1. Introduction

Localized Surface Plasmon Resonance (LSPR) of noble metal nanoparticles has been proven to be very important in chemical and biological sensor applications [‎1], [‎2]. Due to the strong field confinement effects and large extinction cross-section, metal nanoparticles are very sensitive to the dielectric environment close to the surface, with the field decay length in the nanometer scale. They have already been widely investigated in detecting the shift of the LSPR peak owing to the change in the refractive index of a surface bound layer, such as binding of target biomolecules to receptors immobilized on the surface [‎3]. The LSPR of metal nanoparticles depends on the geometry, composition, dielectric environment and the interactions among particles. Among metal nanoparticles of different materials, gold nanoparticles were chosen due to their resistance to surface oxidation to guarantee stable sensor performance over a long period. Gold nanoparticles of specially engineered geometries have shown near-infrared LSPR which is more sensitive than their visible counterparts [‎4]. Gold nanoparticles with symmetric shapes, such as gold nanoshells [‎5] and nanorings [‎6], have been previously demonstrated with tunable LSPR. A type of asymmetric shaped gold nanoparticles, named gold nanoscrescent, has been demonstrated with highly tunable near-infrared resonances and strong field confinement at the tips [‎7-‎9].

Hydrogels are polymeric networks produced by the reaction of one or more monomers which are able to retain water within their structure without dissolution. Hydrogels are used in a wide range of applications including drug delivery systems, absorbents, soil water retainers, contact lenses, and membrane materials [‎10]. Over the past decade, hydrogels have also been demonstrated to show promise in the realm of sensing [‎11-‎14]. They swell or contract upon hydration due to the hydrophilic nature of the polymer chains, and their swelling is limited by the degree of cross-linking of the polymer network. Due to its aqueous environment, hydrogels are biocompatible by nature. This biocompatible property is highly attractive for biosensing applications, which may involve the immobilization of biological recognition elements such as enzymes [‎11], antibodies, nucleic acids, etc.

Combing LSPR spectroscopy with hydrogel thin films open new possibilities for LSPR sensor applications, because the hydrogel can be engineered to change their swelling properties in response to a wide variety of chemical and physical stimuli such as pH, ionic strength, solvent composition, buffer composition, temperature, pressure, electromagnetic radiation, and photoelectric stimulus [‎15]. Towards this goal, the optical response of the hydrogel thin film on LSPR transducers must be well studied. Mack et. al. have demonstrated a pH sensor based on a hydrogel-coated gold plasmonic structure [‎16]. The swelling or shrinking of the hydrogel, when exposed to different pH solutions, is transduced into a change of the integrated transmitted intensity. Here we report the fabrication (by electron beam lithography) of a close-packed periodic array of gold nanocrescent structures (Fig. 1(a) ) and its demonstration as a transducer for very sensitive detection of the swelling of a pH-sensitive hydrogel thin film. Both the wavelength-shift sensing and integrated transmission sensing in the near-infrared regime due to the refractive index change of hydrogel thin film have been demonstrated.

 

Fig. 1 Fabrication of the gold nanocrescent array. (a) SEM micrograph of the fabricated gold nanocrescent array structure. (b) SEM micrograph at a tilted angle of the developed photoresist thin film coated with gold. (c) Schematic of the fabrication process of the gold nanocrescents.

Download Full Size | PDF

2. Fabrication of the sensor device

Electron beam lithography is used to fabricate gold nanocrescents with sharp tips and the fabrication process is schematically demonstrated in Fig. 1(c). A glass slide is used as the substrate, which has its surface pretreated with a 25nm coating of indium tin oxide in order to make the surface conductive. Electron beam lithography is used to expose periodic circular lines on a 200nm thick layer of photoresist (ZEP 520A, ZEON corporation) coated on the substrate. The lattice constant is 330nm and the diameter of the circle is 220nm. When developing the photoresist, the thin film turns into a pattern of circular posts centered inside the periodic array of circular holes. Due to the proximity effects in electron beam exposure and the undercutting sidewall profile of ZEP, each post has a very narrow base in contact with the substrate. From the surface tension during the drying of the sample, each post collapses as a consequence onto one edge of the hole which yields a crescent-shape opening, as can be clearly observed in Fig. 1(b). A 2nm layer of titanium and a 50nm layer of gold are sequentially deposited by physical vapor deposition followed by metal lift-off in solution.

The geometries of the fabricated nanocrescents are easily controlled by electron beam exposure dose on the photoresist. The fabricated gold nanocrescents shown in Fig. 1(a) have an average outer diameter of 260nm, tip-to-tip distance of 150nm and wall thickness of approximately 80nm. The fabricated tips have sub-15nm sharp features, which is very difficult to achieve by directly exposing the crescent shape pattern onto the photoresist. Our method can precisely control both the shapes and the locations of each gold nanocrescent which makes it possible to study the array effects, either in square or hexagonal lattice. The major drawback of our method is that the surface-tension-induced orientation of the fabricated nanocrescent cannot yet be controlled. In comparison, the gold nanocrescents fabricated via nanosphere lithography (NSL) are controlled by sizes and locations of the nanospheres and the tilted deposition angle [‎7-‎9]. Recently, Retsh et. al. has demonstrated a low cost process for fabricating the large area ordered gold nanocrescent array using a shadow mask prepared by reducing sizes of hexagonally close-packed colloidal nanospheres [‎17].

A pH-sensitive hydrogel based on poly(hydroxyethyl methacrylate-co-methacrylic acid) was developed as the sensing layer for the gold nanocrescent structure. Prior to coating the poly(HEMA-co-MAA) onto the surface, the devices were treated with 3-(trimethoxysilyl)propyl methacrylate to promote bonding to the surface using a previously described silanization method [‎18]. Hydroxyethyl methacrylate (HEMA) was mixed with 8 mol% ethylene glycol dimethacrylate (EDMA) and 6 mol% methacrylic acid (MAA) and prepared in an equal volume of isopropanol as the solvent. The MAA monomers allow the hydrogel to change its volume reversibly by swelling or shrinking as a function of the pH of the buffering medium that is used to bathe the gel. 2% w/v of 2,2-Dimethoxy-2-phenylacetophenone (DMPA) photoinitiator was added and slowly mixed into the solution. In order to grow the polymer, the solution was mixed, degassed in a stream of nitrogen, and exposed to UV for 30 seconds to initiate the free radical reaction at room temperature. This pre-polymerization process was repeated once more. The grown polymeric solution was then spin-coated on the gold nanocrescent structure at 4000 rpm for 30 seconds to produce a gel of thickness ca. 150 nm. The device was subsequently placed into a chamber, purged with nitrogen to produce an oxygen-free environment, and was finally cured by UV light for 10 minutes at room temperature. The fabricated hydrogel thin film has been demonstrated to be very robust, as it was effectively studied to show good results even after 30 days of storage.

3. Measurement and simulation of the LSPR

The LSPR of the gold nanocrescent array was measured by normal transmission of focused, unpolarized, near-infrared broadband light (1200nm~2000nm), analyzed by a 0.8m high-resolution spectrograph and InGaAs detector. Unpolarized light was used because of the random orientations of the fabricated nanocrescents. The extinction spectrum Ext(λ) and the transmission spectrum T(λ) are given by:

I(λ) is the transmitted light intensity spectrum through the gold nanocrescent and I_ref(λ) is the reference spectrum.

Before coating the hydrogel, two extinction peaks were acquired from the bare device in the near-infrared spectral range, one located at 1310nm (peak ‘W₀’), the other at 1672nm (peak ‘C₁’), as shown in Fig. 2 . After coating the xerogel (dehydrated hydrogel) on the surface of the device, a change of refractive index around 0.47 is induced. As shown in the dotted line in Fig. 2, the peak C₁ red-shifts to 1828nm, which roughly corresponds to a bulk refractive index sensitivity of 332nm/RIU. This is lower than other gold nanocrescents with similar size [‎7]. We attribute this reduction in sensitivity to the smooth rounded profile of the top edges of the fabricated nanocrescents in our method and the substrate with relatively high index due to ITO. It should be noted that after coating, the peak W₀ also red-shifts by a smaller amount but becomes significantly lower in magnitude and another weak peak ‘W₁’ appears around 1600nm. In addition, the slope of another strong peak ‘U₁’ at shorter wavelength begins to enter the detection range.

 

Fig. 2 Measured and simulated extinction spectra of the fabricated device. The black line represents measurement on the bare device without coating and the red line represents measurement on the device coated with xerogel. The blue line represents simulation of the bare device.

Download Full Size | PDF

To understand these peaks, the LSPR of a single gold nanocrescent is simulated by three-dimensional FDTD simulations incorporating a non-uniform mesh [‎19]. The simulated spectra are plotted together with the measured spectra in Fig. 2. According to the simulations, peak C₁ is excited by the light polarized parallel to the long axis of the nanocrescent (longitudinal resonance). When the light is polarized parallel to the short axis of the nanocrescent (transverse resonance), one strong peak is obtrained at 947nm, which should correspond to peak U₁. However, the existence of the W₀ and W₁ cannot be resolved in FDTD simulations. The origin of these peaks is probably the interactions between the gold nanocrescents. Due to the random orientation of gold nanocrescents, the tips of some neighboring gold nancrescents can be very close with a distance comparable to the field decay length, leading to strong near-field couplings. Such an effect could explain the broad feature of the LSPR peak and the existence of W₀ and W₁ peaks.

4. Sensor response vs. pH

We used the C₁ peak to detect the response of the coated hydrogel thin film. Buffer solutions titrated to various pH levels were introduced onto the device through a fluidic channel built on top of the device. For measurements of each specific pH solution, we waited 15 minutes for the device to reach equilibrium state. We began with pH = 3.08, and progressively increased the pH, which causes the hydrogel to swell.

When the solution is first introduced, the xerogel swells as it absorbs the water, effectively reducing the refractive index of the sensing layer because water has a lower refractive index (n = 1.33) than our hydrogel (n≈1.47). As the pH is gradually increased from 3.08 to 9.08, the hydrogel swells under electrostatic repulsion due to the ionization of carboxylic groups. Swelling occurs when the pH of the swelling medium rises above the pK _a of the moiety, which has been studied to be 4.50 [‎14], causing the hydrogel to become more hydrophilic and absorb more water, which decreases the refractive index.

This decrease in refractive index is transduced into a change in the measured extinction spectrum, as shown in Fig. 3(a) . As the pH increases, the longitudinal extinction peak wavelength λ_max decreases, and the extinction amplitude decreases as well. The latter is equivalent to increasing transmission T(λ). Therefore, the swelling behavior of the hydrogel under increasing pH can be transduced into two characteristic responses, one is λ_max, the other is the transmission T(λ) integrated from 1700nm to 1840nm. This spectral range is selected for optimum signal noise ratio which is strongly affected by near-infrared water absorption. Both responses are plotted in Fig. 3(b) and the details of the trends show that they are very consistent with each other.

 

Fig. 3 Sensor response of the gold nanocrescent array structure coated with hydrogel thin film. (a) Measured extinction spectra of the device for detecting buffered solutions of increasing pH. (b) Response of λ_max (black squares) and the integrated transmission (blue triangles) obtained from the measured extinction spectra. The solid lines are the titration curves fitted into the experimental data. (c) Sensitivity for shift of λ_max (black curve) and sensitivity for transmission (blue curve) vs. pH calculated from the fitted response curves.

Download Full Size | PDF

The two responses were fitted into titration curves using the Henderson-Hasselbalch equation (Eq. (2) for the cross-linked polymers [‎20] [‎21]:

pK _a is the apparent dissociation constant of the cross-linked polymer, n is the correction for the distribution of the MAA groups in the polymer and α is the degree of polymer dissociation which is assumed to be linearly related to the sensor responses. The best fit of the titration curves for both responses are plotted as the solid lines in Fig. 3(b), which yields the fit parameters of pK _a = 5.45, n = 1.25. These values are consistent with previous studies of poly(HEMA-co-MAA) polymers [‎14] [‎21].

The sensitivities calculated from the fitted response curves are plotted in Fig. 3(c). The most sensitive detection range of our hydrogel lies in the pH range centered at the point of pH = pK _a, which is the inflection point of the titration response curves. At pH = pK _a, the sensitivity of the extinction peak shift is 11.1nm/pH unit while the sensitivity of the transmission is 1.16/pH unit. The sensitive range is thus determined to be 4.5~6.4 with the boundaries where the sensitivity decreases by half. As the hydrogel swells in this range, λ_max blue-shifts from 1772nm into 1755nm, and the integrated transmission response increases from 56.8 into 58.6. It should be noted that when pH is increased beyond 8.0, no further significant shift is acquired because all ionizable groups have been ionized.

5. Discussions

Due to the random orientations of gold nanocrescents, the LSPR peak is significantly broadened by near-field effects. The Figure of Merit (FOM, peak-shift per refractive index unit divided by peak width calculated in eV) is around 0.4. Future implementations are expected to provide control for the orientations of gold nanocrescent and the optimum design of the lattice constants leading to an improved FOM. The uncertainty in determining the peak-shift is estimated from the experiment to be better than 0.5nm, which yields detection accuracy around 0.045pH unit. It should be emphasized that, the LSPR spectroscopy can reach a detection resolution ~10⁻⁴ nm range. Combining the hydrogel thin film and LSPR spectroscopy can potentially achieve higher resolution than existing hydrogel sensors. The performance can also be further optimized in terms of sensitivity by modifying the ratios of the constituent monomers [‎14]. It should be mentioned that the sensor shows repeatable results even after 30 days of storage. The chemiresponsive nature of pH-sensitive hydrogels may be exploited for biosensing purposes, such as to monitor the concentration of H⁺ produced in the biocatalytic reaction of organophosphorus compound [‎12].