|
Tronics, Europractice's first MPW MEMS partner, talks about thick SOI metrology challenges |
|
|
|
Mar 11, 2008 at 12:54 PM |
As one of the premier purveyors of MEMS fabricated on thick silicon-on-insulator (SOI) wafers, Tronics Microsystems knows a thing or two about the metrology challenges involved with those fat, bonded substrates.
The Crolles, France-based foundry just signed on with IMEC and the Europractice IC Services multiproject wafer (MPW) program as the first MEMS tech partner. Tronics's expertise in processing "60-micron-thick SOI high-aspect ratio micromachining with hermetic wafer-level packaging" will be leveraged to support the MPW effort's global academic customer base, in those cases when a university or research outfit needs protoyping or other assistance while investigating new sensor, accelerometer, gyroscopic, and related micromechanical structures and devices.
I interviewed Jean-Marc Gilet, Tronics' process and supply chain manager, via email for a feature on MEMS metrology that was first published in the January/February 2008 issue of Small Times and then reprinted in the current (March 2008) edition of Solid State Technology. As is often the case, only a fraction of the interview ended up in print. Since certain unpublished portions of my November 2007 Q&A with Gilet focus on the special challenges of thick SOI metrology, here are a few relevant excerpts to add a little technical color to the Europractice MPW news.
"Suppliers of SOI wafers deliver their wafer with reliable specifications (e.g., top and handle SOI thicknesses and tolerances or oxide thickness)," explained Gilet. "Nevertheless, monitoring the wafer bow/warp is an important parameter for the process stability. Stress measurements are performed to check this aspect.
"For the release of structures using HF, you need to have a very good knowledge of the BOX [buried oxide] properties provided by the different suppliers and select the right ones in function of your product requirements. You also then need to assess precisely what has been etched and not etched.
"In thick etching of SOI by deep-reactive ion etching, or DRIE (typically 20-, 60- and 100-micron-thick are run at Tronics), controlling the etch quality--- e.g. surface smoothness of the etch walls and profile control---by metrology is quite often not the best strategy. Indeed, here the metrology control can only be destructive using SEM and done on test wafers or on samples or structures taken out from the wafer. We found out years ago that electromechanical tests strategies on proprietary well-known test structures and electrical or functional tests on the products can bring far more information when correlated adequately (e.g. surface roughness, scalloping, profiles)."
As for some of the most difficult metrology issues that Tronics faces in manufacturing the Geosensor device (which has a buried hermetic cavity) and microfluidic devices, Gilet had this to say.
"All or most of the inertial accelerometers and gyros we develop or produce present the same difficult metrology issues. The performance of those devices are mostly linked to two factors: The deep etch quality on the thick SOI structures and the vacuum value and stability over time.
"For the first point, visual and dimensional CD controls are performed on the surface of the etched structure, and additional in-line electrical tests are performed to validate the quality of the etched profiles and are augmented by visual inspections. The vacuum value and stability is mostly controlled by a 100% measurement of the Q-factor measurements of the products as this value is directly linked to the vacuum. Our five years' track record in manufacturing vacuum-packaged devices gave the necessary return-on-experience to establish the right process controls and visual inspections required to guarantee the hermeticity."
"Stiction in capacitive or electrostatic structures was of course an issue that we had to deal with," he notes. "No metrology equipment can easily detect such stiction. The solutions here were to first establish robust designs and antisticking process solutions and then set up final burn-in procedures to remove from the line components that, despite the design and front-end solutions, are still presenting some risks of stiction.
"In a microfluidic actuator you also encounter stiction risks but with different sources of failures. Here you need to check that the valves are closed tightly (and not leaking) but not stuck together in a close mode and opened under the right conditions. Here again the metrology does not help; only specific test and characterization benches and protocols can allow the precise control of your production quality."
|
Blogs 
