Engineers Are Using AI to Predict How New, Unknown Materials Will Perform
Materials engineering is critical to the modern world in ways we almost never stop to think about. Understanding how solid structures behave at the nanoscale level is critical to modern advances in many fields, including semiconductors.
Researchers working at MIT and in Russia and Singapore are using AI to predict how strain will impact material performance and to explore which types of strain will create which effects. Some of you may be familiar with the term “strained silicon,” which refers to the process of stretching a layer of silicon over a substrate of silicon-germanium. Strained silicon, which was introduced in modern microprocessor manufacturing during the P4 era, improves overall CPU performance compared with non-strained silicon. But, as the MIT author notes, finding the exact degree and type of strain to use is exceedingly difficult.
Strain can be applied in any of six different ways (in three different dimensions, each one of which can produce strain in-and-out or sideways), and with nearly infinite gradations of degree, so the full range of possibilities is impractical to explore simply by trial and error. “It quickly grows to 100 million calculations if we want to map out the entire elastic strain space,” Li says.
The Wikipedia page for strained silicon itself indirectly hints at the complexity of these changes. It notes that while initial strained silicon work took the form I just described, later improvements to the technique involve additional complex processing steps. Finding the precise tools to further improve the overall performance of these materials is clearly a slow, painstaking process. This is also part of why the development time for new features and capabilities in semiconductor manufacturing (or, say, battery capacity improvements) tends to be as long and slow as it is. Many of the improvements we discuss when we talk about batteries or improved semiconductor technology are fundamentally material engineering improvements.
According to the research team, their neural network model for predicting strain was highly accurate. The team focused on diamond, which has a number of positive traits that would make it an excellent semiconductor if some of its negatives could be ameliorated. There’s also the potential for introducing higher amounts of strain in products that already use the approach, potentially transforming the base material in the process.
Whereas this study focused specifically on the effects of strain on the materials’ bandgap, “the method is generalizable” to other aspects, which affect not only electronic properties but also other properties such as photonic and magnetic behavior, Li says. From the 1 percent strain now being used in commercial chips, many new applications open up now that this team has shown that strains of nearly 10 percent are possible without fracturing. “When you get to more than 7 percent strain, you really change a lot in the material,” he says.
“This new method could potentially lead to the design of unprecedented material properties,” Li says. “But much further work will be needed to figure out how to impose the strain and how to scale up the process to do it on 100 million transistors on a chip [and ensure that] none of them can fail.”
The most interesting thing about approaches like this is whether or not they can scale to the point of becoming fundamentally new approaches to the way we perform materials research. In theory, an AI-powered research engine could tear through material permutations that would take a dedicated research team weeks or months to test. But ensuring that our models can properly anticipate how materials would deform under various conditions is challenging — the data set to “train” the AI would seem to be formidable, in the best of scenarios.
Still, even a model that could pare down the list of ideas to investigate from hundreds of millions to thousands would be a major breakthrough. There could come a time when the use of AI resources like this isn’t just expected but has become a functional requirement of continuing scientific advance.
Continue reading
Chromebooks Gain Market Share as Education Goes Online
Chromebook sales have exploded in the pandemic, with sales up 90 percent and future growth expected. This poses some challenges to companies like Microsoft.
Protect Your Online Privacy With the 5 Best VPNs
Investing in a VPN is a smart choice right now, but the options are vast. To help narrow things down a bit, we've rounded up five of our very favorite consumer services.
AMD’s New Radeon RX 6000 Series Is Optimized to Battle Ampere
AMD unveiled its RX 6000 series today. For the first time since it bought ATI in 2006, there will be some specific advantages to running AMD GPUs in AMD platforms.
New Intel Rocket Lake Details: Backwards Compatible, Xe Graphics, Cypress Cove
Intel has released a bit more information about Rocket Lake and its 10nm CPU that's been back-ported to 14nm.