Sublithographic (Molecular-Scale) Architecture

Lithographically etched silicon has been our primary substrate for the past 20 years, and it looks like it will continue to serve this role for, at least, the next 10 years. What happens beyond that?

We are rapidly approaching the molecular scale, and it's clear many of the features we've relied on for bulk VLSI will no longer work at this scale. One alternative is to turn to molecular-scale phenomena and electronics itself and see how to build computations directly from these building blocks. This will, undoubtably, result in a very different cost structure than we've seen in lithographic silicon. Consequently, we may need to find very different architectures to exploit this new medium.

Shown here is a design which exploit the structures (e.g. self-assembled nanowire arrays and non-volatile diode crosspoints) which it looks possible to build in the next few years. Here, the crossbar structure gives us the computing blocks (in the form of PLAs) and the interconnect (small crossbar blocks), on a single, master, segmented, crossed-wire array. Molecular arrays are likely to require significant regularity so they can be built hierarchically in bulk, and significant defect-tolerance to handle inevitable defects during construction. Post-fabrication repair and personalization will be essential for this class of devices.

• Article describing how we can build programmable systems from nanowire building blocks: Nanowire-Based Programmable Architectures
• Article describing how these, and other nanoscale technologies, can tolerate high defect rates (5-10% defects in individual wires and switches): Seven Strategies for Tolerating Highly Defective Fabrication

For a more complete list, see: André roundup of papers on sublithographic architecture.