1. Introduction

Localized surface plasmons (LSPs) as a collective oscillation of free electrons at metal/dielectric interface in metallic sub-wavelength structures have provided a strong platform for highly-resolved sensitivity to the change in the ambient dielectric constant [1,2]. The high sensitivity of LSPR relies on nanoparticle shape, size, periodicity and dielectric constant of the metal which encourages the researchers to optimize these parameters to enhance bulk and surface sensitivities either in red- or blue-shift regimes [2,3]. So far, researchers have mostly focused on increasing the bulk/surface sensitivity of red-shift structures and limited studies have been reported on the physical origin of blue-shift occurrence and their capabilities in providing ultra-sensitivity. Roxworthy et al., [4] attributed the blue-shift to the larger separation of opposing charges which resulted in coulombic screening reduction according to plasmon hybridization model and the coupling between electric and magnetic dipole modes in the system. In fact, the restoring force in one particle was reduced by the positive charge accumulation in the other and subsequently lowered the resonance frequency and led to red-shift. Hence, the increase in nanoparticle distance augmented the restoring force and caused the blue-shift [4]. They have shown that red and blue-shift can occur simultaneously in some structures like split-ring resonator geometries respectively in visible and near infrared spectra due to the coupling of induced magnetic and electric dipoles. Similarly, Lu et al., [5] have observed the blue-shift by increasing the inter-particle distance to weaken the coupling between plasmon of particles. Some groups have studied the role of substrate in blue-shift emergence. Raza et al., [6] have studied the blueshift of the surface plasmon (SP) resonance energy of isolated Ag nanoparticles with decreasing particle diameter on a thin Si₃N₄ substrate. They have shown that in comparison with the free-space environment, the presence of the dielectric substrate should induce a larger blueshift due to modification in the dipole mode [6]. The main role of the substrate on the electron density inhomogeneity and SP resonance frequency originates from the interaction of the dipole mode of the nanoparticle with the induced dipole mode in the substrate which depends on the thickness and the refractive index of the substrate [1]. Due to the symmetry breaking caused by the contrast of the dielectric functions at the particle/substrate interface, metallic nanoparticles interact with their induced images [7]. It is interpreted as the interaction and coupling of the metallic nanoparticles with their induced images on the substrate [7]. This EM coupling between the particle and its image becomes stronger for substrates with higher permittivity and causes substrate-induced Fano resonance [7]. Similarly, Beck et al., [8] have shown that in the vicinity of a high refractive-index substrate, the light was scattered considerably and anonymously over a wide range of angles. The scattering behavior was modified by the polarizability that was dependent on the electric field leading to the resonance. The nanoparticle shape was designed to provide the effective dipole moment induced in the particle close to the interface. The blueshift occurrence can also be interpreted that the induced dipole fields are all out of phase, leading to blue shifts [9]. Jung et al., [10] reported a blue-shift in transmission from gold nanoslit array upon adsorption of a self-assembled monolayer of molecules. They related this blue-shift behavior to the anomalous dispersion of surface plasmons because of the strong negative dispersion on the high-index metal/dielectric interface dominated over the positive dispersion on the metal/air interface. Lee et al., [11] improved the surface sensitivity in capped nanoslit array via decreased plasmonic evanescent field originated from two effects of Fano coupling and blue-shifted resonance. In this study, we proposed an array structure of rotated gold nanoparticle cubes, i.e. rhombus, which benefit from more faces in comparison with a disk-shaped array on thin Si₃N₄ membrane substrate. This precisely designed substrate provides the symmetry breaking due to the contrast of the dielectric functions at the particle/substrate interface. On the other hand, metallic nanoparticles can interact with their induced images and increase the coulombic restoring force required for blue-shift emergence. Undoubtedly, highly-resolved blueshift sensing can open new avenues in real-time and label-free detection applicable in biosensing and medical diagnostics, chemical sensors, food safety and environmental monitoring [2,11].

2. Materials and methods

2.1 Structure design and fabrication

When gold nanoparticles are placed adjacent to the high-refractive-index substrate, the scattering resonances are localized at the interface of gold/substrate and the scattered light is coupled with the SPP modes at the interface. Due to the high scattering cross-section, these structures are attractive for light trapping applications reported to have 7.5 times larger cross-section in comparison with a dipole-like scatterer with an equivalent free space resonance [8]. Noguez [12] has mentioned the relationship between the shape and the faces of the nanoparticle. Based on these facts, we proposed a rhombus-shaped plasmonic crystal on a Si₃N₄ membrane, which benefits from multi-face structure besides high refractive-index substrate providing blue-shift sensing. For the structure design, various configurations of the triangle structure including separated/connected back-to-back triangles, overlapped triangles and separated face-to-face triangles were investigated and finally, our proposed structure was a rhombus structure that was designed by contacting two face-to-face triangles as shown in Fig. 1(a). Other simulated structures had multiple SPRs or their fabrication was not cost-efficient which was not the goal of our design.

 

Fig. 1. (a) The evolution of the structures before reaching the optimized rhombus array. Fabrication of a rhombus-shape plasmonic crystal on the Si₃N₄ membrane: (b) The process flow of the fabrication of Si₃N₄ membrane from double Si₃N₄-coated wafer using photolithography, dry and wet etches. (c) The process flow of the fabrication of rhombus-shaped plasmonic crystal on this Si₃N₄ membrane. (d) Structural parameters of the fabricated structure. (e) Real-photo of the fabricated chip and the SEM of the chip showing the average diagonal of 125 nm for the gold nanoparticle.

Download Full Size | PDF

Due to the high resolution of electron beam lithography (EBL), we used this fabrication technique to pattern rhombus shape plasmonic crystal on the Si₃N₄ membrane [13]. In order to fabricate the proposed structure, 100∼500 nm Si₃N₄ was deposited on the double top and bottom surfaces of the 500 µm thick silicon wafer using low pressure chemical vapor deposition (LPCVD) followed by RCA clean at room temperature. With AZ1512 photoresist in Ritetrack 88 coater and developer (center of micronanotechnology, EPFL), and the mask with a critical dimension of 800 µm, the pattern was created using photolithography (Süss MA6). Afterwards, using dielectric etcher machine of SPTS APS, a Si₃N₄ etch of 100-500 nm was performed following by the resist strip. Then KOH in Plade Six Sigma machine (center of micronanotechnology, EPFL) was utilized for 500 µ wet etching of Si following the resist stripping. The process flow of membrane chip fabrication is shown schematically in Figs. 1(b) and 1(c). After membrane chip preparation and second RCA clean of the chip at room temperature, MMA/PMMA was spin coated followed by hot plate baking. Then 5-10 nm gold was sputtered using a sputtering machine of Alliance-Concept DP 650 (center of micronanotechnology, EPFL) and electron beam lithography (EBL) was carried out. After wet etching of the gold, the chip was developed with MIBK:IPA. Using Evaporator of Leybold Optics LAB 600H (center of micronanotechnology, EPFL), 5 nm Cr and 100 nm gold were evaporated following the lift-off with acetone and IPA wash and O₂ plasma cleaning.

The structural parameters are shown in Fig. 1(d). Real photo and the scanning electron microscopy (SEM) image of the fabricated chip are shown in Fig. 1(e). Using the Finite Difference Time Domain (FDTD) solver in FDTD solutions pack of Lumerical software, the design parameters were optimized as a square lattice of 45° rotated gold cubes around z-axis with the dimensions of 100×100×30 nm³ with the periodicity of 300 nm on Silicon nitride (Si3N4) substrate. FDTD region of 300×300×700 nm³, mesh region of 160×160×50 nm³ with a fine meshing of 2 nm, source size of 450×450×200 nm³, monitor size of 600×600×250 nm³ and frequency points of 800. For verifying the simulation, the same structure was simulated in CST Microwave Studio software with the thicknesses of substrate and the buffer layer set as 200 nm and gold mesh size of 5 nm. The substrate had the permittivity of 4.4 (n ≈ 2.1) that hardly varied for the frequency range under our consideration [1], however for precision we obtained its wavelength-dependent refractive index from [14]. One of the properties of blue-shift structures is the appearance of Fano resonance in their reflections which is called substrate-induced Fano resonances [7]. In Figs. 2(a) and 2(b), we simulated the reflection response of the proposed structure under the sweep of the incident angle for dry (air) and wet (water) cases. As seen in these figures, by increasing the incident angle from 30° to 60°, the Fano resonance becomes more conspicuous.

 

Fig. 2. (a)-(b) The simulated reflection spectrum of the proposed structure in dry and wet cases using FDTD solutions, Lumerical. By increasing the incident angle from 30° to 60°, the Fano resonance becomes more conspicuous in both dry and wet cases. (c) The schematic of the inverted microscope used for recording the normal transmission responses. (d) The sensitivity of the fabricated structure for water (n=1.33297) and the different concentrations of sucrose solutions with refractive indices of n=1.33721 (50 mg/1.5 ml total vol.), n=1.33938 (75 mg/1.5 ml total vol.) and n=1.34210 (100 mg/1.5 ml total vol.). The quantitative data of the resonance wavelengths are mentioned in Table 1. By increasing the concentration of the sucrose, the restoring columbic force increased that was required for the blue-shift emergence.

Download Full Size | PDF

Table 1. Plasmonic resonance wavelength for patterned Si₃N₄ membrane. There were four rows of membranes, and two membranes were patterned in each row. The columns were named “a” and “b”.

View Table

3. Results and discussion

The scattering spectra of single Au NPs were taken using a Nikon Ti-u microscope under normal incidence transmission regime, where the light from a halogen lamp was focused on the Au NPs through a dark-field condenser (NA 0.85). The scattered light by the nanoparticles was collected with a 100× NA 0.8 objective (variable NA 0.8−1.3) and directed to the entrance slit of a spectrograph (IsoPlane SCT 320, Princeton Instruments) equipped with a CCD camera (PIXIS 1024BR, Princeton Instruments). Scattering spectra were corrected by signal collected from a nearby region without nanoparticles and normalized by the lamp intensity profile. The schematic of the setup is shown in Fig. 2(c). We have chosen air, DI water (n₂=1.33297) and several sucrose solutions with refractive indices of n₃=1.33721 (50 mg/1.5 ml total vol.), n₄=1.33938 (75 mg/1.5 ml total vol.) and n₅=1.34210 (100 mg/1.5 ml total vol.). The refractive index differences between water and these sucrose solutions are 0.004, 0.006 and 0.009, respectively. The related bulk sensitivity of the fabricated structure is shown in Fig. 2(d). Considering the error bar and its linear fit, we had the blue-shift bulk sensitivity of 618.82 nm/RIU. The related quantitative data of the resonance wavelengths are brought in Table 1. In Fig. 3, the transmission spectra of the patterned rhombus plasmonic crystal normalized to non-patterned Si₃N₄ substrate for water and various concentrations of sucrose are demonstrated. There were four rows of membranes and two membranes were patterned in each row. In sum, we have eight repetitions of the results that are shown in Figs. 3(a)–3(h). By increasing the concentration of the sucrose, more particle trapping occurred and restoring columbic force increased that was required for blue-shift emergence. These experimental results had good match with simulation as shown in Fig. 4(a). Figure 4(b) shows the enlarged image of the resonance location for different sucrose concentrations in simulation results. Figure 4(c) shows the simulated unit cell of the rhombus structure. Subsequently Fig. 4(d) shows the corresponding experimental results.

 

Fig. 3. (a)-(h) The transmission spectrum of the patterned rhombus plasmonic crystal normalized to non-patterned Si₃N₄ substrate for water and various concentrations of sucrose. There were four rows of membranes and two membranes were patterned in each row. The columns were named “a” and “b”.

Download Full Size | PDF

 

Fig. 4. (a) Simulated transmission spectrum for various sucrose concentrations (b) enlarged image of A showing the splitting of resonance wavelengths (c) the simulated unit cell of the rhombus structure (d) the corresponding experimental results to the simulated structure.

Download Full Size | PDF

In order to investigate the applicability of this structure for lower refractive index changes, we have investigated the analytes of fresh and primary cell secreted culturing medium (i.e. conditioned medium) with refractive indices of 1.33541 and 1.33556, respectively, where the difference in their refractive index was 0.00015. For fresh medium and secreted medium with an order of 0.00015 difference in their refractive indices, we had a blue-shift of 1 nm which gave us a large surface sensitivity. In Fig. 5, the transmission spectra are shown for fresh and conditioned primary cell medium where we had eight repetitions.

 

Fig. 5. (a)-(h) The transmission spectrum of the patterned rhombus plasmonic crystal normalized to non-patterned Si₃N₄ substrate at the presence of fresh and conditioned primary cell medium. There were four rows of membranes and in each row, two membranes were patterned. The columns are called “a” and “b”. The analytes of fresh and primary cell secreted culturing medium (i.e. conditioned medium) with refractive indices of 1.33541 and 1.33556 had the difference in their refractive index around 0.00015. For fresh medium and secreted medium with this low level of difference in their refractive indices, we generally had a blue-shift of 1 nm which gave us a large surface sensitivity.

Download Full Size | PDF

This blueshift behavior can be interpreted using the characteristics of the designed structure. Our proposed structure was an array of rotated gold nanoparticle cubes, i.e. rhombus on thin Si₃N₄ membrane substrate with high permittivity of 4.4 (n ≈ 2.1). Under an external electromagnetic field, a charge polarization is induced on the particle that subsequently, gives rise to an image charge distribution on the substrate. When the size of the nanoparticle is large (>40 nm), the radiation effect becomes more significant. In this case, the displacement in the electronic cloud is not homogenous and large multipolar charge distribution is induced. Using the dielectric substrate with high permittivity (i.e. Si₃N₄), there is a strong interaction of the dipole (multipolar) mode of the nanoparticle with the dipole (multipolar) mode in the substrate that increases the coulombic restoring force required for blue-shift emergence. When gold nanoparticles are placed adjacent to the high-refractive-index substrate, the scattering resonances are localized at the interface of gold/substrate and the scattered light is coupled with the SPP modes at the interface. This strong interaction causes a substrate-induces Fano resonance as seen in Fig. 2(a). In addition, there is a symmetry breaking due to the contrast of the dielectric functions at the particle/substrate interface.

One main reason for the occurrence of the blueshift is that the induced dipoles (multipoles) are all out of phase. According to [12], as the number of nanoparticle faces increases, the main resonance is subjected to a blues-shift due to the augmentation in coulombic restoring force. On the other hand, by truncation of sphere nanoparticle into a multi-face structure, secondary SPRs may overlap and increase the full width at half maximum (FWHM). Therefore, the blueshift in optical responses are observed in Figs. 3 and 5 with FWHM larger than 100 nm. Regarding these experimental results, the bulk and surface blue-shift sensitivities of the structure were 618.82 nm/RIU and 666.66 nm/RIU. FWHM for the bulk sensitivity ranged from 134.48 nm to 169.23 nm with the standard error ranged from 1.05 to 2.32 nm. FWHM for the surface sensitivity ranged from 153.26 nm to 159.62 nm with the standard error ranged from 1.65 to 2.00. Figure of merit (FOM) of the device can be calculated with the refractive index sensitivity divided by the FWHM of the resonant dip. Therefore, the best FOMs were 4.60 and 43.50 for the bulk and surface sensitivity, respectively.

4. Conclusion

In conclusion, we have fabricated an ultrasensitive multifaced rhombus structure for enhancing the blue-shift sensitivity (∼619 RIU). This behavior was not only confirmed by different sucrose solutions but also confirmed for biological media with very low refractive index difference; normal and conditioned primary cell medium. For fresh medium and secreted medium with 0.00015 difference in their refractive indices, we have a blue-shift of 1 nm which gives us a large sensitivity. We see a blue-shift for surface sensitivity which is proved by simulations and experiments. The existence of the blueshift criterion was proved by the emergence of Fano resonance in the reflection regime by increasing the incident angle. In other words, for normal incidence, we observed the Gaussian shape while for oblique angles we had the Fano resonances which were red-shifted by increasing the incident angle. As it is seen, by increasing the incident angle the Fano resonance becomes more conspicuous.