Scientists at the U.S. Department of Energy’s (DOE) Brookhaven National Laboratory have successfully demonstrated that autonomous methods can discover new materials. Artificial intelligence (AI)-driven techniques have led to the discovery of three new nanostructures, including a first-of-its-kind nanoscale “ladder.”The study was published today in scientific progress.
The newly discovered structure forms through a process called self-assembly, in which the molecules of the material organize themselves into unique patterns. Scientists at Brookhaven’s Center for Functional Nanomaterials (CFN) are experts in directing the self-assembly process, creating material templates to form ideal arrangements for applications in microelectronics, catalysis and more. The nanoladders and other new structures they discovered further broaden the range of applications for self-assembly.
“Self-assembly can be used as a nanopatterning technique, which is a driver of advances in microelectronics and computer hardware,” said CFN scientist and co-author Gregory Doerk. “These technologies have been pushing for higher resolution using smaller nanopatterns. You can get very small and tightly controlled features from self-assembled materials, but they don’t necessarily obey the kind of rules we have for circuits, such as …by directing self-assembly using templates, we can form more useful patterns.”
Staff scientists at CFN, a DOE Office of Science User Facility, aimed to build a library of self-assembled nanopattern types to expand their range of applications. In previous research, they demonstrated that new types of patterns can be achieved by mixing two self-assembling materials together.
“The fact that we can now create a ladder structure that no one has dreamed of before is amazing,” said CFN group leader and co-author Kevin Yager. “Traditional self-assembly can only form relatively simple structures, such as cylinders. bodies, flakes and spheres. But by mixing the two materials together and using the right chemical gratings, we discovered that entirely new structures are possible.”
Mixing together self-assembling materials has allowed CFN scientists to discover unique structures, but it has also created new challenges. As more parameters need to be controlled during self-assembly, finding the right combination of parameters to create new useful structures is a race against time. To accelerate their research, CFN scientists took advantage of a new AI capability: autonomous experimentation.
In collaboration with DOE’s Lawrence Berkeley National Laboratory’s Center for Applied Advanced Mathematics for Energy Research (CAMERA), Brookhaven scientists at CFN, and the National Synchrotron Light Source II (NSLS-II), another Brookhaven Lab science user facility The U.S. Department of Energy’s office has been developing an artificial intelligence framework that can autonomously define and execute all steps of the experiment. CAMERA’s gpCAM algorithm drives the frame’s autonomous decision-making. The latest study is the first time the team has successfully demonstrated the algorithm’s ability to discover new materials.
“gpCAM is a flexible algorithm and software for autonomous experiments,” said Berkeley Lab scientist and co-author Marcus Noack. “In this study, it was used particularly cleverly to autonomously explore different features of the model.”
“With the help of our colleagues at Berkeley Lab, we prepared the software and method to use, and we have now successfully used it to discover new materials,” Yager said. “We now know enough about the science of autonomy that we can easily translate materials problems into autonomous problems.”
To accelerate materials discovery using their new algorithm, the team first developed a complex sample with a range of properties for analysis. The researchers fabricated samples using a CFN nanofabrication facility and performed self-assembly in a CFN material synthesis facility.
“An old-school way of doing materials science is to synthesize a sample, measure it, learn from it, then go back and make a different sample and iterate the process,” Yager said. “Instead, we made a single sample with gradients for each parameter we were interested in. So a single sample is a large collection of many different material structures.”
The team then took the sample to NSLS-II, which produces ultra-bright X-rays to study the material’s structure. CFN operates three experimental stations in cooperation with NSLS-II, one of which is used for this study, the Soft Matter Interface (SMI) beamline.
“One of the strengths of the SMI beamline is its ability to focus the X-ray beam onto the sample with micron-scale precision,” said NSLS-II scientist and co-author Masa Fukuto. “By analyzing how these microbeams of X-rays are scattered by the material, we learn about the local structure of the material at the point of illumination. Measurements at many different points can then reveal how the local structure varies across gradient samples. In this work, we let the AI The algorithm chooses on the fly the point to measure next to maximize the value of each measurement.”
Since the sample is measured on an SMI beamline, the algorithm can create models of many different structural sets of the material without human intervention. The model updates itself with each subsequent X-ray measurement, making each measurement deeper and more accurate.
Within hours, the algorithm identified three key regions in the complex sample for CFN researchers to study more closely. Using the CFN electron microscopy facility, they fine-tunedly imaged these critical regions, revealing the rails and rungs of the nanoscale ladder, among other new features.
From start to finish, the experiment took about six hours. The researchers estimate that it would have taken them about a month to make the discovery using traditional methods.
“Autonomous approaches can dramatically accelerate discovery,” Yager said. “It’s essentially ‘tightening’ the usual loop of scientific discovery so that we cycle between hypothesis and measurement more quickly. However, in addition to speed, autonomous methods expand the scope of what we can study, meaning we can Solving more challenging scientific problems.”
“Going forward, we want to study the complex interplay between multiple parameters. We performed simulations using a CFN computer cluster that validated our experimental results, but they also suggest that other parameters, such as film thickness, can also play an important role ,” Dolk said.
The team is actively applying their autonomous research approach to more challenging materials discovery problems in self-assembly as well as in other classes of materials. Self-discovery methods are adaptable and can be applied to almost any research problem.
“We are now deploying these methods to the broad user community that came to CFN and NSLS-II for experiments,” Yager said. “Anyone can collaborate with us to accelerate their exploration of materials research. We expect this will lead to a range of new discoveries in the coming years, including in national priority areas such as clean energy and microelectronics.”
This research was supported by the U.S. Department of Energy’s Office of Science.
