Green and Clean

Using design software to minimize the environmental impact of tools and processes

Sara Thurwachter and Terry Francis, Applied Materials

New analytical software can help engineers quantify the environmental effects of key process and hardware alternatives and factor in economic considerations early in the design process.

As the semiconductor industry continues to grow and to adopt new process technologies, both chipmakers and their equipment suppliers have exhibited an increasing concern for the consumption of natural resources and the environmental impact of manufacturing effluents. Thus, the development of analytical software tools that would aid the industry in understanding, measuring, and improving the environmental performance of manufacturing equipment and processes is being pursued as a collaborative effort between the environmental solutions and products division at Applied Materials (Santa Clara, CA) and the Consortium on Green Design and Manufacturing (CGDM) at the University of California, Berkeley.

The CGDM was formed in 1993 to encourage multidisciplinary research on environmental management and pollution prevention issues in critical industries. It is based in the university's department of mechanical engineering but also involves faculty and students in the industrial engineering department, the School of Public Health, the Energy and Resources Group, and the Haas School of Business.

Because the most significant improvements can be made early in product development, when there is the most flexibility for adopting changes, the focus of the joint effort has been on developing software tools that can predict environmental performance along with more-classic metrics during equipment, process, and facility design.

In semiconductor equipment and process development, a design project typically evolves in different product groups and companies. Most design decisions are made by the tool supplier's product team. However, these decisions are made at many different levels—such as process engineering, component design, platform engineering, environmental safety and health (ESH), marketing, and general management. They are also influenced by the tools' end-users, who communicate new technology requirements, such as facility designs and international regulations. Because few equipment and process design teams have any background in environmental engineering, much less in ESH disciplines, having a software tool available to analyze the environmental effects of key process and hardware alternatives early in the design phase can help teams to factor in economic considerations and clearly understand the trade-offs available through different approaches. This article describes the environmental analysis software (EnV) that the Applied Materials/CGDM collaborators have developed for this purpose.

Software Design and Analytical Structure

Using process information to evaluate manufacturing functions and interactions, the EnV software characterizes the performance of manufacturing systems in order to determine the optimal opportunities for design improvements during equipment development. It can be used without special training by design team members with differing backgrounds and concerns and by different groups at various design stages. Because economic and ESH terminology is translated into the language customarily used by design and manufacturing engineers, results are easy to understand and apply. By using this design tool, chip equipment suppliers can maximize the environmental performance of their systems while meeting their customers' technical requirements. Chipmakers, in turn, can reduce the potential environmental impact of manufacturing by purchasing "green" production tools designed from a complete-facilities perspective.

Because the software tool will be used to make technical and business decisions, its designers were faced with an unusually diverse set of considerations. Common decision drivers such as economic and technical performance, quality, and customer satisfaction had to be integrated into the tool's analytical capabilities. Moreover, the environmental and health impact of different processes, which is difficult to translate into classic economic values, and a range of other qualitative issues, such as community perceptions and regulatory and resource constraints, had to be taken into consideration. Some scientific data also needed to be integrated into design decisions.

Current semiconductor industry economic practice is to aggregate many ESH costs into facility costs, disguising their significance. The EnV software separates the costs of permitting, monitoring, and treating effluents and of disposing of a manufacturing process tool and its process outputs. This information provides equipment manufacturers with clear performance metrics. Several established models were evaluated when formulating the software's analytical framework, including Sematech's standard cost of ownership (COO) model.^1­5 However, the framework was customized to satisfy the specific concerns of semiconductor equipment and process development.

The EnV software is structured around the following three basic concepts:

• Analysis is based on process models in order to clarify the trade-offs and the largest driving factors of a system's hardware and process characteristics. Using a process-modeling foundation, the software can relate costs, environmental impact, and performance metrics back to process parameters for evaluation of a process's sensitivity to cost and ESH issues.
• A combination of manufacturing and facility-scale environmental parameters is utilized to give users a systems-level look at the overall environmental impact of a manufacturing process and the secondary flows between processes. This comprehensiveness is critical to determining the most significant issues. In addition, unlike most design-for-environment methods, the software does not require large amounts of data, which are often unavailable, to run an evaluation.
• By separating process and facility data into quantitative per-wafer-pass values, the software translates all expenses and emissions data into a common language and incorporates ESH impacts into the overall value. Non-value-adding costs, such as idle time and facility costs, are also included in the analysis, since these expenses are often a large portion of overall costs (and have not typically been included in other cost models).

Because the usefulness of the environmental value analysis depends on its comprehensiveness and predictive capability, the software is layered in two dimensions. First, it manages the interdependencies of the process equipment by expanding out from the process tool to the facility. Then it characterizes the multiple design trade-offs by quantifying the relevant data. As shown in Figure 1, the analytical framework consists of three steps: evaluating the system architecture, developing process models, and characterizing the resulting information into three categories: cost of ownership, environmental and health impact, and process performance.

Figure 1: Schematic of the software tool's analytical framework.

System Architecture and Process Modeling. In semiconductor manufacturing, each primary manufacturing process (chemical vapor deposition or etch, for example) triggers a string of interacting secondary processes, as diagrammed in Figure 2. The software tool takes the outputs of all of these processes and combines the information in the context of a process chain, with the primary and associated secondary processes strung together. Modeling the inputs and outputs of each process allows for both the optimization of individual tools and overall system improvements.

Figure 2: Schematic example of the system architecture and process-modeling approach.

Although an equipment supplier's perspective is understandably focused on its own product, the boundary of what is considered the supplier's responsibility is expanding. Understanding all of the environmental effects and costs arising from a primary processing step enables equipment designers to identify potential improvements. The EnV software also helps them to optimize their tool's design over the bigger-picture "system box" in Figure 2, rather than the much smaller "tool box." Taking this larger view can reveal how a potential solution to an environmental problem may simply shift it to somewhere else in the production flow. For example, although the installation of point-of-use abatement devices for CVD chamber cleans reduced gas emissions associated with global warming, this solution shifted the industry's burden to liquid effluents containing aqueous hydrofluoric acid.

Cost of Ownership. The software's COO calculations are given in a spreadsheet format, similar to the Sematech COO model, and use input cells to estimate costs over the process tool's production lifetime. Inputs are organized into fab, system, equipment, and production data.

One way the COO analysis has been customized for the semiconductor industry is by segregating the costs that can be linked to process models (power per pump, gas panel exhaust, and water per polishing platen, for example). The software also details environmental costs by identifying the expenses incurred for the use, handling, monitoring, treatment, and disposal of process inputs, outputs, and secondary flows (i.e., energy, water, cleaning agents, treatment chemicals, dilution streams, exhaust, and treatment by-products). The separation of these costs encourages process and tool designers to improve a fab's environmental performance by reducing consumables and choosing treatment methods that are the least environmentally risky or burdensome. The analysis also emphasizes how significant ESH-related costs are, compared with standard operating costs. An example of COO results is seen in Figure 3; in this case, treatment-related costs ranged from 20 to 40% of process costs.

Figure 3: Example of COO results generated by the analytical software. Segregation of the various cost factors enables designers to choose alternatives that improve environmental performance.

ESH Impact. In addition to operating costs, which are fairly clear-cut, the impact of such ESH factors as chemical toxicity and global warming, which are typically difficult to quantify, also require evaluation. The software's ESH component uses a multicriteria hazard (MCH) evaluation technique, developed at UC Berkeley, that categorizes the potential health and safety effects of materials or manufacturing processes into six categories: acute toxicity; systemic toxicity; reproductive and developmental toxicity; carcinogenicity, mutagenicity, and genotoxicity; physical hazards; and standards and regulations.² Figure 4 presents an example of the MCH results for four process gases and a reference material. In this figure, the higher the number, the more hazardous the potential effect is considered to be.

Figure 4: Example of MCH results generated by the analytical software (10 represents the most hazardous and 0 the least hazardous effect).

The software also evaluates a tool's or process's impact using an abbreviated version of a life-cycle assessment classification and characterization technique.³ This technique includes sorting potential impacts by their effects, and then characterizing them in terms of the degree to which they contribute to an effect. For example, the emission of 1 g of C_²F_⁶ is considered 9200 times more significant than that of 1 g of carbon dioxide in contributing to a greenhouse effect over a 100-year time horizon.⁶ The results of these analyses are expressed in terms of element-equivalent weights on a per-wafer-pass or a per-layer basis.

The software can characterize several environmental effects, but the initial efforts have focused on a few impact categories of particular concern to the semiconductor industry, namely global warming (greenhouse) effects, hazardous air pollutants, water pollutants, solid waste, and natural-resource depletion. Examples of results from analyses of the first two categories are shown in Figures 5 and 6.

Figure 5: Example of life-cycle assessment results generated by the analytical software for global warming effects.

Figure 6: Example of life-cycle assessment results generated by the analytical software for hazardous air pollutants.

Process Performance. The software uses technical performance data to evaluate engineering objectives that are difficult to translate into costs. These measurements vary more than the other categories since they depend heavily on the specific process or equipment being evaluated. The criteria, ranging from typical technical measurements to business metrics to environmental friendliness, include resource efficiency (dissociation percentage, utilization), influence on yield (throughput, scheduled and unscheduled downtime), abatement effectiveness (destruction efficiencies, output concentrations), and regulatory compliance (requirements for threshold limit values, pH).

Preliminary Results

A series of case studies, which ranged from materials evaluations to process sequences, have contributed to the development of the EnV software. One materials selection case study, for example, evaluated the effects of using CF^₄, C^₂F^₆, and NF^₃ for CVD chamber cleaning.⁷ Although NF^₃ can be 50­250% more expensive than C^₂F^₆, the study showed that a more environmentally friendly, remote plasma NF^₃ process costs only slightly more than a radio-frequency plasma C^₂F^₆ clean because of its more-efficient gas utilization, its lower particle counts, and its softer clean. The environmental values analysis quantified how cost-effective the NF^₃ clean is and how its excellent gas dissociation correlates with reduced emissions.

Although this evaluation demonstrated the benefits of NF^₃ utilization, evaluating the trade-offs of the various cleaning gases independent of other factors had limitations. For example, the overall costs and impacts of these gases also depend on the abatement devices that can be used to treat the process output streams. Therefore, the study was expanded to include several comparative analyses of abatement devices used to treat CF^₄, C^₂F^₆, and NF^₃ emissions.

Using the EnV software to evaluate the process outputs along with the required abatement processes yielded results that differed from those obtained from the assessment of the primary-process economics alone, revealing additional opportunities for improving CVD chamber cleans. The analysis demonstrated that the total cost of NF^₃ abatement is approximately 20% less than that of CF^₄ and C^₂F^₆. The analysis also showed that abatement costs can contribute up to 25% of total operation costs for a CVD chamber cleaning process when the water-flow rate of the abatement unit accounts for up to 50% of total abatement costs because of the need for hydrofluoric acid treatment. These results established the hidden significance of abatement process costs and provided a motivation for expanding the study further.

The next step was to evaluate the trade-offs of running the cleaning processes on different process chambers mounted on two different equipment platforms (designated A and B). One evaluation, for example, showed that running the NF^₃ clean on a process chamber on platform B is less expensive than using a C^₂F^₆ cleaning process on a similar CVD chamber on platform A.⁸ The analyses modeled a cross section of the effects of the CVD-clean process step, from the inputs into the facility to the final outputs to the environment. The results of one comparison, shown in Figure 7, included capital, operation, and treatment costs.

Figure 7: Example of results from a case study that used the analytical software to compare various CVD chamber cleaning processes.

The NF^₃ clean on platform B also resulted in 88% less global warming potential than the C^₂F^₆ chamber clean on platform A after point-of-use abatement. While the NF^₃ clean resulted in a small increase in hazardous air pollutant emissions, this was offset by a reduction of almost 265,000 lb of carbon equivalents per year for a facility with 6000 wafer starts per week. The mere 4.5-lb increase in HF equivalents contributes less than 0.025% to a 10-ton site limit. In this case study the software was able to evaluate entire facility systems, leading to ideas for point-of-use abatement and tool-platform design improvements.

Current Work

A cooperative project to extend the EnV software is under way. The new software, designated EnV-S for environmental value systems analysis, will enable designers to consider other primary process steps that can influence a particular system or technology in a fab. All primary processes performed at a facility emit waste products into the same treatment streams, for example, and thus can affect the overall performance of an abatement system (such as its capacity, concentration thresholds, and efficiencies). UC Berkeley and Applied are setting up the initial framework for the fab-level evaluation tool. Work on primary-process modeling is being done by the UC Berkeley Precision Engineering Laboratory and Small Feature Reproducibility Group, and treatment-process modeling is being performed primarily through the National Science Foundation/Semiconductor Research Corp. Engineering Research Center for Environmentally Benign Semiconductor Manufacturing at the University of Arizona (Tucson).

The EnV-S software will apply the EnV approach to a broader systems view with dynamic process models and a design and decision interface. The interface will facilitate rapid analyses of an overall system's environmental impact to identify areas that require more-detailed modeling and characterization. The new tool will also enable the process models for an individual system to be linked together and a discrete time simulation of the system to be performed. An early version is being applied to effluent recycling and treatment systems for the chemical-mechanical polishing process in order to evaluate alternative treatment scenarios. This particular systems analysis is an appropriate stepping stone toward a multiple-primary-process evaluation.⁹

Conclusion

The industry's increasing awareness of and concern for the environmental impact of semiconductor manufacturing has stimulated the development of analytical software to help both fabs and their equipment suppliers to understand, measure, and improve the environmental performance of chipmaking equipment and processes. The software, which was developed by Applied Materials and UC Berkeley's CGDM, quantifies environmental performance in conjunction with process, equipment, and facility design. It is especially useful early during the equipment development process, when design modifications are relatively easy to adopt, but can also improve design decisions throughout the product life cycle.

Acknowledgments

The authors would like to thank professors Paul Sheng and David Dornfeld of the University of California, Berkeley, and Nikhil Krishnan, who is working on the expansion of the analytical software through both UC Berkeley and Applied Materials. Current information can be obtained at http://greenmfg.me.berkeley.edu/green/home/index.html.

References

1. Cost of Ownership for Semiconductor Manufacturing Equipment Metrics, SEMI E35-95A (Mountain View, CA: SEMI, 1995).
2. S Thurwachter, "Multi-Criteria Hazard Evaluation," Masters thesis, University of California, Berkeley, 1998.
3. J Fava et al., A Conceptual Framework for Life-Cycle Impact Assessment (Pensacola, FL: Society of Environmental Toxicology and Chemistry, 1993).
4. A Veltri, "ESH Cost Model Development Report, " Sematech Doc. No. 97093350A-ENG (Austin, TX: Sematech, 1997).
5. W Lashbrook et al., "Design for Environment Tools for Management Decision Making: A Selected Case Study, " in Proceedings of the 1997 IEEE International Symposium on Electronics and the Environment (Piscataway, NJ: Institute of Electrical and Electronics Engineers, 1997), 99­104.
6. "Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990­1997, " EPA 236-R-99-003 (Washington, DC: Environmental Protection Agency, 1999, 1­7; available from Internet: http://www.epa.gov/globalwarming/inventory/1999-inv.html.
7. S Thurwachter, J Schoening, and P Sheng, "Environmental Value (EnV) Analysis," in Proceedings of the 1999 IEEE International Symposium on Electronics and the Environment Conference, vol. 7 (Piscataway, NJ: Institute of Electrical and Electronics Engineers, 1999), 70­75.
8. S Thurwachter, J Schoening, and P Sheng, "A Design Tool for Semiconductor Process Tools," in Proceedings of the SEMI Environmental Impacts of Process Tools Technical Programs (Mountain View, CA: SEMI, 1999).
9. N Krishnan et al., "The Environmental Value Systems (EnV-S) Analysis: Application to CMP Effluent Treatment Options," in Proceedings of Improving Environmental Performance of Wafer Manufacturing Processes (San Jose: SEMI, 2000).

Sara Thurwachter, PhD, is the environmental systems core technologist in the environmental solutions and products division at Applied Materials (Santa Clara, CA) and a researcher in the Consortium on Green Design and Manufacturing at the University of California, Berkeley. Thurwachter has also worked at Intel and Sematech and has published several papers in manufacturing and semiconductor journals. She received a BS in mechanical engineering in 1996 from the University of Texas, Austin, and a management of technology certification in 1998 from the University of California, Berkeley. In 1998 she received an MS and in 2000 a PhD in mechanical engineering from UC Berkeley. (Thurwachter can be reached at sara@greenmfg.me.berkeley.edu.)

Terry Francis is the chief technical officer of the environmental solutions and products division at Applied Materials. He has over 30 years of experience in semiconductor processing, electronic chemicals, microcontamination, and environmental technologies and holds several patents in these fields. In addition to being a technical advisor to standards organizations, he is a member of the editorial advisory board of MICRO and the industrial advisory board for university research programs. He has also collaborated on the National Technology Roadmap for Semiconductors. Francis has chaired symposiums and workshops, and authored many papers. He received a BS in chemistry in 1970 from Oregon State University (Corvallis) and an MBA in 1982 from the National University in San Diego (Francis can be reached at 408/748-5410 or Terry_francis@amat.com.)

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit