The reasonably good status of 300mm wafer-processing tools
10/01/2000
SPECIAL REPORT: State-of-the-art manufacturing
James A. Irwin, Irwin Consulting, Austin, Texas
overview
It seems apparent that the previous delay in 300mm plus the work of I300I and Selete have done a lot to ensure the existence of an adequate 300mm toolset, at least one that is better than previous industry transitions to larger wafers. Experts at Selete, for example, foresee 100% tool compliance, up from recent 85% value evaluations, when these fabs begin coming on line in early 2001. The remaining minor and isolated problems are mostly related to 300mm's need for full automation and individual user propensity to specify special needs. What the industry is learning in this process is that, for the future, industrial engineering will become a crucial element in further advancing semiconductor manufacturing.
By many accounts, wafer process technology is ready to take on those tasks assigned to the 300mm toolset. At the recent Semicon/West 2000, one could see a huge availability of very good-looking solutions for 300mm.
 Winbond Electronics Corp.'s fab in Hsinchu, Taiwan uses 300mm tools. Courtesy Winbond. |
In one of the most truly amazing announcements for any semiconductor equipment tradeshow, Applied Materials introduced 21 new products with approximately 80 applications, all with 300mm capability and all reportedly extendible to the 100nm technology node. Dan Hutcheson, president of VLSI Research, San Jose, CA, says, "Company officials rightfully expect to get huge orders from every 300mm fab to be built, justifying something over an estimated $2 billion invested in 300mm tools technology so far."
But is an array of individual process tools enough in itself? Larger questions are: Will the tools integrate smoothly enough to fulfill industry requirements, especially in the area of automated, hands-off operation? Will the amazingly high throughput claims of some tools be met? If so, will this throw the rest of an automated fab processing line so far out of balance that overall economic power is lost?
Fortunately, the term "process" is finally beginning — more holistically — to include materials-flow control, wafers, people, and information throughout the factory to a far greater extent than ever before. In addition, "process defect" is beginning to include any glitch in these flows and any flaws in circuit element formation processes. But these changes in semantics mean that process tools must fit into a much larger context to be characterized for process readiness. This context includes mechanical and logical (i.e., software) interfaceability and interoperability with the rest of the fab system.
Various Semi standards identify and specify many of these needs (see table). But some questions remain to be answered:
- How widely are these standards being met?
- How effectively will the standards ease the integration of these tools into the 300mm fabs under construction?
At a recent Semi STEP program, Chris Brandson, an integration engineer at Brooks Automation, said, "There is a whole mess of new standards; don't underestimate how many. And the scope is totally different and much greater than GEM (E-30)." Brandson has actually measured the height of software standards documentation and found that 300mm documents are five times thicker than 200mm documents [1].
At the same program, chairperson Margaret Pratt of International Sematech, said, "[For previous fabs] SECS-GEM was sold as an equipment attribute, but because it was often not implemented at installation, it was not crucially shaken out. 300mm's need for automation changes all that. The success of 300mm [fabs] is crucially linked to application of the standards" [1].
 Brooks Automation's Gemini 6000 Express cluster tool with the ability to process 200 and 300mm wafers. |
"In addition, the standards are still changing and will for a number of years," Brandson said. Consider just one example in the complex of standards, as reported by Erik Seidel, an integration engineer at Intel Corp [1]. "From early Intel results from integration of 300mm overhead hoist vehicles and passive equipment, we found that the E84 standard could be misunderstood. We experienced problems during the single carrier handoff sequence where the active equipment changed the state of more than one signal at a time." This one problem resulted in the submission of five line items to clarify the problem in the E84 standard. "The message here," said Seidel, "is that equipment supplier resources must be E84 knowledgeable, and field service engineers need to be sufficiently trained on E84 to ensure that passive equipment is properly configured."
Pretty good shape
While there are still troublesome caveats, as several notable industry watchers indicate below, there are strengths to the 300mm toolset and its automation. The overall view is that compared to the "gunslinger" approaches to the industry's 2-3 in., 4 in., and 5 in. conversions — even the conversion to 200mm — the pending leap to 300mm is in better shape for success by orders of magnitude.
 |
According to an analysis by Hutcheson, "There are no impossible bottlenecks to overcome to prevent a company building a 300mm fab these days. A complete set of tools is available, the processes are there, and the automation is capable. Proof of this is the fact that companies are finally building fabs. It is not just Infineon acting as the pioneer. UMC, TSMC, Intel, and Samsung are all well into the 300mm club now."
The biggest issue that most of these semiconductor manufacturers have is the lack of breadth in tool choice. "This is due partially to the slowing of 300mm tool development in the recent downturn and to the fact that 300mm tool development prior to the slowdown was always targeted one generation behind the state-of-the-art," says Hutcheson.
Since 1998, the equipment industry has been focused on "bridge tools," tools that represent cutting-edge technology and tools that are getting field experience with 200mm. An added advantage is that chipmakers buying these tools for 200mm will get longer productive lives because they are leading-edge and they can be upgraded without much difficulty.
Today's 300mm toolset is also different in another regard. Hutcheson notes that "it is gratifyingly that Applied Materials is selling information along with process in its tools. For example, integrated metrology monitors how a tool is running and helps ... achieve higher overall equipment effectiveness. Such information helps to greatly implement the latest process technology, but, perhaps most importantly, should give customers some comfort with their decisions."
Seemingly, good roots
Paul Patruno, who served as an I300I director of technology and is an ST Microelectronics assignee to International Sematech, applauds the laying of foundations for 300mm wafer manufacturing by the early I300I and Selete programs, which date back to 1996. "Through various joint activities, these consortia gave the industry demonstration and test methodologies (e.g., starting with the "Equipment Performance Metrics for 0.25mm Technology"), guidelines and standards, silicon wafers, automation technology, and metrology and process equipment evaluations and improvement," he says.
Out of these efforts, a comparison of Selete and I300I maturity indicators showed quite similar results, despite different approaches in the calculation methods used. "Overall, by the end of 1999, good estimates by the end of the I300I and Selete Phase I programs demonstrated an 80% to 85% figure in process performance and productivity capability for 300mm tools compared to the guidelines; the figures were based on the jointly published Unified Metrics for 180nm Technology," Patruno says.
(We must hope that this does not follow a corollary to the Pareto principle where the last 20% of the gain requires 80% of the work.)
Recently Selete and International Sematech published an update, the 1999 Unified Equipment Performance Metrics for 130nm Technology [2]. This document is a very solid piece of information, detailing desired 300mm equipment characteristics, process, and productivity parameters for 40 tool types capable of supporting a full 130nm-process flow for DRAM and logic devices.
There is also CIM Joint Global Guidance, a consensus document developed jointly by International Sematech and Japan's EIAJ 300mm group, J300E. This document sets international agreements on high level requirements for AMHS and production equipment capabilities needed for 300mm fabs [3].
A continuing effort by International Sematech and Selete Phase II programs in 300mm equipment demonstration and "tool hardening" programs and related automation programs will try to bring those indicators up close to 100%. "Part of the work is planned to be done jointly starting in late 2000 and could meet the target along with early 300mm pilot line and production fabs starting up in 2002," says Patruno.
The value of the various documents for 300mm-tool characterization can be clearly seen in the tool evaluations taking place in Japan, as recently reported by Selete's Hisakazu Miyatake [4]. "Selete is using the metrics indicator (MI) to describe tool maturity evaluation result; this is a simple ratio of performance values obtained in our evaluations to the target specification sets for process and productivity performances." At Selete, the overall average MI for 180nm process performance reached over 80% in March 2000 (Fig. 1). Beyond this the organization is well into an evaluation of 300mm tools for the 130nm technology node. "We see 300mm fabrication lines being aggressively installed for 130nm or even smaller process levels. The purpose of our 'new program,' which we began in April 2000, is to evaluate and accelerate 300mm tool development for 130nm technology and the fabrication line infrastructure," he said.
Remaining issues
According to Patruno and other watchers of the 300mm toolset, there are a few issues that still need "a bit more attention" by the entire semiconductor community before 300mm fully arrives. Consider yield, for example. There is a rather widely ignored attribute from the Equipment Performance Metrics that specifies a not-so-aggressive backside contamination level of 200 0.2mm-particles/wafer. Patruno notes, "This requirement has been, rather dangerously, largely neglected by equipment suppliers in their designs and tests. Corrective effort must happen immediately so as to not jeopardize the very fine backside polishing benefit associated with 300mm-wafer fabrication.
 Figure 1. Selete's evaluation, to "metrics indicators," of 300mm tools for 180nm-production processing [4]. |
Another issue involves productivity. The creation of Semi E79 (Standard for the Measurement of Equipment Productivity) established OEE as the standard high-level metric. "While it was widely promoted by Sematech, it was not used and understood the same way by all users," says Patruno. He explains that E79 was recently revised and supplemented with intrinsic equipment efficiency (IEE, see Fig. 2) and with valuable variations on OEE, as well as a crisp and applicable definition of OEE for cluster tools. All of this was aimed at providing suppliers and users with the tools needed to measure productivity with the same rule and to derive from the results gaps to fill and priorities to address. Moreover, by using the more detailed approaches for calculating OEE and its component factors, tracking or monitoring software packages has become available to better measure OEE and IEE and track down sources and causes of tool inefficiencies and utilization inefficiencies.
Availability, operational, quality and rate efficiencies have been thoroughly explained and standardized in the revised E79, and examples are given to guide users in applying these methods. "Widespread early adoption of E79 will give chipmakers and equipment suppliers a common language with which to discuss equipment productivity issues and smooth everyone's comfort with future standards for measuring fab productivity," says Patruno.
 Figure 2. Intrinsic equipment efficiency is a recent addition to OEE that helps communications between supplier and user. (Source: International Sematech) |
Tool automation
One of the crucial ongoing actions concerning 300mm-tool readiness is centered on standards. The guideline compliances for tool automation, FOUP interface and load ports, stockers, and equipment and automated material handling system (AMHS) communications and control are all a very important part of the needed fab productivity improvement associated with 300mm. This whole area is identified loosely by activity in computer integrated manufacturing (CIM), which is continuing at International Sematech and Selete, again in a joint effort to minimize risks and augment the impact of the work across the industry [3].
 Semitool's LT-308 300mm FOUP. |
Here, the key to equipment certification is compliance with guidelines in all aspects, including environment, safety and health (ESH) issues. More activity is likely to continue in this area, led by the consortia and fully supported by their member companies, as the industry is switching from 200mm to 300mm as fast as possible.
Patruno says, "High levels of intelligent utilization of all resources is needed and bridge tools as they are defined are offering a viable path for the upgrade to 300mm, allowing early test and utilization of very similar equipment in existing 200mm fabs and starting 300mm fabs, with easy on-site conversion. This concept is receiving wide support from users as well as from most suppliers, who are able to offset cost through a quicker and broader new generation of equipment introduction to their respective markets."
Patruno says, "In the early days of the 300mm effort, a target has been set for the equipment suppliers to prepare for new equipment having a (ratio relative to 200mm) 1x footprint associated to 1x throughput and 1.3x capital cost. This combination has proven to be achievable and viable as shown by the results of I300I and Selete demo programs. This and bridge tools will certainly help us meet the vision of a profitable conversion for the semiconductor industry."
 Novellus Systems' INOVA xT 300mm metalization system. |
Within tools
Within 300mm tools there exists another whole universe of opportunities for success — and for problems. "Consider, for example, scheduling issues, within the material control system software," says Norm Powell, strategic business manager for OEM sales at Brooks Automation. Short interval scheduling of wafer lot transfer is a big issue for the providers of material control systems software. The heavier FOUPs require automated lot handling, resulting in much more complex intrabay AMHS routing issues. In addition, newer high-throughput tools require more frequent visits by the AMHS. This results in a system of greater complexity, requiring faster and more frequent servicing, compounding material control software system issues.
According to Powell, "There are also issues inside supplier companies in responding to customer special requests by competitors and customers alike. These are taking just enough liberties with standards to prevent plug-and-play substitution of otherwise identical modules, such as FOUP loaders and unloaders. Each fab has its own spin on what is right. Thus, they will depart from standards wherever they feel that their internal requirements are justified. This takes away some of the 'costs savings' intent of standardization." As an example, Powell cites a 300mm fab start-up in Asia that has specified very detailed changes to the FOUP loadport that are not specified in — or disagree with — the standards. "This ends up being a customer specific configuration that adds to logistical and incremental costs," he says.
In addition, customers are creating new (and, of course, variable) requirements for wafer edge gripping end-effectors, creating some new design documentation and tracking challenges for suppliers of intra-tool wafer-handling automation. "Such 'specials' are well recognized as cost risers, but are less appreciated for the cycle time hits and increased rate of mistakes they cause by being different as they travel through tool manufacturing operations," Powell says. "This is a development cost impact and will exist as a higher cost option unless it is standardized."
Consider also that nitrogen consumption could become a serious issue in a fab because many users are requiring nitrogen-purged handling (FOUPs, etc.), seemingly without counting cost. When the costs and logistics of providing this much nitrogen are rolled up fab-wide, this may become problematic. Suitable solutions may have to be created quickly. Powell says, "Nitrogen usage is one of the top consumables costs in any semiconductor factory operations. Sometimes, fear of an unknown leads to an overly aggressive design consideration, for example, nitrogen purging, without fully understanding the impact to yield vs. cost. It drives an automation configuration with the additional costs, without an acceptance of that cost to the delivered product."
Material ID standardization
With efforts to deliver 300mm equipment at Brooks Automation, Powell sees three types of material identification sensors being specified and used: infrared, barcode, and radio frequency (RF). "Their selection and application is not even standardized between fabs of some user companies. This creates different though similar systems at various locations, and enterprise-wide documentation, tracking, and maintenance problem," he says.
 KLA-Tencor's 8400 series 300mm in-line CD SEM tool. |
Powell explains that for something like the 300mm FOUP loadport, the end-user-specified material ID system can create logistical issues for an OEM. "When an end-user specifies a 'non-standards' compliant material ID solution, it requires the end-user to be intimately involved in an integration issue with the OEM and loadport supplier. This cost of integration is hardly ever totally recouped in the sale of a supposedly "commodity" product such as the loadport."
Powell sees a more subtle issue, too. Some bridge tools may not have been designed to extend below 180nm processing. Therefore, some processes designed for 300mm tools may not be implementable in these bridge tools, thus leaving a gap in the migration path. Powell's expectation, then, is that some 300mm fabs will end up costing more than originally estimated while these remaining issues are sorted out in the market place.
In another area, according to Peter Gise, OEM systems marketing manager at PRI Automation, "The FOUP/FIMS Interoperability Study Group recently released its recommendations for the process tool to AMHS interface, and 'compliance' now has taken on a new meaning." Loadport front-opening interface mechanical standard (FIMS, E62) and FOUPs will now have to be more forgiving in the sense of being able to operate with total interchangeability. This means a change in the way vendors of these products approach their designs. Compliance to the Semi standards in the strictest sense may not ensure success. Flexibility and being able to operate with different automation components will be the deciding factors.
Gise says, "Turn-key tool automation solutions will be needed by semiconductor equipment manufacturers who provide a highly efficient bridge from the process tool to the fab floor, while addressing major cost concerns. For example, an integrated front end, when combined with a factory automation system from the same vendor, provides a solution to the semiconductor manufacturer's concerns of integrating process tools to the transport system of a fully automated 300mm factory. Fab managers worry about every interface in the factory, and a fully automated 300mm factory will add more than 200 process tool interfaces. It is essential that these complex interfaces operate seamlessly and efficiently."
Conclusion
Attendees saw many 300mm tools at Semicon/West, most designed for processing down to <100nm with many bridge platform strategies. Today, most process development is being done on 300mm tools. Powell says, "In the first round of 300mm tools, the process capability was fairly focused on sub-0.25µm processing. With the delay of 300mm, the requirement for this process capability for 200mm was then thrust on the advanced 300mm platforms, because that process capability was not in existing 200mm process tools. Suppliers had to scramble to put this 200mm wafer and process handling capability into what was originally a 300mm platform. This created "bridge tools" that are a 300mm-designed platform used for advanced processing of 200mm wafers."
All in all, the industry seems to be arguably much more ready for the 300mm wafer-size generation than ever before. In years past, it was up to one company to take the lead and develop tools and processes for the next wafer size. The kinds of issues we are addressing now suggest a maturing manufacturing industry, with a new set of priorities.
From the highly referenced Competitive Semiconductor Manufacturing Program directed by Professor Rob Leachman at UC Berkeley, which is a comprehensive study of leading fab economic issues, we can extract three meaningful average economic impacts (i.e., dollar loss/wafer) that a semiconductor manufacturing operation stands to suffer for each day's delay in various operation categories [5]. In other words, these are the costs for one day's delay for 100% yielding wafer over the entire life of a fab (assumptions include 200mm wafers and an initial sales price of $10,000/100% yielding wafer, declining 25%/yr). A one-day delay in:
- time to market (i.e., fab construction, process development, and qualification time) equals $3.44 loss/wafer,
- manufacturing cycle time equals a $3.04 loss/wafer, and
- yield ramp time equals a $1.35 loss/wafer.
How similar would be the results of a similar study of the equipment industry? How much more cost-effective and profitable would be the equipment business if design, introduction, ramp-up, etc. cycle times were drastically reduced?
It is clear, then, that companies in the race to build 300mm fabs have a lot to gain by engineering out, as much as possible, anything that creates a delay. They already knew enough to wait for a suitable set of process tools. Now it remains to be seen how efficiently these can be integrated into a production engine.
As semiconductor fabrication matures further, it will become ever more necessary to address productivity issues with manufacturing science and industrial engineering disciplines, more than the traditional image shrinks and yield improvements of old. The high productivity fab of the future will concentrate on the flows of all resources (i.e., wafers, materials, information, and operators) and the minimization of total cycle time and inventories to become a winner in the market.
References
- From Semi STEP program: Semi 300mm Software Standards, Semicon/West 2000.
- See http:www.sematech.org/public
esources/300mm/index.htm. - Version 5 of the CIM GJG document is available at the Sematech web site www.sematech.org/public/docubase/document/3534deng.pdf.
- H. Miyatake, Selete overview of 300mm program (Phase 1) and new programs, presented at Semi Technical Programs, Semicon/West 2000.
- STEP: Quantifying Equipment Productivity - The Semi E79 Standard, July 10, 2000, Semi, Rob Leachman's talk, "Understanding Fab Economics." Also see www.euler.berkeley.edu/esrc/csm.
James Irwin received his BS from Texas A&M University in 1969, and continued his education at Stanford, UC Berkeley, and Sematech through short courses in process and automation technology, project management, TQM, TPM, and related topics. He has been employed by AMD, GCA, IBM, Sematech, Texas Instruments, Ultratech Stepper and Veeco Instruments. Today, Irwin is the principal at Irwin Consulting, Austin, TX; ph 512/502-0797, e-mail [email protected].
300mm Progress Report
300mm fabs: who's building?
After being sidelined a couple of years, the shift to 300mm wafers is in full swing. According to plans announced by several of the world's leading IC producers, mass fabrication on the larger wafers should start before the end of 2001.
Renewed interest in 300mm has been driven by several factors, including better economic times in the IC industry, improved toolset productivity, and the success of Semiconductor300, an R&D-pilot fab based in Dresden, Germany, that is jointly operated by Infineon and Motorola and which has reached many milestones since it began 300mm operations in 1998.
In April 2000, Semiconductor300 announced that it achieved 90% production yields on 64Mb DRAMs. With those yields, Infineon said it was likely that it could cut the per-die cost of 64Mb DRAM production by 30% compared to similar memory densities produced on 200mm wafers. That 30% mark is the target that manufacturers have been striving to reach in order to make 300mm technology a cost-effective alternative to 200mm.
Before the end of 2000, Semiconductor300 will no longer be the only 300mm fab in operation. Intel recently opened a pilot fab in Hillsboro, Oregon, and TSMC expects to start up a pilot fab in Tainan, Taiwan, in the second half of the year.
The first volume production 300mm fabs will be opened by TI, UMC, and TSMC in the second half of 2001. At least seven more 300mm fabs should be in volume production by the end of 2002. Companies operating these fabs include Trecenti (a joint venture of UMC and Hitachi), Infineon, ST/Philips, TSMC, Intel, Samsung, ProMOS (a joint venture of Infineon and Mosel Vitelic), NEC-Hitachi Memory, and Powerchip Semiconductor (a joint venture of Vanguard, Mitsubishi, and others). Several additional 300mm fabs are rumored to open in 2002 and 2003, but are still in the planning stage.
Among the first 300mm wafer users, most are either manufacturers of high-volume, standard-type devices such as DRAMs, DSPs and MPUs, or foundry service providers. Only these types of companies can fill the large wafers with enough chips and keep the fabs running at a high enough utilization rate to afford the $1.5-$3.0 billion needed to build a 300mm fab.
Noticeably absent from the list are many of the biggest Japanese IC producers. Having been stung badly by the industry's downturn in 1996-1998, the Japanese are taking a "wait-and-see" approach to 300mm wafers. Those that are making an early move to the larger wafers are doing so through joint venture partnerships in order to share the cost and risk burden.
The recent strength in the IC market has enabled manufacturers to loosen their purse strings on 300mm investment budgets. However, given that IC Insights sees the possibility for an industry downturn starting in 2002, the start of so many 300mm fabs over the next few years may be ill-timed. Hence, several of the new fabs may end up being partially filled shells until the market begins its next upward cycle in 2004.
Trevor Yancey is VP of Technology at IC Insights, 13901 N. 73rd St. #205, Scottsdale, AZ 85260; ph 480/348-1133, fax 480/348-9745, e-mail [email protected].