Experimental determination of the principal dielectric functions in silver nanowire

metamaterials
Jyotirmayee Kanungo and Joerg Schilling
Citation: Applied Physics Letters 97, 021903 (2010); doi: 10.1063/1.3462311
View online: http://dx.doi.org/10.1063/1.3462311
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/97/2?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Determination of plasma frequency, damping constant, and size distribution from the complex dielectric
function of noble metal nanoparticles
J. Appl. Phys. 116, 233105 (2014); 10.1063/1.4904349
Magnetic resonance on core-shell nanowires with notches
Appl. Phys. Lett. 99, 101907 (2011); 10.1063/1.3637042
A metal-wire/quantum-dot composite metamaterial with negative and compensated optical loss
Appl. Phys. Lett. 93, 193106 (2008); 10.1063/1.3013331
Negative index metamaterials based on metal-dielectric nanocomposites for imaging applications
Appl. Phys. Lett. 93, 123117 (2008); 10.1063/1.2945647
Density functional theoretical determinations of electronic and optical properties of nanowires and bulks for
CdS and CdSe
Appl. Phys. Lett. 90, 031904 (2007); 10.1063/1.2432170

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:
202.43.95.117 On: Fri, 27 Mar 2015 07:01:00

APPLIED PHYSICS LETTERS 97, 021903 2010

Experimental determination of the principal dielectric functions in silver

nanowire metamaterials
Jyotirmayee Kanungo1 and Joerg Schilling1,2,a
1

Queens University Belfast, University Road, Belfast BT7 1NN, United Kingdom
ZIK SiLi-nano, Karl-Freiherr-von-Fritsch Str. 3, 06120 Halle, Germany

Received 13 January 2010; accepted 5 June 2010; published online 12 July 2010
The polarization dependent optical properties of silver nanowire arrays are investigated by angular
resolved transmission measurements. The corresponding spectra show clear FabryPerot
oscillations, which exhibit an unusual shift toward longer wavelengths for the extraordinary waves.
From the peak shift both principal dielectric functions of the metamaterial are determined and
compared with effective medium theories. Their different signs lead to a hyperbolic wave vector
surface for the extraordinary waves and cause the observed peculiar peak shift. 2010 American
Institute of Physics. doi:10.1063/1.3462311
Employing wire meshes and split ring resonators the effective permittivity and permeability of metamaterials
can be shifted from positive to negative values1 and exciting
phenomena such as hyperlensing2 and cloaking3 could be
observed. However, the fabrication of similar extended
metamaterials for the visible and near infrared spectral region becomes much more difficult and time consuming
as it involves several steps of nanofabrication techniques
like electron beam lithography. Because of these complications and limitations metamaterials of much simpler
design are favored in this spectral range. In particular metal
nanowire arrays, which are embedded in a dielectric host
matrix, gained considerable attention recently. These metaldielectric composites are considered as uniaxial anisotropic
metamaterials,46 whose two principal dielectric constants
have opposite sign indefinite tensor. This property should
lead to a hyperbolic equifrequency surface in wave vector
space for the extraordinary, transverse magnetic TMpolarized rays allowing the propagation of waves with unusual large wave vectors. The predicted negative refraction
of these rays was already demonstrated experimentally.7,8
Here we present a direct experimental determination of
the two principal effective dielectric functions and of a
silver nanowire array in a wide spectral range in the visible
and near infrared 6001600 nm. Our method is based on an
analysis of the spectral shift in FabryPerot-peaks in angular
resolved transmission measurements for both principal polarizations TE and TM.
The fabrication of the parallel aligned silver nanowire
arrays is based on nanoporous alumina templates. These templates were created from high purity 99.999% aluminum
foils by a two-step electrochemical-anodization process in
0.3 M H2SO4 at a constant voltage of 25 V.9 A CuCl2 solution was used to remove the Al layer below the oxide layer
and the pore bottoms were finally opened using a 5 wt %
H3PO4 solution at 30 C for 15 min resulting in a through
hole alumina membrane with pore diameter 32 nm and pore
distance center to center of 60 nm. After the deposition of a
15 nm thin gold contact layer on one side of a membrane the
silver nanowires were grown for 1 h within the nanopores
via electrochemical deposition using an electrolyte mixture
a

Electronic mail: [email protected].

of AgBr 0.1 M, Na2S2O3 0.25 M, and Na2SO3 0.2 M at

a constant voltage of 700 mV versus a saturated
Hg/ HgSO4-electrode.
To prepare a 40 30 m2 planar film, which contains
silver nanowires of equal length penetrating it from top to
bottom, we applied focused ion beam FIB milling FEI 200
workstation at 30 kV, 5 nA beam current. Using FIB the
original top and bottom parts of the membrane were removed, clearing also the previously deposited gold contact.
The resulting silver nanowire film had a thickness of 1.7 m
and a metal filling factor of f = 0.26 Fig. 1.
Angle resolved transmission measurements in the range
of 0 to 50 on the cut out films were performed using a
tungsten halogen lamp as a collimated white light source.
After passing through a polarizer the linearly polarized light
was incident at the flat surface of the membrane. The angle
of incidence was varied by rotating the sample Fig. 2a.
The s-polarized light whose electric field vector was perpendicular to the nanowire axis excited the ordinary, TEpolarized waves, while the p-polarized light coupled to the
extraordinary, TM-polarized waves. The transmitted light
was collected by a microscope objective and was either fed
into a Triax 322 spectrometer with attached liquid nitrogen
cooled Si-charge coupled device CCD for detection in the
spectral range 400900 nm or into a Triax 320 spectrometer
with an attached InGaAs-CCD for detection in the IR 900
1600 nm.
An example of a measured transmission spectrum is
shown in Fig. 2b. The transmitted intensity rises from 500
nm toward the infrared. On top of this general trend the
FabryPerot oscillations are visible Fig. 2b. To determine
the exact spectral position of the FabryPerot peaks for our
analysis, a rising baseline is subtracted. The combined, base-

FIG. 1. a SEM image of the slab after FIB milling. b SEM image of the
wire distribution in the slab. The filling factor is f = 0.26.

0003-6951/2010/972/021903/3/$30.00
97,is021903-1
2010 American InstituteDownloaded
of Physics to IP:
This article
is copyrighted as indicated in the article. Reuse of AIP content
subject to the terms at: http://scitation.aip.org/termsconditions.
202.43.95.117 On: Fri, 27 Mar 2015 07:01:00

021903-2

Appl. Phys. Lett. 97, 021903 2010

J. Kanungo and J. Schilling

Since is isotropic and close to unity in the visible and

near infrared, the dispersion relations for ordinary TE polarized and extraordinary TM polarized waves in a uniaxial
crystal are simply
kz =

kz =

FIG. 2. Color online a Measurement geometry for angular resolved

transmission measurement employing a wide collimated incident beam and
local detection by a microscope objective. b Example experimental transmission curve at normal incidence. The vertical dotted black line represents
the stitching range where the two separate measurements IR and visible
were combined. The rising red dashed line represents the baseline.

line subtracted transmission curves represent the oscillations

more clearly and are shown in Fig. 3 for both TM and TE
polarized light.
The spectra exhibit a striking qualitative difference. For
the TE polarized light the FabryPerot peaks shift clearly to
shorter wavelengths with increasing angle of incidence indicated by the black dashed curve in Fig. 3, which is the
classic expected behavior. On the contrary for the TM polarized light the FabryPerot peaks shift toward larger wavelengths with rising angle of incidence. To explain these differences and determine and , the condition for the
occurrence of the FabryPerot peaks and the dispersion of
light inside the nanowire films has to be considered.
The FabryPerot oscillations are caused by the constructive and destructive interference of the partially transmitted
waves, which occur at each inner reflection at one of the
films surfaces.
The condition for the interference maxima is

2
k2x
c2

2 2

k
c2 x

TM polarized,

where and describe the two principal dielectric functions perpendicular and parallel to the optical axis which
coincides with the wire axis z-direction.
Combining Eqs. 1 and 2, using kx = / csin and
/ c = 2 / and squaring the resulting equations leads to
the following relations for the wavelength of the mth transmittance maximum for TM and TE polarized light:
2
=
m

4d2 4d2 2

sin
m2
m2
2
=
m

TM polarized

4d2 4d2 2
2 sin
m2
m

TE polarized.

Since changing the angle of incidence from 0 to 50 results only in a small shift in the resonance wavelengths m,
we can treat and as piecewise constant functions over
each of these small spectral ranges. Under this condition the
Eqs. 3 can be represented by a linear form y m = Am Bmx for
2
, x = sin2 , and Am
each resonance peak, where y m = m
2
2
= 4d / m .
The constant Bm determines the slope of the linear function and differs for both polarizations as follows:
TM
=
Bm

4d2
m2

and

TE
Bm
=

4d2
.
m2

Dividing the constants Am and Bm for each resonance peak m

separately, finally leads to the following values of and :
TM
= ,m
Am/Bm

2m = 2kzd,

TE polarized

TM polarized light

TE
Am/Bm
=

where d is the film thickness, m is the order of the transmittance maximum, and kz describes the wave vector component perpendicular to the surface.

,m

TE polarized light.

Performing these calculations for several resonance peaks in

a spectrum allows to trace the values of and across the
whole spectral range.
To determine the spectral peak position of the Fabry
Perot peaks from the transmission measurements accurately,
the baseline corrected experimental transmission spectra
Fig. 3 were fitted by multiple Lorentzian line shape functions, where the center wavelength of each Lorentzian was
taken as peak position. Subsequently, the determined peak
wavelength was squared and plotted against the squared sinus of the angle of incidence for both polarizations. Figure 4
shows an example of such a plot representing the spectral
shift in the single resonance which is indicated by the black
dashed curve in Fig. 3.
The linear fit of the data reveals, that the relationship
between the squared peak wavelength and the squared sinus
of the angle of incidence can indeed be described by a linear
function as predicted in Eq. 3. The peak specific constants
FIG. 3. Color online Angular resolved transmission measurements baseTE
TM
, Bm
are determined from the linear fits in Fig. 4 and
Am , Bm
line subtracted for a TE-polarization ordinary ray and b TMfinally and are calculated from Eq. 5. Since the propolarization extraordinary ray. The black dashed lines show the different
cedure
repeated
for all nine FabryPerot peaks ofDownloaded
the com- to IP:
shiftsisincopyrighted
the resonances
increasing
of incidence.
This article
as with
indicated
in theangle
article.
Reuse of AIP content is subject
to theisterms
at: http://scitation.aip.org/termsconditions.
202.43.95.117 On: Fri, 27 Mar 2015 07:01:00

021903-3

Appl. Phys. Lett. 97, 021903 2010

J. Kanungo and J. Schilling

FIG. 4. Color online Spectral dependence of a single peak from angle of

incidence, a TE and b TM polarized light. For TE-polarized spectra the
peak shifts to smaller wavelengths with rising angle of incidence while for
TM-polarized spectra the opposite happens. The solid line is a linear fit of
the presented experimental data.

bined spectra, the experimentally determined dielectric functions cover the whole spectral region from 600 to 1600 nm
Fig. 5.
In order to compare our experimental results with theory,
we calculate the effective permittivity values in the same
spectral range using the following WienerGarnett-formulas
and MaxwellGarnett-formulas:10
= fm + 1 fd

and

= d

1 + fm + 1 fd
,
1 fm + 1 + fd
6

where f is the fraction of metal filling, m is the dielectric

constant of the metal silver wire,10 and d = 2.56 is the dielectric constant of the dielectric alumina matrix.
The experimentally determined values for and
agree very well with the theoretical values from the effective
medium theories. These results confirm, that a silver nanowire array represents an indefinite metamaterial, whose two
principle dielectric functions have opposite sign within the
investigated spectral region. Furthermore they are confirming

that the MaxwellGarnett and Wiener formulas yield very

good approximations for and in this spectral range.
The different sign for and is also directly responsible for the observed shift in the peak wavelength toward
larger values for the TM-polarization. Since the ratio /
is negative, an increase in leads directly to a larger peak
wavelength in Eq. 3. On a more general level this behavior
follows from the hyperbolic dispersion between kz and kx for
TM-polarization.
Note, that the initial condition for the resonance in Eq.
1 is based on the real part of kz as this determines the phase
relationship of the partially transmitted waves. To carry out
the evaluation of the resonance shifts and avoid ambiguity,
we had to limit ourselves to real dielectric functions and
. The experimentally derived values in Fig. 5 can therefore
only represent the real parts of the principal dielectric functions. However using the effective medium theories at least
theoretical values for the relatively small imaginary parts of
and can be obtained, which are shown in the supplementary material for information.11
Since the method is based on the spectral shift in extrema in the transmission spectra and not on the absolute
values of transmission or reflection, the impact of scattering
losses on the results e.g., due to surface roughness or inhomogeneity of the sample is strongly reduced. Furthermore,
and could be determined without the knowledge of the
exact sample thickness or the absolute order m of the considered resonances. This simplifies the procedure and further
reduces the impact of potentially erroneous values e.g., inaccurate thickness measurements.
In conclusion we directly derived both principle dielectric functions of the silver nanowire arrays from the angle
dependent spectral shifts in the FabryPerot transmission
peaks. A comparison with MaxwellGarnet effective medium theory showed very good correspondence. The composite is highly anisotropic. The positive values for and
the negative values for lead to the hyperbolic dispersion of
TM polarized light which causes the unusual peak shift toward longer wavelengths with increasing angle of incidence.
J.S. acknowledges the support by the Royal Society and
the BMBF Grant No. FKZ:03Z2HN12 for this work.
R. Shelby, D. R. Smith, and S. Schultz, Science 292, 77 2001.
Z. Liu, H. Lee, Y. Xiong, C. Sun, and X. Zhang, Science 315, 1686
2007.
3
W. Cai, U. K. Chettiar, A. V. Kildishev, and V. M. Shalaev, Nat. Photonics
1, 224 2007.
4
J. Elser, R. Wangberg, V. A. Podolskiy, and E. E. Narimanov, Appl. Phys.
Lett. 89, 261102 2006.
5
W. T. Lu and S. Sridhar, Phys. Rev. B 77, 233101 2008.
6
A. Fang, T. Koschny, and C. M. Soukoulis, Phys. Rev. B 79, 245127
2009.
7
J. Yao, Z. Liu, Y. Liu, Y. Wang, C. Sun, G. Bartal, A. M. Stacy, and X.
Zhang, Science 321, 930 2008.
8
M. A. Noginov, Y. A. Barnakov, G. Zhu, T. Tumkur, H. Li, and E. E.
Narimanov, Appl. Phys. Lett. 94, 151105 2009.
9
O. Jessensky, F. Mller, and U. Gsele, J. Electrochem. Soc. 145, 3735
1998.
10
E. D. Palik, Handbook of Optical Constants of Solids Academic, San
Diego, 1998.
11
See supplementary material at http://dx.doi.org/10.1063/1.3462311 for
theoretical values of the imaginary parts of both principal dielectric functions.
1
2

FIG. 5. Color online Real parts of the principal dielectric constants in the
visible and IR. The symbols represent the experimentally derived values.
The lines show the theoretical values, which were calculated using effective
medium theory for a metal filling factor of f = 0.26.

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:
202.43.95.117 On: Fri, 27 Mar 2015 07:01:00