Hyper-selective plasmonic color filters
D
AGNY
F
LEISCHMAN
,
1
L
UKE
A. S
WEATLOCK
,
1,2
H
IROTAKA
M
URAKAMI
,
3
AND
H
ARRY
A
TWATER
1,*
1
California Institute of Technology, 1200 E California Blvd, Pasadena, CA 91125, USA
2
Northrop Grumman NG Next, One Spac
e Park, Redondo Beach, CA 90250, USA
3
Sony Semiconductor Solutions Corporation, 4-14-
1 Asahi-cho, Atsugi-shi, Kanagawa, 243-0014, Japan
*haa@caltech.edu
Abstract:
The subwavelength mode volumes of pl
asmonic filters are well matched to the
small size of state-of-the-art active pixels in
CMOS image sensor arrays used in portable
electronic devices. Typical plasmonic filters exhibit broad (> 100 nm) transmission
bandwidths suitable for RBG or CMYK color f
iltering. Dramatically reducing the peak width
of filter transmission spectra would allow for the realization of CMOS image sensors with
multi- and hyperspectral imaging capabilities. We find that the design of 5 layer metal-
insulator-metal-insulator-metal structures gives rise to multi-mode interference phenomena
that suppress spurious transmission features and give rise to single transmission bands as
narrow as 17 nm. The transmission peaks of these multilayer slot-mode plasmonic filters
(MSPFs) can be systematically
varied throughout the visibl
e and near infrared spectrum,
leading to a filter that is CMOS integrable,
since the same basic MSPF structure can operate
over a large range of wavelengths. We find that MSPF filter designs that can achieve a
bandwidth less than 30 nm across the visible and demonstrate experimental prototypes with a
FWHM of 70 nm, and we describe how experimental structure can be made to approach the
limits suggested by the model.
© 2017 Optical Society of America
OCIS codes:
(250.5403) Plasmonics; (050.6624)
Subwavelength structures.
References and links
1. M. Khoso and K. Kamra, “Smartphones:
A Supercomputer in Your Pocket,”
http://www.northeastern.edu/levelblog/2016/04/21/
smartphones-supercomput
er-in-your-pocket/.
2. “Sensors Overview,” https
://developer.android.com/guide/topics/sensors/sensors_overview.html.
3. S. Yokogawa, S. P. Burgos, and H. A. Atwater, “Plasm
onic color filters for CMOS
image sensor applications,”
Nano Lett.
12
(8), 4349–4354 (2012).
4. S. P. Burgos, S. Yokogawa, and H. A. Atwater, “Col
or imaging via nearest neighbor hole coupling in plasmonic
color filters integrated onto a complementary me
tal-oxide semiconductor image sensor,” ACS Nano
7
(11),
10038–10047 (2013).
5. T. Xu, Y.-K. Wu, X. Luo, and L. J. Guo, “Plasm
onic nanoresonators for high-resolution colour filtering and
spectral imaging,” Nat. Commun.
1
, 59 (2010).
6. B. Y. Zheng, Y. Wang, P. Nordlander, and N. J.
Halas, “Color-selective and CMOS-compatible photodetection
based on aluminum plasmonics,” Adv. Mater.
26
(36), 6318–6323 (2014).
7. T. W. Ebbesen, H. J. Lezec, H. F. Ghaemi, T. Thio,
P. A. Wolff, T. Thio, and P. A. Wolff, “Extraordinary
optical transmission through sub-
wavelength hole arrays,” Nature
86
, 1114–1117 (1998).
8. J. A. Dionne, L. A. Sweatlock, H. A. Atwater, and
A. Polman, “Plasmon slot wa
veguides: Towards chip-scale
propagation with subwavelength-s
cale localization,” Phys. Rev. B
73
, 03547 (2006).
9. Q. Chen, X. Hu, L. Wen, Y. Yu, and D. R. S. Cumming, “Nanophotonic Image Sensors,” Small
12
(36), 4922–
4935 (2016).
10. H. Wang, X. Wang, C. Yan, H. Zhao, J. Zhang, C. Sa
ntschi, and O. J. F. Martin
, “Full Color Generation Using
Silver Tandem Nanodisks,” ACS Nano
11
(5), 4419–4427 (2017).
11. A. Kristensen, J. K. W. Yang, S. I. Bozhevolnyi, S.
Link, P. Nordlander, N. J. Halas, and N. A. Mortensen,
“Plasmonic colour generation,” Nat. Rev. Mater.
2
, 16088 (2016).
12. L. Frey, P. Parrein, J. Raby, C. Pellé, D. Hérault, M.
Marty, and J. Michailos, “Color filters including infrared
cut-off integrated on CMOS image sensor,” Opt. Express
19
(14), 13073–13080 (2011).
13. T. Hirayama, “CMOS Image Sensor Evolution: Past, Present, and Future,” in
SEMICON Europa
(2016).
14. W.-Y. Jang, Z. Ku, J. Jeon, J. O.
Kim, S. J. Lee, J. Park, M. J. Noyola, and A. Urbas, “Experimental
Demonstration of Adaptive Infrared
Multispectral Imaging using Plasmonic Filter Array,” Sci. Rep.
6
, 34876
(2016).
Vol. 25, No. 22 | 30 Oct 2017 | OPTICS EXPRESS 27386
#296003
https://doi.org/10.1364/OE.25.027386
Journal
©
2017
Received
3 Aug
2017;
revised
5 Oct
2017;
accepted
11
Oct
2017;
published
24
Oct
2017
15. A. A. Gowen, C. P. O’Donnell, P. J. Cullen, G. Down
ey, and J. M. Frias, “Hyperspectral imaging – an emerging
process analytical tool for food quality and
safety control,” Trends Food Sci. Technol.
18
, 590–598 (2007).
16. E. K. Hege, D. O’Connell, W. Johns
on, S. Basty, and E. L. Dereniak, “Hyperspectral imaging for astronomy and
space surveillance,” Opt. Sci. Technol. SPIE’s 48th Annu. Meet.
5159
, 380–391 (2004).
17. S. S. Wang and R. Magnusson, “Theory and applications of guided-m
ode resonance filters,” Appl. Opt.
32
(14),
2606–2613 (1993).
18. D. B. Mazulquim, K. J. Lee, J. W. Yoon, L. V. Muniz, B.-H. V. Borges, L. G. Neto, and R. Magnusson,
“Efficient band-pass color filters enabled by resonant
modes and plasmons near the Rayleigh anomaly,” Opt.
Express
22
(25), 30843–30851 (2014).
19. Z. Li, S. Butun, and K. Aydin, “Large-area, Lit
hography-free super absorbers and color filters at visible
frequencies using u
ltrathin metallic films,” ACS Photonics
2
, 183–188 (2015).
20. H. Gao, W. Zhou, and T. W. Odom, “Plasmonic crystals
: A platform to catalog resonances from ultraviolet to
near-infrared wavelengths in a plasm
onic library,” Adv. Funct. Mater.
20
, 529–539 (2010).
21. D. Pacifici, H. J. Lezec, L. A. Sweatlock, R. J. Walt
ers, and H. A. Atwater, “Universal optical transmission
features in periodic and quasiperi
odic hole arrays,” Opt. Express
16
(12), 9222–9238 (2008).
22. A. Mazor, D. J. Srolovitz, P. S. Hagan, and B. Bukiet, “Columnar growth in th
in films,” Phys. Rev. Lett.
60
(5),
424–427 (1988).
23. A. A. Maradudin and D. L. Mills,
“Scattering and absorption of electromagnetic radiation by a semi-infinite
medium in the presence of
surface roughness,” Phys. Rev. B
11
, 1392–1415 (1975).
24. O. J. Martin, K. Thyagarajan, a
nd C. Santschi, “A New Fabrication Meth
od for Aluminum and Silver Plasmonic
Nanostructures,” in
Workshop on Optical Plasmonic Materials
(The Optical Society of America, 2014).
25. A. Joshi-Imre and S. Bauerdick, “Direct
-Write ion beam lithography,” J. Nanotechnol.
2014
, 170415 (2014).
1. Introduction
Today’s state-of-the-art mobile
electronics are as powerful as
larger computers [1], and are
equipped with a variety of sensors includ
ing accelerometers, gyro
scopes, CMOS image
sensors, and magnetometers [2]. CMOS image sensors have a particularly broad landscape of
potential new functions: optical data on mobile platforms today consists primarily of three-
color imaging, but a wide vari
ety of applications could be accessed by the collection of high
resolution spectroscopic information. Plasmonic structures have been demonstrated as a color
filter platform well-suited for CMOS integration due to the small mode volumes of plasmons
and the CMOS compatibility of many of the ma
terials that support them. Plasmonic hole and
slit array color filters have been demonstrated as
a viable alternative to dye-based filters for
RGB and CMYK color-filtering [3–8]. In additio
n to hole and slit array filters, many other
geometries have been explored as potential platforms for commercially viable plasmonic
color filters [9–12]. Recently, this
spate of development has been
translated into industrial
CMOS image sensor prototypes and are activel
y being considered for full commercialization
[13].
Previous work has demonstrated plasmonic hole and slit array color filters capable of
filtering the visible spectrum into three or four broad spectral bands (>100 nm) [3–8]. By
reducing the transmission bandwidth of the filters to less than 30 nm, CMOS image sensors
would gain the ability to perform multi- and hyper-spectral imaging without needing to rely
on algorithmic post-processing [9,14]. Multi- and hyper-spectral imaging is utilized in a wide
range of terrestrial and space applications, and pr
oviding portal devices
with this functionality
would have implications spanning food quality
control to space exploration [15,16]. Guided
mode resonance filters have been shown capable of generating high intensity, narrowband
filtering [17]. However, the millimeter-scale
of these filters is impractical for CMOS
integration [18]. Narrow bandwidth responses have also been reported in thin filmed multi-
layer plasmonic structures [19]. The peak wavelength of these structures is tuned via
changing the thickness of the intermediate dielectric layer, so the many lithographic steps
needed to patterns the tens to hundreds of sp
ectral bands make these structures infeasible for
multi- and hyper-spectral filtering. In this work, we have designed and prototyped a micron-
scale plasmonic color filter capable of narrow band-pass transmission (<30 nm). The filters
span the visible and near infrared spectrum and all utilize the same planar stack of materials
where the peak transmission wavelength is dete
rmined through a single lithographic step.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27387
Plasmonic color filters utilizing periodic arrays of subwavelength holes or nanoslits in
metal films enable efficient co
nversion of optical energy
between incident photons and
surface propagating two-dimensional charge
density waves, surface plasmon polaritons
(SPPs). Due to the permittivity
discontinuity at metal-dielect
ric surfaces, SPPs have an in-
plane momentum k
SPP
greater than that of light in free space k
o
. Patterned metal surfaces
including gratings, or arrays of holes or slits, allow the matching of momentum and thereby
enable efficient conversion of light into SPPs via scattering. The strength of interaction
between photons and SPPs can be tailored by chan
ging geometric factors such as the shape of
the scattering elements, and the symmetry and periodicity of the array as well as by selecting
the permittivity of the constituent materials [20].
In particular, periodic arrays of subwavelength apertures passing through a metal film
exhibit enhanced transmission exclusively at conditions corresponding to constructive mutual
interference between incident
light and SPPs traveling alo
ng the surface between adjacent
slits. In the case that the metallic layer is thick enough to be substantially opaque to incident
photons, the SPP mediated process is the dominant mode of transmission and the surface acts
as a band-pass color transmission filter. Such aperture arrays have been the topic of
substantial scientific interest due to these re
markable optical properties and their utility as a
testbed for studying fundamental light-matter interactions in plasmonic systems [7,21].
The dispersion of plasmonic propagating modes can be further engineered using metal-
clad slot waveguides, often realized as mult
ilayer stacks with a metal-insulator-metal (MIM)
configuration [8]. Such MIM stacks may suppor
t a multitude of polaritonic modes which lie
either inside or outside the “light cone,” that is
, with in-plane momentum either greater or less
than that of a photon with equal energy. This additional degree of freedom enables
substantially more complex optical transmission filter spectra enabling narrow bandwidth
suitable for multi- and hyperspectral color filtering applications [5].
2. Designing plasmonic color filters
Finite difference time domain methods (FDTD) were used to determine the transmission
spectra of different filter structures. Figure 1
illustrates the different types of transmission
filters and their spectral behavior. MIM have been used to make RGB color filters [5]. These
structures can be optimized to have narrowband
transmission, but as the structure is optimized
to minimize FWHM of the transmission peak, the intensity of the next highest order mode
increases. This trade-off can be lifted by intr
oducing a second MIM m
ode into the structure
that couples with the original MIM mode, leading to the suppression of the spurious
transmission. The multilayer slot-mode plasmoni
c filter (MSPF) investigated demonstrates a
narrow transmission bandwidth and spurious peak suppression, as shown in Fig. 1(b), and by
changing the periodicity of the slits, this filter can be swept across the entire visible spectrum.
Fig. 1. (a) Schematics of MIM and MIMIM filter st
ructures. All dark grey metal layers are Ag
and 70 nm thick, except for the 50 nm center me
tal layer of the MIMIM filter. All light grey
insulating layers are 70 nm of SiO2 (b) Comparison between MIM and MIMIM transmission
behavior shows similar FWHM
but enhanced suppression of the secondary peak in the
MIMIM case.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27388
The MSPFs were optimized using an incident
TM plane wave source. Optimizations were
conducted via a series of parameter sweeps th
at considered both the thicknesses and optical
indices of all the insulating and metallic layers, as well as the width and spacing of the milled
slits. The initial values for the thicknesses of the metallic layers were determined by
considering the skin and penetration depths of various metals. For a successful filter, the top
and bottom metallic layers of the structure must be sufficiently thick to be opaque across the
visible and near IR parts of the electromagnetic spectrum. Using data from Rakic et al. for Ag
as an example, the 1/e penetration depth (d
p
) of Ag was calculated to range from 12.9 nm –
16.8 nm across the visible spectrum. To prevent 98% of light from penetrating the structure,
the top and bottom layers must be at least 4 times d
p
. Therefore 68 nm was used as the initial
parameter sweep value when optimizing the system that utilized Ag.
Likewise, the starting point for the thickness of the insulating layers was approximated by
considering the propagating modes guided laterally within the structure. Numerically
determined dispersion curves derived from experimental optical constants of Ag and SiO
2
can
be used to determine the available modes within an MIM [8]. For SiO
2
thicknesses of less
than 100 nm, traditional photonic waveguide modes are cut off in an Ag/SiO
2
/Ag system, so
the waveguide only supports high-momentum surface plasmon modes. Therefore, the
parameter sweeps used 100 nm as the upper value restriction for the SiO
2
thickness of each
waveguide.
Iterating over the parameter sweep led to the final device structure, with alternating layers
of Ag and SiO
2
. Both SiO
2
layers were optimized to 70 nm, the top and bottom Ag layers are
70 nm and the spacer layer is 50 nm. The width of the slit is 50 nm for all filters and the slit
periodicities investigated vary from 250 nm to 550.
The position of the transmission peak varies linearly with the periodicity of the slits and,
as shown in Fig. 2(a), peak position can be sw
ept across the visible and near IR spectrum.
Therefore, just by varying the inter-slit pitch, a
series of MSPFs with the same layer materials
and thicknesses can be used as a color filter
across a wide range of the spectrum. The FWHM
of the transmission spectra are about 20 nm
on average with no peak exceeding 28 nm, as
shown in Fig. 2(b). Additionally, the overall transmission of the side-lobe peak does not
exceed 11% of that of the prim
ary peak in the visible portion of the spectrum, and does not
exceed 25% of the primar
y peak intensity in all filters in
vestigated. While it is possible to
design MSPF filters using other materials systems, the Ag/SiO2
system was found to be most
optimal. For example, the transmission spectrum
of an optimized Al/SiO2 MSPF possesses a
FWHM of 34 nm and peak transmission of just over 20%.
Fig. 2. (a) Superposition of the tr
ansmission behavior of filters with
varying slit pitches. As slit
pitch increases the narrowband transmission peak
is controllably shifted to longer wavelengths
(b) The relationship between FWHM, peak position,
and sideband to peak ratio. The blue axis
illustrates the ratio of FHWM to the peak position. The dashed line sets the threshold of a
30nm FWHM, and the dotted line illustrates the ratio
of the transmission peak’s FWHM to the
peak position. The dotted line is beneath the
dashed line for the entire visible spectrum,
indicating that all filters fulfill the criteria for
hyperspectral imaging. The orange dotted line
illustrates the ratio between the sideband and main
intensity peaks, showing the best filters are
also in the visible part of the spectrum.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27389
3. Experimental verification
MSPFs were fabricated by depositing
alternating layers of Ag and SiO
2
in an electron beam
evaporator onto a solvent cleaned fused silica slide and then subsequently milled using a
focused ion beam (FIB). The 50 nm slit milled into a 330 nm structure is a prohibitively
demanding aspect ratio for a FIB trench mill. For a set of proof-of-concept filters, these
demanding design conditions can be relaxed by
considering filters only towards the lower
energy portion of the visible spectrum. For a slit width of 120 nm, the suppression of the
spurious transmission peak is retained and th
e FWHM of the primary transmission peak only
takes a 25 nm hit.
When Ag is deposited on SiO
2
in an electron beam evaporat
or, the Ag films grow with a
columnar growth mechan
ism [22]. These films are rough, which increases plasmonic loss,
thereby reducing overall transmission intensity
of the filter [23]. The roughness of Ag
deposited on SiO
2
is even more problematic in a multilayer structure like the MSPF because
the roughness of each Ag layer compounds. A r
ough substrate increases the roughness of the
film deposited on it due to differences in atomic flux received by areas of the film with
positive and negative curvatures that are larger
than can be compensated for by surface
diffusion [22]. Because the SiO
2
conformally deposits on the underlying Ag layer, each Ag
layer sees a progressively rougher substrate, l
eading to a very rough top surface of the MSPF.
By utilizing a seed layer of AgO deposited onto each SiO
2
surface, a much smoother Ag
film can be deposited [24]. The AgO is deposited by slowly electron beam evaporating Ag at
a rate of 0.1 A/s in a chamber with an O2 pressure of 9.5x10
−
5
torr. Once 2 nm of AgO are on
the surface of the SiO
2
, the deposition is paused and the AgO is held under vacuum for 10
minutes. Because AgO is not vacuum stable, the oxygen is pumped out of the film, leaving a
thin Ag layer on the surface of the SiO
2
[24]. The deposition is then resumed and the rest of
the Ag layer is deposited at in a chamber with pressure 2.3x10
−
6
torr and no oxygen flow. The
roughness of Ag films deposited with this method was measured to have an RMS of 2.56 nm
and the top Ag surface of a multilayer deposited
with the AgO growth me
thod has an RMS of
2.92 nm. Each Ag layer was deposited using this method.
To further protect the integrity of the filter,
a sacrificial layer was put on the top Ag
surface. First a 90 nm layer of 950 A2 PMMA
was then spun onto the top surface of the
MSPF and then another 70 nm layer of Ag was deposited on top of the PMMA. The
sacrificial layer protects the top Ag film of th
e MSPF by confining the worst of the ion beam
damage to the surface of the sacr
ificial layer, rather than th
e surface of the MSPF. To utilize
the best possible resolution of the ion beam, the 130 nm wide slits are milled in FEI Versa
FIB, at 30 kV and 1.5 pA [
25]. The high accelerating voltage
and low beam current help
compensate for the high aspect ratio of the filter structure. Multi-pass milling is used to
reduce the taper of the slits—first a rectangl
e is milled, followed by a frame around the
perimeter to better define the edges and clean off reposition within the slit. The sacrificial
layer is then removed using a heated solvent bath.
The fabricated filters are then measured us
ing a supercontinuum laser with monocrometer
set-up that allows for the sample to be illuminated with a narrow bandwidth of incident
radiation. A 5X objective takes the collimated light and focuses it down to a 10 um spot size
with a 5.7° angular spread from the incident no
rmal. The incident beam is shined on the 30
um x 30 um MSPFs. Squares equal in size to
the filters were milled 100 um away from each
filter and are used to determine the intensity of
the laser. All transmitted power was collected
by a Si photodiode that was affixed behind the substrate in which the filters and normalization
squares have been milled. The experimental response of each filter was determined by
normalizing the light transmitted through the filter by that transmitted through its
corresponding normalization square.
Figure 3(a) shows the experimental transmission response of a prototype filter that has an
inter-slit pitch of 475 nm. A cross section of th
e prototype filter was milled using the FIB.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27390
The micrograph of the filter’s cross-section, shown in Fig. 3(c) reveals that there is a slight
taper to the filter structure. Using FDTD simulations, we can compute the transmission
behavior of filters with a progressively in
creasing sidewall taper. The results of these
simulations, shown in Fig. 3(b) illustrate the importance of the slit sidewalls on the overall
behavior of the structure. Using the information gathered from the FDTD simulations, it was
determined that to maintain filtering behavior with side lobe suppression, the sidewalls of the
slit could not possess greater than a 5° taper. The side lobe in the experimental transmission is
due in part to the 13.7° taper in the fabricated filter. In future work, this taper can be
eliminated by utilizing nanoimprint lithography couple with advanced thin film deposition
techniques to produce MSPF filters with le
ss taper. Additionally, diffractively coupled
resonances are angle sensitive, so the angular spread of the incident beam also led to slight
broadening and side lobe enhancement.
The polarization response was also experimentally confirmed to match the simulated
predications, as shown in Figs. 4(a) and 4(b).
While the fabricated filters were polarization
sensitive, Figs. 4(c) and 4(d) illustrate that
the polarization dependence of MSPFs can be
eliminated entirely by introducing a second arra
y of slits perpendicular to the original slit
array, allowing any incident k-vector to eff
ectively couple into th
e plasmon modes of the
filter.
Fig. 3. (a) Experimentally determined tran
smission of a single MIMIM filter (b) Simulated
dependence of transmission on taper of slits (c
) Top down and cross-s
ectional SEMs of the
MSPF filter (d) TEM micrograph showing the la
yer thicknesses and r
oughness of the five
layers of the filter.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27391
Fig. 4. (a) Simulated polarization response varying from 0° (blue) to 90° (green) (b) The
simulated polarization response was confirmed e
xperimentally, with a 0° measurement (blue),
a 30° measurement (teal), and a 45° measurement (green) (c) A transmission spectrum for a
crossed MSPF structure with incident polarizati
on oriented at 30° from the original grating
normal (d) Superposition of transmission curv
es spanning 0° to 90° for a crossed MSPF
structure showing complete polarization insensitivity.
4. Analytical analysis
A series of FDTD simulations was performed in order to study the electromagnetic modes
that propagate in the MIMIM device stack. To study fundamental properties of these guided
mods, light is coupled into the stack using plane wave illumination of a single slit, as opposed
to the periodic array of slits that comprise the MSPF structures. A series of 211 single
frequency TM plane wave excitations were used to sweep over the en
tire visible spectrum.
Complex vector field data was collected by fine
ly meshed monitors capturing the EM time
evolution of each of the FDTD simulations. This data set can be compressed by taking a
discrete-time Fourier transform at runtime whic
h transforms the time-harmonic field data in
the frequency domain.
In Fig. 5(a), a single electric field component is plotted from the compressed data set of a
sample simulation conducted at an excitation energy of 1.88 eV. The spatial mapping of the
electric field depicts light scattered by a slit at z = 0 into multiple modes propagating in the z-
direction. These modes have high field intensity within both SiO
2
layers as well as the top and
bottom Ag surfaces. The sp
atial mapping of the electric field
indicates that slit preferentially
couples energy into the topmost SiO
2
layer. Coupling between the two dielectric layers is also
apparent, as a characteristic beating pattern is
observed indicating that power is oscillating
between the two MIMs. This result was expected physically—the spacer layer between the
two insulating layers is thinner than the skin depth of Ag at this photon energy.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27392
Fig. 5. (a) Fields resulting after a TM plane wave,
incident in the negative x direction, scatters
at a single slit and couples into MIM modes propa
gating in the z direction (b) An FFT of the
fields at a single excitation frequency shows
multiple modes resolved by wavenumber (c)
Cross section through the FFT data at a select
wavenumber reveals the spatial mode profile.
To better determine the natures of the various modes within the MSPF, a second Fourier
Transform was performed. Using Eq. (1), an FFT was taken over the propagation direction of
the modes (z), thereby moving the phasor dir
ect space data set into momentum space (i.e. “k-
space”).
()
()
ikz
f
kFzedz
−
=
(1)
The results of this FFT can be plotted, as shown in Fig. 5(b), to illustrate the various modes
excited by the single frequency source. The ver
tical streaks in the sp
ectral power map indicate
there are modes at multiple propagation wavevectors (k
z
) at this excitation frequency. Vertical
cross-sections through this data set describe the spatial field profile along the x-axis
transverse to the propagation di
rection, and can provide insight into the nature of particular
modes. For example, the dark red streak at k
z
= 2.7
μ
m
−
1
is localized on the metallic surface
of the MSPF filter, and therefore this featur
e can be identified as an SPP. Indeed, this
corresponds to the expected wa
venumber for an SPP on the surface of an Ag film at this
excitation frequency.
Close inspection of the intensity patterns of the hot spots in Fig. 5(b) reveals that the
region of high spectral power on the surface of the filter at wavenumbers less than 2.7
μ
m
−
1
corresponds to unbound modes that do not contribute to filter behavior. After discarding the
evanescently reflecting quasi-modes, the nature of the remaining modes can be determined by
plotting the linear intensity variations at each wavenumber that possesses a streak of high
spectral power, superimposing these linear intensity variations on the same axis yields the
plot shown in Fig. 5(c). This plot depicts the two strongest plasmon modes in the structure:
the mode on the top surface of the filter and the mode spanning the insulating layers
contained within the filter.
The predominant surface mode, shown in blue in Fig. 5(c),
corresponds to SPPs excited at th
e top Ag surface of the filter. The other excitation is a super-
mode corresponding to a coupling of the two MIM modes generated within each of the two
insulating layers in the structure. This is the mode that was implied in the spatial field map in
Fig. 5(a) is now clearly depict
ed in Fig. 5(c), which reveals that the two MIM modes within
the super-mode are coupled because of strong field overlap within the 50 nm Ag spacer.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27393
The behavior of the energy propagating through the filter can be determined by the FFT
analysis, but it does not indicate how these modes contribute to the overall behavior of the
filter. The contribution of each mode to the unsuppressed transmission peak can be
determined by normalizing the transmission curv
es of the periodic grating MSPF structures
over the dispersion curve of each mode and the pitches of the filter gratings.
The dispersion curve of each plasmon mode can be determined by constructing a
dispersion curve for the MSPF. The dispersion curve shown in Fig. 6(a) was constructed by
using the Fourier Transformed k-space data sets
and plotting the power of the modes at each
spectral frequency as a function of energy. The two branches on curve correspond to the
bottom side SPP and the metal-insulator-meta
l-insulator-metal (MIMIM) super-mode, and
can be mapped to the frequency va
lues that correspond
to these modes in Fig. 5(c). The lower
intensity signal plotted to the left of the branches corresponds to the unbound quasi-modes
bouncing off the
surface of the filter. From the data contained within this plot, the dispersion
curve for each of the two modes can
be analytically determined.
Fig. 6. (a) By taking FFTs of a sweep of singl
e frequency excitations, a dispersion curve can be
constructed that illustrates the
behavior of both active modes
(b) Universal curve analysis
confirms that the SPP mode on the top surface of
the MSPF filter is predominantly responsible
for the filters transmission behavior. The various
colors of the transmission curves correspond
to different peak intensity positions that have all been normalized by the SPP dispersion curve.
G1 is the lowest order recipr
ocal lattice vector and a corres
ponds to the pitch of the slits.
The transmission behavior was first normalized over the pitch of the filter grating to lift
the dependence of the transmission curves on that
characteristic of the f
ilters [21]. Figure 6(b)
shows the transmission curves no
rmalized by the SPP dispersi
on curve, leading to curves
aligned between 0.5 and 1, with the transmission minima collapsing at 1 on the normalized
axis [21]. This behavior illustrates that th
e SPP mode satisfies the momentum matching
condition required for it to contribute to the transmission behavior of the filter, with the
MIMIM super-mode acting as a supplementary suppression to remove the second highest
order peak. This analysis confirms the origin of the spurious peak in the experimental results.
As the slits are tapered, the difference between the lengths of the two channels increases,
which affects the interference between the two
modes. The change in interference behavior
reduces the filtering efficiency of this mode and allows multiple orders of modes to propagate
through the structure, leading to a reduced suppression of the spurious peak.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27394
5. Conclusions
A plasmonic color filter with a single narrowband transmission response was designed using
FDTD and fabricated to confirm the simulated
response. The filter is readily amenable to
device integration, with a size well-matched to
state of the art CMOS image sensors. The
plasmonic filter utilizes a geometry that flexibly allows for precise selection of the spectral
bands of interest, allowing for portable electronic devices to be capable of multi- and
hyperspectral imaging. The behavior of this filter was analytically determined to arise from a
combination of SPP excitations–the surface SPP
mode leads to the enhanced transmission
behavior associated with subwavelength plas
monic filters, while the slightly asymmetric
MIM super-mode leads to the suppression of the spurious transmission peak that arises in
other narrowband plasmonic filter geometries. The
MSPF is inherently gated, and this feature
will be capitalized on in future work by incorporating transparent conducting oxides into this
geometry to create tunable narrowband color filters spanning both the visible and near
infrared parts of the spectrum.
Funding
Sony Corporation; the Hybrid Nanophotonics
Multidisciplinary University Research Initiative
Grant (Air Force Office of Scientific Research FA9550-12-1-0024); Northrop Grumman
Corporation.
Acknowledgments
LAS acknowledges support from the Resnick Sustai
nability Institute at Caltech. DF gratefully
acknowledges helpful discussions with Michelle Sherrott, Max Jones, Matt Sullivan, and
Charles Shaw. The facilities of the Kavli Nanoscience Institute (KNI) at Caltech are
gratefully acknowledged.
Vol.
25,
No.
22
| 30
Oct
2017
| OPTICS
EXPRESS
27395