1. Introduction

Optical microscopy is one of the most important inventions in the history of mankind. The noninvasive probe of bio-specimens renders optical microscope as an irreplaceable tool for biological and biomedical researches. A major problem, however, associated with the standard optical microscopy is the intrinsic limit of resolution, known as the diffraction limit. This limit is always considered as the ultimate barrier of optical microscope to pursue higher resolution. The smallest resolvable distance between two objects is ~λ ₀/2NA, where λ ₀ is the wavelength of emission light from sample and NA is the numerical aperture of objective lens. Objects or features with separation smaller than that cannot be resolved. However, there are increasing demands of super-resolved microscopy for imaging sub-100nm scale or even molecular structures. During the past decade, several state-of-the-art imaging techniques have been reported which can achieve resolution beyond this limitation, such as confocal laser scanning microscopy (CLSM) [1,2 ], near-field scanning optical microscopy (NSOM) [3,4 ], stimulated emission depletion microscopy (STED) [5,6 ], structured illumination microscopy (SIM) [7,8 ], and saturated structured illumination microscopy (SSIM) [9]. Nevertheless, the studies on the dynamics of bio-samples require not only the noninvasive and super-resolved imaging, but also the real-time imaging with sufficient contrast. Among the aforementioned imaging techniques, CLSM, NSOM and STED suffer from the relatively low speed due to the point scanning approach, preventing them performing the real time imaging. The photobleaching of florescence signals in an SSIM is a major challenge under saturating light intensities. SIM, as a wide-field imaging technique, has been proven especially useful for real-time super-resolved biological imaging.

In a conventional fluorescence microscopy, the optical transfer function (OTF) of the objective in Fourier space can be represented by a circle with a radius of f_cutoff = 2NA/λ₀, defined as the maximum resolvable spatial frequency. Utilizing the so-called “Moire effect”, SIM is able to code higher frequency information into the detectable low frequency region to improve the resolution [8]. A novel SIM technique combining with total internal reflection fluorescence (TIRF) microscope has shown its potential for near-field fluorescence imaging with lower out-of-focus noise and weaker photobleaching of fluorophores outside the focal plane [10]. In order to enlarge the excitation field, surface plasmons (SPs), the collective charge density waves propagating on a metal-dielectric interface, were introduced to SIM. The so-called SP-SIM possesses a signal-to-noise ratio as good as TIRF-SIM owing to the strong confinement of SPs near metal surface. Despite the tremendous successes of these methods, each has its own practical limitation, e.g., the bulky and sophisticated beam control system [11], the complex substrate fabrication process [12,13 ] and the non-uniform intensity distribution with only one-dimensional enhanced resolution [14].

In this Letter, we propose a structure-free plasmonic structured illumination microscopy (PSIM), which combines SP standing-wave illumination with TIRF-SIM technique to obtain sub-100nm resolution in water-environment (refractive index n = 1.33, close to refractive index of most bio-samples). Optical vortices (OVs) with fractional topological charge are employed to generate SP standing-wave patterns with precise phase shifts of {-2π/3, 0, 2π/3}. Although Berry has pointed out that OVs with fractional topological charge are unstable during propagation because of the interaction of OVs with some integer orbital angular momentum (OAM) states [15], here we merely utilize the phase profiles of OVs to yield the desired phase shifts while the other characteristics of OVs associated with the fractional topological charges are not concerned. In the experiment, a spatial light modulator (SLM) is used to control dynamically the topological charges of incident OVs. Therefore, there is no need for the mechanical moving of optical components and the rotation of specimens.

2. Generation of plasmoinc standing-wave

Optical vortices can be described as beams with donut-shaped intensity profiles due to their phase singularities. When OVs are focused onto a thin metal film by a TIRF objective lens, the p-polarized incident radiation at the SP resonance (SPR) angle θ _SPR will couple to SP waves. Consequently, SP standing-wave patterns are formed on the metal surface due to the interference of the counter-propagating SP waves with wave vector of ± k _SP, as illustrated in Fig. 1(a) . In Figs. 1(b)-1(d) we show the calculated near-field distributions of SPs on the metal surface under the illumination with OVs of various topological charges. A series of SP standing-wave patterns can clearly be seen, of which the fringes can be shifted by tuning the topological charges of OVs. However, such kind of cylindrical-symmetric excitation configuration of SPs will give rise to arc-shaped interference fringes as well as centralized standing-wave patterns. Both of them will impair the imaging performance of a SIM system.

 

Fig. 1 Tight-focusing configuration for the excitation of SP standing-waves with OVs. (a) The schematic diagram and (b)-(d) the calculated SP standing-waves excited by linearly-polarized OVs with topological charges of 1, 2 and 3.

Download Full Size | PDF

In order to generate uniform SP standing-wave patterns, a bow-tie-shaped amplitude filter was employed as an aperture to modify the OVs, as shown in Fig. 2(a) . The aperture was fabricated on a raw polished-aluminum disk (1-mm-thickness) by a commercial diamond lathe. Such a thick disk of aluminum ensures blocking all of the undesired light beams. The diameter of the disk is 1 inch that matches a precision rotation mount used in the experiment. The diameter of internal aperture is set to be 1.5cm, which is larger than the beam size of OVs. Figure 2(b) and 2(c) show the intensity distributions of OVs (with topological charge l = 1) captured by a CCD camera before and after passing through the aperture, respectively. Focused by a high NA TIRF objective lens, the modified OV beam couples to SP waves at a pair of small arcs confined by angle of θ, as shown in Fig. 2(d). In order to generate a standard SP standing-wave pattern, the expansion angle θ should be carefully selected. When the angle is too large (θ>40°), the SPs field distribution becomes non-uniform due to the focusing issue. When it is too small (θ<5°), the interference field of SPs is similar to the two-point-source interference pattern. By considering this trade-off, the value of 20 ^o was selected in this work to get a relatively high quality interference pattern with uniform periodicity, as shown in Fig. 2(e).

 

Fig. 2 SP-standing-wave patterns generated by the modified OV. A bow-tie shaped amplitude filter (a) was employed for shaping a full-intensity OV (b) to a bow-tie shaped intensity distribution (c). In the tightly focus configuration, the excitation position was cut to a smaller pair of arcs (green solid arcs in (d)). A standard SP-standing-wave pattern (e) with uniform periodicity could be generated by two counter-propagate SPs waves toward the silver film center. The successful excitation of SP waves could be confirmed by the dark lines (f) in the image obtained at the back focal plane.

Download Full Size | PDF

The excitation of SPs can be confirmed by the reflected beam captured at the back focal plane of the objective lens (Fig. 2(f)). An oil-immersion objective lens with NA = 1.49 was used in our imaging system, corresponding to a maximum allowed incident angle of 79 degrees, well covering the SPR angle (74 degrees) for a water-metal interface. As shown in Fig. 2(f), a pair of dark arcs can clearly be identified in the image obtained at the back focal plane, indicating the excitation of SP waves at the water-metal interface. In contrast, s-polarized beam is unable to excite SP waves. Under the illumination with s-polarized beam, the dark arc disappears, indicating that no SP wave is excited.

3. Experiment setup

Figure 3 shows the schematic diagram of the proposed structure-free super-resolved imaging system. A solid-state laser of wavelength of 532nm is served as the excitation source. After expansion, the laser beam is incident onto a phase-type SLM, which encodes a computer generated hologram (as shown in Fig. 3(a)) into the wavefront of a linearly polarized beam to generate OV beam. The polarization direction of OVs is aligned by using a half waveplate to match the orientation of bow-tie shaped aperture. When the modulated OVs are focused by a high NA objective lens, SP standing-waves are formed on the metal surface and sequentially illuminate the fluorescent beads dropped on the silver film. The emitted light from the fluorescent beads is able to couple back through the metal film, a phenomenon known as the surface plasmon coupled emission (SPCE), and is collected by the same TIRF lens. Previous studies indicated that the SPCE could enhance the fluorescence collection efficiency to nearly 50% meanwhile suppress the background noise [16]. Though SPCE leads to a donut-shaped point spread function (PSF), a deconvolution algorithm is available to convert this donut-shaped PSF into a conventional Airy disk shaped PSF. A long-pass edge filter is utilized to filter the reflected laser beam into CCD camera. A typical image collected by CCD camera is shown in Fig. 3(d).

 

Fig. 3 Schematics of the PSIM system. A SLM, a half waveplate and an amplitude filter (θ = 20°) were used for dynamically controlling the phase (topological charges), polarization direction and amplitude of the incident OVs. The fluorescent beads were deposited onto the silver film. The emission light from the fluorescent beads was coupled back through the silver film and was collected via the same objective. Due to the SPCE phenomenon, the doughnut shaped PSFs were obtained. Insert: (a) computer generated hologram (CGH), (b) amplitude filter, (c) calculated intensity distribution of SP-Standing-wave pattern on silver film, (d) typical SPCE image of florescent beads.

Download Full Size | PDF

In order to obtain a super-resolved image, at least three intermediate SPCE images under the illumination with three different SP standing-waves with relative phase shifts are required. To achieve this, the SLM-controlled illumination scheme offers a more flexible way than by using manually-controlled optical path difference scheme. The phase shift of the standing-waves is linked to the initial phase of incident beam at the excitation arcs (as shown in Fig. 2(d)), which can approximately be expressed as Δl × π, where Δl is the differential topological charge between two sequential OVs. To achieve the precise phase shifts of {-2π/3, 0, 2π/3}, OVs with fractional topological charges of {1, 1.66, 2.34} were used. Compared to the OVs with integer topological charges, the fractional OVs possess cracks at their intensity profiles because of the step discontinuity of phase. Figures 4(a)-4(c) presents the intensity profiles of OVs with topological charges of 1, 1.66 and 2.34, respectively. As can be seen, the donut-shaped profile is broken due to the non-integer topological charges (Figs. 4(b) and 4(c)). Therefore, it is important to select the orientation of the bow-tie-shaped aperture to avoid the influence from the intensity crack, as illustrated in Figs. 4(b1) and 4(c1). By illuminating with the modified OVs, a set of quasi-standard SP standing-wave patterns with phase shift of {-2π/3, 0, 2π/3} were generated on the metal surface, as shown in Figs. 4(a2) - 4(c2). Figure 4(d) shows the cross section of the SP standing-wave patterns generated by the modified OVs.

 

Fig. 4 Precise phase shifts of {-2π/3, 0, 2π/3} achieved by OVs with fractional topological charges. (a)-(c) Full-intensity distributions of OVs with topological charges {1, 1.66, and 2.34}. (a1)-(c1) Intensity distribution of modified OVs after amplitude filter. (a2)-(c2) Calculated intensity distribution of SP-standing-wave patterns excited by (a1)-(c1). (d) Intensity cross sections of (a2)-(c2).

Download Full Size | PDF

4. Experimental result

In the experiment, florescent beads with diameter of 28nm (peak emission wavelength at 575nm) were deposited on a thin silver film (thickness of ~45nm), which is subsequently covered with deionized water (refractive index ~1.33) to create the aqueous environment. The experimental results are shown in Fig. 5 . As shown in Fig. 5(a), the image with donut-shaped PSFs was captured by CCD camera, when the fluorescent signal coupled through the silver film and collected by the objective lens. In order to convert an original donut-shaped PSF to a conventional Airy disk shaped PSF, Richardson-Lucy (R-L) deconvolution algorithm was applied. Reasons for using this algorithm are the rapid convergence rate by using fast Fourier transforms, the non-negative constraint during iterations and the availability of a built-in R-L deconvolution function ‘deconvlucy’ in MATLAB (The MathWorks, Inc.) image toolbox. The signal-to-noise ratio could also be improved slightly by using this algorithm. The converted image is shown in Fig. 5(b). Three intermediate images were obtained by changing two times the topological charge of incident OVs as mentioned above. Subsequently, followed by the algorithm of SW-TIRF [17], the super-resolved image was obtained as shown in Fig. 5(c). Figure 5(d) shows the corresponding cross-sectional distribution of fluorescence intensity along the red dashed line in Fig. 5(c). Two beads with distance of 149nm are resolved. The full-width at half-maximum (FWHM) of one bead was measured to be ~95nm, indicating that our proposed imaging system is capable of obtaining sub-100nm imaging resolution. To get the 2D lateral super-resolved images, another set of SP standing-wave patterns in the orthogonal direction is required. It can simply be achieved by rotating both the half waveplate and the bow-tie shaped aperture with 90 degrees. Meanwhile, the holograms loaded into SLM should be modified accordingly to avoid the influence from the “intensity cracks” that are generated by the OVs with fractional topological charges.

 

Fig. 5 Demonstration of PSIM system with sub-100nm resolution. (a) Original SPCE image with doughnut-shaped PSFs. (b) Image after R-L deconvolution algorithm processing. (c) Super-resolved image by applied reconstructed algorithm to three intermedia images. (d) The corresponding fluorescence intensity cross-section in (c).

Download Full Size | PDF

5. Conclusion

In summary, we propose and demonstrate a structure-free PSIM system. By using modified OVs with fractional topological charge, a set of quasi-standard SP standing-wave patterns with phase shifts of {-2π/3, 0, 2π/3} are generated to excite florescent beads. We experimentally demonstrate that the imaging system is capable of achieving sub-100nm resolution in aqueous environment. Due to its structure-free configuration and wide-field imaging capability, the demonstrated PSIM technique has potential for low-cost biological dynamic imaging applications.