Advanced Materials

Large-Area Low-Cost Plasmonic Nanostructures in the NIR for Fano Resonant Sensing

Authors


Abstract

Large-area low-cost plasmonic nanostructures with high-quality Fano resonances have been fabricated for the first time. Hole-mask colloidal nanolithography is utilized and the degree of asymmetry of double split-ring resonators is varied to optimize the modulation depth of the Fano resonances for refractive index sensing. In situ optical spectroscopy during thermal annealing monitors the spectral change to control improvement of sample quality. Liquids with different refractive indices yield experimental sensitivities of up to 520 nm/RIU and figures of merit of 2.9.

original image

Plasmonics has recently received a tremendous amount of attention, in particular, due to its potential for generating exceptionally large optical field enhancements in cubic-nanometer volumes. Among the many unique features of plasmonic structures, this field enhancement phenomenon has inspired researchers to construct refractive index sensors that utilize localized surface plasmon resonances (LSPRs).1–3

Complex metallic nanostructures are able to support bright as well as dark optical modes.4 When designed with the appropriate geometry and symmetry, metallic nanostructures have the potential to exhibit narrow Fano resonances due to destructive interference of the different modes.5–8 These Fano resonances can provide enhanced sensitivity and figure of merit of localized plasmon sensors,9–12 which easily reach or even exceed the best known LSPR sensors in single plasmonic particles.13–15 This is due to the fact that the complex geometry and the many degrees of design freedom allow for tailoring the resonant near-fields in a very small volume. Additionally, the spectrally narrow Fano resonances provide steep spectral features, which shift strongly upon refractive index change of the surrounding material and hence the local field. Therefore, measuring transmission and reflection intensity gives a strong response upon change of the environment.

One suitable geometry to generate sharp resonances is the asymmetric double split-ring resonator, where two crescent shaped nanoparticles with different lengths are placed facing each other. In this structure, two gaps are present, and these are placed at angles other than 180°.16–18

Examples for sensors based on such asymmetric split-ring resonators include biosensors in the GHz and THz region,19, 20 as well as sensors in the mid-infrared.21, 22 The asymmetric plasmonic geometry provides also the potential for other nanophotonic applications, such as lasing spasers,23 coherent plasmon emitters,24 and tunable metamaterials with narrow linewidth and varying coupling strength. However, the large scale application of Fano sensors has been hampered by two issues: one is the ability to utilize low-cost visible or near-infrared lasers as light sources, and the other are pathways to mass-fabricate Fano sensors over cm2 areas at low cost. Nanostencil-type fabrication has been suggested25 but has yet to prove its potential. Nanoimprint lithography might provide another option,26 which to date has not been applied to fabricate Fano sensors.

In this paper, we demonstrate how to overcome these two major obstacles, which would pave the road towards wide utilization of this cutting-edge technique. To the best of our knowledge, this work represents the first large-area low-cost plasmonic Fano sensing LSPR structure.

We fabricate asymmetric double split-ring-resonators (ADSRRs) by hole-mask colloidal nanolithography27–30 and vary the degree of asymmetry to study the evolution of the Fano resonance. Thermal annealing31 is utilized to improve the modulation depth of the Fano resonance or to turn off the coupling. In particular, we carry out time-dependent in-situ studies of the plasmonic spectra during heating. Refractive index sensing with different liquids on our low-cost substrate around the Fano resonance in the near infrared is demonstrated.

Figure 1a shows the schematic diagram of the fabricated ADSRRs with gap angle δ, width of the ring w, radius of the ring r, arc angles l1 and l2. We use a constant radius r in the experiment, so the gap δ, as well as the arc angles l1, l2 are given as angle for easy reproducibility. Both arcs are aligned along the y-axis. The degree of asymmetry is defined as a = l1/l2. Normally incident light with different polarizations, with the y-axis defined as 0° is used in all measurements. Figure 1b is a photograph of one of our 1 cm2 samples, which is covered completely and homogeneously with the plasmonic nanostructures, except for the little dark spot near the center, which was caused due to a problem with our spin coater.

Figure 1.

(a) Schematic diagram of planar asymmetric double split-ring resonators, with gap angle δ, width of the ring w, radius of the ring r, arc length l1 and l2. The degree of asymmetry is defined as a = l1/l2. All measurements are taken with normally incident light with different polarizations with 0° polarization parallel to the y-axis. (b) Photograph of one of the samples, size 1 cm2. The sample is fully covered with asymmetric double split-ring resonators, except for a little hole near the center.

ADSRRs with different geometrical parameters and their optical properties are shown in Figure 2. Scanning electron microscopy (SEM) images of large-area ADSRR samples A to D (Figure 2a) demonstrate the low defect concentration. Higher magnification micrographs (Figure 2a inset) show the geometric details of the ADSRR, the low size and shape variation, as well as the sufficient spacing between neighboring elements (i.e., no dominant near-field interaction in the few 10 nm range). The cobweb-like structures and the nanoparticle in the center of the ADSRRs in the insets arise from residual PMMA that was not removed by oxygen plasma etching and the subsequent lift-off process. This residual PMMA does not affect the optical properties of the structures very much, except for a small red shift due to its refractive index. It can be removed by additional oxygen plasma cleaning (see supplement). The gap size δ = 30° (∼45 nm) is kept constant, while the left arc length l1 is increased and the right arc length l2 is decreased from sample A to D. Therefore, the asymmetry parameter a also increases. The width of the ring w depends on the hole-size of the mask and the evaporation angle, which are shown in methods section. For all samples, we use the same hole-mask and a constant evaporation angle. The gold thickness of all samples is about 15 nm. Optical characterization of structures was carried out using a Fourier-transform infrared (FTIR) spectrometer (Bruker Vertex-80). Optical spectra were obtained using normally incident light with polarization 0° (black), 45° (blue), and 90° (pink), with 0° parallel to the arcs. The required geometry parameters are simulated using an FIT code (CST Microwave Studio), as shown in Figure 2c. The refractive index of glass is assumed to be n = 1.5, and the permittivity of bulk gold is described by a Drude model with plasma frequency ωp = 1.37 × 1016 Hz and a damping constant κ = 1.2 × 1014 Hz. The used SRR structure in simulation is rectangular, with semicircular shapes on the two ends. The simulation agrees very well with our experiments. For 0° polarization, we observe an asymmetric Fano resonance at around 220 THz in sample A, due to the coupling of a subradiant dark mode with a superradiant bright mode. This coupling becomes stronger with increasing asymmetry parameter a. In the spectrum of sample C, the Fano resonance becomes very sharp and deep. As the asymmetry parameter increases more, the subradiant hybridized mode changes from dark to bright. In sample D, the Fano resonance disappears and only hybridization and dipole-dipole coupling is observable. For 90° polarization, we observe a weak Fano resonance at around 410 THz in simulation, which also becomes stronger with increasing a, and disappears for too high degree of asymmetry a in sample D. Due to the weak signal near to the limit of near infrared range of the light source in FTIR spectrometer, this Fano resonance for 90° polarization is not visible in experimental results. We only observe an asymmetric broadening of the resonance at around 400 THz in the spectra for 90° polarization of sample B and C. For 45° polarization, all plasmon modes for 0° and 90° polarization are excited simultaneously. To confirm these assessments, we calculate the electric field distribution for all the numbered spectral positions and respective polarizations of Figure 2.

Figure 2.

(a) SEM images of samples A, B, C and D with constant gap angle δ = 30° (∼45 nm) and varying arc lengths l1 and l2. The asymmetry parameter a increases from A to D. The gold thickness of all samples is about 15 nm. The cobweb-like structures and the nanoparticle in the center of the asymmetric double split-ring resonators in the insets of the images arise from residual PMMA. (b) Transmittance spectra of samples A-D, measured after 15 min annealing at 150 °C using normally incident light with different polarization. 0° polarization is parallel to the arcs. (c) Numerically simulated transmittance spectra. The field distribution of the numbered plasmon resonances is shown in Figure 3.

Figure 3 displays these electric field distributions for the samples C and D. Image 1 and 3 show a superradiant hybridized dipole mode (depicted by black arrows) in sample C at 0° polarization. The two parts of the SRR oscillate in phase. Image 2 shows a subradiant hybridized dark mode (i.e., the antisymmetric mode distribution does not couple well to the incident light field), which is located at the Fano peak, and the two SRRs oscillate out of phase. Image 4 and 6 shows the superradiant coupled 2nd-order SRR modes of sample C at 90° polarization, and image 5 shows the dark hybridized subradiant mode. For sample C at 90° polarization, we observe a relatively strong plasmon mode at 480 THz. This mode is a hybridized dipole mode perpendicular to the arcs due to the large width w of the SRRs. For sample D at 0° polarization, the Fano resonance disappears because of the too large asymmetry parameter a. We only observe a subradiant hybridized bright mode (8) at the transmittance dip, which is exited directly from the incident light, due to a relatively large residual aggregate dipole moment in the structure. In mode 9, only the right SRR is exited.

Figure 3.

The simulated electric field Ez at z = 15 nm (component perpendicular to structure) with color code (yellow = positive value, blue = negative value) at the respective resonances in Figure 2c with corresponding incident light polarization. Images 1–3 show the plasmon modes (represented by black arrows) in sample C at 0° polarization. A subradiant hybridized dark mode (image 2) is observable at the Fano peak. Images 4-7 show the plasmon modes in sample C at 90° polarization. In sample D the Fano resonance disappears due to high asymmetric parameter. Only a bright subradiant hybridized mode (image 8) and a fundamental SRR mode from the shorter arc (image 9) are exited.

From the experiment and simulation results, we find that the modulation depth of the Fano resonance depends on the degree of asymmetry a. The Fano resonance appears first with broken symmetry and becomes stronger with increasing a. The structure geometry changes also a lot, while we increase the asymmetry parameter a. The difference of the arc lengths l1 and l2 becomes larger, which influences the position of the plasmon modes of the two individual SRRs and reduces the coupling strength. Therefore, to achieve a pronounced Fano resonance, we should use the optimum asymmetry parameter a, as well as suitable geometry parameters l1 and l2.

In order to improve the modulation depth of the Fano resonances, we introduce an additional controllable process, namely time-dependent annealing of our samples at various temperatures and simultaneous measurement of their optical behavior by in-situ spectroscopy. A different approach, such as electropolishing of plasmonic structures, was investigated recently by other groups.32, 33 During the annealing, the gold coalesces due to high surface tension.34 If the sample is annealed too long, especially for very thin films, the structure can be destroyed, and the desired optical properties are not observed. This can be avoid with our in-situ measurement by annealing.

Annealing is performed using a small hot plate on the sample holder in the FTIR spectrometer and adjusting the temperature accordingly. Hence only reflectance measurements are possible. A spectrum is recorded every minute, until no changes are observed. In Figure 4a, no Fano resonance in the reflectance spectrum of sample C at room temperature (black curve) is observed, and the surface of the structures in the upper inset is rather rough. As the temperature increases to 150 °C, a rather unpronounced Fano dip appears in the reflectance spectrum (red curve). With longer annealing at 150 °C, the modulation depth of the Fano resonance becomes even larger, and all the resonances shift blue, because the arc lengths become shorter by about 10° during annealing. After 15 minutes, the Fano resonance becomes very sharp and deep (orange curve), and the structure surface is also much smoother, as shown in the lower inset.

Figure 4.

In-situ annealing of asymmetric double split-ring resonators. (a) The measured reflectance spectra of sample C, while it is annealed at 150 °C for 15 min. (b) The measured reflectance spectra of sample D, while it is annealed at 140 °C for 14 min. The upper insets in Figure 4a,b show the SEM images of the structures before annealing. The lower insets show the structures after annealing.

Figure 4b shows the in-situ annealing of sample D. At room temperature, only a rather unpronounced Fano resonance is visible in the spectrum, which does exist due to the relatively small gaps before annealing and sufficient coupling. However, because of the high degree of asymmetry, the residual dipole moment in the subradiant mode is relatively large, and this mode is not really dark. During the annealing, the gap size becomes larger by about 10° (∼15 nm), which reduces the coupling strength. Hence, the Fano resonance disappears quickly, and the coupling is turned off. We only observe a bright hybridized mode at lower energy and a fundamental dipole mode of the right SRR at higher energy. Using this in-situ spectroscopy during thermal annealing, we can monitor the improvement of sample quality.

Plasmonic Fano sensors have shown tremendous potential for LSPR sensing due to their large field enhancement as well as their extremely small mode volume, which can reach atto- or zeptoliters.22 They also show very high sensitivity around the Fano resonance.10–12 In order to evaluate the suitability of our low-cost Fano structures for refractive index sensing, we use sample C, which shows the best modulation depth, and measure its transmittance spectra upon exposure to ethanol (n = 1.354), butanol (n = 1.387) and glycerin (n = 1.460), refractive indices at ∼1500 nm (Figure 5a). We observe the expected redshift of all spectral features. Figure 5b shows the simulated transmittance spectra, which agree very well with the experiments. To quantify the sensitivity of the Fano resonance for sensing, we use the figure of merit (FOM) introduced by van Duyne and co-workers35

equation image
Figure 5.

Fano resonance in sample C used for LSPR sensing application. (a) Transmittance spectra measured by FTIR for different surrounding media. Air, n = 1.000 (black); ethanol, n = 1.354 (red); butanol, n = 1.387 (blue); glycerin, n = 1.460 (orange). (b) Simulated transmittance spectra.

Here Δλ/Δn is the resonance shift per refractive index unit. For the Fano peak, Δλ/Δn is around 460 nm/RIU in experiment and 560 nm/RIU in simulation. The full width at half-maximum (FWHM) is defined as the distance between the antipeak on the short wavelength side and the peak,36 which is around 170 nm in measurement and 196 nm in simulation, and the corresponding FOM is around 2.7 and 2.9, respectively. The dip in the transmittance spectra at longer wavelength is even more sensitive than the Fano peak, giving a sensitivity of Δλ/Δn=520 nm/RIU in experiment and 605 nm/RIU in simulation. The FWHM of this dip is around 180 nm in measurement and 125 nm in simulation, and the resulting FOMs are 2.9 and 4.8, respectively. Previous single particle LSPR sensors such as gold nanostars have reached sensitivities up to 665 nm/RIU and FOMs of 5.4.13, 15 The best electron beam lithographic structures exhibited FOMs up to 5.7.37 Regarding our minimal fabrication effort, our sensitivities and FOMs are suitable and could be easily improved in the future by using thicker structures, different geometries, and variation of evaporation conditions.

In summary, we presented the application of hole-mask colloidal nanolithography for low-cost large-area metallic nanostructures, in particular asymmetric double split-ring resonators. Our high-quality samples exhibit sharp and narrow optical Fano resonances in the near infrared around λ = 1.3–1.5 μm, which are well suited for localized surface plasmon resonance refractive index sensing with atto- or zeptoliter volumes. We achieve experimental sensitivities of up to 520 nm/RIU and experimental figures of merit of 2.9. Annealing the sample at a temperature of 150 °C for 15 min can lead to a strong improvement of the shape and the modulation depth of the Fano resonances due to the reshaping of the metal during heating. Our method will allow for large-area low-cost fabrication of high-quality complex plasmonic structures, such as oligomers38, 39 and chiral structures.40 This technique will pave the road for real-life plasmonic applications like optical gas sensors,41, 42 biosensors,22, 43 leak detectors,44 artificial nano-noses,45 as well as chiral plasmonic enantiomer sensors46 that could for example detect specifically glucose, which is a handed molecule. Our method is very flexible towards other structure geometries, scalable to even larger areas, very reproducible, and adaptable for other substrate materials and metals.

Experimental Section

Hole-Mask Fabrication: We use a 0.5 mm thick glass substrate, which is cleaned with acetone and iso-propanol in an ultrasonic bath for about 5 min and then dried using nitrogen. The glass substrate is then functionalized using isobutyl(trimethoxy)silane in toluene (10 mM) for 8 h. This process helps adhere the PMMA film to the glass substrate, which is important for the later lift-off the polystyrene spheres (PS). A 150–250 nm thick PMMA film (Allresist, 200k 4 wt% in chlorobenzene AR-P 641.04) is spin coated onto the glass substrate at 4000 rpm and hard baked at 165 °C for 2 min. To decrease the hydrophobicity of the polymer film, we treat the sample in an oxygen plasma for 12 s (TECHNICS GMBH, PAV TePla 100-E, 250 mTorr, 50W). We then immediately pipette a droplet of a polyelectrolyte solution (Sigma Aldrich, poly(diallyldimethylammoniumchloride), 0.2 wt% in water) on top of the sample, to provide a net positive charge to the PMMA layer. This is followed by rinsing the sample with deionized water and drying in a nitrogen stream. A droplet of water suspended polystyrene spheres (Polysciences, Polybead Microspheres 100 nm diameter, 0.02 wt% in water) is deposited on the sample and rinsed away with deionized water after 1 minute. In the next step, we put the sample into a hot water bath (98 °C, 3 min) and dry it with nitrogen. The PMMA layer is now covered with well separated and randomly arranged PS spheres. Subsequently, a thin oxygen plasma resistant film (2 nm Cr and 20 nm Au) is evaporated on top of the sample. We used an ultrasonic bath with controllable power (90 V, 12 min) to remove the PS spheres afterwards and obtained a perforated gold layer on top of the thick PMMA layer. Finally, an isotropic etching step (oxygen plasma, 7 min, 250 W, 1 Torr, same machine as above) creates extended holes in the PMMA layer underneath the holes in the gold mask. After the SRR evaporation process, the sacrificial PMMA layer is removed by immersing the sample into acetone in an ultrasonic bath for a few seconds.

Evaporation of SRRs: The evaporation process is carried out in a modified thermal evaporation machine (Edwards E306). The sample can be rotated inside the evaporation chamber in polar and azimuthal directions using two vacuum-compatible stepper motors. The control of the azimuthal angle θ and the polar angle l1 and l2 for the two SRRs in combination with the PMMA thickness h0 determines the radius r and the gap angle δ of the ADSRRs. The radius r is given by

equation image((1))

and the gap angle δ is determined by

equation image((2))

The width of the SRR w is determined by the mask-hole diameter, namely, the diameter of the used nanospheres. A layer thickness monitor (crystal oscillator) is used to determine the structure height. First a thin adhesion layer of chromium is evaporated onto the glass substrate and subsequently the gold. The obtained structure height h1 and h2 are calculated using a geometrical model[28]

equation image((3))

Here h is the evaporated material thickness, which is displayed on the layer thickness monitor.

Sensing: We drop-coat the different liquids for sensing onto the sample and covered it with an infrasil cover slide. Between different liquids, the sample was rinsed with acetone and blow-dried with compressed nitrogen.

Supporting Information

Supporting Information is available from the Wiley Online Library or from the author.

Acknowledgements

We acknowledge R. Taubert for critical reading of this paper. We acknowledge the Alexander von Humboldt foundation (Bessel Prize) for PVB and a US DOE EFRC grant (DE-SC0001293) for PVB and CZ. This work was financially supported by BMBF (3D Metamat and NANOGAS), DFG (FOR 730, SPP1391), and BW-Stiftung.