MEMS devices are typically joined to CMOS ICs in order to integrate with larger electronic systems; MEMS doing the job of sensing and physical control while CMOS ICs provide the interface functions such as analog to digital conversion, amplification, signal filtering, and information processing and communication. Many of the processes used for fabricating CMOS devices are also used for building MEMS devices with some slight processing variations.
The two devices can be integrated either by separate processing and subsequent hybridizing and packaging side by side or stacked, or monolithically on the same substrate using consecutive or interlaced processing schemes. The prominent processes in our discussion today are Polysilicon and Silicon Nitride. The needs of the MEMS device, because they are the mechanical portion of the package, must be significantly free of stress whereas the CMOS IC used as an oxidation mask or gate dielectric has less of a stress requirement.
For example, a CMOS device can use stoichiometric Nitride at a stress level of ~1,200 MPa with no deleterious effects whereas a cantilever in a MEMS device is required to be almost stress free and remain straight in action. To accomplish this, we make our Silicon Nitride rich in Silicon. That means we reverse the ratio of Dichlorosilane or DCS (SiH2Cl2) to Ammonia (NH3) flow rates from 1/3 to 3~6/1. The added silicon in the film relieves the stress as shown in the table below:
Table: Examples of residual stress in LPCVD polysilicon vs. process temperature.