Begun Investigating Any Alternatives

Barriers to Best Practice

BARRIERS TO BEST PRACTICE The Interplay of Technological, Social, Economic, and Political Impediments

Professor Lionel Kimerling Michele Mlynarczyk Department of Materials Science and Engineering Massachusetts Institute of Technology

© 1999 Massachusetts Institute of Technology. All rights reserved This case is intended as a basis for classroom discussion, rather than to illustrate either effective or ineffective handling of a management situation.

BACKGROUND

On April 3, 1994, Marcio Valdez began his first assignment in the Technology Development Division (TDD) of Semicon, a large semiconductor manufacturer specializing in advanced integrated circuits such as microprocessors. TDD had recently received the results of an extensive study performed at Teratech, a leading research institute. The study indicated that a solution of methanol-iodine out-performed the current industry standard, dilute hydrofluoric acid, in cleaning and passivating wafer1 surfaces. Valdez was asked to investigate the feasibility of using the methanol-iodine solution in wafer cleaning stations2 and to present his recommendations to the TDD approval board on May 15, 1994. His recommendation, if accepted, could significantly impact the cost and capability of the next generation process technology.

Teratech had produced sufficient data to prove that the methanol-iodine solution was technically superior to hydrofluoric acid (HF). Yet, technical superiority was not a sufficient condition for TDD to commit resources to integrate the new solution into the next generation process technology. Sara Briggs, a TDD manager, explained to Valdez:

The TDD engineers propose multiple ground-breaking process improvements at each new technology review; yet, even the best of them may be rejected because implementation would be too tricky. It seems that the technology approval board is more likely to accept incremental changes than revolutionary changes. The research data suggests that the use of the methanol-iodine solution will result in considerable improvements in wafer surface properties. It might also provide us with processing advantages that would give us a competitive advantage in years to come. Some experts say that HF solutions may be obsolete in just a few years. Others say that improvements in the HF cleaning process will

1 A wafer is a round, thin slice of silicon upon which integrated circuits are fabricated. A finished wafer will contain numerous integrated circuits, which will be separated and individually tested in subsequent processing. 2 A wafer cleaning station contains a bath of cleaning chemicals. Wafers are submerged into the chemical bath to remove wafer contaminants, such as residual metals and dust particles.

1 Barriers to Best Practice

make it feasible for many more years. I’m not sure who is correct or whether or not this is the right time to introduce a HF substitute. The semiconductor consortium has not even begun investigating any alternatives. We need evidence that the substitution is worth the costs and risks involved before the approval board will accept it.

With that advice, Valdez began his investigation of the methanol-iodine solution and what it would take to implement it into the wafer cleaning process.

THE WAFER CLEANING PROCESS

The process of manufacturing integrated circuits can be described by a sequence of fabrication steps that are performed in a cleanroom, where the air is continuously filtered to control the number of particles and impurities that could contract that wafer surfaces. The process of fabricating integrated circuits begins with bare silicon wafers. Each processing step contributes to the formation of electrical components on the wafer surface in order to connect individual devices and form integrated circuits.

The most critical feature of an integrated circuit is the gate oxide, an insulating layer of material that transmits applied gate voltage to the underlying silicon. The electrical conductivity of the silicon surface is modulated by the gate voltage to provide a digital switching response. The oxide is typically 75-100 angstroms (10 billionths of a centimeter) in height and is grown on to the silicon wafer surface using chemical/thermal processing called diffusion.

Wafers are cleaned prior to growing the gate oxide in order to provide a smooth, particle- free, and chemically stable surface. The cleaning process involves multiple steps. The first step, called Standard Clean 1 (SC1), rids the wafer surface of particles, such as microscopic dust. The second step, Standard Clean 2 (SC2), consists of an acidic bath that dissolves residual metals on the wafer surface. The last step involves a water-diluted HF rinse that passivates, or chemically stabilizes, the wafer surface.

IMPETUS FOR CHANGE: TECHNICAL CONSIDERATIONS

Each new generation of process technology (approximately every three years) introduces a scaling down of integrated circuit dimensions. Next generation gate oxide dimensions were forecasted to be as small as 40-50 angstroms. The ability of the gate oxide to meet performance specifications is defined by a gate oxide integrity, GOI, metric. The key requirements are (1) very low contamination levels of the pristine surface and (2) atomistically flat surfaces.

Teratech researchers believed that dilute HF cleaning solutions would most likely not meet requirements even five years out. The ranges for each cleaning process variable, such as immersion time, solution temperature, and water purity, were forecasted to shrink beyond

2 Barriers to Best Practice what was controllable. Furthermore, dilute HF is mildly corrosive to the silicon surface, and extended exposure can roughen the silicon surface on a 10-40 angstrom scale. However, incremental improvements in HF cleaning processes, such as insitu3 purification of the HF solutions, might enable the continued use of HF solutions in wafer cleaning applications for years to come. But, the cost and feasibility of such improvements were not yet known.

Methanol and other alcohols were investigated as alternatives to dilute HF because they are less reactive than water. Methanol, CH3OH, is a polar molecule that shields the silicon surface by physisorption4. The iodine in the solution serves the same purpose as the hydrogen in HF solutions; yet, the resulting surface stability is better. The heavier iodine is less volatile than hydrogen, while providing the same passivating chemical reactions. In addition, methanol, unlike HF, does not corrode the silicon surface. Thus, the methanol- iodine treatment creates atomistically flatter surfaces and enhances the stability of the wafer surface. These advancements would reduce the number of defects and simplify preparation of thinner gate oxides.

In addition to these technical advantages, the methanol-iodine solution appeared to offer other benefits as well. Integrated circuit fabrication requires the use of several hazardous chemicals, and HF is one of the most toxic. Handling and disposal of the HF solutions requires extreme care and significant expense. According to the Teratech study, the use of the methanol-iodine solution would make the workplace safer by eliminating the handling and disposal issues associated with HF. Valdez decided to investigate these aspects of implementation first.

SOCIAL CONSIDERATIONS: WORKPLACE SAFETY AND ENVIRONMENTAL CONTROL

Hydrofluoric acid is an inorganic, corrosive material. Both the liquid and the vapor can cause severe skin and eye burns that may not be immediately visible or painful. In most wafer cleaning applications, 48% concentrated HF is diluted with distilled water to achieve 50:1 (water volume to HF volume) solutions. Even at these low concentrations, HF solutions are highly toxic and require special handling and working procedures.5

Methanol, on the other hand, is safe enough that it is commonly used as an industrial solvent (see Table 1 for a comparison of the properties of hydrofluoric acid and methanol). Other common uses of methanol, also known as methyl alcohol or wood alcohol, include antifreeze for automotive radiators and air brakes, fuel for picnic stoves, and octane

3 Insitu purification involves continuous filtration of the HF solution within the wafer cleaning station. This eliminates the need to dispose of and refill the HF solution since the purity level is constantly maintained at the target level. 4 Physisorbed species adhere to a surface without reacting chemically with the surface. 5 Flow hoods are necessary to extract HF vapors and maintain exposure below threshold limit values (TLVs). Protective gloves (such as neoprene or polyvinyl chloride) are needed to safe guard against skin contact, and safety glasses are required to protect the eyes from chemical splashes.

3 Barriers to Best Practice booster for gasoline. Methanol is safe if handled properly; however, it is toxic if inhaled or swallowed.

Table 1: Properties of Methanol and Hydrofluoric Acid Characteristic Methanol Hydrofluoric Acid (HF) (at 1% concentration) Health risk severe - 3 extreme - 4 Contact damage Slight - 1 Extreme – 4 Reactivity with other chemicals Slight - 1 Moderate – 2 Flammability Severe - 3 None – 0 Permissible exposure limit (a) 2000 parts per million 3 parts per million Evaporation rate 4.6 (butyl acetate = 1) Slower than ether Source: Material Safety Data Sheets (a) The permissible exposure limit is a government-enforced limit on the amount of chemical vapors that can be present in the workplace air.

Safety in the workplace was a top priority at Semicon; therefore, Valdez thought the safety engineers at Semicon would be excited to hear about this new, safer solution. However, they were not as optimistic as he had though. Terry Derkin, a safety specialist at Semicon, informed Valdez:

For the most part, we have eliminated the safety hazards associated with working with HF. We placed automated material-handling systems on our wafer cleaning stations in order to minimize worker contact with HF solutions. As a result, HF burns and safety incidents are rare. We still have some risks though. Preparation of the HF solutions and potential cleaning station leaks present safety risks that we do not have a remedy for yet.

Another safety specialist explained that the methanol-iodine solution might actually compromise workplace safety.

The high flammability and evaporation rate of methanol could cause problems. Special care would have to be taken so that methanol or methanol-contaminated items did not contact acidic substances or materials, which could result in fire or explosions. The wafer cleaning equipment would also have to be modified to prevent the methanol from evaporating. While the permissible exposure limit of methanol is much higher than dilute HF, the high evaporation rate makes it more difficult to prevent methanol from permeating the workplace air.

Since methanol vapors corrode plastics, the frames of flow hoods for wafer cleaning stations would have to be retrofitted with a metal such as stainless steel rather than the typical polypropylene. Unfortunately, metal- framed hoods are more expensive than their plastic counterparts. It’s very likely that additional equipment modifications would be required to reduce evaporation. For example, the holding tank may have to be lowered, and

the wafer pull-out rate may have to be reduced in order to prevent any turbulence at the methanol/air interface. The equipment engineers may also need to place cooling coils above the tank surface and add a top cover to keep the methanol vapors below the permissible exposure limits. So, you see, switching solutions would require extensive and expensive equipment development efforts. As an old equipment engineer, I know the development costs involved in making some of these changes could be huge.

Valdez contacted Semicon’s wafer cleaning equipment supplier and inquired about the impact of the changes. The supplier verified that the retrofits Valdez had described would require a large development effort. He explained that wafer cleaning stations are million dollar pieces of equipment and that they were still recouping their investment in the current design. The supplier made it clear that either Semicon would have to bear all the equipment development costs in the price of the equipment or they would have to find partners to share the costs and any potential benefits as well.

Valdez also spoke with Sandy Kimball, an Environmental Health and Safety engineer at Semicon. She also voiced concerns about substituting the HF solution with the methanol- iodine solution:

The use of iodine, even in small quantities, also elicits safety concerns. Iodine is a poison that can be fatal if swallowed or inhaled. It is corrosive, and skin contact causes severe irritation. Processing wafers requires the use of many dangerous chemicals, but it’s always a chore to figure out how to control chemicals that we have never dealt with before.

Furthermore, the iodine would cause problems with the methanol incineration or recycling. Methanol is a volatile organic compound.6 We are already near the government enforced VOC limits; therefore, we can’t approve anything that would increase the amount of pollutants leaving our plant. In other words, incineration, the least expensive way to dispose of methanol, would not be an option given the volume of production we run here.

We are trying to be an environmentally friend company. Semicon’s goals are not just to meet government regulations but to reduce the environmental impact of our manufacturing as much as possible. We want to design manufacturing processes that eliminate harmful emissions. The emphasis is on life-cycle analysis, which requires thinking about how to avoid emissions and hot to encourage recycling. If the methanol-iodine solution could be recycled, then it might be a more attractive option.

6 Volatile organic compounds (VOCs) combine with NOx to produce ground-level ozone, which can cause respiratory problems in humans.

Valdez had not realized that methanol was a VOC. This would complicate any implementation plan. The federal government limits VOC emissions to 100 tons/year at each site. The local governments of cities where the Semicon plants were located imposed even stricter emission standards.

An EPA specialist also informed him that methanol and hydrofluoric acid are hazardous air pollutants7 that are regulated by the federal government under the Hazardous Air Pollutants (HAP) Act of 1991 and Title III of the Clean Air Amendments Act. The 189 chemicals on the HAP list are considered toxic and hazardous to human health and the environment. Any facility that emits over 10 tons/year of a single chemical on the HAP list, or 25 tons/year of any combination of chemicals on the list, must procure a permit, which often takes 12-18 months to obtain.

Semicon’s method of disposing hydrofluoric acid did not contribute to HAP emissions. The current process involved adding lime to diluted HF solutions to form calcium fluoride cakes. The chemical reaction was as follows:

2HF + Ca(OH)2  Ca2F + 2H2O

The calcium fluoride was allowed to dry into cakes that were usually disposed of in landfills. Since the caking process did not involve air emissions and the neutralized cakes were non-hazardous, the government did not regulate the number of inert cakes that were sent to landfills. The HF treatment process was usually performed on-site and was fairly inexpensive relative to other chemical disposal processes. To minimize disposal costs, two of Semicon’s plants transported the cakes to nearby cement manufacturing companies for use in their manufacturing processes.

Valdez had heard that efforts to recycle HF solutions were already underway. However, he could not obtain more information since the firms who were developing HF recycling systems considered it confidential information. Recycling would potentially eliminate the HF disposal process and would reduce the amount of water used in the manufacturing process. Certain Semicon plants were already constrained by local water availability and conservation measures. However, some experts were skeptical that HF recycling could provide the solution purity that would be needed in the next generation manufacturing process. State-of-the-art processing required inputs with fewer than 10 parts per billion (ppb) metallic impurities. Next generation processes would require purity standards of 1 ppb or less.

Several methods for solvent disposal and recycling methanol were currently available. Some of the most popular options included: outsourcing the whole process, shipping used solvent to cement manufacturers, and recycling used solvent on-site. Recycling methods included: adsorption, distillation recovery, membrane systems, and catalytic treatment (see appendix for a description of each). Yet, the feasibility of each method for methanol recycling depended on factors such as the amount of methanol to be recycled, the ability to achieve high solution purity, and the ability to recover the iodine separately. Membrane

7 Hazardous air pollutants pose health threats to humans; many are carcinogens and are toxic.

6 Barriers to Best Practice technologies appeared to be the most promising method, but considerable research would be necessary to prove its feasibility and cost-effectiveness. The efficiency with which methanol-iodine could be recycled or disposed of would significantly impact its cost- effectiveness relative to the HF solution.

ECONOMIC CONSIDERATIONS

Valdez knew that, if he recommended the methanol-iodine solution, he would have to demonstrate the approval board that it would be economically comparable to the HF solution. However, he did not have all the data needed to do a full cost-benefit analysis. Recycling data could change the cost comparisons significantly, but Valdez did not know which recycling method would prove feasible or how the recycling system scale affected the costs. Given his time constraints, he decided to at least compare the current material and disposal costs for the methanol-iodine solution and the HF solution (see Table 2 and Table 3).

Table 2: Costs for HF Solutions Grade 2 HF (48%) $155.00/gallon Distilled Water $ .04/gallon Solution (50:1) $ 3.14/gallon Disposal (outsourced) $ 1.50/gallon Based on 1994 average prices. Note that the price of HF is highly dependent on the grade (1,2,3). Prices may vary by 50% to 100% between grades. In addition, quantity discounts can significantly reduce the price.

Table 3: Costs for Methanol-Iodine Solutions Electronic Grade Methanol $ 25.36/gallon Iodine (99.99+ pure) $ 1.52/gram Iodine (0.37 gram/gallon) $ 0.58/gallon solution Solution (0.1 g/L iodine) $ 25.94/gallon Disposal (outsourced) $ 2.50/gallon Note: A 5 x 10-4 molar (0.1 g/L) solution of iodine in methanol is effective at passivating wafer surfaces.

POLITICAL CONSIDERATIONS

It seemed that everyone Valdez had consulted with had his/her own opinions on whether or not the methanol-iodine solution was a good option to pursue. The environmental engineers were risk averse. They did not want to be responsible for methanol recycling. If the containment efforts were unsuccessful, Semicon might exceed the VOC limits and be forced to limit production. The operations managers were concerned that the equipment modifications might reduce the wafer cleaning station throughput and increase the need for more capital investment in wafer cleaning solutions. Their bonuses were based on increasing output, reducing capital costs, and improving throughput time. Thus, they were less willing to take a risk as well.

On the other hand, other groups were pushing for TDD to develop new and better cleaning solutions. The quality and reliability engineers were desperately seeking ways to reduce defects, and the methanol-iodine solutions appeared to be a good option for this. The process engineers were also hopeful that the new solutions would eliminate the difficulties of keeping the HF process variables within tight tolerances.

It was impossible to satisfy everyone’s preferences, but Valdez knew that he would have to address both sets of arguments. All groups were represented on the approval board; therefore, it was not clear who would have the most decision-making authority. Besides, he felt that his job was to recommend what was best for the company and not what pleased any particular party.

TAKING A STANCE

At the end of his six-week study, Valdez felt overwhelmed by all the factors involved in his decision of whether or not to recommend the methanol-iodine solution. Somehow he had to pull all of his information together into a cohesive proposal for the TDD approval board. He still had many unanswered questions, but the decision of whether or not to pursue this solution for the next generation process technology had to be made now. On one hand, the methanol-iodine solution offered a technical advantage. The ease-of- handling of the solution and water savings were also clear benefits of making the substitution. On the other hand, some major changes, such as modifying the cleaning station equipment and implementing a recycling system, would be necessary to use the new solution. These changes would require large investments that may not be recouped, particularly if the recycling efforts failed.

Valdez felt that it was important to consider industry trends in making his recommendation. There was a good chance that the HF solution would not meet Semicon’s future needs, and it seemed that recycling was going to be necessary in future years regardless of whether Semicon used dilute HF or methanol-iodine solutions. Valdez was leaning towards advocating the methanol-iodine solution, but he wondered if he could support his argument with the data he had. As a newcomer to Semicon, he did not want to appear unrealistic about the ramifications of making a change of this magnitude. He also knew that members of the approval board tended to be risk averse. Yet, he wondered if it was riskier to invest in the methanol-iodine solution now or wait until other firms pioneered this avenue.

APPENDIX

METHODS OF SOLVENT DISPOSAL AND RECYCLING

Outsourcing Disposal/Recycling The option to have an outside company pick up, treat, and dispose of hazardous waste is always available. Many solvent suppliers supply virgin or recycled solvent and remove spent solvent for a fee. They then reprocess the spent solvents at a centralized facility and resell it to clients. For small-scale waste disposal, this may be a good option. The economies of scale achieved by the reprocessing firm may result in prices that are lower than small-scale, on-site treatment. However disposal services are usually more expensive than on-site treatment when firms generate larger volumes of used solvent. Companies can expect to pay approximately $2.50/gallon for waste disposal services. The price will depend on scale, proximity to receiving sites, and the type of substance to be disposed.

Shipping Hazardous Solvent to Other Manufacturers Solvents can be collected in tanks and transported to other manufacturers, such as cement producers, to be used as an input fuel. The cement manufacturers benefit from this free source of input fuel, and the spending company benefits because the EPA considers this recycling. However there are two requirements with this method. The first is that chemicals must be reprocessed within 90 days or the EPA considers it storage of hazardous waste. Expensive permits are required for waste storage, thus administrative controls would be necessary to ensure compliance with EPA regulations. Second, the halogen treatment process must be used to remove some of the halogens (iodine is a halogen).

Adsorption Adsorption is a separation process based on the ability of certain solids to remove gaseous components from a flow stream. The vapor molecules present in a waste stream collect on the surface of the solid material. The adsorbents used for air pollution control include activated carbon, alumina, bauxite, and silica gel.

Adsorption is one of the most effective and economical methods of controlling emissions of VOCs. The process is typically able to recover many VOCs for reuse, which lowers overall costs for a firm. In general, solvent recovery using adsorption is a logical consideration for any industrial process exhausting sizable quantities of solvent. There are many variations of adsorption systems, and much R&D is taking place to improve these systems. Yet, the applicability of adsorption for methanol-iodine solutions may be complicated by the fact that halogenated materials may oxidize and form acidic counterparts. Thus additional pollution control equipment and costly corrosion-resistant materials for construction may be required.

Distillation Recovery

Distillation consists of heating a solvent to vaporization and then condensing it. The condensed vapor is reused. On-site distillation systems may be a viable option for almost any amount of solvent use. Many commercial stills have capacities of 0.5 to 100 gal/hour. Larger units are often quite complex and require a steam supply. The capital cost is generally $5,000, plus $1,000 per gal/hour capacity. The operating costs include labor, energy, cooling water and maintenance parts. A moderately skilled operator should tend the apparatus about 10% of the operating time. Distillation may be difficult if solvents cannot be segregated. If two or more solvents are mixed, an off-the-shelf still system will often be unable to separate them. Many firms have a single solvent drain, thus separation of solvents may be difficult. In addition, distillation systems require permits that often cost over $250,000 and take 3-5 years to obtain.

Membrane Systems These systems are based on highly selective membranes that are more permeable to organic compounds than to air. They can achieve solvent recovery as high as 99.99% and can virtually eliminate VOC emissions. A typical configuration consists of a compressor, a condenser, and an array of spiral-wound membrane modules. The two-step recovery process combines condensation with a vapor separation performed by highly selective membranes under the driving force of a pressure differential. By efficiently recovering solvents under non-extreme conditions, membrane systems reduce the energy requirement and hence the cost of recovery operations. The most economical applications for these systems are those containing VOC concentrations of about 1,000 ppm. The systems create no secondary waste, recover VOCs for reuse without a steam or inert-gas-regeneration step, and circumvent the problems of condenser icing and the high energy costs associated with low-temperature condensation. In addition, they are compact and are easily integrated into existing processes. Typical annual operating costs are $10-35 per cubic fluid meter.

Membrane systems have been considered for recycling of other chemicals used in the semiconductor industry such as sulfuric acid. However, the ability to achieve parts per million to parts per billion purity levels with such as technique does not look promising. Today, the cost of using such a technology to achieve the purity levels demanded by industry is prohibitively expensive.

Catalytic Recovery Catalytic systems are frequently combined with thermal treatment to lower the temperature needed to burn VOCs. To destroy 95% or more of the total VOC content in any given stream, incineration alone requires temperatures above 1,200 (F. Adding a catalytic process cuts the temperature requirements in half. Catalytic systems also allow for safer incineration of compounds such as halogenated organics.