Flexibility, Interoperability, Contamination Control among Keys to Highly Automated Factories

No technological development has pushed the state of fab automation and integration as hard and as far as the transition to 300-mm chip manufacturing. Manual operation and handling have little or no place in the realm of the big wafer. State-of-the-art process tools and fabs include heretofore-unseen levels of automated intrabay and interbay wafer-handling equipment, factory control systems, and other assorted software. Some pundits have gone so far as to forecast that the long-sought-after "lights-out fab" is just around the corner, but those working up close and personal with the new robotic gear and exotic software know that the industry still has a long way to go. In this installment of The Hot Button, we asked the experts for their take on the critical issues facing fab automation and integration, and their answers may surprise you.

BALA SUBRAMANIAM (manager, component automation systems, Intel): Device makers have invested heavily in automating material handling in 300-mm factories. Stockers used for storage of FOUPs are the single largest component of factory automation cost, contributing about 25% of the cost category that includes automated material-handling systems (AMHS), data automation, and support headcount. A combination of stockers and carousels —both interfaced to overhead transports (OHTs) using existing industry standards—may be a viable solution to lowering cost without sacrificing system-level performance.

Carousels could become the next disruptive technology for conventional stockers for 300-mm automated material handling. —Bala Subramaniam

Carousels have lower cycle-time performance and lower cleanliness, and potentially subject front-opening unified pods (FOUPs) to higher levels of vibration. However carousels are substantially less expensive than stockers: budgetary estimates have been 50% of stocker cost per FOUP stored. While FOUPs holding product wafers need to be moved at the highest possible velocity to and from equipment, there are nonproduction wafer-carrying pods (numbers could be as high as 50%) that are sedentary or slow moving by design. Existing material-control systems could easily move these sedentary lots to carousels but proactively stage them prior to need.

The minienvironment and more-secure wafer capture in FOUPs (relative to 200-mm cassettes) allow 300-mm factories to relax cleanliness and vibration requirements. Since cost pressures are only going to increase, carousels could well become the next disruptive technology for conventional stockers for 300-mm automated material handling.

COURT SKINNER (Semiconductor Research Corp.; ITRS factory integration technology working group): The major theme in factory automation continues to be the move from the dual-track system with an "interbay" track for lots, boxes, pods, standard mechanical interface (SMIF), carriers, cassettes, or whatever you'd like to call them, and an "intrabay" track for wafers that are in the bay and moving through the equipment. Getting to this level of control requires a higher level of software reliability and a new set of standards, both of which are issues for future factories. At the extreme, this trend leads to tracking chips, but in the interim a significant challenge is to allow unique processes for each wafer. This is a clear complexity nightmare, of the same nature as solving the road traffic problem by attempting a freeway 75 lanes wide. New approaches to communication and reliability are clearly required.

We're not quite to the point of robot emotions, but clearly emulators are going to be crucial in the design and implementation of future factories. —Court Skinner

Another issue that these approaches raise is recipe validation. Not only are routing issues crucial, but also there has to be good assurance that wafers are processed with the correct process sequence. Even with today's high-speed processors, the validation time can be excessive with legacy architectures.

Another requirement is to monitor process consistency independent of detailed knowledge of the process recipes, which are often "classified" or trade secrets with chemistries that can't be disclosed. One approach is to "fingerprint" the process chemistry using something like electrochemical impedance spectroscopy (EIS). There are other methods that might be more useful in gas or plasma process environments.

Increasingly people are recognizing the "Gaia principle" in factories. Everything is connected to everything else: the factory is "alive." We're not quite to the point of robot emotions, but clearly emulators are going to be crucial in the design and implementation of future factories. We need valid answers to a plethora of "what if" types of questions, and we need them way ahead of the investment in the factory itself. Further, as the Gaia principle suggests, these factories are more than the traditional "sum of the parts" or even the "integrated whole." They are a reflection of the chip technology that they are being required to produce and are therefore no less sophisticated to create.

Further, everyone in the factory needs to know what's going on in a language that's natural to them. The Internet model has been adapted to Ethernet for Control Automation Technology, known as EtherCAT, but other approaches will likely be developed as well, such as special adaptations of HTML to drive factorywide Web connectivity: Imagine just linking up to fww.etchandclean.fab to find out what's up.

Beyond the factory, the supply chain, or more likely, the Internet, is another trend that can't be ignored. In many cases, even something as simple to conceive as a run card has to cover multiple continents, cultures, and languages. Much of the traditional transportation costs, which are much more energy intensive than we'd like to believe, will move via electrons rather than jet fuel.

With all the focus on cost reduction and speed improvement of both chips and their production, the pace and intensity of novel approaches won't diminish. Increasingly the technology will be profitable because it solves social as well as economic problems, environmental as well as financial problems. Since modern communication technology will become as ubiquitous as the phone and a lot cheaper means the Moore's law trend in wafer-level factories will keep the same pace as that of the chips themselves. Finding waste will become a significant occupation for all of us, of seconds, pennies, and very small fractions of each, as subsidies that shift the burden of costs to the customers are increasingly reduced or eliminated.

REX WRIGHT (director, interoperability, Asyst Technologies): Because 300-mm fabs must rely on automated delivery of FOUPs between stockers, AMHS, and process tools, the interoperability of automation components (FOUPS, loadports, AMHS, etc.) is critical. Interoperability is defined by International Sematech as the ability of two or more systems to exchange information and to mutually use the information that has been exchanged. Although SEMI has defined software and hardware standards for many of these interfaces, interoperability problems can arise even when the components are SEMI compliant and work to specification. For example, a properly functioning AMHS system can time-out in its attempts to dock a FOUP to a properly functioning loadport, resulting in an E84 communication error. These failures are particularly common in the initial phases of production ramp and can significantly impact productivity and time-to-volume in a 300-mm fab.

Fabs need to collect data about the health and performance of the automation components. —Rex Wright

A systematic approach combining data collection and specific expertise is required to identify root causes and remove these issues from the fab. First, to categorize these kinds of issues properly, fabs need to collect data about the health and performance of the automation components. This includes data about the loadport, carrier ID reader, E84 connection, and AMHS. There also needs to be available information about the service and maintenance records for other equipment that will provide clues about potential sources of disruption to the layout and positioning of automation components. Coupled with expertise on proper mechanical operations and procedures (human operator and factory protocols), root causes can then be identified and dealt with quickly and directly. This combination of data collection and understanding is the key to optimizing fab automation.

CHRISTOPHER W. LONG (IMD contamination/ESD control, IBM-Burlington, VT): With the advent of 300-mm manufacturing and the resulting fab automation requirements, a common misconception has been that contamination concerns will be greatly reduced, if not eliminated. This attitude could lull the unprepared into a false sense of security. A number of issues must be taken into consideration. The primary lines of defense against contamination incursion on the 300-mm wafer are the FOUP and the integrated minienvironment (iMEV). The FOUP acts as a carrier that allows for transfer of wafers from tool to tool. The iMEV enables the transfer of wafers from FOUP to tool via a loadport into the tool through a clean environment, typically rated at ISO Class 2 or better.

It is critical that the iMEV environment be properly tested to a specific set of criteria before a tool is released to manufacturing. Stipulations for some tests are clearly defined in ISO 14644-2 (Cleanrooms and Associated Controlled Environments—Part 2: Specifications for Testing and Monitoring to Prove Continued Compliance with ISO 14644-1). It is important to balance airflow while maintaining the required differential pressure level between tool and cleanroom. Interior-surface particle levels and interior static-charge levels should also be addressed. Standard 200-mm techniques such as ionization can be applied to help control both factors. Additionally, a methodical retest program is a good idea to ensure continued performance of the environment; again, ISO 14644-2 can provide guidance with respect to frequency.

What is a clean FOUP? How clean is clean? How often and at what process steps does it make sense to change to a clean FOUP? —Christopher Long

There are some key questions that each manufacturer must address with respect to FOUPs. What is a clean FOUP? How clean is clean? How often and at what process steps does it make sense to change to a clean FOUP? At what process steps or technology node should FOUP purging be considered? These are crucial issues as manufacturers ramp up 90-nm production and drive to the 65- and 45-nm technology nodes.

Finally, FOUPs and iMEVs are well and good, but they provide no protection to the tool that is opened for planned or unplanned maintenance. It is critical that the environment around an open tool be maintained at the cleanest possible levels and that strict protocols be observed when working in a tool.