Electrostatically actuated micromechanical switches using surface micromachining

Introduction

In many electronic applications, there exists a need for a device which acts as nearly like an ideal switch as possible. Currently a designer's choices are limited to the MOS analog switch or a relay. MOS analog devices have a size and speed advantage and can be integrated into switching arrays. Their main disadvantages are that the have a relatively high on impedance and low off impedance, and suffer from stray capacitance. Reed relays on the other hand are large and slow, but behave much more like an ideal switch. In addition, reed relays can be designed to minimize insertion losses and maintain signal fidelity. This attribute is particularly important for automated test equipment (ATE).

In this paper, we report our work on a surface micromachined micromechanical switch whose performance combines some of the best attributes of these two competing technologies. We show that it is feasible to make devices as small as one micron wide by 20 microns long. If configured as a three terminal device, its characteristics are close to those of a near ideal FET switch. A voltage greater than the threshold voltage applied between the source and the gate produces a low impedance connection between the source and the drain. Below this threshold voltage, the drain and source are isolated by a high impedance (>1010 Ohm). The device is fabricated on an insulating substrate (glass).


Principle of Operation

Figure 1. Basic configuration of a micromechanical switch.

Figure 1 shows the basic switch geometry being fabricated. The operation of this device is very simple. When a voltage is applied to the gate electrode, the beam is pulled down by electrostatic force until the switch closes. When the gate voltage is removed, the restoring force on the beam returns it to its original position.

Figure 2. Lumped mechanical model of a microswitch.

To help understand the performance of our device the deflection of the beam can be crudely analyzed using the lumped mechanical model shown in Figure 2. An electrode which represents the source (beam) is suspended a distance d above a second electrode (gate) by a spring. Figure 4. The voltage as a function of beam deflection.

The voltage required to deflect the beam the distance shown on the x-axis is plotted in Figure 4, using typical values for the various parameters. The plot shows that for deflections from zero to one-third of the initial gate-to-beam spacing d, the gate voltage required to hold the beam in its deflected position increases monotonically with the deflection . Further increases in deflection of the beam result in the deflected position requiring a monotonically decreasing gate voltage to hold the beam. Therefore, the system is unstable, and the beam snaps down to close the switch at this voltage, which is in effect the threshold voltage.

The threshold voltage according to our simplified model can be shown to be

When the switch is in its closed position, the gap between the beam and the gate is smaller than it was at the threshold. The gate voltage must therefore be reduced to a voltage lower than the threshold voltage for the switch to open. If we approximate the average gate to beam spacing in the closed position as d/2, the voltage at which the switch opens is given by The switch therefore appears to have hysterisis with respect to the gate voltage.

For some switching applications, it may be appropriate to achieve the fastest switching possible. The model presented above provides a simple means to approximately determine the conditions for high speed operation. Clearly, the stiffer the structure, the higher the switching speed. However, the threshold voltage will also increase with stiffness. The charts shown in Figures 5 and 6 show the relationship between this two competing effects. It can be seen that with a ten micron long by 0.4 microns thick beam the threshold voltage will be about 30 volts and the device will operate at between 2 and 3 MHz.

The lumped mechanical model discussed so far ignores the curvature of the beam (either cantilever or bridge) under the electrostatic field due to the gate. A more general model follows, in which the deflection of the beam under the field due to the gate voltage is solved numerically.

Figure 7. Schematic of cantilever showing internal shear forces and bending moments.

Referring to Figure 7 the electrostatic force (per unit length) acting on the beam is given by

Deflection of beam at a position x along the beam is obtained by solving The internal bending moment is calculated using the following boundary conditions; The system of equations is solved iteratively until convergence is achieved. This is repeated at different voltages to obtain a deflection profile for the beam , as a function of the applied voltage V (Figure 8). The numerical solution confirms the instability in the position of the beam at the threshold voltage (Figure 4).


Design

The primary goals for the design of the relay or switch are a low contact resistance, low threshold voltage, and high switching speed. The threshold voltage may be reduced by increasing the area of the gate, (thus increasing the electrostatic force acting on the beam), reducing the spring constant of the beam and by reducing the gap between the beam and the gate electrode. Reducing the spring constant will limit the maximum switching speed. Reducing the gap will result in an increase in the capacitive coupling between the gate and signal line. Another way of reducing the threshold voltage would be to reduce the gap between the contact tip and the gate (Figures 1 and 2) by increasing the size of the contact tip. By this method, the switch could be made to close before the point of instability was reached (Figure 4). Such a switch would not exhibit the hysteritic behavior discussed in the previous section.

Reduction in on-resistance can be achieved by increasing the width of the beam and increasing the thickness of the contact layer. The latter measure would, as previously stated, limit the switching speed by decreasing the stiffness of the structure. Finally, for a reliable contact to be obtained, the contact area should be very small relative to the gate area, resulting in a high force of contact.

An array of devices was designed using several different geometries, including a simple cantilever, doubly-supported bridge, and a centrally pivoted structure. For each configuration, geometric parameters were varied. These include the number and size of the contacts and the layout of the contact pads. The cantilever structures were approximately 65 microns in length, and 30 microns wide. The thickness of the cantilevers is approximately 2 microns and the beam-to-gate spacing is about 1.5 microns.


Fabrication

The fabrication sequence begins with the deposition and patterning of the first metal layer (chrome-gold) to define the gate and contact electrodes on the glass substrate (see Figure 9). A sacrificial metal layer (copper) approximately 2 microns thick is then deposited. This is patterned in 2 steps. In the first step, the sacrificial layer is partially etched to define the contact tips for the beam. In the second etch step, the sacrificial layer is etched all the way down to the source contact metal to define the beam supports. Subsequently, photoresist is spun on top of the sacrificial layer and patterned to define the mask for the beam structure. The beam consists of a 2 micron thick layer of nickel on top of a 200 nm thick layer of gold. Both these layers can be formed either by electroplating or by electroless plating. The gold layer serves as the contact material with the gold contact pads when the switch closes. Finally, the sacrificial layer is removed by a suitable wet etching process to release the free-standing beam.

Figure 9. Schematic of the fabrication sequence.


Results

......

Figure 10. SEM micrograph of a completed three terminal switch. On the left, the base end of the beam which contacts the source metallization is seen. The gate is visible beneath the beam and at the free end of the cantilever, the contact tips are apparent.

A test die containing about 200 different switch configurations has been designed. Switches include three terminal devices as described above and four and more terminal devices with isolated contacts. Our work has concentrated on the evaluation of three terminal devices. Figure 10 shows a SEM micrograph of a completed three terminal switch. The support end of the beam contacts the source electrode and the beam cantilevers over the drain. At the drain end of the beam, two indentations are visible. These are the contact points for the switch. After actuating switches for a large number of cycles (>106), the beams were turned over using a micromanipulator and the tips were examined. Figure 11 shows two views of the tips on a such a beam. No degradation of the tip is apparent. Examination of the drain showed no evidence of degradation.

...... 

Figure 11. SEM micrographs of overturned switches showing the tips im more detail. The photograph on the left is shows the position of the tips relative to the end of the beam. The photograph on the righ is a detail of a tip which was taken from a position rotated 180o from the photo on the left.

Tests have been conducted to determine the potential lifetime of the devices. In particular, the switch was configured in a circuit consisting of a power supply and a series resistor. The value of the voltage and resistance were varied and the device was hot switched at different conditions. A square wave signal was applied to the gates of the devices and the current flowing between the source and the drain was monitored. All switching was performed in air. Table 1 shows the results achieved. The switches operated over 1e7 cycles before failure. Failures occurred for a variety of reasons including contact stiction and gate to beam shorts.

Variation in the threshold voltage was monitored on a second set of three terminal devices. These results ( Table 2 ) indicate a threshold voltage whose mean is 80.8 volts and standard deviation is 6.5 volts. These devices do not exhibit the hysteritic behavior discussed earlier. This leads us to believe that they operate in the stable portion of the deflection versus gate voltage curve (Figure 4). This large variation in threshold voltage is probably caused by variations in the mechanical properties of the beams. In particular, the gap between the contact tips and drain, the beam thickness and the gate to beam spacing are all sources of variation in the threshold voltage.

Contact resistance could not be measured accurately because the on-resistance was found to be dominated by the resistance of the on-chip metallization (20 Ohms). However, variation in the resistance of the closed switch was found to be much less than 1 Ohm. Figure 12 shows the response of the switch to a 2.5 kHz square wave while switching a signal current of 1 mA (a voltage source of 1V in series with a 1 KOhm resistor). The spikes at either edge of the output square wave may be the result of capacitive coupling between the gate and the signal line.

We have also fabricated micro-relays. We distinguish micro-relays and micro-switches by the relationship between the actuator and the contacting functions. In a relay, a minimum of four terminals are required. The actuator and switching functions are separated. In such a device, the drain function is split between two terminals and the actuator brings a contacting member into contact with the two drains shorting them together and completing an electrical path. The electrical isolation between the actuator and the contactor is achieved with a high quality dielectric material which mechanically connects them. The figure below shows a completed micro-relay.

Figure 12. SEM micrograph of a micro-relay.


Conclusions

A simple process has been developed for the fabrication of a micromechanical switches. The devices are actuated electrostatically and exhibit both hysteretic and non-hysteretic operation. Several configurations have been fabricated and testing of three terminal devices has been conducted. Beam and lead resistance are the primary sources of on-impedance which has been measured at about 20 ohms. Variation in the contact resistance is less than one Ohm. The threshold voltage for the range of devices fabricated varied from a low of 70 volts to a high of 200 volts. The threshold voltage variation for a fixed geometry averaged 80.8 volts with a standard deviation of 6.5 volts. Devices were switched for over 107 cycles with as much as 10 volts applied between the source and drain and the current limited at as high as 5 ma.