A Terahertz Point Source Meta-Sensor in Reflection Mode for Trace-Amount Bio-Sensing Applications

Abstract

Biosensors in the Terahertz (THz) region are attracting significant attention in the biomedical and chemical analysis fields owing to their potential for ultra-trace sensing of various solutions with high sensitivity. However, the development of compact, highly sensitive chips and methods for easy, rapid, and trace-amount measurements have been significantly hindered by the limited spatial resolution of THz waves and their strong absorption by water. In this study, we developed a nonlinear optical crystal (NLOC)-based compact THz sensor chip, and a near-field point THz source with a diameter of ~ϕ20 μm was locally generated via optical rectification. Here, only the single central meta-atom was excited. The reflective resonance responses highly depend on the array number and period of the meta-atom structures. The sensing performance was examined with several liquid biological samples, such as mineral water, DNA, and human blood. 1 μL of samples was directly dropped onto the meta-surface with an effective sensing area of 0.32 mm² (564 μm × 564 μm). Obvious resonance frequency shifts were clearly observed. This research holds significance in advancing liquid bio-sample sensing methodologies by facilitating easy, rapid, and trace-amount measurements and promoting the development of compact and highly sensitive THz sensors tailored for liquid biological samples.

1. Introduction

Biosensing in the terahertz (THz, 0.1–10 THz) region has received considerable attention because of its advantageous properties. THz waves provide vital insights into biological reaction dynamics and have comparatively lower energy than infrared absorption [1,2,3]. It has great potential for noninvasive and label-free detection, particularly for assessing functional expression and conformational change in biomolecules [4,5,6,7]. THz time-domain spectroscopy (THz-TDS), a powerful technique, plays a major role in elucidating distinctive biomolecular collective motions, including intermolecular vibration, rotation, and hydrogen bonding [8,9,10,11,12,13,14]. Practical THz sensor chips are urgently needed and are of great significance for the development of THz technology for biosensing devices in real-life applications. However, trace-amount measurements with THz waves are critically challenging because of their strong absorption in polar solvents and low spatial resolution (1 THz ≈ 300 μm), which are the major obstacles to developing highly sensitive and compact THz sensor chips.

Metamaterials, which are artificially developed arrays at the subwavelength scale [15,16,17,18,19,20], have been increasingly employed in biosensing as promising tools for extracting valuable biological information [21,22,23,24]. By utilizing metamaterial technology, high sensitivity can be attained owing to the strong confinement of electromagnetic fields in metamaterials for detecting small changes in the surrounding dielectric environment, and more feasible and convenient fabrication procedures using conventional photolithography technology are available. Metamaterial-based biosensors have demonstrated great success in the accurate measurement of a wide range of biological samples, including viruses [25,26,27,28], bacteria [29,30], microorganisms [31,32], DNA [33,34,35], proteins [36,37,38], and cells [39,40,41]. However, these studies primarily employed dried samples and overlooked the intricate properties of the samples owing to the wavelength of the far THz fields, necessitating a sensing area larger than the diffraction limit of the THz waves and complex measurement processes. In numerous instances, the physical state of the sample, whether solid or liquid, significantly affects the measurement outcomes, making the feasibility of conducting the measurement questionable. At present, numerous metamaterial-based biosensors operate in transmission mode [27,37,39,40,42,43,44,45,46]. When it comes to polar liquid samples, systems working in transmission mode are less capable of completing the measurement because of the strong absorption and complex optical refraction caused by the irregular shape of the liquid samples. To address this problem, microfluidic technology is typically adopted to confine a liquid sample in a well-developed channel with a suitable thickness to allow THz waves to interact effectively and be transmitted [47,48,49,50,51]. Measurements in the reflection mode exhibit superior performance with the advantages of easy and rapid measurement of liquid biological samples, as evidenced by several related studies [52,53]. Consequently, the reflective properties of THz biosensors are critical for the advancement of THz sensor chips that offer enhanced sensing capabilities. In addition, the spatial resolution of conventional time-domain microscopes is constrained by the diffraction limit, which is determined by the optical beam wavelength. Subwavelength-scale measurement is quite challenging to achieve, and the effective sensing area is relatively large, resulting in considerable difficulty in trace-amount measurement. In contrast to far-field technology [37,47,54,55], near-field technology [48,49,50], which surpasses the diffraction limit, can better implement trace-amount measurements with unique advantages. Within the near-field region, the divergence of the THz beam is minimized, resulting in a substantial reduction in the effective sensing area; therefore, much smaller sample quantities are required. This is critical for achieving trace-amount measurements, which are much more favorable for the development of more compact and practical THz sensor chips.

In this study, reflective near-field THz meta-sensors were developed for easy, rapid, and trace-amount measurements of liquid biological samples. Initially, the reflectance and transmittance responses of a specific meta-atom structure were calculated and compared to confirm the resonance characteristics in the reflection mode, including both the response spectra and electric field distribution. Subsequently, a series of meta-atom structures with various array numbers and periods were developed and fabricated, as illustrated in Figure 1. The reflective response properties of the array number and period dependences were investigated using a near-field point THz source microscope, as depicted in Figure 2. THz pulses were locally generated, and only the central single meta-atom was excited. A meta-atom-based THz sensor with an array number of 5 × 5 and a period of 120 μm, determining an effective sensing area of 564 μm × 564 μm, was chosen to estimate its sensing capability and characteristics. The reflective THz meta-sensor has the superior advantage of sensing a wide variety of liquid biological samples. Mineral water, DNA, and human blood were used as the test samples. A trace amount of each sample (1 μL) was dropped on the meta-sensor to monitor the resonance frequency shift. The reflectance of various concentrations of mineral water was measured, and the sensitivity was estimated. Differences in the resonance frequency shift were compared between double-stranded DNA (dsDNA) and single-stranded DNA (ssDNA) at the same concentration (100 μg/mL). Changes in blood glucose levels (BGLs) before and after a meal were measured using our meta-sensors and compared to the results measured by a commercial glucose meter. Compared to conventional THz metamaterial sensors, the spatial resolution has been dramatically improved, making it possible to perform biosensing in minute quantities. Additionally, compared to the transmission-type chip previously developed by our group, the meta-sensor operating in reflection mode exhibits better sensing capabilities of larger Q factor and higher sensitivity for samples of various forms, including highly viscous blood samples and liquid biological samples that can be directly and simply dropped onto the meta-sensor for easy and rapid measurement. A near-field point THz source is utilized for the trace-amount measurement with a small sensing area. This work is essential and necessary for the development of easy, rapid, and trace-amount measurement methods and compact, practical, and highly sensitive meta-sensor chips operating in reflection mode for real-life applications.

2. Materials and Methods

2.1. Meta-Atom Structure Arrangement and Fabrication

We prepared several arrays of meta-atom structures, shown in Figure 1, known as SRRs, to investigate the resonance response and electric field coupling effect between the meta-atoms by exciting a single central meta-atom with locally generated THz pulses. A single meta-atom had a size of 84 μm × 84 μm (l × l), a linewidth w of 10 μm, and gap G of 20 μm. The meta-atoms were periodically arrayed in 3 × 3 and 5 × 5 patterns with periods p of 100, 120, and 180 μm, respectively. These structures were fabricated on a 500 μm-thick (t_g), (110)-oriented GaAs substrate using a conventional photolithographic technique and by radio-frequency sputtering a 1 nm-thick adhesion layer of titanium, followed by 200 nm (t_m) of gold for the meta-atom structures. These parameters were selected because the fundamental resonance, known as the LC resonance, can be easily exhibited in the THz frequency region. While preparing the experimental system, we gradually adjusted the position of the meta-atom structures to ensure that the laser irradiation spot was located at the center of the central meta-atom, as indicated by the green spots in Figure 1.

In the simulation, the same geometrical parameters and configurations of the meta-atom structures, as shown in Figure 1, were utilized. The material for the meta-atom structures was modeled as a Perfect Electrical Conductor (PEC). The thickness and refractive index of the GaAs substrate were set to 500 μm and 3.59, respectively. A point THz source was positioned at the center of the central meta-atom, with a depth of 50 μm from the upper surface of the GaAs substrate. The point THz source was configured as a “Gaussian” type, with an amplitude of 100 and a wavelength range from 333 μm to 2998 μm. The simulation was conducted at a temperature of 300 K, with boundary conditions set as Perfectly Matched Layer (PML).

2.2. Experimental Setup

Figure 2a illustrates the schematic of the employed near-field point THz source reflection experimental system with a femtosecond fiber laser (TOPTICA FEMTOTIBER pro: maximum power of 400 mW, wavelength of 1.56 μm, pulse width of 100 fs, and repetition rate of 80 MHz) as the optical source. The laser beam emitted from the laser source was divided into two linearly polarized beams, the pump and probe beams, using a beam splitter. The pump beam was modulated by an optical chopper, which provided a reference frequency for the lock-in amplifier, and eventually focused on a 2D THz emitter, which was a piece of a 500 μm-thick, (110)-oriented nonlinear crystal GaAs wafer. The laser pump beam was focused near the upper surface of the GaAs substrate using a high numerical aperture (NA) lens to minimize the laser beam spot size, generating a point THz pulse emission at the laser irradiation spot via an optical rectification process with a diameter of ~ϕ20 μm, similar to the laser beam spot size [56,57], as Figure 1b shows. The emitted THz pulses were reflected by the upper surface of the GaAs, on which the meta-atoms were fabricated. A 175 μm thick indium tin oxide (ITO)-coated PET sheet was used to transmit the laser beam and reflect the THz beam. Finally, the THz beam equipped with sample information was focused and detected by a spiral-shaped low-temperature-grown (LT-) GaAs photoconductive antenna detector. Meanwhile, the detector was excited by the probe beam under running conditions. Before arriving at the detector, the wavelength of the probe beam was converted to 780 nm from 1560 nm by a periodically poled LiNbO3 (PPLN) nonlinear crystal via second-harmonic generation. The optical path length of the probe beam was altered by changing the position of the time-delay stage, and the amplitudes and phases of the THz pulses excited by the pump beam were monitored and recorded. Eventually, the expected information of the measured samples was extracted from the time- and frequency-domain spectra. To position the laser beam at the center of the central meta-atom precisely and quickly, as shown by the red spot in Figure 2b, a photodiode, and a galvano mirror were employed to visualize the laser reflection images around the meta-sensors.

2.3. Sample Measurement

During the measurement, the GaAs substrates with the meta-sensors were placed on the sample holder (X-Y moving stage) with the correct orientation [48]; the gap direction of the meta-atom structures was aligned with the direction of the electric field E of the pump beam, which was at a 54° angle to the [001] crystal orientation of the GaAs substrate, as illustrated in Figure 1 and Figure 2b. After fixing the orientation of the GaAs substrate, a reference signal was first obtained when the laser beam illuminated on the area without meta-atom structures, and another signal was sequentially obtained when the laser beam was focused on the center of the central meta-atom. The experimental conditions were kept unchanged; the biological samples were directly dropped on the meta-sensor with complete coverage, and the reflected signals of the samples were measured and recorded. After Fourier transform and reflectance calculations, the resonance frequency shifts before and after dropping the samples were examined and analyzed. The reflectance was defined as R(ω) = E_R,MM(ω)/E_R,ref(ω), where E_R,MM(ω) and E_R,ref(ω) are the frequency-dependent reflected THz amplitudes with and without meta-atom structures, respectively. If the sample is dropped on the meta-sensor, E_R,MM(ω) is replaced by E_R,S(ω), which is the frequency-dependent THz amplitude reflected by the sample. The measurement method varies depending on whether the biological sample can be cleaned. If a liquid biological sample can be cleaned with ultrapure water, meta-sensors can be reused. After dropping the sample, the reflected signal was promptly measured once. Subsequently, the sample was cleaned using ultrapure water until the meta-sensor reverted to its initial state. This drop–clean procedure was repeated ten times, and the obtained ten resonance responses were averaged to improve reliability. However, if the liquid biological sample cannot be cleaned, each meta-sensor can only be used once. The measurements were conducted continuously until the samples were dry. The first data in these measurements were determined to be reliable because the values changed during the drying process (Figure S3 in the Supplementary Material). This developed sample measurement method, operating in reflection mode, simplifies the measurement process by directly dropping liquid biological samples on the meta-sensor, resulting in easy and rapid measurement.

3. Results and Discussion

3.1. Comparison between Experiments and Simulation

In our previous study, we observed clear LC resonances in transmission mode [48]. In this study, the reflective resonance responses and properties of meta-atom structures were investigated prior to their sensing applications. Initially, the resonance responses in both transmission and reflection modes were compared, as illustrated in Figure 3, in which a 5 × 5 arrayed meta-atom structure with a period of 100 μm was considered an example. All computational responses were studied using finite-difference time-domain (FDTD) software (ANSYS Lumerical FDTD 2024 R1.2). Figure 3a shows the calculated reflectance and transmittance spectra, including the original data points and corresponding smoothed lines. During the simulations and experiments, the sampling time was configured to be sufficiently long to obtain a useful response signal. For the Fourier transformation procedure, we extended the sampling time to an amplitude of zero (Figure S1 in the Supplementary Material). All calculated and experimental results were smoothed to guide the eyes. Notably, the resonance peak manifests at approximately 0.26 THz in the reflection mode and at approximately 0.4 THz in the transmission mode, which are attributed to the LC resonance mode [48]. Transmittance has a similar definition as reflectance, as described in T(ω) = E_T,MM(ω)/E_T,ref(ω), where E_T,MM(ω) and E_T,ref(ω) is the frequency-dependent transmitted THz amplitudes with and without meta-atom structures, respectively. In transmission mode at 0.4 THz, as depicted in Figure 3b, in the LC resonance mode, the electric field was intensified significantly in the gap region within the black box, enhancing the interaction between the meta-atom structures and the sample for heightened sensitivity [48]. A similar electric field distribution was observed in the reflection mode at approximately 0.26 THz (Figure 3c), showing a concentrated electric field in the black box. In the developed meta-structure, known as a split-ring resonator (SRR), the gap and gold arm function as a capacitor and inductor, respectively. The resonance frequency can be calculated by the equation $f_r=1/\left(2\pi \sqrt{LC}\right)$, where L and C are the equivalent inductance and capacitance, respectively. Because the THz beam is perpendicular to the meta-atom plane, the meta-atoms are electrically coupled to the electric component of the THz beam, with the capacitance component playing a dominant role. When the refractive index of the material in the gap region changed, the equivalent capacitance changed, leading to a change in the resonance frequency [17,48].

In this study, the resonance response characteristics under near-field THz point-source excitation differed significantly from those under far-field excitation. The near-field region exerted a dominant influence on the response, resulting in notable differences between the transmission and reflection modes. The coupling effect on the air and GaAs sides played significant roles in the transmission and reflection modes, respectively. Owing to the higher refractive index of GaAs, the resonance frequency occurs at a lower frequency, as shown in Figure 3a, which is also demonstrated by the refractive index dependence in Figure 3e. The reflection resonance mode, arising from the strong coupling between the central meta-atom and adjacent meta-atoms, exhibited greater sensitivity to the surrounding dielectric environment and a superior sensing potential. The Q factor, which can be calculated by the formula $Q=f_r/∆f$, where f_r is the resonant frequency, and ∆f is the full width at half maximum, was estimated to be 2.51, which surpasses that in transmission mode (1.73). Additionally, Figure 3d shows the resonance-frequency dependence on the refractive index of the surrounding material in which the meta-atom structures are immersed. The thickness of the material is set as 10 μm in the simulation. As the refractive index increases, the resonance frequency gradually shifts to lower values. Figure 3e shows the relationship between the resonance frequency of the meta-structure and the refractive index of the surrounding material. The black spheres represent the resonance frequencies extracted from the spectra in Figure 3d, and the red line is the corresponding linear fitting result. The linear fitting equation used is $RF=s\bullet RI+a$, where RF and RI are the datasets of resonance frequency and refractive index, respectively. a is the fitted intercept. The slope s of the fitting result is the calculated sensitivity of 27.77 GHz/RIU, as shown in Figure 3e. Equations of the same form are employed for all the sensitivity calculations in this work. Compared to some other metamaterial sensor studies with higher sensitivity and larger Q factor operating in a far-field way [37,46], although the sensitivity and Q factor reported in this work are smaller, the effective sensing area is much smaller, which is only a few hundredths of that of the compared far-field sensors, highly advantageous for trace-amount measurement for liquid biological samples. The resonance response in the reflection mode demonstrates robust and effective sensing performance.

To delve deeper into the reflective responses and properties of meta-atom structures, the reflectance spectra of meta-atom structures with various array numbers and periods were examined and analyzed. This analysis is crucial for establishing a fundamental understanding of the reflective resonance response of meta-atom arrays and for advancing the development of practical THz sensor chips. Figure 4 illustrates the array-number dependence of the developed meta-atom structures with a fixed period of 100 μm. From the measured results depicted in Figure 4a, distinct resonance dips in arrays 3 × 3 and 5 × 5 are evident, as indicated by the red and blue arrows. These resonances arise from the coupling between meta-atoms, which amplifies the resonance responses of the meta-atom structures [48]. Moreover, as the array number increases from 3 × 3 to 5 × 5, changes in the resonance frequency and intensity occur, signifying variations in the coupling state between the meta-atoms. The resonance response is sensitive to the array number, implying that it can be manipulated by controlling the coupling effect between the meta-atoms by adjusting the array number. Figure 3b shows the calculated results, which clearly explain the measured results.

The period of the meta-atoms, which represents the spacing between them, is another critical parameter that significantly affects the coupling between the meta-atoms. Figure 5 shows the corresponding measured and calculated results for a fixed array number of 3 × 3, which is the smallest array number with sufficient resonance response. Both the experimental and simulation data revealed noticeable resonance peaks, as indicated by the colored arrows. This period also affected the coupling state between the meta-atoms, resulting in shifts in the resonance frequencies of the structures. As the period increased, the resonance frequency gradually shifted towards lower frequencies. This is because the resonance frequency was inversely proportional to the square root of the SRR area [58]. Therefore, adjusting the period of the meta-atom is also an effective method for modulating the resonance response of the developed meta-atoms. Note that differences in the reflectance spectra between the experiment and simulation results can be observed in the exploration of the array number dependence and period dependence. These discrepancies arise because the geometrical parameters of the fabricated meta-atoms do not perfectly match the settings used in the simulations. These parameters include the size of a single meta-atom unit, the linewidth, the gap size, and so on. Additionally, these variations can slightly alter the period of the meta-atoms, further contributing to the differences. Moreover, the laser irradiation spot needs to be focused on the center of the central meta-atom, which cannot be precisely positioned as accurately as in the simulation. This imprecision is also one of the contributing factors to the observed differences. Based on both simulation and experimental findings, the reflective resonance response typically demonstrates superior sensing capabilities in terms of a larger Q factor and higher sensitivity, notably influenced by the array number and period of the meta-atom structures. This is particularly favorable and beneficial for the practical application of THz meta-sensor chips.

3.2. Applications

3.2.1. Mineral Water Measurement

For sample measurements, the meta-atom structure with an array number of 5 × 5 and a period of 120 μm, providing an effective sensing area of 0.32 mm² (564 μm × 564 μm), was chosen as the meta-sensor to assess its sensing efficacy. The reason why the array number of 5 × 5 was chosen is that the resonance response of the 5 × 5 array is stronger than that of the 3 × 3 array, owing to the increased number of meta-atom units. The selection of a period of 120 μm is a compromise between ensuring a relatively larger Q factor and keeping a relatively smaller sensing area suitable for trace-amount measurements. Initially, we used commercial mineral water (Contrex with 1468 mg/L of minerals) at concentrations of 13, 122, 158, and 203 mg/L as liquid test samples. During the measurement process, an adequate volume of 1 μL of mineral water was dropped using a micro-syringe (Sartorius Biohit, 0.1–3 μL) to ensure complete coverage of the meta-sensor. Consequently, the concentration information of the mineral water was encoded into the frequency response of the meta-sensor. After the deposition of mineral water, the reflected THz signal was promptly measured once. Subsequently, the meta-sensor was cleaned with ultrapure water to restore it to its initial state. This drop–clean procedure was repeated ten times, and the obtained resonance frequencies were averaged to enhance the reliability of the measurement of one concentration of mineral water. Figure 6a shows the reflectance spectra measured from a single measurement across various concentrations of mineral water. The resonance frequency was observed at 0.339 THz when no sample was present on the meta-sensor and shifted to lower frequencies because of the higher refractive index of mineral water compared to air. As the mineral concentration increased, the resonance frequency shifted to higher values owing to the decreased refractive index associated with increased water hardness [48], which is consistent with the calculated results shown in Figure 3e. Figure 6b shows the relationship between the resonance frequency and the concentration of the mineral water. The pink spheres are the mean of the ten measured resonance frequencies for each concentration of mineral water, and the red line is the corresponding linear fitting result, the slope of which shows a sensitivity of 101 MHz/(mg/L), which is more sensitive than that in the transmission mode [48]. In ascending order of concentration, the standard deviations for each concentration of mineral water are 2, 2.6, 1.4, and 2.5 GHz, respectively. The mineral water measurement results confirm the validity of the reflective method with the advantages of easy and rapid measurement, as the liquid sample is directly and simply dropped on the meta-sensor. The developed meta-sensor also exhibited exceptional detection capabilities with trace amounts of mineral water in a near-field manner under point THz source excitation.

3.2.2. dsDNA and ssDNA Analysis

DNA and human blood samples were analyzed to estimate the feasibility and effectiveness of the developed reflective method and meta-sensor for real biological sample applications. First, the frequency shifts of dsDNA and ssDNA were compared. DNA samples were procured from Promega Corporation and frozen to prevent denaturation. The storage buffer for DNA samples is 10 mM Tris-HCl (pH 7.5 at 25 °C), 10 mM NaCl, and 1 mM ethylenediaminetetraacetic acid (EDTA). The original concentrations of dsDNA and ssDNA were 584 and 100 μg/mL, respectively. Upon thawing, the dsDNA solution was diluted to match the ssDNA concentration of 100 μg/mL. As these DNA samples could be cleaned with ultrapure water, the measurement process was identical to that for mineral water. Figure 7a shows the measured reflectance spectra of each DNA sample in a single measurement. The solid and dotted lines represent the reflectance spectra without and with DNA samples, respectively, with red and blue indicating dsDNA and ssDNA, respectively. Before depositing the DNA samples on the meta-sensor, resonance peaks appeared at the same position of 0.334 THz, indicating that the meta-sensor exhibits consistent reflective properties and effective cleaning. Any slight differences in the reflectance amplitude stem from the system noise, which does not affect the position of the resonance peak. After directly and simply dropping the DNA samples, a remarkable resonance frequency shift to the lower-frequency regions was observed. Because the refractive index of dsDNA exceeds that of ssDNA [59], its resonance frequency shift is greater than that of ssDNA, which agrees with the calculated results shown in Figure 3e. Figure 7b shows the average resonance frequency shifts of ten measurements for dsDNA and ssDNA, which were 119 and 110 GHz, respectively. The error bars display the standard deviations, which were calculated to be 4.3 and 3.5 GHz, respectively. These disparities in resonance frequency shift enable the differentiation between dsDNA and ssDNA based on their distinct refractive indexes. These findings suggest that our method and meta-sensor operating in reflection mode can easily and rapidly distinguish between dsDNA and ssDNA using extremely small amounts of liquid biological samples.

3.2.3. Blood Glucose Level Analysis

For further estimation of liquid biological samples, human blood samples with notable variations in BGLs before and after a meal were examined. Using the meta-sensor, we are able to monitor the health status related to the BGL. Given the complexity of blood samples, cleaning the meta-sensor to its original state after use is challenging. Hence, each meta-sensor was employed only once, unlike in the case of mineral water and DNA measurements, where meta-sensors were reusable. To minimize measurement errors arising from the use of different meta-sensors, a sufficient number of meta-sensors were fabricated on the same chip, and all were expected to have consistent sensing properties. Accordingly, on a piece of GaAs substrate, 25 meta-sensors were fabricated in a 5 × 5 array pattern with 0.9 mm spacing between sensors both horizontally and vertically. This spacing was chosen to prevent interference from neighboring meta-sensors when liquid samples were applied (Figure S2 in Supplementary Material). During the experiment, fresh blood samples were collected from the fingertips of an individual and immediately placed on the sensing area of the meta-sensor. Meanwhile, the BGL was measured using a commercial glucose meter (NIPRO, FreeStyle Freedom Lite). The same micro-syringe was employed to ensure the consistent delivery of 1 μL of blood for each measurement, minimizing errors related to varying blood volumes. The effective sensing area is 0.32 mm², the same as that of the mineral water and DNA measurements. For each measurement, a blood drop completely covered the entire effective sensing area. Once the blood was applied to the meta-sensor, measurements commenced promptly with continuous monitoring of the THz time-domain waveforms until the blood dried, necessitating 15 measurements. Only the first measured THz waveform was regarded as an accurate and reliable representative of the properties of fresh blood and was analyzed. Eventually, the reflected THz waveforms and spectra of fresh blood samples with varying BGL were collected, and evident resonance frequency shifts were observed, revealing changes in the BGL.

Figure 8 shows the results of the BGL measurement. Figure 8a shows one of the measured THz reflectance spectra. In the absence of the blood sample, the resonance peak occurred at approximately 0.35 THz and shifted to approximately 0.24 THz after the deposition of the blood sample with a BGL of 217 mg/dl, as indicated by the red arrow. Figure 8b shows the plots of the resonance frequency shifts, and BGL changes over time. The BGL was low, and one single measurement was taken prior to a meal, followed by four subsequent measurements after a meal as the BGL increased and then gradually decreased. The resonance frequency shifted to a higher value with an increase in BGL owing to the elevated refractive index associated with increased BGL in the blood [60]. The tendency measured by our meta-sensor is in good agreement with that measured using a commercial glucose meter. Errors may arise from the actual dropping conditions, and the meta-sensor may become dirty and non-reusable, as previously analyzed (Figure S4 in the Supplementary Material). As plotted in Figure 8c, the relationship between the resonance frequency shift and the BGL was estimated. The black spheres represent the resonance frequency shifts for the blood of each BGL, and the red line represents the linear fitting result with a fitting sensitivity of 320 MHz/(mg/dL). For validation, similar trends were observed in the measurements conducted on different days (Figure S5 in the Supplementary Material). In summary, the measurement method operating in reflection mode is feasible and allows for easy and rapid measurement by directly and simply dropping liquid biological samples on a meta-sensor. The developed meta-sensor also demonstrates significant sensing behavior for actual liquid biological samples, which paves the way for the development of more practical and sensitive THz meta-sensor chips operating in reflection mode.

4. Conclusions

A series of meta-atom structures were developed and fabricated. A near-field point THz source was locally generated to excite a single-center meta-atom and explore the coupling effect between meta-atoms in reflection mode. Compared with the LC resonance mode in the transmission mode, the resonance frequency occurs at a lower resonance frequency in the reflection mode. Remarkable resonance responses were observed for arrays 3 × 3 and 5 × 5 owing to the coupling effect between the exciting meta-atom and its neighboring meta-atoms. This coupling effect is highly dependent on the array number and the period of the meta-atom structure, which effectively modulates the resonance frequency response. Based on the results of these studies, a reflective meta-atom-based THz sensor was developed. Mineral water, DNA, and human blood were examined as liquid test samples to evaluate the sensing performance. The effective sensing area of the meta-sensor was 0.32 mm² (564 μM × 564 μM) and could be fully covered by a trace amount of liquid biological sample, as little as 1 μL. Clear resonance frequency shifts were observed when employing these samples, owing to the variations in their refractive indices. The resonance frequency was dependent on the concentration of the mineral water, with a calculated sensitivity of 101 MHz/(mg/L). The distinct resonance frequency shifts of the dsDNA and ssDNA were 119 and 110 GHz, respectively. The tendency of the resonance frequency shifts measured by our meta-sensor was in good accordance with that measured by a commercial glucose meter in the BGL measurement, with a calculated sensitivity of 320 MHz/(mg/dL). All the obtained results underscore the feasibility of the measurement methodology and the excellent sensing capabilities and performance of the developed meta-sensors. This work is important for developing a liquid biological sample sensing method to realize easy, rapid, and trace-amount measurements in a near-field manner and to accelerate the development of more compact, practical, and highly sensitive THz meta-sensor chips for real biological applications.