Lin
Pang
*,
H. Matthew
Chen
,
Lindsay M.
Freeman
and
Yeshaiahu
Fainman
*
Jacobs School of Engineering, University of California, La Jolla, California 92093-0407, San Diego. E-mail: lpang@ece.ucsd.edu; fainmain@ece.ucsd.edu.; Fax: 1-858-534-1225; Tel: 1-858-534-7208
First published on 1st June 2012
Optofluidics integrates the fields of photonics and microfluidics, providing new freedom to both fields and permitting the realization of optical and fluidic property manipulations at the chip scale. Optofluidics was formed only after many breakthroughs in microfluidics, as understanding of fluid behaviour at the micron level enabled researchers to combine the advantages of optics and fluids. This review describes the progress of optofluidics from a photonics perspective, highlighting various optofluidic aspects ranging from the device's property manipulation to an interactive integration between optics and fluids. First, we describe photonic elements based on the functionalities that enable fluid manipulation. We then discuss the applications of optofluidic biodetection with an emphasis on nanosensing. Next, we discuss the progress of optofluidic lenses with an emphasis on its various architectures, and finally we conceptualize on where the field may lead.
Lin Pang | Lin Pang is a research scientist in the Department of Electrical and Computer Engineering, University of California, San Diego. He received his Ph.D. degree from Sichuan University, Sichuan, China, in 1998. His current interests include hybrid holographic lithography, nanofabrication, micro, nanofluidics, and sensing based on coupled plasmonic nanoresonators. |
H. Matthew Chen | Haiping Matthew Chen received his Master’s degree from the University of Illinois, Urbana-Champaign. Currently, he is completing his Ph.D. in electrical engineering at the University of California, San Diego. His interests include computer simulations of plasmonic and optofluidic phenomena and fabrication and characterization of plasmonic nanostructures for biological sensing applications. Presently, he is also with Navair working on defense microwave and photonic applications. |
Lindsay M. Freeman | Lindsay M. Freeman received the B.S. degree in chemical engineering from the University of Southern California, Los Angeles, in 2010. She is currently working towards her Ph.D. degree in chemical engineering at the University of California, San Diego, under the supervision of Prof. Y. Fainman. |
Yeshaiahu Fainman | Y. Fainman is a Cymer Professor in ECE at the University of California, San Diego. He is directing the Ultrafast and Nanoscale Optics Group and has made pioneering contributions to near field optical phenomena, inhomogeneous and meta-materials, nanophotonics and plasmonics, optofluidics and non-conventional imaging. The research applications target information technologies and biomedical sensing. He is a Fellow of the OSA, Fellow of the IEEE, Fellow of the SPIE, and recipient of the Miriam and Aharon Gutvirt Prize, Lady Davis Fellowship, Brown Award and Gabor Award. He has contributed to over 200 manuscripts in peer review journals and over 350 conference presentations and conference proceedings. |
Optical devices are channels whose materials interact with light and result in the alteration of the light path. Snell's law, or more precisely Maxwell's equations, governs the outcome of the interactions. In optical devices, the refractive index describes the material property that directly affects how light interacts with the different media, such as the degree in which light would be reflected, transmitted, or absorbed. In other words, the distribution of the refractive indices of the media would control how the light passes through the media, resulting in components such as lens, mirrors, waveguides, etc.
The dependence of the behaviour of light on the index distribution affects the performance if the index distribution varies. When fluidic control matured to the extent of on-chip configuration, the marriage between optics and fluidics rapidly occurred. The term “optofluidics” was coined in 2004 to reflect a newly formed research initiative supported by the Defence Advanced Research Projects Agency (DARPA) of the U.S. Department of Defense.13,14 Since then, optofluidics has become one of the most active research areas with publications increasing exponentially.15–18 Optofluidic devices have evolved substantially from controlling a device's property via flow or optical manipulations to revolutionary applications in both sensing and imaging.
In this review, after briefly reviewing the background on microfluidics, we will summarize optofluidic devices in terms of their functionalities: property manipulation, sensing, and imaging. Readers will witness the development of the field, starting with a discussion of optofluidic devices controlled by flow, to an overview of liquid photonics. This review will emphasize the inter-disciplinary applications created by optofluidics, leaving readers amazed by the potentials of this promising research field.
Fig. 1 Microfluidic devices made from PDMS; (a) push-up valves; (b) mixer; (c) pump; (d) integrated system. Part (a) is adapted from ref. 3; part (b) from ref. 11, with permission from the American Institute of Physics. Part (c) and (d) are adapted from ref. 4 and 5, respectively, with permission from AAAS. |
In microfluidics, viscous forces become more important than inertial forces, resulting in fluids that do not mix convectively when they meet in a microchannel, but rather will flow in parallel. The only mixing that occurs is the result of diffusion across the interface between two fluids. This laminar flow property preserves refractive index contrast and index distributions in PDMS microchannels, providing a series of special functionalities that have aided the development of novel optofluidic devices.
Fig. 2 (a) Schematic drawing of the three-layer elastomer chip with four optical quality facets. (b) Microphotograph of the switch device. (c) Microphotograph showing the mirror channel reflecting the light with water running through; (d) Light goes through the mirror channel with index matching fluid filling the channel. (e) Images of the bypass mode (upper) and exchange mode (down); (f) Power of transmitted and reflected beams as a function of time. Part (a), (b) and (e) are produced from ref. 25 with permission from the American Institute of Physics. |
Fig. 2(c) shows the mirror channel filled with water for total reflection, while Fig. 2(d) shows the channel filled with index matching solution for transmission. Fig. 2(e) shows the transmitted (upper) and reflected (bottom) beam spots captured on a CCD camera. Fig. 2(f) shows the time response of the switch with a control channel pressure of 15 psi, corresponding to roughly 10 ms switching speed. The switching speed depends on the flow rate, which depends on the dimension of the channel, pressure difference and the viscosity of the flow. Increasing the pressure difference and the channel height while reducing the channel length are potential approaches to increasing the switching speed. Fig. 2(e) also shows the crosstalk, which was caused by the index mismatch. This phenomenon could be used to make an optofluidic variable attenuator by controlling the reflectivity of the fluid–PDMS interface while adjusting the index in the mirror channel.26 This configuration can also be used to determine the index value by measuring the reflected power.
Following the same principle, a waveguide-based, on-chip light routing optofluidic switch is achieved with a tunable air-gap mirror, by which the light is deflected via total internal reflection in the bypass state.27
The optical beam switch can also be achieved based on diffraction from a blazed grating when different fluids flow over it. The diffracted angles are proportional to the mismatch between the refractive indices of the material of the grating and the liquid in the channel.28 Optical switching was also achieved in the in-plane liquid-core/liquid-cladding (L2) waveguides by flow rate manipulations29 and pneumatic control.30
Unlike the abrupt index shift of previously discussed examples, the optical response can be precisely manipulated by fine tuning the required fluidic index by mixing two fluids with any concentration, generating a range of refractive indices between the two initial fluids. A fine tuning of an optical microring resonator device (MRD) by dynamic variation of the refractive index of the medium surrounding the waveguides was demonstrated by manipulating the upper cladding of the solid MRD waveguides.31 This fine tuning of the refractive index of fluids can also be employed in the fine tuning of the surface plasmonic polariton (SPP) mode to achieve designed SPP interaction.32
Fig. 3 (a) Schematic drawing of the liquid-core/liquid-cladding waveguide laser. (b) Whispering gallery mode laser from droplets. (c) Liquid-core/solid-cladding micro ring laser. (d) Distributed feedback laser of solid-core/liquid-cladding wavequide. Part (a) is produced from ref. 33 with permission from American Chemical Society. Part (b) is adapted from ref. 34. Part (c) and (d) are produced from ref. 37 and 41, respectively, with permission from the Optical Society of America. |
On the other hand, the liquid core/solid cladding or solid core/liquid cladding waveguide based lasing configurations proved to be alternatives for mode selections. Li et al.36 and Suter et al.37 used a microring as a liquid core and PDMS as the cladding and demonstrated a single mode liquid-core waveguide based microring cavity, as shown in Fig. 3(c).37 A glass capillary based micro-bubble38 and a micro-ring laser39 were also demonstrated; additional work exhibited lasing based on Forster resonance energy transfer of gain medium with labeled DNA.40 A feedback distributed grating in solid core/liquid cladding configuration was also employed to select a longitudinal lasing mode. Fig. 3(d) shows a typical feedback distributed optofluidic laser.41 The laser cavity consists of a single mode liquid core/PDMS cladding channel waveguide and a phase shifted 15th order distributed feedback (DFB) structure. Its wavelength can be continuously tuned over a large range with a simple mechanical deformation.42 Similarly, a solid circular distributed feedback grating with a dye solution acting as the cladding layer was demonstrated by Song et al.43 These optofluidic dye lasers will have great potential for highly parallel multiplexed on-chip spectroscopic applications.
Fig. 4 (a) Process to form embedded microchannels in n-borophosphosilicate glass; (1) PECVD of template layer, (2) formation of template ridges-photolithography RIE, (3) PECVD of BPSG, (4) high-temperature anneal to form microchannels. (b) ARROW waveguide cross section consisting of SiN and SiO2 dielectric layers and a liquid core. Part (a) is adapted from ref. 45 with permission from Optical Society of America. Part (b) is reproduced from ref. 46 with permission from the American Institute of Physics. |
The low index core configuration enables the solution to operate at a single optical waveguide mode, which will find many applications in bioengineering where aqueous based environments are desired.
Plasmonic sensing is used mainly due to its high sensitivity, which results from the fact that the maximum optical field is confined to the surface, where the biological binding reactions take place. Surface plasmon polariton (SPP) is the surface charge density oscillation, a consequence of the coupling between a photon and free electrons on the interface of a metal and dielectric. Due to the momentum mismatch, SPP waves cannot be excited by direct illumination of light,54 but by (1) prism coupling via attenuated total reflection (Kretschmann configuration) or by (2) a diffractive grating. Surface plasmon resonance (SPR) was first introduced for sensing by Nylander and Liedberg.55 Since then, due to its real-time analysis of biospecific interactions without the use of labeled molecules, SPR sensing has continued to improve the sensor designs, sensing surface characterizations,56 and further sensor commercialization.57 In 1998, Ebbesen et al. reported the observation of enhanced transmittance through optically thick metallic films with subwavelength features.58 Since then, Brolo et al. demonstrated the use of subwavelength hole arrays for sensing applications.59 Different interrogation approaches have been explored, such as incident angle,60 intensity,61 polarization,62 as well as multiple modes.63
Although SPR based sensing's highest sensitivity results from maximum field confinement at the surface, the high propagation loss of a metal surface heavily damps the resonance, leading to a broader resonant linewidth compared with its dielectric counterparts. The broader linewidth of the SPR makes the small resonant shift undetectable. Two orthogonally aligned polarizers sandwiching the nanohole substrate were employed to filter out the directly transmitted nonresonant background and changed the SPR from a Fano shape to a Lorentzian shape.60 The resonant linewidth was reduced from 5.5 nm to 2.5 nm from parallel (PP) to orthogonal polarizer-analyzer (OP), as shown in Fig. 5(a). This improvement of signal-to-noise ratio enhances the LOD. Enhancement of sensitivity was also conducted by coupling localized surface plasmonic resonators (LSPR) with propagating SPR. As shown in Fig. 5(b), there is a layer of gold film on top of a perforated substrate, in which the film protrudes out, forming nanovoids. The coupling between the propagating SPR and LSPR enhances the amplitude of the electric field, as shown in Fig. 6(b). The field penetration depth decreased, leading to surface sensitivity enhancement.64
Fig. 5 Surface Plasmon Resonance (SPR) based detection (a). Conceptual diagram of the 2D nanohole-array-based SPR sensor. (b) Near-field intensity distribution of the electric field on the nanoresonant substrate resulted from coupling between LSPR and SPR in a 2D metal film perforated with nanohole. Adapted from ref. 60 and 64, copyright 2006, 2009 Optical Society of America. |
Fig. 6 a) SEM of fabricated upward nanoresonator with opening of 50 nm. b) A sample uniformity plot of 5 nanotorches at different locations within the same substrate. |
The interrogation area of a propagating SPR based sensor is tens of microns. Nanoparticle or nanoresonator based SPR sensors were also explored, including a nanoparticle array65 and a random nanohole array.66 A single nanoparticle based sensor was demonstrated in a hydrogen sensor at the single-particle level.67 The single nanoresonator based sensor avoids any statistical effects, possibly enabling low LOD detections.
The sensitivity of the nanosensor depends on the strength of the local field. Coupled nanoresonant designs were applied to enhance the local field by almost one order of magnitude,68,69 but the resolution requirement makes them difficult to be fabricated by current top-down nanofabrication approaches with electron beam and focused ion beam lithography. Bottom-up methods were employed to make a nanocrescent resonator shape,70,71 where coupling among nanotip to nanotip (edge–edge), tip–nanocavity, and nanocavity–nanobody were demonstrated. The random orientation of 3D nanocrescent was overcome to produce reproducible nanoresonators, as shown in Fig. 6. The reproducibility of 80%, on par with standard commercial surface enhanced Raman scattering substrate, was verified by measuring different single nanoresonators within the same substrate.72
In optofluidic sensors, the mass transport of the target molecule on the sensing area directly affects the detection speed. For most sensing platforms, recognition molecules are immobilized on the sensing surface, and fluid delivers the target molecule through the channel over the surface. Due to the laminar flow nature of microfluidics, the flow rate at the surface is zero, and target molecules are not delivered convectively, but diffuse to the surface to bind with their recognition molecules. Commercial microfluidic systems use interval injections of fluid and air to improve the mass transport.73 Flow through microfluidics was introduced for both a plasmonic sensor74 and photonic crystal sensor,75 in which the substrate is back-etched so that the liquid could flow through the metal or dielectric membrane. Orders of enhancement in the mass-transport rate were demonstrated compared to the traditional flow-over method.
The lens' focusing power is determined by its index gradient, which can be manipulated by controlling the shape, curvature or index difference of the lens. Optofluidic lenses were first designed with a chamber covered by a flexible membrane, which deforms when a pressure is applied to pump the liquid in and out of the chamber. The pressure in the chamber changes the curvature of the spherical membrane and consequently the focal length.76–80 In this fashion, a plano-convex78 and a bi-convex design79 were proposed, and two-lens chambers80 were employed to vary the curvatures of the two lenses independently. Modulation of the wavefront of light occurs at the interface between the two liquids and is proportional to the difference between their refractive indices and to the curvature of the membrane, which is controlled by the difference in pressure between the chambers. A complex microfluidic channel design was introduced to enable the combination of three-lens chambers with independent controls.81 The optically functional area of the device is a 3 × 3 mm square window in the centre where the two chambers overlap, as shown in Fig. 7. The low refractive index liquid (water, nD = 1.33) is filled in chambers 1 and 4, while the connected chambers 2 and 3 are connected with high refractive index liquid (immersion liquid, nD = 1.58) reservoir. Adjusting these three pressure differences allows for not only different lens combinations, but also for forming different lens shapes, ranging from a normal circular lens to independently orthogonal cylindrical lenses. Fig. 7(b) shows the focal length dependence on the pressure difference in one chamber while keeping another chamber under no deformation, illustrating the focal length adjustments between −40 and 23 mm. Fig. 7(c)–(e) show the intensity profile of a focused laser beam, showing the independent control of the lens shapes and its focusing performance.
Fig. 7 (a) A schematic drawing of the device. (b) Focal lengths of the cylindrical lenses focusing light along the x-axis (squares) and y-axis (crosses) as functions of the pressure differences. (c), (d) and (e) Demonstration of the laser mode. (c) Focusing in both directions; (d) Focusing in y direction; and (e) Focusing in x direction. Adapted from ref. 81, copyright 2005 Optical Society of America. |
Refractive index distribution of the liquid can also be modified by taking advantage of the fluid diffusion at the microscale level to form a refractive index gradient in the liquid medium, as shown in Fig. 8.82 A high-refractive index solution (nD = 1.455) was injected adjacent to a low refractive index solution (nD = 1.33). The diffusion between the fluids of the co-injected flows can form a refractive gradient, which can focus light in a way similar to the traditional gradient refractive index (GRIN) lens. The 2D refractive index profile and subsequently the focal length of the liquid GRIN (L-GRIN) lens can be tuned by changing the ratio of the flow rates of the injected solutions.
Fig. 8 (a) The 2D L-GRIN lens structure and the diffusion of CaCl2 inside the L-GRIN lens chamber. (b) The side view of the refractive index distribution in the L-GRIN lens chamber. (c) The fluid injection setup of the 2D L-GRIN lens. (d) A microscopic image of the 2D diffusion pattern in the lens chamber. Reproduced from ref. 82. |
Pressure driven force is usually used to manipulate liquid flow, in which pressure is applied from outside the chip by either compressed air for fast response or height difference of the liquid for slow tuning. In these cases, reservoirs outside the chip are used for principle demonstration. However, for production operation, on-chip manipulation would be desired, such as electric actuation to modulate the chamber pressure (electroactive polymer,83,84 electromagnetic,85 or stimuli responsive hydrogel).86 However, for droplet formed lens, the effective manipulation would be electrowetting, which employs electrostatic forces to change the hydrophilicity of the solid–liquid or liquid–liquid interface. In this case, the surface tension changes the shape of the liquid surfaces87,88 and the sidewall electrode configuration,89,90 making the electrowetting lens a practical application.
Due to the many applications in biology, e.g. on-chip flow cytometry,91 in-plane adaptive lens were widely produced. Instead of membranes, fluids with different indices were used to form the geometric lens shape, while the pressure or flow difference was used to change the shape for adjusting the focal path. This is referred to as a liquid-core liquid-cladding (L2) lens, as shown in Fig. 9.92 Two streams of liquid with a lower refractive index (the cladding) sandwich a stream of liquid with a higher refractive index (the core). As the core stream enters the expansion chamber, it widens and becomes biconvex in shape for specific flow rates. This biconvex fluidic element focuses light. Manipulating the relative flow rates of the streams reconfigures the shape, and therefore adjusts the focal distance of the L2 lens. By offsetting the inlet and the outlet of the lens chamber to the chamber axis, Song et al.93 achieved a radius of curvature smaller than the limiting chamber radius, allowing for better focusing ability than the symmetric design.
Fig. 9 (a) Schematic of the liquid-core liquid-cladding (L2) lens. The channel for the formation of the L2 lens contains a square expansion chamber. The solid lines show the walls of the channel, and the dashed lines show the interfaces between the core and the cladding streams. (b) Bright-field image of the L2 lens in operation taken using a CCD camera to show the traces due to the fluorescence of the dye. Reproduced from ref. 92. |
Most of the optofluidic lenses are not tested for their imaging performance, but rather only for their dynamic focusing or beam shaping abilities. In optics, lens design and fabrication are complicated tasks because the imaging quality of a lens is strongly affected by the lens' surface, shape, curvature, size, material dispersion, as well as its variation with temperature and pressure.
Optofluidic focusing is a slow process due to the slow response of the fluid feeding/exchanging process. To increase the response time, frequency generated pressure control was introduced for high speed focusing, including 3D imaging generation. Campbell et al.94 designed adaptive lenses with millisecond response time that are based on transparent flexible elastomer membranes affixed to a plano-convex glass lens and a diaphragm between the membrane and the lens, as shown in Fig. 10. The membrane is optically flat. Instead of using a liquid, the application of vacuum to the interior of the mount pulls the membrane inward. Both the rise and fall transitions had very short onset time switches between flat segments and steep rise and fall segments occurring within 0.3 ms, suggesting that the lenses can be driven at frequencies up to 1600 Hz.
Fig. 10 (a) Schematic of an adaptive compound camera lens with a flexible membrane, (b) the same camera lens with the membrane pulled inward when a vacuum is applied. Reproduced from ref. 94 with permission from the American Institute of Physics. |
Instead of moving a membrane, López et al.95 oscillated a drop of water trapped in a hole drilled in a Teflon plate to increase its focusing power. An audio speaker generated a sinusoidal pressure wave that drove the oscillations of the lens. In a period of oscillation the focal length evolved through its full range of values, and a synchronized high-speed camera captured sharp images at different focal planes. Speed of 10 ms was achieved. The combination of oscillating-focal-length lenses with high-speed cameras opens new possibilities for three-dimensional imaging devices.
It is certain that the complete integration of functionalities discussed here will be achieved and commercialized for on-chip source, manipulation, and detection for biochemical and biological applications. Furthermore, new research directions can also be envisioned.
As we have seen at the micrometer scale, current microfluidics exhibits no difficulty interfacing with the photonic design. However, as the photonic length scale reaches the nanometer region with the functional area down to tens of nanometers, mismatches between nanophotonics and nanofluidics become problematic, such as the issues of co-localization of bioreagents on nanometer sensors. The required alignment resolution is beyond current fabrication and processing capabilities. Therefore, novel approaches have to be formulated to align nanoresonators with nanofluidics.
The emergence of electronics and fluidics has been playing an important role in research and everyday life, such as electrowetting89 and liquid crystal display,96 assisted nanoparticle assembling,97 and nanopole DNA sequencing.98–100 Fully integrated opto-electro-fluidics at the micro and nanometer level will definitely enable more accurate and faster controls with added functionalities.
Current optofluidic sensors are designed to be employed for in vitro applications. Both photonics and microfluidics are designed for operation in a low background or interference environment. In in vitro operation, background interference can be eliminated by washing steps before signal readout. For in vivo configuration, no washing steps can be applied with biofouling being an inevitable problem. Serious breakthroughs have to be made in optofluidics before in vivo point of care diagnostics can become a reality.
Other prediction may also become reality within the next few years, like co-localization of single quantum dots in a nanoresonator for quantum analysis101,102 or optofluidics for energy generation and storage.103
Footnote |
† Published as part of a themed issue on optofluidics |
This journal is © The Royal Society of Chemistry 2012 |