1. Introduction

Optical properties of highly ordered, interacting metal nanoparticles have recently come to the forefront in the merging of nanotechnology and photonics [1–16]. The rudimentary system of two interacting metal nanoparticles is described as two interacting dipoles [1]. As the gap between the two interacting particles is decreased, the coupling strength will increase and give rise to an enhanced electric field in the vicinity of the gap. The coupling effect has been studied between two interacting circles [1, 4, 6, 15], ellipses [2, 11–14], triangles [3, 7–10], and rectangles [5, 16] using far-field, polarization specific light. The coupling effect is strongly dependent upon wavelength and polarization of incident light: for pair separation on the order of the particle size the resonance wavelength red-shifts from the single particle resonance with decreasing gap distance due to coupling that occurs for light polarized parallel to the center-to-center axis [1–4, 6, 12, 15]. Far-field measurements have demonstrated enhanced fields corresponding to proper resonant wavelengths and polarization of incident light [5, 7, 12]. Antennae have been fabricated on NSOM tips [9, 10] and laser diodes [11] with demonstrated uses of photolithography [9], improved near-field imaging [10] and ideas for high density data storage [11]. Nanoantennae also hold applications for increased light extraction from localized emitters [13, 15] such as fluorescent dyes, quantum dots and carbon nanotubes as well as the concept of a true subwavelength sized laser like device [17–19].

Our design for nanoantennae consists of two gold nanoparticles: elliptic right cylinders separated by a small gap along their long axis. Large (100 μm by 100 μm) two dimensional arrays, embedded in a quartz substrate exhibit a strong resonance that is experimentally characterized using far-field spectroscopy and near-field scanning optical microscopy (NSOM) [20] with the support of finite element modeling. These techniques allow a comprehensive study on the local field response of the arrayed nanoantenna system. Illumination mode NSOM excites the sample at different positions and the response is detected in the far-field. This gives a mapping of antenna efficiency to local excitation for high resolution observation of the plasmon mode in the nanoantenna. The resolution is high enough to excite the nanoantenna via only one particle or its part but does not allow mapping of the efficiency distribution in the gap. Such an efficiency distribution to local illumination has been compared with the intensity distribution under the plane wave illumination simulated with finite element method (FEM). It is shown that the nanoantennae respond in a similar manner to both local and plane wave illumination. This is an important result for the use of antennae in light emitting devices or with point-like light emitters. Additionally, this is one of the first demonstrations of patterned antenna arrays embedded inside the substrate. The embedding dielectric layer is important for continued development as future devices will require a host dielectric for structural support and to host light emitters.

2. Experimental technique and results

Electron beam lithography (100 kV, JEOL JBX-9300FS) is used to define the antenna pattern in a resist (ZEP520A) on a quartz substrate. Reactive ion etching is used to create 25 nm deep holes in the quartz. With the sample placed in a vacuum deposition chamber, 25 nm thickness of gold is deposited; a lift off technique leaves gold ellipses embedded in the quartz. The result is pairs of coupled elliptic cylinders with a typical dimension of 155 nm for the long axis, 65 nm for the short axis, a height of 25 nm and a gap of 65 nm. The particle pairs are patterned using a 600 nm period in the X direction and 300 nm in the Y direction.

A field emission scanning electron microscope (FESEM) image of the sample (XY plane) is presented in Fig. 1(a) and representation of the XZ cross section is shown in Fig. 1(b). The shape of the voids defined by the electron beam writing is a consistently smooth elliptic cylinder. However, FESEM imaging shows that the gold does not completely fill the holes in the quartz substrate. Each particle has a unique roughness around its perimeter due to limitations in electron beam evaporation uniformly filling the holes.

Far-field spectroscopy is used to characterize the resonance of the nanoantenna sample. The sample is set in the path of linearly polarized light from a lamp source; transmitted or reflected light is collected and delivered (with a SpectraCode device) to a CCD based spectrometer. Spectroscopy for light polarized across the nanoantenna gap along with the orthogonal polarization is presented in the broken lines of Fig. 1(c). Light polarized across the gap shows a large resonance centered at 810 nm; the orthogonal polarization shows a smaller resonance centered at 550 nm. The large resonance is a plasmon resonance of the two coupled particles, which is enhanced and red-shifted relative to the long axis resonance of the single particle. The weak resonance, at the shorter wavelength, of the orthogonal polarization is characteristic of the short axis plasmon resonance.

 

Fig. 1. The nanoantenna sample. (a) FESEM image; (b) XZ cross section; (c) Experimental (broken) far-field transmission and reflection spectra compared to far-field simulation (circled) data, INSET Absorption for the double and single particle (X polarized light).

Download Full Size | PDF

Illumination mode NSOM provides the ability to locally excite different positions on the nanoantenna geometry and observe the field response. A commercial NSOM [Nanonics] is used with a single mode fiber tip drawn into a conical shape, terminated with a 100 nm diameter aperture. The fiber is cantilevered close to the aperture to allow for tuning fork based tapping mode operation and the sides are coated with chromium and gold. Light exiting the fiber is linearly polarized (measured to have a 14:1 extinction ratio) due to the tapering and bending of the fiber tip.

Given the strong electric field resonance of the sample in the near-infrared region, a wavelength of 785 nm is chosen for near-field investigation of the resonance. With the fiber probe illuminating the sample in the near-field, the resultant signal was collected using a 50x magnification, 0.45 NA objective lens in both reflection and transmission configurations. As the polarization of the light exiting the aperture is fixed, two orthogonal sample orientations are used to obtain images with light polarized in the X direction, parallel with the long particle and center-to-center axis, and for light polarized in the Y direction, parallel to the short particle axis. Four resulting near-field images are shown in Fig. 2. These four images are taken from the same corner of the array sample for proper comparison between the four imaging configurations as well as to compare each NSOM signal for both on and off the array. The corresponding atomic force microscope images were used to determine the position of the array edge; lines are drawn on the NSOM images to illustrate this position.

 

Fig. 2. Illumination mode NSOM images taken with a 785 nm light source of the same corner of the sample array: (a) Reflection image with polarization across the gap (X direction), INSET the orthogonal polarization (Y direction); color bar taken with 1 being the reflection value away from the sample; (b) Transmission image with polarization across the gap, INSET the orthogonal polarization; color bar taken with 100 being the transmission away from the sample. Inset images are shown with a size reduction of 50%.

Download Full Size | PDF

The four NSOM images clearly show several effects occurring on the sample. As expected from the far-field spectroscopy result, NSOM images taken with a wavelength near the resonance wavelength show a strong dependence upon polarization. With the incident light polarized across the gap a very strong contrast is seen in both the reflection and transmission NSOM images while the orthogonal polarization shows non-resonant light scattering for both reflection and transmission images.

For light polarized across the gap, reflection mode [Fig. 2(a)] shows the best visualization of paired particles as the topography allows visualization of individual particles that make up the nanoantenna pair. The highly reflective nature of the nanoantennas is indicative of what is expected at the resonance condition. For individual particles, this signal is confined in the X direction but spreads out in the Y direction between adjacent pairs of particles. The transmission image [Fig 2(b)] for polarization across the gap shows a strong extinction of the near-field signal at the position of the paired particles. In Fig. 2(b), features in the X direction are well defined, while the columns of nanoantenna pairs in the Y direction do not provide a good visualization of individual pairs. The strong vertical lines with high transmission are the spaces between adjacent antenna columns while a close examination inside the high extinction regions reveals several small areas where light is leaking through the paired particles.

It needs to be noted that FESEM imaging of this corner revealed that ∼ 10 percent of the individual elliptic voids were empty. Some of these were just a single particle missing, but in some areas, several adjacent particles were missing. For device operation, such voids would be unwelcome, but for initial characterization, such defects help significantly. In overlaying the FESEM image with the NSOM images, the voids helped to directly match particle positions with the NSOM signals. For resonant reflection NSOM, most voids match up with points inside of the array with a very low NSOM signal. In resonant transmission NSOM, most of the void positions match up with the large areas of higher transmission. Additionally, when the reflection and transmission images are superimposed, the low signal regions of the reflection image match with high signal regions in transmission.

The imaging effects presented are not scanning artifacts due to the NSOM tip. The trends of the presented images are identical when the scanning direction of the fast axis is changed. If the NSOM tip were to perturb the resonant system, it would turn from a dimer into a trimer with significantly different resonant properties. NSOM measurements taken away from the resonance condition (by changing the wavelength and polarization) show moderate scattering similar to Fig. 2 (inset) and are consistent with expectations from the far-field spectral results of Fig. 1. Any influence of the tip on system resonance is a weak perturbation at most.

3. Finite element simulations

Three dimensional finite element modeling of the sample geometry is performed using a commercial software package [COMSOL MULTIPHYSICS]. Due to the symmetry of the sample, a unit cell containing one quarter of the particle pair is set up with a Z range of 3000 nm. A single wavelength plane wave is introduced at the top of the simulation space. With a wavelength spacing of 10 nm, the spectral response is obtained for polarizations parallel to the gap and the orthogonal case. These results match well with the experimental results and are shown using the circled data points in Fig. 1(c) for a direct comparison. Simulation with the same particle geometry, but for the case of an equal X period between each particle (Fig. 1(c) inset) shows that the coupled resonance results in a 20 nm red-shift in wavelength, a broadening of the resonance and an increase in absorption from the single particle case.

The local nature of the resonance for the paired particles is demonstrated in Fig. 3. Figure 3(a) shows an electric field intensity mapping produced from the simulation. This image shows the reflection signal in the near-field of six nanoantenna pairs illuminated with a wavelength of 790 nm; the finite element data is averaged from 25 to 35 nm above the sample plane and a running average is performed in the XY plane to give one pixel for every 10×10 nm² for a more realistic comparison with the volume averaging that occurs with experimental NSOM imaging. This image is to be compared with Fig. 3(b) which shows a reflection NSOM image taken at 785 nm. A FESEM image taken at the exact location of the NSOM measurement, showing both particle and void positions is presented in Fig. 3(c). Figure 3(d) shows an XY vector plot of the simulated electric field. The vector plot, produced at the resonance wavelength, is taken half way between the top and bottom of the ellipses. With the incident electric field polarized across the gap, electric field vectors couple from the left particle to the right one. This allows placement of charges on the gold particles with the electric field coupling from a positively charged region on the left particle to a negatively charged region on the right one. As the incident electric field oscillates, the generated surface plasmons forces the charge regions on each particle to oscillate too.

 

Fig. 3. Finite element visualization of the nanoantenna effect in paired particles compared with reflection NSOM. (a) The electric field intensity above the paired particles taken at 790 nm with light polarized parallel to the gap (X axis); (b) Illumination NSOM in reflection of the paired particles; (c) FESEM image at the exact location of the NSOM measurement; (d) XY electric field vector plot half way between the top and bottom surfaces; taken at 810 nm.

Download Full Size | PDF

Further simulations show that the coupling effect between the two particles increases with decreasing gap; this is becomes significant for gaps below 50 nm. The strong coupling increases the electric field intensity between the particles and continues to shift the system resonance to the red. Modeling also shows that embedding the particles inside the dielectric results in a slight red-shift of the resonance wavelength.

4. Discussion

The far-field spectroscopy result shows good comparison between the real sample and the finite element simulations. For polarization parallel to the center-to-center axis, simulations show a stronger and narrower resonance than experiments. This is explained by considering geometrical differences: the simulation space contains a perfect elliptic cylinder in a perfectly matching hole while there is a distribution of shapes in the gold structures on the real sample. This shape disparity leads to a distribution of resonance wavelengths and a reduction in the collective resonance strength seen from the array. Weak resonance coupling in the sample is verified by comparison with simulations for the case of equal spacing on both sides of the long ellipse axis. The coupled particle case shows an increase in absorption and a 20 nm red shift from the single particle case which indicates plasmon coupling between particles.

Local excitation at a wavelength of 785 nm shows a strong spectral resonance with high reflectance and extinction contoured around the elliptical sample. The resolution, illumination spot area is comparable to the particle area. The experimental and simulated field mappings indicate a mostly single particle resonance (which is expected) as the gap between the paired particles is just beyond the threshold for observing strong enhancement centered in the gap.

The comparison of the efficiency of the local excitation [Fig. 3(b)] to a simulated intensity distribution on the sample surface under the plane wave illumination [Fig. 3(a)] suggests an interesting conclusion. The highest efficiency of the local excitation is observed at the position of the highest local field intensity under plane wave illumination. We note similar features of these two images shown in Fig. 3(a) and Fig. 3(b), a high reflection mapping that is well confined along the direction of the resonant polarization, but spreads out beyond the elliptic boundary in the short axis direction. The same antenna resonances excited by a plane wave can be excited by illuminating part of a particle, one particle or part of the particle pair.

There are differences seen in comparing the experimental field enhancement surrounding each particle to the simulation mapping. Counting for the two missing ellipses in Fig. 3(c), some particle positions show higher enhancement than others. This further demonstrates the resonance dependence on the geometry of the individual particles. Small deviations in the particle geometry can lead to significant differences in field enhancement from the structures.

Comparison of the presented results with previous work can be performed on a qualitative level as published results often vary from group to group. The far-field spectral properties of our embedded structures in experiment and simulation both agree well with the gap resonance trends established in the literature for two interacting metal nanoparticles [1–4, 6, 12, 15]. For observation of individual nanoantennas, confocal microscopy has been used to demonstrate strong resonance dependence upon the sample geometry, which shows general field enhancement in the vicinity of an isolated antenna when resonance conditions are met [5, 7, 12]. Scattering based NSOM has been used to locally probe a singular nanoantenna fabricated on the facet of a diode laser and illustrates a highly confined field in the nanoantenna gap [11]. The presented near-field measurements agree with general wavelength, geometry and polarization requirements demonstrated previously. These results build upon the previous literature by presenting high resolution mappings of the resonant response of the antenna to a local source and showing that this response is consistent with the antenna response to a far-field light source. This demonstration and direct visualization of high reflection and extinction due to the particle plasmon resonance is an important step in full nanoantenna characterization, especially for device applications involving antennae and their response to localized light emitters.

The engineered, predictable nature of the localized field enhancement in response to both local and far-field sources is important for future applications. These include improved sensing and tagging applications involving Raman scattering and other molecular fingerprinting, increasing light emission from dyes, quantum dots and carbon nanotubes as well as the basis for a subwavelength sized nanolaser.

5. Summary

Highly ordered arrays of paired gold ellipses have been fabricated. Experimental far-field spectroscopy and near-field imaging show a strong resonance in the NIR for polarization across the gap. Finite element modeling supports these results by showing similar field effects and provides verification of weak coupling between the two particles though the single particle resonance dominates the field mappings. Peak efficiency of the local excitation is observed at the position of the highest local field intensity under plane wave illumination. High field confinement is observed for each ellipse in the long axis direction while the field spreads beyond the short axis ellipse boundary. Despite the large gap, this work provides concrete evidence of field enhancement surrounding a nanoantenna sample (which will increase with a smaller coupling distance) and shows that the antenna response is similar to both local and far-field illumination.