Chemical Management in the Fab of the Future
Marc van den Berg, Jay Jung, Joe Curcio Microbar Inc., Sunnyvale, Calif. -- Semiconductor International, 8/1/2000
 |
At a Glance | | |
|
| ||
| |||
Generally speaking, process control suggests by
its name that a process can be implemented to turn measurements into
information, which can be used in turn for decision-making. Decision-making can
be executed in the form of a local PID loop for the control of temperature in a
process chamber, all the way up to a fab manager's decision to treat a CMP
wastewater stream instead of discharging to POTW due to the conclusions he or
she draws from the information on water mass balancing received from the fab's
water usage database. Fundamentally, there is no difference between the two
decision loops, except for the level of instrumentation required to control the
process and the path the information needs to take. Either the information
requires human involvement, or it doesn't.
Historical approaches
Semiconductor tool manufacturers are required, due to the complex nature of their equipment, to spend an inordinate amount of time closing the control loops in their toolsets in order to optimize wafer performance and throughput. Tool-to-tool communication often is implemented as a means of supporting this requirement. Auxiliary equipment tool suppliers target their control technology similarly — usually following the lead of the connected larger tool to establish reliable communication in order to support the performance objectives. Early communication between tools was attempted through the implementation of standardized communication protocol standards such as SECS or GEMS. Today's electronic hardware for embedded controllers as well as tool controllers offers very affordable industrial networking options.
Current process control systems include, among other things, a data collection computer (also known as a data concentrator) to collect, store and process the vast array of information from the equipment regarding the equipment's real-time status. The data concentrator also may act as the pathway through which an operator and/or engineer can monitor the status of the manufacturing process. The equipment manufacturer may have remote access to the data concentrator to assist the customer with the decision-making processes. In most cases, this is the most economical approach since the equipment manufacturer can provide a wealth of both process and equipment knowledge.
The semiconductor fab environment is a collection of many different tool sets from many suppliers, each with a specific set of local automation requirements. In the database world, a significant number of enterprises have emerged to create middle-ware software that will allow the multiple databases to speak with each other. During the '80s and '90s, huge MIS departments grew to address this dilemma. In accordance with Moore's Law, the semiconductor manufacturing environment has witnessed a constant change of technologies and associated tool sets at an alarming pace, such that it is nearly impossible to devote the necessary resources to true factory process control. In response, the industry sought to identify and publish standards to ensure alignment to the factory host model such that each tool supplier could create a tool set front-end and communication capability.
Future
A local-level "middle-ware" solutions provider can provide the tool interconnect, the data concentration and information presentation to the ultimate fab host. In addition to real-time data monitoring and archiving, this can include materials allocation and flow control, remote problem diagnostics and resolution (up to and including quality control and root cause analysis), toxic monitoring and life safety interface, and "smart tool" coordination. All associated information and interface can be available either via direct connection or through secured Internet access. Outside the semiconductor world, this practice is quite common. Throughout the 1980s, other manufacturing environments adopted this approach and began utilizing supervisory control and data acquisition (SCADA) packages to enable interconnectivity and information sharing. A middle-ware solution can be achieved through the implementation of a SCADA approach in both hardware and software to allow its interconnected tool set to measure, modify, certify and share its collected information to the host of choice. This approach can then address the aforementioned demands in a modular and, therefore, site-specific manner.
An example is shown in the Figure. The control system platform includes TCP/IP Ethernet communication capability for each of its toolsets. The figure demonstrates a distribution and collection/treatment of wet chemistries that, when combined with the middle-ware provider, can consolidate measurement of critical information for usage, mass balance, uptime, environmental concerns and inventory concerns. By utilizing the multi-language capabilities of its hardware platform and a connected SCADA system, the consolidated information can be sent, displayed and shared with fab-wide ERP systems. The technology exists to automate business decisions that today require human intervention. Therefore, as an example, procurement of chemicals directly from the tool in need to the associated supplier is a matter of desire not technology.
Microbar Inc. (Sunnyvale, Calif.), for example, offers a solution called ChemManager. It can reside anywhere that makes the most sense for the information-gathering, analysis, display and reporting requirements. ChemManager is a platform-independent data collection and user interface program tailored to present information to fab hosts by collecting data from equipment toolsets of any origin. The types of information that are typical include:
• Histograms of equipment uptime and performance.
• Maintenance and service logs for reference and predictive maintenance.
• Security for chemical validation by process engineering.
• Mass balancing of chemical distribution loops for environmental and operational expense measurements.
• Automated customer support.
• Automated procurement link for consumables.
• Internet tracking of process data and equipment performance data from remote sites.
• Virtual database and user interface access through the Internet.
The Figure graphically describes the ChemManager implementation as a middle-ware package. The green icons represent tool sets for a given manufacturing environment as well as the middle-ware provider's interconnect TCP/IP lines. The bright blue icons represent ChemManager. Although shown here they physically can reside on the fab floor, in the office of a fab manager or MIS department, or at the provider accessible via the Internet. The yellow icons, again location-independent, represent the use of the system's database by another end user. That could again be the fab manager or a technical support department of a fab or equipment company. Notice that the database is virtual, and useful information is simply an application away. Naturally, security systems are important to address local as well as remote-access security and are integrated into the package.
Tool-to-tool communication hardware strategies need to model themselves directly from desktop to server-based interconnect strategies. Many suppliers' tools employ Ethernet TCP/IP hardware and protocol capable of sending electronic messages in a form that travels across the worldwide web as easily as an e-mail. Tool-to-tool communication handshaking for safety and certain process interlocks will be with us for quite some time. Internet reliability has improved such that the suppliers of both the consumer and industrial tool sets have the confidence to send and receive status and data information over the Internet. In combination with readily available software packages, data easily can be collected, analyzed, reported and acted upon from a collection of different tool sets in such a way that the response can be broad but process-appropriate. The SECS/GEMS model quickly will become extinct.
Implementation
Many forward-thinking companies outside the semiconductor industry have integrated enterprise systems and factory floor equipment for factorywide automation of manufacturing and overall business operation. The supervisory control and data acquisition system typically is used to establish communication with factory floor equipment. Add-on modules for SCADA provide recipe (batch processing), inventory management, maintenance management and other functionality to aid in manufacturing execution. A database such as Microsoft SQL Server can bridge the gap between enterprise systems and factory floor equipment information.
The implementation begins with communication between each equipment unit and the SCADA system. Semiconductor equipment manufacturers in theory have standardized on SECS/GEM; but implementation of this standard is limited. Many semiconductor equipment manufacturers have utilized programmable logic controller for controls, which is the norm outside the semiconductor industry. However, it is both difficult and costly to embed SECS communication in the programmable logic controller. Nonetheless, SECS/GEM capability is necessary for successful SCADA implementation. In addition, the capability to communicate with various programmable logic controller brands and models is invaluable.
The typical "cell controller" in a fab includes SECS/GEM capability and targets SCADA functionality. "Off-the-shelf" SCADA packages are used exclusively and successfully as a "cell controller" outside the semiconductor industry.
A typical SCADA package out of the box is equivalent to an AutoCAD package, as it provides tools but does not automatically guarantee successful implementation. The key ingredient for successful implementation in the semiconductor industry is a system integrator who understands the needs of the semiconductor industry and equipment set. Although implementation tools and software infrastructure can be purchased, selection must be done wisely, as there are a variety of available packages. The focus should be on fulfilling the requirements, rather than squandering 70% of effort on developing tools and infrastructure.
The "supervisory control" part of the SCADA term indicates there is a path to command the equipment how to produce the product. The "data acquisition" part of the term indicates there will be a huge quantity of available data from the equipment. During the 1980s and 1990s, the semiconductor industry invested in business automation systems such as enterprise resource planning (ERP), manufacturing resource planning and manufacturing execution, so that order processing, production scheduling and other activities have been largely automated. Fab floor information must be available to the enterprise system in order to fully realize the benefit (increased productivity).
Enterprise system experts typically fail to understand semiconductor fab equipment. Conversely, equipment manufacturers fail to understand enterprise systems. Integration with an enterprise system is much easier when the equipment/process data is placed in a database. Since an enterprise system is essentially a database, enterprise vendors understand database implementation. Cooperation between equipment automation suppliers and ERP system providers can be promoted and channeled by locating the fab floor information in a database.
Significant benefits can be realized with SCADA and database implementations. Current and historic information is readily available for utilization. The appropriate maintenance personnel will be notified at the first sign of trouble. Once a problem is solved, a report can be archived. If the problem recurs, the report can be retrieved to aid resolution. In the case of a high frequency of occurrences, further analysis may be ordered to identify the root cause. The information for analysis — such as process data, operator interaction, system sequence, maintenance history, recipe and alarm history — will be available.
Typical ERP systems use historic data to forecast future activity. However, algorithms require empirical data to be applicable. Upon implementation, fully integrated enterprise systems will be able to predict and modify equipment performance and be fully effective.
Conclusion
The semiconductor manufacturing industry is ripe for adoption of automation and control technology commonly available outside the industry. Previous efforts have been characterized by large expenditures with individual equipment and facility providers each supplying a local and independent solution associated with their product area, with little focus on the overall integration of the data, analysis, monitoring and control. Third-party integration has been difficult as each supplier has delivered myopic solutions using proprietary software and protocols. The problem was exacerbated only with the advance of technology and associated upgrades of equipment and facilities. A middle-ware provider can integrate the functional areas (process, facility, life safety, administration, materials management, engineering, sales and others) to lower costs, improve productivity and increase value to the end user. •
Marc van den Berg is vice president of engineering for Microbar Inc., responsible for directing the company's research and development efforts. Prior to joining Microbar in 1995, he was director of engineering for Purus Inc., a San Jose-based supplier of capital equipment.Jay Jung is control engineering manager at Microbar Inc., where his responsibilities include managing electrical and software engineering personnel, as well as developing strategies for control system design. He has experience in building capital equipment.
Joe Curcio, director of marketing for Microbar Inc., joined the company in June 2000, and is responsible for product marketing and marketing communications. His career includes senior management positions with Air Products and Chemicals, and most recently at Matheson Tri-Gas, where he was director of marketing.
