Is a discontinuity ahead? Is a discontinuity ahead?

08/01/1998

Robert Haavind

Editor in Chief, [email protected]

Is a discontinuity ahead?

Growth industries often get on a steady track, with only incremental improvements. There may be ups and downs linked to the economy, and some competitors may grab up market share while weaker ones fall by the wayside or move into narrower niche markets, but the basic products and manufacturing methods change only gradually as the industry expands.

By contrast, sometimes industries move rapidly from one growth track onto an entirely new one - a discontinuity. Some examples from the past include diesel engines replacing steam locomotives on the railroads, or jet-powered aircraft replacing prop-driven models.

While progress has been incredibly swift in semiconductor manufacturing, it has been more like a steady growth industry than the discontinuity model. Is this about to change? This is a vital question for companies (and their bankers) putting down investments of over $1 billion for each new manufacturing plant, and spending hundreds of millions on new process tools. A healthy return-on-investment must continue long enough to amortize these huge outlays. So far these investments have paid off handsomely, with Moore`s Law providing some predictability. The steady shrinking of circuit features on silicon wafers allows more functionality at less cost. Myriad new and expanded applications for electronics have resulted, providing thriving markets. Even if a new technology or manufacturing approach appeared on the scene, it took a few years to displace its predecessor - such as CMOS gradually overtaking bipolar as the dominant device technology.

Should we be concerned about a discontinuity as progress accelerates? What could cause one in semiconductor manufacturing? There are some candidates, but in our view, none of them is likely to result in a sudden shift. Here are a few, with some comments.

 Copper and new dielectric materials take over. The transition from aluminum to copper and from SiO2 to other dielectrics, either low-k or high-k, is proceeding rapidly, with new tools and other parts of the infrastructure, such as CMP, adapting to new requirements. These new materials are only needed for the fastest, most high-performance circuits, so plenty of work will remain for existing tools.

 The demise of optical lithography. This has been predicted for so long that extensive work has gone into alternatives, such as x-ray, extreme UV, and electron beam methods. Replacing this fundamental technology (masks, resists, energy sources, etc.) causes many problems, and as a result chipmakers are pushing very hard to stretch photolithography beyond its "limits," using OPC, phase shift masks, off-axis illumination, etc. A gradual transition is more likely than a sudden shift.

 Silicon spheres. In this strange new approach to IC production, circuitry is deposited onto tiny polysilicon spheres (1 mm diameter). At Ball Semiconductor, in Allen, TX, this technology is in the prototype stage. Faster production and fuller use of crystalline silicon are among the advantages claimed. So much infrastructure would be needed, however, that it would take years to displace today`s flat chips even if the claims are valid.

 Quantum devices. If we could make very fast picocircuits using quantum effects with a cheap, reproducible process, it might quickly displace conventional microcircuit production. So far, most of these devices are two-terminal, which never caught on in the past (tunnel diodes, Josephson junctions, etc.). Transistor structures have proven elusive, and quantum effects tend to be highly sensitive to slight variations in fabrication. The promise is intriguing, but the delivery must await a few more breakthroughs.

A few other possibilities, including growing circuits using DNA, seem even more remote. No immediate discontinuity seems imminent, but in earlier cases, the best "experts" in the field were sometimes the biggest nay-sayers. So, keep an open mind and stay alert to intriguing ideas, even if they do seem strange. Meanwhile, the shrink goes on.