Surface plasmon resonance (SPR) sensing is a key technique for real-time and label-free detection of molecular interactions on a gold surface.1–3 The attenuated total internal reflection (ATR), known as the Kretschmann configuration, is most commonly utilized to excite the SPR on a gold surface. In addition to the prism coupling method, metallic nanostructures offer a simpler way for SPR excitation.4–7 Recently, periodic gold nanohole arrays or nanoslit arrays have been utilized for biosensing applications.8–19 By measuring changes in the resonant angle, wavelength or intensity, the amounts of surface binding events are quantitatively estimated. The detection limit of the SPR sensor is usually characterized by minimum refractive index unit (RIU). It is defined by δ/S, where S is the RIU sensitivity and δ, the resolution of the measurement system. Increasing S or lowering δ can enhance the SPR detection. The common detection limit of a nanostructure-based sensor is on the order of 10−5 RIU for 0.1 nm wavelength resolution,9, 12 10−6 RIU for 10−4 degree angle resolution,9 or 10−5 RIU for 0.2% intensity resolution.18, 19
Different from the conventional analysis methods, such as wavelength interrogation, intensity measurement and angular interrogation, a spectral integration method has been proposed and applied in quasi-3D plasmonic crystals and various 2D gold nanostructures to increase the detection limit.17, 20, 21 The periodic gold nanostructure exhibits complex resonant behaviors in the transmission spectrum due to the coupling of localized plasmon resonance (LSPR) in the subwavelength apertures,22, 23 and Bloch wave surface plasmon polariton (BW-SPP)24 on the outside surfaces. By summing up all the transmission changes in the spectrum, the signal-to-noise ratio (SNR) of the system is improved and the sensing capability can be enhanced. In the analysis method, the integrated response (R) is expressed as follows:
where R is the integrated response, T(n,λ) is the transmission spectrum under different external refractive index (n), n0 is the referenced refractive index (n0 = 1.3330 for pure water), λ1 and λ2 are the integrated wavelength range and Δλ is the resolution of the spectrometer. The spectral integration method not only has the BW-SPP mode but has other resonances taken into the measurement. The more SPR-related signals can improve the SNR and thus improve the detection limit. Recently, polarization-dependent sensing using biaxial nanohole arrays was proposed. For a particular wavelength, one linear polarization shows an increased intensity, while the other shows a decreased intensity, which allows separation of SPR sensing resonance effects from spurious polarization-independent changes in the intensity25 and improves the sensing capability.26 In this paper, we proposed a multi-polarization spectral integrated method to increase the SNR by considering the all SPR-related signals at different polarization states. If the nanostructures were made with different SPR wavelengths in different polarization directions, the introduction of polarization in the spectral integration method can efficiently increase the SNR and reduce the detection limit. The integrated response (Rp) is defined by the absolute differences of the normalized transmission spectra.
Where T(n,λ,θ) is the transmission spectrum under various external refractive index (n) and polarization angle (θ) of incident light, n0 is the reference refractive index, λ1 and λ2 are the integrated wavelength range, θ1 and θ2 are the polarization angle range, respectively. In this study, we fabricated dual-period nanogrid arrays and experimentally compared the sensing capabilities using intensity measurement, single-polarization and multi-polarization spectral integration methods. For a dual-period nanogrid structure, the proposed multi-polarization method increased SNR and refractive index detection limit about 8 times larger than intenisty method. The SNR is greatly improved because the nanostructure has SPR wavelengths at different polarization states. By using the multi-polarization integration method, the nanogrid structures can achieve a detection limit of 1.92 × 10−6 refractive index unit when the stability is 0.2%. It is comparable with the bulky ATR sensors using complicated angular detection method. An antigen-antibody interaction experiment in aqueous environment was conducted to verify the detection improvement in surface binding event.
Figure 1a shows the measured transmission spectra of dual-period nanogrid arrays in water for various polarization angles of normally-incident light. The periods of nanogrid arrays were 600 and 590 nm in x-axis direction and in y-axis direction, respectively. The polarization angle of the incident light versus the structures is depicted in the inset. There are some resonant peaks in the spectra. These resonances are Fano-type resonances formed by the interference between spectrally overlapping broad resonance and a narrow discrete resonance.28–30 The LSPR occurring within the nanoslit plays a role as the broad resonant state. On the other hand, the discrete state is associated with BW-SPP on the periodic slits. The BW-SPP occurs when the Bragg condition is satisfied. For a normally incident light, the condition for a 1D array is described as
Where i is the resonant orders, P is the period of the nanostructure, ϵm is the dielectric constant of the metal and n is the external refractive index. The interaction between direct slit transmission (a continuum state) and BW-SPPs (a discrete state) creates a Fano-like resonance profile consisting of a minimum, close to the position predicted by equation 3, and an adjacent maximum. For a TM polarized wave (θ = 90°), a Fano-like resonance in the spectrum was observed. The profile consisted of a peak and dip resonances at 782 and 810 nm, respectively. When the polarization angle of the incident light was gradually increased to 40°, the superposition of all the resonances from the x component (600-nm-period nanoslits) and y component (595-nm-period nanoslits) in the spectrum was observed. The corresponding wavelength of the extra Fano resonance for the profile were 808 and 819 nm. When the polarization angle 0° was chosen, a Fano-like resonance from the x component in the spectrum were observed. The corresponding wavelengths for the peak and dip resonances were 807 and 823 nm. From equation 3, the BW-SPP resonant wavelength of the x component at the water/gold interface is 825 nm (ϵm = -29.1 + 1.56i for gold at 825 nm, i = 1, n = 1.3330 and P = 600 nm). For the y component, the resonant wavelength of the BW-SPP are 813 nm at the water/gold interface (ϵm = -27.9 + 1.50i for gold at 813 nm, i = 1, n = 1.3330 and P = 590 nm). Obviously, the experimental wavelengths were close to the theoretical wavelengths. Figure 1b shows the transmission spectra of nanogrid arrays in water and various water/glycerin mixtures for various polarized light. For clarity, only spectra with polarization angles of 0°, 20°, 40° and 90° were shown. When the concentrations of glycerin increased, the spectra near the SPR wavelengths were red-shifted. The intensity has a large change near the resonant wavelength. By utilizing equation 1 and choosing an integrated wavelength range from 600 to 850 nm, the integrated responses (R) as a function of refractive index for various single polarization angles, from 0° to 90°, are shown in Figure 1c. In Figure 1c the transmission spectrum in water (n0) for each polarization angle was set as a reference. The average noise for different polarization angles extracted from Figure 1c were around 0.49–0.67 nm. Figure 1d shows the linear correlations between the responses and refractive indexes for each single polarization angle. From these correlations, we obtained the RIU sensitivities for different polarization angles (see Figure 1f). The sensitivities for polarization angles of 0°, 10°, 20°, 30°, 40°, 50°, 60°, 70°, 80° and 90° were 7762, 7686, 7939, 8494, 9188, 9939, 10943, 12557, 14315 and 15051 nm/RIU, respectively. It is obvious that the sensitivity was improved when the polarization angle was increased from 0° to 90°. We deduced the higher sensitivity for the polarization angle of 90° is due to the narrower slit width for 590-nm-period slit arrays. The narrower slit width has a sharp bandwidth, which results in a large intensity change when the refractive index of the environment is changed. The multispectral analysis integrates all the absolute intensity changes over the wavelength. Therefore, the slit array with a narrower slit width has a higher integrated sensitivity.
We further analyzed the measured results using the multi-polarization spectral integration method. By utilizing equation 2 and choosing an integrated wavelength range from 600 to 850 nm and an integrated angle range (the number of the integrated polarization angles), the integrated responses (Rp) as a function of refractive index for various numbers of the integrated polarization angles are shown in Figure 2a. The chosen numbers of integrated angles were 1 (90°), 2 (80° and 90°), 3 (70°, 80° and 90°), 4 (60°, 70°, 80° and 90°), 5 (50°, 60°, 70°, 80° and 90°), 6 (40°, 50°, 60°, 70°, 80° and 90°), 7 (30°, 40°, 50°, 60°, 70°, 80° and 90°), 8 (20°, 30°, 40°, 50°, 60°, 70°, 80° and 90°), 9 (10°, 20°, 30°, 40°, 50°, 60°, 70°, 80° and 90°), 10 (0°, 10°, 20°, 30°, 40°, 50°, 60°, 70°, 80° and 90°). We also integrated the spectra at two orthogonal polarizations (0° + 90°) to check whether taking the data over multiple polarizations is better than taking the data at two orthogonal polarizations. Figure 2a shows that the noise floor decreased with the increase of the number of the integrated angle. Figure 2b shows the linear correlations between the responses (Rp) and refractive indexes for various numbers of the integrated angles. Increasing the numbers of polarization angles resulted in a decrease of response. Figure 2c shows the noise and RIU sensitivity as a function of the number of the integrated angle. Compared to the sensitivity for polarization angle of 90°, the sensitivity for the multi-polarized spectrum method was lower. However, if we considered the noise level, the noise is dramatically reduced. The noises were 0.596, 0.455, 0.374, 0.279, 0.231, 0.221, 0.2, 0.186, 0.164, 0.154 and 0.351 nm for the numbers of integrated angles of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 and 2 (0° + 90°), respectively. The noise floor for the integrated polarization from 0° to 90° was 3.9 times lower than that without polarization integration and 2.3 times lower than that for two orthogonal polarizations (0° + 90°). On the other hand, the RIU sensitivities were 15051, 14672, 13852, 12971, 12240, 11684, 11215, 10880, 10614, 10434 and 11402 nm/RIU. The sensitivity decreased when the integrated number increased. The RIU sensitivity for integrated number of 10 only decreased ∼1.4 times than that of integrated number of 1. We further compared the SNR performance and detection limits of various integrated numbers. Figure 2d shows the average SNR for various fractions of glycerin and detection limit as a function of the number of the integrated angle. By utilizing the proposed method, the SNR is increased about 3 times larger than the single-polarization spectral integration method (i.e. the number of integrated angle is 1) and ∼2.3 times than that of integrated number of 2 (0° + 90°). Therefore, the detection limit for the integrated polarization from 0° to 90° is better than that for two orthogonal polarizations and improved about 3 times than that of previous spectral integration method.
We further compared the detection limit performance of nanogrid structures using the conventional intensity method and the proposed multi-polarization spectral integration method. Figure 3a shows the absolute intensity changes over the full spectra for TM-polarized light (θ = 90°), where the transmission spectrum in water (n0) was set as a reference. The intensity has a large change near the resonant wavelength. It increases with the increase of the surrounding refractive index. The most significant change in transmission was at a wavelength of 791 nm, close to the wavelength of BW-SPPs at the water/gold interface. Figure 3b shows the intensity change as a function of different index mixtures at this wavelength. The correlation between the intensity change and refractive index change is shown in the inset. It indicates that the intensity sensitivity was 12963%/RIU. The intensity noise was 0.85%. Hence, the RIU resolution was 6.55 × 10−5 RIU for the intensity method. On the other hand, for the same nanostructures and same measurement setup, the sensitivities and noises were greatly improved by the multi-polarization spectral integration method as seen in Figure 2d. The sensitivity and noise obtained using equation 2 were 10434 nm/RIU and 0.154 nm, respectively. It indicates that the detectable RIU was 1.47 × 10−5, better than that of single-wavelength intensity measurement.
It was noted that equation 2 using the linear deviation for the integration signal. It weights most strongly the points of the spectrum with the lowest signal and would reduce the SNR value. M. Das et al. had proposed a more appropriate formulation for sensing application by using root mean square (RMS) deviation.21 The concept can also be applied to the multi-polarization spectral integrated method to increase the SNR. Here, the integrated response (RPRMS) is rewritten as follows:
Where is the average of the difference signal over the entire spectrum. We analyzed the measured results utilizing equation 4 and chose the same conditions for the previous analysis method. Figure 3c shows the integrated responses (RPRMS) as a function of refractive index for various numbers of the integrated polarization angles. Compared to the linear deviation method (Figure 2a), the RMS method had a lower noise floor. Figure 3d shows the average SNR for various fractions of glycerin and the detection limit as a function of the number of the integrated angle. The SNR acquired using RMS deviation is higher than the SNR using equation 2 (Figure 2d). It is improved 4.7 times than that for linear deviation. The detection limit using RMS method is 8.13 × 10−6 RIU. It is 8 times lower than that of single-wavelength intensity measurement and 1.8 times lower than that using linear deviation. If the intensity stability can be improved to 0.2% by using an intensity-stabilized light source and low-noise spectrometer, the detection limit for a single-wavelength intensity measurement is 1.54 × 10−5 RIU. Meanwhile, the detection limit using the multi-polarization spectral integration method will reach 1.92 × 10−6 RIU. This value is better than that of biaxial nanohole arrays (6.4 × 10−6)26 and close to the best detection limit of a commercial ATR-based sensor, 10−7 RIU in intensity (angular resolved) measurements. The trade-off of the detection improvement is the decrease of bandwidth. A data point needs a scan of 0° to 90° polarization angles. In our experiments, it took about 10 seconds.
An antigen-antibody interaction experiment in aqueous environment was condcuted to verify the detection improvement in surface binding event. The interactions between bovine serum albumin (BSA) and anti-BSA were measured using dual-period nanogrid arrays. In the experiment, the buffer solution, 10 mM phosphate-buffered saline (PBS), was first injected to the fluidic channel to clean the gold surface. The transmission spectra of the nanostructure in buffer solution for various polarization angles, from 0° to 90° with 10 steps, were measured. Then 500 μg/mL BSA was injected on the structure surface. Due to the physical adsorption of BSA on gold surface, the BSA will be coated on the structure surface. The BSA was flowed for two hours to make sufficient BSA immobilized on the gold surface. Next, the sample was washed by PBS buffer to remove the unbound BSA proteins. We then measured the time-lapsed transmission spectra with different polarization angles after injecting 375 μg/mL anti-BSA into the sample. After four hours protein-protein interactions, the unbound anti-BSA was washed away by the PBS buffer. For clarity, only the measured spectra with polarization angles of 0° and 90° were shown in Figure 4a and Figure 4b. Significant changes in wavelength shift and transmission were observed in the spectra when BSA and anti-BSA were bound on the gold surface. In Figure 4a, the BSA on a gold surface made a 0.39-nm red shifts. The shift was increased to 3.54 nm when anti-BSA was bound to BSA. The 150-kDa-sized anti-BSA resulted in a 3.15-nm wavelength shift. We further compared the detection limits for anti-BSA by using the intensity measurement and the multi-polarization spectral integration method. Figure 4c shows the signal changes for the BSA/anti-BSA interaction as a function of time. The wavelengths for the intensity measurement were 818 and 794 nm for polarization angles of 0° and 90°, respectively. The wavelength range for the multi-polarization spectral integration method was from 600 to 850 nm. It is obvious that all the signals increased with time and then saturated. For the single-wavelength intensity measurement, the mean signal change was 86.5% and mean noise was 2.72%. The detection limit for anti-BSA is 13.06 μg/mL. On the other hand, the signal change was 57.39 nm and the noise was 0.562 nm for the multi-polarization spectral integration method. The detection limit is 3.67 μg/mL. Compared to the simple intensity measurement, the detection limit is enhanced by 3.5 times.
In Summary, a multi-polarization spectral integration method to increase the refractive index detection limit of gold nanostructure-based SPR sensor was proposed. The noise level is reduced or the sensitivity is increased by incorporating more information contained in the spectra with various polarized incident light. We experimentally compared the sensing capabilities of gold nanogrid arrays using the intensity measurement, spectral integration approach and the proposed method. For a dual-period nanogrid structure, the proposed multi-polarization method increased signal-to-noise ratio and refractive index detection limit about 8 times larger than the simple intensity method. We attributed the low detection limit of dual-period nanogrid structures to the increase of SPR peaks over the integration wavelength. If the nanostructures can be made with more SPR wavelengths in different polarization directions, the introduction of polarization in the spectral integration method can increase the SNR and reduce the detection limit more efficiently. Currently, the dual-period nanogrids can reach a detection limit of 1.92 × 10−6 RIU if the intensity stability is 0.2%. Such detection limit value by the proposed simple integration method would be comparable with the bulky ATR sensors using complicated angular detection method.