AI helps materials scientists focus on promising new materials

AI helps materials scientists focus on promising new materials

AI and automation are accelerating science and chemistry by helping scientists choose which experiments to conduct and focus on promising new materials.

Why is this important: There is pressure on these areas to produce new materials faster and more cheaply to support and fuel technologies that could transform industries and economies.

The big picture: New materials and molecules are needed for the batteries, drugs, and semiconductors envisioned to support green networking, precise medicine, and the next generation of computing and communications.

  • “Ultimately, this new technology is driving the next revolution, perhaps the next big scientific revolution,” says Olexander Isayev, professor of chemistry at Carnegie Mellon University.
  • The US, China, EU and Japan all have initiatives underway to spur materials development by creating libraries of compounds that can be tested and potentially developed into new materials.
  • The United States led the world in publications in the field two decades ago, but China now ranks first in materials science research by this measure.

What is happening: It can take decades to bring a new material to market in a process that involves “almost artisanal science”, says Isayev.

  • But as some lab tasks are automated and AI is integrated into the analysis of scientific data, materials scientists and chemists are using machine learning and other tools to perform calculations and simulations that can help them. point to candidates for new catalysts, polymers and other materials with unique properties. Properties.
  • They also use AI models to remove noise in data generated in experiments or direct microscopes to areas of interest, reducing the time researchers can spend on them.

Enlarge: In a new study, researchers combined machine learning, theories and calculations of physical properties and experiments to identify new alloys.

  • There are so many possible combinations (10^50) of the elements typically used in alloys, including nickel, iron, cobalt, and copper, that it would be nearly impossible to go through them all by trial and error.
  • The researchers were interested in a particular type of alloys, called high entropy alloys, made up of several elements in similar proportions. They were also looking for alloys called invars that don’t expand or contract when the temperature changes, making them ideal for transporting and storing natural gas.

How it works: The team integrated data on different alloys – some dating back over 100 years – into an AI model that determines correlations between the properties of alloys and the elements they contain and generates hundreds of thousands of candidate materials. . A neural network then narrows that down to around 1,000 remaining candidates.

  • These are then evaluated on the basis of physical theories and calculations on the behavior of alloys, and around 20 compositions are proposed.
  • The three best compositions are selected by researchers and physically measured.
  • This data is fed back into the AI, which tries to improve it after learning the underlying physics, creating an active learning loop.

What they found: The researchers identified two new alloys six times through the loop.

  • It took two to three months, compared to the years of experiments typically needed to find a new alloy, says Ziyuan Rao, postdoctoral researcher at the Max Planck Institute for Iron Research in Germany and co-author of the article, published Thursday in the newspaper Science.

Yes, but: Finding a material or a chemical is an obstacle. In fact, making it is another.

  • It’s much harder to train AI models to predict how to synthesize a material, in part because there’s no data on what can’t be synthesized, says Keith Butler, senior lecturer at the Queen Mary University of London.
  • Researchers are starting to use AI to try to optimize the manufacturing of materials like perovskites, which are used for advanced solar cells.
  • The National Science Foundation has given $20 million over five years to the new Center for Computer-Aided Synthesis at Notre Dame University, which aims to solve the problem.

What they say : “Despite the pace of progress in this area, the revolutionary potential of these approaches has yet to be realized,” reads a description of a conference on the subject being held this month.

  • Materials science lacks the large datasets that power AI advances in genomics and other fields.
  • The high cost of time and money to conduct experiments means that less data is available to train AI systems – and much of what is available is collected from different experiments or under various conditions, and is distributed among institutions or locked away in proprietary databases.
  • The work on alloys is “impressive” because they were able to achieve results with sparse data, says Isayev, who uses AI to identify new materials and predict the properties of chemicals for solar energy technology and drug design.

What to watch: Another AI model – large language learning models that can write text – could be coming to materials science.

  • “I think next year is what’s going to be very hot in materials science and physical science,” says Gabriel Gomes, a CMU professor who uses machine learning to develop new chemical reactions and new catalysts.

#helps #materials #scientists #focus #promising #materials

Leave a Comment

Your email address will not be published. Required fields are marked *

Prendre rendez-vous en ligne