Moh's scale of hardness goes from 1 to 10 - talc to diamond. Materials with a hardness that goes all the way up to 11 were originally referred to as "superhard"; nothing yet fits that description. There were hints of superhard, harder than diamond materials over the last few decades but none has stood up to close inspection by that definition. In the 1980s covalent substances containing boron, carbon, nitrogen, and oxygen atoms instead of just the carbon of diamond were investigated. Later transition metals were added to the blend in an effort to nudge a material's incompressibility beyond 40 gigapascals (GPa) when measured by the Vickers hardness test (the more modern, accepted definition of superhard). Nothing made it to the 70-150 GPa of diamond.

Now, researchers at the University of Buffalo, New York, USA, have taken a computational approach in the quest to find new superhard materials. Their study reveals 43 previously unknown allotropes of carbon that ought to be stable and superhard. There is even the tantalizing possibility that some of these structures might be as hard as or slightly harder than diamond. [Avery, P. et al., Computat. Mater., (2019); DOI: 10.1038/s41524-019-0226-8]

The work reported in Computational Materials combined computational predictions of crystal structure with machine learning to hunt for novel materials.

"Diamonds are right now the hardest material that is commercially available, but they are very expensive," explains Buffalo chemist Eva Zurek. "I have colleagues who do high-pressure experiments in the lab, squeezing materials between diamonds, and they complain about how expensive it is when the diamonds break." An inexpensive replacement for genuine diamond could have a wide range of uses not least in those high-pressure laboratories, but also as protective coatings, electronic heat-sinks and other applications.

Among the 43 structures predicted on is partially diamond and partially so-called hexagonal diamond. Others have would have more intriguing crystal structures should it be possible to synthesize them.

The team describes their approach as a theoretical way to carry out "accelerated material development". Team member Stefano Curtarolo explains that the "The algorithms learn, and if you have trained the model well, the algorithm will predict the properties of a material - in this case, hardness - with reasonable accuracy." With the predictions in hand, it is then the over to the experimentalists to find ways to synthesize them from scratch in the lab starting with the most promising.

David Bradley blogs at Sciencebase Science Blog and tweets @sciencebase.