1. Introduction

With the development of nanoscience and technology, improving the lateral resolution of optical microscopes under direct observation becomes highly demanding [1–3]. However, the resolving power of a conventional optical microscope is limited by Abbe diffraction, which is approximately half of the incident wavelength. A variety of efforts has been devoted to imaging subwavelength objects in the past decades. One approach is the scanning near-field optical microscopy (SNOM) [4], which collects near-field information point-by-point using an aperture with a diameter smaller than the excitation wavelength. With this method, the resolving power is usually limited by the aperture size. In contrast, with the method of stimulated emission depletion (STED) [1, 5], the point spread function (PSF) is reduced by using a second doughnut-shaped laser beam. Through bleaching the excitation back to the ground state, only molecules in the center of the doughnut beam are able to emit fluorescence. This can physically narrow the PSF down to a subwavelength size, and hence, increase the resolution beyond the diffraction limit. Both SNOM and STED are scanning based methods and are therefore time consuming. Structured illumination microscopy has been widely used for life science imaging [1, 6, 7], by which a spatially structured illumination is employed to transfer object information containing higher spatial frequencies. However, multiple image captures and post processing are still required and the resolution is at most doubly enhanced [8].

Other than the methods mentioned above with post possessing or point scanning which are time consuming, plasmonic superlenses have been envisioned as a promising approach for real-time subwavelength imaging [2, 3]. Nevertheless, some intrinsic limitations significantly restrict their applications. For instance, a thin silver film can image deep subwavelength objects, but only from near field to near field [9]. Plasmonic hyperlenses with cylindrical or spherical multilayer structures were proposed to achieve subwavelength imaging with magnification [10–12]. Yet the hyperlenses suffer a high optical loss due to the stack of a large number of metal and dielectric films. The object plane lying inside the curved lens surface makes it further impractical [10, 11]. Alternatively, a metal slab with a periodic surface corrugation was proposed for far-field subwavelength imaging [13]. Due to an incomplete wave transformation in terms of spatial frequency, the image needs to be reconstructed through combining both s- and p-polarized waves. Furthermore, the lens works only at a wavelength close to the surface plasmon frequency of the metal, which is ultraviolet in the case of silver, the metal of which is also used in this report.

Recently, we have proposed a cascaded plasmonic superlens to achieve subwavelength imaging in the far field under direct observation at visible wavelengths. The superlens is composed of two metamaterial slabs, designed based on a suggestion from Ma et al [14–16]. The first part is a metallic meander cavity structure (MCS) used to couple and support the propagation of evanescent waves emanated from subwavelength objects. The second part is a planar plasmonic lens (PPL) used for phase compensation.

In this report, we demonstrate fabrication and imaging results of the cascaded superlens. First, fabrication and characterization results of each element constituting the lens are reported. Then the imaging property of the complete superlens is presented. As a result, the superlens achieves a lateral resolving power of 180 nm at $\lambda$ = 640 nm. The key element for this achievement is the PPL, while MCS can be replaced by other metasurfaces. To our knowledge, this is the first experimental demonstration to employ a PPL for this aim. We expect that our results would initiate further studies on plasmonic superlenses. Due to its compact size and fabrication feasibility for two-dimensional arrays, this superlens can find potential applications for miniaturized photonic devices and planar integrated circuit technologies [17, 18].

2. Design and numerical calculation results

By designing and calculating the superlens, an in-house developed software MicroSim based on rigorous coupled wave analysis method was used [19]. A cross-sectional schematic of the cascaded plasmonic superlens is shown in Fig. 1(a), which is composed of a metallic meander cavity structure (MCS) and a PPL slab.

 

Fig. 1 (a) Cross-section schematic of the cascaded plasmonic superlens with D_PL = 400 nm, D_C = 50 nm, D_OM = 70 nm, D_O = 100 nm, D_spa = 70 nm, P_x = 400 nm, W_r = 170 nm, d = 30 nm, and t = 50 nm. The grating in the y-direction is infinite. (b) Calculated far-field image intensity in the xz-plane from a double-slit object with X_D = 200 nm and a slit width of 100 nm from the lens shown in (a). (c) Image profiles along the x-axis of the field shown in (b) for two slit distances with the superlens. Also a calculated image profile for an object with X_D = 360 nm in the absence of the superlens (SL) is shown for comparison.

Download Full Size | PDF

Stacked metallic meander structures have been demonstrated to possess extraordinary optical properties at visible wavelengths due to the excitation of short and long range surface plasmons polaritons (SPPs), and complex mode interactions in both the near-field and the far-field regimes [20–22]. A broadband optical response in terms of spatial frequency suitable for near-field wave transformation can be induced, and with them subwavelength imaging was demonstrated numerically [15, 21–23]. In our current approach, however, only two layers of the metallic meander structure are used. Although a larger bandwidth of optical response can be generated by stacking more meander layers, the transmission will be reduced dramatically. Furthermore, it is sufficient to use the two-layer structure to obtain an angular dispersion suitable for subwavelength imaging, as we will show below. The geometrical parameters of the MCS shown in Fig. 1(a) are optimized for the observation wavelength of 640 nm. Among them, the corrugation depth is t = 50 nm, the periodicity is P_x = 400 nm, and a film thickness is d = 30 nm. All other parameters are listed in the figure caption.

The design of the PPL is based on the principle that SPPs propagate through nanometer slits in a waveguide mode, and the wave propagating constant depends on the slit width [24, 25]. The transverse magnetic mode of the slit follows the dispersion relationship [24]

The performance of the lens is calculated as it is measured, i.e., the complete structure shown in Fig. 1(a) is observed under a microscope with NA = 1.3. The object is illuminated by a plane wave in the xz-plane with p-polarization (with the electric field along the x-axis) at normal incidence at λ = 640 nm. In the calculation, the permittivity for Ag was taken from [26] and the refractive indices for spin-on-glass (SOG, Futurrex ICI-200) and MgF₂ is 1.41 and 1.385, respectively. Field distributions in the xz-plane for objects with different X_D were calculated and the image field from an object with a slit width of 100 nm and X_D of 200 nm is presented in Fig. 1(b). A far-field image with magnification around z = 0 – 450 nm is clearly shown. Imaged field profiles along the x-axis are plotted in Fig. 1(c) for the object with two different X_Ds. When a contrast criterion of 0.77 is taken, which is sufficient for our CCD camera, the resolvable lateral resolution with the superlens turns out to be 180 nm. With the same criterion, the calculated resolution limit in the absence of the superlens is 360 nm, as shown by the blue curve in Fig. 1(c). This indicates that the resolving power of the microscope in the presence of the superlens is doubled.

The key element for achieving this improvement is the PPL. Although lens behavior of PPLs has been demonstrated almost decade ago [24, 25], they alone cannot be applied for near-field imaging due to their strong interaction with the object. In our design, however, near field interaction is greatly suppressed when the PPL is combined with a MCS slab. To clarify this statement, near field transmissions as a function of k_x/k₀ were calculated, in which k_x is the lateral wave vector and k₀ is the wave vector in free space. As shown schematically in Fig. 2(a), the calculations were performed with the elements in a medium environment as they are in the lens. Near-field transmission was obtained by normalizing the near field 70 nm behind the element to the incident field and the results are plotted in Fig. 2(b). A passband for both propagating wave and evanescent waves are shown for the PPL slab. However, a huge transmission peak is manifested around k_x = 1.6k₀ = 0.5K_g, in which K_g is the wave vector of the slit array. This explains why the PPL is not suitable for near-field imaging. Similarly, there is also a strong transmission peak with the MCS around k_x = 1.8k₀. In contrast, when the PPL is combined with the MCS, a flattened and broadened dispersion is generated without reducing transmission, as shown by the blue curve in Fig. 2(b). It should be emphasized that in our design, the PPL behaves more like a high-index immersion lens than a conventional geometrical lens, except for that it has a flat design.

 

Fig. 2 Near field transmission dispersion of the sub-components and the cascaded superlens. (a) Configuration for the calculation at λ = 640 nm and (b) near-field transmission curves of different elements as a function of k_x/k₀.

Download Full Size | PDF

3. Fabrication and characterization of sub-components of the superlens

Here we describe the fabrication and characterization of the two sub-components of the cascaded superlens: the metallic meander cavity structure the planar plasmonic lens. To validate the numerical calculations shown in Fig. 1(b)-1(c), we fabricated the superlens using a focused ion beam (FIB) milling machine (Helios NanoLab600 DualBeam from FEI). An electron-beam evaporation machine was used to deposit metals (Cr, Ge and Ag) or MgF₂ films. Before fabricating the complete superlens, sub-components in the lens were first fabricated and characterized. This ensures that each component works as is predicted by the numerical calculations.

Process schematic for the fabrication of the MCS is shown in Fig. 3. First, a layer of 100 nm SOG is spin-coated on a glass substrate (Step 1). Then 20 nm Ag and 10 nm Cr are subsequently deposited onto the SOG, which work as sacrificial layers for FIB milling (Step 2). The SOG-grating is then milled by FIB (Step 3) according to the parameters shown in Fig. 1(a). Afterwards, the Cr/Ag layers are removed by wet etching (Step 4). To obtain a single-layer Ag meander structure, films in a sequence of Ge (1 nm)/Ag (30 nm)/MgF₂ (5 nm) are deposited. Here Ge is used as a wetting layer [27], while MgF₂ is used to protect the Ag film from oxidation (Step 5 without the second Ag layer). In calculations, Ge is not considered since with it the optical property of the deposited Ag film approaches that of its bulk material. To obtain a double-layer meander structure, films in a sequence of Ag/MgF₂/Ag/MgF₂ (5 nm) were deposited atop the SOG grating (Step 5).

 

Fig. 3 Illustration of the fabrication process for the MCS slab.

Download Full Size | PDF

Cross-section images of the fabricated single-layer and double layer-meander structures prepared by FIB milling are shown in the inlets of Fig. 4(a) and 4(b) correspondingly. The bright meander lines are from the corrugated Ag films. Geometrical parameters approaching the designed ones with respect to the metal thickness, grating height and aspect ratio are demonstrated. In addition, a good conformity of the top layer with the bottom layer in the double layer meander structure is obtained.

 

Fig. 4 Measured (red curves) and simulated (blue curves) transmittance spectra (a) for a single layer and (b) double layer Ag meander structures (with a distance of 70 nm) fabricated on glass substrates. The periodicity of the grating is 400 nm, the grating height is 50 nm, and the thickness of the Ag films is 30 nm. Inlets show the corresponding SEM cross-sectional images of the two structures.

Download Full Size | PDF

To estimate the quality of the fabricated meander structures, transmittance spectra were measured using a Fourier Transform Infrared Spectrometer (Bruker Vertex-80). For this aim, gratings with an area of 50 × 50 µm² were fabricated, and the surface of both single-layer and double-layer meander structures was planarized with 80 nm SOG in order to obtain a homogeneous dielectric environment for surface plasmons, which is critical for obtaining the predicted optical properties [28]. As plotted in Fig. 4(a) for the single layer and in Fig. 4(b) for the double layer structure, the measured spectra agree roughly with those of the calculated. The noisy signal at shorter wavelengths is due to the weak sensitivity of the silicon photo detector. The transmittance peak with the double-layer structure is around 4%, which is much lower than 45%, that from the single-layer structure. This phenomena is attributed to the near-field interaction as we have shown in [20], since the cavity has only a distance of 70 nm. Once the two layers have a distance within the far-field interaction regime, the transmittance of the double-layer structure would be approximately the square of the transmittance from the single layer. Actually, an ideal near-field lens should have possibly a smooth field transmission over a large bandwidth in spatial frequency domain [29]. Therefore, in our case it is necessary to suppress the strong transmission of the propagating waves. The discrepancy between the measured and calculated curves can be explained by different illumination configurations. With the FTIR spectrometer, an objective lens with NA = 0.4 was used. Hence, the sample was illuminated by a focused light, but not a plane wave as with the simulation. As a result, the measured spectra from the meander structures are rounded at the passband edges. This also red-shifts the spectra since the dispersion of the plasmonic meander structure is negative [20]. When a well collimated light source was used for the measurement, as we have performed in [30], a much better agreement between the simulated and measured curves could be obtained. However, for that measurement a much larger structured area (0.8×0.8 mm²) was required. Nevertheless, based on our experience with previous studies, current results demonstrate that the quality of the fabricated MCS is sufficiently good.

The PPL was realized by ion milling in a 400nm-thick silver film. It was rather challenging to mill the slits without deforming the sidewalls due to the extreme high aspect ratio. Nevertheless, by optimizing the FIB milling parameters including beam current and milling rate for slit with different widths, PPLs with parameters close to the designed were fabricated. An SEM image of a fabricated PPL viewed from the cross-section side at a tilted angle of 52° is shown in Fig. 5(a). We can see that the slits were milled through the Ag slab with almost a vertical sidewall and smooth surfaces. The slit width is homogeneous along both the length and into the depth, which is precisely controlled within several nanometers. To further examine the lens function of the PPL, an aerial image scanning microscope was built to measure its focusing property [25]. The sample was mounted on an xyz-piezo stage (Piezosystem Jena, Tritor 102SG), which is configured for transmission measurement. A photodetector (APD120A2/M) with an effective aperture size of 130 nm in diameter was used to detect the field intensity. Through scanning the xz-plane with steps of 5 nm in the x-direction and 39 nm in the z-direction, field intensity distribution behind a fabricated PPL was measured, as shown in Fig. 5(b). A well-defined focus with the designed focal length of 2 µm is manifested. Comparing it with the calculated field intensity shown in Fig. 5(c) using the parameters in design, we see that a very good agreement is achieved, manifesting a precise nano-fabrication of the lens.

 

Fig. 5 (a) SEM cross-sectional image (tilted view at an angle of 52°) of a fabricated PPL structure milled into a 400 nm-thick Ag layer. Slit widths measured under SEM are also labeled. (b) Field distribution along the xz-plane behind a fabricated PPL illuminated by a plane wave and measured by an aerial image scanning microscope. (c) Calculated field distribution of the PPL with designed slit widths of 88, 64, 57.5, 47.5, 42, 34, 42, 47.5, 57.5, 64, and 88 in nanometer with a pitch of 200 nm.

Download Full Size | PDF

4. Fabrication and characterization of the cascaded superlens

Before fabricating the composed superlens, a double-slit object was first milled in a 100 nm-thick chromium layer deposited on a glass substrate. Then SOG with a thickness of D_OM, which works as a spacer between the object and the superlens, was spin-coated on the object. Fabrications of the MCS and PPL slabs were then proceeded according to the design shown in Fig. 1(a). The two plasmonic elements were aligned with respect to the center of the object and this was achieved by using several deeply etched alignment marks with a size of 5 × 2 µm² arranged around the pattern region. The etching of these marks was repeated before the fabrication of the PPL and an alignment error around 100 nm can be obtained. For each object, the MCS was structured with an area of 20 × 20 µm² and it is only required to align with the object symmetrically with respect to any two neighboring grating grooves. To achieve this, objects with the same size were fabricated in an array with a distance of 20 µm plus an additional lateral shift of n × 20 nm, in which n represents the nth fabricated object in the array. With this additional step, the alignment error can be reduced to 20 nm.

Since plasmonic structures are very sensitive to fabrication defects, structure shapes and especially the quality of metal films, we have to ensure that the cascaded superlens behaves as calculated. A relative large object was first fabricated for this aim, since more light can transmit through the lens. As shown in Fig. 6(a) by the cross-sectional image of the fabricated structure, the object has a slit width of 400 nm and a slit distance of 800 nm. Sequential layers constituting the cascaded superlens can also be seen. Due to a larger slit width and a smaller D_OM, which is 30 nm in this case, the whole structure sinks slightly towards the substrate. Nevertheless, the sink of the MCS at the center does not influence the optical property of the lens as we will show by comparing the numerical results with the experiments. Figure 6(b) shows the calculated field distribution in the xz-plane using the structural parameters designated in Fig. 1(a), but with D_OM = 30 nm. Image from the object is clearly shown at z $\ge$ z₁. The calculated field profiles along the white dashed lines will be compared with the measured results later.

 

Fig. 6 (a) SEM cross-sectional image of a fabricated cascaded superlens with a two-slit object, which has a slit width of 400 nm and a slit distance of 800 nm. (b) Field distribution in the xz-plane of the lens shown in (a) illuminated by a plane wave from the substrate side at $\lambda$ = 640 nm. The dashed lines z₁-z₃ designate the positions where the calculated fields will be compared with the measured ones.

Download Full Size | PDF

Field intensities in the xy-plane transmitted through the structure shown in Fig. 6(a) were measured by a CCD camera via a microscope (Zeiss objective with NA = 1.3). The measured fields at three different z-positions are shown in the top row of Fig. 7. Due to a small alignment angle between the object and the MCS, the intensity peaks in the measured image in Fig. 7(a) are broadened at the upper side. Nevertheless, when we compare the field profiles along the red dashed lines with the corresponding calculated ones taken from Fig. 6(b) as shown in the bottom row, overall good agreements at different z-positions are obtained. The results demonstrate that the fabricated superlens works exactly as calculated, although complex plasmonic structures are involved.

 

Fig. 7 The top row shows images measured in the xy-plane at different z-positions using a CCD camera and the bottom row shows the field plots along the red dashed lines designated in the top row. They are compared with the calculated field intensities along the dashed lines in Fig. 6(b) at (a) z₁ = 1.5 µm, (b) z₂ = 1.0 µm and (c) z₃ = 0.5 µm, respectively.

Download Full Size | PDF

To explore the resolving power of the superlens, we further reduced the double-slit object to a size below the diffraction limit, which has a slit width of 100 nm and a slit distance of 180 nm (from slit center to center). An SEM image of the fabricated object milled in a 100nm-thick Cr layer is shown in Fig. 8(a) and a cross-sectional image of the fabricated lens is shown in Fig. 8(b). A nearly perfect alignment between the lens and the object has been achieved (the two slits are symmetrically positioned with respect to two neighboring grooves of the MCS slab). Thereby, a magnified far-field image with a high contrast in the xy-plane is obtained as shown in Fig. 8(c). The image position along the z-axis should be taken around z = 0 – 450 nm when we compare it with the simulated result shown in Fig. 1(b). The distance between the peak intensities (image size) is about 520 nm, indicating a magnification factor of 2.9. Additionally, the imperfection of the fabricated slit is also clearly imaged by the lens. As a control experiment, a microscope image from a double-slit object with a slit width of 100 nm and a slit distance of X_D = 300 nm in the absence of the superlens is shown in Fig. 8(d). The two slits cannot be resolved by our microscope as has been predicted by the numerical simulation.

 

Fig. 8 (a) SEM image of a double-slit object with a slit width of 100 nm and a slit distance of 180 nm. (b) Cross section of the superlens fabricated on the object shown in (a). (c) Image in the xy-plane from the object shown in (a) captured by a CCD camera. (d) CCD camera image from a double-slit object with X_D = 300 nm in the absence of the superlens.

Download Full Size | PDF

Taking the measured field profile shown in Fig. 8(c) along the red dashed line and comparing it with the calculated result as shown in Fig. 9(a), again a good agreement between them is obtained. This validates the resolving power of the superlens predicted by the numerical calculation shown in Fig. 1(a). Further experiments for objects with different slit distances were carried out. The relationship between the object size and the image size is plotted in Fig. 9(b) for both calculated and experimental results. In the observation range, the measured data follow approximately the calculated data. However, the lens shows a slightly sub-linear behavior, which may be induced by a near-field interaction between the MCS and the object. This can be improved by replacing the center of the MCS by a flat plasmonic cavity, which will be reported elsewhere.

 

Fig. 9 (a) Comparison of the measured image field (bar plot) along the red dashed line shown in Fig. 8(c) with the calculated field profile (red curve). (b) Measured and calculated peak distance (image size) as a function of slit distance (object size) from the lens shown in Fig. 8(b).

Download Full Size | PDF

5. Discussion and summary

In this report, we have presented the fabrication and characterization of a novel cascaded plasmonic superlens, which is composed of a metasurface slab of MCS and a PPL slab for near-field imaging with magnification. Nano-fabrication of the superlens with high precision has been accomplished through a step-by-step optimization procedure. This means that before fabricating the whole lens, each part of it was fabricated and characterized separately. When a good agreement between the measured and calculated results has been achieved, the cascaded superlens was then fabricated. Finally, cascaded superlens with imaging properties as predicted by numerical calculations has been obtained due to the well-controlled nano-fabrication procedures. The achieved lateral resolution of the superlens at the wavelength of 640 nm is 180 nm thanks to the image magnification mechanism introduced by the PPL. Furthermore, the optical loss of the lens is acceptable although several layers of metals are present.

Our numerical studies have shown that the PPL along cannot be used as a near-field imaging device. However, when it is combined with the MCS slab, near field interaction between them suppress the strong grating diffraction from the PPL. Nevertheless, the imaging of the cascaded superlens is very sensitive to the relative position between the MCS and the object. This can be attributed to the near-field interactions between the object and the grating in the MCS. To solve this problem, we will modify the center of the MCS structure into a flat FP cavity and the results will be reported separately. Furthermore, the MCS can be replaced by any other plasmonic metasurfaces when they have an effective material dispersion suitable for near-field wave transformation, namely a relative smooth bandlimited near-field transmission in a possibly large spatial frequency range [14, 29, 31].

The advantage by using the PPL for phase compensation is that it has a compact size and enables the fabrication of lens arrays using current nano-fabrication technologies. Our results may open a way for further investigating far field superlens. New types of planar plasmonic lens have been proposed very recently [32–34] and it would be of great interest to combine them with other plasmonic structures for direct subwavelength imaging. The compact design would have many potential applications for innovative portable instruments.