Indian Journal of Pure & Applied Physics

Vol. 45, April 2007, pp. 387-394

Studies on RF MEMS shunt switch

Preeti Sharma, Shiban K Koul* & Sudhir Chandra

Center for Applied Research in Electronics, Indian Institute of Technology Delhi, Hauz-Khas, New Delhi 110 016
*E-mail: [email protected]

Received 7 June 2006; revised 12 December 2006; accepted 14 February 2007

Radio Frequency Micro-Electro-Mechanical-Systems (RF MEMS) based switches are expected to play a key role in the
field of microwave switching. Traits like low-loss performance; zero-power consumption and very low inter-modulation
distortion have made these switches well suitable for high-performance microwave and millimeter-wave circuits. The
design of RF MEMS switches, however, require sophisticated modeling techniques because of their complicated 3-D
geometries with critical aspect ratios and strong in-homogeneities. In this paper, the effect of various geometric dimensional
parameters on the electromagnetic and mechanical behaviour of a shunt capacitive MEMS switch using finite element
method based Ansoft High Frequency Structure Simulator (HFSS) software tool has been presented. This exercise will
provide a better insight into designing a reliable and high-performance MEMS shunt switch.

Keywords: MEMS shunt switch, Finite-element method, Coplanar Waveguide, RF MEMS

IPC Code: B81B7/02

1 Introduction 2 Electrical and Mechanical Model of the Switch

Micro-electro-mechanical Systems (MEMS) 2.1 Equivalent circuit model of the MEMS shunt switch
electrostatically actuated shunt switches have been The MEMS shunt switch (shown in Fig. 1) is
recently demonstrated for low-loss microwave modeled by two short sections of transmission line
and a lumped CLR model as shown in Fig. 2(a). The
and millimeter-wave applications. The structure of a
capacitance value varies in up-and-down-state of the
shunt-capacitive MEMS switch (Fig. 1) consists of a
bridge with Cu being the up-state capacitance and Cd,
thin metal membrane bridge that is suspended over
the down state capacitance. The transmission line
the center conductor of a coplanar waveguide (CPW)
sections are of length (w/2) + l where l is the distance
and fixed on the ground conductor at either end, as
(a)
shown in Fig. 1(a). When a dc voltage is applied to
the lower electrode, the electrostatic force causes the
suspended metal membrane to snap down onto the
lower electrode and dielectric surface, forming a low
impedance RF path to ground. As the operating
frequencies are shifting to higher regimes and
fabrication conditions need to be more stringent to (b)
fulfill the performance demands, the design needs to
be more exhaustive and robust to reduce the
prototyping time. In this paper, we have carried out
full-wave electromagnetic analysis to predict the
behavior of the switch under geometric dimensional
variations. The geometric parameters investigated in
the present study are air gap height (g), bridge-width
(w), bridge-thickness (t) and dielectric layer thickness
(td). Mechanical behaviour as a function of each Fig. 1Schematic of typical MEMS shunt switch in up state
geometric parameter is also presented and discussed. (a) cross-section and (b) top view
388 INDIAN J PURE & APPL PHYS, VOL. 45, APRIL 2007

(a) − jωCu Z 0
S11 = … (4)
2 + jωCu Z 0

and for S11<-10 dB or ωCuZ0 << 2

ω2 Cu2 Z 02
2
S11 = … (5)
4

Cu is the up-state capacitance that comprises of

parallel plate capacitance and the fringing
capacitance1. Typical up-state capacitance for shunt
switches operating at microwave and millimeter-wave
(b) frequencies is ~ 35-160 fF whereas inductance is of
the order of few nH. The LC-series resonance
frequency is given by:
1
f0 = … (6)
2π LCu

It can be seen that in up-state position, the resonant

frequency is essentially decided by up-state
capacitance.

Governing equations during OFF-state: Capacitance

Fig. 2(a) Equivalent circuit model and (b) schematic defining l
and inductance
from the reference plane to the edge of the MEMS The down state capacitance of the switch is
bridge (Fig. 2 (b)). calculated by Cd = ε0εrA/td. The resonant frequency in
The switch shunt impedance is given by: the down state is given by Eq. (2) by using Cd in place
of Cu.
1
Z s = Rs + jωL + … (1)
j ωC 2.2 Mechanical model
For a beam over a CPW line having center
With C = Cu or Cd depending on the position of the conductor width as one third of its length and with
switch. The LC series resonant frequency of the shunt force distributed uniformly above the center
switch is: conductor, the spring constant and pull in voltage are
given by1:
1
fr = … (2)
2π LC 3
 t   27   t  3 
k = 32 Ew     + 8σ (1 − ν ) w    … (7)
The impedance of the shunt switch can be  l   49   l  5 
approximated by;
8k
Vp = g3 … (8)
1/ jωC f = f0 27ε0Ww

Z s =  Rs f = f0 … (3)
 j ωL f ? f where, t is the thickness of the bridge (µm), ε0 the air-
 0
permittivity, W the width of the CPW center
Governing equations during ON-state: Capacitance conductor (µm), E the Young’s modulus of Gold, w
and inductance the width of the membrane (µm), ν the Poisson’s ratio
In the up-state position of the switch, the reflection of Gold, L the length of the membrane (µm), σ the
coefficient is given by: residual stress (MPa) and g is the air-gap (µm).
 SHARMA et al.: STUDIES ON RF MEMS SHUNT SWITCH 389

3 Effect of Geometric Dimensional Variation on 1-40 GHz for different geometrical parameters of the
the Performance of the MEMS Shunt Switch switch. The substrate is assumed to be gallium
In an effort to study the effect of various geometric arsenide with relative dielectric constant of 12.9. The
parameters on the switch performance, a 3-D thickness of the substrate is 200 µm. The CPW
parametric structure (shown in Fig. 3) was created and conductor material is assumed to be gold with
simulated using High Frequency Structural Simulator thickness of 4.0 µm. The bottom electrode of the
(HFSS). Each parameter is varied one at a time. In switch and the metallic membrane consist of 1 µm-
order to ensure accuracy of the results, each model thick gold. The bottom electrode of the switches is
was converged for at least three consecutive coated with silicon nitrate (Si3N4) having relative
iterations. In the simulation a box size 600 µm × 600 dielectric constant of 7 and thickness of 0.1 µm. The
µm × 1200 µm is used and boundary radiation multilayer switch structures used in the simulation in
conditions are imposed on the six sides of the box. S- ON- and OFF-states are shown in Figs. 3(a-b),
parameters are studied in the frequency range from respectively. Typical geometrical dimensions used in
the simulation of the switch are listed in Tables 1-2.
(a)
3.1 Effect of bridge-width
For a fixed value of the centre conductor
width=100 µm, Figure 4 shows the simulated
insertion loss and return loss in the ON-state (air-
gap=2µm) for various bridge heights. It is observed
that varying only the bridge width, while keeping all
other parameters fixed, effectively varies the up-state
capacitance and thus changes the insertion loss of the
switch. From Eq. (5), it is clear that an increase in
bridge-width will increase the magnitude of S11 and
hence decrease S11 in decibels. The same trend is
evident from the plots shown in Fig. 4 obtained from
full-wave analysis.
(b)
In the OFF-state, both inductance and capacitance
of the bridge determine the response of the switch.
Figure 5 shows the simulated return loss and isolation
of the switch in the down state. It is observed that on
varying the bridge-width, the resonant frequency
varies. A special point that is worth mentioning here
is that when the width of the bridge is varied
uniformly along the length (as shown in Fig. 4), both
the inductance and the capacitance of the bridge
varies. This is attributed to the fact that the bridge
inductance is mainly determined by the portion of the
bridge over the CPW slots and is independent of the
portion of the bridge over the center conductor.
Fig. 3Simulated switch geometry (a) ON-state and (b) OFF- Moreover, the portion of the bridge over the center
state conductor determines the bridge capacitance. Hence,
Typical dimensions of the shunt switch used in simulation (Ref. Fig. 1)
Table 1

Substrate thickness, h (µm) 200 Dielectric constant of the substrate 12.9

Length of the bridge, L (µm) 320 Thickness of the bridge, t (µm) 0.8 @ 17 GHz
Width of the bridge, w (µm) 80 Air-gap, g (µm) 2
s+2w (µm) 120+(2×100) Reference plane distance, l (µm) 40
Silicon nitride thickness (µm) 0.1 Dielectric constant of silicon-nitride 7
390 INDIAN J PURE & APPL PHYS, VOL. 45, APRIL 2007

the shift in resonant frequency as a function of bridge- increase the bridge inductance) or center-conductor
width observed in Fig. 5 is due to change in both the width (to increase the bridge capacitance) cannot be
inductance and the capacitance of the bridge. As a increased beyond a limit where the length of the
thumb rule, wider CPW slots and narrow bridge- bridge and higher-order mode excitation becomes
width results in large inductance value. So, in case problematic issues.
where the switch has to be designed for lower
frequency (< 10 GHz), the bridge structure shown in Mechanical behaviourIt should be noted that
Fig. 6 can be used, as width of the CPW slot (to although pull-in voltage given by Eq. (8) shows a
dependence on the beam width w, the pull-in voltage
Mechanical properties of the shunt switch used in
Table 2 is actually independent of the beam width since spring
simulation
constant k varies linearly with w. Thus changing the
Beam material Gold Young’s modulus (E) 78 electrode area by varying the bridge width is most
Residual stress favorable modification to tune the switch to different
20 Poisson’s ratio (ν) 0.44 frequencies while maintaining low actuation voltage.
(MPa) (σ)

(a) (b)

Fig. 4Simulated (a) return and (b) insertion loss characteristics of the switch in its ON-state (air-gap = 2 µm) for various bridge widths,
center conductor width = 100 µm

(a) (b)

Fig. 5Simulated (a) return loss and (b) isolation characteristics of the switch in its OFF-state for various bridge widths, center
conductor width = 100 µm
 SHARMA et al.: STUDIES ON RF MEMS SHUNT SWITCH 391

Optimum design considerations The width of the completely dependent on fabrication process and can
bridge is practically limited to 200 µm so as to result be measured as discussed in literature2. Theoretical
in a flat contact area between the MEMS bridge and pull-in voltage as a function of bridge-thickness for
the center conductor of CPW. Hence the only design unstressed and stressed beam is listed in Table-3. The
criterion to choose a specific value (should be at least dimensions of the CPW and MEMS bridge used in
< 200 µm) of bridge-width is decided by the resonant calculations are the same as those listed in Tables 1-2
frequency of the switch. accept for bridge-thickness that is varied in the
present study.
3.2 Effect of bridge-thickness Optimum design considerations Though the
Electromagnetic behaviour The resistance of the bridge thickness can be kept 3-4 times the skin depth
switch comprises of two components, Rs1 and Rs but one must restrict it between 1-2 skin depths
where Rs1 is due to the transmission-line loss and can (which in present case is between 0.6-1.2 µm) for
be calculated using: proper operation of the switch. The upper value is
weakly limited by the variation in series resistance of
Rs1 = 2αZ 0 l … (9) the bridge with frequency but strongly due to the large
pull-in voltage required to pull down a thicker
α is the line loss in Np/m. The second component Rs membrane.
is due to the MEMS bridge only. If the bridge
3.3 Effect of silicon nitride thickness
thickness is smaller than two skin depths, the switch
resistance is constant with frequency and for thickness Electromagnetic Behaviour Silicon-nitride (εr =
greater than two skin depths, the switch resistance 7) layer is deposited on CPW center conductor to
have a capacitive coupling between the MEMS bridge
changes with frequency as f due to the skin-depth
and CPW center conductor. In this section we have
effect. The skin depth (at 20°C) is computed by: investigated the effect of silicon-nitride thickness on
the loss characteristics of the switch in its two
2
δ= … (10) operational states. Figures 7-8 show the simulated
ωµσ reflection and transmission losses as a function of
frequency for various dielectric layer thicknesses in
where, ω = 2πf0, f0 is the resonant frequency, µ0 = 4π ON-state and OFF-state, respectively.
× 10-7 H/m is the free-space permeability and σ is the From the plots, it is clear that variation in dielectric
metal conductivity. For gold, σ is 4.098 × 107 S/m layer thickness alters only the resonant frequency of
(20°C). As an example, 17 GHz switch with gold the switch and that too a slight variation in silicon-
bridge will have the theoretical skin depth (20°C) nitride thickness shifts the resonant frequency
equals to 0.6 µm. considerably as compared to the shift observed as a
Mechanical behaviour It should be noted from function of bridge-width. This implies that the
Eq. (7) that spring constant k is a function of t3 and dielectric deposition process must be strictly
thus pull-in voltage increases drastically with increase controlled to ensure minimum deviation of switch
in bridge thickness t. In practice, typical biaxial stress resonant frequency from the designed one.
is between 10-20 MPa but the exact value is Mechanical behaviour According to the
mechanical model of the switch, the properties of
Theoretical Pull-In voltage for different bridge-
Table 3
thickness

Bridge-thickness Pull-in voltage (volts) Pull-in voltage

(µm) Biaxial stress = 0 (volts)
MPa Biaxial stress =
20 MPa

0.5 10.8 17.45

1.0 30.6 36.21
Fig. 6Alternative geometry for low-frequency MEMS shunt 1.5 56.2 61.00
switches shown in OFF-state 2.0 86.5 90.77
392 INDIAN J PURE & APPL PHYS, VOL. 45, APRIL 2007

(a) (b)

Fig. 7Simulated (a) return loss and (b) insertion loss characteristics of the switch in its ON-state (air-gap = 2µm) for various silicon
nitride layer thickness, center conductor width is 100 µm

(a) (b)

Fig. 8Simulated (a) return loss and (b) isolation of the switch in its OFF-state for various silicon nitride layer thickness, center
conductor width is 100 µm

dielectric layer decide the bridge-capacitance (and down state. With a higher εr material, one can achieve
thus the pull-in voltage) and the reliability factors a higher theoretical capacitance ratio that results in
such as dielectric charging. Hence, while designing high-isolation and small size switches3. Mechanically,
the switch, both electrical and mechanical issues have the type of dielectric layer chosen must be hard to
to be considered simultaneously and in general a ensure long-term reliability, free from any sort of
trade-off has to be made between the thickness and crystallographic defects and pinholes and preferably
the dielectric constant of the dielectric layer. should have high-dielectric breakdown. As far as the
thickness is concerned, typical value between 0.1- 0.5
Optimum design considerations In general, the µm can be selected. Here, the lower value is limited
choice of dielectric layer and its thickness is crucial by quality of dielectric layer and dielectric breakdown
for the switch design. Electrically, the dielectric for typical actuation voltage (20-50 V). The upper
constant and thickness of the dielectric layer value is limited by dielectric charging, the thicker
determine the resonant frequency of the switch in the dielectric is more susceptible to dielectric charging4.
 SHARMA et al.: STUDIES ON RF MEMS SHUNT SWITCH 393

3.4 Effect of air-gap height Mechanical behaviour A comparative study is

Electromagnetic behaviour The return loss, done by calculating the theoretical pull-in voltage as a
insertion loss and isolation characteristics of the function of different air-gap heights for both
switch in its two operating states for different air-gap unstressed and stressed beam and the results are
heights is plotted in Fig. 9. In the OFF-state, the presented in Table 4. The dimensions of the CPW and
resonant frequency is completely determined by Cd MEMS bridge used in calculations are same as listed
and L. With an increase in the air-gap height, up-state in Tables 1-2, accept for air-gap height that is varied
capacitance decreases thereby decreasing the reflected in the present study.
power from the switch S11 ( 2
). Optimum design considerations In practice, the
released MEMS bridge is always under stress and the
(a) (a)

(b)

(b)

Fig.10(a) Multi-layer capacitive switch given in Ref. [5] in its

up-position and (b) Top-view showing the conductor pattern

Theoretical Pull-In voltage for different Air-gap

Table 4
heights
Air-gap Pull-in voltage (volts) Pull-in voltage (volts)
height Biaxial stress = 0 MPa Biaxial stress = 20
(µm) MPa
1 7.74 9.87
Fig. 9Simulated (a) return loss (b) insertion loss and isolation 2 21.90 27.92
characteristics of the switch for various air-gap heights, other 3 40.23 51.29
dimensions used in simulation are listed in Table 1 4 61.94 78.96

Typical dimensions of the shunt switch of Ref.(5) that was used to validate our study
Table 5

Substrate thickness, h (µm) 400 Dielectric constant of the substrate 11.9

Resistivity of the silicon (Ωcm) 3000 Dielectric constant of silicon-dioxide 4
Length of the bridge, L (µm) 300 Thickness of oxide layer, t (SiO2) (µm) 0.4
Width of the bridge, w (µm) 80 Thickness of the bridge, t (µm) 2
s+2w (µm) 100+(2×60) Air-gap, g (µm) 1.5
Silicon nitride thickness (µm) 0.14 Dielectric constant of silicon-nitride 7.6
394 INDIAN J PURE & APPL PHYS, VOL. 45, APRIL 2007

(a) given in literature5 with our model and 3D FEM full-

wave simulation. For reference the details of the
silicon-switch shown in Fig. 10 are listed in Table 5.

The silicon nitride thickness that was quoted in the

literature5 is in the range of 0.1-0.2 µm. The present
study showed that the thickness of silicon nitride is
very crucial in determining the resonant frequency of
the switch. The exact thickness of the silicon-nitride
used in the fabricated switch presented in the
literature5 can be effectively determined by carrying
out a parametric study of the switch shown in Fig. 10.
Figure 11 (a) shows that the measured data presented
in Ref. (5), pp. 1052 for 80 µm wide bridge. This
(b) switch will only resonate at 35 GHz when the
thickness of silicon nitride is 0.14 µm, which lies
within the range mentioned in the reference5. For
comparison, the measured S21 (dB) in down-state
(OFF-state) of the switch extracted from the
reference5 and simulated S21 (dB) from our study are
plotted in Fig. 11 (b). From the plot, it can be
observed that the measured and simulated data agree
well.

5 Conclusion
We have presented the electromagnetic and
mechanical performance of a typical MEMS shunt
switch as a function of different geometrical
parameters. Various aspects of switch design are
discussed. The presented parametric results should be
Fig. 11(a) Simulated S -parameters for the switch dimensions
given in Ref (5) in its OFF-state for various silicon nitride layer useful to a designer to develop a high-performance
thickness and (b) simulated S21 (dB) for the switch dimension MEMS shunt switch. As an example, we have
given in Ref. (5) compared with its measured results validated the measured results of an earlier work by 3-
D parametric analysis.
value of its residual stress is entirely fabrication
process dependent. Although an increase in air-gap
References
will increase the ON-OFF capacitance ratio of the 1 Rebeiz G M,(Wiley, New York), 2003.
switch, the gap height is restricted to 1.5-2 µm 2 Chen S, Baughn T V, Yao Z J & Goldsmith C L, J Micro-
because of larger actuation voltage required to pull electro-mechanical Systems, 11 (2002) 309.
down the bridge. 3 Yu L, Taylor T R, Speck J S & York R A, IEEE MTT-S
International Microwave Symposium Digest, 1 (2002) 227.
4 Comparative Studies 4 Rotten X, 12th GaAs symposium, Amsterdam, 2004.
To show the usefulness of the detailed parametric 5 Muldavin J B, IEEE Trans Microwave Theory and
studies carried out, we compared the measured result Techniques, 48 (2000) 1045.