1
Supporting Information
GateTunable Conducting Oxide Metasurfaces
YaoWei Huang
1,2,3,†
, Ho Wai Howard Lee
1,2,†,+
, Ruzan Sokhoyan
1
, Ragip A. Pala
1,2
,
Krishnan Thyagarajan
1,2
, Seunghoon Han
1,4
, Din Ping Tsai
3,5
, and Harry A. Atwater
1,2,*
1
Thomas J. Watson Laboratories of Applied Physics, C
alifornia Institute of Technology,
Pasadena, California 91125, United States
2
Kavli Nanoscience Institute, California Institute o
f Technology, Pasadena, California 91125,
United States
3
Department of Physics, National Taiwan University,
Taipei 10617, Taiwan
4
Samsung Advanced Institute of Technology, Samsung E
lectronics, Suwonsi, Gyeonggido
443803, Republic of Korea
5
Research Center for Applied Sciences, Academia Sini
ca, Taipei 11529, Taiwan
* Corresponding author. Email: haa@caltech.edu
† These authors contributed equally to this work.
+ Current address: Department of Physics, Baylor Un
iversity, Waco, Texas 76798, United States;
The Institute for Quantum Science and Engineering,
Texas A&M University, College Station, Texas
77843, United States
2
1. Optical Properties of ITO in Experiment
To characterize ITO, we perform ellipsometric and H
all measurements. These measurements
confirm that the dielectric permittivity of ITO can
be well described by the Drude model
ε
ITO
=
ε
∞
ω
2
p
/(
ω
2
+
iωΓ
), where
ω
p
is the plasma frequency which is related to the ca
rrier density
N
and
electron effective mass
m
*
as
ω
p
2
=
Ne
2
/(
ε
0
m
*
). Here
ε
0
is the dielectric permittivity of vacuum,
e
is the electron charge, and
Γ
is the damping constant. Using literature values f
or the constants
(
Γ
=1.8×10
14
radHz,
m
*
= 0.35
m
e
, and
ε
∞
= 3.9),
14
we relate the charge distribution in ITO to the
dielectric permittivity of ITO
ε
ITO
. The fact that
ε
ITO
can be related to the carrier concentration in
ITO via the Drude model explains the key physical m
echanism behind metasurface phase and
amplitude modulation.
The ITO films in our experiments are deposited by R
F magnetron sputtering, where the
carrier concentration of ITO is varied by controlli
ng the ratio of the Ar and O
2
flow rates during
deposition.
1,5
We have two valves of gas to the chamber of sputte
r. One is Ar flow valve that we
keep with a flow of 20 sccm. The other is the Ar+O
2
valve with the ratio of Ar and O
2
is 3:1. The
optical properties of the ITO are derived from Hall
and ellipsometry measurements.
6
We
fabricate ITO films with background carrier concent
rations ranging from
N
o
= 8×10
19
cm
3
to
7×10
20
cm
3
. To obtain ITO with predesigned carrier concentra
tions (these carrier concentrations
should yield Re(ε
ITO
) = 1.52.5 at the wavelength of interest), we vary
by a small range, the
Ar+O
2
flow from 0.4 sccm to 1.6 sccm while depositing 23
nm ITO film on several different
quartz substrates (for Hall measurement) and silico
n substrates (for ellipsometry measurement).
The carrier concentration
N
and electron mobility
1
(thus the resistivity
ρ
) of the ITO films are
measured using the Hall method. The complex dielect
ric permittivity of the ITO films in the near
infrared region was measured by ellipsometry, and t
he raw data for the complex permittivity was
fitted by a single Drude function, using the carrie
r concentration and mobility
obtained from Hall
measurements. The high frequency permittivity (
ε
∞
) and damping rate (
τ
1
) are determined via
fitting the Drude model to the measured data. Altho
ugh, for ITO in the ultraviolet to visible
region, the contribution of Lorentzian oscillators
cannot be ignored in fitting the ellipsometry
data, the effect of these oscillators is smaller in
the near infrared and can be absorbed in the
constant
ε
∞
. The real and imaginary part of the dielectric per
mittivity for different samples
3
deposited with different Ar+O
2
flows are shown in Fig. S1. As can be seen in Fig.
S1, for the
assumed experimental conditions, the real part of t
he dielectric permittivity of ITO is between
1.5 and 3 at the target wavelength of 1550 nm. This
indicates that for experimentally achieved
ITO doping levels, one can ensure via gating that t
he epsilonnearzero (ENZ) condition occurs
in the ITO accumulation layer. In addition, the sim
ulated permittivity (black dashed curve)
obtained from a standard Drude model with commonly
used constants indicates agreement both
with the slope of the dispersion and the value of t
he permittivity for the ITO deposited with a 1.2
sccm Ar+O
2
flow. Using these fitting parameters and the follo
wing relations:
*
2
2
2
2
2
*
0
1
,
( )
,
, 1/ ,
e
p
ITO
p
e
m m
q N Nq
Nq
i
m m
ρ
τ
ω
ε ω ε
ω
τ
ω ω
ε
∞
=
=
= −
=
Γ =
+ Γ
we find that the electron effective mass (
m
*
) takes values ranging from 0.3
m
e
to 0.23
m
e
,
depending on fabrication conditions which is consis
tent with previously reported results.
7
The
optical constants obtained from Hall and ellipsomet
ry measurements are listed in Table S1.
Figure S1.
Solid lines represent measured real and imaginary
parts of dielectric permittivity of
ITO for different Ar+O
2
flow during the deposition. The dashed line corres
ponds to the dielectric
permittivity of ITO obtained by using literature va
lues for Drude parameters.
13
The constants
used in the figure are listed in the Table S1.
4
Table S1.
Electrical measurement results as well as optical
Drude parameters for ITO films
deposited with different Ar+O
2
flows. The data is obtained from the Hall and elli
psometry
measurements. The literature values of ITO constant
s are also included in the table.
13
The ITO of our device was fabricated at Ar+O
2
flow rate of 1.2 sccm. As one can see from
Table S1, the carrier concentration of ITO in this
case is
N
0
= 2.3×10
20
cm
3
. It has been
previously observed that the carrier concentration
of ITO increases during ALD process due to
leakage of oxygen from ITO.
8
Since Hall measurements have been taken before ALD
process,
we expect the carrier concentration of ITO in our f
inal device to be higher than the measured
Hall value. To identify the ITO carrier concentrati
on in our final device, we have performed
extensive numerical simulations. Our numerical simu
lations show that the carrier concentration
of
N
0
= 2.8×10
20
cm
3
provides an excellent agreement between the measur
ed and calculated
positions of the reflectance dip. Hence, in what fo
llows we assume that the carrier concentration
of ITO is
N
0
= 2.8×10
20
cm
3
. On the other hand, the electron affinity of ITO d
etermines the
voltage that provides 180° phase shift of the refle
cted light. Figure S2 shows the phase shift as a
function of applied bias for different electron aff
inities of ITO (
χ
ITO
). When increasing the
electron affinity and, consequently, the work funct
ion of ITO, the carrier concentration of ITO at
the Al
2
O
3
/ITO interface increases. This results in reduction
of voltage necessary to achieve 180°
phase shift. We use electron affinity of ITO as a f
itting parameter to match the bias that provides
180° phase shift of the reflected light to the expe
rimentally observed value of 2.5 V. Our
simulations show a perfect match with experimentall
y measured values of the phase shift (Figure
3c, blue line with green spheres) when the electron
affinity of ITO is taken as
χ
ITO
= 5 eV.
5
Figure S2.
Simulated phase shift as a function of applied bias
for different assumed electron affinities of
ITO (
χ
ITO
). The operation wavelength is 1573 nm.
2. Optical Properties of ITO in Simulation
We performed electrostatic calculations to determin
e the electron distribution in ITO as a
function of applied bias. We model ITO as a semicon
ductor with the bandgap of
E
bg
= 2.8 eV,
9
electron affinity of
χ
= 5 eV
,
effective electron mass of
m*
= 0.25
m
e
. In our calculations DC
permittivity of ITO is chosen as 9.3.
10
Depending on fabrication conditions, electron mobi
lity of
ITO ranges from 16 cm
2
/Vs to 38 cm
2
/Vs as confirmed by Hall measurements. We assume th
at
the electron distribution in the ITO, in case of th
e metasurface geometry, can be approximated by
that in the case of a simple planar MOS structure c
onsisting of infinite Au/Al
2
O
3
/ITO layers (Fig.
2b in the main text). For numerical modeling we use
d a commercial software (Device
Lumerical Solutions, Inc.) that numerically solves
the Poisson and driftdiffusion equations. In
the software we define Al
2
O
3
as an insulator that implies that no charge flows
through it. This is
a valid assumption since our leakage currents are s
mall. First, we calculate capacitance of the
Au/Al
2
O
3
/ITO/Au planar structure assuming that the thicknes
s of the dielectric spacer
d
is 5 nm,
and DC permittivity of Al
2
O
3
is
ε
dc
= 9. In the device physics calculations, we used t
he mesh size
6
of 0.05 nm which has been validated by performing c
areful convergence tests. Figure S3a shows
capacitance of the Au/Al
2
O
3
/ITO/Au device as a function of applied voltage for
different
background carrier concentrations of ITO. As one ca
n see, the capacitance of the MOS capacitor
increases when increasing background carrier concen
tration of ITO. The dependence of the
capacitance on applied bias becomes weaker when ITO
is more doped. If we further increase
carrier concentration of the semiconductor in the M
OS structure, the screening length will
decrease and the capacitance of the device will asy
mptotically approach the capacitance of an
ideal parallel plate capacitor. The plates of the i
deal parallel plate capacitor are made of perfect
electric conductor that ensures that electric field
does not penetrate into them. Capacitance of the
ideal parallel plate capacitor is given by the foll
owing wellknown formula:
dc
ideal
0
,
A
C
d
ε
ε
=
where
A
is the area of the capacitor plate,
ε
dc
is the dc permittivity of the dielectric spacer,
ε
0
is
the vacuum permittivity and
d
is the thickness of the dielectric spacer (see Fig
. S3b). If thickness
of the Al
2
O
3
spacer is 5 nm then for the plate area
A
= 1 Om
2
,
C
ideal
= 15.93 fF/Om
2
. Evidently,
for a given capacitor plane area, the capacitance o
f the MOS capacitor is lower than the
capacitance of the ideal parallel plate capacitor.
Indeed, using Gauss’s law one can readily show
that the charge accumulated at the capacitor plates
is proportional to the voltage drop across the
dielectric spacer. Since in case of the MOS capacit
or there is also additional voltage drop inside
of the semiconductor, then for any value of the app
lied voltage
V
, there will be more charge
accumulated at the plates of the ideal parallel pla
te capacitor as compared to the case of the MOS
capacitor (see inset of Fig. S3b).
7
Figure S3.
(a) Capacitance of the MOS capacitor as a function
of applied bias for different
background doping of ITO. (b) Capacitance of the id
eal parallel plate capacitor as a function of
the thickness of Al
2
O
3
spacer. In both cases (a)(b) gate dielectric is 5
nm thick Al
2
O
3
. Inset of (b)
shows schematics of the parallel plate capacitor.
Figure S4 depicts carrier concentration
N
0
, in ITO in logscale as a function of distance
from the Al
2
O
3
/ITO interface at an applied bias of 2.5 V for expe
rimentally accessible values of
the background carrier concentration. As it is in t
he case of fabricated structures, the thickness of
Al
2
O
3
is 5 nm. As one can see, the spatial volume over wh
ich noticeable charge density
redistribution occurs increases with decreasing bac
kground carrier concentration. Based on the
calculated values of the carrier concentration as a
function of applied voltage and distance from
Al
2
O
3
/ITO interface, we calculate the dielectric permitt
ivity of ITO
ɛ
via Drude model. Figures
S5aS5b plot real and imaginary parts of the dielec
tric permittivity
ɛ
as a function of the distance
from Al
2
O
3
/ITO interface at the wavelengths of 1550 nm and fo
r applied bias voltages of 2.5 V
and 3.5 V. Each curve in Fig. S5 corresponds to the
marked background carrier concentration of
ITO. Apparently, for a given background carrier con
centration, one observes larger variation of
ɛ
at longer wavelengths. This is due to the fact that
the second term of the Drude model is
proportional to 1/(
ω
2
+i
ωΓ
).
8
Figure S4.
Log
10
of the carrier concentration in ITO for different
background doping levels (
N
0
= 9×10
19
, 3×10
20
, and 6×10
20
) as a function of distance from the Al
2
O
3
/ITO interface.
Figure S5.
Calculated dielectric permittivity of ITO as a func
tion of distance from the
Al
2
O
3
/ITO interface at 1550 nm. (a)(b) Real part of the
dielectric permittivity of ITO at applied
bias of (a) 2.5 V and (b) 3.5 V. (c)(d) Imaginary
part of the dielectric permittivity of ITO at
applied bias of (c) 2.5 V and (d) 3.5 V.
9
Our calculations show that for the ITO background c
arrier concentration of
N
0
= 3×10
20
cm
3
,
the real part of the dielectric permittivity in the
ITO accumulation layer changes its sign, going
from positive to negative, when a large enough bias
is applied. It is also shown that the carrier
concentration of the ITO at the Al
2
O
3
/ITO interface increases by five times with applied
bias
from 0 to 6 V (see Fig. 1b).
Based on this knowledge, we have developed general
design principles of gatetunable
conducting oxide metasurfaces. When designing a met
asurface that would ensure large phase
modulation of the incoming light we take the follow
ing steps: a) we choose the operation
wavelength; b) we identify the background carrier c
oncentration that will ensure ENZ transition
in the accumulation layer in ITO when applying volt
age; c) finalize the design of the metasurface
by identifying ITO thickness and dimension of the t
op electrode and antenna that would ensure
magnetic dipole resonance at the chosen wavelength.
3. Fabrication of tunable metasurfaces
The metasurfaces were fabricated via standard ebea
m lithography and multiple thinfilm
depositions. Schematic of the tunable metasurface i
s depicted in Fig. S6a where voltage is
applied between the antenna via the external gold p
ads and the gold back plane. The inset of Fig.
S6a shows a closeup look of the stripe antenna tha
t were designed and fabricated. Fig. S6b
shows the fabrication process of our gatetunable m
etasurface. A 5nmthick Ti film was
deposited, as an adhesion layer, by thermal evapora
tion on a quartz glass substrate, followed by
thermal evaporation of 80 nm of Au. Then 18 nm of I
TO was sputtered on the sample by using
RF magnetron sputtering in an oxygen/argon plasma (
ITO targets were purchased from
Plasmaterials Inc.). All depositions were done at r
oom temperature at a pressure of 3 mTorr at a
power of 48W, while the oxygen/argon flow rates wer
e varied to achieve different carrier
concentrations of the ITO (see ITO characterization
in Supporting Information part1). A 5 nm
thick Al
2
O
3
was then grown by atomic layer deposition (Fiji F2
00, Ultratech/Cambridge
NanoTech).
10
Figure S6.
(a) Schematic of the tunable metasurface where volt
age is applied between the
stripe antenna via the external gold pads and the b
ottom gold mirror. (b) The crosssection of the
fabricated metasurface after the liftoff process.
(c) Fabrication process of the gatetunable
metasurface.
Figure S6c shows a schematic of the ebeam lithogra
phy process. After the previous steps,
the sample was further coated with 300 nm of bilaye
r PMMA (MicroChem Corp., PMMA495K
A4 and PMMA950K A4). Subsequently, the stripe stru
ctures with 40 × 40 Om
2
area, gold
connections, and gold pads were patterned using an
ebeam lithography system (Leica Vistec
EBPG 5000+) at an acceleration voltage of 100 keV w
ith 100 pA current (for antenna structure)
and 50 nA current (for connections and pads). After
exposure and development, a 50 nm Au film
was deposited by ebeam evaporation. A combination
of bottom and top layer resists is selected
in such a way that there is a large difference in t
he dissolution rates of the layers at the
development step, leading to a reversed trapezoid r
esist profile. The samples were eventually
soaked in acetone in which the unpatterned regions
were lifted off. The SEM image of the
crosssection of the metasurface sample is depicted
in Fig. S6b. Note that the gold antennas were
11
designed to connect to two isolated left and right
pads so that two different voltages can be
applied on the pads for demonstrating a onedimensi
onal 2level phase grating with periodicity
of 2.4 Om (see Fig. 1).
4. Field analysis at 0 V
When performing electromagnetic calculations the as
sumed mesh size is 0.27 nm. This
mesh choice has been validated by performing carefu
l convergence tests. Our electromagnetic
calculations show presence of largemagnitude
x
component of the magnetic field |
H
x
| localized
in the gap region between gold antenna and gold bac
k plane (see Fig. S7c) that indicates that the
considered resonance is a magnetic plasmon resonanc
e. Figure S7a (which is the same as Fig. 2d
of the manuscript) shows that the z component of th
e electric field
E
z
around right edge of the
antenna (blue, at
y
= 125 nm,
z
= 10 nm) is antiparallel to
E
z
around left edge of the antenna (red,
at
y
= 125 nm,
z
= 10 nm). On the other hand, Fig. S7b shows that
at the bottom of the antenna
the
y
component of the electric field
E
y
is oriented in –y direction (light blue) whereas a
t the top
of the back plane
E
y
is oriented +ydirection (light red). The considere
d resonance supports an
antiparallel or curl electric field, resulting in a
curl current density shown in red cones (Fig. S7b)
.
This curl of the electric field and current density
is consistent with large x component of
magnetic field |
H
x
| shown in Fig. S7c.
Figure S7
. Spatial distribution of the
z
component of the electric field
E
z
(a),
y
component of the
electric field
E
y
with current density shown in red cone (b) and abs
olute value of the
x
component of the magnetic field |
H
x
| (c) for applied bias of 0V.
12
5. Reflectance measurements
The experimental setup used to measure the metasurf
ace reflectance spectrum is illustrated in Fig.
S8. An uncollimated light from halogen lamp is col
limated by using lens and directed through a
polarizer and 50/50 nonpolarizing beam splitter. T
he incident light is then focused onto the
metasurface sample by using a 40
×
, 0.75 NA microscope objective (see inset).
The light reflected from the metasurface is diverte
d by a beamsplitter and sent to nearIR
camera or an optical spectrometer through a multim
ode fiber. Note that the position of the
metasurface is preadjusted by imaging with the cam
era so that only the light reflected from the
center of the metasurface (~ 20
×
20 Om
2
) is collected by the spectrometer. The measured
reflectance is normalized relative to reflectance i
n the area without metasurface structure
(reflected directly from the gold back plane reflec
tor just next to the metasurface).
Figure S8.
Optical setup for measuring the reflectance spectra
of the metasurfaces. The
polarization of the incident light is aligned to th
e
y
axis of the sample, providing polarization
sensitive measurements.
13
6. Phase measurements on metasurfaces
The optical phase of light reflected from the metas
urface is measured by a Michelson
interferometer (Fig. S9a). A narrowband coherent n
earinfrared light source (emitted by a
tunable laser) is directed onto a beam splitter. On
e beam is focused onto the mirror mounted on a
piezostage, while the other beam is aligned to the
edge of the metasurface so that half of the
beam is reflected from the metasurface and the othe
r half from the multilayer of Al
2
O
3
/ITO/Au
backplane (reference) (Fig. S9b). The beam reflect
ed from the sample (metasurface and
reference) separately interferes with the beam in t
he other path that has been reflected from the
mirror, resulting in the two regions of interferenc
e fringe recorded by the nearinfrared camera.
The optical path length between two paths can be pr
ecisely adjusted with the piezostage until
the interference fringes are seen in the camera.
Another method to measure the phase modulation with
applied bias is to capture the
interference fringes for both the metasurface and r
eference at different positions of the
Michelson mirror, using the piezo stage. Such a mea
surement method has been demonstrated in
the past.
11
However since the piezo stage can only provide lim
ited position accuracy, several
measurements would need to be acquired (more than 1
00 times) to measure an accurate phase
shift. We believe our method provides an easy and d
irect way to measure the phase modulation
from the metasurfaces.
To obtain the phase shift induced by applied bias (
Fig. 3c), the shift of the interference
fringes under bias is measured. Figure S10a shows t
he interference fringes recorded by the near
infrared camera at applied bias voltages from 0 V t
o 2.5 V. An offset of fringes between the light
reflected from the metasurface and reference is cle
arly seen in the case of 0V, but disappears in
the case of 2.5V. To analyze the fringes, we select
two crosssections shown in Fig S10a for each
image (Fig. S10b, black lines) and translate them t
o the smooth curves by using a moving
average filter (Fig. S10b, magenta lines for metasu
rface and green lines for reference).
Furthermore, we fit a function (sinusoidal function
times a Gaussian) to each smooth curve. We
use the obtained sinusoidal function to retrieve th
e phase at each applied bias. The fitted
sinusoidal prefactors for each smooth curve are sh
own in Figure S10c, where red and olive lines
represent the sinusoidal prefactors originating fr
om the metasurface and reference, respectively.
14
The phase is defined as the ratio of the distance b
etween the wave crests of metasurface and
reference (UΛ) and the period of the wave (Λ). The
phase shift calculated in such a way will be
given in units of radian. The nonzero phase shift i
mplies that the phase under applied bias is
different from the phase with no applied bias (0 V)
. Table S2 summarizes the values of phases
and phase shifts retrieved for each applied bias. T
he retrieved phase shift as a function of applied
bias is shown in Fig. 3c. It should be mentioned th
e slight variation of fitted Λ value between the
two crosssections for each interference fringes re
sult in the phase measurement errors. The
region of upper and lower error bars of phase shift
(Table S2 and Fig. 3c) corresponds to the
standard deviation of Λ smaller than 7%. We totally
recorded 4 images for each voltage and
chose several different positions of crosssections
for fitting for each image. All of them are
considered for the error bars.
Figure S9.
(a) Schematic of the optical phase measurement setu
p. (b) Schematic showing the
position of the incident light where half of the be
am is reflected from the metasurface and
another half is reflected from the reference (the m
ultilayer stack of Al
2
O
3
/ITO/Aubackplane).
15
Figure S10.
(a) Measured interference fringes. The upper and l
ower fringes originate from the
metasurface and reference, respectively. (b) The ra
w data (black line) and smooth fitting curves
(magenta line for the metasurface and green line fo
r the reference). (c) The fitted sinusoidal
waves for each smooth curve. Red line: metasurface.
Olive line: reference. Λ: the period of the
sinusoidal waves. UΛ: the distance between the peak
s of the sinusoids obtained via fitting
reference fringes and metasurface fringes.
16
Table S2.
The phase and phase shift induced by applied bias
from 0 to 2.5 V.
7. The structural non-uniformity and simulation
The wavelength at which phase measurements are perf
ormed (1573 nm) is shifted with respect to
the resonant wavelength obtained via reflection mea
surements (compare Figs. 3c, 2c, and 3a).
This difference is due to the fact that the reflect
ance measurements have been taken by
illuminating the center of the sample while the pha
se measurements have been taken by
illuminating the edge of the sample. Figure S11 sho
ws that the resonance shifts from 1560 nm to
1585 nm depending on the part of the sample probed
during the measurement (center vs. edge).
During the reflectance position measurement, spot s
ize with diameter of 5 Om is used. The taken
SEM images confirm that when moving to the edge of
the sample the interference fringes
become wider (Fig. S11). In the simulations shown i
n Fig. 2a, the resonance dip is at 1557 nm
yielding operation wavelength of 1550 nm. Note that
the operation wavelength, that is the
wavelength at which large phase shift is observed,
is shifted with respect to the resonant
wavelength. In the phase shift simulations shown in
Fig. 3c, the operation wavelength is 1573
nm.
17
Figure S11.
Measured reflectance spectra correspond to differe
nt positions of the metasurface.
The SEM images show that the width of the stripes i
n the center of the sample is 252 nm while at
the sample edge the stripe width is 258 nm.
In simulation, we keep all the parameters of materi
als and geometry the same except for the
width of stripe antenna. The width is changed from
250 nm (center of the sample) to 257.5 nm in
simulation for the edge of the sample. Figures S12a
and S12b shows the reflectance and phase
shift spectra as a function of applied bias in this
case. The dashed line marks the position of a
reflectance dip at 1580 at no applied bias (0V), sh
owing a god match to the experimental
reflectance spectra displayed in Fig. S11 (edge of
the sample, cyan line). The phase shift under
applied bias at 1573 nm shown in Fig. 3c (simulatio
n, red dashed line) is obtained from the Fig.
S12b (crosssection of wavelength of 1573 nm). Our
calculations show good agreement with the
experimental phase shift (Fig 3c ).
18
Figure S12.
Simulated (a) reflectance and (b) phase shift due
to gating as a function of
wavelength and applied voltage. The dotted lines sh
ow points in Vλ parameter space for which
permittivity of ITO equals from 1, 0 or 1 at the A
l
2
O
3
/ITO interface, representing the ENZ
region in the accumulation layer. The green dashed
line marks the position of a reflectance dip
corresponding to the magnetic dipole resonance. Not
e that the width of stripe antenna
w
is
chosen as 257.5 nm corresponding to the wider strip
e antennas at the edge of the sample.
8. Reflectance spectra of the metasurface under neg
ative bias
Reflectance spectrum of the metasurface was also me
asured under a negative voltage bias
between 0 V to 1.5 V as shown in Fig. S13. It is c
lear from the figure that the reflectance dip
shifts to longer wavelengths when increasing absolu
te values of the applied negative voltage.
This occurs due to further depletion of the ITO at
the Al
2
O
3
/ITO interface as compared to the
case when no voltage is applied. The shift of the r
esonance is less significant in this case as
compared with the positive bias since applying nega
tive bias results in increase of
ε
r
ITO
at
19
Al
2
O
3
/ITO interface, and shifts ENZ regions in the deple
tion layer further away from the
magnetic dipole resonance.
Figure S13.
Measured reflectance spectrum (a) and relative refl
ectance change (b) from the
metasurface for different negative biases.
9. High frequency AC modulation
The reflectance spectrum of the sample we used in t
he frequency modulation experiments is
depicted in Fig. S14a (the width of the stripe ante
nna is
w
= 250 nm). As can be seen from the
figure, the magnetic and electric resonances of the
metasurface are located at ~ 1650 nm and ~
1020 nm, respectively. By applying 2 V bias, the ma
gnetic resonance shifts to shorter
wavelengths, as expected from modeling, thus reduci
ng the reflectance at the target wavelength
of 1515 nm. Different biases are applied to the met
asurface with a modulation speed of 2 MHz
and the results are shown in Fig. S14b. It can be s
een from the figure that up to ~ 15% amplitude
modulation is obtained with an applied voltage of 2
V. The modulation speeds of the device with
as high as 10 MHz were demonstrated (see Fig. 3d of
the main text), clearly showing the
ultrafast modulation based on the fieldeffect dyna
mics. The detection speed is currently limited
by our detection method while we expect that modula
tion speeds up to >1 GHz may be
achievable with appropriate highfrequency intercon
nect designs, due to the small device
footprint and capacitance.
1,12
20
Figure S14.
(a) Measured reflectance spectra of the sample for
high frequency measurement at 0
V and 2 V bias. (b)
Measured time response at 2 MHz modulations for dif
ferent applied voltages.
10. IV characteristic of metasurface sample
A typical currentvoltage response measurement for
a fieldeffect tunable metasurface is
given in Fig. S15. The voltage is applied between t
he gold back plane and the two top gold
electrical pads. The current flowing between 5 nm o
f Al
2
O
3
and 20 nm of ITO remains low (<
200 nA) under an applied bias of 2.5 V. The breakdo
wn voltage for the devices varies from
sample to sample taking values between 2.5 and 3 V,
which correspond the breakdown electric
fields of 56 MV/cm (with 5 nm insulating Al
2
O
3
layer). These values are of the same order of
magnitude as the values reported elsewhere.
13
Note that the breakdown voltage for the
metasurface is lower than the planar structure with
out integrated nanostructures. We anticipate
that the higher breakdown field of the metasurface
device could be achieved by further treatment
on the deposition and nanopatterning/lift off proc
esses. It is worth mentioning that IV curves
do not demonstrate significant hysteretic behavior
thus excluding the possibility of diffusion of
gold into ITO or any ionic movement giving rise to
this modulation.