Since its inception, the semiconductor industry has successfully delivered a continuous cost reduction per device or device function, historically averaging about 35% per year. As a result, transistor cost (per million transistors), memory cost (per megabit), and computing cost (per MIPS) have steadily declined over the past 45 years, while on average the industry has enjoyed revenue growth with good profits. This phenomenal achievement, driven by Moore's Law, has been possible because of manufacturing cost reductions through shrinking device linewidths, improving yields, larger wafer sizes, and increasing productivity and manufacturing efficiencies. Costs have been reduced despite the rising tool and fab costs for the increasingly complex wafer-fabrication processes.

By virtue of doubling the number of components per unit area in each new technology node, shrinking linewidths have been the primary driver of this chip-manufacturing economics trend. This has been accompanied by the excellent progress made in process control and yield management. Mature die yields of more than 90% are regularly achieved in advanced high-volume production. Furthermore, the economic benefits of the transition to larger wafers have been leveraged periodically, a trend recurring with the 300-mm transition. Over the past two decades, tool and fab productivity have steadily improved. But relatively speaking, this area has not realized its full potential and still needs major improvements.

For instance, overall equipment effectiveness (OEE) and overall factory effectiveness (OFE) fall far below the theoretical maximums, with OEEs often less than 50%—a relatively huge efficiency loss. Furthermore, average cycle times exceed the minimum possible (typically about 3 to 10 times the minimum theoretical cycle times), and tool utilization factors often remain quite low. The industry has long recognized and proven that both equipment and fab cycle time and productivity can be improved through properly designed, well-implemented, targeted automation solutions. Other benefits of fab operations automation, such as improved contamination control and prevention, higher yields, improved delivery times, and better capacity utilization, are well established, understood, and documented.

After years of industrywide efforts, the time for migration toward cost-effective, fully automated fabs has arrived; it is an essential requirement for the success of 300-mm manufacturing. Automation addresses the challenges posed by stringent process control requirements for shrinking process windows, as well as the real need to gain the additional cost savings garnered by cycle time and productivity improvements. The International Technology Roadmap for Semiconductors (ITRS) has identified e-manufacturing and the fabwide (unified interbay and intrabay) fully automated materials handling system (AMHS) for product transport, flow, storage, and management as key factors in "realizing 300-mm conversion efficiencies." An AMHS is also required to handle relatively heavy front-opening unified pods (FOUPs), which can weigh as much as 25 pounds and may contain very expensive wafers.

An advanced e-manufacturing system encompasses equipment engineering capabilities, real-time advanced equipment/process control (AEC/APC) systems, e-diagnostics, and integrated yield management tools. The ITRS also calls for "the ability to track and run different recipes for each wafer within a carrier for operational flexibility." This capability is particularly crucial for fabs running a high product mix (such as foundries) and development pilot lines.

Combining the factory information and control systems with the fab's computer-integrated manufacturing (CIM) network is critical for an operationally successful implementation. This requires the integration of process tools and metrology and inspection equipment with e-manufacturing systems, AMHS, stockers, buffers, wafer sorters, reticle management systems, material tracking systems, work-in-process (WIP) management systems, material control systems (MCS), and data collection/storage/retrieval systems. The integration process also includes other factory software, such as the manufacturing execution system (MES), the factory operations system, and the planning and forecasting systems.

A 200-mm fab could spend from $50 million to $70 million for fab automation. For a 300-mm fab, this amount more than doubles, approaching the $130 million to $180 million range. Although the future growth of the nearly $2.4 billion fab automation market is primarily driven by the new 300-mm fabs (which account for about 8% of total equipment spending in a typical fab), existing or future 200-mm fabs can also benefit from the advantages of full fab automation and present an additional market growth opportunity.

Software plays an increasingly important role in automation. However, many fabs traditionally have had difficulty accepting and digesting the price tag and allocating sufficient investment for software solutions. This situation has started to change as fabs continue to examine and recognize the tremendous value of good software solutions.

A fully integrated automation solution would effectively unify the process and yield engineering aspects (dealing with product quality and yield) with its operational side (which is responsible for wafer moves and cycle time)—two parts of the fab that have sometimes been out of sync and at odds with each other. It would enable automated decision making in the management of preventive maintenance of tools and fab systems. Next-generation AMHS networks with distributed local buffers and direct tool-to-tool delivery directed by real-time MCS could lead to significant cycle-time reductions and increased factory throughput. This capability would result in shorter queue (wait) times—down to a few minutes from several hours—and require fewer stockers.

However, both the process equipment and fab layout must be designed and optimized to accommodate full fab automation requirements and effectively leverage its benefits. To further increase OEE and OFE, several key areas such as tool utilization, tool availability, and capacity loss caused by setup or tool qualification must be significantly enhanced.

With the introduction of third- and fourth-generation 300-mm tools, most experts agree that many important 300-mm problems have been solved. However, tool reliability—which is critical for the successful operation of fully automated fabs—remains a major concern. Although tool uptimes are generally fine, mean time between failures (MTBF) and particularly mean time between interrupts (MTBI, which has emerged as a critical issue) must be improved.

Integration is the most important challenge facing fabwide automation. It may take a long time to implement (sometimes more than a year) and can pose a relatively high degree of risk. Operational factors, scenarios, and events such as error recovery, resolution of production anomalies, maintaining lot integrity, lot dispositioning, hardware and software interfaces, and (real-time) scheduling issues all contribute to the difficulties of integration. To address this challenge, many large fabs have taken over the direct overall responsibility for integrating automation hardware and software systems. Recently, with further industry consolidation, some suppliers offer a broad product portfolio with major components needed for full fab automation, including hardware and software solutions as well as integration and support services.

Virtually all new high-volume 300-mm manufacturing fab designs have included a unified interbay and intrabay AMHS, which generally functions well. However, current 300-mm fabs are far from being highly integrated and fully automated, and have yet to realize the full potential of the productivity enhancements promised by automation. Fault detection and correction, e-diagnostics, and other elements of e-manufacturing remain in their infancy. AEC/APC solutions and industry standards are evolving, with many chip manufacturers developing in-house solutions. In addition, some operations continue to be performed manually.

Certain situations (such as an out-of-control status) involve a decision-making process that may require human intervention. Therefore, a practical approach would factor in the fab's operational realities and the maturity levels of various automation solutions. An evolutionary strategy featuring a suitable hybrid mix of automation with limited but essential and well-managed manual operations would be more successful. As more-mature and proven automated solutions for needed but missing applications come on-line, they can be integrated into the fabwide automation system, driving down the levels of manual intervention and operations.

The elimination of people from fabs and realization of the "lights-out fab" is impractical, unrealistic, and unachievable. In fact, the goal itself is unnecessary and can be quite misleading. Regardless of the degree of automation, humans will always be needed to oversee, manage, and address fab operations, processes, and events. In other words, the fab lights will be on for a long time to come. Although more-automated fabs will need fewer operators, engineers and technicians with a diverse mix of skill sets, including computer science and industrial engineering, will still be required. The real goal should be to achieve the shortest cycle times and the highest levels of productivity and manufacturing efficiencies possible in a cost-effective manner with the right amount of automation that is operationally sound, practical, and risk-free.

The fully automated fabs of the future will be quite different from what we know now. They will deliver unprecedented levels of productivity, with great cycle times and vastly improved process control capabilities. The industry must participate to help bring the vision of an integrated and fully automated fab to fruition as soon as is feasible.