A Liquid Sensor Based on Frequency Selective Surfaces

A novel, simple and easy-to-fabricate liquid sensor using frequency-selective surfaces (FSSs) is proposed. The new sensor concept is based on modifying the capacitance between adjacent FSS elements when materials of different electrical characteristics are inserted. The change in capacitance produces a change in resonant frequency. The FSS design consists of a 9 <inline-formula> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula> 9 array of square loops on <inline-formula> <tex-math notation="LaTeX">$0.31\lambda \times 0.31\lambda $ </tex-math></inline-formula> square unit cells with trenches between the loops. The trenches are filled with liquids under test (LTUs). The structure operates at 4.6 GHz without any liquid. When liquids are inserted in the trenches, the resonance frequency varies in relation to the dielectric constant of the liquid. This is observed by measuring the transmission coefficient (<inline-formula> <tex-math notation="LaTeX">$S_{21}$ </tex-math></inline-formula>). Butan-1-ol, ethanol, methanol, propan-2-ol, and xylene are used to demonstrate the sensing function. A maximum sensitivity of 8.65% for xylene was achieved. Furthermore, very low differences were observed between the measured and expected dielectric constant and loss tangent, thus validating the design. The device is inexpensive, compact, and easy to make and scalable for large-area operations in liquid detection for microwave sensing applications. This technique has potential applications in reconfigurable FSS.

in the transmission/reflection response. At single or discrete set of frequencies, resonant technique offers the potential for accurate measurements. Waveguide, dielectric, and coaxial cavity resonators have traditionally been employed for materials characterization [4], [5]. The material to be characterized is inserted in the cavity location where the electric field is at its maximum causing cavity perturbation. The introduction of material results in changes in the resonant frequency of the cavity [3]. Exploiting this phenomenon, several sensing devices have been developed. In [6], a planar substrate integrated waveguide cavity resonator is presented where microfluidic capillaries are used for the insertion of the material. This method is also employed in [7] and [8] with the implementation of a split-ring resonator with a microfluidic channel. The split-ring and capillary concepts are combined in [9] where a complementary split-ring resonator (CSRR) is presented for dielectric characterization of liquids. This is further developed in [10] where an open complementary split-ring resonator is employed with a slot container. In [11], dual mode, quarter ring microstrip resonators microfluidic sensor with capillary are used. In [12], a sensor that exploits the excitation of a resonant mode within a cavity containing a minute sample of a liquid is presented. In [13], an RFID sensor is presented for identification and evaluation of liquid chemical based on the shift in the resonant frequency of an applied UHF RFID chipped tag.
Metamaterials are artificially constructed composites that exhibit electromagnetic properties that are difficult to achieve using conventional or naturally occurring materials [14] and include electromagnetic bandgap (EBG) structures [15] and FSS [16]. They have recently been used for sensing liquids.
In [17], [18], and [19], metamaterial absorber sensors are proposed where an air gap between a copper plate and backside resonator is used to fill chemicals' liquids. EBGs have recently been employed in wireless sensor design in [20] by creating trenches between adjacent cells. This concept was advanced in [21] using a Casero Fractal design element.
In general, these liquid sensors provide good sensitivity and accuracy, but they also present design and operational complexities, and bulkiness, while others are delicate and require intricate assembly. Their design structure requires complicated measurement setups followed by complex design construction. Further costs accrue due to additional drilling and cutting for microfluidic subsystems. Extra parts, e.g., capillaries, capsules, and liquid containers, would then be needed, which further increases the costs. Others EBG-based [20], [21] are two parts designs, which must be intricately assembled beside This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ the design complexity. A single unit, which is easy to make and use sensor, implies simplicity, lower cost, and ease of assembly.
Frequency-selective surfaces (FSSs) are electromagnetic filtering structures usually consisting of arrays of periodic conductors [22], [23] on a supporting dielectric material. They are a type of metamaterial [24]. FSSs modify the incoming electromagnetic signal in relation to their instrinsic resonant frequency. Their responses are dependent on the geometry and spacing of the element, dielectric material thickness, and properties as well as that of its surroundings [23]. This property can be used to develop contactless sensors. FSS sensors have been designed to monitor structural health [25], [26], and [27], sensing of temperature and strain [28], breathing [29], strain [30], and dielectric characterization [31], [32].
In this article, an FSS-based sensor for sensing liquids falling onto a surface is proposed. In the new concept outlined in this work, the sensing function is achieved by altering the capacitance between adjacent FSS elements when materials of different electrical characteristics are inserted in trenches. The change in capacitance produces a change in the frequency resonance, which is detected as transmission response of the FSS. Trenches were dug in the dielectric around the conductors. The basic principle of operation of the proposed sensor is the excitation of a resonant mode of the FSS with the trenches filled with different liquids under test (LUTs). The FSS sensor operates at about 4.6 GHz with a broad range of frequencies in relation to the LUT. To the best of the authors' knowledge, this is the first design based on the FSS structure for sensing of liquids. The proposed sensor is scalable and can be redesigned to operate in different frequency bands. The motivation is the creation of a wireless sensor for the detection of liquids falling into a surface. The main application is on the detection on medium-to-large surfaces that can be adjusted by the array size. A curve-fitting approach is employed to estimate the LUT complex permittivity and sensitivity analysis based on the shifted parameters. The rest of this article is organized as follows. Section II details the geometry of the unit cell, the integrated trenches, and its characteristic behavior. Section III details the fabrication and measurements of the design, while Section IV discusses the performance evaluation of the proposed sensor. Finally, Section V discusses the conclusion.

A. FSS Sensor Unit Cell Design
The FSS was designed, simulated, and tested using CST Microwave Studio 1 . For simplicity, the FSS was simulated as an infinite structure using a unit cell with periodic boundaries. Fig. 1(a) shows the geometry of the unit cell of the FSS sensor design. A band stop FSS was selected as changes in the sensors could be observed through nulls in the transmission responses. In terms of element, a square loop FSS was used because it offers dual polarization and good angle of incidence behavior.
In the design, the square loops were made of copper and the substrate was RT/Duroid 5880 of thickness 3.175 mm, 1 Trademark.  dielectric constant ε r of 2.2, and loss tangent (tan δ) of 0.0009. The cell had 1 mm wide and 2 mm deep trenches around its edges. The trenches were introduced through the partial removal of the substrate around the square loop, as shown in Fig. 1(b). The trenches reduce the amount of dielectric material, which decreases the permittivity and losses of the FSS in free space. The unit cell and the square loop dimensions were optimized to operate at 4.6 GHz. The gap × between the conductor loop and the edge of the unit cell is 1 mm. The periodicity, P, of the square loop array is the sum of g, the gap between adjacent loops, and d 2 , the length of the conductor. d 1 is the length of the unit cell element. The design dimensions are shown in Table I. Fig. 2 shows the simulated transmission response of the FSS for angle of incidence at 0 • incidence, 45 • TE, and 45 • TM incidence. TE denotes the E-plane response, whereas TM denotes the H-plane response. Resonance of the unit cell occurred at normal incidence with a frequency of 4.6 GHz with no appreciable frequency drift at TE 45 • and TM 45 • , i.e., 45 • off broadside illumination for TE and TM polarization.

B. Square Loop Operations and Equivalent Circuit (EC)
The square loop FSS has band reject frequency response dependent on its physical dimensions [33]. It is a relatively simple shape and ideal to build a prototype for performance assessment or application. The EC of the square loop FSS has been investigated in [34], [35], and [36]. Fig. 3(a) presents the EC of the square loop band stop filter. The EC model provides a simple and fast method of FSS analysis that is supported by transmission line analogy in which equivalent capacitive (C i ) and inductive (L i ) lumped components construct of the FSS can be calculated. η is the characteristic intrinsic impedance of free space equal to 377 . The proposed FSS array is symmetrical with periodicity P and a gap g between two adjacent conductor loops. It can be represented as lumped elements consisting of capacitance (C i ) and inductance (L i ). Gap g represents the capacitive elements, while the length of the conductor represents the inductive element. Resonance occurs when the perimeter of the loop is approximately one wavelength. The inductance and capacitance values determine the resonance frequency ω as The dielectric material affects the reflection or transmission responses. With a finite thickness, the dielectric material effect on the FSS structures can be explained using the EC analysis. The designed inductive FSS is a square loop of width t. Its intrinsic capacitance C i and inductance, L i , for TE incident wave can be approximated from path capacitance and strip inductance using (2) and (5), respectively, a conducting strips approximation developed by [37] that enables the calculation of L i and C i values. For transverse electrical (TE) incidence wave, the vertical strips of the FSS act as a L i impedance in the EC and the horizontal gratings as a C i impedance [33] of where C i is the intrinsic capacitance between adjacent loops, determined by periodicity P and the gap g between adjacent loops. G(λ, g, P) is a correction term expressed by and the factor Q 2 is given by where the value of n in this case is 2. To approximate the strip inductance, L i , g in the correction term (3) is substituted by t, and L i results in where L i is the strip inductance determined by conductors of periodicity P, width t, c is the speed of light, and η is the impedance of free space.
From (1) and (5), the theoretical component values for freestanding FSS are L i = 3.8114 nH and C i = 214.26 fF. This capacitance needs to be adjusted due to the effect of dielectric. In a one-sided substrate, the capacitance increases proportional to the effective relative permittivity [23], [38] given by Although (6) is a general equation for the cases when the substrate is sufficiently thick (>λ/5) [33], it can be a good approximation in the case of square loops even in substrates of thicknesses of about 0.05λ [39]. Therefore, using (6), the resulting capacitance when compensated by the dielectric substrate is C i = 342.9 fF. The resonant frequency for this LC circuit is approximately 4.5 GHz, which is within the expected 5% error margin of the frequency of 4.3 GHz for the FSS obtained from the simulation of the dielectric-loaded FSS (no trenches). In liquid sensing, the dielectric losses of the liquids in the trenches can be high. These losses can be represented as a resistor in parallel with the capacitor, as shown in Fig. 3(b).

C. Parametric Analysis
Parametric study was conducted to determine the behavior of the FSS for varying empty trenches' depth (h 1 ), from 0.0 to 3.0 mm, at 0.5 mm intervals. Fig. 4 shows the effect of changing trench depth (h 1 ) on resonant frequency and the corresponding capacitance. In our model, the inductance is constant (L i = 3.8114 nH) [22]. The capacitance has been calculated using the frequency obtained in simulations and (1). As expected, removing dielectric material decreases the effective permittivity and capacitance and increases the resonant frequency. The capacitance as a function of depth, h 1 , follows the fit equation:  To assess the potential behavior of the proposed FSS structure (see Fig. 1) as a liquid sensor, a simulation was carried with the trenches filled with materials of various dielectric constants and loss tangents. Fig. 5 shows the FSS structure indicating the sections of the design with liquid. In the initial study, shown in Figs. 6 and 7, the dielectric permittivity was varied from 1 to 20, and the loss tangent was fixed at 0. The resonant frequency decreased with increase of the dielectric constant. From (1), capacitance C i due to liquids of varying ε r values in the trenches was obtained and is also shown in Fig. 7. C i increases linearly with ε r as C i = 33.312ε r + 288.89 fF.
The effect of changes in loss tangent on the transmission response and resonance frequency is shown in Figs. 8 and 9,   respectively. The permittivity of the liquid was kept at 8, while the loss tangent was varied between 0 and 0.9. The depth of the null in the transmission coefficient curve decreases with the increase in loss tangent. A small decrease in resonant frequency is also observed, with a maximum of about 6% shift for a value of tan δ of 0.9.

D. Sensor Study for Readily Available Liquids
Five chemical liquids were used to authenticate the sensing/detecting functionality of the proposed design. These were Butan-1-o1, Propan-2-o1, ethanol, and methanol whose dielectric properties at 5 GHz and 20 • were obtained from [40] and xylene from [41] and are shown in Table II. Fig. 10 shows S 21 of the unit cell elements when the trenches are filled with the different LUT.

A. Fabrication
The FSS unit cell with dimensions as in Fig. 1 and Table I was extended to a 9 × 9 array to create a prototype of the sensor. The model of the sensor is shown in Fig. 11 with the liquid-filled trenches depicted in red. It has a 5 mm wide edge to hold the liquids inside and for mechanical purposes, making a total dimension of 192 × 192 mm. The prototype was fabricated at Printech Circuit Laboratories [42] using standard printed circuit board (PCB) procedures. It was then milled using a high-precision milling machine to create the trenches around the loops. The fabricated sensor is shown in Fig. 12(a) while Fig. 12(b) shows the milled trenches.

B. Measurements and Results
Using a Marconi Instruments microwave test set 6204B and in a plain wave chamber, the insertion loss of the FSS was measured with empty trenches. Fig. 12(c) shows the measurement setup. The trenches were then filled using a syringe with the correct amount of liquid (14 mL) and the transmission response (S 21 ) measured. The FSS structure was kept horizontal using a bubble level to ensure that the liquid was uniform throughout the structure. The S 21 results are shown in Fig. 13. Table III tabulates both the simulated and measured resonance frequency ( f s ) and the frequency shift  ( f ) from the reference of the LUT. The resonance frequency for both the simulated and measured shows the frequency shift toward lower frequencies from the reference frequency, i.e., air resonance. This agrees with similar observations in [43] and [44]. The resonance points shift to the lower frequencies when the trench is filled with liquids of high dielectric constant. The differences between simulated and measured results could be due to fabrication errors, potential loss of liquid when inserting into the trenches, quality of the LUT, and possible differences in the permittivity values of LUT from those given in [40].

A. Sensitivity Analysis
The proposed design performance evaluation was conducted with the integrated fluidic trenches on the FSS filled with the five different liquids of different electrical properties described in Section II. The sensitivity (S) as a performance parameter was used to evaluate the sensor performance according to the criterion proposed in [21] and [44]. To estimate the sensitivity (S) of the device, air-filled trenches were chosen as the reference. The proposed criterion gives the mathematical expression of the sensitivity, S, of where f o and f s are the resonance frequency of the FSS when filled with air and the LUT, respectively. ε is (ε s − ε o ), where ε s and ε o are dielectric constants of the LUT in the trench and air, respectively. Table IV shows the frequency shift and the computed sensitivity (S) of the various LUTs.

B. Material Dielectric Characterization
The complex permittivity (ε r ) and loss tangent (tanδ) of the LUT were calculated using mathematical models derived using the polynomial curve-fitting technique [2] and [21]. The necessary equations that best fit the given data set were generated to a 3 • for greater accuracy. The resonating frequency ( f s ) and bandwidth of frequency shifting ( f ) were used to determine the dielectric constant and loss tangent of the LUT, respectively, to a good degree of accuracy. The relationship between the resonating frequency variations with respect to the change in the dielectric values of LUT was expressed using the third-order polynomial expression ε r = −1.6763 f 3 s + 24.8613 f 2 s − 121.5387 f s + 198.7187 (10) where ε r is the calculated real permittivity and f s is the LUT resonating frequency. The relationship between the bandwidth of frequency shifting and loss tangent (tan δ) was mathematically derived as a third-order polynomial as the best fit of the given dataset and is expressed by where tan δ is the calculated loss tangent and f is the LUT frequency shift from the reference. From (8) and (9), the dielectric constant and loss tangent of samples of LUT of unknown permittivity and loss tangent were calculated and compared to the actual published values to validate the data. The calculated values of ε r and tan δ are shown in Table V. The calculated results demonstrate a good agreement with actual permittivity and loss tangents values of all the LUT. The root-mean-square error between the actual and the calculated permittivity and the loss tangent was also calculated. The highest error of 1.7% and 3.3% for dielectric tangent and loss tangent, respectively, validates the performance of the device. The error could be due to the scattering parameters of the microwave sensor being influenced by the electric and magnetic properties of its surroundings, which can be reflected in the insertion loss S 21 of the resonator. Impurities and analytes in the trench can have an effect on the amplitude tab3

V. DISCUSSION AND CONCLUSION
A new concept for microwave liquid sensors based on FSS has been demonstrated. The sensor uses trenches between FSS elements to modify the transmission response when liquids with different permittivity are inserted. A mathematical model was developed for determination of the dielectric constant and loss tangent of the LUT. The evaluated complex permittivity and loss tangent are in good agreement with the actual values. The highest errors in the measured values of the dielectric constant and dielectric loss are within 1.7% for Propan-2-o1 and 3.3% for xylene, respectively. This demonstrates that the proposed microwave sensor is suitable as a low-cost platform for the detection of liquids with good sensitivity and low detection error. Table VI summarizes the performance analysis of the reported metamaterial EBG detectors closest to the proposed design. It shows that, besides its simplicity, it offers higher sensitivity and thus improved the detection performance.
The developed sensor is ideal for instances requiring robust, real-time monitoring at low cost and complexity, low power consumption, and simple fabrication techniques. The independent resonant nature due to each chemical allows the device to be reused for different and various liquids. The contactless nature of the detection process makes it suitable for a safe working environment. For example, it can be used in industrial processes where dangerous chemicals may need to be identified. Furthermore, the size of the FSS sensor can be scaled up or down as required. Previous work [45] has indicated that even a unit cell is able to produce a transmission response by adjusting the distance of the antennas used for testing. Fig. 14 shows a potential setup integrated in a wireless sensor network. The sensor is set in such a way that liquids can fall into the FSS structure. To ensure the uniformity of the liquid during the measurements, a bubble level should be used during the installation to prevent erroneous readings. The sensor has a flat surface, which means that some droplets may remain on the surface outside the trenches, which may affect measurements. A future version of the design is envisaged where cells could be three-dimensional with a slight slope to allow droplets to be guided toward the trenches.
Note that this is the first concept and the reading setup is the standard for FSS measurements. However, a more compact arrangement could be developed in the future where the antennas and system only target the specified frequency band of operation. In addition, in an industrial environment, a dedicated space would be desirable.
Even though the main application described here is on liquid sensing, the structure and design have potential applications in reconfigurable FSS using low-loss liquids such as the one employed in [46].