Plasmonic nanoring fabrication tuned to pitch: Efficient, deterministic, and large scale realization of ultra-small gaps for next generation plasmonic devices

One of the main promises of plasmonic research has always been the possibility to enhance light-matter interactions at the nanoscale for a large frequency range. The enormous impact of plasmonic structures on the dynamics of close-by quantum systems has been demonstrated in groundbreaking experiments¹ and new effects have been predicted.² Especially, the use of plasmonic structures functionalized by spectrally tunable quantum dots offers new applications in photovoltaics,³ nonclassical light generation,⁴ and unprecedented sensing devices.⁵ However, to turn ideas into devices either scientifically or from an application perspective, it is of paramount importance to provide high-throughput reproducible fabrication processes for the plasmonic nanostructures of interest. These processes also have to cope with the extreme precision required by small feature sizes. Often, deviations of just a few nanometers in the geometry spectrally shift a resonance by hundreds of nanometers.⁶ This shift would significantly degrade the ability of these plasmonic nanostructures to couple in the near-field to quantum systems that have a well-defined transition energy. With the purpose to provide a material platform that is as functional and as versatile as possible, the plasmonic properties of the nanostructures have to be deterministically scalable across an extended spectral domain.

A metallic nanostructure that has already proven its ability to serve as a strongly dispersive plasmonic building block with excellent tuning properties is a nanoring.^7,8 Nanoring arrays can be characterized by four geometrical parameters, i.e., height, inner and outer diameter, as well as period. Together with the choice of the plasmonic, substrate, and filling material, nanorings offer a great tuning flexibility and a pinpoint customization capability to a broad range of applications.

In this letter, we report on a scalable and efficient multi-step process to fabricate plasmonic nanoring arrays with geometrical features and gaps below 20 nm. Utilizing character projection,⁹ (CP) electron beam lithography (EBL), and a special double patterning (DP) technique allows the definition of multiple square decimeters in a reasonable timespan of hours (3 min per cm²). Thus, this process is extremely appealing for large scale fabrications. Furthermore, a cost-efficient replication is possible by nanoimprint technology.¹⁰ The process is aiming for patterning of nanostructures, which are characterized by small features, geometric fidelity, and purity of the noble ring material.

The fabrication process is described as follows: Sample fabrication begins on a polished fused silica substrate. To improve the poor adhesion of gold, a 3 nm thin titanium layer is deposited by physical vapor deposition. The deposition of a 50 nm thick gold film follows subsequently in vacuum sequence. The sample is annealed at 200 °C in a furnace for 30 min to allow recrystallization and improvement of its plasmonic properties. Afterwards, it is spin-coated with 10 nm Al₂O₃ by atomic layer deposition (ALD). In the next steps, the substrate is subsequently coated with 120 nm Resist (Clariant AZ1505), covered with 10 nm chromium by ion beam deposition (IBD) and spin-coated with 110 nm chemically amplified negative tone resist (TOK CAN034). The chromium layer is primarily used as hard etching mask for the underlying photoresist AZ 1505 and secondarily as a conduction layer required for the e-beam exposure. The exposure is carried out by a Vistec SB350 OS EBL system. After development of the resist, the nanostructures were subsequently etched into the chromium layer by ion beam etching (IBE) and the AZ 1505 layer by oxygen reactive ion beam etching (O₂-RIBE). During this process, the CAN034 layer was completely removed and the template for the following self-aligned DP process¹² is achieved (Figure 1(a)). The template is homogeneously coated with 30 nm chromium, i.e., angle-deposited at 45° by IBD, while continuously rotating the substrate. The chromium ring is yield by removing the chromium from all horizontal surfaces and the inner resist template with IBE and oxygen plasma (Figure 1(b)). Fabrication is continued by transferring the chromium ring into the Al₂O₃ layer by means of IBE. Afterwards, we remove the chromium ring in a NiCr solution (Figure 1(c)). The selectivity between Al₂O₃ and gold for O₂-RIBE is sufficient to mask 50 nm gold with 10 nm Al₂O₃. After transferring the Al₂O₃ mask into the gold layer by O₂-RIBE, it is wet chemically removed in phosphoric acid (Figure 1(d)).

Within this process, EBL is the essential limiting factor concerning the scalability for large volume fabrication, because it implies rasterization of the written sample surface. We overcome this disadvantage and partially parallelize this process by using CP, i.e., projection of a fixed mask into the resist plane. The mask contains one or more unit cells of the written pattern with a maximum area of 2.5 μm × 2.5 μm. By this way, the raster is increased to 2.5 μm, rasterization becomes independent from the pattern, and the overall writing time can be reduced by three orders of magnitude compared to conventional EBL.⁹ 

Introducing an Al₂O₃ layer has multiple advantages for the whole process. First, it is a diffusion barrier between gold and chromium. Second, the etching ratio to gold and chromium is high (5:1), if oxygen is utilized as reactive etchant during IBE. Therefore, it is an etching barrier, which protects the underlying gold layer from etching into it during formation of the chromium ring (Figure 2(i)). Later, this Al₂O₃ layer becomes an etch mask for the following gold layer. By using the low aspect ratio Al₂O₃ mask instead of the chromium ring minimizes impure re-deposition on the sidewall and broadening of the mask. The overall resolution of the etching process is increased.

The position and inner radius of the nanorings are defined by EBL. Hence, an arbitrary arrangement and inner radius of the nanorings are easily achievable within the same sample. It is for example possible to fabricate isolated nanoring dimers, trimers, or structures with spatially varying geometrical properties side-by-side. The definition of all other geometrical parameters are distributed among ion beam deposition, quartz crystal controlled electron beam evaporation, ALD and (reactive) IBE with in-situ optical emission spectroscopy for end point detection. All these processes are known to be highly reproducible and accurate on the nanometer scale. In total, we expect an overall fabrication accuracy of a few nm for pattern position, thickness, height, and spacing.

To verify this sophisticated fabrication process, we performed several experiments that serve the characterization of the fabricated samples. First, the geometry of the samples has been investigated by different imaging techniques (Figure 2). To determine the height profile, cross-sections were prepared by a Focused Gallium Ion Beam and imaged by scanning electron microscopy (SEM) (Figures 2(e) and 2(f)). The radii and distance between the nanorings were determined by evaluating helium ion micrographs (Figures 2(a)–2(d)). This technique is superior to SEM, especially for imaging non-conducting samples.¹¹ From those measurements, we derived a model of the nanorings obtained in the experiment that allows for an easy parameterization (Figure 2(g)). The gold ring itself is parameterized with the unit cell size Λ, the outer diameter D₁, the inner diameter D₂, and the height h. Besides the rings, the most noticeable features in the cross-sectional views (Figures 2(e) and 2(f)) are a socket at the bottom and round edges at the topside of the rings. Both properties are fully taken into account in our model. The rounding, caused by erosion of the mask edges during IBE is approximated by a radius of curvature is estimated to equal half the ring thickness (D₁–D₂)/4. The depth on the outer (h₁) and the inner (h₂) side addresses the socket, caused by poor selectivity between gold and fused silica. The measured parameters, given in Table I, are emphasizing the flexibility and accuracy of our process. The gap and the inner diameter were tuned independently from 16 nm to 83 nm and 65 nm to 143 nm, respectively. The difference between h₁ and h₂ is most likely attributed to trenching, i.e., a localized higher etch rate near the sidewalls of a nanostructure.

Higher aspect ratio structures can be realized by increasing the thickness of the gold, chromium, and aluminum oxide layers, and the corresponding etching times. However, this might cause geometric shadowing while evaporating chromium under 45°. In that case, another coating technique like chemical vapor deposition or ALD should be chosen, which allows a conformal coating. Based on the gained experience from our experiments, we assume that the features of the demonstrated structures are close to the realizable minimum with utilized tools, materials, and methods. For example, a smaller separation will lead to random interconnections between adjacent rings. A smaller inner diameter will not be properly resolved by the applied electron beam resist.

To evaluate the optical properties, we measured several spectra using unpolarized light in a PerkinElmer Lambda 950 spectrometer and compared the results to numerical simulations (COMSOL Multiphysics®). Since the spectra are strongly depending on the geometry of the rings, their exact placement and purity of the noble metal, an agreement between measurement and simulation demonstrates the suitability of the fabrication process. Notably, an agreement will support a reliable design of future nanodevices exceeding the complexity of a periodic nanoring array. The geometry of the rings has been simulated according to the model illustrated in (Figure 2(h)) and described above. The permittivity of gold was taken from ellipsometric measurements of a bare layer. The comparison between experimental and numerical results is outlined in Figure 3. The plasmonic resonances of the nanoring arrays are in the near infrared region. The field inside the core of the rings is relatively homogeneous and commonly found in plasmonic nanorings.⁷ By varying the period, the spectral resonance position was shifted from 750 nm to 1050 nm. Decreasing the gap between the rings leads to broadening of the resonance and a stronger field enhancement between the rings. Measurement (solid lines) and simulation (dashed lines) are in good agreement with insignificant deviations regarding the spectral resonance position, amplitude, and width. Structural fluctuations and the measurement error of the geometrical dimensions only amount to few nanometers, but might be accounted for those deviations. However, the overall agreement between measurement and simulation implies that impurities, which are usually introduced during etching processes, are marginal. Otherwise, we would expect a considerable red shift and broadening of the measured resonances with respect to the simulation.⁸ 

In summary, we provided a reliable, deterministic, and scalable double patterning process for realizing plasmonic nanorings at a writing speed of dm² per hours. The fabricated nanorings exhibit resonances in the near infrared regime and gaps below 20 nm. The found agreement between measurement and simulation will allow precise design of upcoming devices in future applications.

Because of the large flexibility of the geometrical parameters, the investigated nanorings offer a highly adjustable platform for next generation plasmonic devices at visual, near- and mid-infrared wavelengths. On the one hand, a gap-size down to 15 nm will allow the realization of strong coupling effects like energy splitting or level crossing if the rings are arranged as dimers⁷ or in higher order superstructures.² On the other hand, the enhanced light-matter-interaction inside gaps and the provided homogeneous field distribution inside the nanorings promise new insights and applications when combined with quantum systems to form hybrid plasmonic quantum systems.

The authors gratefully acknowledge support by the German Federal Ministry of Education and Research (PhoNa 03IF2101A, OpMiSen 16SV5577) and by the Thuringian State Government (Thuringian Pro Excellence imitative, MeMa PE116-1). They also wish to thank D. Voigt, N. Sergeev, W. Gräf, M. Banasch, H. Schmidt, H.-J. Fuchs, W. Rockstroh, S. Ratzsch, and T. Käsebier for their assistance during sample fabrication.