AI algorithm discovers a new nanostructure, reducing research time from one month to six hours

AI has achieved a new achievement!

It only takes 6 hours to discover new nanostructures. Using traditional methods, it would take at least 1 month to complete this task.

This result was published in the Science sub-journal Advance.

△ Scanning electron microscope image depicts new nanostructure discovered by AI

Experiment from U.S. Department of Energy (DOE) Brooke Black At Wen National Laboratory, researchers used AI-driven technology to discover three new nanostructures.

One of the structures is still a very rare "ladder" type.

They used an algorithm-driven framework called gpCAM for the entire process, which can independently define and execute all steps of the experiment.

After the CEO of a digital product start-up company read the paper, he boldly made some remarks to stimulate the popularity of ChatGPT:

I bet that in the next five years, AI will transform engineering and materials science. , pharmaceuticals, will dwarf ChatGPT’s influence.

Discover three new nanostructures

Discover three new nanostructures, all through a process called self-assembly ) formed through the process.

Self-assembly refers to a technology in which basic structural units, such as molecules, nanomaterials, microns, etc., spontaneously form an ordered structure.

The structure formed is stable and the geometric appearance has certain rules.

The properties of self-assembling materials are small and rigorous Control, using this technique, enables smaller nanopatterns with improved resolution.

△Co-authors Kevin Yager (left) and Gregory Doerk (right).

Let me introduce CFN. The goal of this organization is to establish a library of self-assembled nanopatterns to expand its application scope.

Previously, researchers demonstrated that new types of nanopatterns could be formed by mixing two self-assembling materials.

However, traditional self-assembly has only been able to form relatively simple structures, such as cylinders, sheets or spheres.

But this time, the researchers discovered that among the three new nanostructures, there is a ladder structure!

In other words, once the appropriate chemical grating (spectrum splitter) is used, it is completely possible to discover new structures by mixing two self-assembly materials.

New discoveries bring surprises, but also new challenges in the experimental process:

The entire self-assembly process requires the control of many parameters, and a suitable combination of parameters must be found to create new and useful Structure.

This process is often very long.

To accelerate research, CFN researchers have introduced a new AI capability:

Autonomous Experimentation.

Accelerate from 1 month to 6 hours to complete

You might as well listen to how the traditional method finds the appropriate parameter combination~

First, the researchers will synthesize a sample, then measure it and learn useful information from it.

Then, make a different sample, measure it, learn from it...

In short, just keep repeating this process until you solve the problem you want to solve.

Why not give AI such a tedious and repetitive task a try?

In fact, CFN and the National Synchrotron Light Source II (NSLS-II), the Office of Science User Facility in the same laboratory, have been developing an AI framework that can automatically define and execute all steps of an experiment.

Time was tight, and CFN finally chose to cooperate with the Center for Advanced Mathematics in Energy Research and Applications (CAMERA) of the U.S. Department of Energy.

CAMERA’s gpCAM algorithm-driven framework enables autonomous decision-making. During the collaboration, gpCAM was used to autonomously explore different features of the model.

The latest research is the first time the team has successfully demonstrated the algorithm's ability to discover new materials.

After gpCAM joined, the research team first used CFN's nanoprocessing equipment to prepare a complex sample with a series of characteristics; then it self-assembled and analyzed in CFN's material synthesis equipment.

This sample has spectral properties and also contains gradients for each parameter of interest to the researchers.

In this way, a single sample becomes a huge collection of many different material structures.

This sample was sent to NSLS-II for structural study using ultra-bright X-rays.

ΔX-ray scattering data (left) showing key areas of the sample with corresponding scanning electron microscopy image (right)

## When the #ray is run, gpCAM creates models of multiple different structures of a material without human intervention.

gpCAM also needs to make measurements more insightful. Simply put, it uses AI algorithms to select which point to measure next, making each measurement more accurate.

△Soft Matter Interface (SMI) beamline of NSLS-II.

From start to finish, the AI ​​algorithm took a total of 6 hours.

Assuming traditional methods are used, researchers would have to spend at least a month in the laboratory.

In these 6 hours, the algorithm has identified three key areas in the complex sample.

The researchers used CFN electron microscopy equipment to image these three areas in precise detail, revealing nanorails and gradients, among other new features.

"Autonomous experimentation can greatly accelerate discovery." Kevin Yager, CFN researcher and co-author of the new study, "This is in the process of 'tightening' the usual discovery cycle in science, reducing the time between hypothesis and measurement ."

Yager also said that in addition to speed, autonomous experiments increase the scope of research, which means that more challenging scientific problems can now be attempted.

In other words, the independent experimental method is adaptive and can be applied to almost all research questions.

Researchers are already looking forward to studying the complex interactions between multiple parameters. What do you expect from this?

Reference link:

[1]​

​//m.sbmmt.com/link/8e5231f0eadafd174b670e838e42d97d​​

[2] ​https://www.science.org/doi/10.1126/sciadv.add3687[3]​https://www.bnl.gov/newsroom/news.php?a=120993#:~:text=The artificial intelligence (AI)-,published today in Science Advances.​

The above is the detailed content of AI algorithm discovers a new nanostructure, reducing research time from one month to six hours. For more information, please follow other related articles on the PHP Chinese website!