1. Introduction

Surface plasmon polariton (SPP) is an oscillation of free electrons inside a metal surface. When the frequency of the SPP is coincident with the external electromagnetic waves, there is a strong absorption of the electromagnetic waves. This resonance condition is very sensitive to the refractive index change of the surrounding material of the metal film. These properties of SPP can be used as a sensor for surface chemistry [1]. For a few decades surface plasmon resonance (SPR) sensors have been important in detecting the interaction of biomolecules. SPR sensors do not need any optical, electrical, or radioactive labels for detection. In proteomics, label-free detection is a very attractive characteristic and SPR sensors are major candidates for high-throughput screening solutions in the field of proteomics [2].

Conventional SPR sensors use the attenuation total internal refraction (ATIR) method to excite the SPP. The Kretschmann configuration is usually adopted. This angle incident configuration prohibits these SPR sensors from being used to measure large sample arrays. In an SPR imaging system the sensitivity increases with longer wavelength, but surface plasmon propagation length is proportional to the wavelength, which sets the limit for the unit cell size in an array [3]. In the Kretschmann configuration angle dependence of sensitivity is so great that a portable application is difficult to achieve.

Nobel metals such as gold or silver show extraordinary optical characteristics when the size is order of subwavelength (or sub 100 nanometer size) [4]. Optical absorption and scattering characteristics of nanosize metal varies with the size, shape, array type and the dielectric constant of surrounding materials [5]. This phenomenon is called localized surface plasmon resonance (LSPR). Metallic nanoparticles have been used as labels to detect biochemical reactions in colloidal status using scattering spectrum analysis methods [6]. A Northwestern University group developed a nano-sphere lithography (NSL) method for mass production of nano-triangular silver on a substrate. This NSL method produces sensors with a limited uniform area of a few tens of micrometers due to defects during the self-assembling process. LSPR sensors with a simple configuration can be made at low cost compared to conventional SPR sensors. Such a simple optical configuration is also attractive for portable applications.

Absorption or scattering spectrum analysis has been used for nanoparticle based LSPR sensors. In LSPR sensors, the resonance wavelength shift due to a change in the refractive index of surrounding materials is smaller than that of conventional SPR sensors. Additionally, full-width at half-maximum (FWHM) values of resonance peaks is broader which leads to sensitivity degradation. However, 1-D or 2-D arrays of silver nanoparticles are expected to have a narrow resonance peak [7]. When the target samples are turbid or opaque, the absorption or scattering spectrum can be affected by the samples themselves.

In this study, we propose a new approach to LSPR sensors that would be easily implemented into a portable system with high sensitivity. In this approach a nanosize grating is incorporated into the surface propagating type SPR sensors. Nano-grating surface plasmon (NGSPR) sensors use vertical illumination to excite localized surface plasmon polarition (LSPP). The use of nano-grating can enhance the refractive index sensitivity of LSPR sensors by large resonance wavelength shifts with the surrounding materials and sharp reflection resonance peaks. The influence of structural factors of nano-gratings on the sensitivity of NGSPR sensors was investigated using the rigorous coupled waveguide analysis (RCWA) method. The RCWA method is widely used for the calculation of optical properties of nanostructures [8]. The purpose of this computational study is to predict the optimal design of such a device.

2. Sensor structure and simulation parameters

A schematic diagram of the proposed NGSPR sensor is shown in Fig. 1. The sensor chip is composed of a thin metal film on a transparent substrate with nano-grating on it. Nano-grating can be made by either etching the substrate or molding the substrate. The structural parameters of the NGSPR sensors are shown in Fig. 1. Where Λ is the period of grating, u is the width of the upper region of the grating, d is the depth of the grating, and t is the depth of metal layer.

The transparent substrate is assumed to be SF10 glass and optical properties were obtained from manufactures. A thin gold film is deposited on top of the nano-grating carved substrate. Optical properties of the gold at the incident wavelength were taken from [9]. Sample materials on the thin gold film were assumed to have a refractive index range from 1.2 to 1.5. Biochemical reactions take place in the liquid solutions with typical refractive index values in this range. A TM-polarized light source is assumed. The range of wavelengths in the simulation was set from 500 nm to 850 nm. Below 500 nm the gold film absorbs the incident light and reflectance is very low. Above 850 nm silicon photodiodes or imaging devices such as CCD or CMOS sensors have little sensitivity. The reflection spectra were obtained using the RCWA method with normal incidence of light from the substrate side.

 

Fig. 1. Schematic diagram of a nano-grating based reflection type SPR sensor

Download Full Size | PDF

3. Results and discussion

We investigated the effects of each parameters on the shape and wavelength resonance reflectance peak. First grating period, was varied from 400 to 600 nm. Figure 2 shows the reflection spectrum of NGSPR with Λ=400 nm, u=50 nm, d=40 nm, and t=40 nm. There are 2 peaks, one in a shorter wavelength, and the other in a longer wavelength. The resonance wavelength of a metallic nanoparticle array is related to the period of the array and refractive index of surrounding materials. A simple relation exists that the resonance wavelength is proportional to the period of the array times the refractive index of the surrounding materials [10]. There are two surrounding materials in this sensor; one is the substrate, and the other is the sample to be studied. The SF10 glass used as a substrate has a refractive index of 1.72 around the visible wavelength range.

For a sensor with a period of 400 nm, the expected resonance wavelength due to the substrate-gold interaction is about 680 nm. Peak 2 in Fig. 2 is 710 nm, which is almost the same wavelength expected value from the simple estimation. Peak 2 is not influenced by the change of the refractive index of sample materials on the gold film, and can be identified as the substrate peak. Peak 1 is dependent on the change of refractive index of sample materials on the gold film, and can be identified as the sample peak. Due to the absorption of light by the gold film, the sample peak is not sharp enough to be distinguishable below 550 nm.

 

Fig. 2. Calculated reflectance spectra of a NGSPR sensor in various refractive index conditions (Λ=400 nm, u=50 nm, d=40 nm, t=40 nm)

Download Full Size | PDF

In Fig. 3, the calculated reflectance spectra of the NGSPR sensor with a grating period of 500 nm are shown. The substrate peaks shifted into more than 850 nm. The sample peaks are in the simulated wavelength range. FWHM of the resonance peak is less than 10 nm and the refractive index sensitivity is about 440 nm/RIU. Calculated reflectance spectra of a NGSPR sensor with a period of 600 nm are shown in Fig. 4. The refractive index sensitivity is more than 500 nm/RIU but the FWHM is wider than that of the 500 nm period sensor and the resonance dips are shallower. Previously published values of refractive index sensitivities are 400 nm/RIU for gold nanohole arrays [11], 328.5 nm/RIU for gold nanoshells [12], and 150 nm/RIU for NSL fabricated LSPR sensors [13].

A “figure of merit” (FOM) was defined in [14] to compare the overall performance of optical sensors as

where m is the slope of wavelength over the refractive index. Overall the refractive index sensitivity of optical sensors is affected by the FWHM. A wide FWHM makes it difficult to distinguish small changes in the peak value. The deeper and narrower the resonance reflection peak is, the more accurate the determination of minimum reflection wavelength. NGSPR sensors have a very narrow resonance reflection peak with FWHM of 0.03 eV. The calculated FOM value is about 60 for a grating period of 500 nm. Usual nanoparticle based refractive index sensors have a FOM value of less than 10 due to a wide FWHM of absorption or scattering spectrum.

 

Fig. 3. Calculated reflectance spectra of a NGSPR sensor in various refractive index conditions (Λ=500 nm, u=50 nm, d=40 nm, t=40 nm)

Download Full Size | PDF

 

Fig. 4. Calculated reflectance spectra of a NGSPR sensor in various refractive index conditions (Λ=600 nm, u=50 nm, d=40 nm, t=40 nm)

Download Full Size | PDF

The effects of grating feature size on reflectance spectra of NGSPR sensors with a grating period of 500 nm are shown in Fig. 5. There are sharp resonance peaks when u, the width of the upper region of the grating, is 25 nm or 50 nm. When u is larger than 50 nm, the resonance peaks are broad. We also found that the NGSPR sensors with grating period of 500 nm have similar sharp resonance reflection peaks when u is wider than 450 nm. That is, there should be a region whose feature size is less than 50 nm to have a sharp resonance reflection peak. Some conventional grating coupled SPR (GC-SPR) sensors have similar structure with the NGSPR sensors. In GC-SPR sensors, u is usually 50 % of the grating period, which means u is more than a hundred nanometers. A conventional GC-SPR sensor doesn’t have sharp reflection peak when light is incident upon it vertically.

 

Fig. 5. Calculated reflectance spectra of NGSPR sensors (Λ=500 nm, d=40 nm, t=40 nm) with a feature size of 25 nm (a), 50 nm (b), 100 nm (c), and 250 nm (d)

Download Full Size | PDF

 

Fig. 6. Calculated reflectance spectra of NGSPR sensors with various profiles and widths (Λ=500 nm, d=50 nm, t=50 nm, sample material is water)

Download Full Size | PDF

Grating geometry effects were simulated with different u values. As shown in Fig. 6 when u=70 nm, grating geometry has a great effect on the resonance peak shape. No sharp resonance reflection peak can be found when the feature geometry is rectangle or rounded. In the case of u=50 nm, the peaks are always sharp, though the peak wavelength is slightly different. These simulation results indicate that a grating feature size of less than 50 nm should result in a sharp resonance reflection peak for any grating geometry. Geometrical deformations during the fabrication process will not degrade the sensitivity of the sensors if the minimum feature size of the grating is less than 50 nm.

To investigate the impact of grating depth, this parameter was varied from 30 nm to 60 nm, while the thickness of metal layer was set to 50 nm, period to 500 nm, and upper layer width to 50 nm. When the grating depth is shallow, the resonance peak is not deep [Figs. 7(a), 7(b)]. The reflection spectra are almost identical when the depth of the grating is equal to or greater than the thickness of the metal layer [Figs. 7(c), 7(d)].

 

Fig. 7. Calculated reflectance spectra of NGSPR sensors (Λ=500 nm, u=50 nm, t=50 nm) with a grating depth of 30 nm (a), 40 nm (b), 50 nm (c), and 60 nm (d)

Download Full Size | PDF

A comparison of reflectance spectra obtained with different metal layer thickness is shown in Fig. 8. The metal layer thickness was varied from 30 nm to 60 nm. The depth of the grating was equal to the thickness of the metal layer. An NGSPR sensor with 30 nm metal layer thickness shows shallow resonance reflection peak in the short wavelength region. When the thickness of the metal layer is between 40 nm and 50 nm, the resonance peaks are sharp and deep. When the thickness of the metal layer becomes more than 60 nm, the depth of the resonance reflection peaks become shallow and have large FWHM values. The resonance reflection peak comes from the interaction between the metal film and the sample materials. The thicker the metal layer is, the less the amount of light coupled into the sample materials, which reduces the coupling effects. When the metal film is too thin, radiation damping from the metal film increases, which also reduces the coupling effects [15]. Optimum thickness for NGSPR sensors from this simulation study is between 40 nm and 50 nm.

This work has shown different effects of structural factors on the reflectance spectra from the work by Hick et. al [16], which used the NSL method to make nanowell structure with the light source from the sample side. The NGSPR sensor with optical illumination from the substrate side is less effected by a buffer solution that is turbid or opaque to the wavelength of the light used for a measurement. Resonance reflection peaks are deeper when the light comes from the substrate side. Additionally, the optical setup can remain stationary when one changes the samples. In [16], the authors found that resonance reflection peak wavelength changes with the etching depth. In comparison, NGSPR sensors are predicted to be not sensitive to the grating depth, if the grating depth is deeper than the metal layer thickness as shown in Fig. 7.

 

Fig. 8. Calculated reflectance spectra of NGSPR sensors (Λ=500 nm, u=50 nm) with a metal layer thickness of 30 nm (a), 40 nm (b), 50 nm (c), and 60 nm (d)

Download Full Size | PDF

4. Conclusion

Here we estimate the optimal design of a novel type of sensor. NGSPR sensors with a minimum grating feature size less than 50 nm and a period of 500 nm produce relatively sharp resonance reflection peaks. The resonance wavelength of the sample peak varies with the refractive index of the sample materials. NGSPR sensors have a simple structure for mass fabrication and less sensitive to the grating geometry. The predicted refractive index sensitivity of NGSPR sensors is larger than 400 nm/RIU and the FOM value is expected to be more than 60 due to the narrow FWHM, which is much better than alternative methods.

NGSPR sensors can be made into robust equipment using visible wavelength spectroscopy in reflection geometry. This simple vertical illumination reflection spectroscopic configuration can be easily implemented in small size systems without any moving parts and refractive index matching mechanisms. Recent developments in nanolithography and nanofabrication methods can make it possible to fabricate NGSPR sensor chips at low cost in mass manufacture. The size of unit cells in NGSPR sensors can be made smaller compared to conventional SPR imaging sensors in which the unit cell size is limited by the surface plasmon propagation length. NGSPR sensors can be used as an economic tools for highthroughput screening (HTS) application.