Investigation of graphene‐on‐metal substrates for SPR‐based sensor using finite‐difference time domain

Abstract

In this study, the authors investigated the effects of a single layer graphene as a coating layer on top of metal thin films such as silver, gold, aluminum and copper using finite‐difference time domain method. To enhance the resolution of surface plasmon resonance (SPR) sensor, it is necessary to increase the SPR reflectivity and decrease the full‐width‐half maximum (FWHM) of the SPR curve so that there is minimum uncertainty in the determination of the resonance dip. Numerical data was verified with analytical and experimental data where all the data were in good agreement with resonance angle differing in <10% due to noise present in components such as humidity and temperature. In further analysis, reflectivity and FWHM were compared among four types of metal with various thin film thicknesses where graphene was applied on top of the metal layers, and data was compared against pure conventional metal thin films. A 60 nm‐thick Au thin film results in higher performance with reflectivity of 92.4% and FWHM of 0.88° whereas single layer graphene‐on‐60 nm‐thick Au gave reflectivity of 91.7% and FWHM of 1.32°. However, a graphene‐on‐40 nm‐thick Ag also gave good performance with narrower FWHM of 0.88° and reflection spectra of 89.2%.

1 Introduction

Surface plasmon resonance (SPR) is an excellent way to measure the biomolecular interaction in real‐time label‐free environment. The common technique for plasmon excitation is the Kretschmann [1] and Otto [2] configurations where wide studies of analyte–ligand interactions have been performed on surfaces functionalised with a variety of biomolecules such as protein, ribonucleic acid and deoxyribonucleic acid. SPR phenomenon is represented by a sharp dip of the resonance curve which represents the minimum reflectance. It is based on the total internal reflection at the metal–dielectric interface. Excitation of surface plasmon (SP) is required to get the resonance oscillation. The excitation of SP can be accompanied either by varying the incident angle (angular interrogation) or by varying the wavelength (wavelength interrogation) of the optical laser beam. Thus, the monitoring of biomolecular interactions can be made either by measuring the shift in the incident angle or the optical wavelength at resonance [3].

The graphene sheet, which forms a dielectric layer, has been proposed to prevent the oxidation of silver, aluminium and copper surface that will modify its sensing capability. It provides good support for biomolecules because of their large surface area and rich π conjugation structure, making graphene a suitable dielectric top layer for SPR‐based sensors. It has been theoretically investigated and experimentally demonstrated that graphene can be used for exciting and propagating SPs [4, 5, 6, 7, 8]. Due to its unique electric, mechanical and thermal properties, graphene has found applications in a very wide area, such as active plasmonic switch, perfect absorber, photodetectors, sensors and plasmon waveguiding [9, 10, 11, 12].

Today, various kinds of SPR‐based sensor designs have been proposed for sensing applications. In 1983, Liedberg et al. [13] first reported the SPR sensing method for gas detection. For optical fibre‐based sensors, Rifat et al. [14] explored copper‐graphene‐based plasmonic biosensor with monolayer and multilayer graphene coating on top of a 30 nm‐thick copper. Jing Zhou et al. [15] investigated the characteristics of a silver‐coated sensor when exposed to the air and in liquid environment and they obtained full‐width‐half‐maximum (FWHM) results ranging from 39.3 to 81.0 nm. Other than that, bimetallic layer of rhodium–silver with silicon layer and graphene deposition was characterised using Kretschmann configuration for gas detection at 632 nm wavelength. The researchers found that the use of graphene over silicon not only protects the silicon layer from oxidation but also enables the sensor to sense a wide range of hazardous gases [16]. Numerical analysis on a variety of gold thicknesses in optical fibre based sensor [17] and bimetallic stack of gold/silver in waveguide coupled plasmonic sensor [18] were investigated in an effort to reduce the bandwidth of the SPR curve to enhance sensing resolution. Performance parameters for SPR sensors (i.e. sensitivity and detection accuracy) are influenced by the depth and width of the SPR response curve where the deeper the curve, the higher the sensitivity and the narrower the width, the higher the accuracy.

In SPR biosensor applications, the choice of the metal film is crucial. The SPR active metals must have conduction band electrons that are capable of resonating with the incoming light at a suitable frequency. Such metals are silver (Ag), gold (Au), copper (Cu), aluminium (Al), sodium (Na) and indium (In). Of these metals, Na is too reactive and In is too expensive whereas Cu, Al and Ag are too susceptible to oxidation and thus degrades the performance of the SPR biosensor. Hence, In and Na are neglected in this analysis. Among the factors that affect the performance of SPR‐based sensors are the thin film thicknesses and the type of metal used. The thickness of the metal thin film must be optimised to enhance the reflection spectra and the FWHM. FWHM is the full width at half minimum of the typical resonance spectrum, and defined as the corresponding wavelength width at the half percentage of the normalised resonant dip $(\mathrm{FWHM}=\delta {\theta }_{0.5})$. A smaller FWHM and bigger reflection spectra are desired in the SPR‐based sensor because a narrower and deeper resonance peak allows detecting the resonance shift effectively.

The novelty of this work lies in the fact that we analysed various thicknesses of graphene‐on metal versus conventional metal‐based SPR sensors using 670 and 785 nm illumination wavelengths to measure the resonance conditions such as total attenuation reflection and FWHM. At the end of this analysis, the finite‐difference time domain (FDTD) simulation continues with increase in refractive index of the dielectric layer to measure the angular displacement that occurs in two different illumination wavelengths. This is to see whether the illumination wavelength affects the performance of the SPR sensor apart from the geometrical parameter. This work explores the early stages in designing SPR‐based biosensors utilising graphene and various metal coatings combination before the knowledge is subsequently deployed in optical fibre based SPR sensor for portable and cost‐effective applications.

2 Numerical, analytical and experimental methods

The numerical analysis using Lumerical FDTD software was performed as a parameter sweep over the source incident angle (36–80°) in an effort to seek the source angle that excites the SPR mode (this happens when the y‐ component of the source wave vector matches the wave vector of the SPR mode) or in other words, the SPR occurs when the wave vector components are parallel to the metal surface. The perfectly matched layer profile is set to a ‘steep angle’ to better absorb light propagating at large angles away from the normal incidence. A mesh override region is used to force a finer mesh step size in the y‐ direction (normal to the metal film). The optical parameter was set as a plane‐wave source (bloch/periodic type) at illumination wavelength of 670 nm. This work investigated the optimal thickness of four types of metal layers ranging from 20 to 80 nm. The result of the conventional metal thin film was then compared with the presence of a single layer of graphene sheet (thickness, L  = 0.34 nm) on the metal thin films. As shown in Fig. 1, borosilicate glass (BSG) substrate with refractive index of 1.503 and thickness of 795 nm was used and the simulated dielectric medium was set as air (refractive index, n  = 1.00).

Fig. 1.

FDTD simulation diagram for metal thin film on BSG substrate SPR‐based sensor with Kretschmann configuration using angular interrogation technique with XY, XZ and YZ view

The material parameter that was used for simulating a single graphene layer was Falkovsky while the chemical rubber company (CRC) parameter model was used to depict all four metal types in the Lumerical FDTD software. Table 1 shows the dielectric permittivity for the metal thin films (Ag, Au, Al and Cu) while the surface conductivity for a single graphene layer is stated in Table 2 [19].

Table 1.

Dielectric permittivity for metals (CRC model)

Metal	Wavelength, nm	Dielectric permittivity, ɛ
		Re	Im
Ag	670	−22.1575	2.5149
	785	−30.8666	2.9866
Au	670	−13.3891	0.7223
	785	−21.6448	0.7433
Al	670	−61.5268	25.8277
	785	−65.9432	45.2282
Cu	670	−14.7258	1.6885
	785	−24.3702	2.4188

Table 2.

Surface conductivity for a single layer graphene (Falkovsky model)

Wavelength, nm	Conductivity, σ
	Re, μS	Im, μS
670	60.8373	−6.1572
785	60.6199	‐13.5021

To obtain credible numerical data, analytical and experimental methods were also carried out to monitor the SPR curve pattern. Using Lumerical FDTD software, the analytic transfer matrix method was used to calculate the reflection of a plane‐wave source through a 50 nm‐thick Au thin film on glass with 670 and 785 nm illumination wavelengths. The values of dielectric permittivity for the respective metals as portrayed in Table 1 were used. The ‘stackrt’ script command uses Fresnel equations to calculate the reflection and transmission coefficients. By specifying the thickness and refractive index of each layer, as well as the angle of incidence and frequency range of the illumination, this function returns the fraction of reflected and transmitted power, and the complex reflection and transmission coefficients, for both S and P polarisations.

The SPR Navi 200 instrument (BioNavis Ltd) which was used to run the experimental work is based on the Kretschmann configuration where the incident light is coupled via a prism and then the wave number of the incident light is enhanced to match that of the excited SP. This SPR instrument enables real‐time observation without labelling of molecular interactions, structural changes and layer properties in both wet and dry states. Due to its availability, a 50 nm‐thick Au thin film was coated on a glass slide and was chosen to run the experiments using 670 and 785 nm illumination wavelengths. SPR is excited by a p‐polarised (transverse magnetic mode) light, where once the refractive index of an analyte on the sensing region (dielectric layer above the metal layer) changes, the measured SPR spectrum will shift accordingly.

3 Results and discussion

Fig. 2 shows the simulated reflection spectra of 50 nm‐thick Au thin film on glass as a function of incident angle in comparison with the analytic solution for illumination wavelength of 670 and 785 nm. The analytical curve is plotted in dotted lines whereas the numerical curve is represented by straight lines. Thus, the numerical results are almost 100% matched to the results obtained analytically.

Fig. 2.

SPR curve of conventional 50 nm‐thick Au on glass with illumination wavelength of

(a) 670 nm, (b) 785 nm for numerical (solid curve) and analytical (dashed curve) experiments

Fig. 3 depicts the comparison of experimental and simulated SPR curve, where the experimental curve is plotted in solid lines and the simulated data is depicted in dashed lines. The simulation analysis of 50 nm‐thick Au thin film coated on glass shows good agreement with the experimental results with only a slight right shifted (0.5–0.8°) of resonance angle and difference of <10% in reflectivity results. The difference could be due to the noise factors present during the experimental measurements such as humidity, temperature and so forth.

Fig. 3.

SPR curve of 50 nm‐thick Au between experimental (solid curve) versus FDTD simulation (dashed curve) using

(a) 670 nm, (b) 785 nm illumination wavelengths

Upon validation of the simulated data with analytical and experimental data, further analysis is undertaken on the numerical investigation of the SPR sensor in three different parameter configuration: types of metal, metal thin film thickness and the presence of a single layer of graphene sheet, keeping all other parameters constant. In all configurations, the metal thickness will be numerically optimised with respect to the maximum reflectivity and minimum FWHM. Highest value of reflection spectra (equivalent to the lowest value of the reflection dip) ensures maximum transfer of energy from the incident light to the SP wave that leads to the maximum enhancement of the SPR field in resonance condition.

Fig. 4 illustrates the variation of the reflection spectra with incident angle for Ag, Au, Al and Cu (thickness ranging from 20 to 80 nm) at 670 nm optical wavelength. Results show that for the metal Ag, a thickness of 30 nm shows the best reflection spectra with 92.3% (FWHM = 1.51°) whereas graphene‐on‐40 nm‐thick Ag exhibits the highest reflection spectra of 89.2% (FWHM = 0.88°). Meanwhile, 60 nm‐thick Au thin film gave the best reflection spectra of 92.4% (FWHM = 0.88°) and graphene‐on‐60 nm‐thick Au thin film results highest reflection spectra of 91.7% (FWHM = 1.32°). 40 nm‐thick conventional Cu thin film shows best reflection spectra of 90.7% (FWHM = 2.64°) whereas graphene‐on‐40 nm‐thick Cu demonstrates highest reflection spectra of 91.2% (FWHM = 3.08°). In the case of Al metal thin films, it can be observed that there are no SP oscillation in metal layer thicknesses of 40 nm and above. The SPR curve is formed in Al films with thickness of 30 nm and below but gave results of a very shallow SPR curve which will eventually lead to the decline of sensor sensitivity and detection accuracy.

Fig. 4.

Variation of reflection spectra with incident angle for

(a, b) Ag, (c, d) Au, (e, f) Al, (g, h) Cu

Conventional metal thin film configuration (solid curve) and graphene‐on‐metal thin film configuration (dashed curve)

The reflection spectra and FWHM trends of various thicknesses of the conventional metal thin film versus graphene‐on‐metal thin film configuration are illustrated in Fig. 5. From this figure, we can see that when the thickness is above the optimum value, the SPR curve becomes shallow (low reflection spectra) but when it is below the optimum thickness, the SPR curve becomes broader (high FWHM). One can observe that FWHM is decreased with increase in metal layer thickness in both conventional thin film and graphene‐on‐metal configurations for Ag, Au and Cu except Al. The shape of the reflectivity curve and FWHM for these three metals are similar. This shows that Cu is also a good potential for plasmonic biosensor applications as it is known that Cu exhibits higher conductivity and is less expensive than Au and could be an excellent choice as a plasmonic material with graphene coating layer to prevent from oxidation.

Fig. 5.

Reflection spectra (solid curve) and FWHM (dashed curve) results in different thickness of metal thin film of

(a) Ag, (b) Au, (c) Al, (d) Cu, Conventional (full‐coloured box) versus graphene‐on‐metal (empty box) configuration

In SPR sensors with angular modulation of a monochromatic light, various wavelength sources can be utilised for illumination. Sensitivity is defined as the shift in resonance angle per unit change in the refractive index (n). Fig. 6 shows a master plot showing the variation of sensitivity of the sensing layer from n  = 1.00 to n  = 1.49 for conventional versus graphene‐on‐60 nm‐thick Au at two different illumination wavelengths: 670 and 785 nm.

Fig. 6.

Sensitivity of SPR‐based sensor with the sensing layer exposed from n = 1.00 to n = 1.49 of

(a) Conventional metal thin film, (b) Graphene‐on‐60 nm‐thick Au at two different wavelengths: 670 and 785 nm

It is demonstrated that the proposed sensor sensitivity of 51.18°/RIU with 670 nm wavelength is higher than 28.73°/RIU obtained using 785 nm wavelength light source, which means the sensitivity is degraded with the increment of the incident wavelength of the light source. Sensitivity of Au‐based SPR sensor improved (1.79°/RIU) with the presence of a single layer of graphene. As for the detection accuracy of the SPR sensor, it is observed that the curve width remains almost the same with or without the presence of the graphene layer. The detection accuracy was degraded with increase in refractive index of the sensing medium and also decreased when the illumination wavelength is lowered from 785 to 670 nm.

There is an increasing activity in chemically modifying graphene to enable analyte recognition molecules, e.g. antibodies, to be linked to it, e.g. Pumera (2011), Saleem et al. (2016) [20, 21], and the analysis presented in this paper could be extended to include this protein layer and any associated bound analyte layer as found in typical SPR biosensors [22, 23, 24, 25, 26, 27, 28]. Asymmetric line‐shape or Fano‐type resonances were observed to arise in all the reflectance spectra curve which is due to interference between the background and a resonant scattering process. Future research will emphasise on this interference phenomenon that produces the Fano‐type resonance in plasmonic nanostructures and their effect to the performance of SPR‐based sensor. Our future work will also involve analysing the SPR curve using a wavelength interrogation technique for medical/biosensing applications where devices with moving parts are avoided and deployment to optical fibre‐based SPR biosensors can be easily realised.

4 Conclusion

In summary, we have successfully executed an investigation of graphene‐on‐metal substrates for SPR‐based sensor using FDTD method. Numerical data has been verified analytically as well as experimentally and shows a close match. Subsequently, the performance of various metal thin film thicknesses in conventional and graphene‐on‐metal configuration was numerically investigated and analysed. The effect of graphene sheet on Ag, Au and Cu thin film was found to increase the FWHM by 30–50% at below optimum metal thicknesses whereas reflectivity has only a slight difference for the case of Au and Cu but it increased 18% for graphene‐on‐40 nm‐thick Ag compared with conventional 40 nm‐thick Ag. The analysis of different refractive index solution in sensing medium on conventional versus graphene‐on‐60 nm‐thick Au SPR‐based sensor has been verified that the 670 nm illumination wavelength slightly reduces the detection accuracy but enhances the sensor sensitivity about 56–58% compared with 785 nm wavelength.