Performance Evaluation of Microfluidically Tunable Microwave Filters

Abstract

This paper presents a contribution to evaluating the performances of tunable devices devoted to RF applications. It is based on reconfiguration by fluids of a capacitor/inductor associated in a monolithic substrate. Indeed, the association of two microfluidic passive devices on the same wafer allows us to increase the total frequency response of microwaves structures. The study evokes the presence and displacement of different conductive and dielectric liquids in the structure microchannels. The theoretical analysis concerns the association of microfluidic devices, a capacitor and inductor, in parallel topology. The obtained results show a good agreement between electrical parameters and the microwave response. Furthermore, a significant frequency variation from 370 MHz to 1720 MHz is achieved, with a tuning range that reaches 364.8%. The experimental part exhibits the fabrication and characterization of two structures in order to evaluate the response of microfluidic actuation for two architectures: a pass-band filter (presented in prior work) and a stop-band filter. The obtained results are in good agreement with the modeled behavior and demonstrate a large tuning range for the stop-band filter.

1. Introduction

Tunable filters are crucial features of modern millimeter and microwave systems. These filters are classified into three main categories according to the used configuration: volume filters, planar filters (microstrip, coplanar, multilayer, etc.) and other innovative technologies (Substrate Integrated Waveguide (SIW) [1,2], Surface Acoustic Wave (SAW) [3,4], Bulk Acoustic Wave (BAW) [5,6]).

Volume filters use either waveguides of different shapes, often rectangular or cylindrical, resonant cavities or dielectric resonators. They provide a narrow bandwidth with high-quality factors and good power handling. On the other hand, this technology has several drawbacks, particularly in terms of size and weight. They are more difficult to reconfigure and often require adjustments after manufacturing, which increases manufacturing costs.

Given the constraints of volume technology mentioned previously, filter designers have moved towards planar technologies. Planar filters have the advantages of being compatible with MMIC technology, compact, easy to integrate into systems, easily reconfigurable and simple to produce at low costs. Their disadvantage is the low-quality factor, which strongly contributes to a reduction in selectivity and an increase in insertion losses. In particular, the SIW technique combines volume and planar technologies to design compact filters that offer quality factors of a few hundred at 10 GHz. It consists of generating waveguides in a dielectric substrate in which a wave can propagate. Other innovative technologies require a good mastery of the manufacturing processes (Film Bulk Acoustic Resonator (FBAR) [7,8]) or the integration of new materials (superconductors [9,10] and piezoelectric [11,12]), nevertheless, this greatly increases the cost of manufacturing.

On the other hand, low-cost components such as varactors [13,14] and PIN [15,16] diodes are used in tunable microwave structures. One or more varactors can be integrated to modify the central frequency of the resonators and PIN diodes are frequently used to provide the switching function (ON/OFF states). Using this functionality, several works propose designing discrete variable capacities. However, these variable capacitance diode-based filters have a low-quality coefficient and therefore high insertion losses. To solve this problem, variable capacitors can be realized using MEMS technology. RF MEMS are the most popular tunable elements for the design of variable devices due to their performance. Different MEMS-based bandpass filter architectures (ohmic, capacitive or varactor) are presented in the literature. These filters can then operate at very high frequencies and have much lower insertion losses due to the high coefficient of quality of the RF MEMS components.

Each of these technologies has its own specificities, so the choice of using one or the other depends on the actual need of the system in which the filters will be integrated. For non-demanding specifications in terms of power handling and insertion losses, planar technology offers a very good cost/performance compromise. It is in this scenario that we have positioned ourselves throughout this work in order to evaluate a new tuning feature for variable microwave filters. Indeed, in this work, we associate two microfluidic tunable devices in the same wafer: an inductor [17,18,19] and a capacitor [20,21]. The advantage of these tunable devices lies in their ability to operate at different frequency bands using the simplest possible structure. Our study allows us to catch the resonant phenomena in response to frequency variation and thus understand the behavior of the filter. Furthermore, we analyzed the liquid displacement effect of both conductive and dielectric liquids on the frequency response of the microwave structure.

2. Fluidic Variable Parallel RF MEMS Structure

In a prior study [17,18,19,20,21], we proposed two RF MEMS tunable devices that allow for the incorporation of a microfluidic channel with double-turn electrodes. It is thus necessary to be able to change either their central frequency or their passband or both at the same time. The modification of their characteristics makes it possible to select the useful signals and to eliminate those that are undesirable and parasitic, providing a new microfluidic tunable microwave filter. The three-dimensional view of the proposed filter structure is presented in Figure 1:

3. Analysis of Fields Distributions

3.1. Structure without Fluid

During this part, we study the frequency response of the structure that associates the capacitor and inductor in parallel using FEM software “HFSS”. We started by evaluating the distribution of magnetic field lines when the structure is empty. Figure 2 illustrates the obtained results in response to frequency:

For low-excitation frequencies and up to 500 MHz, the distribution of the magnetic field is maximum on the inductor side and is almost zero on the capacitor side. As the excitation frequency increases, we notice a gradual decrease in the magnetic field in the inductor as shown in Figure 2b. From 1500 MHz, the magnetic field becomes minimal in the inductor and begins to increase in the capacitor branch (Figure 2c). The magnetic field continues to increase in the capacitor and reaches its maximum at 2000 MHz (Figure 2d). With increasing excitation frequencies, the magnetic field starts to decrease in the capacitor as shown in Figure 2e. We can then conclude that, when the inductor is associated in parallel with the capacitor, magnetic field lines alternate between the two branches. Subsequently, we studied the distribution of the electric current as shown in Figure 3:

From Figure 3a,b, we notice that the current density is higher in the inductor branch up to 500 MHz and that it continues to increase throughout the structure and more visibly in the inductor, where it reaches its maximum at 1000 MHz. However, for the same frequency range (500 MHz–1000 MHz), the magnetic field decreases in the inductor and becomes almost zero at 1000 MHz. We can then conclude that, in this frequency range, the magnetic field is not proportional to the current density in the inductor. Then, with the increase in the excitation frequency, we observe a progressive decrease in the current density in the inductor followed by a significant increase in the current density at the capacitor branch (Figure 3c–e). These results show that the transition of the magnetic field to the capacitor branch is followed by the transition of the current density. Figure 4 shows the obtained results of dB (S11) and dB (S21) gains:

This figure reveals the existence of two frequencies for which the gains are extreme; they express the presence of two distinct operating modes. The first mode is obtained at 1400 MHz when the structure behaves like a notch filter. After this frequency, there is a change in the shape of the gains dB (S11) and dB (S21), giving rise to another operation mode. Indeed, the value of the gain dB (S11) reaches its minimum at 2200 MHz, where the gain dB (S21) becomes maximum and the structure then behaves like a resonator.

3.2. Metal Conductor Liquid Effects

This part investigates the response of the parallel RF MEMS structure when the inductor is totally filled with Galinstan. We started by analyzing the distribution of magnetic field lines in response to frequency.

As shown in Figure 5a–c, at 500 MHz, the magnetic field is higher on the inductor side and decreases with the frequency up to 1500 MHz. The magnetic field rises in response to frequency and transits to the capacitor branch. The value becomes maximum on the capacitor side for an excitation frequency of 2200 MHz (Figure 5d). From this frequency, the magnetic field decreases in the capacitance branch as shown in Figure 5e. Next, Figure 6 illustrates the electric current density distribution for frequencies between 100 MHz and 2500 MHz:

As we can see in Figure 6a, the electric current distribution is higher in the inductor branch at the frequency range between 1 MHz and 500 MHz but is minimal in the area filled with Galinstan. This result explains the magnetic field cancellation in the inductor core at 500 MHz as shown in Figure 5a. When the excitation frequency increases, the current density continues to increase and attains its extreme distribution at 1500 MHz (Figure 6c). For the same value of the excitation (1500 MHz), the distribution of the magnetic field reaches its minimum value (Figure 5c). According to Figure 6d,e, the current density remains maximum next to the capacitor and gradually decreases in the inductor branch. These results show that the transition of the magnetic field distribution on the capacitor branch is in agreement with the electric current density for the excitation frequency between 2000 MHz and 2500 MHz. In order to study the effect of the conductive liquid insertion between the inductor turns, we plotted the gain curves in Figure 7:

The gain curves show the same profile as that attained when the structure is empty. However, the cutoff frequency increased and reached 1720 MHz, with a small tuning range of Tr = 13.15%. These results indicate that the insertion of Galinstan makes it possible to vary the cutoff frequency without changing the RF behavior of the parallel structure. In addition, the resonant frequency is obtained at around 2200 MHz, a value similar to the first case without liquid.

3.3. Dielectric Liquid Effects

We would like to mention that the magnetic field transition between the inductor and capacitor is obtained for low values of excitations frequencies. To better assess the distribution of magnetic field lines in the structure, we changed the sampling frequencies as follows: {100 MHz; 300 MHz; 400 MHz; 500 MHz; 1000 MHz; 2500 MHz}. The results obtained are illustrated in Figure 8:

Figure 8a,b show a progressive growth of the magnetic field distribution on the inductor branch for an excitation frequency covering 100 MHz and 300 MHz. When the excitation frequency rises, the magnetic field lines pass through the capacitor and intensify there up to 500 MHz (Figure 8c,d). From 1000 MHz (Figure 8e), the magnetic field becomes weak and is evenly distributed throughout the structure up to 2500 MHz (Figure 8f). Moreover, we studied the electric current density at the same value of frequencies {100 MHz; 300 MHz; 400 MHz; 500 MHz; 1000 MHz; 2500 MHz}.

According to the obtained results at low frequencies and up to 300 MHz (Figure 9a,b), the current density is maximum in the inductor and expands gradually in the capacitor. When the excitation frequency increases and reaches 500 MHz, the current density remains high in the capacitor and decreases in the inductor. These results show that, between 100 MHz and 500 MHz, the magnetic field lines distribution is in agreement with the current density repartition. From 1000 MHz, we notice a difference between the magnetic field spreading, which is weak, and the current density distribution, which is still visibly high. This study leads us to visualize the field and the current, which are linked to the incident and outgoing energy of the structure. Therefore, we have plotted the gains’ curves in reflection and transmission in Figure 10:

Referring to the curve of dB (S21), a magnetic field transition between two branches is obtained at 370 MHz. This transition spreads over a narrow range of frequencies, indicating an abrupt change. We also notice that the resonance value decreased by 1000 MHz in reference to the empty case. The obtained tuning range reaches Tr = 310.8% and is greater than that obtained with the presence of conductive liquid in the inductor.

4. Fluidic Variation

4.1. Conductive Liquid Displacement Effects

The circulation of Galinstan between the inductor turns changes the resonant frequency of the parallel structure. According to the study carried out previously, when the inductor’s microchannels are fully filled with the conductive liquid, the resonant frequency of the structure reaches its maximum value. To visualize the effect of the amount of conductive liquid in the channel on resonance frequencies, we investigated six cases as shown in Figure 11. Furthermore, the variation in the gains for each Galinstan position is presented in Figure 12.

According to Figure 12a, the resonant frequency increases with Galinstan penetration in microchannel turns. At the first Galinstan position, the resonant frequency is 1360 MHz, which rises to 1690 MHz when the inductor is fully filled. The total variation ensured by the metal liquid displacement is considered to be low and allows for a tuning range of Tr = 24.2%. Furthermore, the transmission gain shape (Figure 12b) reveals that dB (S21) depends on the Galinstan position in the microchannel. In fact, the gain begins at −22.67 dB and increases to −16 dB at the last Galinstan position, and this result shows a degradation in the device performance. Subsequently, we studied the effect of salt water on frequency variation in the parallel structure, and Figure 13 below illustrates the gain curves:

We notice that salt water allows us to obtain a similar behavior response instead of Galinstan. Indeed, at the first position of salt water, the resonance frequency is obtained at 1320 MHz, which rises to 1620 MHz when the inductor is completely filled (Figure 13a). We deduce that the total variation provided by the displacement of salt water allows for a tuning range of Tr = 22.7%. Nevertheless, we observe a strong degradation of the dB gain (S21) depending on the position of the salt water (Figure 13b). We can conclude that liquid conductivity is the main parameter in the performance of a notch filter. As was known, the stability of the filter’s gain value is a paramount parameter when designing RF filters.

4.2. Dielectric Liquid Effects

The displacement of fluid in the capacitor’s microchannels influences the medium permittivity and then changes the frequency behavior. Indeed, we analyzed the response of the parallel RF MEMS structure for different cases of distilled water position as presented in Figure 14, and the obtained gains’ curves are illustrated in Figure 15:

Cut-off frequencies are between 370 MHz and 1060 MHz and allow for an important tuning range that reaches Tr = 194.5% as shown in Figure 15a. However, Figure 15b illustrates that dielectric liquid displacement deteriorates the gain dB (S21). Indeed, as we can see, the gain is −18.95 dB at the first position and decreases to −8.02 dB when the capacitor is fully filled with distilled water, making a large difference of −10 dB.

5. Experimental Characterization

5.1. Essential Fabrication Steps

The fabrication process began with the deposition of a Ti layer necessary for the attachment of a 7 µm gold layer on the BR33 substrate. Then, a film of positive resin AZ4562 was deposited and structured by a photolithography process. After ensuring the homogeneity of the metal structure, the AZ4562 molding layer was removed by chemical attack. The Ti/Cu layer was released by wet etching. The microfluidic structure was covered by a dry SU-8 photosensitive film with a 20 µm thickness. The film was laminated at a 2-bar pressure and a temperature of 65 °C with a speed v = 1 m/min. The structure is now finalized and the access holes can be connected to another level of the microchannels. The process of stacking the different layers of cover was carried out by the laminator.

Figure 16 presents two fabricated microfluidically tunable microwave filters. The first associates the inductor and the capacitor in serial topology and behave as a pass-band filter studied in previous work [22]. The second one is a stop-band filter and is analyzed in the paragraphs above.

The measurement system was based on automated syringe pumps for microfluidic characterizations, and the microwave setup was constituted by an Agilent network analyzer 8510. The measurement bench is presented in Figure 17 and the frequencies were performed up to a 2500 MHz range using Microtech GSG probes:

5.2. Dielectric Liquid Measurements

We characterized two structures, and we present the extracted practical results for six positions of DI water in capacitor microchannels as shown in Figure 18:

First, Figure 18a illustrates the obtained results from the pass-band filter that exhibits a low insertion loss between −0.5 dB and −0.9 dB. Furthermore, the resonant frequency variation is between 420 MHz and 690 MHz, allowing for an important tuning range of Tr = 64%. In concordance with the pass-band filter, the stop-filter exhibits a near cut-off frequency that is between 380 MHz and 620 MHz, with a tuning range of Tr = 63%. The obtained variations are lower than those obtained by simulation, which exceeds Tr = 190% due to losses in real dielectric liquid, which decrease the electric field distribution between two capacitor electrodes and deteriorate the total filter capability.

Furthermore, the circulation of Di water between the capacitors’ electrodes in two cases decreases the frequency response by 1000 MHz from the empty state, and the two ranges are too close and are between 400 MHz and 700 MHz.

5.3. Salt Water Displacement Characterization

The obtained results from extracted parameters were modeled in HFSS. Therefore, the practical analyses are illustrated in Figure 19:

The pass-band filter ensures a tuning range of Tr = 27.5% for a frequency range between 1450 MHz and 1850 MHz, and the insertion loss is low and does not exceed −2.9 dB. Moreover, the stop-band cut-off frequency is close to the first one and is between 1390 MHz and 1710 MHz, with a tuning range of Tr = 23%, which is close to the simulation results, where Tr = 22.7%. This result shows that the presence of ionic liquid has a partial effect on the response of the microfluidic filter. Indeed, the electrical current density distribution on the inductor core is partially influenced by the presence of salt water.

On the other hand, the results show a lower frequency variation compared to capacitor actuation. Indeed, the capacitance value depends on the electric field distribution between capacitor electrodes affected by the presence of the high permittivity of DI water compared to the lower electric conductivity of salt water and the ability to change the medium permeability and/or the inductor core impedance.

6. Conclusions

Throughout this study, we have discussed the behavior of the RF MEMS microfluidic tunable parallel structure in response to the presence and displacement of different types of liquids (conductor and dielectric). We have shown that the parallel association of the two microfluidic components (capacitor and inductor) generates two different operating modes. Each mode has a corresponding frequency range defined by the response of the dB (S11) and dB (S21) gains. The transition between the two modes is delimited by the weakening of the current density in the structure. The crossover frequency is determined from the minimum value of the gain dB (S21). In addition, the displacement of dielectric liquid between the capacitor electrode provides a higher variation than that obtained with the inductor using Galinstan, and the obtained results are, respectively, Tr = 194.5% and Tr = 24.2%. These results were then evaluated in the experimental part and a good agreement with the behavior of the microwave parallel structure in response to two different liquids was found. Moreover, the experimental results demonstrate the effect of microfluidic actuation in two different structures that associate the capacitor and inductor in serial (previous work) and parallel topologies. Indeed, the presence of dielectric liquid in capacitor microchannels decreases the resonant frequency but ensures a high tunability. On the other hand, when using conductive liquids with high and low conductivity, the resonant frequency increases but presents a low variation.