© 2021 Wiley-VCH GmbH
Supporting Information
for
Adv. Optical Mater.,
DOI: 10.1002/adom.202100230
Near-Infrared Active Metasurface for Dynamic
Polarization Conversion
Pin Chieh Wu
,* Ruzan Sokhoyan
, Ghazaleh Kafaie
Shirmanesh
,
Wen-Hui Cheng, and Harry A. Atwater
*
1
Near
-
Infrared Active Metasurface for Dynamic Polarization Conversion
Pin Chieh Wu
1,2
,
Ruzan Sokhoyan
2
, Ghazaleh Kafaie Shirmanesh
2
,
Wen
-
Hui Cheng
2
, and Harry A. Atwater
2,3
1
Department
of Photonics, National Cheng Kung University, Tainan 70101, Taiwan
2
Thomas J. Watson Laboratory of Applied Physics, California Institute of Technology, Pasadena, California
91125, USA
3
Kavli Nanoscience Institute, California Institute of Technology, Pasad
ena, California 91125, USA
Corresponding author:
pcwu@gs.ncku.edu.tw
;
haa@caltech.edu
2
1.
Active
reflectance
modulation
with
an
Al metasurface
Figure S1.
Simulated reflectance of the
(a)
co
-
and (b) cross
-
polarized light as a function of wavelength and
applied voltage. The incident light is
y
-
polarized.
The structural dimensions are listed in
the caption of
Fig.
2.
Figure S2
.
(a) Real part of
the dielectric permittivity of the 5 nm
-
thi
ck ITO
layer
Re[
ε
ITO
] as a function of
the spatial coordinate
and
applied voltage.
In the considered Al/HAOL/ITO heterostructure, the spatial
coordinate corresponds to the
distance from the
Al
electrode
so that t
he spatial coordinate of 5 nm corresponds
to
the
ITO/HAOL interface. In (a), the wavelength is fixed
to
1580 nm
, the applied voltage is negative,
indicating that (a) corresponds to the case of the reverse bias.
(b) Simulated reflectance of the co
-
and (c)
cross
-
polarized light as a function of
wavelength and applied voltage
in the case of the reverse bias
.
The
incident light is
y
-
polarized.
The structural dimensions are listed in the caption of Fig. 2.
3
2.
Tunable polarization conversion
with an Al metasurface
at different wavelengths
Figure S
3
.
The voltage
-
dependent path of reflected light
’
s polarization state on the Poincar
é
sphere
at
wavelengths of
(left) 1456 nm and (right) 1632 nm
.
The incident light is
y
-
polarized.
The structural
dimensions are listed in the caption of Fig. 2.
The orange dot indicates the polarization state of
the
incident
light.
3.
T
unable polarization with Au metasurface
s
We developed an alternative design of the dynamic polarization converter based on
a
gate
-
tunable reflectarray metasurface
with
Au patch na
noa
n
tennas
and an Au back reflector. Note that
in the main text, we use Al nanoantennas and Al back reflector rather than Au
.
Our Au
-
based
design
consists of a 150 nm
-
thick Au back reflector, a 9.5 nm
-
thick HAOL gate dielectric, followed by a
5
nm
-
thick ITO layer, and an array of Au patch
nano
antennas. The patch
nano
antennas are 210 nm
-
long a
nd 180 nm
-
wide (
see
Figure
S4a
). In case of
the
Au
-
based polarization converter, the
nano
antenna
s’
dimensions are slightly smaller as compared
with
the case of
the
Al
-
based
polarization converter. The assumed thickness of the Au patch antennas is 140 nm. The period of
the optimized metasurface is 350 nm for both
x
-
and
y
-
directions.
The dielectric permittivity of ITO
is described using the Drude model with consta
nts
휀
∞
= 4.2345,
훤
= 1.7588×10
14
(rad
/s
),
N
=
2.8×10
20
(cm
-
3
),
푚
∗
= 0.2525
m
e
[1]
.
For the device physics simulation, the work function of Au is set
to
5.1
eV.
Figures S4b and S4c
show the s
imulated
reflectance
spectra
of
co
-
and cross
-
polarized
light
, respectively.
We observed significant
reflectance
modulation
at
telecommunication
wavelengths
for both co
-
and cross
-
polarization components when
the
electrical bias is applied.
We
observe strong reflectance
modulation
when t
he bias voltage is varied between 2 V and 4 V.
The
4
observed active reflectance modulation
indicates that we should be able to observe an active
polarization control in the voltage range between 2 V and 4 V.
Figure S4.
(a)
Schematic
of
the tunable
metasurface based on Au nanoantenna
s and Au back reflector
. The
metasurface consists of a 150
nm
-
thick Au back reflector, a 9.5
nm
-
thick HAOL
layer
, a 5
nm
-
thick ITO
layer
, and an array of
140 nm
-
thick
Au
patch antennas
. The
patch antenna
dimensions are de
fined as
L
= 210
nm,
w
= 180 nm, and
θ
= 45°. The periodicity is
p
= 350 nm.
Simulated reflectance of the
(b)
co
-
and (c)
cross
-
polarized light as a function of wavelength and applied voltage. The incident light is
y
-
polarized.
The phase difference betwee
n
the
two orthogonal
polarization
states (
refer to
as
x
-
and
y
-
components
in this work
)
plays a key role in
the
polarization control.
Figure
S
5a shows the phase
difference
between
x
-
and
y
-
components
∆
φ
=
φ
yy
–
φ
xy
as a function of
the
wavelength
of the
incident light
and applied bias voltage.
The white curve
shows
the condition
when
the
x
-
and
y
-
components
of
the
reflected light have equal amplitude
s
.
According to Fig
ure
S5a, a linear
-
to
-
circular polarization conversion can be realized at a wavelength of 1520 nm. Thus,
in what follows,
we will fix the wavelength to 1520 nm
.
Figure
S
5b
plots the reflectance and
the relative
phase shift
of
x
-
and
y
-
polarized components
of
t
he reflected light
(
∆
φ
=
φ
yy
–
φ
xy
)
as a function of voltage
at
a
wavelength of
λ
= 1520 nm. When the applied bias is about 2 V, a
y
-
to
-
x
polarization conversion is
realized because the intensity of the
y
-
polarized light is negligible in this voltage range. For the bias
voltages of
~3
V, the designed Au metasurface converts the incoming linearly polarized light into a
circularly polarized light because of the 90° phase difference and equal reflectance of th
e
x
-
and
y
-
polarized light. For the other voltages, our metasurface realizes a linear
-
to
-
elliptical polarization
conversion.
Figure S5c
plots
Stokes parameters
as function of wavelength and applied voltage
.
Similar to what we observe
d
in the Al
-
based tunab
le polarization converter, the
S
1
exhibits a
5
significant change when the applied electrical bias
crosses
a
particular
value
(~2.5V in this case).
This is consistent with the result of
the
amplitude modulation in which a significant change
can be
observed when the electrical bias
crosses the voltage value of
~2.5 V.
The
parameter
S
3
also show
s
a dramatic change
and reaches about +1 at this voltage value
, revealing the
possibility o
f
the
circular
polarization
conversion
.
As compared to the Al
-
based
metasurface, this Au metasurface requires
much lower gate voltage to achieve linear
-
to
-
circular polarization conversion that significantly
facilitates the experimental demonstration. Next, we plan to experimentally demonstrate this Au
-
based polarization co
nverter.
Figure S5.
(a)
Phase difference between
x
-
and
y
-
components as a function of wavelength and applied bias.
The white solid line marks the parameter values that yield equal reflectance values for
x
-
and
y
-
polarized
components of light.
(b
)
Simulated reflectance (orange line: cross
-
polarized light; olive line: co
-
polarized
light) and phase difference (phase of the cross
-
polarized reflected wave minus the phase of the co
-
polarized
reflected wave) as a function of applied bias. The metasurface
is illuminated by a
y
-
polarized normally
incident light.
The cyan and magenta shadow
ed regions
indicate the
voltage
range where desired polarization
states are obtained.
The operation wavelength is 1520 nm.
(c)
Plot of the Stokes parameters versus applied
electrical bias
.
To experimentally evaluate the performance of
the
proposed metasurface polarization
converter, we have fabricated
the described Au metasurface
-
based active polarization converter
.
6
For sample preparation,
we first patte
rned some alignment markers using
electron beam
lithography
[VISTEC electron beam pattern generator (EBPG) 5000+] at an
acceleration voltage of 100 keV
and
with
a
current of 50 nA
. Then
the bottom contact
(150
-
nm
-
thick Au with an adhesion layer of
5
-
nm
-
thick Ti)
is fabricated.
Subsequently, a gate dielectric HAOL is deposited through a shadow
mask by using a thermal recipe in the ALD tool (Fiji G2 Plasma Enhanced Atomic Layer
Deposition System) at 150°C us
ing the recipe described in Ref
[2]
.
The thickness of the HAOL
film was measured by transmission electron microscopy (TEM). Figure S6 shows the TEM image
of the HAOL film deposited on a Si substrate. According to the TEM image, the thickness of the
de
posited HAOL layer is very close to what we numerically designed. Once the area of HAOL is
determined, the ITO layer is subsequently defined by electron beam lithography with alignment
markers and deposited using room
-
temperature RF magnetron sputtering in
an Ar/O
2
plasma
environment. The deposition pressure and the applied RF power were set to 3 mTorr and 48 W,
respectively. The plasma was struck by using argon (Ar) gas with a flow rate of 20 sccm. A mixture
of argon and oxygen (O
2
) gases (Ar/O
2
: 90/10) wa
s used to control the amount of oxygen deficiency,
and hence, the charge carrier concentration of the ITO layer. We characterized the ITO layer by a
combination of Hall measurements and spectroscopic ellipsometry on the ITO films deposited on
quartz and Si
substrates, respectivel
y
[2]
. The fitted thickness of the ITO layer obtained through the
ellipsometry measurement was 4.8 nm.
Finally, the metasurface block
along with the top contact
s
are patterned with
the fourth
electron beam
writing process
at an
acceleration voltage of 100 keV
with a
beam
current of 300 pA
.
Figure S6.
TEM image of the HAOL control sample deposited on
a
Si substrate
via
ALD
.
7
Figures
S
7
a
and S
7
b
show the schematic of the proposed device
and
the SEM image
of the
fabricated Au metasurface
, respectively
.
To optically characterize the polarization state of reflected
light, a quarter
-
wave
plate
paired with
a linear polarizer is
inserted
in front of the spectrometer
, as
shown in Figure S
7
c
.
A visible camera is utilized to visualize the position where the spectrometer
will
m
easure
.
Figure
S
7
d
shows the
measured
reflectance
as a function of wavelength and voltage
for six polarization states
when the incoming light is
y
-
polarized
.
As expected,
we observed a
resonant dip in the LP
-
V spectrum
associated with a polarization conversion peak in the LP
-
H
spectrum
w
hen the electrical bias is absent
.
When the electrical bias is increased
from 0 V to +10
V
, w
e can see that
the intensity of
the
cross
-
polarized light
slightly increased
at a wavelength of
~1.5 μm
, then saturate
d
at ~60% when the applied bias is greater than
a
certain level
.
Interestingly,
the
reflectance
in the LP
-
H spectrum
shows a significant
decrease when the bias voltage
is
changed
f
rom 0 V to
-
4 V.
Similar intensity variation
can be observed
for
the
case
s
of
RCP, and
LP+45
°
.
We also
found
that
the resonant features in RCP and LP+45
°
spectra become narrower
when
the
electrical bias
region varied from negative to positive
, while the
LP
-
H spectrum becomes broader
.
Contrarily,
for the cases of LP
-
V, LCP, and LP
-
45
°
,
the reflectance
spectrum
shows
slight
changes
when bias voltage
is
varied
. Based on these results, we
conclude
that
when illuminated by the
linearly polarized light
, the
metasurface can dynamically control the polarization state of the
reflected light
.
Figure S
7
.
(a)
Schematic illustration
of
the
tunable polarization converter sample based on Au metasurface
.
(b) Scanning electron
microscopy
image of the fabricated sample.
The
beam splitter
with dashed
-
boundary
is
moved back and forth for either visualizing the sample surface or measuring spectral signal.
(c)
Schematic
for the measurement optical setup. M: mirror, L: lens, BS: beam splitter, λ
/4: quarter
-
wave plate, P: linear
polarizer, O: objective. (d)
Experimental
spectra
for different polarization states of reflected beam
under
different applied biases.
The legend of LP
-
H, LP
-
V, RCP, LCP, LP+45
°
, and LP
-
45
°
correspond to
x
-
8
polarization,
y
-
p
olarization, right
-
hand circular polarization, left
-
hand circular polarization, linear
polarization along +45
°
, and linear polarization along
-
45
°
, respectively
.
To c
learly distinguish the
generated
polarization state
of
the
reflected light
, we
calculate
the
Stokes parameters
using the results
from
Figure S
7
d
.
Figure S
8
shows the measured Stokes
parameters
. The parameters
S
1
,
S
2
, and
S
3
are normalized to
S
0
so that the parameter values
vary
between
-
1 and +1
.
We
experimentally
observe that both
S
2
and
S
3
are electrically modulated
at a
wavelength of ~
1.5 μm
.
Although
the
measurement
results show
that
the
fabricated metasurface
can
electrical
ly
modulate the
Stokes parameters, the
intensity variations are
less
significant
as
compared
with
our theoretical
prediction
s
(cf.
Figure S5c
)
.
The
general trends observed
in
measurements
are quite
different from
those
obtaine
d in simulation
s
. Besides, we
haven’t been able
to experimentally identify the
situation
where
S
3
is unity and
S
1
=
S
2
= 0,
which implies that we
haven’t been able to experimentally achi
e
ve
liner
-
to
-
circular polarization conversion.
T
his
difference
actually arose from the mismatch between the ITO
Drude
parameters used in simulation
s
and
the ones
in the
real
sample
.
Figure S
8
.
Experimental
ly measured
Stokes parameters versus incident wavelength and applied electrical
bias.
Based on our recent work
s
in which we
have
experimentally characterize
d
the optical
properties of
the
ITO film
[2,
3]
, the
parameters of
the
deposited
I
TO
can are given by the following
Drude parameters:
휀
∞
= 3.9,
훤
= 1.8×10
14
(rad/s
),
N
= 3×10
20
(cm
-
3
),
푚
∗
= 0.35
m
e
. But the Drude
parameters used in the design of the tunable polarization converter
are
based on the values obtained
in
our
earlier
work
[1]
.
Next, we
assess how the
variation of the Drude parameters
affects the optical
performance of our tunable polarizer. We fix the geometrical
parameters of our metasurface to
those defined in Figure S4a and change the Drude parameters of the ITO layer, adopting the values
reported in our recent work
[3]
.
Figure S
9
plots
the
simulated
Stokes parameters
. When comparing
9
Fig. S
8
and
Fig. S
9
we observe that our experimental and simulation results qualitatively match.
Please notice that
here
we also revise our device physics simulations according to the
modified
Drude parameters. For example, the effective electron mass of ITO is now taken as
푚
∗
= 0.35
m
e
.
Similarly, we can see all three
Stokes
parameters exhibit much w
eaker modulation
as compared
with
simulation
results shown in Figure S5c
,
indicating that
the accurate determination of the Drude
parameters of
ITO
is crucial for obtaining the optimized performance of the
tunable polarization
converter.
Figure S
9
.
Simulated Stokes parameters
as a function of wavelength
applied electrical bias. The structural
parameters of Au metasurface are from Fig. S4a, and the Drude model parameters of ITO are from Ref.
[3]
.
Next,
we assume that the Drude parameters of the ITO layer are given in Ref.
[
2
, 3]
and
optimize the structural dimensions of the Au metasurface
-
based
tunable polarization converter
.
The
geometrical motif
of the metasurface is
the same as
the one shown in Figure
S4a.
Figure S
10
shows
the simulated results after structural optimization.
T
he metasurface consists of a 150
-
nm
-
thick Au
back reflector, a
15
-
nm
-
thick HAOL, a 5
-
nm
-
thick ITO, and an array of
100 nm
-
thick
Au
patch
antennas
. The dimensions
of the metasurface unit cell
are
given as
L
= 210 nm,
w
= 180 nm,
θ
=
45°
, and
p
= 350 nm.
First, we
evaluate the
reflectance
modulation
capability
of the designed
metasurface
before characterizing its polarization conversion performance
.
As
shown in Figure
s
S
10
a
and S
10
b
, w
hen the electric bias is 0 V, a strong linear cross polarization conversion
effect
can be observed around the telecom wavelength
s
.
The highest cross polarization conversion
efficiency is ~15%, which is higher than
what we observed in
the previous design.
W
hen the
electrical bias is increased
up
to ~4 V
,
the
conversion peak
firstly
blu
e
-
shifts and
then re
d
-
shifts
when
the bias voltage is further increased.
Overall, the
spec
tral features are very similar to the
results shown in Figures S4 and S5,
indicating
that we are able to control the polarization of the
reflected light when applying electrical bias.
10
Figure S
10
.
Simulated reflectance of the
(a)
co
-
and (
b
) cross
-
polarized light as a function of wavelength and
applied voltage. The incident light is
y
-
p
olarized.
The metasurface consists of a 150
-
nm
-
thick Au back
reflector, a 15
-
nm
-
thick HAOL
layer
, a 5
-
nm
-
thick ITO
layer
, and an array of
100 nm
-
thick
Au
patch
antennas
. The
geometrical parameters of our unit cell are given
as:
L
= 210 nm,
w
= 180 nm, and
θ
= 45°.
The
metasurface
period is
p
= 350 nm.
(c) Phase difference between
x
-
and
y
-
components as a function of
wavelength and applied bias. The white solid line marks the parameter values that yield equal reflectance
values for
x
-
and
y
-
polarized compone
nts of light. (d)
Simulated reflectance (red line: cross
-
polarized light;
olive line: co
-
polarized light) and phase difference (phase of the cross
-
polarized reflected wave minus the
phase of the co
-
polarized reflected wave) as a function of applied bias. T
he metasurface is illuminated by a
y
-
polarized normally incident light. The cyan and orange shadows indicate the range of voltage
values
where
desired polarization states are obtained. The
operating
wavelength is 1520 nm. (e) Stokes parameters
as a
functio
n of wavelength and
applied electrical bias.
11
To identify the optimal wavelength for linear
-
to
-
circular polarization conversion, we
investigate the phase difference between
x
-
and
y
-
polarized components of the reflected light. In
Fig. S
10
c,
the solid line labels the condition where
the
x
-
and
y
-
polarized components yield equal
reflectance (see Figure S
10
c). Our analysis shows that we can obtain three polarization states at a
wavelength of 1520 nm. The active polarization conversion performan
ce is summarized in Figure
S9d. Again, three polarization states are successfully realized when the electrical bias is applied in
an appropriate voltage range:
0 V ≤
V
a
≤ 2.8 V
–
linear to cross
-
polarized linear polarization conversion,
9.2 V ≤
V
a
≤ 11.1 V
–
linear
-
to
-
circular polarization conversion,
Other
voltage values
–
linear
-
to
-
elliptical polarization conversion,
where
V
a
is the applied electrical bias. The simulated Stokes parameters exhibit very similar
trends
to the previously discussed cases (Figure 5b in the main article and Figure S5c):
S
1
and
S
3
vary significantly when the applied bias goes across ~5 V. When the condition
S
1
=
S
2
= 0 (
S
2
=
S
3
= 0) and
S
3
= 1 (
S
1
= 1) holds linear
-
to
-
circular (
y
-
to
-
x
cross) polarization conversion occurs.
Moreover,
we found that structural dimensions of our redesigned metasurface are quite close to
our orginal Au
-
based metasurface design. When comparing our original Au metasurface
-
based
design and the redesigned polari
zation converter, we observe that only the thicknesses of the Au
patch antenna and the HAOL layer are slightly changed if the ITO parameters shifted, indicating
that the proposed design motif is quite robust.
4.
Linear cross
-
polarization conversion for diffe
rent thicknesses of Al
nanoantennas
Figure S
1
1
.
Simulated co
-
polarized (blue lines) and cross
-
polarized (red lines) reflectance spectra of the
tunable metasurface for different thicknesses of Al nanoantennas. The co
-
polarized reflectance shows a non
-
zero intensity at a wavelength of 1580 nm when the thi
ckness of Al nanoantennas is less than 80 nm. Thus,
to realize a complete conversion to a linear cross
-
polarized state and to minimize the fabrication difficulty
12
d
uring
the lift
-
off process , we choose the Al nanoantenna thickness as 80 nm. The co
-
polariz
ed and cross
-
polarized reflectance spectra for metasurfaces with 80
-
nm
-
think Al nanoantennas can be found in Fig. 3a.
5.
Linear cross
-
polarization conversion performance for different polarization
angle of incident light
As discussed in the main article, t
he polarization conversion is achieved by simultaneously
exciting two gap plasmon modes that are supported by the long side and short side of the Al
nanoantenna, respectively. Thus, a nonideal incident polarization angle can influecne the
polarization conv
ersion performance. In this section, we briefly discuss the effect of the
misalignment between the incident polarization angle and the Al nanoantenna orientation on the
polarization conversion performance. For simplicity, the applied bias is set to 0 V. Fi
gure S12a
shows the simulated x
-
component (blue dots) and y
-
component (navy blue dots) reflectance spectra
as a function of polarization angle
θ
pol
. At no applied bias
w
e expect to
obtain an x
-
polarized
reflected light of high purity. Therefore, when ananl
ysing the the polarization characteristics of the
refelcted light, we expect that the intensity of the x
-
component has to be significantly higher than
the intensity of the y
-
component. To quantitatively analyze the linear cross
-
polarization conversion
perf
ormance, we utilize a figure of merit named the intensity ratio, which is defined as a ratio of
the reflectance intensity of the y
-
component and the reflectance intensity of the x
-
component. As
seen in Fig. S12a (red circles), the intensity ratio acqures r
elatively low values when the polarization
angle is around 0°. We observe that the intensity ratio is less than 0.05 when the polarization angle
is varied between 5° and 10°, indicating that a highly pure x
-
polarized light is obtained in reflection.
Th
e
intensity ratio increases very significantly when the polarization angle is far away from
0°. The
intensity ratio even reaches unity when the polariztion angle equals to 45° or
-
45°. This is because
only one single plasmon mode is excited under such condit
ions, leading to a linearly
-
polarized
reflected light along either 45° or
-
45°. To further evaluate the polarization conversion performance,
we plot the polarization angle
-
dependent path of the polarization states on the Poincar
é
sphere. As
expected, a ne
arly ideal x
-
polarized light can be realized when the polarization angle is close to 0°.
The polarization state of the reflected light deviates from the target one when the polarization angle
increases. As seein in Fig. S12(b), the reflected light becomes
linearly polarized light along 45° and
-
45° when the polarization angle is 45° or
-
45°, respectively. In summary, to avoid a siginificant
deviation of the resulting polarization state of the reflected light from the target one, the polariztion
angle of inc
ident light has to be within <5°.
13
Figure S12.
(a)
Simulated reflectance of the x
-
component (blue dots) and
y
-
component (navy blue dots) as
a function of incident polarization angle
θ
pol
for the Al tunable metasurface at 0 V. The red circles show the
ratio of the reflectance of the y
-
polarized and x
-
polarized light. The inset shows the orientation of the Al
patch nanoantenna with respect to the
y
-
axis. The incident polarization is aligned
to the long axis(short axis)
of the Al nanoantenna when the polarization angle
θ
pol
= 45° (
-
45°). (b) The polarization angle
-
dependent
path of the polarization state of the reflected light on the Poincar
é
sphere. Incident wavelength: 1580 nm.
Reference
[1]
Y.
-
W. Huang, H. W. H. Lee, R. Sokhoyan, R. A. Pala, K. Thyagarajan, S. Han, D. P. Tsai, H. A. Atwater,
Nano Lett
.
2016
, 16, 5319.
[2]
G. Kafaie Shirmanesh, R. Sokhoyan, R. A. Pala, H. A. Atwater,
Nano Lett
.
2018
, 18, 2957.
[3]
G. K.
Shirmanesh, R. Sokhoyan, P. C. Wu, H. A. Atwater,
ACS Nano
2020
, 14, 6912.