New Twists on 300 mm Fab Design and Layout

MIchael Weiss,
PRI Automation Inc., Billerica, Mass. -- Semiconductor International, 7/1/1999

'O.K.,you are saying to yourself, 'here is another article telling me about the promise and opportunities associated with the conversion to 300 mm wafers. I think I'll skip over this and see what's on the next page.'

Whoa, slow down there. I know you have 'heard it all before.' What I want to do in this article is open your mind to some different ways of looking at the possibilities. Stretch those creative muscles; go crazy; try to really take advantage of the opportunity; make yourself a corporate hero. Stick with me through the requisite background and introductory information, and I promise I will make it worth your while.

The setup

One of the reasons that you have heard so much about the design of 300 mm facilities already is because an impressive international effort has been under way to develop 300 mm manufacturing standards before the facil ities are designed. Two consortia have led this effort: I300I and J300. I300I, the International 300 mm Initiative, part of International SEMATECH, was made up of chipmaker from the United States, Europe, Taiwan and Korea. J300 is made up of chipmakers and suppliers from Japan. Both groups have been active in initiating discussion and developing standards for 300 mm manufacturing related issues.

Fig. 1. In the traditional bay and chase layout, a central corridor flanked by process bays is arranged orthogonally.

The two consortia agreed to use SEMI to administer the international standards to be developed out of these discussions. Included are standards on wafers, materials, carriers and factory interfaces. I300I and J300 published joint guidelines to be used in the design of new 300 mm facilities. The existence of these guidelines will speed development of the tools required for handling 300 mm wafers and carriers and provide semiconductor manufacturers with multiple supplier options.

The guidelines were developed with an eye toward increasing the overall effectiveness of the fab as well as capital productivity of the facility. The following are some of these guidelines and their implications on facility design, layout and operation.

Wafers will be transported and loaded horizontally

Traditionally, wafers have been transported in cassettes of 25, with the wafers aligned near vertical. As the cassette was loaded in the process tools, in most cases the cassette had to be rotated to orient the wafers horizontally. Wafers were transported vertically to protect them from shock, vibration and contamination. Unfortunately, the rotation of the carrier could itself result in damage to the wafers or injury to the operators. The use of SMIF pods demonstrated that wafers could be handled horizontally in an enclosure and not suffer damage or contamination, while reducing operator difficulties.

Fig. 2. The ballroom layout is basically the bay and chase layout shown in Figure 1 without the walls.

Approximately 85% of process tools handle wafers horizontally. The elimination of the 'extra' carrier rotation at these tools itself can contribute to increased factory throughput.

Strict dimensional standards

SEMI standards have been developed for 300 mm cassettes and carriers. While these standards cover a large number of items, the following features are the most important. The carriers will transport either 13 or 25 wafers. It was also determined that cassettes and carriers will all make use of a kinematic coupling to register the carrier at the tool load port. This will allow for interchangeability between carriers. Another feature of the kinematic coupling mating feature is that it should allow the carriers to be placed more precisely, again, speeding the operation of the tool automation system. Another side benefit of this precise location will probably be reduced contamination caused by wafers' interactions with the carriers during removal and insertion.

Carrier type

I300I and J300 both determined they would use horizontally oriented carriers with a kinematic coupling loading feature. At first it appeared as though the J300 was going to recommend open cassettes while I300I was advocating closed pods to be used with minienvironments. Instead, both organizations called for the eventual use of front opening unified pods (FOUPs).

Tool interfaces

All tools will be required to meet the pending SEMI E15.1 standard. Fundamentally, this standard requires all tools to be loaded at a height of 900 mm from the floor and within certain horizontal constraints. E15.1 also attempts to address the need for clearances for the tool loading systems (referred to as 'easements'). By combining this with the standard carriers and horizontal transport, the task of automating the loading and unloading of the tools is greatly simplified.

Additionally, the guidelines call for the capability of buffering enough work at the tool to allow for continuous tool use, even while waiting for lot delivery and removal.

Measuring success

I300I and J300 did all the work. By following their guidelines, the fab designer knows what kinds of carriers and tool interfaces he will use, and all the tools will be supplied with integrated minienvironments. The consortia have even recommended the factory layout: bay and chase design with narrow aisles if using overhead material transport and wider aisles if using automated guided vehicles (AGVs). All the fab designer has to do is fit the specified process tools into the available space.

Wait a minute. If it is that easy, how can one manufacturer differentiate himself from another? Consider the costs of manufacturing a single wafer; the cost of the wafer is made up of the following components:

• Raw materials
• Process materials/consumables
• Capital cost of tools amortized over wafers
• Cost of maintaining WIP
• Facilities related costs.

Assuming that everyone pays the same for materials, consumables, energy, bricks and mortar, then the opportunities for reducing manufacturing costs (that is, increasing fab productivity, wafer/$) can only be a result of reducing the amount of WIP and the number of tools (assuming fixed yield, die size, etc.). In order to reduce the amount of WIP in the fab, we need to shorten the manufacturing cycle time for a lot. That is, we need a lot to spend less time in transit and storage waiting to be processed. The greatest productivity gain can be seen by increasing tool utilization or OEE (thus reducing the number of tools required). Unfortunately, the standard approach to increasing tool utilization is increasing the amount of WIP queued in front of the tool, which increases WIP cost. Ideally, then, we need to optimize the amount and distribution of WIP in the factory with the number of tools required to provide the required throughput, that is, minimize WIP while maximizing throughput.

Fig. 3. The use of automated lot handling requires the use of transit lanes in this ballroom layout.

Stated mathematically, manufacturing costs that can be affected by factory design are represented by the following:

Costs = F(WIP, 1/OEE, Cycle Time).

By reducing WIP and cycle time while increasing OEE, we decrease costs. What opportunities do the new 300 mm standards give us that we have not had before?

Buffers and minienvironments

Every tool is being specified with a minienvironment to allow it to interface to the FOUPs. Numerous studies have been done showing some of the facility savings that can be achieved by using minienvironments in 200 mm fabs. What the use of minienvironments also allows us to do is break the bay and chase paradigm and locate the tools optimally to reduce the amount of time materials spend in transit instead of meeting the constraints required to build a clean environment. We will look at that in some of the coming examples. Also, the size of the buffer in front of the tool must correspond to the process time of the tool and the transit time required to fill the buffer. Again, by reducing transit time, we can reduce WIP.

So, it appears that one element we can concentrate on is reducing travel time and distance. Can we also eliminate stockers for buffering large amounts of material? Since the tools will have local buffers, we can eliminate some stocker requirements in

the fab. Unfortunately, because of the granularity of the available tools, the mix of products and the unanticipated events, there will still be some need for stockers to provide 'bubble management' that will vary from fab to fab and from time to time.

The baseline

As stated earlier, the baseline design that we considered first is the 'bay and chase' layout, a central corridor flanked by process bays arranged orthogonally. This allowed the architect to design a fab with a minimum cleanroom footprint, service access to the tools and a transport corridor for operators and wafers. Figure 1 illustrates a generic version of this layout.

Typically, these fabs are the size of a football field, so for our baseline, imagine the central corridor is 100 m long, and each bay is 30 m long. Also assume that this fab can handle a mix of 200 process tools with 10 tools per bay. Materials enter the bay for one or two operations, and then are transported to another bay. For simplicity, imagine that each trip through the central corridor averages half the corridor's length, or 50 m. In a 200-step process, the wafers will travel 200 3 50 = 10,000 m just going up and down the central corridor. That equals 10 km of travel, where absolutely no value can be added to the wafers. Let us see if we can improve upon this baseline design.

Fig. 4

Step 1: Lose the walls

Take the fab of Figure 1, delete the walls, but lay out the tools in the same locations, and you have the ballroom shown in Figure 2. Operators can now transport materials directly from tool to tool without having to go through the central corridor. In the case of moves between proximal bays, this reduces the up-and-down-the-bay travel. In the case of moves between distant bays on opposite sides of the facility, it also reduces travel distance if we allow diagonal moves instead of purely orthogonal moves. This brings the question of how we transport lots. For purely random, point-to-point transport, the simplest device is the human operator with a cart. Unfortunately, allowing random path selection by all operators that would be required for efficient transport would require a large number of operators and large spaces between the tools. If we employ automated techniques, such as guided vehicles or monorails, then we will have to restrict transit along 'lanes' (Fig. 3), resulting in an effective layout similar to the baseline design.

Step 2: Rearrange the tools

Figure 4 is an illustration of what I like to call the 'pinwheel' layout. In this facility, a central core replaces the central corridor. All bay-to-bay moves are made via the central core. Since the core length is minimal, we immediately see a gain of 10 km in travel distance and an average travel per move of one bay length (half-way per bay). If we add a couple circular paths (Fig. 5), we have a 'cartwheel' layout with a potential for even shorter average moves.

Fig. 5 A circular path at the perimeter is added to the 'pinwheel' layout to form a cartwheel.

Step 3: How do I do that?

'Mitchell,' you say, 'conceptually I follow you, but how do I get all those moves through that central core without operators bumping into each other?'

Glad you asked. First, look some more at the numbers. In a typical fab today, we see 200-600 interbay moves/hour, so about 400 moves/hour is a nice average. Also, we see about 50-100 intrabay moves/hour. Simply put, that means we have to move materials at least once a minute within the bay and seven times a minute through the central core.

One option is to install a high-speed stocker in the center core. The fastest stockers today can load or unload about four times a minute, but since we need to enter and exit the core at a rate of seven per minute, we need the equivalent speed of 27/4 = 31/2 stockers. So this mythical central stocker has to have the capacity to buffer all the materials we need to buffer to provide smooth operation and be able to perform more than seven transactions per minute.

How about transport up and down the bays? I300I suggested overhead hoist vehicles, which provide transport and delivery to the tools while requiring a minimal aisle width. Another concept presented was the 'flat intrabay stocker' as used in the 'sea of lots' concept. This machine is a stocker at the ceiling level that combines local storage, transport and delivery. Again, the design challenge will be developing a stocker with adequate throughput. Finally, we can use ground-based delivery systems.

More ideas

Now that we have reduced transport distance to a minimum just by changing the layout of the fab from rectilinear to radial (and invented some very fast stockers), what other opportunities are there for productivity gains? One option might be to integrate sorting or inspection with the central stocker, or intrabay transport machines. In this way we would make these operations occur concurrently with the transport steps. We could put CMP or implant on a separate level, near the central core, so that travel distances are shortened even further, and environments are better isolated. We could do a multilevel cartwheel.

Summary

The advent of 300 mm presents us with a newly rationalized set of factory interfaces. We can take advantage of this to optimize the layout of our factories and minimize transport and cycle times and WIP. All that is required is the ability to break out of old paradigms that we have fallen into that are no longer justified (the rectilinear bay and chase model) and find the new ones that best suit our needs.

Suggested Additional Reading

1. Weiss, M., 'Overall Factory Effectiveness (OFE) and its Implications for 300 mm Tool Automation,' Semiconductor Fabtech, Fifth Edition, p. 57.
2. Weiss, M., 'Using Analytical Methods to Accelerate Material Handling System Design Optimization,' SEMI Taiwan Technical Symposium, September 1995.
3. VanLeeuwen, C., 'Implications of 300 mm for Fab Design and Automation,' Semiconductor International, April 1996.
4. I300I Guidelines on 300 mm Process Tool Mechanical Interfaces for Wafer Lot Delivery, Buffering and Loading (Rev D. 9/3/96).
5. 2nd Lecture, IC's Factory Design for 300 mm Wafer Line Standardizing Study, Japan 300 mm Semiconductor Technology Conference (J300), Dec. 3, 1996.
6. Plata, J.J., '300 mm Fab Design, A Total Factory Perspective,' IEEE 6th International Symposium on Semiconductor Manufacturing, San Francisco, Calif., October 1997.
7. Subramanian, B., Kryder, K. D., 'Automation Challenges in the Next Generation Semiconductor Factory,' Semiconductor Fabtech, 7th Edition, ICG Publishing.
8. Griessing, J., Ortner, J., 'Sea of Lots Concept,' European Semiconductor, September 1997.