we know
'
12 22
1 '
11 22
1
2
S S
a a
S S
+
=
+
and
9
11 12
2 '
11 22
1
2
S S
a a
S S
+
=
+
.
Finally, the reflection coefficient of the overall array is derived as
11 22 23 12
12
11 22 23
2
2
2
S S S S b
S
a S S S
+ +
I = = +
+
. (18)
Assuming
11 22
S S = ,
12 23
S S = , (18) is approximated for small value of
12
S
11 12
12 11 12
12
2 3
2
2 2
S S
S S S
S
I + +
. (19)
Equation (19) shows the similar property as (13) that the reflection coefficient of the array fed through a
corporate feed is the linear superposition of
11
S and
12
S , which can be generalized to the active reflection
coefficient of the Nelement linear array with all elements excited simultaneously.
Generally, we assume all the elements to be fed with equal powers,
1 2
/
N
a a a a N = = = = , all the
reflection coefficients of antenna ports be equal,
11 22 NN
S S S = = = , the mutual coupling between the
adjacent elements be equal,
12 23 1, N N
S S S
= = = , and all the other mutual coupling be neglected.
Under these assumptions, we get
11 12
11 12
( ) / 1,
( 2 ) /
i
S S a N i N
b
elsewhere
S S a N
+ =
~
+
.
The active reflection coefficient of the overall array is derived as
1
11 12 11 12
/
2( 1)
2 ( )
N
i
i
b N
b N
S S S S N is large
a a N
=
I = ~ ~ + ~ +
. (20)
Equation (20) gives a general closedform expression for the active reflection coefficient of the N
element linear array, which keeps the form of linear superposition of element passive reflection
coefficient
11
S and mutual coupling
12
S .
2.3 Mutual coupling effect on array bandwidth
Taking a Nelement linear array fed through a corporate feed as a single port antenna, it can be
known from (20) that the array could be tuned at a frequency
0
e where
10
11 0 12 0
2( 1)
( ) ( )
N
S S
N
e e
~ (21)
and
0
( ) 0 e I = . Due to the mutual coupling effect,
0
e is not equal to the resonant or antiresonant
frequency [5] of the single elements
0
e , but normally ( )
0 0 0
/ 1 e e e .
We substitute (20) into (10) to obtain the matched VSWR bandwidth for the array at a given o as
0
' ' '
0 0 0 11 0 12 0
2 2
( )
( ) ( ) (2 1) ( ) /
V
FBW
S N S N
o o
e
e e e e e
~ ~
I +
. (22)
For a single element,
'
11
( ) S e does not change greatly nearby
0
e . Therefore, when comparing (22) and
(10), we know that the bandwidth of the overall array system should be affected by the relation between
derivatives of element passive reflection coefficient
11
S and mutual coupling
12
S .
Equations (21) and (22) imply two conditions which should be satisfied to achieve bandwidth
enhancement for the whole array. Firstly, the element passive reflection coefficient
11
S and mutual
coupling
12
S should be out of phase and be cancelled as shown in (21), so that the array can be tuned.
Furthermore, the frequency derivatives
'
11
S and
'
12
S
should be out of phase to cancel each other. We
shall conduct numerical simulation and experimental verification to demonstrate this bandwidth
enhancement.
3 Simulated and measured results
The closedform expressions derived in the previous section imply that the bandwidth enhancement of
the overall array can be achieved by invoking appropriate mutual coupling between elements. In this
section numerical simulation and experimental verification are conducted to investigate mutual
coupling effect on array bandwidth. Two types of antenna are under analysed. Firstly, we consider slot
antenna arrays, which are normally narrowband and only have one antiresonant frequency. Then, a
broadband twoelement Vivaldi array is investigated, which has a combination of two or more
resonances and antiresonances that are so close together to build a broad bandwidth. To optimize the
array bandwidth, we shall vary element spacing d to match the active array with appropriate amplitudes
11
and phase differences for
11
S and
12
S as shown in (21). Then, for a given the element spacing d ,
particular parametric optimizations are conducted to achieve bandwidth enhancement for the arrays by
cancelling the derivatives of S parameters as shown in (22).
3.1 Slot antenna array
We consider a single slot antenna and an array of two identical slots shown in Fig. 3. Their resonant
frequencies are about 5.8 GHz. The substrate used is Rogers RO4230 ( 3
r
c = ) of thickness 0.762 mm.
The slots are cut on the ground plane of the substrate and the microstrip feed lines printed on the top
side. Array #1, which is fed through two independent microstrip lines, is used to obtain Sparameters of
the array. Array #2, which is fed through a Tbranch power divider, is the one we hope to investigate.
Obviously, it is the element spacing d that mainly influences the mutual coupling between elements.
So the phase differences between
11
S and
12
S of array #1 with different spacings are firstly calculated
and shown in Fig. 4. Note that the element spacing should be less than one wavelength to avoid grating
lobes and evoke mutual coupling strong enough to achieve bandwidth enhancement. Fig. 4 indicates
that the phase difference between
11
S and
12
S changes with the element spacing and has a sharp turn at
the resonant frequency. Therefore, we can evaluate the spacing to achieve appropriate amplitudes and
phase difference of t for S parameters approximately at a small deviation from the resonant frequency,
meanwhile, physical parameters are optimized to ensure proper phase relation for their derivatives with
respect to frequency. In particular, we found the optimized spacing 14 d = mm (about a quarter
wavelengths at the resonant frequency), as well as the optimized physical parameters, with which the
bandwidth of the antenna array obviously improved. The corresponding simulated results for phase
difference and magnitude of S parameter are shown in Figs. 4 and 5, respectively. It can be seen that
11
S and
12
S are approximately equal in magnitude and opposite in phase near the frequency of 5.98GHz,
so that the active array can be expected to tuned near this frequency. Fig. 6 shows the optimized results
of phase difference between derivatives of S parameters, which show phase differences about t have
been obtained to strive for bandwidth enhancement of array. The single antenna and array #2 are
designed and fabricated to verify our analysis, and simulated and measured results of return loss are
shown in Fig. 7. We can see that the resonant frequency of the array
0
5.98 e = GHz is slightly different to
12
that of the single slot
0
5.70 e = GHz, just as what we analysed above. It should be more emphasised that
obviously bandwidth enhancements have been demonstrated. It is seen from Fig. 7 that the bandwidth is
improved from 0.50 GHz (6.00 GHz5.50 GHz) to 0.80 GHz (6.30 GHz 5.50 GHz) according to simulated
results, as well as from 0.40GHz (5.95 GHz5.55 GHz) to 0.76 GHz (6.31 GHz5.55 GHz) according to
measured ones. Measured results are in very good agreement with simulated ones. Moreover, in order
to evaluate mutual coupling effect on array performance in general, we further provide realized gain
results of a single slot antenna and array #2, which are shown in Fig. 8. We can see that moderately
improved gain results have been achieved as well as the bandwidth enhancement.
Next, we consider arrays consisting of four identical slots shown in Fig. 3 to further demonstrate our
analysis. Array #3 that is fed by four independent microstrip lines is used to obtain the Sparameters of
the array. Array #4 fed through Tbranch power dividers is the one we hope to investigate. The same
substrate and element dimension are adopted as the twoelement slot array. Particularly, we evaluate
the optimized spacing d=14 mm again for approximately achieving optimal S parameters and their
derivatives. The comparison of return loss results between a single slot antenna and the fourslot array
#4 is shown in Fig. 9. It is seen that the bandwidth of array #4 is improved from 0.50 GHz (6.00 GHz5.50
GHz) to 1.09 GHz (6.69 GHz5.60 GHz) according to simulated results and from 0.40GHz (5.95 GHz5.55
GHz) to 1.38 GHz (6.78 GHz5.40 GHz) according to measured ones. The realized gain results are also
provide in Fig. 10 ,which show that moderately improved gain results have been achieved as well as the
bandwidth enhancement.
13
Figure 3: The layout of slot antennas.
4 4.5 5 5.5 6 6.5 7
220
200
180
160
140
120
100
80
60
40
Frequency, GHz
P
h
a
s
e
(
S
1
1
)

P
h
a
s
e
(
S
1
2
)
,
d=10 mm
d=15 mm
d=20 mm
Figure 4: Calculated phase differences between
11
S and
12
S of array #1.
14
4 4.5 5 5.5 6 6.5 7
25
20
15
10
5
0
Frequency, GHz
M
a
g
n
i
t
u
d
e
o
f
S

p
a
r
a
m
e
t
e
r
s
,
d
B
S
11
S
12
Figure 5: Simulated magnitudes of
11
S and
12
S of array #1.
4 4.5 5 5.5 6 6.5 7
200
150
100
50
0
50
100
150
200
250
Frequency, GHz
P
h
a
s
e
(
S
, 1
1
)

P
h
a
s
e
(
S
, 1
2
)
,
Figure 6: Calculated phase differences between
'
11
S and
'
12
S of array #1.
15
4 4.5 5 5.5 6 6.5 7 7.5 8
25
20
15
10
5
0
Frequency, GHz
R
e
t
u
r
n
L
o
s
s
,
d
B
Simulated single slot
Simulated array #2
Measured single slot
Measured array #2
Figure 7: Simulated and measured return loss results of single slot and twoslot array #2 (d=14 mm).
5 5.5 6 6.5 7
2
3
4
5
6
7
8
Frequency, GHz
R
e
a
l
i
z
e
d
G
a
i
n
,
d
B
Simulated single slot
Simulated array #2
Measured array #2
Figure 8: Simulated and measured gain results of single slot and twoslot array #2 (d=14 mm).
16
4 4.5 5 5.5 6 6.5 7 7.5 8
35
30
25
20
15
10
5
0
Frequency, GHz
R
e
t
u
r
n
L
o
s
s
,
d
B
Simulated single slot
Simulated array #4
Measured single slot
Measured array #4
Figure 9: Simulated and measured return loss results of single slot and fourslot array #4 (d=14 mm).
5 5.5 6 6.5 7
1
2
3
4
5
6
7
8
9
10
Frequency, GHz
R
e
a
l
i
z
e
d
G
a
i
n
,
d
B
Simulated single slot
Simulated array #4
Measured array #4
Figure 10: Simulated and measured gain results of single slot and fourslot array #4 (d=14 mm).
17
DSL
RST
WST
WSL
R
L
d H
z1 z2
y1
y2
(a) A twoelement Vivaldi array (b) Definition of element configuration
Figure 11: Schematic graph of Vivaldi antenna array .
3.2 Vivaldi antenna array
The tapered slot antenna (TSA) was initially introduced as an array element by Lewis et al. in 1974
[11]. Gibson [12] later named the exponentially tapered slot antenna as Vivaldi antenna. Because of its
potential for wideband and wide scan arrays, ease of fabrication, and convenient feed techniques,
Vivaldi antenna array is a prime candidate for highperformance phasedarray systems. In this paper,
Vivaldi array is chosen as an example for broadband arrays having a combination of two or more
resonances and antiresonances
1 2
, , , , , ( 2)
n N
N e e e e > . The bandwidth enhancement can be
achieved at these frequencies by utilizing mutual coupling between array elements, so that the total
bandwidth can be improved for a broadband array.
We consider a twoelement Vivaldi array shown in Fig. 11. For each element, the stripline feed is
coupled to bilateral slotline cut on the ground plane. Element parameters can be subdivided into the
stripline/slotline transition, the tapered slot, and the stripline stub and slotline cavity [13]. The bilateral
slotline is terminated at one end with a circular slot cavity of diameter D
SL
and at the other end it opens
into an exponentially tapered slotline with an opening rate of R. The stripline feed is terminated in a radial
line stub with radius R
ST
and angleu. The stripline/slotline transition is specified by W
ST
(stripline width)
and W
SL
(slotline width). Distances from the transition to the slotline cavity and the taper can be properly
adjusted to accommodate the transition. Broadband behavior has been obtained by appropriately
18
designed W
ST
, W
SL
, D
SL
, R
ST
and u with coupling from the dominant stripling mode to the traveling wave in
tapered slot. The broadband characteristic of antenna itself mainly attributes to the exponential taper
profile, which is defined by the opening rate R and two points
1 1 1
( , ) P z y and
2 2 2
( , ) P z y
1 2
Rz
y c e c = + , (23)
where
2 1
2 1
1
,
Rz Rz
y y
c
e e
2 1
2 1
1 2
2
Rz Rz
Rz Rz
y e y e
c
e e
.
The taper length L is
2 1
z z and the aperture height H is
2 1
2( )
SL
y y W + .
The Vivaldi antenna element has theoretically unlimited bandwidth, and the upper operating frequency
is mainly limited by the transition. However, for a Vivaldi array, the upper operating frequency is limited
by the onset of grating lobes, which is determined by the element spacing. Therefore, we usually
prescribe the element spacing and optimize antenna parameters to achieve desired bandwidth of the
array. When the element spacing is prescribed, the mutual coupling between elements
12
S varies slightly
and the bandwidth enhancement is mainly achieved by optimizing the elements passive reflection
coefficient
11
S to be well cancelled with
12
S .
Here, we prescribe the element spacing d=10 mm, which is also the element width. The substrate used
is RO4230 ( 3
r
c = ) of thickness 1.524 mm. After some optimization following [1415], the physical
parameters are chosen: H=8 mm, W
ST
=W
SL
=0.5 mm, D
SL
= R
ST
=2.6 mm, u=70, L=22 mm and R=0.3 mm
1
.
To clarify the relationship between
11
S and
12
S , the variations of Sparameters with respect to different
opening rates R are provided and are shown in Fig. 12. The corresponding active reflection coefficients I
are shown in Fig.13. It is shown that with R=0.3 mm
1
, the minimum operating frequency is about 5.5 GHz,
while the frequency of one element is about 10.7 GHz. For a specified upper operating frequency 18 GHz,
an optimum 3:1 bandwidth of the array has been achieved. Fig.14 shows the phase difference between
11
S and
12
S with R=0.3 mm
1
, and Fig.15 shows the phase difference between
'
11
S and
'
12
S , which are all
about t. Simulation results show that the bandwidth of the Vivaldi array has been successfully
improved by properly adjusting the passive reflection coefficient of elements along with mutual coupling.
19
The phase differences between
11
S and
12
S and between
'
11
S and
'
12
S can also be useful indicators for
design of Vivaldi array.
4 6 8 10 12 14 16 18
30
25
20
15
10
5
0
Frequency, GHz
S

p
a
r
a
m
e
t
e
r
s
,
d
B
S
11
R=0.15 mm
1
S
11
R=0.30 mm
1
S
11
R=0.45 mm
1
S
12
R=0.15 mm
1
S
12
R=0.30 mm
1
S
12
R=0.45 mm
1
10.7 GHz
Figure 12: Sparameters with various opening rates R.
4 6 8 10 12 14 16 18
30
25
20
15
10
5
0
Frequency, GHz
I
,
d
B
R=0.15 mm
1
R=0.30 mm
1
R=0.45 mm
1
5.5 GHz
Figure 13: The active reflection coefficient I with various opening rates R.
20
4 6 8 10 12 14 16 18
250
200
150
100
50
0
50
100
150
200
250
Frequency, GHz
P
h
a
s
e
(
S
1
1
)

P
h
a
s
e
(
S
1
2
)
,
Figure 14: The phase difference between
11
S and
12
S with R = 0.3mm
1
.
4 6 8 10 12 14 16 18
300
200
100
0
100
200
300
Frequency, GHz
P
h
a
s
e
(
S
, 1
1
)

P
h
a
s
e
(
S
, 1
2
)
,
Figure 15: The phase difference between
'
11
S and
'
12
S with R = 0.3mm
1
.
4 Conclusions
In this paper, mutual coupling between array elements has been utilized to achieve bandwidth
enhancement based on the formulations for the matched VSWR bandwidth and the reflection coefficient
21
of arrays with corporate feed. Instead of sticking to analysis of the mutual impedance, the active
reflection coefficient of the array has been investigated to directly guide the optimization of bandwidth.
Our theoretical analyses show that bandwidth enhancement of the overall array can be achieved when
the element passive reflection coefficient
11
S and mutual coupling
12
S are well cancelled, as well as the
frequency derivatives
'
11
S and
'
12
S
also cancel each other. Both numerical simulated and experimental
results successfully demonstrate our analysis.
5 References
[1] Ludwig A.C.: Mutual coupling, gain, and directivity of an array of two identical antennas, IEEE Trans.
Antennas Propag., 1976, 24, (6), pp. 837841
[2] Gupta I.J., Ksienski A.A., Effect of mutual coupling on the performance of adaptive arrays, IEEE
Trans. Antennas Propag., 1983, 31, (5), pp. 785791
[3] Malherbe J.A.G., Analysis of a linear antenna array including the effects of mutual coupling, IEEE
Trans. Education, 1989, 32, (1), pp. 2934
[4] Balanis C. A., Antenna TheoryAnalysis and Design (John Wiley & Sons, 2005, 3rd edn.)
[5] Yaghjian A. D., Impedance, bandwidth, and Q of antennas. IEEE Trans. Antennas Propag., 2005,
53, (4), pp. 12981324
[6] Wang M., Shen Z., Mutual coupling rffect on the performance of antenna arrays with corporate feed,
AsiaPacific Symp. Electromagnetic & Compatibility, 2008, pp. 546549
[7] Stein S., On cross coupling in multiplebeam antennas, IRE Trans. Antennas Propag., 1962, 10, (5),
pp. 548557
[8] Li Q., Shen Z., Inverted microstripfed cavitybacked slot antennas. IEEE Antennas Wireless
Propog. Lett., 2002, 1, (1), pp.98101
[9] Latif S.I., Shafai L., Sharma S.K., Bandwidth enhancement and size reduction of microstrip slot
antennas, IEEE Trans Antennas Propag., 2005, 53, (4), pp.9941003
[10] Garg R., Bhartia P., Bahl I., Ittipiboon A., Microstrip antenna design handbook (Artech House,
Norwood, MA, USA, 2001)
22
[11] Lewis L.R., Fassett M., and Hunt J., A broadband stripline array element, IEEE Int. Symp.
Antennas Propag. Society, 1974,12, pp.335337
[12] Gibson P.J., The Vivaldi aerial, Proc. European Microwave Conf., 1979, 9, pp.120124
[13] Shin J., Schaubert D.H., A parameter study of striplinefed Vivaldi notchantenna arrays, IEEE
Trans Antennas Propag., 1999, 47, (5), pp.879886
[14] Chio T., Schaubert D.H., Parameter study and design of wideband widescan dualpolarized
tapered slot antenna arrays, IEEE Trans Antennas Propag., 2000, 48, (6), pp.879886
[15] Pozar D.M., Microwave Engineering (John Wiley & Sons, 2005, 3rd edn.)