require dramatic improvements in the anisotropy, selectivity, and uniformity of etching; improved planarization and conformality in deposition; new materials to meet device performance and reliability requirements; and reduced process damage and contamination. To meet these challenges, new processes and new reactors are needed.

Plasma processes in use today have been developed using a combination of intuition, empiricism, and statistical optimization. Design tools such as numerical simulation codes or even expert systems do not exist for plasma processes because of fundamental gaps in understanding. With the unprecedented demands now being placed on plasma processes, it is unlikely that the traditional approaches used in process development will continue to satisfy our needs. Tools are needed to relate process variables to wafer attributes, modify existing processes, and design new processes.

Plasma reactor design is intimately intertwined with plasma processes, but we again lack computer-aided design tools for new plasma reactor design. We are also unable to transfer processes from one plasma reactor to another or to scale processes from a small to a large plasma reactor. Until we understand how geometry and electromagnetic design affect material properties, the choice of reactor for a given process will not be obvious. Without physical understanding leading to computer-aided design tools, only costly empirical, trial-and-error experimentation will be available to provide the necessary guidance.

The United States has lost the lead in new plasma reactor and process development, and the health of the American electronics industry is seriously threatened as a consequence. One reason for this precarious situation is the relationships between chip manufacturers and their processing equipment suppliers. In Japan, most equipment development occurs by collaboration between chip manufacturers and equipment vendors. In the United States, on the other hand, semiconductor processing equipment suppliers are small and typically unaligned with chip manufacturers. As a result, critical process information is not efficiently fed back to equipment builders, and R&D funding for next-generation processing equipment is strained.

As custom-designed, custom-manufactured chips (application-specific integrated circuits, ASICs) gain a larger market share, the future microelectronics market will no longer be dominated by memory chips. But customization at low cost means that the ASIC manufacturer cannot afford to invest heavily in new processes and new processing equipment each time a new order is placed. The future ASIC market will belong to the flexible manufacturer who uses a common set of processes and equipment to manufacture many different circuit designs. Such flexibility in processing will result only from real understanding of processes and reactors.

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