Terahertz virus-sized gold nanogap sensor

1 Introduction

Detecting and characterizing biomolecules such as virus, protein, and DNA has become a vital part of the response to the recent COVID-19 pandemic. Various testing methods such as the polymerase chain reaction (PCR) and several alternative antigen and antibody tests have been further developed and improved in recent years [1], [2], [3], [4]. The PCR tests are considered as the gold standard because the high-sensitivity PCR tests are almost 100% accurate in spotting infected people, however, such tests take hours to provide the results [3]. Antigen tests give results in less than 30 min, but the tests require tens of thousands of virus particles per microliter to produce a positive result [5]. For another alternative diagnostic method, various light-based detection methods have been developed in a broadband wavelength range from X-ray to microwave [6], [7], [8], [9], [10], [11], [12]. Especially, terahertz sensing of chemical and biological molecules has attracted strong attention in optical sensing communities due to their advantages of fast, non-destructive, label-free, and ultra-small quantity detection [7, 8, 13], [14], [15], [16], [17]. However, due to the size mismatch between target viruses with a diameter below 100 nm [18], [19], [20] and the terahertz wavelength of a few hundreds of microns, it still remains a challenge to detect a tiny amount of viruses.

To overcome the size mismatch issue, terahertz metamaterials [21], [22], [23], [24], such as split-ring resonators [7, 14, 25], slot antenna [8, 26, 27], and graphene metasurface [28], have been suggested for sensing the target viruses. These subwavelength structures enable perception of the dielectric constant change within localized electric field areas and allow extreme light–matter interaction. However, it might be difficult to reduce the size of the gap to a few nanometers and simultaneously pattern over a large area, using standard lithography techniques, such as electron-beam lithography and photolithography. Recently, atomic layer lithography that can resolve the aforementioned challenges was developed for the fabrication of nanogaps with a width below 1 nm and packing 150,000 such devices on a 4-inch wafer [29], [30], [31]. Furthermore, it allows decreasing the size of the gap down to 1 nm (∼λ/10⁶) leading to extreme field confinement and enhancement near the gap in the terahertz frequency regime [29] and detecting the change of refractive index caused by 1 nm-thick dielectric overlayers [32].

Here we demonstrated terahertz sensing of a tiny amount of PRD 1 which is double-stranded DNA bacteriophage and prototype of the Tectivirus with 10 species in five genera [33]. We utilized the high-density gold nanogaps with a gap width of 20 nm, which is about three times smaller than the size (∼60 nm) of the target virus, as depicted in Figure 1(a). In the vicinity of optical hotspots near the nanogaps, strong light–matter interaction is promised, and the nanogaps were uniformly extended to a hundred micrometers-long loop (l _x = 10 μm and l _y = 40 μm) to sustain strong terahertz resonances (Figure 1(a) and (b)). As shown in Figure 1(c), our nanogap structure has a slanted wall on the top of gap region, virus particles can tightly fit as a bridge, even though the virus is larger than the gap width (See the details in Figure S1). To experimentally verify the terahertz response when putting the virus onto the nanogaps, we employed a terahertz time-domain spectroscopy over the frequency range from 0.2 THz to 1.2 THz and observed the resulting redshifts in the resonance peak of the nanogap-loop array coated with the target viruses. To predict the redshifts qualitatively and quantitatively with the coated viruses, we performed full-wave numerical simulation and an analytical calculation based on the modal expansion method [13, 34, 35].

Figure 1:

Terahertz virus-sized gold nanogap. (a) Schematic of the gold nanogap-loop array (l _x = 10 μm, l _y = 40 μm, d _x = 100 μm, d _y = 50 μm, and h = 100 nm) coated with PRD1 viruses. The electric field polarization of the terahertz wave is perpendicular to the long axis of each loop. (b) Enlarged view of the PRD1 virus particle onto the etched gap with a width (w) of 20 nm. Strong near-field confinement in the vicinity of nanogaps called an optical hotspot facilitates the sensitive detection of gap-sized PRD1 viruses. (c) Top-view scanning electron microscope (SEM) image of a 20 nm gold nanogap.

2 Results and discussion

We performed the atomic layer lithography [29, 30] to manufacture a periodic array of gold nanogap-loops with a 20 nm gap width and a 100 μm perimeter. As described in the previous works [29, 36, 37], the detailed fabrication steps are introduced as shown in Figure 2. Conventional photolithography is performed to create patterns on an undoped silicon substrate with a thickness of 500 μm. The n-type photoresist (ma-N-1410) is spin-coated at 3000 RPM for 30 s and subsequently baked at 100 °C for 90 s. After that, the sample is irradiated with ultraviolet light (365 nm wavelength, 10 mW/cm² intensity, 40 s exposure time) using an MA-6 mask aligner and is developed with ma-D 533/S. Cr (5 nm) and Au (95 nm) layers are deposited on the photoresist patterns using an electron-beam evaporator and a lift-off process is performed using an NMP solution. Using an atomic layer deposition, an insulating layer of Al₂O₃ with a thickness of 20 nm is conformally covered on the patterned metals, and the thickness of the Al₂O₃ layer deposited on the sidewall of the patterns determines the nanogap width. Through secondary Au deposition, the slots are subsequently filled with gold. Finally, Ar-ion milling is performed with an oblique angle of 85° to peel off the excess gold layer more easily with adhesive tape. Further, we performed an additional etching process to remove remained Al₂O₃ layer on the top surface of gold using a 6:1 buffered oxide etchant (BOE) solution for 20 s.

Figure 2:

Schematic diagram of the fabrication process of the gold nanogap-loop array. (a) Photoresist patterns are fabricated by conventional photolithography on a silicon wafer. (b) Subsequent Au deposition and lift-off processes are performed to obtain the rectangular shape metal patterns. (c) The patterned metal substrate is conformally coated with a 20 nm thick Al₂O₃ layer using atomic layer deposition. (d) The secondary Au deposition is performed to fill inside the rectangular hole arrays. (e) Ar-ion milling is performed before peeling off. The excess metals atop the patterned metal substrate are exfoliated using adhesive tape. (f) Partially remaining Al₂O₃ layers are removed by etching with a 6:1 BOE solution for 20 s.

To experimentally characterize the gold nanogap-loop arrays, we employed a terahertz time-domain spectroscopy in the frequency range from 0.2 THz to 1.2 THz [38], [39], [40]. We irradiated optical pulses with 800 nm wavelength at GaAs photoconductive antenna for a single-cycle terahertz pulse generation. With a series of parabolic mirrors, terahertz pulse is focused with ∼10 mm² spot area onto nanogap. The transmitted terahertz pulse is guided by two parabolic mirrors and detected by an electro-optic sampling method with a 1 mm-thick (110) oriented ZnTe crystal. Figure 3(a) shows terahertz time traces transmitted through an undoped silicon wafer (black) with a high resistivity (>1000 Ω · cm) as a reference and the 20 nm-wide gap arrays before (red) and after (blue) the BOE etching process. By Fourier-transforming the measured time traces, we converted them to the frequency spectra, which are already normalized by the reference spectrum, as shown in Figure 3(b). It is noted that the resonance frequency of the etched nanogap is slightly blue-shifted because of the reduced thickness of the existing Al₂O₃ layer on top of the metal pattern and in the nanogap (Figure 2(f)) [41, 42]. The field enhancement (FE) at the exit of the gap is extracted from the far-field transmitted amplitude using the Kirchhoff integral formalism [43]. It is following the equation, FE = t/β, where t is the far-field transmitted amplitude and β is the coverage ratio, resulting in a 580-fold enhancement near the gap at the resonance frequency of 0.77 THz (Figure 3(b)). We also used the modal expansion [13, 34, 44] to theoretically predict the terahertz transmission spectra of the nanogap-loop array. The calculated resonance frequency at 0.73 THz is in good agreement with the experimental results. In addition, the calculation revealed that the dip at 0.88 THz is the result of the interaction between the nanogap resonance and the Rayleigh minima in the periodic array of the nanogap-loop [45, 46]. We note that the spectral broadening in experiments is due to the electric field penetration into the Au film near the nanogap and its resultant absorption; meanwhile the modal expansion method employed the perfect electric conductor without absorption. Furthermore, using the high-density nanogap-loop array [36], the total coverage ratio of 0.032% as an active area (optical hotspot) for sensing the viruses was only 2.5 times smaller than that of the THz split-ring resonators (THz-SRR) in the previous work [7], even though the gap size of 20 nm is 10 times smaller than 200 nm in the THz-SRR structure (Figure 3(c)).

Figure 3:

Terahertz time-domain spectroscopy of the gold nanogap-loop array. (a) Time-domain signals passing through a bare silicon substrate (black), the nanogap-loop array before (red), and after (blue) the etching process. (b) Fourier-transformed transmission amplitude and field enhancement spectra of the unetched (red) and the etched (blue) nanogaps. The black dashed curve shows the theoretical calculation using the modal expansion. (c) Schematic diagrams of the coverage ratio (β) of the sensing area of the current nanogap-loop, and the split-ring resonator which was already reported [7].

To demonstrate the detection of the virus, the PRD1 virus with an approximate diameter of 60 nm was dispersed on the nanogap-loop array and the terahertz transmission spectrum was measured by varying the surface density of the virus. To uniformly disperse the virus on the surface of nanogap-loop array, PDMS walls with a cross-section area of 10 mm² were built around the nanogap-loop array (Figure 4(a)). The PRD1 virus solution with a volume of 10 μL was dropped onto the nanogap-loop array with PDMS walls and, subsequently, the dropped virus solution was dried at 40 °C for 15 min. The volume density of the virus solution was fixed at 10⁹/mL, and the surface density of the virus was controlled by the number of the repetitive dispersion with the same amount of the solution volume. To figure out the actual number of viruses in the sample, we converted the volume density to the surface density in the same way as in the previous work [7]. With the surface density of 1.0/μm² and the virus solution covering area of 10 mm², we estimate the molar concentration to ∼20 amol for the entire array, corresponding to only ∼10 zmol per one unit cell. Combined with THz near-field imaging spectroscopy with a few micrometers resolution [47], [48], [49], [50], we should be able to observe the resonance peak shift from even smaller quantities of virus particles down to sub-zeptomole. Furthermore, considering only the actual sensing area, the average number of viruses participating in sensing is estimated to be below 100 (For the details, see the following full-wave numerical simulation results).

Figure 4:

Spectral changes of the virus-coated nanogap-loop array. (a) Schematic diagrams of the drop-castingmethod with the PRD1 viruses. After dropping and drying the PRD1 virus solution inside the PDMS wall, the virus-coated nanogap samples are measured. This cycle (virus coating and THz measurement) is repeated to observe the terahertz spectral changes by increasing the virus density. Terahertz transmitted amplitude spectra for (b) unetched nanogaps and (c) etched nanogaps depending on the selected surface density of the PRD1 virus in the range between 1/μm² and 12/μm². (Insets) Schematic diagrams of the PRD1 virus on the unetched and etched nanogaps. (d) Resonance frequency shift (Δf/f ₀) of each sample as a function of the surface density (number of virus particles/μm²).

Figure 4(b) and (c) show the terahertz transmission amplitude spectra for the unetched and etched nanogaps as increasing the surface density of the PRD1 virus from 1/μm² to 12/μm². We clearly observed redshifts in the resonance peaks of the 20 nm-gap loop array even with the virus density of 1/μm². The resonance peak shifts are because the virus particles, which are tightly fitted into the nanogap, strongly perturbed the resonant mode of the rectangular nanogap-loops [32, 51]. In addition, we verified that the resonance frequency of the transmission amplitude spectrum returned to its original frequency by applying a fungicide treatment [14] with sodium hypochlorite solution which easily removes the PRD1 virus particle from the surface (See the details in Figure S2).

Figure 4(d) shows 3–4% redshifts after coating a virus solution with the surface density of 1/μm², and the resonance frequency shifts (Δf/f ₀) gradually saturate to maximum values of 4.9% and 8.9% for the unetched and etched nanogaps, respectively. This difference originates from the rapidly decaying evanescent field on the surface of the nanogap region, resulting in less sensitivity to the change of the local index of refraction in the unetched nanogap with the 20 nm thick Al₂O₃ layer [32, 52]. Compared to the THz-SRR sensing with a gap width of 200 nm [7], the virus sensing capability of the etched 20 nm nanogap-loop array has been enhanced to 160% of the maximum frequency shift and 400% of the frequency shift per single virus particle. These results indicate that tightly attaching the virus particle to the nanogap region forming the bridge between the two metal surfaces shows better sensitivity compared to the one with the wider gap with the loose-fitting particles [7, 8, 51]. It is worth noting that our new nanogap platform provides a two-dimensional-like optical hotspot, facilitating the detection of ultrasmall quantity of viruses (Figure 4(c)). The opening of the two-dimensional-like (2D-like) optical hotspot allows us to access the resonance mode by effectively deforming the electric field on the surface of the nanogap, resulting in the shift of the resonance frequency. Furthermore, using advanced optical or electrical tweezing techniques [53], [54], [55], it is possible to place viruses in the desired nanogap region and enhance the virus detection sensitivity in sub-zeptomole range. In the current drop-drying method, the virus particles were randomly distributed on the nanogap-loop array including the weak electric field area. However, the tweezing techniques may allow us to place the virus in the nanogap region with strong electric fields which facilitates sub-zeptomolar detection.

To elucidate the sensitivity enhancement in the nanogap-loop with ultrasmall amount of the virus particles, we performed the full-wave numerical simulations by using the Finite-element method (See the details for optical simulations in the method). The tightly attached virus particles in the etched nanogap region contact with three faces; the top surface of Al₂O₃ gap-filling material and the two slanted gap surfaces of the metal (Figure 4(c)). To consider such a structure, we modeled the spherical shape virus to a hexahedron shape with a bump fit to the nanogap (Figure 5(a)). The refractive index of the virus particle was set to 1.87 + 0.13i from the earlier result [7]. Figure 5(b) shows the calculated transmission amplitude spectrum with varying the density of the virus particles and the resultant redshifts of the resonance frequency with the increase of the virus density. With the fully covered case with the maximum density, the resonance frequency shift is saturated as shown in the experiments. It is emphasized that although the virus particles were uniformly drop-cast over the whole area, the actual amount of virus particles participating might only be a few hundreds near the nanogap-loop.

Figure 5:

Full-wave numerical simulations. (a) Schematic illustrations of the simulation modeling. The spherical shape of the PRD1 virus has been changed to a T-shape in order to minimize the computing time of the simulation. (b) The transmission amplitude spectra of the 20 nm gap depending on their PRD1 virus densities. Here the density of the virus particle is defined by the number of viruses per unit length in the nanogap-loop.

We verified this noticeable result analytically by placing a layer with a refractive index of 1.87 + 0.13i, as same as that of viruses, over the nanogap (See the details in Supporting Information). By the virus layer, the analytically calculated resonance frequency shift was 30 GHz and transmission amplitude decreased by 12.5% (Figure S3). Although the partial refractive index change inside the nanogap (Δn ₂ ) was not considered in the analytic calculation, we emphasize that there is not a big difference between the numerical simulations with the virus particles and the analytic calculations with the virus layer, indicating that the local index changes near the 2D-like optical hotspot mostly determines the change of the resonant transmission spectrum.

3 Conclusions

We fabricated virus-sized gold nanogaps with a gap width of 20 nm using the atomic layer lithography and implemented an ultra-sensitive virus detection platform working in the THz frequency region. The large-area nanogap-loop array allowed a 580-fold field enhancement, enabling extreme light–matter interactions near the gap. Through the etching process, it was possible to remove the existing Al₂O₃ layer on top of the first metal pattern and part of the inner layer between the two metals, wherein the virus particle can be tightly attached to the nanogap as bridging the gap between the two metals [56]. We consequently observed a factor of two increase in the redshift of the resonant peak compared to the unetched gap sample, leading to the giant sensitivity enhancement of 400% of the resonance frequency shift per single virus particle compared to the previous work [7]. In full-3D simulations and modal expansions, the results were in good agreement with the results of the experiments and showed that the refractive index change near the nanogap, forming an optical hotspot area in two dimensions, largely determined the resonant transmission spectrum. Further sensitivity could be improved by matching the nanogap-loops to one of the viruses’ absorption frequencies, associated with their intermolecular vibrational modes, and fingerprinting of the target virus [13, 55, 57, 58]. With our high-sensitivity, rapid-detection virus sensing technique, we offer a new method of mass-screening for deadly and contagious viruses, including COVID-19.

4 Methods

Optical simulations. The full-wave optical simulations were performed to analyze an ultrasmall quantity virus detection in the nanogap-loop array by using the finite-element method (COMSOL Multiphysics, RF module). The nanogap-loop array was formed in the 100 nm thick gold film on a SiO₂-coated silicon substrate. The lengths of the loop nanogap in the x- and y-direction were 10 and 40 μm, respectively. The gap width was 20 nm. The pitches of the array were 50 and 100 μm in the x- and y-direction, respectively. The gap region with a height of 80 nm from the substrate was filled with Al₂O₃, and the rest of the gap region was filled with air for modeling the etched and wider gap region. The refractive indices of Al₂O₃, SiO_2, and Si in the terahertz region were 2.35, 2.1, and 3.4, respectively, and the dielectric constant of gold was estimated by the Drude model, ε(ω) = 1 – ω _p ²/(ω ²+iωγ _p) with ω _p of 2730 THz and γ _p of 19.4 THz. The x-polarized plane wave was impinging from the substrate, and the transmitted amplitude was measured from the air.

To examine the virus detection, the hexahedron shape virus particles with a bump fit the nanogap were placed on top of the gap region. The width, height, and depth of the virus particles were 60, 60, and 40 nm, respectively (Figure 5(a)). In addition, the air-filled gap region was filled with the bump in order to model the sphere-shaped virus particle embedded into the slanted gap region. The width, height, and depth of the bump were 20, 60, and 20 nm, respectively. To analyze the resonance peak shift with varying the density of the virus particle, the virus particles were periodically placed along the loop of the nanogap, with the inter-distance of 0, 100, 200, and 1000 nm. The refractive index of virus material was 1.87 + 0.13i [7].

The different length scales in the gap width (20 nm), the length of nanoresonators (40 μm), and the wavelength (600 μm at 0.5 THz) demand huge amounts of computational resources. To resolve it, we applied a multi-scale meshing method based on the physical origin of the resonance.