Manufacturing Bits: April 17

Finding metallic glass
Using machine learning techniques, a group of researchers have accelerated the discovery of an alloy called metallic glass.

Northwestern University, the Department of Energy’s National Accelerator Laboratory and the National Institute of Standards and Technology (NIST) have devised a shortcut for discovering and improving metallic glass.

In metallic glass, the atoms in the materials are arranged like glass in a window. The material is stronger and lighter than steel. It also stands up to corrosion and wear, making metallic glass ideal for protective coatings and alternatives to steel.

But it’s difficult to find and develop these materials. Over the years, scientists have investigated about 6,000 combinations of ingredients that make up metallic glass, according to researchers.

In response, the group used advanced metrology techniques at SLAC’s Stanford Synchrotron Radiation Lightsource (SSRL). It combines that with machine learning technology, which are fast pattern matching algorithms.

Researchers were able to screen hundreds of sample materials at a time, allowing them to discover three new blends of ingredients that form metallic glass. They were able to find them 200 times faster than it could be done before.

The group started with a trove of materials. This included the results of 6,000 past experiments. Then, they sifted through the data with advanced machine learning algorithms. This, in turn, enabled researchers to craft two sets of sample alloys. These alloys were scanned using SLAC’s SSRL X-ray beam.

They fed the results in the machine learning algorithms again. The process was then repeated two more times. With this method, the success rate for finding metallic glass had increased from one out of 300 or 400 samples tested to one out of two or three samples tested.

All told, researchers found three different combinations of ingredients. Two of those had never been used before. The combination formed the basis of metallic glass.

“It typically takes a decade or two to get a material from discovery to commercial use,” said Chris Wolverton, the Jerome B. Cohen Professor of Materials Science and Engineering in Northwestern’s McCormick School of Engineering, on the university’s Web site. “This is a big step in trying to squeeze that time down. You could start out with nothing more than a list of properties you want in a material and, using AI, quickly narrow the huge field of potential materials to a few good candidates.”

Apurva Mehta, a staff scientist at SSRL, added: “We were able to make and screen 20,000 in a single year. The unique thing we have done is to rapidly verify our predictions with experimental measurements and then repeatedly cycle the results back into the next round of machine learning and experiments.”

Designing chiral crystals
In a separate breakthrough, Hiroshima University has used machine learning to design chiral crystals.

Chirality is derived from the Greek word for hand. If one holds up their hands, they are identical in structure, but are mirror opposites.

The same is true for molecules. If a molecule is chiral, then the molecule has two enantiomeric forms. They are identical, but are actually different molecules.

Hiroshima University has found a way to make chiral molecules. Researchers used logistic regression analysis, which predict which chemical groups are best to design chiral molecules. Logistic regression is a statistical method, which can discern two objects.

Researchers analyzed 686 chiral crystals and 1,000 achiral crystals from a database. From there, they calculated which chemical groups could coexist in a chiral crystal.

“The most difficult part of making a chiral crystal is knowing how to design them,” said Katsuya Inoue, a researcher at the Graduate School of Engineering at Hiroshima University.

“We started from two atoms. In reality, though, many crystals are made with three or four,” Inoue said. “We have to extend this model to fit for these cases.”