See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/51395278

Perfect Metamaterial Absorber

Article  in  Physical Review Letters · May 2008

DOI: 10.1103/PhysRevLett.100.207402 · Source: PubMed

CITATIONS READS

1,993 1,033

5 authors, including:

Willie Padilla
Duke University
206 PUBLICATIONS   21,564 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

THz metamaterials View project

All content following this page was uploaded by Willie Padilla on 27 May 2014.

The user has requested enhancement of the downloaded file.

 A Perfect Metamaterial Absorber

N. I. Landy
Boston College, Department of Physics, 140 Commonwealth Ave., Chestnut Hill, MA 02467.

S. Sajuyigbe, J. J. Mock, D. R. Smith

Department of Electrical and Computer Engineering, Duke University, Durham, NC 27708 USA.

W. J. Padilla
Boston College, Department of Physics, 140 Commonwealth Ave., Chestnut Hill, MA 02467.
(Dated: March 13, 2008)
arXiv:0803.1670v1 [cond-mat.mes-hall] 11 Mar 2008

We present the design for an absorbing metamaterial element with near unity absorbance. Our
structure consists of two metamaterial resonators that couple separately to electric and magnetic
fields so as to absorb all incident radiation within a single unit cell layer. We fabricate, charac-
terize, and analyze a metamaterial absorber with a slightly lower predicted absorbance of 96%.
This achieves a simulated full width at half maximum (FWHM) absorbance of 4% thus making
this material ideal for imaging purposes. Unlike conventional absorbers, our metamaterial consists
solely of metallic elements. The underlying substrate can therefore be chosen independently of the
substrate’s absorptive qualities and optimized for other parameters of interest. We detail the de-
sign and simulation process that led to our metamaterial, and our experiments demonstrate a peak
absorbance greater than 88% at 11.5 GHz.

The nascent field of electromagnetic metamaterials has

produced exotic effects such as negative index of re-
fraction [1, 2], and devices such as an electromagnetic
cloak [3]. The realization of such properties lies in the
ability of metamaterials to create independent tailored
electric [4] and magnetic [5] responses to incident radia-
tion. Electromagnetic metamaterials are also geometri-
cally scalable which translates into operability over a sig-
nificant portion of the electromagnetic spectrum. To date
metamaterials have been demonstrated in every tech-
nologically relevant spectral range, from radio [6], mi-
crowave [7], mm-Wave [8], THz [9], MIR [10], NIR [11],
to the near optical [12]. These designer electromagnetic
materials are an ideal platform for the investigation of
novel emergent physical phenomena while also holding
great promise for future applications.
FIG. 1: Electric resonator (a) and cut wire (b). Dimension
As an effective medium [13], metamaterials can be notations are listed in (a) and (b). The unit cell is shown
characterized by a complex electric permittivity ˜(ω) = in (c) with axes indicating the direction of propagation of a
1 + i2 and magnetic permeability µ̃(ω) = µ1 + iµ2 . TEM wave.
Much of the work in metamaterials has been focussed
on the real part of  and µ to enable the creation of a
negative refractive material. However, the oft-overlooked
loss components of the optical constants (2 and µ2 ) have tinct metallic elements as shown in Fig. 1 (a) and (b).
much potential for the creation of exotic and useful ma- Electric coupling was supplied by the electric ring res-
terials as well. For example, they can be manipulated onator (ERR), similar to the design presented by Padilla
to create a high absorber. By manipulating electric and et al. [14], and is shown in Fig. 1 (a). This element con-
magnetic resonances independently, it is possible to ab- sisted of two standard split ring resonators connected by
sorb both the incident electric and magnetic field. Ad- the inductive ring parallel to the split-wire. We used this
ditionally, by matching  and µ, a metamaterial can be design instead of a conventional split-wire design because
impedance-matched to free space, minimizing reflectiv- of the limitations of straight wire media [15], i.e. a split-
ity. In this paper, we show that metamaterials can be wire design has limited tunability beyond the addition of
fashioned to create narrow-band perfect absorbers, with more wires per unit cell to increase inductance [16]. The
the potential to be used in devices such as in bolometers. magnetic coupling required a more complicated arrange-
A single unit cell of the absorber consisted of two dis- ment, and thus in order to couple to the incident H-field,
 2

an impedance near the free space value. While 100%

asborbance is theoretically possible, it can only occur
when the metamaterial layer is impedance-matched to
free space such that the reflectance is zero.[19] One may
then minimize transmission such that the addition of
multiple layers ensures T (ω) → 0. The simulated meta-
material had the dimensions, in millimeters, of: a1 =4.2,
a2 =12, W=3.9, G=0.606, t=0.6, L=1.7, H=11.8, and the
metamaterials elements were separated by 0.65 in the ẑ
direction. In Fig. 2 we show the simulation results for
the metamaterial perfect absorber. Reflectance and ab-
sorbance are plotted from zero to 100% (left axis) and
the transmission is plotted on the right axis from zero
to 5%. The reflectance is large ∼ 97% near the bounds
of the plot, 9GHz and 14GHz, but there is a minimum
of 0.01% at ω0 ≡ 11.65GHz. The simulated transmis-
FIG. 2: Simulated perfect metamaterial absorber. Re- sion also undergoes a minimum near ω0 and yields a
flectance (green curve) and absorbance (red curve) are plotted value of ∼ 0.9%. Thus we achieve a best simulated ab-
from zero to 100% (left axis). The transmission is plotted on sorbance A(ω) = 1 − T (ω) − R(ω) slightly less than unity
the right axis as the blue curve on a scale from zero to 5%. A(ω) = 99% with a FWHM of 4% compared to ω0 .
While our simulated absorber achieves narrow-band
high absorbance, we were restricted by the minimum line
widths available to us (250 µm) and other various fab-
we needed flux created by circulating charges perpendic- rication tolerances. These limitations were incorporated
ular to the propagation vector. We created this response into the design process and we fabricated a metamate-
by combining the center wire of the electric resonator rial which deviated slightly from the ideal absorber and
with a cut wire (Fig. 1(b)) in a parallel plane separated had the dimensions, in millimeters, of: a1 =4.2, a2 =12,
by a substrate (Fig. 1(c)). This design is similar to the W=4, G=0.6, t=0.6, L=1.7, H=11.8. Each metalliza-
so-called “fishnet” and paired nanorod structures[11, 12], tion was fabricated on a FR4 substrate with a thickness
in the sense that they derive a magnetic response from of 0.2mm. Metamaterials were fabricated using standard
driving two antiparallel currents in conducting segments. optical lithography. Photosensitized half-ounce copper-
We were then able to tune the magnetic response by clad FR-4 circuit board formed the substrate material
changing the geometry of the cut wire and the separa- with an 17 µm copper thickness. A mask was designed
tion between the cut wire and electric resonator. By ma- and printed in high resolution on a transparency. The
nipulating the magnetic coupling without changing the photosensitized copper clad FR-4 was covered by the
geometry of the electric resonator, we able to decouple  mask and exposed to ultraviolet light then developed in
and µ, and therefore individually tune each resonance. a sodium carbonate solution and etched in ferric chlo-
We performed computer simulations of an ideal – but ride. Finally, the board was placed in a stripper so-
realizable – metamaterial absorber using the commercial lution to remove the remaining photoresist. Both the
finite-difference time domain (FDTD) solver Microwave ERR layer (Fig. 1(a)) and the cut-wire layer (Fig. 1(b))
Studio by CST.[17] The program simulated a single unit were fabricated in such a fashion, see inset to Fig. 3 (b).
cell as shown in Fig. 1 (c) with appropriate bound- These boards where then sandwiched (using an adhesive
ary conditions, i.e. perfect electric (ŷẑ-plane) and per- with 0.06mm thickness) about another 0.2mm thick FR4
fect magnetic (x̂ẑ-plane). Waveguide ports on the other blank substrate to obtain the correct spacing. The end
boundaries simulated a TEM plane wave propagating results was that the metamaterials elements were sepa-
through the medium. Simulation produced the complex rated by 0.72mm in the ẑ direction. This design allowed
frequency dependent S-parameters, Se11 and Se21 , where us to couple to both magnetic and electric resonances
T (ω) = |S21 |2 and R(ω) = |S11 |2 are the transmission without using out-of-plane elements typical with “wine-
and reflectance, respectively. From the S-parameter data crate” designs,[2] thus greatly simplifying construction of
we inverted the fresnel equations to extract the complex the absorber.
optical constants[18]. We then examined the behavior of We experimentally verified the behavior of the ab-
the surface current density, magnetic and electric fields sorber by measuring the complex S-parameters of a large
at ω0 to verify that we were coupling to the correct res- planar array of pixels (outer dimensions of 15 × 15 cm).
onant mode of each metamaterial element. By tuning We used an Agilent vector network analyzer that pro-
ω0 of the  and µ resonances (not shown), due to the duced microwaves in the range of 8-12 GHz. One mi-
two metamaterial elements, we achieved  = µ and thus crowave horn (Rozendal and Associates) focused the GHz
 3

FIG. 4: Main panel shows the simulated (red) and measured

(blue) absorbance curves. Significant broadening in the reflec-
tion curve contributes to the deviation of the experimental ab-
sorbance curve from simulation. The dashed gray absorbance
curve is a Gaussian weighted average which approximates as-
sembly error of the metamaterials absorber, described in the
text. Inset shows the simulated angular dependence of the
FIG. 3: Simulated (red curves) and measured (blue curves) absorbance at at ωmax .
T(ω) and R(ω) for the microwave absorber. (a) displays the
simulated and measured reflectance from zero to 100%. A
schematic of the reflection experiment is shown as the inset
to (a) where BS is the beam splitter. The simulated and
measured transmission, shown in (b), are displayed from zero red curve in Fig. 3 (a). At low frequencies the reflectivity
to 10%. Both reach a minimum of approximately 0.6% near is high and yields values near 95%. A large feature oc-
11.5 GHz. Inset to (b) shows photographs of the individual curs in R(ω) and has a minimum of ∼ 3% near 11.5 GHz.
components which make up the metamaterial absorber. The Likewise the simulated transmissivity, shown as the red
sample consists of the electric ring resonator (left) and the curve in Fig. 3 (b), has 4% T (ω) at low frequencies and
split wire (right) and is joined with an FR4 spacer of 0.72mm then undergoes a minimum near 11.25 GHz. The simu-
thickness.
lated absorbance of our metamaterial unit cell thus has
a maximum near where both R(ω) and T (ω) have their
minima. We plot A(ω) as the red curve in Fig. 4 from 8
beam on the sample and another horn served as a detec- to 12 GHz. The simulated A(ω) peaks at 96% at 11.48
tor. Both horns coupled to linearly polarized light and GHz and has a FWHM of 4% with respect to its center
had parallel polarization directions. To measure trans- frequency.
mission, we set the horns in a normal incidence confocal By performing simulations for transmissivity and re-
configuration. We performed reflectance measurements flectivity, we are able to account for the form of the ex-
at normal incidence. However, in order to eliminate large perimental data. The measured R(ω) and T (ω) are plot-
voltage standing wave ratio (VSWR) which would over- ted as the blue curves in Fig. 3 (a) and (b) respectively.
whelm the data, we used a beam splitter configuration, We find excellent agreement between simulated and mea-
shown as the inset to Fig. 3 (a). Radiation first traveled sured T(ω) over all frequencies characterized. The mea-
from the transmitting horn through the beam splitter, sured reflectance is consistently 8% lower than the sim-
reflected from the sample and returned to the beam split- ulated R(ω) at low frequencies. Both the simulated and
ter where it was reflected at 45% to the receiving horn. experimental R(ω) reach a minimum at approximately
We set the normalization and phase in the transmission 11.5 GHz, but experimentally the minimum is 11%, as
configuration by simply removing the sample. For re- opposed to the simulated value of 3%. There is also sig-
flection, we replaced the sample with a perfect reflector nificant broadening in the experimental reflectivity curve
for normalized measurements. Transmission characteri- with respect to simulation. This is most likely due to
zation techniques utilized here have been described else- tolerances in the fabrication and assembly.
where in detail[20]. The experimental absorbance calculated from T (ω)
Computer simulations were performed for this meta- and R(ω) is shown as the blue curve in Fig. 4. The
material structure and are presented for comparison to simulated and experimental curves have maximum ab-
experimental data. The simulated R(ω) is shown as the sorbance at the same frequency, ωmax =11.5 GHz. How-
 4

that although an impedance matched structure may yield

zero R(ω0 ), this does not necessarily mean the trans-
mission will be small. In this study we are interested
in achieving R(ω0 )=T(ω0 )=0 in a single unit cell in the
propagation direction. Thus our metamaterial structure
was optimized to maximize the absorbance with the re-
striction of minimizing the thickness. If this constraint is
relaxed, however, an impedance matched condition can
be obtained and with multiple layers a perfect absorbance
can be achieved.
We further investigated the origin of the loss in our
metamaterial structure, as indicated in the inset to Fig.
5. The Ohmic “surface” loss originates mainly from the
center conducting region of the electric ring resonator,
and dielectric losses primarily occur in-between the two
metamaterial elements where the electric field is large.
FIG. 5: Simulated absorbance with increasing metamaterial The major component responsible for the absorbance in
layers, red curve is a guide to the eye. Insets show the simu- our metamaterial structure is the dielectric loss and simu-
lated losses at resonance. The Ohmic loss (surface) is shown
lation indicates this is an order of magnitude greater than
in the left inset panel and the dielectric (volume) loss is on
the right. Ohmic losses. This is consistent with previous studies of
frequency selective surfaces (FSSs) where it was found
that metallic absorption was relatively insignificant in
comparison to dielectric losses.[21]
ever, the simulated curve reaches a maximum of 96% and We now discuss the potential application of a perfect
the experimental curve only reaches 88%. The differ- metamaterial absorber as a bolometer. An ideal bolo-
ence between the simulated and measured curves is due metric detector is one in which all photons that fall upon
primarily to deviations in R(ω). Better fabrication and its surface are absorbed, converted to heat, and thus
assembly tolerances will permit absorbances and band- sensed.[22] A prominent feature of commercially avail-
widths similar to that shown as the red curve in Fig. able bolometers is their extreme bandwidth making them
4. We performed computer simulations to verify the ori- ideal candidates as detectors, especially in the far in-
gin of the broadening of A(ω). From a theoretical view- frared frequency regime where other efficient detection
point it is expected that the frequency location of the methods are largely unavailable. However, bolometers
absorbance peak depends significantly upon the separa- which achieve a narrowband response also have signifi-
tion between the electric ring resonator and the cut-wire, cant application as focal plane array detectors for imag-
since this distance determines the frequency location of ing. The metamaterial design presented here achieves
the magnetic resonance. Computer simulations verified simulated absorbtivities near unity making them ideal
this conjecture. We simulated various spacings centered candidates for bolometric pixel elements. The natu-
about the designed value of 0.72mm, and computed a ral narrowband metamaterial resonance is a salient fea-
Gaussian weighted average A(ω) with a standard devi- ture for focal plane array detectors as it is naturally
ation of σ =20 µm. The absorbance calculated in this apodizing and operates at room temperature. Since
manner agrees very well with the experimental curve and the elements which constitute our bolometer are sub-
is plotted as the grey dashed curve in the main panel of wavelength, metamaterials can inherently image at the
Fig. 4. Thus assembly errors in the metamaterial spac- diffraction limit. Further, here we have used only a sin-
ing of only 5% can easily account for the disagreement gle unit cell in the propagation direction, (with a thick-
between the simulation and experiment shown in Fig. 4 ness significantly smaller than the wavelength λ0 /35), yet
as the red and blue curves, respectively. achieved an experimental absorbance of 88%. By adding
An investigation into the effect of adding multiple multiple layers one can achieve narrowband absorptivi-
metamaterial absorbing layers is shown in the main panel ties of unity. Although the design is planar and A(ω)
of Fig. 5. As can be observed the absorbance rises should fall off relatively rapidly, we investigated the an-
sharply with additional layers and is asymptotic to unity, gular dependence (inset to Fig. 4) and found the meta-
within computational noise. Two layers of the metama- material still achieved 50% absorbance at an incident
terial absorber achieve a value of ∼99.9972%. At this full-angle of 16◦ . Finally, the scalability of metamateri-
thickness the entire metamaterial is only ∼ λ0 /2 at res- als permits the usage of these perfect absorbers at other
onance. For the structure shown in Fig. 1, A(ω0 ) can wavelengths of interest such as the possibility of room
never reach unity since the metamaterial is not perfectly temperature high resolution imaging at mm-Wave and
matched to free space. However, it is important to note THz frequencies. This is particularly interesting owing to
 5

the many possible applications. We have designed simi- direction of the incident wave to couple to the magnetic
lar metamaterials (not shown) as in Fig. 1 that achieved resonator. This resulted in variations between unit cells
comparable A(ω) to that shown in Fig. 2 operating at and less distinct absorbance peak.
94 GHz and 1 THz. NIL and WJP acknowledge support from the Los
While the creation of a metamaterial absorber is novel, Alamos National Laboratory LDRD program, DRS ac-
the technology to create highly absorptive materials in knowledges support MDA Contract W9113M-07-C-0078,
the microwave frequency range is well-established. The and DRS, SS, and JJM acknowledge support from a Mul-
material “chiroshield” is capable of very high absorption, tiple University Research Initiative sponsored by the Air
and reduces backscatter by 15 to 25 dB [23]. Chiroshield, Force Office of Scientific Research Contract No. FA9550-
however, has geometries on the order of one wavelength, 06-1-0279.
which is 35 times larger than our design at a given center
frequency. Some ferrites show reflection loss of -30 dB
[24]. These ferrites, however, lack the inherent and pre-
cise tunability of metamaterials. Other work has been
done on the application of frequency-selective materials [1] V. G. Vesalago, Sov. Phys. Usp. 10, 509 (1968).
to bolometric devices. These bolometric designs, how- [2] R. A. Shelby, D. Smith, and S. Schultz, Science 292, 77
ever, require cryogenic temperatures to operate and only (2001).
[3] D. Schurig, J. J. Mock, B. J. Justice, S. A. Cummer, J. B.
achieve absorbances of about 50%[25]. At optical wave-
Pendry, A. F. Starr, and D. R. Smith, Science 314, 977
lengths, metal colloides are known to have large absorp- (2006).
tive properties,[26] due to the geometric specific surface [4] J. B. Pendry, A. J. Holden, W. J. Stewart, and I. Youngs,
modes[27]. This suggests and interesting possible appli- Phys. Rev. Lett. 76, 4773 (1996).
cation of artificial plasmonic wire media[4] shaped into [5] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J.
particular geometries to function as perfect absorbers[16] Stewart, IEEE Trans. Microwave Theory Tech. 47, 2075
operating at microwave frequencies. (1999).
[6] M. C. K. Wiltshire, J. B. Pendry, I. R. Young, D. J.
In conclusion, we have demonstrated that metamate-
Larkman, D. J. Gilderdale, and J. V. Hajnal, Science
rials can be highly absorptive over a narrow frequency 291, 849 (2001).
range. This stems from the ability to design metamate- [7] D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-
rial elements which can individually absorb the electric Nasser, and S. Schultz, Phys. Rev. Lett. 84, 4184 (2000).
and magnetic components of an incident electromagnetic [8] M. Gokkavas, K. Guven, I. Bulu, K. Aydin, R. S. Penciu,
wave. In contrast previous experimental results at THz M. Kafesaki, C. M. Soukoulis, and E. Ozbay, Phys. Rev.
frequencies using a single type of metamaterial element B 73, 193103 (2006).
[9] T. J. Yen, W. J. Padilla, N. Fang, D. C. Vier, D. R.
(electric) yielded values of only A(ω)=∼ 20%.[28] The
Smith, J. B. Pendry, D. N. Basov, and X. Zhang, Science
resonant frequency of the metamaterial elements may 303, 1494 (2004).
be tuned throughout some range of frequencies [29, 30] [10] S. Linden, C. Enkrich, M. Wegener, J. Zhou, T. Koschny,
thus enabling hyperspectral imaging. A further benefit and C. M. Soukoulis, Science 306, 1351 (2004).
afforded by metamaterials is the ability to construct a [11] S. Zhang, W. Fan, N. C. Panoiu, K. J. Malloy, R. M.
single unit cell with µ(ω) = (ω) over an extended fre- Osgood, and S. R. J. Brueck, Phys. Rev. Lett. 95, 137404
quency range. Thus this unit cell can achieve zero re- (2005).
[12] G. Dolling, M. Wegener, C. M. Soukoulis, and S. Linden,
flectance since it can p have an impedance equal to the
Opt. Lett. 32, 53 (2007).
free space value Z = µ/ = 1. This is similar to the [13] D. R. Smith and J. B. Pendry, J. Opt. Soc. Am. B 23,
well known theoretical construct, the Perfectly Matched 391 (2006).
Layer (PML) Absorbing Boundary Condition (ABC),[31] [14] W. J. Padilla, M. T. Aronsson, C. Highstrete, M. Lee,
which splits waves incident upon a boundary into electric A. J. Taylor, and R. D. Averitt, Phys. Rev. B 75, 041102
and magnetic components to obtain near perfect absorb- (2007).
tion. The PML, however, requires gain and additionally [15] D. Schurig, J. J. Mock, and D. R. Smith, Appl. Phys.
Lett. 88, 041109 (2006).
is on the order of λ0 /2 in thickness[32].
[16] D. R. Smith, D. C. Vier, W. Padilla, S. C. Nemat-Nasser,
The design presented here can still be improved. By and S. Schultz, App. Phys. Lett. 75, 1425 (1999).
incorporating a substrate with a highly consistent dielec- [17] CST Studio Suite 2008, CST of America.
tric constant, we will be able to optimize the design at http://www.cst.com (2008).
the correct resonant frequency. Similarly, tighter fabri- [18] D. R. Smith, S. Schultz, P. Markoš, and C. M. Soukoulis,
cation tolerances would allow us to bring the impedance Phys. Rev. B 65, 195104 (2002).
closer to unity and therefore closer to 100% absorption. [19] S. G. Johnson, P. Bienstman, M. Skorobogatiy,
M. Ibanescu, E. Lidorikis, and J. D. Joannopoulos, Phys.
More progress could also be made on the design itself. Rev. E 66, 066608 (2002).
Currently, the absorber is polarization sensitive, which [20] A. F. Starr, P. M. Rye, D. R. Smith, and S. Nemat-
is not ideal for some applications. Additionally, the as- Nasser, Phys. Rev. B 70, 113102 (2004).
sembly is complicated due the restrictions placed on the [21] J. E. Raynolds, B. A. Munk, J. B. Pryor, and R. J.
 6

Marhefka, J. Appl. Phys. 93, 5346 (2003).

[22] P. L. Richards, J. App. Phys. 76, 1 (1994).
[23] I. V. Lindell, A. H. Sihvola, S. A. Tretykov, and A. J. Vi-
itanen, Electromagnetic waves in Chiral and Bi-Isotropic
Media (Artech House, 1994).
[24] P. Singh, V. K. Babbar, A. Razdan, S. L. Srivastava, and
T. C. Goel, Materials Science and Engineering B 78, 70
(2000).
[25] T. A. Perera and et al., App. Opt. 45, 7643 (2006).
[26] W. Lamb, D. M. Wood, and N. W. Ashcroft, Phys. Rev.
B 21, 2248 (1980).
[27] J. B. Pendry, J. Mod. Opt. 41, 209 (1994).
[28] J. F. O’Hara, E. Smirnova, H.-T. Chen, A. J. Taylor,
R. D. Averitt, M. L. C. Highstrete, and W. J. Padilla, J.
Nanoelectron. Optoelectron. 2, 90 (2007).
[29] W. J. Padilla, D. R. Smith, and D. N. Basov., J. Opt.
Soc. Am. B 23, 404 (2006).
[30] H.-T. Chen, J. F. O’Hara, A. K. Azad, A. J. Taylor,
R. D. Averitt, D. Shrekenhamer, and W. J. Padilla, Na-
ture Photonics – To appear (2008).
[31] J.-P. Berenger, J. Comp. Phys. 114, 185 (1994).
[32] A. Taflove and S. C. Hagness, Computational Electro-
dynamics: The Finite-Difference Time-Domain Method,
Third Edition (Artech House, 2000).

View publication stats