In the Materials Age, What Happens to Economy of Scale?

The semiconductor industry is looking to a variety of new materials to help boost chip performance and reduce power consumption and leakage. The push to copper and low-k dielectrics is well documented, and there's considerable interest in high-k materials for gate dielectrics and capacitors, metal gates, strained silicon, silicon-on-insulator (SOI), electroless CoWP (to cap copper layers) and even more exotic techniques such as germanium-on-insulator (GeOI).

The rapid pace of introduction of new materials is startling compared with what the industry has done in the past. In the 1970s, the materials used by the semiconductor industry could be counted on one hand. By the end of 2010, some experts believe more than 35 different materials will be in use.

Most of these materials have been introduced fairly recently. Going from the 1 µm process technology used in 1986 to 0.5 µm processing in 1992 required no material changes — only a switch from g-line to i-line lithography. The Materials Age really started in 1999 with the 0.18 µm generation, which saw a widespread switch from aluminum to copper. This was followed with a push to lower-k dielectrics, such as SiOC, which occurred in 2003. Along with these changes came others in the silicide, metal capping and dielectric barrier areas. 45 nm generation front-end processing could well employ hafnium-based high-k gate dielectrics, metal gates, fully depleted SOI and strained silicon.

There are, of course, many technical challenges to be overcome before these and other new materials can be deemed production-worthy. One of the goals of the ITRS is to allow semiconductor manufacturers and their suppliers to evaluate various options and narrow those options over time. Although the ITRS doesn't attempt to factor in the cost of developing new technologies, it's basically assumed that it will be too expensive to develop in parallel three or four options. Picking one or two options early on is essential so that equipment and material suppliers can develop commercially available solutions by the time they are needed for production.

What I question is what happens to the economy of scale that the industry has enjoyed to date. Economy of scale is basic economics, formally defined as a reduction in cost per unit resulting from increased production, realized through operational efficiencies. Economies of scale can be accomplished because, as production increases, the cost of producing each additional unit falls. This is true within a company and across the semiconductor industry.

By defining "pre-competitive" areas, through the ITRS and through consortia and joint-development programs, companies are in fact standardizing and commoditizing many aspects of manufacturing. This helps enable good economies of scale, reduces risk, and gives equipment and material suppliers a clear goal, enabling them to commit the resources necessary to develop a particular technology. That's the good news.

The bad news is that the increased complexity of the semiconductor manufacturing process — and the greater number of materials required — is going to make it more difficult to achieve the same overall economy of scale the industry enjoyed to this point by sticking to silicon and SiO₂.

It is probably impossible to calculate exactly how much this will cost the industry. Most likely, it will show up in increased equipment and material costs, as suppliers are forced to pass their costs on to their customers. Or it could be that more process steps and more equipment will be required and fabs will have to be bigger.

Alternatively, it might turn out that some kinds of materials prove just too costly for anyone to develop. Millions if not billions have already been spent on low-k dielectrics without the kind of success that had been originally planned and hoped for (the new ITRS pushes back low-k by more than three years). This would make it the Materials Age That Almost Was But Wasn't Because It Cost Too Much.