We explore a new confocal microscope for the detection of surface-generated fluorescence. The instrument is designed for high resolution imaging as well as for the readout of large biochips. Special feature is the separated collection of two different fluorescence emission modes. One optical path covers the emission into the glass at low surface angles, the other captures high angles, exceeding the critical angle of the water/glass interface. Due to the collection of the supercritical angle fluorescence (SAF) the confocal detection volume is strictly confined to the interface, whereas the low angles collect much deeper from the aqueous analyte solution. Hence the system can deliver information about surface-bound and unbound fraction of fluorescent analyte simultaneously.
©2004 Optical Society of America
Many applications in chemistry and biochemistry are based on the measurement of the interaction between a surface-bound receptor and a ligand. Due to its high sensitivity the fluorescence detection is often the method of choice when minute amounts of material need to be analyzed.
In order to record a binding reaction at the surface in real-time it is of major importance to detect the fluorescence from bound ligand with high efficiency and at the same time to discriminate against the signal of unbound ligand. The most commonly used approach to confine the fluorescence detection volume to an interface is total internal reflection fluorescence (TIRF) excitation . Hereby illumination above the critical angle produces an evanescent field which excites fluorophores only in direct vicinity to the interface. In this context a very promising approach for the study of surface binding reactions is to combine TIRF with fluorescence correlation spectroscopy (FCS) [2–4].
The high relevance of achieving minimum detection volumes for single molecule analysis at high concentrations has lead to the development of so called zero-mode waveguides.  With diameters of typically 50 nm these holes in a thin metal layer are too small to sustain propagating light modes, providing detection volumes down to 10 zeptoliters (1 zl = 10-21 L).
In contrast to the use of sophisticated metal nanostructures or high angle illumination the confocal microscope investigated in this paper achieves its superb surface confinement by collecting the fluorescence at very high angles, exceeding the critical angle of total internal reflection. Hereby the extreme surface selectivity is obtained on the basis of the fluorescence emission properties near the interface [6,7]. This powerful alternative to TIRF, we refer to as the supercritical angle fluorescence (SAF) collection method.
2. Supercritical angle fluorescence (SAF)
Figure 1 shows the emission direction of fluorophores with isotropically orientated dipole moments at the water/glass interface, which is in biological microscopy the case of highest practical relevance. With the refractive indices of water, nw = 1.333, and glass, ng = 1.523, the critical angle amounts to 61.1°. The simulations of the angular distribution of radiation for emitters located directly at the surface and located at a surface distance equal to a third of the emission wavelength were performed according to Ref. . Obviously only surface-bound emitters send a substantial part of their intensity into supercritical angles in glass direction. This part amounts to ~34% of the emission into all directions.
The surface selectivity of the SAF method is even better than achieved in most TIRF systems. With TIRF the penetration depth of the detection volume into the analyte depends on the incident illumination angle . A high selectivity is obtained only if the illumination angle exceeds the critical angle significantly, whereas the SAF method achieves its resolution independently from the largeness of the captured angles, provided that these lie above the critical angle. A comparison of the penetration depths of the observation volumes is shown in Fig. 2. The TIRF method achieves a comparable resolution to SAF only at very large illumination angles, which are technically difficult to realize.
By illuminating the interface at supercritical angles, with TIRF the observation is restricted to the interface, whatever surface angles are chosen for the fluorescence collection. Here another advantage of the SAF technique comes into play. Since the surface can be illuminated at moderate angles, below the critical angle, the exciting beam can propagate into the analyte, opening up the possibility to additionally detect inside the aqueous solution. Consequently information about the surface-bound and unbound fractions of fluorescent analyte can be gathered simultaneously, by additionally collecting the signal below the critical angle.
For a confocal detection scheme it is necessary to produce a laterally confined excitation spot at the interface. This task is quite intricate when only supercritical angles may be used, but facile to accomplish at moderate excitation angles.
2. Composition of the microscope
We have demonstrated that parabolic glass collectors are highly efficient elements for the collection of surface-generated fluorescence [8–12]. In the proposed microscope such element is used to collect the SAF emission. The assembly of the instrument is shown in Fig. 3. A Gaussian laser beam of 0.8 mm waist diameter is enlarged by a factor of six using a beam expander consisting of two achromatic lenses (focal lengths: f1 = 20 mm, f2 = 120 mm). The expanded beam is focused with an aspheric lens onto the surface of the sample. The asphere, embedded in the parabola, has a focal length of f3 = 4 mm and a numerical aperture of 0.62. It is designed to focus a parallel beam to the diffraction limit through a glass window of 1.2 mm thickness. An asphere with exactly these specifications is commercially available from LightPath Technologies (Lens Code 350340). In the microscope the planar glass window of 1.2 mm thickness consists of a 0.15 mm thick microscope coverslip and a 1.05 mm thick bridge in the center of the parabola. Coverslip and parabola are connected with index matching liquid. The asphere captures the fluorescence emitted into the glass with the range of surface angles between 0° and 24°. Approximately 6% of the overall fluorescence is emitted into these angles. This portion of the signal is refocused with an achromatic lens (f4 = 60 mm) onto an avalanche photo diode (APD#1). The small diameter (180 μm) of its photosensitive area acts as a confocal detection aperture.
The parabolic element features a focal length of 7.1 mm and an outer diameter of 47.6 mm, which covers the range of surface angles down to 62°. An opaque disc below the parabola of 37.2 mm limits the high marginal collection angle to 75°. The neglect of higher angles is appropriate since the fluorescence emission approaches zero for surface angles approaching 90°. With a fluorescence collection efficiency of ~30% for collecting between 62° and 75° (compared to ~34% between 62° and 90°) the opaque disc will suppress background from scattered light without giving away much of the signal. The fluorescence collected by the parabola is refocused with an achromatic lens (f5 = 300 mm) onto a second avalanche photo diode (APD#2). A microscope stage can be used to scan the sample along the front face of the parabola and obtain surface images.
It should be noted that the uncommon dimensions of the optics make the use of standard microscope platforms rather unfavorable. The realization of the microscope involves the use of customized mechanical mounts, which however represents only minor technical challenge and has no effect on the performance.
We found that coverslips feature a high macroscopic planarity (≤0.5 μm per centimeter lateral translation) but their thickness usually varies by a few microns from one coverslip to the next . Accordingly the position of the submicron focus is kept at the surface at sizeable lateral translations of the coverslip, but the optics needs to be adjusted onto the surface of each coverslip individually. To balance the focus displacement the beam expander lens (f1 = 20 mm) is relocatable along the optical axis (±15 mm). A CCD camera is used to verify the focus position at the surface. The positions of the beam expander lens (f2 = 120 mm) and the asphere are fix, their distance amounts to 144 mm. The required displacement of the lens in dependence of coverslip thickness was found by raytracing calculations and is shown in Fig. 4(a). For instance a 155 μm thick coverslip (5 microns higher than the standard thickness) requires an adjustment travel of the lens of 5 mm from its home position towards the second expander lens. Due to its chosen distance of 144 mm from the rear expander lens the illumination of the asphere is scarcely influenced by the adjustment of the front expander lens. Moreover the raytracing simulations of the spatial collection efficiency demonstrate that the position and shape of the detected volume element is unaffected by the adjustment on different coverslip thicknesses. Figure 4(b) shows the collected volume element which is representative for all coverslip thicknesses within 150±10 μm.
The low angle path is analog to conventional confocal epifluorescence microscopes, because the same optical components are used for excitation and fluorescence collection. An advantageous side-effect of such configuration is that, once the laser focus is adjusted to the interface, the secondary focus is in the proper plane of confocal detection. This does not apply for the separated fluorescence collection with the parabola. When the position of the interface is altered by using glass substrates of varying thickness the secondary focus produced by the large achromatic lens travels along the optical axis. If a compensation of the high angle path is neglected and the position of the confocal plane of detection is kept constant the collection efficiency at the interface may easily even shrink down to zero like shown in Fig. 5. The use of a 160 μm instead of a 150 μm thick coverslip produces a collection volume element completely located inside the glass.
The most convenient way to accomplish consistent results with the parabolic detection channel independent from substrate thickness is to adjust the confocal detection aperture, i.e. to travel the detector along the optical axis. We found that moving the detector by 0.18 mm per micron deviation from the standard coverslip thickness of 150 μm delivers remarkably good results, provided that in doing so the detector is moved towards the sample for thinner and away from the sample for thicker coverslips. As shown in Fig. 6 this procedure generates nearly identical collection efficiencies at the interface for a wide range of coverslip thicknesses.
The adjustment of the high angles assumes to have knowledge about the particular coverslip thickness. This information is obtained directly by reading out the position of the beam expander lens. In practice both expander lens and APD#2 should be conducted by motorized translation stages and synchronized by software.
4. Resolution and sensitivity
The lateral resolution of the microscope is determined by the size of the laser focus at the interface and is identical for both fluorescence collection paths. Consequently the focusing performance of the asphere is the limiting factor of the image resolution. In order to testify the real focus properties of the asphere and to verify the new scan mechanism, the sample slides on a thin layer of immersion liquid, we developed a prototype system comprising only the low angle detection channel. A detailed description of this instrument can be found in Ref.13. It was demonstrated that indeed a high lateral resolution is achieved and that the scan mechanism allows for error-free scanning of large surface segments, on the scale of millimeters, without the requirement of a dynamic autofocus control.
Figure 7(a) shows a scanned image of 20 nm fluorescence beads immobilized at the coverslip surface using only the asphere for the collection of the signal. Figure 7(b) gives the simulated diffraction limited point-spread function (PSF) and Fig. 7(c) the experimental PSF, which is only slightly larger than predicted by theory.
The two fluorescence detection paths of the microscope generate two detection volumes of completely different shape as shown in Fig. 8. Consequently it is made possible to measure with two different resolutions along the optical axis simultaneously. The signal measured with APD#1 gives access to analyte molecules deep inside the aqueous solution, whereas APD#2 acquires data of analyte directly in front of the interface.
It has been demonstrated that the fluorescence collection only at supercritical angles allows for highly sensitive measurements  and even single molecule sensitivity . Moreover we showed that with the low angle collection of the asphere the intensity fluctuations generated by individual dye labeled proteins are measurable . But it shall be pointed out that as a result of the limited aperture of the asphere the achieved signal-to-noise ratio is clearly lower than with high aperture microscope objectives. Due to an approximately five times larger collection efficiency of the parabola however, single surface bound molecules are detectable with high signal-to-background ratios.
A direct improvement for both lateral resolution and collection efficiency is obtainable by increasing the numerical aperture of the low angle detection. Ideal would be to use the full angular spectrum up to the critical angle, corresponding to a NA of 1.33. This value is hardly achievable, as the integration of such optics will geometrically interfere with the high angle collection. Technically doable is in our opinion the use of a lens system with a NA of ~1.0. Compared to the performance of the asphere (NA = 0.62) this would more than triple the collection efficiency of the subcritical angles and nearly halve the lateral size of the PSF. However for the development of a first prototype we have chosen to fall back on the standard asphere.
5. Discussion and conclusion
With the introduction of the SAF microscope we have proposed a new microscopy technique, which can be a powerful alternative to the widely used TIRF method. There are three main advantages of SAF over TIRF. Firstly the SAF detection achieves an excellent surface confinement with any selected angles above the critical angle whereas TIRF requires very high angle illumination. Secondly the possibility to illuminate the sample at moderate surface angles allows for simultaneous detection within an aqueous analyte solution. Thirdly it can be considered to be very complicated to produce an excitation spot of submicron dimensions at the interface using TIRF. Such small focusing is important to efficiently excite the fluorophores and obtain high count rates per single molecule. Single molecule detection using widefield TIRF microscopy is a well established laboratory technique, but a confocal TIRF microscope with single molecule sensitivity we have introduced only recently [11,12]. This is indeed a quite tricky instrument concerning its implementation and operation. We are certain that the confocal SAF microscope will be by far more convenient to use.
The importance of producing very small detection volumes in order to measure very weak pair interaction has recently been pointed out  as well as the great potential of FCS experiments at an interface . Both are potential fields of application of the confocal SAF microscope. Additionally the microscope can be applied to imaging, e.g. cell membranes, and is especially promising for measurements of parallel binding kinetics on microarrays by scanning a biochip in course of the reactions.
The authors wish to thank Prof. Dr. Stefan Seeger for his continuous support. This research has been supported by EU-Grant No. QLK3-CT-2001-00278.
References and links
1. D. Axelrod, “Total internal reflection fluorescence microscopy,” in Methods in Cellular Imaging, Ammasi Periasamy, ed. (Oxford University Press, New York, 2001), pp. 362–380.
2. R. L. Hansen and J. M. Harris, “Total internal reflection fluorescence correlation spectroscopy for counting molecules at solid/liquid interfaces,” Anal. Chem. 70, 2565–2575 (1998). [CrossRef] [PubMed]
4. A. M. Lieto, R. C. Cush, and N. L. Thompson, “Ligand-receptor kinetics measured by total internal reflection with fluorescence correlation spectroscopy,” Biophys. J. 85, 3294–3302 (2003). [CrossRef] [PubMed]
5. M. J. Levene, J. Korlach, S. W. Turner, M. Foquet, H. G. Craighead, and W. W. Webb, “Zero-mode waveguides for single-molecule analysis at high concentrations,” Science 299, 682–686 (2003). [CrossRef] [PubMed]
6. T. Ruckstuhl and S. Seeger, WO 9946596 (1999).
7. J. Enderlein, T. Ruckstuhl, and S. Seeger, “Highly efficient optical detection of surface-generated fluorescence,” Appl. Opt. 38, 724–732 (1999). [CrossRef]
13. T. Ruckstuhl, A. Walser, D. Verdes, and S. Seeger, “Confocal reader for biochip screening and fluorescence microscopy,” Biosens. Bioelectron. (to be published).