Using combinatorial chemistry, researchers have designed a new class of metallic glasses that works well at high temperatures and should be easy to form into useful components (Nature 2019, DOI: 10.1038/s41586-019-1145-z).
Metallic glasses are alloys in which atoms occupy random positions, as opposed to the neat crystalline arrangements in conventional metals. They can be shaped like plastics and are strong and corrosion resistant, which makes them ideal for fashioning gears, precision molds, and electrochemical devices. But they’re tricky to make because they can be molded only within a narrow temperature window. When heated, the alloys become viscous fluids that can be molded. But apply too much heat and they crystallize and become brittle.
“We wanted to develop metallic glasses with both high glass-transition temperature and wide liquid region,” says YanHui Liu, a materials scientist at the Chinese Academy of Sciences. Liu and his colleagues first identified elements with the desired physical properties, such as high melting points. They chose iridium, nickel, and tantalum.
Then they used plasma sputtering to make a library of thousands of alloys, each with slightly different ratios of the three metals, on a silicon wafer. They could make glassy rods from seven of the most promising alloys and could pull the rods into nanowires and mold them into tiny gears.
The new materials had a liquid window of 136 °C, wider than most current alloys. They can also withstand extremely corrosive conditions: rods made with them did not corrode for up to 112 days in aqua regia, while those made of zirconium-based metallic glasses corroded within an hour. They are in many ways superior to today’s brittle, hard-to-process iridium-based alloys, the researchers say.
The new metallic glasses have good properties, says Christopher Wolverton, a materials scientist and engineer at Northwestern University, and he says there’s potential to make the screening method even more powerful by combining data generated through the combinatorial experiments with machine learning.