Major Semiconductor Innovations




Point contact transistor

Bell Labs-Western Electric


Single-crystal growing

Western Electric


Zone refining

Western Electric


Grown junction transistor

Western Electric


Silicon junction transistor

Texas Instruments


Oxide masking and diffusion

Western Electric


Planar transistor and process



Integrated circuit

Texas Instruments, Fairchild


Gunn diode



Ion implantation



Plasma processing



E-beam technology




SOURCE: John Tilton, Brookings Institution, Washington, D.C.

technology has several interesting characteristics. The interconnect is still principally aluminum, although it is now alloyed with copper and titanium. The interlayer dielectric is still silicon dioxide. You will note that this plasma-enhanced chemical vapor deposition of silicon dioxide leaves very uneven layers. There is a complex metallurgy in the vias that consists of titanium and titanium nitride with tungsten as the principal conductor.

Today, the semiconductor industry is roughly a $200 billion-per-year industry; $150 billion of this is in the sale of semiconductor devices, roughly $30 billion in the sale of processing equipment, and approximately $20 billion in the sale of manufacturing materials, including silicon, mask blanks, photoresists, and production gases and liquids (Table 4.2). In the 1990s, the industry has grown at an extremely rapid rate, averaging more than 30 percent per year over the past three years. The industry is projected to grow at an average rate of about 20 percent per year for the next 15 years.

If we compare the growth rate of the semiconductor industry (about 20 percent per year) with the gross domestic product (about 2 percent per year), by the year 2019 semiconductor sales will be equal to the U.S. gross domestic product of $11 trillion. This leads me to prediction 1: The increase in semiconductor sales will flatten before the year 2019.

The U.S. semiconductor industry has developed a road-mapping process that looks at technology needs for the industry 15 years into the future if costs are to continue to decline. An example of the content of this road map is shown in Table 4.3. The road map considers memory, high-volume logic, and low-volume logic. The year represents the first shipment of a product with the given technology. A major consideration of the road map is cost. The cost per transistor for each of these products is predicted to decline with each generation of technology. This leads to prediction 2: There is no physical barrier to the transistor effect in silicon being the principal element in the semiconductor industry to the year 2010.

Although the road map projects technology generations for 15 years in the future and the technology required for that projection, it does not define solutions for all of the technology requirements. Let us look at some of the challenges that must be addressed to meet the road map projections to the year 2010.

Will it be possible to design 100 million transistor logic circuits and 70 to 75 billion transistors on memory chips? Figure 4.2 describes what the semiconductor community has called the design productivity crisis. The upper curve shows the compound complexity growth rate in silicon integrated circuits at roughly 60 percent per year. The lower curve shows the compound design productivity growth at about 20 percent per year. This leaves a considerable gap in 1995 and a growing gap into the twenty-first century, which is the consensus of a group of design experts in industry, government, and universities. There is no complete solution to this design dilemma. There is cooperative work on building an open infrastructure that will allow interoperability of commercially available design tools and of those tools that are developed as competitive capabilities in integrated circuit

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement