1. Introduction

The spectral characteristics of metallic nanostructures due to the localized surface plasmon resonance (LSPR) have been known for decades [1,2]. The corresponding resonance spectrum changes with morphologies, materials, and environments of the nanostructures due to variations of phase-matching conditions between the incident light and induced localized surface plasmon polaritons [3–5]. The LSPR spectrum is widely applied in spectroscopy such as sensing [6,7] and colorimetric devices [8]. Among the plasmonic noble metals, gold is preferable in LSPR-based applications due to its stability and biocompatibility [9,10]. Since the induced collective charge on the metallic nanostructure is aligned with the polarization of incoming field, the spectral characteristics reveal the polarization-dependent energy coupled into the nanostructure [11]. For a symmetric sphere, the resonant condition does not vary with the polarization of incident field [12,13]. However, for anisotropic structures such as nanodimers, nanorods or particle-on-film systems, the field enhancement varies with the polarization directions [14–19]. Therefore, the resonant behaviors of the nanostructure can be tuned by controlling the polarization of excitation field.

Due to the weak scattered light from a single nanostructure, the conventional microscope imaging relies on the dark-field (DF) excitation [14,19]. The main optical path of the excited light is blocked by a DF shutter with oblique incident light focused onto the sample through a condenser lens. The scattered signal from the nanostructure is then collected with an objective lens of high numerical aperture [12,20,21]. However, this requires a different system setup to measure either the transmission or reflection signal. The prism-based total-internal-reflection excitation under oblique incidence enables the direct collection of scattered light, but it is limited to transmission measurements [15,22]. As an alternative, the waveguide (WG) excitation scheme enables oblique incident light to be coupled to nanostructures through the evanescent tail of the WG mode in the substrate [23]. Both transmission and reflection spectra can be measured by simply altering the substrate orientation without modifying the measurement system. Therefore, the system requirement based on WG excitation is less complicated.

Being able to detect the spatially-resolved spectral signal from the nanostructure, an imaging spectroscopic system offers an efficient way to collect both microscopic images and spectral information of a sample [24]. A hyperspectral imaging (HSI) system records both spatial and spectral signals of a detected object into a data cube with different scanning methods such as wavelength scanning, time-division, and so on. [25–27]. Medical imaging using a HSI system has been widely developed [28]. Recently, the HSI detection of nanostructures has been demonstrated with LSPR in, for example, the detection of inhomogeneous nanostructures [29], protein binding on metallic nanoparticles [30], near-field detection on a plasmonic antenna [31], and scatterometry of silicon nanopillar arrays [32]. Although the excitation and detection schemes vary differently in aforementioned applications, simultaneous acquisition of spatially-resolved spectral information in HSI system enables high-throughput detection of various samples.

In this work, a polarization-dependent HSI system using WG excitation is proposed for the detection of spatially-resolved single-particle scattering. The system is built on an inverted microscope with a Michelson interferometer. Gold nanoparticles, nanorods, and particles on film are detected with a polarization-controlled broadband light source. Microscope images of the nanostructures with interference patterns are recorded by a camera into an image sequence. The detected spectrum is obtained by Fourier transforming the interferogram on selected pixels of the recorded image. Both the WG and DF excitations are utilized. The WG excitation facilitates both schemes of the transmission and reflection measurements, while the dark-field excitation is demonstrated only in the transmission mode. The polarization-induced WG coupling enables efficient energy delivery to the nanostructure under a controllable polarization direction. As a platform for nanostructure detection, various single nanostructures can be detected simultaneously under the same microscope view. The real-time and high-throughput detection capabilities of this polarization-controlled HSI system open up a versatile detection platform for nanostructures.

2. Methods

2.1 Measurement system with waveguide and dark-field excitation schemes

Figure 1(a) shows the polarization-controlled HSI system. The detection system is built on an inverted microscope (Olympus, IX73) and a broadband light source is used for nanostructures excitation. For the WG excitation scheme, light is coupled from the end facet of a transparent substrate and is obliquely incident onto the metallic nanostructure through the evanescent tail of WG mode. A cylindrical focusing lens reshapes the light into a line beam. Polarization of the incident light is controlled at the entrance facet of WG using a linear polarizer. For comparison, the conventional DF excitation scheme is also presented. Light is obliquely incident onto the sample from the top through a linear polarizer and a DF condenser. For both excitation schemes, the scattered light from the nanostructure is collected with an objective lens and then sent to a Michelson interferometer. The objective lens (Olympus, LMPLFLN 50X/0.5 BD) is long-working-distance, semi-apochromat, and infinite corrected with a numerical aperture of 0.5 under the immersion of air. Since the system is built on an inverted microscope, there is an equipped tube lens after the objective lens inside the microscope, as indicated in Fig. 1(a). A charge-coupled-device (CCD) camera is on the imaging plane without a tube lens in front of it. However, for other applications that require a full field of view, an additional tube lens in front of the camera may be essential. The optical path difference ($\textrm{OPD}$) of the interferometer is controlled by a function generator that sends out a ramp wave and moves one of the mirrors equipped with a piezoelectric actuator. The voltage signal controls the moving speed and distance of the mirror. The interferogram of the nanostructures is recorded by a CCD camera as an image sequence. By analyzing selected pixels on the recorded sequence of images and Fourier transforming them based on home-built MATLAB scripts, we obtain multiple selected imaging spectra simultaneously.

 

Fig. 1. (a) Measurement setup of the proposed HSI system with both the DF and WG excitation schemes. The measured interference pattern is shown on the camera screen. Sample position of (b) DF and (c) WG excitation schemes in the transmission mode. (d) Sample position of WG excitation scheme in the reflection mode.

Download Full Size | PDF

In this measurement system, the transmission mode can be only operated in the DF excitation scheme. However, both the transmission and reflection modes can be conducted in the WG excitation scheme without modifying the system setup. Figure 1(b) to 1(d) show the positions of samples in various light excitation and light collection schemes. The direction of linear polarizer is also indicated as black dotted double-arrows or blue dots. Figure 1(b) shows the DF excitation in transmission mode. Light is obliquely incident from the top of the substrate with controlled polarization. The scattered light from nanostructures, after passing through an intermediate cover glass, is collected directly below the substrate. The transmission mode using WG excitation is shown in Fig. 1(c). Light is incident onto the end-facet of the substrate with a polarization parallel (blue dot) or perpendicular (black dotted double-arrow) to the sample surface. The nanostructures are excited by the evanescent tail of the WG mode which has a zig-zag trajectory in the substrate. Scattered signals from the nanostructure are collected with an objective lens below the intermediate cover glass. For the reflection mode with WG excitation, we simply reverse the sample upside-down. As shown in Fig. 1(d), light is incident through the end-facet of another transparent substrate covering the sample. In the WG excitation scheme, both the transmission and reflection signals from nanostructures can be measured without altering the system setup, while in the DF excitation scheme, only the transmission mode can be implemented under the same setup. In the following experiments, both the transmission and reflection modes will be demonstrated. The polarization-dependent spectrum as well as field distribution due to coupled resonance will be discussed in both the aspects of experiments and simulation.

2.2 Spectra obtained from images

We obtain the spectra from recorded image sequence as follows. Firstly, a three-dimensional (3D) data cube is recorded. The detected images with interference pattern varying with OPD are recorded by a CCD camera in a time sequence. The recorded 3D data cube contains the image intensity $I({x,y,\textrm{OPD}} )$ as a function of position $({x,y} )$ and $\textrm{OPD}$. The data points of maximal $\textrm{OPD}$ is 2300 in this work. Secondly, the pixels in the regions of interest are selected. The interferogram of each selected pixel can be obtained as $I({{x_i},{y_j},\textrm{OPD}} )$, where ${x_i}$ and ${y_j}$ corresponds to pixel position in the x and y directions, respectively. In this work, an image with 696 × 520 size in pixels is recorded. Since the measured single nanoparticle image is larger than one pixel, only one center pixel at a bright spot is selected on the recorded image to analyze the spectrum. Since the OPD varies with time, the recorded interferogram can be considered as a time t sequence as $I({{x_i},{y_j},vt} )$, where v is the ramp speed of OPD. The fast Fourier transform converts the time sequence signal into the frequency-domain ($\omega $) spectrum $\tilde{{I}}({{x_i},{y_j},\omega } )$. The transformed spectrum is further converted into a function $\tilde{{I}}({{x_i},{y_j},2\pi c/\lambda } )$ of the wavelength $\lambda $ by introducing the speed of light c.

In practice, the spectral axis is first confirmed with narrow band light sources before performing the measurements of nanostructures. The spectrum of selected pixel is finally obtained by performing the fast Fourier transform on the measured interferogram using the home-built MATLAB scripts.

2.3 Sample preparation

Gold nanoparticles of diameter 100 (103 ± 10) nm and gold nanorods of length 45.5 (±6.3) nm with aspect ratio of 2.6 (nanoComposix) are prepared in de-ionized (DI) water with concentration of 52 µg/mL and 31 µg/mL, respectively. The nanoparticle solution is further diluted to 1:100 in DI water for the regional sparse distribution. The diluted concentration is controlled with a micropipette for single structure detection under the microscope view.

For substrate cleaning, quartz plates of size 2.5 × 2.5 × 0.1 cm³ are immersed into piranha solution for 15 minutes and then cleaned by DI water and ethanol. For surface modification, the cleaned substrates are further immersed into a mixed solution with 3-aminopropyl-triethoxysilane (APTES) and ethanol for one hour. The substrates are then cleaned with DI water and dried using nitrogen gas, followed by hot-plate bake at 80°C for one hour. The quartz substrates are then functionalized with amine-terminated silane. The prepared citrate-stabilized gold nanoparticles and nanorods are dispersed on the clean and functionalized quartz substrates using a micropipette as immobilized nanostructures [33,34].

For the particle-on-film resonator, a 60 nm gold film is first deposited on a cleaned quartz substrate using a thermal evaporation system. An aluminum-oxide layer of 3 or 4 nm as the oxide gap is subsequently deposited with an atomic layer deposition system. The 100 nm-size gold nanoparticles are then dispersed on the fabricated substrate as a particle-on-film system.

2.4 Simulation

The simulations are performed with the finite-difference time-domain (FDTD) method (Lumerical FDTD software module). The total-field scattered-field (TFSF) source as a plane wave confined inside a box is used to excite the nanostructure, where the source is extracted outside the source box at the collecting monitor. A 30° excitation angle is set to match the experimental condition of waveguide excitation. For the nanoparticles on quartz substrates, both the cubic monitors for absorption and scattering are set in between the light source and the particle as well as outside around the light source, respectively. The extinction spectrum of the nanoparticle is obtained by collecting both the absorption and the scattering spectra. For the nanoparticles on the gold film, a plane monitor is set above and below the structure to mimic the detection setups in the reflection and transmission modes, respectively. The polarization angle of the light source is tuned from 0° to 90°, depending on the experimental condition. The 3D simulations are performed. All the computation boundaries are set as perfectly matching layers.

3. Results and discussions

3.1 System check for spectral and spatial resolutions

The system is first tested under various maximal $\textrm{OPD}$s to estimate the spectral resolution. A ramp wave of 6 mHz, 5 V, and duty cycle of 100% from a function generator is amplified to 75 V through a piezoelectric controller. The amplified voltage wave is then sent to a piezo actuator (Thorlabs, DRV517) with a travel distance of 30 µm that translates to the $\textrm{OPD}$ of the interferometer. A CCD camera (PCO, pixelfly usb) records the image sequence under the exposure time of 70 ms at the speed of 14 frames per second. In one recorded sequence, 2300 images are taken with a 21 nm step size of $\textrm{OPD}$. The spectral resolution of the system is tested using a He-Ne laser at 633 nm. Figure 2(a) and (b) shows the measured spectrum of a light source at maximal $\textrm{OPD}$ of 5 and 45 µm, respectively. The estimated spectral resolution increases from 231 nm to 10.8 nm with increasing maximal $\textrm{OPD}$, as shown in Fig. 2(c). In this work, the maximal $\textrm{OPD}$ is chosen at 48.3 µm with an estimated spectral resolution less than 10 nm.

 

Fig. 2. Measured spectrum of a He-Ne laser with a maximal OPD of (a) 5 µm and (b) 45 µm. (c) The spectral resolution of the system under various maximal OPDs.

Download Full Size | PDF

The pixel size of the camera is important in determining the spatial resolution of the system. The standard pixel size of the CCD camera in the experiment is 6.45µm. However, we use 2×2 binning in the camera setting to raise the image contrast in the case of low scattered light. The resulting pixel size is doubled to 12.9 µm, with total pixels of 696 × 520. Figure 3(a) shows a standard ruler under a microscope view with a 50× objective lens. The pixel resolution of the projected image is estimated to be 0.3 µm/pixel after examining Group 6-Element 2 on the recorded image of USAF 1951 resolution target shown in Fig. 3(a). The spatial resolution of the system is estimated by analyzing the on-off ratio at the pattern edge from 10% to 90%, as indicated by the red dotted line in Fig. 3(a) and the cross-sectional line intensity in Fig. 3(b). The enlarged region as the blue rectangle indicated in Fig. 3(a) is redrawn in Fig. 3(c). The spatial resolution is then estimated to be 1.2µm. The field of view under the 50× objective lens, as shown in Fig. 3(a), is 209 ×156 µm².

 

Fig. 3. (a) Field of view of a standard ruler under a 50× objective lens. (b) The cross-sectional line intensity as indicated in red dotted line in (a). (c) The zoom-in line intensity of (b) with the region indicated in a blue rectangle in (a).

Download Full Size | PDF

To detect a single nanoparticle, the nanoparticle size on the image plane can be estimated from the diameter of Airy disk by assuming it as a point source. The diameter of Airy disk can be expressed as ${d_{\textrm{Airy}}} \approx 1.22{\lambda _{\textrm{max}}}/\textrm{NA}$, where ${\lambda _{\textrm{max}}}$ is the maximal wavelength of the spectral range, and $\textrm{NA}$ is the numerical aperture of the objective lens [30]. By taking ${\lambda _{\textrm{max}}}$ as 800 nm and $\textrm{NA}$ as 0.5, the estimated image size of a nanoparticle is around 1.95 µm, which covers roughly 6.5 pixels on the camera. Based on the estimation of spatial resolution, pixel resolution, and field of view, single nanoparticle is resolvable with this optical system.

3.2 Light source spectrum

In the experiment, we use a 150 W broadband halogen fiber-coupled light source (Thorlabs OSL2, 400–1600 nm) to excite gold nanostructures. Figure 4(a) shows the detected image of the light source with interference pattern recorded by the CCD camera. By selecting a pixel on the image of Fig. 4(a), the measured interferogram and the corresponding Fourier transformed spectrum of the light source are obtained and shown in Fig. 4(b) and 4(c), respectively. The measured spectrum (Measured) is compared with the one taken with a commercial compact spectrometer (Commercial) (BWTEK BRC642E) using the same microscopic system. The spectrum of light bulb provided by the company (Spec.) is also presented for comparisons. Note that the near-infrared band has been filtered out by the microscope during measurements.

 

Fig. 4. (a) The light-source image with interference pattern recorded by the camera. (b) The interferogram of the light source and (c) the corresponding spectrum (Measured). The measured spectrum is compared with that taken by a commercial compact spectrometer (Commercial) and the one provided by the company (Spec.).

Download Full Size | PDF

As shown in Fig. 4(a), the period of the interference pattern is around 20µm. The interference pattern moves with the piezo-electric actuator, and it can be seen clearly on the whole image. This leads to the variation of intensity with OPD for each pixel. However, when it comes to nanostructures in this study, only the bright spots on the image blinks. It is therefore difficult to observe the interference pattern on the whole image since it appears only at the bright regions. As estimated before, the image of a single nanoparticle on the image plane is roughly 1.95µm. Therefore, the observation of single nanostructures is not influenced by the interferometric signals. We can only observe the interferogram from the selected pixel and analyze it to obtain the spectrum of the nanostructure.

3.3 Polarization-independent transmission spectrum of gold nanoparticles

Gold nanoparticles of diameter 100 nm dispersed on a quartz substrate are detected under the proposed imaging spectroscopic system. Figure 5(a) shows a microscope image using the WG excitation. Several single nanoparticles and some nanoparticle clusters are displayed as the white spots on the image, and they are indicated with yellow or orange spheres for clearer views. The enlarged image of an example nanoparticle is shown in Fig. 5(b). The scanning-electron-microscope (SEM) image in Fig. 5(c) shows how nanoparticles (with 1:10 diluted concentration) are distributed on a quartz plate (the inset shows the actual size of a single nanoparticle). After the interference image sequence is recorded, the corresponding spectra at chosen pixels can be analyzed with the Fourier transform. Figure 5(d) shows the detected transmission spectra of a chosen single gold nanoparticle under the WG excitation in horizontal (Horz), vertical (Vert), and non-controlled polarization (No Pol) directions, where the corresponding spectral peaks lie at 579 nm, 586 nm, and 593 nm, respectively. The measured spectra are all compared to the simulation (Sim) with a resonance peak at 571 nm. The simulated spectra are close to the measured ones in spectral peak and linewidth. Due to the symmetric morphology of a single nanoparticle, there is no major differences among the three spectra.

 

Fig. 5. (a) The microscope image of the 100 nm gold nanoparticles under the WG excitation (scale bar: 10 µm). (b) The enlarged image of a nanoparticle (scale bar: 1 µm). (c) The SEM image of 100 nm gold nanoparticles on a quartz substrate (scale bar: 1µm). Inset: single nanoparticle (scale bar: 100 nm.) (d) The measured and simulated transmission spectra of a gold nanoparticle under various excitation conditions of polarization.

Download Full Size | PDF

3.4 Polarization-dependent reflection spectrum of particle-on-film system

The resonance coupling of a particle-on-film system is detected by the polarization-controlled HSI system using the WG excitation. Gold nanoparticles of size 100 nm are dispersed on a 60 nm gold film with an intermediate aluminum oxide layer of 3 and 4 nm in thickness. The excitation scheme is depicted in Fig. 1(d) for the reflection measurement. Figure 6(a) shows the measured (Meas) and simulated (Sim) reflection spectra. For the sample with 3 nm oxide gap, two spectral peaks lie at 592 nm and 701 nm in the measurement and 592 nm and 694 nm in the simulation. When the thickness of oxide gap increases to 4 nm, two spectral peaks blue shift to 571 nm and 683 nm in the measurement and 579 nm and 685 nm in the simulation.

 

Fig. 6. (a) The measured (Meas) and simulated (Sim) spectra of the gold nanoparticle on a gold film with intermediate oxide layers of 3 and 4 nm in thickness. (b) The measured spectra of the sample with 4 nm oxide gap under controlled polarizations (No Pol, Horz, Vert) and their comparisons with simulations (Sim). (c) The simulation model. Simulated field distribution of the sample with 4 nm oxide gap at spectral peaks of (d) 579 and (e) 685 nm, respectively.

Download Full Size | PDF

We further inspect the sample with 4 nm oxide gap under the WG excitation scheme with controlled polarization. Figure 6(b) shows the measured (No Pol) and simulated (Sim) reflection spectra in the absence of controlled polarization. The measured spectrum with two spectral peaks located at 571 and 683 nm are close to the simulated ones at 579 and 685 nm. The two spectral peaks can be separated under the excitations with vertical (Vert) and horizontal (Horz) polarizations, as also shown in Fig. 6(b). The resonance wavelengths under the horizontally-polarized and vertically-polarized excitations are 574 nm and 684 nm, respectively, which are close to the two measured spectral peaks at 571 and 683 nm when there is no controlled polarization.

The resonant behaviors at both resonance wavelengths are examined through the simulated field distributions. Figure 6(c) shows a schematic view of the simulation model. The magenta arrow indicates the direction of inclined incident light, and the blue double-arrow indicates the polarization direction. The reflection spectrum is evaluated at the position of green line above the particle while the green box monitors how the field is distributed in the spectral range of interest. When the horizontal excitation polarization is adopted, only the LSPR of the particle can be excited. The simulated electric field distribution of LSPR at the particle edge (at 579 nm) is shown in Fig. 6(d), which is close to the detected spectral peak at 574 nm in the measurement of horizontally-polarized excitation. For the vertically-polarized excitation, the measured spectral peak at 684 nm is close to the simulated one at 685 nm, where only the gap LSPR mode is excited. The simulated electric field distribution of LSPR at the gap (at 685 nm) is shown in Fig. 6(e). We have therefore demonstrated that the WG excitation scheme can effectively control the resonant condition and the spectral characteristics of metallic nanostructures. Both the reflection and transmission spectra are shown under the same system setup using the WG excitation method.

3.5 Polarization dependent transmission spectrum of gold nanorod

We demonstrate the scheme of polarization-controlled excitation with a gold nanorod which has an anisotropic structure. The gold nanorods are dispersed on a clean and functionalized quartz substrate. The WG and DF excitation schemes are compared, whose setups are shown in Fig. 7(a) and 7(b), respectively. Figure 7(c) shows a schematic view of the simulation layout. The magenta arrow indicates the incident direction of light excitation while the blue double arrow indicates the polarization direction. The outer and inner green box monitors the scattered and absorbed lights, respectively. Both the scattered and absorbed signals contribute to the extinction spectrum of nanorods. Due to the anisotropic structure of the nanorod, the excited lights with different polarizations can induce the corresponding resonant dipoles aligned with various rod orientations. The polarizers in both experiments and simulation are rotated from 0° to 180° to observe the spectral change.

 

Fig. 7. The (a) WG and (b) DF excitation schemes of a gold nanorod. (c) The schematic view of simulation layout.

Download Full Size | PDF

By selecting specific pixels on the recorded image sequence, we obtain the image spectrum of a gold nanorod. Figure 8(a) and 8(b) shows the measured transmission spectra compared with the simulation (Sim) when the polarization is aligned with and 60° off the long axis of gold nanorod, respectively. Both the WG and DF excitation spectra are measured, and they are compared to the simulation (Sim). The insets in Fig. 8(a) and 8(b) show the microscope images of a gold nanorod under the WG excitation. Slight differences on color and brightness can be observed between two images resulted from the two excitations with distinct polarization angles. As shown by the spectra in Fig. 8(a), resonance peaks are located at 665, 641, and 658 nm, for the cases of Sim, WG, and DF, respectively, which are close to the resonance wavelength of 660 nm provided by the manufacture of gold nanorods. From Fig. 8(b), when the polarization is 60° off the long axis, both resonance peaks corresponding to the short and long axes of the nanorod can be seen. The simulated LSPR peak associated with the short axis lies at 516 nm, while those for the WG and DF excitations are at 549 and 532 nm, respectively. Since the dispersed nanorods may not lie flatly on the substrate, slight deviations of the LSPR wavelengths occur. The clean resonance corresponding to the short axis in simulations is not observed in measurements. Figure 8(c) shows the intensity variations at the wavelengths of the long-axis resonance obtained with different methods. The simulation results (Sim) are in accordance with theoretical predication (Theory), where the intensity $I \propto \textrm{co}{\textrm{s}^2}\theta $ is a function of angle $\theta $ off the long axis. Note that in experiment, we only rotate the polarizer from 0 to 150°, and the data from 180° to 330° is a replica. The intensity variations in both experiments (DF and WG) are normalized for comparisons with the simulation. The spectral intensity is maximal at polarization angles of 0° and 180°, indicating a dipole induced along the long axis. The minimal spectral intensity occurs at 90° and 270° when the induced dipole is aligned to the short axis of the nanorod. The polarization-dependent excitation and control over the resonance spectrum of gold nanorods is thus demonstrated.

 

Fig. 8. Simulated (Sim) and measured transmission spectra of a gold nanorod using the WG and DF excitation when the polarization is (a) aligned with and (b) 60° off the long axis of nanorod. (Inset: microscope images of the gold nanorod under the WG excitation.) (c) The intensity of the spectrum at the wavelength of long-axis LSPR under different polarization angles of the excitation. The measured data with the WG and DF excitations are compared with the simulation (Sim) and theory (Theory).

Download Full Size | PDF

4. Conclusions

In this work, an imaging spectroscopic system using the polarization-controlled WG excitation scheme is proposed. The image spectrum of a single nanostructure can be obtained by Fourier transforming the signal at selected pixels of the recorded image sequence under the microscope view. Transmission or reflection spectra of gold nanostructures such as nanoparticles, particles on film, and nanorods are retrieved without modifying the system setup through the WG excitation scheme. The transmission spectra under the DF excitation scheme are also taken for comparisons. The polarization-dependent spectral characteristics of the nanostructures are verified with simulations. The field distribution around nanostructures at spectral peaks of the LSPR are discussed. The proposed HSI system with the polarization-controlled WG excitation demonstrates versatile detection capabilities of metallic nanostructures.