INDUSTRY NEWS

Sematech Watch

Innovative energy study designed to trim fab costs, maintain yields

Sematech is wrapping up an innovative study that aims to help participating semiconductor manufacturers cut their operating costs by reducing energy consumption at their fabs without sacrificing high yields. Working with a provider of energy management services, the consortium completed the first phase of the study last year. Data collection was to end last month, and Sematech expects to complete its final report by mid-1998, says Joe White, manager of the international energy project.

Two consortium members, Advanced Micro Devices and National Semiconductor, participated in the pilot program. AMD's Fab 25 in Austin, TX, and National's plant in Arlington, TX, were both assessed under the audit program developed by Sematech and Planergy, an energy management services firm based in Austin. Fab areas that were examined included the cleanroom, support areas such as ultrapure water operations, and office space. At Fab 25 a total of 150 process tools and 120 sample points were examined, with energy use measured on at least one tool from each step in the process. HVAC operations, exhaust, and nitrogen supplies were also spot-checked.

The project has two primary goals. The first goal is to establish a "replicable template" for assessing energy efficiency at any fab around the world. The second is to offer feedback to equipment suppliers that can be used to build more energy-efficient tools. According to Sematech and Planergy, the study breaks new ground because participants are measuring energy use in an actual fab operating environment, producing a "comprehensive snapshot" of energy consumption. The fact that participants are sharing the information with toolmakers also makes the project unique.

AMD discovered that modifications made to HVAC, lighting, compressed air, and tool support in Fab 25 could result in annual energy cost savings of $579,615, minus an initial implementation cost of $252,500. After paying the setup costs, AMD would realize the full savings annually. Planergy reported that National found it could save up to $700,000 annually based on an implementation cost of $2 million.

The six-month pilot study "pretty much validated that Fab 25 is a state-of-the-art facility," says David Bennett, AMD's technology transfer manager. One of the main conclusions was that "the extra money we put into making the fab efficient from a design standpoint was probably well spent." The study did generate "four or five specific recommendations" for improving energy consumption, although he declined to identify them.

"Most of the fabs are obviously extremely concerned with defects and their yields," emphasizes Richard Patton, vice president of engineering for Planergy. "Because of that, they really are prone to install whatever possible equipment they think will improve their yields. Sometimes that can lead to inefficiencies in manufacturing."

Although not necessarily yield related, these inefficiencies can lead to increased operating costs, which in turn can raise product costs and reduce profits, Patton asserts. A typical example, he adds, is "dramatic oversizing" of facility equipment such as HVAC. "It might also apply to things like the size of the vacuum pumps on process tools. Those vacuum pumps run almost all the time at almost a constant load."

Project manager White, an assignee from Rockwell who works in the consortium's environmental, safety, and health department, says it's not clear how chipmakers could reduce energy consumption in tools as they could in their facilities "because the tools use whatever [amount of energy] they're going to use, and they're probably going to do that for the life of the tool." The study's aim "is to drive down energy use by the next generation of tools," White points out, adding that equipment suppliers participate in the study through a technical advisory board.

"I think there's a tendency by most of the users to maintain tools at their operating environment, whether it's vacuum or temperature or the like, even when the tool is not being used," White says. "You're actually consuming energy there. You're not building a product. You're doing that because changing the tool and bringing it back to the proper process ambient raises the fear that there could be a yield impact by not maintaining the tool."

Survey participants are scrutinizing energy use in nine different equipment categories: patterning, thermal, thin film, dry etch, metrology, material handling, CMP, ion implant, and wafer cleaning. Tools consuming the most energy, White says, have tended to fall in the plasma process equipment areas such as thin film, etch, and ion implant. Cleaning and thermal processing appear to consume a large amount of energy as well.

The project manager says there is a "huge discrepancy" between the design parameters for a tool coming into the fab and the actual measured tool loads. In some cases the difference is a factor of three or more, asserts White. "In practically all cases it's two times or higher. The magnitude of the difference was a big surprise to a lot of people."

Bennett of AMD concurs with White about this finding from the pilot program. "One of the more significant points was the actual demands of the wafer processing equipment within the fab. The exhaust rates and the power rates of some of the equipment didn't match up with the supplier-furnished specifications for those tools. It gave us a much better insight into the actual demands and usage of the tools in the fab." This information will allow AMD to better design the infrastructure and support facilities for the equipment, he points out, adding, "strictly from a cost-savings standpoint, we benefited."

White sees no direct correlation between fab energy consumption and defect densities, although he asserts, "I'm sure there is a relationship there." The connection is probably most evident in the photolithography area of the fab. "One of the things we're seeing is that certainly the environmental conditions in the litho area have a substantial bearing on the variability of the process and ultimately on yield."

He explains that humidity control in the lithography area is "typically very critical." Maintaining relative humidity "in the area of 40%" depends a great deal on outside air conditions. "Energy consumption in the fab comes down to how much exhaust you have. When you have exhaust you have makeup air, and when you have makeup air you're bringing in air from the outside, which has to be conditioned for humidity, contaminants, and temperature.

"There's a relationship between maintaining the environmental conditions for your litho area versus keeping your energy costs of the facility in line," White continues. He adds that some chipmakers "have gone so far as to split the airflow for the litho area away from the rest of the fab." One additional objective of the study is to assess the sensitivity of energy consumption to climate in the hope of determining a global benchmark suitable for fabs around the world.

White declined to say how much the project costs. Fourteen international fabs are participating in the current phase of the study. Sematech published a guideline document "based on some of the experience gained from the pilot surveys." Best-practices teams or users groups are in the process of forming and will continue to do so through 1998 and into 1999, White says.

The project "is open in terms of the information being fed back to the fabs and equipment suppliers," White says. Because the results are proprietary "there is no clear decision at this point how to make this information available to the industry at large or even whether we will share it. How that information gets released in a general way is not clear."

AMD is already incorporating the early findings into the design of a fab in Dresden, Germany, that will run its first silicon in the latter part of 1998, Bennett says. "One of the issues in that country is the quality of the power that you can get from the government power plants. Typically, what you have to do is build your own cogeneration facilities. We're in the process of doing that, and we're able to more precisely design that system now that we know the actual energy demands inside the fab."

Meanwhile, back in the States, AMD can point to the environmental benefits of having an energy-efficient fab, Bennett states. "Less energy means fewer emissions from the power plants creating that energy. That's one obvious benefit. The city of Austin power plant is emitting less, if you will, because Fab 25 is more efficient."

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