• metamaterials;
  • sensors;
  • plasmonics;
  • THz;
  • subwavelength materials

One of the most notable application of terahertz (THz) radiation is that of sensing. This is because THz waves, being non-ionising, do not damage biological samples usually sensitive to visible light. More importantly, many complex molecules have vibrational and rotational modes in the THz regime, resulting in unique absorption spectral features.[1, 2] This opens up THz technology to a myriad of real-world sensing applications in biological and environmental fields as well as industries such as oil and gas. However, the widespread use of THz spectroscopy in sensing applications is greatly hindered by the scarcity of good sources and the limited sensitivity of detectors which lead to poor system sensitivities.[3] These limitations can be circumvented in THz sensing by using methods such as strongly confined electromagnetic fields to maximize light–matter interactions and sharp spectral features to enable the detection of small changes in the dielectric environment, similar to surface plasmon polariton (SPP) sensing at visible wavelengths.[4]

THz metamaterials have been the object of intense scientific research as platforms for generating high field enhancements as well as extremely narrow resonances. In particular, it has been recently shown that metal surfaces corrugated at subwavelength scales can support highly confined electromagnetic surface modes.[5-11] These are analogous to SPPs in the visible regime, and are usually termed ‘spoof plasmons’. The tightly confined nature of spoof plasmons[7] means that they could be very sensitive to their dielectric environment, similar to SPP sensing. Comprehensive studies have been conducted on the waveguiding capabilities of spoof plasmon surfaces (SPS).[12-14] However, very little has been done in the experimental implementation of spoof plasmon concepts to sensing applications.[15] It would be interesting to see if the unique characteristics of spoof plasmons would translate into good sensitivity to refractive index changes in a sensing scenario and perhaps even aid in side-stepping problems such as strong absorption by polar liquids.

In this article, we make use of a spoof plasmon metamaterial and its surface sensitivity to demonstrate refractive index sensing of various fluids in an Otto prism setup. Prism coupling configurations have been previously used to excite surface modes on semiconductor and metallic surfaces in the THz regime.[10, 16-20] However, in the case of semiconductors, the resonances probed this way tend to be quite broad due to high inherent material losses, thus limiting their ability to discern small refractive index changes.[16, 17] Here we couple to THz spoof plasmons via a wax prism and experimentally demonstrate refractive index sensing of various fluids by monitoring sharp changes in both amplitude and phase of the THz radiation, assessing the efficacy of SPSs for refractive index sensing. The ability to tailor the optical properties of SPSs by properly designing their geometric parameters means that SPSs could be a very flexible platform on which to conduct THz sensing. Broadband information can potentially be extracted via alternative coupling methods such as edge coupling.[7] At the same time, the SPS could be functionalised for high specificity chip-scale sensing and SPSs of different geometries can pave the way for novel THz sensing methods.

Our SPS consists of a linear array of grooves[21] as shown in Figure 1(a). A periodic groove array with an overall area of 18 mm × 34 mm is defined on a thick layer of AZ 9620 positive photoresist via conventional UV photolithography on a glass substrate. A 600 nm thick layer of gold is then sputtered onto the photoresist. As the gold layer is much thicker than the skin depth of THz frequencies (≈80 nm at 1 THz), the final structure closely approximates a corrugated perfect electrical conductor (PEC) surface. Figure 1(b) shows an optical microscope image of the fabricated groove array with the following dimensions as labeled in Figure 1(a): h = 27 μm, wt = 37 μm, wb = 25 μm, and d = 60 μm. A period of 60 μm implies that diffraction effects are only manifested above 2.5 THz. Below this frequency the SPS features are subwavelength and the SPS operates within the so-called metamaterial regime.


Figure 1. (a) 3D image of spoof plasmon metamaterial. The red region depicts the photoresist grooves and the yellow region depicts the gold film. (b) Optical microscope image of the fabricated groove array. (c) Schematic of the Otto Prism setup.

Download figure to PowerPoint

Characterisation of the SPS is carried out via THz Time-Domain Spectroscopy (THz–TDS) in a nitrogen purged chamber with relative humidity less than 5%. The frequency spectrum is then obtained from the time-domain signal via a Fourier Transform. A wax prism in the conventional Otto prism configuration is used for phase-matching, coupling to the spoof plasmon mode evanescently at the base of the prism with a coupling gap, g, between the SPS and prism base, as shown in Figure 1(c). We then monitor the changes in the reflectivity R, and phase change spectra inline image as the grooves are filled with various fluids namely nitrogen (n = 1.00), gasoline (n = 1.41), liquid paraffin (n = 1.49), glycerin (n = 1.82) and water (n = 2.1). This gives a good spread of refractive index values, enabling us to investigate the efficacy of THz spoof plasmon sensing with various sample fluids.

Reflectivity measurements were performed by detecting the signal reflected off the prism base with the SPS at a coupling gap g, beneath the prism, E(w,g), normalized to the detected signal in the absence of the SPS beneath the prism, inline image. This way, the experimental THz reflectivity is defined as inline image. The emergence of dips in the reflection spectrum can be linked to the destructive interference amongst the incident light, light coupled into spoof plasmons and the out-coupled light from the spoof plasmon into free space.[22, 23] They are clear indications of the coupling between the THz radiation and the SPS modes. The phase change is given by the difference in phase of each frequency component of the sample inline image and reference inline image spectra, i.e., inline image. For each fluid sample, we take measurements with a coupling gap of g = 70 μm, which yields the strongest spoof plasmon coupling in the case of nitrogen (See Figure 2). Measurements were taken with a frequency resolution of approximately 7 GHz. The samples were washed with deionised water and an additional frequency spectrum was taken between different fluids to ensure that the spoof plasmon coupling point remains the same as the nitrogen case (control sample).


Figure 2. Experimental reflectivity spectra for g = 20 μm (black), 40 μm (blue), 60 μm (red), 70 μm (green), 80 μm (cyan), 100 μm (magenta) and 120 μm (yellow). Simulated (dashed-cross, black) reflectivity spectrum at strongest spoof plasmon coupling point (g = 70 μm) is also shown for comparison. The fluid in the grooves is nitrogen. Inset: Simulated field distribution at 1.70 THz for g = 70 μm. The white line demarcates the prism-coupling gap interface.

Download figure to PowerPoint

A figure-of-merit can be defined for the sensing performance of the SPS as inline image, where inline image is the spoof plasmon frequency, inline image is the change in resonance frequency per refractive index unit (RIU) and inline image is the width of the resonance.[24] This FOM was calculated for each sample to evaluate the performance of our SPS for refractive index sensing.

2D simulations were carried out to obtain the reflectivity spectrum using COMSOL Multiphysics with a single groove as the unit cell and Floquet periodic boundary conditions applied along the direction of periodicity. The gold layer was simulated by applying a surface impedance boundary condition at the groove surface and using a Drude model to model the permittivity of gold in the THz regime.[25] Spoof plasmon dispersion curves were calculated using CST Microwave Studio Eigenmode Solver and used to predict inline image when the grooves are filled with dielectrics of different refractive indices.

Figure 2 shows the experimental reflection spectra for various g values as well as the simulated reflection spectrum at g = 70 μm with only nitrogen filling the grooves. Firstly, it can be noted that the reflectivity dips corresponding to coupling to spoof plasmons red shift as the coupling gap is decreased. This observation can be linked to the increase in effective refractive index in the vicinity of the SPS. Thus as the prism is brought closer to the metamaterial surface, the spoof plasmon frequency lowers due to the presence of the prism. Secondly, the strongest spoof plasmon coupling occurs at the spoof plasmon frequency inline image for a gap of g = 70 μm. This is in good agreement with simulations, which predict inline image. However, note that the experimental spectrum (green, solid) in Figure 2 is broader and deeper than its simulated counterpart (black, dash-cross), presenting a full width at half maximum (FWHM) inline image of 40 GHz and a Q-factor of approximately 43. The broadening of the resonance indicates the shortening of the spoof plasmon lifetime. Thus, we can conclude that there are experimental damping mechanisms which are not reflected in our calculations. The lowering of the spectral dip reveals that these damping channels do not contribute to the SPS reflectance. Therefore, we can relate them to an increase in absorption losses, due to the formation of hot spots at the structural imperfections in the samples, and to diffraction effects, caused by the scattering of the spoof plasmon modes at the edges of the experimental SPS. The inset of Figure 2 shows the simulated field distribution at 1.70 THz for g = 70 μm. As it can be seen, the fields are tightly confined to the SPS at the spoof plasmon coupling point (inline image 1.70 THz), well removed from the prism-coupling gap interface demarcated by the white line. The relatively high Q-factor of the reflection dip could potentially be used for sensing purposes in analogy to previous proposals for surface plasmon sensing at visible frequencies.[4]

Before we proceed on to the SPS refractive index sensing results, we draw attention to an oft-neglected source of information in metamaterials research that is available via THz-TDS, the phase of the electromagnetic fields. Figure 3 shows the change in phase between the sample and reference spectrum inline image (a) and the absolute gradient of phase change inline image (b) for g = 70 μm. Importantly, a sharp change in phase takes place at inline image, whereby inline image follows a Fano lineshape (see Figure 3(a), blue line). This sharp phase change can be used as a readout response by which refractive index sensing can be conducted. inline image (Figure 3(b), green line) exhibits a sharp peak at inline image corresponding to the sharp phase change. The width of this peak is approximately 10 GHz giving a inline image value of 170, where inline image and inline image are the peak position and the width of peak in inline image, respectively. Note that this peak is four times sharper than its counterpart in amplitude measurements, making phase change potentially a very effective indicator for the SPS to detect small changes in the refractive index of the material filling the grooves.


Figure 3. Phase change inline image (a), and absolute gradient of inline image, inline image (b) at g = 70 μm.

Download figure to PowerPoint

Figure 4 plots the reflection spectra for the various fluids mentioned above. As can be seen, there is a significant red-shift in the reflection dip as the refractive index of the fluid filling the grooves increases. The resonance points for nitrogen, gasoline, liquid paraffin, glycerin and water are 1.71, 1.53, 1.48, 1.30, and 1.17 THz, respectively. For low loss fluids, such as gasoline and liquid paraffin, the width of the resonance, inline image are 90 GHz and 50 GHz, respectively. However for fluids with higher loss, such as glycerin, the reflection dip broadens significantly to 280 GHz, yielding a Q-factor of approximately 4.6. With a fluid of very high loss such as water filling the grooves, inline image further broadens to approximately 390 GHz. Thus, refractive index sensing with amplitude measurements is not a good option for highly absorbing fluids, since a larger refractive index change would be required to properly discern any spectral shifts. Nonetheless, as compared to refractive index sensing via resonant metamaterials in transmission mode, our proposal signifies a relevant improvement since the THz signal after passing through a high loss fluid layer would be greatly attenuated.[26-32]


Figure 4. Reflection spectra of sample fluids (nitrogen (blue), gasoline (cyan), liquid paraffin (red), glycerin (green) and water (magenta)) at g = 70 μm.

Download figure to PowerPoint

Figure 5 shows that the inline image spectra, obtained using fitted inline image data, exhibits a phase change maximum for all fluids tested. The inline image spectra (with the exception of the nitrogen spectrum) are magnified to enable easy reading. This is because losses in the fluids result in a smaller overall phase jump at resonance and hence a smaller inline image peak. Remarkably, all the peaks in Figure 5 are narrower than the reflectivity dips in Figure 4. The inline image values for nitrogen, gasoline, liquid paraffin, glycerin and water are approximately 170, 44, 78, 18, and 7, respectively. In particular, inline image values for glycerin and water are 4 and 2 times as high as its equivalent for amplitude measurements. The broadening of the inline image spectra for fluids with higher loss, namely glycerin and water, is again due to their absorbing nature. Nonetheless, our experimental results indicate that the sharp phase changes characteristic of resonance phenomena mitigates the broadening, making refractive index sensing possible on fluids with higher loss.


Figure 5. Absolute gradient of phase change spectra inline image for nitrogen (blue), gasoline (cyan), liquid paraffin (red), glycerin (green) and water (magenta) obtained from fitted inline image data. The inline image spectra for fluids other than nitrogen are magnified for clarity.

Download figure to PowerPoint

The resonance frequencies of the various fluids are plotted against their respective refractive indices in Figure 6. The blue solid line is a linear fit given by inline image. This gives a sensitivity (defined as the change in inline image per RIU) of 0.49 THz RIU-1 and a detection limit of 0.02 RIU assuming a detection resolution of 10 GHz. The sensitivity featured in our SPS is higher than that previously reported in similar devices.[32] Hence using inline image as our sensor readout response, the FOM values for nitrogen, gasoline, liquid paraffin and glycerin are 49, 15, 25, and 7, respectively (see Table 1). This is much higher than previously reported values[33, 34] and makes our sensing approach very promising for application in areas where the fluids generally present low absorption losses, such as the petroleum industry.[35] The high FOM values here are made possible by the narrow spectral features in the inline image spectra. This is in contrast to the generally broad SPP resonances on semiconductor surfaces, where absorption losses significantly decrease surface plasmon lifetimes.[16, 17] For water, our SPS sensing platform yields approximately FOM = 3 (see Table 1). Although this value is lower than the other fluids tested here, it demonstrates that refractive index sensing can still be reasonably carried out in high loss fluids using our scheme. The experimental results are well corroborated by simulated results from the CST Microwave Studio Eigenmode Solver given by the red open circles in Figure 6. The data points are extracted from the crossings between the calculated spoof plasmon dispersion curves for the respective values of n and the prism light line, giving the predicted values for inline image. A linear fit of the simulated result is given by inline image. The resultant sensitivity of 0.52 THz RIU-1 is in good agreement with experiment.


Figure 6. Resonance frequencies of various fluids, inline image at g = 70 μm plotted against refractive indices for experiments (blue solid, cross), simulations (red dashed, open circle) and analytical expressions with (black, dotted) and (black, dash-dot). (Inset) Electric energy density distributions for n = 1 and n = 2.1 at inline image.

Download figure to PowerPoint

Table 1. Summary of Figures-of-Merit for the various fluids tested
Liquid Paraffin25

The relationship between inline image and n of an idealised SPS consisting of a linear array of square grooves with groove width inline image, i.e., inline image, is analytically given by solving the following equation for inline image,

  • display math(1)

where inline image is the prism refractive index, inline image is the angle of incidence of light in the prism, c is the speed of light and n is the refractive index of the dielectric filling the grooves.

Equation (1) equates the spoof plasmon dispersion of the SPS (RHS)[36] with the prism light line (LHS) and is solved for the cases inline imageand weff = wb = 25 μm with inline image. The solutions for weff = 37 μm (black, dotted) and weff = 25 μm (black, dash-dot) are plotted in Figure 6. As it can be seen, the experimental curve (blue solid, Figure 6) falls within the limits of the idealized cases. The minimum sensitivity over the range of sampled values of n, calculated by taking the numerical gradient of the analytical curves, are 0.40 THz RIU-1 and 0.47 THz RIU-1 for weff = 37 μm and inline image, respectively.

It is noted that the analytical curve follows a 1/n trend due to the tangent term in Equation (1). However, in our experiments, the trend observed is quite linear, as can be clearly seen in Figure 6. This is because the slope in the groove walls introduces a varying groove width inline image, with the depth of the grooves. The insets of Figure 6 shows the electric energy density distributions for n = 1 and n = 2.1 at their corresponding inline image. As it can be seen, for n = 1 the electric energy is concentrated mainly at the top of the groove and for n = 2.1, the electric energy resides predominantly within the groove itself. This explains the observation that in the vicinity of n = 1 , the experimental results are similar to the case of weff = 37 μm while near n = 2.1 the SPS behaviour is closer to that of weff = 25 μm since the inline image experienced by the electromagnetic fields is smaller at n = 2.1. Lastly, it is noted that in the vicinity of n = 1, the sensitivity values given by the analytical results can be higher than those experimentally obtained. This means that the performance of SPS-based sensors can potentially be improved and higher FOM values can be obtained.

In this work, we propose a novel high FOM refractive index sensing platform based on SPSs. The SPS consists of a periodic array of subwavelength grooves with a conventional Otto prism setup used for the excitation of spoof plasmons. Experiments were performed validating this approach for the sensing of lossy liquids. We highlight here the use of phase information readily available from THz TDS, particularly the characteristic phase jumps on our SPS, as a readout response to conduct refractive index sensing that out-performs amplitude-based measurements usually seen in works concerning resonant THz metamaterials sensing. A FOM as high as 49 was achieved for lossless fluids while FOM values in the vicinity of 20 can be expected for low loss fluids such as gasoline and liquid paraffin. The sensing of high loss fluids like water is tenable under this scheme and theoretical results suggest further improvements to sensing performance can be achieved. Good agreement was achieved between simulations and analytical results, and a sensitivity close to the theoretical limit was achieved. Depending on the sensing scenario, the proposed methodology could be further optimized by methods such as tailoring coupling gaps to the sensed fluid. SPSs of other geometries, such as annular hole arrays[8] or thin meta-films[37] could also be exploited for THz sensing. The sensing mechanism proposed is compatible with other sensing methods such as Attenuated Total Reflection spectroscopy and could be used in conjunction to increase sensitivity.[38] Moreover, further work such as obtaining broadband data via edge-coupling, sensing via interferometric means[39] and functionalising the surface for high specificity sensing can be done. We expect that our work here may pave the way for novel spoof plasmon-based sensing devices in the THz regime.

Experimental Section

  1. Top of page
  2. Experimental Section
  3. Acknowledgements

SPSs consisting of a linear array of subwavelength grooves were fabricated via conventional UV lithography. A thick layer of AZ 9620 positive photoresist was first coated onto a glass substrate via a two-step spin coating process. In the first spin coating step, the AZ 9620 was spun at 1350 RPM for 1 min followed by a soft bake at 110 °C for 80 s. The second spin coating step was performed at the same spin speed followed by a soft bake at 110 °C for 180 s, giving a final photoresist thickness of approximately 28 μm. The photoresist is then exposed to UV light (λ = 405 nm) via a pre-fabricated photomask in a mask and bond aligner (Karl Suss, MA8/BA6) with a dosage of 1000 mJ cm–2 and developed with AZ 400K developer to reveal the photoresist groove array. Lastly, 600 nm of gold is sputtered onto the photoresist groove array (AJA International Inc., ATC 1800V) to give the final SPS.

The Drude model parameters used to calculate the simulation results presented in Figure 2 are: ωp = 1.367 × 1016 rad s–1 and γ = 4.072 × 1013 rad s–1, where inline image is the plasma frequency and inline image is the collision frequency.

The dispersion curves used to calculate inline image in Figure 6 were obtained using the Eigenmode solver in the commercial software CST Microwave Studio. The unit cell consists of a single PEC groove filled with a dielectric of refractive index n = 1, 1.41, 1.49, 1.82, 2.1. Periodic boundary conditions were applied along the direction of periodicity. The intersection points between the resulting dispersion curves and the prism light line inline image at the various values of n, were then obtained numerically.


  1. Top of page
  2. Experimental Section
  3. Acknowledgements

The authors would like to acknowledge the funding provided by the Leverhulme Trust and EPSRC for the research work published in this paper. Binghao Ng would like to express his gratitude for the support from the A*STAR–Imperial Partnership (AIP) scholarship program. The author affiliations were corrected on August 23, 2013.