1. Introduction

The demand for higher spatial resolution in optical microscopy has substantially promoted the development of novel techniques for super-resolution beyond the diffraction limit. At present, state-of-the-art techniques based on, for example, stimulated emission depletion [1], molecular photoactivation [2,3], and structured illumination [4] enable significantly enhanced resolution in biological imaging. Even though these microscopes are commercially available, the requirements for specialized and expensive instruments, intricate optical systems, or additional data processing may hinder the widespread use of the techniques.

On the other hand, confocal laser scanning microscopy (CLSM) is now routinely used as one of the essential instruments by which to investigate the structure of biological samples. If the spatial resolution is enhanced in conventional CLSM without any complexities, bioscience researchers can derive the full benefit of CLSM. Since scanning of a focused laser beam is used in the CLSM illumination system, the size of the focal spot primarily restricts the lateral resolution of images. Therefore, CLSM using a conventional laser beam, namely, a linearly or circularly polarized Gaussian beam, cannot resolve structures smaller than the focal spot size, which is commonly known to be approximately half the wavelength. Thus, in order to improve the spatial resolution in CLSM, the focal spot size must be below the diffraction limit.

The longitudinal electric field produced by a tightly focused, radially polarized (RP) beam [5] is expected to show such a smaller focal spot [6]. R. Dorn et al. experimentally demonstrated that the longitudinal field can be considerably enhanced by the use of an annular aperture in front of a focusing lens, resulting in a focal spot size significantly smaller than that for a linearly polarized beam [7]. However, an annular aperture generally gives rise to severely low transmittance of light. The insertion of an annular aperture in front of an objective lens in CLSM will cause the attenuation of not only the excitation light but also the fluorescence signal. Recently, we have numerically revealed that the focal spot by the longitudinal field decreases without loss when an RP beam possesses several π phase shifts in the beam cross section [8]. This characteristic has also been explored numerically by several research groups [9–11].

In addition to numerical studies, a few studies using an RP beam have recently demonstrated enhanced resolution in fluorescence imaging with the assistance of annular beam focusing [12], surface plasmons [13], and two-photon excitation at a dielectric interface [14]. Although an RP beam is quite attractive as an excitation beam for CLSM, previous studies were carried out under limited experimental conditions and with specialized systems. Furthermore, these techniques require modifying certain parts of an existing microscope system. Thus, the more convenient approach to readily enhance the spatial resolution in conventional CLSM is favorable for practical achievement.

In this paper, we developed a pair of liquid crystal devices (LCDs) that convert a linearly polarized (LP) Gaussian beam into an RP beam with six concentric rings, which can be referred to as a higher-order radially polarized (HRP) beam. We investigated the role of the confocal aperture (CA) for the tight focusing of an HRP beam. The side-lobes of an HRP beam can be effectively suppressed by a practical size of the CA used in conventional CLSM. Owing to the smaller focal spot characteristics of an HRP beam, we experimentally demonstrated the enhanced lateral resolution in fluorescence imaging without any modifications of the system except for the addition of the LCDs.

2. Higher-order radially polarized beam excitation

2.1 Mode conversion

A schematic diagram of beam conversion to produce an HRP beam is shown in Fig. 1 . The converter consists of two LCDs. One is a binary phase mask (device 1 in Fig. 1) to modulate the phase distribution of an input LP Gaussian beam. We designed a phase mask device bearing six concentric zones and applied a driving voltage so as to produce a π phase shift between the adjacent zones. The other device (device 2 in Fig. 1) operates as a polarization converter that behaves as a duodecimally segmented half-wave plate. An LP Gaussian beam is thus changed into an RP beam with six concentric rings.

 

Fig. 1 Schematic of beam conversion using LCDs.

Download Full Size | PDF

The intensity distribution near the focus depends strongly on the phase pattern of device 1. Several phase mask designs have been numerically investigated to obtain a smaller focal spot composed by the longitudinal field [9–11]. The higher-order transverse mode of a Laguerre-Gaussian RP beam, which has a multi-ring-shaped intensity pattern with intrinsic π phase shifts in the beam cross-section, is also a promising candidate for generating a smaller focal spot under the tight focusing condition [8]. In general, however, the decrease in the size of the center lobe at the focal plane comes at the expense of an increase in the strength of the side-lobes in a high numerical-aperture (NA) focusing. Therefore, in order to produce a sharp focal spot with weak side-lobes, we determined the radius of each zone, indicated by r_n (n = 1, ..., 6), in accordance with the node spacing of a sixth-order Laguerre-Gaussian RP beam mode, which is expressed by a Laguerre polynomial and a Gaussian function [8,15]. In the present study, the device 1 was designed for a water immersion lens with NA = 1.2, whose pupil radius is r ₆ = 3.6 mm (UPlanApo60xW, NA 1.20, Olympus). The ratios of the radii r_n are r ₁ = 0.15 r ₆, r ₂ = 0.28 r ₆, r ₃ = 0.41 r ₆, r ₄ = 0.55 r ₆, and r ₅ = 0.72 r ₆, respectively. The driving voltage (square wave with a frequency of 1 kHz) was applied to the second, fourth, and sixth zones of the device 1.

In the polarization converter (device 2), each rubbing direction of the twelve segments was designed to behave as a half-wave plate to achieve radial polarization. The fast axis of each segment is indicated by a double-headed arrow in Fig. 1. When the driving voltage is applied so as to produce a π phase difference between the fast and slow axes in each segment, the polarization device can transform a linearly polarized beam into a radially polarized one. As a result, a beam converted through these LCDs behaves as an HRP beam.

The phase mask (device 1) and polarization converter (device 2) were inserted between the objective lens and the objective turret of an upright confocal laser-scanning microscope system (FV1000 and BX61WI, Olympus). When an LP beam was preferred, zero phase shift was applied to the LCDs by adjusting the driving voltage. Therefore, thanks to the convenience of LCDs, the conversion of polarization and phase patterns for an input laser beam is readily electrically-controllable. In the experiments, a diode laser with a wavelength of 473 nm was used for excitation.

2.2 Numerical simulations for the tight focusing of an HRP beam

Figure 2 shows calculated intensity distributions near the focus using vector diffraction theory [16]. In the case of an LP beam [Figs. 2(a) and 2(d)], the intensity pattern in the focal plane is no longer cylindrically symmetric and represents the elongation along the polarization direction (x-axis) due to the longitudinal component in high-NA focusing [16]. The full-width half-maximum (FWHM) values for the focal spot along the x- and y-axes are 0.76λ and 0.59λ, respectively, where λ is wavelength of the excitation beam. In contrast, when an HRP beam

 

Fig. 2 (a) and (b) show the calculated intensity distributions in the focal plane (z = 0) for focusing of the LP (polarized along the x-axis) and converted HRP beams, respectively. In these calculations, NA = 1.2 and the refractive index of the surroundings (n = 1.33) were used. (c) The corresponding intensity profiles along the radial direction (x- and y-axes). The dashed and dashed-dotted curves indicate the intensity profiles of a focal spot of an LP beam in (a) along the x- and y-axes, respectively. The red solid curve shows the intensity profile of the HRP beam focusing. The inset in (c) shows the magnified profiles near the focus. (d) and (e) show the calculated intensity distributions in the z-x plane for the focusing of the LP and HRP beams, respectively (f) The corresponding intensity profiles along the z-axis in (d) and (e). The dashed and solid curves in (f) are the intensity profiles for LP and HRP beams, respectively. The scale bars in (a), (b), (d), and (e) are 1λ ( = λ₀/n, where λ₀ is the wavelength of a focused light in a vacuum).

Download Full Size | PDF

converted by the LCDs is tightly focused [Figs. 2(b) and 2(e)], the center focal spot is cylindrically symmetric and elongated along the z-axis, as compared to the results for LP beam focusing. The FWHM of the focal spot is 0.50λ, which is reduced by 34% from that of an LP beam along the polarization direction. Thus, a sharper focal spot can be produced at the cost of elongation in the axial direction [8], which is also referred as a light needle [11]. The outer side-lobe seems to exhibit 12th-order discrete rotational symmetry caused by the duodecimally segmented pattern of the polarization converter. Although the local maximum intensity of a side-lobe is 17% of the peak intensity of the center spot, we expect to suppress the undesirable fluorescence signal from the outer side-lobes due to a CA of CLSM.

3. Results and discussion

Figure 3 shows the horizontal profiles of the imaged fluorescence yellow-green beads having a diameter of 0.17 μm (Invitrogen). In order to estimate the size of the imaged beads, we averaged the measured profiles of ten isolated beads under each condition. For the CAs of 800 and 100 μm, the FWHMs of the center peak for an LP beam were determined to be 350 and 292 nm, respectively, where the LP beam was polarized along the horizontal axis in Figs. 3(a) and 3(b). In the case of an HRP beam, the FWHMs for CAs of 800 and 100 μm were reduced to 194 and 188 nm, which reveals a sharper lateral width than in the case of an LP beam. Figure 3(e) plots the variation of the intensity profile for HRP beam excitation with different CA sizes. Although the first (arrow head) and second (blue arrow) side-lobes appear for the CAs of 200 μm and larger, these side-lobes are effectively suppressed for the CA of 100 μm. Consequently, the CA of 100 μm (which is nearly identical to one airy unit in this system) can sufficiently eliminate the side-lobes in HRP beam excitation.

 

Fig. 3 Images and intensity profiles of a 0.17 μm fluorescent bead for the LP [(a) and (b)] and HRP [(c) and (d)] beams with CA sizes of 800 and 100 μm. Measured intensity profiles along the horizontal axis were plotted as gray dotted curves. Red curves indicate the numerically predicted profiles. The insets show the representative images of the bead for each condition. (e) The variation of the intensity profile for HRP beam excitation with different CA sizes.

Download Full Size | PDF

The expected profiles estimated by simulating an image of a 0.17 μm bead under confocal imaging were also shown by solid lines in Figs. 3(a)–3(d). The numerical simulation was based on a theoretical model derived by various authors [17–20]. In the present study, for simplicity, we assumed that the fluorescence emission is incoherent with respect to the focused excitation beam and is therefore randomly polarized. Based on this assumption, the point spread function of the confocal laser scanning microscopy (PSF_conf) can be expressed as ${\text{PSF}}_{\text{conf}}={\text{PSF}}_{\text{ex}}\cdot \left({\text{PSF}}_{\text{em}}\otimes D\right)$, where ⊗ denotes the convolution and PSF_ex and PSF_em are the point spread functions for excitation and fluorescence emission, respectively. Here, D represents the sensitivity function of the detector, which is assumed to be uniform within the aperture of a rectangular confocal pinhole. The PSF_ex, which is equivalent to the intensity distribution near the focus of an excitation beam with a wavelength (λ_ex) of 473 nm, was calculated based on the vector diffraction theory, as shown in Figs. 2(a) and 2(b). We assumed that the beam size of an incident LP beam corresponding to the 1/e ² width of the Gaussian profile coincides with the pupil radius of the objective lens. On the other hand, for the PSF_em, the intensity distribution in the detector plane for the incoherent and randomly polarized emission from a fluorescent molecule was numerically simulated based on the mathematical formula proposed in [19,20]. The wavelength for the emission was chosen to be 1.1λ_ex, corresponding to the center wavelength of the emission spectrum. Finally, the simulated intensity profile for the image of a fluorescent bead with a diameter of 0.17 μm was estimated by taking into account the 2D convolution of the PSF_conf and a finite-sized fluorescent bead, the distribution of which is assumed to be a planar disk with the same diameter as the fluorescent bead.

A comparison of the measured profiles with the simulation results revealed good agreement for both CA sizes, indicating that the sharp focal spot by the longitudinal component of an HRP beam was successfully obtained at the focus. The subtle difference between the experimental and theoretical profiles, which is especially notable in LP beam excitation [Figs. 3(a) and 3(b)], may be partly attributed to the slight polarization-dependent aberration of the objective lens. Such an aberration is, however, expected to be almost negligible for HRP beam focusing due to all the incident rays being p-polarized on the surface of the objective lens if the optical axis of the objective lens coincides with the direction of the incident beam.

The combination of an HRP beam having an appropriate CA size enabled finer images to be obtained. In order to verify the resolution improvement, we observed aggregated fluorescent beads using LP and HRP beams with a CA of 100 μm as shown in Fig. 4 . Confocal images obtained by an LP beam were very blurry and individual beads could not be distinguished because the PSF_conf for an LP beam was larger than the bead diameter [Fig. 4(a)]. Using an HRP beam, however, each of aggregated 0.17 μm fluorescent beads was finely distinguished and countable [Fig. 4(c)]. The resolution of the proposed novel microscopy is thus much higher than that of conventional microscopy. The weak ripples around each bead in the confocal image by an HRP beam [Fig. 4(c)] can be attributed to a residual side-lobe for the HRP beam focusing with a CA of 100 μm. These ripples will be further reduced if a smaller CA size is employed in the confocal microscope system. For the direction along the z-axis, the intensity pattern became slightly longer than in the case of an LP beam [Figs. 4(b) and 4(d)]. The elongation effect is thought to have been smaller than that shown in Figs. 2(e) and 2(f) because the smaller confocal pinhole also contributed to the suppression of the elongation along the z-axis [21].

 

Fig. 4 Confocal images of aggregated fluorescent 0.17 μm beads by LP [(a) and (b)] and HRP beams [(c) and (d)]. (a) and (c) are images in the focal plane. (b) and (d) are images in the y-z plane at the position indicated by the green arrow head. The scale bar (500 nm) in (a) is valid for all of the figures.

Download Full Size | PDF

Next, we applied the proposed resolution enhancement technique to the visualization of intracellular fine structures. The microtubule is a kind of cellular skeletal structure with a diameter of approximately 25 nm and is used for the evaluation of resolution [22]. For the imaging of microtubules, COS-7 cells fixed by formaldehyde and treated by detergent to permeabilize were stained with anti-α-tubulin antibody (clone DM1A, Cell Signaling) and AlexaFluor488-conjugated anti-mouse IgG antibody (Invitrogen).

Figure 5 shows the confocal images of microtubules in a COS-7 cells using LP and HRP beams with a CA of 100 μm. Although the images were obtained using an LP beam, the fibers in the magnified images [Figs. 5(b) and 5(d)] were blurred and their widths were measured as approximately 300 nm. Meanwhile, confocal images obtained using an HRP beam clearly revealed fine mesh structures that cannot be identified using an LP beam in the same field of view [Figs. 5(c) and 5(e)]. Compared to the results in Fig. 4, no apparent ripples caused by the residual side-lobe for the HRP beam focusing were observed in the confocal images. This is partly due to a complex three-dimensional structure of the microtubules. Because the intensity distribution along the z-axis for the HRP beam focusing shows slightly longer than that of an LP beam [Figs. 4(b) and 4(d)] even under the confocal imaging condition, the fluorescence signal from out-of-focus planes may obscure the resolution enhancement of the confocal image by an HRP beam. Nonetheless, for the intensity profiles of a single fiber of a microtubule, the resolution in biological specimens for the case of an HRP beam was approximately 30% higher than that for the case of an LP beam [Fig. 5(f)]. These results revealed the superior ability of an HRP beam to finely image small intracellular structures.

 

Fig. 5 Images of microtubules in a COS-7 cell with LP [(a), (b) and (d)] and HRP [(c) and (e)] beams. (b) and (c) show magnified images obtained using the LP and HRP beams, respectively, at the position indicated by the rectangle in (a). The scale bar (2 μm) in (b) is valid for (b) and (c). (d) and (e) show the magnified images at the positions indicated the rectangles in (b) and (c), respectively. The scale bar (1 μm) in (d) is valid for (d) and (e). (f) The intensity profiles along the short green line in (d) and (e) are indicated by the blue and red lines, respectively.

Download Full Size | PDF

The results of the present study demonstrate that an HRP beam enabled super-resolution imaging. Note that HRP beam imaging does not require modification of the fundamental design of a microscope, but only the attachment of the LCDs. Besides being convenient, this allows several resources accumulated in laboratories to be used without modification, and HRP beam imaging is applicable to biological specimens as well as the imaging of living cells and in vivo imaging in living animals. Among the most important features of this newly developed technology are its applicability to various excitation wavelengths and a lack of dependence on fluorescent spectra. The capability for multi-color imaging using multiple

wavelengths enables visualization of two or more types of small fine structures in the same specimen. This feature is very important in biological research in order to investigate the relationships between signaling molecules and changes in intracellular structures. The possibility of using LCDs for various excitation wavelengths enables the proposed method to be applied to even two-photon excitation. The simplicity and inexpensive setup of the LCDs allow HRP beam imaging to be integrated in various medical and industrial instruments. Thus, the proposed method has the potential for application to a wide range of fields, such as clinical medicine, semiconductor device physics, and nano-photonics.

4. Conclusion

We proposed a novel method to enhance the spatial resolution in CLSM using a HRP beam excitation. A HRP beam was generated simply by putting a pair of LCDs in front of an objective lens of a conventional microscope. Owing to the smaller focal spot characteristics of a tightly focused HRP beam, aggregated 0.17 μm fluorescence beads were visualized individually. Furthermore, the capability for biological imaging was also demonstrated experimentally. Hence, the method presented in this study enables us to achieve the resolution enhancement in CLSM without any modification of the system except for the addition of compact LCDs.