Highly Sensitive Terahertz Dielectric Sensor for Liquid Crystal

Abstract

This paper presents the design and process of two highly sensitive sensors working in the terahertz band. The sensors comprise the quartz substrate, medium, reflection plate, and metal resonant layer with a symmetrical single-slot patch array. The devices help study the electrically induced permittivity of two liquid crystals in different frequency bands and at different voltages, and the experimental data verify that both liquid crystals have a large birefringence. Based on experimental results, the sensitivity of the fabricated sensor is 47.03 GHz/RIU in the frequency range 90–140 GHz. Similarly, the other fabricated sensor has a sensitivity of 112.47 GHz/RIU in the frequency range 325–500 GHz. The results show that both sensors have superior sensing properties and potential applications in biological and chemical liquid sensing.

1. Introduction

Terahertz technology has shown good application prospects and value in biomedical science, wireless communication, non-destructive testing, aerospace, and other fields [1,2,3,4,5]. In recent years, terahertz (THz) technology has attracted the attention of many researchers because of its particular position in the electromagnetic spectrum and its unique advantages different from other electromagnetic waves [6]. THz waves have low photon energy, which does not easily destroy the structure of the material under investigation and still has intense penetration like x-rays [7,8]. Notably, many organic molecules and biomolecules have unique signatures in the THz band, which is helpful in chemical and biosensing detection [9,10,11]. Thus, it enables label-free, non-destructive, non-contact detection through the analysis of the THz frequency spectrum [12,13,14].

For detecting trace analytes, the metamaterial (MM) sensors introduced in THz technology can achieve high sensitivity and non-destructive examination [15,16]. Metamaterials (MMs) are artificially constructed electromagnetic materials having negative refraction and stealth capability, and symmetrical array of artificial structures is the main way to construct THz metamaterials [17,18]. MM-based THz sensors are symmetrical periodic arrays of artificially designed symmetrical or asymmetrical resonant elements and significant sensitivity to minute amounts of trace analytes [19]. The trace analyte with a particular dielectric constant is placed on the sample layer of the MM-based THz sensor. The resonant frequency of the trace analyte changes when the permittivity, concentration, or thickness of the trace analyte placed on the sample layer of the metamaterial-based THz sensor, is changed [20,21,22].

The rapid development of THz technology provides efficient modulating techniques for THz waves. Liquid crystals (LCs) play an essential role in the THz tunable devices, such as phase shifters [23,24], absorbers [25,26,27], filters [28], and resonators [29], which are a few to mention. As a tunable material, the LCs are very sensitive to changes in the applied external fields, such as electric and magnetic fields. High-birefringence LCs can achieve more efficient dynamic regulation of terahertz waves [30]. Accurately obtaining the electromagnetic parameters of LCs is the basis of developing tunable parts. Hence, it is indispensable to study the characteristics of LCs in the THz frequency range.

In this paper, based on the symmetrical array consisting of slot unit cells, two highly sensitive THz metamaterial sensors for liquid dielectric detection are designed and processed. The working principle of this sensor is to convert the permittivity of the liquid into an electromagnetic parameter that can be measured by the sensor, such as the S parameter or the offset of the resonant point. The measurement relationship between the permittivity and the electromagnetic parameter is established to represent the permittivity by the electromagnetic parameter. The sensor structure based on THz band resonance has the advantages of small size, high test accuracy, and an extensive range of applications for the detection of objects. In order to make more liquid crystals deflect at the bias voltage, the copper area needs to be larger, which is why we set the resonant structure into a single-slotted shape. We tested two nematic liquid crystals, C–09–2 and M10, including their permittivity. The test results show that the sensors have good sensing characteristics and verified that two new liquid crystals, C–09–2 and M10, have high birefringence and stronger tuning ability for terahertz waves.

2. Sensor Design, Theoretical Analysis, and Manufacturing

Figure 1a is a schematic structural diagram of the THz metamaterial sensor designed to detect the dielectric properties of a liquid crystal, divided into seven layers with 30 × 30 units. From top to bottom, the following layers were used: quartz substrate, an array of single-slotted resonant structures, Polyimide (PI), LC, PI, metallic reflectors, and a quartz substrate. Figure 1b is a unit schematic diagram of the metal resonant layer. In order to get a better orientation of the liquid crystals induced by the bias voltage, the copper needs to occupy a larger proportional area. So, we designed the resonant structure into a single, symmetrical slotted shape [31].

Sensor-1 works at 90–140 GHz and sensor-2 works at 325–500 GHz and have exactly the same structure, with only the size being different. The symmetric metal resonant structure unit designed in this research is shown in Figure 1b. For the sensor-1, the optimized size of the resonant layer is found by numerical simulation, with cell cycles of p₁ = 1150 μm, slotted dimensions Lx₁ = 110 μm, and Ly₁ = 650 μm. The metal resonant layer and the metal reflector thicknesses Hc₁ = 0.5 μm. Moreover, the LC layer thickness Hlc₁ = 45 μm. Similarly, for the sensor-2, the dimensions of the p₂, Lx₂, and Ly₂ are 300 μm, 20 μm, and 150 μm, respectively. The thicknesses of both the resonant and reflective layers Hc₂ = 0.5 μm, and the thickness of the LC layer Hlc₂ = 45 μm. In addition, the quartz substrate has a thickness of Hq = 490 μm, a permittivity of 3.75, and a loss tangent of 0.002.

The performance of the sensors was analyzed by using the finite element method (FEM), assuming that the resonant layer unit is located in a periodic environment, illuminated by an x-polarization and normal incidence plane wave [32]. Since copper is not 100% covered, the force on the liquid crystal molecules far away from the copper-covered area is feeble, and crystal molecules cannot undergo deflection. Therefore, in the simulation, the refractive index of the liquid crystal layer in the region below the single-dipole slotted structure does not change as shown in Figure 1b. It is worth noting that the thickness of the polyimide (PI) layer is minimal and has a negligible effect on the simulation results. So, we ignore the PI layer in the simulation.

Figure 2 is a simulation diagram of the reflection properties of the cells at different refractive indices. It can be seen from Figure 2 that as the refractive index changes from 1.5 to 2, the resonance frequency of the sensor in the 90–140 GHz frequency band moves from 125 GHz to 101.65 GHz. Additionally, for the 325–500 GHz frequency band, the sensor’s resonance frequency shifts from 483.2 GHz to 427.7 GHz. The resonant frequency and amplitude change is more obvious in high frequency band. Simulation results show that both sensors have good sensing properties. We performed linear fitting to the simulation data, as shown in Figure 3. Figure 3a shows that the sensor has a sensitivity of 47.68 GHz/RIU in the 90–140 GHz, and Figure 3b shows that the sensor has a sensitivity of 111.43 GHz/RIU in the 325–500 GHz band. Since the accuracy of the equipment used for the fabrication is limited, we cannot just pursue high sensitivity in the design process but also need to consider the process accuracy.

The 0.5 μm thick single-slot array coated on the quartz substrate is prepared with evaporation, rotary coating, ultraviolet exposure, developing, etching, and other processes. The PI is spin-coated on the surface of the array and the reflector is orientation layer. In order to control the direction of the liquid crystal molecules as parallel to the substrate, we mechanically rubbed the PI along the x-axis direction, and 45 µm diameter polystyrene microspheres fixed the gap between the two quartz substrates. The upper and lower substrates are intentionally staggered to expose the two electrodes and conveniently fill liquid crystals. It is should be noted that the PI at the electrodes must be wiped off with alcohol. Liquid crystal is filled using the principle of siphon. To prevent the liquid crystal from leaking, we use epoxy resin to encapsulate the four sides of the liquid crystal layer.

The two fabricated single-slotted sensors are shown in Figure 4a,b. We can see that the sensor size for the 90–140 GHz band is 4 cm × 4 cm, and the size for the 325–500 GHz band is 2 cm × 2 cm. The images of the metal resonance layer under the microscope are shown in Figure 4c,d. We observed that the actual size of Lx1 in the 90–140 GHz frequency band is 109.635 μm, and the production error is 0.365 μm (0.3%). We built the model, simulated it with the actual measured size, and found that this 0.3% error shifted the sensor’s resonance frequency by 0.05 GHz towards the lower frequency end. For the sensors in the 325–500 GHz frequency band, the actual size Lx2 of the sensor is 19.924 μm, and the production error is 0.076 μm (0.38%). Similarly, we modeled and simulated with the measured dimensions and found that the 0.38% error shifted the sensor’s resonance frequency to the left by 0.1 GHz. Therefore, to improve the precision of the sensor, we subtracted this offset of the resonant frequency in advance for subsequent tests.

3. Measurement and Experimental Analysis

As shown in Figure 5, the test system was built with the vector network analyzer (Agilent N5224A), signal generator, voltage amplifier, horn antenna, and VNA extension module (N5262AW08), which measures the reflection loss of the liquid crystal at different bias voltages. The sample was placed in the far-field position of the antenna and was be covered with the absorbing material. The liquid crystals used in the experiment are C–09–2 and M10, with large birefringence. Liquid crystal mixture C–09–2 are formulated by isothiocyanatotolane liquid crystal components and M10 are specially designed for microwave wavelength [33]. Fitting the experimentally measured data with the simulation model gives the equivalent permittivity of the liquid crystal at different bias voltages.

Figure 6 shows the reflection losses of the liquid crystals C–09–2 and M10 at different voltages in the 90–140 GHz frequency band, where the liquid crystal molecules are parallel to the PI layer at the respective unbiased voltage. As the voltage increases, the deflection of the liquid crystal molecules increases, but the resonant frequency decreases until the full biased voltage. From Figure 6a, we can see that the resonant frequency is 120.1 GHz at 0 V, and the resonance frequency moves to the left as the voltage increases. At the voltage of 18 V, the resonant frequency is 104.8 GHz, and from this point onwards, the resonant frequency does not change with the increase in the voltage; hence, the full biased voltage is reached. Similarly, Figure 6b shows the reflection losses of the liquid crystal M10 at different voltages in the 90–140 GHz frequency band. With the unbiased voltage, the resonance frequency is 121.2 GHz. The resonance frequency moves to the left as the voltage gradually increases up to 16 V. The resonance frequency is 107.3 GHz at the full bias voltage of 16 V.

Similarly, we also tested the two liquid crystals in the 325–500 GHz frequency band. As shown in Figure 7a, for liquid crystal C–09–2, the resonant frequency is 470.58 GHz at 0 V voltage. As the voltage increases, the resonant frequency gradually moves to the left. At 18 V, the resonance frequency is 448.156 GHz, and for the voltage values greater than 18 V, the resonant frequency remains at 448.16 GHz. As shown in Figure 7b, for liquid crystal M10, the resonant frequency is 471.02 GHz at 0 V voltage. As the voltage increases, the resonant frequency gradually moves to the left. At 20 V, the resonance frequency is 449.14 GHz. From 20 V onwards, the resonant frequency remains unchanged as the voltage increases, which means the liquid crystal is wholly deflected.

To further understand the dielectric properties of the liquid crystals at different voltages, we fitted the test results to the simulation results. To see the fitting results more intuitively, we show the fitting of test data to the simulation data in Figure 8. The fitting method used is described in the work presented in the reference [34].

As shown in Figure 8a the fitting results, in 90–140 GHz frequency bands, the liquid crystal C–09–2 has a permittivity ε_min = 2.57 at a resonant frequency of 120.1 GHz. It has a permittivity ε_max = 3.73 at a resonant frequency of 104.8 GHz. Similarly, the liquid crystal M10 has a permittivity of ε_min = 2.50 at a resonant frequency of 121.2 GHz. It has a permittivity ε_max = 3.52 at a resonant frequency of 107.3 GHz, as shown in Figure 8b. The method presented in Figure 8 was used for the other fitting results as shown in Table 1 and Table 2.

Figure 7. Reflection losses of the two LCs in the high-frequency band. (a) C–09–2; (b) M10.

Figure 7. Reflection losses of the two LCs in the high-frequency band. (a) C–09–2; (b) M10.

Figure 8. Fitting of the simulation and test data. (a) C–09–2; (b) M10.

Figure 8. Fitting of the simulation and test data. (a) C–09–2; (b) M10.

From fitting results, we summarized the data of C–09–2 in different frequency bands and at different voltages, including resonant frequencies, permittivity, and refractive indices, in Table 1 and Table 2.

Table 1. The resonant frequencies, permittivities, and refractive indices of C–09–2 at different voltages in the 90–140 GHz frequency band.

Voltage	0 V	1.3 V	1.6 V	1.9 V	2.5 V	5 V	9 V	18 V
Resonance frequency (GHz)	120.1	118.6	115.9	113.45	110.9	108	106.8	104.8
Permittivity	2.570	2.684	2.873	3.040	3.238	3.475	3.562	3.720
Refractive index	1.603	1.638	1.695	1.744	1.799	1.864	1.887	1.929

Table 1. The resonant frequencies, permittivities, and refractive indices of C–09–2 at different voltages in the 90–140 GHz frequency band.

Voltage	0 V	1.3 V	1.6 V	1.9 V	2.5 V	5 V	9 V	18 V
Resonance frequency (GHz)	120.1	118.6	115.9	113.45	110.9	108	106.8	104.8
Permittivity	2.570	2.684	2.873	3.040	3.238	3.475	3.562	3.720
Refractive index	1.603	1.638	1.695	1.744	1.799	1.864	1.887	1.929

Table 2. The resonant frequencies, permittivities, and refractive indices of C–09–2 at different voltages in the 325–500 GHz frequency band.

Voltage	0 V	0.9 V	1.3 V	1.8 V	2.2 V	3 V	4 V	6 V	10 V	18 V
Resonance frequency (GHz)	470.58	468.28	465.55	462.05	460.08	457.56	455.05	452.97	450.78	448.15
Permittivity	2.820	2.912	3.011	3.129	3.168	3.220	3.290	3.367	3.446	3.580
Refractive index	1.679	1.706	1.735	1.769	1.780	1.794	1.814	1.835	1.856	1.892

Table 2. The resonant frequencies, permittivities, and refractive indices of C–09–2 at different voltages in the 325–500 GHz frequency band.

Voltage	0 V	0.9 V	1.3 V	1.8 V	2.2 V	3 V	4 V	6 V	10 V	18 V
Resonance frequency (GHz)	470.58	468.28	465.55	462.05	460.08	457.56	455.05	452.97	450.78	448.15
Permittivity	2.820	2.912	3.011	3.129	3.168	3.220	3.290	3.367	3.446	3.580
Refractive index	1.679	1.706	1.735	1.769	1.780	1.794	1.814	1.835	1.856	1.892

As shown in Figure 9, we linearly fit the resonant frequency and refractive index in Table 1 and Table 2 to obtain the sensitivity of the fabricated sensors. In Figure 9a, in 90–140 GHz band, the relation between the resonant frequency Y and the refractive index X is Y = −47.03X + 195.55. The sensitivity S (S = Δf/Δn, Δf refers to the change in the resonant frequency with respect to the change in the refractive index Δn) of the sensor is 47.03 GHz/RIU, and the linear correlation coefficient is 0.999. The sensor’s average value of Q (Q = Δf₀/FWHM) is 31.93. In Figure 9b, for the 325–500 GHz band, the relation between the resonant frequency Y and the refractive index X is Y = −112.47X + 659.98. Hence, the sensor’s sensitivity in the 325–500 GHz frequency band is 112.47 GHz/RIU, with a linear correlation coefficient of 0.990 and an average Q-value of 46.9.

Figure 3a and Figure 9a show that the sensitivities of the simulated and actual sensors developed in this research are compared with errors of only 1.3% in the range of 90–140 GHz, and Figure 3b and Figure 9b show that the sensitivities of the simulated and actual sensors developed in this research are compared with errors of only 0.9% in the range of 325–500 GHz. It was found experimentally that the sensors in the high-frequency band are about twice the sensitivities of sensors in the low-frequency band, with higher Q-values and larger frequency shift ranges. Improvements in fabrication accuracy and test environment will improve sensor performance.

Figure 9. Linear fitting of the test data (a) in 90–140 GHz frequency band and (b) in 325–500 GHz frequency band.

Figure 9. Linear fitting of the test data (a) in 90–140 GHz frequency band and (b) in 325–500 GHz frequency band.

To show the dielectric properties of the two liquid crystals for the two frequency bands, a summary is shown in

Table 3. Summary of the dielectric properties of the two liquid crystals for different frequency bands.

LC	Frequency (GHz)	Frequency Shift Range (GHz)	εmin	εmax	Δε	Birefringence
C-09-2	120.10–104.80	15.30	2.57	3.72	1.16	0.33
M10	121.20–107.30	13.90	2.50	3.52	1.02	0.30
C-09-2	470.58–448.16	22.43	2.82	3.58	0.76	0.21
M10	471.02–449.14	21.88	2.81	3.51	0.70	0.20

below.

The test results found that, in 90–140 GHz, the Δε of the liquid crystals C–09–2 and M10 are 1.16 and 1.02, respectively, and in 325–500 GHz, the Δε of the C–09–2 and M10 are 0.76 and 0.70, respectively. Moreover, in 90–140 GHz, the birefringence of C–09–2 and M10 are 0.33 and 0.30, respectively, and in 325–500 GHz, the birefringence of the C–09–2 and M10 are 0.21 and 0.20, respectively. Both the liquid crystals C–09–2 and M10 have a high birefringence. The frequency shift range of liquid crystals in the low-frequency band is smaller than in the high-frequency band. The birefringence of liquid crystals in the high-frequency band is about two-thirds of the birefringence of the liquid crystals in the low-frequency band. Both the liquid crystals C–09–2 and M10 have a high birefringence, which means they are well-performing tunable materials and have good application prospects.

4. Summary

In this paper, two highly sensitive THz sensors for liquid permittivity measurement were designed and processed. Using the electromagnetic characteristics of the sensors, different liquid crystals were tested. We obtained the dielectric properties of the two liquid crystals in different frequency bands and at different voltages. It verifies that both sensors have a remarkably high birefringence and are extremely meaningful for the design and fabrication of tunable terahertz devices. The experimental results show that both sensors of this design have excellent sensing characteristics and minor errors, which can be well utilized for the permittivity measurement of the liquids. The fabricated sensor has a 47.03 GHz/RIU sensitivity in the frequency band of 90–140GHz, with a linear correlation coefficient of 0.999 and a Q-value of 31.93. In the 325–500 GHz frequency band, the fabricated sensor has a sensitivity of 112.47 GHz/RIU, a linear correlation coefficient of 0.990, and a Q-value of 46.9. Therefore, by comparing the data of sensor-1 and sensor-2, we also found that the designed sensor in the higher the frequency band has the higher the sensitivity.