An automation route to upgrading fabs toward 300mm
10/01/2000
SPECIAL REPORT: State-of-the-art manufacturing
Michael Brain, Asyst Technologies Inc., Fremont, California
overview
Today's wafer fab automation capabilities, particularly those associated with minienvironments, enable competitive upgrading of existing facilities. Indeed, in a number of cases, a minienvironment-based fab upgrade strategy has proven to be a low-risk means to improve the performance of an existing facility. This approach is a viable alternative to a new facility that still keeps semiconductor manufacturers on a route to an eventual 300mm fab.
 Figure 1. Example payback of a fab upgrade project as a function of die yield improvement. |
Competition in today's semiconductor industry is driving product pricing lower in search of market share. This requires a fab to continuously lower the cost of finished wafers. The challenge is that in this environment of cost reductions, the industry has to improve the performance of semiconductor products. Semiconductor manufacturers must quickly introduce new technologies such as copper interconnections and innovative dielectric materials to their production operations, and yet maintain high yields. Consistent yield performance is no longer a fab goal, but is now a baseline requirement.
A semiconductor manufacturer in need of more capacity will try to choose the most economical approach. The options are to build a new 300mm fab, build a new 200mm fab, or upgrade an existing fab. The device manufacturer must also consider buying capacity from a wafer foundry, which can offer state-of-the-art wafer-processing capacity at a globally competitive price without investing more than a billion dollars in capital.
 |
The transition to 300mm processing now appears to be progressing rapidly, but this serves to complicate the decision for many fabs. A new 300mm fab is expected to offer tremendous financial leverage when fully ramped to capacity. The 300mm surface area is 2.5 times that of 200mm wafers, yet may be produced at about 1.5 to 1.8 the costs incurred in current fab generations. Obviously, current fab owners would like to build their next fabs at 300mm to take advantage of this economic leverage.
The capital expenditures are large enough to make the risks substantial. But the aggressive manufacturer asks, "Do I want to be the last manufacturer to build a 200mm fab?" And yet the risks are compounded when building a new fab with a new wafer size (300mm) and new technologies (copper, etc.) all at one time.
Indeed, such an approach may not be the best use of capital. As an alternative, a faility upgrade might improve yield and add enough capacity to bridge a surge in demand. Many fabs operating today are producing semiconductors using 0.18-1mm-design rules on 150mm or 200mm wafers. A facility upgrade that allows process development for 130-100nm processes with copper and new materials within an existing 200mm facility could be a viable answer.
Indeed, several fabs have adopted this incremental migration path to 300mm. Some have upgraded from 150mm to 200mm, adding SMIF and minienvironments to enable aggressive process development in the same fab as current production. Foundries in Taiwan have achieved highly successful examples of this approach.
Below, we discuss wafer-handling and automation issues, and associated economic ramifications facing facilities that want to implement an upgrade to provide advances in process capability on a possible route to a 300mm fab. In essence, these are very viable strategies to extend the useful life of an existing facility.
The overriding concern
When a fab introduces a process that uses linewidths below 180nm, contamination is of great concern. Indeed, manufacturers are becoming concerned about additional types of contamination as they transition to sub-quarter-micron processing. International Sematech has performed a comprehensive survey of fab environmental conditions. Results show that a variety of airborne molecular contaminants exist in fab airflow, with most facilities exceeding the recommended levels suggested by the SIA Roadmap for 0.25mm processes. The effects of these contaminants on yields and lifetime performance of semiconductor products are not fully known at this time. This problem is expected to become a more significant issue as gate oxides become even thinner.
Transition strategies
An obvious fab transition approach involves building a new facility, buying new equipment, and ramping up production in the new fab with new technologies. The old fab is relegated to less demanding products and, often, eventually phased out. Unfortunately, this approach delays the availability of the new processes until a "green-field fab" can be built. Valuable time needed to debug and stabilize the processes is lost waiting for the new fab. The cost of new construction is very high, and it is compounded by the costs of closing down the old fab once the new facility is online. So, if a viable strategy can be found to upgrade the existing facility, these costs can be eliminated and time to market can be dramatically reduced.
Various strategies for fab upgrade offer different advantages and disadvantages. Some methods can minimize the disruption of wafer production during the upgrade, while others minimize the time that is required to complete the upgrade. To determine the best course of action for a specific facility, one must evaluate a variety of factors. For example, a full audit of the current facility is typically performed to establish the current baseline. Then, the goals for the upgraded facility are determined and a set of alternative upgrade plans is developed. With the evaluation of these alternatives, one can establish an optimal upgrade plan.
Fab audit
To perform an audit of a wafer fab facility, interpret the results, prepare alternative plans, and evaluate these alternatives, a cross-functional team is most effective. Team members should represent a diverse skill set drawn from facility staff and suppliers (see table).
The audit team needs to first establish the goals of the upgrade program. What are the products to be made? What are the predicted critical dimensions, die sizes, and recovery strategies (such as redundant RAM cells)? What is the time schedule for product development, introduction, and production ramp? What technologies are to be used (copper, new dielectric materials, etc.)?
With the short- and long-term goals in mind, the team characterizes the facility's capabilities. All factors that might impact the ability to perform at the desired technology level should be examined, including laminar airflow and velocity; material-handling protocols; materials installed within the fab; chemical supplies and distribution; vibration isolation; exhaust; DI water; chilled water; heating; waste treatment; process-cooling water; house and specialty gases; electrical power; and the cleanroom HVAC system.
Current product mix and yield sensitivities must also be determined and documented. Information about probe yield history is gathered, including photo-limited yield data. Defect Paretos are developed to determine the major sources of defects. A yield model is established that relates current yields with current defect densities as accurately as possible. Many fabs use a negative binomial yield model as suggested in the International Technology Roadmap for Semiconductors (ITRS).
While this model is commonly used, we have found that photo-limited yield often follows the Murphy-Seeds model:
Yield = (1 + DIA)-8, where DI represents defect density/layer and A represents the chip area corrected for critical (particle-contamination-sensitive) area.
In the audit phase, some time should be spent understanding the protocol and consistency of material handling by the operations staff. The use of carriers and wafer-handling devices or systems and an understanding of the potential contributors to contamination, breakage, and scratches are essential.
Financial criteria to be gathered include "the cost of money," manufacturing capacity and value, and revenue associated with work-in-process, equipment depreciation schedules.
 Figure 2. Wafer-scratch reduction measured during an upgrade to isolation-based wafer handling. |
Upgrade options
Armed with an accurate model of the existing status and desired goals, the team can define the areas that will require an upgrade. In each category, the requirements to support the desired level of processing are defined and then compared to the existing facility. In each area that requires upgrading, the potential techniques are identified. Now the team can evaluate and estimate the potential impact of the various upgrade options (Fig. 1). To compare options, the total cost and estimated benefits of each option are quantified. These values are used to perform a return on investment calculation allowing us to compare the payback period of different options.
The net present value of a potential investment should also be examined to demonstrate the long-term value of more expensive investments.
In Fig. 1, we see an example upgrade project that converted a fab from 150mm to 200mm wafers and introduced SMIF and minienvironments. Adding minienvironments was expected to increase the net die yield by at least 4% (as indicated by the red arrow). This would have resulted in a payback of <12 months. The results of this project showed a much higher yield gain and therefore a faster payback.
The role of minienvironments
Because of the overriding concern for increasing levels of contamination, use of isolation or minienvironment technology has proven to be a highly effective alternative to upgrading a facility's air-handling system when upgrading existing fabs using open 150mm and 200mm cassettes. Proper application can maximize probe and line yields by reducing wafer scratching (Fig. 2) and particle contamination (Fig. 3), both of which can distract yield management personnel from the pursuit of more systemic yield problems.
In Fig. 2, we see the progressive reduction of scratches each week during a 1-year upgrade project. As the number of wafer-handling operations increases over the period of 52 weeks, the number of scratches that are found on the wafers decreases proportionately.
In Fig. 3, we see a double effect of the introduction of minienvironments. First, the overall level of contamination is greatly reduced. (Notice the data are not even plotted on the same scale.) In addition to the average cleanliness supported by the minienvironment, note the elimination of "particle bursts" in the data of Fig. 3. This is what provides a consistent manufacturing environment.
As shown previously in Fig. 1, the resulting yield improvement typically supports a return on investment of <1 year. In addition to improved yields, the upgraded fab can support the production of larger die sizes and decreasing critical geometries ("CD shrink"). If a shrink were attempted in the nonupgraded facility, the yields would be prohibitively low.
Upgrading a facility's air-handling systems can also be considered. However, this approach requires extensive facilities remodeling, substantial investments in additional fan-filter units, retraining of operators, and improved gowning protocols. These can result in lost production time and increased project costs.
In the end, use of state-of-the-art isolation technologies can achieve a tighter level of contamination control than traditional air-handling techniques and does not require the facility to be shut down during the upgrade. By continuing production during the upgrade, a manufacturer saves many weeks of production revenues and can achieve critical production targets.
Isolation technologies concentrate on critical areas for material handling: equipment input and output and wafer handling. By carefully engineering the airflow and maintaining positive pressure in the minienvironment, the cleanest possible environment is maintained for the wafers. Because the airflow is concentrated in a smaller volume, significant savings in operating costs result from the reduced electricity demands.
A number of studies have shown that a wafer spends about 80% of its time in a wafer fab waiting in a buffer location or in transit to the next process location. In many fabs, the corridors used for lot transportation are less clean than the areas dedicated to process equipment. Storage racks are often placed against a wall in a corridor where the airflow is not laminar and the traffic movement is high. This puts lots in jeopardy unless they are in sealed storage and transport containers.
 Figure 3. a) Sample airborne particle counts in a Class 1 fab without minienvironments and b) in a Class 1000 fab with minienvironments. |
With a fab using minienvironments, the cleanliness of the ambient environment becomes less important. This means that additional equipment can be added in a fab that is out of floor space by using nontraditional space. Examples include adding equipment to the sub-fab area, conversion of equipment maintenance gray areas, or test areas for additional equipment. When wafers are enclosed in minienvironments, these areas have been proven to be very acceptable process tool locations.
Operator cassette-handling protocols vary widely from fab to fab. When performing a transfer of a cassette from a lot box to the cassette loadport of equipment, the wafers are exposed to the facility airflow, as well as potential breakage. Upgrade of these aspects can be attempted with training and monitoring programs. Use of minienvironments and automated SMIF input-output devices, however, eliminates these as significant factors.
Automated material handling
Automated material-handling systems (AMHS) can be used to advantage in fab upgrade situations. If a fab is short of space, a stocker might improve the density of storage, thereby freeing up some floor space for more process equipment. Innovative new AMHS techniques offer zero-footprint buffering and storage of material above the fab equipment or away from the main fab area. If minienvironments, such as SMIF pods, are used for storage and transport, additional areas can be used for storage such as sub-fab or remote building areas.
Information automation
In many cases, the most significant improvements and fastest payback of any upgrade element involves the automation of information. For example, automating the identification (ID) of lots can save operator time. More important, linking a lot ID system to the manufacturing execution system (MES) that is controlling the fab can eliminate costly process errors. Real-time visibility of manufacturing issues can make factory management more efficient. If the connection extends all the way from MES to lot tracking and material handling, an unbroken chain of protection is established.
Once material is connected to the MES, the major remaining source of misprocessing is the selection of recipes. If the operator does not need to be concerned with making sure that the lot has been correctly identified and tracked with the MES, and that the correct recipe has been selected, he is free to concentrate on other tasks. This allows the operator to focus on assuring that equipment is operating within process control tolerances, and that the equipment is not allowed to starve for material, causing a loss of equipment effectivity.
Automated ID systems can be an essential element in an AMHS strategy. If a stocker can read the lot ID directly, the storage process is more efficient. If an operator must enter the lot ID manually, time is lost and errors can occur. If a lot is misidentified to a stocker, it can be stored on a shelf and "lost." In many cases, fab efficiency has been improved by adding lot readers in manual storage racks. This provides some benefits associated with AMHS, such as improved lot retrieval time at a lower cost.
Financial analysis
Once the automation upgrade options have been determined, they must be evaluated to determine the return on investment.
Certainly, there are also important decision criteria that are not financial. These might include factors like the capability of the organization to implement the option in question, the availability of capital to make a specific investment, or the risk associated with an approach.
To perform an economic evaluation, however, the team should estimate the total project cost for each option proposed. The cost should include the total cost of ownership of the project. This includes equipment that must be purchased along with services to be rendered, installation, maintenance, and support. Costs of construction and commissioning should be estimated. It should also include the cost of any lost production time and additional or fewer operators or technicians who might be required. The annual costs of power and other utilities must be included for accuracy.
The improvements that result from an upgrade must also be estimated to calculate the return on investment. The largest improvements usually come from increased die yield and increased wafer yield. In our experience, a fab yield can be increased between 4% and 25%. In addition to yield, one fab was able to increase the revenue that it obtained from each wafer due to the inclusion of a higher level of technology.
Overall equipment effectivity improvements also directly impact the bottom line. Productivity is affected by any change in gowning and handling protocols required of the operators. Operator efficiency can be improved by as much as 15% based on simplified gowning protocol.
Using the audit of the current facility as a benchmark, we are able to estimate the specific benefits of an upgrade based on the experience of previous upgrade projects. With an estimate of the costs and the benefits in hand, a return-on-investment calculation can be performed. This puts the project into clear perspective for fab and corporate management.
Fab upgrades have now been performed in more than 18 facilities around the world. A common upgrade is to add capacity by expanding into nearby Class 1000 space such as an old test or even office area. Several fabs have converted an older, 150mm fab to a current state-of-the-art 200mm fab. These conversions have been done in a phased approach, where current production is phased out while new production is ramped up.
We are now moving into the next generation of IC fabrication with 300mm wafers and an increased level of economic pressure. The minienvironment-based fab upgrade strategy has proven to be a low-risk means of improving the performance of an existing facility, while helping to prepare for the future.
A suitable wafer fab audit and upgrade planning team
Individuals who specialize in:
- Equipment engineering
- Manufacturing operations
- Facilities engineering
- Yield enhancement
- Contamination control
- Automated material handling
- CIM and information systems
- Economic modeling
- Industrial engineering
Michael Brain graduated from the University of Cincinnati, School of Electrical Engineering. He has more than 23 years of factory automation experience working for General Motors, National Semiconductor, Hewlett-Packard, and Asyst Technologies. Brain is currently VP of systems and software solutions at Asyst Technologies Inc., 48760 Kato Rd., Fremont, CA 94538; ph 510/661-5975, fax 510/661-5960, e-mail [email protected].