Portland, Ore. - A MEMS-based implant holds the promise of an artificial ear that would let the deaf hear without external electronics. The device mimics the snail-shaped structure of the inner ear, the cochlea, that makes hearing possible.
The fully integrated implant-the size of a real cochlea-has been fabricated with backside through-wafer etching using an inductively coupled plasma deep-reactive ion etch. It was based on a microelectromechanical system designed by University of Michigan associate professor Karl Grosh.
While other researchers have fabricated micromachined devices that mimic aspects of the cochlea, Grosh said, "our design differs from previous work by using a beam-array structure in a fully micromachined, liquid-filled two-duct structure."
Much of the lab work for the artificial cochlea was performed by a University of Michigan doctoral candidate, Robert White. Grosh and White will be honored by the National Academy of Sciences, which will publish the details of their accomplishment next month.
While other artificial cochleas are in use today (see www.eetimes.com/futureofsemis/directions/OEG20030923S0053), they are not integrated devices. Instead, an electrode is implanted in the audio nerve and a signal wirelessly transmitted from an external signal processor and microphone.
Rather than try to persuade the brain to accepts this new type of audio input, Grosh's MEMS cochlea is implanted into the same space occupied by the real cochlea, taking over its operations with mechanical structures that perform the same tasks. No wireless transceivers or signal processors are required.
Grosh's design was preceded by a decade or more of research into the mechanical-to-electrical operation of real cochleas. Others have tried a MEMS structure that either vibrated in the air (rather than the fluid environment inside the ear) or used a silicon-nitride membrane instead of the more cochlea-like beam-array structure in Grosh's design. Grosh's accomplishment was to combine a fluid-environment sensor with a MEMS design using a beam-like structure that imitates the human ear.
In a nutshell, an array of beams-attached only at each end, like a vibraphone-was micromachined in silicon to be the same 3 cm long as the real cochlea. By graduating 3,000 beams from large to small, a frequency analysis can be performed mechanically (since only the beams whose lengths match the wavelength of a corresponding input frequency will oscillate). In the current chip, the vibrations of the correct beams were confirmed by visual observation. In the next-generation device, however, each beam will be attached to a piezoelectric actuator. That will enable just the hair cell in the ear corresponding to a particular frequency to be stimulated on the audio nerve, achieving hearing renewal without requiring the brain to "relearn" the signal from a DSP (and eliminating the need for external parts).
"Our ultimate goal is to create a mechanical sensor with the highest possible localization and sensitivity over the same acoustic bandwidth as a real cochlea, but which can be practically manufactured and packaged using MEMS," said Grosh.
Real cochleas are a spiral-shaped affair with two fluid-filled chambers separated by a flexible membrane. Hairs vibrate in the fluid depending on their frequency, which gets higher as the spiral continuously narrows. For Grosh's artificial cochlea, he flattened out the spiral and replaced the hairs with graduated silicon beams. He modeled his cochlea as a linear orthotropic Kirchoff plate with interaction between the plate and fluid governed by a linearized Euler equation.
It could be years before Grosh's cochlear implants are in use, but his lab's next goal is a fully functional sensor version using integrated piezoelectric actuators attached to the beams driving external electronics. This artificial ear could replace "dumb" piezoelectric sensors used, for instance, by submarines, giving them wider bandwidth, more dynamic range and higher resolution.