Chip Manufacture A Dirty Problem

Lithography lets us see the features of a device we intend to build. However, the diffraction limit has pushed our optical systems towards shorter and shorter wavelengths that still exceed atomic dimensions. Using atomic force microscopes and scanning tunneling microscopes we can even scan a sharp tip over a sample to measure piezoelectrically the deflections of a cantilever or the electron tunneling current to measure the atomic topography.

If we limit manufacture of devices to planar processes, we vastly limit our ranges of motion for our deformable structures. If we can limit motion, a device's shape and behavior become less complex than those of macro-mechanical systems, and for less complex designs we may use photolithography. But, even here we get stuck. As features shrink, dust and other particles get magnified. As with bubbles in small tubes, at chip sizes a very small defect may disable an entire part.

To circumvent the need for ultra-cleanliness, we can design MEMS expecting that most of those we assemble will be faulty. But even so, we can still rewire hierarchies of components into complex modules after we test the components separately and together.

Looking closely at our MEMS devices shows us that devices have progressed from jointed rigid mechanisms towards compliant and deformable mechanisms, so in these respects, our machines may be more like bees. Now it should soon be possible to use functional specifications alone to create multiple systematic designs that range over rigid and compliant structures to support loads and also to control the transmission of force and motion. After we determine components and the displacements of parts within a device, we may then be able to make masks for lithographic reproduction almost automatically.

0 0

Post a comment