"MEMS [devices] are limited to one or two degrees of freedom today because 3-D devices are so difficult to assemble, but we believe that our magnetic actuator can self-assemble 3-D devices in a production environment," said University of Illinois professor Chang Liu. "This would be a perfect addition to the process Sandia National Labs is licensing to foundries."

On-chip optical waveguide self-assembles by magnetically attracting first the metal bars on each side of bull's-eye, then those on the smaller locking flaps.

Essentially, MEMS development is based on the photolithography technologies used in conventional chip fabrication assembly lines to etch out silicon parts for tiny devices. Surface-assembled devices, such as the MEMS airbag sensor found in vehicles in the United States, merely etch free a part with one degree of freedom by surrounding it with two sacrificial layers that are later etched away. By laying down the first sacrificial layer in the pattern of the "bottom side" of the device and the top sacrificial layer in the pattern of the "top side," the two can be etched away to leave just the desired silicon microstructure.

A device with two degrees of freedom can also be made in this way, as long as the two directions in which it will move are within the plane of the chip. But the technology for enabling devices to protrude upward into the third dimension from the substrate has been troubling.

"Other people try to use water flow to make some of the 3-D structures pop up, but it is a pretty random process, with nothing like the 100 percent yields we get," said Liu. "Others have built individually addressable electronic actuators on the chip to make parts pop up, but [the actuators] need to be bigger than the part you are trying to elevate."

The method devised by Liu's team works across the whole chip, with 100 percent yields, by merely applying a linearly increasing magnetic field to the entire substrate after fabrication. The slowly increasing magnetic field causes the 3-D parts to pop up out of the plane of the substrate in the proper order, so that each locks into place before the next pops up.

Basically, the magnetic actuation process starts with the same sort of MEMS structures that other institutions, such as Sandia National Labs, are making available for licensing to conventional foundries. However, instead of five layers, as in the Sandia process, Liu needed just four layers for his prototypes. Another "story" to his current single-story 3-D devices could be added with further layers, but initial prototypes used just four.

Freewheeling hinge

The magnetic actuation process depends on a hinge that resembles the hinge on a regular macro-sized door. One part of the hinge is left fastened down to the substrate's surface like the part of a door hinge that is fastened to the wall, while the other half of the hinge is left free to pop up vertically into the third dimension. Two sacrificial layers were needed, in addition to the two layers for each side of the hinge. The two sacrificial layers were then etched away from the top and bottom of the freewheeling hinge half.

Thus far, the process is nearly identical to Sandia's but adds a layer of metallization atop the freewheeling half of the hinge. Liu's photolithography mask makes sure that the right amount of metal gets deposited on each of the freewheeling parts of a 3-D MEMS, just before the sacrificial layers are etched away. After etching, the three-dimensional parts remain folded down into the surface of the substrate. But by applying a magnetic field to the bottom of the chip, each part is repulsed upward, until each pops up in the correct order to self-assemble a 3-D MEMS — such as a microprobe array, fields of optical waveguides, 3-D accelerometers or even on-chip gyroscopes.

"We pretty much use the Sandia Labs process, which was invented at the University of California at Berkeley almost 10 years ago," said Liu. "First we put down a sacrificial layer, then a structural layer for half the hinge, then a sacrificial layer, then another structural layer for the other half of the hinge. Our addition is that we add metal to the half of the hinge we want to pop up."

For devices that have just a single part type to pop up, a permanent magnet can be used to make all the parts pop up simultaneously. But for devices that have parts that must assemble in order, a linearly increasing magnetic field can be applied. In that setup the part with the most metal for its weight will pop up first, followed by the part with the second most metal for its weight and so forth, until the entire 3-D structure self-assembles and locks into place.

Liu plans to build complete 3-D sensors-on-chip using multiple freewheeling parts that snap together like Legos. Another project, in cooperation with chemists at Northwestern University, aims to create nanometer-scale pens that dispense resist directly onto the surface of a chip during fabrication, resulting in submicron lines that cannot be written as small with conventional lithography.