Surface plasmon (SP) refers to electron density waves produced at metal–dielectric interfaces. SP resonance (SPR) has been used extensively due to the capability of detecting molecular interactions label free and providing quantitative data in real time [1]. SPR has also found applications in microscopy [2,3]: while SP can excite fluorescence for extracting microscopic information based on localization [4–6], SP microscopy (SPM) has a great advantage of imaging events in an evanescent field without label interference and was thus employed in many biological studies, e.g., to investigate binding reactions [7], electrochemistry [8,9], crystalline domains [10], lipid vesicle adsorption [11], cell structures [12–15], cell substrate interactions [16–18], and intracellular kinetics of membrane protein [19,20]. SPM was further used to characterize thin films [21], and to quantify index changes [22,23], local distances [24], and cell-electrode gap [25,26]. In addition, SPM was merged with other modalities such as surface enhanced Raman scattering [27], and spectroscopy [28].

However, SPM is either not as prevalent as expected or limited to specific applications, mainly because SPM suffers from poor image resolution. This arises from the large SP propagation length ($L_{\mathrm{sp}}$) in the range of $L_{\mathrm{sp}}=10–100\mathrm{\mu m}$, which is much larger than the diffraction limit and degrades an image [29]. Many studies thus have been conducted to enhance the resolution of SPM [30,31] and remove artifacts [32], e.g., by optimizing light wavelengths [33] or optics [34], employing localized light excitation [35], and controlling SP propagation based on nanoplasmonic structures [36,37]. SPM also suffers from interference induced by multiple scattering [38].

Here, we describe a method based on spatially activated light switching to improve effective image resolution of SPM. The light switching allows an image to be momentum-sampled for super-resolved fluorescent images [39]. Direct scanning of coherent light in a microscope objective was also attempted under total internal reflection using a spatial light modulator or a scan mirror [40]. For the momentum sampling in the proposed spatially switched SPM (ssSPM), we implemented two-channel switching of light illumination with opposite incident directions, while maintaining the angle of incidence and polarization in order not to affect SP propagation. ssSPM was validated by imaging nanowires as reference targets and biological cells in comparison with conventional SPM and bright-field microscopy (BM).

Image reconstruction from the two-channel ssSPM was performed by minimum filtering, i.e., in the images acquired with an opposite incident direction, the intensity of corresponding pixels was compared individually and the smaller was taken to remove the most dominant scattering component. Compared to particle-based deconvolution [41], ssSPM provides a label-free imaging technique with high image quality, whereby an edge perpendicular to SP propagation can be reconstructed in a way that is much improved to the optical diffraction limit. The sharpness of an edge serves as an indirect measure of image resolution, although the relation may not be linear, because the possibility of imaging an edge can be translated to resolving two neighboring edges with clarity. For this reason, we employ a propagation length ($L_{80/20}$), defined as the distance over which the normalized intensity after SP scattering by an edge falls from 80% to 20% as a measure of image resolution.

Periodic nanowires to experimentally evaluate SPM based on two-channel light switching were prepared on BK7 coverslips (No.1, Duran Group, Wertheim/Main, Germany) as substrate. A 2-nm chromium layer and a 50-nm gold layer were deposited using an e-beam and thermal evaporator, followed by a 400-nm thick resist (AR-P 679.04, Allresist, Strausberg, Germany). For ssSPM, we fabricated resist nanowires of widths ($w$) $w=150\mathrm{nm}$ and 550 nm and larger ones of $w=1.1\mathrm{\mu m}$ and 7.8 μm with the center-to-center period ($\mathrm{\Lambda }$) fixed at $\mathrm{\Lambda }=20\mathrm{\mu m}$. The separation ($s$) between neighboring wires, therefore, is $s=\mathrm{\Lambda }-w$. Nanoparticles were also fabricated in e-beam polymer resist to measure point diffraction patterns in SPM and BM. The particles take a cylindrical cross section with 250-nm lateral diameter and 400-nm height.

For cell imaging, human lung carcinoma cell line A549 was obtained from ATCC and cultured in MEM medium containing 10% fetal bovine serum and penicillin-streptomycin ($100\text{units}/\mathrm{ml}$ and $1\mathrm{\mu g}/\mathrm{ml}$, respectively) at 37°C in a humidified 5% ${\mathrm{CO}}_2$ incubator. A549 cells were cultured at $1\times {10}^5$ cells per dish for 24 h. After incubation, adherent cells on the gold film were washed with Dulbecco’s phosphate-buffered saline. For fixation, adherent cells were treated with Karnovsky’s fixative for 6 h and washed for 30 min in 0.1 M phosphate buffer. Cells were post-fixed with 1% osmium tetroxide (${\mathrm{OsO}}_4$) in 0.1 M phosphate buffer for 1.5 h.

The ssSPM setup shown in Fig. 1(a) is based on the existing SPM with a conventional inverted microscope equipped with high numerical aperture (NA) TIRF objective lens ($\mathrm{NA}=1.49$, oil immersion, UAPON 100XOTRIF, Olympus, Tokyo, Japan) as the backbone. A 20 mW He–Ne laser (05-LHP-991, Melles Griot, Carlsbad, California, USA) was used as a light source, which is attenuated by a neutral density filter (FW2AND, Thorlabs, Newton, New Jersey, USA) and $p$ polarized at the sample stage. A spatial filter with a $20\times$ objective, a 200-mm plane-convex lens, and a 10-μm sized pinhole was used. An achromatic lens was employed to focus light in the back-focal plane while the lens position is controlled by a linear motor stage (M-UTM150PP.1, Newport, Irvine, California, USA) in the direction perpendicular to the optical axis. Because both mirror 2 (M2) and the achromatic lens were located on the same stage, translation of the stage causes the focal spot to be shifted in the back-focal plane along the stage. The switching between the two channels, as schematically described in Fig. 1(b), takes place mechanically by the translation of the stage within 1–2 s which may be shortened by electro-optic modulation. A pellicle beam splitter (CM1-BP145B1, Thorlabs, Newton, New Jersey, USA) was inserted in order to reduce interference and ghosting noise. Coherent waves reflected and scattered by the substrate and the targets pass through the beam splitter and tube lens and are finally collected by an s-CMOS camera (Zyla 4.2, Andor Technology, Belfast, UK).

 

Fig. 1. (a) Schematic of the optical setup of ssSPM (FL, focusing lens with $f=200\mathrm{mm}$; M1 and M2, mirrors; NDF, neutral density filter; SF, spatial filter; P, polarizer; and TL, tube lens). (b) Conceptual illustration of switching in the ssSPM. Ch. 1 and 2 represent laser illumination at a nanowire for momentum sampling in two opposite directions at an identical angle of incidence.

Download Full Size | PDF

Figure 2 shows the images acquired by conventional single-channel SPM, ssSPM, and BM with polymer resist nanowires of $w=550\mathrm{nm}$. The nanowires are perpendicular to the direction of light incidence marked by the red arrows (perpendicular orientation). To eliminate the crosstalk between adjacent nanowires, neighboring nanowires were formed to be 12–20 μm apart with separation to be much longer than SP propagation length ($L_{\mathrm{sp}}=8.3\mathrm{\mu m}$ at $\lambda =633\mathrm{nm}$). Figures 2(a) and 2(b) present single-channel SPM images ($I_1$ and $I_2$) obtained by opposite beam illuminations at a resonance angle of BK7/Au/Air configuration (incident direction marked with a red arrow). As expected, very blurred images of an edge are observed on the opposite side of SP incidence, due to the SP scattering by the nanowires and the associated leakage radiation. The trend coincides well with the diffraction pattern obtained by nanoparticles adsorbed to the gold surface [41], in which patterns have tails along the direction of SP propagation, and the tails can be much larger than a nanoparticle. $L_{80/20}^{\mathrm{os}}=6.54\mathrm{\mu m}$ on average (superscript os and ss denote, respectively, the opposite and the same sides measured with respect to SP incidence). If we define an enhancement factor $E_F$ as a ratio of an average propagation length to that of a reference BM image, $E_{F,\mathrm{BM}}=L_{80/20}^{\mathrm{os}}/L_{80/20}^{\mathrm{BM}}=6.54/0.98=6.7$, i.e., for Ch. 1 (Channel 1), SPM shows 6.7 times worse image resolution compared to the BM image. On the same side, we can observe much less blurred edges, which is not the result of dominant SP scattering but, rather, largely due to imperfect fabrication, i.e., because the edges of the nanowire patterns are not perfectly rectangular. By comparing the images of Ch. 1 in Fig. 2(a) with bright field in Fig. 2(d), the associated length $L_{80/20}^{\mathrm{ss}}$ is almost equal to or even slightly smaller than that of bright field, suggesting no substantial degradation of resolution on this side due to SP scattering and propagation. For Ch. 2, the overall trends are mostly identical to the case of Ch. 1. Measured propagation lengths are moderately different, i.e., $L_{80/20}^{\mathrm{os}}=8.11\mathrm{\mu m}$, so that image quality is degraded by $E_{F,\mathrm{BM}}=8.11/0.59=13.7$ times for Ch. 2. Despite the similarities, the observed disparity between the results of Ch. 1 and 2 is ascribed to the fabrication non-uniformity. An AFM profile of the fabricated nanowire pattern attests a profile that may be approximated as a triangular or a sinusoidal shape, a result of the exposure and development process in lithography. Such a profile produces broadening in the propagation length $L_{80/20}$ beyond 155 nm, even under the diffraction limit.

 

Fig. 2. (a) Images of conventional SPM and ssSPM of nanowires (550 nm width) in comparison with BM. Conventional single-channel SPM: (a) Ch. 1 and (b) Ch. 2. The red arrows indicate the direction of light incidence. (c) ssSPM and (d) bright-field image. (e) Normalized intensity along the horizontal ($y$) direction in (a)–(d). The intensity was averaged vertically ($x$) over the dashed rectangle. os and ss represent, respectively, the opposite and the same sides with respect to the light incidence.

Download Full Size | PDF

In contrast to conventional SPM, image deterioration due to SP propagation can be overcome in ssSPM by minimum filtering. The ssSPM image shown in Fig. 2(c) on average reduces the propagation length, thereby improving the image resolution, by an enhancement factor $E_{F,\mathrm{ssSPM}}=L_{80/20}^{\mathrm{os}}/L_{80/20}^{\mathrm{ssSPM}}=7.32/0.67=10.9$ times (7.32 is the average $L_{80/20}^{\mathrm{os}}$ in Ch. 1 and 2 and 0.67 average $L_{80/20}$ for ssSPM). To the first degree, this clearly confirms the working principle of the ssSPM and the feasibility of obtaining much improved image resolution over conventional SPM. Without degradation due to SP propagation, one may be able to achieve image clarity comparable to BM with enhanced depth resolution.

We now investigate the effect of pattern geometry. Figure 3 shows the images when the wire orientation is parallel to the direction of light incidence (parallel orientation) for $w=1.1\mathrm{\mu m}$. Clearly, SP scattering affects the images visibly, which, however, show little difference between channels because of symmetry. Also, the propagation length is almost identical at both edges due to symmetry. Quantitatively, $E_{F,\mathrm{ssSPM}}=1.15$ and 1.05 for Ch. 1 and 2. At first glance, this result may appear to suggest limited effectiveness of ssSPM in the parallel orientation: in fact, $L_{80/20}$ measured in Fig. 3 is on the order of $L_{80/20}^{\mathrm{os}}$ obtained from Fig. 2, i.e., minimum filtering is not required for parallel orientation, and only two-channel ssSPM would be sufficient to avoid the adverse effects of SP propagation for a rectangular object. Note that normalized intensity of the bright-field image in Fig. 3(d) shows artifacts due to the interference through wires of translucent polymer. We also explore the effect of pattern size, which appears to be negligible in the sense that ssSPM worked well regardless of the size. For example, Figs. 3(g)–3(j) show images of Ch. 1 and 2 in comparison with BM and ssSPM for wires of $w=0.15\mathrm{\mu m}$, 0.55 μm, 1.1 μm, and 7.8 μm. Note that interference fringes are more prominent as the pattern becomes larger, due likely to a larger number of scatterers that may contribute to increased interference between SPP waves. Negligible dependence of ssSPM on the pattern size also appears to be the case for the parallel orientation, compared with Fig. 3.

 

Fig. 3. (a) Images of periodic wires of 1.1 μm width in conventional single-channel SPM: (a) Ch. 1 and (b) Ch. 2 and of (c) ssSPM and (d) BM. Scale bar: 10 μm. The wire orientation is parallel to the direction of light incidence. (e) Normalized intensity and (f) histograms of propagation length $L_{80/20}$. The numbers represent the center and the standard deviation of the distribution of $L_{80/20}$. Images of conventional SPM and ssSPM of reference periodic wires ($w=0.15\mathrm{\mu m}$, 0.55 μm, 1.1 μm, and 7.8 μm from top to bottom rows) compared with BM in the perpendicular orientation. Conventional single-channel SPM: (g) Ch. 1 and (h) Ch. 2. (i) ssSPM and (j) BM image. Red arrows indicate the direction of light incidence.

Download Full Size | PDF

Quantitative analysis was performed for the image data in the perpendicular and parallel orientations using histograms of propagation length $L_{80/20}$ in the case of conventional SPM and ssSPM in comparison with BM [shown in Figs. 2(f) and 3(f)]. Conventional SPM is characterized by two separate distributions, one extremely narrow with $L_{80/20}^{\mathrm{ss}}$ smaller than 1 μm, which is similar to that of BM, and the other with $L_{80/20}^{\mathrm{os}}$ that is fairly broad with a large standard deviation due to SPP propagation and can be extended beyond $\sim 15\mathrm{\mu m}$. The variation is deemed largely to be a result of sample fabrication non-uniformity. On the other hand, ssSPM presents a single distribution that is comparable to that of BM in the case of parallel orientation. For quantitative comparison, Table 1 summarizes an enhancement factor ($E_F$), which stresses that an enhancement by more than 10 folds over conventional SPM was achieved in ssSPM and, in fact, the performance of ssSPM approaches that of BM. Also, the average enhancement does not have clear dependence on the pattern size, which corroborates lack of crosstalk scattering between neighboring wires. If we compare ssSPM with SPM and BM, ssSPM outperforms conventional SPM by more than 15 times, although it is above diffraction limit by $29\%=(21.58–15.35)/21.58$. This is due presumably to SP backscattering across an interface with an index difference and less dominantly imperfect alignment of wire patterns with respect to incident light. Note that the improvement by ssSPM can be more significant at wavelengths characteristic of severe light scattering. Similarly, ssSPM can be most effective at a SPR angle, although it should also work at other incidence angles. Yet, we emphasize the possibility of ssSPM implementing the resolution that approaches optical diffraction limit and beyond, to a degree unimaginable in conventional SPM, without compromising optical parameters.

Table 1. Enhancement Factor ($E_F$) of ssSPM ($n$: Number of Sample Measurements)

View Table

ssSPM has been applied to acquiring images in cell microscopy. Figure 4 shows the images of A549 cells acquired by ssSPM. The images from Ch. 1 and 2 produce significant blur that depends on the direction of SP propagation. Image reconstruction by ssSPM can remove a large part of the blur and give rise to enhanced image resolution. The difference of an ssSPM image from that of BM reflects the image acquisition of SPM within penetration lengths. Figure 4 confirms the visible improvement of resolution in SPM in applications, such as imaging nanoparticles and cells that are more complicated than simple nanowire patterns. For example, Fig. 4(α) presents images of a single resist polymer nanoparticle (diam. $\varphi =250\mathrm{nm}$). Point diffraction due to wave scattering by a nanoparticle under conventional SPM made it difficult for general imaging purposes. As shown in Figs. 4(a) and 4(b) in α row, V-shaped diffraction patterns are formed in the direction of light incidence. In contrast, Fig. 4(c) confirms a much improved image of a single nanoparticle by ssSPM with little trace of point diffraction, although background noise appears in all SPM images due to the low index contrast. Improvement on a similar scale can be observed when imaging different sites of cells in Figs. 4(β)–4(ϵ). For the convenience of comparison, cell membrane is magnified in Fig. 4(δ), which presents better defined membrane structure with ssSPM.

 

Fig. 4. Images of single nanoparticle point diffraction and A549 cells: (a) Ch. 1 and (b) Ch. 2 by conventional SPM. Red arrows of (a) and (b) indicate the direction of light incidence with respect to the text. (c) ssSPM and (d) BM. α-row is scattering images of a single resist polymer nanoparticle (diam. $\varphi =250\mathrm{nm}$). Black arrow in (d) shows a single nanoparticle under BM. (β)–(ϵ) represents different sites of the cell culture that was imaged. δ inset: magnified cell membrane.

Download Full Size | PDF