Scientists model new superhard materials with artificial intelligence

A group of scientists from the University of Buffalo automated 43 new forms of carbon, some of which could be even tougher than diamonds.

Each computer-modeled carbon variety is made of carbon atoms placed with a certain pattern in a kind of crystal lattice. The researchers then published the study on npj Computational Materials, a study that also confirms how artificial intelligence, and in particular the machine learning technique, can also be important for the research of new materials and in general in all fields of materials science.

Among other things, superhard materials can be very useful because they can cut, drill or even grind other materials and objects and can also be used to create scratch-resistant coatings. Currently there is no harder material than diamond, but the latter is also very expensive, as Eva Zurek, a chemist at the University of Buffalo, who was involved in the research, recalled.

Precisely for this reason, there are many laboratories around the world that try to synthesize materials, at least by modeling them, which are harder than diamonds and possibly cheaper. However, these are often long and laborious processes, and this is where the computer and the new and increasingly powerful artificial intelligence algorithms play a role.

With the computer, you can get modeled materials that can exhibit other interesting properties, such as certain interactions with heat or electricity or other properties that diamonds don’t have.
“Few superhard materials are known, so it’s interesting to find new ones,” says the researcher, who suggests that the open-source algorithm they used, called XtalOpt, to generate random and crystalline structures containing carbon, can also be used to discover new structures and new materials in an increasingly efficient and fast way and that some of them can reserve a pleasant surprise.

See also:

Image source:

Leave a Reply

Your email address will not be published. Required fields are marked *