1. Introduction

Recently, high resolution 3D optical microscopy at the nanoscale is highly demanded in various fields of research and application like device inspection and bio-imaging. Some fluorescence techniques e.g. stochastic optical reconstruction microscopy (STORM) [1] or stimulated emission depletion microscopy (STED) [2] show promising improvement in lateral resolution. The higher the lateral resolution is, the smaller the depth of field would be. As a result, to obtain a 3D image some scanning techniques like scanning confocal microscopy [3] or selective plane illumination microscopy (SPIM) [4] have been developed. But, the above mentioned techniques are only capable of imaging fluorescent particles which limits their application and on the other hand the scanning process is time consuming and makes it impractical for imaging dynamic events.

Having the information about the phase of the object gives the possibility to extract the depth information of the given object. Holography [5] has been shown to be a practical technique in recording the phase information of the object in which a reference wave-front interferes with the object wave-front and the fringe pattern is recorded on a film. Exposing the recorded fringe pattern (hologram) to the initial reference wave reproduces the 3D image of the object. Goodman [6] showed that it is possible to record the hologram using an electronic detector and reconstruct the image digitally using a simulated computer model [7], which introduced digital holographic microscopy (DHM).

Unlike scanning systems, DHM has the ability to extract the 3D information of the object by taking only one image and gives the possibility to derive the phase and amplitude information of the object wave-front in any plane [8–13] that makes DHM more feasible and convenient. Moreover, DHM allows to investigate the shape and the displacements of objects [14,15] in the nanoscale with high resolution and it highlights its capability as a promising 3D imaging technique. By improving nanotechnology, the dimension of structures and functional devices is decreasing in size down to the nano-scale which is smaller than the wavelength of the light. The diffraction-limited lateral resolution is a barrier that makes an obstacle for DHM methods to go that deep. This limit introduced by Abbe’s criterion, where the resolution cannot exceed $k\lambda /NA$, in which k is 1 for coherent light, λ is the wavelength of the light and $NA$ is the numerical aperture of the imaging system. To improve the resolution one should increase $NA$ and/or use a shorter wavelength. One approach, to increase $NA$, is to guide light beams with higher angles to the detector. Some advances have been made in this area using synthetic aperture technique [16–21], where the higher spatial frequencies of the light diffracted by the object can be guided to the aperture by means of gratings [17,18] or oblique illumination [19–21]. In this paper we arranged a setup implementing a short wavelength (193 nm) and an objective with high numerical aperture ($NA=0.75$) and using oblique illumination technique.

2. Experiment

We employed an ArF Excimer Laser, ExiStar 200 (TUI), operating at deep UV (193 nm) with pulse duration of ~10 ns, as a light source. Some advantages of using this light source are as follows: 1) the laser has a short coherence length, in the order of 100 µm, which reduces the noise coming from the back reflections in the setup, 2) the setup does not need to operate in vacuum, 3) The optical elements made of fused silica have been used, as fused silica is transparent at 193 nm, and 4) a digital camera is commercially available for this wavelength with high UV sensitivity, which can perform direct imaging with no filter or fluorescent plate in front (PCO Sensicam em680, quantum efficiency: 20% at 193 nm, pixels 1000 × 1024, pixel size 8µm × 8µm).

2.1. Off-axis digital holography setup

Figure 1 shows a schematic of the setup. The laser beam is spatially filtered using some lenses and diaphragms to create a uniform beam profile. A polarizing beam splitter (PBS) separates the laser beam into two parts, one as a reference (dashed line) and the other for illuminating the object (solid line). A moving mirror (M₁) is installed in the reference path to adjust the relative path difference between the reference and the object paths to lie within the coherence length of the laser. To increase the contrast of the image, the interfering beams have the same polarization. In the object path a focusing lens (L₃) is located to illuminate the object. A home-designed objective (section 2.2) collects the light transmitted by the sample. To create a perfect reference, a pinhole is located in the reference arm which is directed to the CCD using a mirror (M₃).

 

Fig. 1 The schematic of the off-axis digital holographic microscope setup. ‘L’ and ‘M’ stand for ‘Lens’ and ‘Mirror’, respectively. Insight: The custom designed objective, right: aspheric lens system, left: the objective holder which is also designed to adjust the reference beam using a built-in mirror.

Download Full Size | PDF

The angle between the reference wave and the object wave should be optimized in a way to produce fringes larger than the resolution of the detecting camera (two pixels are required to record each fringe) and it should satisfy${\alpha }_{\max }\le \lambda /2\mathrm{\Delta }$, where${\alpha }_{\max }=D/d$ is the maximum possible angle (for Drefer to the objective in Fig. 1) and Δ is the CCD pixel size. As a result, the distance d should be larger than 42 cm (the smallest value forD is the diameter of the objective).

2.2. Deep-UV-objective

The custom design of the objective (see insight in Fig. 1) was done to meet the demands of the off-axis setup while having a low price for imaging with a deep UV light source. Light coming from the object plane is collimated, respectively focussed towards infinity. This is done by using a half-ball-lens and a custom-design asphere. The half-ball-lens allows the numerical aperture to reach 0.75 as the asphere is used to mainly correct the spherical aberration introduced by the half-ball-lens. Both lenses consist of fused-silica and are chosen plano-convex, that offers an excellent precondition for fitting them into an assembly with close tolerances. The focal length of the objective was set to 1 mm and it can image a field with the radius of 10 µm.

2.3. Oblique illumination approach

To achieve high resolution, oblique illumination is implemented in the setup. The zero order component is shifted to one of the peripheral sides of the objective, instead of passing through the center of the aperture (which is the case of on-axis illumination). This shifting enables more additional spatial frequencies to enter the imaging pupil, but only the higher frequencies from one side of the zero order get involved in the imaging process and the other side is removed from the view of the objective. Figure 2 shows a simple schematic of the method. The numbers “-2, −1, 0, 1 and 2” qualitatively represent the magnitude and component of the spatial frequency of the light coming from object. In the case of direct illumination (Fig. 2a) the zero order and two components of the frequencies assigned by “1” i.e. ( + 1,-1) are collected by the objective. Using oblique illumination for the given example, instead of the component “-1”, the component “2” is collected by the objective (Fig. 2b). Guiding higher spatial frequencies to the imaging system is somehow equivalent to increasing the$NA$of the imaging system. In addition, the poor visibility in conventional bright field imaging, caused by the symmetrical presence of lower frequencies is suppressed and consequently makes it possible to resolve very fine structures in the specimen.

 

Fig. 2 The diagram shows the principle of oblique illumination method. (a) Direct (on-axis) illumination: in the given example the 0th, 1st and −1st components are collected by the objective, (b) oblique illumination: 2nd component is replaced by the −1st component and consequently, the 0th, 1st, and 2nd components are guided to the aperture of the objective, (c) the schematic of the oblique illumination setup.

Download Full Size | PDF

Although the smaller structures are observable, due to the aberration caused by tilted illumination, some artifacts appear and a single image is not enough to precisely retrieve the structures. To remove this effect, we symmetrically performed imaging through the oblique illumination from four different sides and combine all images together to obtain an image with the more realistic and correct size for the structures. Figure 2c shows the utilized configuration. Using M₅, the collimated beam is redirected to the corner of the illumination lens, L₃ and hits the sample with an angle. By adjusting M₅ it is simply possible to send the beam to the different corners of L₃ and change illumination direction and angle. The angle of incident can be calculated by$\beta =\tan (D^{\prime }/l)$ which was 10° in our case (parameters are shown in Fig. 2c).

2.4. Nano-structured template

According to the Abbe criterion ($R=\lambda /NA$), the high $NA$ of the objective ($0.75$) and the short wavelength (193 nm) leads to a theoretical resolution of $R\approx 250nm$. To be able to examine the resolution, we have designed a nano-structured template (Fig. 3a ). The template is made of 35 nm of gold layer coated by ion-beam sputtering on a fused silica substrate. The patterning of the gold film is performed in FEI NovaNanoLab 600 equipped with a Raith Elphy + lithorgraphy system. The structure is cut through the gold layer using a Focused Ion Beam (FIB) with an accelerating voltage of 30 kV. Owing to the small dimensions a beam current of 1 pA and multiple loop exposure were selected. It includes square and line structures, ranging from 500 to 100 nm in width. The logo of the Institut für Technische Optik “ito” is also included in the template in which its elements have a width of ~300 nm.

 

Fig. 3 (a) The Scanning Electron Microscope (SEM) image of the nano-structured template, (b) a typical recorded digital hologram and (c) its Fourier transform, (d) the reconstructed amplitude and (e) phase of the object illuminated with on-axis illumination, (f) the reconstructed amplitude and (g) phase of the object illuminated with oblique illumination along “y” axis, (h) the reconstructed amplitude and (i) phase of the object illuminated with oblique illumination along “x” axis, (j) the final image obtained by combining the reconstructed amplitude images taken using oblique illumination from four symmetric directions, (k) the image taken by a conventional optical microscope with NA = 0.75, × 100 objective. The scale bar is 3 µm in (a), (d), (j) and (k).

Download Full Size | PDF

3. Result

To test the resolution of the setup we have used our designed template (the SEM image shown in Fig. 3a). Figures 3b and 3c show a typical recorded hologram and its Fourier transform, respectively. One of the lobes which is highlighted with a white square is used for image reconstruction. First, the experiment has been done using on-axis illumination. For this case, the reconstructed amplitude and phase of the object are shown in Figs. 3d and 3e, respectively.

To increase the resolution, the oblique illumination was performed through an incident angle of ~10° relative to the normal axis of the object plane. To avoid artifacts appearing in the image, we have performed oblique illumination from four symmetric directions. The reconstructed amplitude and phase for two selected directions are shown in Figs. 3f-3i. The illumination direction is indicated by an arrow in each image. The amplitude of the object has been separately reconstructed for each of four oblique illumination directions. Then we have combined the reconstructed amplitude images, obtained from each individual direction, by adding the complex amplitude of the raw images and without implementing any further image processing technique. A phase gradient is apparent in Figs. 3g and 3i, which is caused by tilted illumination. Nevertheless, the combined image does not suffer from phase mismatches and as a result, no phase correction process has been taken into account. The combined image is shown in Fig. 3j. In this image the “ito” logo is clear and even the line structures with the width of 250 nm are well-resolved, that confirms a significant enhancement in resolution compare to the result obtained with on-axis illumination (Fig. 3d), in which the line structures with the size of 350 nm are the smallest resolvable structures and the “ito” logo and the square structures smaller than 400 nm cannot be recognized. The image taken by an optical microscope (Zeiss Axiovert 200, with × 100 objective, NA = 0.7) is also shown in Fig. 3k for comparison. To better compare the final combined image (Fig. 3j) with the SEM image (Fig. 3a) the small structures are magnified in Fig. 4 . An inversed intensity profile is plotted for each structure size to better show the visibility of the structures (Figs. 4b and 4e). Figure 4b shows a clear profile of the 300 nm sized structures and Fig. 4e presents the profile of the 250 nm sized structures.

 

Fig. 4 A fine comparison of the SEM image of the template (right-center) and the image obtained using our DHM setup (left-center). (a) The magnified DHM image of the 300 nm sized structures, (b) the inversed intensity profile along the dashed line in Fig. 4a, (c) the magnified SEM image showing the 300 nm sized structures, (d) The magnified DHM image of the 250 nm sized structures, (e) the inversed intensity profile along the dashed line in Fig. 4d, (f) the magnified SEM image showing the 250 nm sized structures. The scale bar is 3 µm in the centered images.

Download Full Size | PDF

4. Conclusion

In this work we have developed an off-axis digital holographic microscope in deep UV, capable of recording 3D image from nanostructures. The setup has been designed with the least possible optical elements in the imaging path to avoid aberration due to the non-perfect optical elements. By implementing the oblique illumination method we are able to detect structures even down to 250 nm. The setup can be implemented in a way to be able to image in the reflection mode.