Optical near-field spots generated at a tiny aperture or end of a sharpened tip can be applied to microscopy with a spatial resolution considerably lower than the diffraction limit. Near-field scanning optical microscopy and its variations have been applied to nanometric optical imaging [1–10], chemical analysis of molecules [11,12], nanofabrication [13], and high-density data storage [14]. As a material inspection method based on the differences in the dielectric function at the surface [8,9], it has advantages in terms of signal intensity and compactness as it does not require the use of a spectrometer to obtain the spectra. A probe capable of broadband operation can be used for accurate inspection of dielectric properties of materials because it captures the whole characteristic spectra.

Various optical probes have been developed, and their spatial resolution and efficiency have been improved [2–5,15–17]. Recent research has focused on expanding the available spectral range of a single optical probe [15–17]. Studies have been conducted on generating broadband near-field optical hot spots whose electric field strength is sufficient for a modulation technique, in which harmonic signals improve spatial resolution [9]. The goal is to simultaneously measure the surface topography and optical properties with lateral spatial resolution of a few nanometers, and to achieve better signal-to-noise ratio over a wider spectral range. Furthermore, broadband near-field hot spots also have the possibility of advancing microfabrication and data storage through multiplex recordings. Microstructures have been researched to generate nanometer scale hot spots to assist magnetic recording [18].

In metrology, generation of broadband near-field optical spots with an electric field of sufficient strength for the harmonic detection will advance nanometric optical measurements. Methods for generating a near-field optical spot at the end of a tip are broadly divided into two categories: direct excitation and indirect remote excitation. In the direct method, the excitation laser light is focused on the end of the probe [7–12]. This method is easier in terms of its construction and tuning but has high background noise [9,10]. The indirect method, in which the excitation laser light irradiates at a distant position from the near-field optical spot [3,4,15], has less background noise. The combination of surface plasmon resonance and optical energy transfer via surface plasmons is one of the best ways to enhance efficiency and suppress background noise. We applied excitation of the surface plasmons using the Kretschmann–Raether configuration to a plasmon propagation probe in order to increase the efficiency of near-field generation [19].

We designed an optical probe that simultaneously generates near-field optical spots in the visible and near-infrared spectral range with high efficiency and theoretically and experimentally investigated its operation. We based the probe on an atomic force microscope made of silicon. This silicon probe offers three advantages. It is compatible with the sophisticated probe control techniques that have been developed for atomic force microscopes. We can precisely fabricate the probe with techniques that have been thoroughly studied and applied to silicon engineering. In addition, we can benefit from the chemical stability of silicon, which directly affects the repeatability of measurements. Our plasmonic probe is composed of a thin-film silicon waveguide coated with gold at the end of a silicon cantilever. The waveguide was fabricated in two steps. A thin gold film was deposited on a silicon trigonal pyramid [Fig. 1(a)]. Then the front part of the silicon trigonal pyramid was removed using a focused ion beam [Fig. 1(b)].

 

Fig. 1. (a) Schematic and geometric configuration of the waveguide. Angles $\phi$ and $\theta$ are incident angles on the silicon–gold interface and the surface, respectively. The dashed line indicates the removed part of the silicon. (b) Scanning electron micrograph of the fabricated waveguide. (c) Reflectance of the three-layer system with the 50-nm-thick gold film depending on the incident angle at wavelengths of 660 nm and 850 nm.

Download Full Size | PDF

We analytically designed the geometry of the waveguide by considering its implementation in advance. First, we determined the thickness of the metal film on the rear side to optimize the excitation efficiency of surface plasmons. We calculated the angle dependence reflectance $R(\phi )$ of $p$-polarized light using Fresnel’s equation for a three-layer system composed of semi-infinite silicon, a gold film of thickness $t$, and a semi-infinite vacuum [20]. The calculated reflectances shown in Fig. 1(c) indicate that surface plasmons can be excited resonantly in both wavelengths with a gold film thickness of $t=50\mathrm{nm}$ [21–23]. The incident angles for resonant excitation ${\phi }_{\mathrm{RES}}$ of 50 nm gold film are 15.7° for visible light and 16.3° for infrared light. In this article, we use a classical electromagnetic treatment for waveguide design because the latest theoretical work [23] verified that for film thickness of 50 nm, the resonant excitation angle, found from the minimum of $|R|$, was close to the resonant excitation of surface plasmons and, the same time, to the maximum of the excited electric near field on the outer surface of the gold film. Examinations were performed at wavelengths of 850 nm and 660 nm, which are at both ends of the wavelength range we intended to use.

Next, we calculated the incident angle $\theta$ to the silicon surface. The relationship between the incident angles $\phi$ and $\theta$ was determined using Snell’s law and the refractive index of silicon [21]. In order to make the excitation beam vertically incident in the application situation, we processed the silicon substrate into a waveguide whose thickness gradually changes with an apex angle of $\alpha$ [Fig. 1(a)]. The relationship between the incident angle $\theta$ and apex angle $\alpha$ can be calculated in terms of the resonant angle ${\phi }_{\mathrm{RES}}$ as

 

Fig. 2. (a) Relationship between apex angle $\alpha$ and incident angle $\theta$, which satisfies the resonant excitation condition of the surface plasmons. (b) and (c) Incident angle dependence of reflectivity for the wavelengths of 850 nm and 660 nm, respectively. For comparison, the origins of the angle are set at the minimum of $|R|$. The solid and dashed lines indicate the reflectivity at the silicon surface $R(\theta )$ and silicon–gold interface $R(\phi )$, respectively.

Download Full Size | PDF

We fabricated the waveguide by processing a commercially available silicon cantilever [Arrow NC (Nanoworld)]. Polycrystalline gold film was deposited on a silicon pyramid with a thickness of 50 nm on the rear surface and 150 nm on the front surface [Fig. 1(a)]. The film thickness of the front surface was determined to thoroughly block the residual excitation light. We shaved the silicon tip at the end of the cantilever by using a focused ion beam whose cutting surface was at $\beta =52.3°$ from the back surface of the cantilever. This cutting angle made an apex angle 2°. Because commercially available atomic force microscopes hold cantilevers with a dip angle of $\gamma \sim 10°$ [Fig. 1(a)], the dip angle of the cutting surface ($\beta +\gamma$) was about 62°. This slanted surface allowed the excitation beam to irradiate the surface from above. Therefore, the waveguide probe can be used simply by placing it in a commercial atomic force microscope. The thickness of the silicon waveguide at the thinnest point, which is the base of it, was about 250 nm [Fig. 1(a)]. By thinning the silicon plate, the transmittance of excitation light of wavelengths 850 nm and 660 nm reached 0.95 and 0.78, respectively. Considering Fresnel reflection at the silicon surface, the total usage rates of the excitation beam were 0.90 for 850 nm and 0.77 for 660 nm.

In order to validate the concept and check the performance, we performed two-dimensional simulations to derive the distribution of the electric field using the finite element method [21,22,24]. In this simulation, a $p$-polarized slightly focusing excitation beam with a wavelength of 850 nm was set several micrometers above the waveguide. An example of the simulated electric field is shown in Fig. 3(a). A periodic electric field on the surface of the gold film accompanied the surface plasmons. The reflectivity at the silicon surface was $\sim 0.05$, owing to the oblique incident of the light, which was close to the Brewster angle. This means that our waveguide utilizes almost all of the irradiated beam for exciting the surface plasmons while we suppress the background noise. Figure 3(b) shows magnified images around the end of the tip. This simulation proved that an incident beam at a specified angle facilitates excitation of the surface plasmons using a three-layer system with a semitransparent plate with gradually changing thickness as the substrate. The simulation also showed that the transferred surface plasmons injected optical energy into the end of the tip and generated a hot spot there. This hot spot had a radius of the full width at half-maximum equivalent to the size to the radius of curvature of the probe [25]. Radius dependency of the strength of the electric field [Fig. 3(c) filled diamond] showed that the electric field exponentially strengthens with the reciprocal of the radius of curvature. Strength of the electric field at 10 nm below the end of the isolated tip showed equivalent to that of the incident beam when the radius of curvature was $r\sim 20\mathrm{nm}$. If we sharpened the radius of curvature of the tip to $r=5\mathrm{nm}$, the electric field was enhanced to 1.6 times compared to the incident one. These were similar trends to other techniques [25,26].

 

Fig. 3. (a) Absolute of the electric field $|E|$ around the silicon waveguide simulated by the finite element method. (b) A magnified image at the end of the tip. (c) The tip radius dependency of the electric field. The electric field intensity of the near field was relative value to that of the incident light $|E_0|$. Dashed lines are fitted by the exponential function. (d) A magnified image at the end of the tip with a flat gold surface. The gap width was 2 nm.

Download Full Size | PDF

We also simulated the electric field around the tip in close proximity to a metallic surface [Fig. 3(d)]. The electric field was concentrated and enhanced inside the gap. The strength of the electric field at the midpoint of the gap dependent on the radius of the curvature $r$ is shown in Fig. 3(c) (filled circle). Any radius of curvature enhanced the electric field 2.3–6.9 times stronger than that without the gold surface, owing to the gap mode. When the radius of curvature was $r=5\mathrm{nm}$, the electric field at the midpoint of the 2 nm gap was enhanced to be $\sim 10$ times stronger than that of the incident light [25,26]. In a similar system, the decay length of surface plasmon was measured to be $\sim 10\mathrm{\mu m}$ [27,28]. This implied that the simulated electric field was suppressed to about half by the decay of the propagation from the excitation point. This propagation loss may be overcome using plasmon focusing, which was not woven into our two-dimensional simulation, on a triangle waveguide whose shape and size are similar to our waveguide [29].

We conducted experiments to confirm the concept and simulation results using a commercially available scanning probe microscope to control the probe. This system held cantilevers at a dip angle of 12° with respect to the horizontal sample stage. The cantilever was oscillated at its resonant frequency of $f\sim 285\mathrm{kHz}$ with a vibration amplitude of about 10 nm [9]. The microscope objective with a numerical aperture of 0.26 focused the excitation beam onto the waveguide. The excitation beam was a $p$-polarized continuous wave with an average power of 10 mW. The spot was about 8 μm in diameter, which is 4 times larger than the diffraction limit.

The distribution of the near-field optical spots was measured using scattered light from the end of the waveguide probe. The scattered light was collected by another microscope objective ($\mathrm{NA}=0.40$), which was held at the obliquely upper of the probe with its elevation angle of about 10° and was guided to a photomultiplier tube (Hamamatsu R9110). The electrical signal from the photomultiplier tube was filtered by a lock-in amplifier (Stanford Research Systems SR844), whose detection frequency was locked to the vibration frequency of the cantilever $f$ and its harmonics. The intensity of the scattered light versus the tip–sample distance $d$ was monitored while the tip (waveguide probe) approached the sample surface using a piezoelectric positioner. We used 100-nm-thick polycrystalline gold deposited on the silica substrate for a sample surface, which scattered the optical near field. Figures 4(a) and 4(b) show the intensity of the scattered light as a function of tip–sample distance $d$ with frequencies $f$ and $2f$ using infrared light (850 nm) and visible light (660 nm). Harmonic signals (solid lines) reflect the distribution of the pure near-field optical spots, although the optical signals at $f$ (dashed lines) contained the interference with a bulk-tip-independent scattering from the probe. Exponential increases in the optical signal of $2f$ below 20 nm were clear evidence of the existence of the optical spots around the end of the waveguide probe [9,25,30].

 

Fig. 4. (a) and (b) Intensity of the scattered light against the distance $d$ between the tip and sample surface. (c) Intensity of the scattered light depending on the incident angle $\theta$. Solid and dashed lines show the incident-angle-dependent reflectance when the divergence of the beam is neglected and considered, respectively. Assuming that Gaussian beam diverges with its cone angle 2°, the effect of divergence was incorporated by convolution with the Gaussian function.

Download Full Size | PDF

We also measured the incident angle dependence of the scattered light intensity to examine the excitation principle of surface plasmon resonance. We used a single-mode infrared laser of wavelength 850 nm and average power 10 mW to excite the near-field optical spot. We monitored the intensity of the scattered light of $2f$ while varying the incident angle around the angle of surface plasmon resonance. We focused the excitation beam on the waveguide through the microscope objective, which is fixed above the waveguide with its optical axis vertical to the horizontal surface illustrated in Fig. 1(a). We varied the incident angle $\theta$ by changing the illuminating part in the input pupil of the microscope objective. In experiments, the diameter of the excitation beam was reduced to $\sim 1\mathrm{mm}$, and its focusing angle corresponded to $\sim 2°$. The incident angle dependence of intensity [Fig. 4(c) filled circle] has a positive peak whose angular spread is $\sim 2°$. The numerical simulation using the finite element method reproduced this angle dependency [Fig. 4(c), red diamonds]. The analytical calculation of the reflectance $R(\theta )$ [Fig. 4(c), solid line] owing to the Fresnel reflection at the surface shows a similar resonant angle value. Angles of peaks coincided within the errors, which was the accuracy of the definition of angle in experiments ($\pm 1°$). Analytically calculated reflectance had narrower peak than others. This difference originated the difference of the divergence of the input beam. Analytical calculation used an ideal plane wave for incident light, although in experiment, we focused the excitation beam with its focusing angle of $\sim 2°$. In the numerical simulations, we focused the excitation beam with its focusing angle $\sim 2°$ to describe the experimental condition faithfully. If this divergence was considered in analytical calculation, the calculated reflectance [Fig. 4(c), solid line] became wider and similar to the experimental and simulated results [Fig. 4(c), dashed line]. This provides clear evidence that the optical near-field at the end of the waveguide originated from the plasmon resonance mentioned above.

In conclusion, we developed and investigated a near-field optical waveguide probe that operates in the visible and near-infrared spectral range. A thin-film silicon waveguide with gradually changing thickness based on a commercially available microcantilever is used to resonantly excite surface plasmons and generate a hot spot at the end of the tip. Our design realizes the nearly vertical incident of the excitation beam, which facilitates its use as an optical probe by just mounting it on a standard atomic force microscope. Numerical simulation estimated the strength of the electric field at the midpoint of the tip–sample gap of a wavelength 850 nm as being $\sim 10\mathrm{times}$ stronger than that of the incident one. We experimentally confirmed the generation of the hot spot at the end of the probe, and it had enough strength to detect the harmonic component of the optical near field. Another advantage is that we can fabricate the waveguide with just two steps by a well-established technique. This waveguide has potential applications in mapping of tip-enhanced Raman scattering. The technique of generating the localized hot spot developed in this work also has potential application for heat-assisted magnet recording (HAMR) technology [18] and microfabrication.