1. Introduction

The terahertz (THz) optical detection techniques are promising as emerging sensing tools for bio/chemical species as small molecules, proteins, and cancer tissues [1–4]. The time-domain spectroscopy is highly advantageous for identifying the complex optical property information in both quantitative and qualitative manners, thus precise determination of intra- and inter-molecular vibrational modes in broad THz spectra was realized [5]. Moreover, advances in nanotechnologies have boosted its capability and application area by providing dramatic sensitivity enhancement, therefore, even such small trace molecules as nucleotides could be measurable [6]. For such signal-enhancing purposes, metasurfaces based on nanoscale gaps [7] and on-chip topological sensors [8,9] have been introduced to remarkably increase the detection sensitivity. For the specific targeting, surface treatment on the nanoscale optical detection chip including receptors, antibodies, or structural perturbations can offer a clue for contribution to the THz studies for health care issues as well [10–14]. Recent studies tried to monitor the aerosols in the atmosphere by viral disasters or air pollution, which severely disrupt the health of humans [15]. However, the aerosol particles are diffusive and extremely small (viruses are 200 nm in diameter or ultrafine dust are 300 nm in diameter), and most importantly, they are extremely dispersed in the air. What is worse, the dispersive properties become accelerated as the target particles become smaller. Although it is very important to investigate to define the physicochemical characteristics of such particles, it is challenging to concentrate them into the sensing hotspots, thus leading to a decrease in detection efficiency and accuracy. For example, the target substances were prepared in solution bases, involving aggregations, random re-crystallization, and a coffee-ring effect unavoidably [16].

Here, we demonstrate a rather simple methodology to study such small-sized particle behaviors. Our developed metasurface techniques have represented a variety of applied studies ranging from extreme trace chemical sensing and selective evaluations of various species [17–19]. Additionally, specially designed metasurfaces with vertically double layers of slots gave a chance to monitor the nanoscale particle dynamics and evolutionary behaviors under the water environment recently [20]. Such a unique trapping method and following clean numerical analysis help to enable accurate optical signal control via much increased detection efficiency and reliability. Based on micro- and nano-structured resonators with various shapes, gap scales, and deposition methods will be discussed as functions of the deposited ultrafine dust particles, here. We measured THz transmission for 300 - 3000 nm size dust particles that are spread over the metasurface sensing chips. A significantly localized and enhanced THz field distribution was distorted by either the adsorption or trapped particles. The changed optical signals were then discussed quantitatively and qualitatively with regard to the reduced transmission signal and distinctive resonance frequency shift, and the results were supported by the numerical simulation as well.

2. Design and sample preparation

The metasurface structures for the THz signal enhancement were fabricated by photolithography technology, which guarantees several hundreds of nanometers of scale etching. Two different types of resonant cases were produced and tested for ultrafine dust particle detections; double split ring resonators (DSRR) and punctured nanoslot resonators (NSR). The split ring resonators are made to produce a very sharp resonance with a high-Q factor in the spectrum, determined by the Fano-type resonance behavior. In that case, the transmitted spectrum has a strong angle dependence between the certain axis of the resonant patterns and the incident THz polarization [21,22], thus the sensitivity can be further controlled by the relative angle as well. The slot structure has more advantages for the sensing purpose because it has a certain volume in two dimensions containing the target samples inside the gap region. It is noted that the mode volume at the resonance frequency was determined by the geometry of the slots and the squeezed mode volume by the boundary conditions in the subwavelength scale, providing the maximal signal change at the same time. From that point of view, the recent paper introduced that the evolutionary particle dynamics in real-time can be cleverly controlled by overlapping the optical hotspot and the trapping capability at the same location [20]. Although such dynamical behaviors can be captured with the active control of both electrical- and optical parameters of the device, here, it is rather discussed that the particle sensing performance and the efficiency are influenced by the geometrical factors and the difference in the resonant mode volume.

To understand such a resonator-based sensing platform, it is critical to check the electric field enhancement behavior through the metasurface. The field enhancement is defined as the coverage normalized ratio E_sam/E₀, where E_sam is the transmitted field and E₀ is an incident field at the near field region. By the finite element method (FEM) simulation, significantly enhanced transmission at the near-field was captured as shown in the cross-sectional view of the field map (Fig. 1(a)). The field distribution of |E_x| demonstrates strong near-field was localized and amplified at the gap position. The THz optical hot spot, in turn, results in dramatic enhancement of the absorption cross-section of the target molecules as well as the sensitivity of the detection platform [23]. The analysis of the measured THz signal was performed in terms of the optical property change of the deposited samples (Fig. 1(b)). The changed spectrum by the deposited material provides both optical properties including the complex refractive indices; the decreased transmission signal, ΔT, is induced by the THz absorption of the materials, and the shifted resonance frequency, Δf_res, is dominantly generated from the changed refractive index, n. By the relation between the THz electromagnetic wave and the complex optical characteristics, the exact refractive index of the deposited materials on the resonators can also be extracted [24]. The algorithm to find a complex refractive index by transmission/reflection measurement through resonant metasurfaces was introduced elsewhere [24]. The optical images before (left) and after (right) the particle dropping were taken for DSRR (Fig. 1(c)) and NSR (Fig. 1(d)) cases, respectively, showing the dispersed particle distribution.

 

Fig. 1. (a) FEM-calculated map of cross-sectional THz near-field amplitude near the nanoslot antenna, creating enhanced local field inside the slot. (b) The measured THz signal in transmission was plotted for the bare nanoslot chip and the sample deposited chip can be analyzed in terms of the change in transmission (ΔT) and the shifted frequency (Δf_res). Optical images of the metachip surfaces before and after ultrafine dust solution dropping onto the metasurface chips in DSRR (c) and NSR (d) cases, respectively.

Download Full Size | PDF

The positively patterned SRR arrays were made by photolithography. A 150 nm thick-gold layer onto a 10 nm thick-titanium adhesion layer was coated in a silicon substrate with a high resistivity (> 10,000 Ω·cm) to completely prevent signal distortion by the substrate. A single unit of the split ring with radius (r = 18 µm) and thickness (t = 4 µm) has a gap of 25° with a gap opening along 0°- and 135°-axis, which are asymmetrically aligned with each other to create Fano-like resonance behavior in the spectrum. The nearest neighbor of the unit is separated with a period of 50 µm. The superlattice elements were periodically repeated in a total area of 10 × 10 mm². Figure 2(a) represents an SEM image after the dust particle dropping process. The particles have a diameter in sub-hundred to several micro-meters and huge randomness in their shape. Even though the random drop-and-dry protocol results in inhomogeneous particle distribution as shown in the image, many particles are located at the optical hot spot (the split gap region), allowing reasonable signal change. The experiments were performed at a certain angle of 67.5° in the azimuthal direction, where the DSRR chip shows a sharp resonance feature (Q-factor = f_resonance/FWHM (Full-Width at Half Maximum) is 10) at 1.2 THz in the transmittance spectrum. The negatively patterned NSRs were fabricated by conventional photolithography as well. A 150 nm thick-gold layer onto a 10 nm thick-titanium adhesion layer was deposited on a silicon substrate with a high resistivity. Then slot arrays were fabricated with void rectangular slots with 500 nm in width and l = 48.9 µm in length, allowing a resonance peak at 1.2 THz. The slot arrays have a period of 40 µm horizontally and 10 µm vertically, and 1,700 slots in total in 2 × 2 mm². By making such many slots, the system noise can be further reduced by the fact that the side lobes of the diffractive THz signals in transmission at the near-field region can be destructively interfered with each other, in turn, the normal directivity can only survive, maintaining the maximum value. This many-slot array effect can further help to enhance the system’s sensitivity as well. The silica particles as ultrafine dust particles (Arizona test dust) were obtained in the form of the aqueous suspension from the micro-to-nano scale particles solution. The particle-contained solution was substituted by the addition of purified water repeatedly and dropped onto the prepared metasurface chips and dried under the vacuum condition. The SEM images after the drop-and-drying process show a similar distribution of the particles onto the NSR metasurface chips, as the random and inhomogeneous location of them (Fig. 2(b)).

 

Fig. 2. (a) A bird’s-eye view of SEM image of the DSRR metachip surface after the dust dropping. (b) SEM image of the NSR metachip surface after the dust dropping. (c) The measured THz transmittances from the DSRR chips were plotted in terms of the increased particle mass. (d) Measured THz transmittances from the NSR chips as the particle mass increases. The right upper color legends represent the dropped particle mass from 0 µg to 140 µg. Insets were added to clearly show the changes in transmission spectra.

Download Full Size | PDF

3. Results and discussion

The time-domain THz spectroscopy was performed using an 800 nm centered femtosecond laser (Maitai HP, Spectra-Physics, 80 MHz in repetition rate and 100 fs in pulse duration). The laser pulses divided by the beam splitter were incident to the THz emitter including a high-voltage photoconductive antenna and to the THz detector including the ZnTe nonlinear crystal for electro-optic sampling. The generated broadband THz spectrum covers 0.2-2.0 THz, and the THz beam propagation was controlled by the parabolic mirrors and focused by the lens pair. At the THz focus, the particle-deposited metasurface chips were placed and the transmissions through the chips were measured under the dry compressed air-purged surrounding to avoid the absorption by humidity. The obtained time-domain signals were further Fourier-transformed to produce the frequency-domain spectra. The DSRR has a 2 µm split gap which can be considered as a capacitor in an LC circuit, thus the optical spectra can be analyzed with the refractive index of deposited material at the gap region. The measured THz transmittance data in terms of the dust mass was plotted and the following signal decreases according to the increased dropped particles were obtained (Fig. 2(c)). The measured transmissions for nanoslot chips with different masses of particles were also plotted in Fig. 2(d) as well. In both graphs, the insets represent the decreased signal trends in terms of the dust mass more clearly.

Because the homogeneous and uniform film-like dielectric media as deposited sample gives the clear redshift in resonance frequency normally [25], it is meaningful to investigate the resonance shift behavior mainly (Fig. 3(a)). A resonance shift changes by simple drop-casting were fitted by the Michaelis-Menten function as y = V_max·x/(k_m+ x) [26], where V_max is the maximum of resonance change, and k_m is also the slope of a curve describing sensitivity. In DSRR cases (green in Fig. 3(a)), for the region with relatively larger mass (40 to 140 µg), the data show a very clear dependence with R²= 0.96. However, the system is more influenced by the deposited particles’ condition such as size or shape, in particular, at the rare density region. A drop-casting onto such a larger gap (as compared to the particle size) clearly lacks a quantitative sensing capability at the low concentration regime as R²= 0.80 with much severe data fluctuation. It can be easily considered that such extremely smaller particles (sub-hundred nanometers) than the THz wavelength and separately located ones in the middle of the gap may lead to missing the meaningful signal change (there are many separately existing particles as shown in SEM image from Fig. 2(a)). Although there is a clear field enhancement, 150 at the maximum, at the gap position of the DSRR, it is difficult to expect that a tiny optical property changes in an area much smaller than the gap directly leads to a measurable change in the signal. It is, therefore, quite limited to apply such larger gap size metasurface to the real samples including broadly distributed particle size and shape samples. From that point of view, it is notable that there is a strong correlation between the gap size of the metasurface and the relative size and amount of the target samples for the sensing purpose.

 

Fig. 3. (a) The changes in resonance frequency for both DSRR and NSR chips and (b) enlarged data plot only for NSR is represented with Michaelis-Menten function fitting for the brown dashed region from (a). (c) The changes in transmittance (intensity value) are plotted in terms of the dust mass for both DSRR and NSR chips. (d) Delta frequency vs. delta transmittance map for several interesting particles using 500 nm width NSR type of metasurfaces.

Download Full Size | PDF

As compared to the DSRR results, however, there is a clear advantage from two perspectives using the narrower gap width in NSR; increased field enhancement and rather squeezed mode volume in space at the same time. Even for such a lower Q value of NSR (Q = 4), more squeezed mode volume (V_eff = 0.39 µm³) leads to the extremely high Q/V_eff (two orders of magnitude higher than DSRR), allowing more efficient detection of ultrafine particles [27]. The narrower gap induces more field enhancement as 300 at the maximum, in turn, greatly increases the system’s sensitivity discussed in earlier research [23]. Also, the reduced mode volume can efficiently target smaller particles even from such mixture particle size distributions, implying better quantitative detection capability as shown in curve fitting with the R²= 0.99 even at the smaller mass regime (4-to-20 µg) (Fig. 3(b)). The calculated field enhancement, |E_x|² is 90,000 at the gap position, and is four times larger than the DSRR case in intensity, allowing much increased readability. The squeezed mode volume by the narrower gap improves the detection scope, enabling tighter focusing. The focusing effect makes smaller particles within several hundred-nanometer particles more efficiently visible. Such a great correlation demonstrates the potential for this NSR type of metasurface to operate splendidly for real samples even with varying size distributions.

Apart from the typical optical sensing platforms like typical plasmonic resonance-based sensors, which relies on the refractive index only, the THz spectrum gives one more degree of freedom in physical parameter, as transmission change. Since the transmission amplitude is related to the absorption property directly, thus, enables the quantitative analysis possible. Similar to the resonance behavior, NSR-type metasurface has a larger dynamic range of up to 12% in the transmission changes even for tiny quantities of dust as well, implying better performance for particle monitoring (Fig. 3(c)). Both important optical parameters including resonance frequency shift and transmission amplitude changes, which are correlated to the complex refractive index values can be mapped as in Δf-ΔT (Fig. 3(d)). All measurements were performed with 500 nm NSR-type metasurfaces, which means similar THz field enhancement values for all chips and re-normalized at the initial bare experimental conditions, respectively. The critical slopes in the map are strongly dependent on the material species. The comparisons with other experimental results from several references, which were measured using the same 500 nm slot nanogaps from the same group, clearly represent that the measured optical property changes in terms of the material components. The gold nanoparticles (AuNP, yellow in the map) have the highest refractive index values among them [28], for example, therefore, showing the highest slope and largest dynamic range in both Δf and ΔT values. It is also reasonable that the AuNP + FITC (Fluorescein isothiocyanate) has a rather smaller data change with the fact that the particles are coated by the dielectric fluorescence materials, in turn, induce decreasing the effective refractive index. The most considerable part is that the dust particles are placed between Polystyrene (PS) nanoparticles (gray in the map) [20] and AuNPs. Given the fact that the dust standard is composed of SiO₂, Al­₂O₃, and Fe₂O₃ mostly which have a refractive index (n) in the range 2 - 3 in THz range [29], the effective optical properties can be estimated as being in the middle of normal synthetic polymers and conductors. Such mapping in Δf and ΔT values has a great advantage, in that point of view, as a promising methodology for material sorting and classification even for such mixture samples in the real world.

Given the versatility and convenience of this sensing platform, the addition of such chemical treatment (attaching dye molecules, receptors, antibodies, or complicated chemical surface treatments) was ruled out, here. As a rather convenient and efficient way, sweeping with a flexible film was tried to effectively collect and trap the target particles into the optical hot spot as described in the schematic illustration (Fig. 4(a)). For the sweeping purpose, a piece of Polydimethylsiloxane (PDMS) film was suggested. The PDMS sweeper was made by mixing Sylgard 184 silicone elastomer (Dow Corning Corp.) under the curing process (10:1 ratio curing agent), followed by 3 hours at 80 °C thermal curing. A slice of 5 mm thick-PDMS film in 5 × 10 mm² was completely dried and then prepared for the sweeping. After dropping the particle solution, followed sweeping process showed completely inserted dust particles into the nanogap volume in the SEM images (Fig. 4(b-c)). Additionally, it can be also expected that random particles can be sorted by size in the swept situation. The sizes smaller than the nanoslot volume (sub-hundred to hundred nanometers) could be only trapped at the gap volume and all the rest were removed out by the sweeper. The measured transmittance data for the dust and repeated sweeping depict clear signal dependency on the sweeping cycles (Fig. 4(c)). The increased transmission amplitudes in terms of the sweeping cycles, implying more localization of the dust particles inside the nanogaps, give an excellent correspondence with the measured THz signal, as R²= 1.00 (Fig. 4(d)). By applying the sweeping process additionally, most of the particles at the surface were completely inserted into the nano volume. Previously introduced swept results showed very robust dependence on the concentration of the dropped AuNPs, accompanied by clean control of the THz transmission. The fundamental THz resonance was suppressed by the swept particles supporting a clear THz transmission control by the filled particle concentrations, as well as the removal of unwanted by-products which occasionally frustrates signal accuracy in sensing experiments.

 

Fig. 4. (a) Schematic illustration of the concept of sweeping particles inside the nanoslot area using a PDMS sweeping method. SEM images for the metasurface after several cycles of sweeping are shown in (b) and (c). (d) Measured THz transmission spectra for metasurfaces after sweeping process in several cycles and (e) change in transmission data. The right upper color legends represent the dropped particle mass from 0 µg to 140 µg.

Download Full Size | PDF

4. Conclusion

In conclusion, we propose efficient and sensitive detection methods for ultrafine dust particles using THz spectroscopy and THz metasurfaces. The resonant behaviors realized by the DSRR and NSR array result in spectral changes in measured THz signals. The further sweeping technique was applied to robust control of the transmission concerning to filling of the nanogaps and then show a highly sensitive sensing performance. A detailed comparison in terms of the resonator geometry and gap size was discussed. By mapping the data in Δf and ΔT values, it was shown that material sorting and classification are feasible in non-contact manners. Based on the studies above, hundred-nanometer size gaps are beneficial in both fields of view including field enhancement factor and appropriate size matching to the ultrafine dust particles. This convenient solution for tracing of ultrafine particles can provide insight into quick and accurate detection in a vast range of bio-applications including nanoscale organs inside a cell, small molecules, proteins assembly, and various nanoscale solid materials.