1. Introduction

In the past years, different types of chemical sensors and biosensors based on ordered arrays of nanopillars (NPs) have been successfully developed and reported in the literature [1–8]. In comparison with a flat sensing surface, NPs arrays considerably increase the available surface for the attachment of chemical or biological elements, which can lead to a higher performance of the sensor. From the optics point of view, and depending on the pillars architecture, arrays of nanoscale pillars benefit from different interactions among light and the pillars, as field confinement or a variety of resonant effects like surface plasmon resonance [9,10], localized surface plasmon resonance [11,12], metal assisted guided mode resonance [13] and surface-enhanced Raman scattering [14–16]. These NPs can be produced using a variety of fabrication methods and materials and can be optically interrogated by using common measurement techniques.

In previous works the development of periodic arrays of dielectric resonant NPs (R-NPs) made with two Bragg reflectors consisting of a stack of Si₃N₄/SiO₂ alternate layers, 5 pairs each, with a central region of SiO₂, was reported [17,18]. Fabrication consisted of an initial deposition of an anti-reflective coating and a photoresist, followed by the deposition of films of SiO₂ and chromium (Cr) using an e-beam evaporator, then a pattern was written by means of laser interference lithography (LIL), the patterns were then transferred using reactive ion etching, and finally the remaining Cr caps were removed with a proper Cr etchant. Normal incidence visible light on the backside of the transparent substrate generates a sharp, well defined resonant mode in reflectance that shifts as a function of refractive index (RI) variations of the environment in contact with the pillars. This effect is certainly used for sensing purposes as label-free biosensing [19]. Other advantages offered by these optical biosensors are the high chemical stability in aqueous media, high optical transparency and the possibility of surface functionalization.

In this work, we go a step forward in the resonant Si₃N₄/SiO₂ NPs fabrication leading to a novel tiered nanostructure of SiO₂ disks linked by Si₃N₄ necks, obtained after a controlled, selective inductively coupled plasma reactive ion etching process (ICP-RIE). The tiered resonant nanopillars (TR-NPs) surface-to-volume ratio is much higher than the already reported conventional R-NPs, and the fabrication complexity is reduced because the metal mask disks are conserved as nanomirrors. The result is a remarkable bulk sensitivity increment offered by these array-based optical sensors suitable for biochemical sensing. To our knowledge, this is the first demonstration of TR-NPs based optical sensors made from a Si₃N₄/SiO₂ multilayer resonant stack, albeit some other nanostructures showing a certain resemblance have been already reported [20], however they were made with different materials, wet-etching instead dry-etching, and aiming to different purposes. Furthermore, the fabrication method shown in this work may be useful for bio-inspired nanoarchitectures replication, in particular structures resembling the nanostructures found on the iridescent Morpho butterfly wings, with interesting photonic properties and possible applications in biosensing as well [21–25].

2. Experimental

The multilayer stack, deposited onto a 0.5 mm thick quartz substrate, comprises two Bragg reflectors with five alternate layers of SiO₂ (thickness ~110 nm) and Si₃N₄ (thickness ~95 nm) each. A ~210 nm thick layer of SiO₂ is located between the Bragg reflectors playing the role of an optical resonant cavity, there are Si₃N₄ layers in contact with the central SiO₂ cavity, and SiO₂ layers are placed both on the top and the bottom of the stack. Total multilayer thickness is around 2200 nm, and the layers were grown by mid frequency magnetron sputtering with additional plasma oxidation for both materials by the Fraunhofer Research Organization [26–28]. RI values in the 580-630 nm wavelength range vary as following: 1.4646-1.4626 for SiO₂ and 2.0197-2.0121 for Si₃N₄, respectively.

A 300 × 300 µm², high aspect ratio TR-NPs array was fabricated by combining electron-beam lithography (EBL) in a Crestec CABL-9000 C high-resolution EBL system with ICP-RIE, using an Oxford Instruments NGP80 ICP65. The structure is a squared periodic array, pitch 500 nm, with ~250 nm in diameter TR-NPs showing a characteristic tiered morphology. The detailed fabrication process is as following and is shown in Fig. 1.

 

Fig. 1 Schematic diagram of the steps involved in the TR-NPs fabrication. The process involves EBL to make arrays of nanoholes, metal deposition, lift-off step, and CHF₃/O₂ plasma etch.

Download Full Size | PDF

First, ZEP-520 (Zeon Chemicals) positive-tone EBL resist was spin-coated on the Si₃N₄/SiO₂ stack at 4000 rpm and immediately baked for 2 min at 160 °C. Once the sample reached room temperature, Espacer 300Z (Showa Denko) was spin-coated at 5000 rpm to obtain a conductive thin layer. Next, 500-nm-period square lattice of 5 nm in diameter circular solid dots were written on the resist thin film by EBL at 50 kV and 50 pA, following the same procedure used in a previously published work [13] involving ultra-small nanocircle patterns layout and out-of-focus beam exposition, this time applying a high dose of 960 mC cm⁻². After EBL process, the sample was thoroughly washed under deionized water flow to remove the conductive layer, then dried with nitrogen gun, then developed using ZED N50 (Zeon Chemicals) at 20 °C during 60 s under gently agitation and, finally, dried using nitrogen flow.

The resulting array of nanoholes in ZEP-520 resist was coated with a 100-nm-thick nickel film by electron-beam thermal evaporation at a deposition rate of 1 Å s⁻¹. Next, a lift-off step using 1-methyl-2-pyrrolidone (Emplura, Merck KGaA) at 80 °C was carried out, removing the EBL resist and leaving an array of nickel nanodisks onto the Si₃N₄/SiO₂ stack as a mask for dry etching.

The ICP-RIE process used to create the TR-NPs was achieved using CHF₃ (50 sccm) and O₂ (3 sccm) gases, RF and ICP power of 150 W each and pressure chamber of 30 mTorr, until a laser interferometer end point detector (Global Laser Technology Solutions, Intellevation LEP 400) indicated a full etch of the Si₃N₄/SiO₂ multilayer. Etching time was around 1000 s, and nickel caps remained unaltered after this ICP-RIE step.

A FEI Inspect F50 scanning electron microscope (SEM) set at 2 kV was employed for morphological characterization. Optical characterization was performed by combining a Stellarnet Inc. visible light source and a Mightex HRS-VIS-005 CCD spectrometer, using an optic fiber setup to interrogate the TR-NPs array through the transparent quartz substrate and also to collect the reflected light in the 370-830 nm range. When the signal intensity is maximized, the optical fibers are almost in contact with the substrate backside and the light spot diameter is 600 µm in diameter approximately when it hits the NPs, meaning that all the NPs of the array contribute to the optical response. Along the interrogation optical fiber, there was another fiber for reflected light collection. In all cases, one spectrum was acquired every second during 100 s, applying a Savitzky-Golay smoothing process (k = 2, f = 201). Figure 2 shows a diagram of the optical characterization setup.

 

Fig. 2 Sketch of the optical measuring setup for liquids. The sample with NPs is placed downside on a cell filled with liquid. Optical interrogation is performed through the transparent quartz substrate using an optical fiber, and reflected light is collected with another optical fiber adjacent to the first one.

Download Full Size | PDF

With the purpose of conducting a bulk sensing experiment, the TR-NPs array was interrogated in air and using the following solvents in contact only with the sample side containing the array: deionized water (DIW, RI = 1.332), ethanol (EtOH, RI = 1.3613), hexane (RI = 1.3754), isopropyl alcohol (IPA, RI = 1.3825), heptane (RI = 1.3885), dimethyl sulfoxide (DMSO, RI = 1.4765) and toluene (RI = 1.4958). After measuring with every solvent, the sample was washed with IPA, gently dried with nitrogen flow and heated at 85 °C for 5 minutes on a hot plate to remove the IPA residues. The resulting spectra show reflectance, with some resonant minimum peaks which shift to the red when the RI of the environment in contact with the TR-NPs increases.

3. Results and discussion

Regarding the fabrication process and the techniques employed, EBL step took 64 minutes for the 300 × 300 µm² array exposition. However, 1 minute is feasible by optimizing a single-shot EBL exposition strategy to speed up this process. The lithography time needed can also be drastically reduced if other techniques like LIL [18] or nanoimprint lithography [29] are used instead EBL.

Concerning etching time, conventional R-NPs etching required around 130 minutes [18] due to a 16.1 nm min⁻¹ etch rate for the Si₃N₄/SiO₂ multilayer. In this work, the dry etch step needed around 15 minutes (etch rate equals 132 nm min⁻¹), and we observed that this time lapse can be reduced to 5 minutes or less playing adequately with the ICP-RIE parameters. The CHF₃ + O₂ gas mixture and an appropriate combination of pressure chamber, flow and power generated a plasma capable, on the one hand, of a relatively fast vertical etch of Si₃N₄/SiO₂ stack and, on the other, of a selective, slow Si₃N₄ horizontal etch at an appropriate rate (6 nm min⁻¹ approximately) which led to the tiered nanostructure in only one etching step. A probable explanation of this phenomenon is the abundant ions collisions in the ICP-RIE chamber under the conditions employed for the plasma generation, conducting a 3D etching process that includes the expected vertical etch plus the capability of a horizontal, much slower etch. The resulting tiered structure emerges from a very high Si₃N₄:SiO₂ etching ratio in the horizontal plane (parallel to the substrate) which allows for the conservation of the oxide layers while the nitride layers are etched.

Figure 3 contains some examples of high aspect ratio TR-NPs made of Si₃N₄/SiO₂ pillars after selective Si₃N₄ etching, and a top view showing the high uniformity of the fabricated arrays. Figure 1(a) shows a highly etched isolated nanopillar of 600 nm in diameter approximately; in fact it is hard to see the Si₃N₄ connecting necks between SiO₂ disks in the upper part. Figure 1(b) shows two parallel rows of TR-NPs from the first fabrication experiments: pillars diameter is 350 nm and the pitch is 875 nm. Figure 1(c) shows part of the TR-NPs 500 nm pitch array used in the bulk sensing experiment. All the TR-NPs are close to 2200 nm in height and have a 100 nm nickel cap. It is clearly observed that lower TR-NPs region is usually less etched than upper part because the exposition time to the plasma is shorter. Figure 1(d) shows a uniform large area of the same array shown in Fig. 1(c).

 

Fig. 3 SEM images of different TR-NPs. (a) Isolated, tiered nanopillar of 600 nm in diameter. (b) TR-NPs are 350 nm wide, pitch 875 nm. (c) TR-NPs are 250 nm wide, pitch 500 nm. (d) Top view of a large uniform area of the same TR-NPs shown in (c). Scale bars equal 500 nm for (a-c) and 10 µm for (d). (a-c) are 85° tilted views.

Download Full Size | PDF

Additional fabrication experiments reveal the probable influence of the lattice pitch in the horizontal etch rate if compared with the isolated TR-NP instance. In the latter case, horizontal etch rate seems to be higher, as expected, probably because the TR-NP is more exposed to the surrounding plasma. In the case of different lattice pitches, seems that there are not important differences in horizontal etch rate and TR-NPs morphology in the 500-800 nm pitch range.

In order to compare the TR-NPs performance, bulk RI sensing tests were conducted with the abovementioned TR-NPs array along with a conventional R-NPs array made like the samples previously reported [17–19]. In both cases, arrays have square lattices of nanopillars of ~250 nm in diameter, pitch 500 nm. The optical response (reflectance) is shown in Fig. 4. The multilayer stack in both cases is the same and was designed to obtain a prevailing resonant dip for previous works [17] involving conventional R-NPs, as can be checked in Fig. 4(a), where the studied peak for bulk sensing is marked in a box. Figure 4(b) shows the same spectra for the TR-NPs with nickel cap array instance. The latter shows a less defined minimum resonant peak because the multilayer stack is not optimized to match the photonic gap of the TR-NPs array, but it is good enough to study the bulk sensitivity and allows for an optical behavior comparison considering the same stack and lattice parameter. The optimization of the array configuration, the metal employed and the multilayer stack will be studied in future works for a better tuning of the studied peak with the photonic gap given by the resonant cavity. Note that TR-NPs with nickel cap spectra show additional and better defined resonances located at longer wavelength which therefore can be related with the resonant cavity, however the focus of this work is to compare the same spectral region.

 

Fig. 4 Spectra acquired with a visible spectrometer considering different solvents in contact with a conventional R-NPs array (a) and a TR-NPs with nickel cap array (b). In both cases, there is a resonant minimum peak in the 580-630 nm wavelength range. Minimum peaks regions are enlarged to improve clarity.

Download Full Size | PDF

Figure 5 shows the linear fitting for conventional R-NPs and nickel capped TR-NPs arrays considering the resonant minimum peak position shift. In both cases, the sensor response is highly linear (adjusted correlation coefficient, R², is 0.9972 for R-NPs and 0.9966 for TR-NPs) over a wide range of RI values, giving bulk sensitivity values of S_B = dλ/dn = 295.93 nm RIU⁻¹ for R-NPs and 379.13 nm RIU⁻¹ for TR-NPs considering the minimum peak associated with the resonant cavity, which is located in the 580-630 nm wavelength range for the solvents employed in the test. This result means a 28.11% bulk sensitivity increment for the nickel capped TR-NPs presented in this work. Standard deviation (SD) of every 100 measurements for every solvent was set in 0.12 nm, which is the resolution of the spectrometer, since the calculated SD is sometimes slightly under that value, demonstrating very high measurement stability.

 

Fig. 5 Linear fitting of the resonant minimum peak position for a conventional R-NPs array (solid line) and a TR-NPs array (dashed line), square lattice and pitch 500 nm in both cases. Bulk sensing tests with different solvents gave values of S_B (R-NPs) ~296 nm RIU⁻¹ and S_B (TR-NPs) ~379 nm RIU⁻¹, showing a 28% sensitivity increment approximately.

Download Full Size | PDF

4. Conclusions

Periodic arrays of high aspect ratio, tiered Si₃N₄/SiO₂ resonant nanopillars with nickel caps working as nanomirrors were fabricated by means of EBL and ICP-RIE through controlled and selective, 3D dry-etching of the nitride/oxide multilayer. The resulting tiered nanostructure involves less fabrication steps and the available specific surface is significantly higher if compared with previously reported conventional nanopillars. The metal capped TR-NPs array showed robustness during a bulk sensing test involving different solvents and drying steps, and demonstrated a remarkable improvement as optical sensor suitable for chemical and biochemical sensing, giving a bulk sensitivity of 379.12 nm RIU⁻¹ considering the resonant peak in the 580-630 nm wavelength range, highlighting a 28% increment in sensitivity comparing with standard Si₃N₄/SiO₂ conventional R-NPs. Future work comprises the study of alternative multilayer configurations, different array parameters and different cap metals in order to find better quality spectral peaks for sensing and to improve the sensor sensitivity. Finally, the development of the selective, 3D dry-etching process employed in this work opens the door to a controlled one-step fabrication of tiered or undercut novel nanostructures.