Creating a data-driven tool architecture


Click here to enlarge image

Larry Bourget, Axcelis Technologies Inc., Beverly MA USA; David Faulkner, Cimetrix Inc., Salt Lake City, UT USA

Data requirements for semiconductor manufacturing equipment have increased dramatically over the past decade. So much, in fact, that tool-control software purchased or built more than five years ago is probably already outdated. Equipment suppliers today are required to support more than a dozen SEMI standards related to factory automation and a host of commonly used substrate-handling components, such as robots and vacuum system hardware. Additional data-driven initiatives by other industry organizations — such as the International SEMATECH Manufacturing Initiative’s 300mm next-generation factory (NGF) program — also focus on the accessibility of high-quality data, using the data to improve efficiency and productivity, and enabling future analysis of large amounts of stored data. Unfortunately, most existing equipment-control software was developed prior to these new requirements.

The industry needs next-generation data-driven tool architectures to provide access to the data demanded. However, creating a data-driven tool architecture presents a couple of significant challenges. First, an adequate supply of high-frequency, high-quality data is required. Second, data must be gathered and presented to users in a way that is useful and easily and quickly understandable.

Axcelis and Cimetrix grappled with both of these challenges recently while undertaking an 18-month program to jointly develop a new tool-control software framework. The focus of the project was to improve the efficiency of 300mm substrate handling and factory automation in a new line of photoresist strip cluster tools.

To address the data-supply problem, the two companies opted to use the SEMI-specified equipment data acquisition (EDA) standards, also known as Interface A. These standards, designed to improve process monitoring and control, provide a virtual fire hose of nearly-real-time data for use by equipment and process engineers.

To date, only a handful of the largest chip makers have developed programs to make use of this new data. Equipment makers also have been cautious about adopting the new interface standards because the vast amount of data they generate can be exposed not only to customers, but also to anyone else the customer wishes to share it with, including competitors. That possibility can be avoided, however, by implementing a system that restricts suppliers’ access to only the data generated by their own tools.

All this new data creates a need for higher-performance data-analysis tools. Much work remains to be done on this front, although a variety of applications appear to be worth pursuing. By more closely monitoring process conditions, for instance, it should be possible to diagnose potential problems and schedule predictive preventive tool maintenance — before the problems become critical.

Over the longer term, we expect the industry to gradually evolve toward web services and XML to transport and store equipment data. However, especially during the current economic slowdown, manufacturers are proceeding slowly on adopting new standards. In the meantime, EDA and other existing data-driven standards offer significant opportunities for improved process control, productivity and efficiency, thanks to their inherent advantages in providing rapid data collection, analysis and control.

Developing new data-driven tool architectures is not an easy undertaking. However, as our recent collaboration suggests, it is achievable and can produce significant benefits.

Acknowledgments: SEMI is a registered trademark of Semiconductor Materials and Equipment International. ISMI is a servicemark of SEMATECH.

Larry Bourget is director of Integra Product Development for Axcelis Technologies Inc., Beverly, MA USA.

David Faulkner is EVP at Cimetrix, Inc., Salt Lake City, Utah USA;