We demonstrate the implementation of fluorescence correlation spectroscopy (FCS) on a chip. Full planar integration is achieved by lithographic definition of sub-picoliter excitation volumes using intersecting solid and liquid-core optical waveguides. Concentration dependent measurements on dye molecules with single molecule resolution are demonstrated. Theoretical modeling of the FCS autocorrelation function in microstructured geometries shows that the FCS behavior can be controlled over a wide range by tailoring the micro-photonic environment. The ability to perform correlation spectroscopy using silicon photonics without the need for free-space microscopy permits implementation of numerous diagnostic applications on compact planar optofluidic devices.
©2007 Optical Society of America
Ultrasensitive measurements on single molecules have become widely used in biophysics, molecular biology, and biomedical applications. They can, for example, be utilized for elucidating dynamics of single biomolecules in the absence of ensemble effects , or for identification or counting of biomolecules at the ultimate sensitivity limit The first observation of fluorescence emitted from single dye molecules in solution at room temperature was reported in 1990 . Since then a number of experimental approaches with varying level of complexity have been developed, including confocal , total internal reflection , and near-field optical microscopy . Fluorescence correlation spectroscopy (FCS)  has emerged as one of the most powerful and versatile optical methods for studying small numbers of molecules. FCS is based on analyzing the autocorrelation function of fluorescence fluctuations collected from a small excitation volume. Because this signal can contain information about numerous experimental parameters and can be obtained both in and out of thermal equilibrium, FCS has found a large number of applications including detection of protein conformational changes, DNA hybridization, membrane binding, or protein association interactions . While numerous variations of single molecule spectroscopy and FCS in particular exist, curiously none of these are fully optically integrated. Nano- and microfabrication have been used for partial integration [6-8], but the optical excitation and/or detection volume were still defined with the help of at least one microscope objective . In light of the emerging field of optofluidics , the integration of single molecule analysis capabilities promises numerous advantages ranging from improvement of existing instrumentation to potentially new capabilities. The main obstacle preventing optical integration was the inability to guide light in fluidic channels small enough to be compatible with single molecule detection requirements. Recently, several ways to overcome this have been developed using either novel approaches to index guiding or dielectric multilayer confinement of inherently leaky modes in hollow waveguides . Among these, antiresonant reflecting optical (ARROW) waveguides are perhaps the most promising for building ultrasensitive, planar optofluidic systems. Aside from merely being able to guide light , liquid-core ARROWs can be efficiently coupled with solid-core waveguides to form planar two-dimensional optofluidic networks. Experimentally, single molecule fluorescence sensitivity has been recently demonstrated in a solid-liquid core waveguide intersection, the most basic element of such a network . Here, we report the first demonstration of fluorescence correlation spectroscopy on a planar optofluidic chip with single molecule resolution and lithographically defined excitation and collection volumes. We also show that the microphotonic environment of the physical interaction region can be flexibly tailored to make FCS-based methods ideally suited for implementation of single molecule analysis on a chip.
2. Optofluidic chip design and characteristics
Implementation of FCS on a chip requires several steps: lithographic definition of a sub-picoliter fluorescence excitation volume, optical waveguides to guide light to the excitation region, and an efficient way to collect the fluorescence and guide it to the detector. Fig. 1(a) shows schematically how this concept is realized in the ARROW-based optofluidic approach. The image shows a cross section through a liquid-core waveguide in which fluorescent molecules can freely diffuse or be moved by electric fields. The liquid core is surrounded by dielectric confinement layers that fulfill the ARROW antiresonance condition for high reflectivity  in the x and y-directions. An excitation beam propagating in the x-direction is coupled into the liquid core via a solid-core ARROW waveguide which is vertically aligned with the center of the liquid core and at the same time serves as the top layer of the liquid core waveguide. This geometry defines a cylindrical excitation volume with a cross section given by the solid-core mode profile as shown in the figure. Molecules within this excitation volume fluoresce and their emission is captured by the liquid-core waveguide and guided along the liquid channel in the z-direction. The ARROW layers are designed to be both highly reflective for the fundamental liquid core mode at grazing incidence and highly transmissive at near-normal incidence to ensure efficient coupling between solid and liquid core sections . This capability is the key for realizing compact and self-contained planar optofluidic devices.
Fig. 1(b) shows how these concepts are implemented in an optofluidic chip. In contrast to earlier versions with completely open channels , the liquid channel is now terminated by a fluidic reservoir that allows for continuous analyte supply. The generated fluorescence can be coupled to another solid core waveguide for transfer to the edge of the chip. The platform is implemented on a silicon substrate with silicon nitride (n=2.1) and silicon dioxide (n=1.46) ARROW layers using a previously described sacrificial layer process . The dielectric layer sequence for the ARROW waveguides was (starting from the substrate; all values in nm): SiO2/SiN/SiO2/SiN/SiO2/SiN – core – SiN/SiO2/SiN/SiO2/SiN/SiO2 (268/100/268/100/268/100/5000/120/337/156/243/156/2976) resulting in low reflectivity for light entering the core through the top SiO2 solid core (see section 4). The liquid core dimensions are 5×12μm, and the solid core waveguide width was 12μm. This leads to an excitation mode area (full width half maximum) of 7.1μm2, and a cylindrical excitation volume of 85fl. SEM images of both waveguide types are shown in Figs. 1(c) and (d).
3. Single molecule FCS on optofluidic chip
Fluorescence correlation spectroscopy measurements were performed on the chip depicted in Fig. 1(b) by filling the hollow waveguides with a solution of water and Alexa 647 dye (Invitrogen) molecules at various concentrations. Y-polarized pulses (pulse width 200fs, wavelength 633nm, repetition rate 76MHz) from an optical parametric oscillator (Coherent OPO) were coupled into the solid-core waveguides via single-mode fiber. The average power of 300μW at the waveguide input resulted in ∼100μW in the excitation volume. The fluorescence emitted into the detection arm was detected (15% efficiency , 700 cts/s/molecule) by a single photon avalanche photodiode (Perkin Elmer) after removing residual excitation light with a filter (Omega Optical). The data were analyzed with a standard time-correlated single-photon counting setup (Time Harp 200, PRT 400 router; PicoQuant Inc.) in time-tagged time-resolved (TTTR) collection mode . The results for the concentration-dependent autocorrelation signal are displayed in Fig. 2(a).
Excellent agreement is found between the experimental data (symbols) and an ARROW-based FCS model (solid lines, see section 4). We observe the characteristic shape and concentration dependence of fluorescence correlation signals, in particular the increase in FCS contrast with decreasing concentration. Dye concentrations as low as 10pM were detected with the ARROW device. For the present excitation power and delay times, nonideal effects such as triplet formation and photobleaching can be neglected, and the average number of molecules N probed by the measurements is determined by G(0)=1/N . The extracted amounts are displayed by the open symbols in Fig. 2(b), showing a resolution of less than one molecule on average (0.35) in the integrated device. The agreement with estimates of the average molecule number obtained using the nominal concentration and effective excitation volume (circles) in time-integrated fluorescence is excellent. However, the demonstration of FCS in this more highly integrated platform opens the way for more advanced and practically relevant biomolecule analysis. As one example, we can extract the diffusion coefficient of the Alexa 647 dye molecules from the FCS traces in Fig. 2(a) using Eqs. (3) and (4) below and find D=250μm2/s in good agreement with a model assuming spherical particles.
4. Microphotonic control of FCS characteristics
In addition to allowing for conventional FCS with single molecule sensitivity, the integrated optical approach offers further possibilities for improvements and novel capabilities by tailoring the microphotonic environment. This requires a quantitative description of the observed signal that takes into account the experimental geometry. The autocorrelation function used in FCS analysis is given by ,
where C is the average concentration, angular brackets denote a temporal average containing the physics of the fluctuation process, and W(r)=Iexc(r)κ(r) is a spatial distribution factor that contains the system geometry via the shape of the excitation profile Iexc and the collection efficiency profile κ. In conventional confocal setups, W(r) is a product of Gaussians in all three dimensions and results in an effective FCS volume in the shape of an ellipsoid of revolution given by Veff=[C•G(0)]-1=π3/2wxwywz, where the wi are the Gaussian beam waists. The situation is different in the case of FCS in an ARROW chip as none of the dimensions are defined using free-space optics. The collection profile is independent of the z-coordinate and can be calculated from the coupling efficiency of dipole fluorescence into the fundamental ARROW mode as a function of the dipole position. These distributions were calculated using a commercial waveguide solver (OmniSim, Photon Design) and can be approximated very well by Gaussian profiles (waists wxc and wy). The excitation profile Iexc(r) requires more careful analysis. Its y-and z-components are given by the solid core mode profile and can again be approximated by Gaussians (waist wz). In the x-direction, the ARROW confinement layers create a cavity that results in a standing wave for the electric field inside the liquid core. This variation of the electric field across the core is depicted in the inset of Fig. 3(a).
We can model the cavity effect by calculating the x-dependence of the excitation intensity as
where L is the ARROW core width along x, nc is the core index, and r and t are the reflection and transmission coefficients of the ARROW layers, respectively. Combining all spatial dependencies, the autocorrelation function is given by
where k0=4πnC/λ and R=∣r∣2. A plot of this autocorrelation function for the parameters of Alexa 647 dye normalized to G(τ=0,R=0) for different values of the ARROW reflectivity R is shown in Fig. 3(a). The cavity effect modifies the FCS signal in two distinct ways. First, G(0) increases with increasing R due to a change in the effective FCS volume. In the limit of wxc ≫ λ/2πnC, the expressions can be simplified and we find an effective FCS volume of
Eqn. (5) is reminiscent of conventional confocal geometries. However, the effective volume in x-direction is now determined by the sum of all regions with large electric field amplitude. These are represented by the gray ellipses in Fig. 3(a). By changing the waveguide layer design, the reflectivity R can be tailored and the effective excitation volume can be reduced continuously by up to a factor of 1.5. Second, with increasing R a shoulder appears in the FCS signal at short times due to the standing wave pattern . The time scale of this shoulder is determined by diffusion out of a single gray ellipse. Consequently, a second diffusion feature appears temporally shifted by the ratio of wxc and the 1/e2 width of the standing wave pattern. This feature can be used to significantly reduce the measurement time required for obtaining a sufficient signal-to-noise ratio. In addition, the relative position of the two diffusion features can be tailored by microphotonic design, i.e. adjusting wxc via the core width. The parameters for the sample used in the experiments described above are wxc=4.68μm, wy=1.07μm, and wz=4.15μm, resulting in an effective FCS volume of 115fl for R=0. The cavity-related diffusion feature occurs around 5×10-6s, 2,700 times more quickly than the ordinary diffusion process as indicated by the black arrows in Fig. 3(a). It does not appear in our data (Fig. 2(a)) because the ARROW was designed for maximum transmission into the liquid core (R≈0).
In addition, the cavity in the x-direction can be coupled favorably with electrokinetic flow (drift) of molecules along the channel (z-direction). It is well known that the presence of drift modifies the FCS signal . However, we find that drift affects only the slow diffusion feature related to the z-component of the effective volume, and not the cavity-induced signal at short times which is due to the structure along x, perpendicular to the flow. This effect is demonstrated in Fig. 3(b), where autocorrelation curves for dye molecules with different drift velocity in z-direction are shown. Therefore, the contributions of diffusion and drift can be extracted in a single FCS measurement by determining the diffusion coefficient from the short time feature and the flow velocity from the slow process.
We have shown that fluorescence correlation spectroscopy can be implemented in a fully planar optofluidic device. Single molecule resolution was demonstrated. It was shown that the FCS signal and volume depend on the microphotonic environment of the ARROW waveguide intersection and can be tuned in a unique way by rational design of the ARROW confinement layers. The demonstration of FCS in an integrated optical and fluidic setting leaves room for future developments in several directions. Further integration with electrical control of molecule behavior is possible. Analysis of electrokinetic motion of liposomes on the single molecule level in integrated ARROW chips will be presented elsewhere. Integrated FCS is also perfectly suited for FCS derivatives such as dual color, dual focus, or multi-photon FCS [17,18]. In particular, the ability to define foci positions and overlap very precisely and a priori by using lithographically defined waveguides is attractive. Finally, the ARROW chips can be expanded into larger arrays of intersecting waveguides that allow for correlation spectroscopy on multiple samples in parallel with a single excitation waveguide.
We thank D. W. Deamer, S. Kuehn, and J. Barber for helpful discussions. Parts of this work were funded by the NIH/NIBIB under grant R01EB006097, the NSF under grant ECS-0528730, and NASA/UARC under an Aligned Research Program (ARP) Award. D.Y. acknowledges support by University of California Systemwide Biotechnology Research & Education Program GREAT Training Grant No. 2005-245.
References and links
01. C. Zander, J. Enderlein, and R.A. Keller, “Single-molecule detection in solution: methods and applications,” 1st ed., Wiley (2002).
02. E. B. Shera, N. K. Seitzinger, L. M. Davis, R. A. Keller, and S. A. Soper, “Detection of single fluorescent molecules,” Chem. Phys. Lett. 174, 553 (1990). [CrossRef]
03. W.E. Moerner and D.P. Fromm, “Methods of single-molecule fluorescence spectroscopy and microscopy,” Rev. Sci. Instrum. 74, 3597 (2003). [CrossRef]
04. S. Kuhn, U. Hakanson, L. Rogobete, and V. Sandoghdar, “Enhancement of single-molecule fluorescence using a gold nanoparticle as an optical nanoantenna,” Phys. Rev. Lett., 97, 017402 (2006). [CrossRef]
05. D. Magde, E. Elson, and W.W. Webb, “Thermodynamic Fluctuations in a Reacting System-Measurement by Fluorescence Correlation Spectroscopy,” Phys. Rev. Lett., 29, 705 (1972). [CrossRef]
07. M. Foquet, J. Korlach, W. R. Zipfel, W.W. Webb, and H. G. Craighead, “DNA Fragment Sizing by Single Molecule Detection in Submicrometer-Sized Closed Fluidic Channels,” Anal. Chem. 74, 1415–1422 (2002). [CrossRef] [PubMed]
09. L. Kastrup, H. Blom, C. Eggeling, and S.W. Hell, “Fluorescence fluctuation spectroscopy in subdiffraction focal volumes,” Phys. Rev. Lett., 94, 178104 (2005). [CrossRef]
10. C. Monat, P. Domachuk, and B.J. Eggleton, “Integrated optofluidics: A new river of light,” Nat. Photonics, 1, 106 (2007). [CrossRef]
11. D. Yin, J.P. Barber, A.R. Hawkins, D.W. Deamer, and H. Schmidt, “Integrated optical waveguides with liquid cores,” Appl. Phys. Lett., 85, 3477 (2004). [CrossRef]
13. H. Schmidt, D. Yin, J.P. Barber, and A.R. Hawkins, “Hollow-core waveguides and 2D waveguide arrays for integrated optics of gases and liquids,” IEEE J. Sel. Top. in Quantum. Electron. 11, 519 (2005). [CrossRef]
14. J.P. Barber, E.J. Lunt, Z. George, D. Yin, H. Schmidt, and A.R. Hawkins, “Integrated Hollow Waveguides with Arch-shaped Cores,” IEEE Photon. Technol. Lett., 18, 28 (2006). [CrossRef]
15. D. Yin, J.P. Barber, A.R. Hawkins, and H. Schmidt, “Highly efficient fluorescence detection in picoliter volume liquid-core waveguides,” Applied Physics Letters, 87, 211111 (2005). [CrossRef]
16. P.F. Lenne, E. Etienne, and H. Rigneault, “Subwavelength patterns and high detection efficiency in fluorescence correlation spectroscopy using photonic structures,” Appl. Phys. Lett. 80, 4106 (2002). [CrossRef]
17. P. Schwille, U. Haupts, S. Maiti, and W.W. Webb, “Molecular Dynamics in Living Cells Observed by Fluorescence Correlation Spectroscopy with 1- and 2-Photon Excitation,” Biophys. J., 77, 2251 (1999). [CrossRef]