Nanofluidic tuning of photonic crystal circuits
David Erickson,
*
Troy Rockwood, Teresa Emery, Axel Scherer, and Demetri Psaltis
Center for Optofluidic Integration, California Institute of Technology, Pasadena, California 91125
Received August 2, 2005; revised manuscript received September 15, 2005; accepted September 16, 2005
By integrating soft-lithography-based nanofluidics with silicon nanophotonics, we demonstrate dynamic,
liquid-based addressing and high
⌬n /n共⬃0.1兲 refractive index modulation of individual features within pho-
tonic structures at subwavelength length scales. We show ultracompact tunable spectral filtering through
nanofluidic targeting of a single row of holes within a planar photonic crystal. We accomplished this with an
optofluidic integration architecture comprising a nanophotonic layer, a nanofluidic delivery structure, and a
microfluidic control engine. Variants of this technique could enable dynamic reconfiguration of photonic cir-
cuits, selective introduction of optical nonlinearities, or delivery of single molecules into resonant cavities for
biodetection.
© 2006 Optical Society of America
OCIS codes: 230.7370, 250.5300.
Optical devices that incorporate liquids as a funda-
mental part of their structure can be traced at least
as far back as the 18th century, when rotating pools
of mercury were proposed as a simple technique to
create smooth spherical mirrors for use in reflecting
telescopes.
1
Modern microfluidics has enabled the de-
velopment of a present-day equivalent of such de-
vices centered on the marriage of fluidics and
optics
2–6
that we refer to as ‘optofluidics.’ Such de-
vices have substantial advantages for creating adap-
tive optical elements including high ⌬n/n, inherently
smooth optical interfaces, and thermal stabilization.
At present, techniques for local refractive index
modulation in photonic structures are limited to the
exploitation of relatively weak nonlinearities,
7
where
⌬n / n is of the order of 10
−3
or lower, and thus require
long interaction lengths, high operational power, or
the incorporation of resonant elements to enhance
the effect. Techniques such as mechanical deforma-
tion, thermo-optics, liquid crystal infusion, and liquid
fluid infusion, offer much higher effective ⌬n/n; how-
ever, they tend to be nonlocalized effects requiring or
resulting in modification over a large area, if not the
entire device. Thus whereas local tunability over
small interaction lengths requires the high ⌬n/n af-
forded by these global approaches, the ability to per-
form such manipulations with the submicrometer-
scale precision required for advanced photonic
devices remains elusive. The development of such a
technique could enable the creation of a new class of
ultracompact adaptable photonic circuits and
sensors.
Nanofluidics provides a solution that enables both
localized control and high refractive index modula-
tion. Here we demonstrate the integration of
multilayer soft-lithography
8
nanofluidics with silicon
nanophotonics and use it to address and tune fea-
tures within planar photonic crystals.
9
Photonic crys-
tals are attractive for controlling optical propagation
by introducing pre-engineered defects into an other-
wise regular lattice to create spectral filters, tight
bend waveguides, resonant cavities, and highly effi-
cient lasers. As a first step in the development of two-
dimensional reconfigurable photonic circuits, here we
demonstrate the nanofluidic addressing of a single
row of holes within a photonic crystal.
Figure 1 outlines our approach, which follows a
three-level architecture: the nanophotonic level, the
nanofluidic delivery level (which delivers liquids di-
rectly into the photonic structure), and the microflu-
idics control engine (which performs all fluidic ma-
nipulations). In this case the photonic level consists
of an array of 30 identical photonic crystal structures
defined through electron beam lithography and dry
etching in a silicon-on-insulator (SOI) substrate.
Each structure had a triangular lattice of holes con-
stant a=434 nm, hole radius r = 140 nm, and height
h=207 nm as shown in Fig. 2(a). A fabrication ap-
proach is used that is modified from the one de-
scribed previously,
10
in that we defined the pattern in
a negative flowable oxide resist that is left on after
processing to enhance bonding with the fluidic layer.
Additionally we do not etch the underside insulator
layer in favor of using it to enhance the mechanical
stability of the structure. Ridge waveguides extend
from the crystal array to the edge of the chip for op-
tical coupling into the crystal. In the experiments
Fig. 1. Nanofluidically tunable photonic structures. (a)
Exploded view of the opto-fluidic assembly showing the
photonic layer on bottom, fluidic layer in the middle, and
control layer on top. (b) Overview of device operation. The
microfluidic control engine mixes and dispenses liquid
plugs to the nanofluidic array. The nanofluidic structure
serves to deliver liquids directly into a targeted row of
holes in the photonic crystal enabling localized, high ⌬n/n
refractive index tuning. (c) Photograph of an assembled
chip. (d) Optical image showing overlay of nanochannels
with photonic crystal.
January 1, 2006 / Vol. 31, No. 1 / OPTICS LETTERS 59
0146-9592/06/010059-3/$0.00 © 2006 Optical Society of America
presented here we increase the radius of the holes
within the central row of the photonic crystal (which
is to be targeted fluidically) to 203 nm in order to in-
troduce a reduced index guided mode into the band-
gap created by the otherwise regular crystal lattice.
This geometry was selected on the basis of a series of
numerical experiments as having the highest spec-
tral contrast given the achievable change in refrac-
tive index.
To create the nanofluidic delivery layer we first de-
fine the nanofluidic system in positive relief on a
separate SOI substrate using processing identical to
that used to define the photonic crystals. To match up
with the photonic structure here we use an array of
200 nm tall, 350 nm wide channels spaced with a
5
m period. The microfluidics control engine is fab-
ricated using the multilayer soft-lithography tech-
niques developed by Unger et al.
8
In this case we su-
perimpose the fluidic layer of the microfluidics
control engine in photoresist on the previously de-
fined nanofluidic substrate, using standard photo-
lithographic processing. Once the master is formed,
the fluidic layer is cast in silicone elastomer. A valve
layer that provides the active control is similarly de-
fined, cast, and bonded to the fluidic layer utilizing
the techniques outlined in Ref. 8. The multiscale ap-
proach used here allows the fluidic operations, spe-
cifically switching and mixing, to be performed at the
tens of microns length scale where they can be ac-
complished more rapidly and the products simply
coupled into the nanofluidic system.
To align the system we varied the spacing between
the photonic structures by one half the lattice con-
stant, thereby ensuring alignment of at least one
nanochannel with the central row of at least one pho-
tonic crystal. A modified mask aligner was used to en-
sure the nanochannels were parallel with the photo-
nic crystals. In addition to manufacturing simplicity,
building the fluidic layer in a soft elastomer enables
conformal sealing over the photonic structures. To
enhance the bond between the fluidic and photonic
layers we exposed the RTV cast to a short duration
air plasma treatment immediately prior to assembly.
To directly examine the channel alignment and
seal integrity at the nanoscale we infused cetyltri-
methylammonium bromide (CTAB) solution of
roughly 5% by mass in deionized (DI) water into the
fluidic system and allowed the solvent to evaporate
overnight. The CTAB surfactant served to reduce the
liquid–vapor surface tension of the solution and fa-
cilitated initial wetting of the device. Scanning
electron microscopy (SEM) images of the photonic
crystal prior to assembly and after the removal of the
RTV fluidics are shown in Fig. 2. As can be seen in
Fig. 2(b), the fluid remains confined to the targeted
row of holes within the photonic structure. Close ex-
amination around the exterior of the photonic crystal
shows some residue that may be the result of a small
leak where the nanochannel first comes in contact
with the photonic structure. As can be seen, however,
this leakage remained confined outside the photonic
crystal and does not appear to have significantly af-
fected the targeted delivery or optical response. The
repeatability of the experiments presented below and
the evidence shown in Fig. 2(b) suggest that the sur-
factant solution, combined with the air-permeable
nature of the RTV, was sufficient to remove trapped
nanobubbles from the photonic structure.
Liquids selected for dynamic modulation of the re-
fractive index within the photonic structure must ex-
hibit relatively low viscosity and high ⌬n
liquids
/
n
substrate
. They must also be miscible to facilitate ex-
change within the nanostructures and exhibit good
compatibility with soft elastomers.
11
As such we se-
lected solutions of aqueous CaCl
2
, which ranges in
composition from DI water 共n = 1.33兲 to5M共n
=1.44兲,
12
as being the most appropriate given the
above requirements. Figure 3(a) shows the normal-
ized quasi-TE mode transmission through the
photonic crystal for DI water and 5 M CaCl
2
. The ex-
periments were carried out using a tunable infrared
laser source the output of which was coupled into the
ridge waveguide corresponding to the fluidically
aligned photonic crystal (which was found by per-
forming a manual prescan of the chip) using a ta-
pered fiber lens. The output from the chip was
coupled back into a second tapered lens fiber and re-
corded on an optical powermeter. The data presented
has been smoothed to remove higher frequency
Fabry–Perot resonances. The results show a shift in
the peak transmission of the guided mode from a/
=0.291 to a/=0.289 (corresponding to =1491 and
1502 nm, respectively) when the lower index liquid is
displaced by the higher index salt solution. The
higher index solution serves to decrease the effective
optical radius of the holes and shifts the guided mode
toward the dielectric band of the regular crystal. As
can be seen from Fig. 3(a), increasing the index dif-
ference between the two fluids would provide a
greater shift in the peak transmission of the guided
mode, but not a higher extinction ratio. The high
⌬n / n afforded by nanofluidic modulation demon-
strated here enables consistently high contrast over a
relatively wide range of wavelengths.
Dynamic modulation of the transmitted power is
demonstrated in Fig. 3(b) at a /=0.291 by fluidically
switching between the DI water and CaCl
2
solutions.
As can be seen the switching time, t
s
, is roughly on
the second time scale. While accurate modeling of the
fluidic transport in systems at these scales becomes
exceedingly difficult (as a result of elastic deforma-
Fig. 2. SEM images showing the photonic crystal (a) prior
to integration with nanofluidics and (b) after removal of the
fluidics. The darkened regions show deposited CTAB after
allowing the solvent to evaporate within the nanochannels,
illustrating precise nanofluidic confinement within the tar-
geted region of the photonic crystal.
60 OPTICS LETTERS / Vol. 31, No. 1 / January 1, 2006
tion of the PDMS, overlapping double layers, and
electroviscous effects, for example) order or magni-
tude estimates are possible from simple laminar flow
analysis. For pressure-driven transport the mini-
mum switching time, t
s
, scales roughly with ⌬PL
2
/D
h
2
(where ⌬P , L, and D
h
are the applied pressure, chan-
nel length, and hydraulic diameter, respectively).
Thus while L could be reduced, potentially bringing
the switching time down into the millisecond range,
D
h
represents the most fundamental limitation on
the speed of such nanoscale optofluidic devices as the
geometry is fixed by the lattice constant of the photo-
nic crystal. As an alternative mechanism, electroki-
netic transport, where flow is induced through the in-
teraction of an applied electric field and the charge in
the electrical double layer, exhibits a more favorable
scaling ratio independent of D
h
,t
s
⬀⌬VL
2
. Thus in
principle it is more amenable to nanoscale transport
and could result in lower switching times. Here, how-
ever, the low electro-osmotic mobility of the CaCl
2
so-
lution made such an approach less attractive. The
diffusive transport time scale into the nanowells 共t
d
⬀d
2
/D兲, where d is well depth and D is the diffusion
coefficient) is of the order of 10
−6
s and thus does not
represent a significant limitation. The reproducible
peaks in output power shown in Fig. 3(b) are a result
of irregular bumps in the transmission spectrum of
the photonic crystal as they pass through the switch-
ing wavelength.
The integration of nanofluidics with nanophotonics
presents, to our knowledge, a new approach for dy-
namic manipulation of optical properties at subwave-
length length scales. Extensions of this technique
could be used to create fully reconfigurable photonic
devices through arbitrary redefinition of fluidically
defined defects (i.e., passive structures could be acti-
vated or defined fluidically when required and then
removed later in favor of alternative functionality).
Other potential functionalities include delivery of op-
tical gain media, nonlinear liquids, or colloidal par-
ticles into arbitrary regions of these structures. Such
integration could also enable a new class of resonant
cavity sensors incorporating targeted delivery of
single or few molecules.
This work was supported by the Defense Advanced
Research Projects Agency through the Center for
Optofluidic Integration funded under the University
Photonics Research program. The e-mail address for
D. Psaltis is psaltis@optics.caltech.edu.
*
Present address, Sibley School of Mechanical
and Aerospace Engineering, Cornell University, Ith-
aca, New York 14853.
References
1. B. Gibson, J. R. Astron. Soc. Can. 85, 158 (1991).
2. S. Kuiper and B. Hendriks, Appl. Phys. Lett. 85, 1128
(2004).
3. D. Wolfe, R. Conroy, P. Garstecki, B. Mayers, M.
Fischback, K. Paul, M. Prentiss, and G. Whitesides,
Proc. Natl. Acad. Sci. U.S.A. 101, 12434 (2004).
4. P. Mach, M. Dolinski, K. Baldwin, J. Rogers, C.
Kerbage, R. Windeler, and B. Eggleton, Appl. Phys.
Lett. 80, 4294 (2002).
5. P. Domachuck, H. C. Nguyen, B. J. Eggleton, M.
Straub, and M. Gu, Appl. Phys. Lett. 84, 1838 (2004).
6. B. Maune, M. Loncar, J. Witzens, M. Hochberg, T.
Baehr-Jones, D. Psaltis, A. Scherer, and Y. Qiu, Appl.
Phys. Lett. 85, 360 (2004).
7. V. Almeida, C. Barrios, R. Panepucci, and M. Lipson,
Nature 431, 1081 (2004)
8. M. Unger, H. Chou, T. Thorsen, A. Scherer, and S.
Quake, Science 288, 113 (2000)
9. J. Joannopoulos, R. Meade, and J. Winn, Photonic
Crystals: Molding the Flow of Light (Princeton U.
Press, 1995).
10. M. Loncar, D. Nedeljkovic, T. Doll, J. Vuckovic, A.
Scherer, and T. Pearsall, Appl. Phys. Lett. 77, 1937
(2000).
11. J. Lee, C. Park, and G. Whitesides, Anal. Chem. 75,
6544 (2003).
12. D. Lide, ed., Handbook of Chemistry and Physics, 79th
ed. (CRC Press, 1999).
Fig. 3. TE-like transmission through a photonic crystal
with an aligned nanochannel. (a) Shift in transmitted
power spectrum when the aligned nanochannel filled with
H
2
O 共n =1.33兲 and5MCaCl
2
共n =1.44兲. (b) Dynamic
switching at a/=0.291.
January 1, 2006 / Vol. 31, No. 1 / OPTICS LETTERS 61
