Artificial intelligence discovers new nanostructures

Scientists at the US Department of Energy’s (DOE) Brookhaven National Laboratory have successfully demonstrated that independent methods can discover new materials. Artificial intelligence (AI) technology has led to the discovery of three new nanostructures, including a first-of-its-kind nanoscale “ladder”. Search was Posted today in Science advances..

The newly discovered structures were formed by a process called self-assembly, where molecules of matter organize themselves into unique patterns. Scientists at the Brookhaven Center for Functional Nanomaterials (CFN) are experts in guiding the self-assembly process, creating templates for materials to form desirable arrangements for applications in microelectronics, catalysis, and more. Their discovery of nanoscales and other novel structures expands the range of self-assembly applications.

“Self-assembly can be used as a nanostreaming technique, an engine for progress in microelectronics and computers,” said CFN scientist and co-author Gregory Duerke. “These technologies are always pushing for higher fidelity with smaller nanopatterns. You can get very small, tightly controlled features out of self-assembling materials, but they don’t necessarily adhere to the kind of rules we make for circuits, for example. By directing self-assembly with a template, we can form more beneficial patterns.”

Scientists working at CFN, a DOE affiliate of the Office of Science, aim to build a library of types of self-assembled nanopatterns to expand their applications. in previous studiesthey showed that new kinds of patterns were made possible by mixing two self-assembling materials together.

“The fact that we can now create a staircase structure that no one had ever dreamed of before is amazing,” said CFN Group President and co-author Kevin Yager. “Traditional self-assembly can only form relatively simple structures such as cylinders, sheets, and spheres. But by blending two materials together and using just the right chemical lattice, we have found that entirely new structures are possible.”

Mixing together self-assembling materials enabled CFN scientists to uncover unique structures, but it also created new challenges. With so many parameters to control the self-assembly process, finding the right set of parameters to create new and useful structures is a battle against time. To speed up their research, CFN scientists have taken advantage of a new AI capability: Independent experimentation.

In collaboration with the Center for Advanced Mathematics for Energy Research Applications (CAMERA) at DOE’s Lawrence Berkeley National Laboratory, Brookhaven scientists at CFN and the National Synchrotron Light Source II (NSLS-II), another DOE Office of Science user facility at the Brookhaven Laboratory, we have It was Developing an artificial intelligence framework He can independently define and implement all steps of the experiment. The camera’s gpCAM algorithm drives the framework’s independent decision-making. The latest research is the team’s first successful demonstration of the algorithm’s ability to discover new materials.

“gpCAM is a flexible algorithm and software for subjective experiments,” said Marcus Nowak, Berkeley Lab scientist and co-author. “It was used particularly ingeniously in this study to independently explore the various features of the model.”

“With the help of our colleagues at Berkeley Lab, we had this program and methodology ready to go, and now we’ve used it successfully to discover new materials,” Yager said. “We have now learned enough about independent science that we can easily take the materials problem and turn it into an independent problem.”

To speed up material detection with the new algorithm, the team first developed a complex sample with a spectrum of properties for analysis. The researchers made the sample using the CFN nanofabrication facility and carried out self-assembly in the CFN material fabrication facility.

“The old-school method in materials science is to collect a sample, measure it, learn from it, and then go back to make a different sample and keep repeating that process,” Yager said. “Instead, we make a sample that contains a gradient for each variable we’re interested in. So this single sample is a broad set of many distinct physical structures.”

Next, the team brought the sample to NSLS-II, which generates ultra-bright X-rays to study the material’s structure. CFN operates three experimental stations in partnership with NSLS-II, one of which was used in this study, the Soft Material Interfaces (SMI) radiation line.

“One of the strengths of the SMI beamline is its ability to focus the X-ray beam on the sample down to microns,” said NSLS-II scientist and co-author Masa Fukuto. “By analyzing how these microbial X-rays are scattered by the material, we learn about the local structure of the material in the bright spot. Measurements at many different points can then reveal how the local structure varies across the gradient sample. In this work, we let an AI algorithm You choose, on the spot, where to measure next to maximize the value of each measurement.”

While measuring the sample on the SMI beamline, the algorithm created, without human intervention, a model of the material’s many and varied structures. The model updated itself with each subsequent X-ray measurement, making each measurement more precise and accurate.

Within hours, the algorithm identified three key areas in the complex sample of CFN researchers to study more closely. They used the CFN electron microscopy facility to image those key regions in exquisite detail, revealing rods and rungs of a nanoscale, among other new features.

From start to finish, the experience took about six hours. The researchers estimate that it would have taken about a month to make the discovery using traditional methods.

“Autonomous methods can greatly speed up discovery,” Yager said. “They are essentially ‘tightening up’ the usual discovery loop of science, so that we can move between hypotheses and measurements more quickly. However, beyond speed, independent methods increase the range of what we can study, which means we can tackle more challenging scientific problems.”

“Going forward, we want to investigate the complex interplay of multiple parameters. We performed simulations using the CFN computer suite that validate our experimental results, but they also suggested how other parameters, such as film thickness, may also play an important role,” said Doerk.

The team is actively applying their independent research method to more challenging material discovery problems in self-assembly, as well as other classes of materials. Independent discovery methods are adaptable and can be applied to almost any research problem.

“We are now publishing these methods to the broad community of users who come to CFN and NSLS-II for experiments,” Yaeger said. “Anyone can work with us to accelerate exploration of their materials research. We expect this to enable a range of new discoveries in the coming years, including in areas of national priority such as clean energy and microelectronics.”