1. Introduction
Due to their low power consumption, high isolation, low insertion loss, and ability to integrate with other electronic devices, RF MEMS (Radio Frequency Micro-Electro-Mechanical System) switches have been receiving increasing attention. Such advantages render RF MEMS switches an attractive alternative to PIN or FET RF switches [
1]. RF MEMS switches are widely used in the fields of aerospace [
2], radar [
3], mobile communications [
4], tunable filters [
5], phase shift networks [
6], and other notable fields. These switches must satisfy stringent reliability requirements of failure-free operation over billions of cycles. However, due to the special structure thereof, such switches are frequently prone to failures.
Long term reliability still remains a matter of concern in the commercial utilization of RF MEMS switches [
7]. The reliability of RF MEMS switches is affected by mechanical and electrical phenomena such as stiction [
8], fatigue [
9], creep [
10], impact force [
11], residual stresses [
12], and dielectric charging [
13]. Among all these reliability issues, the presence of impact force is a major problem. Due to the mechanical nature thereof, the impact force during the suction phase of the switch is considerably high, causing the beam membrane to be prone to failure [
14]. Although significant efforts have been made to develop materials that can maintain high switching speeds while keeping structural failures low, high impact force remains one of the leading causes of device failure [
15]. For capacitive RF MEMS switches, the higher the impact force when the top and bottom electrodes of the switch come into contact, the faster the impact velocity. This results in more damage to the beam of the switch, which leads to a shorter switch life. In practice, there is always a trade-off between impact velocity and pull-in time. The variation of the fall time and impact velocity with the applied voltage was given in a previous study [
11]. Common methods of reducing impact velocity include tailoring the shape of the actuation pulse [
15], applying step voltage [
16], and using charge drive [
17]. A number of researchers have achieved significant results by adjusting the shape of the driving pulse to reduce the impact velocity [
18]. Despite such efforts, the customized pulse technique can only be applied in relatively slow switches with switching times Ts ≥ 10 us, requiring sufficient time to form an effective pulse train [
19]. The use of a charge drive from a constant current source to reduce the impact velocity in previous research [
19] resulted in an 80% reduction in the impact velocity with a bias resistance of 33 MΩ, but caused a 182% increase in the switching time.
In the present study, a new method was established for reducing the impact velocity of RF MEMS capacitive switches while keeping the pull-in time essentially unchanged. A new shunt capacitive MEMS switch was designed based on the support pillars being added to both sides of the signal line of the conventional MEMS switch. The effect of the positioning of the support pillars on the pull-in voltage was analyzed, and the effects of different positions and heights of the support pillars on the electromechanical and electromagnetic performance of the MEMS switch were simulated and compared. The new structure proposed in this paper can be used in satellite, mobile phone communication, radar, and other applications requiring high life.
3. Electromechanical Performance Analysis
Simulation software was used to model and simulate the new MEMS switch. Two support pillars were placed on both sides of the central signal line, as shown in
Figure 3. The theoretical value of
ranged from 30 um to 90 um. The positions of the support pillars were on both sides of the signal line at 90 um and close to the ground line at 30 um. The height
of the support pillars was the equivalent gap
after the beam membrane contacted the support pillars, and the theoretical value ranged from 0 um to 0.9 um. If the height
of the support pillars was 0 um, the pillars were not in contact with the beam membrane during the pull-down process. If the height
was 0.9 um, the pillars were always in contact with the beam membrane and the support pillars became the new anchor points.
Equation (8) shows that when the position
of the support pillars changed, the length of the equivalent beam membrane changed and the pull-in voltage
would also change.
Figure 7 depicts the relationship between the position
of the support pillar and the pull-in voltage
. The black curve is the theoretical value and the red curve is the simulated value. The distance between the top of the support pillars and the insulating dielectric layer was 0.75 um, but since the effect of the insulating dielectric layer is not considered in Equation (8), the initial distance needed to be added to the height of the insulating dielectric layer by 0.1 um. In the case of
= 0.85 um, changing the value of
yielded a different pull-in voltage
.
As the position
increased, the pull-in voltage gradually increased, and the larger
was, the faster the pull-in voltage increased, reaching the maximum pull-in voltage at
= 90 um. Such findings could be attributed to a larger
, indicating that the support column is closer to the centerline. The stronger the hindering effect of the support pillars on the bridge, the larger the equivalent elasticity coefficient of the beam membrane. The more difficulties in pulling down the bridge, the higher the pull-in voltage of the bridge. The indication is that the positioning of the support pillars has a large influence on the driving voltage of the MEMS switch. For the MEMS switch, a suitable pull-in voltage can weaken the ion injection into the dielectric layer to cause shielding failure or adhesion failure [
22]. Therefore, suitable support pillar positioning should be selected to keep the pull-in voltage within an acceptable range. The theoretical values in
Figure 7 are basically consistent with the simulated data curves, indicating that the equivalent model is physically reasonable.
Figure 8 depicts the falling displacement of the beam membrane at different times. The displacement changed slowly until 3 us, and the falling displacement was about 1/3 of the gap g
0 at 3 us. Subsequently, the displacement of the beam membrane increased rapidly and reached the maximum displacement at 3.9 us. The MEMS switch closed completely at 5 us. The simulation performance was as expected, and thus, the simulation software could be used for further analysis.
Figure 9a depicts the impact velocity and descent displacement as a function of time with and without the addition of the support pillars, respectively, with a driving voltage of 50 V. The velocity and displacement of both were the same from 0 to 3 us, indicating that the beam membrane was not in contact with the support pillars. The velocity slowed down and fluctuated around 0.5 m/s after 1 us of contacting the support pillars, and then the velocity increased rapidly. The bridge contacted the dielectric layer at 4.2 us when the instantaneous impact velocity reached a maximum of 1.1 m/s. Compared with the conventional MEMS switch, the pull-in time was only slightly increased, but the impact velocity was significantly reduced.
Figure 9b shows that there was a significant difference in the descent displacement after the beam membrane came into contact with the support pillars, which acted as a buffer for the descent of the bridge.
Figure 7 indicates that when the positioning of the support pillars changed, the MEMS switch pull-in voltage also changed. The impact velocity and pull-in time were not only related to the positioning of the support pillars but also to the applied voltage. If the applied voltage was less than the pull-in voltage causing the switch to fail to close, the impact velocity and pull-in time did not exist. As such, a voltage three to four times higher than the pull-in voltage was applied to weaken the effect of different pull-in voltages in order to observe the effect of the change of the support pillar positioning on the impact velocity and the pull-in time. The simulations were performed by applying 45 V and 50 V, respectively, and the height of the support pillars was 0.75 um.
Figure 10a shows the variation of the pull-in time by different positions of the support pillars, and
Figure 10b shows the variation of the impact velocity by different positions of the support pillars.
The impact velocity and pull-in time of the conventional MEMS switch and the new MEMS switch were considerably close to each other because the obstruction of the beam membrane by the support pillars was considerably small until 30 um.
Figure 10a shows that the pull-in time changed only slightly until 80 um, and the time increased significantly afterwards.
Figure 10b shows that the impact velocity decreased slightly with the increase in the support pillar positioning until 60 um, while the impact velocity decreased significantly between 60 um and 80 um, compared with the conventional MEMS switch. From the aforementioned analysis, a conclusion can be drawn that the increase in the support pillar position led to an increase in the pull-in time and a decrease in the impact velocity. Therefore, a suitable position can be found where the pull-in time remains the same while the impact velocity decreases significantly. As an example, at
x = 80 um, the closing time was 4.2 us and the impact velocity was 1.1 m/s. Compared with the pull-in time of 3.9 us and the impact velocity of 1.8 m/s without the support pillars, the closing time increased by 0.05% and the impact velocity decreased by almost 40%.
As shown in
Figure 11, the simulation was repeated several times when the driving voltage was 50 V and the positioning of the support pillars was changed to 80 um to analyze the change of the positioning of the support pillars on impact velocity and pull-in time.
Since the beam membrane did not come into contact with the support pillars during the pull-down process when the height was less than 0.4 um, the impact velocity and pull-in time were basically the same as when no support pillars were added. After 0.5 um, the pull-in time increased gradually with the increase in height. Further, as the height of the support pillars increased, the increase in the pull-in time became increasingly larger. Because the instantaneous velocity change of the bridge contacting with the dielectric layer was significantly large, the simulation results of the impact velocity had some errors due to the limitation of the mesh division accuracy of the simulation. However, an observation can be made from
Figure 11 that the overall change of impact velocity after 0.5 um exhibited a decreasing trend. A conclusion can be drawn that the higher the height of the support pillars, the longer the pull-in time and the smaller the impact velocity. Therefore, a suitable height can be found to make the best balance between pull-in time and impact velocity. For instance, for
x = 0.8 um, the impact velocity was 0.8 m/s, which was 45% less than 1.8 m/s without the support pillars, and the pull-in time was 4.7 us, which was 0.2% more than 3.9 us without the support pillars.
From the aforementioned analysis, a conclusion can be drawn that the positioning and height of the support pillars had a significant influence on the impact velocity and the pull-in time of the MEMS switch. The closer the positioning of the support pillars to the signal line and the higher the height, the lower the impact velocity and the longer the pull-in time, which is not suitable for high frequency applications. Thus, choosing the right position and height is critical. The positioning and height of the support pillars were not only related to the impact velocity and pull-in time, but also to the electromagnetic performance of the MEMS switch.
4. Electromagnetic Performance Analysis
In the present study, an electromagnetic simulation model of a parallel-connected capacitive RF MEMS switch was constructed. Further, the positioning and height of the support pillars were scanned and optimized by means of simulation software to derive the effect thereof on the electromagnetic performance. A total of nine combinations of x = 70 um, 80 um, and 90 um, and h = 0.7 um, 0.8 um, and 0.9 um were simulated.
As shown in
Figure 12, the curves of different heights at the same position approximately overlap, indicating that the height of the support pillars had basically no effect on the electromagnetic performance of the switch.
Figure 12a,b shows that the curves of S11 and S21 at
= 70 um and
= 80 um in the open-state approximately overlap, and the electromagnetic performance became worse at
= 90 um. Such results show that the increase in the position
of the support pillars decreased the electromagnetic performance of the MEMS switch. An observation can be made that the closer the support pillars are to the center signal line, the larger the S21 in the open-state and the smaller the S21 in the off-state. Such findings can be attributed to the contact between the support pillars and the signal line changing the equivalent capacitance of the MEMS switch, thereby degrading the electromagnetic performance of the switch. Thus, the actual support pillars should not be positioned too close to the center signal line. The influence of the height and position of the support pillars on the electromagnetic performance was considered to be optimal for the MEMS switch at
= 80 um and
= 0.8 um. The insertion loss was less than 0.5 dB up to 32 GHz, and the isolation was more than 50 dB at 40 GHz. Therefore, the effect of adding the support pillars on the electromagnetic performance of the MEMS switch was considerably small, and the suitable positioning of the support pillars could be selected to weaken the effect thereof. Overall, the proposed scheme of adding the support pillars is practical and feasible.
5. Conclusions
The present study provides a new method for reducing the impact velocity of the MEMS switch. Specifically, two symmetric support pillars were added under the bridge to buffer the beam during the falling process, thus reducing the impact velocity of the MEMS switch. The effect of adding the support pillars on the impact velocity and pull-in time of the MEMS switch was also analyzed, and the positioning and height of the support pillars were scanned and optimized. A conclusion can be drawn that the closer the support pillars are to the signal line and the higher the height, the lower the impact velocity and the longer the pull-in time. The electromagnetic performance of the switch was largely unaffected by the height of the support pillars. However, when the support pillars were considerably close to the signal line, the electromagnetic performance became considerably poor. The influence of the height and positioning of the support pillars on the impact velocity, pull-in time, and electromagnetic performance was found to be reduced by 40% from 1.8 m/s to 1.1 m/s at
= 80 um,
= 0.8 um, and 50 V applied voltage. The closing time increased from 3.9 us to 4.2 us, which was only 0.05%, the insertion loss was less than 0.5 dB up to 32 GHz, and the isolation degree was more than 50 dB at 40 GHz. The experimental data suggest that the RF MEMS switch designed in the present study has lower impact speed, shorter closing time, and better electromagnetic performance. Among other methods to reduce the impact velocity, the charge drive of the constant current source is used in the literature [
2] to reduce the impact velocity by 80%. The switching time increased by 182%. In the literature [
3], the impact velocity was reduced by 21.6% by modulating the actuation force with open-loop operations. We provided the method of adding support pillars, which reduced the impact velocity by 40% but increased the pull-in time by 0.05%.
However, the proposed new method is not perfect. The proposed switch reduces the impact velocity and increases the closing time at the same time, although the increased closing time is acceptable. In addition, manufacturing the proposed MEMS switch requires more than one process to manufacture the support column, which means that manufacturing this switch is technically complicated.