Home
About Future-Fab
Download New Issue
Contact Us
Volume Archives
Editorial Panel

FUTURE FAB ARCHIVES


DMOS6: The Shift From 200mm to 300mm
Or: When Is a Conversion Not a Conversion?

(2/2/2002) Future Fab Intl. Issue 12
By John J Plata, Texas Instruments Incorporated
Print Article Print this paper
Send As Email Send as email

When is a conversion not a conversion? Texas Instruments DMOS6 wafer fab was designed and constructed in the mid- 1990s as a 200mm Megafab, the business and technology climate suggesting that its capacity would be required in advance of 300mm production becoming viable.

DMOS6 was conceived, designed, and built in the heyday of wafer fab expansion. The design was a culmination of the collective experience of TI’s internal Worldwide Construction team (part of TI Worldwide Facilities), technology experts from various internal organizations, and resources from external design and construction partners. DMOS6 was designed to be a flexible, extendable, state of the art 200mm fab.

Introduction

When is a conversion not a conversion? Texas Instruments DMOS6 wafer fab was designed and constructed in the mid- 1990s as a 200mm Megafab, the business and technology climate suggesting that its capacity would be required in advance of 300mm production becoming viable. The project was put on hold after building shell completion but prior to actual clean room construction, to allow alignment with market conditions. With the building spaces and major systems literally set in concrete, the decision was made upon restart to execute DMOS6 as a 300mm facility. The resulting ground up review and modification of the facility and infrastructure for 300mm processing highlighted that, from a facility perspective, the move from 200mm to 300mm is more evolutionary than revolutionary.

Figure 1.

DMOS6 was conceived, designed, and built in the heyday of wafer fab expansion. The design was a culmination of the collective experience of TI’s internal Worldwide Construction team (part of TI Worldwide Facilities), technology experts from various internal organizations, and resources from external design and construction partners. DMOS6 was designed to be a flexible, extendable, state of the art 200mm fab. Some key features of the original concept were:

  • Large, single room, ballroom fab.
  • Bay and chase (white area vs gray area) definition via modular interior partition wall system
  • Class .03 bays with open cassette processing, transport, and storage; Class 300 chases
  • Interbay automated material transport with distributed bay stockers
  • Manual material handling within the process bays/manual tool loading and unloading
  • Designed for all processing to be in the main fab/all on one level
  • Cleanroom support areas placed and designed to complement the process tool layout concept

The original DMOS6 project was stopped short of actually building the cleanroom inside of the building shell, installing the complex internal utility distribution systems, or in most cases, the main utility systems (i.e. DI water, exhaust systems, etc.) In this sense, therefore DMOS6 as a 300mm fab is not a conversion as much as a re-direction of effort. Although there were no existing tools to convert or fab infrastructure to modify, the system designs for the 200mm version of the fab had been completed. The re-direction to 300mm at this point allowed the REAL differences in 200mm vs 300mm cleanrooms to stand out, highlighting what did not have to change as well as what did.

Figure 2.

Figure 3.

What Did Not Change

  • The Building
  • Utility Systems
  • Fab Layout Design Concept

DMOS6 was designed and built at the end of a succession of a half dozen fabs designed and project managed by the same dedicated fab teams on both the customer and facilities sides. The accumulated knowledge and experience of these teams was expended on the design and construction of DMOS6. The depth and breadth of planning that went into the original design of DMOS6 helped facilitate a smooth re-direction from 200mm to 300mm.

The Building

Perhaps the greatest ‘asset’ DMOS6 has is its sheer size. In its 300mm incarnation, the cleanroom consumes over 150,000 square feet of the building, making it physically the largest fab at Texas Instruments. The main fab floor alone covers nearly three acres, providing a vast area for flexible layout design. This size was driven in part by the manufacturing customer desire for a large capacity fab, but also as a result of design inputs that indicated manufacturing cost benefits, process flow efficiencies and tool layout advantages to larger, wider areas to work with. Even with a cleanroom of this size, it was deemed necessary to add some additional clean space to accommodate specific new technology issues. Again, the size and flexibility of the overall facility allowed this expansion to be designed into the 300mm plan. Fab adjacent space designed for support and admin was built out as additional cleanroom space in parallel with the main cleanroom build-out. No major changes to the building structure were required.

Even though the building and external spaces for most utility and support systems had been defined and built, most of the actual systems were not yet installed. For the most part, support area sizing was not an issue in the 300mm re-direction, as facilities support systems are not driven by wafer diameter, rather by manufacturing tool utility requirements. Although many of the 300mm tools proved to be ‘thirstier’ for utilities, their larger footprint helped offset this to a great extent.

Utility Systems

The re-direction from 200mm to 300mm sparked a bottoms up review of the facility and infrastructure requirements. In the early days of 300mm, groups like Sematech (forerunner of I300I) put out guidelines that new 300mm tools should use only 1.1x the utilities, 1.2x the space etc. This was in an effort to diminish the perceived huge increase in utility consumption or tool size. These factors proved to be unrealistic and inconsistent measures of what was really happening to the tools. Tool size growth was much greater in many cases, and most tools emerged on entirely new platforms optimized for handling the larger wafers.

For DMOS6, the original utility system designs were sized for a full fab of 200mm tools at a very aggressive tool density. Although the utility consumption on 300mm VS 200mm tools did increase (plus the minienvironment load plus automation system power loads), the larger tool footprints helped to balance the space/utility equation. The wild card is that since so many of the tools for 300mm are on new platforms, actual running rates for usage in a production environment are somewhat speculative. Rather than try to outguess the future requirements of an unknown ultimate tool set, the original utility systems designs were retained. In the shift to 300mm, none of the main utility systems in DMOS6 were upsized (or downsized for that matter) to accommodate the new tools. In terms of space management, there are spots for additional capacity in the future should it be required. Again, the flexibility of the facility design makes provision for future space needs for supply and support systems. Utility distribution may be more the issue as time progresses.

Fab Area Layout Design Concept

The DMOS6 200mm tool layout concept was designed in parallel with the fab building and support areas. The initial design involved process and material flow analysis, and creation of a detailed concept plan of the production floor .The concept plan divided the fab into a ‘transistor factory’ for FEOL (Front End Of Line) processes, and an ‘interconnect factory’ for BEOL (Back End Of Line) processes. The photolithography area, which is common to both ‘factories’, was positioned in the center of the fab.

This detailed concept was developed in conjunction with the designs for key building support spaces. For example, the placement of FEOL wet process areas was designed for close proximity to the DI water polishing station, chemical supply systems, waste treatment plant, and exhaust abatement yard. The BEOL processes were designed for proximity to the specialty gas supply and CMP slurry areas. Sub fab support areas were optimally placed to support specific tool types. In short, a great deal of planning was done to integrate the tools and manufacturing requirements into the facility design, and vice-versa.

The redirection to 300mm changed the entire tool set, but the basic processes and production flows remained the same, the only exceptions being driven by technology changes (i.e. additional CMP, copper). The relative locations of most process areas were retained, allowing the execution of the utility systems in each area to proceed along the original design guidelines, saving time and avoiding costly redesign.

In terms of detailed module layout, the larger size of 300mm tools permitted fewer tools per bay. Expanding on an existing concept where bays were designed in linked pairs, with a common operation and metrology cell, coupled with the higher relative output of the 300mm tools, accommodated the required number of 300mm tools without a significant redesign for most bays.

What Did Change

  • The FOUP Effect
  • Minienvironments
  • The Tools
  • Automation
  • Manufacturing
  • Technology

Changes for 300mm Capability

Most of the significant changes in DMOS6 were based on the effect of physical differences in 200mm vs 300mm tools, their factory interfaces (load ports), and the enhanced material handling automation scheme. The combined effects of these parameters required changes in the final design and structure of the cleanroom, as well as having significant influence on the details of the factory layout design. Changes in manufacturing protocol were implemented as a result of changes in the cleanroom requirements.

The FOUP

No other single factor had as far reaching implications as the implementation of the FOUP (Front Opening Unified Pod) in DMOS6. Driven by the inherit difficulty of safely handling large open cassettes, the FOUP provides a high level of protection to the wafers, while at the same time changes the face of the fab in many areas. The ‘FOUP Effect’ impacted tool design and operation, the layout detail design, the structure of the fab area, and the manufacturing protocol.

Tool Minienvironments

The use of minienvironments on all tools in DMOS6 was a step function technology change that altered the original cleanroom class philosophy. With tool minienvironment specs in the < Class 1 range and FOUPs used for all material transport, the wafers are isolated from the room environment at all times. With the elimination of the room environment as a source of particle contamination, the cleanroom classification was relaxed. The cleanroom spec changed from 1 particle /ft3 @ 0.1u (near ISO Class 1) in the bays and 300 particles /ft3 @.1u (near ISO Class 3) in the tool chases to a uniform 100 particles /ft3 @ .5u ( ISO Class 5). The reduced air flow requirements allowed a design change from 70% filter coverage to 50% filter coverage, and elimination of 20% of the room air circulation fans.

Another impact of implementing minienvironments was the elimination of all bay/chase separation walls and perimeter partition walls from the fab area There is no longer a need to differentiate or separate ‘operator’ from ‘maintenance’ areas. Elimination of walls in turn eliminates the need for tool bulkhead hardware, enhances space utilization, and eases operator and maintenance access around the tools. The ‘cost’ of minienvironments in terms of the cleanroom is their impact on the size and overall design of the tools.

The Tools

Size and Space

Early ‘requirements’ for 300mm process tools optimistically stated that the size of the tools would increase only slightly as a result of their 300mm capability. This has not been the case. Tool ‘sprawl’ is evident in many areas, and is even more pronounced with the addition of minienvironments, multiple FOUP load ports on every tool, and internal FOUP stockers on some tools (more ‘FOUP effect’). This is particularly noticeable on metrology tools, where the measurement unit – the business end of the tool – is often dwarfed by the FOUP loader and robot that services it. Additionally, front loaded tools have an operator interface location on the side of the tools so as to avoid safety/interference issues with AMHS delivery systems. Scale factors typically only comprehend the main tool footprint, and thus may be overly conservative. A tool represented as being 25% larger (wider/deeper) consumes (1.25 x 1.25)=1.56x as much floor space PLUS the FOUP interface and minienvironment.

Access aisle width in the fab came under review with the 300mm re-direction. Not only did the fab aisles need to accommodate moving in the larger tools, but with overhead automation, easement on the main aisles and in front of the tools was required for the necessary overhead tracks. Tool move-in in critical bays was modeled electronically to insure sufficient space.

The number of tool bays decreased by about one third as a result of larger tools, accommodation for FOUP loaders, and wider aisles to accommodate AMHS and tool movement. Many 300mm tools make better use of bottom feed utilities, however, giving some relief in terms of tool packing density. In addition, perimeter service requirements have not increased for most tools, and in many cases manufacturers are designing to require access on only two sides (front +back) or three sides, allowing for denser tool packing.

Table 1.

Weight

Larger tools usually = heavier tools. The fab raised floor system was reinforced in the main aisles to accommodate the load of heavier 300mm tools and subsystems, and additionally reinforced under many tools when installed. Custom-built steel support bases became the norm for many additional types of tools, based on both overall and point load weights.

Based on limited knowledge of what the final 300mm tool selections and specifications would be, the main elevator used to bring tools from the level 1 receiving dock to the level 2 support and level 3 main clean room was modified to a higher capacity. Although the original lifting capacity of 16,000 lbs. was sufficient for the heaviest 200mm tool component expected, the capacity was raised to 30,000 lbs. based on data from potential 300mm tool suppliers. The elevator interior dimensions did not change, however.

Full Fab Automation

DMOS6 was originally designed to implement a 200mm overhead track type interbay Automated Material Handling System (AMHS), with cassette stockers and operator access ports in each operational bay. This was to be linked to a single inter-level transport elevator to move material to a subfab test area. All material movement from the bay stockers to the tools was to be manual and all transport and storage was in open cassettes. The fab layout plan and structural design comprehended these elements.

Figure 4.

The re-direction to 300mm heralded a change from open 200mm cassette based wafer transport to 300mm FOUP based wafer transport. As the 25 wafer 300mm FOUP weight is beyond recommended ergonomic guidelines for repeated manual handling, transport of the FOUP from the stocker to the tool load port and back – intrabay automation – was implemented (more ‘FOUP effect’!). Coupled with a heavier duty interbay – stocker to stocker – automation system, FOUP transport is directly from tool load port to tool load port. The intrabay system chosen is an overhead transport (OHT), which allows better space management while at the same time allowing unobstructed floor access to the tools by operators and maintenance personnel.

Originally conceived as an ISO Class 1, bay and chase style ballroom, the FOUP/minienvironment scenario freed DMOS6 from being in one continuous clean space. The virtual cleanroom could now be as large as the scope of the expanded wafer transport system, with peripheral areas previously relegated to support functions being utilized for wafer processing. The construction of these new areas only needed to accommodate the requirements of the actual tools being installed there. In DMOS6, CMP tools which have no large external subsystems are located in a two-floor (and in the future three-floor) deep clean room space

The advent of full fab material handling automation mandated a microscopic look at the layout of each bay, its process flow, and how its tools were served by the automation system. Rule sets were developed early on to allow maximum flexibility in both automation system design and tool layout design, even before automation or process tool vendors were chosen. The hardware automation was not ‘tacked on’, rather it was ‘designed in’ based on 200mm experience, internal knowledge of AMHS systems, and a concept of how we wanted to execute the manufacturing operations of the fab.

The most pervasive changes in the facility were driven by the 300mm AMHS strategy. The physical and structural changes required for the facility were captured early in the project, allowing them to be designed into the cleanroom by the facilities team. This allowed execution with minimal compromise in most cases. Design decisions that could wait were delayed as long as practical to allow complete definition of the final AMHS system requirements. Some of the major issues addressed are as follows.

  • Ceiling height changes: Whereas the original cleanroom ceiling design height of 13’2” was more than adequate for a 200mm fab and interbay AMHS system, additional headroom was deemed necessary at 300mm to accommodate the taller payload and to exploit the possibility of two levels of automation. Within the completed building structure, the maximum cleanroom ceiling height attainable was 14’4”, which was sufficient for 300mm with ‘stacked’ or ‘crossed’ automation tracks.
  • Ceiling structural support changes: The original HEPA ceiling structure was reinforced to accommodate a 200mm, open cassette type interbay (bay to bay) AMHS system in the main aisles of the fab. For 300mm, moving the heavier and larger FOUP required upsized automation hardware, and adding intrabay (direct tool delivery) overhead transport required additional reinforcement be designed into the bay ceiling grid. Since final layout design and fine positioning of the tool bays was still in flux, all of the fab ceiling was reinforced to accept track loads. This provided current and future flexibility for automation track routing, and has proved to be a good decision.
  • Fab Growth: Although not a 300mm specific issue, the necessary addition of off-waffle (ex-main fab) wafer processing areas (i.e. CMP, wafer support) required the AMHS system to access locations out of the main fab building. These areas were challenging to material transport in that they involved changes in floor height, ceiling height and the requirement to use fire doors in horizontal and vertical transport system runs.
  • Interlevel transport: The inter-level transport lifter shaft provided for the 200mm system was 1) not the right size for 300mm hardware; and 2) not in the correct fab area. The existing shaft holes in the fab level and support level floors were filled in, and new holes were saw cut and framed in the concrete in the required multi-floor locations.
  • Layout design and implementation: In the 200mm design, AMHS was to be used strictly for interbay material movement, while actual wafer delivery to the process tools was to be handled by fab operators. Tool positioning and placement rationale was to place tools of like function together for manufacturing efficiency (Farm layout) and build bays in process order sequence, and secondary was the actual physical placement accuracy of each tool. A tool located at the end of an aisle might be positioned at 90 degrees in relation to other tools, making it easier for an operator to access. In other areas, cross aisles between bays might be populated with metrology tools, for convenient sharing. Modification of process bay layout was constrained only by available space, and a tool might be placed ‘out of position’ in order to get it installed at all. In short, tool placement was flexible and more free form, creating operational (and visual) chaos in certain areas over time.
Figure 5.

Figure 6.

When intrabay direct to tool automation is used, physical positioning of the tools becomes critical, and random or convenience location of metrology or any tool for that matter is out of the question. Each bay’s position and its tool layout requires enhanced modeling efforts, and final bay designs are coordinated with AMHS requirements. As the AMHS system must access all load ports, tool placement must adhere to rules in terms of

  • Tool alignment on the delivery aisle: Load port centerlines aligned with the automation track. (AMHS system requirement)
  • Tool to tool alignment of load port reference points: Load ports of all tools on a bay are in line with each other. (AMHS system requirement)
  • Tool pitch (spacing) on the delivery aisle: Mandated minimum distance between adjacent tool load ports. (AMHS system dependent)
  • Aisle widths: Sufficient for tool move in, and in conformance with safety guidelines for co-existence of operators, floor-based delivery vehicles, and automated handling systems. (Modeled in CAD)

Precision tool placement and positioning was performed by establishing a series of benchmark reference points in the fab, and developing and implementing use of precision laser alignment tools. A procedure and set of alignment fixtures were developed internally to allow alignment of all loadports in a bay to within 3mm of the standard reference point in each bay.

  • Reticle storage and delivery automation: Although not directly related to the wafer diameter change, the high projected throughput of the 300mm exposure tools and the probability of mixed product lots in a single FOUP pod placed demands on reticle availability best served by an automated delivery system. The reticle management system in DMOS6 controls the automated handling and transport of reticles. The reticles are stored in a series of distributed bare reticle stockers and transported from the stockers to the exposure tools in 150mm SMIF pods The operational goal is to deliver the reticles to the exposure tool ahead of need. A matrix of dedicated overhead delivery tracks added to the ceiling loading and drove vertical height requirements to allow crossing of reticle delivery tracks over the intrabay wafer delivery tracks.

Manufacturing Changes

Changes in manufacturing protocol and procedures in DMOS6 were mainly driven by implementation of FOUPs and tool minienvironments, and are a second order effect of the change to 300mm. As noted earlier, the FOUP isolates the product from the room environment and operators, while at the same time its size and weight preclude repetitive manual handling. The FOUP effect impacted manufacturing protocol as compared to TI’s 200mm fabs by:

  • Simplifying the fab gowning procedures and elimination of fab entry air showers.
  • Eliminated washing of gloved hands
  • Allowing ungowned workers in the subfab areas
  • Elimination of all manual single wafer handling, using wafer sorters instead.
  • Use of special transport carts for any necessary manual movement of loaded FOUPs
  • Eliminated manual handling and loading/unloading of reticles to the exposure tools.

The inability to manually handle individual wafers complicates test/dummy/filler wafer usage. Since all of these wafers still need to be in FOUPs, a dedicated area for cleaning and kitting non-production wafers was set up. This area, which is separated from the main fab operation, has the cleaning, metrology, and wafer sorters necessary to kit up a pod of wafers for tool qualification or other tests. Single wafer tests within the process flow are transported to the metrology station as a single wafer in a pod.

Figure 7.

Figure 8.

The FOUPs also require periodic cleaning, and possibly service in the event of eventual mechanical wear or damage. Since the FOUPs will be opened to the room environment for cleaning, the dedicated FOUP service area retained its Sub ISO Class 1 specification, as it did in the incoming wafer inspection area. These are the only areas where wafers and or the inside of the FOUP can be exposed to the room environment.

Technology Driven Changes

Technology changes in this context are time and process node driven. Although not specifically 300mm issues, each of these had some effect on the execution of DMOS6. During the hiatus in the DMOS6 fab completion, several processing technologies moved into the forefront, requiring a plan on how they should integrate into the manufacturing flow and thus alter the fab design and implementation. Some of the major technology changes and their impact are as follows:

a) CMP process expansion: In the intervening time between fab design and cleanroom build out, CMP planarization processing came into its own as a major process subset. Although the baseline 200mm fab had CMP capability designed in, the widespread use of CMP processing in newer technology nodes presented the single largest challenge in terms of space and infrastructure requirements. The amount of cleanroom space needed for the number of CMP and support tools required, along with the desire to keep CMP isolated, pushed the CMP area out of the main fab ‘envelope’ and into adjacent space on two different levels. This introduced issues with material transport and slurry delivery. Material transport was addressed by extending the main fab AMHS system into these areas. The addition of several different slurry types required the addition of a large slurry dispense room in close proximity to the tools.

b) Copper: Challenging the stability of the original layout concept, the shift to copper processing impacted the fab in several ways. An issue not only for the new tool set itself, but also for the added chemical dispense systems, and waste treatment systems to collect heavy metal waste streams. Additional safeguards were required in the layout and manufacturing protocol to address copper contamination issues (dedicated FOUPS/pilot wafer separation/ metrology tool management/wafer sorters/etc.).

c) DUV Lithography: Use of an increased proportion of Deep UV exposure systems presented challenges for placement and containment of the laser subsystems, along with other added support units for these tools. The lithography tools also exhibited one of the greatest weight gains of any tool category, prompting analysis of structural building loading in the densely packed photolithography area, and driving spacing of tools to some extent.

Vibration: Specs for vibration are same. DMOS6 was designed and built with close attention to vibration control, with an oversized area in the fab having a thicker waffle table for vibration control and stiffness under the photo and critical metrology tools.

Conclusion

The re-direction of DMOS6 to 300mm processing highlighted that, from a facility and infrastructure perspective, the move from 200mm to 300mm is more evolutionary than revolutionary. The facility changes were largely due to the physical attributes of the 300mm tools, the effects of the FOUP, and the automation system, and in a smaller way due to changes in technology. The flexibility designed into the original DMOS6 facility served it well in the re-direction of effort toward the 300mm goal. DMOS6 achieved its plan of first 300mm silicon out in September of 2001.

Going forward, the lessons learned from this first 300mm foray will have a significant influence on the next generation of fab design at TI. Changes in the cleanroom philosophy and requirements for 300mm will receive close study, to allow the facilities team to continue to design and construct flexible, cost effective production facilities

 
 
Search



Published By:
?¡Àcheap #167
38 Miller Ave.
Mill Valley, CA 94941
www.mazikmedia.com
converse@mazikmedia.com
Disclaimer | Privacy Policy