Compound and silicon semiconductors: A technology duet

03/01/2002

William J. Kroll,
Matheson Tri-Gas Corp.

While silicon CMOS devices are by far predominant in today's electronic systems, special needs, especially for applications in the rf and optical sectors, are spurring the growth of compound semiconductors and silicon germanium technology. We believe this trend will continue, broadening the partnership between complementary technologies in the future. The roots of this latest trend go back to the early days of the semiconductor industry, with some fits and starts along the way. Tracing this history will help show how the partnership is growing stronger.

Since the industry's inception, semiconductor markets have experienced cycles of rapid and then slower growth. Overlaid on this market cycle, but less noted, have been shifts in basic technology. In the early days, Group IV materials were used for devices, starting with germanium (Ge) and then evolving toward robust enabling designs using silicon (Si). In those times, the industry became totally dependent on the indirect bandgap technology of silicon.

Then, in the 1960s, the first nonsilicon branch of the family using Group III-V materials was spawned, and the industry would never be the same again. These direct bandgap devices based on compound semiconductors had the ability to emit light, and when operated as electronic circuitry, they offered much higher mobilities than silicon. This allowed them to achieve higher-frequency operation than had ever been thought possible for solid-state devices, making higher-bandwidth end products a reality. Compound semiconductor companies sprouted overnight. Gallium Garden in New Jersey was born with the promise that this new branch of semiconductors would eventually replace silicon because they could be used for higher-speed microprocessors and could also be made to emit light for applications like digital displays in watches.

Two basic problems with this scenario soon became evident. First, silicon semiconductor vendors were not ready to abandon their installed capital. They were able to use process ingenuity to defend their current and future markets. Second, the silicon infrastructure was much stronger. Compound semiconductor tools and materials supplies were embryonic, offering substandard resources and capabilities compared to their silicon counterparts. Only tiny wafers could be processed with those early tools, so it was difficult to make cost-competitive devices with gallium arsenide. To make things worse, the chemicals needed often weren't available in the necessary quantities and purity levels. As a result, GaAs devices were only used in a limited number of high-performance niche applications where price was no object. Military electronics proved to be the only killer application for them.

Compound semiconductors became the perennial "technology of tomorrow" ("always have been, always will be," as the saying goes). They were the answer waiting for a question. Then, in the 1990s, a new array of killer apps emerged in telecommunications, data communication, and entertainment. The hope was that all these needs could be met by one type of III-V compound using the same processes to produce both optoelectronic devices and high-speed transceivers. Demand grew for devices complementary to silicon, focusing the efforts of compound semiconductor technologists on development of processes, materials, and wafers that would enable the economical production of specialized devices for mass markets. The potential market demand for compound semiconductor devices, especially to provide rf circuits for portable devices like cell phones and to serve high-bandwidth optical networks, drove tool and materials suppliers to adopt the same best practices defined in the silicon marketplace.

All this has led to a much better infrastructure. Today's norms are 150mm GaAs wafers and robust and serviceable processing tools adopted from the silicon side. This has raised yields, productivity, and throughput for GaAs chip manufacturing. Wafer growth techniques have improved measurably, and 200mm GaAs wafers are on the horizon. Larger wafer sizes led to bigger "sweet spots," as they did years ago for silicon, helping to dramatically improve yield and throughput. Older 150mm silicon fabs considered obsolete in the age of 200/300mm silicon processing have been turned into fully depreciated, state-of-the-art GaAs fabs. Lessons learned in silicon are driving socket wins for compound semiconductor device makers. Special properties have also led to commercial uses for other types of III-V and II-VI devices, such as gallium nitride and indium antimonide, although on a much smaller scale.

Surprise! Now compound semiconductors are beginning to return the favor by making contributions to the silicon chipmakers. Compound semiconductor processing depends on one tool, the MOCVD/epitaxial technology, which takes a very different approach by depositing the entire device structure, than an additive-processing approach with more than 25 tools. Compound semiconductor technologists learned to grow precise films with atomic layer epitaxy using metallorganic compounds aided by in situ monitors and real emissivity-corrected temperature control. That knowledge provided some enabling technology to their silicon technology counterparts when they needed it most. These techniques are required today with the very thin CVD or epitaxial layers made with the new materials finding their way onto silicon wafers.

Convergence of the technologies to solve emerging problems is increasing. Light can transmit signals at higher speeds than electrons, so for faster circuitry, we may see a merger of lasers and detectors from compound semiconductors with SiO2 optical wave-guides and silicon-based logic. Compound semiconductors have gotten so many design wins for HBT power amplifiers that silicon chipmakers have taken notice and developed their own "compound semiconductor" to compete. Enter SiGe. We now have a IV-IV compound device in the silicon world with epitaxy needed to grow the device structure. SiGe is finding its way into mainstream CMOS device processing. Higher-speed silicon devices combining CMOS and SiGe are made possible through the strained lattice structural techniques practiced in III-V devices. The use of a base epi layer of SiGe (a buffer layer in III-V parlance) prior to Si epi provides stresses that lead to an increased lattice constant for the Si epitaxial layer. This gives higher electron mobility.

Lattice matching is also being explored by chipmakers such as Motorola to allow gallium arsenide and other compound semiconductor devices to be grown right on top of silicon, bringing both technologies together on the same wafers.

Where we go next depends on the evolving technologies of both semiconductor families. While compound semiconductors are roughly only 8% of the devices sold today, this is projected to grow to more than 10% in the next 4-5 years. Broadband applications and new types of solid-state lighting are two killer apps leading the way toward a projected 30-35% CAGR for compound semiconductors.

Where is the cooperation between the different technologies leading? Growth of our industry in the future will depend on our ability to design for energy efficiency while providing for a proliferation of real-time information. Complementary solid-state devices hold promise to help meet those goals effectively. An examination of the past decade shows how advances in compound semiconductor devices can actually help further the growth of more traditional silicon devices, expanding the range of applications for both technologies. The convergence is likely to grow even stronger, to the benefit of the whole industry.

William J. Kroll is president and CEO of Matheson Tri-Gas Corp., 959 Rte. 46 East, Parsippany, NJ 07054; ph 973/257-1100 ext. 334, e-mail [email protected].