C014

Filippo Capolino/Theory and Phenomena of Metamaterials 54252_C014 Page proof Page 1 2009-1-28 #5
14
RF Metamaterials
M. C. K. Wiltshire
Imperial College London
I4.I Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I4-I
I4.2 RF Metamaterials Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . I4-2
Dielectrics

Split Ring Resonators
Spiral Resonators
Ring
Resonators
Swiss Rolls
I4.3 Efective Medium Description. . . . . . . . . . . . . . . . . . . . . . . . I4-7
Permeability

Propagation

Transmission

Numerical
modeling

Comparison with Experiments in the Negative
Regime

Discussion
I4.4 RF Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I4-I3
Flux GuidingInitial Demonstration

Flux GuidingHigh
Performance Material

RF Focusing

Discussion
I4.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I4-2I
RF Endoscope/Faceplate
Yoke

Waveguides

Flux
Compressor
I4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I4-23
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I4-23
14.1 Introduction
Metamaterials are efective media [PHSY96,PHRS98,PHRS99] that can provide an engineered
response to electromagnetic radiation that is not available from the range of naturally occurring
materials. Tey consist of arrays of structures in which both the individual elements and the unit cell
are small compared to the wavelength of operation; homogenization of the structures then allows
them to be described as efective media with the conventional electromagnetic constants of permit-
tivity () and permeability (), but with values that could not previously be obtained. For example,
material with simultaneously negative and can be built which have a negative refractive index
[SPV
+
00,SSS0I,PGL
+
03], and muchattentionhas beengiventothe behavior of suchmedia [SPW04].
Most of the work on metamaterials has been concentrated in the microwave regime and above
(gigahertz to terahertz frequencies), and the majority of these metamaterials has been constructed
from a combination of fne wire grids [PHSY96,PHRS98] to give a dielectric response and split ring
resonators [PHRS99] to provide the magnetic response. Tese are simple to fabricate [SRSNN04],
and are active in the microwave regime, providing negative permeability typically over a bandwidth
of some I0[SSMS02]. In the frst examples (see Figure I4.I), the fne wire grid was constructed from
a 20 m diameter gold-coated tungsten wire [PHSY96] which showed a plasma frequency at 9 GHz,
whereas the split ring resonator (SRR) array was made by etching conventional FR4 circuit board
with patterns that were approximately 3 mm in diameter; these had a resonant frequency of about
3 GHz.
14-1
14-2 Theory and Phenomena of Metamaterials
3.0 mm
(a) (b) (c)
FIGURE I4.I Early examples of metamaterials: (a) schematic of the fne wire grid structure; (b) its implementation;
and (c) the split ring resonator structure.
In a seminal series of papers, Smith and coworkers [SPV
+
00,SSNNS0I] combined the wires and
SRRs into a composite structure etched on circuit board, and were able to demonstrate negative
refractive index behavior [SSS0I]. More recent work has built on this approach, and metamateri-
als have now been made that to operate at terahertz [YPF
+
04], infrared [LEW
+
04], and even optical
frequencies [EWL
+
03]. Tese have all used the SRR structure, or modifcations thereof.
Tere is much to be gained, however, from working at lower frequenciesin the radiofrequency
(RF) regime [Wil07]. Te wavelengthof the electromagnetic radiationis extremely long, sothe condi-
tion that the structure should be much smaller that a wavelength is easily met. Moreover, all distances
are very small compared tothe wavelength, soall measurements are made inthe very near feld, where
the electric and magnetic felds are essentially independent [Pen00,Str4Ia], thus simplifying both the
material requirements and the interpretation of measurements. Finally, there are potential applica-
tions for these materials in magnetic resonance imaging (MRI), which operates at radiofrequencies
(RF). At these lower frequencies, however, the SRR structure becomes impractically large, and dif-
ferent structures are required; in particular, the so-called Swiss roll [PHRS99] has proved to be very
efective.
In this chapter, we frst review RF metamaterial development. We then consider how they can
be described by the efective medium approximation, and develop the mean feld description of how
electromagnetic radiation propagates through them, and compare these predictions with experimen-
tal results. We then describe how these materials can be used to produce high resolution images with
RF radiation, and conclude with an overview of some potential devices and applications.
14.2 RF Metamaterials Design
For metamaterials in the microwave regime and at higher frequencies, the building blocks have been
the SRR and fne wire structure, and much of the practical considerations have focussed on making
these small compared to a wavelength. However, when working at lower frequencies (and we select a
frequency of about 64 MHz, the operating frequency of a I.3 Tesla MRI system, as a specifc example),
a diferent set of issues arises. Te concern that the structure is much smaller than the wavelength,
now 3 m in vacuo, becomes trivial, but the individual elements still need to be physically small in
order frst that the materials can be handled in the laboratory, and second that they may be used
in RF applications. Tus we need to drive the critical frequency of the components down without
making them physically larger. In the following sections, we discuss the suitability frst of the fne
RF Metamaterials 14-3
wire structure and then of the SRR structure and its derivatives for use at RF. We then briefy discuss
the use of discrete capacitors within these structures, and fnally introduce the Swiss roll structure,
which turns out to be ideally suited to this frequency range.
14.2.1 Dielectrics
Te plasma frequency
p
of the fne wire grid structure [PHSY96] is given by
2
p
=
2c
2
0
a
2
l n(a/r)
(I4.I)
where
r is the wire radius
a the grid spacing and
c
0
the speed of light in vacuo
so that 20 m wires placed on a 3 mm grid show a plasma frequency of 9.3 GHz [PHRS98]. Using
Equation I4.I as a scaling relation to design material active at 64 MHz shows that, once the wire
diameter is very much smaller than the unit cell size (r a), the frequency depends most strongly
on the unit cell size a, so we would need to have very thin wires (say I0 m diameter) spaced on a
very sparse grid (a 600 mm), which would not constitute a feasible material.
Smith et al. [SVP
+
99] suggested forming the wire into a loop in a plane normal to its length (see
Figure I4.2): this increases the self inductance and hence reduces the resonant frequency to
2
p
=
2c
2
0
a
2
(l n(a/r) + (2R/l ) [l n(8R/r) 7/4])
(I4.2)
where
R is the radius of the loop and
l the length of the wire
2R
2r
a
FIGURE I4.2 Schematic diagram of the loop-wire dielectric structure: the self inductance of the thin wire is
enhanced by forming a loop of radius R in the unit cell, thus reducing the efective plasma frequency.
Tis lowers the resonant frequency signifcantly, but not sumciently to provide viable material in the
I0s of MHz range. For example, taking the same I0 mdiameter wire as above, and assuming that the
loop size is equal to the spacing (R = a/2) and that there is one loop per unit cell (l = a), we require
a wire spacing of 200 mm. Tis is substantially smaller than the 600 mm found for the straight wire,
but is still not a viable material option. Tus far, no dielectric metamaterial has been reported that
operates in this range. However, as pointed out earlier (see Section I4.I), at RF frequencies and work-
ing in the very near feld (i.e., at distances of a few cm), the electric and magnetic feld components
are essentially independent, and so to manipulate the felds from magnetic sources, the most com-
mon requirement at RF, only requires control of the permeability. Tus good progress can be made
without having access to dielectric metamaterials.
14.2.2 Split Ring Resonators
Te conventional SRR structure has a resonant frequency of [PHRS99]
eff
= I
r
2
a
2
I
3l c
2
0
2
r
3
l n(
2w
d
)
+ i
2l
r0
(I4.3)
where
d is the gap between the rings
w the width of the rings
r is the radius of the inner ring
a is the size of the unit cell
l is the inter-layer spacing
To obtain a resonant frequency of 60 MHz while maintaining a gap of 0.I mm and ring width of
0.3 mm requires a radius of 200 mm. Although this is still small compared to the wavelength of
I0 m, the element is impractically large. To make a viable element, we have to increase either the
self capacitance or self inductance of the structure or, ideally, both.
We consider several approaches. First, we note that the self capacitance of the conventional, edge-
coupled SRR described above (Figure I4.3a) is small; this can be increased by using the broadside
SRR (Figure I4.3b), in which the two rings are on opposite sides of a (thin) membrane. Tis structure
has been discussed in some detail by Marques and coworkers [MMMM03,MMREI02], who show
that an element size of /23 can be achieved using a substrate of I00 mthickness with permittivity
= I0 at a frequency of I0 GHz. Tinner substrates are available, so this size could be further reduced.
Despite this, however, a structure to operate at say 60 MHz would be rather large (60 mm radius),
albeit much smaller than the edge coupled case discussed above, and so this structure will not be
further discussed.
14.2.3 Spiral Resonators
To increase the self inductance of the resonator, a spiral structure can be considered for which
the self-inductance scales as approximately the number of turns in the spiral [MHBL99]. Baena
et al. [BMMM04,BJMZ06] have considered two and three turn elements, and have shown that the
frequency for a three turn spiral is approximately 2
2 times smaller than that of a commensurate

SRR structure; however, even the three turn structure is still resonant in the GHz regime for an ele-
ment size of 4 mm, and so to achieve a 64 MHz response would require it to have 70 mm radius.
Wiltshire [Wil00] has considered a hexagonal, eight turn double spiral structure with an element size
r
(a) (b)

r
t
FIGURE I4.3 Schematic diagram of split ring resonator structures: (a) the edge coupled structure consists of two
concentric split ring on a dielectric substrate, whereas (b) the broadside coupled form has the two rings disposed on
opposite sides of the (thin) substrate thus enhancing the self-capacitance of the structure.
of 40 mm which is resonant at around I00 MHz. Tus using spiral structures does give a magnetic
response from a planar structure in the required frequency range, but the elements are still too big
to be incorporated in a material.
14.2.4 Ring Resonators
Adding a discrete capacitance to the ring structure produces a simple LC resonant loop [SKRS02a,
SKRS02b] whose frequency and size can be selected at will. One such element, investigated by Wilt-
shire et al. [WSYS03] consisted simply of twoturns of I mmdiameter copper wire, wound ona 9.6 mm
diameter Delrin rod, with a I00 pF capacitor connected across the ends of the wire to produce an ele-
ment that was resonant at 60 MHz. In a similar way, Sydoruk et al. [SRZ
+
06] connected 330pF
capacitors to a 20 mm diameter split pipe to produce resonators at 46 MHz. Both these elements are
constructed macroscopically, but there is no a priori reason why circuit board technology cannot be
used to fabricate elements, and this was in fact done by Syms et al. [SYS06], who constructed minia-
ture spirals combined with surface mounted capacitors to produce 8 mmdiameter elements resonant
at I30 MHz. Te behavior of ring resonator structures is fully described in the chapters by Sydoruk
et al., and so will not be further discussed here.
14.2.5 Swiss Rolls
Te Swiss roll structure [PHRS99] (see Figure I4.4) is particularly suitable for use at RF frequencies
up to 200 MHz, because it has inherently large self-inductance and self-capacitance. Te elements
consist of a number of turns N of insulated conductor wound onto a central mandrel. In practice,
this is achieved by using a metaldielectric laminate that is wound spirally onto a dielectric rod.
Te self-inductance is governed by the number of layers of conductor, whereas the self-capacitance is
determined by the thickness and permittivity of the dielectric. For a compact, lowfrequency element,
we require the dielectric layer to be thin and of low loss.
Te magnetic permeability of an array of such rolls is given by [PHRS99]
() = I
r
2
a
2
(I
dc
2
0
2
2
r
3
(N I)
2
) + i
2
r0 (N I)
(I4.4)
20
10
0
R
e
a
l

p
a
r
t
Frequency/MHz (c)
I
m
a
g
i
n
a
r
y

p
a
r
t
10
20
15.0 20.0 25.0 30.0
0
10
20
30
40
(a) (b)
1 cm
FIGURE I4.4 Te Swiss roll metamaterial: (a) schematic diagram of an individual element, showing the spiral-
wound structure; (b) a photograph of an element; and (c) the measured permeability of a material designed to operate
near 2I.3 MHz. Full line and lef-hand scale show the real part; dashed line and right-hand scale show the imaginary
part.
where
N is the number of turns in each roll of diameter r
is the (complex) permittivity of the dielectric between the conducing layers
is the resistivity of the conductor
Te rolls are packed with a unit cell spacing of a. Typical values for resonance at 2I.3 MHz are r =
3 mm, N = II, using the material Espanex, which consists of a I2.3 m polyimide sheet with I8 m
of copper laminated to it, manufactured as fexible printed circuit board material. Te polyimide has
a permittivity of 2.3 and a loss tangent of 0.0I, and the resonance has Q 60 [WHP
+
03].
Te frequency-dependent permeability (Equation I4.4) is anisotropic, and usually written
[WHP
+
03]
zz
() = I
F
(I

2
0
2
) + i

,
xx
=
y y
= I (I4.3)
in which
F is the flling factor, given as r
2
/a
2
in Equation I4.4
0
is the resonant angular frequency, given by
0
=
dc
2
0
2
2
r
3
(N I)
(I4.6)
and is determined by the construction of the element
is the damping, and includes both the resistance of the conductive layer and the dissipative part of
the permittivity of the dielectric layer.
Te damping plays a critical role in determining the performance of these materials: strong damp-
ing leads to broad resonances with low quality factor Q and weak magnetic efects. In the very frst
system studied [WPY
+
0I], the metal layer was extremely thin and the resistivity term dominated the
loss. For all other systems, based on fexible printed circuit board (PCB) materials, the dielectric loss
dominates. It has therefore been important to seek fexible PCB systems which are adhesiveless, and
based on low loss dielectric. In most of the work discussed here, the Swiss rolls had a Q 60 at a
resonant frequency near 2I.3 MHz. An optimized Kapton base (Novaclad) can provide a Q I00,
whereas a PTFE-based material (CuFlon) has a Q 300. AQI
14.3 Effective Medium Description
Tere are two approaches to describe the behavior of ensembles of the metamaterial elements: the
microscopic and the macroscopic. In the microscopic approach the ensemble is considered as an
array of coupled resonators, each having a resonant frequency and being coupled to neighboring ele-
ments (or indeed to all other elements). Te currents and voltages fowing in the elements can then
be calculated explicitly, given some excitation, and the resulting current distribution in the ensem-
ble found. It is found that there are wave-like solutions for the current equations, which describe
magneto-inductive waves [SKRS02a,SKRS02b,WSYS04]. From this analysis, the feld distributions
may be obtained for comparison with measurement [ZSS03]. Tis approach is described in detail in
the chapters by Sydoruk et al., and will not be discussed further here.
In the macroscopic view, on the other hand, we assert that, since the element size is much smaller
than the wavelength at the frequency of operation, we can ignore the detail in the material and
describe it as a homogeneous, efective medium whose electromagnetic response is defned by an
efective permittivity and permeability whose values need not be confned to those available in nat-
ural materials. Based on these parameters within the efective medium approach, we investigate the
interaction of the mediumwith electromagnetic felds, and calculate their behavior either analytically
or numerically.
14.3.1 Permeability
Te efective medium prescription was given by Pendry et al. [PHRS99] for a variety of structures,
and considered in more detail by Smith and Pendry [SP06]. Te efective medium parameters arise
from consideration of average felds
B
ave
=
eff
0
H
ave
and D
ave
=
eff
0
E
ave
(I4.7)
We write Maxwells equations in integral form
C
H dl = +

t

S
B dS and

C
E dl =

t

S
D dS (I4.8)
where the line integral is taken along the curve C which bounds the surface S. Ten when the
magnetic felds are very inhomogeneous, and vary rapidly within the structure, the averages for
Hand Bare quite diferent, sothat
eff
becomes signifcantly diferent fromunity. Tis situationarises
in the resonant structures discussed above, and following this prescription, Pendry et al. [PHRS99]
calculated the efective permeability of the Swiss roll structure to be that given in Equation I4.4.
Whereas the permeability of SRR metamaterials has to be derived from the inversion of trans-
mission and refection data in the microwave regime, it can be measured directly for the Swiss roll
structure because the elements are long thin, needle-like elements. Accordingly, the demagnetizing
factor (see e.g., [Str4Ib]) for a single roll or a small bundle of long rolls is quite small and can be
approximated to that of an ellipsoid with the same axial ratio. Te material is inserted into a solenoid
and the change in its self-inductance and resistance is measured. Te complex permeability is then
obtained by applying a volume correction and the demagnetizing factor. An alternative method is to
modify the mutual inductance between two loops by introducing a material sample along their com-
mon axis. By this means, the efective permeability of the medium can be obtained, and is shown in
Figure I4.4 for the material based on Espanex (see above). It is clear that this has the resonant form
of Equation I4.3, and the values of
0
and can be obtained from a least squares ft to the data. Fur-
thermore, by measuring samples with diferent numbers of turns, it is possible [WPY
+
0I] to deduce
both the permittivity of the dielectric (taken as unity in the theory above) and the efective radius of
the roll expressed as r = r
0
+ N, where is a parameter to take into account the thickness of the
laminate (the theory assumes that the layers are infnitesimally thin so that Nd << r
0
, whereas in
reality the thickness of the winding Nd and the radius r may be comparable).
14.3.2 Propagation
We consider the behavior of electromagnetic felds in a highly anisotropic efective medium, and
describe the felds by a Fourier expansion of the form [WHP
+
03]
E(r, t) =
k
E(k) exp i [k r t] (I4.9)
with an analogous expression for H. Maxwells equations for a monochromatic feld with angular
frequency and wave vector k are
ik E = i
0
H, ik H = i
0
E (I4.I0)
Te medium is magnetically active but dielectrically inactive, so on eliminating E we fnd
k k H =
2
c
2
0
H or k (k H) + k
2
H = k
2
0
H (I4.II)
where k
0
is the free space wave vector, /c
0
.
Te efective medium is isotropic in the x y plane, so we can confne our calculations to the x z
plane without loss of generality. It is convenient to rewrite Equation I4.II in terms of B, whereupon
expanding gives
[
k
2
z
/
xx
k
x
k
z
/
zz
k
x
k
z
/
xx
k
2
x
/
zz
] [
B
x
B
z
] = k
2
0
[
B
x
B
z
] (I4.I2)
Te condition for solution is
k
2
x
zz
+
k
2
z
xx
= k
2
0
=

2
c
2
0
(I4.I3)
and the associated eigenvectors are
[
B
x
B
z
] = [
k
2
x
/
zz
k
2
0
k
x
k
z
/
xx
] (I4.I4)
Equations I4.I3 and I4.I4 allow us to explore the behavior of felds both inside the medium and at its
interfaces. In free space, the eigenvalue equation becomes
k
2
x
+ k
2
z
= k
2
0
=

2
c
2
0
(I4.I3)
so that k
z
=

2
/c
2
0
k
2
x
, and only those felds with k
x
< k
0
can propagate; for k
x
> k
0
the felds
decay evanescently along the propagation direction, z.
Within the medium, however, consideration of Equation I4.I3 shows that there are several
possibilities of interest:
I. In what Smith et al. [SKS04] have termed an indefnite medium, that is one with opposite
signs of
xx
and
zz
, propagation is not restricted to k
x
< k
0
. If
zz
< 0, Equation I4.I3
has real solutions for k
z
for all values of k
x
.
2. In the very near feld in an indefnite medium, when k
x
, k
z
>> k
0
, Equation I4.I3
reduces to
k
z
=
k
x
zz
(I4.I6)
so that the felds propagate with a conical wavefront [Bal64,BLK03].
3. In the limiting case on resonance,
xx
=I and
zz
. Ten the eigenvalues (Equation
I4.I3) reduce to
k
z
= k
0
(I4.I7)
and the eigenvectors (Equation I4.I4) become
[
B
x
B
z
] = [
k
0
k
x
] (I4.I8)
We see that k
z
is now independent of k
x
, so all the transverse Fourier components of an
object propagate along the z-axis with the same relative phase: if we measure the intensity
we see a perfect image. In the electric feld equivalent of this situation [RPWS03], an
incident electric feld distribution is transported through the material as if the faces of
the slab were connected by perfectly conducting wires. By analogy in the present case, we
can imagine magnetic wires, composed of a perfect magnetic conductor, transporting
the magnetic image information across the material slab.
14.3.3 Transmission
Te eigenvectors fromEquationI4.I4 (afer using EquationI4.I3 and removing the redundant factors)
are found to be
[
B
x
B
z
] = [
k
z
k
x
] (I4.I9)
so we can match the felds at the boundaries between the prism and free space to obtain the interface
transmission and refection coemcients, t
kx
and r
kx
, as a function of the transverse wavevector k
x
.
Hence the transmission of a semi-infnite slab can be calculated in the conventional manner [PP64]
to obtain
B
z
(r, z) = 2
0
[c
+
kx
J
0
(k
x
r) exp(i k
z
(z d))]dk
x
(I4.20)
where
c
+
kx
= [cos (k
z
d) +
I
2
[
xx
k
x
k
z
k
z
xx
k
x
] sin(k
z
d)]
I
(I4.2I)
is the overall transmission coemcient of semi-infnite slab as a functionof k
x
and J
0
(k
x
r) is the zeroth
order Bessel function of the frst kind.
To take into account the fnite size of sample, we note that there are internal refections at the
entrance and exit surfaces, and also at the sides of the material. For the square prism, this can be
treated by considering a unit cell with periodic boundary conditions and folding all the higher order
components back intothe central zone todeduce the total feld. We combine this with Equations I4.20
and I4.2I for the transmission coemcient to calculate the output pattern, which should consist of a
central ring whose radius is given by Equation I4.I6 along with additional structure arising from
internal refections [WPWH07,Wil03].
14.3.4 Numerical modeling
No numerical modeling of the individual elements has yet been carried out. Unlike the SRR struc-
ture, where the minimum feature size (i.e., the gap between the rings or in the ring itself) is typically
I/I000 of the wavelength and detailed modeling can be performed, in the Swiss roll, this ratio is
nearer I/300,000 and the gridding problem is formidable. Accordingly, the only simulations that
can be performed treat the sample as a homogeneous efective medium having an anisotropic, fre-
quency dispersive permeability. Numerical simulations can then be carried out using, for example,
the transient solver of CST MicroWave Studio. Here, a short pulse of radiation is launched into the
model, whose time evolution is calculated. Tis is then Fourier transformed to provide the frequency
response of the system. Te transverse permeabilities are set to unity, and the axial permeability can
be described using the Lorentzian dispersion in MicroWave Studio (MWS), which sets
(f ) =
s
+
(
s

)f
2
0
(f
2
0
f
2
) i f
(I4.22)
where
s
and
are the low-frequency and high-frequency limiting values, respectively, of the

permeability,
f
0
is the resonant frequency
is the damping
By comparing this with Equation I4.4, we see that the metamaterial requires
s
= I and
= I F;
the resonant frequency and the damping have the same values in both equations.
Te source of the magnetic feld in the calculation can be a plane wave or, to model realistic situa-
tions, a 3 mm diameter wire loop placed in the space behind the slab and excited by a current source
in the loop, acts as a point source. Te metamaterial is embedded in a background medium, typically
vacuum, and so-called openboundaries (i.e., perfectly matched layers) were placed approximately
/8 away from the region of interest; this distance being set by the sofware itself. Although this can
lead to an extremely large model, the gridding is required to be fne only across the metamaterial
region; away from the material, quite a coarse grid can be used so that the overall model is a
manageable size.
14.3.5 Comparison with Experiments in the Negative Regime
To explore the behavior of magnetic RF metamaterials in the regime above their resonant frequency,
when the permeability is negative, transmission experiments have been performed [WHP
+
03,
WPWH07]. In this work, Swiss rolls were assembled into either a square or a hexagonal prism, and
their transmission was measured by placing a small loop acting as a magnetic dipole point source
on one face while scanning a similar loop acting as a detector across the other face. It is clear from
the theoretical discussion above that the transmission is not just a simple number: it depends on the
transverse wavevector, and so a variety of feld patterns are created on the output face. Te challenge
for a mean feld theory is to reproduce these.
Accordingly, detailed comparisons between measured, analytical calculations and numerical mod-
eling needed to be made [WPWH07]. An early observation was that the calculated results showed
much fner detail than was observed experimentally. In part, this diference is a consequence of the
fnite size of the elements in the metamaterial: no component of the feld pattern with a spatial wave-
length smaller than the size of an element can be sustained, so there is an efective cut-of at high
transverse spatial frequency or wavevector. Moreover, the fnite size of the prism itself leads to a
minimum for the transverse wavevector. Tus for accurate modeling, the range of wavevector needs
to be restricted. Tis can be implemented in the analytical model with some success, but it is not
possible to impose such limits directly in the numerical model, so the efect of the upper limit on
wavevector was approximated by moving the source further away from the prism, thus making the
incident feld pattern more difuse and reducing the high spatial frequency components of the feld
on the input face.
14.3.5.1 The Square Prism
Te measured results for the magnetic feld patterns on the output face of the square prismare shown
in the central column of Figure I4.3, starting at the highest frequency. Te frst point to note is
that in the negative
z
regime, the boundary conditions at the edges of the prism require H
z
= 0.
Tus, at the highest frequency (28.3 MHz), where we observe a uniform mode, it has H
z
= 0 at the
edges. As the frequency is reduced, all intensity fades and thena sequence of resonant patterns appear,
with frst fve and then nine high intensity spots, and then increasingly complicated patterns evolve
as the frequency is reduced towards f
0
. We also note in Figure I4.3 that the measured patterns have
a granularity: this is due to the size of the individual elements of the metamaterial and we cannot
expect any calculation based on an efective medium model to reproduce this.
In the lef-hand column of Figure I4.3, we show the feld patterns calculated using the analytical
theory of Section I4.3.3, along with the periodic boundary conditions discussed there, implemented
for the situation when the source was taken to be 20 mm behind the rear face of the prism and the
wavevector integral inEquationI4.20 truncated at a spatial frequency corresponding to the roll diam-
eter. First, we note that this model does not produce a result at the highest frequency. Here
z
0,
so there is an extremely large mismatch between the medium and the vacuum, and hence little feld
penetration. Tus the predicted transmission in this frequency region is essentially zero. Te corre-
lation between the other calculated and measured patterns is better than when the source lies 3 mm
behind the prism, and we can clearly see the basic structure of the feld patterns is correctly produced.
However, the agreement is not particularly good: overall, the features are rather smaller and sharper
than those observed.
In the right-hand column are the results of the numerical simulation, for the case when the source
is 20 mmbehind the prism. Here we see an excellent agreement between the model and the measured
data: not only is the basic structure of the patterns correctly given, but also the size and shape of the
(a)
(b)
(c)
(d)
(e)
FIGURE I4.3 Comparison of (center column) the feld patterns observed 3 mm from the exit face of the square
metamaterial prism at (a) 28.30 MHz, (b) 24.70 MHz, (c) 23.63 MHz, (d) 23.23 MHz, and (e) 22.30 MHz, with the
patterns calculated using the analytical model (lef column) and the numerical model (right column).
high intensity regions are well described. Tis model is able to calculate the correct pattern at the
high frequency end (Figure I4.3a, 28.3 MHz). Between the two highest resonances, however, some
extra structure was calculated except when a plane wave source was used. At the lower frequencies,
below23 MHz, the agreement between the measured and calculated feld patterns is extremely good,
and the whole progression from one pattern to another down and through the resonance frequency
is correctly described by the numerical simulation. Te results are shown for 24.7, 23.63, 23.23, and
22.30 MHz as in Figure I4.3b through d, respectively.
14.3.5.2 The Hexagonal Prism
Magnetic feld patterns have also been measured for a hexagonal prism, for both the axial and radial
felds. At the highest frequency, 29.7MHz, a simple drum-head-like resonance is observed, with H
z
being maximum at the center and zero at the edges of the prism. Conversely, the radial feld is zero at
the center and maximum at the edges, and points uniformly outward. As the frequency is reduced,
the intensity fades until the next resonance at 24.7 MHz, where a central peak and a ring of intensity
is seen in which the sign of H
z
is reversed. In the radial feld, we see the complementary pattern.
As the frequency is further reduced, additional rings of intensity, modulated by the hexagonal
symmetry of the prism, appear. Te results are shown in Figure I4.6, as the lef-hand frames in each
set of data.
29.7 MHz 24.7 MHz
23.2 MHz 22.6 MHz
FIGURE I4.6 Comparison of measured and modeled resonant feld patterns for the hexagonal prism sample. In
each frequency group, the lef frames are the measured patterns and the right frames are calculated. Te upper pair of
frames in each case showthe axial feld amplitude, H
z
, whereas the lower pair showthe radial feld amplitude, H
rad
,
whose direction is shown by the arrows.
Te right-hand frames of the data sets in Figure I4.6 show the results of the numerical simulation,
again with the source placed 20 mmbehind the prism, as described in the previous section. It is clear
that the agreement between the measured patterns and the simulated results is extremely good, both
for the axial and radial feld components. As for the square prism, in the high frequency regime,
between the frst two resonances (at 29.7 and 24.7) the numerical simulation shows additional struc-
ture that is not observed in the measurements. As pointed out above, this region is better described
by simulations using a plane wave source rather than a fnite sized loop source. As the frequency
is reduced towards the resonance of the individual rolls, however, the sequence of patterns and the
progression from one to another is well described by the numerical simulation.
14.3.6 Discussion
Te results above show that a numerical simulation based on an efective medium description of a
magnetic metamaterial is able to give a very good description of the observed spatial resonances in
the feld patterns around the material samples. However, this was achieved by modifying the actual
experimental layout: with the feld source in its correct position, additional structure was present in
the calculations that did not appear in the measurements. Tis indicated that high spatial frequency
components arising from the fnite size of the source continued to be present in the calculation,
although they were not observed in the measurements. Clearly, spatial frequencies greater than that
set by the unit cell cannot be sustained in the real material, but are present in the modelthere is
no cutof mechanism in an efective medium model. To some extent, this restriction on high spatial
frequencies was simulated by moving the source further away from the sample, to a distance several
times the source diameter, so that the highest spatial components were attenuated before impinging
on the material. As shown above, this approach has been very successful in the lower frequency
regime, but less so between the frst two resonances; here the plane wave excitation (i.e., launching a
uniform magnetic feld in the model) gives the best result.
We have also considered an analytical approach within the efective mediumframework, but again
there is additional, sharper structure in the calculation than is observed in experiment. Attempts to
impose a spatial frequency cutof in the analytical models (e.g., by constraining the upper limit of
integration in Equation I4.20) have been partially successful: the pattern details are indeed smeared
out, and the characteristic features of the spatial resonances are reproduced, but the details of the
patterns are not correct. Nevertheless, overall this simple model gives a surprisingly good account of
the experiments.
Te alternative description of the Swiss roll medium as an array of coupled resonators has been
investigated out by Zhuromskyy and coworkers [ZSS03], using data extracted from a linear array
of these rolls [WSYS04], and the response of a hexagonal prism as a function of frequency was cal-
culated. Tis calculation showed very similar features to those described here. In particular, as the
frequency is reduced from well above f
0
, a frst, uniform resonance is predicted. As the frequency is
reduced, the intensity falls, rising again at the next resonance; this has the central peak and a further
ring of intensity, as seen in our hexagonal prism at 24.7 MHz. Tere is no structure in the pattern
between these two resonances. Similarly, no extra structure is predicted between here and the next
resonance, corresponding to the measured pattern at 23.2 MHz. Tereafer, however, much detailed
structure is predicted: indeed, this persists below f
0
, and this is not observed experimentally. Tus,
the situation regarding additional structure is reversed: whereas in the efective medium model this
appears at the higher frequencies and the behavior near f
0
is correctly predicted, in the coupled
resonator approach the reverse is true.
A possible explanation for these observation may be seen by considering the dispersion relation,
which is plotted over the frequency region of interest as Figure I4.7 for two values of the transverse
wavevector, k
x
, corresponding to the prism size and to the element size. Tis fgure shows that at
a given frequency, a higher k
x
demands higher k
z
. However, one might expect that the efective
medium models would not be accurate for very large k
z
, especially for those values corresponding
to wavelengths much smaller than the thickness of the prism, i.e., k
z
30 or /0.06, shown as the
dashed vertical line in Figure I4.7.
Although the efective medium certainly can support large k
z
, the actual Swiss rolls probably
cannot: no variations in amplitude or phase were observed along the length of 200 mm long rolls
excited by a loop at one end as was done in [WPY
+
0I]. Te impact of such a restriction is that the
efective medium model is good for the lower k
z
, and hence for the lower frequencies, but breaks
down at higher frequencies when there is no mechanism within the model to restrict the k
z
and
hence the k
x
. Te converse appears to be the case for the coupled resonator description.
Finally, because the wavelength of electromagnetic radiation at these frequencies is so long com-
pared toany lengthscale inthe experiment, we expect the electric and magnetic felds to be essentially
independent of one another. Accordingly, an equivalent dielectric model, with the same dispersion
parameters but with an electric dipole excitation, should show the same results. Tis situation, of
course, corresponds to the better known plasmon resonances, but on an interface that lies between a
dielectric ( positive) and a metal ( negative). We have carried out the MWS calculations for such
a system, and indeed fnd that the results for the electric feld are the same as those for the magnetic
system. Accordingly, we can think of the resonances that we measure in the feld patterns as being
due to magnetic plasmons [PO02].
35
33
31
29
27
25
23
21
19
17
15
0 20 40
Wavevector k
z
/m
1
F
r
e
q
u
e
n
c
y
/
M
H
z
60 80 100
FIGURE I4.7 Dispersion curves of frequency vs. k
z
plotted for k
x
= 3 (full line) and I2 (dashed line), correspon-
ding to the two limiting values imposed by the roll diameter and the sample size. Te vertical dashed line corresponds
to the prism thickness being half a wavelength.
14.4 RF Imaging
Tere are two approaches to imaging with RF metamaterials, which we will denote as nonfocussing
(the endoscope) and focussing (the lens). Inthe former case, the material is used onresonance, sothat
the permeability is very large, and signal is guided according to Equations I4.I7 and I4.I8 from one
face of the material to the other, so that the image of the fux pattern on the entrance face appears on
the exit face. In the second case, the material is used at the frequency where = I, the prescription
for the perfect lens [Pen00], and an image (that would be perfect if the material were lossless) is
formed in free space at a distance from the object equal to twice the metamaterial thickness. Both
approaches have been reported, and will be discussed below.
14.4.1 Flux GuidingInitial Demonstration
In the frst demonstration of Swiss rolls used for fux guiding [WPY
+
0I], the bulk material was made
up of a bundle of I9 rolls in a hexagonal close packed array. Tese initial Swiss roll structures were
constructed using ProFilm Chrome [a proprietary aluminized mylar flm, about 30 mthick, with
a thermosetting glue layer], which was wound on mandrels 200 mm long, made of glass-reinforced
plastic (GRP) rod. Following initial characterization, a material was designed and made for use at
the 2I.3 MHz operating frequency of a Marconi Medical Systems (Cleveland, Ohio) Apollo 0.3T MRI
machine.
Te coupling (S
2I
) between two short coils, linked by one of the Swiss rolls, was measured as a
function of their separation. Figure I4.8 shows the coupling between the coils (S
2I
) at 2I.3 MHz, plot-
ted as a function of the separation of the two coils. Te dashed line shows the result without the Swiss
roll present. When a Swiss roll was inserted so that the drive coil was 23 mm from the end of the roll,
the full line was obtained. It is clear that the Swiss roll acts as a fux-guiding medium, providing
70
0 50 100 150 200 250
Coil separation (mm)
C
o
u
p
l
i
n
g

(
d
B
)
60
50
40
30
20
FIGURE I4.8 Measured coupling, S
2I
, as a function of the separation of two coils with a Swiss roll inserted between
them (solid line) and with the Swiss roll removed (dashed line). Te fxed coil was placed 23 mm from one end of the
Swiss roll; the extent of the Swiss roll is indicated by the shading.
Thumb
Coil
200 mm
Swiss
rolls
Water
phantom
(a) (b) (c) (d)
FIGURE I4.9 Te MRI imaging experiment: (a) Schematic of setup. A small coil (diameter 37 mm) acts as the
receiver, and a thumb is the object to be imaged. Te water phantom provides a reference plane. Te 200 mm space
between the phantom and the thumb is flled either with an inert plastic block (not shown) or with Swiss rolls.
(b) Areference image obtained with the body coils that are built into the structure of the magnet, showing the thumb
and the reference plane. (c) Te image from the small receiver coil when the thumb is supported on an inert plastic
block. Only the phantomis visible. (d) Te image from the same coil when the Swiss rolls are inserted. Now the image
of the thumb can be clearly seen.
linkage between coils that may be up to I30 mm apart in this case. Note that there is little fux
leakage along the length of the core, which is qualitatively diferent from what would be observed
for a conventional magnetic core with the same permeability of 2.
Tis Swiss roll metamaterial was then applied in the MRI environment [WPY
+
0I]. Te bundle
of rolls was used to duct fux from an object to a remote detector, with the results being shown in
Figure I4.9. Since the metamaterial used inthese experiments was lossy, all the positional information
inthe image was provided by the spatial encoding systemof the MRI machine (see also SectionI4.3.I).
Nevertheless, it was clear from this work that such metamaterials could perform a potentially useful
and unique function.
14.4.2 Flux GuidingHigh Performance Material
To improve the fux-guiding performance, a material with a lower loss is necessary. To achieve this, a
thicker metal layer is required, along with a low loss dielectric, and hence adhesiveless construction.
Very thin fexible printed circuit board (pcb) provides a good material for this purpose, and Swiss
rolls have been made with a Q 60 at a resonant frequency near 2I.3 MHz, by rolling approximately
II turns of the material Espanex SC-I8-I2-00-FR, which consists of an adhesiveless laminate of an
I8 mcopper sheet with a I2.3 mpolyimide layer, onto a I0 mmdiameter Delrin mandrel. Te efec-
tive permeability of the Swiss roll medium was determined as described in Section I4.3.I by inserting
a roll into a long solenoid, and measuring the changes in the complex impedance that result. On
resonance the peak imaginary value of
= 33 was found. Tese rolls were assembled as a hexag-

onal array in a balsawood box to create the prism of material whose behavior above resonance was
described in Section I4.3.3.
At 2I.3 MHz, the wavelength in vacuo is about I4 m, so the length scales are very much less
than a wavelength and the losses will dominate in Equation I4.I3. At resonance, we can write the
permeability as
z
(
res
) = i
2
(I4.23)
Assuming that k
2
x
/
2
>> k
2
0
, Equation I4.I3 gives
k
2
z
i k
2
x
/
2
(I4.24)
Tus for fnite loss, k
z
has an imaginary component, and the material does not transport the image
perfectly: the higher Fourier components degrade faster with distance. For a material thickness d the
attenuation will become signifcant when Im(k
z
)d I or, using Equation I4.24 , when k
x
(max)
/d. Te resolution is therefore limited to
I/k
x
(max) d/ (I4.23)
In the present case, 6 and d = 60 mm, so that I0 mm, approximately equal to the diameter
of the individual rolls. Tus, we do not expect loss efects to degrade the resolution of any transmitted
structure beyond the intrinsic granularity of the Swiss rolls.
Figure I4.I0 shows both the cross-section and the plan view of both the measured and simulated
results. Tese plots demonstrate convincingly that the face-plate behavior is obtained in a homo-
geneous (but strongly anisotropic) efective medium, and is not just due to guiding through the
individual Swiss rolls. Moreover, in Figure I4.I0e, we plot a comparison of the measured profle
(dots) with an analytic calculation based on the efective medium formalism that we reported pre-
viously (dashed line) and the profle obtained from the present numerical calculation (full line).
Te detailed structure in the measured data arises because we sample discrete rolls, and fux is
trapped inside the individual elements. Clearly, this efect cannot be represented in an efective
medium approximation, so the comparison should be made between the envelope of the data points
and the calculated profles. Te agreement between the two calculated profles is very good over a
wide intensity range (note the logarithmic scale), and both are accurate envelopes for the measured
points.
To test the two-dimensional imaging performance of the material, an antenna was constructed
from a pair of antiparallel wires, bent into the shape of the letter M (Figure I4.IIa). Tis gener-
ated a line of magnetic fux, so providing a characteristic feld pattern for imaging. It was placed
0
(e)
10
20
30
40
50
50 40 30 20 10 0
Position/mm
I
n
t
e
n
s
i
t
y
/
d
B
10 20 30 40 50
(a)
(c) (d)
(b)
FIGURE I4.I0 Axial magnetic feld (H
z
) patterns from the hexagonal prism on resonance: (a) Measured intensity
(dB) in the XZ plane; (b) amplitude in the XYplane, 3 mmabove the prism; (c) modeled intensity (dB) in the XZ plane,
showing the jet of fux propagating through the material; (d) amplitude in the XY plane, 3 mm from the prism; and
(e) Comparison of measured and calculated profles: the points are measured data with the dotted line being a guide
to the eye, the dashed line is the analytical profle and the full line is the profle from the numerical calculation.
100
80
60
40
20
0
0 20 40 60
X Distance/mm
Y

D
i
s
t
a
n
c
e
/
m
m
80 100 120
FIGUREI4.II Te M-shapedantenna, constructed fromtwo antiparallel wires heldI mmapart, and the feld pattern
observed at 2I.3 MHz in a plane approximately 2mmabove the surface of the metamaterial slab. Te Swiss roll structure
is overlaid.
horizontally, and the material was positioned on top of it. Te transmitted feld was measured
by scanning a small probe above the surface of the material, and the pattern thus observed at
2I.3 MHz is shown in Figure I4.IIb, in which the Swiss roll structure is overlaid on the feld pattern
[WHP
+
03].
Figure I4.II clearly shows that the material does indeed act as an image transfer device for the
magnetic feld. Te shape of the antenna is faithfully reproduced in the output plane, both in the
distribution of the peak intensity, and in the valleysthat bound the M. Tese mimic the minima
in the input feld pattern either side of the central line of fux. Te upper right arm of the M itself
was twisted, so that the fux pattern was launched with a much reduced vertical component. Tis is
reproduced in the weaker intensity observed in this region.
14.4.3 RF Focusing
To observe focusing efects [Pen00], the material should be isotropic and have a refractive index
n = I. If we confne our attention to the (x, z) plane (where z is the direction of propagation and
x is a transverse direction) such material can be constructed by stacking alternate layers axially and
transversely to make a 2-Dlog-pile [WPH06]. Moreover, it was also pointed out in [Pen00] that when
all relevant lengths are very much less than the wavelength, the electric and magnetic components of
electromagnetic radiation are decoupled. In this very near feld regime, therefore, a magnetic signal
can be focused in the very near feld using material with permeability = I, but materials with
this property are not found in nature, so this focusing had not previously been observed. However,
using metamaterials allows us to construct materials with the specifed = I, and, by working at
radiofrequency (RF), the requirements that boththe material elements and the measurement distance
should be much smaller than a wavelength are readily satisfed.
Te resolution enhancement, R, that can be achieved with a negative index slab was calculated ana-
lytically in [PR02,Ram03,SSR
+
03], where it was shown that the limit of resolution, , is determined
by the loss in the material, which here is
, and by the length of the sample, d, to be

R =

=
I
2
ln(
/2)

d
or =
2d
ln(
/2)
(I4.26)
Using the measured values, we obtain
= 0.I4 when = I, so that I40 mm and R 90.

Alog-pile was made [WPH06] fromSwiss rolls, arranged so that the transverse and axial permeabil-
ities were equal. Magnetic feld sources consisting of long thin loops that each generated a line of
magnetic fux were placed 20 mm behind the slab, and a similar loop was used to measure the
magnetic feld, H
z
, in the output space [WPH06].
Te experimental data were analyzed to determine the frequency at which
= I; this was found

to be 24.33 MHz. Te data were then plotted to show the spatial distribution of the magnetic feld
in the image space Tis is shown in Figure I4.I2. Here, Figure I4.I2a shows the distribution of H
z
arising just from the two sources, spaced I00 mm apart. We note that there is no discernible struc-
ture at I20 mm from the source plane. When the slab of metamaterial is introduced in the position
indicated, the felds in the image plane are enhanced by a factor of I3, and signifcant structure is
obtained (Figure I4.I2b). Near the surface of the metamaterial, there are strong felds with rapid spa-
tial variationnote that the intensity scale is the same in both frames. In the image plane, indicated
by a dashed line in Figure I4.I2a and b, distinct modulation can be seen. Plotting the feld magnitude
as a function of position in the image plane shows two peaks (Figure I4.I2c). As the source separa-
tion is increased, the weakly modulated peak observed at the lowest value of I00 mm, is split into two
peaks with appropriate spacing and increasing contrast. Tis confrms that the structure in the image
plane does indeed arise from imaging the sources.
Te performance of an imaging system is defned by the transfer function, which describes the
(complex) transmission of the system as a function of the spatial frequency. Te formula of [SSR
+
03]
200
160
120
80
40
0
0.10
F
i
e
l
d

a
m
p
l
i
t
u
d
e

|
H
z
|
T
r
a
n
s
f
e
r

f
u
n
c
t
i
o
n

|
|
0.08
0.06
0.04
80 40 0 40 80
120
(a)
(c) (d)
(b)
1.0
0.8
0.6
0.4
0.2
0.0
0 50 100 150
80 0 40 40 80 120 120 80 0 40 40 80 120
X distance/mm
X distance/mm
X distance/mm
Transverse wavevector (k
x
/k
0
)
Z

d
i
s
t
a
n
c
e
/
m
m
X10
FIGURE I4.I2 Measured distributions of H
z
feld intensity (dB) at 24.33 MHz: (a) Fromtwo sources spaced I00mm
apart in free space; (b) as (a) but with the metamaterial slab in place. In these frames, the position of the metamaterial
is indicated by the thin white line, and the image plane by the dashed black line. (c) Te variation of the feld amplitude
in the image plane z = I20 mm without the metamaterial (full line, X I0), and with the metamaterial when the sources
are spaced I00mm (dashed line), I20 mm (dotted line), and I40 mm (dash-dot line) apart. (d) Measured (points) and
calculated (lines) transfer function for a 60 mmslab of metamaterial with
= I at 24.33 MHz: dashed line
= 0.I4,
full line
= 0.26.
was used to calculate the transfer function using the predicted value of
= 0.I4. Tis is shown as the

dashed curve in Figure I4.I2d. Te transfer function was measured and the resulting points were also
plotted in Figure I4.I2d. It is clear that the actual value of
is rather larger than that estimated from

measurements of a single element. Aleast-squares ftted value is
= 0.26, and the transfer function

for this value is plotted as the full line in Figure I4.I2d. Te Rayleigh criterion was applied to estimate
the resolution as /64, although the measurements actually display a rather higher resolution, as
indicated by the high spatial frequency tail in Figure I4.I2d.
14.4.4 Discussion
Te frst point that should be noted for these imaging mechanisms is that the resolution is inde-
pendent of the wavelength of operation, see Equations I4.24 and I4.23. Tus by working at long
wavelength in the RF, it is simple to achieve massivley sub-wavelength resolution. Te absolute reso-
lution (i.e., itself) depends critically on the loss in the material: the smaller the loss, the better the
resolution. In the case of the fux guiding endoscope, the achievable resolution is also limited by the
fnite size of the individual elements, because the image is formed at the surface of the material. In
the case of the lens, on the other hand, the image is removed from the surface, so that the efect of
granularity, which only persists to a distance about equal to the element size, is smeared out. Tus,
to achieve improved performance of the endoscope, the material needs to be made from smaller ele-
ments with lower loss, whereas to improve the lens just needs lower loss in the material. However,
the improvement that can be achieved in the case of the lens is limited because of the logarithmic
dependence of the resolution on the imaginary part of the permeability.
It should also be noted that these mechanisms only show focusing in the transverse plane: there is
no focusing in the longitudinal direction, and conventional three-dimensional image formation does
not occur. Tis is self-evident in the case of the endoscope where the image is formed on the output
face. For the lens, the sub-wavelength features in the image arise from the presence of the evanescent
feld components, which are amplifed by the material in such a way that they combine to form the
image observed. Tus the feld distribution between the lens and the image plane is complicated and
shows no longitudinal focussing. Such longitudinal focussing has been obtained [FM03,MFMB03],
but it relies on the detector perturbing the output feld, which does not occur here, and, indeed, is
generally undesirable.
14.5 Applications
Tere are many potential applications of RF metamaterials, but the key feature that will govern how
well they perform is their dispersion: their permeability is a strong function of frequency. Terefore,
the most promising applications are those with narrow bandwidth demands. One possible use is
in antenna applications, where the ability to tailor the permeability of the substrate opens up a wide
range of novel antenna concepts; these are explored elsewhere in this Handbook. Another application
area is inmagnetic resonance, where the signal bandwidth is small, but there are other magnetic felds
present which must not be perturbed. MRI is a particularly promising area in which metamaterial
components might be used.
In an MRI system, the main magnetic feld (typically 0.33 Tesla) needs to be homogeneous to
a few parts per million, thus ruling out the introduction of any conventional magnetic material.
Nevertheless, it would be very useful to have access to magnetic materials with which to manipulate
the radiofrequency (RF) signals (in the range 20I20 MHz). Functions such as guiding, focussing and
screening could substantially enhance the performance of MRI systems. Metamaterials can achieve
this because they ofer a means of obtaining magnetic properties at RF (for example large positive
or negative permeability) without afecting the other magnetic felds in the system: here we describe
work on a faceplate, a yoke, waveguides and fux compressors, all of which could have application in
an MRI environment.
14.5.1 RF Endoscope/Faceplate
Te frst demonstration of RF metamaterials being used in an MRI environment [WPY
+
0I] was
described earlier (Section I4.4.I). Further experiments on this concept have been carried out with
the improved material described in Section I4.4.2. To do this, the M-shaped antenna was tuned
and used to excite localized feld patterns in an NMR-visible polymer sheet (Spenco), as shown
in Figure I4.I3. Te NMR signal from the Spenco sheet was detected by using the M-antenna in
transmit-receive mode. In the control experiment (Figure I4.I3a), the RF excitation level was set so
that signals were received only from locations close to conductors, so that the conductor pattern
was directly visualized in the image using multi-slice spin echo imaging performed in planes parallel
to the sheet surface, and these revealed the expected fux patterns (Figure I4.I3c). When the meta-
material was placed between the excitation coil and Spenco sheet (Figure I4.I3b), no change of RF
excitation amplitude was required to maximize signal, and the fux pattern was directly transferred
with geometry preserved (Figure I4.I3d) [WPL
+
03].
Excited spins
Excited spins
M Antenna
(a)
(b) (c) (d)
Reference
Through Swiss Rolls
M Antenna
Swiss Roll Slab (60 mm)
Spenco
Spenco
FIGURE I4.I3 Te MRI experiment with the metamaterial faceplate. (a) Te reference layout with the Spenco
placed directly on the M-antenna; the measured image is shown in (c). (b) Te Swiss roll prism is inserted between
the antenna and the Spenco; the image obtained using the M in a transmit-receive mode through the Swiss rolls is
shown in (d).
Behr et al. [BHJ04] also used a Swiss roll as a fux guide in an NMR system. It was used to couple
a small solenoid that acted as both source and detector for the RF signal, in this case form the
3I
P
resonance at 34.23 MHz in a 2 Tfeld. Tey found that the signal was transmitted with minimal prop-
agation loss, although there was a signifcant insertion loss. Te potential advantage of using such a
guide is that the signal coils does not have to optimized for each diferent sample geometry.
14.5.2 Yoke
A metamaterial yoke made of Swiss rolls could provide a low reluctance pathway that might assist in
excitation delivery or signal reception in magnetic resonance imaging and spectroscopy applications
[WHYH04].
A set of Swiss rolls designed to resonate at 2I.3 MHz with a Q of 30 were used in the frst experi-
ments. Preliminary tests were made on single rolls, by injecting a signal through a coupling loop at
one end, and recording the detected signal through a second loop that could be moved along the
roll. As shown in Figure I4.8, the detected signal was independent of the position of the receiver.
However, it did depend on the length of the roll, being smaller for longer rolls. Tus the rolls act as
good magnetic fux conductors, and the signal is determined by the reluctance (i.e. the length) of the
fux return path.
Joining the individual rolls together introduces extra loss, so a yoke constructed frombutt-coupled
single rolls would not be viable. To reduce the corner losses, bundles of seven rolls of diferent lengths
were used, so that the corners were mitred at 43 degree, as shown in Figure I4.I4a. Tis arrangement
was much less sensitive to alignment, and signifcantly reduced the corner losses, so that a full yoke
became viable.
Te performance of the yoke was tested by using a remote receiver loop to detect fux circulating
through the metamaterial bundles from a source loop placed between the pole pieces. Te reference
level was defned when the two loops were in a co-planar confguration without the metamaterial,
and the signal being guided around the yoke was then measured (Figure I4.I4b). At frst sight, this
result shows perfect coupling, but it must be recalled that this is a resonant system, so we expect the
signal on resonance to be much higher than the reference, so although this result is encouraging, it
also shows that the device must be much improved to be truly useful.
1.0
0.8
0.6
0.4
0.2
0.0
20.0
(a) (b)
21.0 22.0
Frequency/MHz
S
i
g
n
a
l

t
h
r
o
u
g
h

Y
o
k
e
/
R
e
f
e
r
e
n
c
e
23.0 24.0 25.0
FIGURE I4.I4 (a) Schematic diagram of an RF yoke, constructed from bundles of Swiss rolls; the bottomarm con-
sists of two sets of pole pieces with the sample volume between them. Te star in the sample volume denotes either the
source point or the required excitation volume within a sample, and the lines on the top and bottom arms represent
alternative positions for exciting or receiver coils. (b) Te signal recorded by the remote detector demonstrates that the
fux is being guided from the source through the Swiss roll bundles.
TABLE I4.I Comparison of Junction
Losses for Straight and 90
Joints in
Pyralux, Espanex, and Coupled Systems
Signal (dB) Straight Joint 90
Joint
Pyralux 3.9 II.3
Espanex 2.7 6.8
Espanex + coupler I.I 0.9
Te higher performance material described in Section I4.4.2 has also been tested. Tis has a per-
meability and Q that are at least a factor of 2 higher than in the Pyralux material above. Te signal
down a 200 mm roll was increased, showing much improved fux ducting. However, the losses at
joints, while reduced compared to those in the Pyralux system, are still unacceptable (see Table I4.I),
so the efect of additional couplers in the form of two connected loops that link the end of one roll
with the next, was investigated. Tis signifcantly improves the fux linkage, as shown in Table I4.I,
but further work is necessary to optimize this approach.
Allard et al. [AWHH03] modeled the performance of such a yoke in an MRI system using an
efective medium model. Tese calculations showed that signifcant signal gain should be obtained
when the yoke cross-section is approximately the same as the width of the gap in the yoke, and the
permeability is as large as possible. With achievable values of the permeability ( I00), a signal
gain of 23 dB was predicted. In further work [AH06], they used an FDTD model to calculate the AQ2
currents in a circuit model of the Swiss roll structure, and concluded that there would in fact be little
gain, due mostly to the fnite Q of the rolls. Tey did point out, however, that there is almost no
leakage of magnetic fux from the rolls, a point noted in [WPY
+
0I] even for low-Q material.
14.5.3 Waveguides
Te concept of the magneto-inductive waveguide was introduced by Shamonina et al. [SKRS02a],
who derived the dispersion relations and considered the current distribution and power fow
through the line. Te impact of unmatched ends and the coupling between two guides was treated
by Shamonina and Solymar [SS04], and further waveguide device concepts were presented by
Syms et al. [SSS06].
Syms et al. [SYS06] discuss the performance of waveguides constructed from arrays of up to 70
metamaterial elements, formed into a planar ring resonator at I30 MHz, and reduce the propagation
loss to 0.I2 dB per element. Consideration also has to be given of the termination of such guides
[SSS03], so that the insertion loss is minimized. Sydoruk et al. [SRZ
+
06] built waveguide structures
from their split-pipe elements and considered the coupling between the lines. By assembling the
elements in a ring, Solymar et al. [SZS
+
06] proposed a device with a rotational resonance, which
potentially could be used in detection of NMR signals.
14.5.4 Flux Compressor
To use our metamaterials successfully in the MRI environment, it will be necessary to develop various
components which can be incorporated into the metamaterial assemblies. One such component, a
fux compressor, is intended to collect signal from a signifcant area and output it to a much smaller
one or vice versa. A prototype device [WSSY04], shown in Figure I4.I3a, was made from I9 turns of
I mmdiameter wire, wound on a tapered mandrel, with maximumand minimumdiameters of I0 and
3 mm, respectively. Its length was 20 mm. Te compressor was tuned with 6.8 pF to give a frequency
of 67 MHz. Te quality factor was Q I63.
Te compressor was tested by placing the wide end on a I0 mm diameter transmitter loop, and
measuring the transmission to a receiver loop (3 or I0 mm in diameter) at the other end. Te dif-
ference between the 3 and I0 mm loop measurements at the compressor tip was only I dB, whereas
the diference in reference signal in this plane was I7 dB. Tus it is clear that the fux was confned
to the 3 mm exit diameter and the compressor concept is valid. On resonance, we see a signifcant
enhancement of the compressed signal, and such a device could play a role in coupling elements of
unequal size (for example coupling a pole-piece to a MM yoke).
(a) (b)
FIGURE I4.I3 Prototype fux compressors: (a) Te tapered solenoid device and (b) the resonant ring structure .
.
An alternative device can be built from a sequence of resonant loops, wound on diferent diame-
ter formers and all tuned to the same frequency (see Figure I4.I3b). CoilCoil interactions result in
magneto-inductive (MI) waves which propagate along the coupled coils. Measurements show that
the device can operate over an appreciable bandwidth, and that the overall transmission levels are
little afected by the precise spacing of the coils, so that the design of these devices appears quite
robust.
14.6 Conclusion
Inthis chapter, we have reviewed the development and properties of metamaterials in the RF band. At
these lowfrequencies, the wavelength of electromagnetic radiation is very long, so that the condition
for homogenization, i.e., that the structure and its unit cell should be much smaller than the wave-
length, is easily satisfed. Moreover, because measurements can be performed in the very near feld,
it is only necessary to manipulate the permeability of the material in order to control the behavior of
RF magnetic felds.
Te basic magnetic element that is used at higher frequency, the split ring resonator, is not suitable
for use at RF, but an alternative structure, the Swiss roll is ideal. It is compact, can be made to resonate
at frequencies as low as 3 MHz, and displays intense magnetic activity: when assembled into a bulk
material, the negative permeability region extends to a bandwidth /
0
40. Moreover, there is
a wide range of permeability, from large positive through zero to large negative values, to explore.
Te applicability of the mean feld or efective medium approach has been tested by measuring
the feld patterns that are induced on the surface of metamaterial prisms when excited by a point
source on the opposite surface, and comparing the data to those calculated assuming the prism to be
a homogeneous block of material with an efective permeability derived frommeasurements. At high
spatial frequency, mean feld theory breaks down, because the material is granular and the theory
contains no mechanism for limiting the spatial frequencies that propagate. For all other cases, the
agreement between the measured and the calculated distributions is excellent.
Te materials have been used to demonstrate sub-wavelength imaging, both as an endo-
scope/faceplate and as a lens. In the former case, the material is anisotropic and used on resonance,
when the permeability is large, and acts to transfer the feld pattern from the input face faithfully to
the output face. Te lens requires an isotropic material with a permeability of = I that focusses
both the propogating and evanescent waves to produce a perfect image. Te performance of both
mechanisms is in accord with theory, and is dominated by the losses in the material.
Finally, possible applications in the felds of magnetic resonance imaging have been explored, and
potential device prototypes have been tested, to show that metamaterials can indeed perform useful
functions. However, there is much work still to be done before fully practical devices can be realized.
References
[AH06] M. Allard and R. M. Henkelman. Using metamaterial yokes in nmr measurements. Journal
of Magnetic Resonance, I82(2):200207, 2006.
[AWHH03] M. Allard, M. C. K. Wiltshire, J. V. Hajnal, and R. M. Henkelman. Improved signal detec-
tion with metamaterial magnetic yokes. Proceedings of International Society of Magnetic
Resonance in Medicine, I3:87I, 2003. AQ3
[Bal64] K. G. Balmain. Te impedance of a short dipole in a magnetoplasma. IEEE Transactions on
Antennas and Propagation, AP-I2:6036I7, I964.
[BHJ04] V. C. Behr, A. Haase, and P. M. Jakob. Rf fux guides for excitation and reception in (3I)p
spectroscopic and imaging experiments at 2 tesla. Concepts in Magnetic Resonance Part
B-Magnetic Resonance Engineering, 23B(I):4449, 2004.
[BJMZ06] J. D. Baena, L. Jelinek, R. Marques, and J. Zehentner. Electrically small isotropic
three-dimensional magnetic resonators for metamaterial design. Applied Physics Letters,
88(I3):I34I08, 2006.
[BLK03] K. G. Balmain, A. E. A. Luttgen, and P. C. Kremer. Power fow for resonance cone phenom-
ena in planar anisotropic metamaterials. IEEE Transactions on Antennas and Propagation,
3I(I0):26I226I8, 2003.
[BMMM04] J. D. Baena, R. Marques, F. Medina, and J. Martel. Artifcial magnetic metamaterial design
by using spiral resonators. Physical Review B, 69(I):0I4402, 2004.
[EWL
+
03] C. Enkrich, M. Wegener, S. Linden, S. Burger, L. Zschiedrich, F. Schmidt, J. F. Zhou,
T. Koschny, and C. M. Soukoulis. Magnetic metamaterials at telecommunication and visible
frequencies. Physical Review Letters, 93:20390I, 2003.
[FM03] M. J. Freire and R. Marques. Planar magnetoinductive lens for three-dimensional subwave-
length imaging. Applied Physics Letters, 86(I8):I82303, 2003.
[LEW
+
04] S. Linden, C. Enkrich, M. Wegener, J. F. Zhou, T. Koschny, and C. M. Soukoulis. Magnetic
response of metamaterials at I00 terahertz. Science, 306(3700):I33II333, 2004.
[MFMB03] F. Mesa, M. J. Freire, R. Marques, and J. D. Baena. Tree-dimensional superresolution in
metamaterial slab lenses: Experiment and theory. Physical Review B, 72(23):233II7, 2003.
[MHBL99] S. S. Mohan, M. D. Hershenson, S. P. Boyd, and T. H. Lee. Simple accurate expressions for
planar spiral inductances. IEEE Journal of Solid-State Circuits, 34(I0):I4I9I424, I999.
[MMMM03] R. Marques, F. Mesa, J. Martel, and F. Medina. Comparative analysis of edge- and broadside-
coupled split ring resonators for metamaterial designtheory and experiments. IEEE
Transactions on Antennas and Propagation, 3I(I0):2372238I, 2003. Part I.
[MMREI02] R. Marques, F. Medina, and R. Rafi-El-Idrissi. Role of bianisotropy in negative permeability
and lef-handed metamaterials. Physical Review B, 63(I4):I44440, 2002.
[Pen00] J. B. Pendry. Negative refraction makes a perfect lens. Physical Review Letters, 83(I8):3966
3969, 2000.
[PGL
+
03] C. G. Parazzoli, R. B. Greegor, K. Li, B. E. C. Koltenbah, and M. Tanielian. Experimental
verifcation and simulation of negative index of refraction using snells law. Physical Review
Letters, 90(I0):I0740I, 2003.
[PHRS98] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart. Low frequency plasmons in thin-
wire structures. Journal of Physics-Condensed Matter, I0:47834809, I998.
[PHRS99] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart. Magnetism from conductors and
enhanced nonlinear phenomena. IEEE Transactions on Microwave Teory and Techniques,
47:20732084, I999.
[PHSY96] J. B. Pendry, A. J. Holden, W. J. Stewart, and I. Youngs. Extremely low frequency plasmons
in metallic mesostructures. Physical Review Letters, 76:47734776, I996.
[PO02] J. B. Pendry and S. OBrien. Very-low-frequency magnetic plasma. Journal of Physics-
Condensed Matter, I4(32):740974I6, 2002.
[PP64] W. K. H. Panofsky and M. Phillips. Classical Electricity and Magnetism, 2nd edition, Chapter
3.8. Addison-Wesley Reading, MA, I964.
[PR02] J. B. Pendry and S. A. Ramakrishna. Near-feld lenses in two dimensions. Journal of Physics-
Condensed Matter, I4(36):84638479, 2002.
[Ram03] S. A. Ramakrishna. Physics of negative refractive index materials. Reports on Progress in
Physics, 68(2):44932I, 2003.
[RPWS03] S. A. Ramakrishna, J. B. Pendry, M. C. K. Wiltshire, and W. J. Stewart. Imaging the near feld.
Journal of Modern Optics, 30(9):I4I9I430, 2003.
[SKRS02a] E. Shamonina, V. A. Kalinin, K. H. Ringhofer, and L. Solymar. Magneto-inductive waveg-
uide. Electronics Letters, 38(8):37I373, 2002.
[SKRS02b] E. Shamonina, V. A. Kalinin, K. H. Ringhofer, and L. Solymar. Magnetoinductive waves in
one, two, and three dimensions. Journal of Applied Physics, 92(I0):6232626I, 2002.
[SKS04] D. R. Smith, P. Kolinko, and D. Schurig. Negative refraction in indefnite media. Journal of
the Optical Society of America B-Optical Physics, 2I(3):I032I043, 2004.
[SP06] D. R. Smith and J. B. Pendry. Homogenization of metamaterials by feld averaging (invited
paper). Journal of the Optical Society of America B-Optical Physics, 23(3):39I403, 2006.
[SPV
+
00] D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Schultz. Composite
mediumwithsimultaneously negative permeability andpermittivity. Physical ReviewLetters,
84:4I844I87, 2000.
[SPW04] D. R. Smith, J. B. Pendry, andM. C. K. Wiltshire. Metamaterials andnegative refractive index.
Science, 303(3683):788792, 2004.
[SRSNN04] A. F. Starr, P. M. Rye, D. R. Smith, and S. Nemat-Nasser. Fabrication and characterization of
a negative-refractive-index composite metamaterial. Physical Review B, 70(II):II3I02, 2004.
[SRZ
+
06] O. Sydoruk, A. Radkovskaya, O. Zhuromskyy, E. Shamonina, M. Shamonin, C. J. Stevens,
G. Faulkner, D. J. Edwards, and L. Solymar. Tailoring the near-feld guiding properties
of magnetic metamaterials with two resonant elements per unit cell. Physical Review B,
73(22):224406, 2006.
[SS04] E. Shamonina and L. Solymar. Magneto-inductive waves supported by metamaterial ele-
ments: components for a one-dimensional waveguide. Journal of Physics D-Applied Physics,
37(3):362367, 2004.
[SSMS02] D. R. Smith, S. Schultz, P. Markos, and C. M. Soukoulis. Determinationof efective permittiv-
ity and permeability of metamaterials fromrefection and transmission coemcients. Physical
Review B, 63(I9):I93I04, 2002.
[SSNNS0I] R. A. Shelby, D. R. Smith, S. C. Nemat-Nasser, and S. Schultz. Microwave transmission
through a two-dimensional, isotropic, lef-handed metamaterial. Applied Physics Letters,
78:48949I, 200I.
[SSR
+
03] D. R. Smith, D. Schurig, M. Rosenbluth, S. Schultz, S. A. Ramakrishna, and J. B. Pendry.
Limitations on subdifraction imaging with a negative refractive index slab. Applied Physics
Letters, 82(I0):I306I308, 2003.
[SSS0I] R. A. Shelby, D. R. Smith, and S. Schultz. Experimental verifcation of a negative index of
refraction. Science, 292:7779, 200I.
[SSS03] R. R. A. Syms, L. Solymar, andE. Shamonina. Absorbing terminations for magneto-inductive
waveguides. IEE Proceedings-Microwaves Antennas and Propagation, I32(2):778I, 2003.
[SSS06] R. R. A. Syms, E. Shamonina, and L. Solymar. Magneto-inductive waveguide devices. IEE
Proceedings-Microwaves Antennas and Propagation, I33(2):IIII2I, 2006.
[Str4Ia] J. A. Stratton. Electromagnetic Teory, Chapter 8. McGraw-Hill, NewYork, I94I.
[Str4Ib] J. A. Stratton. Electromagnetic Teory, Chapter 3.27-8. McGraw-Hill, NewYork, I94I.
[SVP
+
99] D. R. Smith, D. C. Vier, W. Padilla, S. C. Nemat-Nasser, andS. Schultz. Loop-wire mediumfor
investigating plasmons at microwave frequencies. Applied Physics Letters, 73(I0):I423I427,
I999.
[SYS06] R. R. A. Syms, I. R. Young, and L. Solymar. Low-loss magneto-inductive waveguides. Journal
of Physics D-Applied Physics, 39(I8):3943393I, 2006.
[SZS
+
06] L. Solymar, O. Zhuromskyy, O. Sydoruk, E. Shamonina, I. R. Young, and R. R. A. Syms.
Rotational resonance of magnetoinductive waves: Basic concept and application to nuclear
magnetic resonance. Journal of Applied Physics, 99(I2):I23908, 2006.
[WHP
+
03] M. C. K. Wiltshire, J. V. Hajnal, J. B. Pendry, D. J. Edwards, and C. J. Stevens. Metamate-
rial endoscope for magnetic feld transfer: near feld imaging with magnetic wires. Optics
Express, II(7):7097I3, 2003.
[WHYH04] M. C. K. Wiltshire, R. M. Henkelman, I. R. Young, and J. V. Hajnal. Metamaterial yoke for
signal receptionan initial investigation. Proceedings of International Society of Magnetic
Resonance in Medicine, I2:43, 2004. AQ4
[Wil00] M. C. K. Wiltshire. Structure with switchable magnetic properties. UK Patent Application
GB2360I32A, 2000. (also unpublished work).
[Wil03] W. Williams. Bessl Beams, Resonances and Meatamaterials. PhD thesis, Department of
Physics, Imperial College, London, 2003.
[Wil07] M. C. K. Wiltshire. Radio frequency (rf) metamaterials. Physica Status Solidi B-Basic Solid
State Physics, 244(4):I227I236, 2007.
[WPH06] M. C. K. Wiltshire, J. B. Pendry, and J. V. Hajnal. Sub-wavelength imaging at radio frequency.
Journal of Physics-Condensed Matter, I8(22):L3I3L32I, 2006.
[WPL
+
03] M. C. K. Wiltshire, J. B. Pendry, D. J. Larkman, D. J. Gilderdale, D. Herlihy, I. R. Young, and
J. V. Hajnal. Geometry preserving fux-duxcting by magnetic metamaterials. Proceedings of
International Society of Magnetic Resonance in Medicine, II:7I3, 2003.
[WPWH07] M. C. K. Wiltshire, J. B. Pendry, W. Williams, and J. V. Hajnal. An efective mediumdescrip-
tion of swiss rolls, a magnetic metamaterial. Journal of Physics-CondensedMatter, I9:4362I6,
2007.
[WPY
+
0I] M. C. K. Wiltshire, J. B. Pendry, I. R. Young, D. J. Larkman, D. J. Gilderdale, and J. V. Hajnal.
Microstructured magnetic materials for rf fux guides in magnetic resonance imaging.
Science, 29I(3303):84983I, 200I.
[WSSY04] M. C. K. Wiltshire, E. Shamonina, L. Solymar, and I. R. Young. Development of metamaterial
components for use in mri and nmr systems. Proceedings of International Society of Magnetic
Resonance in Medicine, I2:I382, 2004.
[WSYS03] M. C. K. Wiltshire, E. Shamonina, I. R. Young, and L. Solymar. Dispersion characteristics of
magneto-inductive waves: comparison between theory and experiment. Electronics Letters,
39(2):2I32I7, 2003.
[WSYS04] M. C. K. Wiltshire, E. Shamonina, I. R. Young, and L. Solymar. Experimental and theoreti-
cal study of magneto-inductive waves supported by one-dimensional arrays of swiss rolls.
Journal of Applied Physics, 93(8):44884493, 2004.
[YPF
+
04] T. J. Yen, W. J. Padilla, N. Fang, D. C. Vier, D. R. Smith, J. B. Pendry, D. N. Basov, andX. Zhang.
Terahertz magnetic response from artifcial materials. Science, 303:I494I496, 2004.
[ZSS03] O. Zhuromskyy, E. Shamonina, and L. Solymar. 2d metamaterials with hexagonal structure:
spatial resonances and near feld imaging. Optics Express, I3(23):92999309, 2003.
AUTHOR QUERIES
AQI Please explain PTFE.
AQ2 Please explain FDTD model:
AQ3 Please provide location for this reference [AWHH03].
AQ4 Please provide location for this reference.

C014

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

C014

Загружено:

Авторское право:

Доступные форматы

Filippo Capolino/Theory and Phenomena of Metamaterials 54252_C014 Page proof Page 1 2009-1-28 #5

2 times smaller than that of a commensurate

are the low-frequency and high-frequency limiting values, respectively, of the

= 33 was found. Tese rolls were assembled as a hexag-

, and by the length of the sample, d, to be

= 0.I4 when = I, so that I40 mm and R 90.

= I; this was found

= I at 24.33 MHz: dashed line

= 0.I4. Tis is shown as the

is rather larger than that estimated from

= 0.26, and the transfer function

Вам также может понравиться