Efab enables MEMS to be forged at the desk

Efab enables MEMS to be forged at the desk
LOS ANGELES — Physicist Adam Cohen is leading a team at the University of Southern California's Information Science Institute that seeks to easily produce micromachined parts from a desktop. The team has built a prototype micromachine fabrication unit that's small enough to fit on a desk, and which can reproduce complex shapes at micron dimensions.

The prototype desktop fab has so far created such test parts as a 12-layer microchain with 14 independently movable links, monolithically fabricated in nickel without the need for assembly. The potential complexity of parts produced by the process could go far beyond what is possible with state-of-the-art micromachine processes, Cohen said.

"We're trying to create the milling machine or the lathe of the micro world," he said. To describe the process, Cohen has coined the term "Efab," short for "electrochemical fabrication."

The USC group is now working on a fully automated version that would take CAD output from a computer and mass-produce 3-D devices with thousands of layers. The unit would exist as a self-contained turnkey machine that runs on electrical power and compressed air, and sits on a desk.

Cohen came up with Efab because he believed microfabrication processes were too limiting. "There had to be a better way," he said. Cohen blended his knowledge of rapid prototyping, also known as "solid free-form fabrication," with his understanding of semiconductor fabrication and applied it to MEMS technology.

Rapid prototyping has become a standard commercial process that's used in many industries to quickly generate complex models and prototypes on the macro scale. The technique is used to make prototypes for automotive parts, jet engines, medical devices and appliances, to name a few. Using rapid prototyping, designers go from 3-D CAD to the device being built automatically in a machine that acts like a 3-D printer, called an SFF machine.

Cohen wanted to take the same principle and create an inexpensive way to make MEMS devices, without a clean room and without limitations like surface micromachining and the Liga deep X-ray lithography technique. One problem with MEMS is the lack of a standard process for every design, Cohen said. Instead, each design needs a custom process — for instance, Texas Instruments Inc. created a custom process for its digital micromirror MEMS devices.

The MEMS technique of surface micromachining — depositing thin layers of material on silicon wafers — is limiting, only allowing the creation of up to five cross-sections. The process also requires an expensive clean room, high-temperature processing (which can damage the integrated circuit) and a high skill level that requires researchers with PhDs, Cohen said. And MEMS processes can take a long time — up to eight to 10 weeks — increasing time-to-market for products using MEMS devices.

Efab removes these limitations, he said, and creates a way to produce thousands of layers of arbitrary 3-D shapes hundreds of microns tall on a chip. Materials are deposited by electroplating, using a new selective process that Cohen and his team call "instant masking."

The mask is patterned onto a conducting anode plate using an electrically insulating material. A series of such masks, each of which represents a thin cross-section through the part being built, can be fabricated in a separate process before the formation of the part.

The anode plate with its mask is simply inverted and pressed against the substrate, and the assembly is then immersed in an electrochemical bath. The prototype uses nickel as the first plating material. After the nickel plating, the mask is removed, and the substrate with its nickel pattern is copper-plated, filling in areas left by the insulator. The two materials are then polished flat, providing a smooth layer for the next mask.

The process can be repeated indefinitely, which creates the potential for defining arbitrarily complex shapes. Once the final plating has taken place, the copper is removed by chemical etching, leaving a three-dimensional nickel structure.

Using electrodeposition to define the structure has several advantages over conventional semiconductor methods, such as chemical vapor deposition. The process can be done at close to room temperature and does not need a clean room. In addition, the process is far simpler than semiconductor masking, and the fact that the masks themselves can be made prior to electrochemical definition has allowed the USC researchers to shrink the Efab to desktop proportions.

Efab is faster than most MEMS processes. In the eight weeks it would take most MEMS processes to develop five layers, Cohen said, Efab could build 1,500 layers. While it takes a few days to make one layer using traditional MEMS methods, Efab could take 40 minutes, Cohen said. In 24 hours, Efab can produce 30 to 35 layers. Cohen said he and his team believe Efab could be a standard process for MEMS, with one caveat: "We cannot deposit insulating materials and some MEMS require this."