FlexFab - A New Approach to Fab Design

Lindsay Leveen - SLS Partners, Inc. 

Introduction

Modern semiconductor wafer Fabrication facilities producing sub-micron IC’s are rapidly becoming prohibitively expensive to build and operate. The escalation in costs is driven by the industries ongoing need to reduce cost per bit in a competitive market. This demands smaller device geometry fabricated on larger wafers in factories producing greater quantities. The prevalent method of Fab design employs a bay and chase or ballroom clean room layout in a large rectangular building under a single roof.

This paper will describe an improved design approach offering significant advantages.

This Fab design is significantly different from current practice, so much so that the US Patent office recently awarded a patent to SLS Partners [1].

Wafer Fab Trends

The trend data of Table 1 illustrate the escalation of Fab capital cost over time. The Fab cost includes factory construction and processing equipment, of which construction cost is typically 30 - 35% [2].

Table 1

Period 80’s 90’s 2000’s
Feature Size (m ) 10 - 2 1.5 - 0.25 0.18 - 0.07
Mask Layers 10 - 15 15 - 20 20 - 25
Wafer Size (mm) 100 - 150 200- 300 300 - 450
Clean room Size (SF) 25 - 50 K 50 -100 K 100 - 200 K
Clean room Air Class 1 - 10 Class 0.1 - 1 Class 0.01 - 0.1
Fab Cost ($ Million) 50 -300 400 - 2,000 2,000 - 5,000

The projected cost of up to $5 billion for the Fabs of the next decade creates a major problem for the chip industry. Much effort is being spent on ways to reduce the cost escalation trend, including ways to optimize Fab design and construction. Recent proposals in this regard include a revolutionary ‘lights out, hands off’ design [3] out of SEMATECH and an evolutionary design [4] from a consortium of industry participants, which makes extensive use of cluster tools and minienvironments. These designs have not been adopted, perhaps because of the conservatism in the wafer Fab design community against revolutionary changes and the trend away from over-use of cluster tools due to the risk of increased equipment downtime. Here we will describe a new Fab design having several unique advantages.

Conventional Fab Design

The current state-of-the-art Fabs utilize a 3 story building with a structural system covering a very wide span (Figure 1). The Fab clean room occupies the middle story. All equipment in the Fab is installed on a raised floor system which is required to facilitate air return to the recirculating air handlers. In a bay and chase layout air is returned via the chases, in a ballroom layout a return air plenum under the raised floor is employed. As the ballroom is wide and long, the perforated raised floor serves to keep laminarity of clean room air flow.

The ground level functions as an area for locating support pumps and utilities, chemicals, and gases needed to operate the process equipment. The upper level of the building houses air recirculation fans. Scrubbers and exhaust treatment systems are also located on the upper level together with the stacks that expel treated exhaust to the atmosphere. Make up air handlers are located in a rooftop penthouse. Safety codes require emergency corridors on all levels of the Fab as well as a separation of the corridors used for chemical delivery and emergency exiting in case of alarm conditions. A central utilities building (CUB) for the process and clean room HVAC support systems is located along one side of the rectangular shaped Fab.

There are a number of drawbacks associated with this approach.

The FlexFab Approach

The FlexFab departs from the conventional approach. It incorporates a multiple building structure in a hub and spoke or finger arrangement (Figure 2), where each spoke contains the clean processing area (Figure 3). Air recirculation fans are located in the attic space above the clean room. Makeup air handlers and where needed, gas or chemical vaults and scrubbers are located alongside an exterior wall of the spoke (Figure 4). Support facilities which are common to all spokes (electrical, bulk gas, chillers, process cooling water, etc.) are centrally located within the first floor of the hub or proximate thereto. Processing equipment of similar types are grouped within the spokes and those support services predominantly used by those equipment types (ultra pure water, specialty gases, bulk chemicals, liquid waste, exhaust, etc.) are located adjacent to the appropriate spokes. Exterior walls of the spokes are demountable for direct installation of process equipment.

The advantages of this approach are immediately apparent.

As a result of these advantages, and despite the tradeoff of extra land and more exterior walls, the FlexFab offers construction and equipment fitup savings in the 15 to 20% range and will be ready for wafer production three to five months earlier than conventional Fabs.

Automation

The FlexFab design has other significant advantages. The need for automated handling systems is mandatory at the 300 mm wafer size. The SEMATECH [3] patent as well as several Japanese groups have postulated that a central hub facilitates improved automated wafer handling and processing. The central hub in the FlexFab layout acts as the terminus for routing wafers to their next processing step. The radial distribution to and from process locations lends itself well to automated handling. One of the spokes can be used for WIP storage. Wafer inspection and metrology can be co-located with the WIP area. Fabs producing thin film displays will also benefit significantly from the FlexFab design. The large glass substrates require automated handling, the process equipment is extremely large and their placement and hookup in a traditional Fab is extremely difficult.

Expandability

Perhaps the most significant advantage of the FlexFab design is that it is incrementally expandable. Fab processing capacity is easily expandable by radial extension of the spokes (Figure 5) which can be accomplished with little to no impact on existing operations. Industrial Engineers attempt to forecast Fab equipment throughput. This can never be done exactly but is merely a simulation of an expected outcome. In the design of a traditional Fab, the tool layout is fundamental to the overall design. During design and construction, changes are often made to the equipment set and layout which causes significant changes and delays in the Fab design and construction in the traditional approach. However, changes to the anticipated tool set do not adversely affect the FlexFab design. If for example it is determined that additional plasma etchers are required, the plasma etch spoke can simply be expanded without changing any other process tool layout..

This scalability feature offers an attractive option to hold off on incrementing building size until process yields are maximized and market demand is better determined. Phasing capital expenditure in this way, coupled with the overall savings and rapid time to production delivered by the FlexFab leads to a much improved return on capital employed.

A Fab is analogous to an airport in that, just like an airport, construction of a Fab is never fully complete. When aviation was in its infancy, major city airports terminals were laid out in simple rectangles. Now that air travel is popular, out of necessity the layout of airport terminals mimics the FlexFab approach. The analogy of the arrival and departure of planes, travelers and their luggage in an airport to that of microelectronic circuit processing is strong. However, unlike an airport, the FlexFab can be expanded with negligible impact on existing operations.

The FlexFab approach comprehends and addresses the inherent flexibility that is needed in component fabrication, if firms are to continue to prosper. The evolving FlexFab will be constructed in several phases, with radial and lateral expansions appended to the original Fab (Figure 6). The ultimate area coverage ratio (building to land) in the FlexFab approach will be slightly less (5% to 8%) than in a conventional layout, since much of the open space between the processing spokes will be used to house process support systems.

Future expansion needs are never well defined and certainly change over the economic life of a Fab, therefore, the obvious benefit of the FlexFab approach is that it prevents one from painting oneself into a corner.

FlexFab and Minienvironments

There is a trend toward the use of minienvironments and SMIF pods in modern wafer Fabs [5]. Asyst Technology claims to have sold their systems into as many as 43 Fabs. Air cleanliness requirements of Class 0.1 or better at the wafer surface (Table 1) are difficult to achieve without minienvironments, and the operational savings resulting from improved wafer yield and energy savings from the simplified clean room air handling system are becoming more attractive. Clean room personnel are more productive by eliminating the need for overly restrictive clean room attire.

The FlexFab is particularly well suited to the use of minienvironments. The narrower widths of the clean room means that adequate and laminar air flow can be maintained with through the wall air return. Thus, the FlexFab employing minienvironments can be designed without any raised floor, which is not possible in the large ballroom approach. Elimination of a raised floor system reduces the Fab construction cost exclusive of process tooling, by a further 5 - 8%. In the extreme case, for a prototyping or development Fab, the spokes could be single floor structures with no Sub-Fab reducing building costs even further.

Conclusions

The need to mitigate the disturbing upward trend in wafer factory costs has been addressed. The wafer Fab design described here avoids a revolutionary approach and applies common sense to a problem, using structural simplification and process support system optimization as the drivers for improvement. Process layout flexibility is not compromised as the device manufacturer can choose a bay and chase layout, a ballroom layout or the use of minienvironments within the process spokes. The design reduces capital cost by 15 - 20 %, allows incremental scalability as market demand dictates with little or no impact on existing operations and provides for easier process equipment installation and changeout. Further construction time is reduced by an estimated 3 months, operator safety is enhanced and the risks for catastrophic losses are reduced. In combination these features significantly improve return on capital employed and reduce technical and financial risks for the owner.

References

  1. Patent No. 5,795,356, Issued Aug 18, 1998, assigned to SLS Partners.
  2. Art, M. O’Halloran, and B. Butler, "Wafer Fab Construction Cost Analysis & Cost Reduction Strategies" presented at the IEEE/SEMI Advanced Semiconductor Manufacturing Conference, 1994.
  3. Patent No. 5,334,365, Issued September 6, 1994, assigned to SEMATECH.
  4. Castrucci, "The Future Fab, The Changing Paradigm", Solid State Technology, January 1995.
  5. C-Y Shu, "The Impact Of Minienevironments On Fab Capacity And Yield", Proceedings, Symposium On Minienvironments, 41 Annual Technical Meeting, Institute Of Environmental Sciences, 1995.

 About The Author and SLS Partners

Prior to starting SLS Partners, two and a half years ago, with Paul Reneau and Len Slater, Leveen spent six years as a senior project manager with, Industrial Design Corporation, the leading Fab design and construction firm in the USA where he managed the design and construction of several world class Fabs. Leveen, who graduated from Iowa State University as a chemical engineer, has considerable expertise in bulk and specialty gas applications for microelectronics fabrication. Prior to working in Fab design, he was the Vice President of Corporate Planning for L’Air Liquide in Paris, France.

Reneau recently retired from a long and successful career in Silicon Valley after having worked as a senior purchasing executive first at National Semiconductor and then at Cypress Semiconductor. Reneau is regarded as one of the pioneers in the Valley and was recently recognized by the President of Air Products and Chemicals for having entered into the first contract to buy nitrogen from the now famous pipeline that serves most of the Silicon Valley. Slater also a Valley veteran, was previously the CEO of Arrowhead Industrial Water a leading supplier of ultra pure water systems and services to the microelectronics industry. SLS Partners specializes in providing strategic and operations planning services to industry and government agencies involved in high technology.

Back to top

SLS Partners Inc. 1998.