Diversity is good, but is more diversity better? Diversity enables a species to survive, it weeds out dead-end developments, and it's a source of seemingly infinite improvement. So is more better? Not always, at least from a processing/production point of view. Let's look at MEMS: how many processes already exist or are proposed as production technologies? Does every MEMS product have its own technology as some believe? Is it good to have such diversity in MEMS processing? Is the MEMS world becoming more or less diversified in terms of technology? To find answers to these questions, we analyzed over 110 MEMS processes from established MEMS suppliers, start ups, universities, etc. Surprisingly, from all these products only a few (7) can be regarded as product/company unique. The rest can be more or less easily divided into standardized process techniques.
So what are the differences between these processes? You don't need much to make a MEMS device: a free-hanging structure, a construct to detect the movement of the free-hanging part, and, as always, some electronics. To make a free-hanging structure you need a suitable material: a structural layer and a process to remove the material around that layer (sacrificial etching). If material is removed from the substrate, the process is called bulk micromachining (BMM), and if deposited layers are removed, the process is called surface micromachining (SMM).
The process you will choose is of course very much product-based. Pressure sensors are still mainly created using BMM, accelerometers and gyroscopes by polysilicon-based SMM (or low-temperature alternatives), RF MEMS products by metal SMM, and microphones by dielectric-based SMM. If you decide to build a micropump with MEMS technologies, using multi-stack wafers structured with BMM is the technology of choice.
So it's more or less clear what the status of MEMS processing is, but what are the trends? BMM is in some ways a technology of the past; it is difficult to produce low-cost products with BMM or to combine it with CMOS, and, even more importantly, it needs a lot of expensive wafer space. However, many products made using BMM are still on the market and will stay there for many years to come.
These products are often characterized by having high-precision performance, and they tend to be highly robust. Using standard CMOS layers as structural layers would be ideal, but this technique still has many problems and is not yet seen in products on the market. Polysilicon is CMOS-compatible, but it is difficult to combine with CMOS due to the differences in processing temperatures. That's why low-temperature alternatives are under investigation, like IMEC's SiGe process. A route that is attracting a growing band of supporters is using silicon on insulator (SOI) wafers. The top silicon layer is used as a structural layer, and the underlying oxide layer acts as a sacrificial layer. SOI processing is easy to combine with IC processing, since you can process the IC part before you etch the MEMS part. The growing popularity of SOI in MEMS and its popularity in the IC industry (i.e. for power devices) are behind the growing demand for those wafers.
So in short, the trends in MEMS processing are integration with CMOS, from bulk to surface micromachining and, as always, the smaller the better!
-Patric SalomonenablingMNT, GermanyPatric Salomon (Patric@enablingMNT.com) offers services with a focus on marketing, public relations, and business strategy in micro- and nanotechnology. He is the vice-chair of NEXUS, the European Microsystems Association.
-Henne van HeerenenablingMNT, The Netherlands
Henne Van Heeren (Henne@enablingMNT.com) is a specialist in MNT manufacturing, packaging, and technology in business-related issues. He is a member of the NEXUS Task Force on market research and an author of the MANCEF MNT roadmap.
Salomon and van Heeren advise industry and public bodies on MNT product commercialization and business development. They publish a report series called enablingMNT Industry Reviews, www.enablingMNT.com.
