APPLIED PHYSICAL
SCIENCES
Metasurface-generated complex 3-dimensional optical
fields for interference lithography
Seyedeh Mahsa Kamali
a,b,c,d
, Ehsan Arbabi
a,b
, Hyounghan Kwon
a,b
, and Andrei Faraon
a,b,1
a
T. J. Watson Laboratory of Applied Physics, California Institute of Technology, Pasadena, CA 91125;
b
Kavli Nanoscience Institute, California Institute of
Technology, Pasadena, CA 91125;
c
Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, CA 94720; and
d
Department of Bioengineering, University of California, Berkeley, CA 94720
Edited by Naomi J. Halas, Rice University, Houston, TX, and approved September 18, 2019 (received for review May 15, 2019)
Fast, large-scale, and robust 3-dimensional (3D) fabrication tech-
niques for patterning a variety of structures with submicrometer
resolution are important in many areas of science and technology
such as photonics, electronics, and mechanics with a wide range
of applications from tissue engineering to nanoarchitected mate-
rials. From several promising 3D manufacturing techniques for
realizing different classes of structures suitable for various appli-
cations, interference lithography with diffractive masks stands
out for its potential to fabricate complex structures at fast speeds.
However, the interference lithography masks demonstrated gen-
erally suffer from limitations in terms of the patterns that can be
generated. To overcome some of these limitations, here we pro-
pose the metasurface-mask–assisted 3D nanofabrication which
provides great freedom in patterning various periodic structures.
To showcase the versatility of this platform, we design meta-
surface masks that generate exotic periodic lattices like gyroid,
rotated cubic, and diamond structures. As a proof of concept,
we experimentally demonstrate a diffractive element that can
generate the diamond lattice.
metasurface
|
3D printing
|
interference lithography
|
beam shaping
|
nanophotonics
I
n the 20th century, nanofabrication techniques have truly rev-
olutionized the electronics and photonics industries. As the
trend continues, currently there is high interest in fast fabrication
of large-scale 3-dimensional (3D) lattices with nanoscale resolu-
tion. These structures have applications in various areas includ-
ing novel engineered materials (1), microelectromechanical
systems (2), nanoarchitected materials (3, 4), microelectronics
(5), tissue engineering and biomedical engineering (6), micro-
fuel cell development (7), optics (8), and micro- and nanofluidics
(9). Different 3D manufacturing techniques have been pro-
posed for different applications, including approaches based
on self-assembly methods (10, 11), holographic lithography (5,
12–16), multiple-exposures lithography (17), controlled chem-
ical etching (18), and various additive manufacturing meth-
ods (19) like stereolithography (20) and laser- or ink-based
direct writing (21) among many others. Each of these tech-
niques provides new capabilities for fabricating different classes
of 3D structures for different applications beyond traditional
2D photolithography steppers. Nevertheless, none could reach
the performance of traditional steppers in simultaneously pro-
viding a high-speed, large-scale and scalable lithography, a
simple and robust experimental setup, high yield, and defect-
free structures. Here, we introduce the concept of large-scale
metasurface-assisted 3D lithography, schematically shown in
Fig. 1, to circumvent some of these shortcomings. The method
is based on using metasurfaces as photolithography masks to
generate exotic 3D structures and also take advantage of the
traditional stepper technique in fabricating fast, large-scale, 3D
patterns with nanometer resolution through a relatively sim-
ple and robust process. The metasurface mask (which we call
metamask from here on) provides control over the complex
coefficients of 2 orthogonal polarizations for various diffraction
orders, resulting in the realization of exotic 3D patterns like the
gyroid, diamond, or cubic. As a proof of concept, we experi-
mentally demonstrate the diamond pattern through design and
fabrication of a metamask. We should note that conformal masks
have been used before for fabricating 3D patterns in photoresists
(22, 23). However, they suffer from limited diffraction efficien-
cies (which result in low-contrast 3D structures) and have limited
degrees of freedom in generating desired 3D patterns.
Optical metasurfaces are 2D arrangements of scatterers that
are designed to modify different characteristics of light such
as its wavefront, polarization, intensity distribution, or spec-
trum with subwavelength resolution (24–28). By proper design
of the scatterers, different characteristics of the incident light
can be engineered, and therefore different optical elements like
gratings, lenses, holograms, waveplates, polarizers, and spectral
filters can be realized (29–37). Furthermore, a single metasur-
face can provide novel functionalities, which, if at all possible,
would require a combination of complex optical elements to
implement (38–43). Here, we demonstrate that by exploiting the
metasurface capabilities in modifying the phase, intensity, and
polarization of the optical wavefront, different metamasks can
be designed to generate 3D patterns like the gyroid, cubic, or
diamond structures.
Fig. 2
A
schematically shows a metamask and how it gener-
ates different diffraction orders that interfere to realize a specific
Significance
Fast submicrometer-scale 3D printing techniques are of inter-
est for various applications ranging from photonics and elec-
tronics to tissue engineering. Interference lithography is a
versatile 3D printing method with the ability to generate com-
plicated nanoscale structures. Its application, however, has
been hindered by either the complicated setups in multi-
beam lithography that cause sensitivity and impede scal-
ability or the limited level of control over the fabricated
structure achievable with mask-assisted processes. Here, we
show that metasurface masks can generate complex volu-
metric intensity distributions with submicrometer scales for
fast and scalable 3D printing. These results push the limits
of optical devices in controlling the light intensity distribu-
tion and significantly increase the realm of possibilities for 3D
printing.
Author contributions: S.M.K. and A.F. designed research; S.M.K., E.A., and H.K. performed
research; S.M.K. contributed new reagents/analytic tools; S.M.K. and E.A. analyzed data;
and S.M.K., E.A., and A.F. wrote the paper.
y
Competing interest statement: S.M.K., E.A, and A.F. are inventors of US patent applica-
tion US20190173191A1 that covers the use of metasurface masks for 3D beam shaping.
The authors declare no other competing interests.
y
This article is a PNAS Direct Submission.
y
This open access article is distributed under
Creative Commons Attribution-NonCommercial-
NoDerivatives License 4.0 (CC BY-NC-ND)
.
y
1
To whom correspondence may be addressed. Email: faraon@caltech.edu.
y
This article contains supporting information online at
www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1908382116/-/DCSupplemental
.
y
First published October 7, 2019.
www.pnas.org/cgi/doi/10.1073/pnas.1908382116
PNAS
|
October 22, 2019
|
vol. 116
|
no. 43
|
21379–21384
Fig. 1.
Concept of large-scale metamask-assisted 3D fabrication. Shown is
a schematic illustration of large-scale metasurface-assisted 3D printing. A
large metamask (
∼
1 cm
2
) is designed and used as a photolithography mask
to create the desired 3D pattern in the photoresist. The large (
∼
1 cm
2
wide
and
∼
10
μ
m thick) 3D pattern is generated inside the photoresist through
single-photon lithography. Similar to a stepper, a linear stage could be used
here to create large-scale 3D periodic patterns. The rendering just serves to
demonstrate the concept and the sizes are not to scale.
desired intensity pattern inside a transparent photoresist. It was
previously shown that all 14 Bravais lattices can be formed by
interference of 4 noncoplanar beams (44). Therefore, the meta-
mask is in principle capable of generating all Bravais lattices in
the resist. In Fig. 2, the illumination is assumed to be a 532-nm
laser with linear polarization. The photoresist is assumed to be a
sensitized SU-8, with the ability to form a solid structure under
532-nm photoexposure (13, 22, 45). The metamask is assumed
to provide the desired amplitude and phase masks for the
x
- and
y
-polarized incident light.
The metamask generates different plane waves (diffraction
orders) with different propagation directions that are deter-
mined by its lateral periods. The electric field associated with the
n
th plane wave can be written as
~
E
n
e
−
j
~
k
n
·
~
r
. The overall electric
field resulting from the interference of these plane waves can be
written as
~
E
=
E
x
~
x
+
E
y
~
y
+
E
z
~
z
,
[1]
where
E
x
=
N
∑
n
=1
E
nx
e
−
j
~
k
n
·
~
r
E
y
=
N
∑
n
=1
E
ny
e
−
j
~
k
n
·
~
r
E
z
=
N
∑
n
=1
(
−
~
k
n
·
~
x
~
k
n
·
~
z
E
nx
e
−
j
~
k
n
·
~
r
+
−
~
k
n
·
~
y
~
k
n
·
~
z
E
ny
e
−
j
~
k
n
·
~
r
)
.
[2]
The 3D intensity profile is defined as
I
=
1
2
η
|
~
E
|
2
,
[3]
where
η
=
√
μ/
is the characteristic impedance of the prop-
agating medium. In single-photon lithography the photoresist
polymerization is proportional to the exposure intensity, and
therefore the 3D structure is generally formed for intensities
above a specific threshold value, defined here as
I
th
.
The main advantage of using metamasks is that the complex
coefficients of different
x
- and
y
-polarized diffraction orders (
E
nx
and
E
ny
) can be controlled independently and at will. Therefore,
it provides more degrees of freedom to define more exotic 3D
structures like the gyroid and diamond patterns.
The design process of the metamask to generate a spe-
cific 3D periodic lattice is as follows: First, the lateral peri-
ods of the 3D structure must be properly selected such that
the intensity interference pattern is fully periodic in all 3
dimensions. These lateral dimensions play a critical role as
they define the number of diffraction orders and their direc-
tions, as well as the in-depth periodicity. After selecting
the appropriate lattice constants, the corresponding diffrac-
tion order coefficients are optimized to generate the desired
3D pattern. Finally, the metamask is designed and imple-
mented through a high-contrast dielectric transmittarray. To
showcase the capability of metamasks, we demonstrate meta-
masks that generate different lattices like the gyroid, cubic,
and diamond patterns. The in-plane dimensions and the corre-
sponding in-depth periods are noted for these structures in
SI
Appendix
, Fig. S1
.
The level-set representation of the 3 structures is given in
SI Appendix
, in the form of
f
(
x
,
y
,
z
)
−
t
>
0
. In these equa-
tions, the parameter
t
is used to control the volume fraction of
the structure, as it is assumed to be solid for
f
(
x
,
y
,
z
)
−
t
>
0
.
This parameter can be controlled experimentally by adjusting the
exposure threshold of the photoresist. Here, we have assumed it
to be 0.25, 0.15, and 0.3 for defining the gyroid, rotated cubic,
and diamond patterns, respectively. See
SI Appendix
, Fig. S2
for the defined target patterns. For realizing the patterns with
amplitude and phase masks, we used a global optimization tech-
nique to find the complex coefficients of different diffraction
orders,
E
nx
and
E
ny
, which are given in
SI Appendix
, Fig. S2
.
The optimized amplitude and phase masks for each pattern,
shown in Fig. 2
B
–
D
,
Left
, are calculated from the optimized
diffraction order coefficients using Eq.
2
at
z
=
0 plane. The
3D intensity patterns are then calculated using Eq.
2
and are
shown in Fig. 2
B
–
D
,
Center
. The corresponding 3D structures,
assuming
I
th
= 0
.
5
, are shown in Fig. 2
B
–
D
,
Right
. To deter-
mine the degree of similarity between the achieved and desired
patterns, we used a fitness factor (
FF
) defined as the frac-
tion of voxels in 1 unit cell that match the 3D target structure.
The fitness factors for the gyroid, rotated cubic, and diamond
patterns are 97
%
, 82
%
, and 93
%
, respectively (see
Materials
and Methods
for simulation details). It is worth noting that the
gyroid lattice discussed here is an example of a chiral structure,
showing the capability of the developed technique in genera-
tion of such chiral patterns (46). Other chiral structures with
interesting optical properties [such as spiral lattices (47)] can
also be designed using this platform as shown in
SI Appendix
,
Fig. S6
.
To realize the diamond metamask for 532-nm wavelength,
we used a metasurface platform composed of cuboid-shaped
crystalline silicon (cSi) nanoposts embedded in an SU-8 pro-
tecting layer and resting on a quartz substrate. A schematic
of the metasurface platform is shown in Fig. 3
A
. Transmis-
sion phases of the
x
- and
y
-polarized light can be fully and
independently controlled from 0 to 2
π
by changing the in-
plane dimensions of the nanoposts (39). The cSi nanoposts are
291 nm tall and fully embedded in the SU-8 layer, and the
lattice constant is 250 nm. A periodic array of such cuboid-
shaped nanoposts was simulated to find the transmission phases,
which are plotted in Fig. 3
B
(see
Materials and Methods
for
simulation details and
SI Appendix
, Fig. S3
B
for transmis-
sion powers). The diamond phase masks shown in Fig. 2
D
are sampled at 4 points with a 250-nm period, and the cor-
responding nanoposts are shown in Fig. 3
B
with black circles
(see
SI Appendix
, Fig. S3
A
for the sampling points). The input
21380
|
www.pnas.org/cgi/doi/10.1073/pnas.1908382116
Kamali et al.
APPLIED PHYSICAL
SCIENCES
t
y
|
t
y
|
|
t
x
|
t
y
479 nm
|
t
y
|
t
x
rad/(2
S
)
1
0
Transmission
1
0
Gyroid pattern
Intensity pattern
Final 3D structure
2
u
479 nm
Top view
x
z
|
t
x
|E
x
e
i
t
x
E
x
E
y
|
t
y
|E
y
e
i
t
y
|
t
x
|
t
x
u
408 nm
408 nm
rad/(2
S
)
1
0
Transmission
1
0
Intensity pattern
Final 3D structure
u
408 nm
u
408 nm
Top view
2
u
500 nm
Intensity pattern
Final 3D structure
Top view
0
th
-1
st
+1
st
Interference
region
532 nm planewave
illumination
Rotated cubic pattern
Diamond pattern
AB
C
D
Meta-mask
Side view
Photoresist
|
t
x
|
t
y
500 nm
|
t
y
|
t
x
rad/(2
S
)
1
0
Transmission
1
0
Intensity (a. u.)
01
I
th
I
th
=0.5
Intensity (a. u.)
0
1
I
th
I
th
=0.5
Intensity (a. u.)
01
I
th
I
th
=0.5
x
y
x
y
x
y
z
x
y
x
y
x
y
z
x
y
x
y
z
x
y
Fig. 2.
Design of metamasks for generating desired 3D periodic patterns through interference. (
A
,
Top
) Schematic illustration of a metamask generating
specific diffraction orders with designed complex coefficients to make a desired 3D pattern in the photoresist. (
A
,
Bottom
) Schematic of a metamask
with amplitude, polarization, and phase control to make a desired 3D pattern. (
B–D
)
x
- and
y
-polarized transmission amplitudes and phases of dif-
ferent metamasks designed to create the gyroid, rotated cubic, and diamond patterns in the photoresist. The input light is assumed to be linearly
polarized with
|
E
x
|
/
|
E
y
|
=
0
.
85,
|
E
x
|
/
|
E
y
|
=
0
.
97, and
|
E
x
|
/
|
E
y
|
=
1 for the 3 different patterns, respectively. (
B–D
,
Left
) Transmission amplitude and
phases of the designed metamasks. (
B–D
,
Center
) generated 3D intensity patterns in the sensitized SU-8 photoresist under 532-nm laser illumination.
(
B–D
,
Right
) Bird’s-eye view and top view of the expected 3D structures formed in the negative photoresist (sensitized SU-8) assuming a specific inten-
sity threshold. Here, it is assumed that the regions with intensity values above 0.5 will be polymerized in the resist, and areas below this level are
developed.
polarization is chosen such that
|
E
x
|
/
|
E
y
|
is equal to
<
|
t
x
|
>
/<
|
t
y
|
>
, where
<
·
>
denotes averaging over a unit cell area.
The full-wave simulated 3D intensity distribution and the cor-
responding periodic 3D structure are shown in Fig. 3
C
for
this initial design. The fitness factor of this initial design is
84
%
and the total simulated transmission efficiency is 74
%
(see
Materials and Methods
for full-wave simulation details and
SI
Appendix
, Fig. S4
A
for the phase and amplitude masks). To
improve the degree of similarity of the achieved and desired
structures, we used this design as a starting point and fur-
ther optimized the nanoposts’ widths through a global opti-
mization method. Considering the diagonal symmetry of the
diamond metamask, the optimization parameters were reduced
to 4 values (widths of the 2 different nanoposts). The opti-
mized values are shown in Fig. 3
B
with black stars. The full-
wave simulated 3D intensity profile and the resulting structure
are shown in Fig. 3
D
in a volume equal to
2
×
2
×
2
periods
(see
SI Appendix
, Fig. S4
B
for the corresponding phase and
Kamali et al.
PNAS
|
October 22, 2019
|
vol. 116
|
no. 43
|
21381
x
z
|
t
x
|E
x
e
i
t
x
E
x
E
y
|
t
y
|E
y
e
i
t
y
Phase/(2
S
)
0
1
t
y
t
x
Optimization
500 nm
2
u
500 nm
500 nm
2
u
500 nm
I
th
=0.5
I
th
=0.5
x
z
y
D
x
D
y
cSi
SU-8
Quartz
Initial design
Optimized design
CD
A
B
Intensity (a. u.)
0
1
I
th
Intensity (a. u.)
0
1
I
th
60
180
120
120
180
D
x
(nm)
D
y
(nm)
60
60
180
120
120
180
D
x
(nm)
D
y
(nm)
60
Fig. 3.
Realization of the diamond pattern metamask with nanoposts. (
A
) Schematic drawing of different views of a uniform array of rectangular cross-
section cSi nanoposts arranged in a square lattice resting on a quartz substrate and covered by an SU-8 layer. Tuning the in-plane dimensions of nanoposts,
D
x
and D
y
, allows for independent control of the transmission phases of
x
- and
y
-polarized light at 532 nm. (
B
) Transmission phases of the
x
- and
y
-polarized
light at 532 nm for the uniform array shown in
A
, as functions of the nanopost widths. The nanopost’s height is 291 nm and the lattice constant is 250 nm. (
C
)
The initial diamond metamask is designed through sampling the phase diagrams shown in Fig. 2
D
at 4 points. The corresponding nanopost dimensions are
shown in
B
with black circles. The full-wave simulated 3D intensity pattern and the corresponding 3D structure demonstrate an 84
%
similarity compared to
the target diamond pattern. (
D
) The nanopost dimensions are further optimized to realize a 90
%
similarity with the target diamond pattern. The optimized
nanopost dimensions are shown in
B
with black stars. All simulations are performed at the wavelength of 532 nm. cSi: crystalline silicon. See
Materials and
Methods
for simulation details.
amplitude masks). The fitness factor of the final optimized struc-
ture is 90
%
and the total simulated transmission efficiency is
82
%
. It is worth noting that the optimized metamask solution
is not unique and various initial points or optimization tech-
niques can result in different optimized designs. Details of the
simulation and optimization steps are discussed in
Materials
and Methods
.
The metamask is fabricated using standard nanofabrication
techniques (see
Materials and Methods
for fabrication details).
Fig. 4
A, Top
shows an optical image of the final fabricated device.
A scanning electron micrograph of a part of the fabricated meta-
mask before being capped with the SU-8 protecting layer is
shown in Fig. 4
A, Bottom
.
To characterize the fabricated metamask we used a confo-
cal microscopy setup with an oil immersion objective lens that
captures all of the excited diffraction orders. The sample was illu-
minated by a 514-nm laser beam, which was the closest available
laser line to 532 nm in the microscopy setup. The optical inten-
sity distribution was captured in multiple parallel planes with
∼
45-nm depth steps. Fig. 4
B
,
Right
shows the measured intensity
profiles at 2 sample cross-sections (
xy
and
xz
planes as schemati-
cally shown in Fig. 4
B
). See
Materials and Methods
for details of
the measurement procedure, and see
SI Appendix
, Fig. S5
for
measurement results over a larger area. The measured inten-
sity profiles are in good agreement with the simulated results
(simulated with the same illumination wavelength of 514 nm) as
shown in Fig. 4
B
,
Left
. We attribute nonuniformities and small
drifts in the
z
stack to sample vibrations and sample mount
tilt angles.
The metamask-assisted 3D fabrication platform enables a fast,
large-scale, and robust system for realizing exotic 3D structures.
The realized 3D structures with nanoscale resolution could have
properties with great potential. For example, the gyroid, spiral,
and diamond lattices show interesting optical properties (like
optical chirality and photonic band structures) (46–48). Also,
triply periodic gyroid, diamond, and cubic surfaces that have
superior mechanical properties can be fabricated (49, 50) using
the developed technique.
In this paper, we have focused on the generation of 3D
lattices with wavelength-scale periodicities as we envision that
the developed techniques will be mostly influential in this
area. Nevertheless, the same concept using similar or differ-
ent design methods can be extended to aperiodic 3D structures
with the use of aperiodic metasurface masks (51). Specifically,
one can think of these aperiodic metasurface masks as an
extension of the superpixels used in the design of the periodic
3D lattices. Aside from 3D printing purposes, such aperiodic
3D patterns would expand the applications of this platform
to other areas of science and technology like particle trap-
ping, 3D structured light illumination, holography, optical
microscopy, etc. Moreover, it is worth noting that high diffrac-
tion efficiencies provided by the metamasks result in high-
contrast well-defined 3D structures in the photoresist even
under fast single-photon lithography. Therefore, the use of
metamask-assisted platforms could eliminate the limited inten-
sity contrast issue faced in single-photon lithography that has
previously been addressed through multiphoton lithography
(52, 53). Furthermore, here we showcased the capability of
this platform through a single-layer metasurface, while cas-
caded metasurface layers could also be designed to provide
full and precise control over the complex coefficients of the
2 orthogonal polarization diffraction orders or provide addi-
tional control over the degrees of freedom like wavelength or
illumination angle.
21382
|
www.pnas.org/cgi/doi/10.1073/pnas.1908382116
Kamali et al.
APPLIED PHYSICAL
SCIENCES
2 mm
500 nm
Intensity
pattern
Meta-mask
x
y
z
514 nm laser
illumination
xz plane
xy plane
Simulation
xy plane
xz plane
Measurement
xy plane
xz plane
1
0
Intensity (a. u.)
A
B
Fig. 4.
Experimental characterization of the diamond metamask. (
A
,
Top
) Optical image of the fabricated optimized diamond metamask. A 5
×
5 array of
masks is fabricated and shown on top. (
A
,
Bottom
) Scanning electron microscope image of a portion of the mask before spin coating the SU-8 layer. (
B
) The
metamask is characterized under 514-nm laser illumination using a confocal microscopy setup (514-nm laser was the closest available laser line to 532 nm
in the confocal microscopy setup). Two measured cross-sections of the captured 3D intensity pattern (
Right
) are in good agreement with the simulated
results (
Left
).
In conclusion, here we introduce the concept of metamask-
assisted interference lithography, which could provide a fast
and robust technique for fabrication of exotic 3D periodic pat-
terns at large scales. We demonstrated the versatility of this
platform through designing different exotic 3D patterns like
the gyroid, rotated cubic, and diamond. Moreover, as a proof
of concept, we experimentally demonstrate the diamond pat-
tern through design and fabrication of the metamask. Besides
large-scale interference lithography with nanoscale resolution,
the presented concept can be used to generate complex
3-dimensional light fields for various applications including
structured light illumination, microscopy, particle trapping, and
holography.
Materials and Methods
Simulation and Optimization Procedure.
To find the optimized complex
coefficients of different diffraction orders and the input polarization for
generating the target 3D periodic pattern, we used a global particle swarm
optimization method. For the diamond and rotated cubic structures, we
forced the coefficients of unwanted diffraction orders to be zero. The tar-
get 3D patterns were defined with voxel sizes of
∼
10 nm
3
and
∼
13 nm
3
for
the rectangular (gyroid and diamond) and triangular (rotated cubic) lattices,
respectively.
To find the transmission powers and phases of a uniform array of
nanoposts under
x
- and
y
-polarized illumination, the rigorous coupled-
wave analysis (RCWA) technique was used (54).
x
- and
y
-polarized incident
plane waves at 532 nm wavelength were used as the excitation, and
the transmission powers and phases of the
x
- and
y
-polarized transmit-
ted waves were extracted. The subwavelength 250-nm lattice constant
in the SU-8 propagating medium results in the excitation of only the
zeroth-order diffracted light. The cSi layer was assumed to be 291 nm
thick. Refractive indexes at 532-nm wavelength were assumed as follows:
cSi, 4.136
−
1
j
0.01027; SU-8, 1.595; and quartz, 1.4607. The nanopost in-
plane dimensions (D
x
and D
y
) were swept such that the minimum feature
size and the gap size remain larger than 60 nm for relieving fabrication
constraints.
We used the finite-difference time-domain method (Lumerical) for sim-
ulating the metamasks realized with cSi nanoposts. The electric fields were
extracted on an
xy
plane
∼
20 nm above the nanoposts. We used the plane-
wave expansion (PWE) technique (55) to generate the 3D intensity profiles
and the 3D structures.
To optimize the nanoposts’ in-plane dimensions for the diamond meta-
mask, we used a global particle swarm optimization method with the fitness
factor (
FF
) target function. To find the 3D structure, we used the same
Lumerical simulation package with the PEW technique.
Sample Fabrication.
To define the pattern in cSi on quartz wafers, a Vis-
tec EBPG5200 e-beam lithography system and an
∼
300-nm-thick layer of
ZEP-520A positive electron-beam resist were used (spin coated at 5,000
rpm for 1 min). The pattern was developed in the resist developer (ZED-
N50 from Zeon Chemicals) for 3 min. The pattern was then transferred
into an
∼
50-nm-thick deposited Al
2
O
3
layer, by a lift-off process. The pat-
terned Al
2
O
3
hard mask was then used to dry etch the cSi layer in a
mixture of SF
6
and C
4
F
8
plasma. Finally, the Al
2
O
3
mask was removed in
a 1:1 solution of ammonium hydroxide and hydrogen peroxide at 80
◦
C.
Finally, a 2-
μ
m-thick layer of SU-8 protecting layer was spin coated on the
metamask.
Measurement Procedure.
The diamond metamask was measured using a
confocal microscopy setup (Zeiss LSM 710). A 100
×
oil immersion objective
lens (alpha Plan-Apochromat Oil DIC M27, numerical aperture [NA] = 1.46)
was used to capture all of the excited diffraction orders, as the diamond
mask has NA
∼
1.45 at 514-nm wavelength. We used Zeiss Immersol 518 F
with a refractive index of 1.518, which was the closest allowed oil in the
microscopy setup to the refractive index of SU-8 (1.595). We captured 3D
image stacks with in-plane pixel sizes of
∼
28 nm
2
and in-depth pixel sizes
of
∼
45 nm. We captured in-plane images as large as 15
μ
m
2
, shown in
SI
Appendix
, Fig. S5
. We should note that the resolution of the system is set
by the objective lens and is
∼
176 nm and
∼
482 nm in-plane and in-depth,
respectively.
ACKNOWLEDGMENTS.
This work was supported by the Department of
Energy (DOE) “Light-Material Interactions in Energy Conversion” Energy
Frontier Research Center funded by the US DOE, Office of Science, Office
of Basic Energy Sciences under Grant DE-SC0001293. The device nanofab-
rication was performed at the Kavli Nanoscience Institute at California
Institute of Technology (Caltech). Confocal imaging was performed at
the Caltech Biological Imaging Facility with the support of the Caltech
Beckman Institute and the Arnold and Mabel Beckman Foundation; we
thank Dr. Andres Collazo for his help in the confocal microscopy. We
thank Prof. Julia Greer, Dr. Travis Blake, Prof. Amir Arbabi, and Daniel
Bacon-Brown for fruitful discussion. We also thank Sana Kamali for the
artistic renders.
Kamali et al.
PNAS
|
October 22, 2019
|
vol. 116
|
no. 43
|
21383