Researchers at the Brookhaven National Laboratory of the U.S. Department of Energy (DOE) have successfully shown that autonomous approaches can find novel materials. Three new nanostructures were found as a result of the artificial intelligence (AI)-driven approach, including a first-of-its-kind nanoscale “ladder.” Science Advances published the research.
The newly identified structures were created by a procedure known as self-assembly, in which the molecules of a substance arrange themselves into distinctive patterns. The Center for Functional Nanomaterials (CFN) at Brookhaven is home to scientists who are experts at controlling how materials assemble themselves into the kinds of configurations that are desirable for use in microelectronics, catalysis, and other fields. The breadth of self-uses assembly’s is further expanded by their discovery of the nanoscale ladder and other novel structures.
Gregory Doerk, a CFN scientist and co-author, said: “Self-assembly can be employed as a tool for nanopatterning, which is a driver for developments in microelectronics and computer hardware.” “These technologies are constantly aiming for smaller nanopatterns and higher resolution. Self-assembling materials can produce incredibly small and precisely controlled features, but they are not always subject to the same set of principles as, say, circuits. We can create more useful patterns by employing a template to guide self-assembly.
Building a library of self-assembled nanopattern types is a goal of the staff scientists at CFN, a DOE Office of Science User Facility, in order to increase the scope of their potential uses. In earlier research, they showed that combining two self-assembling materials enables the creation of novel designs.
The CFN group leader and co-author Kevin Yager commented, “It’s incredible that we can now build a ladder structure, which no one has ever conceived of before. The only relatively basic structures that can be created with traditional self-assembly are cylinders, sheets, and spheres. The correct chemical grating and the combination of two materials, however, enable the creation of completely new structures, as we have discovered.
CFN researchers have discovered novel structures by combining self-assembling materials, but this has also presented new difficulties. Finding the ideal combination of factors to generate novel and useful structures requires a race against time because there are so many additional parameters to control in the self-assembly process. Scientists at CFN used a brand-new AI feature called autonomous experimentation to advance their studies.
The National Synchrotron Light Source II (NSLS-II), another DOE Office of Science User Facility at Brookhaven Lab, and the Center for Advanced Mathematics for Energy Research Applications (CAMERA) at DOE’s Lawrence Berkeley National Laboratory have collaborated to create an AI framework that can autonomously define and carry out all the steps of an experiment. The autonomous decision-making of the framework is controlled by CAMERA’s gpCAM algorithm. The most recent study represents the group’s first effective demonstration of the algorithm’s capacity to find novel materials.
Marcus Noack, a scientist at Berkeley Lab and a co-author of the paper, described gpCAM as “a flexible algorithm and software enabling autonomous experimentation.” It was employed especially cleverly in this work to allow the model to independently explore various characteristics.
We were able to successfully employ this software and methods to uncover novel materials thanks to assistance from our colleagues at Berkeley Lab, according to Yager. We have enough knowledge of autonomous science now to be able to quickly transform a materials problem into an autonomous one.
The team first created a complex sample with a spectrum of attributes for investigation in order to speed up materials discovery using their novel method. The sample was created by researchers using the CFN nanofabrication facility, and the self-assembly process was completed at the CFN material synthesis facility.
In the past, material scientists would create a sample, measure it, and learn from it before creating another sample and repeating the process, according to Yager. Instead, we created a sample that contains a gradient of each parameter that is relevant to us. Therefore, that one sample is a huge collection of numerous unique material structures.
The team next took the sample to NSLS-II, a facility that produces ultrabright x-rays for analysing the structural properties of materials. In collaboration with NSLS-II, CFN runs three experimental stations, one of which, the Soft Matter Interfaces (SMI) beamline, was utilised in this investigation.
According to NSLS-II researcher and co-author Masa Fukuto, “one of the SMI beamline’s strengths is its ability to focus the x-ray beam on the sample down to microns.” We discover information about the local structure of the material at the lit point by examining how these microbeam x-rays are scattered by the substance. The local structure’s variations within the gradient sample can then be seen by measurements taken at numerous locations. In order to optimise the value of each measurement in this work, we let the AI programme decide instantly which spot to measure next.
Without any assistance from a human, the algorithm built a model of the material’s extensive and varied collection of structures while the sample was measured at the SMI beamline. With each consecutive x-ray measurement, the model modified itself, improving the accuracy and insight of every reading.
Within a few hours, the algorithm had pinpointed three crucial regions in the complicated sample that the CFN researchers should pay more attention to. They were able to photograph those important regions with the CFN electron microscopy facility in incredibly fine detail, revealing, among other unique features, the rails and rungs of a nanoscale ladder.
The experiment took around six hours to complete from beginning to end. According to the researchers, it would have taken around a month to achieve this finding using conventional techniques.
According to Yager, “autonomous approaches can significantly speed up discovery.” It basically involves ‘tightening’ the typical scientific discovery loop to make the transition between hypotheses and measurements happen faster. However, autonomous approaches broaden the area of what we can investigate in addition to being faster, allowing us to take on more difficult scientific challenges.
“Going on, we aim to look at the intricate interactions between many aspects. The CFN computer cluster was used to do simulations that supported our experimental findings while also suggesting that other factors, such film thickness, might also be crucial, according to Doerk.
The team is actively using its autonomous research methodology to tackle increasingly difficult material discovery issues in other kinds of materials as well as self-assembly. Autonomous discovery techniques are flexible and can be used to solve almost any research issue.
We are currently making these approaches available to the large user base that visits CFN and NSLS-II to conduct research, according to Yager. Anybody can collaborate with us to quicken the investigation of their material research. In the upcoming years, we anticipate that this will enable a plethora of new discoveries, particularly in high-priority fields like renewable energy and microelectronics.
