Helium focused ion beam fabricated plasmonic antennas with sub-5 nm gaps

Optical antennas have been extensively studied in the last decade due to their unique ability to overcome diffraction limits and manipulate light at a nanometer scale [1]. This property makes them attractive for various applications [2–5], in particular for biosensing [6, 7] and Raman spectroscopy [8–10]. Plasmonic dipole antennas are structures formed by a pair of metal particles where the electromagnetic field is concentrated in the gap between them, which leads to a high field enhancement [11]. The degree of the local field enhancement in the gap is increased exponentially by reducing the gap size and can reach values of several orders of magnitude. It was experimentally confirmed that this property can increase the surface enhanced Raman scattering (SERS) signal by a factor of 10⁸ for gap values of 3 nm, which is enough for single molecule detection [12]. On the other hand, it was experimentally demonstrated with AFM-based measurements that molecules with a low quantum yield can be best detected with a 5 nm molecule–antenna separation [13]. Thus, it is important to have a capability to produce reliable plasmonic antennas with gaps in this range. However, the reliable fabrication of plasmonic antennas with gaps in this dimension range on a substrate is very demanding.

Electron beam lithography (EBL) is perhaps the most popular way to produce nanostructures of various kinds, because the electron beam can be focused down to a nanometer size and can be controllably scanned to define custom patterns. In reality, however, the resolution is reduced at best to ≈2 nm for isolated structures and ≈5 nm for dense structures in the resist [14] and is further reduced to ≈10 nm after the pattern is transferred from resist to gold antennas [15, 16]. All of the reported values above were achieved on thin (10–30 nm) silicon nitride membranes. The disadvantage of membranes is that they are very fragile and have limited size. In the case of bulk substrates the lithography is limited to ≈10 nm due to an effect of back scattered electrons during the EBL step [17]. After the pattern transfer the antennas are reliably fabricated with gap sizes down to 15–20 nm [8]. A method to produce antennas down to 3 nm by a two step lithography process, with defining an oxidized mask between the steps, was recently demonstrated [12]. However, this method is limited to a single gap value per substrate.

Focused ion beam (FIB) milling is another well established technology which has already shown its great convenience for nanostructure fabrication. Typical FIB systems use a beam of gallium (Ga) ions, which allows production of plasmonic antennas with gaps down to 15 nm in single crystalline gold flakes [18]. FIB provides better shape control, however it is a slow process compared to EBL. Both methods can be combined in order to reduce fabrication time, which was demonstrated for the fabrication of plasmonic antennas with a gap of 20 nm [19].

Recently, a helium (He) ion microscope providing sub-nm imaging resolution became commercially available [20]. It has been found that this instrument has high potential for nanostructuring at the 5 nm level by direct milling of the material. However, the sputter rate in case of He-FIB is much lower compared to Ga-FIB, as He ions have a much lower mass. In this paper, we demonstrate a simple and robust way to produce plasmonic antennas with gaps down to 3.5 nm by combining electron beam lithography and helium focused ion beam (He-FIB) milling. While the EBL was used to produce single gold particles, He milling was used to cut them in half and to reproducibly define gaps in a range from 20 nm down to 3.5 nm with a reproducibility of about 1 nm.

2.1. Simulations

Simulations were performed with Comsol multiphysics: a commercial finite element method (FEM) software product. The geometry of plasmonic antennas is shown in figure 1. The structures were simulated in 3D on a glass substrate. The computational domain had a radius of 500 nm with a perfectly matched layer (PML) of 50 nm. In order to reduce the computational domain, the two axes of symmetry of the antenna were used in order to model only one quarter of the structure. A perfect electric conductor (PEC) was used in the symmetry plane perpendicular to the long axis of the antenna, and a perfect magnetic conductor (PMC) was used in plane with the long axis of the antenna and perpendicular to the substrate. The antenna edges were rounded with a radius of 5 nm in order to avoid field singularities. Structures were excited with a monochromatic plane wave from the air side at normal incidence with respect to the substrate. The polarization was set to be linear with an electric field orientation in the direction of the long axis of the antenna. The simulations were performed for a range of wavelengths between 400 and 1100 nm on a dipole having a total length of 130 nm, a width of 45 nm, and a thickness of 20 nm.

Zoom In Zoom Out Reset image size

Wavelength independent refractive indices of 1.5 for the substrate (glass) and 1.0 for the environment (air) were used. The refractive index of gold, used for the antenna, was wavelength dependent and was interpolated on empirical measurements [21]. The titanium adhesion layer was also included in the simulation. However, we assumed that the layer was completely oxidized, as the simulation with a TiO₂ layer instead of titanium leads to a better match with regards to the measurement results. We used the Sellmeier equation,

$$\begin{eqnarray*} &&n=\sqrt{A+\frac{B}{{\lambda }^{2}-C}} \end{eqnarray*}$$

for the refractive index dependence on the wavelength with empirical values A = 5.913, B = 0.2441 and C = 0.0803 μm² for TiO₂ [22]. The mesh size was set to 2 nm for the antenna itself and was increased progressively to 20 nm for the environment and to 50 nm for the substrate (see figure 1).

2.2. Fabrication

The fabrication process of the plasmonic antennas is schematically illustrated in figure 2. We used a standard microscopy glass cover slip (borosilicate D 263 M, Schott) as a substrate. A 2 nm titanium layer was deposited by electron beam physical vapor deposition (EBPVD) in order to avoid charging during the EBL and He-FIB steps. A 50 nm layer of PMMA 950 k (Allresist GmbH) was used as a resist. The EBL step was performed with a Raith 150TWO system using a beam energy of 5 kV and a current of 18 pA (figure 2(a)). Rods were created using single pixel line exposure with a length of 100 nm, a step size of 1 nm and a line dose of 500 pC cm⁻¹. The sample was developed for 5 min in pure isopropanol at room temperature (figure 2(b)). After development a 2 nm titanium adhesion layer and 20 nm gold were deposited by EBPVD (figure 2(c)). Lift-off was performed in acetone using ultrasonic agitation (figure 2(d)). This process resulted in elliptically shaped gold islands of about 50 × 130 nm² in size.

Zoom In Zoom Out Reset image size

The gaps of the plasmonic antennas were produced by direct milling with focused He ions using a Carl Zeiss SMT Orion Plus helium ion microscope (figure 2(e)). We used a beam energy of 30 keV, a beam current of ≈1 pA and a step size of 1 nm for milling. We varied gap sizes by defining different widths of the milled area. Depending on the width of the milled area we adjusted the dose in different ranges by varying the dwell time (40–700 μs) and number of loops (50–100).

2.3. Characterization

The optical response of the antenna was recorded with an Olympus IX-81 microscope connected to an Andor Shamrock 301i spectrometer equipped with an iDUS DV420A camera. A reference spectrum (S_0,BF) of the lamp was taken in bright field (BF). Scattering spectra of the antennas (S_a,DF) and the scattering spectra of the substrate (S_bg,DF) were taken in dark field (DF). The resulting background corrected spectra of the antennas presented throughout this paper were calculated by equation (1), where $\overline{S}$ is the dark signal of the camera,

$$\begin{eqnarray} &&S=\frac{\left ({S}_{\mathrm{a},\mathrm{DF}}-{\overline{S}}_{\mathrm{a},\mathrm{DF}}\right )-\left ({S}_{\mathrm{bg},\mathrm{DF}}-{\overline{S}}_{\mathrm{bg},\mathrm{DF}}\right )}{{S}_{0,\mathrm{BF}}-{\overline{S}}_{0,\mathrm{BF}}}. \end{eqnarray} \tag{ 1 }$$

A halogen lamp (with its internal infrared filter removed) was used as illumination source in the wavelength range from 450 to 1100 nm. One has to note that the lamp is not efficient approaching the limit of 450 nm, while the detector becomes insensitive approaching the limit of 1100 nm. Thus, results are only shown in the range from 500 to 1000 nm.

He-microscope images of fabricated plasmonic antennas are shown in figure 3. The top row shows the nanostructures produced by EBL and lift-off. The rectangles over the structures correspond to the areas to be milled by He-FIB. The bottom row shows the result after cutting the gaps by He-FIB with the gap value measured from the image.

Zoom In Zoom Out Reset image size

The He-microscope images shown in figure 3 are based on secondary electrons (SEs) produced by He ions. The gap was defined as a full width at half maximum (FWHM) of the SE-signal, obtained from the cross-section through the antenna (see figure 4(a)). The width of the SE images taken with He-microscopy is 300 nm with a resolution of 512 pixels. Thus, the pixel size is 0.586 nm giving an accuracy of the measurement for the gap of ≈1.2 nm (one pixel on each side). A higher zoom would result in a larger degradation of the antenna through the imaging process. The minimum gap achieved was 3.5 nm, which corresponds to a milling using a single pixel scan with a current of 0.9 pA, a dwell time of 200 μs and 50 loops (see figure 4(a)).

Zoom In Zoom Out Reset image size

We fabricated antennas of different gap sizes by adjusting the dwell time τ, the beam current I, the nominal (milled) width and the number of loops N. The FWHM of the gap of the antennas as a function of the milled width is shown in figure 4(b). A width of 1 nm corresponds to a single pixel line scan with focused helium ions. Different colors in the plot correspond to different doses, which were calculated according to equation (2), where sz is the step size,

$$\begin{eqnarray} &&\text{dose}=\tau \cdot N\cdot I\cdot {\mathit{sz}}^{-2}. \end{eqnarray} \tag{ 2 }$$

One can see that for gaps above 10 nm the width of the obtained gaps is well matched with the width of the milled area. For the smaller lines the obtained width tends to saturate at about 5 nm. This is caused by the physical limitation of the sputtering process with helium ions at this energy [23]. With a careful adjustment of the dose we were able to get some antennas with smaller gaps. However, the gap values there do strongly depend on the dose. Also, the images do not provide information regarding whether the antennas are cut all the way through the 20 nm thick gold layer due to the high aspect ratio of the produced trench; the optical response of antennas with the lowest gaps indicates that this is likely to be the case. One of the main limiting factors for reproducible gap milling in this range is the stability of the system during the milling time, as even a tiny drift may eventually cause a widening of the milled lines, which explains a slight spreading of the obtained gap. The example scattering spectra measurements of antenna presented in figure 3 are shown in figure 5.

Zoom In Zoom Out Reset image size

In the case of gaps larger than 5 nm, the resonance wavelength value of the measurements matches well with the simulations (see figures 5 and 6(a)). Simulations were performed for a perfect geometry, a homogeneous material and a wavelength step of 10 nm. In reality the fabricated antennas suffer from defects, mostly related to the material inhomogeneities due to the polycrystalline structure of gold. These nanocrystals have different crystal orientation and grain sizes, and form roughness both in lateral and vertical directions. The deviations from perfection add an extra dampening, which broadens the measured resonance. Nevertheless, the resonance position is in good agreement with the measurements.

Zoom In Zoom Out Reset image size

In order to compare the performance of antennas we show the resonance wavelength (figure 6(a)) and a relative scattering intensity at resonance (figure 6(b)) as a function of the measured gap size of the antennas. One can clearly see a red shift of the resonance wavelength and an increase of the scattering intensity of antennas when reducing the gap size. The simulation results show good agreement with measurements. However, the experimentally measured resonances for smaller gap sizes are spectrally more red shifted, compared to simulations. This can be caused by several factors.

First, the ion milling process is always associated with a certain degree of redeposition of the sputtered material. This redeposition loads the antenna with a dielectric material and thus shifts its resonance wavelength toward the red. As the spectral sensitivity of the plasmonic antennas is higher for smaller gap sizes [11], one can expect a stronger spectral shift due to redeposition.

Second, we have to note here that the curves of figures 4(b) and 6(a) contain measurements from two different samples. They were milled using a different length of the cut rectangles—70 nm for sample 1 and 100 nm for sample 2. One can assume that as this causes more substrate milling for sample 2, it would result in a stronger redeposition effect of dielectric material onto the antenna, especially in the vicinity of the gap and into the gap itself. This argument is supported by observing a larger (on average) red shift of the resonance wavelength for antennas of sample 2 (blue circles) compared to sample 1 (red circles).

Third, the milling process tends to produce structures with sloped sidewalls. Antennas with tapered gap sidewalls concentrate light in the region with the closest spacing between antenna arms. This means that the definition of the antenna gap size at FWHM, as shown in figure 4(a), is too conservative. In reality, the antennas possess smaller gap sizes at the bottom of the cut, which defines their optical response. Unfortunately, due to the high aspect ratio of the gap an accurate measurement of the gap geometry is not possible. Further investigations are required to confirm those effects.

To confirm the refractive index sensing capability of He-FIB fabricated antennas, we performed spectral measurements of the samples in water. For this, we put a drop of water onto the substrate and covered it with a 250 μm thick glass cover slip. This allowed us to prevent the water from evaporating and to keep the water layer on the sample flat, which avoids light distortions. The measurements are summarized in figure 7(a). Different colors correspond to the resonance wavelength of the antennas when the sample is in air (red) and water (blue). Circles and triangles correspond to measurements on samples 1 and 2 respectively. The dashed curves are simulation results. We can see a clear red shift for all the antennas within a gap size range of 3.5–21 nm. The shift of the plasmonic resonances, in units of nanometer per refractive index unit (nm/RIU) as a function of antenna gap size when the sample is immersed in water, is plotted in figure 7(b). The wavelength shift is about 100 nm/RIU for gap sizes above 5 nm and increases to values of about 250 nm/RIU for gap sizes smaller than 5 nm.

Zoom In Zoom Out Reset image size

Simulations on the resonance wavelength of the antennas were performed with different refractive index values for the environment (see figure 7(c)). The calculated points (open circles) are fitted by a straight line in order to define the sensitivity of the antennas. The sensitivities of antennas are defined as the slope of the straight lines in the units of nanometers per refractive index unit (nm/RIU) and are marked above the lines, corresponding to antennas with different gaps. As shown, sensitivity increased with the decrease of the gap size.

In conclusion, we have demonstrated a robust process for the fabrication of gold dimer plasmonic antennas by cutting gaps in the range of sizes 3.5–20 nm in a controllable and reproducible way with a precision of ≈1 nm by helium ion milling. Scattering spectra of the antennas reveal an increase of the intensity and a red shift of the peak of the scattered light with a decrease of the gap size. The difference of measured plasmonic resonances compared to simulations for the smallest gap sizes is explained by redeposition of the substrate material and by conservative definition of the gap size at FWHM, as the size at the bottom of the gap has smaller value due to the sloped sidewalls. We have measured an increase of the sensitivity of the antennas, with smaller gap sizes, to values of ≈250 nm/RIU. We believe that the demonstrated technology to produce plasmonic antennas with He-FIB will find applications in biosensing, Raman analysis and optical communications.

We acknowledge the Swiss National Science Foundation for funding (research grant no. 200021-125162 and R'equip grants nos 206021-121306 and 206021-133823). We would like also to acknowledge Prayanka Rajendran (ETH, Zürich) and Dr Takumi Sannomia (Inst. of Technology, Tokyo) for fruitful discussions.