1. Introduction

Terahertz (THz) wave lies between microwave band and infrared band in the electromagnetic spectrum, and it shares some of their characteristics. In recent years, the development in THz source and detection technologies has led to a growing demand for THz devices in the fields of communication, global environmental monitoring, non-destructive evaluation, security screening [1–4]. THz frequency selective surface (FSS) is typically a periodic structure with micro structural unit cells, which can be used in a wide range of applications in the fields of radars, absorbers, and radomes [5,6]. Low insertion loss in THz bandpass FSSs indicates that more wanted signal is expected to be transmitted to the desired target. It is known that the loss performances of THz devices are closely associated with the employed dielectric substrates, while it shows an increase of dielectric dissipation factor for these dielectric substrates in THz band, which is considered as one of the main challenges for the development of THz technology. Numerous studies have attempted to apply various dielectric substrates for component designs in THz field. As the dimensions in THz devices are usually small, almost all of the previously reported THz devices have been realized by microfabrication process, where silicon (Si) was widely used as substrate or supporting materials. Therefore, it appears clear that Si was used in many THz designs [7,8]. While in these designs, adhesion layers might be required during metallization process. For instance, the narrowband THz bandpass filters proposed in [8] used a high-resistivity Si with refractive index of 3.42, and a 10-nm-thick titanium (Ti) was applied as the adhesion layer before depositing a 200-nm-thick gold film. Similarly, silicon dioxide (SiO₂), or named quartz, is another attractive and widely used dielectric substrate in THz filed, whose dissipation factor is considered as 0.0004 in several hundred GHz [4,9,10]. Another material that has been used as substrates in THz designs is glass, which has been presented in [11,12], and the described dissipation factor at 440 GHz was 0.05. In [13], a 100 um thick sapphire substrate with the dielectric constant of 9.0 was employed to support the VO₂ thin film for tunable terahertz bandpass filters, where the dielectric loss tangent of sapphire was considered as 0.01 at ∼400 GHz. In [14], a metamaterial design was fabricated on a silicon-on-sapphire wafer, with a thickness of 600 nm and 530 um for the silicon and sapphire layer, respectively. Both of the silicon and sapphire were modeled as lossless dielectrics, with dielectric constant of 11.7 and 10.5 at ∼1 THz, respectively. More recent attention has focused on the development of new materials in THz applications. A significant analysis and discussion on this topic was presented in [15], where SU-8 was exploited and examined due to its low dissipation factor, ability to form thick and high-resolution structures. However, the measured dielectric dissipation factor at 30 GHz and 1000 GHz reaches ∼0.02 and ∼0.05, respectively. In an investigation into a second-order terahertz bandpass filter design, Amir Ebrahimi et al. reported the use of Polydimethylsiloxane (PDMS) as its substrate, whose dielectric dissipation factor at 240 GHz was 0.048, and the dielectric constant in this frequency was 2.35 [16]. In a higher frequency application, PDMS was applied in THz reflectarray and nonuniform metasurface designs, where the dielectric constant and dissipation factor at 1 THz were 2.25 and 0.06, respectively [17]. In a different example, Ludovic et al. proposed a study of textile metamaterials produced by a weaving of metallic and dielectric yarns for filtering operations at around 500 GHz [18]. In this work, polyethylene terephthalate (PET) monofilaments were considered to simulate the dielectric yarns with a dielectric parameter of ${\varepsilon _r} = 3.7({1 - 0.02i} )$. In a more recent work, published in 2019, benzocyclobutene (BCB) was demonstrated as the dielectric substrate of a terahertz FSS, where the dielectric constant and dissipation factor are assumed as 2.48 and 0.01, respectively [19]. In some other work, the dissipation factor of BCB was also considered to be 0.012 (1-220 GHz) and 0.015 (660 GHz), or ∼0.01 (0.3-1.3 THz), which was mainly derived based on transmission line methods, while the value measured through Fourier Transform Spectroscopy technique was ∼0.007 (1 THz) [20,21]. The difference of dielectric parameters may be explained by various material processing technology/condition and error of the applied extraction techniques. We can see that there are not many dielectric substrates that are suitable to use or own low dielectric dissipation factor in THz region. Therefore, it is of great significance to exploit new dielectric substrates with remarkable characteristics in THz band.

In recent years, a new type of substrate polymers, cyclic olefin copolymer (COC), has attracted the attention of researchers because of its advanced dielectric properties in millimeter-wave and THz bands. Two monomers, norbornene and ethylene, constitute its chemical structure, which contributes to the extraordinary characteristics of COC materials [22,23]. High chemical resistance to acids and polar solvents is a significant contributory factor to make COC easily compatible with microfabrication methods and realize flexible thin films [22]. To date, several studies have gradually investigated the use of COC materials in electromagnetic field. One of the early attempts at using COC materials in THz applications was reported in 2009 by Kristian Nielsen et al., who proposed COC based THz fibers with unprecedented low loss and low material dispersion in THz band through a drill-and-draw fabrication technique [24]. A propagation loss less than 0.1 dB/cm at 0.6 THz was demonstrated in this study. In 2011, the first COC based thin-film microstrip lines and grounded coplanar waveguides were fabricated and measured up to 220 GHz [25]. A COC rod was characterized in free space in the frequency band of 220-325 GHz by means of WR-3 frequency extenders coupled to horn antennas, which were driven by a vector network analyzer (VNA). The derived dielectric loss tangent in 220-325 GHz band was 0.0005, and the attenuation was as small as 0.6 dB/mm at ∼220 GHz for the 22um-thick COC film based grounded coplanar waveguide structure. It is worthy mentioning that the transmission lines were fabricated on low-resistivity Si substrates by depositing a 0.2-um-thick Ti and 1-um-thick Au as the conductive layer. In a more recent case, Fabio Pavanello et al. designed, fabricated, and characterized high-pass metal mesh filters operating in THz band based on a type of COC film, with a dielectric constant of 2.34 and dissipation factor of 0.0023 (at 2.5 THz) [26]. In this work, COC film was firstly spin coated on a Si substrate, then cured at 140 $\mathrm{\circ{C}}$ for 2 minutes. For the conductive mesh, chromium (Cr) was deposited as the adhesion layer before evaporating 300-nm-thick gold. One year later, the authors proposed another COC based Yagi-Uda antenna operating at around 300 GHz with a high gain of 15.4 dB [27]. There are other small amount of work related to COC materials, such as microstrip transmission lines, subharmonic mixer [28] [29]. However, the metallization experiment as well as the design and fabrication of low-loss THz FSS based on flexible COC film substrate are rarely involved in the literature.

This study therefore set out to further assess the properties of the newly synthesized and fabricated COC materials in THz band from the device level by characterizing COC based THz devices in experiment. Different from the commonly used spin coating technique for realizing COC film substrates, we propose a new and simple temperature-control film manufacturing method without the use of solvent, and this method can easily achieve COC substrates with the thickness of millimeter scale, where no solvent is involved in the process. Furthermore, we fabricated a THz bandpass FSS operating at ∼550 GHz to verify the substrate properties. To the authors’ knowledge, this is the first COC based THz bandpass FSS working at ∼550 GHz in literature. The paper has been organized in the following way. The first section of this paper gives a brief overview and discussion of dielectric substrates used in THz spectrum and an introduction of COC materials. It will then go to the fabrication and characterization of proposed COC substrates in THz band, including a comparison with that of Rogers RT5880 and quartz. Section 3 is concerned with metallization, design, and fabrication of the proposed COC film based THz bandpass FSS, followed by an experimental and discussion section. Finally, in Section 5 the conclusion is provided.

2. Fabrication and characterization of the proposed COC substrate

COC dielectric films are often obtained by spin coating technique, where COC pellets have to be dissolved in an appropriate solvent [25,26,28]. Other preparation methods proposed in recent years, such as Vacuum Solvent Desorption (VSD) technique and Thermic Solvent Desorption (TSD) technique, are also based on COC solution for the film fabrication [22]. Due to the limitation of micromachining technology, the thickness of COC dielectric films prepared by these solution methods is mainly between hundreds of nanometers and tens of micrometers. Here, we propose a new and simple temperature-control film manufacturing method without the use of solvent and solution, and a schematic of COC film casting process is presented in Fig. 1(a). It can be seen that COC pellets need to be pre-dried and pre-heated using hot air to remove the dissolved oxygen. In step 3, the pre-heated COC pellets are added into a film casting machine through a hopper, and they are transported to a drying chamber for further drying with the help of pump and conveyor belt. The pellets are then sent to a high-temperature chamber for melting, where fluid-state COC slurry forms. The COC slurry ejected from a nozzle is cooled down by cooling rollers to form COC thin films, which are then manually pulled out. The temperature of drying and high-temperature chambers is around 120 $\mathrm{\circ{C}}$ and 300 $\mathrm{\circ{C}}$, respectively, and the colling roller temperature is about 140 $\mathrm{\circ{C}}$. The temperature of each part and COC slurry speed can be regulated and optimized for millimeter thick films with good quality.

 

Fig. 1. (a) Schematic of COC film casting process. (b) Photograph of the used THz TDS measurement setup in glass sealed chamber. (c) The time-domain results of the reference (air) and sample signals. The extracted (d) refractive index $n(\omega )$ andextinction coefficient $\kappa (\omega )$, and (e) dielectric constant ${\varepsilon _r}(\omega )$ and dielectric loss tangent $\tan \delta $. (f) A comparison of $\tan \delta $ between the proposed COC material at 550 GHz and other commonly used THz substrates, and data for selected substrates are from Quartz [4,9,10], Glass [11,12], Sapphire [13], SU-8 [15]^, PDMS [16,17], PET [18], BCB [19–21] .

Download Full Size | PDF

Dielectric properties of substrate materials in frequency band of interest are very important in component designs. As the frequency spectrum we are interested in is around 550 GHz, a typical THz time-domain spectroscopy (TDS) was used to characterize the proposed COC substrate material for its THz performance [30]. Photograph of the used THz TDS measurement setup in glass sealed chamber is shown in Fig. 1(b). Rogers RT5880 and quartz are among the most commonly used materials with low dielectric loss tangent in millimeter-wave and terahertz frequency range. For example, the RT5880 datasheet shows that its dielectric constant and dielectric loss tangent at 10 GHz are 2.2 and 0.0009, respectively, and the RT5880 is considered as one of the smallest among microwave and millimeter-wave substrates. Quartz is also a substrate material with extremely low dielectric loss tangent at high frequencies, with a loss tangent of around 0.0004 at several hundred GHz [4,9,10]. Therefore, in order to explore the low-loss characteristics and potential applications of this novel COC material, we chose to compare it with quartz and RT5880 as references and conducted measurements using the same THz-TDS testing system. Refractive index $n(\omega )$ and extinction coefficient $\kappa (\omega )$ (or absorption coefficient ${\; }\alpha (\omega )$) are two typical parameters for characterizing material properties in the frequency band of THz and visible light. While in microwave/ millimeter-wave/THz spectrum, we tend to use dielectric constant ${\varepsilon _r}(\omega )$ and dielectric loss tangent (or dissipation factor) $\tan \delta $ to describe dielectric substrate properties. According to certain algorithm, THz TDS system could extract $n(\omega )$ and $\kappa (\omega )$ of the tested dielectric samples by using a reference signal, then it can be further converted to ${\varepsilon _r}(\omega )$ and $\tan \delta $ by exploring the relations between $n(\omega )$/$\kappa (\omega )$ and ${\varepsilon _r}(\omega )$/$\tan \delta $ described in following equations [31]. The measured time-domain results of the reference (air) and sample signals are shown in Fig. 1(c). It is worth mentioning that due to the use of TM polarization when measuring Rogers RT5880 and quartz substrates, and the use of TE polarization for COC measurement, their time-domain signal amplitudes are different. This is due to the fact that the emission and detection efficiencies of photoconductive THz transmitters and receivers are sensitive to the polarization of the optical grating pulses, and the signal amplitudes vary with the rotation of the excitation pulse polarization [32]. Nevertheless, as long as the polarizations used for the measured samples and referenced air are consistent, the samples’ dielectric parameters can be accurately extracted. The extracted parameters of the dielectric COC, quartz, and RT5880 samples are illustrated in Fig. 1(d)-(e), including the refractive index $n(\omega )$, extinction coefficient $\kappa (\omega )$, dielectric constant ${\varepsilon _r}(\omega )$, and dielectric loss tangent $\tan \delta $.

It can be seen that the extracted $n(\omega )$ and $\kappa (\omega )$ of the COC sample own an average value of 1.5 and 0.002 in the frequency range of 0.2-2.2 THz (Fig. 1(d)), respectively, due to its high transmittance of around 90%. According to the described conversion formulas, the dielectric constant ${\varepsilon _r}(\omega )$ and dielectric loss tangent $\tan \delta $ could also be calculated, as shown in Fig. 1(e). The obtained results indicate that the dielectric constant is very stable in a whole frequency band of 0.2-2.2 THz, with a value between 2.31 and 2.33. The relatively small dielectric constant can lead to a low impedance mismatch with free space for COC based designs. The dissipation factor $\tan \delta $ of the COC sample tends to slightly increase, ranging from 0.0001 to 0.003 in 0.2-2.2 THz band. What is more, the dielectric constant and dissipation factor of the proposed COC sample at the working frequency (550 GHz) are 2.31 and 0.0002, respectively.

If we now turn to the comparison of these three dielectrics, what can be clearly seen in Fig. 1(e) is the gradually increase of dielectric loss tangent with the increase of frequency for all the three samples, while among which the proposed COC material has the most gentle rise in the THz band. Overall, the average dielectric loss tangent of the COC material is much lower than that of RT5880 and quartz. To be specific, at 550 GHz the measured dielectric loss tangent of RT5880 and quartz are ∼0.0018 and ∼0.0012, respectively. This indicates that at 550 GHz the dielectric loss tangent of the COC material proposed in this work is 8 times smaller than that of Rogers RT5880, and 5 times smaller than that of quartz. For the dielectric constant, it is also noticeable from Fig. 1(e) that the variation for the samples of RT5880 and quartz is larger than that of COC material, which means the COC material has a lower frequency dispersion in THz band. The possible extraction errors and fluctuations in the tested results are mainly caused by the uneven thickness of dielectric samples and the measurement environment. Moreover, the multiple reflections caused by these thin samples were included when performing Fast Fourier Transform to obtain the frequency domain spectra. Due to the relatively lower signal-to-noise ratio (SNR) on both sides of the frequency band, the extracted results in low and high frequencies might be less reliable. A comparison of dissipation factors between the proposed COC material at 550 GHz and the commonly used THz substrates described in Section 1 is shown in the Fig. 1(f), which indicates the superiority of our fabricated COC material in realizing extremely low-loss THz devices. The smaller dielectric loss tangent of COC over other materials can be attributed to its unique structure, which is composed of polyethylene-norbornene chains, and the compound is incorporated randomly on the carbon main chain, making the COC structure totally amorphous. Due to the amorphous structure of the nonpolar COC, the proposed COC material possesses a lower dielectric loss tangent when compared with other materials.

3. Metallization, design, and fabrication of COC based THz FSS

The next step is to realize the metallization on this new type of COC substrate, before fabricating and experimentally validating its excellent properties. Metallization and device design were carried out on a first proof-of-concept COC film with a sufficiently smooth surface. To prevent circuits from being oxidized, in this work we used gold (Au) as the conductive layer. For surface metallization of various substrates, transition conductors such as titanium (Ti), Platinum (Pt), or chromium (Cr) have been widely utilized as adhesion layers [9,16,17,25]. As presented in [17], Ti was employed as an adhesion layer before depositing Pt film on a supporting wafer, where Pt film operates as the ground plane in order to ensure better selectivity during wet chemical etching process when comparing Pt with other metals with higher conductivity such as Au/copper. In this work, Cr was also used as an adhesion layer for a PDMS substrate, followed by a deposition of Au onto the PDMS layer via electron beam evaporation. As metallic layer thickness is required to be larger than the corresponding THz skin depth, which is related to the working frequency (described in the next section), a variety of schemes were investigated to realize metallic films with different thickness onto the proposed COC substrate. The experiment results indicate that if deposited Au thickness is larger than 200 nm, it is suggested to have a coating of Ti or Cr first as the adhesion layer to achieve a robust and reliable metallic Au film. Additionally, it is not easy to realize double-sided gold structure on the COC substrate. This is due to the fact that double-sided operation involves several times of acetone immersion and baking, which leads to severe bending for the COC sample. Uneven and curved surface makes the reworking process difficult, which aims to remove excessive positive photoresist (PPR) and peel off the COC substrate from supporting structures. Therefore, in the reworking step, the patterned metallic layer is likely to be damaged and destroyed. When the required Au thickness is equal or less than 200 nm, the metallization process can be carried out directly on the COC sample by magnetron sputtering without the need of adhesion layers. Moreover, it is feasible to realize double-sided metallic patterns on the 0.1 mm thick COC substrate due to the thinner Au layer.

According to the COC metallization process described in previous section, a double-sided FSS with its Au thickness of 200 nm on a 0.1 mm thick COC substrate is proposed. The design starts by considering the relationship between skin depth and the corresponding working frequency, showing in the following formula [33,34]:

Now, we will demonstrate the design and fabrication of a COC based THz bandpass FSS to verify the potential of the proposed COC material in THz field. Figure 2(a) depicts a three-dimensional structure of the proposed THz bandpass FSS with 16 unit cells, composing of two periodic metallic arrays separated from each other by a dielectric COC substrate with a subwavelength thickness. Figure 2(a) also includes the upper and lower metallic layers of the unit cell. The upper metallic layer is made of a wire loop surrounding a square patch. On the basis of the upper layer structure, a cross shaped resonator cavity is excavated in its metallic square patch to form the lower metallic layer. The FSS unit cell with four periodic boundary conditions on four sidewalls and two Floquet ports on the top and bottom is constructed to simulate and optimize in a commercial finite element method solver-High Frequency Structure Simulator (HFSS) developed by ANSYS. The parameters of the optimized FSS design centered at 550 GHz are illustrated, and the unit cell size ($a = 0.12\; mm$) is equal to $0.22{\; }{\lambda _0}$, where ${\lambda _0}$ is the wavelength in free space at the working frequency. Other dimensions are listed as follows: $g = 0.008\; mm$, $w = 0.005\; mm$, $b1 = 0.02\; mm$, $b2 = 0.07\; mm$, and the substrate thickness $h = 0.1\; mm$. The operation principle and lumped element equivalent circuit can be analyzed through [16]. The unit cell is repeated to produce a periodic array covering $1.8 \times 1.8\; \textrm{c}{\textrm{m}^2}$ area. The red and grey areas in the figure represent the deposited Au film and COC substrate, respectively.

 

Fig. 2. (a) Three-dimensional structure of the proposed COC THz FSS design, and the upper/lower layer of the unit cell. (b) The fabrication process of the COC based THz FSS design. (c) The fabricated COC THz FSS sample. The micrographs of the (d) upper, and (e) lower metallic layers in optical microscope, and the size unit in inset is microns (um).

Download Full Size | PDF

Once the COC based THz FSS design was optimized and finalized, we fabricated the device by using microfabrication techniques with the steps illustrated in Fig. 2(b). These steps can be summarized as follows:

Figure 2(c)–2(e) show the photographs of the fabricated COC based THz FSS sample, the upper and lower metallic layers in optical microscope, respectively. The sample in Fig. 2(c) presents four identical FSS designs, each of which is $1.8\; \textrm{cm}\; \times 1.8\; \textrm{cm}$ in size. It is noticeable that the proposed COC device owns very good flexibility. As the size of each unit cell is $0.12\; \textrm{mm}\; \times 0.12\; \textrm{mm}$, every FSS is designed with 150 units in horizontal and vertical directions.

4. Experimental results and discussion

As the main purpose of the measurement is to verify the low-loss characterization and potential application of the proposed COC materials in THz field, here we focus on the transmission response. The used THz TDS setup to characterize the transmission spectrum of the fabricated THz FSS design based on the proposed COC substrate is shown in Fig. 1(b). The broadband THz beam generated from a photoconductive emitter is collimated and focused by lens 1 and lens 2, and then passes through the hole on the shelf for loading FSS samples. The THz signal is then sent to a photoconductive detector via lens 3 and lens 4. Through time-domain measurement and then Fourier transformation of the temporal data, transmission spectrum of the THz FSS can be acquired. Firstly, a reference signal was required, and here we used air as the reference (measured without any samples). So a THz time-domain data $\textrm{R}(\textrm{t} )$ can be obtained through the measurement. Secondly, the fabricated FSS sample was placed on the shelf, where THz pulses from the emitter were focused on the FSS sample, and then to the photoconductive detector to acquire the THz time-domain data $\textrm{S}(\textrm{t} )$. Finally, by Fast Fourier Transformation of the THz time-domain data R(t) and S(t), the extracted frequency dependent complex transmission spectrum can be obtained according to the following formula, where $\textrm{S}(\mathrm{\omega } )$ and $\textrm{R}(\mathrm{\omega } )$ are the frequency-domain representations of $\textrm{S}(\textrm{t} )$ and $\textrm{R}(\textrm{t} )$, respectively.

During the measurement process, it should ensure that the humidity of the test environment is low and stable, which can be maintained by continuously filling nitrogen into a glass sealed chamber with the THz TDS setup, and the humidity during our measurement was kept below 5%. TE and TM polarizations can be manually switched by rotating the emitter and detector around the propagation axis. In order to evaluate the transmission response of the proposed THz FSS at oblique angles of incidence in TE and TM polarizations, we adjust the positions of the emitter and detector through an assembled control system, while the shelf loading FSS samples remains stationary. The measurement results indicate that THz signal intensities from the photoconductive emitter in TE and TM polarizations are not the same, which can be seen in Fig. 3(a). Figure 3(a) shows the measured time-domain results in TE and TM polarizations with/without the FSS sample under normal incidence. From this figure, we can see that the THz signal amplitudes in TM polarization are always larger than that in TE polarization. Moreover, we found that the measured THz signals passing at various incident angles are different as well. Therefore, when calculating the transmission response of the proposed FSS at different incidence angles and polarizations, it is essential to measure the received air signal, which is served as the reference $\textrm{R}(\textrm{t} )$, at various angles when the photoconductive antennas are in TE and TM polarizations, respectively.

 

Fig. 3. (a) Measured THz time-domain signal in TE and TM polarizations with/without FSS sample under normal incidence. (b) Comparison between the measured and simulated transmission spectrum at normal incidence, with gray representing the -3 dB bandwidth area.

Download Full Size | PDF

As shown in Fig. 3(b), the measured transmission spectrum of the fabricated FSS sample at normal incidence is plotted together with the numerical result, where a dielectric dissipation factor of 0.0002 was applied. Figure 3(b) reveals that the measured transmission spectrum is in very good agreement with the numerical result, especially the passband center frequency and the passband transmittance. The comparison between the experimental and simulation results is summarized in Table 1. The measured center frequencies ${f_0}$ are 559 GHz and 541 GHz for TE and TM polarizations, respectively, with a negligible 9 GHz (1.6%) difference when comparing with the targeted value (550 GHz) in simulation. As seen, the measured passband transmittances at their center frequencies in TE and TM polarizations are considerably consistent (-1.36 dB vs -1.43 dB), which are in agreement with the simulated -1.57 dB. It is also apparent from the table that the passband peak transmission in numerical analysis is ∼-1.05 dB, and this value is located between the measured transmission peaks in TE and TM polarizations, which are -1.22 dB and -0.68 dB, respectively. In addition, very close 3 dB lower frequency can be observed for the simulated and measured results in TE and TM polarizations (420 GHz vs 400 GHz vs 411 GHz). By contrast, the difference in their 3 dB upper frequency is relatively larger. The measured transmission spectrum shows slight discrepancies when compared to the simulation results. Specifically, the measured transmission values at the center frequency are slightly better than the simulation results, and the simulated transmission peaks in the passband are located between the measured TE and TM values, although these values are very close. These discrepancies are attributed to errors in modelling, fabrication, and measurement. During the measurement process, the FSS sample needs to be manually taken out and placed into a glass-sealed chamber with nitrogen to maintain humidity levels below 5%. Additionally, the polarization of the emitter and detector needs to be manually rotated, and oblique angle measurements under different polarizations are also required. This long and complicated process can result in certain deviations and fluctuations of the laser output power and the measured reference signal, leading to measurement errors. Moreover, the dimensional error of the substrate film thickness and metallic structure, as well as the surface roughness would also contribute to deviations in the center frequency and corresponding transmittance. These factors can introduce errors in the measurement. In our study, we made efforts to minimize these errors by carefully preparing the samples and ensuring their uniformity. However, these discrepancies are within an acceptable level.

The performance of the proposed FSS design based on COC substrate at oblique angles of incidence is also studied. The simulated and measured transmission responses of the FSS design at different angles of incidence under TM polarization are presented in Fig. 4(a)–4(b), which illustrate that the measurement and simulation are in good consistency, and the proposed FSS design in working frequency band exhibits a stable frequency response in TM polarization under various incidence angles. Figure 4(c)–4(d) provide the transmission spectrum of the device at various incidence angles under TE polarization. Both of the simulation and measurement results indicate that the passband insertion loss increases with an increase of incidence angles, while it shows a decrease for its bandwidth. What stands out in these figures is that the response stability of TE polarization at various incidence angles is worse than that of TM polarization. Similar to the results under normal incidence, there are certain deviation between the measurement and simulation results at oblique angles of incidence, though the deviation is small and within acceptable range. As discussed, the discrepancies in the measured and simulated transmission spectrum can be attributed to inevitable errors in modelling, fabrication, and measurement, including the complex and time-consuming measurement process and dimensional errors in the samples. As a whole, the measured results are still within an acceptable range, and they are inconsistent with the simulation, thus verifying the FSS design based on the proposed COC substrate.

 

Fig. 4. Transmission spectrum of the proposed FSS design at oblique angles of incidence, with gray representing the -3 dB bandwidth area. (a)-(b) Simulation and measurement of the sample in TM polarization. (c)-(d) Simulation and measurement of the sample in TE polarization.

Download Full Size | PDF

Table 1. Comparison between measured and simulated transmission spectrum of the FSS at normal incidence^a

View Table | View all tables in this article

In order to further investigate the performance of our proposed FSS under flexible conditions, we bend FSS samples on cylindrical objects with various radii, exploring their transmission response and compare it with the unbent condition. Due to the required huge computing resources and lengthy computing time for electrically large and flexible FSS structures, a FSS sample consisting of 225 unit cells (15 × 15) is analyzed, after ensuring that its transmission response at this scale is close to the result by using planar Floquet port modeling of the FSS unit cell. Figure 5 shows the device transmission response under bending radii of 1 mm, 2 mm, 4 mm, and 10 mm, corresponding to bending angles of approximately 103°, 52°, 26°, and 10°, respectively. The results demonstrate that the flexible FSS transmission response under flexible conditions follows a similar trend to the planar FSS measurement results conducted at various incidence angles and polarizations showing in Fig. 4. However, under flexible conditions, there is more pronounced transmission amplitude variations and bandwidth changes within the passband. Nonetheless, the transmission in the passband remains consistently high, indicating low insertion loss even under flexible conditions. This implies that our proposed flexible FSS samples have the potential to operate effectively around cylindrical surfaces and curved objects, including bendable radomes and antennas, conformal sensors, wearable electronics, as well as flexible communication systems, and offer enhanced functionality, detection capabilities and adaptability in these potential applications.

 

Fig. 5. Transmission spectrum of the flexible FSS design wrapped on a cylinder under various bending radii. (a) TM polarization. (b) TE polarization.

Download Full Size | PDF

Table 2 demonstrates a comparison between the proposed COC based THz FSS design and other THz bandpass FSS filters reported in literature. Quartz was used as a low-loss substrate for the THz filter designs described in [9] and [4], where its dielectric constant and dissipation factor were 4.4 and 0.0004, respectively. However, the fabricated design in [9] showed a transmittance of around -3 dB at the center frequency of 255 GHz, and also the metallic layer was composed of multiple Ti/Pt/Au films by the process of electron beam deposition. In [4], the insertion losses in the first passband (centered at 315 GHz) and the second passband (centered at 480 GHz) of the THz filter were -1.9 and -6 dB, respectively. However, the reference spectrum in measurement was based on the quartz substrate without metallic patterns, so if using our measurement approach (Air as reference), the passband insertion loss in [4] was expected to be larger. For the metallic layer, 200 nm thick aluminum (Al) was deposited, while a SiO₂ film produced by plasma-enhanced chemical vapor deposition is required during the metallization process. This work also suggested that Ti/Pt/Au films could be used as its metallic layer to further reduce the device insertion loss. In [35], a Rogers RT5880 substrate with a low dielectric constant was used to produce a THz FSS filter, and the measured insertion loss at the center frequency of 250 GHz was approximately 3 dB by the use of microwave measurement system. The dielectrics separating two metallic layers for the THz bandpass filters described in [16] and [11] were Polydimethylsiloxane (PDMS) and air, respectively. Additionally, PDMS and borosilicate glass were used as the top and the bottom layers in these two studies with five-layer structures. Though the operating frequencies in these THz bandpass filters are smaller, their insertion losses are larger than that of the proposed COC based THz FSS (measured peak value was -1.22 dB at TE polarization). The experimental results of the raw COC sample and the fabricated COC based FSS design verify that the proposed COC material possesses a very small dissipation factor in THz band, which is much lower than other commonly used dielectrics, showing a very huge application prospect in THz field.

Table 2. Comparison between the proposed and related THz bandpass FSS filters^a

View Table | View all tables in this article

5. Conclusion

In this paper, we investigated a new type of dielectric film substrate (COC) with extraordinary properties that can be used in THz field, and a simple solvent-free synthesis of temperature-control method for casting COC films with the thickness of millimeter scale was proposed. The raw COC substrate sample without metallic patterns was characterized in THz band, and the measured dielectric properties were then compared with that of Rogers RT5880 and quartz. It shows that the COC substrate possesses a lower frequency dispersion with a stable dielectric constant (∼2.3) in an extremely wide THz band. The relatively small dielectric constant is beneficial to achieve a low impedance mismatch with free space for COC based designs. The dielectric dissipation factor is in the order of 0.0001 in a broad THz range, and the value at 550 GHz reaches 0.0002, which is 8 times smaller than that of Rogers RT5880 and 5 times smaller than that of quartz. The surface metallization on COC films was also studied. The experimental results indicate that for thin COC films (i.e. 0.1 mm), direct metallization without adhesion layers is feasible, and the Au thickness of 200 nm is possible. Moreover, the detailed fabrication process was presented, and double-sided metallic patterns can be realized. In order to further verify the remarkable characteristics of the proposed COC substrate, a COC based double-sided THz bandpass FSS has been successfully designed, fabricated and measured. The measured transmission spectrum was in very agreement with the numerical result in terms of the center frequency, transmittance level, and 3 dB bandwidth. The comparison between the proposed COC THz bandpass FSS and other related filters in literature indicates that the proposed COC substrate with extraordinary properties (small dielectric constant, low frequency dispersion, low dissipation factor, good flexibility, etc.) through solvent-free fabrication technique has a great application prospect in THz field.