This article is part of an exclusive IEEE Journal Watch series in partnership with IEEE Xplore.
Many technology industries are rapidly adopting AI and virtualization tools, but there are some concerns about the amount of energy required to run these tools. But is it possible that, in some situations, computer simulations have smaller environmental impacts than real-world experiments?
Lam Research, one of the world’s largest semiconductor manufacturing companies, conducted a detailed internal analysis of its R&D operations to answer this question. As a result, despite the large amount of energy required for AI modeling and virtualization, depending on what aspects of semiconductor research are being done, the carbon footprint is lower compared to actual experimentation. It has been shown that this can reduce the amount of water by 20-80%.
In 2021, Lam Research set a goal to operate on 100% renewable electricity by 2030 and achieve net-zero emissions by 2050.
David Fried, corporate vice president of Semiverse Solutions at Lam Research, said of his team in achieving this net-zero goal: “This is the information we need to test different processes and measure the real carbon footprint of our R&D activities to make the most meaningful changes to our business.”
In particular, the company looked at carbon emissions associated with research and development of equipment and processes needed for nanoscale manufacturing of semiconductors such as integrated circuits.
“The fabrication of these microscopic devices is a highly complex manufacturing process spanning hundreds of specialized steps that have been extensively refined through the research and development process, nearly half of which involve etching with nanoscale precision. and complex chemical plasma processes such as and deposition,” Freed says. “This requires large amounts of resources in the form of materials, chemicals, gases and high energy consumption.”
Replace lab tests with virtual tests
Fried’s team aims to reduce the carbon footprint associated with traditional research and development conducted in physical laboratories using AI and virtualization, which are used to replace laboratory experiments wherever possible. We tried to compare it to the same study. They evaluated various areas of hardware prototyping, process optimization, and recipe development for wafers, which are round, thin disks used in semiconductor manufacturing.
In each case, they compared the carbon footprint of the computational modeling with that of the physical experiment. The results are described in a study published in the November issue of IEEE Transactions on Semiconductor Manufacturing.
The research team reports that while their simulations yielded the same results as their laboratory experiments, their simulations significantly reduced carbon dioxide emissions. In many research scenarios it was around 20%, but in some scenarios, such as plasma ion simulations, carbon emissions were as high as 80%. percent.
“Simulations appear to be less resource-intensive than physical experiments in most cases, so researchers are encouraged to minimize experiments and look for opportunities to use computational methods to solve problems. I recommend it,” Freed said.
Remarkably, the study revealed that manufacturing just one full-loop wafer produces approximately 1,500 kilograms of carbon dioxide over its lifetime. By comparison, virtualization and AI tools running on high-end computers would need to perform 27,000 hours of calculations, or just over three years, to achieve the same level of carbon emissions.
Freed said the environmental benefits of the simulation include reducing carbon emissions by conserving other critical resources such as water and chemicals widely used in semiconductor manufacturing, and by stopping the release of air pollutants. He points out that this is not just a reduction.
When it comes to running the company, Freed says his team has also found that simulation saves time, reduces costs and increases collaboration between teams.
“[Taken together]these promising results further demonstrate that virtual twinning and simulation will be a game-changer for semiconductor manufacturing,” he says.
From an article on your site
Related articles on the web
