of 18
1
Millivolt modulation of a plasmonic metasurface via
ionic conductance
Krishnan Thyagarajan,
1, 2
, Ruzan Sokhoyan
1
, Leonardo Z. Zornberg
1
and Harry A. Atwater
1, 2
1
Thomas J. Watson Laboratory of Applied Physics, California Ins
titute of Technology,
Pasadena, California 91125, USA
2
Kavli Nanoscience Institute, California Institute of Technolog
y, Pasadena, California 91125,
USA
KEYWORDS:
Active plasmonics, metasurface
, field effect, plasmonic nanoca
vities, resistive
switching, ion conductance, transparent conducting oxide
ABSTRACT:
We report here and experimentally demonstrate an actively cont
rolled gate-tunable
plasmonic metasurface operating in the visible region of the el
ectromagnetic spectrum, where –
strikingly – the operating voltages for reflectance modulation
are much less than 1V. The
electrically tunable metasurface c
onsists of inverse dolmen str
uctures (iDolmen) patterned on
silver and chromium on a quartz substrate and subsequently cove
red with a 5 nm thin layer of
Al
2
O
3
followed by a 110 nm indium tin oxide (ITO) layer, which acts
as a transparent electrode.
Our designed structures show up to 78% change in reflection upo
n applying small voltages (<1V).
We explain this behaviour via ion conductance of silver through
Al
2
O
3
and ITO, leading to active
2
resistive switching. Interesting complementary effects such as
decreased reflection in the same
structures over a broadband of wavelengths is also seen on reve
rsing the applied bias. The results
provide an insight into the use of the resistive switching for
electrical control over light-matter
interaction in plasmonic metasurfaces.
ARTICLE:
Controlling the configuration of nanostructures in the mesoscop
ic domain, permits targeted light-
matter interaction allowing the creation of systems including o
ptical modulators, switches, phase
plates and wave-front shaping components. Metasurfaces are two
dimensional analogues of
metamaterials, and are comprised of nanostructures on a surface
that modifies the way incident
light interacts with the surface. Many examples of such metasur
faces have been explored in the
past and have given rise to the field of flat optics, where con
ventional optical functionalities are
retained while shrinking the macroscopic optical component itse
lf, into a two dimensional surface.
Such a modification of the surface has been demonstrated in ear
lier work to effectively control
various facets of light-matter interaction including the phase,
1
polarization,
2
reflection and
transmission amplitude,
3
the shaping of a complex wavefront
4
etc. However such metasurfaces are
passive in nature, implying that their response is fixed once t
hey are fabricated. Another class of
metasurfaces includes those that are active in nature. Such sys
tems can be made to dynamically
change their response to the incident light by using an externa
l mechanism. Many different
mechanisms have been exploited to make such active metasurfaces
. These include the use of liquid
crystals,
5
the modulation of fermi levels in graphene by the application
of an external voltage bias,
6
the change in the phase of a material upon the application or r
emoval of heat,
7
the field effect in
materials such as transparent conducting oxides which exhibit l
arge changes in their complex
3
refractive indices upon the application of a bias
8
etc. It is hoped that a few of these mechanisms
may be used for the active optoelectronic control of elements o
n a chip since the inclusion of
plasmonic structures will permit the realization of ultracompac
t active optical components such as
optical transistors, photodetectors and modulators. However mos
t of these mechanisms require
conditions such as the application of heat or large external vo
ltages that are often not practical
when a device on chip needs to be fabricated for practical appl
ications. Here, we propose a novel
class of tunable metasurfaces in which the modulation is based
on ionic conductance, permitting
millivolt bias requirements apart from being applicable across
plasmonic and dielectric
metasurfaces.
Furthermore, if the scope of using plasmonic elements in future
nanophotonic chips is to be
increased, the problem of energy efficiency needs to be address
ed. Not being able to reduce the
inherent material losses, reducin
g the physical dimension of a
device active volume will partially
help alleviate this problem, however the need for attojoule lev
el modulation is becoming a
necessary requisite as well. It is known that the power consump
tion in a standard CMOS device
goes as P = CV
2
f , where C is the capacitance of the device, V is the voltage
bias applied and f is
the clock frequency. Hence an ideal nanophotonic CMOS device wi
ll be significantly aided by
displaying a low capacitance as well as very low operating volt
ages. Recent demonstrations of low
energy optical modulators include a plasMOStor as a metal-oxide
-Si field effect plasmonic
modulator,
9
a femtojoule electro-optic modul
ation using a silicon-organic
hybrid device,
10
atomic
scale plasmonic switches,
11
and memristor based effects in plasmonic waveguides.
12
An interesting
mechanism that some of the examples take advantage of, is that
of ion conductance reminiscent of
memristors.
4
Ion conductance arises in material systems that permit the move
ment of ions of a certain species
across a host medium, especially under the application of an ex
ternal bias. While permitting the
movement of ions, such media prevent the movement of electrons,
thus allowing for insulating
behaviour despite the movement of ions. When the ions move, the
y cause a significant change in
the resistance of the medium, thus giving rise to the mechanism
of resistive switching. Resistive
switching denotes a reversible phenomenon in two terminal eleme
nts, which change their
resistance reversibly, often in
a non-volatile manner, upon ele
ctrical stimuli. When the resistance
states are non-volatile in nature, it is as if the stimulus aff
ects an internal state variable of the
element, which controls the resistance. This is akin to the res
istance being
memorized
by the
element, therefore comparing this mechanism to phenomena first
proposed in the so-called
memristive elements
. Such a switching of resistance can be brought about by a wide
variety of
different phenomena including but not restricted to nanomechani
cal phenomena,
13
magnetoresistive effects such a
s spin-transfer torque (STT),
14
electrical effects such as leakage
current through flash gate stacks in which trapping/de-trapping
phenomena occur,
15
phase change
between amorphous and crystalline phases,
16
and nanoionic redox phenomena or electrochemical
metallization.
17
The particular mechanism which is utilized in this present wor
k is based on the
nanoionic redox phenomenon/electrochemical metallization. A typ
ical system involves a simple
metal-insulator-metal configuration. One of the metals is an ac
tive electrode that is oxidizable
(soluble), such as silver or copper, while the counter electrod
e is inert (insoluble). An oxide,
chalcogenide, or halide material in between these two electrode
s serves to transport the metal
cations. When a positive voltage is applied to the oxidizable e
lectrode, its constituent metal starts
to dissolve and results in the deposition of a metallic filamen
t at the opposite inert electrode. In the
extreme case, this metallic filament ultimately bridges the rel
atively insulating ion conductor and
5
causes a very large change in the resistance of the device. Upo
n reversing the bias, the filament
starts to dissolve and the system tends back towards its origin
al state. Commonly used active
electrodes include those with mobile ions, such as silver ions
in silver halides and chalcogenides
– AgI, Ag
2
S, Ag
2
Se, and Ag
2
Te. It is well known that although the low-temperature crystall
ine
phases of such materials are less conductive with often mixed i
onic and electronic contributions,
their high-temperature polymorphs are excellent conductors. How
ever, what is also known is that
the temperature at which they show excellent ionic conductance
can be dramatically brought down
to room temperature or lower simply by scaling down the size of
the silver nanoparticles
involved.
18
As for the insulator sandwiched in between the two electrodes,
it is well known that
the movement of ions in a medium is strongly influenced by inte
rstitial channels in certain
directions in crystalline materials, long-range disorder in amo
rphous, nanoscopically porous
materials as well as the presence of defects including in conve
ntional oxides such as alumina
(Al
2
O
3
), silica (SiO
2
) and titania (TiO
2
).
19-22
Alumina deposited using atomic layer deposition has
been shown to be conductive to specific sets of ions.
19
Finally, the counter electrode has in
particular no other role to play, other than permitting the app
lication of a bias.
Such a resistive switching based phenomenon has been attempted
to be integrated with plasmonic
systems, allowing the authors to explore switching in waveguide
s
12
as well as make atomic scale
switches.
11
With the general mechanism of memristive switching requiring o
nly a few 100 mV of
biasing and continuing to go lower,
23
incorporating this mechanism in nanophotonics can perhaps
help significantly lower the power consumption in optical chips
. In this work, we experimentally
demonstrate the integration of such a resistive switching mecha
nism in a plasmonic metasurface,
permitting even lower operating voltages as small as 5 mV.
6
Our device consists of 80 nm of silver evaporated using electro
n beam evaporation with a 1 nm
chromium adhesion layer onto a quartz substrate (SPI Inc.). Foc
used ion beam lithography is
carried out on the deposited metal film to create a metasurface
of inverse dolmen structures. The
typical dimensions of these features involve the two lower para
llel rods of lengths 150 nm and
width 100 nm separated by a 30 nm gap. Both of these are separa
ted by another 30 nm gap to the
top rod of length 230 nm and width 100 nm. An array of 200
μ
m x 200
μ
m is milled onto the film
and a subsequent 5 nm atomic layer deposition of Al
2
O
3
is undertaken. Thereafter, 110 nm of
indium tin oxide (ITO) is sputtered using RF magnetron sputteri
ng (Ar+O
2
flow rate of 0.5 sccm)
(see Fig. 1). All the deposition steps are undertaken with appr
opriate face masks made of stainless
steel, at every step to permit an easy bias application configu
ration in the final device. Two sets of
samples were made, one set was milled until the chromium adhesi
on layer was reached, and
thereby retaining a thin layer o
f chromium still at the bottom
of the milled inverse dolmen legs,
while the other set was milled through till the quartz substrat
e was reached. We will refer to these
two sets of samples as
with
and
without chromium
respectively in future references.
Materials characterization of the indium tin oxide was undertak
en to determine the carrier
concentration and permittivity of the deposited film, using the
Van der Pauw Hall measurement
technique (instrument make) and spectroscopic ellipsometry (VWA
SE, J. A. Woollam)
respectively. The resulting carrier concentration was 0.5x10
20
/cm
3
and the obtained permittivity is
shown in the supporting information. The 5 nm ALD of alumina wa
s repeatedly tested for its
insulating properties and it was
seen that most of the samples
were displaying very good insulation
properties (for sample curve see SI). Those samples that did no
t exhibit a good insulating layer
were discarded. A transmission electron microscopy (TEM) image
of an identical deposition on a
silicon substrate was seen to show a uniform layer of roughly 1
.2 nm of chromium as the adhesion
7
layer (see SI for images) and hence did not create disconnected
islands. Spectroscopic ellipsometry
was conducted on a dummy sample of silver deposited on silicon,
however it was observed that
there was a temporal evolution of the permittivity. This was ve
rified by undertaking spectroscopic
ellipsometry every day for a week (see SI). It was observed tha
t the trend in the permittivity would
predict a slow red-shift of the intended plasmon resonance in o
ur structures. Furthermore, energy-
dispersive X-ray spectroscopy (EDX) was undertaken on the sampl
es and it was seen that there
was a finite amount of Sulphur present throughout, wherever sil
ver was present (see SI). It is
believed, that this was due to the general contamination of sil
ver (tarnishing), when exposed to air.
With various forms of materials characterization undertaken on
our samples, optical measurements
were undertaken to observe reflection, transmission and absorpt
ion.
All optical measurements were undertaken at normal incidence wi
th incident polarization along
the two parallel bottom legs (s
ee Fig.). A microscope objective
5 X (Olympus, add NA) was used
to focus down the incident light from a supercontinuum laser li
ght source (Fianium) to a spot of
size roughly xxx micron and the signal measured using a silicon
detector (make specifications).
Measurements were taken for a wavelength range from 450 nm to 8
50 nm. For the bias dependent
measurements, a voltage was applied using a Keithley Source met
er (specify make), and a bias
applied between the top layer of ITO and the bottom layer of si
lver at appropriate locations far
from the actual device area so as not to damage it. This was pe
rmitted by the design of the masks
that created a pad large enough to be biased externally using e
ither contact probes or conducting
wires/tape.
The results of the optical measur
ements are shown in Fig.2. The
se results involve applying a
positive bias to the silver elect
rode and a negative bias to th
e ITO electrode. This has two
consequences. Firstly, this permits the movement of silver ions
through the alumina, as was
8
described in the earlier part of the manuscript and secondly, i
t creates a charge accumulation or
active layer at the ITO-alumina interface. One can clearly see
from Fig.2a, that upon the
application of a bias of 5 mV (in 10 steps, leading to each inc
remental graph representing a bias
of 0.5 mV), the reflection exhibits a broadband change in ampli
tude by nearly 25%. There is
however no noticeable wavelength shift observed. Similarly, the
measured transmission shows a
decrease for the same set of bias voltages. And the consequent
change in absorption shows an
increase or decrease depending on the wavelength region looked
at. The complexity of the curve
can be attributed to the inherent complexity of the inverse dol
men structure itself. Also shown in
the same figure are the full-wave electromagnetic numerical sim
ulation results using finite
difference time domain. It can be seen that the experimental pl
ot is red-shifted with respect to the
numerical results, by an amount which can be attributed to the
temporal degradation of silver itself,
as was discussed earlier in the manuscript. Another reason for
this mismatch could be the non-
uniformity of the milled structures across the focused spot. To
qualitatively probe the phenomenon
of resistive switching in our system, we conceptualize the silv
er ions as having moved into the
alumina and possibly even the ITO, creating an effective medium
in the region, whose complex
permittivity depends on the fill-fraction of the silver ions in
the ‘host’ medium. The presence of
the silver ions at the ITO alumina interface also contributes t
o an increase in the decay parameter
(decrease in the scattering time) of the ITO itself, which can
be modeled by changing the ‘gamma’
parameter used in the Drude part to model the ITO. These effect
s are used to create an effective
medium in the alumina region using the Bruggeman model of the e
ffective medium theory. An
increase in bias is modeled as an increase in the fill-fraction
of the silver in the alumina, leading
to a change in the complex permittivity. However at the same ti
me, the gamma parameter of the
inherent ITO is also changed (details in SI). As can be seen in
Fig.2, probably the actual mechanism
9
occurring in this system is a complex combination of the two ab
ove mentioned phenomena. The
competition of these dif
ferent factors determines the eventual
trend observed experimentally. A
complete thorough description of the exact modeling in our syst
em is beyond the scope of this
present work and will be undertaken in future work. However, wh
at can be seen is that just the
modification of the complex permittivity in the active region o
f ITO is not sufficient to explain our
observed phenomenon, for with these small voltages, the index c
hange is insufficient to observe
what we experimentally obtain. Secondly, the modeling also hint
s that there are several competing
factors contributing to our final experimental results. Neverth
eless we can qualitatively see that
these may be potential explanations for what we obtain.
A second interesting observation in our systems involves the di
stinction between samples with and
without chromium. The
obtained experimental reflection for the
same polarization and set of
applied voltages for the two sets
of samples in seen in Fig.3.
It is surprising to see how large an
impact the chromium has on the observed phenomenon. As mentione
d earlier, it is known that
long-rage disorder as well as defects and stress in the insulat
ing oxide can dramatically increase
ionic conductance.
19
In particular, the presence of defects and stress on the behav
iour of the ionic
conductance of alumina (deposited using ALD) deposited onto chr
omium is known to (a) increase
the oxygen vacancies as well as (b) reduce the surface resistan
ce at the interface.
19
Apart from
ionic conductance, the same resistive switching mechanism is al
so known to occur in systems
exhibiting oxygen vacancies, such as in the TiO
2
based memristive systems.
24
Secondly, the
reduction in surface resistance might be permitting quicker and
easier movement of the silver ions
in to the alumina layer. Therefore both these observed effects
of the ionic conductance properties
of alumina in the presence of chromium and chromium oxide, hint
at the reason why such a large
difference is seen in the two sets of samples. Therefore, it is
imagined that the chromium is actually
10
making the ionic conductance of the alumina much better and the
reby permitting us to see this
optical effect at such small voltage biases.
Upon the increase of the external bias, it is seen that the ref
lection increases until it reaches a
maximum saturation value of 90% upon the application of 60 mV.
Thereafter, there is a steep
reduction in the reflection amplitude, tailing down at 12% beyo
nd which there is a dramatic
increase in the leakage current. This gives an amazing dynamic
range of 78% with the application
of just 100 mV before the breakdown occurs. It is believed that
with this small 5 nm alumina layer
and under these fabrication conditions, the filament formation
is complete with as little as 100 mV,
leading to a shorting of the insulating layer and increasing ab
sorption and decreasing reflection
significantly.
The metasurface was tested for its modulation speed by measurin
g the modulation of the reflection
amplitude at normal incidence at a peak reflectance (lambda = x
xx nm), with polarization along
the bottom legs. It was seen that the measured reflection faith
fully reproduced th
e modulation of
the input bias up to 650 Hz. Since the modulation here is induc
ed by the movement of ions, it is
expected that it will cap out at a few kilohertz. However, sinc
e the same resistive switching
mechanism can also occur in the form of other more mobile confi
gurations such as oxygen vacancy
defects, it is expected that modulation can go up to the GHz re
gime in other such similar systems.
In conclusion, in this letter we have experimentally demonstrat
ed the first of its kind ultralow
power consumption in an optical modulator exhibiting up to 78%
modulation in reflection
amplitude with a bias of just 0.1
V (and up to 25% with a bias
of just 5 mV). Our system exploits
the mechanism of resistive switching in metal-insulator-metal s
ystems and improves upon
previous work by pushing down the required bias voltage require
d for modulation by incorporating
11
additional external defects that
allows for higher ionic conduc
tance. A complete understanding of
such a mechanism can have a significant impact on the design of
future attojoule optical
modulators, highly sensitive optical readout of surface phenome
na, optical display technology and
even non-volatile optical memories.
AUTHOR INFORMATION
Corresponding Author
*kthyagar@caltech.edu
,
haa@caltech.edu
12
Figure 1.
13
Figure 2.
14
Figure 3.
15
Figure 4.
16
References
1.
Yu, N. and Capasso, F.
Nature Materials
2014,
13
, 139-150.
2.
Arbabi, A.; Horie, Y.; Bagheri, M.; Faraon, A.
Nature Nanotechnology
2015,
10
, 937-943.
3.
Yu, N.; Genevet, P.; Kats, M. A.; Aieta, F.; Tetienne, J.-P.; C
apasso, F.; Gaburro, Z.
Science
2011,
334, 333-337.
4.
Veksler, D.; Maguid, E.; Shitrit, N.; Ozeri, D.; Kleiner, V.; H
asman, E.
ACS Photonics
2015, 2(5), 661-667.
5.
Buchnev, O.; Podoliak, N.; Kaczmarek, M.; Zheludev, N. I.; Fedo
tov, V. A.
Adv. Opt.
Mater.
2015,
3
(5), 595.
6.
Brar, V. W.; Jang, M. S. ; Sherrott, M. ; Lopez, J. J.; Atwater
, H. A.
Nano Lett.
2013,
13
(6),
2541-2547.
7.
Dicken, M. J.; Aydin, K.; Pryce, I. M.; Sweatlock, L. A.; Boyd,
E. M.; Walavalkar, S.; Ma,
J.; Atwater, H. A.
Optics Exp.
2009,
17
(20), 18330-18339.
8.
Huang, Y.-W.; Lee, H. W.; Sokhoyan, R.; Pala, R.; Thyagarajan,
K.; Han, S.; Tsai, D. P.;
Atwater, H. A. 2015, arXiv:
1511.09380 [physics.optics].
9.
Dionne, J. A.; Diest, K.; Sweatlock, L. A.; Atwater, H. A.
Nano Lett.
2009,
9
(2), 897-902.
10.
Koeber, S.; Palmer, R.; Lauermann, M.; Heni, W.; Elder, D. L.;
Korn, D.; Woessner, M.;
Alloatti, L.; Koenig, S.; Schindler, P. C.; Yu, H.; Bogaerts, W
.; Dalton, L. R.; Freude, W.;
Leuthold, J.; Koos, C.
Light: Science & Applications
2015,
4
e255.
17
11.
Emboras, A.; Niegemann, J.; Ma, P.; Haffner, C.; Pedersen, A.;
Luisier, M.; Hafner, C.;
Schimmel, T.; Leuthold, J.
Nano Lett.
2016,
16
(1), 709-714.
12.
Emboras, A.; Goykhman, I.; Desiatov, B.; Mazurski, N.; Stern, L
.; Shappir, J.; Levy, U.
Nano Lett.
2013,
13
(12), 6151-6155.
13.
Akarvardar, K. and Philip Wong, H
.-S. (2012) in Nanoelectronics
and information
technology, 3
rd
edition, (ed R Waser), Wiley-VCH Verlag GmbH, pp 375-387.
14.
Dieny, B.; Sousa, R.; Nozieres, J.-P.; Redon, O.; and Prejbeanu
, I. L. (2012) in
Nanoelectronics and information technology, 3
rd
edition, (ed R. Waser), Wiley-VCH
Verlag GmbH, pp 655-668.
15.
Zhirnov, V. and Mikolajick, T. (2012) in Nanoelectronics and in
formation technology, 3
rd
edition (ed R. Waser), Wiley-VCH Verlag GmbH, pp 621-634.
16.
Raoux, S. and Wuttig, M. (2012) in
Nanoelectronics and informat
ion technology, 3
rd
edition (ed R. Waser), Wiley-VCH Verlag GmbH, pp 669-682.
17.
Waser, R.; Bruchhaus, R. and M
enzel, S. (2012) i
n Nanoelectroni
cs and information
technology, 3
rd
edition (ed R. Waser), Wiley-VCH Verlag GmbH, pp 683-710.
18.
Makiura, R.; Yonemura, T.; Yamada, T.; Yamauchi, M.; Ikeda, R.;
Kitagawa, H.; Kato,
K.; and Takata, M
Nature Materials
8
, 476-480 (2009).
19.
Samarao, A. K.; O’brien, G.; and Feyh, A., patent WO 2015021368
A1.
20.
Lei, Y. J and Xiao, D. Q.
IEEE Transactions on dielectrics and electrical insulation
, Vol.
13, No. 1, February 2006.
21.
Lei, Y. J. and Xiao, D. Q.,
J. Electroceram.
(2008), 21: 214-217.