Researchers develop ‘scientific’ AI to self-pilot materials discovery experiments

A new AI-powered method for efficient data collection could help scientists overcome complex challenges in materials discovery and design, enabling unprecedented accuracy and speed.

The method could also pave the way for “autonomous experiments,” in which the intelligent algorithm can take a set of data and then set the parameters for the next set of measurements to be made. This could enable the rapid discovery of new materials, the researchers say, leading to greater advances in combating climate change, advancing quantum computing, and speeding up drug design.

The research, published in npj computer materialsis the result of a collaboration between computer science and materials science researchers at Stanford University and the Department of Energy’s SLAC National Accelerator Laboratory.

Enable the rapid discovery of new materials

The advent of high-powered computers and computer modeling software capable of running large-scale molecular dynamics simulations was a huge boost for materials scientists. Suddenly, instead of spending thousands of human hours on manual trial-and-error experiments, it became possible to model these materials and simulate their potential behavior using computers and extensive theoretical calculations.

Despite this progress, materials discovery remains a very long and expensive process. The scale of possible materials is staggering: it is estimated that with just four elements, it is possible to create more than 10 billion possible materials.

For scientists interested in targeted materials discovery – that is, finding materials with certain desired properties – traditional discovery techniques are still very slow, especially if the researcher has a goal more complex than maximizing or minimizing a simple property.

In their new paper, the Stanford and SLAC researchers present a new approach that can address more complex design goals, such as discovering the conditions for synthesizing nanoparticles of a certain size, shape, or composition. It can also learn and improve from each experiment it observes, using AI to suggest next steps based on the data it has read so far.

The approach is based on the so-called Bayesian Algorithm Analysis (BAX), developed by the study’s author Willie Neiswangerwho was a postdoctoral researcher in computer science at Stanford during the research period and is now an assistant professor of computer science at the University of Southern California.

With this approach, a researcher can transform their complex design goal into a “recipe” or “shopping list” type filtering algorithm that is automatically translated into one of three BAX-based data collection strategies. This avoids many of the time-consuming challenges associated with previous methods, resulting in a process that excels in situations where multiple physical properties need to be considered.

“Our method allows for the specification of complex objectives, enabling automatic optimization over a large design space, which increases the probability of finding amazing new materials,” said Sathya ChitturiPhD student at SLAC and Stanford who led the research. “The Bayesian algorithm execution framework allows you to capture the subtleties of materials design tasks in an elegant and simple way.”

Better materials for a better world

To demonstrate the usefulness of this method, the research team applied it to a variety of custom targets for nanomaterial synthesis and magnetic material characterization. The results suggest that this new method is significantly more efficient than other popular modern techniques, especially in complex scenarios.

“By combining advanced algorithms with targeted experimental strategies, our method makes the process of discovering new materials easier and faster,” said Chris Tassone, director of the Materials Science Division at the Stanford Synchrotron Radiation Lightsource (SSRL) at SLAC, said: “This can lead to new innovations and applications in many industries.”

Continue reading below…

The team suggests that this method could enable the more efficient design of new materials with specific catalytic properties, which could improve the chemical processes used in manufacturing, making them more energy-efficient and sustainable while producing less waste. Similarly, the method could be applied to the creation of tailor-made drug delivery systems that can improve the targeting and release of therapeutics, which could result in increased efficacy and reduced side effects.

The researchers say their new method is designed to be user-friendly and open source, allowing diverse groups of scientists around the world to use it and easily adapt it to their own research.

As the research team continues to look for ways to integrate this framework into more experimental and simulation-based research to further demonstrate its effectiveness, scientists in SLAC’s Machine Learning Initiative (MLI) have begun investigating its application in larger-scale materials simulations. Neiswanger, alongside collaborators in the MLI, has already published an additional document describing how the application of BAX could also help optimize the performance of particle accelerators.

Reference: Chitturi SR, Ramdas A, Wu Y, et al. Targeted materials discovery using Bayesian algorithm execution. npj Comput Mater. 2024;10(1):156. doi: 10.1038/s41524-024-01326-2

This article is a rehash of a Press release issued by the SLAC National Accelerator Laboratory The content and length of the document have been modified.