Super hard materials are indispensable for many tools. Image credits: pixabay.com.
Superhard materials are in great demand in modern technological applications. In several laboratories, a large number of studies are conducted with the intention of producing increasingly harder materials, using various processes such as mixing, alloying, impurification, formation of new compounds, among others, because these materials are essential for both tools, armament parts and even aerospace applications.
However, these materials are basically developed through trial and error, practicing structures that resemble that of other known superhard materials, such as diamond. But this is about to take a big turn, as researchers from the University of Houston and Manhattan College have reported for the journal Advanced Materials a machine learning model that can accurately predict the hardness of new materials, so they can more easily find suitable compounds for a wide variety of applications.
The materials that are super hard are those that have proven to be as strong or stronger than diamond, according to the Vickers scale are those that need more than 40 gigapscals of pressure to leave a crack on its surface, so they are scarce materials. This group of scientists has demonstrated that a learning method in conjunction with machines is capable of finding superhard materials by directly predicting the load-dependent Vickers hardness based on chemical composition alone.
The new model of machine learning seeks materials harder than diamond. Source: image designed by the author using the public domain images 1, 2.
A total of 1062 load-dependent Vickers hardness data extracted from the literature were used to train a supervised machine learning algorithm. This model was then tested by synthesizing and measuring the load-dependent hardness of several unreported disilicides and analyzing the predicted hardness of several classic superhard materials, obtaining 97% accuracy. The hardness model is then combined with the data-based phase diagram generation tool to extend the limited number of reported high hardness compounds.
Machine learning has traditionally been used to predict a single hardness variable, but that left the complexities of this property, such as its dependence on load, unexplained, making this new model a great tool.
Thanks to this, researchers have reported to have found more than 10 new and promising stable phases of borocarbons with thermodynamically favorable compositions and with a hardness higher than 40 GPa; which demonstrates the capacity of this set model to find previously unknown materials with outstanding mechanical properties.
Although machine learning still has many limitations in this field, they have collected a good set of data with which to obtain fairly accurate statistical predictions, although it still cannot provide the exact composition for the best superhard material, it certainly helps them to guide the experimental search.
Thanks for coming by to read friends, I hope you liked the information. See you next time.