Fast, accurate, point-of-care COVID-19 pandemic diagnosis enabled through advanced lab-on-chip optical biosensors: Opportunities and challenges

The emergence of novel respiratory tract infections in the 21st century has been a growing concern that has turned into a major cause of concern, hospitalization, and death. The growing alert turned into a full-scale, worldwide deadly pandemic in 2020 with SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) following predecessor respiratory pandemics, including SARS (severe acute respiratory syndrome) in 2003, H1N1 influenza (swine flu) in 2009, and MERS (Middle East respiratory syndrome) in 2012. The coronavirus disease 2019 (COVID-19) pandemic and its rapid growth rate have driven an unprecedented worldwide demand for emergency measures to mitigate its fast rate of spread.^1–3 In addition to inevitable policies like social distancing, laborious sterilization measures, and protocols for coping with infected patients, it is paramount to have a system of detection and cure in place for a catastrophic pandemic of which very few humans alive have experienced. It is also well known that pressures of war have always stimulated advances in engineering, science, and medicine. Therefore, the new, invisible battle against SARS-CoV-2 infection can stimulate major breakthroughs in the development of diagnosis and treatment systems. The highly contagious SARS-CoV-2 infection is hard to detect, as patients present with clinically inapparent symptoms, including fever, cough, and shortness of breath.⁴ The worldwide high morbidity and mortality of SARS-CoV-2 plus no guaranteed vaccine or treatments on the horizon as of mid- to late-2020 are a call to action for scientists and researchers to probe various medical interventions. Immediate, cost-effective, point-of-risk measures, including identification, diagnosis, and isolation of the infected individual, are still regarded as the single best viable solution to slow down this epidemic pneumonia.

SARS-CoV is an enveloped, single-stranded RNA virus that exists in humans and animals and is mainly transmitted through aerosols and nearby interpersonal contact.^5,6 RNA viruses like SARS-CoV-2 usually have the length of 2–32 kb, and SARS-CoV-2 possesses the largest genome size of any known RNA virus with a length of 30 kb and an S protein trimer of 600 kDa. Once the virus enters the body, it sticks to primary target cells that provide plenty of virus receptors, the angiotensin-converting enzyme 2 (ACE2).^5,7 Its genome RNA infusion into the cell results in the formation of protein building blocks consisting of spike, envelope, membrane, nucleocapsid, and proteins.^8–10 Thus, relying on ACE2, human SARS-CoV infuses into the target cell for which S glycoprotein trimeric spikes on the surface mediate the entrance into the host cell.¹¹ The S glycoprotein of SARS-CoV is therefore a main target for neutralizing antibodies.¹² Similar SARS-CoV and SARS-CoV-2 amino acid identity in their S proteins makes them prone to have analogous immunogenic surfaces on these antigens.¹³ SARS-CoV-2 demonstrates a complex pattern for receptor recognition that results in its trimeric spike protein attachment to ACE2 receptors on human cells^7,14–18 [Fig. 1(c)]. Attempts to block the infusion of virus have been carried out through targeting mainly the spike protein of SARS-CoV-2 and the receptor binding domain (RBD).¹⁹ Developed antibodies targeting these regions can expand the potency, power, and chance of success against the infusion of SARS-CoV-2 in the host cell. Once the virus enters the body through the cells, it replicates, and virions are then set free to infect new target cells^16,20 [Fig. 1(b)]. SARS infectious virion particles can be easily found in respiratory secretions, saliva, and sweat within the early days of effective infection.^21–26 SARS-CoV-2 infection harms lung tissues, resulting in pneumonia with rapid respiratory deterioration, failure, and death in almost 5% of cases.^27,28 Applying an effective vaccine against SARS-CoV-2 can be done through the S protein and especially the RBD as they induce a highly potent neutralizing antibody to block virus binding and its membrane infusion or form an immunity protective layer against viral infection.⁵ 

The most well-known and commonly used procedure for identifying pathogens like SARS-CoV-2 relies on clinical real-time RT-PCR (reverse transcription polymerase chain reaction).²⁹ The high accuracy and precision of the RT-PCR clinical test at early stages of infection in a symptomatic patient makes it the most reliable method to detect SARS-CoV-2 to date. The test is specifically done for a qualitative analysis of nucleic acid of the pathogen in people who meet SARS-CoV-2 clinical infection signs and symptoms. For infected patients who are asymptomatic or identified as false negative with a viral load below the detection limit of RT-PCR assays, serological tests are the second-best alternative and can be used to further control the virus spread. However, the aforementioned procedures and similar clinical diagnosis all require advanced laboratory testing, expensive equipment, and expertise, which are counterproductive for the purpose of point-of-care (POC) fast detection and hard to be found in low- and middle-income countries (LMICs) that are more prone to suffer from the outbreak.^22,29–32 The increasing number of potentially positive people and the inability to test on a mass scale due to the high-cost and time-consuming nature of clinical diagnosis procedures have imposed the need to develop methods that are more conducive to fast, cost-efficient detection of SARS-CoV-2.

POC biosensors can provide the next best alternative reliable solution to clinical diagnosis with a much faster real-time detection without compromising sensitivity and accuracy.^33,34 The World Health Organization has defined and specified the standards and requirements of a convenient biosensor to be used throughout the world as ASSURED: affordable, sensitive, specific, user-friendly, rapid and robust, equipment free, and deliverable to end-users.^35,36 Compatible with these requirements, there have been many POC nucleic acid biosensor platforms introduced, developed, or modified to detect SARS-CoV-2 by ASSURED standards.^35,37,38 However, a major challenge for all available biosensors is early confirmation of SARS-CoV-2 with a low level of viral RNA at the onset of the infection to avoid further spread.^39,40 Among all the portable platform solutions to this task, optical biosensors (Fig. 2) present a strong conceivable potential to grow expeditiously in healthcare and biomedical fields as they provide a condensed, accurate analytical tool to promote mass-scale screening of a broad range of samples through different parameters.^40–43 The main advantages to optical biosensors are delivering the accuracy of nucleic acid–based clinical tests while eliminating the need to postprocess the extracted sample or collected serum.⁴⁴ As a result, they are very promising to become the future COVID-19 diagnostic tools as they provide sensitive and durable point-of-care testing (POCT) devices, which is imperative for controlling epidemics like SARS-CoV-2.^{39,42,45–48}

Optical biosensing works under different physical transduction principles, such as interferometers, resonators, and plasmonics,⁴⁹ and has been investigated in several studies to monitor many viruses with a good accuracy.^24,45,50 By exploiting the strong light–matter interactions in a POC optical biosensor, an ultrasensitive real-time detection platform for different pathogens, including novel SARS-CoV-2, is achievable. An optical biosensor essentially translates the capture of the virion in a measurable alteration of a light property, such as the refractive index (RI), intensity, or resonance shift, through different methods such as resonators and interferometers (Fig. 2).

A biosensor functioning mechanism, in general, consists of a sample or target analyte that binds to the bioreceptor, after which the transducer helps in converting the biorecognition data into a measurable quantity (Fig. 2). The difficulty in directly detecting biological analytes based on their physical properties has led to labeling techniques in which an additional molecule is attached to immobilized target molecules, viruses, or cells to enhance the quantitative signal.⁴¹ Therefore, the specific immobilized bioreceptors on the chip sensing area will be bound to their targeted labeled pathogens or proteins upon introduction of analytes into the sensing area. Notable examples of labels used in biosensing are dye molecules, fluorescent tags, and enzymes. Various types of bioreceptor–target coupling mechanisms have been demonstrated (Fig. 3), including antibody–antigen binding, enzyme–substrate catalytic reaction, and complementary DNA (cDNA)–DNA hybridization.

These labels require sophisticated reagent selection and modification that in turn can come with the drawback of perturbing the assay and, in some cases, making final detection a challenging task. On top of that, labeling chemistry can be expensive and time consuming. Thus, there have been many intriguing efforts in biosensing systems involving unlabeled or unmodified biomolecules (label-free biosensing)^41,42,51,52 in which native molecular properties, like molecular weight and RI, are utilized for sensing. Label-free detection, on the other hand, has its own shortcomings as well; for instance, it requires low, nonspecific binding⁴⁹ and sufficient signal to be generated upon target binding.⁴² However, its benefits, such as providing real-time analysis by simplifying assays and reducing time and number of steps required as well as eliminating experimental uncertainty, can compensate its limitations provided that the target analyte concentration and its surface adsorption are sufficient enough to enable detection.⁵³ 

In the case of an optical biosensing platform, the sensing transduction signals are usually based on miniscule changes in refractive index, which itself results from the attachment of biomolecules to the immobilized bioreceptors. The biosensor final sensitivity and specificity are also strongly dependent on the immobilized molecules and the accessibility of target analytes to them. In a label-free biosensor, a highly sensitive biorecognition layer on the transducer surface is therefore of vital importance.^49,54,55 Hence, the optimization of sensing surfaces and their biofunctionalization strategies becomes a significant element for any accurate, sensitive, label-free biosensor.⁴⁹ However, the diverse range of target molecules and biosensor applications makes obtaining a universal surface biofunctionalization procedure extremely difficult; consequently, the procedure needs to be custom designed. For instance, in graphene-based field effect biosensing (FEB), the surface is functionalized for protein immobilization with anti-Zika NS1 mouse monoclonal antibody (mAb) 6B1 developed by the Centers for Disease Control and Prevention.⁵⁶ Polyethylene glycol (PEG) has also been studied for removing proteins from the surface of silicon-based biosensors as it provides a stable and anti-absorptive block against undesired, nonspecific interactions.^57,58 Putting the difficulty of abundant surface functionalization methods aside, the main challenge for label-free biosensing compared to label-based biosensing is achieving the desired sensitivity and limit of detection (LOD) without increasing the target analyte concentration. For detecting pathogens like SARS-CoV-2, a variety of factors can determine the effectiveness of labeled vs nonlabeled procedures.

All in all, for a highly sensitive optical biosensor suitable for detecting SARS-CoV-2, labeled or label free, the limit of detection is the determining factor to have a reliable POC substituting the lab-based detection methods. In Secs. II–V, we discuss the most sensitive optical biosensors that can be used as a platform for a highly sensitive fast point-of-care detection system along with the latest advancement as well as associated challenges.

Due to their immunity to electromagnetic interference (EMI), compactness, and high selectivity, optical biosensors have attracted special attention as a biosensing system.^59,60 The interaction of target molecules with bioreceptors on the surface leads to an effective index and absorption coefficient change (Fig. 4). The effective index change is a function of the concentration of biological or chemical targets on the surface.

Effective mode index change for a perturbed waveguide can be calculated through the variation method:⁵² 

where $E(x,y)$ is the electric field $; Z0$ is the free space impedance; $P$ represents the light wave power, and $nc and nm$ are the refractive indexes of the aqueous solution without analyte and molecular adsorption layer, respectively.

Sensitivity (S) and LOD are the two main criteria for evaluating the performance of the optical biosensor, which in turn depends on the strength of the interaction between the substance and light in a solution or on the surface.⁶⁰ Here, we further investigate the most well-defined biophotonic-sensing mechanisms based on interferometer and resonance shift in microcavities.

Integrated interferometer photonics is one of the most practical architectures for sensing applications. It is based on splitting the input beam into two arms through a Y junction: One arm is completely retained as the reference arm, and the other arm contains the target. The interaction of electromagnetic waves on the sensing arm will cause a phase difference with respect to the reference arm, and recombination of the two beams in the output will cause constructive or destructive interference [see Fig. 5(a)].

Output intensity $IOut$ of the Mach–Zehnder interferometer (MZI) is described as follows:^48,61,62

where $Iref$ and $Isen$ are light intensity in reference and sensing arms, and $φ0$ is the initial phase difference between two arms without external perturbation. The sensitivity of the MZI-based sensor is related to the phase sensitivity relative to the length of the sensor arm,

In an imbalanced MZI, considering the phase matching condition, vis-à-vis the wavelength sensitivity, we can approximate the phase sensitivity of the MZI-based sensor,

where $Δλ$ is the free spectral range (FSR) and $SFSR$ is the spectral sensitivity.

Chemical and biosensing via MZIs have been widely exploited,⁶³ and judicious design of imbalanced interferometers based on waveguides and photonic crystal waveguides (PCWs) have been explored by a lot of research groups.^42,64,65 Figure 5(b) shows an MZI biosensor with a silicon nitride strip waveguide as the reference arm and a silicon nitride slot waveguide as the sensing arm. Sensitivity of as low as 18.9 fg/ml (fM) and bulk refractive index sensitivity of 1864 π/refractive index units (RIU) are obtained through increasing light–analyte interaction in the slot waveguide-sensing region thanks to the low refractive index subwavelength-sized slot regions. Liu et al.²⁵ designed a slot waveguide in the sensing arm to maximize the overlap between light and target analyte [see Fig. 5(c)]. They reported a wide range of linear tests with concentrations ranging from 19 fM to 190 nM (R² = 0.979177). The asymmetry of the MZI array is used to detect microRNA (miRNA) in human urine samples by measurement of light phase change due to attachment of complementary DNA capture probe and the target miRNA. Utilizing slot waveguide as the sensing arm plus the elimination of the tedious incubation process make it an integrated, accurate, and sensitive POC optical biosensor. Chalyan et al.⁶⁶ demonstrated detection of aflatoxin M1 through asymmetric Mach–Zehnder interferometer surface functionalized with antibody fragments. They demonstrated a high volumetric sensitivity of 10⁴ rad/RIU, leading to a LOD <5 × 10⁻⁷ RIU enabled through surface functionalization of the biosensor [Fig. 5(d)].

As is shown in Fig. 5(e), Zhao et al.⁶⁷ demonstrated a polymer-based lateral bimodal interferometer. They used two transverse modes in the waveguide to create the interferometer and short wavelength of 890 nm to reduce the absorption loss from the aqueous solution to be sensed. The overall sensitivity of the manufactured interferometer sensor with a sensing length of 5 mm is reported to be 316π rad/RIU, and the extinction ratio can reach 18 dB. However, reliability of the polymer needs to be further studied. An analogous platform with a laser-based bimodal waveguide interferometer is also proposed to detect COVID-19 via changes in the sensor's evanescent light field [Fig. 5(f)]. This nanophotonic POC device can directly examine body respiratory fluids (e.g., saliva) and does show a sensitivity of attomolar (aM) level for direct, specific miRNA targeting.

In contrast to waveguide-based sensors that rely on light wave absorption, resonant displacement in functionalized microcavities provides a wide range of ultrasensitive optical biosensors.^68,69 The magnitude of binding is determined by de Feijter's formula,⁷⁰ which relates the absolute quantity of adsorbed molecules M with the change in the refractive index as

where d_A is the thickness of the adsorbed layer, n_A is the refractive index of adsorbed molecules, n_c is the refractive index of the cover solution, and dn/dt is the change in the refractive index of molecules, which is proportional to the shift dλ in position of the resonance peak. The size of the resonance wavelength shift is proportional to the number of adsorbed biomolecules, thus providing a label-free method to quantitatively determine the target analyte.

Although high-quality (Q) ring resonators can be achieved with a larger radius, the trade-off between the Q and the FSR limits the radius for a given FSR, which should be large enough for effective recognition of the sensing signal from the adjacent interference signals or for large-scale on-chip multiplexing sensing applications. Wang et al.^71,72 proved through experiments that the Q-enhanced subwavelength grating waveguide-based metamaterial ring resonator (SWGMR) was, in particular, designed using a trapezoidal silicon column (T-SWGMR) (Fig. 6). Contrasted with conventional rectangular silicon pillars comprising SWGMRs (R-SWGMRs), an asymmetric effective refractive index distribution is created, which can significantly reduce bending loss and thus increase the Q of SWGMRs.

The experimental results show that the applicable Q value of T-SWGMR with a radius of 5 μm is as high as 11 500, which is 4.6 times the Q value (∼2800) provided by R-SWGMR with the same radius, indicating that the propagation loss is reduced by 81.4%. To go one step farther, Hai et al.⁷³ proposed a T-SWGMR biosensor and demonstrated the unique stable surface sensing characteristics through a demonstration of miRNA detection at a concentration of 1 nm [Fig. 6(b)].

In addition to utilizing the unique stable sensing characteristics of SWGMR and the enhanced Q of T-SWGMR, Chang et al.⁷⁴ showed a pedestal T-SWGMR biosensor that maximizes the mode volume overlap by implementing an asymmetric refractive index distribution along the vertical direction on the silicon-on-insulator (SOI) platform, thereby further improving sensitivity [Fig. 6(d)]. Both theoretic analysis and experimental proofs show that the volume sensitivity and surface sensitivity have been significantly increased by 28.8% and 1000 times, respectively. For streptavidin, a spectrometer with a resolution of 0.01 nm is used, and its LOD is ∼400 fM. Owing to an imperfect manufacturing process, experimental Q estimate of T-SWGMR with a radius of 10 μm and FSR of ∼13 nm is 1800. The optimized SWGMR with symmetric coupling demonstrated by Huang et al.⁷⁵ estimated Q to be 9800.

Microtoroids are resonators with a Q of >10⁸ and a small mode volume that can be fabricated on silicon using standard microelectronics techniques.⁷⁶ However, microtoroids need to be strictly aligned with the tapered fiber waveguide to achieve high coupling and cannot meet our needs for high-throughput multiplexing sensing. Armani et al.⁷⁶ demonstrated the possibility of detecting unlabeled single molecules and higher concentrations on a single platform [Figs. 7(a) and 7(b)]. According to reports, the quality of performing planar lithography is ∼1.83 × 10⁸. The authors reported that by using the interleukin 2 (IL-2) solution, the microring sensor can provide a dose response of 10⁻¹⁹–10⁻⁶ M and a working range of 5 aM to 1 μM. In another report for a microtoroid with a diameter of 90 μm, the authors reported⁷⁷ that a measurement lifetime of 43 ns corresponds to an inherent quality factor of 1.25 × 10⁸ [Figs. 7(c) and 7(d)].

For microcavity photonic crystal (PhC) sensors, we mainly use 2D PhC biosensors, which have the advantages of design flexibility, compact size (surface area of about a few square micrometers), and strong light interaction with the analyte of interest. As shown in Figs. 8(a) and 8(b), Yang et al. recently confirmed the work of practical pancreatic cancer detection by using nanopore-assisted high-Q (22 000) and high-S (112 nm/RIU) L13 PhC cavities.⁷⁸ The detection results showed that a concentration of 0.334 pg/mL (8.8 fM) pancreatic cancer biomarker was successfully detected in patient plasma samples, which is 50 times more diluted than conventional enzyme-linked immunosorbent assay (ELISA). To go farther, by designing and developing multimode interference (MMI) separators,^52,79–83 high-throughput and multiplexed biosensor arrays have been proposed and demonstrated [Figs. 8(a) and 8(b)].^52,79–81 The integrated scheme and array methods proposed and proven improve the multiparameter and multifunction detection capabilities of the sensor and can be used in practical diagnostic applications.

Our team has extensively described the functionalization of silicon surfaces using various probe biomarkers⁵² and their use in the detection of specific conjugated biomarkers using our silicon photonic crystal microarray structure. Previously, we demonstrated a multimode interference coupler architecture [Figs. 9(a) and 9(b)] that shows the series and parallel integration of 64 sensors on the silicon chip. Multiplexed sensing with specificity of lung cancer cell line lysates was demonstrated. We have also demonstrated experimentally that the silicon photonic crystal sensor chips can be fabricated in a commercial foundry for high-volume manufacturing. For COVID-19 testing, the silicon chip manufacturing process and sensor functionalization process will be the same as before.

Label-free microarrays are particularly exciting because they simplify biochemistry significantly when probe–target binding conjugations can be studied without steric hindrance associated with fluorescent or radioactive tags. In Fig. 10, we compare our photonic crystal microarray approach with other research performed using PhCs and show that our microarray has the highest sensitivity to small changes in concentration. It summarizes sensitivities and detection limits demonstrated in our system compared to other label-free methods, including surface plasmon resonance (SPR), optofluidic ring resonator (OFRR), ring resonator (RR), and PhC devices, as a function of sensing area. Sensitivities of PhC microcavity structures demonstrated at Omega Optics (OO) and The University of Texas (UT), Austin, are shown. It is important to note that the PhC sensors with the lowest detection limit of 67 fg/ml (1 fM) was the result of several carefully optimized processes. First, the L13 cavity with high group index was selected based on a series of comparisons, such as with L21, L55 cavities, and optimized (coupling length between the cavity and the guiding waveguide and the waveguide width) to achieve high Q up to ∼10⁴ in experiments. Furthermore, low refractive index modulators (air holes here) were optimized (mainly the radius) and added to the cavity area to enhance the light–matter interaction and, thus, the sensitivities (because of high filling factor). Thus, the minimum detectable concentration could be obtained from the optimal dependence on Q, group index, and filling factor. In fact, compared with the various SPR- or RR-based sensors, the most important advantages of PC sensors are the minimized sensing areas and the ultra-high-Q performances, which significantly contribute to the practical applications of point-of-care biosensors, fulfilling the requirements of miniaturization, microdosage, and low detection limit. It is worth mentioning that as far as the design and fabrication complexities are concerned, the SPR sensors are the simplest, and the RR sensors are simpler than PhC sensors.

Our photonic crystal microcavity not only has high sensitivity and low detection limit but also can achieve dense integration of sensors due to its small geometric size. In Table I, we compare our proposed PhC microarray platform with other optical biosensing methods capable of detecting SARS-CoV-2. We compared the limit of detection advantages of our proposed platform and show that our platform can provide comparable high sensitivity to other methods. The relatively big size of the S protein trimer in SARS-CoV-2 compared to other detected analytes in these works dictate the applicability of the Table I methods as COVID-19 biosensors.^{64,74,78,84,86–90,92–94,96,98–100}

A plasmon can be described as a collective oscillation of a free electrons or a quantum of plasma oscillation. Propagation of electromagnetic waves along the surface of a metallic surface or surface plasmons (SPs) can be understood as a strong interaction between conduction electrons of the metallic surface and electromagnetic waves, which leads to resonance modes trapped on the surface, also known as SPRs.^24,45,50 SPR propagation along the conductor surface produces a charge density distribution, which enhances the light–matter interaction on the nanoscale. Such enhancements, so-called “hot spots,” that occur at the interface between a dielectric and metallic surface offer the higher sensitivity for plasmonic biosensors. Many types of optical biosensors based on plasmonic platforms have been studied as fascinating candidates for biomedical and chemical sensors.^{26,33,51,102–112} The selectivity of the plasmonic-based biosensor can be achieved by using immobilization of the various bioreceptors. Depending on the target analytes, specific bioreceptors can be selected to be immobilized on the surface of the sensor and react or bind only to its counterparts. Based on the device configurations, plasmonic biosensors can be divided into two groups: SPRs and localized SPRs (LSPRs).

Fundamentally, when the phase matching condition between the incident light and the SP wave guided along the metal/dielectric interface is reached, the incident light can be coupled to the surface-guided mode. Note that the resonance condition between the incident light and the conductive electrons at the metal–dielectric interface with a fixed angle of incidence is only achieved at a specific wavelength. The guided light will be absorbed by the conducting electrons that resonate, which will significantly reduce the reflected light at that particular wavelength. Therefore, once the target molecule is attached to the functionalized metal film, the refractive index changes, causing a shift in the resonance wavelength. Consequently, SPR angle alteration can be characterized as the main sensing mechanism. Several coupling methods have been proposed, including a grating coupler, a waveguide coupler, and a prism coupler, but the prism coupling method has been used as a standard configuration based on the Kreichman configuration.¹¹² Figure 11 shows the schematic of the conventional SPR sensor configuration.

The wavevector of the evanescent field of the incident electromagnetic wave propagating along the prism–metal interface is shown in the following equation:^102,112

where $np$ is the refractive index of the prism, $λ$ is the wavelength of the incident light, and $θ$ is the incident angle.

The wavevector of the SP wave propagating along the metal/dielectric interface is as follows:^113,114

where $ω$ is the angular frequency of the wave, c is the speed of light in vacuum, and $nM$ and $nD$ are the refractive indices of the metal and dielectric, respectively. As mentioned before, the resonance condition is met when $kin=kSP$⁠, so we can calculate the SPR angle in the following equation:

The sensitivity of the SPR devices is determined by the resonance shift with respect to the change of the refractive in the absence and presence of the target analyte,^113–116 

where $Δλ$ is the resonance wavelength shift and $Δn$ is the change of bulk refractive index, including the target analyte.

On the other hand, nanostructures in conductive thin films are among the essential building blocks of LSPR plasmonic biosensors (see Fig. 12). These nanoscale geometric/periodic lattice factors bring huge advantages over conventional SPR devices. Contrasted with SPR occurring along the propagation surface, the attenuation length of the local electromagnetic field is much shorter. These strict restrictions, with a shorter subwavelength structure, can achieve ultra-low-mode volume resonance, making it sensitive to environmental refractive index changes, which are particularly helpful for the detection of tiny biological molecules. Also, an incident light can be directly coupled to the SP wave on the conductive structures without any external couplers, e.g., prisms or gratings, which ameliorates the complexity of the entire system and enables sensor miniaturization¹¹⁷ and absorbance, transmittance, and reflectance-based sensing.^{43,106–108,118–120}

Moreover, LSPRs can be utilized for various types of resonance modes and detection methods, including surface-enhanced Raman spectroscopy (SER),^111,112 photoluminescence/fluorescence,^55,110 and mid-infrared spectroscopy,¹⁰⁵ by tuning the resonance wavelength for a specific light–matter interaction. Table II shows the comprehensive comparison between conventional SPR and LSRP biosensors.

Figure 12 shows the basic configuration of the LSPR device based on resonance shift sensing. The metal nanostructure on the dielectric substrate is used as a resonator, and due to the aforementioned advantages, the sensitivity can be further improved compared with the conventional SPR resonance shift device. The sensitivity and the figure of merit (FOM) of LSPR resonance shift sensors follows the same definition of SPR as in Eq. (10). Another important performance factor of resonance-based sensors is the Q factor, which is defined as^121,122

where $λo$ and FWHM are the wavelength and full-width half maximum of the resonance peak, respectively. To enhance the sensing performance, a higher Q value is desirable because sharper peaks with high Q values are much easier to detect. Considering all these factors, the inherent limit of detection (ILOD) of the resonance displacement sensing device can be defined as follows:^69,123

which indicates that both the higher sensitivity (S) and Q factor are required to minimize the limit of detection of the sensors.

Although these so-called hot spots provide higher sensitivity for LSPR biosensors, their performance is greatly limited due to the basic limiting factors of ohmic losses in metal surfaces. In other words, compared to other photonic biosensors, the absorption loss in the conductive nanocavity leads to a low Q value. To enhance the sensing performance (represented by LOD), researchers have achieved a higher sensitivity and Q factor by using advanced materials^142,143 or optimizing the geometry of metamaterials.^136,140

On the other hand, the concept of a plasma perfect absorber (PPA) sensor was introduced to overcome this intrinsic limiting factor.^124–128 Figure 13 shows that the typical configuration of a PPA sensor consists of periodically arranged metallic nanoantennas (metamaterial) on top and a thin metallic “mirror” layer on the bottom separated by a dielectric spacer.¹²⁶ The basic concept is to have a perfect absorbance at the operating wavelength and make a “zero” transmittance by maximizing the metamaterial losses; in other words, the losses provide an advantage in the PPA sensors. In this structure, most of the incident light at the operating wavelength is absorbed by the top nanoantennas operating as a resonator through impedance matching, and the metallic bottom layer acts as a mirror to eliminate the transmittance. As a result, the reflectance of light can be characterized for sensing as in Fig. 13, and the $FOMPPA$ is defined as follows:¹²⁷ 

where $dIλo/I(λo)$ is the relative intensity change of reflected light at a fixed resonance wavelength $λo$⁠, which is induced by a refractive index change $dn(λo)$⁠.

Moreover, it has been shown that the perfect absorption (>99%) of incident light at working wavelength can remain over a wide incident angle and be insensitive to the polarization (TE/TM) of incident light.¹²⁷ 

Another important detecting method using LSPR is surface-enhanced infrared absorption (SEIRA) spectroscopy. Based on molecular absorption spectroscopy and the fundamental vibrational–rotational transitions of chemical bonds in the wavelength of 3–20 μm, mid-infrared (MIR) absorption spectroscopy has been studied vigorously for label-free detection and identification of molecules in the optical sensor domain. In particular, unlike the near-infrared (NIR) wavelength region, the molecular fingerprint region (700–1500 cm⁻¹) in MIR wavelengths contains many absorption bands related to bending and stretching of chemical bonds (such as −C−C−, −C−O−, −C−N−, etc.) that allow the unique identification of biomolecules with high sensitivity and specificity.

Among many types of optical-based molecular absorption spectroscopy platforms, SEIRA spectroscopy for LSPR devices has shown great promise for detecting a thin layer of surface-bound nanomolecules due to its tight confinement of surface plasmons on metallic nanostructures, which can significantly enhance the IR absorption of small molecules. Figure 14 shows the typical configuration of SEIRA spectroscopy using an LSPR sensor. When the plasmonic resonance peak generated by the metallic nanoantennas is matched with the fundamental vibration signatures of chemical bonds in the biomolecules, the coupling of molecular transitions with the LSPR field on the surface allows significant absorption of the corresponding wavelength, so the decrease in transmitted light can be characterized as a sensing result. To obtain more intense IR absorbance, researchers have shown various nanostructures, including nanorod antennas,¹²⁹ coaxial nanogaps,¹³⁰ and nanocavities,^131–133 to enhance the optical confinement and field enhancement of MIR light.

Taking all these device structures and detection methods into account, a number of nanostructure designs have been studied and optimized to apply the plasmonic resonance sensor for the detection of various bioanalytes with high sensitivity, which can be potentially applied for SARS-CoV-2 sensing applications. As shown in Fig. 15(a), Lee et al. showed a multiplex biosensor for cancer biomarker detection based on the resonance shift of the LSPR single gold nanoparticles (AuNPs); the selective sensing results with LODs of 91 fM, 94 fM, and 10 fM for the α-fetoprotein (AFP), carcinoembryonic antigen (CEA), and prostate specific antigen (PSA) analytes, respectively, are reported by antibody–antigen binding.⁴³ Haes et al. reported the detection of Alzheimer disease biomarkers from clinical samples [Fig. 15(b)] using surface-confined Ag nanoparticles and sandwich assay; the LOD of < 100 fM for amyloid β-derived diffusible ligand (ADDL) detection with the specific anti-ADDL antibodies is reported by the LSPR-induced wavelength shift.¹⁰⁸ Chen et al. reported multiplex serum cytokine analysis by immunoassay that was enhanced using nanoplasmonic biosensor microarrays [Fig. 15(c)]. Periodically arranged gold nanorod microarray conjugated with corresponding antibodies of each cytokine species [IL-2, IL-4, IL-6, IL-10, (IFN-γ), and (TNF-R)] results in LODs of 5–20 pg/ml from a 1-μl serum sample within 40 min.¹¹⁸ Integration of graphene with an SPR sensor, as reported by Zeng et al., has shown ultrasensitive sensing [Fig. 15(d)].¹³⁴ They reported an LOD of 1 aM for 7.3-kDa 24-mer single-stranded DNA (ssDNA). Moreover, an ultrasensitive SPR sensor based on halloysite nanotubes (HNTs)/MoS₂/black phosphorous (BP) atomic layers on gold films have been introduced by Jia et al., with the angular and phase detection sensitivities up to S_A = 77.1 RIU⁻¹ and S_P = 1.61 × 10⁵ RIU⁻¹, respectively¹³⁵ [Fig. 15(e)]. Most recently, a theorical study of a novel phase modulation plasmonic biosensor working in the NIR region, which can be employed for sensitive detection of SARS-CoV-2 and its S glycoprotein, was proposed using 2D van der Waals heterostructures, including tellurene and carboxyl-functionalized molybdenum disulfide (MoS₂) layers, with transparent indium tin oxide (ITO) film [Fig. 15(f)];¹⁷⁶ the highest detection sensitivity of 8.41 × 10⁴°/RIU and an excellent linear detection range of 0∼301 nM and 0∼67.9 nM for S glycoprotein and SARS-CoV-2 specimens were explained.

Here, we review the most up-to-date advances, especially for the coronavirus sensors in the plasmonic domain, and introduce well-established plasmonic SARS-CoV-2 biosensing systems. Researchers have demonstrated that using SPR/LSPR-based sensors and corresponding binding biological receptors can effectively and selectively detect coronavirus.^26,102,104 Moreover, several researchers have already reported experimental SARS-CoV-2 sensing results as shown in Table III and Fig. 16.^33,120

A recent article reported the dual-functional plasmonic photothermal biosensors for SARS-CoV-2 detection [Fig. 16(a)].³³ The authors demonstrated a highly sensitive, fast, and reliable SARS-CoV-2 virus detection capability by integrating the plasmonic photothermal (PPT) effect and conventional LSPR sensing transduction on a single gold nanoisland (AuNI) chip. The two-dimensional AuNIs functionalized with cDNA receptors (RdRp-COVID-C) can perform a selective detection of the RdRp-COVID through DNA hybridization, and a LOD down to the concentration of 0.22 pM was reported. Ahmadivand et al. demonstrated a femtomolar-level detection of SARS-CoV-2 spike proteins using toroidal plasmonic metasensors [Fig. 16(b)].¹³⁶ Based on the unconventional features of the toroidal metasensors, which can focus down to the incident electromagnetic radiation within a tiny hot spot, these metamolecules support resonances that possess much higher sensitivity to the refractive index perturbations in the surrounding media. Moreover, to improve the binding properties, functionalized colloidal AuNPs conjugated with the respective antibody and captured the S proteins present in the sample. The resonance shifts for diverse concentrations of the S protein is monitored with the LOD of ∼4.2 fmol (0.28 nM). Other research with a similar concept but different structures was reported using nanoplasmonic resonance sensors.¹³⁸ The low-cost nanoplasmonic sensor chips consist of a Au-TiO₂-Au nanocup array that allow observation of the plasmon resonance wavelength and intensity change on the virus-capturing event by transmission light spectroscopy without any external coupling optics. The sensitivity is enhanced by applying a AuNP-enhanced sandwich plasmonic resonance immunoassay method, as shown in Fig. 16(c), and the authors demonstrated the minimum detection of 30 virus particles (vp) in one step within 15 min with specificity.

Furthermore, several researchers reported that the sensitivity and the signal-to-noise ratio (SNR) of conventional ELISA or fluorescence-linked immunosorbent assay (FLISA) tests can be significantly improved by applying the “add-on” plasmonic particles without altering their workflow.^26,55 As the ELISA test is widely used for precise SARS-CoV-2 detection,²⁶ plasmon enhanced ELISA/FLISA tests can be applied to COVID-19 sensing as well.

Although the aforementioned SARS-CoV-2 sensing applications^33,120 have shown great specificity and performance of sensitive and selective specificity and sensitivity up to a few fM, a huge potential for more sensitive, accurate, and fast on-chip sensing with a less complex system still remains in the LSPR biosensor domain due to smaller sensing volume and electromagnetic decay length. For example, the sensing systems in Fig. 16(a) require the prism coupler to couple the incident light into an SPR device with an accurate incident angle. It requires very sensitive alignment of optical devices, which makes the overall system complex and hard to integrate with sources and detectors. However, as described in Table II, the incident light can be coupled into LSPR sensors directly without external couplers, and this normal incident angle can make the alignment easier, in turn mitigating the complexity of the system and making fully integrated on-chip sensing possible. Moreover, due to the capability of sensor miniaturization through LSPR nanostructures, label-free, real-time, and parallel detection with multiple channels with high specificity is achievable. Furthermore, improving the sensitivity by applying advanced materials like graphene and 2D materials (2DMs) has incited a great interest for various optical biosensor applications. For plasmonic biosensors, the ultrasensitive graphene and 2D material–enhanced SPR devices have been reported [Figs. 15(d)–15(f)], and the experimental sensing result with LOD value approaching 1 aM has been shown.^134,139 Accordingly, the LSPR biosensors enhanced with advanced materials are anticipated to enable the possibility of a highly sensitive, accurate, and fast point-of-care lab-on-chip–integrated sensor with unprecedented high sensitivity and limit of detection of up to sub-fM. A detailed discussion of emerging nanomaterials, such as graphene and graphene oxide, for optical biosensors are described in Sec. V.

To develop an accurate estimate of a COVID-19 biosensing functioning mechanism, a simulation model first needs to be designed. Here, in a proposed simulation model, COVID-19 is approximated to be a solid sphere core containing RNA covered with a membrane protein with radii of r₁ and r₂, respectively [Fig. 17(a)].⁵⁰ Thus, the effective RI of the virus is calculated by taking a volume-weighted sum of the two refractive indices,

where n₁ (V₁) and n₂ (V₂) are the total RI of the RNA and the membrane protein volume, respectively. As the RI of the virus is determined mainly by material composition rather than by its geometrical size, η is a constant value for the same kind of virions [η_COVID-19 = 1.25 average value of several measurements of transmission electron microscopy (TEM) pictures¹⁴⁰]. We use the SWGR design to simultaneously take advantage of the enhanced binding surface and strong light–substance interaction. As shown in Figs. 17(c) and 17(e), the energy mode is distributed between the gratings as well. To further improve the SWG waveguide functioning in the subwavelength range, the grating period Λ, the waveguide width w, and the fill factor are designed to be 230 nm, 1.23 μm, and 0.5 μm, respectively. For the SWGR, the radius is set as 5 μm with the corresponding FSR¹⁴¹ of 25 nm at 1550 nm. Here, the simulation system includes a 220-nm silicon top layer with a 3-μm buried oxide (BOX) wafer and a liquid solution, with the refractive index of n_clad of 1.35.⁴⁵ Adopting our previous designs features,^142,143 we optimized a high-Q SWGR by utilizing a trapezoidal (T) silicon pillar and reducing bending loss by ∼50% compared to a conventional rectangular silicon pillar. We therefore set the SW waveguide width to 0.5 μm (correlated with the fundamental TE mode) and studied the effect of the trapezoidal width. Note that to obtain the lowest bending loss of the T-SWG waveguide, we employed the particle swarm method for the optimization process. Three parameters (w, A₁, A₂) are optimized and defined as the width and the tuning factor of the outer and inner filling factor of the SWG, respectively [as shown in the inset of Fig. 17(b)]. Considering the limitations of the design for fabrication, the slot between gratings is preset to be >60 nm; thus, A₁ and A2 are limited to be (1, 2) and [0, (1–60 nm/Λ)/f], respectively. At the same time, to keep the SWG working in the subwavelength regime (Λ ≪ λ / 2n_eff), Λ is safely set to be 230 nm, and f is simply set to be 0.5 with no optimization. Furthermore, to make the SWG waveguide work as a single or few-mode waveguide, the width of the gratings is set to <2 μm. All in all, the ranges of w, A₁, and A₂ are set to be (0.5 μm, 2 μm), (1, 2), and (0, 0.522), respectively. The FOM is defined to achieve the lowest bending loss, with a bend radius of 5 μm. Based on the optimization measures taken, we finally achieved a bending loss as low as 0.0279 dB/cm with the optimized (w, A₁, A₂) = (1.23 μm, 1, 0.522). By adjusting the coupling gap between the insertion SWG waveguide and the designed SWGR, the Q can be as high as ∼50 000 (the resonance at 1557.6 nm) with a broad FSR of 25 nm, as shown in Fig. 17(a). We also optimized the 10-μm-radius SWGR [not shown in Fig. 17(f)], achieving a loaded Q of ∼75 000 (the resonance at 1552.1 nm) with an FSR of 11 nm with (w, A₁, A₂) = (1.23 μm, 1, 0.522), at the same waveguide–ring cross-coupling coefficients. Needless to say, as the quality factor of the ring becomes higher, the fabrication tends to be more challenging. Bulk RI sensitivity [shown in the inset of Fig. 11(f)] in the buffer solution is calculated to be $Sres=ΔλresΔnclad=400 nm/RIU.$ Thus, the instrument detection limit (iDL) can be calculated to be as low as ∼7.5 × 10⁻⁵ RIU. Note that iDL performances can be further improved by exploiting a larger radius ring or by further achieving the critical coupling condition given the predictable higher Q while making a trade-off between the performance and the sensor size or the resonance peak extinction ratio.

To evaluate the specific sensing ability of the proposed device for COVID-19, surface sensing performance is analyzed by considering the device immersed in buffer solution bonded by several surface layers (generated in the sensing preparation process), including the ∼2–3-nm surface oxide layer, ∼10-nm functionalization layer with bonded antibody (protein layer), and virus particles layer, in the detection process. In simulations, the layers generated during the preparation process are further simplified to be a uniform layer (RI = 1.45) with a thickness of 15 nm, and the bonded virus layer is simplified as a uniform layer with a thickness of 125 nm (the maximum diameter of the COVID-19 virus) [Fig. 17(a)]. It is noted that the equivalent RI of the virus layer (n_binding) depends on the number of bonded virus layers, which is a function of the virus concentration and binding processing time, and is dominated by the concentration in real sensing process with a given binding time. Thus, the SWGR sensing performance can be evaluated by calculating the n_binding response of the device, with n_binding ranging from 1.35 (no binding) to 1.5 (full binding). Simulation results in Fig.17(f) show that the functionalization and the full-binding process induce a shift of 3.41 nm and 1.14 nm, respectively. The obvious simultaneously measurable shifts in the FSR range (⁠$Δλres<FSR$⁠) and experimental values (⁠$Δλres≫1pm$⁠) indicate the promising potentials of the proposed device in detecting the COVID-19 virus or simply as a chemical/biosensor in future practical applications.

Ever-increasing advances and developments in the preparation and utilization of nanostructured materials have facilitated the extraction of an unprecedented range of properties from nanomaterials. Biosensing and amplification of targeted samples at extremely low concentrations for detection are among the most intriguing properties.⁴⁸ Two-dimensional materials , as one of the most discussed groups of nanostructured materials, are composed of a variety molecular structures [Fig. 18(a)]. They possess favorable properties such as high surface-to-volume ratio and tunable physicochemical features for detection of biomolecules and various analytes, including virions.^144–149 Also, a wide range of surface chemistry methods like amino silane and plasma chemistry can be done to immobilize biomolecules on the surface for selective sensing of different target analytes.^150,151 Therefore, a sensitive 2D material structure with the right biofunctionalization strategy is favored to be employed for a successful early, rapid therapeutic biosensing system.^144,152,153

Among the 2DM family, graphene possesses a special electron band structure showing high carrier mobility and zero bandgap characteristics, exceptional high-energy transfer efficiency, and large surface area.⁴⁶ Other than that, graphene biosensing properties are easier to exploit due to the sophisticated fabrication process, ease of functionalization, and biocompatibility.¹⁴⁹ 

As a result, graphene and its derivatives, like graphene oxide (GO), have been used in various biomedical areas, including but not limited to DNA sequencing, tissue engineering, stem cell research, and biosensing.¹⁵³ More specifically, there have been many works on utilizing graphene and graphene oxide for their biosensing properties.^{144,155–157} For instance, Jin et al. demonstrated a functionalized graphene oxide wrapped around SiO₂ that possessed superior RNA sensitivity and a limit of detection up to 1 fM, with the potential to show even higher sensitivity values.¹⁵⁸ They showed that high electron conduction and larger surface area in spherical morphology are, in particular, effective at improving the sensitivity and limit of detection. However, as far as pathogen detection is concerned, very few works and reviews based on graphene exist so far to the best of our knowledge. Most are based on various designs, such as field-effect transistors (FETs) and electrochemical systems based on graphene.^{143,149,152,159–161} The working principle of graphene-based FET biosensing design as a nanoelectronic biosensor relies on charge detection for sensing in which an electrical signal is generated upon attachment of biomolecules to the surface of the sensor as a result of charge-density change.^149,162 Although the FET-based graphene biosensor provides easier mass-scale production with satisfying sensitivity, its limited sensing capability along with being damaging to living cells make its application restricted compared to analogous optical ones.^{46,148,163,164} The superb optical properties of graphene, such as broadband and saturation absorption, and great affinity to adsorb biomolecules as a result of π–π stacking on its surface, make graphene a very promising candidate for development of any optical 2DM–based biosensor^165,166 [Fig. 18(b)]. We briefly discuss the potential of graphene-based optical biosensors toward development of a fast, accurate, point-of-care pathogen detection system.

Optical biosensors based on 2DMs and, in particular, graphene can generally be divided into three main groups for sensing: fluorescence, SPR, and SERS [Fig. 19(a)]. Here, we briefly discuss each method and improvement of its sensing properties by utilizing graphene and graphene oxide.

The working mechanism of fluorescence biosensors is based on fluorescence quenching, fluorescence enhancement, or fluorescence resonance energy transfer (FRET).^149,167 Graphene and its dioxides possess fluorescent quenching abilities that have enhanced visualization and detection of the target analyte through labeling biomolecules, such as ssDNA with fluorescent dye. The fluorescent-labeled DNA can be immobilized on the sensing area through π–π interaction between DNA probe and graphene surface.^149,168 Jung et al.⁹⁹ have demonstrated a GO-based immune biosensor to detect pathogens with high sensitivity and selectivity by GO photoluminescence quenching realized by FRET between AuNPs and GO [Fig. 19(b)]. Liu et al.¹⁶⁹ and He et al.¹⁷⁰ have also shown different complex sensing systems based on the high quenching ability and different affinity of GO toward DNA for complex sequence-specific DNA detection [Fig. 19(c)].

An SPR and LSPR biosensing platform conjugated with graphene has been developed to enhance the optical biosensor's sensitivity through amplifying detectable signal compared to conventional plasmonic materials.¹⁴⁹ The conventional methods rely heavily on using biosensors based on gold and silver materials; however, the limitation of these metallic materials, like silver oxidization and poor adsorption of biomolecules on gold surface, have turned attention to utilizing a graphene-based optical SPR biosensor. The gold-coated graphene has a better biological sensing ability by adsorbing biomolecules better through π–π stacking. The effect of adding a graphene layer to the gold surface in a prism-based SPR has been shown by Wu et al.¹⁷¹ [Fig. 19(e)]. It has been demonstrated that the addition of graphene layer (L) would increase the sensitivity by $1+0.025 L$ × γ (where γ > 1), where γ is the adsorption efficiency enhancement factor. It has been shown that graphene increases the sensitivity to refractive index change by 25% for L = 10 and, therefore, S increases five times in the case where graphene adsorbs four times more biomolecules (γ = 4).

A Raman spectroscopy working mechanism is based on measuring inelastic scattering of light, which provides a footprint regarding the molecular or material structure. However, the Raman scattering from organic molecules is weak and needs to be amplified through applicable methods to be used as a biosensing mechanism. The enhancement of Raman signal can be accomplished by adding metal nanostructures due to the SPR effect. The graphene can be used as a very effective SERS active substrate as it can extinguish photoluminescence of fluorescent dyes and eliminate fluorescence background. Xie et al.¹⁷² have shown the photoluminescence suppression effect in a graphene substrate grown on a SiO₂/Si surface due to electron transfer and energy transfer between the graphene and R6G dye molecules [Fig. 19(f)].

As mentioned, a wide range of graphene-based optical biosensors based on fluorescence, SERS, and SPR has been developed in recent years, among which few possess sufficient sensitivity and specificity to be utilized for SARS-CoV-2 detection (Table IV).^174–179 Lee et al. showed a highly sensitive plasmonic/magnetic hybrid nanomaterial graphene-based platform possessing an attractive combination for virus detection: plasmonic and magnetic effects. They used a magnetic field for separation of target virus from impurities along with the plasmonic substrate for fluoroimmunosensing to detect influenza virus (H1N1) with LOD of 7.27 fg/ml.¹⁸⁰ Graphene oxide was also used by Jeong et al. as a base for a facile fluorometric system for detection of influenza viral genes.¹⁷⁷ Furthermore, an optical biosensor based on functionalized graphene oxide and cadmium sulfide quantum dots (CdS-NH₂GO) has been developed to function through the surface plasmon resonance principle. The high binding affinity of targeted dengue virus envelope (DENV) E-proteins to the surface resulted in a high detection sensitivity of 0.08 pM. The low absorption of graphene and small photocurrent are a driving force to use graphene quantum dots (GQDs), which show a significant photoluminescence that can be improved by functionalization and doping of graphene.⁴⁰ All the aforementioned methods show the potential of combining graphene-based materials with nanomaterials in COVID-19 optical biosensors. The final sensitivity can be improved by increasing the graphene-based sensing surface and proper surface functionalization with SARS-CoV-2 antibody or nucleic acid.

Emerging pandemic and epidemic diseases, like COVID-19, bring out a high demand for advancement and research in medical detection and treatment methods. The optical biosensors provide a fast detection (<1 min) of such a virus at very low concentrations (∼1 fM). However, they need to be designed and functionalized to be the most absorptive to the target analyte. The ideal label-free biosensor is cost-effective, disposable or reusable, compact, and semiautomatic. Although most efforts in biosensors have been focused on protein biomarkers, other targets, such as small molecules and nucleic acids, are crucial in expanding the application of biosensors, including optical ones. A common challenge for optical biosensors is to reach the capability of performing the measurement in real complex samples, avoiding, or limiting, the sample preparation phase. Developing a label-free biosensor is aligned with that purpose because the need for onsite detection techniques is increasing as the world after the COVID-19 pandemic will never be anything like before.

A.A., C.W., and K.M.Y. contributed equally to this work.

The research team is supported by the AFOSR Multidisciplinary Research Program of the University Research Initiative (MURI) Silicon Nanomembrane Research Center (FA9550-08-0394), National Institutes of Health (HHSN261201200043C), U.S. Army Research Office (ARO) (W81XWH-14-C-0029), NASA Small Business Technology Transfer (STTR) phase I program (80NSSC18P2146), AFOSR STTR phase I program (FA9550-19-P-0004), AFOSR Small Business Innovation Research (SBIR) phase II program (FA9550-19-C-0003), and the State of Texas. A.A. and K.M.Y. are supported by a UT, Austin, Research Assistantship, and C.W. is supported by scholarship funding (No. 201806470008) for his basic Ph.D. research in UT, Austin.

The data that support the findings of this study are available from the corresponding author upon reasonable request.