Abstract
The conventionally designed
300mm IC fabricator in an automated environment has many drawbacks such as:
high capital outlay and high cost of ownership, long fab
construction time, low flexibility for multi-product and multi-process
operation, long process cycle time and high work in process. These drawbacks have an enormous impact on
the economics of a 300mm fab. This paper proposes a new fab
architecture and operational philosophy to improve the economics of 300mm/450mm
fabs. The new fab architecture and design will reduce capital outlay and fab construction time by 50%, increase equipment
utilization by 100%, reduce COO by 30%, and reduce both process cycle time and
work in process by at least 30%, and cut annual operating expense by
30%. Integral to this new approach is
the cooperative alignment of IC makers and equipment suppliers throughout the
equipment and fab life cycle. By closely working together, both will reap
the economic rewards.
Keywords
IC Manufacturing Economics IC Fabricator Architecture IC Fab Design Semiconductor Equipment Design Cost of Ownership IC Fab Operations
1.
Introduction
There are nine 300mm fabs in existence in the world today. These fabs cost
billions of dollars to build; it is therefore imperative for these fabs to run constantly and smoothly to stay ahead of their
depreciation curve. This means that a
300mm fab must be reliable, flexible, and fully
automated requiring minimum human intervention.
Yet today’s 300mm fabs are not necessarily
more reliable or flexible than 200mm fabs, because
the 300mm fabs employs a similar equipment and fab design as that of 200mm fabs,
only larger in size. In addition, the
automation of 300mm fabs also requires more manpower
in order to maintain both equipment and the systems in operation. A staffing level of 700 plus for a 300mm fab is not unusual.
In an environment of ever changing market demand versus the capacity
planning of IC manufacturers/foundries, both the IC manufacturers and the
equipment suppliers are paying high prices caused by the mismatch of forecast
versus actual capacity demand. In the
fourth quarter of 2001 there was 40% more fab
capacity than actual demand, which means many fabs
are running way behind their depreciation curve. This paper proposes a new innovative fab design and operational approach for 300mm and 450mm IC fabs to mitigate the above situation. It calls for close cooperation among IC
manufacturers and equipment suppliers to literally think outside of the box.
The payoff is great. A joint innovative effort could reduce capital outlay and fab
construction time by 50%, increase equipment utilization by 100%, reduce COO by
30%, reduce both process cycle time and work in process by at least 30%, cutting annual operating costs by 30%,
By standardizing common equipment and facility interfaces as well as physical
and software protocols, the lifetime of a fab may be
extended from the 300mm to the 450mm generation. This innovative approach will
enable the semiconductor industry to adapt to future steep demand variations in
the world market place.
2. Conventional 300mm Fab Drawbacks
The conventional 300mm fab design is similar to that of a 200mm fab with an automated material handling system (AMHS)
housed in an enlarged ballroom encased in a 3 or 4 level structure. The drawbacks are:
High Capital Cost: Unless sufficient clean room space and facilities are constructed for
the current and expected future capacity demands of the fab
initially, it will be costly to add additional capacity afterwards. That means extremely high capital outlay for fab construction to the tune of $1.5B to $2.0B.
Long Fab Construction Time: The fab
building time ranges from 18 to 24 months because of the sequential nature of
building construction, facilities and equipment installation activities.
Low Flexibility: The physical layout in the clean room area constrains the fab flexibility for multi-product and multi-process
operation. The slow AMHS speed caused by vibration and contamination concerns
in the ballroom also limits fab flexibility. Any
modifications in equipment or adaptation of new process would risk the
interruption of fab operation.
Long Cycle Time and High WIP: Short MTTF and long MTTR of semiconductor
equipment together with the slow AMHS speed cause a long production cycle time
and a large amount of work in process (WIP) stored in expensive stockers.
High COO: Semiconductor equipment by its nature has a short lifetime and is
subject to the fast change of technologies; it is difficult for their
reliability to equal that of mature industrial equipment. This intrinsic low reliability of
semiconductor equipment contributes to the low average fab
equipment utilization around 35% in 200mm fabs, and
the average equipment utilization of a conventional 300mm fab
is not likely to exceed that of the 200mm ones.
Short Fab Life Cycle: Despite the great progress
made in SEMI standardization efforts in 300mm equipment and fab
architecture, it is difficult to design a conventional fab
to cover both 300mm and 450mm generation wafer sizes in a single fab’s lifetime.
- Proposed
Innovative Fab Design
The proposed new
IC fab design consists of hundreds of transportable,
environmentally controlled process modules placed on a fab
superstructure with a facilities supply system installed in the superstructure
instead of the conventional brick and mortar fab
building. The process modules are
connected by an independent automated rapid material transit system. Figure 1. shows an overview of a fab
designed with the new architecture. A pancake shaped superstructure houses
hundreds of rectangular shaped process modules and five dome-shaped control
centers for accommodating operating personnel. The modular units are arranged
in 3 concentric rings in the super-structure.
The pancake shaped super-structure shown in Figure 2 serves as the main fab structure for foundation and facility support of the
whole IC fab. The construction of this
super-structure and the facility support system can be built on-site parallel
to the construction of the process modules offsite at equipment supplier
locations. The two construction activities do not interfere each other as often
happens in the case of the conventionally designed fabs. This will cut overall construction time from
the present minimum of 18 months to 9 months.
It means the new Fab will bring IC product to
the market 9 months earlier than a conventional fab.
Figure 1: New Architecture Fab Design
Overview
Figure 2: Super-structure
of the New Fab Design
Figure 3 shows the
independent rapid material transport system (which consists of four monorail
rings) sandwiched between rings of process modules on the super-structure.
Figure 4 shows one of the many multi-pod carriers mounted on the monorail. This rapid transportation system moves wafer
pods (FOUPs) among process modules. It is like the inter-bay transportation of a
conventional 300mm fab. Because this rapid material transport system
is built on foundations separated from the main fab
super- structure, it can operate at high speed to achieve short cycle time
without generating vibrations in the modular process areas. The fast inter-module transportation combined
with the granular modular manufacturing unit form a powerful, flexible IC Fab that meets the requirements for multi-product and
multi-process operation. The fab also can handle fast process routing changes as well as
process equipment modifications to extend fab
life.
Monorails
Figure 3: Rapid material transport
system (4 rings of monorail on their support posts)
Environ’t chamber FOUPS
Figure 4: A monorail multi-pod
carrier
Figure 5 shows the schematic of a process module; its profile resembles the conventional 3-level fab design. Each module is made of an upper sub-module and a lower sub-module connected at the equipment floor level. The upper sub-module consists of fan-filter units for environment control; an overhead material transportation serves as an intra-module (intra-bay) wafer or FOUP mover. The bottom floor of the upper sub-module is made of standard clean room floor plates placed on steel beams. Each process module can accept 3 or 4 pieces of processing or metrology equipment. This upper sub-module is removable by disconnecting the supply/facility piping through quick connectors from the lower sub-module for fast replacement of a calibrated standby or modified upper sub-module. This design feature is to shorten fab downtime due to major equipment repair or modification. The lower sub-module contains the supply lines (gas, chemical, water, and electricity) and support equipment such as pumps or electrical boxes. They are semi-permanently connected to the facility supply network housed by the fab super-structure. The unit construction cost of the manufacturing module is estimated to be 2/3 of that of a conventional IC Fab of equal floor space with the same environment clean class. The operating energy cost of the new IC Fab design is estimated to be 2/3 of that of a conventional fab of the same production capacity because of the smaller clean modular manufacturing area required. The transportable process modules are designed to take advantage of the well-developed worldwide container transportation infrastructures. The manufacturing module itself is both the equipment-shipping containers as well as the manufacturing environment in which the equipment is housed. They can be constructed and loaded at the equipment supplier’s location for shipment to the Fab construction site, and can be placed on the super-structure connected to the facility system. It saves equipment installation and calibration time as well as shipping costs. Prototype process modules have been built and tested. Results show that the modules can be installed on a fab within 5 hours [1]. Fully loaded modules of a similar design were shipped across the ocean to their fab site; and all installed equipment in those modules started functioning properly when powered on in their fab [2].
Facility supply system Interface to monorail Upper
sub-module Lower sub-module
Figure 5:Schematic of a transportable process module
4. Advantages of the New Fab Design
The proposed new fab
design with transportable, manufacturing modules addresses most of the
drawbacks of the conventional 300mm fab with economical
advantages, yet the new fab design requires no
scheduled technology invention. All
employed technologies are available with proven industrial reliability and
infrastructure. The new fab design advantages address each of the conventional
300mm fab drawbacks as follows:
High Capital Cost Reduction: The new fab design enables both
the fab superstructure including facilities and the
manufacturing modules to be constructed in parallel. The number of modules for
the fab can be tailored for the necessary capacity
required by the market by the time of installation of the modules at the fab site. The new fab design and
operation paradigm provides leasing options that may not have previously
existed. Instead of owning the equipment outright, which accounts for more than
70% of a fab’s cost, the IC Maker can lease the
equipment and sub-contract total maintenance of the modules and equipments to
their respective supplier. Thus the savings on both initial capital investment
and interest expenses can easily be reduced by 50% with respect to that of
conventional 300mm fab. By properly structuring the
charges there will be a steady flow of income from equipment leasing and
contracted maintenance for the suppliers to pursue profitable new product
development and production. It is a win-win proposal for both the makers and
suppliers.
Reduction of Long Fab Construction Time: The conventional 300mm fab building time ranges from 18 to 24 months due to the
sequential nature of fab construction. The new fab design reduces construction time by 9 to 12 months
through parallel processing of its building activities. The construction of
manufacturing modules and the building of infrastructure and facilities are
progressing simultaneously without interference from each other. In addition to
capital savings, the Maker reaps a huge economic and strategic advantage by
getting his product to the market 9 to 12 months ahead of his competitors.
Improved Fab Flexibility: The new modularized fab with high speed AMHS provides an ideal agile
manufacturing system for multi-product and multi-process operation. Any substantial
modifications in equipment or adaptations of a new process will only affect one
module without risking interruption of the entire fab
operation. The modular architecture also reduces the risks of environmental,
health and safety exposures of a conventional fab.
Furthermore, R/D works of new process/equipment can be completed in a real
manufacturing module offsite. After completion of R/D, the module can be
delivered to the site and incorporated into the fab.
For example, the CMP environment contamination concerns present during
adaptation to a conventional fab would be greatly
reduced by the modular fab architecture.
Short Cycle Time and Low WIP: The combination of rapid independent AMHS among
modules and the short intra-module wafer mover facilities within each module
provides the new fab with a high speed multi-path
AMHS. Optimized product routings from the fab
manufacturing execution system can easily be carried out by the multi-path
agile AMHS. Short MTTR is achieved by
replacing out of order module with calibrated standby unit. This enables the fab
to run with just local buffer storage on equipment instead of large costly
stockers. Since up to 90% of the wafer
cycle is in stocker waiting time, the elimination of stockers with fast,
flexible AMHS and local equipment buffers will cut production cycle time by at
least 30%. Low WIP is a natural
consequence of shorter cycle time and the elimination of large expensive stockers
in a conventional fab.
Low COO and Fab Operation Expenses: The new fab design replaces the traditional brick and mortar fab with modules on a superstructure; its estimated cost is
2/3 of that of a conventional fab with the same
capacity. The new fab
design also recognizes the intrinsic limited nature of semiconductor equipment
by not requesting unreachable equipment reliability in its planning. Instead,
the fab’s modular, quick replacement capability will
improve average fab equipment utilization from 35% to
70%. By leasing process modules
populated with equipment and sub-contracting out equipment maintenance to the
suppliers, the IC maker does not have to hire an expansive team of equipment
maintenance experts. All these
advantages result in the reduction of COO by 30%. The new 300mm equipment is highly automated
requiring minimum human intervention to the operation of equipment installed in
process modules. By eliminating the main
source of contamination resulting from human operation the clean environment
inside the module will need only very little energy to maintain its
cleanness. The overall fab energy cost can easily be reduced by 30% from that of
an equivalent conventional ballroom fab.
Long Fab Life Cycle: In the new fab
design the sub-systems like inter-module AMHS and facilities support can be
designed to accommodate both 300mm and 450mm wafers with small incremental fab costs. The
manufacturing modules can also be designed to accommodate both wafer sizes at a
moderate cost. The key to a successful
implementation of a long life fab design is to
accelerate the great progress made in SEMI standardization activities in 300mm
wafer fabs.
Not only should this activity be expanded to both 300/450mm wafers, but
the hardware and software interfaces for equipment and facilities should also
be agreed upon and adopted by all suppliers.
Most large wafer process steps will be done in single wafer chamber; a
family of standard facility backbones can be designed to accommodate both size
wafers with similar physical interfaces and reusable supporting software. Because of the decentralized design and
building process, it is relatively easy to implement a long fab
life strategy in a modular fab, compared to designing
for both 300mm and 450mm wafer sizes in a conventional 300mm fab environment. If
a fab’s life can be extended to two generations of
wafer size, the cost of manufacturing would be reduced by almost half.
All the drawbacks of a conventional 300mm fab mentioned in section 2 have been addressed with
substantial economical savings by the proposed new fab
design and operation philosophy.
5. Discussions
In the era of 300mm and 450mm Silicon wafer
manufacturing large process equipment manufacturers will play a pivotal role in
the standardization of process equipment and its interface design. This in
turn, will affect the fab design. The users are left to tailor their individual
process variations through operational parameters. Standardization in equipment
does not diminish industrial competition, since more and more semiconductor
product differentiation will be based mainly on product design. A winner product has the best intellectual
design content to meet the market demand with the most economical manufacturing
cost. The 30% to 50% manufacturing cost
advantage offered by the new fab design certainly is
a competitive tool for any IC maker. Now
is the time for IC makers and equipment suppliers to think outside of the box
by jointly working out a coherent program to implement the new fab concept based on SEMI standardization to assure the
semiconductor industry a continuous growth along
Figure 6: A Transportable Semiconductor Processing Module on ITRI campus, Hsinchu,
Taiwan
6. Conclusions
This paper proposes a new fab
architecture and operational philosophy to improve the economics of 300mm/450mm
fabs. The new fab architecture and design will reduce capital outlay and fab construction time by 50%, increase equipment
utilization by 100%, reduce COO by 30%, and the reduction of both process cycle
time and work in process by at least 30%, cutting annual operating expense by
30%. Early feasibility studies showed
promising results. This new fab design and its new
operational approach require both the IC makers and the equipment suppliers to
closely cooperate throughout the equipment and fab
life cycle to reap mutual economic benefits.
It is believed that this fab design will
assure the world semiconductor industry a continuous growth along the Moore’s
curve without steep upheavals resulting from demand and capacity mismatch.
References
[1]. Bevan P. Wu: Modular
Clean Room for Effective Integrated Circuit Manufacturing – MIRL/ERL Joint
Accomplishment, ITRI Today, No. 16, Spring (1999) pp. 1-4, Industrial Technology
Research Institute, Hsinchu, Taiwan, R.O. C.
[2]. Private communications
[3]. Kishore
Potti: Survey of 300mm Wafer Fab
Designs, Semiconductor Fabtech, Edition 6, May (1997) pp. 101-106
**Selected
pages of No. 18 - Wu, B. "The First Automated Semiconductor Manufacturing System in
the Industry - The FMS Feasibility Line"
IBM Research Report RC 14203, November 28, 1988
