1. Introduction

Localized surface plasmon resonance (LSPR) with a wavelength-specific light extinction and a strong electromagnetic field enhancement near metal nanostructures have been extensively utilized in many biological and chemical sensors [1–6]. Because the effective sensing distances of LSPR sensors, which are determined by the decay length of the enhanced electromagnetic field, are generally about 10 nm from the nanostructure surface [3,7,8], LSPR sensors are sensitive to the binding of small analytes to the nanostructure surface, and they are inherently advantageous in excluding false-positive signals from the bulk solution [3,4]. However, such a short sensing distance makes LSPR sensors unsuitable for detecting analytes of several hundred nanometers in size. Viruses and exosomes (extracellular vesicles containing biomarkers for diagnosing and prognosing diseases) are examples of such mesoscale analytes, and the development of sensitive and simple platforms for mesoscale analytes is of great scientific and social interest [9–13].

Here, we propose a three-dimensional (3D) gold nanobowl (GNB) array as an LSPR sensor to extend the effective sensing distance and to detect mesoscale particles. Although similar nanostructures to the GNBs have been fabricated by coating spherical nanoparticles with a gold shell and subsequently etching the inner nanoparticles [14–18], the non-uniformity and irregular arrangements of the nanostructures limit their use in LSPR sensors whose prerequisite is a reproducible extinction peak from reproducible plasmonic nanostructures. Therefore, we prepared GNBs using nanoimprint lithography (control the overall size of the GNB) and vacuum deposition (control the thickness of the GNB) to obtain uniform GNBs that could be perfectly arranged on a wafer-scale substrate, which resulted in reproducible and reliable LSPR signals. Furthermore, the proposed concave structure of the 3D GNBs made it possible to detect mesoscale analytes, as verified by electromagnetic simulations of extended sensing distances and by experimental measurements of extinction spectra.

2. Experimental

Figure 1(a,b) shows the proposed sensing platform using 3D GNBs to contain and detect mesoscale analytes. The mesoscopic analytes inside the GNB perturb the local plasmon field around the GNB as a result of which the LSPR wavelength of the GNB is shifted as illustrated in Fig. 1(b). The fabrication process of the proposed 3D nanostructure is shown in Fig. 1(c), and similar nanofabrication was reported in our previous studies [19–21]. First, the array of polymer pores was prepared by nanoimprint lithography: after thermally nanoimprinting a resist layer (mr-I 8010R, micro resist technology, Germany) and etching the residual resist with O2 plasma, the sacrificial under-layer (PMGI SF5, MicroChem Corp., USA) was developed by chemical solution (AZ MIF300, AZ Electronic Materials, USA). A gold film was then thermally evaporated, and the polymer layers and gold film deposited outside of the nanoimprinted pores were removed from the sample. During the gold deposition, the sample was loaded at an oblique angle to the gold source and rotated. Therefore, the sidewall and bottom surfaces of the polymer pores could be coated by the gold film, and the 3D GNBs were formed on a glass substrate as shown in Fig. 1(d). The periodicity, outer diameter, and height of the GNBs were 600, 400, and 180 nm, respectively. The overall patterned area determined by the nanoimprint mold was 40 × 40 mm² in which approximately 5 × 10⁹ GNBs with uniform shapes and sizes were formed.

 

Fig. 1. Schematic illustration of (a) the proposed sensing platform using 3D GNBs to contain the mesoscale analyte and (b) the extinction spectra of empty and filled GNBs. (c) The overall process of GNB fabrication: preparation of the polymer template, oblique-angle deposition of the gold film, and removal of the polymer layers. (d) SEM image of the 3D GNBs fabricated on a glass substrate. The inset shows a magnified view of the GNB observed by AFM.

Download Full Size | PDF

3. Results and discussion

To assess the ability of GNBs to sense mesoscale analytes, we calculated the electric field contours around the proposed 3D nanostructure using a 3D finite-difference time-domain simulation software (FDTD Solutions 8.21, Lumerical, Inc.) and compared with a conventional two-dimensional (2D) gold nanodisk array. The hexagonal GNB array was modelled with the dimensions obtained from the scanning electron microscopy (SEM) and atomic force microscopy (AFM) images in Fig. 1(d). The dimensions of the gold nanodisks (periodicity 600 nm, diameter 300 nm, thickness 30 nm) were set so that its resonance wavelength (940 nm) was similar to that of the GNB array (960 nm). For both of the nanobowls and the nanodisks, perfectly matched layer and periodic boundary conditions were used for the vertical and the lateral direction in the FDTD simulations, respectively. As shown in Fig. 2(a), in the case of the 2D nanodisks, the electric fields enhanced by surface plasmon were distributed only at the side edge of the disk. In the 3D case, however, the enhanced fields extended from the sidewall of the GNB to the interior space of the GNB. Therefore, a mesoscopic spherical analyte located inside the 3D GNB can disturb the plasmon field much more effectively than one placed on the 2D disk. The right-side images of Fig. 2(a) demonstrate that the 200-nm-sized spherical particle, which was assumed to have a constant refractive index of 1.6, induced a substantial change in the electric field contours in the case of the 3D GNB but not in the 2D disk. This effective interaction between the 3D analyte and the 3D plasmonic nanostructure increased the LSPR peak shift. The calculated absorbance curves in Fig. 2(b) prove that the spherical nanoparticle caused the resonance wavelength of the 3D GNB to shift by approximately 40 nm; the analogous shift in the 2D disk case was less than 4 nm. As the size of the spherical nanoparticle increased, the enhanced fields inside the GNBs were more substantially disturbed by the loaded nanoparticles, which resulted in a greater shift in the GNB resonance wavelength. The simulation results in Fig. 2(c) show that the shift in the LSPR peak of the GNBs increased from approximately 15 nm to 150 nm when the size of the loaded nanoparticle increased from 150 nm to 300 nm. The LSPR peak shift of the 2D disk could be further increased by placing the analytes in positions where the electric fields were more concentrated, such as the edge of the nanodisk. However, as shown in Fig. 2(c), the proposed 3D GNB was more advantageous in detecting mesoscale analytes than 2D nanodisk in all cases simulated.

 

Fig. 2. (a) Squared amplitudes of the local electric fields near (upper) a 2D gold nanodisk and (lower) 3D GNB (left) without and (right) with 200-nm-sized spherical particle. The wavelength of the incident light was (upper) 940 nm and (lower) 960 nm. Light directions (single-sided arrow) and polarization angles (double-sided arrow) are marked. The Interface interfaces between air, glass and the particle are marked by dotted lines. The scale bars are all 100 nm. (b) Calculated extinction curves of the arrays of (black) a 2D nanodisk and (red) a 3D GNB (solid) without and (dotted) with a 200-nm-sized spherical particle. The spherical nanoparticle was located at the center of the nanodisk and nanobowl. (c) Calculated LSPR peak shifts of the (black) 2D nanodisk and (red) 3D GNB arrays with spherical particles with diameters from 150 nm to 300 nm. Maximum and minimum peak shifts of each data point were obtained from the cases where the beads were located at the edge and the center of the nanostructures, as illustrated in the inset.

Download Full Size | PDF

For experimental demonstration of the proposed GNB array as LSPR sensor for mesoscale analytes, 200-nm-sized polystyrene beads (3200A, ThermoFisher Scientific, USA) were used as model analytes. A 2 μL drop of an aqueous solution containing polystyrene beads at concentrations of 360 fM to 36 nM was drop-casted onto a 3 mm × 3 mm sample of the GNB array, and the solvent was evaporated at 60 °C to minimize the non-uniform distribution of the beads due to the coffee-stain effect [22]. Then the surface of the sample was analyzed by AFM. As shown in the AFM images in Fig. 3(a), the filled GNBs were clearly distinguishable from the empty GNBs, and the population of filled GNBs increased with the bead concentration. Figure 3(a) also shows that some beads were positioned outside the GNBs due to the lack of specific interaction between the GNBs and the beads. Coating the inside of the GNBs with target-specific ligands could improve the selectivity of the sensor and help guide the analytes into the GNBs. Compared to the AFM images, which show visible evidence of the analytes but are limited to local information, the extinction curves in Fig. 3(b) obtained using a UV-Vis spectrophotometer (UV-2600, Shimadzu, Japan) show mean LSPR signals of GNB arrays at the millimeter scale. The series of extinction curves in Fig. 3(b) clearly demonstrate that the LSPR peak shift is dependent on the bead concentration. The peak shift was detected at 360 fM compared to the blank signal and eventually saturated at concentrations above a few nanomolar. Considering the three times the standard deviation (inset of Fig. 3(b)), the detection limit was estimated to be approximately 400 fM (corresponding to 800 zeptomole). Depositing polystyrene beads from a solution with a nanomolar concentration resulted in a thick film of beads forming on the surface of the GNB arrays. At such a high concentration, it is therefore thought that the refractive index around the GNB arrays is no longer changed. When the refractive index of the surrounding medium was set to 1.6, the LSPR peak shift of the GNB array was calculated using the FDTD simulation to be approximately 190 nm (data not shown here), which was consistent with the experimental results in Fig. 3(b). Although not applied in this study, it would be possible to monitor the change in extinction intensity at a particular wavelength (e.g. initial resonant wavelength) as another signal read method. In addition, LSPR measurements of individual nanostructures are expected to be possible with the aid of hyperspectral dark-field microscopy to lower the detection limit.

 

Fig. 3. (a) AFM images of the GNBs containing 200-nm-sized polystyrene beads. The polystyrene beads were dispersed in an aqueous solution and drop-casted onto the GNB array. The polystyrene bead concentrations were 0, 0.36, 3.6, and 7.2 pM, and the corresponding images are shown in a clockwise direction from the upper left AFM image, as indicated by the white arrow. (b) (upper) The extinction curves and (lower) the peak wavelengths of the GNBs treated with polystyrene beads at different concentrations. The dash-dot line indicates the initial peak wavelength (988 nm) of the blank GNBs and the dotted curve is a guide to the eye. The inset of the lower graph shows a magnified view for evaluation of the detection limit with the three times the standard deviation (σ).

Download Full Size | PDF

4. Conclusion

We have proposed using 3D GNB arrays as LSPR sensors to detect mesoscale analytes. The structural advantage obtained from the concave shape of the 3D GNBs makes it possible to extend the LSPR sensing distance and consequently to detect mesoscale analytes. Reliable nanofabrication of uniform GNBs and their perfect arrangement were realized via nanoimprint lithography and vacuum deposition. FDTD simulations confirmed the substantial changes in the plasmon resonance of the 3D GNBs by the loaded nanoparticle with a size of 150 nm to 300 nm. In addition, experimental measurements of the LSPR peak shifts successfully demonstrated the zeptomole-level detection of the mesoscale polystyrene beads. Considering the present theoretical and experimental results, we expect that the proposed GNB arrays will be useful as LSPR sensors for the detection and study of mesoscale analytes.