Pushing the limits of deep-ultraviolet scanning near-field optical microscopy Open Access

In the framework of classical optics, the lateral resolution R and the depth of focus (DOF) are given as¹ 

where k₁ and k₂ are experimental factors, λ is the optical wavelength, and NA is the numerical aperture.

The lateral resolution of lithography has been enhanced by shortening the wavelength of illumination sources from the g-line of a mercury lamp (λ = 436 nm) to the i-line (λ = 365 nm), a krypton-fluoride laser (λ = 248 nm), and an argon-fluoride laser (λ = 193 nm). The state-of-the-art immersion argon-fluoride technique (i-ArF: k₁ = 0.28, λ = 193 nm, and NA = 1.35) has achieved R as small as 40 nm.² A leading manufacturer has stated that R could reach the sub-10 nm by a combination of i-ArF and multiple patterning techniques. Furthermore, extreme-ultraviolet (EUV: λ = 13.5 nm) lithography holds promise as a next-generation sub-7 nm process, where the extremely shallow DOF could be a challenge to overcome.

Likewise, lateral resolution plays an essential role in the field of spectroscopy.^3–5 One spectroscopic trend is to shorten the excitation and detection wavelengths from the visible (VIS) into the deep-ultraviolet (DUV) spectral region, pushing the lateral resolution limit. It should be noted that DUV absorption, scattering, and fluorescence spectroscopies have their own advantages. These include enabling label-free live-cell imaging,^6,7 selective molecular analysis,⁸ and optical characterizations of ultrawide bandgap materials.^9–11 Unfortunately, the lateral resolution of DUV spectroscopy is far from that of DUV lithography (i-ArF). This is because lithography utilizes a single wavelength, whereas spectroscopy involves multiple wavelengths. When the spectral bandwidths are narrow as in DUV monochromatic absorption imaging and Raman scattering spectroscopy, R = 200 nm^8,12 and R = 140 nm¹³ have been achieved, respectively. On the other hand, the lateral resolution of DUV luminescence (fluorescence) spectroscopy is limited at R > 300 nm,^6,7,14 which does not surpass the lateral resolution limit of conventional VIS fluorescence microscopy or microspectroscopy (e.g., R ∼ 200 nm by k₁ = 0.5, λ = 500 nm, and NA = 1.3). This is because luminescence spectroscopy involves broader spectral bandwidths. In the DUV spectral region, it is generally difficult to design an objective lens with a high NA and broadband chromatic-aberration correction.¹⁵ 

One approach to resolve this issue is luminescence spectroscopy using a scanning near-field optical microscope (SNOM). SNOM can overcome the diffraction limit of light.^16,17 Additionally, the DOF problem, which is serious in EUV lithography, does not exist in scanning near-field optical microscopy because the probe-sample distance is controlled (one drawback of SNOM is the relatively low signal-to-noise ratio). Table I summarizes the development history of DUV-SNOM. Aoki et al. visualized DUV fluorescence signals from organic and biological materials with R = 50 nm.¹⁹ Marcinkevičius et al. constructed a DUV-SNOM with an excitation wavelength of 227 nm and observed emissions from ultrawide bandgap materials with R = 100 nm.²³ Previous studies could not investigate luminescence signals below λ = 240 nm due to the constraints of previous DUV-SNOM apparatus.

Presently, there are increasing demands to develop nanospectroscopic tools, which further shorten the excitation and detection wavelengths. For example, solid-state light sources emitting below λ = 240 nm suffer from an extremely low external quantum efficiency compared with that in the VIS spectral region.²⁴ Nanospectroscopic optical characterizations (e.g., luminescence spectroscopy) should elucidate the underlying physics behind such device operations. The realization of solid-state light sources with such short wavelengths will open up emerging application areas such as gas (e.g., NOx) sensing²⁴ and inactivation of bacteria without harming human skin.²⁵ 

Various problems have hampered the development of nanospectroscopic tools operating at shorter wavelengths. These include temporal deterioration of DUV transmission and reflection optics, degradation-induced low pointing stability of an excitation source, and a low throughput of DUV transmission optics. Herein, we overcome these problems and develop a DUV-SNOM with an excitation wavelength of 210 nm.

Unlike a pulsed laser, a continuous-wave (CW) laser realizes a luminescence measurement with a high signal-to-noise ratio even under weak excitation conditions.²⁶ Therefore, we use the fourth harmonic generation (FHG) of a CW Ti:sapphire laser (M SQUARED, SOLSTIS ECD-X-Q) as the excitation source. The wavelength is adjusted to 210 nm. Because the laser consists of three enhancement cavities, the oscillation is quite sensitive to vibrations and shocks. To stabilize the laser oscillation and align optical systems, two optical tables are used: one for the excitation system (first optical table) and one for a detection system (second optical table). The excitation beam is steered by two types of flat mirrors: aluminum-coated mirrors and dielectric-coated mirrors.

One of the most serious problems in DUV spectroscopy is the temporal deterioration of transmission and reflection optics. For example, the optical output of our excitation source gradually decreases with time due to the degradation of a FHG crystal (β-Ba₂B₂O₄). Degradation of reflection and transmission optics induces the power and pointing instabilities of the excitation beam. Separating the optical systems into two optical tables also induces the pointing instability. To resolve these issues, we constructed a power and pointing stabilizing optical system (Fig. 1). This system allows DUV scanning microscopy to be performed over a long period of time.

The principle of the power stabilization is as follows. A half-wave plate and a linear polarizer can adjust the transmission power because the excitation beam is linearly polarized. Additionally, part of the transmission beam is picked up by a wedge plate and focused by a planoconvex lens composed of synthetic fused silica. The optical power is measured by a photodiode and stabilized by a negative-feedback control (TEM-Messtechnik, NoiseEater).

Two aluminum mirrors mounted on piezoelectric (fine tuning) and motorized (coarse tuning) adjusters change the pointing of the excitation beam. The pointing is detected by two position sensitive detectors and regulated by a negative-feedback control (TEM-Messtechnik, Aligna). It should be noted that the pointing stabilizing optical system is installed on the second optical table.

Two aspheric planoconvex lenses (Edmund optics, 48 537 and 33 950) and a 15 μm precision pinhole expand and adjust the excitation beam shape, respectively (Fig. 2).

After passing through a spatial filter, the excitation beam is steered to a reflective objective by a custom dichroic mirror. Figure 3 shows the transmission spectrum of our custom dichroic mirror. Compared to the commercially available ones, our dichroic mirror has a definite threshold around 220 nm.

Figure 4 schematically depicts the objective optical and scanning probe system. We used a reflective objective (Edmund optics, 89 722) with an NA of 0.23 to efficiently couple the excitation beam onto the optical fiber probe (NA = 0.2). The reflective objective does not introduce chromatic aberrations. Compared to a conventional SNOM, the optical fiber probe of our DUV-SNOM has some striking features. The short (3 mm) and double-tapered²⁷ optical fiber probe realizes a high-throughput transmission of DUV light. In addition, the optical fiber probe of our DUV-SNOM is not coated by metallic materials. Previous studies have theoretically^28–30 and experimentally^28,31,32 shown that a subwavelength resolution can be obtained using an uncoated optical fiber probe in the VIS spectral region. Our DUV-SNOM uses a scanning probe system developed by JASCO (NFS-510UV). The distance between the optical fiber probe and the sample is controlled by the shear-force method,³³ where the force is detected via the deflection of near-infrared (NIR) light.

The illumination source is an incandescent lamp (Edmund optics, 59 235). The light is coupled to a liquid light guide (Edmund optics, 53 691) with a core diameter of 5 mm. The beam is subsequently collimated by a VIS achromatic lens with a focal length of 25 mm and focused onto an optical fiber probe by a reflective objective (Figs. 2 and 4). The illumination optical system has a magnification ratio of 0.77. Due to the liquid light guide, the sample image is clearly projected despite the critical illumination optical design.

An ultraviolet-ray shooting lens (Tochigi Nikon, UV-105mmF4.5) is used in an imaging optical system (AL₄ in Fig. 2). The imaging lens has a focal length of 105 mm and a small chromatic aberration from the DUV to the NIR spectral region. The magnification ratio of the imaging optical system is 5.4. The optical fiber probe and sample are imaged by a charge-coupled device camera (Artray, ARTCAM-407UV-WOM).

A long pass filter (custom) is used to cut an excitation beam. Figure 5 shows the transmission spectrum of the long pass filter. Our filter has a steep threshold compared to the commercially available ones. An apochromatic lens with a focal length of 50 mm focuses the collimated luminescence signals onto the entrance slit of the monochromator (Princeton Instruments, SP2500). The chromatic aberration of this refractive lens is collected from the DUV to the NIR spectral region. The magnification ratio of this detection optical system is 2.6. The luminescence signals are subsequently dispersed by a grating with 600 grooves/mm and electrically detected by a LN₂-cooled charge-coupled device (Princeton Instruments, SPEC-10).

The probe of our DUV-SNOM was fabricated by chemical etching of a silica-based optical fiber similar to that of a conventional aperture-type SNOM.³⁴ DUV irradiation tends to create color centers in silica, which gradually degrade the transmission properties of silica-based optical fibers and consequently, optical fiber probes.^35,36 This phenomenon, which is called the “solarization effect,” causes a serious problem for long-term DUV optical measurements.

We prepared a custom optical fiber and measured the DUV transmission properties over a period of several hours. The optical fiber (NA = 0.2) has a length and core diameter of 5 mm and 6 μm, respectively. The power- and pointing-stabilized incident beam has a wavelength of 210 nm and a power of 350 μW. The incident beam is coupled onto the front face of the optical fiber using a reflective objective (NA = 0.23). The output from the end face of the optical fiber was measured by a calibrated photodiode (Ophir, PD300-UV-193). It should be noted that this transmission experiment was performed after an aging process. In the aging process, the front face of the optical fiber was exposed to the above condition for an hour to minimize the effect of the initial deterioration.

Figure 6 shows the DUV (λ = 210 nm) transmission of our optical fiber as a function of time. Although the DUV transmission is not very high, the optical fiber is solarization-resistant within our experimental conditions. We also evaluated the solarization effect on our optical fiber probe by monitoring the luminescence intensity of the sample of interest. Our optical fiber probe is also nearly solarization-free after the aging process (the results are not shown).

In the following experiments, we used an uncoated optical fiber probe with a curvature radius less than 100 nm and a length of 3 mm. Figure 7 shows a scanning electron microscope image of the optical fiber probe. A double-tapered structure is clearly observed. Photoluminescence (PL) spectroscopy was performed using our DUV-SNOM for two quantum well (QW) structures of aluminum gallium nitride (AlGaN), which is an ultrawide bandgap material.

To demonstrate the features of our DUV-SNOM, we measured two samples: Al_0.79Ga_0.21N/AlN multiple QW structure and Al_0.8Ga_0.2N/AlN single QW structure. The former is used to confirm that our DUV-SNOM can detect luminescence signals below λ = 240 nm with an excitation wavelength of 210 nm, while the latter demonstrates the high lateral resolution. The first sample has a period and well/barrier width of 10 and 2.5/15 nm, respectively. The growth condition is detailed elsewhere.^37,38 The second sample has a 1.5 nm well width. The growth conditions are detailed elsewhere.^39,40 The strong emission inhomogeneity in the single QW sample originates from crystal imperfections (screw dislocations). The inhomogeneity was shown by the combination of atomic force microscopy and cathodoluminescence spectroscopy.⁴⁰ The estimated areal density is 4–6 × 10⁸ cm⁻².

Figure 8 shows the DUV-SNOM PL spectrum of the first sample at room temperature. The experimental configuration was the illumination-collection mode. The incident power of an excitation beam onto a reflective objective is 100 μW. The PL spectrum is produced by the subtraction of the far-field spectrum (the probe is not approached) from the near-field one (the probe is approached). The fluorescence of the optical fiber probe generates noise at longer wavelengths. Our DUV-SNOM can detect luminescence signals below λ = 240 nm with an excitation wavelength of 210 nm.

Figures 9(a)–9(c) show spectrally integrated PL intensity mapping images (2 μm × 2 μm) of the second sample at room temperature. Figures 9(a) and 9(b) are the micro-PL results taken using NA = 0.23 and NA = 0.65 (Beck optronic solutions, 5007-190) reflective objective, respectively. Figure 9(c) is the acquired image using our DUV-SNOM under the illumination-collection mode. It should be noted that these images are not from the same region. A previous study has shown that several tens of emissive localization centers exist in 4-μm² area.⁴⁰ Although emissive localization centers are not detected in Figs. 9(a) and 9(b) due to the diffraction-limited images, the fine structure and individual emissive centers are well resolved in Fig. 9(c). The experimental results demonstrate the superior lateral resolution of our DUV-SNOM. It should be also noted that the bright/dark contrast in Fig. 9(c) is not due to the spectral response of our DUV-SNOM system. Figure 9(d) shows SNOM-PL spectra of the second sample at room temperature, where the color (blue, green, and red) of the spectra corresponds to the positions in Fig. 9(c), respectively. The green and red spectra have similar shapes. Meanwhile, the blue one vastly differs as it has another peak at shorter wavelength. A similar result was observed in the previous cathodoluminescence study.⁴⁰ 

We experimentally estimated the lateral resolution of our DUV-SNOM by the line profile of a PL mapping image [Fig. 9(c), yellow broken line]. Figure 10 shows the spectrally integrated PL intensity of the second sample as a function of the X position of an XYZ piezoelectric scanner. In Fig. 10, a local maximum has a relative deviation of 6.1% with the adjacent minimum. On the other hand, the standard deviation (σ) of the excitation power is estimated to be 2.8% from Fig. 6. Therefore, it is unlikely (beyond 2σ error) to consider that the relative deviation in Fig. 10 originates from noise in our experimental setup. As a result, the estimated lateral resolution of our DUV-SNOM is better than 150 nm. It should be noted that a previous study²³ did not experimentally deduce the lateral resolution of their DUV-SNOM but speculated it from the curvature radius of their uncoated optical fiber probe. The curvature radius of our uncoated optical fiber-probe is also less than 100 nm. Therefore, we believe that the lateral resolution of our DUV-SNOM is similar to that of previous studies^22,23 in Table I.

In conclusion, we developed a DUV-SNOM with a 210-nm excitation wavelength. The PL mapping image of an ultrawide bandgap material clearly indicates that emissive localization centers are individually visualized via our DUV-SNOM with a lateral resolution exceeding 150 nm. We believe that these results should push the limits of DUV scanning near-field optical spectroscopy.

DUV-SNOM has many future prospects. In the VIS and the NIR spectral region, SNOM measurements can be performed under extreme environments, including under cryogenic temperatures⁴¹ and magnetic fields.⁴² The multimode (hybrid mode),^27,43 the dual-probe mode,^44,45 and time-resolved⁴⁶ scanning near-field optical microscopy and microspectroscopy have been proposed in the VIS spectral region. Applying these techniques into DUV-SNOM should bring exciting information in the fields of biology, chemistry, and materials science.

The authors would like to thank K. Aizawa and T. Inoue at JASCO Corporation for contributing to the construction of DUV-SNOM. This work was partly supported by JSPS KAKENHI Grant Nos. JP15H05732, JP16H06426, and JP17H04810.