Controlling fab tool costs with improved equipment productivity

Neal Marmillion, Internationsl Sematech Manufacturing Initiative (ISMI)

The cost of new semiconductor fab equipment, both in terms of dollars and floor space, continues to be a large incentive driving semiconductor manufacturing productivity improvements. Models show that a 200-mm fab manufacturing 20,000 wafers per month at the 130-nm node requires an equipment investment of ~$900 million. In contrast, a 300-mm fab manufacturing 20,000 wafers per month at the 90-nm node requires an equipment investment of ~$1.4 billion. Finally, a 300-mm fab manufacturing 20,000 wafers per month at the the 65-nm node is expected to require an ~$1.8 billion equipment investment.

This investment is impelling semiconductor manufacturers to remain focused on maximizing equipment productivity. Industry leaders are realizing that on their own, they cannot afford to do the learning necessary to maximize equipment investment. Hence, partnerships, collaboration, and the sharing of a combined knowledge set are becoming the norm as manufacturers struggle with the economics of today's fabs. Further optimization of equipment productivity through the use of automation and advanced control schemes will lead to reduced cycle times and defects, thus shortening time to market. Improving equipment productivity and its associated costs are key areas in which International Sematech Manufacturing Initiative (ISMI) member companies are focusing their efforts and support.

Industry Collaboration

Collaboration is not a new concept to the semiconductor industry. In the 1990s, it became very apparent to leading-edge technology providers that their development costs were escalating dramatically as technology was adapting to the requirements of the International Technology Roadmap for Semiconductors. To help defray the escalating costs, companies formed joint development programs and alliances with customers and other technology leaders in which all involved parties shared the costs of their projects. Companies also used their membership in consortia such as IMEC and Sematech to leverage a pool of development monies supplied by 10 to 15 companies.

Typically, a joint development effort or alliance arrangement involves two or three similar companies whose agreement is fairly narrow in scope. A consortium, on the other hand, has to satisfy the needs of sometimes very diverse businesses. Either way, the pooling of funds is used to pursue an outcome that is shared by all parties involved.

In the area of equipment productivity, ISMI, a wholly owned subsidiary of Sematech, has taken collaboration to the next level. While the fundamentals of a collaborative effort are still in place, where every ISMI member provides funding for the unified effort to drive equipment productivity improvements, members combine their manufacturing experiences to help shape the continuous improvement program (CIP) solutions provided by equipment suppliers. A membership champion is appointed to test the solution in the fab and then share the conclusions with all the other participants. This procedure results in a win-win situation for the consortium and the supplier. Membership resources are conserved by centralizing the effort to test the solution, and suppliers receive a single, prioritized response from ISMI's members as to where improvement is needed.

At the first annual ISMI Symposium for Manufacturing Effectiveness in 2004, Varian Semiconductor Equipment Associates (Gloucester, MA) demonstrated how collaboration with ISMI had been integrated into the company's CIP. Illustrating the CIP input process, the schematic diagram in Figure 1 shows how an ISMI/Varian equipment productivity improvement team (EPIT) has direct input into the company's CIP activities. The team, relying on input from ISMI members, sought to make software improvements (beam tuning and bug identification), hardware improvements (reliability and wafer handling), and process improvements (yield enhancement, source life, particle reduction).¹

Figure 1: Collaborative learning integration process for continuous equipment improvement. (Illustration courtesy of Varian Semiconductor Equipment Associates.)

Two improvements, the E-series roplat cable and Productivity Plus, were in part a result of the EPIT's productivity and cost of ownership (COO) reduction efforts. The E-series roplat cable improvement involved reengineering the two electrical cables that connect to the roplat cable into a single flexible ribbon cable to eliminate particle or reliability issues that were caused when the cables came loose or broke. While cable issues arose once a month before the modification, they were reduced to once every 18 months thereafter.

Productivity Plus achieved a needed throughput improvement without increasing tool footprint. It was accomplished by installing an optional additional pick arm on a tool to reduce the mechanical wait time associated with wafer movement from the cassette to the orienter, the orienter to the platen, and the platen to cassette. This option is especially advantageous in short-duration implant processes, providing up to 30% greater throughput. Modeling indicated that a 20% throughput improvement could be achieved with normal fab implant loading at a member company's site.

These examples of win-win collaboration demonstrate how suppliers can provide needed improvements that make their products more cost competitive while all ISMI member companies can enjoy the benefits of mutual support to solve technology issues, enabling them to become early adopters of new hardware. Varian highlighted four key advantages of collaborating with ISMI:

• Access to open discussion on best-known methods.

• A return on investment (ROI) in the form of "golden nuggets."

• Exposure to technology issues from nonmember companies through CIP presentations.

• The establishment of EPIT members' equipment performance as a benchmark for the industry as a whole.

The issues addressed in this collaborative environment are both strategic and tactical. They encompass software and hardware inadequacies, where solutions will benefit multiple member companies. Some examples include:

• Improving mean time between wet cleans by extending part lifetimes. This goal can be achieved by using different materials or redesigning parts to improve particle performance. For example, by optimizing the slit-valve O-ring material used in chemical vapor deposition processes, coupled with adjusting slit-valve operation, a 50% reduction in baseline particle performance was achieved.

• Providing software solutions that enable equipment to recover from processing errors automatically.

• Addressing across-wafer uniformity. A supplier developed integrated hardware and software solutions to improve across-wafer uniformity issues, which reduced edge-die yield loss to within 1% of the center-wafer average.

• Improving electrostatic chuck (ESC) critical dimension (CD) uniformity. Another supplier improved ESC CD uniformity by providing a best-known method that reduced CD shift from 0.03 to <0.01 µm on a 0.4-µm feature.

Collaborating to develop a multicompany consensus and encouraging supplier engagement has proven to be a most effective tool for driving equipment productivity. When multiple end-users jointly decide where to focus their activities, equipment suppliers can justify particular CIP activities. Although each company has to deal with its own protocol for incorporating solutions into their fabs, the bottom line for suppliers is that a solution has been provided that addresses the needs of several end-users.

Information Sharing

Belonging to a consortium enables participants to leverage their combined knowledge. Benchmarking results of member companies' equipment performance are a valued commodity for driving equipment productivity. Protected by legal restrictions on further distribution, member companies share equipment performance data using the SEMI E10 reporting methodology to understand best-of-breed manufacturing capabilities. This process is supported by equipment experts from the member companies who work in teams to seek equipment productivity opportunities.²

The teams focus on equipment productivity issues while keeping cost-, process-, and recipe-specific information confidential. This procedure enables participants to share equipment software and hardware experiences. At ISMI, these stories are called golden nuggets, concrete information gathered by participants in team meetings (as pictured in Figure 2).

Figure 2: A 2004 EPIT team face-to-face meeting.

The nuggets are stored in a database that is available to all ISMI members. That database contains more than 2300 entries from 33 different equipment projects undertaken in the past six years. Entries cover tools used in all processes, including furnace applications, wet processing, etch, deposition, lithography, ion implant, chemical-mechanical polishing, and metrology. To make them easy to find, the data are categorized. Every entry contains information on the advantages expected from a modification, including COO, throughput, availability, reliability, and uptime improvements. Table I lists some examples of golden nuggets.

After the effectiveness of a modification is demonstrated, the results are relayed to the team and compiled in a searchable database for subsequent use. This golden nuggets database has become a benchmarking instrument that highlights the power of information sharing in the program.

Pushing the Cost-Reduction Envelope

As they compete in the marketplace, fabs are constantly pressured to reduce costs. Once a fab's equipment depreciation base is set, cost-reduction efforts focus on the outflow of "green" dollars—actual cash being spent to support the equipment set. Consumable parts quickly become a focus of cost-reduction measures, driving the effort to improve part lifetimes and leading to improved equipment productivity. At this stage, fabs either purchase parts from the original equipment manufacturer (OEM) or eventually turn to second-source suppliers. Increased equipment output, improved defect density, and higher overall equipment effectiveness are common leverage schemes for lowering costs. While all IC manufacturers rely on these methods, they do not have the capacity to address all of them independently, leading to missed opportunities.

Traditionally, an effective method for improving equipment productivity has been the use of beta sites, where suppliers and end-users collaborate to test solutions. Eventually, successful work performed at beta sites makes its way into the marketplace in the form of products. Confidentiality agreements between suppliers and evaluators often postpone the availability of a solution in the marketplace.

Although positive results often migrate into marketing literature, the evaluation source is not identified, limiting other suppliers' ability to collaborate directly with the evaluator about the project.

In contrast, when OEMs or second-source suppliers within the consortium perform work at beta sites, they can share both positive and negative results with other consortium members in a timely fashion. Over the course of an evaluation, the participants share written reports that describe the project, chart the expected results, catalog observations from installation through execution, and rate the evaluation.

In one case, a supplier-sponsored improvement package involving a new chamber body design and a new valve to improve particle performance on an aluminum/copper 0.5% metal etch tool was evaluated. The beta site results were compared with a control group of seven other tools that were configured without the upgrade at an ISMI member company's facility. To determine the effectiveness of the tool reconfiguration, data were gathered on in-line blanket-wafer particle levels, product-wafer particle levels, mean time between wet cleans (MTBCs), and mean time between failures (MTBFs). Product-wafer particle results were obtained using an AIT II patterned-wafer inspection system from KLA-Tencor (San Jose). The evaluation, which received a rating of 8 out of 10, concluded:

• The in-line blanket-wafer particle data baseline degraded as a result of the beta site installation, the reasons for which were detailed in the report.

• Depending on the product type, in-line product-wafer defect density improved by approximately 35%.

• MTBC improved from 200 to 300 hours.

• MTBF doubled.

Another advantage of the beta site program is that participants also gain access to reports on unsuccessful beta site evaluations, especially those based on input from consortium members. For example, a supplier proposed replacing the turbo molecular pump in an etch tool that was thought to be a key source of particles streaming back into the metal etch chamber. However, participants in the evaluation could not collect particle data because of interface problems at the site. After multiple attempts were made, other data from the site became available and the test was aborted.

This and the previous example highlight the advantage of sharing beta site information, both good and bad. The credibility of beta site evaluations is enhanced when participants share objective evaluation results of beta equipment performance, making them available to other consortium members for discussion.

Once end-users cannot continue to lower costs by buying replacement parts directly from OEMs, they seek to identify and qualify lower-cost, second-source parts suppliers. The marketplace eventually reacts, and companies emerge to fill the demand. That trend can be seen most clearly in the automobile industry, where common high-turnover parts such as bulbs, belts, filters, and brakes quickly enter the market and become alternatives to high-cost dealer offerings. This trend has also occurred in the semiconductor equipment replacement-parts sector. In contrast to the auto industry, however, specialized technical proficiency is required to manufacture IC tool replacement parts, leading to higher cost margins.

This is another area in which consortium membership is advantageous. Although information on parts costs is not shared among members, shared knowledge of the areas in which companies are purchasing spare parts from second-source providers can be a powerful cost-reduction tool. This activity also provides visibility to qualified second-source parts suppliers. The OEMs and their OEM parts suppliers are reacting to the loss of market share by making price concessions and providing end-users with previously unavailable customized solutions, such as reworked-parts programs.

Striving to become best of breed can help to improve equipment productivity in other areas as well. The use of in-house organizations versus service contracts to support equipment maintenance, quartz cleaning, parts stocking, pump rebuilding, and gas services can have an impact on overall cost and equipment productivity.

Internal ISMI workshops have been formed in which consortium members can discuss their companies' operational strategies. The end result is a benchmarking tool that is available to ISMI member companies as they strive to make their fabs best of breed for equipment performance. These workshops have focused on the advantages of single-source procurement and maintenance of support equipment such as pumps, the use of complete outsourcing versus in-house activities, OEM service contracts and future-use strategies, and second-source parts use.

E-Manufacturing

The integration of metrology data systems, statistical process control (SPC), and fault detection and classification (FDC) into the fab will play a key role in identifying equipment problems and reducing mean time to detection. Residual-gas analysis, arc detection, optical emission spectroscopic monitoring, and downstream sensors for detecting particle excursions have become the norm for extracting consistently high-quality data from process equipment.

In addition, by using feed-forward and feed-backward control schemes that enable engineers to compensate for or fine-tune process recipes, advanced process control (APC) maximizes the value of equipment output, increasing process and die yields, eliminating the need for send-ahead wafers, and reducing rework. Figure 3 illustrates the interaction between SPC, metrology, process equipment, and process adjustment.

Figure 3: Role of statistical process control in detecting process equipment issues.

Data outputs are also being used to predict maintenance activities and equipment performance trends, reducing nonscheduled downtime and extending lifetime estimates of preexisting parts. Multivariate analysis of equipment parameter data has proven to be an effective technique for providing additional visibility to equipment performance trends in complicated control schemes. Translation of the multivariate output signal into a concise set of corrective actions will further improve equipment effectiveness.

IC manufacturers' ability to make automated data-driven decisions will be key to realizing the productivity roadmap.³ However, with its data-sampling limitations, the use of the SECS/GEM interface to collect and transmit data without affecting equipment processors' ability to maintain maximum throughput is beginning to limit fabs' ability to catch perturbations in equipment performance.⁴ The development of Interface A, which has a data-collection rate 10 times greater than that of SECS/GEM, has been aided by the adoption of four standards that will create a new dataport for external access to equipment data. The new interface is scheduled for deployment on 300-mm equipment.⁵

Conclusion

The huge tool investments required by the semiconductor industry continue to fuel the never-ending drive to improve equipment productivity. Multicompany alliances are an effective way to lower individual companies' investments in the learning process, which is necessary for maximizing equipment performance. Groups of companies can share costs, resources, and knowledge in order to define, execute, and verify equipment productivity improvements. Collaboration between equipment suppliers and end-users on CIP projects creates a win-win business environment. By sharing information on equipment problems, performance benchmarks, beta site evaluations, second-source suppliers, and equipment-related success stories, joint company collaboration can be leveraged to improve equipment productivity and lower bottom-line costs. Further productivity gains can be achieved by integrating APC in the fab through the establishment of equipment SPC, FDC, and run-to-run rules.

References

1. P Tarvin, "How the EPIT Process Has Helped VSEA to Improve Equipment Productivity" (paper presented at the First ISMI Symposium on Manufacturing Effectiveness, Austin, TX, October 26, 2004).