Virus recognition with terahertz radiation: drawbacks and potentialities

A virus is a pathogenic agent that is able to replicate only inside host living cells, parasitizing the cellular biosynthetic system. Its diameter typically ranges between ∼20 and ∼300 nm [1–3]. When a host cell is infected, an avalanche effect arises because the infecting virus forces the replication of its genetic material. Virus infection can spread in many ways, for instance by coughing and sneezing [4, 5], by ingesting infected food and/or water [6, 7], by sexual contact [8–10], by punctures of blood-sucking insects [11] etc. Therefore, rapid and effective techniques for virus sensing and recognition are becoming of vital importance. A fast and accurate monitoring can help to prevent wide transmission, spreading of viral infection and to carry out timely therapeutic treatment, disease control and surveillance. Nowadays, this is a crucial point, especially during the Covid-19 pandemic situation that is affecting our lives. This emergency, in fact, has imposed a severe burden on society and has led to the urgent need of effective virus diagnostic methods. Actually, many molecular and biological techniques are conventionally used. They include reverse transcription polymerase chain reaction (RT-PCR) [6, 12], reverse transcription loop-mediated isothermal amplification (RT-LAMP) [13, 14] and CRISPR–Cas13-based SHERLOCK system [15, 16]. In addition to those techniques, microfluidic biosensors are also considered [17–20]. In box 1 we report a brief summary concerning the most conventional molecular virus-detection techniques.

However, the above mentioned assays are time-consuming, labor-intensive and no reagent-free [21, 25]. Therefore, a complementary, fast, sensitive and non invasive alternative is highly desirable. In the following sections, an overview concerning the most promising terahertz (THz) technology approaches virus monitoring and detection will be presented and discussed.

In the last decades, technology has experienced a wide diffusion of customizable and flexible THz systems for spectroscopic and imaging applications [34–38]. These systems found challenging applications in many science fields, like biology and medicine [37, 39–41], gas sensing [42, 43], material science [44, 45], chemical analysis and identification [46, 47] etc. THz radiation is largely appealing for biomedical issues because its low photon energy (4 meV @ 1 THz [48]) and its high penetration in anhydrous targets enable nondestructive and non-ionizing sensing. This is in contrast with other optical techniques, including ultraviolet or x-rays, where the high energy photons ($\gt\gt$eV) can cause direct biological damage on the sample [48–50]. In addition, many rotational and vibrational energy levels are located within the THz spectral range [48, 51] providing chemical specificity to imaging experiments, which can be efficiently done in label-free and non-contact modes [34, 37, 52]. However, THz radiation suffers from poor spatial resolution due to its large wavelength (λ = 300 µm @ 1 THz) [53]. Detecting tiny microorganisms such as bacteria and viruses is very challenging because their lateral dimensions are strongly sub-diffraction [25]. In addition, the extreme sensitivity of THz radiation to polar molecules, specifically water, limits THz waves penetrability from tens to hundred microns in hydrated samples, preventing wide technological spread in biological fields [35, 48, 51, 52]. Regardless of these constraints, several THz investigations have been performed on biological materials [39, 52, 54]. For instance, Lee et al [50] exploited THz spectroscopy for measuring THz absorbance of H9N2 virus sample, in the spectral range from 0.2 to 2 THz. As it can be seen in figure 1 the freeze-dried pellet virus sample absorption is practically identical to the substrate, showing no identifiable spectral features.

Zoom In Zoom Out Reset image size

Authors assigned the case to superposition of many protein vibration modes and inhomogeneous broadening of absorption features [50]. This suggests that the weak THz detection sensitivity and low chemical specificity are the main obstacles for THz sensing applications in pathogenic monitoring.

Interestingly, Sun and coauthors [55] used their THz-time domain spectroscopy (THz-TDS) system in transmission mode to assess an indirect virus detection modality. They investigated the binding properties of H9 HA protein and its specific human monoclonal antibody F10. H9 HA protein is the main surface glycoprotein of the Avian influenza (AI) virus and F10 is its specific antibody. Sun and the authors demonstrated that THz spectroscopy is sensitive to the antibody-antigen reaction, by looking at the changes in the hydration shell around H9 HA molecules, in absence or in presence of F10. They measured the absorption coefficient for three different samples in liquid phase (∼µm thickness): H9 HA, H9 HA with F10 (specific antibody) and H9 HA with irmAb (non specific antibody). Figure 2 shows the absorption coefficient vs ratio of binding sites (that takes into account both protein and antibody concentrations) for the three samples [55]. The absorption coefficients for H9 HA and H9 HA-irmAb show a non monotonous behavior indicating a hydration shell formation. The H9 HA-F10 sample, instead, present a monotonous behavior with no evidence of hydration. This trend highlights the consequence of binding sites of the antibody-antigen reaction on the protein configuration. Thus, using the capabilities of THz spectroscopy for probing antibody-antigen binding, the optical properties provide a sensitive approach to monitor the H9 HA-F10 interaction. The concentration detection limit was 15 µg ml⁻¹, compared to 113 µg ml⁻¹, determined by ELISA assay. THz spectroscopy approach resulted then ∼7.5-fold more sensitive compared to standard ELISA [32, 33].

Zoom In Zoom Out Reset image size

Although a more extensive research should be performed, previous results [50], suggest that a direct virus THz detection is very difficult due to the virus low THz absorption coefficient and the absence of specific THz spectral features. On the contrary, indirect methods can be successfully used to investigate antibody-antigen reactions, as reported by Sun and coauthors [55]. Those results also suggest that efficient devices, for example based on graphene [56] or on micro-fluidic chips [57], and novel materials like meta- and nano-materials should be developed for increasing sensitivity and chemical specificity of THz radiation in this vibrating research field [56]. As reported by Zhou et al [56], graphene exhibits very fascinating electrical and optical properties, with the possibility to increase the sensitivity by graphene plasmonics and also it can be efficiently used for detecting biomolecules. Moreover, micro-fluidic chips [57] are of great interest because they allow to use extremely small liquid volumes, limiting the strong water absorption in the THz range. In addition, molecules can be trapped within a very small region in the microfluidic channels, providing very concentrated measurements.

Metamaterial based chips are artificially structured devices with a wide variety of metallic structures with sub-wavelength [56, 58–62]. They gained popularity because of their operational simplicity, compactness and point-of-care suitability. They show attractive electromagnetic properties, mostly related to the excitations of surface plasmon polaritons (SPPs) [63–65], including the capability to extremely localize and enhance the electric field associated to the incoming radiation [50]. Sample deposition on metamaterial devices produces a variation in their dielectric properties, changing the SPPs propagation and interaction with the electromagnetic radiation [64]. In particular, THz metamaterials chips may have single- or multi-resonance frequencies $f_0^{\,\,n}$ exploiting their frequency shift Δf due to sample deposition. $f_0^{\,\,n}$ strictly depend on the unit cell geometry and on the substrate refractive index [25, 50, 66–68]. The local dielectric changes caused by virus, proteins, nucleic acids and other biological samples can be efficiently detected by using metamaterials. Indeed, as very thin water layers (∼20 µm) are needed, the limitations imposed by the strong water absorption coefficient in the THz range can be easily overcome [69].

Sample discrimination takes into account (1) the resonance frequency shift Δf and (2) the relative variation of optical response (OR) (typically transmittance, absorbance or reflectance) at the resonance frequencies $f_0^{\,\,n}$. This variation is generically labeled with $\Delta OR = OR_{SUB}-OR_{SAMP}$, where OR_SUB and OR_SAMP are the substrate and sample optical responses at the initial and final (after the interaction) resonance frequencies, respectively [25, 50].

The sensitivity of these metamaterial devices is commonly defined through:

(a) in refractive index n unit (RIU) [70]

$$\begin{equation} S_1 = \frac{\Delta f}{\Delta n} \left[\frac{GHz}{RIU}\right], \end{equation} \tag{ 1 }$$

(b) in virus surface number density unit $\displaystyle \left(N_{av} = \frac{N_{virus}}{AREA}\right)$ [25, 69]

$$\begin{equation} S_2 = \frac{\Delta f}{N_{av}} \left[\frac{GHz\cdot\mu {\text{m}}^2}{particle}\right].\end{equation} \tag{ 2 }$$

Being the sensitivity in equations (1) and (2) proportional to Δf, an appropriate design of the metamaterial unit cell area is crucial to enhance the detection capability, shifting the resonant peaks to the desired modes [71]. Many shapes and substrate refractive indexes have been proposed to efficiently fabricate these devices. For instance, periodically repeated split-rings resonators [25], rectangular slots [50], planar Jerusalem crosses [64], H-shaped graphene resonators [70] have been proposed. In figure 3, a schematic picture of the metamaterial unit cell used in [25, 50, 64, 70] is reported.

Zoom In Zoom Out Reset image size

Recently, nanostructures such as nanogaps have been embedded in metamaterials by nanolithography. Researchers demonstrated that the reduction of the substrate refractive index and the gap width leads to an improvement of sensitivity [69, 72]. However, the sensor specification can be only obtained with functionalization: the biomolecules and/or other biological components are anchored on the meta- and nano-materials. For example, functionalized metamaterials [73–75] with alkanethiol molecules of well-ordered covalent bonded monolayers, or surface chemical modification using silane and silanol chemistries [76–78], CO_x -H modification, anchoring and/or decorating with antigens [79], have been proposed. Moreover, gold nanoparticles (GNPs) have been used, in the THz spectral range, also in combination with metamaterial chips [56, 80]. Ahmadivand et al [80] exploited GNPs to effectively detect ZIKV-envelope proteins at low concentration, with a 100-fold enhanced sensitivity.

Hong et al [69] proposed their hybrid slot antenna with silver mono-dimensional nanowires. They demonstrated a 2.5-fold sensitivity enhancement, compared to the one without silver nanowires on it.

All these advances in material technology set the scene for extremely promising performance of THz biosensors.

Park et al [25] fabricated THz split-ring resonators (see figure 3(a)) by e-beam lithography and photolithography. Their metamaterials were prepared both on quartz and on silicon substrate, followed by metal evaporation of Cr (3 nm) and Au (97 nm). They used them for detecting small MS2 and PRD1 viruses (30 and 60 nm, respectively), at low density. They analyzed a 40 µm thick layer for both viruses and obtained information on the viruses dielectric constant (_r ), lacking in literature.

Successively, the same group performed measurements on both viruses, changing their density and the metamaterial gap widths. They demonstrated that the resonance frequency shift increases with virus surface density until saturation and the sensitivity is higher for smaller metamaterial gap widths (200 nm instead of 3 µm).

Lee et al [50] fabricated rectangular nano-antennas (see figure 3(b)) by e-beam lithography, using a silicon wafer patterned with gold (150 nm). They tested their device with periodic rectangular slots for sensing AI virus subtypes: H1N1, H5N2 and H9N2. Authors demonstrated that each virus sample produces a different transmission change and resonance frequency shift, with respect to substrate only. This opens the way to discrimination among different virus types, by looking at ΔOR and Δf. Moreover, Lee et al [50] studied H9N2 transmittance, by varying its concentration. Four different concentration values have been chosen (buffer and three virus increasing concentrations). A linear decrease in transmittance value was found, at the resonance frequency, as the concentration increased (see figure 4).

Zoom In Zoom Out Reset image size

Figure 5 shows resonance frequency (green triangles) and transmittance changes (pink circles) as a function of virus concentration [50]. Note that ΔT linearly changes as a function of sample concentration (see figures 4 and 5), while Δf keeps constant [50].

Zoom In Zoom Out Reset image size

Finally, Hong et al [69] manufactured their hybrid slot antennas on quartz substrate garnishing them with silver nanowires (with 20 nm diameter and 1–5 µm length). They tested their metamaterial on PRD1 virus droplets. Authors observed sensitivity enhancement due to the presence of nanowires. The hybrid chip exhibit a ∼2.5 factor enhanced sensitivity S, if compared to the bare chip (S∼33 GHz·µm²/particle instead of ∼13 GHz·µm²/particle). This value can be further improved by performing multi drop-casting and by changing geometrical factors and used substrate materials.

Some groups [50, 64, 70] performed finite-difference time-domain simulations on various metamaterial unit cell chip geometries. Three AI virus samples (H1N1, H5N2 and H9N2) have been modeled as ∼µm thick layers, composed by homogeneous dielectric clads. Each virus complex refractive index N has been written as follows:

$$\begin{equation} N = \alpha n+i\beta k, \end{equation} \tag{ 3 }$$

where α and β are frequency independent parameters. (α, β) pairs are distinctive of different viruses and their concentrations. n and k are the real and imaginary part of $\tilde{n} = \sqrt{\tilde{\epsilon}}$, respectively [50, 64]. In particular, viruses have been described using the parameters listed in table 1.

Lee and collaborators [50] found a great accordance between experimental and simulated data on the three virus types. H1N1 and H5N2 share the same Δf, while H1N1 and H9N2 the same ΔT. This finding is directly related to the (α, β) values. α is responsible of resonance frequency shift, while β of transmittance changes. Therefore, as H1N1 and H5N2 share the same α, they will experience the same Δf. The same applies for β and ΔT.

Cheng and coauthors [64], with their Planar Jerusalem Cross structures (figure 3(c)) confirmed Lee et al [50] results by varying (α, β) pairs and looking at resonance frequency shifts and absorption changes. They modeled their 5 µm thick samples and the results are reported in figure 6. In figure 6(a) they kept β = 1.4 and varied α between 1.0 and 1.4. They demonstrated that, as α increases, the resonance frequency linearly decreases, therefore the resonance frequency shift ΔF linearly increases, while absorption (Δ A) values remain almost constant (see figure 6(b)). The opposite trend has been verified for β, as shown in figures 6(c) and (d).

Zoom In Zoom Out Reset image size

In figure 7(a) the simulated absorption response for three AI virus types is reported. The viruses are modeled by different (α, β) values, as reported in table 1. The histogram in figure 7(b) shows absorption differences and resonance frequency shifts. Thereby, all these virus models confirm the efficient discrimination by looking at Δ F and Δ A. Finally, more interestingly, Cheng et al [64] simulated how the increasing sample thickness (from 1 to 8 µm) affects the resonance frequency shift. They found a saturation effect (see figures 7(c) and (d)), verifying the saturation-thickness behavior exploited by Park et al [25] in their measurements.

Zoom In Zoom Out Reset image size

The demand for large-scale tools to detect viruses in quick and effective modalities has grown and is continuously increasing, in the light of pandemic events of the last two decades. Although many molecular techniques (RT-PCR, bDNA, RT-LAMP etc) are currently available and successfully used, they are still complex and difficult to apply on a large-scale. In this framework, THz technology has some characteristics that make it suitable for biosensing applications, such as pathogens, complementing or enhancing conventional solutions. This short review emphasizes the possibility to sense and identify viruses with THz radiation. In the first part, we focused on traditional THz sensing, such as THz-TDS, which does not reach a good sensitivity to monitor and/or discriminate the virus presence, unless indirect investigations are preferred. In the second part of this review, we dealt with the effective THz virus sensing using meta- and nano-sensors. Both experiments and simulations were reported to demonstrate the great potentialities of virus detection with metamaterial chips. Different unit cells geometries were studied in order to push over the detection limit. Meta- and nano-materials, operating in the THz regime, are an emerging alternative with a great potential for high-speed and on-site virus detection. The recent technological improvements in manufacturing techniques promise to extend the performances of meta- and nano-sensors, achieving sensitivity values much higher than traditional/conventional tools. Consequently, this research topic, although started merely few years ago, has grown and reached a discrete maturity, and it is destined to hold a relevant position in the future.

No new data were created or analyzed in this study.

We thank for funding the Italian Ministry of Foreign Affairs and International Cooperation (MAECI) in the framework of the Great Relevance Bilateral Project PGR01020 between Italy and China. We also thank BRIC-INAIL project ID12. The work was also supported by the 'Multi-Messenger and Machine Learning Monitoring of SARS-CoV-2 for Occupational Health & Safety' (4M SARS-CoV-2) project, funded by the Special Integrative Fund for Research (FISR), Italian Ministry of University and Research (MUR).