OPEN ACCESS You may also like

- An Introduction to Classical
A simple route to a tunable electromagnetic Electromagnetic Radiation
Minh Quang Tran
gateway - Special issue on 'Interaction of
electromagnetic waves with low-
temperature plasmas'
To cite this article: Huanyang Chen et al 2009 New J. Phys. 11 083012 Osamu Sakai and Shahid Rauf

- Calculation of electromagnetic field for

very fast transient process in GIS
Yang Zhuangzhuang
View the article online for updates and enhancements.

This content was downloaded from IP address 190.136.114.87 on 14/12/2023 at 21:58

 New Journal of Physics The open–access journal for physics

A simple route to a tunable electromagnetic gateway

Huanyang Chen1 , Che Ting Chan1,3 , Shiyang Liu2
and Zhifang Lin2
1
Department of Physics and the William Mong Institute of NanoScience and
Technology, The Hong Kong University of Science and Technology,
Clear Water Bay, Hong Kong, People’s Republic of China
2
Surface Physics Laboratory, Department of Physics, Fudan University,
Shanghai 200433, People’s Republic of China
E-mail: phchan@ust.hk

New Journal of Physics 11 (2009) 083012 (13pp)

Received 6 May 2009
Published 13 August 2009
Online at http://www.njp.org/
doi:10.1088/1367-2630/11/8/083012

Abstract. Transformation optics is used to design a gateway that can block

electromagnetic waves but allows the passage of other entities. Our conceptual
device has the advantage that it can be realized with simple materials and
structural parameters and can have a reasonably wide bandwidth. In particular,
we show that our system can be implemented by using a magnetic photonic
crystal structure that employs a square array of ferrite rods, and as the field
response of ferrites can be tuned by external magnetic fields, we end up with
an electromagnetic gateway that can be open or shut using external fields. The
functionality is also robust against the positional disorder of the rods that make
up the photonic crystal.

3
Author to whom any correspondence should be addressed.

New Journal of Physics 11 (2009) 083012

1367-2630/09/083012+13$30.00 © IOP Publishing Ltd and Deutsche Physikalische Gesellschaft
2

Contents

1. Introduction 2
2. A new ‘superscatterer’ 2
3. An invisible electromagnetic gateway 4
4. The implementation of a tunable electromagnetic gateway 7
5. Conclusion 9
Acknowledgments 9
Appendix A 9
Appendix B 11
References 12

1. Introduction

Transformation optics [1]–[3] has paved the way for the development of optical devices that
can realize functionalities that were thought to be possible only in science fiction [4]–[18].
Luo et al [19] have recently proposed a conceptual device that has attracted great public interest.
This device is a gateway that can block electromagnetic waves but that allows the passage
of other entities. It can be viewed as an implementation of a ‘hidden portal’ mentioned in
fiction [20]. However, the feasibility of such devices is limited by the very complex material
parameters and the narrow bandwidth. Here, we show that gateway-type devices can actually
be realized with simple parameters and they can have wider bandwidths such that the concept
is closer to reality [21]–[26] than previously thought. The structure can be implemented by
using magnetic photonic crystal structures that are field tunable, resulting in an invisible
electromagnetic gateway that can be open or shut using magnetic fields.

2. A new ‘superscatterer’

We start from a simple implementation of transformation optics. In figure 1(a), an object

(colored in blue) is placed to the left side of a double negative medium (DNM) with ε = µ = −1.
Let the object be of permittivity ε0 and permeability µ0 . The detailed shape and the length scale
of the structure are shown in figure 1(a). From the viewpoint of transformation optics, the whole
structure is optically equivalent to another object described in figure 1(b) for far-field observers.
↔ ↔ ↔ ↔
The equivalent permittivity and permeability tensors are ε = ε0 c and µ = µ0 c with a constant
↔
tensor c . We introduce some parameters for ease of reference following the coordinates x1 , x2 ,
y1 , y2 and the angle α in figure 1. Let 1 = x2 − x1 be the waist of the object in the x-direction,
r = 2x11+1 be a coordinate compression ratio and
1 2(x1 + 1) y2 − y y2 − y

 , y1 < y < y2 , −x2 < x < x1 ,
α 1




 tan 2x 1 + y 2 − y1 y2 − y1

p = 0, |y| < y1 , −x2 < x < x1 , (1)


1 2(x1 + 1)

 y2 + y y2 + y
, −y2 < y < −y1 , −x2 < x < x1 ,

−

tan α 2x1 + 1 y2 − y1 y2 − y1

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

3

Figure 1. A simple electromagnetic structure comprising of two homogeneous

regions (left panel) and an equivalent structure that gives the same
scattering (right panel). (a) The schematic plot for the scatterer that appears
electromagnetically larger than its physical size. The red region is the DNM, the
blue region is the original isotropic material that is to the left of the DNM, and the
green region represents the air background. x1 , x2 , y1 , y2 and α are the detailed
coordinates of the structure. The dashed lines in the region (x > 0) give the
virtual boundary for the equivalent transformation medium. (b) The equivalent
transformation medium is indicated by the yellow region, whose parameters are
implemented by equation (2). The dashed lines distinguish the three different
regions for the definition of p.

where (x, y) are the position coordinates of the equivalent transformation medium in
figure 1(b). The inequalities for the definition of the parameter p come from three different
coordinate transformations and the three regions are distinguished by the two dashed lines in
figure 1(b).
It can be proved that (see appendix A):
 1+ p2 
↔ r
− p 0
c =  − p r 0 . (2)
0 0 r

The coordinate transformation is based on the recent work on transformation optics with
complementary media (or folded geometries) [27, 28]. The original object is mapped into the
equivalent transformation medium. The DNM together with its mirror image (the boundary is
marked by the dashed line in figure 1(a)) of air form a pair of complementary media, in the sense
that the phase accumulated in one segment is exactly canceled by another. A more detailed
discussion of the optical properties of this specific geometry and the associated coordinate
transformation can be found in appendix A. The structure in figure 1(a) can be treated as a
form of a scattering amplifier [29] because its scattering cross section can be much larger than
its geometric cross section. The amplification effect originates from the excitation of surface
plasmons in the DNM and air interface.
As a concrete example, we suppose that the original isotropic material (the blue region in
figure 1(a)) has a large value of permittivity, ε0 = −10 000 and µ0 = 1, which may be treated

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

4

Figure 2. The scattering pattern for the scattering object and the equivalent
scatterer. (a) The scattering pattern of the structure depicted in figure 1(a). (b)
The scattering pattern of the equivalent material but with the PEC replacing the
material implemented by equation (2).

approximately as a perfect electric conductor (PEC). Figure 2(a) shows the scattering field
pattern of the structure in figure 1(a). The plane wave is incident from the top to the bottom
and has transverse electric (TE) polarization, for which the E field is along the z-direction.
In this paper, we only consider the TE modes for simplicity. Note that the same idea works
for transverse magnetic (TM) modes as well. The frequency is 2 GHz and x1 = y1 = 0.1 m,
x2 = y2 = 0.2 m and α = π/4. The structure behaves like an equivalent material with parameters
described by equation (2) to the far-field observers. Due to the large mismatch of the impedance
of the equivalent material with the air background, we expect that the equivalent object (the
yellow domain in figure 1(b)) should scatter like a PEC, and for that reason we choose for
comparison in figure 2(b) the scattering field pattern of a perfect conductor filling up the entire
domain of the virtual object in figure 1(b). The similar far-field scattered patterns between
figures 2(a) and (b) confirm the strong scattering effect. The permittivity and permeability of the
DNM are actually chosen as −1 + 0.0001 × i in the simulations to avoid numerical divergence
problems [28]. The PEC-like scatterers are always chosen as materials of permittivity −10 000
and the permeability is taken to be 1 in this paper for simplicity.

3. An invisible electromagnetic gateway

The amplified scattering effect can be utilized to make an invisible gateway [19]. Suppose that a
PEC wall separates the whole space into two regions, the upper domain and the lower domain.
If there are channels (or gateways) opened in the PEC wall, people in the two different spaces
can communicate with each other, both physically and through electromagnetic (EM) waves.
However, if we replace the doors with the above-described configuration at a specific frequency,
the communication for that frequency will be blocked because the systems behave like PECs.
The most amazing fact is that the channel is in fact physically empty. There is nothing but air
in the channel so objects can ‘walk through’ but the channel is blocked as perceived by the eye
because light at the designated frequency cannot penetrate. Figure 3 is a schematic plot of such a
gateway based on the above idea. With the same scale as the one in figure 2(a), we demonstrate

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

5

Figure 3. The computation domains for the electromagnetic gateway. The yellow
regions are the perfectly matched layers (PMLs), the red region is the DNM, the
green region is air, and the blue regions are PECs. The point source is located at
above the gateway and it is denoted by a brown circle. The region between the
DNM and the right PEC is the so-called ‘invisible gateway’.

the properties of such a gateway. We suppose that there is a line source located at (0.05 m, 0.4 m)
with a frequency of 2 GHz in the upper domain. Figure 4(a) shows that the waves cannot pass
through the gateway and are excluded from the lower domain. However, without the DNM, the
waves can propagate into the lower region as shown in figure 4(b).
We now consider the bandwidth of such a device. To be more concrete, we consider the
following dispersions for the DNM:

f p2
ε = 1− ,
f ( f + i0)
(3)
Ff2
µ = 1− 2 ,
f − f 02 + iγ f

where f p = 2.828 GHz, 0 = 0.1 MHz, F = 1.5, f 0 = 1 GHz and γ = 0.075 MHz. When f =
2 GHz, the relative permittivity and permeability return to −1 + 0.0001 × i used in the above
simulations. When f = 1.6 GHz, the impedances do not match at the interface of air and the
DNM, while the refractive index of the DNM is about −1.76. Figure 4(c) shows the electric
field pattern, whereas figure 4(d) shows the field without the DNM. The DNM can reduce
the penetration of the waves from the upper region when compared with figures 4(c) and (d).
However, as n < −1, there is still a ‘slit’ between the virtual boundary and the right PEC,
which allows the waves to propagate somewhat into the lower region. If n = −1, the virtual

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

6

Figure 4. The functionalities of the gateway (with or without the DNM) at

different frequencies. (a) The electric field pattern for the gateway at 2 GHz. (b)
The electric field pattern for the gateway without the DNM at 2 GHz. (c) Same
as (a) but for 1.6 GHz. (d) Same as (b) but for 1.6 GHz. (e) Same as (a) but for
2.4 GHz. (f) Same as (b) but for 2.4 GHz.

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

7
boundary is simply the mirror of the interface of the DNM and PEC on the left with x = 0
as the mirror. The position of the virtual boundary can be obtained heuristically from the
image-forming principle. When f = 2.4 GHz, the impedances are still mismatched while the
refractive index of the DNM is about −0.56. As the absolute value of n decreases, the virtual
boundary of the image expands and there will be no passage for the waves to penetrate because
the virtual boundary overlaps with the PEC on the right-band side. For example, we plot the
electric field pattern in figure 4(e) and the case without the DNM in figure 4(f). We find that
the DNM can eliminate the penetration of the waves from the upper region. That means that
the negative band of the DNM beyond f = 2 GHz (or −1 < n < 0) is the working frequency
of the designed gateway. However, we also find that the wave blocking effect will become
weaker with higher frequencies (in this case, about 2.5 GHz) in our simulations. As negative
index media are intrinsically dispersive, the bandwidth has to be finite as the parameters will
deviate progressively from those required by transformation optics. But the present gateway is
shown to be relatively robust and has a broad operation bandwidth of about 20%, which can
be regarded as a broadband device. To have a broader bandwidth, we can simply reduce the
distance of the DNM and the right PEC to enhance the overlapping while the functionality of
the gateway is sacrificed. As there are extensive designs of the DNMs at various wavelengths
(both theoretically and experimentally) [21]–[26], it would be reasonably feasible for the present
gateway to be realized in the future.
For higher frequencies, the working bandwidth will become narrow again like the situation
described in [19]. Since the functionality of the invisible gateway comes from the evanescent
waves in the DNM interface, the effect will disappear if we employ a geometric optics
description. If the losses of the metamaterial are large (i.e. the imaginary parts of the parameters
of the DNM are larger than 0.1), the bandwidth of the gateway will be small and the functionality
will be compromised. The absorption of the DNM is the key difficulty for the gateway described
here as for other devices such as the perfect lens. All our simulation results are calculated using
the COMSOL Multiphysics finite element-based electromagnetics solver.

4. The implementation of a tunable electromagnetic gateway

Very recently, it has been demonstrated that the DNM can be realized with a simple array of
ferrite rods without any metallic components [30], which can be used to implement the present
gateway. One of the unique merits of this magnetic metamaterial is that its optical properties are
magnetically tunable. As such, the gateway can be manipulated using external magnetic fields.
The magnetic metamaterial is a periodic square array of subwavelength ferrite rods with radii
of r = 3.5 mm and a lattice constant of a = 10 mm. The permittivity is taken to be ε = 25, and
the permeability of the ferrite rods has the form

µr −iµκ 0
 
ωm (ω0 − 2π iα1 f )
µ̂= iµκ µr 0 with µr =1 +
0 0 1 (ω0 − 2πiα1 f )2 − 4π 2 f 2

and
2π ωm f
µκ = ,
(ω0 − 2π iα1 f )2 − 4π 2 f 2

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

8

50

Y (a) 25

0 –2.0

–25

–50 (a) (b)

–0.8
50

25
Y (a)

0.5
0

–25

–50 (c) (d)

–25 0 25 –25 0 25
X (a) X (a)
Figure 5. The implementation of the invisible gateway using magnetic photonic
crystals. Here, we show the electric field intensity pattern for different geometries
working at different values of the external magnetic field. (a) H0 = 500 Oe,
(b) H0 = 475 Oe, (c) without magnetic metamaterials and (d) with magnetic
metamaterial replaced by PEC.

where α1 is the damping coefficient, ω0 = γ1 H0 is the resonance frequency with γ1 being the
gyromagnetic ratio; H0 is the sum of the external magnetic field and shape anisotropy field along
the z-direction, ωm = γ1 Ms is the characteristic frequency with Ms the saturation magnetization
which is taken as Ms = 1750 G, typical for single-crystal yttrium–iron–garnet (YIG). As the
absorption of single-crystal YIG is extremely low, we can set α1 = 0.
Numerical simulations of the gateway under different conditions were performed by using
the multiple scattering method [30]. Figure 5 shows the electric field intensity in logarithmic
scale. The bulk PEC is replaced by a discrete system of the periodically arranged PEC rods
(marked with dark green solid circles in the figure), and the ferrite rods are denoted as hollow
circles. Under H0 = 500 Oe and the frequency f = 2.55 GHz, the effective refractive index of
the magnetic metamaterial is n = −1 (εeff = µeff = −1). In figure 5(a), it can be seen that the
electric field is excluded from the air channel, so that the open channel appears to be blocked to
the eye at this frequency. This demonstrates that the gateway can be implemented by a photonic

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

9
crystal type structure in which each element (each ferrite rod in the periodic array) is identical.
By changing the external field to H0 = 475 Oe, the refractive index of the magnetic metamaterial
becomes n = 1 (εeff = µeff = 1). The electric field intensity under such conditions is shown
in figure 5(b), which shows that the channel is open for EM wave passage, or the passage
appears to be open to the eye. Our results demonstrate that the gateway can be implemented
by a very simple configuration and the effect is tunable by an external field. As shown in
appendix A, the effect is fairly robust to the disorder of the position of ferrite cylinders.
Figures 5(c) and (d) illustrate the fact that the channel is electromagnetically open if the
ferrite rods are replaced by air and by PEC rods. In addition to the simple geometry and weak
absorption, the magnetic metamaterial has a reasonably broad bandwidth as a DNM.

5. Conclusion

In conclusion, we showed that a robust and tunable electromagnetic gateway can be constructed
using simple material parameters. The idea is conceived through consideration of transformation
optics and can be realized using a photonic crystal type structure.

Acknowledgments

We thank Drs Yun Lai and Jack Ng for their helpful discussions. This work was supported
by Hong Kong Central Allocation grant no. HKUST3/06C. Computation resources were
supported by the Shun Hing Education and Charity Fund. SYL and ZFL were supported
by CNKBRSF, NNSFC, PCSIRT and MOE of China (B06011), and Shanghai Science and
Technology Commission.

Appendix A

In this appendix, we give the detailed coordinate transformation to produce the folded geometry
used in the text. Let us first consider the following coordinate transformation:
x2 + x1

× (x 0 + x2 ), −x2 < x 0 < −x1 ,
y y y y

 −x2 +

 x2 − x1
x = −x 0 , −x1 < x 0 < 0,
y

 (A.1)
x,

 0
else,
y 0 = y, z 0 = z,
where  
y2 − y y2 − y
, x2 , y1 < y < y2 ,


 x1



 y2 − y1 y2 − y1

(x1 , x2 ) = (x1 , x2 ),
y y
−y1 < y < y1 , (A.2)


  
 y2 + y y2 + y
, x2 , −y2 < y < −y1 .

 x1


y2 − y1 y2 − y1
This coordinate transformation maps the virtual space (x-space) described in figure 1(b)
into the physical space (x 0 -space) described in figure 1(a). If the virtual space is a vacuum, we

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

10

Figure A.1. The folded geometry used in the text. (a) A obliquely incident
TE plane wave in free space. (b) The structure with the material described by
equation (A.4) and the DNM are invisible.

can obtain the corresponding material parameters in the physical space. It contains two kinds of
materials in the vacuum background, one a DNM with ε = µ = −1 (shown in red in figure 1(a)),
the other an anisotropic medium (shown in blue in figure 1(a)) whose parameters are derived in
the following section.
We can easily obtain the Jacobian transformation matrix,

∂x0 ∂x0
 
∂x ∂y
0
3= 0 1 0 (A.3)
 
0 0 1

from equations (A.1) and (A.2) for the mapping between the blue region in figure 1(a) and the
yellow region in figure 1(b), and we can also prove that ∂∂xx = r and ∂∂xy = p in the text. From
0 0

transformation optics, we can obtain the material parameters for the blue region,
 r 2 + p2 p

r r
0
↔ ↔
εp = µp =  p 1
0 . (A.4)
 
r r
1
0 0 r

Figure A.1(a) shows the propagation of the incident TE polarization plane wave from the left
to the right but at an angle of 45◦ from the x-direction in free space. The frequency is 2 GHz.
We consider the object described in equation (A.4) together with a DNM domain. The scales
of the structure in our simulations are the same as the ones used in the text. Figure A.1(b)
shows the scattering pattern of such a structure and demonstrates that the structure itself is
invisible to far-field observers.
A modification of the above-folded geometry can lead to a variety of conceptual devices.
For example, if the virtual space is not a vacuum but an object of permittivity εs and
↔
permeability µs , then the parameters in the blue region should be a permittivity tensor, εs ε p ,
↔
and a permeability tensor, µs µ p . The object in the blue region together with the DNM

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

11

Figure A.2. The scattering pattern of an equivalent scatterer and the real
scatterer. (a) The scattering pattern of an object of permittivity εs = 2.25. (b)
The corresponding scattering pattern of the scattering amplifier.

domain will have a scattering cross section that can be much larger than its geometric cross
section. Figure A.2(a) shows the scattering pattern of an object with permittivity εs = 2.25 and
permeability µs = 1. Figure A.2(b) shows the scattering pattern of the structure including the
↔ ↔
object with permittivity tensor εs εp and permeability tensor µs µ p and the DNM. The two similar
far-field patterns in figures A.2(a) and (b) confirm the enhanced scattering effect. In the text, we
apply this structure in an inverse manner. We suppose that the object in the blue region is of
permittivity ε0 and permeability µ0 . Then, the whole structure is equivalent to another object
↔ ↔ ↔ ↔
with permittivity and permeability tensors, ε = ε0 c and µ = µ0 c (described in the text). We
can easily obtain equation (2) by performing an inverse transformation.

Appendix B

In this appendix, we demonstrate the robustness for the implementation of the gateway using the
magnetic photonic crystal structure. We introduce disorder by displacing the ferrite rods from
their original positions with

1r = (a − 2r ) × s × rn , (B.1)

where a is the lattice constant, r is the radii of the ferrite rods, rn is a uniform random variable
in the interval (−0.5, 0.5) (assures no overlap of cylinders), s is the strength of disorder.
Figure B.1(a) corresponds to figure 5(a) but with a disorder of s = 0.5 introduced. The gateway
remains blocked to the eye for this strength of disorder. When H0 changes to 475 Oe, the
gateway is open again as shown in figure B.1(b). Figure B.1(c) corresponds to figure 5(a) but
with a strong disorder of s = 1.0 introduced. The functionality of the gateway is now slightly
compromised but the blocking is at least partially effective at such a strength of disorder. With
H0 = 475 Oe, the gateway is open again as shown in figure B.1(d). The results indicate that the
functionality is robust to structural perturbations.

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

12

50

Y (a) 25

0 –2.0

–25

–50 (a) (b)

–0.8
50

25
Y (a)

0.5
0

–25

–50 (c) (d)

–25 0 25 –25 0 25
X (a) X (a)
Figure B.1. The robust effect from the disorder. The electric field intensity
pattern for the gateway with different levels of disorder working at different
external magnetic fields. (a) s = 0.5, H0 = 500 Oe; (b) s = 0.5, H0 = 475 Oe;
(c) s = 1.0, H0 = 500 Oe; and (d) s = 1.0, H0 = 475 Oe.

References

[1] Leonhardt U 2006 Science 312 1777

[2] Pendry J B, Schurig D and Smith D R 2006 Science 312 1780
[3] Greenleaf A, Lassas M and Uhlmann G 2003 Math. Res. Lett. 10 685
[4] Schurig D, Mock J J, Justice B J, Cummer S A, Pendry J B, Starr A F and Smith D R 2006 Science 314 977
[5] Cai W, Chettiar U K, Kildishev A V and Shalaev V M 2007 Nat. Photonics 1 224
[6] Rahm M, Schurig D, Roberts D A, Cummer S A, Smith D R and Pendry J B 2008 Photonics Nanostruct.:
Fundam. Appl. 6 87
[7] Chen H Y and Chan C T 2007 Appl. Phys. Lett. 90 241105
[8] Chen H Y, Hou B, Chen S, Ao X, Wen W and Chan C T 2009 Phys. Rev. Lett. 102 183903
[9] Greenleaf A, Kurylev Y, Lassas M and Uhlmann G 2007 Phys. Rev. Lett. 99 183901
[10] Kildishev A V and Narimanov E E 2007 Opt. Lett. 32 3432
[11] Zhang X, Chen H Y, Luo X and Ma H R 2008 Opt. Express 16 11764
[12] Bergamin L 2008 Phys. Rev. A 78 043825
[13] Yu G, Jiang W and Cui T J 2009 Appl. Phys. Lett. 94 041904
[14] Li J and Pendry J B 2008 Phys. Rev. Lett. 101 203901
[15] Liu R, Ji C, Mock J J, Chin J Y, Cui T J and Smith D R 2009 Science 323 366

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)

13

[16] Leonhardt U and Tyc T 2009 Science 323 110

[17] Ng J, Chen H Y and Chan C T 2009 Opt. Lett. 34 644
[18] Lai Y, Chen H Y, Zhang Z Q and Chan C T 2009 Phys. Rev. Lett. 102 093901
[19] Luo X D, Yang T, Gu Y, Chen H Y and Ma H R 2009 Appl. Phys. Lett. 94 223513
[20] Rowling J R 1998 Harry Potter and the Chamber of Secrets (London: Bloomsbury)
[21] Shelby R A, Smith D R and Schultz S 2001 Science 292 77
[22] Smith D R, Pendry J B and Wiltshire M C K 2004 Science 305 788
[23] Soukoulis C M, Linden S and Wegener M 2007 Science 315 47
[24] Lezec H J, Dionne J A and Atwater H A 2007 Science 316 430
[25] Yao J, Liu Z, Liu Y, Wang Y, Sun C, Bartal G, Stacy A and Zhang X 2008 Science 321 930
[26] Valentine J, Zhang S, Zentgraf T, Ulin-Avila E, Genov D A, Bartal G and Zhang X 2008 Nature 455 376
[27] Leonhardt U and Philbin T G 2006 New J. Phys. 8 247
[28] Pendry J B and Ramakrishna S A 2003 J. Phys.: Condens. Matter 15 6345
[29] Yang T, Chen H Y, Luo X D and Ma H R 2008 Opt. Express 16 18545
[30] Liu S, Chen W, Du J, Lin Z F, Chui S T and Chan C T 2008 Phys. Rev. Lett. 101 157407

New Journal of Physics 11 (2009) 083012 (http://www.njp.org/)