Letters
https://doi.org/10.1038/s41566-018-0098-3
Universal impedance matching and the perfect
transmission of white light
Ku Im 1,2
, Ji-Hun Kang 1,2
and Q-Han Park1*
Light is reflected at the interface between heterogeneous
media due to the mismatch of impedance1–3. Removing this
mismatch using additional materials, a technique known
as anti-reflection, has so far been restricted to specific fre-
quencies and incidence angles3–7. The anti-reflection of white
light, which requires the simultaneous matching of imped-
ance over extremely wide angular and spectral ranges, has
until now been considered impossible. Here, we develop
a theory of universal impedance matching and introduce
a matching layer that enables the perfect transmission of
white light. The ability of a matching layer to assist in omni-
directional and frequency-independent anti-reflection has
been confirmed analytically and numerically. We explain the
feasibility of a universal matching layer using metamateri-
als, and demonstrate a transmission rate of over 99% for
white light in the visible range with a double-layered dielec-
tric metamaterial. This is confirmed experimentally by dem-
onstrating the omnidirectional anti-reflection of microwaves
in heterogeneous media.
When moving from air into a dielectric medium, light will be
reflected unless it enters at the Brewster’s angle with a particular
polarization2. Maximizing light transmission in optical compo-
nents and devices by removing reflection is an important consid-
eration in optical science and technology. Rayleigh’s discovery3
of reduced reflection in tarnished glass prompted the application
of an additional layer to suppress reflection—an anti-reflection
coating4. The anti-reflection coating, which relies on destructive
wave interference, only works for specific frequencies and inci-
dence angles. Significant efforts, mostly based on trial and error,
have been expended to effectively extend single-frequency quarter-
wave anti-reflection to broader frequency ranges by either adopting
multilayers4, graded-index layers5, surface textures such as moth-
eye structures6,7 or nanoparticle scatterers8. However, these tech-
niques still tend to only function over a narrow range of incidence
angles or frequencies. Universal anti-reflection is typically repre-
sented in numerical analysis by the perfectly matched layer9, which
is used as a perfect absorber, and the invisibility cloak, which avoids
reflection by deforming the optical path10,11. However, in these cases,
impedance is constant over the entire region, so there is no imped-
ance mismatch to resolve. A non-reflecting potential has been
identified12 in applications related to soliton theory13, and recently
other types of non-reflecting but absorbing potential have also been
proposed using the spatial Kramers–Kronig relations14. This poten-
tial is defined over the entire space and is not sufficient to remove
any impedance mismatch. So far, no systematic approach has been
found to remove the impedance mismatch between two transparent
media over a full angular range and the entire visible spectra.
Here, we prove that universal impedance matching—matching
that is independent of frequency, incidence angle and polarization—
is in fact possible. We derive the principles of universal impedance
Department of Physics, Korea University, Seoul, Korea. 2These authors contributed equally: Ku Im and Ji-Hun Kang. *e-mail: qpark@korea.ac.kr
1
Nature Photonics | VOL 12 | MARCH 2018 | 143–149 | www.nature.com/naturephotonics
© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
matching from Maxwell’s equations and identify the requirements
for a matching layer at the interface between two transparent media.
Of the infinitely many possible matching layers that satisfy these
requirements, we present a unique non-graded matching layer, called
the universal impedance matching layer (UIML), which is an immedi-
ate generalization of the well-known quarter-wave anti-reflection layer
(QAR). The UIML possesses spatially and temporally dispersive per-
mittivity and permeability tensors, enabling the perfect transmission
of white light independent of the initial state of the incoming light.
We verify the ability of the UIML analytically via the direct evaluation
of reflection and transmission spectra in relation to incidence angle,
and numerically via the pseudo-spectral time-domain numerical
simulation15 of white light and a light pulse entering a dielectric
medium. We explain how the UIML can be physically realized using
metamaterials, and propose a specific metamaterial structure that can
achieve light transmission of over 99% over the entire visible spectra and
nearly the full angular range up to high glancing angles. Experimental
verification of the UIML is achieved using microwave reflection and
transmission measurements with a double-layer metamaterial, with the
results conforming to our predictions based on UIM theory.
Given the material parameters of an inhomogeneous medium,
the scattering of electromagnetic waves, including reflection as a
special case, can be described accurately using Maxwell’s equations.
However, the problem of inverse scattering—determining the mate-
rial parameters of an unknown inhomogeneous medium based on
its scattered waves—is far less well understood. Here, we show that
all of the material parameters of an unknown medium with planar
geometry can be expressed in terms of two admittance functions,
which allows the problem of inverse scattering to be solved for UIM.
Consider an electromagnetic wave incident on the planar inter-
face between two different dielectric media with isotropic permit-
tivities ε1, ε2 and permeabilities μ1, μ2, respectively. When the wave
impedance does not match (μ1/ε1 ≠​  μ2/ε2), reflection occurs accord-
ing to the Fresnel equations. We now seek to completely remove the
impedance mismatch by adding a layer (0 <​  z <​  d) between medium
I (z <​  0; ε =​  ε1, μ =​  μ1) and medium II (z >​  d; ε =​  ε2, μ =​  μ2), as shown
in Fig. 1a. Because the system is rotationally symmetrical about the
z axis, we assume that the permittivity and permeability tensors
of the matching layer are uniaxial and graded with non-vanishing
components given by
εxx = εyy ≡ εr (z ) ≠ εzz ≡ εz (z ), μxx = μ yy ≡ μr (z ) ≠ μzz ≡ μz (z ) (1)
Within the layer, the propagation of a monochromatic plane
wave of the form
→ ⎯→
⎯
E = e ik xx +ik yy −iωt E (z ), H = e ik xx +ik yy −iωt H (z ) (2)
is governed by the reduced Maxwell’s equations
143
Letters NaTuRe PHoTonics

a b
z=0 z=d
7.7 5.8
10 6.9 10 5.2

ε1, μ1 ε2, μ2 8 ε r/ε o 5.3 8 ε r/ε o 4.0

μ /μ o 6 μ r/μ o
6 =α =α 2.7
r 3.7
UIM
layer 4 4
2.1 1.5
2 2
0.56 0.28
0 10 0 10
ω, k 0 8 ) 0 8 )
ω0 An /ω 0
An
gle 30 6 (ω/ gle 30 6 y (ω
of 4 cy of 4 e nc
inc 60 en inc
ide 60 qu
ide 2 qu nc 2 fre
nc re
e( 90
la rf e(
de
90 lar
de
gu g) gu
g)
An An

c UIM layer d QAR layer

Total (a.u.) Scattered (a.u.) Total (a.u.) Scattered (a.u.)
10 10
(μm) 1.8 1.8 (μm) 1.8 1.8
1.6 1.6 1.6 1.6
1.4 1.4 1.4 1.4
1.2 1.2 1.2 1.2
5 1.0 1.0 5 1.0 1.0
0.8 0.8 0.8 0.8
0.6 0.6 0.6 0.6
0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2
0.0 0.0 0.0 0.0
0 0
0 5 (μm) 0 5 (μm) 0 5 (μm) 0 5 (μm)
Transmission = 0.995 Reflection = 0.002 Transmission = 0.859 Reflection = 0.129

Fig. 1 | Perfect transmission of white light with a universal impedance matching layer. a, A UIML of thickness d is placed at the interface between
two transparent media possessing isotropic permittivity and permeability ε1, μ1 and ε2, μ2, respectively, to remove the reflection of incident light with
angular frequency ω and wavevector k. b, Spatially and temporally dispersive permittivity and permeability of the UIML. The uniaxial permittivity tensor
(εr =​ εxx =​ εyy ≠​ εzz =​ εz) is proportional to the permeability tensor (μr =​ μxx =​ μyy ≠​ μzz =​ μz) such that ε/ε0 =​ αμ/μ0, where α = ε 1ε 2 ∕ μ1μ2 . Contour plots
show the spatial and temporal dispersion of the UIML material parameters for ε 1 = ε 0, ε 2 = 4ε 0, μ1 = μ2 = μ0, ω0 = π∕8d ε 0μ0 . c,d, Numerical simulation
of random incoherent light entering a dielectric medium with a UIML (c) and with a QAR layer (d). Incoherent light is simulated with a group of dipole
sources with randomly chosen phase and polarization. Total reflectance, calculated using the Poynting vector, shows that the UIML completely removes
reflection while the QAR layer only reduces substrate reflectance by half.

∂zEx = ikx Ez + iωμr Hy , ∂zE y (a) iωεr = ∂zY + (iωμr −ik 2 ∕ ωεz )Y 2 (b) ∂zlnΦE
(5)
2
= ik y Ez −iωμr Hx , ikx E y −ik y Ex = (iωμr −ik ∕ ωεz )Y

= iωμz Hz , Similarly, in transverse electric (TE) mode where Ez =​  0, field

(3)
∂zHx = ikx Hz −iωεr E y , ∂zHy components can be expressed in terms of the two functions Z(z),
ΦH(z) such that
= ik y Hz + iωεr Ex , ikx Hy −ik y Hx
= −iωεz Ez Ex = k y Z Φ H , E y
= −kx ZΦH , Hx
In transverse magnetic (TM) mode where Hz =​ 0, the rest of the
= k x ΦH , Hy (6)
field components can be expressed in terms of the two functions
Y(z), ΦE(z) such that = k y ΦH , Hz
= −(k 2 ∕ ωμz )ZΦH
E x = −k x Φ E , E y
= −k y Φ E , Ez and the Maxwell’s equations become

= (k 2 ∕ ωεz )YΦE, Hx (4)

(a) ωμr = ∂zZ + (iωεr −ik 2 ∕ ωμz )Z 2 (b) ∂zlnΦH
(7)
= k y Y ΦE , Hy = (iωεr −ik ∕ ωμz )Z 2

= −k x Y Φ E
The critical observation leading to UIM is that functions
Y, Z represent wave admittance and impedance, respectively, and
where k 2 ≡ kx2 + k y2, and the Maxwell’s equations become that they determine four material parameters through the simple

144 Nature Photonics | VOL 12 | MARCH 2018 | 143–149 | www.nature.com/naturephotonics

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe PHoTonics Letters
a Reflectance (QAR vs. UIM)
5 0.8 5 1 × 10–10
4 4
0.7 1.3 × 10–11
3 3
ω/ω0

ω/ω0
2 0.6 2 1.6 × 10–12
1 1
0.5 1.8 × 10–13
0 15 30 45 60 75 0 15 30 45 60 75
0.4 TM angle of incidence (deg) 1.8 × 10–14
TM angle of incidence (deg)
5 0.3 5 1.8 × 10–15
4 4
3 0.2 3 1.6 × 10–16
ω/ω0

ω/ω0
2 2 1.3 × 10–17
0.1
1 1
0 1.0 × 10–18
0 15 30 45 60 75 0 15 30 45 60 75
TE angle of incidence (deg) TE angle of incidence (deg)
b Light pulse propagation (QAR vs. UIM)
Total Scattered Total Scattered

t=0 t=0

t=1 t=1

t=2 t=2

t=3 t=3

(a.u.) (a.u.) (a.u.) (a.u.)

t=4 1.8 1.8 t=4 1.8 1.8
1.6 1.6 1.6 1.6
1.4 1.4 1.4 1.4
1.2 1.2 1.2 1.2
1.0 1.0 1.0 1.0
0.8 0.8 0.8 0.8
0.6 0.6 0.6 0.6
0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2
0.0 0.0 0.0 0.0
0 10 mm 0 10 mm 0 10 mm 0 10 mm
Transmission = 0.873 Reflection = 0.129 Transmission = 0.994 Reflection = 0.001

Fig. 2 | Omnidirectional and frequency-independent anti-reflection and transmission. a, Reflection spectra of TM- and TE-polarized incident plane waves
for different incidence angles with a QAR layer (left) or a UIM layer (right). A multilayer matrix was used to calculate the reflection spectra. The UIM layer
produced no reflection. b, Perfect transmission of a photon torpedo (that is, an optical pulse). The optical pulse is generated inside a small waveguide
and sent horizontally in both directions, as shown by the time-lapse shots in the figure. The rightwards-moving pulse with a broad frequency band and
incidence angle range enters the dielectric medium with either a QAR layer (left) or an approximate UIM layer (right). Perfect transmission with no
reflection is observed for the UIM layer. See Supplementary Videos 1 and 2.

relationships given in equations (5a) and (7a) (see Supplementary of the tangential field components requires the matching admit-
Section 1 for derivation). Therefore, instead of solving the direct tance function Y (see Supplementary Section 1 for derivation):
scattering problem, that is, determining Y, Z from the given mate- ωε1 ωε2
rial parameters, the inverse scattering problem of constructing the Y (0) = Y1(k) ≡ , Y (d ) = Y2(k) ≡ (8)
2 2
material parameters from a given Y, Z can be solved. As shown in ε1μ1ω −k ε 2 μ 2 ω 2 −k 2
the following, the specific material parameters for UIM can be con-
structed from Y, Z to satisfy reflection-less impedance (admittance) Additional boundary conditions16,17 are required to fully deter-
matching boundary conditions. If the TM-polarized incident plane mine the non-vanishing reflection and transmission coefficients,
wave in region I experiences no reflection at the interface (z =​  0) and which for our purposes are not needed. Similarly, for the TE mode,
transmits completely into region II (z >​  d), maintaining the continuity reflection-less impedance matching requires that

Nature Photonics | VOL 12 | MARCH 2018 | 143–149 | www.nature.com/naturephotonics 145

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letters NaTuRe PHoTonics

a b Universal matching layer

700
εc 1
600
d 2
εc 500

Wavelength (nm)
3
400
0 30 60
4
Double-layer metamaterial
700
5
600
d ε(ω,k) 6
500

400 7
0 30 60
Angle of incidence (deg)
d c
5
80
ε/ε 4
70 Air 3
0.03
0.03 50 nm 400 500 600 700 (nm)
60 0.02
0.01
ε = 2.5~ 25 nm
Angle (deg)

50 0.01
Air
0.01 105 nm
40 0.005 0.03
0.02
30 ε = 10 25 nm
0.001 55 nm
20

10 0.03 ε=4

0
400 450 500 550 600 650 700
Wavelength (nm)
e f
0.03
1.0
80 0.03
0.01 0.02
Reflectance
70
0.8 Transmittance
60 0.005 0.005 Reflectance (QAR)
0.02 0.03
Angle (deg)

50 0.6

40 0.02
0.4
30 0.01 0.01
20
0.005 0.2
0.00
10 0 20 40 60 80

005 0 0.0
0
0 20 40 60 80
400 450 500 550 600 650 700
Angle of incidence (deg)
Wavelength (nm)

Fig. 3 | Metamaterial realization of the UIML. a, A thin double layer constructed from constant-dielectric materials with a subwavelength gap can be
regarded as a metamaterial possessing spatiotemporal dispersion. b, Spatiotemporal dispersion of the permittivity εr of the UIML in Fig. 1 (above) and the
effective permittivity of the double-layer metamaterial (below) with d = 80 nm, d1 = 13 nm, ε c = 8ε 0. c, Structure of the double-layer metamaterial used
for the UIML. The top layer is made of a dielectric material possessing normal dispersion. d,e, Reflection spectra of visible light of varying incidence angles
with a QAR layer (d) and with the double-layer metamaterial in c (e). f, Reflection (black) and transmission (red) spectra of TM waves at a wavelength of
425 nm with a metamaterial UIM layer (vertical dashed line in e). Inset, Zoomed plot of reflectance.

ωμ1 ωμ 2 required of the UIM layer. Because any incoming light is a linear
Z (0) = , Z (d ) = (9) superposition of its TE and TM components, this UIM layer works
2 2 2 2
ε1μ1ω −k ε 2 μ 2 ω −k
regardless of the characteristics of the incoming light, such as fre-
quency, polarization and incidence angle. If the goal is the complete
Because the TM and TE modes are the same in terms of the nor- transfer of wave energy without any loss or reflection, the material
mal incidence limit18 (k = 0), an additional condition imposed on parameters for region I, region II and the UIM layer must be real.
Y, Z is This restriction is valid within a finite frequency range. The inevi-
table losses incurred by dispersive media will be addressed later.
Y (z )∣ k=0 = Z (z )−1∣ k=0 (10) By requiring parameters εr, εz, μr, μz to be real and constant, a
solution to Maxwell’s equations in equations (5) and (7) can be
When an otherwise arbitrary Y, Z satisfy the matching conditions found that satisfies the impedance matching conditions in equa-
in equations (8) to (10), the material parameters constructed from tions (8) to (10) (see Supplementary Section 2 for derivation). This
Y, Z according to equations (5a) and (7a) describe the characteristics solution characterizes UIML for white light with the conditions

146 Nature Photonics | VOL 12 | MARCH 2018 | 143–149 | www.nature.com/naturephotonics

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe PHoTonics
π ⟨Y ⟩
εr = , μr
2ωd ⟨R⟩
π k2
= , 2
2ωd ⟨Y ⟩ ⟨R⟩ ω εz
 1  k2
π
=  −⟨R⟩  , 2
2ωd ⟨Y ⟩  ⟨R⟩  ω μz
π⟨Y ⟩  1  quency band and the incidence angles of the optical pulse.
=  −⟨R⟩  , (11)
2ωd  ⟨R⟩ 
⟨Y ⟩ ≡ Y1(0)Y2(0) , ⟨R⟩ ≡ cosθ1cosθ2 , Yn(0)
εn
= , cosθn
μn
k2
= 1− ;n
εnμn ω 2
= 1, 2
Here, 〈​Y〉​ and 〈​R〉​are the geometric mean admittance and direc-
tional cosine of the propagating waves in media 1 and 2. For nor-
mal incidence (k =​ 0) and fixed ω, this reduces to the well-known
anti-reflection condition for a quarter-wave matching layer. Thus,
the condition in equations (11) is an immediate generalization of
the conventional anti-reflection condition that incorporates all
incidence angles and frequencies. It should be emphasized that
the UIML, determined by the condition in equations (11), is non-
absorbing and allows the perfect transmission of white light into an
inhomogeneous medium. This is clearly different from the perfect
absorber, such as the perfectly matched layer in numerical simula-
tions, which removes not only reflection but also transmission.
Figure 1 presents a numerical characterization of the UIML.
Figure 1b is the contour plot of the dispersive permittivity and per-
meability tensor of the UIML in equation (11) applied to an air–
substrate interface with ε1 =​  ε0, μ1 =​  μ0, ε2 =​  4ε0 and μ2 =​  μ0. Near the
grazing angle of incidence, the directional cosine factor ⟨R⟩ becomes
small. Subsequently, εz, μz of the UIML shrinks and vanishes (εr, μr
grows and diverges, respectively) at 90°, where reflection ceases to
have meaning. The omnidirectional, perfect transmitting behaviour
of the UIML is demonstrated in Fig. 1c through numerical simula-
tion of the propagation of incoherent light radiating from a group
of dipole sources with a randomly chosen position and polariza-
tion near the air–substrate interface. With a UIM layer (Fig. 1c)
or a QAR layer (Fig. 1d) placed at the air–substrate interface, trans-
mittance and reflectance are calculated in terms of the total energy
flow using the time-averaged Poynting vector. Compared to the
quarter-wave conventional anti-reflection layer, which only reduces
reflection from the substrate by half due to the wide range of inci-
dence angles for incoherent light, the UIM layer demonstrates per-
fect transmission by completely removing reflection.
Note that the UIML parameters in equation (11) depend on
ω and k, so they possess spatiotemporal dispersion leading to non-
local responses in time and space19. This dispersion is required for the
complete removal of reflection over the entire range of frequencies
and incidence angles, as shown in the reflectance spectra of Fig. 2a
obtained with the multilayer matrix calculation20. However, the cau-
sality requirement in temporally dispersive materials, expressed by
the Kramers–Kronig relation, demands that the UIML parameters
cannot be real over the entire frequency range but that non-zero
imaginary components of the material parameters should arise.
Nevertheless, within a finite frequency band, the parameters can
be approximated in terms of analytically continued functions and
assumed to be real and causal. An example of the Drude-type approx-
imation of UIM respecting causality (see Supplementary Section 3)
Nature Photonics | VOL 12 | MARCH 2018 | 143–149 | www.nature.com/naturephotonics
© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letters
is used for the numerical simulation of the optical pulse propaga-
tion shown in Fig. 2b. A short pulse comprising a wide-band fre-
quency is generated inside a small waveguide (not shown) and sent
horizontally in both directions. The left-moving pulse moves in the
empty space while the right-moving pulse enters into the dielectric
medium with a wide range of incidence angles. In contrast to the
QAR (left), the approximated UIM layer (right) enables nearly per-
fect transmission with negligible reflection, despite the wide fre-
To this point, we analytically confirmed universal imped-
ance matching and numerically tested its validity. The feasibility
of the UIML in terms of practical use could be doubted because
it requires an unusual form of spatiotemporal dispersion that is
difficult to realize with naturally occurring materials. Even for a
specific angle of incidence (that is, fixed k), dispersion requires
an anomalous temporal dispersion relation that is difficult to
find in lossless transparent materials. Here, we explain how this
dispersion can be achieved using metamaterials. It is known that
metamaterials such as metallic nanowires21 or metal–dielectric
multilayers22 cause strong spatial dispersion due to structural
inhomogeneity. To design a metamaterial structure for a UIML, we
first observe that the effective material parameters of a multilayer
metamaterial obtained by the Nicolson–Ross–Weir parameter
retrieval method23,24 are functions of the wavevector components
k zm = εmμ m ω 2−k 2 in each layer m (see Supplementary Section 4
for details). This k2-dependence resembles that of the UIML. For
instance, the metamaterial constructed from constant dielectric
layers presented in Fig. 3a possesses the effective permittivity
shown in Fig. 3b. Surprisingly, this effective permittivity exhibits
anomalous temporal dispersion without loss, and also increasing
behaviour as the incidence angle increases, which is a character-
istic of the UIML described in Fig. 3b. This suggests that variants
of a double-layer metamaterial, with the right choice of constitut-
ing materials and structural optimization, can realize a UIML for
practical applications. Figure 3c presents an example of an opti-
mized metamaterial that realizes UIM for TM-mode white light.
It consists of two periodically grooved dielectric layers, with the
upper dielectric layer having normal dispersion. When applied to
a substrate with ε =​  4ε0, it achieves near-perfect transmission of
white light into the substrate by reducing the reflectance to less
than 1% in the visible frequency range and for incidence angles
reaching up to 75° (Fig. 3e). This may be compared with the
quarter-wave layer in Fig. 3d, which covers only a narrow band of
frequencies and incidence angles. A more elaborate and optimally
designed metamaterial structure could further improve perfor-
mance. In particular, a doubly periodic two-dimensional groove
structure can be used for the UIM of white light in both TE and
TM modes simultaneously.
To experimentally confirm universal impedance matching, we
conducted microwave reflection measurements. As shown in Fig. 4a,
two patch antennas were used as the source and the detector inside
a waveguide formed by two parallel aluminium plates, support-
ing single-TM-mode propagation. A 10-mm-thick slab of high-
density polyethylene (HDPE), with refractive index of n =​  1.50 (or
waveguide effective index neff of 1.40) at 2.4 GHz, was selected as
the dielectric substrate. The optimized design for the metamaterial
UIML is given in Fig. 4b; this consists of two periodically grooved
layers separated by Styrofoam. We placed the metamaterial UIML
symmetrically on both sides of the slab to eliminate reflection.
Measurements of reflectance were made by measuring the scat-
tered field, which was achieved by subtracting the empty space
background field from the total measured field and normalizing the
reflectance using the reference value (see Supplementary Section 5
for details). Measurements were made over a frequency range of
2.2 GHz to 2.6 GHz and incidence angles up to 75°, with varying
positions of the source and detector. The slab substrate alone
147
Letters NaTuRe PHoTonics

a d QAR (simulation) e Substrate (exp.)

Aluminium plate 80 70
70
60

Angle of incidence

Angle of incidence
60
112.5 mm 50
50
40 40
2.0
Patch antenna 30
30
y 20
20 1.6
x z 2,440 mm 10
1,220 mm

Reflectance (%)
0 10
2.2 2.3 2.4 2.5 2.6 2.2 2.3 2.4 2.5 2.6 1.2
Frequency (GHz) Frequency (GHz)
b Substrate f g
UIML (simulation) UIML (exp.) 0.8
UIML UIML
Unit cell
80 70
142.9° 0.4
Material 70 60

Angle of incidence

Angle of incidence
3.0 mm 60
HDPE 2.0 mm
50 0
: 17.6 mm 50
40 40
Styrofoam 154.5° 30
: 30
2.0 mm 20
20
x 8.0 mm 10
0 10
z 17.6 mm 2.2 2.3 2.4 2.5 2.6 2.2 2.3 2.4 2.5 2.6
10 mm 30 mm Frequency (GHz) Frequency (GHz)
15 mm
h 5 i 20 Substrate (exp.)
c Substrate (simulation)
UIML (exp.)
4 Substrate (exp.) UIML (simulation)
15

Reflectance (%)
Reflectance (%)

Substrate (simulation) QAR (simulation)

UIML (exp.)
3 UIML (simulation)
QAR (simulation) 10
2
5
1

0 0
2.2 2.3 2.4 2.5 2.6 0 20 40 60 80
Frequency (GHz) Angle of incidence

Fig. 4 | Microwave measurement of reflectance using a metamaterial UIML. a, Schematic of the microwave measurement set-up. Inside the parallel plate
waveguide, two patch antennas, which act as a source and a detector, are located at various positions to measure reflection spectra for different incidence
angles. b, Optimized structure of the double-layer metamaterial UIML. A 10-mm-thick slab is used as a substrate and the UIML is attached symmetrically
to both sides of the slab to remove reflection. c, Images of the measurement unit protected by microwave absorbers and the metamaterial UIML sample.
d, Reflection spectra according to incidence angle with the QAR layer applied to the substrate. e, Measured reflection spectra of the substrate. f, Simulated
reflection spectra for the metamaterial UIML. g, Observed reflection spectra for the metamaterial UIML. h, Comparison of reflectance at a 45° incidence
angle with and without the UIML. i, Comparison of reflectance at 2.4 GHz for various incidence angles with and without the UIML.

produced a reflectance of over 10% at normal incidence, with reflec-
tion disappearing at the Brewster angle (Fig. 4e,i), which is in good
agreement with theoretical predictions. The QAR layer applied to
the substrate removed reflection for a normally incident wave, but
reflectance rapidly increased for incidence angles larger than 20°
and ceased to work as an anti-reflection layer. On the other hand,
Fig. 4f–i shows that UIML removes reflection almost completely for
an extremely wide range of incidence angles of up to 70° for the
entire spectral range of 2.2 GHz to 2.6 GHz. The measured reflec-
tance spectra in Fig. 4g are in good quantitative and qualitative
agreement with theoretical predictions, confirming the validity of
a metamaterial UIML. To check that the vanishing reflection with
the aid of UIML is indeed accompanied by complete transmission
rather than absorption, we conducted a microwave transmission
measurement. The measured transmittance reached almost 100%,
showing that UIML does indeed induce almost perfect transmission
(see Supplementary Section 5 for details).
We have established a theory of universal impedance match-
ing, and have presented a universal impedance matching layer that
enables the perfect transmission of white light by removing reflec-
tion in a lossless manner. The feasibility of a universal impedance
matching layer using metamaterials has been demonstrated via
experimental confirmation using linearly polarized microwaves.
Despite our confirmation with microwaves, it is still a challenge
to show perfect transmission of white light using metamaterials.
Because metamaterials use effective material parameters for struc-
turally and compositionally inhomogeneous materials, practically
realizing a UIML incorporating them limits the optimization pro-
cess to their constituent materials and patterns. Future research
needs to confirm that UIMLs based on our theory can be physically
created in practice.
Data availability. The data that support the plots within this paper
and other findings of this study are available from the correspond-
ing author upon reasonable request.
Received: 6 September 2017; Accepted: 10 January 2018;
Published online: 26 February 2018

148 Nature Photonics | VOL 12 | MARCH 2018 | 143–149 | www.nature.com/naturephotonics

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe PHoTonics
References
1. Born, M. & Wolf, E. Principles of Optics (Cambridge Univ. Press,
Cambridge, 2003).
2. Brewster, D. On the laws which regulate the polarisation of light by reflection
from transparent bodies. Phil. Trans. R. Soc. Lond. 105, 125–159 (1815).
3. Rayleigh, L. On the intensity of light reflected from certain surfaces at nearly
perpendicular incidence. Proc. R. Soc. Lond. 41, 275–294 (1886).
4. Macleod, H. A. Thin-film Optical Filters (McGraw-Hill, New York, 1989).
5. Raut, H. K., Ganesh, V. A., Nair, A. S. & Ramakrishna, S. Anti-reflective
coatings: a critical, in-depth review. Energ. Environ. Sci. 4, 3779–3804 (2011).
6. Clapham, P. B. & Hutley, M. C. Reduction of lens reflection by the ‘moth eye’
principle. Nature 244, 281–282 (1973).
7. Huang, Y. F. et al. Improved broadband and quasi-omnidirectional
anti-reflection properties with biomimetic silicon nanostructures.
Nat. Nanotech. 2, 770–774 (2007).
8. Spinelli, P., Verschuuren, M. A. & Polman, A. Broadband omnidirectional
antireflection coating based on subwavelength surface Mie resonator.
Nat. Commun. 3, 692 (2012).
9. Berenger, J. P. A perfectly matched layer for the absorption of electromagnetic
waves. J. Comput. Phys. 114, 185–200 (1994).
10. Pendry, J. B., Schurig, D. & Smith, D. R. Controlling electromagnetic fields.
Science 312, 1780–1782 (2006).
11. Leonhardt, U. Optical conformal mapping. Science 312, 1777–1780 (2006).
12. Kay, I. & Moses, H. E. Reflectionless transmission through dielectrics and
scattering potentials. Appl. Phys. 27, 1503–1508 (1956).
13. Ablowitz, M. J. & Clarkson, P. A. Solitons, Nonlinear Evolution Equations and
Inverse Scattering (Cambridge Univ. Press, Cambridge, 1992).
14. Horsley, S. A. R., Artoni, M. & La Rocca, G. C. Spatial Kramers–Kronig
relations and the reflection of waves. Nat. Photon. 9, 436–439 (2015).
15. Taflove, A. & Hagness, S. C. Computational Electrodynamics: The Finite-
Difference Time-Domain Method (Artech House, London, 2005).
16. Halevi, P. & Fuchs, R. Generalised additional boundary condition for
non-local dielectrics. J. Phys. C 17, 3869–3888 (1984).
17. Churchill, R. J. & Philbin, T. G. Electromagnetic reflection, transmission and
energy density at boundaries of nonlocal media. Phys. Rev. B 94, 235422 (2016).
18. Kim, K. H. & Park, Q. H. Perfect anti-reflection from first principles. Sci. Rep.
3, 1062 (2013).
Nature Photonics | VOL 12 | MARCH 2018 | 143–149 | www.nature.com/naturephotonics
© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letters
19. Landau, L. D., Lifshitz, E. M. & Pitaevskii, L. P. Electrodynamics of Continuous
Media (Pergamon, Oxford, 1960).
20. Yeh, P. Optical Waves in Layered Media (Wiley, Hoboken, NJ, 2005).
21. Belov, P. A. et al. Strong spatial dispersion in wire media in the very large
wavelength limit. Phys. Rev. B 67, 113103 (2003).
22. Chebykin, A. V., Orlov, A. A., Simovski, C. R., Kivshar, Y. S. & Belov, P. A.
Nonlocal effective parameters of multilayered metal–dielectric metamaterials.
Phys. Rev. B 86, 115420 (2012).
23. Nicolson, A. M. & Ross, G. F. Measurement of the intrinsic properties of
materials by time-domain techniques. IEEE Trans. Instrum. Meas. 19,
377–382 (1970).
24. Weir, W. B. Automatic measurement of complex dielectric constant and
permeability at microwave frequencies. Proc. IEEE 62, 33–36 (1974).
Acknowledgements
This work was supported by the Samsung Science and Technology Foundation under
project no. SSTF- BA1401-05. Q.-H.P. thanks Y. Kivshar and W. Choi for comments and
encouragement.
Author contributions
Q.-H.P. developed the theory with numerical tests and wrote the paper. K.I., J.-H.K. and
Q.-H.P. conducted the microwave experiments and analysed the data.
Competing interests
The authors declare no competing financial interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/
s41566-018-0098-3.
Reprints and permissions information is available at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to Q.-H.P.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
149