Integrated Pressure Monitor Technology for RF MEMS

Many MEMS devices, like accelerometers and mechanical resonance devices, require a vacuum for proper operation. Verifying that a package cavity has the required vacuum, though, can be a trick. Researchers in Khalil Najafi's group at the University of Michigan (Ann Arbor) have developed methods for micromachining Pirani gauges in package lids that could outperform helium leak testing for the package, according to work published in the IEEE Transactions on Advanced Packaging .

The performance of an RF MEMS device, for example, may depend on the properties of a mechanical resonating structure. The presence of air or any gas would not only hinder the motion of such a device, but it would also make it less predictable.

Another possibility is that trapped air or moisture could react with MEMS structures, altering their mechanical characteristics. The inaccuracy caused by the proof-mass in an accelerometer changing may not be significant, but a change in the deflection properties of the beam used to measure the force on that proof-mass could be significant. Also, a change in the distribution of mass in a mechanical resonance structure could significantly change the RF properties of the device.

One way to detect a leak in an RF MEMS package is Q-factor extraction, in which the resonant structure characteristics are calculated from its electrical characteristics. By the time a leak is detected using this technique, however, the device may have already gone bad.

Helium leak testing is a standard method, but it is difficult and expensive to use on an individual packaged device. It would probably be simpler and less expensive to wait until a MEMS device performance starts to degrade and then replace it. In cases where this would be unacceptable, an inexpensive, in situ pressure monitor would be welcome.

Pirani gauges serve well in the milliTorr range. A Pirani gauge has a chamber that contains a wire (or resistor) and a heat sink. A current is passed through the wire, and its resistance is measured. If the pressure is very low, then very little heat is carried from the wire to the heat sink by convection, and the wire has the resistance value associated with a higher temperature. At a higher gas pressure, the temperature of the wire is lower because of convective cooling, which gives a lower resistance measurement.

Ideally, Pirani gauges are operated in the molecular regime, where the distance from the wire to the heat sink is smaller than the mean free path of a gas molecule. This puts an upper limit on the useful range of a given Pirani gauge. The lower limit is set by the need for heat transfer through the gas to be greater than heat transfer by radiation or through supporting structures.

The group developed two designs involving dual heat sinks. The traditional vertical design has resistor to heat sink gaps of 0.4 µm and a dynamic range of 20 mTorr to 2 Torr. The lateral design has 1 µm gaps and a dynamic range of 50 mTorr to 5 Torr. Both are machined out of heavily doped p-type (p++) silicon, and both are compatible with MEMS packaging and have suitable performance. The lateral design, however, requires only two masking steps, while the vertical design requires six.

To make the lateral gauge, a silicon wafer is first blanket doped with boron (p++), from which the resistor and heat sinks are to be formed. Deep reactive ion etching is used to make 1 µm wide trenches to isolate the p++ regions (Figure ). A glass wafer is etched to form the cavities, and the two wafers are bonded face to face using anodic bonding. The undoped silicon is removed using a selective etch, typically ethylene-diamine pyrocatechol (EDP).

Lateral Pirani gauges can be manufactured in a simple, two-mask dissolved wafer process.

Though this dissolved wafer process uses a lot of silicon, it is not really expensive. Silicon is not expensive until circuitry is made on it, and the alternatives for measuring leak rates are significantly more expensive.

The work did not include the integration of a Pirani gauge in a package, but the main components are made of p++ silicon, a widely used material for structural components of MEMS devices. Including a Pirani gauge inside a MEMS package might only require simple adjustments to a MEMS design layout.

At last year's IEEE MEMS conference, the group published work on Pirani gauges using various beam and ladder structures made of polysilicon for the sensing resistor, some of which had a reported useful range from atmospheric pressure to <10 mTorr.

The group has also published work on encapsulating these devices using gold-silicon eutectic processes and integrating getters from NanoGetters, a subsidiary of Integrated Sensing Systems (Ypsilanti, Mich.). Integrating sensors of this type might seem a bit much for an RF MEMS device for now. With their increased use, though, it may be necessary to identify and replace a failing device before its performance degrades too far.

Find more information on semiconductor packaging.