1. Introduction

Localized surface plasmon resonance (LSPR), collective motions of conduction electrons, surppoted by metallic nanoparticles, has attracted wide attention in the past few years for its promising applications such as surface enhaced spectroscopies, nanolasers, and chemical or biosensing [1–3]. Specially, LSPR has become a powerful analytical tool for label-free biomelecule sensing because its high-sensitivity to the local dielectric ambient and direct excitation by incident light [4–11]. The optical properties of these structures depend sensitively on their geometry parameters, making it possible to engineer their electric and magnetic responses over a broad specral range [12]. The strong field localization, arising in metallic nanostructures at their resonance position, has extensive applications in the fields of chemical and biomolecular sensing [13–17]. Well-controlled structures relying on nanotechnique mean that level of more insight and control has been gained on localized surface plasmon resonances. A wide variety of complex plasmonic nanostructures have been realized by utilizing conventional nanofabrication technologies or chemical methods. Traditionally, plasmonic nanotructures are fabricated on planar substrates by top-down processes, such as focused ion beam milling or electron beam lithography. With these methods, geometries can be prepared precisely to support electric, magnetic and Fano-like resonances, such as metallic nanoring resonantors [9,10] and planar metamaterials with rectangle hole [18–20]. Although these fabrication routes allow for a high spatial resolution (∼10 nm), high-cost hampers the large-area fabrication of nanostructures. Furthermore, all these fabrication strategies are not well suitable for the fabrication of complex three-dimensional structures, further limiting the optical coupling, and the processes tend to genarate surface defects and roughness resulting in scattering losses. Other alternative methods by means of bottom-up approaches include chemical self-assembly [21], laser printing, and transfer nanopriting [22,23]. With the innovative method of ‘transfer’, template-stripping nanofabrication technique has realized successfully to produce two-dimensional arrays of large nanoparticles over wafer-scale area [22]. More recently, a gold nanohole array is adhered to an optical fiber with a beveled face by a silicon template [23], and a highly ordered nanodisk array is successfully prepared onto a conductive layer based an ultrathin anodic aluminum oxide (AAO) template [11]. These methods provide a low-cost route to constuct two-dimensional optical materials. However, it remains challenge to fabricate structures with complex shapes, or percolated features such as waveguides and resonators.

Solid state dewetting of ultralthin films is the most immediate approach of fabricating substrate-supported metal nanostructures, which utilizes the fluid instabilities to realize self-assembled nanostructures [24–29]. The dewetting phenomenon is driven by an interplay between thermodynamics and kinetics. When a continuous ultralthin metal film deposited on a substrate is heated at a metastable state, thermodynamics drive tends to reduce the surface area of metal. But the metal atoms lack enough kinetic energy to move significant distance across the surface, leading to a continuous metastable film. Such a film is heated toward equilibrium state, and it will agglomerate into metal nanostructures when temperatures are fall well below the melting point of the metal. Ultrathin films are often deposited onto the flat substrates at room temperature followed by their subsequent agglomerated at elevated temperatures in the past few years [24,25]. The widespread use of this thermal solid state deweeting process is largely due to its easy preparation of nanostructures over large areas. However, the nanostructures produced are unsatifactory as there is a lack of control over nanoparticle size distribution, spacing and position on substrate. Dewetting technology is gradually marginalized with the the stringent requirements, such as the tunning of size, shape, and spatial arrangement of nano-objects.

In this paper, we propose a new approach to fabricate plasmonic metamaterials on the free-standing porous anodic aluminum oxide AAO template by manipulating the dewetting effect. The interplay between texture reflow, spinodal instabilities and capillary break-up is effectively controlled by inherent nanopattern of AAO template to generate two sizes of metallic nanoparticles over the AAO membrane. The optical characteristics and sensing performance of the fabricated plasmonic metamaterials are systematically investigated, which demonstrated that the structure has excellent biosensing performances including bulk, surface sensing and protein detection.

2. Materials and methods

The AAO templates with large length-diameter ratio are successfully obtained by the two-step anodization method. A high purity aluminum sheet is firstly annealed for 2 hours in a H₂/Ar (volume ratio = 1:5) to eliminate internal stress. After that, it is electrochemically polished (voltage = 7 V) for 20 minutes in a mixture of perchloric acid and absolute ethyl alcohol (volume ratio = 1:5) at the temperature range of 13 ∼ 14 °C. Next, the pretreated aluminum flake is performed by first time constant current anodizing (constant current = 8 mA/cm², temperature = 0 °C) for 6 hours in oxalic acid electrolyte with 0.3 M. Therewith, the aluminum sheet is immersed in the mixture of phosphoric acid (0.4 M) and chromic acid (0.2 M) for 2 hours at a temperature of 60 °C. After the pellumina that is formed at first anodizing process is cleared away with chemical methods, the second anodizing is carried out under same conditions. Finally, after the second anodizing, the aluminum substrate is dissolved by using the mixture of cupric chloride and hydrochloric acid solutions, and the barrier layer at other side is eliminated by phosphoric acid solution (5 wt%), obtaining the bi-pass AAO template. In the above fabrication process, the pore size of the AAO templates is controlled by the corrosion time in phosphoric acid. And the thickness of AAO template is controlled by the second anodizing time. When the second anodized time increases to 18 hours, the template thickness rapidly grows above 120 µm.

The sketch map of fabrication method of the plasmonic metamaterials and corresponding SEM images are depicted in Fig.  1. Firstly, a 20 nm gold film is deposited on the AAO membrane by electronic beam evaporation. Finally, the process of annealing is performed for 5 hours at temperature of 600 °C with the protection of Ar gas environment. Annealing induces the dewetting of the metal layer, generating two types of Au nanoparticles.

 

Fig. 1. (a) Schematic of the fabrication process of the plasmonic metamaterials via the solid states dewetting technic. Left, the free-standing AAO template fabricated by using two-step anodization method. Middle, the deposition of 20 nm gold film. Right, the dewetting of Au film induced by annealing. (b) The corresponding SEM images of every stage. Scale bar from left to right, 500 nm, 3 µm and 1 µm. Scale bars in the insets are 300 nm.

Download Full Size | PDF

The formation of Au nanoparticles is attributed to the template dewetting process, which is the result of two competing processes, namely template reflow and spinodal dewetting. Template-induced reflow, driven by the Laplace pressure at the optical materials/air interface, tends to break up the film to reduce its surface area. The template-based spinodal dewetting effect, which represents an effective molecular interaction between the film and underneath texture, can accelerate thinning, leading to instabilities and eventually breaking-up of the film [29]. The ultimate morphology of the nanostructure is considerably affected by the thickness of the deposited metallic film and the texture of the underlying template. When the gold film is too thick, it is difficult to form nanoparticles on the AAO template by high temperature annealing. However, if the gold film is too thin, large size of nanoparticles is not easy to form. For the AAO template with cylindric pores, the distance between the holes is relatively spacious, leaving material that has not reflowed into the concave region, forming ultimately large size metallic nanoparticles on top of mesas. Moreover, both the inherent polycrystalline microstructure in metals and the surface defects of AAO membrane lead to the formation of small nanoparticles.

3. Results and discussion

3.1 Optical responses of plasmonic metamaterials

Figure  2(a) shows the experimental extinction spectrum of the fabricated nanostructure at normal incidence, demonstrating two resonant peaks within the spectral region of interest. For experimental measurements, we collect the transmission data using an angle-resolved system (R1, Ideaoptics), which is equipped with a UV-vis-NIR light source (iDH2000H-B, Ideaoptics) with wavelengths ranging from 430 nm to 2500 nm. Data collection is performed under a saline environment with refractive index of 1.3341. Herein, numerical simulation is performed by using of a commercial finite element method (FEM) package (COMSOL Multiphysics) to illustrate the origin of these two plasmon modes. In our simulations, two sizes of metallic nanoparticles are modeled, the diameters of which are 150 nm and 50 nm, respectively. The AAO substrate only introduces a weak scattering peak. The two sizes of metallic nanoparticles are embedded in a homogenous medium with refractive index of 1.33. The dielectric constant of gold is taken from Johnson and Christy [30]. The linearly polarized light with wavelengths ranging from 400 nm to 800 nm impinges perpendicularly on the metal surface.

 

Fig. 2. (a) Experimental and (b) FEM calculated normalized extinction spectra of the plasmonic metamaterials. The resonant peak position is indicated by λ₁ and λ₂, respectively.

Download Full Size | PDF

The calculated result plotted in Fig.  2(b) is in good agreement with the experimental spectrum. It is clearly seen that the spectral superposition of nanoparticles gives rise to two distinct resonant peaks that stem from the optical response of the small and large nanospheres, respectively. The difference between experiment and calculation may stem from the nonuniform of prepared nanoparticles.

3.2 Bulk refractive index sensitivity

The ability of the fabricated sensor to transduce changes of surrounding refractive indices (RIs) is firstly examined. The bulk refractive index sensitivity S_bulk = ΔR/Δn is often used to quantitatively evaluate the sensing performance of LSPR-based sensors, where ΔR is the change in the resonant wavelength or intensity of plasmon modes and Δn is the change of refractive index of the environment. In order to real-timely monitor the evolution of the signal with changing ambient RIs, a microfluid channel is assembled on the top of the structure, where the real-time data collection is performed using a home-made labview program. Six gradients of sodium chloride (NaCl) solutions with different mass percent are injected automatically into the microfluid channel by using a peristaltic pump at a constant flow rate of 0.2 mL/min. The RI of the NaCl solutions is calibrated experimentally by using an Abbe refractometer as 1.3341, 1.3427, 1.3496, 1.3556, 1.3636 and 1.3735. Figures  3(a) and 3(b) show the time responses of two resonant peaks of the structure with intensity measurement strategy with respect to successive NaCl solution, which exhibit the step-down decrease of the normalized intensity with the step-wise increase of RI. The average signal intensity of these two resonant peaks under each step RI change in time responses process as a function of the corresponding RI are shown in Figs.  3(c) and 3(d). All fitting curves suggest the negative correlation between normalized intensity and the RI of the injected solution. The slope of linear regression is 1.3652 and 1.4959 RIU⁻¹ for resonant peaks λ₁ and λ₂, respectively. Therefore, the bulk refractive index sensitivity for plasmon mode λ₂ is larger than that of mode λ₁.

 

Fig. 3. Sensorgrams for resonant peaks (a) λ₁ and (b) λ₂. Step 1 to 6 respectively represents the solution RI of 1.3341, 1.3427, 1.3496, 1.3556, 1.3636 and 1.3735. (c, d) The corresponding calibration curve of two resonant peaks in response to different surrounding RI solutions. The error bars show the standard deviation for three repeated measurements. (e, f) RI resolution of two resonant peaks estimated by employing relationship between signal-noise ratio ΔI/N and RI change Δn, where ΔI and N refer to intensity change and experimental noise, respectively.

Download Full Size | PDF

Based on these results in Figs.  3(a) and 3(b), the experimental noise N of resonant peak λ₁ is 3.95×10⁻⁴, 6.61×10⁻⁴, 4.40×10⁻⁴, 3.86×10⁻⁴, 6.73×10⁻⁴ and 5.06×10⁻⁴. For resonant peak λ₂, the experimental noise N is 4.53×10⁻⁴, 4.12×10⁻⁴, 4.22×10⁻⁴, 4.35×10⁻⁴, 7.06×10⁻⁴ and 3.45×10⁻⁴, respectively. On the basis of above experimental noises and the bulk sensitivity of resonant peaks λ₁ and λ₂, Figs.  3(e) and 3(f) show the signal-noise ratio (S = ΔI/N) versus refractive index change. Herein, the RI resolution is defined as the ratio of the experimental noise and bulk refractive index sensitivity, that is, r=|N/S|=|Δn/(ΔI/N) |. According to the above definition, the RI resolution of resonant peaks λ₁ and λ₂ is calculated to be 3.27×10⁻⁴ and 2.53×10⁻⁴ RIU, demonstrating that resonance peak λ₂ is possible to detect and distinguish a tinier change of surrounding medium than the case of resonance peak λ₁.

3.3 Surface sensitivity

The sensitivity of the LSPR sensor not only depends on the dielectric properties of surrounding, including changes induced by binding of molecules to the metal structures, but also on the distance from the metal surface since the electric field of the metal nanostructures is highly confined and decays away from the surface. Therefore, the commonly studied bulk sensitivity and RI resolution are not optimal parameters to represent the biosensing performance. In this section, surface sensitivity is investigated by using self-assembled monolayers of polyelectrolytes with known thickness. Employed poly(allylamine)hydrochloride (PAH, 65 kDa) with positive charge and poly (styrene sulfonate) (PSS, 75 kDa) with negative charge are assembled on the top surface of the nanoparticle array by using a layer-by-layer (LBL) alternate deposition. The alternate deposition process of PAH and PSS polyelectrolytes is schematically illustrated in Fig.  4(a). The thickness of each polyelectrolyte bilayer is about 2.9 nm measured by ellipsometry [31]. According to the LBL method, we obtained the spectral evolution of the nanosheet with the increasing number of PAH/PSS bilayer, as shown in Fig.  4(c). It is demonstrated that optical intensity change tends to plateau when the number of PAH/PSS bilayer increases to 17 layers, stemming from the electric field of Au nanoparticle decays away from the surface.

 

Fig. 4. (a) Schematic of assembling the PAH-PSS bilayer on the plasmonic structure. Exponential fitting of the intensity changes at the peaks (b) λ₁ and (d) λ₂, as a function of polyelectrolyte multilayer thickness. (c) Representative spectra shift of the sensor with the number of PAH-PSS bilayer increasing. (e) Comparison of the second order surface sensitivity curve for two modes. The shadowed areas show the thickness range in which the curves in corresponding to colors have higher surface sensitivity.

Download Full Size | PDF

The surface sensitivity strongly depends on the electric field distributions, which decays away from the metal surface. Therefore, to further analyze quantitatively the surface sensitivity of this structure, the decay length of optical field near the metal surface is described reasonably well on the basis of the following equation [32,33]

3.4 Real-time detection of specific binding between protein molecules

To further evaluate the sensing ability of the plasmonic metamaterials, we make use of the specific binding of the glycoprotein concanavalin (Con A) and the ribonuclease B (RNase B) to carry out the sensitive measurement of the Con A protein. This is depicted in Fig.  5(a), where the fabricated nanostructure is functionalized with RNase B that is immobilized on the metal surface to capture protein molecule Con A s with different concentrations. The more detailed immobilization process of the RNase B is as follows: First of all, the sensing nanochip is cleaned with ultrapure water and ethanol. After cleaned, the sensing layer is immersed in an ethanol solution of 11-meraptoundecanoic (MUA, 10 mM) at room temperature for 24 hours to obtain an alkanethiol monolayer self-assembled on the metal surface. The unreacted thiol molecules are washed away with ethanol. After dried with N₂ gas, the sensing nanosheet, at temperature of 4 °C, is dipped into a mixed aqueous solution of 1-ethyl-(3-dimethylamino-propyl) carbodiimide hydrochloride (EDC, 0.55 M) and N-hydroxysuccinimide (NHS, 0.5 M) for 30 minutes to activate the alkanethiol monolayer on the metal surface. Subsequently, the sensing chip is rinsed with deionized water and dried in a N₂ environment, and then the sensing chip is equipped with the microfluidic channel. RNase B with mass concentration of 0.1 mg/mL in phosphate buffered saline (PBS) is injected into the sensor microchannel for 30 minutes to form a stable monomolecular layer. After the rinsing in PBS buffer, bovine serum albumin (BSA, 0.1 mg/mL) is used for 30 minutes to deactivate the remaining activated carboxyl sites that did not combine with the RNase B.

 

Fig. 5. (a) Schematic of the specific binding between Con A and RNase B. The sensor responses to the Con A solution at concentrations of 0.9, 1.8, 4.5, 9, 18 μM for resonance peaks (b) λ₁ and (c) λ₂. (d, e) Corresponding calibrations curves of panels b and c.

Download Full Size | PDF

By injecting Con A with different concentrations of 0.9, 1.8, 4.5, 9, 18 μM, the change of these two resonance peaks of the sensor are monitored in real-time, which represents the dynamical process of the specific binding between RNase B and Con A. The whole process is summerized as follows: First, PBS solution is injected into the flow channel to obtain the baseline. Then, Con A solution is pumped to bind with RNase B on the sensing surface. Next, PBS buffer solution removes the unbound Con A molecules. Finally, the urea solution (0.8 M) is used to strip the surface-bound Con A molecules to effectively regenerate the sensing region, and the next concentration can be tested sequentially. Figures  5(b) and 5(c) show that resonant intensity has an abrupt change at time T₁ due to the specific binding of Con A and RNase B, whereas tends to return an initial state at time T₂ resulting from the buffer washing. Each concentration of Con A solution is carried out three times to guarantee the reliability and repeatability of experimental results. The relationship between the signal intensity and the concentration of Con A are displayed in Figs.  5(d) and 5(e). Determined from the experimental noise and the relationship between the intensity response and Con A concentration, the limit of detection (LoD) of 68 and 79 nM for plasmon modes λ₁ and λ₂ are estimated, respectively. The experimental results confirm that the plasmonic metamaterials offer a high-performance biosensing platform.

4. Conclusion

In conclusion, we introduce a novel nanotechnology based on the template dewetting method which provide a low-cost and large-scale fabricated route, as well as fabricate successfully a metallic nanoparticle array over a large length-diameter AAO template. The system supports dual-band resonant modes due to the spectral superposition between nanoparticles with two sizes, and the bulk and surface sensitivity of these two plasmon modes are systematically investigated. Moreover, we further demonstrate the sensing ability of real-time detection of Con A of these two resonance peaks, demonstrating respectively a limit of detection as low as 68 and 79 nM, which is highly advantageous for label-free biosensing at ultralow analyte concentrations.