PHYSICAL REVIEW B 79, 033101 共2009兲

Wide-angle perfect absorber/thermal emitter in the terahertz regime

Marcus Diem,1,* Thomas Koschny,1,2 and C. M. Soukoulis1,2

1Department of Physics and Astronomy and Ames Laboratory, Iowa State University, Ames, Iowa 50011, USA
2Department of Material Science and Technology and Institute of Electronic Structure and Laser (IESL),

Foundation for Research Technology Hellas (FORTH), University of Crete, 71110 Heraklion, Crete, Greece
共Received 9 November 2008; published 7 January 2009兲
We show that a perfect absorber/thermal emitter exhibiting an absorption peak of 99.9% can be achieved in
metallic nanostructures that can be easily fabricated. The very high absorption is maintained for large angles
with a minimal shift in the center frequency and can be tuned throughout the visible and near-infrared regime
by scaling the nanostructure dimensions. The stability of the spectral features at high temperatures is tested by
simulations using a range of material parameters.

DOI: 10.1103/PhysRevB.79.033101 PACS number共s兲: 44.40.⫹a, 78.20.Ci

Since the beginning of the last century it is known that a the specific needs of the final application, e.g., to obtain good
perfect thermal emitter follows Planck’s law of blackbody heat transport to sensors or to obtain a better stability. On top
radiation.1 Realistic structures, however, generally do not of the metal plate a spacer layer of silicon nitride SiN is
follow Planck’s law but exhibit a smaller emission. The deposited with a thickness Dt. The structure is terminated by
properties of these emitters strongly depend on the materials an array of metallic stripes with a rectangular cross section.
and their shapes. From the absorption spectra of a structure Their arrangement is described by a lattice constant a and
the emission properties can be deduced since Kirchhoff’s law
their shape is given by a width Ww and a thickness Wt. In this
directly relates the absorption with the emissivity. The emis-
sion is then determined by multiplying the emissivity with setup a strong resonance with a large field enhancement in
the blackbody radiation spectrum. Using photonic crystals,2,3 the dielectric spacer layer and in between the stripes can be
it has been shown that this approach is also valid for peri- obtained, as will be shown later. Adjusting the size of the
odically structured materials. For a number of applications metal stripes on the top, the coupling to this resonance can
such as thermophotovoltaic converters, it is necessary to con- be tuned and the reflection can be minimized.
trol the spectral properties to achieve, e.g., selective emitters Due to the scalability of Maxwell’s equations, in prin-
in a narrow frequency band corresponding to the band gap of ciple, the structure can be simulated using dimensionless
solar cells.4 In the case of structured metallic surfaces, the units by dividing all sizes by the lattice constant and using
changes in the emission spectra are based on surface waves ␻⬘ = a / ␭ as frequency. However, the Drude model used to
coupled to the external radiation through the periodic describe the metal requires frequencies in terahertz and
surface.5,6 Alternatively, microcavity resonances can also be
used to create narrow-band thermal radiation.7 Unfortu- therefore the lattice constant must be assigned in the simula-
nately, most of the recent designs6,8 for perfect absorbers/ tion. If a shift in the frequencies of the spectral features by
emitters only work for one incident angle and one polariza- adjusting the lattice constant is intended, a different simula-
tion. So, there is a need for wide-angle perfect absorber/ tion must be done since changes in the dielectric constant
emitter nanostructures. In this Brief Report, we suggest a would not be considered.
structure which exhibits a large absorption in the terahertz In the simulation frequency-dependent material param-
regime for a wide range of angles with respect to the surface. eters are required. We calculate those using standard methods
We show that the absorption characteristics are maintained and adjust their values to take into consideration the high
even if the uncertainties in the estimated changes in the ma- temperatures. The tungsten parts 共plate and stripes兲 are de-
terial parameters, due to high temperatures, are considered. scribed by a Drude model
The proposed structure can be easily manufactured with to-
day’s planar microfabrication techniques. We also comment
on the impact of deviations in the geometrical parameters θ
caused by fabricational tolerances. The small size of the Ww a
structure, in comparison to the wavelength together with the
relatively straightforward fabrication, allows for easy inte- Wt
gration into various devices, such as perfect thermal emitters, Dt
perfect absorbers, bolometers, and very effective light ex-
traction light-emitting diodes 共LEDs兲.
The suggested structure is shown in Fig. 1. It consists of a FIG. 1. Definition of the structure and parameters. An incoming
metal back plate 共black兲 with a thickness larger than 200 nm. plane wave with an angle ␪ to the surface normal is considered. The
This is much larger than the typical skin depth in the tera- structure consists of an array 共lattice constant a = 2 ␮m兲 of tungsten
hertz regime and avoids transmission through the structure. wires 共Wt = 0.2 ␮m , Ww = 0.4 ␮m兲 on top of a silicon nitride sub-
In this case the reflection is the only factor limiting the ab- strate 共Dt = 0.65 ␮m兲. The tungsten plate at the bottom must be
sorption. The thickness of the back plate can be adjusted to thicker than 200 nm. For material parameters, see text.

1098-0121/2009/79共3兲/033101共4兲 033101-1 ©2009 The American Physical Society

BRIEF REPORTS PHYSICAL REVIEW B 79, 033101 共2009兲

␻2p 1
⑀=1−
␻ 共 ␻ + i ␻ c兲 R
0.8 Abs
with the plasma frequency ␻ p = 1448 THz and the collision
frequency ␻c = 13 THz at room temperature.9 In order to ac- 0.6
count for the higher temperatures, we use an increased value
for the collision frequency. Since the resistivity of tungsten
increases linearly with temperature, a linear dependence of 0.4
the collision frequency is assumed, leading to an estimated
increase in the collision frequency by a factor of 3–5 共␻c 0.2
= 50 THz兲. The plasma frequency is assumed to be constant.
The simulations are repeated for several values of the colli- 0
sion frequency 共13, 50, and 100 THz兲 to ensure the stability 50 60 70 80 90 100
of the emission spectra in a wide range. ν [THz]
The index of SiN depends on the actual fabrication
FIG. 2. 共Color online兲 Absorption and reflection for perpendicu-
process10 and shows a strong wavelength dependence below
lar incidence. The tungsten back plate is thicker than the skin depth
1 ␮m but is roughly constant for longer wavelengths until and does not allow for transmission. The absorption reaches more
about 8 ␮m.11 For our simulation we use experimental val- than 99.9% for ␯ = 69.24 THz 共␭ = 4 , 3 ␮m兲.
ues for room temperature obtained from Sandia National
Laboratory12 and fitted a Cauchy model for dielectrics in the
cavity geometry and the coupling to it lead to a substantial
infrared to calculate the required values. In order to account
improvement. We find that the most important criteria for
for temperature increase, we assume a linear temperature de-
obtaining a very weak angular dependence are to avoid cou-
pendence as found, e.g., in silica13 and add a constant value
pling to propagating first negative diffraction orders. In the
共+0.1兲 to the imaginary part of the experimental data. Again,
simulations we choose a lattice constant of a = 2 ␮m, so that
we perform several simulations with different values of the
the first negative diffraction order at large angles 共␪ ⬇ 90°兲 is
imaginary parts of SiN 共0.05, 0.1, 0.15, and 0.2兲 for compari-
given by ␭ / a ⬇ 2 corresponding to 75 THz. Any absorption
son.
peak at higher frequencies couples to several reflected dif-
For the simulations we use our own implementation of a
fraction orders and shows a strong angular dispersion at the
Fourier-modal method with a scattering matrix approach,
corresponding frequency/angle combination. For a structure
also known as rigorous-coupled wave analysis 共RCWA兲.14
with the parameters given in Fig. 1, we find a strong absorp-
Special care is taken in use of the correct Fourier-
tion peak for perpendicular incidence fulfilling the above cri-
factorization rules to ensure a fast convergence.15 This ap-
teria. The corresponding reflection and absorption spectra are
proach assumes incoming plane waves onto a periodic sur-
shown in Fig. 2.
face defined by the dimensionless frequency ␻⬘ = ␻a / 2␲c
The obtained absorption peak has a center frequency of
= a / ␭ and the angle ␪ to the surface normal 共Fig. 1兲. For the
␯0 = 69.24 THz with a half-width at half-maximum of 10.27
material parameters the frequency is calculated using a fixed
THz corresponding to 14.8% of the center frequency. Simu-
lattice constant of a = 2 ␮m. The absorption is determined by
lations with MULTIPHYSICS showed that the field is able to
subtracting the total reflection of all diffraction orders given
couple into the dielectric layer in contrast to regions outside
by the individual Poynting vectors. The transmission is on
the resonance. A strong enhancement of the electric field can
the order of the numerical error. By Kirchhoff’s law the ab-
be found in the region between the metallic stripes mostly
sorption equals emissivity and the emission can be calculated
located in the dielectric 共Fig. 3兲. Numerically, we find an
by multiplying the absorption with the blackbody radiation,
eigenmode showing the same field pattern at an eigenfre-
even in the case of structured materials.3
To confirm the results and to study the field distributions/
energy flow and resistive heating, we use the commercially
available finite-element-based program, MULTIPHYSICS, by
COMSOL with periodic boundary conditions in the direction
parallel to the surface. In this case the absorption is deter-
mined by A = 1 − 兩S21兩2 − 兩S11兩2. This program is also used to
determine possible eigenmodes and their corresponding
complex eigenfrequencies. In this simulation, the material
parameters are set to the values at the resonance frequency.
For wide-angle absorption in a small frequency window,
the center frequency of the absorption peak should not shift
for oblique incoming plane waves at different angles. It was FIG. 3. 共Color online兲 Magnitude of the z component of the
shown before that tungsten microcavities can exhibit a weak electric field at the resonance. The linear scale ranges from
angular dependence;16 however, both the angular range and 0 V m−1 共green online, white in print兲 to 25 000 V m−1 共red on-
maximum absorption are smaller than in our case. The addi- line, dark in print兲. The maximum of the field is located between the
tional dielectric layer offers the possibility to optimize the metal stripes in the dielectric.

033101-2
BRIEF REPORTS PHYSICAL REVIEW B 79, 033101 共2009兲

150 1
(a) 0.9
0.8 0.8
125
0.7
0.6
0.6
ν [THz]

100
0.5 0.4
0.4
75
Black Body Spectrum
0.3 0.2
Absorption
0.2 Emission
0.1
0
50 40 50 60 70 80 90 100
ν [THz]
0 20 40 60 80
θ FIG. 5. 共Color online兲 Blackbody radiation 共dotted curve兲, ab-
1 sorption 共dashed curve兲, and emission 共solid curve兲 of the structure,
(b) if the temperature is adjusted so that the maxima of the blackbody
0 spectra and the absorption peak coincide using Wien’s law 共T
0.8 50 = 1176 K兲. Emission and absorption are roughly the same due to
60
Absorption

high value of the blackbody radiation.

0.6 70
80
88 parameters are chosen, so that the lowest peak has a higher
0.4 center frequency 共above 75 THz for the given lattice con-
stant兲, it shows a weak angular dispersion until the first nega-
0.2 tive starts propagating and then the absorption vanishes
quickly.
Finally, we compare the emission properties of our struc-
0
40 50 60 70 80 90 100 ture with a regular blackbody emitter in Fig. 5. We plot the
ν[Thz] blackbody emission using Planck’s law at a temperature of
1176 K, so the maximum of the emission corresponds to the
FIG. 4. 共Color online兲 共a兲 Angular peak dispersion of the ab- center frequency of the absorption peak. The emission of the
sorption peak. The additional peaks around 110 and 150 THz show structure given by the product of the emissivity and the
a strong dispersion although they are caused by cavity modes as blackbody radiation is also plotted in Fig. 5 as a solid line.
well. 共b兲 For 0°–40° the absorption is above 99.7%. Even at 70°
The combined effect of the two aspects suppresses the sides
共80°兲 still 83.3% 共58.1%兲 of the incoming energy is absorbed.
of the peak even stronger, so a selective emission is
quency of 70.05 THz 共real part of the complex eigenfre- achieved.
quency兲. In this simulation the material parameters are fixed For these high temperatures, the material parameters are
to the values at ␯0. Changing the thickness of the metal plate not known exactly and must be estimated. Since estimates
or the air space in front of the structure changes the eigen- are always a possible source of errors, we ensured that the
frequency by less than 0.01 THz once the air part in front of spectral properties exist in a wide range of parameters. We
the structure is chosen large enough. Although losses in the tested the results for different adjustments in the collision
dielectric layer occur due to the imaginary part, the energy frequency of tungsten 共␻c = 13, 50, 100 THz兲 and the imagi-
flow described by the Poynting vector reveals that the main nary part of the SiN dielectric 共0.05, 0.1, 0.15, and 0.2兲. For
absorption takes place at the surface of metallic back plate all combinations of these parameters, the absorption reaches
and the rear side of the metal stripes. values of 93% or more, except for the case with ␻c
If the structure is used as a perfect absorber, e.g., in sensor = 13 THz and Im ⑀ = 0.05 for which only 82% are achieved.
applications, it is important to absorb as much of the radia- The center frequency of the absorption peak remains un-
tion as possible, independent of the direction of incidence. In changed. However, an increase in the imaginary part of SiN
Fig. 4 we show the angular dispersion of the peak together leads to a broadening of it. In general, an increase in the
with the spectra for several angles. For angles up to 40° the collision frequency allows for a wider range of possible ad-
peak becomes slightly narrower and the center frequency in- justments in the imaginary part of the SiN.
creases by 0.5 THz, but the maximum absorption is above Although we do not present the data in this Brief Report,
99.7%. For larger angles the peak starts to drop, but is still we also conducted several simulations to determine the sen-
above 80%, even for waves impinging on the structure with sitivity of the absorption peak to fabricational tolerances. In
an angle of ␪ = 70°. In Fig. 4共a兲, two additional absorption this study, we varied all three parameters 共Ww , Wt , Dt兲 by
peaks at higher frequencies are visible. Both peaks are ⫾50 nm individually and combined. In all cases an absorp-
caused by resonant modes in the dielectric as well, but their tion of more than 98.8% was obtained. Wider 共narrower兲
angular dependence shows a very strong dispersion. If the stripes lead to a small increase 共decrease兲 in the frequency of

033101-3
BRIEF REPORTS PHYSICAL REVIEW B 79, 033101 共2009兲

less than 1 THz. For thicker stripes, the coupling became be built with planar fabrication techniques, it offers an inter-
weaker and the absorption dropped by 1%. Changes in the esting approach to wide-angle perfect absorbers/emitters.
thickness of the dielectric layer caused the strongest change First studies using a lattice constant of several millimeters
in the absorption spectra, shifting the center frequency of the showed that it is possible to obtain similar effects also for
peak by about ⫾4 THz. Since the field was concentrated frequencies in the gigahertz range. Due to the wide angular
mainly in this area, deviations in the spacer thickness absorption this can offer a new way to avoid reflections in
changed the effective size of the cavity and, therefore, the
microwave experiments with an easy-to-build structure on
frequency of the cavity mode. This makes it important to
control the thickness of the dielectric layer very precisely in length scales smaller than the wavelength.
the fabrication process.
In summary, we presented a design for a perfect absorber/ M.D. gratefully acknowledges financial support from the
thermal emitter with a very high absorption peak over a wide Alexander-von-Humboldt Foundation 共Feodor-Lynen Pro-
range of angles. We demonstrated that the absorption peak is gram兲. Work at Ames Laboratory was supported by the De-
stable with respect to the estimated changes in the material partment of Energy 共Basic Energy Sciences兲 under Contract
parameters at high temperatures. We also studied the impact
No. DE-AC02-07CH11358. This work was partially sup-
of deviations in the geometry caused by fabricational toler-
ances. From this it can be expected that the spectral features ported by the Office of Naval Research 共Grant No. N0014-
are also present in realistic samples. Since the structure can 07-1-0359兲.

*[email protected] 8 H. Tao et al., Opt. Express 16, 7181 共2008兲; N. I. Landy, S.

1
M. Planck, Ann. Phys. 309, 553 共1901兲. Sajuyigbe, J. J. Mock, D. R. Smith, and W. J. Padilla, Phys. Rev.
2 J. G. Fleming et al., Nature 共London兲 417, 52 共2002兲; S. Y. Lin
Lett. 100, 207402 共2008兲.
et al., Appl. Phys. Lett. 83, 380 共2003兲; S. Y. Lin et al., Opt. 9
M. A. Ordal et al., Appl. Opt. 22, 1099 共1983兲.
Lett. 28, 1909 共2003兲; S. E. Han, A. Stein, and D. J. Norris, 10 A. Piccirillo and A. L. Gobbi, J. Electrochem. Soc. 137, 3910
Phys. Rev. Lett. 99, 053906 共2007兲. 共1990兲; B.-H. Jun et al., J. Mater. Res. 14, 995 共1999兲; C. C.
3 D. L. C. Chan, M. Soljačić, and J. D. Joannopoulos, Phys. Rev.
Lee, H. L. Chen, J. C. Hsu, and C. L. Tien, Appl. Opt. 38, 2078
E 74, 036615 共2006兲.
4 H. Sai and H. Yugami, Appl. Phys. Lett. 85, 3399 共2004兲; H. Sai 共1999兲.
11
N. Ravindra et al., IEEE Trans. Semicond. Manuf. 11, 30
et al., J. Opt. Soc. Am. A 22, 1805 共2005兲.
5 J.-J. Greffet et al., Nature 共London兲 416, 61 共2002兲; M. Laroche 共1998兲.
12 I. El-Kady 共private communication兲.
et al., Opt. Lett. 30, 2623 共2005兲; I. Puscasu and W. L. Schaich, 13 A. D. McLachlan and F. P. Meyer, Appl. Opt. 26, 1728 共1987兲.
Appl. Phys. Lett. 92, 233102 共2008兲; F. Marquier, K. Joulain, J. 14
P. Mulet, R. Carminati, J. J. Greffet, and Y. Chen, Phys. Rev. B G. G. Moharam and T. K. Gaylord, J. Opt. Soc. Am. 71, 811
69, 155412 共2004兲; T. V. Teperik et al., Nat. Photonics 2, 299 共1981兲; D. M. Whittaker and I. S. Culshaw, Phys. Rev. B 60,
共2008兲. 2610 共1999兲.
6 M. Laroche, R. Carminati, and J.-J. Greffet, Phys. Rev. Lett. 96, 15 L. Li, J. Opt. Soc. Am. A 13, 1870 共1996兲; 14, 2758 共1997兲; L.

123903 共2006兲. Shen and S. He, ibid. 19, 1021 共2002兲.

7 S. Maruyama et al., Appl. Phys. Lett. 79, 1393 共2001兲; I. Cel- 16 H. Sai, Y. Kanamori, and H. Yugami, Appl. Phys. Lett. 82, 1685

anovic, D. Perreault, and J. Kassakian, Phys. Rev. B 72, 075127 共2003兲; F. Kusunoki et al., Jpn. J. Appl. Phys., Part 1 43, 5253
共2005兲. 共2004兲.

033101-4