Marie Mertens, Maede Chavoshi, Olivia Peytral-Rieu, Katia Grenier, Dominique Schreurs
©SHUTTERSTOCK.COM/TUMSTATION
Accurate characterization of biological matter, for example, in tissue, cells, and biological fluids, is of high importance. For example, early and correct detection of abnormalities, such as cancer, is essential as it enables early and effective type-specific treatment, which is crucial for mortality reduction [1]. Moreover, it is imperative to investigate the effectiveness and toxicity of pharmaceutical treatments before administration in clinical practice [2]. However, biological matter characterization still faces many challenges. State-of-the-art imaging and characterization methods have drawbacks, such as the requirement to attach difficult-to-find and costly labels to the biological target (e.g., COVID-19 rapid tests), expensive equipment (e.g., magnetic resonance imaging), low accuracy (e.g., ultrasound), use of ionizing radiation (e.g., X-rays), and invasiveness [3]. The characterization of biological matter using microwave (${\mu}{\text{W}}$), millimeter-wave (mmW), and terahertz (THz) spectroscopy is a promising alternative: it is label-free, does not require ionizing radiation, and can be noninvasive. Moreover, there is a significant difference in how different biological materials absorb, reflect, and transmit electromagnetic (EM) waves [4] that is due to the difference in their dielectric properties. The dielectric properties are described by the frequency-dependent material parameter called the complex permittivity ${\varepsilon}\left({f}\right),$ which expresses how the material responds to an external oscillating electric field. The complex permittivity of a material determines how the material absorbs, reflects, and transmits EM waves at different frequencies (Figure 1). Since each biological material’s permittivity spectrum is different, it acts as an EM fingerprint. A material’s complex permittivity can be calculated from the reflection and transmission of EM waves through the material, described by the S-parameters, which can be measured using a vector network analyzer (VNA) transmitting and receiving EM waves over a range of frequencies. The amplitude and phase of the transmitted and reflected EM waves at different frequencies are influenced by different underlying biological effects at different scales. That causes the entire spectrum to provide information from the supracellular to the molecular and even atomic scale.
Figure 1. A conceptual illustration of how biological tissues (e.g., healthy and cancerous tissue) absorb, reflect, and transmit EM waves differently because of the difference in their dielectric spectra.
Spectroscopy using ${\mu}{\text{W}}$, mmW, and THz EM waves has shown potential to sense the concentration of important biomarkers, such as glucose in diabetes patients, DNA and proteins that inform about genetic disorders, and cancer biomarkers to distinguish healthy matter from cancerous matter [4]. These applications will be further elaborated on in the section “Dielectric Spectroscopy Techniques.”
In this section, we explain the meaning and interpretation of the dielectric spectrum, representing complex dielectric permittivity as a function of frequency. The physical phenomena underlying the spectrum’s most important features are elaborated on in the sections “Permittivity Relaxations” and “Permittivity Resonances.”
As opposed to most nonbiological materials, the dielectric permittivity depends heavily on the frequency of the applied electric field for biological matter. Indeed, at different frequencies, the material will absorb, reflect, and transmit EM waves differently. Moreover, permittivity is a complex material parameter: \[\varepsilon\left({f}\right) = {\varepsilon'}\left({f}\right){-}{j}{\varepsilon''}\left({f}\right){.} \tag{1} \]
A schematic representation of both the real and the imaginary parts of the dielectric permittivity is shown in Figure 2. The real part of permittivity describes how well the electric dipole moments inside the material align to the electric field. The imaginary part of the permittivity describes the EM losses in the surrounding media. These EM losses can be attributed to the friction caused by the movement of the charges aligning to the electric field. More fundamentally, the permittivity describes polarization, and the magnitude and phase describe its extent and delay, respectively. The dielectric spectrum shows relaxation and resonance phenomena. Figures 3 and 4 illustrate this relaxation and resonance. At the frequencies of the relaxation phenomena, also called dispersions, there is a decrease in the real part of permittivity, whereas at the frequencies of the resonance phenomena, a sharp resonance is seen in the real part of the permittivity. These phenomena will be further explained in the following section.
Figure 2. The complex permittivity of biomatter (similarly represented in [4]).
Figure 3. A conceptual illustration of a dispersion/resonance in permittivity spectrum: as the frequency of the electric field oscillations rises, dipole moments start to be unable to align to the electric field, and their contribution to permittivity lowers, seen in a decrease in the real part of permittivity.
Figure 4. A conceptual illustration of a dispersion/resonance in permittivity spectrum: as the frequency of the electric field oscillations reaches the dipole’s resonance frequency, its bound charges resonate to the electric field, stretching out and aligning to it with maximal polarization as they are being pulled apart the farthest.
An intuitive explanation for the relaxations (or dispersions) in the real part of the permittivity is given here: an external oscillating electric field causes the alignment of dipole moments (due to the motion of ions, atoms, and molecules) inside the material to the electric field. The larger this alignment, the larger the total electric field and the larger the real part of the permittivity. When the charges’ inertia causes their inability to follow the electric field’s oscillations, their contribution to the total electric field drops to zero. This can be seen in the spectrum as a drop in the real part of the permittivity and is called a relaxation [3], which is visualized in Figure 3. At each dispersion, there is a peak in the dielectric losses, which can be seen in the imaginary part of the permittivity. This frequency range is indeed where the charges are most out of phase, delayed to the external electric field, causing the presence of an imaginary component in the polarization. The relaxation event does not occur at one single frequency but over a range, as biological matter is highly heterogeneous, and the different ions, atoms, and molecules have slightly different relaxation frequencies. The various relaxation mechanisms (${\alpha}$-, ${\beta}$-, and ${\gamma}$-dispersion) are determined by distinct polarization mechanisms for the different constituents of matter (cells and cell membranes, ions, molecules, atoms, and electrons) [5]. As such, the spectrum informs us about the different constituents of matter in different frequency bands. It is essential to understand the fundamentals behind the relaxation phenomena to understand how one could extract information on the biological and physiological state of a material from the frequency spectrum.
The first relaxation is called ${\alpha}$-dispersion. It is caused by the inability of ions to respond to a fast-changing field and occurs at Hz to kHz frequencies. Unfortunately, due to high measurement uncertainty through unwanted measurement electrode polarization, ${\alpha}$-dispersion can often not be interpreted accurately in terms of permittivity [6].
The second relaxation is called ${\beta}$-dispersion. It is caused by the disappearance of Maxwell–Wagner polarization at the frequency of the dispersion at kHz to MHz frequencies. This Maxwell–Wagner polarization is caused by the presence of physical interfaces (e.g., cell membranes) that constitute the border between two different media, such as the intracellular and the extracellular fluid. At frequencies lower than the ${\beta}$-dispersion and upon the effect of an external electric field, charges accumulate at the border and create a space charge region, which contributes to the permittivity. At frequencies higher than the ${\beta}$-relaxation, these interfaces are shortened, and they become invisible to the electric field, which can now penetrate to the insides of the cell [4], [6] (see Figure 5). As such, when investigating permittivity at frequencies beyond the ${\beta}$-relaxation, examination of the cells’ contents (e.g., intracellular fluid and particles) becomes possible. Because of the disappearance of the space charge region, there is a decrease in polarization and as such a decrease in permittivity.
Figure 5. An illustration of the concept of ${\beta}$-relaxation, where the Maxwell–Wagner polarization disappears because the electric field oscillations bypass the cell membranes at frequencies beyond the ${\beta}$-relaxation.
Finally, ${\gamma}$-dispersion occurs at frequencies near 24 GHz. This dispersion is highly influenced by the water content of the material, which is often high for biological matter. At frequencies below the ${\gamma}$-dispersion, the dipole constituted by the asymmetric shape of a free water molecule vastly contributes to the induced polarization when it aligns to the external electric field. However, around 24 GHz, these molecules are unable to align with the external electric field [4], [6]. The higher the free water content, the higher the orientation polarization and the higher the permittivity before the ${\gamma}$-relaxation. Given that water content varies among different organs, tissues, and cells, measuring permittivity around the ${\gamma}$-relaxation can help in differentiating between different organs for ${\mu}{\text{W}}$ imaging, between live and dead cells, or between cancerous and healthy cells, and so forth. A separate dispersion is attributed to water molecules that are bound to other molecules. Given the complexity of this ${\delta}$-dispersion, it will not be elaborated on here.
The physical events underlying the resonances in the THz part of the spectrum are rather complicated. However, we will attempt to describe the fundamental idea: when the external electric field is oscillating at THz frequencies (1,000,000,000,000 oscillations per second), only a few very small biological structures are still able to align to the electric field and can as such contribute to the polarization. These small structures are 1) the electrons in the electron cloud surrounding their positive core and 2) atoms that are attached to each other in a molecule. The electric field not only rotates them but also pulls apart the positive and the negative parts, after which a restoring force acts to bring them back together; the structures then can be considered as if they were springs [2]. At their resonance frequency, the distance between the positive and the negative parts reaches its maximum, first in the direction of the external electric field (local permittivity maximum), then, because of delay to the electric field, in the opposite direction (local permittivity minimum). This is illustrated in Figure 4. There is a peak in losses because of the energy absorption that comes with the separation of the charges. Changes in the concentration of some important biomolecules can be observed as a shift in resonant frequency in the dielectric spectrum as their vibrational frequency lies in the frequency range of 300 GHz to 1 THz [4].
To conclude, the ${\alpha}$-, ${\beta}$-, and ${\gamma}$- dispersions and the electronic and atomic resonances that appear in biological matter’s frequency spectrum are influenced by different biological events. Researchers can measure the whole frequency spectrum and deduct information about all these biological events at the same time. Indeed, researchers can characterize biological matter based on fundamental characteristics of the spectra, such as how broad each dispersion is and at which frequency each relaxation occurs. Examples of how scientific research used these characteristics for biological matter characterization are given in the following section. However, in practice, not many have succeeded in really linking these parameters to specific biological events, and so the field still needs to advance in this area. Moreover, the frequency of characterization can be chosen according to which biological events are of interest to the research (e.g., ${\beta}$-frequencies to obtain information on the number of cells and cell membranes and ${\gamma}$-frequencies when water content is of interest).
Finally, before diving into the practical aspects of permittivity measurements, we ought to clarify one thing: one might wonder why we choose to consider dielectric constant rather than impedance. Some [7] prefer to consider impedances which, as opposed to dielectric permittivity, are directly measurable, and no difficult mathematical transformations or calculations are required to obtain them. For the characterization of two or more cells, which always comprise the same geometry and are measured with the exact same method, one is completely right to consider impedances. However, as “dielectric electrical (oxymoron) engineers” [7], we consider dielectric permittivity to be the fundamental parameter. It is a material characteristic and should as such, after de-embedding, be independent of the measurement method and size and shape of the material under test (MUT), which is not true for impedances. Even though still to be improved in practice, the dielectric permittivity of cells, tissues, and liquids of different sizes should thus reflect only the content of the biological matter rather than the size, shape, and volume of the cell.
Dielectric spectroscopy (DS) is the measurement of a material’s fundamental dielectric properties as a function of frequency. These properties can be interpreted as explained above and employed to characterize biological matter. The general methodology consists of 1) the measurement of the reflection and/or transmission of EM waves as S-parameters through the MUT, followed by 2) a mathematical conversion to dielectric permittivity. The former will be discussed below. The latter depends on the geometrical setup of the measurements, as well as the size and the shape of the MUT and the calibration method, and will not be explained in detail further in this article. The basic principle is that the measurement results will be de-embedded, i.e., the effects of the setup will be removed, until only the effect of the MUT’s permittivity is visible. Alternatively, an analytical description of the effect of the sample’s permittivity in the measurement results can be employed, from which the permittivity can directly be extracted.
The most common methodologies for measurement of S-parameters of biological matter for reflection and transmission are explained below and illustrated in Figure 6 and Table 1. Some applications are given as well, and it is shown how the methods have been selected and adapted for measurements on biological matter. For example, biocompatibility of the measurement device must be guaranteed, and possible EM-heating effects must be prevented. The preferred method depends on the frequency, the physical and chemical attributes of the biological MUT (liquid/solid, size, etc.), and the requirements in terms of sensitivity and limit of detection among others. Moreover, the operating bandwidth of the measurements plays a key role in selecting the technology. Broadband characterization provides a global measure on a broad range of frequencies and thus lets us know about the constituents of the matter on different scales but requires technology that is appropriate for the whole spectrum. The technology can be hard to develop, and the instrumentation can be costly and fragile. Narrowband characterization, however, focuses on one to a few discrete frequencies so that the device can be designed to be optimized to these frequencies using cheaper and simpler components. These measurements are often more precise and easier to develop. Below we will first elaborate on broadband DS, and then the main approaches and applications for narrowband characterization will be explained.
Figure 6. An illustration of different types of ${\mu}{\text{W}}$ sensors for dielectric characterization of biological matter. CPW: coplanar waveguide; CSRR: complementary SRR; PGL: planar Goubau line; SRR: split ring resonator.
Table 1. An overview of dielectric characterization methods for biological matter and their applications.
In this subsection, the most common approaches for broadband DS of biological matter are discussed. More specifically, coaxial probes and coplanar waveguides (CPWs) are first explained in depth, after which we summarize the upcoming technologies for sub-THz and THz broadband characterization.
An open-ended coaxial probe is commonly used for measurements of the reflection of EM waves (S 11) from a biological MUT in the frequency range from $\pm{0}{.}{1}{to}$10 GHz (and sometimes even for measurements up to 110 GHz). As the coaxial transmission line is cut off at the border with the MUT, its field fringes into the liquid or semisolid biological matter, which affects the reflection of the EM waves [8]. The basic concept is illustrated in Figure 7. The conversion method from reflection to permittivity is well established, and the method requires minimal sample preparation and is easy to use [9]. The coaxial probe has mainly been used to characterize biological tissue, from normal, benign, and malignant breast or liver tissue to renal calculi and muscle [10], [11], [12]. The measurement methodology had to be adapted largely for biological tissue when researchers noticed that the practices were not proper for biological matter. Indeed, it was found necessary to control the precise position and pressure of the probe on the tissue as compressing the semisolid would change its dielectric properties [6]. Moreover, in a large-scale breast tissue investigation performed by Lazebnik et al. [10], it became clear that an analysis of the content of excised tissue before measurement was of critical importance. A large difference was found between dielectric properties of normal breast tissue and malignant tissue, but it appeared to be largely determined by the adipose (fatty) content of the tissue sample, which was higher in the excised healthy samples. The differences between healthy and malignant tissue were far smaller when this adipose content was accounted for [10]. Nevertheless, other studies still showed promising results: O’Rourke et al. [11] showed that the dielectric properties of ex vivo malignant liver tissues are 19% to 30% higher than normal tissue. Moreover, the renal calculi (kidney stone) category can be statistically differentiated by dielectric properties over the frequency range from 500 MHz to 18 GHz [12].
Figure 7. Open-ended coaxial probe method.
Transition from nonplanar setups to planar ones enables measurements on smaller samples and as such is interesting for the measurement of extracted biological liquid and tissue samples and cell suspensions. Dielectric matter characterization can be performed using many types of transmission lines, such as CPWs, microstrip lines (MLs), substrate integrated waveguides (SIWs), and planar Goubau lines (PGLs) for higher frequencies. However, most of these methods, such as, for example, MLs and SIWs, are mainly used for characterization of the dielectric substrate on which they are fabricated, as their modes of propagation show maximum electric field distributions in the substrate (often between a metallic ground plane and the transmission line). They thus rely on the deposition of metallic lines onto the MUT. Characterization using CPWs (or PGLs; see next section) can be performed by putting the sample, contained in a biocaptor, on top of and thus directly in contact with the metallic lines that are deposited onto another low-loss dielectric material. Gold is often used for these metallic lines as it allows for maximum biocompatibility, resistance against oxidation when measuring ionic substances, and good transmission of the EM waves.
In CPWs, the electric field lines in the dominant mode of propagation, quasi-TEM mode, are oriented as shown in Figure 8(a) [14]. The propagation of the electric field will be affected by any biological matter put on top of the transmission lines. The effective permittivity of the CPW lines is a mathematically complex combination of the permittivity of the dielectric and the permittivity of the material on top of the lines (the unknown editor of Microwaves101 [15] provides us with the rule of thumb that it is approximately the average). In most devices, a microfluidic channel, sometimes with hydrodynamic traps, functions as biocaptor and is constructed on top of the CPW lines [Figure 8(b)], mainly on the signal line and gaps where the field is largest, to allow the electric field to be affected by the biological matter that is pumped through the channel [16], [17]. In fact, as the position of the biological object (e.g., cell or organelle) must be controlled to maximize the sensitivity, “trapping” structures can be added to the microfluidic structure [18], [19]. As the CPW operates in the GHz frequency range, where the electric field can probe the inside of the cells through the shortening of the capacitive effect of the cell membrane, it is used extensively in investigating the intracellular components of different cell lines and their reaction to chemical and physical treatment [18], [20], [21], [22]. Moreover, CPW lines can be adapted to measure the dielectric properties of biological matter of different sizes, from in vivo tissue characterization down to single cells, making it a promising technique to monitor the effects of drugs on different biological entities in a label-free, noninvasive, and nondestructive way [16], [17]. Different studies have shown the possibility to monitor cell electroporation, electrochemotherapy, and the effect of saponin and mitochondrial activity at the single-cell or cells-in-suspension level as represented respectively in the first and second images of Figure 9 [21], [22], [23], [26]. For instance, in [23], it was found that saponin, which affects the cell membrane and allows for exchange of contents through the membrane, effectively caused a reduction in dielectric contrast between the extracellular and intracellular medium. Furthermore, mitochondrial activity can be monitored. This is important as it can be an indicator for how cells will react to therapy. Moreover, melanomas (skin cancers) have been studied directly in vivo and can be detected using ${\mu}{\text{W}}$ spectroscopy with a sensor in contact with the skin of the subject, as represented in the fourth image of Figure 9 [27]. Finally, the dielectric characterization of microtissue (3D cell aggregates as a model for cancer tissue) is emerging with the device represented in the middle of Figure 9 [19], [28]. In conclusion, the CPW allows studies in multiple biosample scales as summarized in Figure 9 [19], [23], [27], [28], [29].
Figure 8. (a) The propagation of electric field in quasi-TEM mode. (b) Microfluidic channel on top of CPW line for liquid characterization [13].
Figure 9. Examples of ${\mu}{\text{W}}$ transmission-based sensors depending on biosample scale for dielectric characterization showing the versatility of CPW sensors in biological applications [19], [23], [27], [28], [29].
In this section, we list interesting technologies for broadband sub-THz and THz measurements (0.1 THz to 10 THz). These frequency ranges show great potential for dielectric characterization. THz time domain spectroscopy (TDS) has been shown to be able to detect drug effectiveness through comparison of absorption coefficients and refractive indices in [30]. It was found that these coefficients and indices showed the trend toward healthy tissue when a drug was administered and the negative effect on the healthy cells, respectively. Moreover, in [31], promising results were obtained showing significant differences in the THz absorption spectrum for different histological types, pathological grades, and glioma-specific biomarkers in measurements of healthy and cancerous brain tissue. In [32], myocardial amyloidosis (deposition of specific protein causing heart stiffness) was detected using THz TDS, combined with convolutional neural networks. Finally, as mentioned in the explanation of dielectric permittivity, the frequency spectrum overlaps with the resonance frequencies of important proteins and biomolecules in this range. To extend these methods for a more accurate frequency domain analysis with higher frequency resolution, there is a need to establish frequency domain spectroscopy methods at THz frequencies. Technologies are being developed to cope with the high loss of EM power into lossy biological MUT and the high noise that is inherent to these frequency bands. For instance, [33] showed the use of a microwire to detect resonating proteins at 0.314 THz excited by visible light. Moreover, a protein concentration change effectively caused a change in the measured resonance frequency in [34]. Planar alternatives are being researched as well, such as the PGL, a fine ML with no ground onto which the Goubau mode can be excited through the CPW to PGL transition. This mode shows high confinement around the line and can be used for measurement of extremely small structures close to the line. In [35], transmission through lysozymes was successfully measured using PGL. More research is necessary to investigate the further capabilities of PGL. Finally, research is ongoing for the use of free space measurements of biological cells and tissues. Interestingly, the sub-THz to THz measurements are hovering on the border between THz spectroscopy and visible light spectroscopy, slightly farther from our field of knowledge but definitely not less important. Researchers are exploring the borders of both the ${\mu}{\text{W}}$ and light (toward quantum) theories and verifying how they can be crossed.
The technologies and applications discussed so far are mainly used for broadband DS of biological matter, i.e., in a wide frequency range. However, broadband permittivity information is not necessarily needed in many cases; for example, if one is interested in studying the presence or absence of a specific substance, solution concentrations, and viability and growth of cells or organisms like bacteria, among others, at a single frequency or in a small frequency range, narrowband techniques can be more precise and faster to investigate dielectric properties of materials.
Narrowband measurements often employ a resonance-based ${\mu}{\text{W}}$ sensor that concentrates the electric field at frequencies near a specific frequency, depending on the application. The term “resonance” in this section is different from the one used in the prior sections; here we mean the resonance of a circuit, due to its components. Depending on the structure of a sensor, it can be modeled as a resonant circuit, including an inductance L, capacitance C, and a term relating to conductivity (or inversely, resistance) G or R. Depending on modeling, they can be connected in a series or parallel form.
This is how these resonant sensors come into play: the sensor behaves as a series or a parallel RLC circuit with its components tuned by the MUT’s properties. For instance, the resonance frequency ${f}_{\text{res}}$ and the quality factor ${Q}$ define the resonance behavior of the sensor. Although the precise relation depends on sensor configuration and calibration, we can approximately assume that the resonance frequency (determined by the equivalent inductance L and capacitance C) is related to the permittivity as \[{f}_{\text{res}} = \frac{1}{{2}{\pi}\sqrt{LC}}\propto\frac{1}{\varepsilon} \tag{2} \] and the quality factor can be inversely related to the losses as \[{Q}\propto\frac{1}{\tan{\delta}}{.} \tag{3} \]
Thus, the interaction of the highly concentrated field and the MUT determines the system’s frequency response by changing \[{f}_{{\text{res}}_{0}}\,\rightarrow\,{f}_{{\text{res}}_{1}},\,{\text{and}}\,{Q}_{0}\,\rightarrow\,{Q}_{1}{.} \tag{4} \]
Nevertheless, challenges in resonance-based biological matter measurements arise from the fact that a considerable portion of them consists of a highly lossy substance: water. The loss decreases the quality factor, flattening the resonance dip and making it difficult to find the resonance frequency.
Various types of planar and nonplanar ${\mu}{\text{W}}$ resonant sensors are used for DS of different biological matter, ranging from aqueous biological solutions, like glucose or saline solutions, to cells and tissues. Here we subdivide the narrowband technologies into nonplanar cavity resonators and planar resonators.
In the case of nonplanar cavity resonators, devices such as coaxial resonators have been developed to study the concentration of solutions. As the name implies, a cavity consists of a hollow space filled with the sample, and the resonance behavior changes according to the sample’s permittivity.
As discussed previously, in wideband open-ended coaxial probe measurements, the sample is in contact with the end of the probe, changing the impedance. Instead of relying only on the conventional commercially available probes, the idea of designing a closed-ended coaxial resonator by replacing the inner dielectric of the coaxial probe with the MUT is proposed in [36]. In both cases, the sample is in contact with the probe, which makes the probe susceptible to contamination and the measurements prone to error. To avoid any direct contact with the sample, cavities for contactless measurement have been introduced [37], [38]. In these cavities, a container is placed inside or close to the cavity, and the electric field is perturbed when the container is loaded with the sample. Figure 10 illustrates a cavity resonator with a container for contactless liquid characterization.
Figure 10. A cylindrical resonator with a container for loading the sample for contactless measurement of liquid concentration [35].
For instance, a cylindrical resonator for contactless measurement of NaCl solution (0%–5%) was proposed in [37]. This sensor includes a probe inserted into a cavity and a container to be filled with the saline solution. Among the main propagation modes (TE111, TM010, and TE011), the one with the highest quality factor and the largest separation from the other modes is chosen (TE011). Based on the perturbation theory, the variation in permittivity ${\Delta}{\varepsilon}$ is obtained.
Despite having a higher quality factor, nonplanar resonant sensors are often bulky, whereas in biomedical applications of DS, an appropriate platform for samples with a volume as small as nL or ${\mu}{\text{L}}$ is required. Although bulk sensors can be integrated with microfluidics [39], miniaturized planar sensors are more easily compatible with various microfluidic systems such as lab-on-a-chip, organ-on-a-chip, cell sorting and trapping, and flexible electronics and wearables.
In fact, basically any structure that forms an RLC circuit on a planar technology, such as CPW or microstrip, can be utilized as a planar resonant sensor, including but not limited to split ring resonators (SRRs), their dual counterparts complementary SRRs (CSRR), and interdigitated capacitors combined with other mentioned structures or solely with an inductance.
In this section, we will discuss how various ${\mu}{\text{W}}$ planar resonators have been introduced to the fields of chemistry and biology, focusing on more recent advances in applications of SRRs and CSRRs.
Planar SRRs make it possible to observe the occurrences close to the surface of a sensor, thanks to interaction of a biomatter with the highly concentrated electric field in the slit of a split ring. One newly developed example is a hairpin resonator for melanoma cell detection [40]. As the cell is trapped in a channel, the EM wave goes through the growth medium, enters the cells, and passes through the growth medium again to reach the other electrode. Although it is usually preferred to evaluate the change in transmission amplitude, the reflection (S11) change is more pronounced than the transmission (S21) change in this case. A higher variation in reflection due to the nonhomogeneous interface between the cell and the medium explains this difference.
While trapping techniques often accommodate single-cell DS, Watts et al. [39] introduce a biosensor that consists of a double SRR that detects the free-flowing live cells moving in the microfluidic channel along the gap. This sensor cannot be considered entirely planar as it has a copper cavity for stimulation, and the gold ring resonator is deposited on a glass coverslip. Nevertheless, this hybrid configuration produces a large quality factor, even with lossy water.
Following the studies on DS for cells and bacteria [41], [42], [43], SRR-based sensors have been developed for real-time monitoring of bacteria concentration and growth [44], [45], [46] and the effect of glucose or antibiotics on this growth [47], [48]. As illustrated in Figure 11(a), the resonance frequency changes the most when the sample is placed over the ring gap, and so the bacteria growth medium is placed over the sensitive gap in SRR, and the corresponding capacitance and conductance gradually change over time, which is measured through the change in amplitude of the reflection or the transmission as well as the resonance frequency shift. The time-dependent variation of the resonance amplitude for two different volumes of Escherichia coli [Figure 11(b)] corresponds to the bacterial growth. As the simulation results in Figure 11(c) and (d) confirm, the changes in the loss tangent contribute to changes in the amplitude, which can be translated into bacterial growth.
Figure 11. The planar SRR sensor for bacterial growth monitoring in a real-time manner with (a) effect of putting the sample on the coupling gap or the ring gap, (b) shift in the amplitude of resonance during time due to bacterial growth and permittivity variation, and (c) and (d) simulation results that show resonance frequency and amplitude changes for different matter with different permittivity [44].
SRRs have also been implemented in tissue studies, focusing on the tissues’ dielectric characteristics as a key feature of their physiological state. A microstrip ring resonator (MRR) with CPW access has been devised to measure the dielectric parameters of animal tissues [49]. The sensor is made up of an MRR on the top surface with CPW feed lines on the other side, and the sample is loaded as a superstrate on the MRR. The advantage of the proposed configuration compared with a simple MRR with a microstrip feedline is that the sample-induced losses do not affect the signal propagation in feedlines. The same as the applications of Goubau line for broadband measurement of biomatter, at higher frequencies (∼600 GHz), a THz-imaging technique based on SRR integrated with PGL is proposed in [50]. With a bare sensor resonance frequency of 596 GHz, healthy and cancerous colonic tissue models cause 130- and 155-GHz frequency shifts to lower frequencies, respectively. Nevertheless, the effective sensing depth of the sensor is limited to a few micrometers, limiting its performance in deeper analysis of a tissue.
CSRRs are one of the most popular planar resonators because of the strong electric field perpendicular to the CSRRs’ surface, improving the penetration depth. This is important especially when a sample is not sensed in direct contact to the sensor as, for example, sensing the blood beneath multiple layers of skin or in a container to investigate blood glucose concentration or in vivo tissue characterization. Omer et al. have developed several glucose sensors in a hexagonal CSRR form (honey cell) [51], [52], [53] and three circular CSRRs [54], [55] for wearable applications. An illustration is given in Figure 12. These studies not only present a novel structure compared with a single CSRR but also develop 1) a dipole antenna tag reader for distant sensing and 2) a radar-driven portable sensor to overcome the need for high-cost and bulky VNAs. The sensor’s selectivity and specificity remain the main constraints in using DS for glycemia. The presence of other substances such as ions, proteins, and other molecules can negatively affect the repeatability and reproducibility of the measurements. However, the glucose concentration in the blood is dominant compared with these other components [51], enabling the design of ${\mu}{\text{W}}$ sensors for glycemia monitoring as a potential noninvasive replacement for currently available invasive and minimally invasive solutions.
Figure 12. A triple CSRR sensor for noninvasive glucose measurement and equivalent electrical model of a single CSRR loaded with MUT [54].
Similar to the case of blood glucose biosensors, noninvasive in vivo tissue spectroscopy requires a good penetration depth such that the signal passes through different layers. Furthermore, flexibility and spatial resolution are essential. For instance, Maenhout et al. [56] developed a flexible tubular device using CSRRs with high penetration depth to characterize colorectal tissue. The prototype is designed to be deployed in patients’ colons and evaluate the tissue’s health or malignancy in six directions (Figure 13).
Figure 13. A flexible tubular device for colon tissue DS using three CSRRs and their equivalent electrical models [56].
Skin tissue and its abnormalities have also been considered as a possible application of resonant ${\mu}{\text{W}}$ sensors [57] and CSRRs; as in [58], the performance of a CSRR-based sensor on a detection of cancerous cells in a skin tissue phantom is validated using simulations. Variation in the amplitude and phase of reflection and transmission coefficients in the frequency range of 2–18 GHz is correlated to the presence of cancer cells with higher water content in the epidermis of their developed skin model.
The resonant sensors and the characterized biological matter are not limited to what we discussed here; the research in this area is ongoing, aiming to improve the specifications such as quality factor and sensitivity, for a more precise detection.
In conclusion, the field of DS is large and growing. This nonionizing, noninvasive, and label-free characterization technique is already employed for biological matter in the detection of cancer biomarkers, glucose sensing in diabetes patients, and many more areas and can be adapted for the characterization of any type of biological matter on a continuously lowering scale as the operating frequencies rise through the development of new measurement technologies and equipment. The biomedical applications of DS are not limited to the ones mentioned here, neither are the technologies only valuable on their own. Combining different DS techniques as well as combining DS techniques with other characterization methods, e.g., electrochemical analysis, can allow us to achieve measurements in a broader frequency range [59], [60]. Moreover, artificial intelligence has yet to improve this field as drastically as it did a lot of scientific domains in the last decades: machine learning algorithms can be applied in the design process [24] or in data analysis, such as for classification [25] or personalized sensor calibration, to improve the performance of DS. Despite the advances in sensor miniaturization as well as measurement techniques and instruments, bringing DS to point-of-care applications remains challenging. Large, complicated, and costly VNA setups which require calibration before measurements restrain the application of the sensors for untrained users. Moreover, it causes the devices to be a chip-in-a-lab rather than a lab-on-a-chip. Recent progress in the development of VNA-on-a-chip is considered a game changer in that regard and is expected to bring the devices into the mobile testing domain. Finally, DS combined with ${\mu}{\text{W}}$ heating brings up new possibilities in cancer theranostics; the tumor can be detected through DS (diagnostics) and specifically targeted as its dielectric properties alter its sensitivity to ablation techniques (therapeutics).
[1] “Cancer,” World Health Organization, Geneva, Switzerland, Feb. 2022. [Online] . Available: https://www.who.int/news-room/fact-sheets/detail/cancer
[2] Y. Pan et al., “3D cell-based biosensor for cell viability and drug assessment by 3D electric cell/matrigel-substrate impedance sensing,” Biosensors Bioelectronics, vol. 130, pp. 344–351, Apr. 2019, doi: 10.1016/j.bios.2018.09.046.
[3] M. A. Aldhaeebi, K. Alzoubi, T. S. Almoneef, S. M. Bamatraf, H. Attia, and O. M. Ramahi, “Review of microwaves techniques for breast cancer detection,” Sensors, vol. 20, no. 8, Apr. 2020, Art. no. 2390, doi: 10.3390/s20082390.
[4] P. Mehrotra, B. Chatterjee, and S. Sen, “EM-wave biosensors: A review of RF, microwave, mm-wave and optical sensing,” Sensors, vol. 19, no. 5, Feb. 2019, Art. no. 1013, doi: 10.3390/s19051013.
[5] H. P. Schwan, “Electrical properties of tissue and cell suspensions,” Adv. Biol. Med. Phys., vol. 5, pp. 147–209, Jan. 1957, doi: 10.1016/B978-1-4832-3111-2.50008-0.
[6] G. Maenhout, “Dielectric detection and ablation techniques for endoscopic cancer theranostics,” Ph.D. dissertation, KU Leuven, Leuven, Belgium, 2022.
[7] J. C. M. Hwang, “Label-free noninvasive cell characterization: A methodology using broadband impedance spectroscopy,” IEEE Microw. Mag., vol. 22, no. 5, pp. 78–87, May 2021, doi: 10.1109/MMM.2021.3056834.
[8] “Basics of measuring the dielectric properties of materials,” Keysight Technologies, Halasuru, India, Aug. 2020. [Online] . Available: https://www.keysight.com/in/en/assets/7018-01284/application-notes/5989-2589.pdf
[9] C. Aydinalp, S. Joof, I. Dilman, I. Akduman, and T. Yilmaz, “Characterization of open-ended coaxial probe sensing depth with respect to aperture size for dielectric property measurement of heterogeneous tissues,” Sensors, vol. 22, no. 3, Jan. 2022, Art. no. 760, doi: 10.3390/s22030760.
[10] M. Lazebnik et al., “A large-scale study of the ultrawideband microwave dielectric properties of normal, benign and malignant breast tissues obtained from cancer surgeries,” Phys. Med. Biol., vol. 52, no. 20, pp. 6093–6115, Oct. 2007, doi: 10.1088/0031-9155/52/20/002.
[11] A. P. O’Rourke et al., “Dielectric properties of human normal, malignant and cirrhotic liver tissue: In vivo and ex vivo measurements from 0.5 to 20 GHz using a precision open-ended coaxial probe,” Phys. Med. Biol., vol. 52, no. 15, p. 4707, Jul. 2007, doi: 10.1088/0031-9155/52/15/022.
[12] T. Yilmaz et al., “Microwave dielectric spectroscopy of renal calculi: A large scale study on dielectric properties from 500 MHz to 18 GHz,” IEEE Trans. Dielectr. Electr. Insul., vol. 26, no. 5, pp. 1425–1433, Oct. 2019, doi: 10.1109/TDEI.2019.008029.
[13] S. Liu et al., “Hybrid characterization of nanolitre dielectric fluids in a single microfluidic channel up to 110 GHz,” IEEE Trans. Microw. Theory Techn., vol. 65, no. 12, pp. 5063–5073, Dec. 2017, doi: 10.1109/TMTT.2017.2731950.
[14] R. N. Simons, Coplanar Waveguide Circuits, Components, and Systems. New York, NY, USA: Wiley, 2001.
[15] “Coplanar waveguide,” Microwaves101, Tucson, AZ, USA. [Online] . Available: https://www.microwaves101.com/encyclopedias/coplanar-waveguide
[16] F. Artis et al., “Microwaving biological cells: Intracellular analysis with microwave dielectric spectroscopy,” IEEE Microw. Mag., vol. 16, no. 4, pp. 87–96, Mar. 2015, doi: 10.1109/MMM.2015.2393997.
[17] K. Grenier et al., “Recent advances in microwave-based dielectric spectroscopy at the cellular level for cancer investigations,” IEEE Trans. Microw. Theory Techn., vol. 61, no. 5, pp. 2023–2030, May 2013, doi: 10.1109/TMTT.2013.2255885.
[18] X. Du, C. Ferguson, X. Ma, X. Cheng, and J. C. M. Hwang, “Ultra-wideband impedance spectroscopy of the nucleus in a live cell,” IEEE J. Electromagn., RF, Microw. Med. Biol., vol. 6, no. 2, pp. 267–272, Jun. 2022, doi: 10.1109/JERM.2021.3121258.
[19] O. Peytral-Rieu, K. Grenier, and D. Dubuc, “Microwave sensor dedicated to the determination of the dielectric properties of 3D biological models from 500 MHz to 20 GHz,” in Proc. IEEE MTT-S Int. Microw. Symp. (IMS), Jun. 2021, pp. 222–225, doi: 10.1109/IMS19712.2021.9574794.
[20] A. Zedek, D. Dubuc, and K. Grenier, “Microwave permittivity extraction of individual biological cells submitted to different stimuli,” in Proc. IEEE MTT-S Int. Microw. Symp. Dig. (IMS), Jun. 2017, pp. 865–868, doi: 10.1109/MWSYM.2017.8058718.
[21] A. Tamra, M. P. Rols, D. Dubuc, and K. Grenier, “Impact of a chemical stimulus on two different cell lines through microwave dielectric spectroscopy at the single cell level,” in Proc. IEEE MTT-S Int. Microw. Biomed. Conf. (IMBioC), May 2019, pp. 1–4, doi: 10.1109/IMBIOC.2019.8777745.
[22] A. Tamra, A. Zedek, M.-P. Rols, D. Dubuc, and K. Grenier, “Single cell microwave biosensor for monitoring cellular response to electrochemotherapy,” IEEE Trans. Biomed. Eng., vol. 69, no. 11, pp. 3407–3414, Nov. 2022, doi: 10.1109/TBME.2022.3170267.
[23] F. Artis, D. Dubuc, J.-J. Fournie, M. Poupot, and K. Grenier, “Microwave dielectric spectroscopy of cell membrane permeabilization with saponin on human B lymphoma cells,” in Proc. IEEE MTT-S Int. Microw. Symp. (IMS), Jun. 2014, pp. 1–4, doi: 10.1109/MWSYM.2014.6848520.
[24] B. X. Wang, W. S. Zhao, D. W. Wang, J. Wang, W. Li, and J. Liu, “Optimal design of planar microwave microfluidic sensors based on deep reinforcement learning,” IEEE Sensors J., vol. 21, no. 24, pp. 27,441–27,449, Dec. 2021, doi: 10.1109/JSEN.2021.3124294.
[25] J. A. Osterberg et al., “Microwave sensing of yeast cell species and viability,” IEEE Trans. Microw. Theory Techn., vol. 69, no. 3, pp. 1875–1886, Mar. 2021, doi: 10.1109/TMTT.2020.3048176.
[26] G. Poiroux et al., “Label-free detection of mitochondrial activity with microwave dielectric spectroscopy,” Int. J. Biotechnol. Bioeng., vol. 6, no. 5, pp. 1–14, Jul. 2020.
[27] D. Dubuc, K. Grenier, F. Morfoisse, and B. Susini-Garmy, “In vitro and in vivo investigations toward near-field microwave-based detection of melanoma,” in Proc. 1st IEEE MTT-S Int. Microw. Biomed. Conf. (IMBioC), May 2017, pp. 1–4, doi: 10.1109/IMBIOC.2017.7965789.
[28] O. Peytral-Rieu, D. Dubuc, and K. Grenier, “Microwave microfabricated sensor dedicated to the dielectric characterization of biological microtissues,” in Proc. 3rd URSI Atlantic Asia Pacific Radio Sci. Meeting (at-AP-RASC), Jun. 2022, pp. 1–4, doi: 10.23919/AT-AP-RASC54737.2022.9814189.
[29] T. Chen, F. Artis, D. Dubuc, J.-J. Fournie, M. Poupot, and K. Grenier, “Microwave biosensor dedicated to the dielectric spectroscopy of a single alive biological cell in its culture medium,” in Proc. IEEE MTT-S Int. Microw. Symp. Dig. (MTT), Jun. 2013, pp. 1–4, doi: 10.1109/MWSYM.2013.6697740.
[30] S. Nourinovin, M. M. Rahman, R. C. Jones, M. P. Philpott, and A. Alomainy, “Impact of drug treatment on the electromagnetic properties of basal cell carcinoma cancer in the terahertz band,” in Proc. 16th Eur. Conf. Antennas Propag. (EuCAP), Apr. 2022, pp. 1–4, doi: 10.23919/EuCAP53622.2022.9769353.
[31] X. Wu et al., “Biomedical applications of terahertz spectra in clinical and molecular pathology of human glioma,” Spectrochimica Acta A, Mol. Biomolecular Spectrosc., vol. 285, Jan. 2023, Art. no. 121933, doi: 10.1016/j.saa.2022.121933.
[32] P. Yu et al., “Myocardial amyloidosis detection with terahertz spectroscopy,” IEEE Sensors J., vol. 22, no. 3, pp. 2389–2398, Feb. 2022, doi: 10.1109/jsen.2021.3133294.
[33] I. Nardecchia et al., “Out-of-equilibrium collective oscillation as phonon condensation in a model protein,” Physical Rev. X, vol. 8, no. 3, Oct. 2018, Art. no. 031061, doi: 10.1103/PhysRevX.8.031061.
[34] M. Lechelon et al., “Experimental evidence for long-distance electrodynamic intermolecular forces,” Sci. Advances, vol. 8, no. 7, Feb. 2022, Art. no. eabl5855, doi: 10.1126/sciadv.abl5855.
[35] S. Laurette, A. Treizebre, and B. Bocquet, “Co-integrated microfluidic and THz functions for biochip devices,” J. Micromechanics Microengineering, vol. 21, no. 6, May 2011, Art. no. 065029, doi: 10.1088/0960-1317/21/6/065029.
[36] A. M. Mohammed, Y. Wang, and M. J. Lancaster, “3D printed coaxial microwave resonator sensor for dielectric measurements of liquid,” Microw. Opt. Technol. Lett., vol. 63, no. 3, pp. 805–810, Mar. 2021, doi: 10.1002/mop.32679.
[37] Y. J. Mao, Y. J. Zhang, Z. R. Chen, and M. S. Tong, “A noncontact microwave sensor based on cylindrical resonator for detecting concentration of liquid solutions,” IEEE Sensors J., vol. 21, no. 2, pp. 1208–1214, Jan. 2021, doi: 10.1109/JSEN.2020.3016290.
[38] R. S. Hassan, S. I. Park, A. K. Arya, and S. Kim, “Continuous characterization of permittivity over a wide bandwidth using a cavity resonator,” J. Electromagn. Eng. Sci., vol. 20, no. 1, pp. 39–44, Jan. 2020, doi: 10.26866/jees.2020.20.1.39.
[39] C. Watts, S. M. Hanham, J. P. K. Armstrong, M. M. Ahmad, M. M. Stevens, and N. Klein, “Microwave dielectric sensing of free-flowing, single, living cells in aqueous suspension,” IEEE J. Electromagn., RF, Microw. Med. Biol., vol. 4, no. 2, pp. 97–108, Jun. 2020, doi: 10.1109/JERM.2019.2932569.
[40] C. F. Liu, M. H. Wang, and L. S. Jang, “Microfluidics-based hairpin resonator biosensor for biological cell detection,” Sensors Actuators B, Chem, vol. 263, pp. 129–136, Jun. 2018, doi: 10.1016/j.snb.2018.01.234.
[41] I. Piekarz et al., “Planar single and dual-resonant microwave biosensors for label-free bacteria detection,” Sensors Actuators B, Chem., vol. 351, Jan. 2022, Art. no. 130899, doi: 10.1016/j.snb.2021.130899.
[42] M. Russel et al., “High-frequency, dielectric spectroscopy for the detection of electrophysiological/biophysical differences in different bacteria types and concentrations,” Analytica Chimica Acta, vol. 1028, pp. 86–95, Oct. 2018, doi: 10.1016/j.aca.2018.04.045.
[43] G. Bridges, T. Cabel, S. Afshar, E. Salimi, D. Thomson, and M. Butler, “Microwave near-field detection of single biological cells and nanoparticles,” in Proc. 18th Int. Symp. Antenna Technol. Appl. Electromagn. (ANTEM), Aug. 2018, pp. 1–2, doi: 10.1109/ANTEM.2018.8572975.
[44] M. C. Jain, A. V. Nadaraja, S. Mohammadi, B. M. Vizcaino, and M. H. Zarifi, “Passive microwave biosensor for real-time monitoring of subsurface bacterial growth,” IEEE Trans. Biomed. Circuits Syst., vol. 15, no. 1, pp. 122–132, Feb. 2021, doi: 10.1109/TBCAS.2021.3055227.
[45] R. Narang et al., “Sensitive, real-time and non-intrusive detection of concentration and growth of pathogenic bacteria using microfluidic-microwave ring resonator biosensor,” Scientific Rep., vol. 8, no. 1, Dec. 2018, Art. no. 15807, doi: 10.1038/s41598-018-34001-w.
[46] S. Mohammadi, R. Narang, M. Mohammadi Ashani, H. Sadabadi, A. Sanati-Nezhad, and M. H. Zarifi, “Real-time monitoring of Escherichia coli concentration with planar microwave resonator sensor,” Microw. Opt. Technol. Lett., vol. 61, no. 11, pp. 2534–2539, Nov. 2019, doi: 10.1002/mop.31913.
[47] M. C. Jain, A. V. Nadaraja, B. M. Vizcaino, D. J. Roberts, and M. H. Zarifi, “Differential microwave resonator sensor reveals glucose-dependent growth profile of E. coli on solid agar,” IEEE Microw. Wireless Compon. Lett., vol. 30, no. 5, pp. 531–534, May 2020, doi: 10.1109/LMWC.2020.2980756.
[48] M. C. Jain, A. V. Nadaraja, R. Narang, and M. H. Zarifi, “Rapid and real-time monitoring of bacterial growth against antibiotics in solid growth medium using a contactless planar microwave resonator sensor,” Scientific Rep., vol. 11, no. 1, Dec. 2021, Art. no. 14775, doi: 10.1038/s41598-021-94139-y.
[49] F. Deshours et al., “Improved microwave biosensor for non-invasive dielectric characterization of biological tissues,” Microelectronics J., vol. 88, pp. 137–144, Jun. 2019, doi: 10.1016/j.mejo.2018.01.027.
[50] S. J. Park, R. Tucker, E. Pickwell-MacPherson, and J. E. Cunningham, “Design of a split ring resonator integrated with on-chip terahertz waveguides for colon cancer detection,” Adv. Theory Simul., vol. 5, no. 9, Sep. 2022, Art. no. 2200313, doi: 10.1002/adts.202200313.
[51] A. E. Omer et al., “Low-cost portable microwave sensor for non-invasive monitoring of blood glucose level: Novel design utilizing a four-cell CSRR hexagonal configuration,” Scientific Rep., vol. 10, no. 1, Dec. 2020, Art. no. 15200, doi: 10.1038/s41598-020-72114-3.
[52] A. E. Omer, G. Shaker, and S. Safavi-Naeini, “Portable radar-driven microwave sensor for intermittent glucose levels monitoring,” IEEE Sens. Lett., vol. 4, no. 5, May 2020, Art no. 3500604, doi: 10.1109/LSENS.2020.2986208.
[53] A. E. Omer, G. Shaker, S. Safavi-Naeini, H. Kokabi, G. Alquie, and F. Deshours, “Compact honey-cell CSRR-based microwave newline biosensor for monitoring glucose levels,” in Proc. 14th Eur. Conf. Antennas Propag. (EuCAP), Mar. 2020, pp. 1–5, doi: 10.23919/EuCAP48036.2020.9135301.
[54] A. E. Omer et al., “Non-invasive real-time monitoring of glucose level using novel microwave biosensor based on triple-pole CSRR,” IEEE Trans. Biomed. Circuits Syst., vol. 14, no. 6, pp. 1407–1420, Dec. 2020, doi: 10.1109/TBCAS.2020.3038589.
[55] A. E. Omer, G. Shaker, and S. Safavi-Naeini, “Wearable CSRR-based sensor for monitoring glycemic levels for diabetics,” in Proc. IEEE 20th Int. Conf. Bioinf. Bioengineering (BIBE), Oct. 2020, pp. 922–928, doi: 10.1109/BIBE50027.2020.00156.
[56] G. Maenhout, T. Markovic, and B. Nauwelaers, “Flexible, segmented tubular design with embedded complementary split-ring resonators for tissue identification,” IEEE Sensors J., vol. 21, no. 14, pp. 16,024–16,032, Jul. 2021, doi: 10.1109/JSEN.2021.3075570.
[57] J. Bai, H. Guo, H. Li, C. Zhou, and H. Tang, “Flexible microwave biosensor for skin abnormality detection based on spoof surface plasmon polaritons,” Micromachines, vol. 12, no. 12, Dec. 2021, Art. no. 1550, doi: 10.3390/mi12121550.
[58] M. K. Sharma, R. Kumari, A. Mittal, M. K. Upadhyay, A. Mittal, and K. Singh, “Design and modeling of the ring resonator-based microwave sensor for skin cancer detection,” in Flexible Electronics for Electric Vehicles, S. Dwivedi, S. Singh, M. Tiwari, and A. Shrivastava, Eds. Singapore, Singapore: Springer-Verlag, 2023, pp. 47–58.
[59] X. Bao et al., “Broadband dielectric spectroscopy of cell cultures,” IEEE Trans. Microw. Theory Techn., vol. 66, no. 12, pp. 5750–5759, Dec. 2018, doi: 10.1109/TMTT.2018.2873395.
[60] G. M. Dittami, H. E. Ayliffe, C. S. King, and R. D. Rabbitt, “A multilayer MEMS platform for single-cell electric impedance spectroscopy and electrochemical analysis,” J. Microelectromech. Syst., vol. 17, no. 4, pp. 850–862, Aug. 2008, doi: 10.1109/JMEMS.2008.921726.
Digital Object Identifier 10.1109/MMM.2022.3233510