In a recent study, researchers used AI to find properties and experiments to identify new alloys. Without AI, sifting through the vast number of possible combinations of elements is a daunting task. “As manufacturing technologies advance, the compositional and processing space of alloys increases,” Mohadeseh Taheri-Mousavi, a professor of materials science and engineering professor at Carnegie Mellon University, who was not involved in the study, told Lifewire in an email interview. “For example, now you find alloys with more than ten alloying elements in the body of, e.g., airplanes. The design of this alloy and choosing the concentration of each alloying element to achieve those properties is beyond the human mind as it is a high-dimensional space.”
Better Gadgets Through AI
AI is already being used to discover new materials. Tech consultant and IEEE Member Yale Fox told Lifewire in an email interview that scientists have found ion-selective membranes using AI that are cheaper and more effective and can be used for many applications like wastewater treatment or biomedical devices. Researchers have also found several new solid-state materials that conduct lithium, which can be used to improve modern battery technology. “The most active space right now is in batteries; how to make them charge faster, hold the charge for longer, be more lightweight, environmentally friendly, etc.,” Fox added. “This has positive implications for electric vehicles such as cars and airplanes.” New semiconductors created with AI could enable faster computing speeds or more powerful artificial intelligence algorithms, Adnan Masood, chief AI architect at UST, said in an email. By better understanding how different materials behave under various conditions, scientists can develop more effective treatments for diseases like cancer or create safer nuclear reactors. Artificial intelligence has also helped build novel alloys, semiconductors, ceramics, and other types of materials by allowing scientists to process large amounts of data quickly and efficiently. “The benefits of new materials include improved performance, lighter weight, and increased strength,” Masood added. “AI can help discover lighter and stronger materials by analyzing data from experiments and simulations, and by optimizing material compositions through simulation of trial-and-error methods, instead of creating these which can take forever.” AI accelerates discovery by taking experimental results and guiding researchers on where and how to look for the most likely conditions that can lead to new materials or better processes, Miguel Modestino, a professor in the Department of Chemical and Biomolecular Engineering of New York University, told Lifewire via email. “This can significantly reduce the time for new discovery, as well as potentially reduce energy utilization and material waste,” he added.
Finding New Materials to Benefit the Environment
Modestino’s lab combines AI with autonomous research tools to run hundreds to thousands of experiments per day to decrease the time from idea to discovery by more than 100 times. The researchers seek to discover new chemical reactions that can produce existing materials in new, more efficient ways that reduce the negative environmental impact of current approaches to manufacturing chemicals. Most of our everyday items are made from materials produced by the chemical industry, Modestino said. His group has been focusing on the manufacture of Nylon 66. More than five million tons of nylon are produced annually, and 2 million of those are Nylon 66, including textile fibers we use in winter jackets and plastic car parts that make our vehicles lightweight and fuel-efficient. Modestino said the chemicals used to make nylon and most chemical products are produced through fossil fuel combustion, resulting in 10% of global energy consumption and 20% of carbon emissions from industry. “Using AI, we discovered a new, more energy-efficient method of manufacturing nylon 66 precursors based on electricity instead of fossil fuel combustion,” he added. “We were able to do this by training machine learning models with information from a limited number of experiments and accelerating the discovery of the best conditions for this process.” Update 11/2/2022: Corrected the spelling and bio link for the source in paragraph three.
