JOURNAL OF APPLIED PHYSICS 102, 014901 共2007兲

Composites with tuned effective magnetic permeability

Alireza V. Amirkhizi and Sia Nemat-Nassera兲
Center of Excellence for Advanced Materials, Department of Mechanical and Aerospace Engineering,
University of California, San Diego, La Jolla, California 92093-0416
共Received 6 November 2006; accepted 17 May 2007; published online 2 July 2007兲
Pendry et al. 关J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, IEEE Trans. Microwave
Theory Tech. 47, 2075 共1999兲兴 and Smith et al. 关D. R. Smith, W. J. Padilla, D. C. Vier, S. C.
Nemat-Nasser, and S. Schultz, Phys. Rev. Lett. 84, 4184 共2000兲兴 have shown that the effective
magnetic permeability, ␮, of free space can be rendered negative over a certain frequency range by
a periodic arrangement of very thin conductors with suitable magnetic resonance properties, the
so-called split-ring resonators. Because of its rather bulky architecture, this structure does not lend
itself to a proper integration into a reasonably thin real composite structural panel. To remedy this
fundamental barrier, we invented a new magnetic resonator consisting of very thin folded plates that
are suitably nested within one another to form folded-doubled resonators 共FDRs兲 that can be
integrated into an actual composite panel. Measurements, using a focused beam electromagnetic
characterization system combined with time-domain numerical simulations of the reflection and
transmission coefficients of such a composite slab have revealed that indeed the composite has a
negative ␮ over a frequency range of about 9.1–9.35 GHz 关S. Nemat-Nasser, S. C. Nemat-Nasser,
T. A. Plaisted, A. Starr, and A. Vakil Amirkhizi, in Biomimetics: Biologically Inspired Technologies,
edited by Y. Bar Cohen 共CRC Press, Boca Raton, FL, 2006兲兴. Thus, it has become possible to
construct a structural composite panel with negative index of refraction by simultaneously creating
negative effective ␧ and ␮ 关V. G. Veselago, Sov. Phys. Usp. 10, 509 共1968兲; R. A. Shelby, D. R.
Smith, and S. Schultz, Science 292, 77 共2001兲; A. F. Starr, P. M. Rye, D. R. Smith, and S.
Nemat-Nasser, Phys. Rev. B 70, 113102 共2004兲兴. © 2007 American Institute of Physics.
关DOI: 10.1063/1.2751084兴

I. INTRODUCTION based on resonating current loops. Another important aspect

is that all of the current loops have a gap, since, in the ab-
The overall electric permittivity of composites is modi- sence of a gap, the restraining electromotive force against the
fied by inclusion of arrays of certain conductive components induced current 共assuming perfect conductors兲 would be
such as thin straight wires or coils.1–4 The influence of the rather small. On the other hand, when a gap is introduced in
scattering elements on the propagation of electromagnetic the loop, the induced current charges a capacitor that in turn
waves in a composite modifies its overall 共effective兲 dielec- creates an opposing electromotive force. Pendry et al.1 stud-
tric constant. In most cases, this is achieved by arrays of ied simple loops and concluded that they have a very small
conductors that create electric dipoles in response to the os- magnetic effect.
cillating electric fields; see Fig. 1. The properties of the elec- While the electric dipoles define the effective inductance
tric dipole array determine the overall response of the mate- of the medium, it is the effective capacitance that, along with
rial. For example, the overall inductance of the medium the overall inductance and the geometry of the scattering
dictates the characteristic resonance frequency of the unit array, controls the magnetic properties of the composite. The
cell and therefore the dispersive behavior of the composite. original split-ring resonators increase the capacitance of the
Similarly, in order to modify the magnetic properties of a medium by having two separate resonators that were ar-
composite, one can embed scattering structures within a
composite such that an oscillating magnetic field would cre-
ate current distributions that produce a magnetic dipole reso-
nator. A simple example is a resonating current loop. Pendry
et al.1 studied various examples of such structures in some
detail. They predicted that, using certain designs, one can
render the overall magnetic permeability of the composite
negative. Smith et al.2 experimentally verified these predic-
tions by fabricating one of the structures suggested by
Pendry,1 and measuring its transmission spectrum.
All such examples have the basic property of being
FIG. 1. Electric and magnetic dipole resonators. 共left兲 Wave polarization.
共center兲 An electric dipole resonator induced in a straight conductor and the
a兲
Author to whom correspondence should be addressed; present address: accompanying current density. 共right兲 A magnetic dipole resonator induced
9500 Gilman Dr., La Jolla, CA 92093-0416; electronic mail: sia@ucsd.edu. in a conductive split ring and the accompanying current density.

0021-8979/2007/102共1兲/014901/7/$23.00 102, 014901-1 © 2007 American Institute of Physics

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
177.20.130.9 On: Thu, 14 Aug 2014 16:44:49
 014901-2 A. V. Amirkhizi and S. Nemat-Nasser J. Appl. Phys. 102, 014901 共2007兲

FIG. 2. Split-ring resonators. 共left兲 Schematics of a split-ring resonator with the wave polarization that can excite it. 共right兲 A typical dispersion curve for
effective magnetic permeability normalized with respect to ␮0. Note that the real permeability becomes negative but the imaginary part is always positive as
expected for a passive medium. 共Drawing and graph courtesy of D. R. Smith.10兲

ranged to constitute the two surfaces of a capacitor. As rameters can be calculated based on the geometry of the unit
shown by Starr et al.,3 one may achieve a similar result using cell and the material properties; for details, see Ref. 2.
single split rings that are modified to have sufficient capaci- The index of refraction, n = 冑␮␧, becomes imaginary in
tance to suitably reduce the effective magnetic permeability the negative magnetic permeability frequency band in which
of the composite. there are no propagating modes within the composite, the
only possible modes being the evanescent waves. All the
II. SPLIT-RING RESONATORS energy of a wave in this frequency regime will be either
reflected or absorbed within a thin layer of the composite.
The first structure that was extensively studied and Smith et al.2 verified the existence of this stop band using a
showed a considerable effect on the overall magnetic re- scalar network analyzer and measuring the transmitted power
sponse of the medium consisted of split-ring resonators. It in a waveguide filled with a split-ring resonator array.
was introduced and analytically studied by Pendry et al.1 and
later fabricated and tested by Smith et al.2 Its basic elements
and a typical dispersion curve for its normalized overall III. ORIGINAL FOLDED-DOUBLED RESONATORS
magnetic permeability are shown in Fig. 2. As is seen, the The split-ring resonator design is easily tunable and
real part of the magnetic permeability becomes negative but demonstrates a relatively wide band with negative effective
its imaginary part is always positive, as expected from a permeability. Consequently, the corresponding permeability,
passive medium. ␮共␻兲, varies slowly with frequency within this band. It per-
Pendry et al.1 gave an analytical prediction for the effec- mits tuning within a wide-band frequency. However, the
tive magnetic permeability based on the line and surface in- split-ring resonators cannot be easily integrated into a real
tegral averaging method.5,6 A somewhat modified formula, structural composite of limited thickness, since the plane on
used by Smith et al.,2 is which the split-rings rest must be perpendicular to the plane
␮共␻兲 ␻20F of the composite panel, parallel to the wave vector, to inter-
=1− 2 . 共1兲 act with the incident waves that are normal to the composite
␮0 ␻ − ␻20 + i⌫␻
panel. Ideally, one would like to be able to stack layers of
Here, F is a geometric quantity that can be interpreted as the magnetic resonators between prepreg layers or fiber bundles.
filling factor; ⌫ is a loss parameter that depends on the con- Therefore, it is important from the manufacturing point of
ductivity of the rings; and ␻0 is the resonance frequency view, to create unit cells that can be embedded in a fiber
which depends on the capacitance of the element. These pa- reinforced composite, compatible with its layered structure
and amenable to common fabrication methods. One such de-

FIG. 3. 共left兲 Schematics of a folded-double resonator.

共right兲 Samples made from adhesive copper tape.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
177.20.130.9 On: Thu, 14 Aug 2014 16:44:49
 014901-3 A. V. Amirkhizi and S. Nemat-Nasser J. Appl. Phys. 102, 014901 共2007兲

Here, F stands for any of the electromagnetic fields. The

boundary condition for the x faces is periodic with a pre-
scribed phase advance that is related to the wavelength
through Eq. 共5兲. A stop band between 6.09 and 6.51 GHz
共negative permeability兲 is predicted by this calculation. The
value of permeability was not directly calculated. However,
the simulation shows that there are no propagating modes in
FIG. 4. Cross section of a FDR made from adhesive copper tape with this frequency range. Following the discussion in Pendry
dimensions. et al.,1 we consider this as an indication that the overall index
of refraction is imaginary. Since in this case the permittivity
sign is the folded-doubled resonator 共FDR兲. It consists of is positive, it follows that the effective permeability is nega-
two folded conductive thin strips with one tucked inside the tive.
other, as shown in Fig. 3 共left兲. The geometry of this con- The fabricated samples were tested in an anechoic cham-
struction is suitable to being embedded into a layered fiber ber. The transmission through one, two, and three layers of
reinforced composite. The electric capacitance is then pro- this composite is measured in the 3–12 GHz range. The pre-
vided by the two folded conductive strips. dicted stop band is verified. The depth of the stop band in-
To examine this idea we fabricated samples using thin crease with the number of layers. The width of this band is
adhesive copper tapes. A 6.4 mm wide tape is folded and considerably larger than the numerical predictions. This is
placed inside another tape of 10 mm width. The outer tape is partially due to imperfect hand fabrication technique. The
then folded over. The adhesive backing of the copper pro- FDRs in this sample are not uniform and the deviation from
vides the necessary insulation. It also modifies the capaci- the perfect geometry 共assumed in simulation兲 has affected
tance of the resulting element. The copper thickness is about the resonance frequency of the unit cells, as is seen from Fig.
0.1 mm and the total thickness of the FDR is about 1 mm; 5. Also shown in Fig. 5 is the transmitted power when the
see Fig. 4. The folded strips are cut in 4.5 mm sections, incoming wave is polarized in such a way not to interact with
giving planar cross sections of 4.5 mm by 4.5 mm and 4.5 the FDRs. The FDR resonates when the incoming magnetic
mm by 3 mm outer and inner resonators, respectively. They field is polarized along the y axis, but it is inactive when the
are then glued on a thin Styrofoam panel and arranged into magnetic field is polarized parallel to the z axis. The pres-
an array of 6 mm by 6 mm by 3 mm unit cells. ence of the FDR affects the magnetic field in the entire unit
We have numerically simulated this structure using cell, but its effect is greater within the volume enclosed by
Ansoft-HFSS finite-element electromagnetic software.7 The the plates. In certain cases, the direction of the magnetic field
simulations are performed in the frequency domain, using a within this volume is opposite to that outside of this region,
unit cell with periodic boundary conditions on its y and z due to the strong induced currents in the FDR.
faces such that
F共x = x+, y, z兲 = F共x = x−, y, z兲ei␸ , 共2兲 IV. X-BAND FDR COMPOSITES
The samples described in the previous section have two
F共x, y = y +, z兲 = F共x, y = y −, z兲, 共3兲 shortcomings. First, they are not structural composites. They
have very low load-bearing strength or stiffness. Second, the
F共x, y, z = z+兲 = F共x, y, z = z−兲, 共4兲 samples have significant imperfections because of the fabri-
cation process. To address these issues, we employed a con-
x+ − x− ventional commercial printed circuit-board fabrication tech-
␸ = 2␲ . 共5兲
␭ nology; see Fig. 6. The dimensions in the thickness

FIG. 5. The transmitted power through one, two, and

three layers of the hand-fabricated FDR. Also shown, is
the transmission when incident wave is polarized in
such a way not to interact with the FDRs.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
177.20.130.9 On: Thu, 14 Aug 2014 16:44:49
 014901-4 A. V. Amirkhizi and S. Nemat-Nasser J. Appl. Phys. 102, 014901 共2007兲

FIG. 6. X-band FDR. 共top left兲 Schematic and dimen-

sions. 共bottom left兲 Cross section of the finished board.
共right兲 Planar view of the board.

directions were selected based on the available material for tromagnetic 共EM兲 properties, it was sufficient to hold the
printed circuit board 共PCB兲 fabrication. The choice of other layers together mechanically. In this manner, we were able to
geometrical parameters was ad hoc; however, the dimensions change the arrangement and the number of layers easily to
were adjusted after a few preliminary numerical simulations. examine the effect on the properties. The resulting structure
The plates of the inner resonator have a rectangular form, was tested using a focused beam setup for electromagnetic
while the outer plates are square. characterization of composites, developed at UCSD’s
In this method, the inside resonator is made out of a CEAM; see Fig. 7. The transmitted wave through the sample
single board of FR-4 material that is plated on both sides by is measured using an Agilent 8510C vector network analyzer.
copper. The copper is etched away into 2.6 mm by 1.9 mm The results are shown in Figs. 8 and 9. The curves shown
rectangles to produce the inside resonators. Then, two vias, represent the magnitude and phase of the transmitted wave
160 ␮m nominal diameter, are drilled on one edge of the through a five-layer slab of FDRs. The difference between
copper rectangles. The inside of the vias are plated with cop- the experimental curves is in the polarization of incident
per to provide electrical connection between the top and the wave. For the solid curve the incident magnetic fields are
bottom parts. The outside resonator is made out of two FR-4 polarized such that it creates a magnetic resonance in the
boards that have copper on one face only. The outer resona- FDRs; see Fig. 1. The dashed curve shows the transmission
tor’s platelets are 2.6 mm by 2.6 mm copper squares and are when the sample is rotated by 90 deg, that is when the inci-
created similarly to the inner resonators. The finished inner dent wave is in inactive polarization, i.e., when the incident
board is sandwiched between the outer boards and treated magnetic field does not excite resonance in the FDRs.
under elevated pressure and temperature to make a single
monolithic board. Finally, two vias are drilled and plated
with copper to provide electrical connection between the out- V. NUMERICAL SIMULATION AND PROPERTIES
side copper squares. The average measured thickness of the EVALUATION
final board is 775 ␮m. A more precise arrangement and
spacing of the FDRs is ensured by this fabrication method. The results of the transmission measurement were also
To test this structure, a layered composite was created by modeled numerically. We simulated two structures; see Figs.
stacking the printed FDR boards with 0.8 mm thick blanks in 8 and 9. First, we assumed that all the material in the com-
between to produce the unit cell with the designed dimen- posite, except for the copper resonators, has a uniform di-
sions. The resulting composite may be processed into a electric constant of 3.6. Second, since the material of the
monolithic sample by elevated temperature and pressure spacing boards has different amount of resin and a process-
treatment, but, for the purposes of the evaluation of its elec- ing history, we estimated and used a dielectric constant of 4.4

FIG. 7. UCSD/CEAM’s focused beam setup.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
177.20.130.9 On: Thu, 14 Aug 2014 16:44:49
 014901-5 A. V. Amirkhizi and S. Nemat-Nasser J. Appl. Phys. 102, 014901 共2007兲

FIG. 8. Comparison between measured transmission

共S21兲 and simulated values of a five-layer printed board
X-band FDR. The triangles are the simulation results
when it is assumed that the entire composite 共except for
the conductors兲 has uniform dielectric constant of 3.6.
The squares are the simulation results when layers of
the printed board with dielectric constant 3.6 are sand-
wiched between FR-4 layers of dielectric constant 4.4.
The solid and dashed lines are the measurements in the
active and an inactive 共obtained by 90 deg rotation兲
polarization, respectively.

for the boards that were sandwiched between the FDR layers. measurement setup. Therefore, its measurement may be un-
We simulated the structure shown in Fig. 10 and calculated reliable for most cases. The transmission measurement alone
the scattering parameters, Sij, is not enough to yield the two complex-valued material pa-

冋册冋
V−1
V−2
=
S11 S12
S21 S22
册冋 册
V+1
V+2
. 共6兲
rameters, n and z. This requires four real quantities, i.e., the
magnitude and phase of both transmission and reflection co-
efficients. Therefore, we used both transmission and reflec-
Here V+i and V−i denote the voltage of the incident and scat- tion data from simulation for this purpose. For a slab of
tered signals at the ith port, respectively. The results of the thickness d, refractive index n, and impedance z, we have
simulations are shown in Figs. 8 and 9. 1

冋 册
To extract the overall material properties of the compos- t= , 共7兲
i 1
ite, we substitute the results of the simulation for S11 and S21 cos共nkd兲 + z+ sin共nkd兲
in the expected reflected and transmitted waves off of a slab 2 z

冋 册
of homogeneous material; see Fig. 11. Note that we chose to
use the results of the numerical simulation with uniform di- i 1
z− sin共nkd兲
electric constant 3.6, since it agrees with the experimental 2 z

冋 册
r= , 共8兲
data very closely. Use of the simulated values has two ad- i 1
vantages. First, we can avoid the noise of the measured data cos共nkd兲 + z+ sin共nkd兲
2 z
in a systematic way. The simulation results follow the experi-
mental data closely but do not include various noise sources. except for a phase factor e±ikd. Here k = ␻ / c is the wave
Second, the reflection measurement, S11, involves many in- number for free space, t is the transmission coefficient, and r
ternal reflections in the experimental setup. The value of the is the reflection coefficient; see Fig. 11. The inversion of
coefficient of reflection off of the sample, though not neces- these equations gives the overall material properties of the
sarily small, is dominated by the internal reflections in the FDR composite. Here we followed the method of Smith

FIG. 9. Transmission phase for a five-layer printed

board X-band FDR. Comparison between simulation
and measurement results.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
177.20.130.9 On: Thu, 14 Aug 2014 16:44:49
 014901-6 A. V. Amirkhizi and S. Nemat-Nasser J. Appl. Phys. 102, 014901 共2007兲

FIG. 11. Transmission and reflection of a monochromatic wave with wave

number k off of a homogenous plate.

tinguished from previous efforts in being the first to achieve

structural integrity for a material with overall negative ␮. We
FIG. 10. Setup for time-domain simulation of a finite-thickness plate of the
FDR composite. Each cube represents a unit cell. based our unit cell on the original idea of split-ring resonator
by Pendry et al.1 However, by introducing the folded-
et al.8 We used a MATLAB macro written originally by C. G. doubled resonator design, we were able to easily integrate
Parazzoli.9 We adapted the macro for the current application this magnetic resonator into structural composites using con-
by applying it to the output of the Ansoft-HFSS solver and ventional composite processing techniques. The FDR unit
de-embedding the results appropriately. cell consists of two nested resonators. The flat shape of
The material parameters calculated based on this inver- FDRs makes them a perfect candidate for embedding inside
sion are given in Fig. 12. The results show a negative per- structural composites. The composite is fabricated from com-
meability band between 9.1 and 9.35 GHz, as well as an mercial lithography printed boards. The measured transmis-
increase in the overall dielectric constant of about 7. The sion through a five-layer slab of FDR indicates a stop band
minimum normalized magnetic permeability is calculated to between 9.1 and 9.35 GHz. The numerical predictions are in
be around −0.65. We did not perform a sensitivity analysis good agreement with experiments. We used the results of
for the material parameters n and z based on the alternate numerical simulation and calculated overall material proper-
values of the dielectric constant in the unit cell. However, the ties by inverting the transmission and reflection coefficients,
numerical results, of the two cases we studied, agree with using a homogeneous slab of comparable thickness. The cal-
one another and with the experimental data. culated magnetic permeability shows a negative band be-
tween 9.1 and 9.35 GHz. This was the first fabrication and
measurement of a structural composite with negative overall
VI. SUMMARY
magnetic permeability. It led to a similar fabrication method
We have designed a structural composite that exhibits that was subsequently used by Starr et al.3 to produce a
negative overall magnetic permeability. Our work was dis- structural composite with negative index of refraction.

FIG. 12. Extraction of material prop-

erties of X-band printed board FDR
from S-parameter simulation. Inver-
sion macro, courtesy of Parazzoli
共Ref. 9兲.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
177.20.130.9 On: Thu, 14 Aug 2014 16:44:49
 014901-7 A. V. Amirkhizi and S. Nemat-Nasser J. Appl. Phys. 102, 014901 共2007兲

ACKNOWLEDGMENTS ego, and recently completed through partial support from

The FDR idea was conceived by one of the authors AFOSR/MURI Grant No. FA9550-06-1-0337 to Kent State
共S.N.-N., October 2002兲, its simulation and design were University; and Subaward No. 444286-PO61719 to Univer-
completed by the first author 共A.V.A.兲, and its EM measure- sity of California, San Diego.
ments were performed by A. F. Starr 共SENSORMETRIX, 1
J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, IEEE Trans.
CA兲. The original version of the FDR was hand-constructed
Microwave Theory Tech. 47, 2075 共1999兲.
by D. Arbelaez 共then an undergraduate, MAE student兲 with 2
D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Schultz,
the help of T. Plaisted 共graduate student兲 and CEAM techni- Phys. Rev. Lett. 84, 4184 共2000兲.
3
cal staff, in early 2003. The X-band FDR composites were A. F. Starr, P. M. Rye, D. R. Smith, and S. Nemat-Nasser, Phys. Rev. B 70,
produced by Hughes Circuits, San Marcos, CA. It led to a 113102 共2004兲.
4
new design by Starr et al.,3 who coined the expression “The S. Nemat-Nasser, S. C. Nemat-Nasser, T. A. Plaisted, A. Starr, and A.
Vakil Amirkhizi, in Biomimetics: Biologically Inspired Technologies, ed-
New Deal” for the new composite that had negative index of ited by Y. Bar Cohen 共CRC Press, Boca Raton, FL, 2006兲.
refraction in about 8.4–9.2 GHz frequency range. The pro- 5
J. B. Pendry, J. Mod. Opt. 41, 209 共1994兲.
6
gram used for the inversion of the S-parameter data and ex- J. B. Pendry, J. Phys.: Condens. Matter 8, 1085 共1996兲.
7
traction of material properties was kindly provided to the Ansoft, Ansoft HFSS 8.0 User Documentation, Ansoft Corporation, Pitts-
authors by C. G. Parazzoli and D. R. Smith and was adjusted burgh, PA, 2001.
8
D. R. Smith, S. Schultz, P. Markos, and C. M. Soukoulis, Phys. Rev. B 65,
for application to the present problem. This work was ini- 195104 共2002兲.
tially supported by DARPA through Grant. No. ARO 9
C. G. Parazzoli, Boeing Phantom Works, Seattle, WA.
10
DAAD19-00-1-0525 to the University of California, San Di- D. R. Smith, Duke University, Durham, NC.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
177.20.130.9 On: Thu, 14 Aug 2014 16:44:49